Metabolic Responses to Potassium Availability and Waterlogging in Two Oil- Producing Species: Sunflower and Oil Palm

Home , Aldaric acid, Citramalate lyase, Perakine reductase, Zerumbone synthase

Metabolic responses to potassium availability and waterlogging in two oil- producing species: sunflower and oil palm

Jing Cui Supervisor: Prof. Guillaume Tcherkez

A thesis submitted for the degree of Doctor of Philosophy The Australian National University Declaration

Except where otherwise indicated, this thesis is my own original work. No portion of the work presented in this thesis has been submitted for another degree.

Jing Cui

Research School of Biology College of Medicine, Biology and Environment

September, 2019

1 Acknowledgements

Primarily my deepest thank goes to my principal supervisor Professor Guillaume Tcherkez (Research School of Biology, The Australian National University, Canberra), who has always given me help, advice and support throughout my PhD, Thank you very much for your enthusiasm, patience, efficience and guidance. One of the major motivations I committed to this PhD is your strong faith in me, I have been truly fortunate to work with you for the last three years. I thank my co-supervisor Dr. Emmanuelle Lamade (CIRAD, France), for your time, thoughtful comments and invaluable advice that they gave me along my project. Thank you also to my PhD committee members Dr. Adam Carroll (Research School of Chemistry and Biology Joint Mass Spectrometry Facility), Dr. Cyril Abadie, Dr. Hilary Stuart-Williams and Prof. Marilyn Ball (Research School of Biology, The Australian National University) for their help and guidance.

I would like to thank Dr Thy Truong (Research School of Chemistry and Biology Joint Mass Spectrometry Facility) for taking the time to share with me her expertise in GC-MS, and look forward to working with her in LC/MSMS, also thank Dr.Marlène Davanture and Dr. Michel Zivy (PAPPSO, France) for their help in proteomics analysis. In particular, I would like to acknowledge the advice from Dr. Juan Chao de la Barca (University of Angers, France) for statistical analysis. I thank Les and Linda (Research School of Earth Science, ANU) for their assistance with ICP-OES analysis and for helpful discussions of digestion protocol. In addition I wish to thank Andrew Higgins (Fenner School, ANU) for welcoming me to do digestion experiment in his laboratory. I am also grateful for the invaluable advice on gas exchange matters offered by Dr. Florian Busch and Ross Dean (both of the Research School of Biology, ANU). I thank Dr Christina Spry and Professor Kevin Saliba (both of the Research School of Biology) for letting me work in their laboratory, so I can finish all my experiment before lab moving to Robertson Building. Many thanks go to the Plant Science HDR convenor, Professor Spencer Whitney for his encouragement and giving guidance of completing all PhD milestone. Also, I want to thank Professor Barry Pogson for giving me opponituty to involve in teaching, which is one of my favourite parts during study. I would like to thank Research School of Biology Plant Service Team (Steven Dempsey, Darren, Gavin, Steve Zabar, Jenny and Christine) for their kind help in the glasshouse. Thank you to Stephanie McCaffery, Ursula Hurley,

2 Dr. Farid Rahimi and Graham Hicks, for helping me solve different kinds of problems in their first time. Thank you to Gagan Bhardwaj and Helen Wong for their kind administrative help, I could not have finished all the headache paperwork without them.

I also would like to say thank you to my group of friends and colleagues in the Tcherkez laboratory and Research School of Biology. It was an absolute luck to work with, especially, Dr. Camille Bathellier, Dr. Illa Tea, Sophie Blanchet, Jean-Baptiste Domergue, Cathleen Mirande-Ney, Sophie Renou, Haydon Siiteri, Peter Groeneveld, Dr. Charles Hocart, Dr. Suan Wong, Thomas Davis, Denis Hawkins, Dr. John Rivers, Dr. Kai Chan, Dr. Xin Hou, Kiran Javed, and Graeme Clemow. This project would not have been done without their friendship, encouragement and cheering me up.

Additionally, I am thankful to receive the Australian postgraduate Scholarship (APA), without which I would not have been able to undertake this PhD. This whole project was partly funded by the Australian Research Council via a Future Fellowship awarded to my principal supervisor, under contract FT140100645. Also, I would like to acknowledge the support from the ANU Vice-Chancellor’s HDR Travel Grant and CSIRO Plant Nutrition Trust Travel Grant, under these grants support, I attended two international conferences and shared my PhD results with other scientists.

Very big thank you to my family, thanks my parents and parents-in-law’ help, support and all of the sacrifices you have made for me, especially during the hard times.

Last but not least goes to my beloved husband, Weihua, for everlasting support throughout this thesis, wherever possible, for patiently during my tantrums, for being such a wonderful husband in so many ways for the last twelve years of my life, and for your unconditional love, I will not be able to go through this amazing journey without you.

3 Abstract

Oil palm and sunflower are important major oil-producing crops at the global scale, with 70 Mt oil produced each year. Although they have been studied for decades for their agronomical aspects, their metabolism is poorly documented, in particular situations such as low potassium. In fact, these species are strong potassium (K)-demanding species cultivated in regions where soil K availability is generally low and waterlogging caused by tropical heavy rains or poor drainage system can limit further nutrient absorption. That is, K deficiency and waterlogging are common stresses that can occur simultaneously and impact on crop development and yield. They are both known to impact on catabolism, with rather opposite effects: inhibition of glycolysis and higher glycolytic fermentative flux, respectively. However, the metabolic effects of K and waterlogging have never been assessed precisely. This is very surprising because most tropical or wet areas where rice, sunflower, oil palm and other important crops are cultivated combine these two environmental constraints.

The aim of this study was to understand the overall impact of waterlogging and limited K availability on metabolism, and to understand how such advert environmental conditions reshape the carbon balance in two-oil producing species (sunflower and oil palm). Here, we examined the metabolic response of sunflower seedlings and oil palm saplings in the greenhouse under controlled conditions (nutrient composition with low or high K availability, with or without waterlogging) using gas exchange, metabolomics, proteomics, elemental quantitation, isotope and major flux analyses at different sampling times.

While separate K deficiency and waterlogging caused well-known effects like polyamines production and sugar accumulation, respectively, waterlogging altered K- induced respiration enhancement and polyamine production, and K deficiency tended to suppress waterlogging-induced accumulation of Krebs cycle intermediates in leaves, stems and roots. Furthermore, the natural 15N/14N isotope composition (δ15N) in leaf, stem and root compounds shows that there was a change in nitrate circulation, with less nitrate influx to leaves and stems under low K availablity combined with waterlogging, and more isotopic dilution of lamina nitrates under high K.

4 Our results also show that both low K and waterlogging have a detrimental effect on photosynthesis but clearly stimulate leaf respiration, thereby impacting the carbon use efficiency. Omics analyses show differential accumulation of typical metabolic intermediates and enzymes not only of the Krebs cycle but also of alternative catabolic pathways. In addition, we found a strong relationship between metabolic composition and the rate of leaf dark respiration. Finally, based the results we have and from literature, in the last chapter of this thesis, we briefly describe potential roles of putrescine, and then review data that help defining the most likely specific role of putrescine under K deficiency.

Overall, advert environmental conditions (here, low K and waterlogging) have an enormous impact on respiration in oil producing species (sunflower and oil palm here). Leaf metabolome and proteome appear to be good predictors not only of K availability but also of CO2 efflux, and this opens avenues for cultivation biomonitoring using functional genomics technologies.

5 Guide

This thesis is structured in five chapters: Introduction, Materials and methods, two Results chapters based on experiments, General discussion and perspectives chapter and Appendix chapters. The Introduction focuses on the aims of this thesis and literature review. The two Results chapters can be read semi-individually, together with the relevant sections in the Materials and methods, since each presents data on separate experimental systems. The general discussion chapter aims to provide an integrated discussion and reviews the major outcomes of this thesis. Below is an overview:

Chapter 1: Introduction - literature review on the topics investigated in this thesis. Chapter 2: Materials and methods - information on the experimental procedures performed. Chapter 3: Responses to K deficiency and waterlogging interact via respiratory and metabolism in sunflower seedlings: Results and Discussion. Chapter 4: Metabolic responses to potassium availability and waterlogging reshape respiration and carbon use efficiency in oil palm saplings: Result and Discussion. Chapter 5: General discussion and perspectives.

6 List of publication from this thesis

Journal publications • Cui J, Davanture M, Zivy M, Lamade E, Tcherkez G. (2019) Metabolic responses to potassium availability and waterlogging reshape respiration and carbon use efficiency in oil palm. New Phytologist 223: 310-322 • Cui J, Abadie C, Carroll A, Lamade E, Tcherkez G. (2019) Responses to K deficiency and waterlogging interact via respiratory and nitrogen metabolism. Plant, Cell & Environment 42: 647-657 • Cui J, Lamade E, Pottosin I, Tcherkez G. (2019) What is the role of putrescine accumulated under potassium deficiency? Plant, Cell & Environment. In press. (doi: 10.1111/pce.13740) • Cui J, Lamade E, Fourel F, Tcherkez G. (2019) δ15N values in plants are determined by both nitrate assimilation and circulation. New Phytologist. In press. (doi: 10.1111/nph.16480) • Cui J, Lamade E, Tcherkez G. Potassium deficiency reconfigures sugar export and induces dopamine accumulation in oil palm leaves. Plant Science (under reviewed) • Cui J, Chao de la Barca JM, Lamade E, Tcherkez G. The metabolomics signature is a reliable tool to assess K nutrition in oil palm saplings. Plant Planet People. (under reviewed) • Cui J, Lamade E, Tcherkez G. Seed germination in oil palm (Elaeis guineensis Jacq.): a review of metabolic pathways and control mechanisms. IJPM (under reviewed) Conference posters • Cui J, Mirande-Ney C, Lamade E, Tcherkez G. (2018) Metabolic responses to potassium deficiency in oil palm: interactions between primary metabolism and potassium availability. [Conference Paper: 1879/PPKS/0.11IX2018] The 6th International Oil Palm Conference (IOPC): Smoothing the Market Disequilibria, July 17-19, Medan, Indonesia • Cui J, Davanture M, Zivy M, Mirande-Ney C, Lamade E, Tcherkez G. (2019) Metabolic responses to potassium availability and waterlogging in oil palm. 15th Annual Conference of the Metabolomics Society, June 23-27, The Hague, Netherlands Related publication by the author not included in this thesis • Mirande-Ney C, Tcherkez G, Balliau T, Zivy M, Gilard F, Cui J, Ghashghaie J, Lamade E. (2020) Metabolic leaf responses to potassium availability in oil palm (Elaeis guineensis Jacq.) trees grown in the field by MRS. Environmental and Experimental Botany. In press. (doi: 10.1016/j.envexpbot.2020.104062) • Tcherkez G, Mariem S B, Larraya L, García-Mina J M, Zamarreño A M, Paradela A, Cui J, Badeck F, Meza D, Rizza F, Bunce J, Han X, Tausz-Posch S, Cattivelli L, Fangmeier A and Aranjuelo I. Despite minimal effects on yield, elevated CO2 has concurrent effects on leaf and grain metabolism in wheat. Journal of Experimental Botany (under revision)

7 List of abbreviations

ATP Adenosine triphosphate CRL Cumulated respiratory loss

CUE Carbon use efficiency

EI Electron Ionisation

FDR False discovery rate

GC Gas Chromatography

GC/MS Gas Chromatography Mass Spectrometry (Spectrometer)

HPLC High-Performance Liquid Chromatography

ICP-OES Inductively coupled plasma - optical emission spectrometry

K Potassium

LC-MS/MS Liquid chromatography-tandem mass spectrometry

M+ Molecular ion

MM Molecular Mass

MS Mass Spectrometry (Spectrometer) m/z Mass-to-charge ratio

N Nitrogen

NMR Nuclear magnetic resonance

NSAF Normalized spectral abundance factor

OPLS-DA Orthogonal Projections to Latent Structures Discriminant Analysis

O2PLS Two-way Orthogonal Partial Least Squares

PCA Principal Component Analysis

RI Retention index

SPE Solid phase extraction

TCA cycle Tricarboxylic acid cycle

PEP Phosphoenol pyruvate

PEPC Phosphoenol pyruvate carboxylase

GABA g-aminobutyric acid

8 Table of contents

Declaration ...... 1 Acknowledgements ...... 2 Abstract ...... 4 Guide ...... 6 List of publications from this thesis ...... 7 List of abbreviations ...... 8

Chapter 1: Introduction 1.1 Overview of potassium abundance and roles in plants ...... 12 1.1.1 Potassium in soils ...... 12 1.1.2 Potassium abundance in plants ...... 14 1.1.3 K+ transporters and channels ...... 14 1.1.4 Roles of potassium ...... 20 1.1.4.1 K+ role on anzyme activation ...... 20 1.1.4.2 K+ role in photosynthesis ...... 27 1.1.5 K+ deficiency alters transcript and protein profiles: overview ...... 30 1.1.6 K+ deficiency effects on respiration ...... 32 1.1.7 K+ effects on metabolite profiles ...... 33 1.2 Overview of waterlogging stress in plants ...... 35 1.2.1 Metabolic acclimation to waterlogging ...... 36 1.3 Cross effect of K+ deficiency and waterlogging ...... 39 1.3.1 Co-occurrence of K+ deficiency and waterlogging ...... 39 1.3.2 Ionic aspects of the interaction between waterlogging and potassium nutrition ...... 39 1.3.3 Metabolic aspects of the interaction between waterlogging and potassium nutrition ...... 40 1.4 Background information on two oil-producing crops, sunflower and oil palm, and their specific K requirements...... 41 1.4.1 Sunflower ...... 42 1.4.2 Oil palm ...... 43 1.5 Research questions and thesis outline ...... 48 1.5.1 Key research question addressed here ...... 48 1.5.2 Specific objectives and impact ...... 49

9 Chapter 2: Materials and methods 2.1 Plant material, growth conditions, and stress treatments ...... 65 2.1.1 Plant material and growth condition ...... 65 2.1.2 Nutrition solution and waterlogging ...... 67 2.1.3 Experimental design ...... 69 2.1.4 Experimental sampling ...... 71 2.2 Relative chlorophyll content...... 75 2.3 Photosynthesis and respiration ...... 75 2.4 Metabolomics ...... 75 2.5 Ionomics ...... 76 2.6 Proteomics ...... 76 2.6.1 Protein extraction and in-solution digestion ...... 77 2.6.2 LC-MS/MS analyses ...... 77 2.6.3 Protein identification and peptide quantification ...... 78 2.7 Estimation of CUE ...... 78 2.8 Isotope analysis and %N ...... 79 13 2.9 CO2 labelling and sampling ...... 79 2.10 Extraction and NMR analysis ...... 80 2.11 Statistical analyses and data presentation ...... 80

Chapter 3: Responses to K deficiency and waterlogging interact via respiratory and metabolism in Sunflower 3.1 Motivation ...... 83 3.2 Context ...... 83 3.3 Key results and implications ...... 83 3.4 Published paper ...... 83

Chapter 4: Metabolic responses to potassium availability and waterlogging reshape respiration and carbon use efficiency in oil palm 4.1 Motivation ...... 104 4.2 Context ...... 104 4.3 Key results and implications ...... 105 4.4 Published paper ...... 105

Chapter 5: General discussion and perspectives 5.1 Comparison of 15N/14N fluxes in sunflower and oil palm under varying K availability and waterlogging (see appendix 1: published in New Phytologist) ...... 137 5.2 The control of respiration by K availability and thus the impact on carbon use efficiency: summary ...... 145

10 5.3 Distinct effect of waterlogging on leaf respiration in oil palm and sunflower….150 5.4 The role of putrescine (see appendix 2: paper on putrescine roles, published in Plant Cell and Environment) ...... 151 5.5 Rationales for the selected K concentration for treatments and their appropriateness in agricultural K fertilization practice ...... 152 5.6 Limitations of this thesis ...... 154 5.7 Future directions ...... 155 Appendix 1: δ15N values in plants are determined by both nitrate assimilation and circulation ...... 158 Appendix 2: What is the role of putrescine accumulated under potassium deficiency ...... 184 Appendix 3: The metabolomics signature is a reliable tool to assess K nutrition in oil palm saplings ...... 206 Appendix 4: Potassium deficiency reconfigures sugar export and induces dopamine accumulation in oil palm leaves ...... 213

11 Chapter 1 - Introduction

In the present thesis, the physiological and metabolic effects of potassium limitation and waterlogging have been investigated on two oil-producing crops, namely, sunflower and oil palm. Therefore, in this Introduction, I first give a brief review of potassium and its roles in plants, followed by an overview of waterlogging stress, and summarize sunflower and oil palm nutritional (K) requirements.

1.1 Overview of potassium abundance and roles in plants 1.1.1 Potassium in soils

Potassium (K) is the eighth most abundant element, constituting about 2% of Earth’s crust. At first glance, it might therefore be expected that soil K+ reserve should be sufficient for plant growth and development. However, a high proportion of agricultural lands of the world are recognized as being K deficient. For example, most equatorial regions, three quarters of paddy soils in China, and two thirds of the wheat belt of Southern Australia are K deficient, these regions being in general also associated with acidic soils (Römheld and Kirkby, 2010) (Figure 1). It is also worth noting that regions where sunflower (USA, Siberia) or oil palm (equatorial region, including Malaysia and Indonesia) are cultivated intensively are notoriously K-deficient (Figure 1).

Figure 1. World map of K-deficient soil regions (red color). Here, K-deficient means that total available K is less than 0.1 mg g-1 soil in the first soil layers (< 1 m depth). From FAO Soil Portal (2020). (http://www.fao.org/soils-portal/soil- survey/soil-maps and%20databases/harmonized-world-soil-database-v12/en)

12 K deficiency in soils can be caused by several reasons. First, inherently low available K. Although K is very abundant in the Earth’s crust, it is not distributed evenly amongst soil horizons (layers) or geographical regions, depending on the geological substratum and the type of soil. As such, several soil types are particularly depleted in exchangeable K. Second, the rate of removal of K from soil (leaching or removal upon crop harvesting) may be higher than that of replenishment (fertilization). It is estimated that the harvest of crops worldwide removes 20 million tons of K from soil (Smil, 1999). In developing countries, crop residues (e.g. straws) are also removed from land and used as house fuel, building materials etc. Crop residues contains 60 million tons K globally (Smil, 1999). This practice further remove K from soil. In addition, the application of K fertilizer is at a much lower rate. It is estimated that less than 50% of K removed is replenished (Römheld and Kirkby, 2010; Zorb et al., 2014). In the case of oil palm agrosystems, usual fertilization practice (e.g. in Sumatra) adds about 180 kg K ha-1 y-1 while harvesting may represent up to 250 kg K ha-1 y-1. Third, the vast majority of K in soil is not readily available to support plant growth because it is not in the form of exchangeable K but remains insoluble in the mineral phase.

Four distinct pools of soil K are recognized: (a) structural K: 90 to 98% of K in soil is present in the crystal lattice structure of minerals such as feldspar and mica, which cannot be absorbed by plants. However, it contributes to the plant available K pools in the long term due to weathering; (b) non-exchangeable K, trapped between 2:1 clay mineral interlayer sites, accounting for 1 to 10% of soil K. It also serves as a reservoir for plants and it is usually included in “available” K pools; (c) exchangeable K: electrostatically bound to the surface of clay minerals and soil organic matter (humic acids), and makes up about 1–2% of total soil K; (d) soil solution K, which is readily available for plants uptake and represents only about 0.1–0.2% soil K (Zorb et al., 2014; Prajapati and Modi, 2012; Romheld and Kirkby, 2010). Typical K concentrations in the soil solution vary between 0.1 and 5 mM (Schroeder et al., 1994; White et al., 2007). Plants primarily absorb K from the soil solution, but K can be released into soil solution from exchangeable K pools. Therefore, both solution K and exchangeable K are considered as being plant-available K. However, the pool size of these K sources is very small. In addition, the absorption of K by plants is often faster than the release of K from exchangeable K pool, often leading to very low K concentration in soil solution (Pretty and Stangel, 1985; Johnston, 2005).

13 1.1.2 Potassium abundance in plants

Potassium ion (K+) is one of the most abundant cation in plants, representing up to 5% of plant dry weight (Leigh and Wyn Jones, 1984). K+ assimilated by roots is translocated to shoots via the xylem transpiration stream and a high porpotion of K+ is circulated back to root via the phloem (Marschner et al., 1997). The high mobility of K+ allows recycling from older to younger developing tissues or reproductive organs, typically when the demand in K+ in these tissues exceed the quantity of K available locally. This is why visible symptoms of K deficiency appear on older leaves first when plants are grown under K-deficient conditions (Marschner, 1995).

The two major K+ pools in plant cell are the cytoplasm and the vacuole. K+ concentration in the cytoplasm is usually rather constant at about 100 mM, and is essential for maintaining the rate of K+-dependent processes. By contrast, vacuolar K+ varies widely (10−200 mM, Lüttge and Higinbotham, 1979; MacRobbie, 1970, Walker et al., 1996). In addition to providing turgor and drive cell expansion, K+ located in the vacuole can play the role of a remobilizable K reserve. When K is sufficient in the external environment, vacuolar K+ accumulates, and may be re-used later to cope with sudden environmental stresses (Kafkafi, 1990). Under K-deficient conditions, plant cells may mobilize vacuolar K+ to feed the cytoplasm and keep a constant concentration in the short- term. However, in the long-term, cytosolic K+ concentration gradually decreases under K deficient conditions, ultimately influencing biochemical and physiological processes in plant cells. Many physiological processes such as electrochemical homeostasis, stomatal aperture and enzyme activities (Wang and Wu, 2013; Anschütz et al., 2014) require relatively high K+ concentrations in the cytoplasm (Ahmad and Maathuis, 2014) (see discussion below). The K+ concentration in mitochondria is also a critical aspect of plant stress physiology, and this will be addressed later (see Appendix on the role of putrescine).

1.1.3 K+ transporters and channels

Cytoplasmic K+ concentration in plant cells is approximately 100 mM, which is much higher than K+ concentration in soil (0.1 mM to 5 mM). Therefore, uptake of K+ by plant root cells from the soil is against the K+ concentration gradient. Early work using barley roots revealed that the uptake of K exhibits a dual affinity (high and low) mechanism (the

14 + affinity if defined by Km): the high affinity K transport system with an apparent affinity

(Km) at about 20 μM operates at low external concentrations range (below 0.2 mM) and + the low affinity K transport system with a Km at about 11 mM opperates at high external concentration range (above 0.2 mM) (Epstein et al., 1963). More recently, molecular and genetic studies demonstrated that the high affinity component is mainly mediated by K+ transporters, while the low affinity component involves K+ channels. It is believed K+ transporters mediates the active transport of K+ into the cell against its concentration gradient which is due to its symport with proton (Britto and Kronzucker, 2008; Szczerba et al., 2009). The passive influx of proton into the cells dependent on the proton motive force that is primarily established and maintained by the H+-ATPase, drives the infux of K+. Transport of K+ mediated by channels relys on the membrane poteintial gradients, including chemical gradient and eletrical gradient, for K+ across the membrane. Though K+ channels usually operate at relative high external K+ concentrations (mM ranges), Hirsch et al. (1998) showed that it can mediate K+ uptake at 0.1 mM external supply. The superfacial contradictory observation can be explained by the membrane potential. Accoridng to Nernst equation, at membrane potential higher than 177.5 mV, K+ can passively move into the cells which have a concentation of 100 mM from external medium containing 0.1 mM K+. Over the last decades, many genes encoding K transporters and channels have been identified and characterised (Figure 2). These transporters/channels are involved not only in the absorption of K+ from soil by epidermis (rhizodermis) cells but also in the loading of K+ into the xylem, K+ phloem loading/unloading as well as K+ fluxes between cell compartments (such as between cytosol and vacuole).

+ 푅푇 [K ]푂 푉 = 푙푛 + 푧퐹 [K ]푖

Nernst equation. R is the universal gas constant and is 8.314 J.K-1.mol-1 (Joules per Kelvin per mole). T is the temperature in Kelvin (°C + 273.15). z is the valence of the ionic species. For K+ z is +1. F is the Faraday's constant and is 96,485 C. mol- 1 + + + (Coulombs per mole). [K ]o is the extracellular K concentration while [K ]i is the intracellular K+ concentration.

Figure 2. Overview of K+ transporters and channels in plants. Blue double ovals represent channel, while yellow cylinder represents transporters. For each protein’s details, see text. Notes: *GORK mediates efflux of K+ from guard cells thereby leads to closing of stomata. #HKT in wheat encodes K+ transporter, while in Arabidopsis encodes Na+ transporter (Schachtman and Schroeder, 1994; Uozumi et al., 2000).

16 A total of 71 K+ transporters and channels have been identified in Arabidopsis and they have been classified into five gene families: three transporter families and two channel families (Wang and Wu, 2010; Sharma et al., 2013), described below.

Transporters. The three family K+ transporters are: K+ uptake permeases (KT/HAK/KUP), the high-affinity K+ transporter (Trk/HKT) family and cation proton antiporters (CPA).

• The KT/HAK/KUP family comprises essential components of high affinity uptake systems. For example, as compared to wild-type, athak5 T-DNA insertion mutants in Arabidopsis miss a major part of the inducible high-affinity Rb+/K+ transport system

(Gierth et al., 2005). The estimated Km for Arabidopsis AtHAK5 is between 15–24 μM (Gierth et al., 2005). However, the lack of apparent growth phenotype in athak5 plants suggests that other K+ transport systems can compensate for the loss of AtHAK5. In fact, the K+ channel AKT1 has been shown to contribute to high affinity K uptake as well (discussed below). In addition, AtKUP7 has also been shown to play important roles in K+ uptake in Arabidopsis root (Han et al., 2016). The disruption of AtKUP7 leads to a dramatically decreased K+ uptake rate, especially under low K conditions. Furthermore, the mutation of AtKUP7 leads to remarkably reduced K+ loading rate into the xylem, thereby decreasing shoot K+ content and suggesting that AtKUP7 in essential for K+ translocation from root to shoot via the xylem.

• Trk/HKT genes are present in bacteria, fungi and plants, but have been lost in the animal kingdom (Gierth and Maser, 2007). Trk/HKT from fungi and bacteria function as K+ symporters, coupling K+ exchange with proton or sodium. In plants, Trk/HKT transporters received more attention as sodium transporters, but some members of this family have been shown to transport K+. For example, wheat HKT1, the first Trk/HKT gene identified from plants, facilitates high-affinity K+ uptake when expressed in oocytes (Schachtman and Schroeder, 1994); however, the Arabidopsis counterpart AtHKT1 is a Na+ selective uniporter (Uozumi et al., 2000).

• The CPA family can be further divided into CPA1 and CPA2 sub-families. The former consists of Na+/H+ exchanger (NHX), and the later includes K+-efflux antiporters (KEA) and cation/H+ exchangers (CHX). Alike the Trk/HKT family, not all of CPA members can transport K+, but some do. Reconstitution of tomato LeNHX2 in proteoliposomes

17 demonstrated that it catalyses K+/H+ exchange although with a low affinity (Venema et al., 2003). Zhao and coworkers (2015) using 86Rb+ as a tracer for K+ demonstrated that the plasma membrane-located Arabidopsis CHX14 functions as a K+-efflux transporter. Based on its expression pattern and transport activities, these authors proposed that CHX14 involves in K+ homeostasis and recirculation in plants. NHX1 and NHX2 genes encodes the two major vacuolar NHX isoforms in Arabidopsis. Disruption of them led to reduced ability to compartmentalize K+ in vacuole (Bassil et al., 2011; Barragán et al., 2012). Arabidopsis AtKEA3 functions as a H+/K+ anti-porter in thylakoid membranes. It mediates the H+ efflux from the thylakoid lumen to the stroma, thereby, accelerating the downregulation of pH-dependent energy dissipation when transited to low light (Armbruster et al., 2014).

Channels. Plant K+ channels include two families: Shaker, and Tandem-Pore K+ (TPK).

• The Shaker family comprises nine members in Arabidopsis, (Pilot et al., 2003; Lebaudy et al., 2007). They can be classified into five groups based on their phylogenetic relationships. Interestingly, the members of each group have roughly the same electrophysiological characteristics, suggesting the phylogenetic grouping correlated with the diversification of the functions of these channels. Group I and II contains five inward rectifying channels including AKT1, AKT5, SPIK (also known as AKT6), KAT1 and KAT2. AKT1 plays important role in mediating K+ uptake from soil by the root cells (Hirsch et al., 1998; Gierth et al., 2005; Pyo et al., 2010). It is a voltage-dependent inward- rectifying K+ channel mediating K+ uptake over a broad range of external K+ concentrations (Rubio et al., 2000). Disruption of AKT1 leads to complete loss of inward rectifying K+ currents in mutant root cells (Hirsch et al., 1998). Compared with wild-type seedlings, the growth of akt1 loss-of-function mutant seedlings is compromised on media containing a K+ concentration of 100 μM or less (Hirsch et al., 1998). In addition, the uptake of 86Rb+ as a tracer for K+ in akt1 loss-of-function mutant is dramatically reduced compared to wild-type, at micromolar to millimolar ranges of K+ concentration in the medium (Gierth et al., 2005; Pyo et al., 2010). Furthermore, T-DNA insertion line in AtAKT1 missed a transport system with an affinity of 0.9 mM for Rb+/K+ (Gierth et al., 2005). Taken together, these data demonstrate AKT1 facilites K uptake in plants at both low and high external concentrations.

18 Group III contains only one member AKT2 forming a weak inward rectifier. Modulating the activity of AKT2 via post-translational regulation taps the “potassium battery” for reloading sucrose into phloem (see below section Phloem loading and transport of photosynthates).

Group IV is composed of one member AtKC1 which is not able to form a functional homotetrameric K+ channel, but functions as a regulatory subunit negatively modulating channels from Group I, II and III (Wang and Wu, 2013).

Group V contains channels (SKOR and GORK) that are outward rectifiers. GORK expressed in guard cells and root hairs (Ivashikina et al., 2001; Szyroki et al., 2001). Its activity modulates stomatal movement by mediating K+ release from guard cells (Hosy et al., 2003). SKOR expresses in root stelar tissues, where it contributes to the K+ loading into the xylem sap (Gaymard et al., 1998).

• The TPK channel family comprises six members in Arabidopsis (TPK1-TPK5 and KCO3) (Sharma et al., 2013). Fusing with green fluorescence protein (GFP) revealed that TPK1, TPK2, TPK3, TPK5 and KCO3 localized in the tonoplast (Voelker et al., 2006), while TPK4 is in the plasma membrane and ER membranes (Becker et al., 2004; Dunkel et al., 2008). Studies with yeast vacuolar membrane containing Arabidopsis TPK1 using patch-clamp revealed that TPK1 is a voltage-independent, Ca2+-activated, K+-selective ion channel (Bihler et al., 2005).

• The K+ channels in mitochondria. The existence of four different types of K+ channels in plant mitochondria have been identified, namely, the ATP-sensitive K+ channels + + (PmitoKATP ); the ATP-insensitive K channel; the large conductance K channel 2+ 2+ activated by Ca (mitoBKCa); the large conductance Ca -insensitive, iberiotoxin sensitive K+ channel (mitoBK) (Trono et al., 2015). The molecular identities of these channels are unclear, but AKT1/AKT1 like channels and KCO3 have been proposed to be channels accounting for K+ transport across inner mitochondria membrane (Trono et al., 2015).

19 1.1.4 Roles of potassium

K+ is important for plants because of its vital roles in photosynthesis, osmoregulation, enzyme activation, protein synthesis, ion homeostasis, and the maintenance of anion– cation balances in plants (Bhandal and Malik 1988; Hafsi et al., 2014; Marschner 1986; Zhao et al., 2001; Kanai et al., 2007). In this section, I will focus on reviewing roles of K+ in enzyme activation and photosynthesis since this thesis mostly deals with K+ effects on metabolic profiles, respiration and photosynthesis.

1.1.4.1 K+ role on enzyme activation

Enzyme activation is one of the most important functions of K+ in living cells. K+ is known to be involved in the activation of a myriad enzymes with a wide range of functions (Suelter, 1970). However not all activation by K+ is K+-exclusive, since other + + + + + monovalent cations such as Na , Li , Rb , NH4 can replace some of the effects of K (Gohara and Di Cera, 2016). However, one has to bear in mind that the replacement by other univalent cations on enzyme activation has been tested in vitro without considering the capacity of plant cells to “tolerate the concentrations of the particular cations that are required for activation of specific enzymes” (Miller and Evans, 1957) and their physiological relevance (in particular, Rb+ and Li+ are not naturally present at significant concentration). K+ activates enzymes by providing the right ionic strength or/and ‘functioning’ (proper electrostatic sphere in the active site or K+-binding site) and thus plays the role of a co-factor. In general, K+ complexation benefits to enzyme-substrate interaction and catalysis via either K+ direct contact with the substrate (in the active site) or binding to a separate site (Di Cera, 2006; Page and Di Cera, 2006; Gohara and Di Cera, 2016). In past years, some progress in the structure of enzymes, especially from mammals and bacteria, in the presence of ions and substrates has shed lights on the molecular mechanism of activation by K+. Details on the requirement for K+ by key activities of primary C, N or S metabolism are provided (Figure 3) below:

Figure 3. Overview of K+-requiring enzymes in metabolic pathways. Enzymes that were shown requiring K+ in plants are shown in green boxes. Enzymes in blue boxes are ubiquitously present in animals, bacteria and plants and were shown to require K+ for activation in animal or bacteria, but not in plants. ASNase; L-Asparaginase; BAM, β-amylase; BCAT, Branched-chain aminotransferase; BCKDC, branched-chain α-keto acid dehydrogenase complex; FBPase, Fructose-1,6-bisphosphatase; PDC, pyruvate dehydrogenase complex; SAMS, S-Adenosylmethionine synthetase (also known as methionine adenosyltransferase). SCS, Succinyl-CoA thiokinase.

21 • Pyruvate kinase. In the glycolytic pathway, pyruvate kinase catalyses the conversion of phosphoenolpyruvate (PEP) and adenosine di-phosphate (ADP) to pyruvate and ATP (Boyer et al., 1942). Pyruvate kinase from rat muscle is the first enzyme reported to be activated by K+ (Boyer et al., 1942; Lardy and Ziegler, 1945). It was later demonstrated that the activity of this enzyme totally depends on univalent cations together with magnesium (Mg2+) or manganese (Mn2+) (Kachmar and Boyer, 1953; Mesecar and Nowak, 1997a, b). Pyruvate kinase from different plant species were also activated by K+, with activation effect of K+ saturating at about 50 mM (Miller and Evans, 1957; McCollum et al., 1958). A structural study of pyruvate kinase from yeast complexed with a substrate analog phospho-glycolate and catalytic metal ions, manganese (Mn2+) and K+, has revealed that K+ interacts closely with residues of the active site and substrates (Jurica et al., 1998). Interestingly, the structure of the Na+-bound pyruvate kinase complex is + similar to that with K (Larsen et al., 1997), while the kcat of pyruvate kinase with Na+ activation is about 8% only of that with K+ (Kayne, 1971), suggesting K+ may have function(s) other than the overall geometry of the active site.

• Asparaginase. Asparagine contains a high N-to-C ratio (one of the highest after arginine) amongst amino acids and is relatively chemically inert (in particular the amide group is neutral in most physiological pH conditions), therefore, it is not a surprise that it plays the role of a major nitrogen transport and storage compound in many higher plants, especially in carbon-limiting conditions or in legumes (Lam et al., 1995; Lea et al., 2007). In leaf phloem exudates of Arabidopsis, from dark-grown or dark-adapted plants, asparagine is the major amino acid exported (Schultz, 1994). Asparagine can be hydrolysed by asparaginase into aspartate and ammonia, while the latter can be subsequently reassimilated by glutamine synthetase thus forming glutamine. Therefore, it has been proposed that asparaginase play an important role in providing nitrogen to sink tissues including under N deficiency conditions, by processing asparagine into available ammonium (Curtis et al., 2018). Accordingly, asparaginase gene expression is mainly observed in developing tissues including apical meristems, expanding leaves, inflorescences and seeds (Grant and Bevan, 1994; Lea et al., 2007). Two classes of asparaginase exist in plants: K+-dependent and K+-independent. The presence K+- dependent asparaginase was first shown in developing seeds of various plant species (Sodek et al., 1980). In the absence of K+, asparaginase activity was hardly detectable in cotyledons and/or testa of Pisum sativum, Pisum arvense, Phaseolus multiflorus, Lupinus.

22 albus, L. mutabilis and Zea mays (Sodek et al., 1980). K+ activation saturated at about 20 mM. Na+ can activate asparaginase as well but is much less effective, e.g. only 27 % of enzyme activity can be achieved relative to that with K+ in the testa of Pisum sativum (Sodek et al., 1980). Though two subtypes of asparaginase can co-occur in one plant species, their contribution to hydrolysis of asparagine are different. For example, Arabidopsis recombinant K+-dependent asparaginase (encoded by At3g16150) displays 80-fold higher catalytic efficiency with asparagine than that of the K+-independent isoform (At5g08100), suggesting the former may metabolize asparagine more effectively in planta (Bruneau et al., 2006). The mechanistic basis of K+ activation in asparaginase catalysis has been studied in details with the French bean (Phaseolus vulgaris) enzyme. Two recent studies using recombinant K+-dependent asparaginase PvAspG1 from Phaseolus vulgaris showed that the addition of K+ leads to an increase in the apparent Vmax and a decrease in the apparent Km, although the two studies exhibit numerical differences in fold changes of Vmax and Km (Bejger et al., 2014; Ajewole et al., 2018). Overall, K+ increases more than ten-fold the catalytic efficiency of PvAspG1 (Bejger et al., 2014; Ajewole et al., 2018). Crystal structures of PvAspG1 in the presence of K+ or/and Na+ revealed the mechanism of requirement of K+ for activation of enzyme (Bejger et al., 2014). There are four metal binding sites in PvAspG1 enzyme with two in stabilization loops (residues Leu58–Arg68) and two in activation loops (residues Val111– Ser118). One of the loops is also present in K+-independent asparaginases (“stabilization” loop), while the latter is not (“activation” loop). The K+ binding sites is outside the enzyme active site, suggesting the activation of K+ is not via K+ facilitating the interaction of the active site with substrates. The binding of Na+ or K+ in the stabilization loop does not influence the enzyme active site, consistent with the fact they are present in both K+- dependent and -independent asparaginases. The enzyme active site is located between the two loops. K+ binding in the activation loop changes the conformation thereby allowing the binding of anchoring of the reaction substrate/product in the active site (by contrast, Na+ binding does not allow such a motion). The importance and sufficiency of the activation loop for K+ activation of plant asparaginase is also supported by mutations of the Ser118 residue to Ile in the K+-dependent asparaginase, PvAspG1, and the corresponding residue (Ile-117) to Ser in the K+-independent asparaginase PvAspGT2 (Ajewole et al., 2018). Changing Ile-117 in PvAspGT2 to Ser results in turning PvAspGT2 into a K+-dependent asparaginase with catalytic efficiency increased 45-fold in the presence of K+. Conversely, changing Ser to Ile in PvAspG1 abolishes K+ activation.

23 • Starch synthase. Starch is one of the most important storage carbohydrate in plants, in particular in leaves (including in the species of interest here, sunflower and oil palm). The degradation of transitory starch provides carbon and energy to support sucrose export and plant growth at night (Graf and Smith, 2011) and in starchy seeds, it provides carbon and energy for germination and seedling establishment (Zeeman et al., 2010). Starch synthase catalyses the biosynthesis of starch by adding glucose from ADP-glucose to the pre- existing glucose chains via α-1,4-linkage. K+ stimulates the activity of starch synthase from a wide range of plants including maize, legumes (bush beans, peas, soybeans), wheat, and potatoes (Nitsos and Evans, 1969). The optimum K+ concentration for starch synthase activity is equivalent to concentrations normally found in plant cells (Nitsos and Evans, 1969). Na+ and Li+ can also activate starch synthase, but are less effective than K+. The mechanism of potassium activation is still unclear. Crystal structure of starch synthase from Arabidopsis (AtSSIV), Cyanobacterium sp. CLg1 (CLg1GBSS), the glaucophyte Cyanophora paradoxa (CpGBSSI) and barley were resolved (Cuesta-Seijo et al., 2013; Nielsen et al., 2018), but these proteins were not crystallized in the presence of K+, thereby no information of the role of K+ could be obtained for the current structures.

• β-amylase. Degradation of starch plays important roles in plant growth and development and in the stress response (Lloyd et al., 2005; Sulpice et al., 2009; Zeeman et al., 2010; Streb and Zeeman, 2012). A series of enzymes mediates the degradation of starch, among which is one of the most important enzyme, β-amylase (BAM) (Stitt and Zeeman, 2012). BAM is an exohydrolase hydrolyzing the α-1,4 glucosidic linkages of starch from non- reducing ends to liberate successive β-maltose units. BAM2 is one of the nine BAMs in Arabidopsis (Monroe and Storm, 2018). BAM2 was initially shown to have negligible activity (Fulton et al., 2008; Li et al., 2009), but Monroe et al. (2017) demonstrated that BAM2 requires K+ for activity with maximal activity at 80 mM and above. That is, the use of assays without addition of salts in previous reports probably explains the observed low activity of BAM2. Other univalent cations such as Li+, Na+, Rb+ and Cs+ have a similar effect as K+. In addition, ions were not present in any crystallized plant BAMs (where the active site residues are well-conserved across species). These led Monroe et al. (2017) to propose that the role of K+ may be to provide ionic strength for BAM2 activity. Still, it is possible that K+ forms part of the enzyme structure: although active site residues are conserved across species, K+ may not bind to the active site, since some K+-activated enzymes are stimulated by K+ via binding outside active site (Gohara and

24 Di Cera, 2016). This is typically the case for asparaginase (discussed above), where K+ binding to the activation loop affects the conformation of enzyme (including the active site) thereby affecting catalytic activity (Bejger et al., 2014). Also, the absence of ion(s) from K+-independent BAMs crystal structure may not be representative of the situation in BAM2, which requires K+. Future studies of the crystal structure of BAM2 in the presence of univalent cations will shed light on whether K+ forms part of the structure of BAM2 or simply provide ionic strength.

• Protein synthesis. K+ is required for protein synthesis (Blaha et al., 2000). In plants, K+ has been shown to be required both in vitro (ribosome-mediated GTP hydrolysis), and in vivo (15N-incorporation into proteins) protein synthesis (Amberger, 1975). As such, when plants are under K-deficient condition, protein synthesis is disturbed despite high nitrogen (N) availability. As a result, protein precursors such as amino acids, amides and nitrate accumulate (Armengaud et al., 2009; Wang et al., 2012), and K+ deficiency tends to inhibit nitrogen assimilation. Similarly, in E. coli, it has been first suggested that “decreased cell K+ limits the rate of cell growth by a specific effect on protein synthesis” (Lubin and Ennis, 1964). In fact, K+ is primarily required for ribosome structure: hundreds of potassium ions have been shown to be present in Thermus thermophilus 70S ribosomes structure and many were in crucial positions within the ribosome such as codon-decoding and peptidyl transferase domains (Rozov et al., 2019). K+ ions associated with ribosomes are believed to be essential to stabilize mRNA binding so as to maintain the correct frame position during elongation, stabilize tRNA ligands, and reinforce rRNA–protein interactions (Rozov et al., 2019).

• Fructose-1,6-bisphosphatase. Fructose-1,6-bisphosphatase is a key enzyme in glycolysis and catalyses the hydrolysis of D-fructose 1,6-bisphosphate to D-fructose 6- phosphate and inorganic phosphate. K+ functions as an activator of the enzyme. It has been shown that K+ can increase the maximal velocity of mammalian Fructose-1,6- bisphosphatase (Hubert et al., 1970). Crystallographic structures of the enzyme from pig kidney showed that three K+ binding sites are present and K+ can facilitate the nucleophilic attack on the phosphorus centre (Villeret et al., 1995).

• Succinyl-CoA thiokinase. Succinyl-CoA thiokinase (also known as Succinyl coenzyme A synthetase, succinate thiokinase) is an essential enzyme in TCA cycle catalysing the reversible reaction of succinyl-CoA to succinate. It is activated by univalent cations with

25 K+ as the most effective activator (Bush, 1969; Besford and Maw, 1976). However, structural insight on how K+ activates succinyl-CoA thiokinase is not yet available.

• Other enzymes. S-Adenosylmethionine synthase, branched-chain α-keto acid dehydrogenase and pyruvate dehydrogenase are involved in essential metabolism and ubiquitous in living cell. These enzymes have been shown to be activated by K+ in mammals or bacteria, but there is presently no study of the requirement of K+ for these enzymes in plants. Due to their importance in plant metabolism, I briefly review K+ involvement in these enzymes below.

S-Adenosylmethionine synthase. S-adenosylmethionine (SAM) is the main methyl donor in myriads biological and biochemical events, such as DNA methylation, or phospholipid synthesis. In addition, SAM functions in the aminopropylation pathway and transulfuration pathway which are involved in the synthesis of polyamines and glutathionine, respectively. The formation of SAM from methionine and ATP is catalysed by SAM synthase, which absolutely require K+ (Mudd and Cantoni, 1958). The study of the crystal structure of Escherichia coli SAM synthase in the presence of an ATP analogue and methionine, Mg2+ and K+ has shown that K+ is present in the active site and likely play a key role in maintaining the active site geometry (Komoto et al., 2004).

Branched-chain α-keto acid dehydrogenase. The branched-chain α-keto acid dehydrogenase complex (BCKDC) is involved in the catabolism of branched-chain amino acids (L-isoleucine, L-valine, and L-leucine). The activity and stability of BCKDC from rat liver strictly depends on K+ (Shimomura et al., 1988). The presence of two K+ binding sites in the crystal structure of branched-chain α-keto acid dehydrogenase (BCKD) (Ævarsson et al., 2000), which is one of the subunits in BCKDC, could explain the requirement of K+ for BCKDC activity and stability. Indeed, impeding K+ binding by mutating the K+ binding site (T166M mutant) in human BCKD results in total loss of enzyme activity (Wynn et al., 2001).

Pyruvate dehydrogenase. Pyruvate decarboxylase (subunit E1) together with dihydrolipoyl transacetylase (subunit E2) and dihydrolipoyl dehydrogenase (subunit E3) form the pyruvate dehydrogenase complex (PDC). PDC catalyses the conversion of + pyruvate to acetyl-CoA and CO2 (coupled to the reduction of NAD to NADH). It links glycolysis to the citric acid cycle. High potassium concentration (100 mM KCl) can

26 increase pyruvate decarboxylation activity by 50−100% compared to low K (5 mM KCl) (Lai and Sheu, 1985). A stimulating role of K+ in catalysis is supported by the fact that two pairs of symmetrical K+ binding sites, similar to BCKD, have been found in the crystal structure of human pyruvate dehydrogenase (Ciszak et al., 2003).

1.1.4.2 K+ role in photosynthesis

+ Photosynthetic CO2 fixation depends on K supply, including under abiotic stress condition (Cakmak, 2005; Tränkner et al., 2018). The involvement of K+ in photosynthesis can be summarized in five aspects:

• Rubisco. Ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco) is the key enzyme responsible for photosynthetic carbon assimilation. It has been reported that there is a decrease in the activity of Rubisco carboxylation in plants grown in K+ deficient conditions (Weng et al., 2007; Hu et al., 2015), thereby contributing to reduced photosynthesis. However, the decreased Rubisco activity in K+ deficient plants was unlikely due to the reason that K+ functions as an activator of Rubisco (Jin et al., 2011; Oosterhuis et al., 2013). In fact, Rubisco activity was determined in crude extracts and was not normalized to the amount of Rubisco. In Hu et al. (2015), Rubisco activity was expressed as per g of fresh weight (FW) without taking into consideration the Rubisco content difference. Peoples and Koch (1979) clearly demonstrated that the activity purified Rubisco isolated from K+-deficient Medicago leaflets was not significantly different to that from K+-sufficient leaflets, showing that K+ has no role in Rubisco catalysis. These authors also showed that the in vitro biosynthesis of Rubisco in extracts prepared from K+-deficient leaflets is only 15% of that of K+-sufficient leaflets, but it could be increased by the addition of K+, indicating that the low Rubisco content under K+ deficiency (Peoples and Koch, 1979; Osaki et al., 1993) was caused by lower protein synthesis (Peoples and Koch, 1979; Osaki et al., 1993). This result is consistent with the general role of K+ in ribosome activity and thus protein synthesis (discussed above). Interestingly, the Rubisco content decreased in a K+-deficiency sensitive soybean variety but not in a K+-deficiency tolerant variety (Wang et al, 2015), suggesting that keeping (up-regulating) protein synthesis capacity is a trait that may favours tolerance to K+ deficiency in plant breeding.

27 • ATP synthesis. Efflux of ion such as K+ is required to counter balance the light-induced proton influx into the thylakoid membrane (Hind et al., 1974) which leads to the establishment of transmembrane pH gradient necessary for ATP synthesis. Increasing K+ concentration supplied to roots generally led to an increase in the amount in chloroplastic electron transport complexes and ATP synthase (expressed on a leaf area basis) thereby promoting photosynthesis (Chow et al, 1990; Zhao et al, 2001).

• Phloem loading and transport of photosynthates. Efficient export and utilization of photoassimilates is central to maintain photosynthesis at high rate (in practice, to avoid feedback inhibition), and adequate K supply is known to facilitate phloem transport of photosynthates (Tränkner et al., 2018). Numerous studies showed that K+ deficiency induces the accumulation of sucrose in source leaves and a decrease in sucrose concentration in roots (Cakmak et al., 1994a, b; Bednarz and Oosterhuis, 1999; Zhao et 14 al., 2001; Hu et al., 2017). Using CO2-labelling, Hartt (1969) demonstrated that the photosynthate export rate from source leaves to other organs in K-deficient sugar cane plants is much lower than that under K sufficient conditions. In oil palm, it is believed that K+ is important to allow remobilization (phloem transport) of sugars from the trunk to developing fruits (Lamade et al., 2014, IOPRI Proceedings). The active transport of sucrose against its concentration gradient across the plasma membranes of phloem companion cells and sieve elements by sucrose transporter requires the proton-driven force (gradient). H+-ATPases create the proton gradient across the plasma membrane at the expense of ATP. Under K+ deficient conditions, less ATP is produced thereby less proton gradient can be generated to allow active sucrose transport. In addition, mobile K+ gradients can be used by plants as an energy source for loading sucrose into phloem cells (Gajdanowicz et al., 2011; Dreyer et al., 2017). AKT2 channel can be switched to a nonrectifying (open) channels via an not fully identified post-translational modifications that allow the efflux of K+ out of the cell at entire physiological voltage range. The release of K+ from the cell results in establishing a transmembrane electrical gradient without ATP hydrolysis. Together with existing pH gradient, the electrical gradient can power sucrose transporters to load sucrose into phloem (Dreyer et al., 2017; Gajdanowicz et al., 2010). Disruption of Arabidopsis akt2 results in reduced efficiency of phloem transport of photoassimilates (Deeken et al., 2002). K+ deficiency decreases K+ concentration in the phloem (Peuke et al., 2002) and thus probably leads to a reduced K-driven electrical gradient effect thereby reducing photoassimilates long distance transport efficiency. In

28 addition to the role of K+ in phloem sucrose loading, K+ contributes to phloem osmotic potential (Marschner, 2012). High osmotic potential in phloem is necessary for maintining a water potential allowing of water from the xylem and thus phloem sap movement (Münch theory). Taken as a whole, K+ is very important for sucrose export from source leaves, and K+ deficiency conditions lead to an accumulation of carbohydrates which in turn may inhibit photosynthesis.

• Stomatal conductance. Photosynthetic reactions only depends on photochemical and biochemical processes, but also rely on CO2 availability. CO2 enters leaves via stomata therefore the size of stomatal aperture determines CO2 movement (stomatal conductance). Decreased stomatal conductance caused by K+ deficiency has been widely reported (Jin et al., 2011; Battie-Laclau et al., 2014; Hu et al., 2015; Wang et al., 2015; Lu et al., 2017). The influx and efflux of K+ and organic acids in guard cells regulate the opening and closing of stomata (Maathuis and Sanders, 1996; Marschner, 1995). Thereby, K+ contributes to the controlling CO2 conductance. Two third of the Shaker channel family members are expressed in guard cells including AKT1, KAT1, KAT2 AKT1, AtKC1 and GORK (Sharma et al., 2013). During stomatal opening, H+ efflux from the cytosol mediated by H+-ATPase hyperpolarizes the plasma membrane which activates voltage- regulated inward K+ channels, such as KAT1, KAT2 and AKT1, leading to K+ influx. 2- - - + Malate , Cl and NO3 are transported as counter-ions to balance the uptake of K . During stomatal closure, membrane depolarization caused by inhibition of H+-ATPase and the + + activation of anion channels produces a force for the efflux of K through the K out channel such as GORK in Arabidopsis mediating K+ release from guard cells (Jeanguenin et al., 2008; Hosy et al., 2003).

• Chloroplast development. K-deficiency is closely associated with low chlorophyll content (per unit leaf area) and disorganized chloroplast ultrastructure (Zhao et al., 2001). Also, at extremely low K+ levels in leaves, malfunction of photosystem II (PSII) occurs, resulting in a loss of photosynthetic quantum efficiency (Ball et al, 1987). This effect is consistent with the sensitivity of PSII activity to K+ deficiency in tomato leaves observed decades ago (Spencer and Possingham, 1960) and the role of K+ in accompanying the H+ flux during the light phase of photosynthesis (see above). However, the underlying mechanisms of how K+ deficiency causes general changes in chloroplast ultrastructure are not clear. These might be related to secondary effects of K+ deficiency, such as

29 changed ion balance (like Ca2+ content) which in turn impacts on PSII structure and thus thylakoid morphology.

1.1.5 K+ deficiency alters transcript and protein profiles: overview

Plants are able to (re)orchestrate the expression of genes and change protein composition in response to potassium availability in the environment. Several studies have been carried out to investigate the transcriptional and protein profiles in response to low K conditions in a wide range of species such as Arabidopsis (Armengaud et al., 2004), rice (Zhang et al., 2017), wheat (Ruan et al., 2015), pear (Shen et al., 2017) and millet (Cao et al., 2019). Distinct sets and numbers of differentially expressed genes (DEGs) regulated by K+ availability were observed in different species, but there are some common DEGs. For example, Cao et al. (2019) found that 248 out of the 1982 DEGs between low K+- and high K+-treatments are common to foxtail millet, rice and Arabidopsis. In this section, I will mainly summarize K+ effects on genes involving in carbon/nitrogen metabolism because of the topic of this thesis.

• K+ channels and transporters. Under K+ deficiency conditions, an increase in the uptake of K+ and a relocation of K+ between cellular compartments and tissues were observed in many plant species. Relocation and uptake of K+ require K+ transporters/channels. In Arabidopsis, AtHAK5 is the sole K+ uptake system when external K+ concentrations are below 10 μM (Rubio et al., 2008, 2010; Pyo et al., 2010). Therefore, it is not surprising to observe the upregulation of AtHAK5 in Arabidopsis (Armengaud et al., 2004), and AtHAK5 homologs in other species such as OsHAK1 in rice (Ma et al., 2012).

• Nitrogen assimilation. When plants are under K deficiency, protein synthesis is disturbed despite the abundance of available nitrogen (N). As a result, protein precursors such as amino acids, amides and nitrate accumulate (Wang et al., 2012). Furthermore, K+ - - and NO3 are coordinated during absorption and translocation, since NO3 is the companion counter anion for K+ uptake. Therefore, K+ deficiency impacts on different aspects of nitrogen assimilation. Accordingly, the regulation of nitrate transporters by K+ availability was observed in several studies. Nitrate transporters in Arabidopsis (AtNRT2;1, AtNRT2;3 and AtNRT2;6), rice (Os02g0112600 and Os02g068990) and wheat (Ta.5174.3.S1_x_at, Ta.8619.1.A1_at and Ta.21127.1.S1_at) were down-regulated by low K+ conditions. In addition, the resupply of K+ to K+ deficient Arabidopsis up-

30 regulated AtNRT2;1, AtNRT2;3 and AtNRT2;6. However, some nitrate transporter were found to be up-regulated by K+ deficiency. For example, the expression level of Arabidopsis NRT1.1/CHL1, encoding a dual-affinity nitrate transporter, was up-regulated 3-fold in roots during K starvation. It is noteworthy that the contribution of NTR1.1/CHL1 to nitrate flux is negligible under normal conditions of nitrogen availability (Glass and Kotur, 2013) and thus NTR1.1/CHL1 has been proposed to be a nitrate sensor for nitrate signalling pathway (Gojon et al., 2011). Consistent with decreased nitrate transporter expression, cotton leaves accumulate less ammonium and nitrate and have lower nitrate reductase activity under K deficiency than under K- sufficient condition (Hu et al., 2017).

• Primary C metabolism. It is well known that in plants, low K+ primarily causes an inhibition of pyruvate kinase (PK) activity (explained above), thereby causing a decreased pyruvate production (Evans, 1963; Besford and Maw, 1976). Pyruvate sits at an intersection between many key metabolic pathways. It links glycolysis to the Krebs cycle (or tricarboxylic acid pathway, TCAP). Therefore, plants acclimate to low K conditions via changes in gene expression to reorchestrate metabolic pathways in order to compensate for the diminished pyruvate synthesis. In particular, one would expect that the inhibition of PK activity would lead to an increased expression of gene(s) encoding PK to compensate for the loss of activity. However, the expression level of PK in pear fruits were significantly down-regulated under low K+ as compared to controls (Shen et al., 2017). Pyruvate could also be synthesized from an alternative pathway involving malate synthesis via phosphoenolpyruvate carboxylase (PEPC) and malate dehydrogenase (MDH) followed by malate degradation via the malic enzyme (ME). ME was upregulated in pear fruit (Shen et al., 2017) and Arabidopsis root (Armengaud et al., 2004) under low K conditions. Also, the malate content has been found to decrease under K+ deficiency (Armengaud et al., 2004; Hu et al., 2017), suggesting pyruvate may be synthesized from malate under K deprivation to feed the TCAP. By contrast, the possible involvement of the synthesis of malate via PEPC and MDH is less convincing, since transcript abundance of PEPC and MDH were downregulated in pear fruit at low K+ (Shen et al., 2017) and PEPC activity was lower under K+ deficiency compared to K sufficient condition (Hu et al., 2017).

31 1.1.6 K+ deficiency effects on respiration

Plant K+ status has been known for several decades (first observations in the 50s) to influence respiration. Although K+ is required for some enzymatic activities of catabolism (such as pyruvate kinase), K+ deficiency generally causes an increase (not a decrease) in the respiration rate (CO2 evolution), in particular in leaves. Although this phenomenon has been known for ages, its metabolic origin is not clear. Several possible reasons can be suggested:

• Firstly, there is an enhanced supply of carbohydrates to respiratory metabolic pathways. It has been proposed that cellular sugar availability participates in controlling dark respiration in leaves (Azcón-Bieto and Osmond, 1983). K+ deficiency enhances the accumulation of soluble sugar in leaves due to limited export of photosynthates, and also a lower starch synthase activity (see above). Thus K+ deficiency may increase respiration by a “source” effect.

• Secondly, K+ may lead to an uncoupling of the TCAP to oxidative phosphorylation (Okamoto, 1967). Dinitrophenol (DNP), an uncoupler used for separating the flow of electrons and the pumping of H+ ions for ATP synthesis, increased the respiration of leaf disks from both K+-deficient- and -sufficient-leaves, but to a lower extent in the former, suggesting a ‘loose’ coupling of electron transfer with oxidative phosphorylation in K+- deficient leaves (Okamoto, 1967). This may lead to an inefficient supply of ATP, thereby increasing the demand in NADH and thus in TCA cycle activity to meet the energy demand of cells.

• Thirdly, under K+ deficiency, the metabolite profile changes. These changes may explain part of the production of CO2. For example, putrescine accumulates, sometimes to high levels. The formation of putrescine via either ornithine or agmatine is accompanied by the formation of CO2 (Figure 4). Therefore, the accumulation of putrescine may contribute to the observed enhanced CO2 evolution.

Figure 4. Simplified metabolic pathway of putrescine synthesis.

• Fourthly, possible alternative metabolic routes may contribute to a larger respiration rate. Although the majority of respiratory enzymes are well-conserved in plants, some species can employ alternative pathways bypassing several of the conventional steps of glycolysis or the TCAP. More generally, alternative respiratory bypasses could allow plants to survive under stressful environments (O'Leary et al., 2019). For example, malate decarboxylation by ME to generate pyruvate produces CO2; if it is not coupled to bicarbonate fixation by PEPC, this would lead to an increase in net CO2 production. Unfortunately, to our knowledge, there is little information on changes in metabolic fluxes under K deficiency and thus, the utilization of alternative pathways that might be associated with CO2 release is not well documented.

1.1.7 K+ effects on metabolite profiles

K+ deficiency has far-reaching effects on carbon and nitrogen metabolism. It is well known that K+ deficiency induces an accumulation of soluble sugars (Evans and Sorger, 1966; Cakmak et al., 1994a, b; Bednarz and Oosterhuis, 1999; Zhao et al., 2001; Armengaud et al., 2004; Hu et al., 2017). The reduction of sucrose export into the phloem caused by K+ deficiency results in the accumulation of sucrose in source leaves, which is likely to be converted to other sugars such as fructose, glucose and possibly, starch (Zhao et al., 2001; Hu et al., 2017). However, under low K+ conditions, starch synthesis activity is low because K+ activates starch synthase catalysis (Nitsos and Evans, 1968; 1969) (see above). Therefore, K+ deficiency is expected to cause a general increase in hexoses and their derivatives.

K+ deficiency also induces an increase in total free amino acids (Armengaud et al., 2009; Hu et al., 2017). This could be attributed to K+ deficiency inhibiting protein synthesis. However, all amino acids are not equally accumulated under K deficiency. Amino acids with high nitrogen-carbon ratio (glutamine, asparagine, and arginine) are amongst the most dramatic increased under K+ deficiency. Negatively charged amino

33 acids, such as aspartic acid and glutamic acid have been reported to be decreased by K+ deficiency (Armengaud et al., 2009; Pi et al., 2014). This suggest that factors other than reduced protein synthesis contribute to reshaping amino acids profile under K deficiency. Altered amino acid synthesis may be amongst them. For example, transcript abundance of glutamine synthetase GLN1-1 (At5g37600) and GLN1-2 (At1g66200) increase (Armengaud et al., 2004) and glutamine synthetase activity also increases under K+ deficiency in Arabidopsis roots (Armengaud et al., 2009). Amino acids may also be directed to other pathways. For example, glutamate can be consumed by the synthesis of putrescine, the accumulation of which is induced by K+ deficiency.

As expected, a visible metabolic effect of K+ deficiency in plants is the decrease in pyruvate content, directly caused by the requirement for K+ of pyruvate kinase. Due to the central role of pyruvate in carbon and nitrogen metabolism, it has been proposed that the inhibition of pyruvate kinase activity explains most metabolic disorders in root cells under low-K (Armengaud et al., 2009). One would expect that the decreased pyruvate kinase activity would lead to an accumulation of the pyruvate precursor, phosphoenolpyruvate (PEP). However, PEP content decreased in roots under K deficiency (Armengaud et al., 2009). As mentioned above, this might reflect a lower glycolytic input or the increase in the flux of PEP to oxaloacetate catalysed by PEPC. However, PEPC activity has been found to decrease in K+ deficient cotton leaves (Hu et al., 2017) and the activity of enzymes involved in glycolysis such as NAD-dependent glyceraldehyde-3-phosphate dehydrogenase, glucokinase and fructokinase increased. To compensate for the inhibition of pyruvate production from glycolysis under K deficient conditions, malate may be converted to pyruvate by the malic enzyme. In fact, NADP- ME activity was up-regulated by K+ deficiency in Arabidopsis roots (Armengaud et al., 2009). Accordingly, leaves of cotton grown under K deficiency have less malate and citrate compared to the control (Hu et al., 2017). Arabidopsis roots have also less malate and 2-oxoglutarate (Armengaud et al., 2009). These organic acids are intermediates of the TCAP and precursors for synthesis of other metabolites such as amino acids, thus their content reflect both the restriction of pyruvate synthesis and the increase in free amino acid content. Pyruvate can also be formed from alanine by the activity of alanine aminotransferase. Alanine content did not change under potassium deficiency, but the activity of alanine aminotransferase increased (Armengaud et al., 2009), suggesting alanine may contribute to pyruvate synthesis.

34 The accumulation of polyamines, especially putrescine, under K deficiency is very well documented and known for ages (Jones, 1961, 1966; Okamoto, 1966; Freeman, 1967; Okamoto, 1967, 1968; Besford and Maw, 1976; Armengaud et al., 2009; Hussain et al., 2011; Jwakyung et al., 2015) but still remains enigmatic. In particular, the metabolic roles of putrescine are still not well defined. The biosynthesis and roles of putrescine accumulation under K deficiency will be discussed specifically in Appendix which has been submitted as an article to Plant, Cell and Environment, under revision.

1.2 Overview of waterlogging stress in plants

Flooding due to excessive rainfall, especially in tropical and subtropical regions, causes waterlogging and in some circumstances, complete submergence of plants. Climate change causes the number and severity of flooding events to increase (Hirabayashi et al., 2013). The term “waterlogging” in this thesis refers to soil water excess impacting roots only. It is one of the major abiotic stresses in many regions, affecting crop growth and productivity (Linkemer et al, 1998; Lone et al, 2018). For example, it has been reported that waterlogging reduces crop yields by up to 80% across about 13 million ha of irrigated land in developing countries (Mancuso and Shabala, 2010). In 2006 and 2007, Australian farmers spent AU$300 million per annum to fight against waterlogging in wheat cultivation alone (Dennis et al., 2000).

When soil is saturated with stagnant water (waterlogging conditions), plant roots faces shortage of oxygen, because the oxygen-rich air in soil pores are replaced by water, and the supply of oxygen to water-saturated soil is very slow due to about 10,000-fold higher diffusion resistance for O2 in water than in air (Armstrong et al., 1994). The metabolic activity of plant roots and soil microbacteria consume remaining oxygen in soil, eventually leading to hypoxia (low-oxygen conditions) or even anoxia (no available oxygen). To alleviate the consequences of oxygen deprivation caused by waterlogging, plants initiate various responses including anatomical (such as formation of aerenchyma and pneumatophores), metabolic, and molecular changes. Here, I will discuss metabolic changes since they are relevant to the topic of this thesis (summarized in Figure 5). The reader may refer to reviews to gain information on molecular signaling and cellular changes under waterlogging (Bailey-Serres and Voesenek, 2008; Mustroph et al., 2010).

35 1.2.1 Metabolic acclimation to waterlogging

• ATP balance. Oxygen, as the final electron acceptor in the mitochondrial electron transport chain, is required for aerobic respiration to create chemical energy stored in the form of ATP. In the absence of oxygen, electron transport chain is blocked, thereby inhibiting ATP synthesis. To compensate for this, there is generally an upregulation of the glycolytic flux in order to enhance substrate-level ATP production. At the same time, alternative reactions occur preferably using pyrophosphate (PPi)-dependent kinases over ATP-dependent kinases in hypoxic tissues (van Dongen et al., 2011; Planchet et al., 2017).

• Sucrose metabolism. Both invertase and sucrose synthase catalyse the degradation of sucrose to feed the glycolytic pathway, but the energetic cost of these pathways differs. For the conversion of one molecule of sucrose to two hexose phosphates, the use of invertase requires 2 molecules of ATP, while the use of sucrose synthase only requires one molecule of PPi (Huber and Akazawa, 1986; Stitt, 1998). Unsurprisingly therefore, under waterlogging, invertase activity is inhibited while sucrose synthase activity is promoted (Germain et al., 1997; Zeng et al., 1999). Considering that PPi is produced as a byproduct of many biosynthetic reactions, the overall cost of using PPi dependent sucrose synthase for sucrose breakdown is even lower. In addition, another PPi dependent enzyme, pyrophosphate-dependent fructose-6-phosphate-phosphotransferase (PFP), can play the role of an alternative pathway in lieu of ATP-dependent phosphofructokinase (PFK) in the conversion of fructose-6-phosphate to fructose-1,6-bisphosphate, a key step in glycolysis (Bologa et al., 2003; van Dongen et al., 2011; Planchet et al., 2017).

• Fermentative metabolism. The increase in the flux associated with glycolysis under waterlogging requires sufficient regeneration of NAD+ and this has to be achieved via alternative pathways since oxygen availability is limited. Under hypoxia, pyruvate decarboxylase and alcohol dehydrogenase are up-regulated (Mustroph et al., 2009), indicating a shift to fermentation pathway. The formation of ethanol allows the regeneration of NAD+ to maintain glycolysis, though at low efficiency. One molecule of glucose produces only two molecules of ATP during fermentation – compared to 36 molecules of ATP in aerobic respiration. Beyond reduced energy production, the formation of ethanol and its leakage into the medium results in a carbon loss. Though pyruvate can also be converted to lactate to regenerate NAD+, pyruvate dehydrogenase

36 prevails over lactate dehydrogenase in plant cells under hypoxia condition so that lactic acid is generally of limited importance (Bailey-Serres and Colmer, 2014).

• The alanine cycle. Under hypoxia, pyruvate can also be converted to alanine instead of ethanol. The accumulation of alanine has been regarded as a hallmark of hypoxic metabolism in many plant species (Planchet et al., 2017). Alanine production is catalysed by alanine aminotransferase (AlaAT) from pyruvate and glutamate (Figure 5). Both transcript abundance and enzyme activity associated with of AlaAT are strongly induced by hypoxia (Rocha et al., 2010a). Diab and Limami (2016) proposed that alanine formation is involved in a “AlaAt/NADH-GOGAT cycle” where NAD+ can be produced from NADH by the NAD-dependent glutamine-2-oxoglutarate aminotransferase (NAD- GOGAT), thus generating glutamate which can in turn be used to synthesize alanine from pyruvate. Therefore, in this cycle, NAD+ is regenerated and organic carbon from the glycolytic pathway is saved in the form of alanine rather than lost in the form of ethanol. However, if glutamine is synthesised de novo from glutamate, it requires an extra ATP, which is limiting under hypoxia. Protein degradation is enhanced upon hypoxia (Planchet et al., 2017). Therefore, under the assumption that the alanine cycle takes place, it is likely that glutamine originates from breakdown of protein rather than de novo synthesis (Planchet et al., 2017).

• Acidification. Due to the energy crisis (i.e., ATP limitation), waterlogging also causes cytoplasm acidification because of the decreased activity of plasma membrane proton- pumping ATPase (Planchet et al., 2017). Low pH activates glutamate decarboxylase (GAD) (Carroll et al., 1994; Crawford et al., 1994), which catalyses the synthesis of gamma amino butyric acid (GABA) from glutamate, and consumes a proton. Therefore, the synthesis of GABA is not only a marker of hypoxia (stress) but also contributes to the proton balance of the cell (Reggiani et al., 2000). Accumulation of GABA has indeed been observed during hypoxia (van Dongen et al., 2011; Planchet et al., 2017). In addition, the GABA shunt may contributes to the NAD+ regeneration (António et al., 2016) as well as the alanine cycle (via a pyruvate-GABA aminotransferase, generating alanine and succinic semialdehyde).

Figure 5. Overview of primary metabolism in plant cells under waterlogging. PK, pyruvate kinase; PEPC, phosphoenolpyruvate carboxylase; MDH, malate dehydrogenase; ME, malate enzyme; AlaAT, alanine aminotransferase; GOGAT, Glutamine oxoglutarate aminotransferase; GS, Glutamine synthetase; GAD, Glutamate decarboxylase; GABA T, GABA transaminase; SSADH, Succinic semialdehyde dehydrogenase.

38 1.3 Crossed effect of K+ deficiency and waterlogging

1.3.1 Co-occurrence of K+ deficiency and waterlogging

K-deficient soils are relatively common in countries of the intertropical region, such as Brazil, Equatorial Africa (wet tropical), China and South-Eastern Asia (such as Malaysia or Indonesia), and Australia. In addition, most of these areas are associated with limited soil drainage, occasional flooding, or waterlogging events (Figure 6). These areas include important agricultural regions where crops such as oil palm, sunflower, cotton, or rice are cultivated (www.fao.org).

Figure 6. World map of areas affected by flooding and poor drainage. From FAO Soil Portal (2020). (http://www.fao.org/soils-portal/soil-survey/soil-maps and%20databases/harmonized-world-soil-database-v12/en)

1.3.2 Ionic aspects of the interaction between waterlogging and potassium nutrition

Waterlogging leads to a deficiency in ATP generation in roots, which is required by PM H+-ATPase to generate the proton motive force to maintain membrane potential and nutrient uptake. Therefore, under waterlogging condition, the uptake of many nutrients is reduced or becomes thermodynamically not feasible. Decrease in root K+ content and increased K+ efflux upon hypoxia treatment has been reported (Buwalda et al., 1988; Alcántara et al., 1991; Pezeshki et al., 1999; Smethurst et al., 2005; Kreuzwieser and Gessler, 2010; Zhao et al., 2010). These tend to lead to potassium deficiency (Phukan, Mishra, & Shukla, 2016). Pharmacological and genetic studies indicate that the efflux of

39 K+ could be mediated by depolarization-activated outward-rectifying K+ (GORK) channels (Pang et al., 2006; Wang et al., 2016). In fact, Arabidopsis with loss-of-function of gork1 displays enhanced hypoxia tolerance. The ability of root in maintaining K+ is well-correlated with hypoxia tolerance for barley (Pang et al., 2006; Zeng et al., 2014; Gill et al., 2018). In addition, it has been shown that foliar application of K+ alleviates waterlogging symptoms (Ashraf et al., 2011; Gautam et al., 2016; Dwivedi et al., 2017). In addition, nitrate application also improves plant tolerance to oxygen deficiency (Trought and Drew, 1981; Allègre et al., 2004). Under oxygen limitation conditions, nitrate assimilation was reduced (Oliveira et al., 2013; Oliveira and Sodek, 2013). It is + - not a surprise that K and NO3 were both down regulated at the same time, because it is + - well documented that K and NO3 are common counter ions and co-regulated (Coskun et al., 2017). The supply of nitrate may contribute to NAD+ regeneration, which is limited under oxygen deficiency, from NADH by the action of nitrate reductase (Limami et al., 2014). Thus application of nitrate improves plant growth under oxygen deficiency conditions (Trought and Drew, 1981; Allègre et al., 2004).

1.3.3 Metabolic aspects of the interaction between waterlogging and potassium nutrition

As discussed above, under waterlogging conditions, due to oxygen deprivation, plants rely on glycolysis for the generation of ATP, while under K deprivation affects glycolysis via the well-known inhibition of pyruvate kinase. I thus expect that the combination of waterlogging and K limitation should be highly detrimental to respiratory metabolism and growth. But quite critically, despite the high co-occurrence probability of K deficiency and waterlogging in the field, their combined effects on plant physiology are not so well- known, and in particular, their interacting effect on carbon balance and metabolism has not been described.

40 1.4 Background information on two oil-producing crops, sunflower and oil palm, and their specific K requirements

1.4.1 Sunflower

Sunflower (Helianthus annuus L.) is an important oil-producing crop (FAO, 2010). It is grown worldwide in temperate and subtropical areas and major sunflower production regions are Ukraine, Russia, European Union and Argentina. Globally, sunflower cultivation represents about 27 million hectares and the total worldwide production of sunflower seed is around 50 million metric tons (MMt) (FAO, 2010). In Australia, sunflower seed production is about 34,000 tons and represents 30,000 hectares mainly in Queensland and New South Wales, and a small area in Victoria (Agri-Futures Australia, 2017).

Amongst oil-producing crops, sunflower is representative as an average K demanding species, with an apparent K use efficiency of about 38 g oil g−1 applied K in the field (compared with 32–53 g oil g−1 K in canola and about 35 g oil g−1 K in oil palm; Foster, 2003; Kaiser, Rosen, & Lamb, 2016). K is the most abundant element in sunflower regardless of growth stage (Figure 7). It can constitute up to 6% of dry weight in stem and petioles at six-leaves stage (Figure 7B). With every 1 t/ha seed yield, sunflower removes 30 kg K/ha including 8 kg K/ha by seed and 22 kg K/ha by stover (Serafin and Belfield, 2008). It means that about 53 kg K/ha are removed from soil considering the yield of sunflower is 1.76 t/ha in 2013 (FAO, 2011). Therefore, the application of K is necessary to achieve high yield. Indeed, it has been well documented that the application of K fertilizer increase sunflower yield (Lewis et al., 1991; Abbadi et al., 2008; Amanullah and Khan, 2011; Li et al., 2018). For example, application of 100 kg K/ha with phosphate application level of 45 kg P/ha increases sunflower yield by 36% compared to that of 50 kg K/ha (Amanullah and Khan, 2011).

(A) Three-row anthesis stage 3.5 roots stem and petioles bottom leaves 3 mid leaves upper leaves capitulum florets seeds 2.5

% DW 1.5

0.5

0 P Mg Ca S K

(B) Six-leaves stage 7

6 roots stem and petioles bottom leaves upper leaves 5

% DW 3

0 P Mg Ca S K

Figure 7. Average mineral content in Sunflower at six-leaves stage (A) and three- row anthesis stage (B). Data was obtained from Hocking and Steer (1983).

42 1.4.2 Oil palm

Figure 8. World map of major oil palm cultivated regions. From Eden Project (www.edenproject.com).

Oil palm (Elaeis guineensis Jacq.), belonging to the Family Arecaceae, is the most efficient oil crop in the world. It produces four to six times more vegetable oil per unit of land than other oil-producing crops, such as soybean, sunflower, and rapeseed (Table 1). Palm oil represents about 36% of the total world vegetable oil production (USDA, 2014). It is widely cultivated in tropical regions including Indonesia and Malaysia which account for about 85% of the world’s palm oil (Tiemann et al., 2018), as well as South American countries (Brazil, Peru, and Columbia) and Congo (Figure 8) where plantations are mostly found at low elevation and therefore susceptible to flooding under wet tropical climate.

Table 1. Oil production and yield of various oil-producing crops.

Area harvested Oil Production Yield Species Year (Mha) (Mt) (t/ha) Sunflower 2013 27.1 14.0 0.5 Palm 2013 18.7 54.6 2.9 Soybean 2013 117.8 54.0 0.5 Rapeseed 2013 43.9 30.1 0.7 Sunflower 2014 26.2 16.1 0.6 Palm 2014 19.5 57.6 3.0 Soybean 2014 124.4 57.8 0.5 Rapeseed 2014 43.9 31.6 0.7

Data was obtained from FAO. Mha, million hectare; Mt, million tons; t/ha, ton per hectare.(http://www.fao.org/economic/the-statistics-division-ess/chartroom-and factoids/chartroom/36-world-cropped-area-yield-and-production-of-oil-bearing- crops/en/)

Oil palm also produce high amounts of dry matters (leaves and trunk) and requires large amounts of nutrients (Tiemann et al., 2018). Even fertile soils (such as those found in Sumatra) cannot meet the growth and production requirement of oil palm over extended periods. Therefore, application of fertilizer is central for achieving high productivity (Tiemann et al., 2018). Through surveying over 300 farmers in Indonesia, Woittiez and coworkers (2018) found that average applications of N and P cover more than the demand, while applications of K are only about 45%−70% of the required amount for optimal growth. In addition, excessive application of N may exacerbate K deficiency (Broschat, 2009) and more generally, optimal K fertilization depends on N and P provision (Hartley, 1988).

K requirements of oil palm in the field have been studied for a long time, and a strong response of fruit bunch production to K has been observed. In fact, agronical trials in the 70s have shown that an increase in tissue K content by 0.5% can lead to an increase in fresh fruit bunches of nearly 70% (reviewed in Hartley 1988). On average, K fertilization in oil palm represents about 2.5 kg KCl tree-1 y-1 and this can be adjusted depending on soil conditions, typically Ca and Mg soil content (up to 5 kg KCl tree-1 y-1 in some agronomical trials). Nevertheless, at least three things (reviewed in Hartley, 1988;

44 Corley and Tinker, 2016) complicate K nutrition in oil palm: First, there is an interaction between K and Mg whereby strong addition of KCl fertilizers is detrimental to Mg absorption (and this effect is even more pronounced when K2SO4 is used instead of KCl). Second, optimal K fertilization depends on P, with a much better response to K when P is high. Third, excess Cl- (when K is applied as KCl) can be detrimental to fruit quality (with more fibers and less oil in mesocarp, and lower mesocarp-to-kernel ratios). Note that the interaction between Mg, Ca, and K is discussed further in the Appendix Chapter submitted to Plant Cell and Environment – under revision (role of putrescine during K deficiency).

Recently, K requirements have been documented further thanks to the international project Isopalm (2011−2015) which involved both oil palm growers (Socfindo, Indonesia) and research labs (CIRAD, France) (Figures 9 to 12). K has been found to represent on average 0.8% of dry matter of oil palm tree, which is even more than N (0.5%) (Figure 9). K supply had an obvious effect on oil palm yield. In fact, fresh fruit biomass (fresh fruit bunches) increased with supplied KCl (Figure 10): for example, fruit biomass production at 0.63 t KCl ha-1 y-1 of KCl application was about two times higher than that of the control (where no K has been added to background soil K content, i.e. 0.2 meq exchangeable K per 100 g soil) (Figure 10).

45 P 6.7-

3.4- Ca

N Fresh fruit amount amount (t/h) fruit Fresh 0.0- K 0 0.63 1.26 KCl application (t/h/y) 0 0.5 1 % DM

Figure 9. Average mineral content in Figure 10. Effect on K fertilization on oil palm. Data released from the yield of oil palm. Potassium chloride was international project Isopalm. applied annually. The fresh fruit biomass (http://agritrop.cirad.fr/564411/) was determined at the 10th year. Data released from the international project Isopalm. (http://agritrop.cirad.fr/564411/)

A B K (% DM) leaflets av. K control rachis + K ++ K glucose (% petiole DM)

meristem

trunk top

trunk middle

trunk base 0.00 2.00 4.00 6.00 roots

0 5

Figure 11. Distribution of potassium and glucose (A) in different tissues, and glucose-to-starch ratio (B) in tree trunk. In A, numbers are expressed on a dry mass basis (%DM). In B, the effect of potassium application is shown, with KCl applied at 0 (K control), 0.63 (+K), 1.26 (++K) t/h/y. Data released from the international project Isopalm. (http://agritrop.cirad.fr/564411/)

Figure 12. Relation between K content and glucose content in different vegetative oil palm tissues. Data released from the international project Isopalm. (http://agritrop.cirad.fr/564411/)

In oil palm, K is not distributed evenly amongst different tissues. In vegetative organs, maximum K content is observed in trunk base while minimal K is found in leaflets and root (Figure 11). It seems that K preferentially localizes in heterotrophic organs such as trunk, petiole and rachis compared to photosynthetic tissue leaflets. Interestingly, maximal glucose content is also found in truck while a minimal content is observed in leaflets (Figure 11). Therefore, there seems to be a relationship between K and glucose allocation (Figure 12). In addition, with the increase of K availability, the ratio of glucose- to-starch also increases in all of the three parts of palm trunk, especially at trunk base (Figure 11). It is possible that enhanced K availability promotes starch remobilization.

Importantly, elemental K content is very high in fruit bunches. The highest K content can be found in stalks (5−8% of dry matter), which account for up to 26% of total bunch K. Therefore, it has been estimated that an enormous amount of K (up to 250 kg K ha-1 y-1) is abstracted during bunch harvesting, and therefore K efflux (harvesting) can exceed K influx (fertilization) in oil palm agrosystems. This also explains why K fertilization is so crucial to oil palm cultivation, since it not only allows proper development but also avoids progressive K depletion in soils and trees.

Taken as a whole, K distribution within oil palm seems to be complicated and coupled to sugars (and thus likely to be influenced by production and the sink strength

47 exerted by bunches). The preliminary results of Isopalm suggest that the positive effect of K fertilization on oil palm growth and yield could be partly related to a stimulation of sugar metabolism. However, the specific mechanisms by which K content changes photosynthesis and metabolic carbon (C) partitioning have never been examined. The relationship between K concentration and C allocation is not very well known either.

1.5 Research questions and thesis outline

1.5.1 Key research question addressed here

As explained above, K deficiency and waterlogging are two severe global problems for agriculture. Waterlogging and K deficiency affect a number of biological and chemical processes, which can affect crop growth in both the short and long term. Critically, K deficiency and waterlogging are two conditions that can co-occur in many agrosystems worldwide.

Nevertheless, metabolic consequences of the combination of K deficiency and waterlogging have not been documented yet. Because K deficiency mostly impacts on glycolysis while waterlogging-induced oxygen shortage stimulates substrate-level ATP generation, a drastic additive effect might be anticipated, at least in roots. By contrast, the overall effect of the combination of K deficiency and waterlogging on primary carbon and nitrogen metabolism in leaves is more difficult to anticipate, because leaves of waterlogged plants do not exhibit typical hypoxic metabolism (Rocha et al., 2010b) and may experience nitrogen shortage due to impeded plant nitrate circulation.

Filling this gap of knowledge is at the heart of this PhD work. I took advantage of two widely cultivated species, which can experience the combination of K deficiency and waterlogging in the field (for example, in Russia and USA for sunflower, and in all South- Eastern Asia and South-America for oil palm). The use of oil palm as a biological model is important here for two reasons: First, oil palm is peculiar in that its K requirements are exceptionally high, so that K fertilization represents an expenditure far larger than N or P fertilization. Therefore, K nutrition in oil palm is currently at the heart of many research teams across the world, and a better knowledge of how to adjust K fertilization practices, including under stress conditions such as waterlogging, is desirable. Second, metabolic aspects of oil palm biology are still in their infancy. Oil palm genome has been sequenced recently (Singh et al., 2013), and thus this species offers genomic resources that can be

48 used. However, it is likely that metabolic consequences of conditions such as K deficiency and waterlogging in oil palm are very different from that in herbaceous species such as Arabidopsis, due to effects on allocation, remobilization of reserves in the trunk, etc.

1.5.2 Specific objectives and impact

• Specific objectives

In this thesis, questions are formulated in terms of metabolism, and my specific objectives are:

-What are the effects of waterlogging and K deficiency on respiration and carbon balance?

-What are the metabolic pathways affected by waterlogging and K deficiency?

-Are there impacts on other pathways such as N metabolism by waterlogging or/and K deficiency?

• Structure of the thesis

The first aim of this thesis was to investigate the effects of K and waterlogging on metabolic pathways in oil palm. However, it took more than nine months before I could harvest the first samples. During the time of oil palm cultivation (from germinated seeds), I chose to study K and waterlogging effect on another oil-producing crop, sunflower. First, sunflower requires a relatively short time to grow, thereby allowing me to practice techniques and workflows required to investigate physiology and metabolism such as the gas exchange system (LI-COR), metabolomics (GC-MS), etc. Second, sunflower is an annual species with a conservatively high stomatal conductance and therefore the effect of K availability and waterlogging on both photosynthesis and carbon allocation likely differs from that in oil palm (in particular in young oil palm plants that have a relatively low stomatal conductance). Therefore, this thesis deals with two plant species. It contains five chapters including one literature review, materials and methods, two original chapters based on experiments, and one general discussion. In the general discussion chapter, the important results from analyses that exploit recent δ15N measurements were also discussed.

Here, all the experiments in this thesis were carried out in glasshouses. Another thesis by a colleague (PhD candidate of the University of Paris-Sud, the work of which is co-

49 supervised by the same lab leader, Guillaume Tcherkez) is in progress on the effect of K availability (no waterlogging) on fruit and leaf metabolism of oil palm trees in the field.

• Potential impact. The aims of this study were to clarify the metabolic effects of K deficiency and waterlogging so as to better understand how they interact with carbon primary metabolism. This question is of prime interest in the field, since K fertilization strategies have to be found to optimize the carbon budget, plant production and yield. In practice, the identification of a metabolic biomarker of K deficiency can be helpful to facilitate plant K-status monitoring in the field. In fact, it is worth noting that answering all the questions above will be useful to provide a better knowledge of metabolic effects in leaves, and this can be instrumental to define a K bio-index (Hochmuth, 1994; Zörb et al., 2014). Because roots sampling is far more difficult than leaf sampling (in particular in oil palm), having a leaf metabolic signature to detect and anticipate, e.g., waterlogging- induced K deficiency would certainly be helpful. This topic of implementable bio-index is of particular importance for oil palm, and this is at the heart of research works conducted by other research teams (for example, Dr. Lamade in CIRAD, France). It is often assumed that K elemental content in leaves can be used as an indicator of the K nutritional status (typically, fertilization may be required if the K content falls below 0.8%). However, since K is involved in multiple processes and interacts with many pathways, it is unlikely that the elemental K content alone represents a reliable bio- indicator. In that context, the definition of metabolic markers would be extremely useful. Of course, the present thesis is far from the implementation of a bio-index since it addresses fundamentals of metabolic responses to K and waterlogging, but it represents the first step in a better knowledge of mineral nutrition in oil palm.

50 References

Abbadi, J., Gerendás, J., and Sattelmacher, B. (2008). Effects of potassium supply on growth and yield of safflower as compared to sunflower. Journal of Plant Nutrition and Soil Science 171, 272-280. Ævarsson, A., Chuang, J.L., Max Wynn, R., Turley, S., Chuang, D.T., and Hol, W.G.J. (2000). Crystal structure of human branched-chain α-ketoacid dehydrogenase and the molecular basis of multienzyme complex deficiency in maple syrup urine disease. Structure 8, 277-291. Agri-Futures Australia. (2017) https://www.agrifutures.com.au/farm-diversity/sunflowers/ Ahmad, I., and Maathuis, F.J. (2014). Cellular and tissue distribution of potassium: physiological relevance, mechanisms and regulation. Journal of plant physiology 171, 708-714. Ajewole, E., Santamaria-Kisiel, L., Pajak, A., Jaskolski, M., and Marsolais, F. (2018). Structural basis of potassium activation in plant asparaginases. The FEBS Journal 285, 1528-1539. Alcántara, E., de la Guardia, M.D., and Romera, F.J. (1991). Plasmalemma Redox Activity and H+ Extrusion in Roots of Fe-Deficient Cucumber Plants. Plant Physiology 96, 1034-1037. Allègre, A., Silvestre, J., Morard, P., Kallerhoff, J., and Pinelli, E. (2004). Nitrate reductase regulation in tomato roots by exogenous nitrate: a possible role in tolerance to long-term root anoxia. Journal of Experimental Botany 55, 2625- 2634. Amanullah, and Khan, M.W. (2011). Interactive Effect of Potassium and Phosphorus on Grain Quality and Profitability of Sunflower in Northwest Pakistan. Pedosphere 21, 532-538. Amberger, A. (1975) Protein biosynthesis and effect of plant nutrients on the process of protein formation. In Fertilizer use and protein production, Proceedings of the 11th colloquium of the International Potash Institute, Ronne-Bornholm, Denmark. pp. 75-89 Anschütz, U., Becker, D., and Shabala, S. (2014). Going beyond nutrition: Regulation of potassium homoeostasis as a common denominator of plant adaptive responses to environment. Journal of Plant Physiology 171, 670-687. António, C., Päpke, C., Rocha, M., Diab, H., Limami, A.M., Obata, T., Fernie, A.R., and van Dongen, J.T. (2016). Regulation of Primary Metabolism in Response to Low Oxygen Availability as Revealed by Carbon and Nitrogen Isotope Redistribution. Plant Physiology 170, 43-56. Armbruster, U., Carrillo, L.R., Venema, K., Pavlovic, L., Schmidtmann, E., Kornfeld, A., Jahns, P., Berry, J.A., Kramer, D.M., and Jonikas, M.C. (2014). Ion antiport accelerates photosynthetic acclimation in fluctuating light environments. Nature Communications 5, 5439. Armengaud, P., Breitling, R., and Amtmann, A. (2004). The Potassium-Dependent Transcriptome of Arabidopsis Reveals a Prominent Role of Jasmonic Acid in Nutrient Signaling. Plant Physiology 136, 2556-2576. Armengaud, P., Sulpice, R., Miller, A.J., Stitt, M., Amtmann, A., and Gibon, Y. (2009). Multilevel analysis of primary metabolism provides new insights into the role of potassium nutrition for glycolysis and nitrogen assimilation in Arabidopsis roots. Plant physiology 150, 772-785. Armstrong, W., Brändle, R., and Jackson, M.B. (1994). Mechanisms of flood tolerance in plants. Acta Botanica Neerlandica 43, 307-358.

51 Ashraf, M.A., Ahmad, M.S.A., Ashraf, M., Al-Qurainy, F., and Ashraf, M.Y. (2011). Alleviation of waterlogging stress in upland cotton (Gossypium hirsutum L.) by exogenous application of potassium in soil and as a foliar spray. Crop and Pasture Science 62, 25-38. Azcón-Bieto, J., and Osmond, C.B. (1983). Relationship between Photosynthesis and Respiration. The Effect of Carbohydrate Status on the Rate of CO2 Production by Respiration in Darkened and Illuminated Wheat Leaves 71, 574-581. Bailey-Serres, J., and Voesenek, L.A. (2008). Flooding stress: acclimations and genetic diversity. Annual review of plant biology 59, 313-339. Bailey-Serres, J., and Colmer, T.D. (2014). Plant tolerance of flooding stress – recent advances. Plant, Cell & Environment 37, 2211-2215. Ball, J., Woodrow, I., Berry, J. (1987). A model predicting stomatal conductance and its contribution to the control of photosynthesis under different environmental conditions. Progress in Photosynthesis Research IV, 221-224.

Balotf, S., Niazi, A., Kavoosi, G., Ramezani, A. (2012) Differential expression of nitrate reductase in response to potassium and sodium nitrate: real-time PCR analysis. Aust J Crop Sci. 6, 130-134. Barragán, V., Leidi, E.O., Andrés, Z., Rubio, L., De Luca, A., Fernández, J.A., Cubero, B., and Pardo, J.M. (2012). Ion Exchangers NHX1 and NHX2 Mediate Active Potassium Uptake into Vacuoles to Regulate Cell Turgor and Stomatal Function in Arabidopsis. The Plant Cell 24, 1127-1142. Bassil, E., Tajima, H., Liang, Y.C., Ohto, M.A., Ushijima, K., Nakano, R., Esumi, T., Coku, A., Belmonte, M., and Blumwald, E. (2011). The Arabidopsis Na+/H+ Antiporters NHX1 and NHX2 Control Vacuolar pH and K+ Homeostasis to Regulate Growth, Flower Development, and Reproduction. The Plant Cell 23, 3482-3497. Battie-Laclau, P., Laclau, J.P., Beri, C., Mietton, L., Muniz, M.R.A., Arenque, B.C., De Cassia Piccolo, M., Jordan-Meille, L., Bouillet, J.P., and Nouvellon, Y. (2014). Photosynthetic and anatomical responses of Eucalyptus grandis leaves to potassium and sodium supply in a field experiment. Plant, Cell & Environment 37, 70-81. Becker, D., Geiger, D., Dunkel, M., Roller, A., Bertl, A., Latz, A., Carpaneto, A., Dietrich, P., Roelfsema, M.R., Voelker, C., Schmidt, D., Mueller-Roeber, B., Czempinski, K., and Hedrich, R. (2004). AtTPK4, an Arabidopsis tandem-pore K+ channel, poised to control the pollen membrane voltage in a pH- and Ca2+- dependent manner. Proc Natl Acad Sci U S A 101, 15621-15626. Bednarz, C.W., and Oosterhuis, D.M. (1999). Physiological changes associated with potassium deficiency in cotton. Journal of Plant Nutrition 22, 303-313. Bejger, M., Imiolczyk, B., Clavel, D., Gilski, M., Pajak, A., Marsolais, F., and Jaskolski, M. (2014). Na+/K+ exchange switches the catalytic apparatus of potassium-dependent plant l-asparaginase. Acta Crystallographica Section D 70, 1854-1872. Besford, R.T., and Maw, G.A. (1976). Effect of Potassium Nutrition on Some Enzymes of the Tomato Plant. Annals of Botany 40, 461-471. Bhandal, I.S., and Malik, C.P. (1988). Potassium estimation, uptake, and its role in the physiology and metabolism of flowering plants, International Review of Cytology. vol. 110 (pg. 205-254).

52 Bihler, H., Eing, C., Hebeisen, S., Roller, A., Czempinski, K., and Bertl, A. (2005). TPK1 is a vacuolar ion channel different from the slow-vacuolar cation channel. Plant Physiol 139, 417-424. Blaha, G., Stelzl, U., Spahn, C.M., Agrawal, R.K., Frank, J., and Nierhaus, K.H. (2000). Preparation of functional ribosomal complexes and effect of buffer conditions on tRNA positions observed by cryoelectron microscopy. Methods in enzymology 317, 292-309. Bologa, K.L., Fernie, A.R., Leisse, A., Ehlers Loureiro, M., and Geigenberger, P. (2003). A Bypass of Sucrose Synthase Leads to Low Internal Oxygen and Impaired Metabolic Performance in Growing Potato Tubers. Plant Physiology 132, 2058-2072. Boyer, P.D., Lardy, H.A., and Phillips, P.H. (1942). The rôle of potassium in muscle phosphorylations. Journal of Biological Chemistry 146, 673-682. Broschat, T.K. (2009). Palm Nutrition and Fertilization 19, 690. Bruneau, L., Chapman, R., and Marsolais, F. (2006). Co-occurrence of both l- asparaginase subtypes in Arabidopsis: At3g16150 encodes a K+-dependent l- asparaginase. Planta 224, 668-679. Bush, L.P. (1969). Influence of Certain Cations on Activity of Succinyl CoA Synthetase From Tobacco. Plant Physiology 44, 347-350. Buwalda, F., Thomson, C.J., Steigner, W., Barrett-Lennard, E.G., Gibbs, J., and Greenway, H. (1988). Hypoxia Induces Membrane Depolarization and Potassium Loss from Wheat Roots but does not Increase their Permeability to Sorbitol. Journal of Experimental Botany 39, 1169-1183. Cakmak, I. (2005). The role of potassium in alleviating detrimental effects of abiotic stresses in plants. Journal of Plant Nutrition and Soil Science 168, 521-530. Cakmak, I., Hengeler, C., and Marschner, H. (1994a). Changes in phloem export of sucrose in leaves in response to phosphorus, potassium and magnesium deficiency in bean plants. Journal of Experimental Botany 45, 1251-1257. Cakmak, I., Hengeler, C., and Marschner, H. (1994b). Partitioning of shoot and root dry matter and carbohydrates in bean plants suffering from phosphorus, potassium and magnesium deficiency. Journal of Experimental Botany 45, 1245-1250. Cao, X., Hu, L., Chen, X., Zhang, R., Cheng, D., Li, H., Xu, Z., Li, L., Zhou, Y., Liu, A., Song, J., Liu, C., Liu, J., Zhao, Z., Chen, M., and Ma, Y. (2019). Genome- wide analysis and identification of the low potassium stress responsive gene SiMYB3 in foxtail millet (Setariaitalica L.). BMC genomics 20, 136. Carroll, A.D., Fox, G.G., Laurie, S., Phillips, R., Ratcliffe, R.G., and Stewart, G.R. (1994). Ammonium Assimilation and the Role of [gamma]-Aminobutyric Acid in pH Homeostasis in Carrot Cell Suspensions. Plant Physiology 106, 513-520. Chow, W. S., Melis, A., Anderson, J. M. (1990). Adjustments in photosysterm stoichiometry in chloroplasts improve the quantum efficiency of photosynthesis. Proc. Natl. Acad. Sci. U.S.A. 87, 7502-7507. Ciszak, E.M., Korotchkina, L.G., Dominiak, P.M., Sidhu, S., and Patel, M.S. (2003). Structural Basis for Flip-Flop Action of Thiamin Pyrophosphate-dependent Enzymes Revealed by Human Pyruvate Dehydrogenase. Journal of Biological Chemistry 278, 21240-21246. Corley, R. and Tinker, P. (2016). The oil palm, 5th edn. Chichester, UK: Wiley- Blackwell. Coskun, D., Britto, D.T., and Kronzucker, H.J. (2017). The nitrogen-potassium intersection: membranes, metabolism, and mechanism. Plant Cell Environ 40, 2029-2041.

53 Crawford, L.A., Bown, A.W., Breitkreuz, K.E., and Guinel, F.C. (1994). The Synthesis of [gamma]-Aminobutyric Acid in Response to Treatments Reducing Cytosolic pH. Plant Physiology 104, 865-871. Cuesta-Seijo, J.A., Nielsen, M.M., Marri, L., Tanaka, H., Beeren, S.R., and Palcic, M.M. (2013). Structure of starch synthase I from barley: insight into regulatory mechanisms of starch synthase activity. Acta crystallographica. Section D, Biological crystallography 69, 1013-1025. Curtis, T.Y., Bo, V., Tucker, A., and Halford, N.G. (2018). Construction of a network describing asparagine metabolism in plants and its application to the identification of genes affecting asparagine metabolism in wheat under drought and nutritional stress. Food Energy Secur 7, e00126-e00126. Deeken, R., Geiger, D., Fromm, J., Koroleva, O., Ache, P., Langenfeld-Heyser, R., Sauer, N., May, S.T., and Hedrich, R. (2002). Loss of the AKT2/3 potassium channel affects sugar loading into the phloem of Arabidopsis. Planta 216, 334- 344. Diab, H., and Limami, A. (2016). Reconfiguration of N metabolism upon hypoxia stress and recovery: Roles of alanine aminotransferase (AlaAT) and glutamate dehydrogenase (GDH). Plants, 5, 25. Di Cera, E. (2006). A structural perspective on enzymes activated by monovalent cations. The Journal of biological chemistry 281, 1305-1308. Dreyer, I., Gomez-Porras, J.L., and Riedelsberger, J. (2017). The potassium battery: a mobile energy source for transport processes in plant vascular tissues. New Phytologist 216, 1049-1053. Dunkel, M., Latz, A., Schumacher, K., Muller, T., Becker, D., and Hedrich, R. (2008). Targeting of vacuolar membrane localized members of the TPK channel family. Mol Plant 1, 938-949. Dwivedi, S.K., Kumar, S., Bhakta, N., Singh, S.K., Rao, K.K., Mishra, J.S., and Singh, A.K. (2017). Improvement of submergence tolerance in rice through efficient application of potassium under submergence-prone rainfed ecology of Indo-Gangetic Plain. Functional Plant Biology 44, 907-916. Eden Project. Oil Palm exhibit in the Rainforest Biome - Eden Project, Cornwall. Retrieved on 05/02/2019. Retrieved from https://www.edenproject.com/visit/whats- here/rainforest-biome/oil-palm-exhibit Epstein, E., Rains, D.W., and Elzam, O.E. (1963). Resolution of dual mechanisms of potassium absorption by barley roots. Proc Natl Acad Sci USA. 49, 684-692. Evans, H.J. (1963). Effect of Potassium & Other Univalent Cations on Activity of Pyruvate Kinase in Pisum sativum. Plant physiology 38, 397-402. Evans, H.J., and Sorger, G.J. (1966). Role of Mineral Elements with Emphasis on the Univalent Cations. Annual Review of Plant Physiology 17, 47-76. FAO. (2010). Agribusiness Handbook: Sunflower Crude and Refined Oils. http://www.fao.org/docrep/012/al375e/al375e.pdf FAO Soil Portal. (2020). Harmonized World Soil Database v 1.2. http://www.fao.org/soils-portal/soil-survey/soil-maps and%20databases/harmonized-world-soil-database-v12/en. Accessed on 10 July 2019. FAOSTAT. (2011). Production of sunflower seed throughout the world. Retrieved from: http://faostat3.fao.org/home/index. Foster, M. (2003), GM Canola: What are its Economics under Australian Conditions? Australian Grains Industry, ABARE, Canberra.

54 Freeman, G.G. (1967). Studies on potassium nutrition of plants. II. Some effects of potassium deficiency on the organic acids of leaves. Journal of the Science of Food and Agriculture 18, 569-576. Fulton, D.C., Stettler, M., Mettler, T., Vaughan, C.K., Li, J., Francisco, P., Gil, M., Reinhold, H., Eicke, S., Messerli, G., Dorken, G., Halliday, K., Smith, A.M., Smith, S.M., and Zeeman, S.C. (2008). β-AMYLASE4, a Noncatalytic Protein Required for Starch Breakdown, Acts Upstream of Three Active β-Amylases in Arabidopsis Chloroplasts. The Plant Cell 20, 1040-1058. Gajdanowicz, P., Michard, E., Sandmann, M., Rocha, M., Corrêa, L.G.G., Ramírez- Aguilar, S.J., Gomez-Porras, J.L., González, W., Thibaud, J.B., van Dongen, J.T., and Dreyer, I. (2011). Potassium K+ gradients serve as a mobile energy source in plant vascular tissues. Proceedings of the National Academy of Sciences 108, 864-869. Gautam, P., Lal, B., Tripathi, R., Shahid, M., Baig, M.J., Maharana, S., Puree, C., and Nayak, A.K. (2016). Beneficial effects of potassium application in improving submergence tolerance of rice (Oryza sativa L.). Environmental and Experimental Botany 128, 18-30. Gaymard, F., Pilot, G., Lacombe, B., Bouchez, D., Bruneau, D., Boucherez, J., Michaux-Ferrière, N., Thibaud, J.B., and Sentenac, H. (1998). Identification and Disruption of a Plant Shaker-like Outward Channel Involved in K+ Release into the Xylem Sap. Cell 94, 647-655. Germain, V., Ricard, B., Raymond, P., and Saglio, P.H. (1997). The Role of Sugars, Hexokinase, and Sucrose Synthase in the Determination of Hypoxically Induced Tolerance to Anoxia in Tomato Roots. Plant Physiology 114, 167-175. Gierth, M., and Maser, P. (2007). Potassium transporters in plants-involvement in K+ acquisition, redistribution and homeostasis. FEBS letters 581, 2348-2356. Gierth, M., Mäser, P., and Schroeder, J.I. (2005). The Potassium Transporter AtHAK5 Functions in K+ Deprivation-Induced High-Affinity K+ Uptake and AKT1 K+ Channel Contribution to K+ Uptake Kinetics in Arabidopsis Roots. Plant Physiology 137, 1105-1114. Gill, M.B., Zeng, F., Shabala, L., Bohm, J., Zhang, G., Zhou, M., and Shabala, S. (2018). The ability to regulate voltage-gated K+-permeable channels in the mature root epidermis is essential for waterlogging tolerance in barley. J Exp Bot 69, 667- 680. Glass, A.D.M., and Kotur, Z. (2013). A reevaluation of the role of Arabidopsis NRT1.1 in high-affinity nitrate transport. Plant physiology 163, 1103-1106. Gohara, D.W., and Di Cera, E. (2016). Molecular Mechanisms of Enzyme Activation by Monovalent Cations. Journal of Biological Chemistry 291, 20840-20848. Gojon, A., Krouk, G., Perrine-Walker, F., and Laugier, E. (2011). Nitrate transceptor(s) in plants. J Exp Bot 62, 2299-2308. Graf, A., and Smith, A.M. (2011). Starch and the clock: the dark side of plant productivity. Trends in plant science 16, 169-175. Grant, M., and Bevan, M.W. (1994). Asparaginase gene expression is regulated in a complex spatial and temporal pattern in nitrogen-sink tissues. The Plant Journal 5, 695-704. Hafsi, C., Debez, A., and Abdelly, C. (2014). Potassium deficiency in plants: effects and signaling cascades. Acta Physiol. Plant. 36, 1055-1070. Han, M., Wu, W., Wu, W.H., and Wang, Y. (2016). Potassium Transporter KUP7 Is Involved in K+ Acquisition and Translocation in Arabidopsis Root under K+- Limited Conditions. Mol Plant 9, 437-446.

55 Hartley, C. (1988). The Oil Palm. 3rd Edition, Editorial Longman, London, 958. Hartt, C.E. (1969). Effect of Potassium Deficiency Upon Translocation of 14C in Attached Blades and Entire Plants of Sugarcane. Plant Physiology 44, 1461-1469. Hind, G., Nakatani, H.Y., Izawa, S. (1974) Light-dependent redistribution of ions in suspensions of chloroplast thylakoid membranes. Proc Natl Acad Sci USA 71, 1484-1488. Hirabayashi, Y., Mahendran, R., Koirala, S., Konoshima, L., Yamazaki, D., Watanabe, S., Kim, H., and Kanae, S. (2013). Global flood risk under climate change. Nature Climate Change 3, 816. Hirsch, R.E., Lewis, B.D., Spalding, E.P., and Sussman, M.R. (1998). A Role for the AKT1 Potassium Channel in Plant Nutrition. Science 280, 918-921. Hochmuth, G.J. (1994). Efficiency Ranges for Nitrate-Nitrogen and Potassium for Vegetable Petiole Sap Quick Tests. HortTechnology 4, 218-222. Hocking, P.J., and Steer, B.T. (1983). Uptake and partitioning of selected mineral elements in sunflower (Helianthus annuus L.) during growth. Field Crops Research 6, 93-107. Hosy, E., Vavasseur, A., Mouline, K., Dreyer, I., Gaymard, F., Porée, F., Boucherez, J., Lebaudy, A., Bouchez, D., Véry, A.A., Simonneau, T., Thibaud, J.B., and Sentenac, H. (2003). The Arabidopsis outward K+ channel GORK is involved in regulation of stomatal movements and plant transpiration. Proceedings of the National Academy of Sciences 100, 5549-5554. Hu, W., Coomer, T.D., Loka, D.A., Oosterhuis, D.M., and Zhou, Z. (2017). Potassium deficiency affects the carbon-nitrogen balance in cotton leaves. Plant physiology and biochemistry : PPB 115, 408-417. Hu, W., Yang, J., Meng, Y., Wang, Y., Chen, B., Zhao, W., Oosterhuis, D.M., and Zhou, Z. (2015). Potassium application affects carbohydrate metabolism in the leaf subtending the cotton (Gossypium hirsutum L.) boll and its relationship with boll biomass. Field Crops Research 179, 120-131. Huber, S.C., and Akazawa, T. (1986). A Novel Sucrose Synthase Pathway for Sucrose Degradation in Cultured Sycamore Cells. Plant Physiology 81, 1008-1013. Hubert, E., Villanueva, J., Gonzalez, A.M., and Marcus, F. (1970). Univalent cation activation of fructose 1,6-diphosphatase. Archives of Biochemistry and Biophysics 138, 590-597. Hussain, S.S., Ali, M., Ahmad, M., and Siddique, K.H.M. (2011). Polyamines: Natural and engineered abiotic and biotic stress tolerance in plants. Biotechnology Advances 29, 300-311. Isopalm Project. Annual report. (2011) Characterization of mineral fluxes N-K for oil palm under Sumatran conditions. http://agritrop.cirad.fr/564411/. Accessed on 10 July 2019. Ivashikina, N., Becker, D., Ache, P., Meyerhoff, O., Felle, H.H., and Hedrich, R. (2001). K+ channel profile and electrical properties of Arabidopsis root hairs. FEBS letters 508, 463-469. Jeanguenin, L., Lebaudy, A., Xicluna, J., Alcon, C., Hosy, E., Duby, G. (2008). Heteromerization of Arabidopsis Kv channel α-subunits: data and prospects. Plant Signal. Behav. 3, 622-625. Jin, S.H., Huang, J.Q., Li, X.Q., Zheng, B.S., Wu, J.S., Wang, Z.J., Liu, G.H., and Chen, M. (2011). Effects of potassium supply on limitations of photosynthesis by mesophyll diffusion conductance in Carya cathayensis. Tree Physiology 31, 1142- 1151.

56 Johnston, A.E. (2005). Understanding potassium and its use in agriculture. Brussels: EFMA. Jones, L.H. (1961). Some effects of potassium deficiency on the metabolism of the tomato plant. Canadian Journal of Botany 39, 593-606. Jones, L.H. (1966). Carbon-14 studies of intermediary metabolism in potassium- deficient tomato plants. Canadian Journal of Botany 44, 297-307. Jurica, M.S., Mesecar, A., Heath, P.J., Shi, W., Nowak, T., and Stoddard, B.L. (1998). The allosteric regulation of pyruvate kinase by fructose-1,6-bisphosphate. Structure 6, 195-210. Jwakyung, S., Yeonkyu, S., Yejin, L., Seongsoo, K., Sangkeun, H., B., K.H., and Taek-Keun, O. (2015). Compositional changes of selected amino acids, organic acids, and soluble sugars in the xylem sap of N, P, or K-deficient tomato plants. Journal of Plant Nutrition and Soil Science 178, 792-797. Kachmar, J.F., and Boyer, P.D. (1953). Kinetic analysis of enzyme reactions: II. the potassium activation and calcium inhibition of pyruvic phosphoferase. Journal of Biological Chemistry 200, 669-682. Kafkafi, U. (1990). The functions of plant K in overcoming environmental stress situations. pp. 81-93. In Proceedings 22nd Colloquium, International Potash Institute, Bern, Switzerland. Kaiser, D.E., Rosen, C.J., and Lamb, J.A. (2016). Potassium for crop production. Nutrient management, university of Minnesota extension, FO-6794-D. Kanai, S., K. Ohkura, J.J. Adu-Gyamfi, P.K. Mohapatra, N.T. Nguyen, H. Saneoka, and K. Fujita. (2007). Depression of sink activity precedes the inhibition of biomass production in tomato plants subjected to potassium deficiency stress. J. Experimental Bot. 58, 2917-2928. Kayne, F.J. (1971) Thallium (I) activation of pyruvate kinase. Arch. Biochem. Biophys. 143, 232-239. Komoto, J., Yamada, T., Takata, Y., Markham, G.D., and Takusagawa, F. (2004). Crystal Structure of the S-Adenosylmethionine Synthetase Ternary Complex: A Novel Catalytic Mechanism of S-Adenosylmethionine Synthesis from ATP and Met. Biochemistry 43, 1821-1831. Kreuzwieser, J., and Gessler, A. (2010). Global climate change and tree nutrition: influence of water availability. Tree Physiology 30, 1221-1234. Lai, J.C.K., and Sheu, K.F.R. (1985). Relationship Between Activation State of Pyruvate Dehydrogenase Complex and Rate of Pyruvate Oxidation in Isolated Cerebro-Cortical Mitochondria: Effects of Potassium Ions and Adenine Nucleotides. Journal of Neurochemistry 45, 1861-1868. Lam, H.M., Coschigano, K., Schultz, C., Melo-Oliveira, R., Tjaden, G., Oliveira, I., Ngai, N., Hsieh, M.H., and Coruzzi, G. (1995). Use of Arabidopsis mutants and genes to study amide amino acid biosynthesis. The Plant Cell 7, 887-898. Lamade, E., Ollivier, J., Rozier-Abouab, Th., Gérardeaux, E. (2014). Occurrence of potassium location in oil palm tissues with reserve sugars: consequences for oil palm K status determination. IOPC conference, 17-19 June 2014, Bali Convention Center. Lardy, H.A., and Ziegler, J.A. (1945). The enzymatic synthesis of phosphopyruvate from pyruvate. Journal of Biological Chemistry 159, 343-351. Larsen, T.M., Benning, M.M., Wesenberg, G.E., Rayment, I., Reed, G.H. (1997) Ligand-induced domain movement in pyruvate kinasestructure of the enzyme from rabbit muscle with Mg2+, K+, and L-phospholactate at 2.7Å resolution. Arch. Biochem. Biophys.345, 199-206.

57 Lea, P.J., Sodek, L., Parry, M.A.J., Shewry, P.R., and Halford, N.G. (2007). Asparagine in plants. Annals of Applied Biology 150, 1-26. Lebaudy, A., Véry, A.A., and Sentenac, H. (2007). K+ channel activity in plants: Genes, regulations and functions. FEBS letters 581, 2357-2366. Leigh, R.A., and Wyn Jones, R.G. (1984). A hypothesis relating critical potassium concentrations for growth to the distribution and functions of this ion in the plant cell. New Phytologist 97, 1-13. Lewis, D.C., Potter, T.D., and Weckert, S.E. (1991). The effect of nitrogen, phosphorus and potassium fertilizer applications on the seed yield of sunflower (Helianthus annuus L.) grown on sandy soils and the prediction of phosphorus and potassium responses by soil tests. Fertilizer research 28, 185-190. Li, J., Francisco, P., Zhou, W., Edner, C., Steup, M., Ritte, G., Bond, C.S., and Smith, S.M. (2009). Catalytically-inactive β-amylase BAM4 required for starch breakdown in Arabidopsis leaves is a starch-binding-protein. Archives of Biochemistry and Biophysics 489, 92-98. Li, S.T., Duan, Y., Guo, T.W., Zhang, P.l., He, P., and Majumdar, K. (2018). Sunflower response to potassium fertilization and nutrient requirement estimation. Journal of Integrative Agriculture 17, 2802-2812. Limami, A.M., Diab, H., and Lothier, J. (2014). Nitrogen metabolism in plants under low oxygen stress. Planta 239, 531-541. Linkemer, G., Board, J. E., and Musgrave, M. E. (1998). Waterlogging effects on growth and yield components in late-planted soybean. Crop Sci. 38, 1576-1584. Lloyd, J.R., Kossmann, J., and Ritte, G. (2005). Leaf starch degradation comes out of the shadows. Trends in plant science 10, 130-137. Lone, A. A., Khan, M. H., Dar, Z. A., Wani, S. H. (2018). Breeding strategies for improving growth and yield under waterlogging conditions in maize: a review. Maydica 61, 11. Lu, Z., Pan, Y., Hu, W., Cong, R., Ren, T., Guo, S., and Lu, J. (2017). The photosynthetic and structural differences between leaves and siliques of Brassica napus exposed to potassium deficiency. BMC Plant Biology 17, 240. Lubin, M., and Ennis, H.L. (1964). On the role of intracellular potassium in protein synthesis. Biochimica et biophysica acta 80, 614-631. Lüttge, U., and Higinbotham, N. (1979). Transport in Plants, Springer Verlag New York Inc. Ma, T.L., Wu, W.H., and Wang, Y. (2012). Transcriptome analysis of rice root responses to potassium deficiency. BMC Plant Biology 12, 161. Maathuis, F.J.M., and Sanders, D. (1996). Mechanisms of potassium absorption by higher plant roots. Physiol Plant. 96, 158-168. MacRobbie, E. A. C. (1970). The active transport of ions in plant cells. Quart. Rev. Biophys. 3, 251-294. Mancuso, S., and Shabala, S. (2010). Preface. In: Mancuso S, Shabala S eds. Waterlogging signalling and tolerance in plants. Berlin Heidelberg, Springer- Verlag. Marschner, H. (1986). Mineral nutrition in higher plants. Wd Ltd. The Greystone Press, Antrim, Northern Ireland. Marschner, H. (1995). Mineral Nutrition of Higher Plants (Second Edition). (London: Academic Press). Marschner, H., E. A. Kirkby, and C. Engels. (1997). Imponance of cycling and recycling of mineral nutrients within plants for growth and development. Bot. Acta 4, 265-273.

58 Marschner, H. (2012). Marschner’s Mineral Nutrition of Higher Plants, 3rd Edn London: Academic Press. McCollum, R.E., Hageman, R.H., and Tyner, E.H. (1958). Influence of potassium on pyruvic kinase from plant tissue. Soil Science 86, 324-331. Mesecar, A.D., and Nowak, T. (1997a). Metal-Ion-Mediated Allosteric Triggering of Yeast Pyruvate Kinase. 1. A Multidimensional Kinetic Linked-Function Analysis. Biochemistry 36, 6792-6802. Mesecar, A.D., and Nowak, T. (1997b). Metal-Ion-Mediated Allosteric Triggering of Yeast Pyruvate Kinase 2. A Multidimensional Thermodynamic Linked-Function Analysis. Biochemistry 36, 6803-6813. Miller, G., and Evans, H.J. (1957). The Influence of Salts on Pyruvate Kinase from Tissues of Higher Plants. Plant physiology 32, 346-354. Monroe, J. D., Breault, J. S., Pope, L. E., Torres, C. E., Gebrejesus, T. B., Berndsen, C. E., & Storm, A. R. (2017). Arabidopsis β-Amylase2 Is a K+-Requiring, Catalytic Tetramer with Sigmoidal Kinetics. Plant physiology, 175(4), 1525- 1535. Monroe, J.D., and Storm, A.R. (2018). Review: The Arabidopsis β-amylase (BAM) gene family: Diversity of form and function. Plant Science 276, 163-170. Mudd, S.H., and Cantoni, G.L. (1958). Activation of methionine for transmethylation: Iii. The methionine-activating enzyme of bakers' yeast. Journal of Biological Chemistry 231, 481-492. Mustroph, A., Zanetti, M.E., Jang, C.J.H., Holtan, H.E., Repetti, P.P., Galbraith, D.W., Girke, T., and Bailey-Serres, J. (2009). Profiling translatomes of discrete cell populations resolves altered cellular priorities during hypoxia in Arabidopsis. Proceedings of the National Academy of Sciences 106, 18843-18848. Mustroph, A., Lee, S.C., Oosumi, T., Zanetti, M.E., Yang, H., Ma, K., Yaghoubi- Masihi, A., Fukao, T., and Bailey-Serres, J. (2010). Cross-kingdom comparison of transcriptomic adjustments to low-oxygen stress highlights conserved and plant-specific responses. Plant Physiol 152, 1484-1500. Nielsen, M.M., Ruzanski, C., Krucewicz, K., Striebeck, A., Cenci, U., Ball, S.G., Palcic, M.M., and Cuesta-Seijo, J.A. (2018). Crystal Structures of the Catalytic Domain of Arabidopsis thaliana Starch Synthase IV, of Granule Bound Starch Synthase From CLg1 and of Granule Bound Starch Synthase I of Cyanophora paradoxa Illustrate Substrate Recognition in Starch Synthases. Frontiers in Plant Science 9. Nitsos, R. E., and Evans, H. J. (1968).The effect of univalent cations on participate starch synthetase; Abstr. Am. Soc. Plant Physiol. Western Section. Program of Meetings, Utah State University, Logan, Utah, June 25-27. Nitsos, R.E., and Evans, H.J. (1969). Effects of univalent cations on the activity of particulate starch synthetase. Plant physiology 44, 1260-1266. O'Leary, B.M., Asao, S., Millar, A.H., and Atkin, Owen K. (2019). Core principles which explain variation in respiration across biological scales. New Phytologist 222, 670-686. Okamoto, S. (1966). Effect of mineral nutrition on metabolic change induced in crop plant roots (III). Soil Science and Plant Nutrition 12, 13-17. Okamoto, S. (1967). Effects of potassium nutrition on the glycolysis and the Krebs cycle in taro plants. Soil Science and Plant Nutrition 13, 143-150. Okamoto, S. (1968). The respiration in the roots of broad bean and barley under a moderate potassium deficiency. Soil Science and Plant Nutrition 14, 175-182.

59 Oliveira, H.C., and Sodek, L. (2013). Effect of oxygen deficiency on nitrogen assimilation and amino acid metabolism of soybean root segments. Amino Acids 44, 743-755. Oliveira, H.C., Freschi, L., and Sodek, L. (2013). Nitrogen metabolism and translocation in soybean plants subjected to root oxygen deficiency. Plant Physiology and Biochemistry 66, 141-149. Oosterhuis, D.M., Loka, D.A., and Raper, T.B. (2013). Potassium and stress alleviation: Physiological functions and management of cotton. Journal of Plant Nutrition and Soil Science 176, 331-343. Osaki, M., Shinano, T., and Tadano, T. (1993). Effect of nitrogen, phosphorus, or potassium deficiency on the accumulation of ribulose-1,5-bisphosphate carboxylase/oxygenase and chlorophyll in several field crops. Page, M.J., and Di Cera, E. (2006). Role of Na+ and K+ in enzyme function. Physiological reviews 86, 1049-1092. Pang, J.Y., Newman, I., Mendham, N., Zhou, M., and Shabala, S. (2006). Microelectrode ion and O2 fluxes measurements reveal differential sensitivity of barley root tissues to hypoxia. Plant, Cell & Environment 29, 1107-1121. Peoples, T.R., and Koch, D.W. (1979). Role of Potassium in Carbon Dioxide Assimilation in Medicago sativa L. Plant Physiology 63, 878-881. Peuke, A.D., Jeschke, W.D., and Hartung, W. (2002). Flows of elements, ions and abscisic acid in Ricinus communis and site of nitrate reduction under potassium limitation. Journal of Experimental Botany 53, 241-250. Pezeshki, S.R., DeLaune, R.D., and Anderson, P.H. (1999). Effect of flooding on elemental uptake and biomass allocation in seedlings of three bottomland tree species. Journal of Plant Nutrition 22, 1481-1494. Phukan, U. J., Mishra, S., Shukla, R. K. (2016). Waterlogging and submergence stress: affects and acclimation. Crit. Rev. Biotechnol. 36, 956-966. Pi, Z., Stevanato, P., Yv, L.H., Geng, G., Guo, X.L., Yang, Y., Peng, C.X., and Kong, X.S. (2014). Effects of potassium deficiency and replacement of potassium by sodium on sugar beet plants. Russian Journal of Plant Physiology 61, 224-230. Pilot, G., Pratelli, R., Gaymard, F., Meyer, Y., and Sentenac, H. (2003). Five-Group Distribution of the Shaker-like K+ Channel Family in Higher Plants. Journal of Molecular Evolution 56, 418-434. Planchet, E., Lothier, J., and Limami, A.M. (2017). Hypoxic Respiratory Metabolism in Plants: Reorchestration of Nitrogen and Carbon Metabolisms. In Plant Respiration: Metabolic Fluxes and Carbon Balance, G. Tcherkez and J. Ghashghaie, eds (Cham: Springer International Publishing), pp. 209-226. Prajapati, K.B., and Modi, H.A. (2012). Isolation and Characterization of Potassium Solubilizing Bacteria from Ceramic Industry Soil. CIB Tech Journal of Microbiology 1(2-3), 8-14. Pretty, K.M., and Stangel, P.J. (1985). Current and future use of world potassium. In: Munson RD, ed. Potassium in agriculture. Madison, Wisconsin, USA: American Society of Agronomy, 99-128. Pyo, Y.J., Gierth, M., Schroeder, J.I., and Cho, M.H. (2010). High-affinity K+ transport in Arabidopsis: AtHAK5 and AKT1 are vital for seedling establishment and postgermination growth under low-potassium conditions. Plant Physiol 153, 863-875. Reggiani, R., Nebuloni, M., Mattana, M., and Brambilla, I. (2000). Anaerobic accumulation of amino acids in rice roots: role of the glutamine synthetase/glutamate synthase cycle. Amino Acids 18, 207-217.

60 Rocha, M., Sodek, L., Licausi, F., Hameed, M.W., Dornelas, M.C., and van Dongen, J.T. (2010a). Analysis of alanine aminotransferase in various organs of soybean (Glycine max) and in dependence of different nitrogen fertilisers during hypoxic stress. Amino Acids 39, 1043-1053. Rocha, M., Licausi, F., Araújo, W.L., Nunes-Nesi, A., Sodek, L., Fernie, A.R., and van Dongen, J.T. (2010b). Glycolysis and the Tricarboxylic Acid Cycle Are Linked by Alanine Aminotransferase during Hypoxia Induced by Waterlogging of Lotus japonicus. Plant Physiology 152, 1501-1513. Römheld, V., and Kirkby, E.A. (2010). Research on potassium in agriculture: needs and prospects. Plant and Soil 335, 155-180. Rozov, A., Khusainov, I., El Omari, K., Duman, R., Mykhaylyk, V., Yusupov, M., Westhof, E., Wagner, A., and Yusupova, G. (2019). Importance of potassium ions for ribosome structure and function revealed by long-wavelength X-ray diffraction. Nature Communications 10, 2519. Ruan, L., Zhang, J., Xin, X., Zhang, C., Ma, D., Chen, L., and Zhao, B. (2015). Comparative analysis of potassium deficiency-responsive transcriptomes in low potassium susceptible and tolerant wheat (Triticum aestivum L.). Scientific reports 5, 10090. Rubio, F., Santa-María, G.E., Rodríguez-Navarro, A. (2000). Cloning of Arabidopsis and barley cDNAs encoding HAK potasium transporters in root and shoot cells. Physiologia Plantarum 109, 34-43. Rubio, F., Nieves-Cordones, M., Alemán, F., Martínez, V. (2008) Relative contribution of AtHAK5 and AtAKT1 to K+ uptake in the high-affinity range of concentrations. Physiol Plant 134, 598-608. Schachtman, D.P., and Schroeder, J.I. (1994). Structure and transport mechanism of a high-affinity potassium uptake transporter from higher plants. Nature 370, 655- 658. Schroeder, J.I., Ward, J.M., and Gassmann, W. (1994). Perspectives on the physiology and structure of inward-rectifying K+ channels in higher plants: biophysical implications for K+ uptake. Annual review of biophysics and biomolecular structure 23, 441-471. Schulz, A. (1994). Phloem transport and differential unloading in pea seedlings after source and sink manipulations. Planta, 192(2), 239-248. Serafin, L., and Belfield, S. (2008). Sunflower production guidelines for the northern grains region-Northern NSW and Southern QLD. NSW Department of Primary Industries. Sharma, T., Dreyer, I., and Riedelsberger, J. (2013). The role of K+ channels in uptake and redistribution of potassium in the model plant Arabidopsis thaliana. Frontiers in Plant Science 4. Shen, C., Wang, J., Shi, X., Kang, Y., Xie, C., Peng, L., Dong, C., Shen, Q., and Xu, Y. (2017). Transcriptome Analysis of Differentially Expressed Genes Induced by Low and High Potassium Levels Provides Insight into Fruit Sugar Metabolism of Pear. Frontiers in plant science 8, 938-938. Shimomura, Y., Kuntz, M.J., Suzuki, M., Ozawa, T., and Harris, R.A. (1988). Monovalent cations and inorganic phosphate alter branched-chain α-ketoacid dehydrogenase—Kinase activity and inhibitor sensitivity. Archives of Biochemistry and Biophysics 266, 210-218. Singh, R., Ong-Abdullah, M., Low, E.T.L., Manaf, M.A.A., Rosli, R., Nookiah, R., Ooi, L.C.L., Ooi, S.E., Chan, K.L., Halim, M.A., Azizi, N., Nagappan, J., Bacher, B., Lakey, N., Smith, S.W., He, D., Hogan, M., Budiman, M.A., Lee,

61 E.K., DeSalle, R., Kudrna, D., Goicoechea, J.L., Wing, R.A., Wilson, R.K., Fulton, R.S., Ordway, J.M., Martienssen, R.A., and Sambanthamurthi, R. (2013). Oil palm genome sequence reveals divergence of interfertile species in Old and New worlds. Nature 500, 335. Smethurst, C.F., Garnett, T., and Shabala, S. (2005). Nutritional and chlorophyll fluorescence responses of lucerne (Medicago sativa) to waterlogging and subsequent recovery. Plant and Soil 270, 31-45. Smil, V. (1999). Crop Residues: Agriculture's Largest Harvest: Crop residues incorporate more than half of the world's agricultural phytomass. BioScience 49, 299-308. Sodek, L., Lea, P.J., and Miflin, B.J. (1980). Distribution and Properties of a Potassium- dependent Asparaginase Isolated from Developing Seeds of Pisum sativum and Other Plants. Plant Physiology 65, 22-26. Spencer, D., and Possingham, J. V. (1960). The Effect of Nutrient Deficiencies on the Hill Reaction of Isolated Chloroplasts from Tomato. Australian J. Biol. Sci.13, 441-455. Stitt, M. (1998). Pyrophosphate as an Energy Donor in the Cytosol of Plant Cells: an Enigmatic Alternative to ATP. Botanica Acta 111, 167-175. Stitt, M., and Zeeman, S.C. (2012). Starch turnover: pathways, regulation and role in growth. Current Opinion in Plant Biology 15, 282-292. Streb, S., and Zeeman, S.C. (2012). Starch metabolism in Arabidopsis. The arabidopsis book 10, e0160. Suelter, C.H. (1970). Enzymes activated by monovalent cations. Science 168, 789-795. Sulpice, R., Pyl, E.T., Ishihara, H., Trenkamp, S., Steinfath, M., Witucka-Wall, H., Gibon, Y., Usadel, B., Poree, F., Piques, M.C., Von Korff, M., Steinhauser, M.C., Keurentjes, J.J., Guenther, M., Hoehne, M., Selbig, J., Fernie, A.R., Altmann, T., and Stitt, M. (2009). Starch as a major integrator in the regulation of plant growth. Proc Natl Acad Sci USA 106, 10348-10353. Szyroki, A., Ivashikina, N., Dietrich, P., Roelfsema, M.R.G., Ache, P., Reintanz, B., Deeken, R., Godde, M., Felle, H., Steinmeyer, R., Palme, K., and Hedrich, R. (2001). KAT1 is not essential for stomatal opening. Proceedings of the National Academy of Sciences 98, 2917-2921. Tiemann, T.T., Donough, C.R., Lim, Y.L., Härdter, R., Norton, R., Tao, H.H., Jaramillo, R., Satyanarayana, T., Zingore, S., and Oberthür, T. (2018). Chapter Four - Feeding the Palm: A Review of Oil Palm Nutrition. In Advances in Agronomy, D.L. Sparks, ed (Academic Press), pp. 149-243. Tränkner, M., Tavakol, E., and Jákli, B. (2018). Functioning of potassium and magnesium in photosynthesis, photosynthate translocation and photoprotection. Physiologia Plantarum 163, 414-431. Trono, D., Laus, M.N., Soccio, M., Alfarano, M., and Pastore, D. (2015). Modulation of Potassium Channel Activity in the Balance of ROS and ATP Production by Durum Wheat Mitochondria—An Amazing Defense Tool Against Hyperosmotic Stress. Frontiers in Plant Science 6. Trought, M.C.T., and Drew, M.C. (1981). Alleviation of Injury to Young Wheat Plants in Anaerobic Solution Cultures in Relation to the Supply of Nitrate and other Inorganic Nutrients. Journal of Experimental Botany 32, 509-522. Uozumi, N., Kim, E.J., Rubio, F., Yamaguchi, T., Muto, S., Tsuboi, A., Bakker, E.P., Nakamura, T., and Schroeder, J.I. (2000). The Arabidopsis HKT1 gene homolog mediates inward Na+ currents in xenopus laevis oocytes and Na+ uptake in Saccharomyces cerevisiae. Plant Physiol 122, 1249-1259.

62 USDA. (2014). Oilseeds: World Markets and Trade (Washington, D.C USDA Foreign Agricultural Service). van Dongen, J.T., Gupta, K.J., Ramírez-Aguilar, S.J., Araújo, W.L., Nunes-Nesi, A., and Fernie, A.R. (2011). Regulation of respiration in plants: A role for alternative metabolic pathways. Journal of plant physiology 168, 1434-1443. Venema, K., Belver, A., Marín-Manzano, M.C., Rodríguez-Rosales, M.P., and Donaire, J.P. (2003). A Novel Intracellular K+/H+ Antiporter Related to Na+/H+ Antiporters Is Important for K+ Ion Homeostasis in Plants. Journal of Biological Chemistry 278, 22453-22459. Villeret, V., Huang, S., Fromm, H.J., and Lipscomb, W.N. (1995). Crystallographic evidence for the action of potassium, thallium, and lithium ions on fructose-1,6- bisphosphatase. Proc Natl Acad Sci USA 92, 8916-8920. Voelker, C., Schmidt, D., Mueller-Roeber, B., and Czempinski, K. (2006). Members of the Arabidopsis AtTPK/KCO family form homomeric vacuolar channels in planta. The Plant journal: for cell and molecular biology 48, 296-306. Walker, D.J., Leigh, R.A., and Miller, A.J. (1996). Potassium homeostasis in vacuolate plant cells. Proc. Natl. Acad. Sci. USA 93, 10510-10514. Wang, F., Chen, Z.H., Liu, X., Colmer, T.D., Shabala, L., Salih, A., Zhou, M., and Shabala, S. (2016). Revealing the roles of GORK channels and NADPH oxidase in acclimation to hypoxia in Arabidopsis. Journal of Experimental Botany 68, 3191-3204.

Wang, N., Hua, H., Egrinya Eneji, A., Li, Z., Duan, L., and Tian, X. (2012). Genotypic variations in photosynthetic and physiological adjustment to potassium deficiency in cotton (Gossypium hirsutum). Journal of Photochemistry and Photobiology B: Biology 110, 1-8. Wang, X.G., Zhao, X.H., Jiang, C.J., Li, C.H., Cong, S., Wu, D., Chen, Y.Qq., Yu, H.Q., and Wang, C.Y. (2015). Effects of potassium deficiency on photosynthesis and photoprotection mechanisms in soybean (Glycine max (L.) Merr.). Journal of Integrative Agriculture 14, 856-863. Wang, Y., and Wu, W.H. (2013). Potassium Transport and Signaling in Higher Plants. Annual review of plant biology 64, 451-476. Wang, Y.i., and Wu, W.H. (2010). Plant Sensing and Signaling in Response to K+- Deficiency. Molecular Plant 3, 280-287. Weng, X.Y., Zheng, C.J., Xu, H.X., and Sun, J.Y. (2007). Characteristics of photosynthesis and functions of the water-water cycle in rice (Oryza sativa) leaves in response to potassium deficiency. Physiologia Plantarum 131, 614-621. White, P.J., Wheatley, R.E., Hammond, J.P., and Zhang, K. (2007). Chapter 34 Minerals, Soils and Roots. In Potato Biology and Biotechnology Advances and Perspectives, D. Vreugdenhil, ed (The Netherlands: Elsevier. Woittiez, L.S., Slingerland, M., Rafik, R., and Giller, K.E. (2018). Nutritional imbalance in smallholder oil palm plantations in Indonesia. Nutrient Cycling in Agroecosystems 111, 73-86. Wynn, R.M., Ho, R., Chuang, J.L., and Chuang, D.T. (2001). Roles of Active Site and Novel K+ Ion-binding Site Residues in Human Mitochondrial Branched-chain α- Ketoacid Decarboxylase/Dehydrogenase. Journal of Biological Chemistry 276, 4168-4174. Zeeman, S.C., Kossmann, J., and Smith, A.M. (2010). Starch: its metabolism, evolution, and biotechnological modification in plants. Annual review of plant biology 61, 209-234.

63 Zeng, F., Konnerup, D., Shabala, L., Zhou, M., Colmer, T.D., Zhang, G., and Shabala, S. (2014). Linking oxygen availability with membrane potential maintenance and K+ retention of barley roots: implications for waterlogging stress tolerance. Plant Cell Environ 37, 2325-2338. Zeng, Y., Wu, Y., Avigne, W.T., and Koch, K.E. (1999). Rapid Repression of Maize Invertases by Low Oxygen. Invertase/Sucrose Synthase Balance, Sugar Signaling Potential, and Seedling Survival. Plant Physiology 121, 599-608. Zhang, X., Jiang, H., Wang, H., Cui, J., Wang, J., Hu, J., Guo, L., Qian, Q., and Xue, D. (2017). Transcriptome Analysis of Rice Seedling Roots in Response to Potassium Deficiency. Scientific reports 7, 5523. Zhao, D., Oosterhuis, D.M., and Bednarz, C.W. (2001). Influence of Potassium Deficiency on Photosynthesis, Chlorophyll Content, and Chloroplast Ultrastructure of Cotton Plants. Photosynthetica 39, 103-109. Zhao, X., Li, T.L., and Sun, Z.P. (2010). Effects of prolonged root-zone CO2 treatment on morphological parameter and nutrient uptake of tomato grown in aeroponic system. Journal of Applied Botany and Food Quality 83, 212-216. Zhao, J., Li, P., Motes, C. M., Park, S., & Hirschi, K. D. (2015). CHX14 is aplasma membrane K‐efflux transporter that regulates K+ redistributionin Arabidopsisthaliana. Plant, Cell & Environment, 38 (11), 2223-2238. Zörb, C., Senbayram, M., and Peiter, E. (2014). Potassium in agriculture – Status and perspectives. Journal of Plant Physiology 171, 656-669.

64 Chapter 2 - Materials and methods

2.1 Plant material, growth conditions, and stress treatments

2.1.1 Plant material and growth condition

Sunflower (Helianthus annuus L.) var. XRQ was provided by the INRA Toulouse (France). Plants were sown directly in sand (washed with distilled water) in the greenhouse, using 7-L pots (Figure 1). Growth conditions were: 9/15 h (August, 2016) photoperiod, 25/18°C air temperature, 70/60% relative humidity day/night.

Day (4) Day (12)

Day (17) Day (28)

Figure 1. Pictures of sunflower growing in glasshouse at various days after sowing.

65 Germinated oil palm seeds (Dura  Pisifera) were obtained from Siam Elite Palm Co. Ltd. (Krabi, Thailand). Plants were sown directly in sand (washed with water) in the greenhouse, using 7-L pots (Figure 2). After 5 months growing, seedlings were transplanted into 10-L pots. K-containing nutrient solution (4 mM KCl) was provided for one month until emergence, and then controlled nutrient conditions at varying K concentration were used. Plants were cultivated in the Research School of Biology plant facility at the Australian National University in Canberra (Australia). Growth conditions were: natural photoperiod (from 13.5/10.5 h at emergence (February, 2018) to 14.5/9.5 h at the end of the experiment (January, 2019), with a minimum of 9.5/14.5 h in June, 2018), 30/24°C air temperature, 70/60% relative humidity day/night.

Figure 2. Pictures of 9-month, 10-month and 11-month old oil palm grown in glasshouse. The K concentration (KCl) in the nutrient solution used to water the plants are indicated.

66 2.1.2 Nutrition solution and waterlogging

2.1.2.1 Sunflower Monofactorial (K) and Crossed effect experiment (K×waterlogging) nutrient solution

The nutrient solution was composed on purpose for this experiment, whereby the amount of K+ was varied by changing the amount of KCl. Two K availability conditions were used here: “low K” (0.2 mM) and “high K” (4 mM). The amount of nitrate and phosphate (in mM) in the nutrient solution was kept constant throughout experiments (nutrient solution composition in Table 1). Under non-waterlogging (control) conditions, 300 mL nutrient solution was provided to each pot every day. Excess liquid was allowed to drain freely from the pot, so that on average 250 mL was retained in pots each day. Waterlogging was performed by filling the pot with the nutrient solution. The volume of water was adjusted everyday to compensate for evaporation and transpiration. In practice, about 200 mL were added each day to each pot. Therefore, the concentration in dissolved oxygen in the pot decreased slowly, generating a progressive hypoxia rather than abrupt anoxia. Nitrate concentration of the nutreitn solution was 15 mM so that the volume of

– nutrient solution added each day ( 3 mmol NO3 ) covered plant N needs. In a separate experiment (Figure 4), no waterlogging was applied and four K availability conditions were used: “low K” (0.2 mM), “medium K” (1 mM), “high K” (4 mM), and “resupply” whereby cultivation started with 0.2 mM K+ and after two weeks changed to 4 mM for two weeks.

Table 1. Chemical composition of nutrient solutions used in the sunflower experiments. All concentrations are given in mM. KCl concentration was fixed at 4, 1 or 0.2 mM to have high, medium and low K conditions, respectively. The composition of nutrient solutions used in potassium nutrition experiments, based on Hewitt’s nitrate type formula (Wong, 1979; Hewitt and Smith, 1975).

Compound Concentration (mM) KCl 4/1/0.2 NaNO3 4 Ca(NO3)2 4 MgSO4 1.5 NaH2PO4 1.33 Fe-Na (EDTA) 0.11 -3 MnSO4 10 -3 ZnSO4 10 -3 CuSO4 10 H3BO3 0.05 -4 Na2MoO4 510 -4 Co(NO3)2 210

67 2.1.2.2 Oil palm Monofactorial (K) and Crossed effect experiment (K×waterlogging) nutrient solution

The nutrient solution composition is in Table 2, the amount of K+ was varied by changing the amount of KCl. Three K availability conditions were used here: “low K” (0.2 mM), “medium K” (1 mM), and “high K” (4 mM) before the waterlogging experiment. Resupplying K to low-K plants was from month 9 or 10 (Figure 5). The amount of nitrate and phosphate (in mM) in the nutrient solution was kept constant throughout experiments. Under non-waterlogging (control) conditions, 500 mL nutrient solution was provided to each pot every day. Excess liquid was allowed to drain freely from the pot. Waterlogging was performed by filling the pot with the nutrient solution. Only two K concentrations were used for waterlogging (low, high). The volume of water was adjusted everyday to compensate for evaporation and transpiration. In practice, about 200 mL were added each day to each pot. Therefore, the concentration in dissolved oxygen in the pot decreased slowly, generating a progressive hypoxia rather than abrupt anoxia.

Table 2. Nutrient solution composition used in oil palm experiments. KCl concentration was fixed at 4, 1 or 0.2 mM to have high, medium and low K conditions, respectively. The composition of nutrient solutions used in potassium nutrition experiments, based on Hewitt’s nitrate type formula (Wong, 1979; Hewitt and Smith, 1975).

Compound Concentration (mM) KCl 4/1/0.2 NaNO3 4 Ca(NO3)2 4 MgSO4 1.5 NaH2PO4 1.33 Fe-Na (EDTA) 0.11 -3 MnSO4 210 -3 ZnSO4 210 CuSO4 210-3 H3BO3 0.1 Na2MoO4 110-3 Co(NO3)2 410-4

68 2.1.3 Experimental design

2.1.3.1 Sunflower experiments design

For sunflower Crossed effect (K×waterlogging) experiment, plants were cultivated for 2 weeks under low or high K conditions without waterlogging, and then waterlogging was started, with half of the plants kept under control conditions (no waterlogging). The experimental design is shown in Figure 3. Samples were havested at four sampling dates: just before the onset of waterlogging (day 0), and then after 2, 7 and 14 days. The day before sampling, photosynthesis and dark respiration (reported in Chapter 3 Figure 1) were measured on the same plants.

For Sunflower Monofactorial (K) experiment design (Figure 4), there are four treatments, namely, low K, medium K, high K or resupply K. For low, medium or high K treatment, plants were cultivated in these conditions for four weeks, while for resupply K treatment, plants were initially cultivated at low K for two weeks followed by resupplying with high K for an additional two weeks.

S WL

LK + WL or HK + WL Growth under LK or HK for two weeks LK or HK (no waterlogging) Days: 0 2 7 14

Four samplings

Figure 3. Diagram of Sunflower Crossed effect experiment. Arrows: S, sowing; WL, onset of waterlogging. Triangles: sampling dates. Five biological replicates were harvested from five individual plants. Abbreviation: LK: Low K, HK: High K, LK+WL: Low K+waterlogging, HK+WL: High K+waterlogging. Measurements including photosynthesis, dark respiration and SPAD were carried out the day before sampling.

Figure 4. Diagram of Sunflower Monofactorial (K) experiment. Triangle (blue): onset of waterlogging. Triangles (black): sampling dates. Five biological replicates were harvested from five individual plants. Abbreviation: LK: Low K, MK: Medium K, HK: High K, RK: Resupply K.

2.1.3.2 Oil palm experiments design

For oil palm experiments, plants were cultivated for 11 months under low, medium or high K conditions without waterlogging, and then waterlogging was started, with half of the plants kept under control conditions (no waterlogging). The experimental design is shown in Figure 5. Data presented here show results obtained upon seven sampling dates: two months, one month or one week before the onset of waterlogging (samplings no. 1, 2 and 3 respectively), and then after 1, 2, 3 or 7 weeks (samplings no. 4, 5, 6 and 7, respectively). The day before sampling, photosynthesis and dark respiration (reported in Chapter 4 Figure 1) were measured on the same plants.

Figure 5. Schematic representation of the experiment carried out in oil palm. Note that the total duration of the experiment was 12 months and 7 weeks, and that there were 7 sampling campaigns, representing 140 trees overall. The experiment was organized in such a way that the effect of K conditions alone (monofactorial experiment) was examined before the onset of waterlogging, such that the age of oil palm trees was not exactly the same between the monofactorial experiment and the crossed effect experiment (maximal difference of 3 months, between first and last sampling dates). Five biological replicates were harvested from five individual plants.

2.1.4 Experimental sampling

For sunflower experiments, on the day before sampling, photosynthesis and dark respiration were measured on the same plants subjecting to be harvested. Total number of plants, replicate numbers were summarised in Table 3 and 4. Root and shoot tissues were harvested separately. Two sets of three leaf discs (size: 1.9 cm2 each) from two young mature leaves (recently fully expanded leaves, the 5th and 6th leaves counted from bottom) of the same plant were harvested and quickly weighed before snap-frozen (metabolic quenching) in liquid nitrogen for metabolites analysis. The leftover materials of those two leaves were weighted and snap-frozen. Leaf 3 and 4 counted from bottom were weighted and snap-frozen. Rest of shoot tissues were measured for fresh weight and snap-frozen in liquid nitrogen.

71 Root tissues were carefully recovered from sand and rinsed in water for three times followed by blot-drying with paper towel. Typical sunflower roots after washing are shown in Figure 6. After determining fresh weight, samples were separated in two parts and snap-frozen in liquid nitrogen within 2−3 minutes. All samples were kept in - 80℃ until use.

All tissues except the two sets of the three leaf discs and one set of fresh root tissue were freeze-dried. All samples were grounded (at -80°C) into fine powder using ANU Cryogenic Grinder System (Labman Automation LTD). A protortion of grinded fresh samples were taken for metabolites anlaysis. A proportion of freeze-dried grinded material were taken for isotope and ionomics analyses.

Table 3. Sunflower Crossed effect (K×waterlogging) experiment sampling plants number. Five biological replicates were from five individual plants.

Treatment Plant Sampling date Treatment number Replicate number 2-week LK and HK 2 5 10 2-week and 2-day LK, HK, LKW and HKW 4 5 20 2-week and 7-day LK, HK, LKW and HKW 4 5 20 2-week and 14-day LK, HK, LKW and HKW 4 5 20 Total plants harvested 70

Table 4. Sunflower Monofactorial (K) experiment sampling plants number. Five biological replicates were from five individual plants.

Plant Sampling date Treatment Treatment number Replicate number 2-week LK, MK, HK 3 5 15 4-week LK, MK, HK and RK 4 5 20 Total plants harvested 35

Figure 6. Typical roots of 28-day-old sunflower under K and waterlogging treatments: (A) Low K well drained; (B) High K well drained; (C) Low K + waterlogging; (D) High K + waterlogging. Seedlings were grown for 14 days under low K or high K before subjecting to another 14 days under well drained or continuous waterlogging treatment with same level of K supplied.

For oil palm Monofactorial (K) and Crossed effect experiments, on the day before sampling, photosynthesis and dark respiration were measured on the same plants subjecting to be harvested. Total number of plants, replicate numbers were summarised in Table 5. Upon sampling, Root, stem, rachis, and leaf tissues were harvested separately. Six leaf discs (size: 1.9 cm2 each) from recently fully expanded leave (the 8th leaf counted from bottom) of the same plant were harvested and quickly weighed before snap-frozen (metabolic quenching) in liquid nitrogen for proteomics analysis. The leftover leaflet materials of leaf 8 were weighted and snap-frozen. Rachis 8 were weighted and snap- frozen. Rest of shoot tissues were measured for fresh weight and snap-frozen in liquid nitrogen.

Root tissues were carefully recovered from sand and rinsed in water for three times followed by blot-drying with paper towel. A typical root after washing is shown in Figure 7. After determining fresh weight, samples were snap-frozen in liquid nitrogen within 2−3 minutes. All samples were kept in -80℃ until lyophilisation.

All tissues except the six leaf discs were freeze-dried. All samples were grounded (at -80°C) into fine powder using ANU Cryogenic Grinder System (Labman Automation LTD). Grinded fresh leaflet samples were taken for proteomics anlaysis. A proportion of

73 freeze-dried grinded material were taken for metabolomics, isotope and ionomics analyses.

Table 5. Oil plam Monofactorial (K) and Crossed effect (K×waterlogging) experiment sampling plants number. Five biological replicates were from five individual plants.

Treatment Plant Sampling date Treatment number Replicate number 9-month LK, MK and HK 3 5 15 10-month LK, MK, HK and RK 4 5 20 11-month LK, MK, HK, RK_1 and RK_2 5 5 25 11-month and 1-week LK, HK, LKW and HKW 4 5 20 11-month and 2-week LK, HK, LKW and HKW 4 5 20 11-month and 3-week LK, HK, LKW and HKW 4 5 20 11-month and 7-week LK, HK, LKW and HKW 4 5 20 Total plants harvested 140

Figure 7. A typical oil palm root after washing.

74 2.2 Relative chlorophyll content

Leaf relative chlorophyll content was determined by SPAD 502 (Konica Minolta®). The device measures the transmittance of the leaf using two wavelengths: at 650 nm where chlorophyll absorbance is high and not affected by carotenes absorbance and 940 nm where the absorption of the light emitted is not only due to chlorophyll pigments but also to the characteristics of leaf structure. After a calibration carried out with an "empty" sample to determine a reference measurement at 100% transmission, the device calculates an index "SPAD" (Markwell et al., 1995).

2.3 Photosynthesis and respiration

Photosynthetic parameters were measured using a portable open system Li-Cor 6400 XT. Net assimilation (A) and conductance reported in Chapter 3 Figure 1 and Chapter 4 Figure -2 -1 -1 1 was obtained under saturating light (1500 µmol m s PAR) at 400 µmol mol CO2 -1 and 21% O2, flow rate of 500 µmol s . Night respiration (Rn) was measured after -1 photosynthesis measurements on dark-adapted leaves (30 min) at 400 µmol mol CO2,

21% O2 and 25°C.

2.4 Metabolomics

Metabolomics were performed by gas chromatography coupled to mass spectrometry (GC-MS, Agilent Technologies, Palo Alto, CA, USA) as in (Abadie et al., 2017). Briefly, 5 volumes of cold (–20°C) extraction medium (methanol:water 90:10 v:v with 0.01 mg/mL ribitol) were added to grounded plant tissue. Samples were then vortexed and incubated at 60°C for 15 min with 1,400 rpm shaking. After centrifugation for 10 min at 14,800 g, 10 µL of the supernatant was transferred to a glass vial and vacuum-dried. Samples were derivatized with methoxylamine and N-methyl-N-(trimethylsilyl) trifluoroacetamide (MSTFA) in pyridine and 1 µL of derivatized extract was injected into a VF-5 capillary column (30m250µm0.25µm). Samples were injected in split and splitless modes. Injector temperature was set at 230°C. Initial GC oven temperature was set at 70°C. 1 min after injection, the GC oven temperature was raised to 325°C at 15°C min-1, and finally kept at 325°C for 3 min. The GC run time was 21 min. Helium was used as the carrier gas with a constant flow rate of 1 mL min-1. Measurements were done with electron impact ionization (70 eV) in the full scan mode (m/z 40−600). 5 µL of alkane mix (C12, C15, C19, C22, C28, C32, C36 alkanes standard at 28 mg/L) were injected to

75 compute the retention index. Peak integration and identification was done using Metabolome Express (Carroll et al., 2010). Briefly, peak detection was carried out using a slope threshold of 200, minimal peak area of 1000, minimal peak height of 500, minimal peak purity factor of 2 and minimal peak width of 5 scans. Peaks were identified against a MSRI library using the retention index (RI) and mass-spectral similarity as identification criteria, with matching parameters as follows: RI window tolerance of ± 2 RI units, MST centroid distance of ± 1 RI unit, minimal peak area of 5000, MS qualifier ion ratio error tolerance of 30%, minimum number of correct ratio qualifier ions of 2, and maximum average MS ratio error of 70%. The primary MSRI library used contained entries derived manually from analyses of authentic standards run under the same GC- MS conditions as well as entries for unidentified peaks that were automatically generated by Metabolome Express while processing the data. Peak areas were normalized to the internal standard (ribitol). As a quality control filter, samples were checked for the presence of a ribitol peak with an area of at least 105 and a deviation from the median internal standard peak area of less than 70%. Metabolomics data were normalized to fresh weight.

2.5 Ionomics

Ionomics (elemental) analyses were done by inductively coupled plasma optical emission spectroscopy (ICP-OES). 10 mg freeze-dried powdered sample were reduced to ashes at 500°C for 3 h in a porcelain crucible (pre-washed with 10% nitric acid). Then digestion was carried out with 1 mL nitric acid (6 M in LC-MS grade water) at 70°C for 1 h. The liquid was evaporated and the sample was re-extracted with 2 mL nitric acid (2%) at 70°C for 3 h. The sample was transferred into a 15-mL tube and nitric acid was added up to 5 mL final volume. A 3 mL aliquot of the digestion was injected in a ICP-OES system (Agilent 720) with a flow rate of 1 mL min-1. Elemental contents were calculated using a calibration curve processed for each batch of samples. ICP-OES data reported in Chapter 3 Figure 2 and Chapter 4 Figure S2 are given in mg element mg-1 DW.

2.6 Proteomics

Analyses were done by a facility (PAPPSO, France) and here is the protocol used.

76 2.6.1 Protein extraction and in-solution digestion

Plant samples were finely ground in liquid nitrogen. Pellets proteins were solubilized in 30 µL per mg of extract of solubilization buffer (6 M urea, 2 M thiourea,10 mM dithiothreitol (DTT), 30 mM Tris-HCl pH 8.8, and 0.1% zwitterionic acid labile surfactant (ZALS I, Proteabio, Morgantown, WV, USA)). Protein concentrations were determined using the PlusOne 2D-Quant Kit (GE Healthcare, Little Chalfont, UK) and adjusted to 0.2 µg µL-1. Digestion was performed in 0.2 mL strip tubes, 20 µg of each sample was alkylated by incubation in darkness for 1 h at room temperature with 20 μL of 300 mM iodoacetamide (in 50 mM ammonium bicarbonate). Proteins were then diluted ten times in 880 μL of 50 mM ammonium bicarbonate and digested using 3 μL trypsin solution (enzyme:substrate ratio of 1:30 w:w). Digestion was stopped with 5 μL of trifluoroacetic acid. Peptides were desalted using C18 solid phase extraction (SPE) cartridges (strata XL 100 µm ref 8E-S043-TGB, Phenomenex, Torrance, CA, USA).

2.6.2 LC-MS/MS analyses

HPLC was performed on a Nano-HPLC (Eksigent). Buffers A and B were prepared with 0.1% formic acid in water, and with 0.1% formic acid in acetonitrile, respectively. A 4 μL-sample of the peptide solution was loaded at 7.5 μL min-1 for 1 min on a Biosphere

C18 trap-column (particle size: 5 μm, pore size: 12 nm, inner/outer diameters: 360/100 μm, length: 20 mm; NanoSeparations) and desalted with 0.1% formic acid, 2% acetonitrile in water. Then peptides were separated on a biosphere C18 column (particle size: 3 μm, pore size: 12 nm, inner/outer diameters: 360/75 μm, length: 300 mm; NanoSeparations). Peptide separation was achieved at 300 nL min-1 with the following steps: equilibration in 95% buffer A for 9 min, separation with a linear gradient from 5 to 30% buffer B for 75 min, linear gradient from 35 to 95% buffer B for 3 min and regeneration with 95% buffer B for 10 min. Eluted peptides were analyzed on-line with a Q-Exactive Plus mass spectrometer (ThermoFisher Scientific, Courtaboeuf, France) using a nanoelectrospray interface. Ionization (1.5 kV ionization potential) was performed with a glass needle (non-coated capillary silica tips, 360/20-10, New Objective). Peptide ions were analyzed using Xcalibur 4.0.27 with the following datadependent acquisition steps: (1) full MS scan (mass-to-charge ratio (m/ z), 350 to 1,400, in profile mode) with a resolution of 70,000 and (2) MS/MS (isolation window = 1.5 m/z, AGCtarget = 1E5, max Ion Time = 120 ms, collision energy = 27%; profile mode,

77 resolution = 17,500). Step 2 was repeated for the eight major ions detected in step 1. Dynamic exclusion was set to 50 s. Only doubly and triply charged precursor ions were subjected to MS/MS fragmentation.

2.6.3 Protein identification and peptide quantification

Xcalibur raw data were transformed to mzXML open source format and centroided using the msconvert software of the package ProteoWizard 3.0.3706 (Kessner et al., 2008). Protein identification was performed using the X!Tandem Alanine (version 2017.2.1.4; www.thegpm.org) by querying MS/MS data against the GCF_000442705.1_EG5_protein.faa protein library together with a custom contaminant database (trypsin, keratins). The following static and possible modification were accounted for: one missed trypsin cleavage allowed, alkylation of cysteine and oxidation of methionine, respectively. Precursor mass tolerance was set to 10 ppm and fragment ion mass tolerance was 0.02 Da. A refinement search was added with similar parameters except that possible N-terminal acetylation with peptide signal cleavage was included. Identified proteins were filtered and grouped using X!TandemPipeline (C++) 0.2.27 (pappso.inra.fr/bioinfo/xtandempipeline) (Langella et al., 2017), with (1) a minimum of two different peptides required with an E-value smaller than 0.01, (2) a protein E-value (calculated as the product of unique peptide E-values) smaller than 10-4. Using a reverse version of the GCF_000442705.1_EG5_protein.faa database as a decoy, the false discovery rate (FDR) was estimated by X!Tandem to be 0.08% and 0.00% for peptide- spectrum match and protein identification, respectively. Peptide quantitation used the normalized spectral abundance factor (NSAF) (McIlwain et al., 2012).

2.7 Estimation of CUE

Carbon use efficiency (CUE) was calculated as the ratio of present carbon in biomass

(Ab) to photosynthetically fixed carbon via leaf net assimilation (gross photosynthesis,

Ab):

A CUE = b Ag

Ab is given by Bp, where B is plant biomass (in g) and p is %C in total organic matter - (40%). Ag was calculated as the cumulated photosynthetic input, as 10

6 -1 -1 12iAm(i)Bliti where Am(i) is average photosynthesis rate in µmol g s during

78 growth period i (comprised between two sampling dates), li is the proportion of leaf biomass in total tree biomass and ti is total photoperiod time (day time integrated length, in s) between two sampling dates. Under our conditions, li was about 0.4−0.5 across all conditions, due to the substantial biomass represented by rachis and stem. 12 represents the molar mass of carbon (12 g mol-1) and 10-6 the conversion factor from µmol to mol.

By definition, total (cumulated) respiratory loss (CRL) was given by CRL = Ag – Ab =

Ag(1 – CUE).

2.8 Isotope analysis and %N (13C and 15N abundance by elemental-analysis coupled to isotope ratio mass spectrometry)

100 mg of powdered freeze-dried material were first extracted with hexane to obtain the total lipid fraction (supernatant that contained chlorophylls). The pellet was extracted with MilliQ water. After centrifugation, the supernatant was collected and heated at 100°C for 5 min. The heat-precipitated protein pellet was collected and the deproteinated supernatant was sub-divided into two halves. One half was freeze-dried and used for total soluble N isotope analysis. The other half was used for nitrate extraction as barium nitrate with barium iodide, after (Huber et al., 2011). Samples were weighed in tin capsules, and δ15N values and %N were measured using an elemental analyser (Carlo-Erba) coupled to isotope ratio mass spectrometer (Isochrom, Elementar) run in continuous flow. All sample batches included standards (glycine, +0.66‰; cysteine, +7.78‰; previously calibrated against IAEA standards glutamic acid USGS-40 and caffeine IAEA-600) each ten samples. When the sample mass was occasionally low (typically < 0.08 µmol N sample-1), the δ15N value was corrected using the response curve of the δ15N value to sample mass using standard glycine and cysteine.

13 2.9 CO2 labelling and sampling

Gas exchange was per-formed in a chamber coupled to a LI-COR 6400-XT device (LI- COR Biosciences, Lincoln, NE, USA), with walls to allow instant sampling by liquid nitrogen spraying, as described previously (Tcherkez et al., 2012). The leaf chamber was allowed to two individual oil palm leaflets with a surface area of 65−90 cm². Light was supplied by an RGBW-L084 LED panel (Walz, Effeltrich, Germany). Gas-exchange -1 conditions were: 400 µ mol mol CO2 photosynthetically active radiation (PAR), 80% relative humidity and 21–23°C air temperature. Isotopic labelling was performed using 13 13 CO2 (Sigma-Aldrich, 99% C) for 4 h after having reached steady photosynthesis using

79 ordinary CO2 (natural abundance) for 40−60 min. In the dark experiment, after 4 h 13 labelling with CO2, labelled with CO2 (natural abundance) for 1 h under dark condition. For all K treatment conditions (low K and high K; light and dark here, two series of 13 12 experiments were done: with CO2 and with natural CO2. Performing experiments with 12 13 natural CO2 was strictly required for % C calculations using NMR data.

2.10 Extraction and NMR analysis Samples were extracted with perchloric acid in liquid nitrogen as previously described (Aubert et al., 1994). Briefly, the sample was ground with 900µl perchloric acid 70%; 500µl maleic acid 0.25 M (i.e. a total of 125 µmol per sample, used as an internal standard) and 50µl methyl-phosphonate 0.1 M (i.e. a total of 5 µmol per sample, used as an internal standard). The powder was poured in a 50 ml centrifuge tube and then10 ml MilliQ water was added. After centrifugation (10 000g, 15 min), the pellet was re-extracted with 3 ml perchloric acid 2% and centrifuged. The two supernatants were combined in the same 50 ml tube, the pH was adjusted to 5 with potassium hydroxide/potassium bicarbonate and the sample was frozen-dried. Resuspended samples in 1.2 ml CDTA 12.5 mM at pH 7, after centrifugation at 10 000g, the supernatant was transferred in an Eppendorf tube, adjusting the pH to 7 with KOH, centrifuge the sample at about 10 000g, In total, 550 µl supernatant was transferred to a new Eppendorf tube, 50 µl D2O was added and the sample was vortexed and poured in a 5-mm NMR tube (Z107373; Bruker Biospin, Wissembourg, France) for analysis (acquisition with CDTA). Samples were analysed with an Advance 700 Mz NMR spectrometer (Bruker Biospin, Alexandria, Australia). NMR analyses were performed at 298 K (25°C) without tube spinning, using a proton-decoupled carbon pulse program (zgig) with 90° pulses for 13C (10µs at 50 W). Acquisition parameters were: decoupling sequence waltz 16, acquisition time 0.9 s, size of flame ionization detector 65 k, and a relaxation delay (D1) of 1.2 s. Twelve thousand scans were done, representing c. 9 h of analysis per sample. NMR data presented in the paper are mean ± SD of n = 4 replicates.

2.11 Statistical analyses and data presentation

Unless otherwise stated, five replicates were done for all conditions. Supervised multivariate analysis of metabolomics data was carried out by orthogonal projection on latent structure (OPLS) with Simca (Umetrics), using K level and/or waterlogging as predicted qualitative Y variables and metabolites as predicting X variables. Also, to gain

80 insight on metabolic mechanisms that determine the rate of leaf dark respiration, an OPLS analysis was performed using protein and metabolite contents as predicting (X) variables and using Rdark as a predicted (Y) quantitative variable. The absence of statistical outliers was first checked using a principal component analysis (PCA) to verify that no data point was outside the 99% confidence Hostelling region. The goodness of the OPLS model was appreciated using the determination coefficient R² and the predictive power was quantified by the cross-validated determination coefficient, Q². The significance of the statistical OPLS model was tested using a ² comparison with a random model

(average  random error), and the associated P-value (PCV-ANOVA) is reported. A permutation test was also performed to check the reliability of the OPLS model, that is, to verify that at maximal permutation (similarity of permuted dataset tending to zero), Q² was always negative. Best discriminating metabolites (biomarkers) were identified using volcano plots whereby the logarithm of the P-value obtained in univariate analysis (two- way ANOVA) was plotted against the rescaled loading (pcorr) obtained in the OPLS. In such a representation, best biomarkers have both maximal –log(P) and pcorr values. Univariate analysis of statistical classes (e.g. in Chapter 3 Figure 3 and 4, Figure S3 and S4; Chapter 4 Figure 2 and 3 , Figure S3 and S6) was performed using a two-way ANOVA (Fisher test), with a threshold of P = 0.05.

81 Reference Abadie, C., Blanchet, S., Carroll, A., Tcherkez, G. (2017). Metabolomics analysis of postphotosynthetic effects of gaseous O2 on primary metabolism in illuminated leaves. Functional Plant Biology 44(9), 929-940. Aubert, S., Gout, E., Bligny, R., Douce, R. (1994). Multiple effects of glycerol on plant cell metabolism. J Biol Chem 269, 21420-21427. Carroll, A.J., Badger, M.R., Millar. H.A. (2010). The MetabolomeExpress Project: enabling web-based processing, analysis and transparent dissemination of GC/MS metabolomics datasets. BMC Bioinformatics 11, 376. Hewitt, E.J. and Smith, T.A. (l975). Plant mineral nutrition. The English University Press, London. Huber, B., Bernasconi S.M., Luster, J., Pannatier-Graf, E. (2011). A new isolation procedure of nitrate from freshwater for nitrogen and oxygen isotope analysis. Rapid Communications in Mass Spectrometry 25(20), 3056-3062. Kessner, D., Chambers, M., Burke, R., Agusand, D., Mallick, P. (2008). ProteoWizard: open source software for rapid proteomics tools development. Bioinformatics 24, 2534-2536. Langella, O., Valot, B., Balliau, T., Blein-Nicolas, M., Bonhomme, L., Zivy, M. (2017). X!TandemPipeline: a tool to manage sequence redundancy for protein inference and phosphosite identification. Journal of Proteome Research 16, 494- 503. Markwell, J., Osterman, J., and Mitchell, J. (1995). Calibration of the Minolta SPAD- 502 leaf chlorophyll meter. In: Photosynthesis Research, Vol. 46(3), 467-472. McIlwain, S., Mathews, M., Bereman, M.S., Rubel, E.W., MacCoss, M.J., Noble, W.S. (2012). Estimating relative abundances of proteins from shotgun proteomics data. BMC bioinformatics 13, 308. Tcherkez, G., Mahé, A., Guérad, F., Boex‐Fontvieille, E.R.A., Gout, E., Lamothe, M., Barbour, M.M., Bligny, R. (2012). Short‐term effects of CO2 and O2 on citrate metabolism in illuminated leaves. Plant, Cell & Environment 35, 2208- 2220. Wong, S.C. (1979). Stomatal behaviour in relation to photosynthesis. PhD thesis.

82 Chapter 3 - Responses to K deficiency and waterlogging interact via nitrogen metabolism and partitioning in sunflower

3.1 Motivation

Experiments in this chapter were carried out for two reasons: first, to take advantage of the time spent in cultivating oil palm (used in following chapters) to assess protocols and methods and second, to have results in an herbaceous oil-producing species. In addition, I planned to see whether the response of sunflower to K availability and waterlogging differed to that in oil palm, since sunflower has a conservatively high stomatal conductance (hence with a potentially high sensitivity of stomatal conductance and water loss to K limitation and waterlogging) while oil palm has a relatively low stomatal conductance in young plants (saplings).

3.2 Context

K-deficient soils are relatively common in countries of the intertropical region, such as Brazil, Equatorial Africa (wet tropical), China and South‐Eastern Asia (such as Malaysia or Indonesia), and Australia. In addition, most of these areas are associated with limited soil drainage, occasional flooding, or waterlogging events. These areas include important agricultural regions where crops such as oil palm, sunflower, cotton, or rice are cultivated. But quite critically, despite the frequent co-occurrence probability of K deficiency and waterlogging in the field, their combined effects on plant physiology are not so well known.

3.3 Key results and implications

I provide evidence that waterlogging and K deficiency do not have additive effects, and mostly impact on N allocation and metabolism. In addition, I take advantage of relationships between respiration rate and metabolomics patterns to show that catabolism involves an alternative, C5-branched acid pathway under K deficiency that is suppressed by waterlogging. This chapter will be interesting for plant physiology but also for potential applications to define a biomarker of waterlogging and K availability using leaf metabolites.

3.4 Published paper

83 Received: 27 June 2018 Revised: 13 September 2018 Accepted: 16 September 2018 DOI: 10.1111/pce.13450

ORIGINAL ARTICLE

Responses to K deficiency and waterlogging interact via respiratory and nitrogen metabolism

Jing Cui1 | Cyril Abadie1 | Adam Carroll2 | Emmanuelle Lamade3 | Guillaume Tcherkez1

1 Research School of Biology, Australian National University, Canberra, ACT, Australia Abstract 2 Joint Mass Spectrometry Facility, Research K deficiency and waterlogging are common stresses that can occur simultaneously School of Chemistry, Australian National and impact on crop development and yield. They are both known to affect catabolism, University, Canberra, ACT, Australia with rather opposite effects: inhibition of glycolysis and higher glycolytic fermentative 3 Unité PERSYST, UPR34, Système de pérennes, Centre de Coopération flux, respectively. But surprisingly, the effect of their combination on plant metabo- Internationale en Recherche Agronomique lism has never been examined precisely. Here, we applied a combined treatment pour le Développement, Montpellier, France Correspondence (K availability and waterlogging) to sunflower (Helianthus annuus L.) plants under con- Guillaume Tcherkez, Research School of trolled greenhouse conditions and performed elemental quantitation, metabolomics, Biology, Australian National University, Canberra 2601, ACT, Australia. and isotope analyses at different sampling times. Whereas separate K deficiency Email: [email protected] and waterlogging caused well‐known effects such as polyamines production and Funding information sugar accumulation, respectively, waterlogging altered K‐induced respiration Australian Research Council, Grant/Award Number: FT140100645; Australia Awards enhancement (via the C5‐branched acid pathway) and polyamine production, and K PhD Scholarship deficiency tended to suppress waterlogging‐induced accumulation of Krebs cycle intermediates in leaves. Furthermore, the natural 15N/14N isotope composition (δ15N) in leaf compounds shows that there was a change in nitrate circulation, with less nitrate influx to leaves under low K availablity combined with waterlogging and more isotopic dilution of lamina nitrates under high K. Our results show that K deficiency and waterlogging effects are not simply additive, reshape respiration as well as nitrogen metabolism and partitioning, and are associated with metabolomic and isotopic biomarkers of potential interest for crop monitoring.

KEYWORDS

isotope fractionation, K deficiency, metabolomics, nitrates, waterlogging

1 | INTRODUCTION occasional flooding, or waterlogging events. These areas include important agricultural regions where crops such as oil palm, sunflower, Potassium (K) is a macronutrient representing up to 5% of plant dry cotton, or rice are cultivated (www.fao.org). But quite critically, weight. K+ is involved in many physiological processes such as electro- despite the high co‐occurrence probability of K deficiency and chemical homeostasis, stomatal aperture, and several enzyme activi- waterlogging in the field, their combined effects on plant physiology ties (Anschütz, Becker, & Shabala, 2014; Wang & Wu, 2013). K are not so well known. deficiency is highly detrimental to plant growth and primary produc- However, there are significant interactions between K nutrition tion by crops, and intense efforts are being devoted to improve K and waterlogging (Ashraf, Ahmad, Ashraf, Al‐Qurainy, & Ashraf, acquisition by plants (Rawat, Sanwal, & Saxena, 2016; Shin, 2014). 2011; Cakmak, 2005; Gautam et al., 2016; Hasanuzzaman et al., K‐deficient soils are relatively common in countries of the intertropical 2018; Wang, Zheng, Shen, & Guo, 2013). Waterlogging leads to a gen- + − region, such as Brazil, Equatorial Africa (wet tropical), China and eral ion leakage and hence K and nitrate (NO3 ) loss by roots South‐Eastern Asia (such as Malaysia or Indonesia), and Australia. In (Demidchik, 2014). In fact, oxygen shortage in roots leads to a defi- addition, most of these areas are associated with limited soil drainage, ciency in ATP generation by respiration as well as a change in

84 648 CUI ET AL. transporter activity (such as an inhibition by ROS) such that both waterlogging, or both under controlled conditions (under nonlimited nitrate and K+ efflux from root cells increases (Barrett‐Lennard & N supply). Amongst oil‐producing crops, sunflower is representative Shabala, 2013; Shabala & Pottosin, 2014; Sharma, Sharma, & Choi, as an average K‐demanding species, with an apparent K use efficiency 2010; Zeng et al., 2014). It is believed that the rate of nitrate reduction of about 38 g oil g−1 applied K in the field (compared with 32–53 g (including in roots) does not decrease at least in the short‐ to mid‐term oil g−1 K in canola and about 35 g oil g−1 K in oil palm; Foster, 2002; so that there is a general decrease in nitrate concentration in plant tis- Kaiser, Rosen, & Lamb, 2016). Our data do not show an additive effect sues, typically in sunflower (Agüera, de la Haba, Fontes, & Maldonado, of K deficiency and waterlogging but rather an interaction whereby 1990). The simultaneous decrease in nitrate and water circulation surprisingly, waterlogging tended to suppress some K deficiency from roots to shoots and the increase in root K+ efflux tend to induce metabolic symptoms, such as increased leaf respiration or polyamine a potassium deficiency (Phukan, Mishra, & Shukla, 2016). As a matter accumulation. δ15N in leaf nitrates show that the balance between of fact, waterlogging symptoms have been shown to be partly allevi- nitrate utilization and import from roots was typically perturbed in ated by foliar spraying of potassium (Ashraf et al., 2011; Dwivedi waterlogged K‐deficient plants. Metabolomics patterns further reveal et al., 2017; Gautam et al., 2016). regulation points most affected by waterlogging and K deficiency Symptoms of K deficiency in plants have been studied for more and how they may interact to define leaf respiration rate. than 50 years and include metabolic effects such as an inhibition of glycolysis due to a decrease in pyruvate kinase activity, as well as accumulation of typical metabolites such as polyamines (Armengaud 2 | MATERIAL AND METHODS et al., 2009; Besford & Maw, 1976; Freeman, 1967; Hussain, Ali, Ahmad, & Siddique, 2011; Jones, 1961; Jones, 1966; Jwakyung 2.1 | Plant material and photosynthesis et al., 2015; Okamoto, 1966; Okamoto, 1967; Okamoto, 1968). Meta- Sunflower (H. annuus L.) var. XRQ was provided by the INRA Toulouse bolomics analyses of K deficiency in Arabidopsis showed a reorches- (France). Plants were sown directly in sand (washed with distilled tration of amino acid and organic acid synthesis, consistent with a water) in the greenhouse, using 7‐L pots. Growth conditions were change in nitrogen assimilation (in favour of neutral amino acids), 12/12‐hr photoperiod, 25/18°C air temperature, and 70/60% relative and an inhibition of the most K‐sensitive enzyme, pyruvate kinase, humidity day/night. Photosynthetic parameters were measured using thereby leading to altered glycolytic metabolism (Armengaud et al., a portable open system Li‐Cor 6400 XT. Net assimilation (A) and 2009). Despite some inconsistency in the effect of K deficiency on conductance reported in Figure 1 were obtained under saturating light root respiration (CO2 evolution rate; Okamoto, 1968), most studies (1,500 μmol m−2 s−1 PAR) at 400 μmol mol−1 CO and 21% O . Night show an increase in amino acid export from roots to shoots, and anal- 2 2 respiration (R ) was measured after photosynthesis measurements yses of xylem composition have indeed shown an increase in gluta- n on dark‐adapted leaves (30 min) at 400 μmol mol−1 CO , 21% O , mine and γ‐aminobutyrate content (Jwakyung et al., 2015). Such 2 2 and 25°C. metabolic consequences partly overlap with waterlogging effects (Orchard, Jessop, & So, 1986; Orchard & So, 1985; Phukan et al., 2016; Wang et al., 2014), and it has been shown that waterlogging 2.2 | Nutrient conditions and waterlogging and K deficiency share common signalling pathways via ethylene (Jung, Shin, & Schachtman, 2009). The nutrient solution was composed on purpose for this experiment, Nevertheless, the metabolic consequences of the combination of whereby the amount of K+ was varied by changing the amount of K deficiency and waterlogging have not been documented yet. KCl. Two K availability conditions were used here: “low K” (0.2 mM) Because K deficiency mostly impacts on glycolysis, whereas and “high K” (4 mM). The amount of nitrate and phosphate (in mM) waterlogging‐induced oxygen shortage stimulates substrate‐level in the nutrient solution was kept constant throughout experiments ATP generation, a drastic additive effect might be anticipated in roots. (nutrient solution composition in Table S1). Under nonwaterlogging By contrast, the overall effect of the combination of K deficiency and (control) conditions, 300ml nutrient solution was provided to each waterlogging on primary carbon and nitrogen metabolism in leaves is pot every day. Excess liquid was allowed to drain freely from the more difficult to anticipate, because leaves of waterlogged plants do pot, so that on average 250 ml was retained in pots each day. not exhibit typical hypoxic metabolism (Rocha et al., 2010) and may Waterlogging was performed by filling the pot with the nutrient solu- experience nitrogen shortage due to impeded plant nitrate circulation tion. The volume of water was adjusted every day to compensate for (see above). It is worth noting that a better knowledge of metabolic evaporation and transpiration. In practice, about 200 ml was added effects in leaves would be useful to define biomarkers for crop moni- each day to each pot. Therefore, the concentration in dissolved oxy- toring in the field (Hochmuth, 1994; Zörb, Senbayram, & Peiter, 2014). gen in the pot decreased slowly, generating a progressive hypoxia In effect, because roots sampling is far more difficult than leaf sam- rather than abrupt anoxia. Nitrate concentration of the nutrient solu- pling, having a leaf metabolic signature to detect and anticipate tion was 12 mM so that the volume of nutrient solution added each − waterlogging‐induced K deficiency would certainly be helpful. day (≈ 3 mmol NO3 ) covered plant N needs. In a separate experi- As an aid in clarifying the metabolic reorchestration of both K ment (Figures S1–S2), no waterlogging was applied, and four K avail- deficiency and waterlogging, we carried out gas‐exchange, GC‐MS ability conditions were used: “low K” (0.2 mM), “medium K” (1 mM), metabolomics, and isotopic (15N natural abundance, δ15N) analyses “high K” (4 mM), and “resupply” whereby cultivation started with in sunflower (Helianthus annuus L.) subjected to K deficiency, 0.2mM K+ and after 2 weeks changed to 4 mM for 2 weeks.

85 CUI ET AL. 649

FIGURE 1 Effects of K availability and waterlogging (WL) on carbon balance: (a) root and leaf biomass; (b) leaf elemental carbon content (% DW); −1 (c) and (d) net CO2 assimilation and stomatal conductance for water vapour under standard conditions (400 μmol mol CO2, 21% O2, saturating light, 25°C); (e) leaf respiration in darkness (at 25°C). The vertical dashed line separates samples before and after the onset of waterlogging. Mean ± SD (n = 5). In (a), error bars are small and partly hidden by symbols

−1 2.3 | Experimental design and sampling C36 alkanes standard at 28 mg L ) were injected to compute the retention index (RI). Peak integration and identification were done Plants were cultivated for 2 weeks under low or high K conditions using Metabolome Express (Carroll, Badger, & Harvey Millar, 2010). without waterlogging, and then waterlogging was started, with half Briefly, peak detection was carried out using a slope threshold of 200, of the plants kept under control conditions (no waterlogging). The minimal peak area of 1,000, minimal peak height of 500, minimal peak experimental design is shown as a separate figure in Scheme S1. Data purity factor of 2, and minimal peak width of five scans. Peaks were presented here show results obtained upon four sampling dates: just identified against an MSRI library using the RI and mass‐spectral sim- before the onset of waterlogging (Day 0) and then after 2, 7, and ilarity as identification criteria, with matching parameters as follows: RI 14 days. The day before sampling, photosynthesis and dark respiration window tolerance of ±2 RI units, MST centroid distance of ±1 RI unit, (reported in Figure 1) were measured on the same plants. Upon sam- minimal peak area of 5,000, MS qualifier ion ratio error tolerance pling, plants were measured for size, fresh weight, and leaf surface of 30%, minimum number of correct ratio qualifier ions of 2, and max- and dissected and kept in liquid nitrogen for metabolomics and isoto- imum average MS ratio error of 70%. The primary MSRI library used pic analyses. Roots were sampled after having removed sand and contained entries derived manually from analyses of authentic stan- washed with water. In practice, roots were washed, rapidly dried with dards run under the same GC‐MS conditions as well as entries for absorbing paper and quenched in liquid nitrogen within 2–3 min. unidentified peaks that were automatically generated by Metabolome Metabolomics analyses were performed on fresh material; isotope Express while processing the data. Peak areas were normalized to the and ionomics analyses were performed on freeze‐dried material. internal standard (ribitol). As a quality control filter, samples were checked for the presence of a ribitol peak with an area of at least 105 and a deviation from the median internal standard peak area of 2.4 | Metabolomics and ionomics less than 70%. Metabolomics data were normalized to fresh weight. Metabolomics were performed by gas chromatography coupled to Ionomics (elemental) analyses were done by inductively coupled mass spectrometry (GC‐MS, Agilent) as in Abadie, Blanchet, Carroll, plasma optical emission spectroscopy (ICP‐OES). Ten milligram and Tcherkez (2017). Briefly, five volumes of cold (−20°C) extraction freeze‐dried powdered sample was reduced to ashes at 500°C for medium (methanol:water 90:10 v:v with 0.01 mg ml−1 ribitol) were 3 hr in a porcelain crucible (prewashed with 10% nitric acid). Then added to grounded plant tissue. Samples were then vortexed and incu- digestion was carried out with 1ml nitric acid (6 M in LC‐MS grade bated at 60°C for 15 min with 1,400 rpm shaking. After centrifugation water) at 70°C for 1 hr. The liquid was evaporated and the sample for 10 min at 14,800 g, 10 μl of the supernatant was transferred to was re‐extracted with 2ml nitric acid (2%) at 70°C for 3 hr. The sample a glass vial and vacuum‐dried. Samples were derivatized with was transferred into a 15 ml tube, and nitric acid was added up to 5‐ml methoxylamine and N‐methyl‐N‐(trimethylsilyl) trifluoroacetamide final volume. A 3‐ml aliquot of the digestion was injected in a ICP‐OES (MSTFA) in pyridine, and 1 μl of derivatized extract was injected into system (Agilent 720) with a flow rate of 1 ml min−1. Elemental con- aVF‐5 capillary column (30 m × 250 μm × 0.25 μm). Samples were tents were calculated using a calibration curve processed for each injected in split and splitless modes. Injector temperature was set at batch of samples. ICP‐OES data reported in Figure 2 are given in mg 230°C. Initial GC oven temperature was set at 70°C. One minute after element mg−1 DW. injection, the GC oven temperature was raised to 325°C at 15°C min−1 and finally kept at 325°C for 3 min. Helium was used as the carrier gas − 2.5 | Isotope analysis and %N with a constant flow rate of 1 ml min 1. Measurements were done with electron impact ionization (70 eV) in the full scan mode (m/z One hundred milligrams of powdered freeze‐dried material was first

40–600). Five microlitres of alkane mix (C12, C15, C19, C22, C28, C32, extracted with hexane to obtain the total lipid fraction (supernatant

86 650 CUI ET AL.

FIGURE 2 Response of macronutrients to K availability and waterlogging (WL) in sunflower: leaves (a), stem (b), and roots (c). Elements were quantified by ICP‐OES, and contents are given in mg g−1 DW. Data shown here were obtained after 4 weeks (i.e., 2 weeks after starting waterlogging). Mean ± SD (n = 5) [Colour figure can be viewed at wileyonlinelibrary.com]

that contained chlorophylls). The pellet was extracted with MilliQ out by orthogonal projection on latent structure (OPLS) with Simca water. After centrifugation, the supernatant was collected and heated (Umetrics), using K level and/or waterlogging as predicted Y vari- at 100°C for 5 min. The heat‐precipitated protein pellet was collected, ables and metabolites as predicting X variables. The absence of sta- and the deproteinated supernatant was subdivided into two halves. tistical outliers was first checked using a principal component One half was freeze‐dried and used for total soluble N isotope analy- analysis to verify that no data point was outside the 99% confi- sis. The other half was used for nitrate extraction as barium nitrate dence Hostelling region. The goodness of the OPLS model was with barium iodide, after Huber, Bernasconi Stefano, Luster, and appreciated using the determination coefficient R2, and the predic- Pannatier‐Graf (2011). Samples were weighed in tin capsules, and tive power was quantified by the cross‐validated determination δ15N values and %N were measured using an elemental analyser coefficient, Q2. The significance of the statistical OPLS model was (Carlo–Erba) coupled to isotope ratio mass spectrometer (Isochrom, tested using a χ2 comparison with a random model (average ± ran-

Elementar) run in continuous flow. All sample batches included stan- dom error), and the associated P value (PCV‐ANOVA) is reported. A dards (glycine, +0.66‰; cysteine, +7.78‰; previously calibrated permutation test was also performed to check the reliability of the against IAEA standards glutamic acid USGS‐40 and caffeine IAEA‐ OPLS model, that is, to verify that at maximal permutation (similar- 600) each 10 samples. When the sample mass was occasionally low ity of permuted dataset tending to zero), Q2 was always negative. (typically <0.08 μmol N sample−1), the δ15N value was corrected using Best discriminating metabolites (biomarkers) were identified using the response curve of the δ15N value to sample mass using standard volcano plots whereby the logarithm of the P value obtained in uni- glycine and cysteine. variate analysis (two‐way analysis of variance [ANOVA]) was plotted

against the rescaled loading (pcorr) obtained in the OPLS. In such a

representation, best biomarkers have both maximal −log(P) and pcorr 2.6 | Statistics values. Univariate analysis of statistical classes (e.g., in Figure 2) was Unless otherwise stated, five replicates were done for all conditions. performed using a two‐way ANOVA (Fisher test), with a threshold Supervised multivariate analysis of metabolomics data was carried of P = 0.01.

87 CUI ET AL. 651

3 | RESULTS 3.3 | Metabolomics pattern in roots

Metabolomics data were analysed using both univariate and multivar- 3.1 | Photosynthesis and carbon balance iate methods (Figures 3 and 4). In roots, there was a clear discrimination of samples using supervised multivariate analysis (O2PLS), with Both root and leaf biomass were affected by K availability and no significant effect of sampling time, the effect of waterlogging along waterlogging (and so was stem biomass, data not shown). Leaf bio- Axis 1 and the effect of K deficiency along Axis 2 (Figure 3a). The mass was maximal at high K (4 mM), intermediate at low K (0.2 mM), −11 O2PLS model generated was highly significant (PCV‐ANOVA = 2.25·10 and low (with a rather small difference between low and high K) upon [K effect] and 6.37·10−19 [waterlogging]), explained 92% of total vari- waterlogging (Figure 1a; mind the very small standard deviations). This ance (R2 = 0.918) and was highly predictive (Q2 = 0.789). The impor- pattern was similar in roots except that low K and high tance of metabolites in the discrimination (ability to represent a K + waterlogging (which had distinguishable effects on leaf biomass) biomarker) was assessed using the combination of univariate and mul- had no effect on root biomass. There was a clear effect of tivariate analyses whereby −log(P) obtained using a two‐way ANOVA waterlogging on %C in total leaf matter (Figure 1b), suggesting that (univariate) was plotted against the loading along the axis of interest waterlogging led to a decrease in other macroelements (N, P, and K). (multivariate; Figure 3b,c). Amongst the best biomarkers of K defi- Net photosynthesis (Figure 1c) showed a complex pattern whereby a ciency were putrescine (increased), and sugar derivatives (threonate, significant effect of K deficiency was visible at the beginning of the xylose, and glucarate) and organic acids (fumarate, malate, glycerate, experiment (in particular, 2 days after the onset of waterlogging), and 3‐phosphoglycerate; decreased; Figure 3b). Amongst the best bio- and then low or high K conditions were similar on average regardless markers of waterlogging were alanine, sugars and polyols, organic of waterlogging (but note the rather substantial variability). This pho- acids (aconitate and citrate), and phosphate (increased), and tosynthetic pattern appeared to be related to stomatal conductance, amino acids (aspartate family) and fumarate (decreased; Figure 3c). with low K conditions having the lowest values (Figure 1d). Leaf night The metabolic ratio glutamine‐to‐glutamate was changed by both respiration was first (2 days after the onset of waterlogging) signifi- waterlogging and K deficiency (Figure 3d), whereas the asparagine‐ cantly affected by K availability, with higher respiration rates at low to‐aspartate ratio was not significantly altered due to considerable K regardless of waterlogging (Figure 1e). This effect lasted for the variability but increased visibly after 14 days under low K (Figure 3 whole experiment except at 7 days where low K under waterlogging e). Both the malate‐to‐fumarate ratio and the relative polyamine con- appeared to be similar to low K conditions. It is worth noting that tent were altered by waterlogging (Figure 3f,g). Nine metabolites although net CO2 assimilation was 10–20% lower in low K conditions appeared to be associated with a significant (P < 0.01) K defi- (Figure 1c), night respiration was higher by 25–60% in the same con- ciency × waterlogging interaction effect (Figure S2 and Table S2): pen- ditions, meaning that low K affected leaf carbon use efficiency, that toses (xylulose and xylitol), aspartate, and 3‐methylglutarate (leucine is, the percentage of fixed C lost by respiration. Similar results were degradation product; more decreased by waterlogging under high K), obtained in a separate experiment where only K availability was and glycerate metabolism (glycerol, glycerate) and polyols (inositol; varied, with a decrease in biomass, leaf elemental K content, less decreased by waterlogging under high K). photosynthesis and stomatal conductance, and an increase in leaf night respiration (of more than 25%) under low K, with all of these effects being reversible after K resupply (Figure S1). 3.4 | Metabolomics pattern in leaves

The same analysis was conducted in leaves using both univariate and multivariate methods. There was a clear discrimination of samples based 3.2 | Elemental contents on metabolomics (Figure 4a), with a good determination coefficient Figure 2 shows elemental contents measured using ICP‐OES in differ- (R2 = 0.924), good predictive power (Q2 = 0.866), and very high signifi- −28 −14 ent organs 14 days after the onset of waterlogging (results are very cance (PCV‐ANOVA = 2.05·10 [K effect] and 1.89·10 [waterlogging]). similar at other sampling times, data not shown). As expected, there There was an effect of sampling time but perfectly aligned with was a clear effect of K conditions on elemental K content in all organs, waterlogging along Axis 1, simply showing that some lag time was neces- with a value of about 1% at low K and up to 5% at high K. Interest- sary to see waterlogging effects on leaf metabolome, typically 7 days ingly, the change in K was compensated for by reciprocal changes in here. The best metabolic biomarkers of K deficiency were putrescine, sodium in roots and stems (Figure 2b,c). It is worth noting that in citramalate, butanoate, and dehydroascorbate (increased), and (pyro) shoots (leaves and stems), calcium, phosphorus, and magnesium con- glutamate, aspartate, and sugar derivatives (decreased; Figure 4b). A tents were altered by both high K and waterlogging, so that the lowest separate experiment where the effect of K only was examined amounts were found under high K combined with waterlogging rather (monofactorial experiment without waterlogging) produced a rather than low K combined with waterlogging. This pattern was roughly sim- similar pattern, whereby putrescine was the best increased marker and ilar in other organs, with the exception of phosphorus. In fact, phos- sugar derivatives and 2‐oxoglutarate (oxoacid associated with gluta- phorus was more abundant in waterlogged compared with control mate) decreased (Figure S3). The best metabolic biomarkers of roots, and stems showed little variation, demonstrating that plant waterlogging were by far sugars and polyols (increased; Figure 4c). The phosphate circulation was impeded under waterlogging, causing a glutamine‐to‐glutamate ratio was decreased by both waterlogging and P‐enrichment in roots and a P‐depletion in leaves. high K, with an additive effect. By contrast, the asparagine‐to‐aspartate

88 652 CUI ET AL.

FIGURE 3 Metabolomics response to K availability and waterlogging (WL) in sunflower roots. (a) Score plot of the O2PLS analysis using K, waterlogging, and time as predicted Y variables. Samples are discriminated along the y axis (K) and x axis (waterlogging). (b) Volcano plot (−log(P value) from two‐way ANOVA versus O2PLS loading) showing best discriminating metabolites (biomarkers) associated with K availability. (c) Volcano plot showing best discriminating metabolites associated with waterlogging. (d–g) Metabolic ratios (relative to that under low K before waterlogging in d–f): glutamine/glutamate, asparagine/aspartate, malate/fumarate, and total polyamines (arginine + putrescine + ornithine + spermidine, in % of total). Abbrevations: Aco, aconitate; Ala, alanine; Asp, aspartate; β‐ala, β‐alanine; Cit, citrate; Cyt, cystathionine; Ery, erythritol; Fum, fumarate; GA, glycerate; Gla, glucarate; Gln, glutamine; Ile, isoleucine; Mal, malate; Malt, maltose; MeGla, methylglucarate; MeGlc, methylglucose; Met, methionine; MG, methylglutarate; Myo, myoinositol; Ono, ononitol; PGA, phosphoglycerate;

Phe, phenylalanine; Pi, phosphate; PI, phospho‐inositol; Put, putrescine; Qui, quinate; Tha, threonate; Thr, threonine; Xol, xylitol; Xyl, xylose. The vertical dashed line in (d–g) separates samples taken before and after the onset of waterlogging. In (e), the asterisk stands for a significantly different value at 14 days under low K compared with other time points [Colour figure can be viewed at wileyonlinelibrary.com] ratio was strongly increased by K deficiency, but this effect was sup- upon waterlogging (Figure S4d,e), whereas waterlogging tended to pressed by waterlogging (Figure 4d). Although the malate‐to‐fumarate increase the spermidine‐to‐putrescine ratio at Day 14 (arrow, Figure ratio did not show significant changes (Figure 4f), polyamines were con- S4f). Univariate analysis of the K deficiency × waterlogging interac- siderably accumulated under low K, but this effect was very clearly tion effect showed that the increase in sugar (such as mannose) suppressed by waterlogging (Figure 4g). A separate analysis was con- caused by waterlogging was suppressed by high K, whereas the accu- ducted at Day 14 (where the waterlogging effects were presumably mulation of amino acids under low K was suppressed by waterlogging maximal) to avoid time as a confounding factor (Figure S4). (Figure S2). Aconitate was by far (P <10−6) the most significant Putrescine, butanoate, citramalate, and dehydroascorbate remained metabolite for the interaction effect: Its accumulation caused by the best increased metabolites (Figure S4b), and sugars were the best waterlogging was suppressed by low K (Figure S2 and Table S2). increased metabolites upon waterlogging. It also showed that waterlogging was associated with an increase in spermidine and man- 3.5 | Correlation between leaf metabolome and nose, and a decrease in phosphate, glycerol, and mannose‐6‐phos- respiration phate. Putrescine tended to be less abundant upon waterlogging. As a result, there was a significant increase of the redox ratio ascor- We took advantage of the subtantial variation in respiration rate bate‐to‐dehydroascorbate and a decrease in sugar phosphorylation (Figure 1) to explore the relationships with metabolites across all

89 CUI ET AL. 653

FIGURE 4 Metabolomics response to K availability and waterlogging (WL) in sunflower leaves. (a) Score plot of the O2PLS analysis using K, waterlogging, and time as predicted Y variables. Samples are discriminated along the y axis (K) and x axis (waterlogging + time). (b) Volcano plot (−log(P value) from two‐way ANOVA versus O2PLS loading) showing best discriminating metabolites (biomarkers) associated with K availability. (c) Volcano plot showing best discriminating metabolites associated with waterlogging. (d–g) Metabolic ratios (relative to that under low K before waterlogging in d–f): glutamine/glutamate, asparagine/aspartate, malate/fumarate, and total polyamines (arginine + putrescine + ornithine + spermidine, in % of total). Abbreviations: Ara, arabinonate; Asp, aspartate; But, butanoate; Citm, citramalate; DHA, dehydroascorbate; F6P, fructose‐6‐phosphate; Fru, fructose; Gal, galactose; Gat, galactonate; Glc, glucose; Glu, glutamate; Lyx, lyxose; Man, mannose; Myo, myoinositol; PG, pyroglutamate; PI, phospho‐inositol; Psi, psicose; Put, putrescine; Sor, sorbose; Tal, talose. The vertical dashed line in (d–g) separates samples taken before and after the onset of waterlogging. A separate analysis with samples collected 2 weeks after the onset of waterlogging is provided in Figure S3a‐c [Colour figure can be viewed at wileyonlinelibrary.com] conditions used here. To do so, we carried out a multivariate analysis 3.6 | Nitrogen content and natural 15N abundance in where the respiration rate was used as a continuous Y variable and leaves metabolites as predicting X variables (Figure S5). The statistical model −18 generated was highly significant (PCV‐ANOVA = 1.03·10 ) showing a To gain further insight into nitrogen metabolism, we analysed elemental strong and predictive covariation of respiration with some metabolites N content and the natural 15N abundance (δ15N) in different leaf frac- (R2 = 0.773; Q2 = 0.682). Of course, it must be kept in mind that this tions (Figure 5): total organic matter, nitrates, heat‐precipitated pro- analysis is potentially associated with confounding factors due to the teins, and total soluble fraction. Soluble organic N was calculated by underlying cause of respiratory changes (K, waterlogging), with typi- mass balance from the difference between total soluble fraction and cally higher respiration rates at low K. The most strongly associating nitrates. Such an analysis was not conducted on roots due to too low metabolites were glycerate and fumarate (low respiration rates) sample size. Waterlogging caused a clear decline in N elemental content and some amino acids, oses (hexoses, pentoses), putrescine, and as well as low K without waterlogging (Figure 5a). The total N demand citramalate (high respiration rates). Interestingly, not all Krebs cycle by leaves was calculated from biomass increment (Figure 1) and %N. intermediates were associated with higher respiration rates: 2‐ Waterlogging led to a drastic decline in the N demand at Day 7 resulting oxoglutarate was not related at all with respiration, whereas malate, from both lower N content and slower growth (Figure 5b). At Day 14, fumarate, and succinate correlated negatively and aconitate and the N demand was higher only at high K without waterlogging. When − citrate correlated positively with respiration. expressed in μmol g 1 DW, there was significantly less leaf nitrate

90 654 CUI ET AL.

FIGURE 5 Time course of N content and δ15N in leaves: elemental N content in % DW (a), average N demand (b), δ15N in total leaf organic matter (c), leaf nitrate (dashed) and proteins (continuous) (d), and soluble organic compounds (dashed) and lipids (continuous) (e). In (b), the average N demand was calculated using the increment of leaf biomass and %N between sampling time points. Time is shown on the x axis in days after the onset of waterlogging. Inset: leaf nitrate content in μmol g−1 DW. Mean ± SD (n = 5). In (a), error bars are small and partly hidden by symbols [Colour figure can be viewed at wileyonlinelibrary.com] under waterlogged conditions (Figure 5, inset) with higher nitrate contents under low K conditions. Under all conditions, nitrate decreased with time. Differences in δ15N in total organic matter were small, with values between −3‰ and 0‰, that is, up to 5.5‰ 15N‐depleted compared with source nitrate (Figure 5c). However, there were important differences between leaf compounds, with nitrates being 15N‐enriched (15–20‰, dashed, Figure 5d), soluble organic nitrogen being relatively depleted (dashed, Figure 5e), and lipids generally slightly enriched (continuous, Figure 5e). Proteins showed significant differences between treatments mostly at Day 7, with low K conditions being isotopically lighter. Because protein N directly derives from amino acids (whereas total organic N contains other compounds such as polyamines), we compared observed nitrates with their expected δ15N just by adding 16‰ (fractionation by nitrate reductase) to the δ15N value of proteins. As expected, datapoints fall relatively close to the 1:1 line (Figure 6), but under high K conditions, nitrates were less enriched than expected, FIGURE 6 Relationship between observed and expected δ15Nin showing that leaf nitrates were isotopically diluted by an influx of 15N‐ leaf nitrate across all conditions. Expected nitrate was calculated as depleted nitrates. Conversely, leaf nitrates under low K + waterlogging 15 δ Nproteins +16‰ (see Figure 5 and main text) [Colour figure can be 15 ‐ conditions were more N enriched than expected, suggesting a viewed at wileyonlinelibrary.com] decline in isotopic dilution thereby enhancing the 15N‐enrichment by nitrate reduction. in leaf phosphorus (Figure 2a) as previously found (Milroy, Bange, & Thongbai, 2009). Here, we further show by metabolomics a decrease

in leaf phosphate (Pi), mannose‐6‐phosphate (Figure S4), fructose‐6‐ 4 | DISCUSSION phosphate, and phospho‐inositol (Figure 4). Because phosphate was amongst the best significantly accumulated molecules in roots (Figure 3c), this suggests that waterlogging simply inhibited root‐to‐ 4.1 | Differential effects of K limitation and shoot circulation of Pi. In addition, leaf metabolism did not reflect oxy- waterlogging in roots and shoots gen shortage under waterlogging conditions (no alanine accumulation Waterlogging caused a general decrease in most macronutrients in for example), and this comes as no surprise because leaves were not roots, except in phosphorus (Figure 2). Variations in K were mirrored under oxygen limitation (in this work, only roots were submerged). by Mg and Na suggesting that they could substitute K for electro- However, waterlogging induced metabolic modifications after some chemical potential homeostatis and play the role of counter‐cation. time (here, visible after 7 days), such as sugar accumulation These variations in Mg and Na differed in leaves where Ca was quan- (Figure 4) and glycerol (Figure S4c), that might have been partly linked titatively more important, K and Ca contents being antagonistic in to the decrease in Pi via limitation of phosphorylating NAD‐dependent leaves under varying KCl fertilization (Dubos, Baron, Bonneau, Flori, glyceraldehyde‐3‐phosphate dehydrogenase activity (Plaxton & & Ollivier, 2017). In contrast to roots, waterlogging caused a decrease Carswell, 1999).

91 CUI ET AL. 655

In roots, waterlogging included metabolic changes typical of oxy- 15N‐nitrates. If we assume that (a) the variations in total leaf nitrates gen shortage, such as alanine accumulation (Limami, Diab, & Lothier, are mostly due to lamina nitrate concentration and that vein water 2014), build‐up of organic acids such as aconitate and citrate (origi- represents ≈25% of total leaf water (Holloway‐Phillips et al., 2016) nating from slow Krebs cycle due to limited NADH reoxidation), and (b) vein water is at ≈5 mM nitrate (Schurr, Gollan, & Schulze, and accordingly a tendency for an increased redox ratio malate‐to‐ 1992) with a δ15N value at, say, 4.5‰—that is, slightly enriched com- fumarate (Figure 3c). There was also an accumulation of sugars and pared with source nitrate due to root reduction, which is rather small polyols and a decrease in amino acids of the aspartate family showing in sunflower (Hocking & Steer, 1983)—then lamina nitrate should have the limited commitment of sugars to phosphenolpyruvate (PEP) pro- been 1‰ to 6‰ 15N‐enriched compared with bulk leaf nitrate. This duction and thus oxaloacetate and malate via PEP carboxylase, con- explains a significant part of the isotopic offset seen in Figure 6 under sistent with previous findings in Arabidopsis (Armengaud et al., high K (blue and orange symbols). By contrast, under low K combined 2009). The general decrease in oxaloacetate and thus aspartate with waterlogging, the isotopic dilution effect would have been lower metabolism also reduced the content in polyamines, because arginine and the imbalance between low input and nitrate consumption would biosynthesis requires aspartate (Slocum, 2005). As expected, the typ- have made 15N‐nitrate richer than expected. In other words, modelled 15 ical putrescine accumulation under K deficiency (Hussain et al., 2011) nitrates (computed as δ Nproteins +16‰) were not sufficiently was observed (Figure 3b). High K conditions led to an accumulation enriched, meaning that accumulated leaf proteins were not of phospho‐glycerate and its derivatives, consistent with a release representative of source nitrate because it was not in the isotopic of the glycolytic restriction (by pyruvate kinase) taking place under steady‐state (becoming progressively richer). This can be simply low K (Figure S2). 3‐Methyl glutarate, a metabolite of leucine turn‐ explained by a proportionally lower input from xylem despite contin- over that is recycled by isoprenoid synthesis, was more abundant at ued N reduction in leaves. high K but suppressed by waterlogging, reflecting the down‐ Isotopic data thus indicate that there was an slight imbalance regulation of volatiles biosynthesis in hypoxic roots (Bracho‐Nunez between xylem nitrate input and nitrate reduction (consumption) in et al., 2012; Copolovici & Niinemets, 2010). leaves, in favour of xylem input under high K and in favour of nitrate Under our conditions, there was a limited effect of high K on alle- consumption under low K + waterlogging. This is consistent with pre- viating the effects of waterlogging on leaf growth or nutrient contents vious data that showed that potassium was essential to facilitate (Figures 1 and 2). However, high K increased photosynthesis— nitrate circulation via a shoot‐root cycle where K+ plays the role of a consistent with previous experiments in cotton (Ashraf et al., 2011)— counter‐cation for nitrate (Casadesús, Tapia, & Lambers, 1995; and decreased respiration and thus was beneficial for the recovery Coskun, Britto, & Kronzucker, 2017; Wang & Wu, 2013). However, of carbon use efficiency under waterlogging. isotopes would not be consistent with an altered nitrate circulation under low K alone, but only under low K combined with waterlogging. This agrees with the decrease in other ions under such circumstances 4.2 | Nitrogen metabolism (Ca2+ and Mg2+; Figure 2) and the concomitant impact on phosphate Our data are consistent with a strong modification of nitrogen metab- (see above). Furthermore, the content in Na+ was unchanged in roots olism and cycling under K deficiency and waterlogging, mostly in but decreased in stems and leaves under low K + waterlogging as com- leaves (in roots, there was no significant change of %N across the pared with low K alone, suggesting that the transport of Na+, which experiment, data not shown). In fact, leaves were less N‐rich under can substitute K+ as a counter‐cation at low K, was inhibited by waterlogging conditions (Figure 5a), consistent with the preferential waterlogging (Figure 2). A decrease of N reduction and assimilation accumulation of sugars (Figure 4) and calculated lower N demand to at low K is unlikely considering the relative increase in glutamine and produce new leaf material (Figure 5b). However, the amount in asparagine, which are primary N assimilates (Figure 4d,e). Further- nitrates did not increase and was only marginally affected by more, low K triggered the synthesis of putrescine that requires waterlogging (Figure 5, inset). The concentration in leaf free nitrate aspartate, glutamate, and glutamine as N donors. The fact that (expressed in mg g−1 DW) was the net result of nitrate “dilution” by waterlogging suppressed polyamine accumulation under low K was organic matter accumulation (growth), input from xylem sap (influx), due to the limitation of the nitrate input and the restriction of and nitrate consumption (reduction). The fact that despite a slower 2‐oxoglutarate synthesis (Figure S2; see below). growth, the leaf free nitrate pool was not affected by waterlogging suggests that there was a change in the balance between nitrate influx and reduction. 4.3 | Leaf respiration and tricarboxylic acid pathway Accordingly, leaf nitrates were not as naturally 15N‐enriched as expected, suggesting an imbalance between the input of 15N‐depleted We find here an increase of leaf night respiration under low K condi- nitrates from xylem and the 15N‐enrichment caused by fractionation tions, and this increase was suppressed by waterlogging. We recog- (14N/15N kinetic isotope effect) during reduction (Rayleigh effect). In nize that these effects could have been due to changes in gene other words, total leaf nitrates were a mixture of xylem (veins) and expression and/or total respiratory capacity such as mitochondrial lamina nitrates, the latter being enriched by the fractionation (16‰) number per cell. However, electron microscope observations have against 15N by nitrate reductase (Ledgard, Woo, & Bergersen, 1985). shown that K‐deficient plants presented no alteration in mitochondrial Under high K, there was a decrease in the 15N‐enrichment thereby number or ultrastructure (Marinos, 1963). Further, little change in causing a proportional increase in the isotopic dilution of lamina gene expression has been found under K deficiency in Arabidopsis

92 656 CUI ET AL.

(Armengaud et al., 2009) so that metabolic regulation (allosteric and alternative pathway involves PEP carboxylase that fixes bicarbonate. direct effects of K) are believed to be crucial. The combination of waterlogging with K deficiency slows down Leaf respiration rate correlated to putrescine and citramalate, two aconitase via NO inhibition and therefore down‐regulates Krebs cycle metabolites also associated with low K (Figure 4). Of course, this cor- decarboxylations and suppresses the stimulation of respiratory CO2 relation is linked to K availability as a causative factor for the variation evolution by K deficiency (for a graphical summary, see Figure S6). in respiration. However, it also reflects the fact that putrescine accumulation was suppressed by waterlogging, indicating a potential link 4.4 | Conclusions and perspectives between respiration and polyamines. In fact, putrescine synthesis pro- Taken as a whole, we show here that both K deficiency and duces CO2, and arginine synthesis (precursor of polyamines) requires waterlogging had a strong effect on nitrogen metabolism and NADPH (Slocum, 2005) produced by pentose phosphates (dark partitioning in sunflower leaves, with changes in the relative impor- metabolism) thereby producing CO2. In addition, leaf respiration was tance of nitrate influx, and nitrogen assimilation to amino acids and highly correlated to citramalate and more loosely to citrate and polyamines. This metabolism is coupled to respiration via carbon skel- aconitate (Figures 4 and S5) and anticorrelated to succinate, malate, eton provision (2‐oxoglutarate) but also via polyamine synthesis that and fumarate (whereas 2‐oxoglutarate did not correlate at all). That requires glutamate and aspartate. We hypothesize here that such an is, acids of the Krebs cycle upstream of isocitrate correlated positively, interaction leads to an aspartate cycle whereby aspartate consump- whereas those downstream anticorrelated, suggesting that some tion generates fumarate, which can be used to regenerate aspartate steps of the cycle between aconitate and succinate were important (Figure S6a). Because (a) aspartate regeneration can be supplemented control points. by succinate oxidation and (b) aspartate can be used to synthesize In fact, both waterlogging and low K influence aconitase and suc- amino acids such as methionine that is required for higher molecular cinate thiokinase and thus the way the Krebs cycle operates. First, an weight polyamines such as spermidine, this aspartate cycle can be eas- inhibition of aconitase can simply explain the accumulation of its inter- ily broken if there is an imbalance caused by, for example, mediate product, aconitate. Such a situation may happen under waterlogging. Further studies are warranted to elucidate this aspect waterlogging conditions because hypoxia leads to increased nitric using isotopic tracing and also assess the usefulness of leaf low‐K bio- oxide (NO) production, which is a potent inhibitor of aconitase (Gupta markers such as citramalate found here. Our study also strongly sug- et al., 2012). In effect, NO inhibits the second step (i.e., conversion of gests that changes in CoA cycling accompany the appearance of aconitate to isocitrate) and not the first step of aconitase catalysis citramalate as a biomarker of low K coupled to polyamines. In addition, (conversion of citrate to aconitate; Castro, Rodriguez, & Radi, 1994; spermidine biosynthesis requires S‐adenosylmethionine (SAM), the Lloyd, Lauble, Prasad, & Stout, 1999; Tórtora, Quijano, Freeman, Radi, production of which by SAM synthase is K+‐dependent (Takusagawa, & Castro, 2007). Second, low K conditions led to a typical accumula- Kamitori, & Markham, 1996). We thus recognize that more work is tion of citramalate, which is one of the best biomarkers. Succinate required on cofactor analysis and quantitation (by LC‐MS) under K thiokinase catalyzes a reversible reaction that has been shown to and/or waterlogging conditions, and this will be addressed in a subse- require K+ for maximal activity (Besford & Maw, 1976; Bush, 1969), quent study. and without K+, the equilibrium favours much less the forward direction, that is, succinyl‐CoA conversion to succinate (Lynn & Guynn, ACKNOWLEDGEMENTS 1978). In addition, pyruvate kinase is a well‐known K+‐dependent enzyme, and low K+ impedes pyruvate formation (Besford & Maw, J. Cui was supported by an Australia Awards PhD Scholarship, and the 1976; Evans, 1963; Nowak & Mildvan, 1972). Therefore, K+ shortage research was supported by the Australian Research Council via a must in principle lead to a drastic change in mitochondrial metabolism. Future Fellowship, under contract FT140100645. The authors thank Experiments in Arabidopsis have indeed suggested that intracellular K+ J. M. Chao de la Barca for advices on biostatistics. We thank W. Chen concentration was impacted by K deficiency (Armengaud et al., 2009). for helping with experiment and critical reading of the manuscript. We Both acetyl‐CoA and succinyl‐CoA metabolism are thus perturbed thank T. Truong (Joint Mass Spectrometry Facility) for advices for GC‐ allowing more CoA cycling by aconitate conversion to itaconyl‐CoA MS analyses. We also thank the RSB Plant Service Team for and citramalate synthesis. The overall balance of aconitate metabolism supporting the work carried out in the glasshouse and Prof. M. Ball + for her kindness and letting us use the Licor tripod. to succinate via itaconate is 2 aconitate + ADP + Pi + 2 NAD → succi- inate + citramalate + ATP + 2 NADH + 3 CO2 (instead of + ORCID aconitate + ADP + Pi + 2 NAD → succinate + ATP + 2 NADH + 2

CO2 via ordinary Krebs cycle reactions). Therefore, this pathway Guillaume Tcherkez https://orcid.org/0000-0002-3339-956X involves, on average, only one conversion to succinate for two aconitate consumed, the other one being converted to citramalate. REFERENCES

In addition, the CO2/ATP ratio is larger, probably explaining why Abadie, C., Blanchet, S., Carroll, A., & Tcherkez, G. (2017). Metabolomics analysis of postphotosynthetic effects of gaseous O2 on primary CO2 evolution increases under K deficiency. K deficiency has also metabolism in illuminated leaves. Functional Plant Biology, 44(9), been shown to be associated with a much higher activity of NADP‐ 929–940. https://doi.org/10.1071/FP16355 dependent malic enzyme (Armengaud et al., 2009) to regenerate pyru- Agüera, E., de la Haba, P., Fontes, A. G., & Maldonado, J. M. (1990). Nitrate vate and thus compensate for the decrease in pyruvate kinase activity. and nitrite uptake and reduction by intact sunflower plants. Planta,

However, this should not in principle increase CO2 efflux because this 182(1), 149–154. https://doi.org/10.1007/BF00239997

93 CUI ET AL. 657

Anschütz, U., Becker, D., & Shabala, S. (2014). Going beyond nutrition: sustainable yields (p. 382). Singapore: Potash and Phosphate Institute Regulation of potassium homoeostasis as a common denominator of of Canada (ESEAP). plant adaptive responses to environment. Journal of Plant Physiology, Freeman, G. G. (1967). Studies on potassium nutrition of plants. II.—Some 171(9), 670–687. https://doi.org/10.1016/j.jplph.2014.01.009 effects of potassium deficiency on the organic acids of leaves. Journal Armengaud, P., Sulpice, R., Miller, A. J., Stitt, M., Amtmann, A., & Gibon, Y. of the Science of Food and Agriculture, 18(12), 569–576. https://doi. (2009). Multilevel analysis of primary metabolism provides new insights org/10.1002/jsfa.2740181205 into the role of potassium nutrition for glycolysis and nitrogen assimi- Gautam, P., Lal, B., Tripathi, R., Shahid, M., Baig, M. J., Maharana, S., … lation in arabidopsis roots. Plant Physiology, 150(2), 772–785. https:// Nayak, A. K. (2016). Beneficial effects of potassium application in doi.org/10.1104/pp.108.133629 improving submergence tolerance of rice (Oryza sativa L.). Environmen- Ashraf, M. A., Ahmad, M. S. A., Ashraf, M., Al‐Qurainy, F., & Ashraf, M. Y. tal and Experimental Botany, 128,18–30. https://doi.org/10.1016/j. (2011). Alleviation of waterlogging stress in upland cotton (Gossypium envexpbot.2016.04.005 hirsutum L.) by exogenous application of potassium in soil and as a foliar spray. Crop & Pasture Science, 62(1), 25–38. https://doi.org/ Gupta, K. J., Shah, J. K., Brotman, Y., Jahnke, K., Willmitzer, L., Kaiser, W. … 10.1071/CP09225 M., Igamberdiev, A. U. (2012). Inhibition of aconitase by nitric oxide leads to induction of the alternative oxidase and to a shift of metabo- ‐ Barrett Lennard, E. G., & Shabala, S. N. (2013). The waterlogging/salinity lism towards biosynthesis of amino acids. Journal of Experimental — ‐ interaction in higher plants revisited focusing on the hypoxia induced Botany, 63(4), 1773–1784. https://doi.org/10.1093/jxb/ers053 disturbance to K+ homeostasis. Functional Plant Biology, 40(9), 872–882. Hasanuzzaman, M., Bhuyan, H. M., Nahar, K., Hossain, S. M., Mahmud, A. J., Hossen, S. M., … Fujita, M. (2018). Potassium: A vital regulator of Besford, R. T., & Maw, G. A. (1976). Effect of potassium nutrition on some plant responses and tolerance to abiotic stresses. Agronomy, 8(3), 1–29. enzymes of the tomato plant. Annals of Botany, 40(3), 461–471. https://doi.org/10.1093/oxfordjournals.aob.a085155 Hochmuth, G. J. (1994). Efficiency ranges for nitrate‐nitrogen and potassium for vegetable petiole sap quick tests. HortTechnology, 4(3), Bracho‐Nunez, A., Knothe, N. M., Costa, W. R., Maria Astrid, L. R., Kleiss, 218–222. B., Rottenberger, S., … Kesselmeier, J. (2012). Root anoxia effects on physiology and emissions of volatile organic compounds (VOC) under Hocking, P. J., & Steer, B. T. (1983). Distribution of nitrogen during growth short‐and long‐term inundation of trees from Amazonian floodplains. of sunflower (Helianthus annuus L.). Annals of Botany, 51(6), 787–799. Springerplus, 1(1), 9. https://doi.org/10.1186/2193‐1801‐1‐9 https://doi.org/10.1093/oxfordjournals.aob.a086530 Bush, L. P. (1969). Influence of certain cations on activity of succinyl CoA Holloway‐Phillips, M., Cernusak Lucas, A., Barbour, M., Song, X., synthetase from tobacco. Plant Physiology, 44(3), 347–350. https:// Cheesman, A., Munksgaard, N., … Farquhar Graham, D. (2016). Leaf doi.org/10.1104/pp.44.3.347 vein fraction influences the Péclet effect and 18O enrichment in leaf Cakmak, I. (2005). The role of potassium in alleviating detrimental effects water. Plant, Cell & Environment, 39(11), 2414–2427. https://doi.org/ of abiotic stresses in plants. Journal of Plant Nutrition and Soil Science, 10.1111/pce.12792 – 168(4), 521 530. https://doi.org/10.1002/jpln.200420485 Huber, B., Bernasconi Stefano, M., Luster, J., & Pannatier‐Graf, E. (2011). A Carroll, A. J., Badger, M. R., & Harvey Millar, A. (2010). The new isolation procedure of nitrate from freshwater for nitrogen and MetabolomeExpress Project: Enabling web‐based processing, analysis oxygen isotope analysis. Rapid Communications in Mass Spectrometry, and transparent dissemination of GC/MS metabolomics datasets. 25(20), 3056–3062. https://doi.org/10.1002/rcm.5199 ‐ ‐ BMC Bioinformatics, 11(1), 376. https://doi.org/10.1186/1471 2105 Hussain, S. S., Ali, M., Ahmad, M., & Siddique, K. H. M. (2011). Polyamines: ‐ 11 376 Natural and engineered abiotic and biotic stress tolerance in plants. + − Casadesús, J., Tapia, L., & Lambers, H. (1995). Regulation of K and NO3 Biotechnology Advances, 29(3), 300–311. https://doi.org/10.1016/j. fluxes in roots of sunflower (Helianthus annuus) after changes in light biotechadv.2011.01.003 intensity. Physiologia Plantarum, 93(2), 279–285. https://doi.org/ Jones, L. H. (1961). Some effects of potassium deficiency on the metabo- 10.1111/j.1399‐3054.1995.tb02229.x lism of the tomato plant. Canadian Journal of Botany, 39(3), 593–606. Castro, L., Rodriguez, M., & Radi, R. (1994). Aconitase is readily inactivated https://doi.org/10.1139/b61‐048 by peroxynitrite, but not by its precursor, nitric oxide. Journal of Biolog- Jones, L. H. (1966). Carbon‐14 studies of intermediary metabolism in ical Chemistry, 269(47), 29409–29415. potassium‐deficient tomato plants. Canadian Journal of Botany, 44(3), Copolovici, L., & Niinemets, Ü. (2010). Flooding induced emissions of vol- 297–307. https://doi.org/10.1139/b66‐036 atile signalling compounds in three tree species with differing ‐ waterlogging tolerance. Plant, Cell & Environment, 33(9), 1582–1594. Jung, J. Y., Shin, R., & Schachtman, D. P. (2009). Ethylene mediates response and tolerance to potassium deprivation in arabidopsis. The Coskun, D., Britto, D. T., & Kronzucker, H. J. (2017). The nitrogen–potas- Plant Cell, 21(2), 607–621. https://doi.org/10.1105/tpc.108.063099 sium intersection: Membranes, metabolism, and mechanism. Plant, Cell & Environment, 40(10), 2029–2041. Jwakyung, S., Yeonkyu, S., Yejin, L., Seongsoo, K., Sangkeun, H., Krishnan, H. B., & Taek‐Keun, O. (2015). Compositional changes of selected Demidchik, V. (2014). Mechanisms and physiological roles of K+ efflux amino acids, organic acids, and soluble sugars in the xylem sap of N, from root cells. Journal of Plant Physiology, 171(9), 696–707. https:// P, or K‐deficient tomato plants. Journal of Plant Nutrition and Soil Sci- doi.org/10.1016/j.jplph.2014.01.015 ence, 178(5), 792–797. Dubos, B., Baron, V., Bonneau, X., Flori, A., & Ollivier, J. (2017). High soil calcium saturation limits use of leaf potassium diagnosis when KCl is Kaiser, D. E., Rosen, C. J., & Lamb, J. A. (2016). Potassium for crop produc- ‐ applied in oil palm plantations. Experimental Agriculture,1–11. tion. Nutrient management: University of Minnesota Extension. Dwivedi, S. K., Kumar, S., Bhakta, N., Singh, S. K., Rao, K. K., Mishra, J. S., & Ledgard, S., Woo, K., & Bergersen, F. (1985). Isotopic fractionation during Singh, A. K. (2017). Improvement of submergence tolerance in rice reduction of nitrate and nitrite by extracts of spinach leaves. Functional through efficient application of potassium under submergence‐prone Plant Biology, 12(6), 631–640. rainfed ecology of Indo‐Gangetic Plain. Functional Plant Biology, 44(9), Limami, A. M., Diab, H., & Lothier, J. (2014). Nitrogen metabolism in plants 907–916. https://doi.org/10.1071/FP17054 under low oxygen stress. Planta, 239(3), 531–541. https://doi.org/ Evans, H. J. (1963). Effect of potassium & other univalent cations on activ- 10.1007/s00425‐013‐2015‐9 ity of pyruvate kinase in pisum sativum. Plant Physiology, 38(4), Lloyd, S. J., Lauble, H., Prasad, G. S., & Stout, C. D. (1999). The mechanism – 397 402. https://doi.org/10.1104/pp.38.4.397 of aconitase: 1.8 Å resolution crystal structure of the S642A:citrate Foster, H. L. (2002). Assessment of oil palm requirements. In T. H. complex. Protein Science, 8(12), 2655–2662. https://doi.org/10.1110/ Fairhurst, & R. Härdter (Eds.), The oil palm, management for large and ps.8.12.2655

94 658 CUI ET AL.

Lynn, R., & Guynn, R. W. (1978). Equilibrium constants under physiological Takusagawa, F., Kamitori, S., & Markham, G. D. (1996). Structure and function conditions for the reactions of succinyl coenzyme A synthetase and the of s‐adenosylmethionine synthetase: Crystal structures of S‐ hydrolysis of succinyl coenzyme A to coenzyme A and succinate. adenosylmethionine synthetase with ADP, BrADP, and PPi at 2.8 Å resolu- Journal of Biological Chemistry, 253(8), 2546–2553. tion. Biochemistry, 35(8), 2586–2596. https://doi.org/10.1021/bi952604z Marinos, N. G. (1963). Studies on submicroscopic aspects of mineral deficiencies. Tórtora, V., Quijano, C., Freeman, B., Radi, R., & Castro, L. (2007). Mito- II. Nitrogen, potassium, sulfur, phosphorus, and magnesium deficiencies in chondrial aconitase reaction with nitric oxide, S‐nitrosoglutathione, the shoot apex of barley. American Journal of Botany, 50(10), 998–1005. and peroxynitrite: Mechanisms and relative contributions to aconitase https://doi.org/10.1002/j.1537‐2197.1963.tb06582.x inactivation. Free Radical Biology and Medicine, 42(7), 1075–1088. Milroy, S. P., Bange, M. P., & Thongbai, P. (2009). Cotton leaf nutrient concentra- https://doi.org/10.1016/j.freeradbiomed.2007.01.007 tions in response to waterlogging under field conditions. Field Crops Wang, C., Fan, L., Gao, H., Wu, X., Li, J., Lv, G., & Gong, B. (2014). Poly- Research, 113(3), 246–255. https://doi.org/10.1016/j.fcr.2009.05.012 amine biosynthesis and degradation are modulated by exogenous ‐ ‐ ‐ Nowak, T., & Mildvan, A. S. (1972). Nuclear magnetic resonance studies of gamma aminobutyric acid in root zone hypoxia stressed melon roots. – the function of potassium in the mechanism of pyruvate kinase. Bio- Plant Physiology and Biochemistry, 82,1726. https://doi.org/ chemistry, 11(15), 2819–2828. https://doi.org/10.1021/bi00765a014 10.1016/j.plaphy.2014.04.018 Okamoto, S. (1966). Effect of mineral nutrition on metabolic change Wang, M., Zheng, Q., Shen, Q., & Guo, S. (2013). The critical role of potas- induced in crop plant roots (III). Soil Science & Plant Nutrition, 12(4), sium in plant stress response. International Journal of Molecular Sciences, – 13–17. https://doi.org/10.1080/00380768.1966.10431948 14(4), 7370 7390. https://doi.org/10.3390/ijms14047370 ‐ Okamoto, S. (1967). Effects of potassium nutrition on the glycolysis and Wang, Y., & Wu, W. H. (2013). Potassium transport and signaling in higher – the krebs cycle in taro plants. Soil Science & Plant Nutrition, 13(5), plants. Annual Review of Plant Biology, 64(1), 451 476. https://doi.org/ ‐ ‐ ‐ 143–150. https://doi.org/10.1080/00380768.1967.10431989 10.1146/annurev arplant 050312 120153 Okamoto, S. (1968). The respiration in the roots of broad bean and barley Zeng, F., Konnerup, D., Shabala, L., Zhou, M., Colmer Timothy, D., Zhang, G., & under a moderate potassium deficiency. Soil Science & Plant Nutrition, Shabala, S. (2014). Linking oxygen availability with membrane potential 14(5), 175–182. https://doi.org/10.1080/00380768.1968.10432762 maintenance and K+ retention of barley roots: Implications for waterlogging stress tolerance. Plant, Cell & Environment, 37(10), 2325–2338. Orchard, P. W., Jessop, R. S., & So, H. B. (1986). The response of sorghum — and sunflower to short‐term waterlogging: IV. Water and nutrient Zörb, C., Senbayram, M., & Peiter, E. (2014). Potassium in agriculture – uptake effects. Plant and Soil, 91(1), 87–100. https://doi.org/ Status and perspectives. Journal of Plant Physiology, 171(9), 656 669. 10.1007/BF02181821 https://doi.org/10.1016/j.jplph.2013.08.008 Orchard, P. W., & So, H. B. (1985). The response of sorghum and sunflower to short‐term waterlogging. Plant and Soil, 88(3), 407–419. https://doi. SUPPORTING INFORMATION org/10.1007/BF02197497 Phukan, U. J., Mishra, S., & Shukla, R. K. (2016). Waterlogging and submer- Additional supporting information may be found online in the gence stress: Affects and acclimation. Critical Reviews in Biotechnology, Supporting Information section at the end of the article. 36(5), 956–966. https://doi.org/10.3109/07388551.2015.1064856

Plaxton, W., & Carswell, M. C. (1999). Metabolic aspects of the phosphate Scheme S1. Diagram of the experimental design. starvation response in plants. In H. Lerner (Ed.), Plant responses to Table S1. Chemical composition of nutrient solutions used in this environmental stresses, from phytohormones to genome reorganization (pp. 1997–2017). Boca Raton: CRC Press. study. Rawat, J., Sanwal, P., & Saxena, J. (2016). Potassium and its role in sustain- Table S2. Metabolites that are significant (P < 0.01) in the K availability able agriculture. In V. S. Meena, B. R. Maurya, J. P. Verma, & R. S. x waterlogging effect with a univariate analysis using a 2‐way ANOVA Meena (Eds.), Potassium solubilizing microorganisms for sustainable agri- (regardless of time). culture (pp. 235–253). New Delhi: Springer India. https://doi.org/ 10.1007/978‐81‐322‐2776‐2_17 Figure S1. Effect of K availability on growth and photosynthesis in Rocha, M., Licausi, F., Araújo, W. L., Nunes‐Nesi, A., Sodek, L., Fernie, A. R., sunflower in a separate experiment where no waterlogging was & van Dongen, J. T. (2010). Glycolysis and the tricarboxylic acid cycle applied. are linked by alanine aminotransferase during hypoxia induced by Figure S2. Significant metabolites for the K availability × waterlogging waterlogging of Lotus japonicus. Plant Physiology, 152(3), 1501–1513. ‐ https://doi.org/10.1104/pp.109.150045 interaction effect using a 2 way ANOVA. Schurr, U., Gollan, T., & Schulze, E. D. (1992). Stomatal response to drying Figure S3. Metabolomics pattern of sunflower leaves under varying K soil in relation to changes in the xylem sap composition of Helianthus availability in a separate experiment where no waterlogging was annuus. II. Stomatal sensitivity to abscisic acid imported from the xylem applied. sap. Plant, Cell & Environment, 15(5), 561–567. https://doi.org/ Figure S4. Additional metabolomics data on leaves. 10.1111/j.1365‐3040.1992.tb01489.x Figure S5. Correlation analysis of leaf night respiration with Shabala, S., & Pottosin, I. (2014). Regulation of potassium transport in plants under hostile conditions: Implications for abiotic and biotic metabolomics. stress tolerance. Physiologia Plantarum, 151(3), 257–279. https://doi. Figure S6. Summary of metabolic steps discussed in main text, show- org/10.1111/ppl.12165 ing (a) regulated points under K deficiency and waterlogging, and (b) Sharma, P. K., Sharma, S. K., & Choi, I. Y. (2010). Individual and combined details of citramalate biosynthetic pathway. effects of waterlogging and alkalinity on yield of wheat (Triticum aestivum L.) imposed at three critical stages. Physiology and Molecular Biology of Plants, 16(3), 317–320. https://doi.org/10.1007/s12298‐010‐0027‐5 Shin, R. (2014). Strategies for improving potassium use efficiency in plants. How to cite this article: Cui J, Abadie C, Carroll A, Lamade E, – Molecules and Cells, 37(8), 575 584. https://doi.org/10.14348/ Tcherkez G. Responses to K deficiency and waterlogging inter- molcells.2014.0141 act via respiratory and nitrogen metabolism. Plant Cell Environ. Slocum, R. D. (2005). Genes, enzymes and regulation of arginine biosyn- – thesis in plants. Plant Physiology and Biochemistry, 43(8), 729–745. 2019;42:647 658. https://doi.org/10.1111/pce.13450 https://doi.org/10.1016/j.plaphy.2005.06.007

95 6XSSOHPHQWDU\PDWHULDO

S WL

LK+WLorHK+WL Growth under LKorHKfortwo weeks LK orHK (nowaterlogging) Days:02714

Foursamplings 6FKHPH 6 Diagram of the experimental design. Arrows: S, sowing; WL, onset of waterlogging. Triangles: sampling dates.

7DEOH6Chemical composition of nutrient solutions used in this study. All concentrations are given in mM. The KCl concentration used depended on the desired treatment (high, medium, or low K). Note that final nitrate concentration was 12 mM.

&RPSRXQG &RQFHQWUDWLRQ PPRO/ KCl 4/1/0.2

NaNO3 4

Ca(NO3)2 4

MgSO4 1.5

NaH2PO4 1.33 Fe-Na EDTA 0.11 –3 MnSO4 10 –3 ZnSO4 10 –3 CuSO4 10

H3BO3 0.05 –4 Na2MoO4 5 10 –4 Co(NO3)2 2 10

96 b elem (b) obtained attheendofexperiment,i.e.4weeksaftersowing.(a)plantbiomass, ingDWper plant; at 4mM blue).All then2weeksresupply data shownherewere (light low Kfortwoweeksand VHSDUDWH D mM,from sowing: lowK(0.2 blue-green),highK(4mM, mM, green),medium K(1 LQ blue) and navy VXQIORZHU LQ SKRWRV\QWKHVLV DQG JURZWK H[SHULPHQWZKHUHQRZDWHUORJJLQJZDVDSSOLHG RQ DYDLODELOLW\ . RI (IIHFW 6 )LJXUH conductance for water vapour (at400 for watervapour conductance high K(navy in darkness(at25°C).NotethereversibilityofK Biomass (g plant-1) 1.0 0.5 0.0 0 1 2 3 4 5 6 7 D Root Shoot Low K KM KLK+HK HK MK LK a ental contentin K determinedental byICP-OES;(c)and(d)netCO b Medium K Medium c blue) andlow K+resupply(lightblue). c

Leaf K content (mg g-1) High K High 10 20 30 40 0 E KM KLK+HK HK MK LK a Low K + resupply b c c 97 mol mol mol A (mol m-2 s-1) 10 15 20 25 30 35 0 5 F KM KLK+HK HK MK LK a -1 ab CO deficiency: there is nosignificadeficiency: . Four nutrientconditions wereusedduringgrowth b 2 , 21%O ab

-2 -1 gh (mol m s ) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 2 , saturatinglight,25°C);(e)respiration G KM KLK+HK HK MK LK a b c 2 assimilation andstomatal c

-2 -1 nt differencebetween Rn (mol m s ) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 H KM KLK+HK HK MK LK a ab b ab )LJXUH 6 6LJQLILFDQW PHWDEROLWHV IRU WKH . DYDLODELOLW\ u ZDWHUORJJLQJ LQWHUDFWLRQ HIIHFW XVLQJ D ZD\ $129$. Heat map representation with hierarchical clustering (Pearson correlation) on left. Time (in days after the onset of waterlogging) is indicated above growth conditions (LK, low K; HK, high K; WL, waterlogging). Colors used here are the same as in figures in main text. To facilitate reading, significant metabolites are also listed in Table S2.

98 7DEOH 6. Metabolites that are significant (P < 0.01) in the K availability u waterlogging effect with a univariate analysis using a 2-way ANOVA (regardless of time). Note that in Fig. S4, some metabolites appear two times because they are represented by several analytes (derivatives) in the derivatized extract. In such a case, we give here the average of the P-value. Metabolites are given by alphabetical order.

/HDYHV 5RRWV Metabolite –log(P-value) Metabolite –log(P-value) 2-oxoglutarate 5.31 3-methylglutarate 2.43 2-piperidine carboxylate 2.12 3-phosphoglycerate 2.27 Aconitate 6.95 Aspartate 3.05 Aspartate 2.55 Glycerate 2.31 Butanoate 2.19 Glycerol-2-pyranoside 3.33 Citramalate 2.37 Myoinositol 3.58 Galactarate 6.94 Phospho-inositol 2.70 Gluconate 4.65 Xylitol 2.17 Glutarate 4.89 Xylulose 5.35 Isoleucine 3.83 Leucine 2.63 Lysine 2.52 Mannose 2.73 Oxalate 3.39 Proline 2.26 Succinate 2.58 Tryptophan 3.35 Tyrosine 3.57

99 )LJXUH 6 0HWDERORPLFV SDWWHUQ RI VXQIORZHU OHDYHV XQGHU YDU\LQJ . DYDLODELOLW\ LQ D VHSDUDWH H[SHULPHQW ZKHUH QR ZDWHUORJJLQJ ZDV DSSOLHG: (a) Score plot of the OPLS analysis showing the discrimination of samples (low K, green; medium K, blue-green; High K, navy blue; low K for two weeks and then resupply, light blue). The OPLS discriminated samples following K level on axis 1 (sampling time on axis 2). It was associated with a good predictive value (Q² = 0.77) and the statistical model was highly –9 significant (PCV-ANOVA = 4.5 10 ). (b) Volcano plot showing log of the P-value in univariate analysis (one- way ANOVA) against the loading of the K effect in OPLS. In this representation, metabolites on the right hand side are accumulated under K deficiency. Best biomarkers are: putrescine (Put), pinacol (Pin) and fructose (Fru) (increased under K deficiency) and glutarate (Glt), xylulose (Xul), threonate (Tha), 2- oxoglutarate (2OG) and malate (Mal) (decreased).

100 )LJXUH6. $GGLWLRQDOPHWDERORPLFVGDWDRQOHDYHV. (a-c) Statistics carried out separately after 2 weeks waterlogging, to look at waterlogging-specific effects: score plot of the O2PLS (a), volcano plot showing best discriminating metabolites for K availability (b) and waterlogging (c). (d-f) Relative metabolic ratios (value under low K before the onset of waterlogging fixed at 1): ascorbate-to-dehydroascorbate (ascorbate reduction index) (d); sugars-to-phosphosugars (e); and spermidine-to-putrescine (f). Same abbreviations as in Fig. 4. Gol, glycerol; M6P, mannose-6-phosphate; Pi, phosphate. Asterisks show significance for K availability (P<0.01): ascorbate-to-DHA after one week (d); sugars-to-phosphosugars under waterlogging (pairwise comparisons at fixed time). The vertical dashed line separates samples taken before and after the onset of waterlogging. Note the opposite effect of waterlogging on spermidine and putrescine and thus the reversion of spermidine-to-putrescine ratio after 2 weeks: arrows in (c) and (f).

101 )LJXUH6&RUUHODWLRQDQDO\VLVRIOHDIQLJKWUHVSLUDWLRQZLWKPHWDERORPLFV. Leaf respiration rate in darkness (Rn) was used as a continuous Y variable in an OPLS model, with metabolites as X variables. (a) Score plot showing the range of respiration rates along axis 1 with the color scale. (b) Same score plot as in (a) but colored to show the distribution of respiration rates by treatment. Note the predominance of low K (without waterlogging) conditions (green) in high respiration rates (right hand side). (c) Relationship between respiration rate predicted by the OPLS model (x axis) and observed rates (y axis). The dashed line represents the 1:1 line. The correlation coefficient of this relationship is given by the R² of the OPLS model and was 0.77. (d) Volcano plot showing the relative correlation coefficient between Rn and the metabolite of interest (VIP) against the loading of the OPLS model. The threshold corresponding to VIP = 1.5 corresponds to R = r0.5 (dotted horizontal line). Metabolites above the threshold are listed on bottom. Metabolites are colored by biochemical class as shown in legend. In particular, note the wide distribution of amino acids (green) and hexoses (purple). TCAP intermediates are labelled in black with arrows. Aspartate and glutamate are shown in blue. Note the tendency for an increase in citrate + aconitate (right) and decrease in fumarate + malate + succinate (left) when respiration increases. 2-oxoglutarate (2OG) had no influence at all, with a loading value pcorr very close to zero. Aconitate is associated with two points corresponding to two analytes (GC-MS derivatives). -18 The OPLS model of respiration rate was very predictive (Q² = 0.68) and highly significant (PCV-ANOVA = 10 ).

102 Spermidine (a) AspartateͲderived amino acids (Threonine,Asparagine,Methionine) Butanoate

GLYCOLYSIS SAM STP Aspartate Putrescine CO 2Glu 2 2OG MTA CO2 NADPH Arginine Ornithine Ͳ Glu 2OG HCO3 OPPP Succinate PEP OAA CO2 Fumarate

SuccinylͲCoA Ͳ NO3 CO2 CO2 N 2OG Glu,Gln Malate CO2

Pyruvate AcetylͲCoA Citrate Aconitate Isocitrate

CO2 CO2 Itaconate Citramalate ACETOACETATE PATHWAY

Leucine,Methylglutarate,Pinacol (b)

6 CO2 7 CO2 5 5 2 CitrateAconitateIsocitrate 2OGSuccinylͲCoA Succinate NAD+ NADH NAD+ NADH ADP ATP 1 CoA Pi CoA CO2 2 3 Itaconate ItaconylͲCoA CitramalylͲCoA Citramalate 4 ATP ADP H2O CoA Pi

Overall sum:

+ 2aconitate+ADP+Pi +2NAD succinate +citramalate +ATP+2NADH+3CO2

1. Aconitatedecarboxylase 2. SuccinylͲCoA ligase(succinate thiokinase) 3. ItaconylͲCoA hydratase 4. Citramalate CoA transferase (citramalate lyase) 5. Aconitase 6. Isocitrate dehydrogenase (NAD) 7. 2Ͳoxoglutaratedehydrogenase

)LJXUH66XPPDU\RIPHWDEROLFVWHSVGLVFXVVHGLQPDLQWH[W, showing D regulated points under K deficiency (red circles) and waterlogging (blue circles), and E details of citramalate biosynthetic pathway. In (a), the pentose phosphate pathway in simply mentioned in grey. The dotted line starting from arginine and leading to putrescine stands for the arginine decarboxylase pathway (mostly in Arabidopsis which lacks ornithine decarboxylase). 2OG, 2-oxoglutarate; MTA, S- methylthioadenosine; PEP, phosphoenolpyruvate; OAA, oxaloacetate; SAM, S-adenosylmethionine; STP, S-adenosyl methylaminopropane. Steps associated with a decarboxylation are shown with CO2 in blue.

103 Chapter 4 - Metabolic responses to potassium availability and waterlogging reshape respiration and carbon use efficiency in oil palm

4.1 Motivation

As already stated in the Introduction, there is currently a gap of knowledge on biochemical aspects of oil palm physiology, despite the fact that oil palm is the most important oil-producing crop in the world. In practice, oil palm agroforestry is associated with specific constraints, such as a particularly high need in K fertilization, and the occurrence of environmental stresses. K fertilization (as well as fertilization with other nutrients such as N and P) is an important topic for current agronomical projects in the field, typically in projects involving Indonesian oil palm growers and the CIRAD (France). An important issue of K fertilization (still not resolved) is to find a good bio- indicator (bio-index) that would respond rapidly to the demand, as determined by development, growth and environment. This would allow proper fertilization monitoring in the field and thus avoid applying excess or insufficient K fertilization levels. The motivation of the present work was thus be an aid in this long-term objective, by providing a metabolic characterization of oil palm.

4.2 Context

The combination of stresses, K deficiency and waterlogging, is of importance in the field but has never been studied, at least in oil palm. In fact, a significant proportion of oil- palm cultivated regions are close to rivers or installed on peat areas, and thus subjected to periodic flooding. Therefore, the exploration of the effects of varying K availability combined with waterlogging has been seen as a priority. In this Chapter, I take advantage of omics techniques (in addition to classical eco-physiological measurements) to highlight metabolic effects of the combination of low or high K and waterlogging, using saplings grown in the greenhouse. It is worth noting that the present work was conducted in the greenhouse with plants grown from germinated seeds in order to have a perfect control of nutritional conditions.

104 4.3 Key results and implications

In addition to expected changes in photosynthesis (stomatal conductance and production of the photosynthetic machinery) and elemental content (decrease in K compensated for by other cations), I find a considerable effect on respiration, with (i) a huge increase in the respiration rate under K deficiency, (ii) changes in enzymes and metabolites involved in catabolism (including via a new alternative pathway I propose here, based on parapyruvate aldolase), leading to (iii) changes in total respiration loss and carbon use efficiency. I also demonstrate that the availability of K (a critical element of oil palm agroforestry) is not associated with a linear response of metabolism such that there is a rather narrow optimal window of K availability. Outside this window, metabolic symptoms of oxidative stress could be detected. These results are highly relevant for the oil palm and plant science community because they provide a comprehensive view of the metabolic reorchestration under a stress combination.

4.4 Published paper

105 Research

Metabolic responses to potassium availability and waterlogging reshape respiration and carbon use efﬁciency in oil palm

Jing Cui1, Marlene Davanture2, Michel Zivy2, Emmanuelle Lamade3 and Guillaume Tcherkez1 1Research School of Biology, ANU Joint College of Sciences, Australian National University, Canberra, ACT 2601, Australia; 2Plateforme d’Analyse de Proteomique Paris-Sud-Ouest (PAPPSO), GQE Le Moulon, INRA, Univ. Paris-Sud, CNRS, AgroParisTech, Universite Paris-Saclay, Ferme du Moulon, Gif-sur-Yvette 91190, France; 3UPR34 Performance des systemes de culture des plantes perennes, Departement PERSYST, Centre de Cooperation Internationale en Recherche Agronomique pour le Developpement (CIRAD), Montpellier 34398, France

Summary Author for correspondence: Oil palm is by far the major oil-producing crop on the global scale, with c. 62 Mt oil pro- Guillaume Tcherkez duced each year. This species is a strong potassium (K)-demanding species cultivated in Tel: +61 2 6125 0381 regions where soil K availability is generally low and waterlogging due to tropical heavy rains Email: [email protected] can limit further nutrient absorption. However, the metabolic effects of K and waterlogging Received: 1 December 2018 have never been assessed precisely. Accepted: 12 February 2019 Here, we examined the metabolic response of oil palm saplings in the glasshouse under controlled conditions (nutrient composition with low or high K availability, with or without New Phytologist (2019) waterlogging), using gas exchange, metabolomics and proteomics analyses. doi: 10.1111/nph.15751 Our results showed that both low K and waterlogging have a detrimental effect on photosynthesis but stimulate leaf respiration, with differential accumulation of typical metabolic Key words: metabolomics, oil palm, intermediates and enzymes of Krebs cycle and alternative catabolic pathways. In addition, we potassium, proteomics, respiration, found a strong relationship between metabolic composition, the rate of leaf dark respiration, waterlogging. and cumulated respiratory loss. Advert environmental conditions (here, low K and waterlogging) therefore have an enormous effect on respiration in oil palm. Leaf metabolome and proteome appear to be good predictors of carbon balance, and open avenues for cultivation biomonitoring using functional genomics technologies.

Introduction Major oil palm cultivated regions include Indonesia and Malaysia, as well as South American countries (Brazil, Peru, and Oil palm (Elaeis guineensis Jacq.) is a world-wide cultivated, Columbia) and Congo where plantations are mostly found at low oil-producing crop presently representing c. 21 Mha (global elevation and therefore susceptible to flooding under wet tropical cultivated surface area) and 300 Mt raw products (global fruit climate. These areas are also often nutrient-poor, have a low production) (www.fao.org). Despite a much lower total vege- potassium-to-calcium ratio, or have a low potassium exchange tation biomass on a surface area basis, oil palm plantations capability, and therefore require expensive mineral fertilisation assimilate more CO2 than tropical forests due to rapid growth with potassium (K) mostly in the form of potassium chloride and fruit production (for a recent review, see Paterson & (KCl). In addition, apparent potassium use is relatively low Lima, 2018). However, oil palm cultivation practices are asso- (c. 35 g oil g 1 K) (Foster, 2002) and physiological efficiency (in- ciated with serious environmental effects such as CO2 produc- crease in fresh fruit bunch per increase in K uptake) is also low (c. tion and it has been recently estimated that deforestation 35%) compared with nitrogen (c. 55%) (Tohiruddin et al., associated with oil palm cultivation is responsible for 6–17% 2007). Therefore, it is common practice to fertilise with consider- 1 of anthropogenic CO2 release (Baccini et al., 2012). Also, oil able amounts of KCl (from 0.3 to 3 kg tree for standard appli- generation is associated with low carbon use efficiency (CUE), cations, up to 8 kg tree 1 y 1 in agronomical trials) even though that is, a low efficiency of conversion of source carbon into tropical soils contain background K+ (partly nonexchangeable) effective oil (0.34 g C g 1 C), simply because lipid synthesis (Ollagnier et al., 1987). With fertilisation, K+ capacity in pore requires considerable amounts of energy and leads to high water may reach c. 4 mM across the first soil layer explored by CO2 loss by fruit respiration (Amthor, 2000). At the tree roots and at the global scale, the KCl input to oil palm planta- scale, oil palm CUE is not very well known but biomass and tions represents c. 3MtKy 1. photosynthesis rates reported so far suggest a CUE of c. 38% Quite surprisingly, although flooding (waterlogging) and lim- (that is, 62% of fixed carbon lost by respiration) (reviewed in ited K availability are two common situations in oil palm agro- Lamade et al., 2016). forestry, their overall effect on carbon balance and metabolism as

Ó 2019 The Authors New Phytologist (2019) 1 New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com

106 New 2 Research Phytologist

well as their interaction are not well documented. That is, the proteomics, and the net effect on carbon balance was deter- overall effect of K fertilisation on fresh fruit bunch and oil pro- mined by gas exchange and biomass measurements. We used duction has been extensively studied (Ochs, 1965; Hartley, 1988; K conditions typical of oil palm plantations, with a maximal Ochs et al., 1991; Corley & Tinker, 2016) but the metabolic K+ concentration of 4 mM. In the present paper, we deliber- response to K availability, its crossed effect with waterlogging and ately put emphasis on leaf composition because of the practi- the effect on net carbon fixation (and CUE) have not been exam- cal significance of leaf diagnostic in oil palm cultivation ined. In addition to their co-occurrence in the field, K limitation monitoring (Chapman & Gray, 1949; Foster, 2002). In fact, and flooding responses are physiologically related. In effect, the elemental composition of standardised samples (collected waterlogging generally leads to oxygen deprivation (root hypoxia) at fixed leaf rank no. 17 and leaflet position called ‘B point’) thereby stimulating substrate-level ATP production, while K lim- is commonly used to estimate oil palm nutrient status in the itation affects glycolysis via the well-known inhibition of pyru- field (Ollagnier et al., 1987). Our results show that both low vate kinase. Therefore, in principle, the combination of K availability and waterlogging have drastic effects on leaf waterlogging and K limitation should be highly detrimental to metabolome and proteome and stimulate leaf respiration. This respiratory metabolism and growth. Past studies on different combined effect was exploited using multivariate statistics to plant species have shown that the metabolic effects of K defi- delineate how leaf respiration is metabolically controlled. ciency not only include an inhibition of glycolysis due to a decrease in pyruvate kinase activity, but also the accumulation of Materials and Methods typical metabolites such as polyamines (Jones, 1961, 1966; Okamoto, 1966, 1967, 1968; Freeman, 1967; Besford & Maw, Plant material and photosynthesis 1976; Armengaud et al., 2009; Hussain et al., 2011; Sung et al., 2015), and amongst them putrescine is a typical marker known Germinated oil palm seeds (Dura 9 Pisifera) were obtained for > 60 yr (Richards & Coleman, 1952). In Arabidopsis, it has from Siam Elite Palm Co. Ltd (Krabi, Thailand). Plants were been shown that amino acid and organic acid synthesis are re- sown directly in sand (washed with distilled water) in the oriented to more neutral amino acids, and that the inhibition of glasshouse, using 7-l pots. K-containing nutrient solution pyruvate kinase effectively leads to altered glycolytic metabolism (4 mM KCl) was provided for 1 month until emergence, and (Armengaud et al., 2009). Metabolic effects of waterlogging are then controlled nutrient conditions at varying K concentration rather similar (Orchard & So, 1985; Orchard et al., 1986; Wang were used. Plants were cultivated in the Research School of et al., 2014; Phukan et al., 2016), and it has been shown that Biology plant facility at the Australian National University in waterlogging and K deficiency share common signalling pathways Canberra (Australia). Growth conditions were: natural photope- via ethylene (Jung et al., 2009). Furthermore, waterlogging symp- riod (from 13.5/10.5 h at emergence (February) to 14.5/9.5 h toms have been shown to be partly alleviated by foliar spraying of at the end of the experiment (January), with a minimum of potassium (Ashraf et al., 2011; Gautam et al., 2016; Dwivedi 9.5/14.5 h in June), 30/24°C air temperature, 70/60% relative et al., 2017). We recently showed that in sunflower, (1) K defi- humidity day/night. Photosynthetic parameters were measured ciency leads to a (modest) decrease in photosynthesis and typical using a portable open system Li-Cor 6400 XT. Net assimilation metabolite accumulation (such as putrescine) but this effect is (A) and conductance reported in Fig. 1 were obtained under mostly suppressed (rather than aggravated) by waterlogging and saturating light (1500 lmol m 2 s 1 PAR) at 400 lmol mol 1 (2) despite the reorchestration of glycolysis, K deficiency leads to CO2 and 21% O2. Night respiration (Rdark) was measured after a considerable increase in leaf respiration (Cui et al., 2018). Sun- photosynthesis measurements on dark-adapted leaves (c. l 1 ° flower is an annual species with a conservatively high stomatal 60 min) at 400 mol mol CO2, 21% O2 and 25 C. conductance and therefore the effect of K availability and waterlogging on both photosynthesis and carbon allocation likely dif- Nutrient conditions and waterlogging fers considerably from that in oil palm (in particular in young oil palm plants that have a relatively low stomatal conductance). The nutrient solution was composed on purpose for this This question is of crucial importance for the global carbon bud- experiment, whereby the amount of K+ was varied by chang- get since palm oil production by fruit metabolism (c. 62 Mt y 1) ing the amount of KCl. Three K availability conditions were represents nearly 0.4 Gt CO2 liberated each year in the atmo- used here: ‘low K’ (0.2 mM), ‘medium K’ (1 mM), and ‘high sphere and overall, crude oil palm production is estimated to lib- K’ (4 mM) before the waterlogging experiment. Resupplying 1 erate on average 2.52 t CO2 equivalents t (RSPO, 2017). In K to low-K plants was from month 9 or 10 (see Supporting other words, a significant change in oil palm respiration might Information Fig. S1 for a schematic representation of the have pervading consequences for tropical agrosystems, atmo- experimental design). The amount of nitrate and phosphate sphere composition, and guidelines on glasshouse gas emission (in mM) in the nutrient solution was kept constant through- for sustainable oil palm production. out experiments (nutrient solution composition in Table S1). Here, we carried out experiments on oil palm saplings cul- Under nonwaterlogging (control) conditions, 500 ml nutrient tivated in the glasshouse under controlled conditions, under solution was provided to each pot every day. Excess liquid varying K availability with or without waterlogging. The was allowed to drain freely from the pot. Waterlogging was metabolic response was followed with metabolomics and performed by filling the pot with the nutrient solution. Only

New Phytologist (2019) Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust

107 New Phytologist Research 3

(a) (b) (c)

∗ ∗ ∗ –1 –1 –1 –2 –2 –2 ∗ ∗

∗ ∗

∗

(d) (e) (f) –1 –1 –1 –2 –2

–2 –1

Fig. 1 Photosynthetic parameters in oil palm leaves under varying K conditions and waterlogging. (a) Net photosynthesis (A). (b) Stomatal conductance for water vapour (gh). (c) Respiration in darkness (Rdark). (d) Leaf elemental K content measured by ICP-OES. (e, f) Relationship (continuous line, linear regression) between photosynthesis and stomatal conductance (R2 = 0.70, e) and relative chlorophyll content as measured by a chlorophyll meter (R2 = 0.63, f). Asterisks (*) stand for signiﬁcance between waterlogged and nonwaterlogged conditions at sampling date 7. In (d), letters stand for statistical classes (P < 0.05), and on the x-axis, ‘before’ and ‘after’ refer to before and after the onset of waterlogging, and ‘early’ and ‘late’ refer to early and late K resupply (see the experimental design in Supporting Information Fig. S1 for more details). In (a–d), each point is mean SD (n = 5). The onset of waterlogging is just after sampling date 3 (11 months cultivation). In (a–c), the difference between low and high K is always signiﬁcant regardless of time.

two K concentrations were used for waterlogging (low, high). upon seven sampling dates: 2 months, 1 month or 1 wk before The volume of water was adjusted everyday to compensate the onset of waterlogging (sampling nos. 1, 2 and 3, respectively), for evaporation and transpiration. In practice, c. 200 ml were and then after 1, 2, 3 or 7 wk (sampling nos. 4, 5, 6 and 7, added each day to each pot. Therefore, the concentration in respectively). The day before sampling, photosynthesis and dark dissolved oxygen in the pot decreased slowly, generating a respiration (reported in Fig. 1) were measured on the same progressive hypoxia rather than abrupt anoxia. plants. Upon sampling, plants were measured for size, fresh weight, and dissected and kept in liquid nitrogen for metabolomics and isotopic analyses. Roots were sampled after Experimental design and sampling having removed sand and washed with water. In practice, roots Plants were cultivated for 11 months under low, medium or high were washed, rapidly dried with absorbing paper and quenched K conditions without waterlogging, and then waterlogging was in liquid nitrogen within 2–3 min. Metabolomics and ionomics started, with half of the plants kept under control conditions (no analyses were performed on freeze-dried material, proteomics waterlogging). The experimental design is shown as a separate were performed on fresh instant-frozen (with liquid nitrogen) leaf scheme in Fig. S1. Data presented here show results obtained samples.

Ó 2019 The Authors New Phytologist (2019) New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com

108 New 4 Research Phytologist

Univariate analysis of statistical classes was performed using a Omics analyses two-way ANOVA (Fisher test), with a threshold of P = 0.05. Metabolomics analyses were carried out by gas chromatography coupled to mass spectrometry (GC-MS) as in (Abadie et al., Estimation of CUE 2017), using methanol : water extraction followed by derivatisa- tion with methoxylamine and N-methyl-N-(trimethylsilyl) triflu- Carbon use efficiency (CUE) was calculated as the ratio of pre- oroacetamide (MSTFA) in pyridine. Further details are provided sent carbon in biomass (Ab) to photosynthetically fixed carbon in Notes S1. Proteins were extracted using TCA-acetone and after via leaf net assimilation (gross photosynthesis, Ab): digestion, peptides were analysed by LC-MS/MS (nano-HPLC A coupled to mass spectrometry via an electrospray interface). CUE ¼ b : Protein identification, filtering and grouping was carried out Ag using the X!Tandem pipeline, after (Langella et al., 2017). See 9 Notes S2 for further details. Ab is given by B p, where B is plant biomass (in g) and p Ionomics (elemental) analyses were done by inductively cou- is %C in total organic matter (40%). Ag was calculated as the 69 9 Σ D pled plasma optical atomic emission spectroscopy (ICP-OES). cumulated photosynthetic input, as 10 12 iAm(i) Bli ti l 1 1 10 mg freeze-dried powdered sample were reduced to ashes at where Am(i) is average photosynthesis rate in mol g s dur- ° 500 C for 3 h in a porcelain crucible (prewashed with 10% nitric ing growth period i (comprised between two sampling dates), li D acid). Then digestion was carried out with 1 ml nitric acid (6 M is the proportion of leaf biomass in total tree biomass and ti in LC-MS grade water) at 70°C for 1 h. The liquid was evapo- is total photoperiod time (day time integrated length, in s) rated and the sample was re-extracted with 2 ml nitric acid (2%) between two sampling dates. Under our conditions, li was at 70°C for 3 h. The sample was transferred into a 15-ml tube c. 0.4–0.5 across all conditions, due to the substantial biomass and nitric acid was added up to 5 ml final volume. A 3 ml aliquot represented by rachis and stem. 12 represents the molar mass of of the digestion was injected in a ICP-OES system (Agilent 720; carbon (12 g mol 1) and 10 6 the conversion factor from lmol Agilent Technologies, Mulgrave, Australia) with a flow rate of to mol. By definition, total (cumulated) respiratory loss (CRL) 1 = = 9 1 ml min . Elemental contents were calculated using a calibra- was given by CRL Ag Ab Ag (1 CUE). tion curve processed for each batch of samples. ICP-OES data 1 reported in Fig. 2 are given in mg element mg DW. Results

Statistics Photosynthesis, respiration and elemental composition Unless otherwise stated, five replicates were done for all condi- Photosynthetic parameters are shown in Fig. 1. There was a clear tions. Supervised multivariate analysis of metabolomics data was effect of K availability on net CO2 assimilation (Fig. 1a), with a carried out by orthogonal projection on latent structure (OPLS) reduction of 40–50% at low K compared with high K. As with Simca (Umetrics, Umea, Sweden), using K level and/or expected, photosynthetic rates were intermediate at medium K waterlogging as predicted qualitative Y variables and metabolites availability, and after growth at low K followed by resupply, pho- as predicting X variables. Also, to gain insight on metabolic mech- tosynthetic rates were close to that at high K. There was a strong anisms that determine the rate of leaf dark respiration, an OPLS detrimental effect of waterlogging on photosynthesis after 7 wk analysis was performed using protein and metabolite contents as (sampling no. 7), with a 50% reduction at low K and a 25% predicting (X) variables and using Rdark as a predicted (Y ) quanti- reduction at high K (asterisks, Fig. 1a). Stomatal conductance tative variable. The absence of statistical outliers was first checked for water vapour was always relatively low, in the 0.05– using a principal component analysis (PCA) to verify that no data 0.2 mol m 2 s 1 range. Both low K and waterlogging caused a point was outside the 99% confidence Hostelling region. The decrease in conductance (Fig. 1b). By contrast, dark respiration goodness of the OPLS model was appreciated using the determi- was clearly stimulated (more than doubled) under low K and nation coefficient R2 and the predictive power was quantified by waterlogging further increased respiration (Fig. 1c): at low K, res- the cross-validated determination coefficient, Q2. The signifi- piration increased by c. 30% and at high K, respiration increased cance of the statistical OPLS model was tested using a v2 compar- more than two-fold (asterisks). The effects on photosynthesis and ison with a random model (average random error), and the respiration were not only driven by K and therefore not due to a associated P-value (PCV-ANOVA) is reported. A permutation test side effect of waterlogging on K tissue concentration, since leaf K was also performed to check the reliability of the OPLS model, elemental was c. 0.5% K at low K regardless of waterlogging that is, to verify that at maximal permutation (similarity of per- (Fig. 1d). However, it is worth noting that at high K, the elemen- muted dataset tending to zero), Q2 was always negative. Best dis- tal K content declined to 1.4% under waterlogging in both stems criminating metabolites (biomarkers) were identified using and leaves, while it did not change in roots. This suggests that volcano plots whereby the logarithm of the P-value obtained in waterlogging did not alter K absorption per se but rather inhib- univariate analysis (ANOVA) was plotted against the rescaled ited K translocation to stems and leaves (Fig. S2). K deficiency loading (pcorr) obtained in the OPLS. In such a representation, was compensated for by an increase in Ca, Mg and Mn mostly in – best biomarkers have both maximal log(P)andpcorr values. leaves (Fig. S2a,d). In roots, it is worth noting that K deficiency

New Phytologist (2019) Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust

109 New Phytologist Research 5

High K Low K High K + WL Low K + WL Medium K Resupply

(a) (b)

30 Decreased by Increased by K deﬁciency –log(P) K deﬁciency 20 14

10 12

10 0 cPGK ACO 1.000108 * t[2] 8 SDH –10 mTHM cTPI cALD mPDH(E2) 6 GABAT

–20 cG6PDH 4

–30 2

–40 0 –60 –40 –20 0 20 40 1.00042 * t[1] –1.5 –1 –0.5 0 0.5 1 1.5 pcorr

–log(P) (c) 1 (d) 14

12 Decreased by Increased by 0.5 waterlogging 10 waterlogging ALAT cPGK ALAT 8 2OGDH 0 ATP(γ) mTHM –1 –0.5 0 0.5 SDH 1 PFP 6

ACO, 4 –0.5 mPDH(E2)

2 LIP cG3PDH 0 –1 CAT2 1 0.5 0 –0.5 –1 pcorr Fig. 2 Proteomics pattern in oil palm leaves under K deficiency and waterlogging. (a) Multivariate O2PLS discrimination of samples using proteins as X 6 variables and K (high or low K) and waterlogging (with or without) as predicted Y variables. The statistical model was highly significant (PCV-ANOVA < 10 (K effect) and 3 9 10 4 (waterlogging)) and predictive (R2 = 0.993, Q2 = 0.904). (b) Volcano plot showing the best K biomarker proteins with –log (PANOVA) against pcorr, where PANOVA is the P-value obtained in the univariate analysis (two-way ANOVA) and pcorr the loading in the O2PLS. (c) Loading plot of a separate OPLS analysis carried out to examine the effect of K alone. Note the nonlinear effect of K from low (green) to medium (dark cyan) and 4.5 high (blue) or resupply (cyan). (d) Volcano plot showing the best waterlogging biomarker proteins. Proteins with PANOVA < 10 (Bonferroni threshold, shown with a dotted red line in (b, d) and involved in catabolism and respiration are indicated with arrows (and also in (c), catalase-2 (CAT2) and linoleate 9S-lipoxygenase (LIP)). Abbreviations: 2OGDH, 2-oxoglutarate dehydrogenase (E1 component, decarboxylating); ACO, aconitase; ALAT, alanine aminotransferase; ATP(c), mitochondrial ATP synthase c subunit; cALD, cytosolic fructose 1,6-bisphosphate aldolase; cG3PDH, cytosolic glyceraldehyde 3- phosphate dehydrogenase; cG6PDH, cytosolic glucose 6-phosphate dehydrogenase; cPGK, cytosolic phosphoglycerate kinase; cTPI, cytosolic triose phosphate isomerase; GABAT, c-aminobutyrate aminotransferase; mPDH (E2), subunit E2 of the mitochondrial pyruvate dehydrogenase complex; mTHM, methyl-THF homocysteine methyltransferase (methionine synthase); PFP, PPi-dependent phosphofructokinase; SDH, succinate dehydrogenase. The colour code for K/waterlogging conditions is as in Fig. 1. Significant proteins in (b) and (d) are listed in Supporting Information Table S2.

Ó 2019 The Authors New Phytologist (2019) New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com

110 New 6 Research Phytologist

caused an increase (two-fold) in iron (Fe) content (Fig. S2f). aconitate, maleate, fumarate and putrescine (increased at low K), Taken as a whole, photosynthesis appeared to be correlated to and glycerol 3-phosphate and fructose 6-phosphate (decreased at both stomatal conductance (Fig. 1e) and leaf chlorophyll content low K). A separate analysis of the K effect alone was broadly con- (Fig. 1f), demonstrating the effect of K and waterlogging on both sistent (best biomarkers: putrescine and maleate at low K, fruc- stomatal aperture and photosynthetic apparatus development. tose at high K) and, like the proteome, showed a nonlinear response to K availability (Fig. S3). Waterlogging tended to suppress the effect of low K on organic acids, with a decrease in Leaf proteome citrate and aconitate. There was also an increase in myoinositol The proteomics analysis allowed us to identify and quantify 2712 and a decrease in amino acids serine and asparagine (Fig. 3c). unique proteins. There was a very clear difference in protein Only three metabolites were associated with a significant composition between conditions (K and waterlogging), with an (P < 3910 4 in the two-way ANOVA) waterlogging 9 K inter- easy discrimination between sample groups in the multivariate action effect: glycine, glucuronate and tyramine (data not analysis (OPLS), both effect being highly significant (Fig. 2a). shown). Using the Bonferroni criterion in univariate analysis to sort sig- The organic acid composition (Krebs cycle intermediates) is nificant proteins (and keep a low FDR), 168 appeared to be sig- shown in more details in (Fig. 3d) where all conditions are repre- nificantly affected by K availability and 43 by waterlogging sented, including medium K availability and K resupply. There (Figs. 2b–d; full list given in Table S2). Most proteins involved was a significant increase in citrate and aconitate at low K (as- in the effect of K availability were associated with photosynthesis terisks) that was suppressed by waterlogging (symbol $). This (photosystems, Rubisco, ATP synthase, Calvin cycle enzymes, depressing effect of waterlogging on citrate and aconitate was very etc.) and photorespiration (glycine decarboxylase, aminotrans- strong at high K (symbol §) and was accompanied by a variable ferases, etc.) showing the detrimental effect of low K on chloro- increase in 2-oxoglutarate. Other organic acids derived from the plastic metabolism and development. The effect of waterlogging Krebs cycle via the C5-branched pathway are shown in (Fig. 3e) was more evenly distributed between cellular functions, and also and replaced into general catabolism in (Fig. 3f). Low K condi- included structural proteins (actin isoforms) (Table S2). tions caused a significant increase in citramalate (asterisks). It is worth noting that catabolism was also affected by low K Under high K, waterlogging caused a decrease in methylmaleate availability, with the up-regulation of several Krebs cycle enzymes (symbol $). As expected, low K was associated with a huge and (succinate dehydrogenase, pyruvate dehydrogenase, and aconi- significant increase in putrescine while other polyamines varied tase) and glycolytic enzymes (enolase, triose phosphate isomerase but did not change significantly (Fig. 3e). Under our conditions, and phosphoglycerate mutase) while phosphoglycerate kinase GC-MS profiling did not give direct access to absolute amounts and aldolase were down-regulated (Fig. 2b; Table S2). When the (in moles) since it was used for semi-quantitative analysis (which effect of K was analysed separately and included medium K con- is adapted to comparison between samples). We therefore carried dition and K resupply, the results were similar, with an increase out 13C-NMR analyses on c. 1.5 g pooled leaf material to have a in aconitase, pyruvate and succinate dehydrogenases and a more quantitative view of carbon pools at low and high K decrease in phosphoglycerate kinase at low K (Fig. 2c). Waterlog- (Fig. S4). The most visible changes at low K were the appearance ging also affected catabolism, with an increase in 2-oxoglutarate of citrate and malate and the increase in putrescine at the expense dehydrogenase and phosphofructokinase, and a typical hypoxic of ornithine (Fig. S4a,b). Putrescine accumulation appeared to be response protein, alanine aminotransferase (Fig. 2d). Alanine similar (in moles of carbon) to citrate accumulation. Also, it is aminotransferase, PPi-dependent phosphofructokinase and worth noting that despite fumarate (and its isomer maleate) was cytosolic glyceraldehyde 3-phosphate dehydrogenase were associ- highly significant in GC-MS metabolomics (Fig. 3b), it repre- ated with a significant K 9 waterlogging interaction effect (data sented a negligible carbon pool (Fig. S4, red arrowheads). not shown). It is worth noting that the effect of K availability on protein composition was not linear, and this appeared clearly in Relationships with dark respiration the multivariate analysis where low, medium and high K could not be aligned along the same axis (Fig. 2c). Amongst the most In order to gain insight on the cause of increased respiration rates important proteins encapsulated by axis 2 that separates low and at low K and under waterlogging, we carried out a multivariate high K from medium K, were linoleate lipoxygenase and catalase, analysis, whereby leaf dark respiration (Rdark) was used as a Y suggesting the occurrence of oxidative stress at extreme (low and quantitative variable. There was a strong and very well supported high) K concentrations. relationship between observed Rdark and Rdark reconstructed from metabolome (Fig. 4a), showing that the metabolic composition reflected the CO respiratory efflux. As respiration was varied Leaf metabolome 2 through K availability and waterlogging, some metabolites sup- GC-MS metabolomics analysis allowed us to identify and quan- porting this relationship could be unrelated to respiration but tify 162 analytes, representing 134 metabolites. The multivariate driven by K availability and waterlogging only. To sort out this analysis showed a clear discrimination between sample groups, potential problem, we highlighted metabolites that were relevant regardless of sampling time (Fig. 3a). Many metabolites were for the relationship with Rdark but not for K or waterlogging affected by K availability, with the best biomarkers being citrate, effects using a selection based on the variable importance for the

New Phytologist (2019) Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust

111 New Phytologist Research 7

High K Low K High K + WL Low K + WL Medium K Resupply 0 wk (a) 1 wk (b) (c) 2 wk Effect of K deficiency Effect of waterlogging 3 wk 4 7 wk Decreased –log(P) IncreasedDecreased –log(P) Increased 18 6 Citrate Putrescine Asn 16 2 T-aconitate 5 14 Maleate Ser T-aconitate Fumarate C-aconitate Glycerate Myoinositol 12 C-aconitate 4 0 Ile Tartarate Citrate 10 1.01152 * t[2] Glycerol Tyramine Val 3 3-phosphate 8 Citramalate –2 Fructose 6 2 6-phosphate 4 –4 1 Ononitol 2

0 0 –6 –10 –8 –6 –4 –20 2 468 –1 – 0.5 0 0.5 1 –0.6 –0.4 –0.2 0 0.2 0.4 0.6 1.00085 * t[1] pcorr pcorr (d) (e) (f) 4 3.0 Citrate Putrescine ∗ Organic acid pathway Cis-aconitate ∗ Spermidine (Krebs cycle and C -branched) Trans-aconitate Ornithine and derivatives 5 Isocitrate 2.5 Itaconate 2-oxoglutarate Citramalate 3 Pyruvate glycolysis Succinate Methylmaleate ∗ Fumarate 2.0 Malate ∗ Acetyl-CoA Citrate ∗ 2 ∗ 1.5 ∗ ∗ Oxaloacetate Aconitate Itaconate

$ 1.0 Relative content Relative $ content Relative Malate Isocitrate $ 1 $ Citramalate

0.5 $ Fumarate 2OG §§§ Methylmaleate 0 0.0 Maleate Succinate LK MK HK Early R Late R LK+WL HK+WL LK MK HK Early R Late R LK+WL HK+WL Tartarate

Fig. 3 Metabolomics pattern in oil palm leaves under K deﬁciency and waterlogging. (a) Multivariate O2PLS discrimination of samples using metabolites as X variables and K (high or low K) and waterlogging (with, without) as predicted Y variables. The statistical model was extremely signiﬁcant 32 14 2 2 (PCV-ANOVA < 10 (K effect) and 10 (waterlogging)) and predictive (R = 0.957, Q = 0.827). Note the effect of time is very small, with a clear overlapping of data points regardless of sampling time. (b, c) Volcano plots showing the best K (resp. waterlogging) biomarker metabolites with –log

(PANOVA) against pcorr, where PANOVA is the P-value obtained in the univariate analysis (two-way ANOVA) and pcorr the loading in the O2PLS. The 4 horizontal dashed line represents the Bonferroni threshold (3.5910 ). (d, e) Changes in Krebs cycle intermediates (d) and polyamine and C5 branched acids (e) relative to the overall average fixed at 1 (red dash-dotted line). Significant differences (P < 0.05) at low K (LK) compared with high K (HK) is shown with asterisks (*). Significant changes due to waterlogging (at fixed K) are shown with symbols $ or §. (f) Summary of organic acid metabolism showing in red discriminating metabolites appearing in (b, c).The colour code is as in Fig. 1. In (d, e), on the x-axis, ‘early’ and ‘late’ refer to early and late K resupply (see the experimental design in Supporting Information Fig. S1 for more details). 2OG, 2-oxoglutarate; Asn, asparagine; C-aconitate, cis- aconitate; Ile, isoleucine; Ser, serine; T-aconitate, trans-aconitate; Val, valine. projection (VIP): only metabolites with low VIP (and loading) and C-skeletons provision (that is growth respiration as opposed values in the K and/or waterlogging discrimination but high VIP to maintenance respiration). Also, as expected, some of the best values in the relationship with Rdark were coloured (in red, Rdark-related proteins were directly involved in respiration, such Fig. 4b). It appears that the best positively related metabolites as mitochondrial pyruvate dehydrogenase, citrate synthase, isoci- were soluble sugars and their derivatives (re-contextualised in trate dehydrogenase, and succinate-CoA ligase (Table S4, red). general sugar metabolism in Fig. 4c) while the best negatively related metabolites were intermediates of the polyamine pathway Effects on other organs and plant carbon use efficiency (uracil, ornithine, b-alanine, Fig. 4b,c), phytol and products of amino acid degradation (3-methyl-oxovalerate) or sugar oxida- The CUE was calculated using the photosynthesis rate, total pho- = tion (glucarate, 2-oxogluconate) (the full list of metabolites is tosynthetic mass and biomass as CUE Ab/Ag where Ab is the given in Table S3). A similar analysis conducted with proteomics amount of C effectively found as biomass and Ag is the cumu- also yielded a good correlation between observed and predicted lated C input by photosynthesis (‘gross’ photosynthesis). CUE respiration rates (Fig. S5). Proteins associated with Rdark (and not was rather variable and not statistically different between low and related to K or waterlogging effects) were numerous and some of high K up to 11 months in the 0.2–0.4 range, although it tended these were related to photosynthesis (such as photosystem II sub- to be lower at high K concentrations (Fig. 5a). After 11 months, units or Calvin cycle enzymes), ribosomes or cell cycle the CUE was significantly lower at high K concentrations and (Table S4). This is unsurprising and suggests that the relationship this was also the case under waterlogged conditions. The CUE between proteins and respiration also encapsulated developmen- was lower during waterlogging for 3 wk. After 13 months’ culti- tal processes, which are effectively linked to respiration via energy vation, the CUE rose to c. 0.35 but remained significantly lower

Ó 2019 The Authors New Phytologist (2019) New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com

112 New 8 Research Phytologist

(a) (b) (c) 1.2 3.0 Polyamine Sugar and aldarate Uracil metabolism metabolism

) 1.0 2.5 -1 – Asp Ornithine Starch

s Catechine Phytol Gentiobiose –2 Gentiobiose CO2 Beta-alanine BST 0.8 2.0 G6P_1 Putrescine Maltose PPA Xylitol Uracil THB Ornithine Maltose_1 Glucose-6-phosphate G6P_2 Spermi(di)ne (μmol m 1.5 0.6 IP Sucrose ark V

d MOV Galactose Glucarate Maltose_2 N-carbamoyl-β-alanine Galactose UDPG R Threonine Glucose 2OU Sorbitol 0.4 1.0 2OU PAB Pipecolate CO2 NH3 Hexitols Glucuronate β-alanine 0.5 Pentitols

Predicted Predicted 0.2 CO2 Glucarate CO2 Pentoses Gluconate 0.0 Asp 0.0 0.0 0.2 0.4 0.6 0.8 1.0 1.2 –0.6 –0.4 –0.2 0.0 0.2 0.4 0.6 2-oxogluconate R –2 –1 r Observed dark (μmol m s )

Fig. 4 Relationship between leaf metabolome and leaf dark respiration (Rdark) using an OPLS statistical model with Rdark as a quantitative Y variable. (a) Relationship between observed and predicted respiration rate. (b) Biomarkers of respiration rate (red discs) shown as a volcano plot (variable importance for the OPLS projection, VIP, against the regression coefﬁcient, r) where biomarkers that were also involved in group (K, waterlogging) differentiation were kept as grey discs. (c) Simpliﬁed polyamine and aldarate metabolism showing some key metabolites found in (b) (in red). 2OU, 2-oxogluconate; Asp, aspartate; BST, b-sitosterol; G6P, glucose 6-phosphate; MOV, 3-methyl-2-oxovalerate; PAB, parahydroxybenzoate; PPA, phenylpropaneamine; THB,

3,4,5-trihydroxybenzoate; UDPG, UDP-glucose. ‘Hexitol’ means polyol in C6 (such as sorbitol), ‘pentitol’ means polyol in C5 (such as xylitol). In (a), the 2 2 37 OPLS model is highly signiﬁcant and predictive (R = 0.917, Q = 0.795, PCV-ANOVA < 10 ). Dash-dotted, 1 : 1 line. The colour code is the same as in Fig. 1. The full list of discriminant metabolites is shown in Supporting Information Table S2.

at high K under waterlogging. These variations simply reflect that of proteins involved in photosynthesis and chloroplast develop- while total plant biomass was larger at high K (Fig. 5b), photo- ment (Table S2). We also found specific biomarkers of K avail- synthesis was even larger and, therefore, the total respiratory loss ability and waterlogging, in addition to the well known was proportionally important (CRL, Fig. 5c). Waterlogging putrescine accumulation at low K (Armengaud et al., 2009; Cui caused a decline in total plant biomass (Fig. 5b), but photosyn- et al., 2018) originating from the increased activity of ornithine thesis was nearly unaffected before the last sampling date (see decarboxylase (Martin-Tanguy, 2001) (except in Arabidopsis in Fig. 1a) and, therefore, the respiratory loss at high K under water- which this enzyme is lacking, Hanfrey et al., 2001). Low K causes logging was very large (Fig. 5c). As plant CUE is influenced by changes in K-dependent enzyme activities such as pyruvate kinase metabolism and respiration of organs other than leaves, we also (decreased pyruvate production) (Evans, 1963; Nowak & Mild- carried out metabolomics analysis in rachis, stem and root. There van, 1972; Besford & Maw, 1976), succinyl-CoA thiokinase were significant differences in the metabolic profile of organs (equilibrium displaced towards succinyl-CoA formation) (Lynn other than leaves between growth conditions (Fig. 5d): leaf rachis & Guynn, 1978) or S-adenosylmethionine synthase (Takusagawa generally had more amino acids at high K and roots were et al., 1996). Therefore, glycolysis and the TCAP were expected depleted in most amino acids and sugars at low K. Most signifi- to be reorchestrated at low K with, for example, the alternative cant metabolites revealed by separate multivariate OPLS analyses pathway for pyruvate synthesis involving malate synthesis via pointed to increased putrescine at low K in all organs, as well as phosphoenolpyruvate carboxylase (PEPC) and subsequent malate an increased Krebs cycle intermediates (fumarate, maleate). degradation by the malic enzyme (ME). However, proteomics Waterlogging caused an increase in data only showed significant changes in glycolysis (in particular, a 2-oxoglutarate in the stem at high K (black arrowhead) and in decrease in cytosolic 3-phosphoglycerate kinase), Krebs cycle several sugar species at low K in the rachis (red arrowheads). A (such as aconitase) and the oxidative pentose phosphate pathway more detailed analysis of the effect of K availability in roots (cytosolic glucose 6-phosphate dehydrogenase) but no change in further show the accumulation of citrate and putrescine at low K PEPC or ME abundance (Fig. 2). Therefore, the PEPC-ME and a decrease in sugars and the glycolytic intermediate 3- route might have contributed to pyruvate provision at a rather phosphoglycerate (Fig. S6). low rate and there must have been other mechanisms to synthe- sise pyruvate and sustain mitochondrial pyruvate dehydrogenase Discussion activity (the content of which was found to increase). Amongst possible mechanisms, our data suggest the involvement of organic acid recycling to pyruvate via the C -branched Metabolic effects of K deficiency and waterlogging 5 acid pathway, because methylmaleate and citramalate are two C5- Both K deficiency and waterlogging had a strong effect on photo- branched intermediates that are highly significant at low K (and synthesis and development, as revealed by the decline in net methylglutamate was also more abundant at low K, with a assimilation and chlorophyll content (Fig. 1) and the abundance P-value of 0.02, data not shown). In addition, the proteomics

New Phytologist (2019) Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust

113 New Phytologist Research 9

1.0 250 (a) High K (b) 200 Low K 150 High K + WL Low K + WL 100 0.8 Shoot 50 0 -10 ot 060.6 -20 ass (g DW per plant) -30 Roo -40 Biom CUE 140 0.4 (c) 120 100 80 020.2 60 g C per plant) 40

20 ( CRL 0.0 0 1234567 1234567 Sampling date Sampling date (d) l to si ine te o er a in S 4 Citr P G3 Myo 3 2 1

Leaf 0

ine inateithine ine cc rn sc gin u Ar S O 4 Putre 3 2 1

e t Rachis 0 a o

a c e e d hin e a it n x e ci n s e Orn e H 4 agi tr r u P 3 spa A 2 1 Stem 0

te a A n AB e co 4 te n u G ci gl lea s xo -o 3 Ma 2 Putre 2 1 Root 0 Amino acids OA P Sugars Other Fig. 5 Biomass and carbon use efﬁciency (CUE) in oil palm under varying K conditions and waterlogging. (a) Calculated CUE. The dashed-dotted line corresponds to CUE = 0.38. (b) Root and shoot biomass (in g DW per tree). (c) Cumulated respiratory loss (CRL) calculated from CUE and biomass. (d) Overview of metabolic proﬁle in different organs at sampling time 7. In each panel, labelled metabolites correspond to best discriminating metabolites in multivariate analysis. Metabolite contents are expressed relative to the overall average in the organ of interest. G3P, glycerol 3-phosphate; OA, organic acids; P, polyamine and associated metabolites (such as b-alanine and uracil); ‘2-oxogluco’, 2-oxogluconate; ‘Hexad’, hexadecanoate. ‘Sugars’ refers to hexoses, disaccharides, and aldaric acids. Red arrowheads, fructose and mannose (rachis); black arrowhead, 2-oxoglutarate (stem). Mean SE (n = 5).

Ó 2019 The Authors New Phytologist (2019) New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com

114 New 10 Research Phytologist

analysis indicated that there was a 4-hydroxy-4-methyl-2- is, interconversion of aconitate and isocitrate. Also, there was a oxoglutarate (parapyruvate) aldolase in oil palm (Table S2), tendency for 2-oxoglutarate to increase under waterlogging which belongs to the C5-branched pathway and converts (Fig. 3e) and accordingly, 2-oxoglutarate dehydrogenase was sig- 4-hydroxy-4-methyl-2-oxoglutarate to pyruvate. Like PEPC or nificantly altered (Fig. 2). We recognise that the increase in 2- ME, the peptide associated with this enzyme was, however, not oxoglutarate could also have originated from a higher activity of significantly affected by K availability (but positively related to glutamate dehydrogenase (GDH) but it is presently less likely Rdark, see the next section). We also note that sulfur metabolism because GDH was not amongst significant proteins under water- could have been involved, since methionine synthesis is associated logging (Table S2). with cysteine conversion to pyruvate. Here, methionine synthase Changes in organic acids were accompanied by a modification (methyl-THF homocysteine methyltransferase) was found to be of amino acid metabolism, with a decrease in asparagine and ser- significantly increased at low K. ine, and in the content of alanine aminotransferase. Accordingly, When K availability was examined separately and included we previously showed that waterlogging inhibited N assimilation medium K conditions, it was clear that the effect of K was far in leaves (Cui et al., 2018). It is worth noting that at high K, from being linear, with metabolites and proteins differentially waterlogging was also associated with generally less amino acids abundant under medium K (Figs 2, S3). Furthermore, both low in rachis, suggesting an inhibition of N assimilates export from and high K conditions were associated with common features, leaves (Fig. 5c). Here, the effect of waterlogging was even more such as catalase, lipoxygenase, erythronolactone (ascorbate deriva- pronounced after 7 wk (sampling date 7), with an important tive), all related to oxidative stress metabolism. This suggests that alteration of net photosynthesis (Fig. 1). This is in contrast with extreme KCl fertilisation led to a stress response driven either by previous experiments carried out on waterlogged young oil palm deficiency (low K) or ion excess (high K). Accordingly, the reac- grown on peat, that showed a rapid recovery after 3–4 wk (Hen- tive oxygen species (ROS)-mediated response has been shown to son et al., 2008), but agrees with results obtained by (Lamade be involved in both potassium deficiency and under salinity stress et al., 1998) on waterlogged seedlings. Presumably, the effect of (Ahmad et al., 2014; Hossain & Dietz, 2016). In particular, K waterlogging depends on age, the ability to form pneu- deficiency leads to ROS generation via membrane-bound peroxi- matophorae (aerial roots), and the biophysical soil environment dases, and ROS in turn stimulate the expression of genes involved (such as pH and dissolved O2). Here, oxygen deprivation was in the low-K response, such as those encoding high-affinity K+ very progressive (because of pot volume and the addition of channels (Shin et al., 2005; Kim et al., 2010; Wang & Wu, nutrient solution everyday to adjust the water level) and under 2013). Oxidative stress at high K could explain why large potas- our conditions, we estimate that oxygen reached 10% of its origi- sium application has been shown to be occasionally associated nal value within c.12d. with a lower yield in oil palm (Ollagnier & Ochs, 1977; Zin et al., 1993) and may even cause a decrease in leaf K content What is the origin of increased leaf respiration rates under (Akbar et al., 1976). The metabolic effect found here in leaves low K and waterlogging? was also visible in roots, with threonate (ascorbate derivative) increased at high K, and a-tocopherol (antioxidant) under both A striking common effect of low K and waterlogging was the low and high K (Fig. S6). In other words, our metabolic data sup- increase in CO2 efflux by leaf dark respiration (Rdark, Fig. 1). The port the idea that high K conditions, and therefore high K fertili- considerable variance in Rdark in the present study represented a sation levels used in the field, may be excessive and can lead to good opportunity to look at potential metabolic drivers of respi- side effects, due to overall salinity or chloride that unavoidably ration. The multivariate analysis (where metabolites involved in comes along with KCl application. the response to K and waterlogging were subtracted) suggested Waterlogging partly suppressed the effect of low K on organic that most represented drivers were sugars (including glucose acids (Fig. 3), with a significant decrease in citrate and aconitate 6-phosphate). This agrees with the original assumption that cel- and, at high K, a variable increase in itaconate. Such changes were lular sugar availability participates in controlling dark respiration likely related to an inhibition of aconitase, which interconverts in leaves (Azcon-Bieto & Osmond, 1983). However, it is citrate and isocitrate via aconitate in two steps (Castro et al., unlikely that sugar availability itself is the sole factor that dictates 1994; Lloyd et al., 1999; Tortora et al., 2007). Two potential the flux in catabolism because several pathways can be involved effectors could have been involved in aconitase inhibition: ROS (not only glycolysis) and many enzymes are controlled posttrans- (such as H2O2) and NO. In fact, plant aconitase has indeed been lationally or via effectors (reviewed in Tcherkez, 2017). Here, it shown to be inactivated by H2O2 (Verniquet et al., 1991) and is likely that other CO2-producing pathways were involved and ROS are known to increase under waterlogging, K deficiency and contributed to the observed CO2 production rate Rdark. salinity. Here, we found a metabolic and proteomic signature of Polyamine metabolism involves decarboxylations, and so does a ROS-mediated response under low and high K (see the the oxidative pentose phosphate/aldarate pathway, and typical previous paragraph). NO is also an inhibitor of aconitase that has metabolites of these two pathways (b-alanine, ornithine, uracil, been demonstrated to accumulate under waterlogging (Gupta glucarate, 2-oxogluconate) were found by our statistical analysis et al., 2012). This gaseous molecule is soluble and believed to dif- (Fig. 4). Protein drivers of leaf respiration were found to include fuse easily from roots to shoots via the xylem (Meyer et al., enzymes of mitochondrial metabolism (pyruvate dehydrogenase, 2005). NO inhibits the second step of aconitase mechanism, that succinyl-CoA ligase, etc.) showing, as expected, the importance

New Phytologist (2019) Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust

115 New Phytologist Research 11

of mitochondrial metabolism and the Krebs cycle for Rdark. In effect, despite the increase in putrescine in rachis, stem and However, we also found an enzyme involved in putrescine roots when K availability was low, the general increase in organic metabolism (putrescine hydroxycinnamoyl transferase) and two acids found in leaves was not seen in stems and roots (Figs 5, S6). other catabolic enzymes, 3-hydroxyisobutyryl-CoA hydrolase Here, plants were in the vegetative stage and it is possible (isoleucine/leucine degradation) and 4-hydroxy-4-methyl-2- that respiratory losses and CUE would have been different if oxoglutarate aldolase (C5-branched pathway, see above). Note- they had been in the productive stage, that is, with fruit worthy, metabolites associated with higher Rdark also comprised production. However, the CUE value found here, c. 0.3, 3-methyl-2-oxovalerate (isoleucine degradation). This shows the matches estimates obtained in adult trees (reviewed in Corley importance of amino acid turn-over in leaf dark respiration, as & Tinker, 2016). This suggests that respiratory losses occur- well as pathways other than the Krebs cycle that can be exacer- ring in developing fruits in adult trees are of similar magni- bated under low K and/or waterlogging. tude than respiratory losses of rapid vegetative development In effect, low K led to a considerable increase in putrescine in young trees. Nevertheless, oleosynthesis itself in fruits is (Figs. 3, S4) and this should be in principle associated with CO2 associated with a low CUE and therefore the overall effect ? + production (ornithine putrescine CO2). Also, as mentioned of K limitation and waterlogging on adult tree CUE may be above, K limitation and waterlogging directly impact on glycoly- even larger than that found here. In addition, K losses repre- sis and the Krebs cycle (altered pyruvate kinase, aconitase, and sented by fruit harvesting are of considerable importance in succinate thiokinase activity) and alternative pathways can lead to oil palm (up to 15% of total palm K is represented by fresh proportionally higher CO2 release per generated ATP. The most fruit bunches collected in 1 yr, Teoh & Chew, 1988) and likely alternative pathway here consists of aconitate recycling to this should exaggerate the effect of K deficiency. citramalate (2 aconitate + ADP + Pi + 2 NAD+ ? succinate + We recognise that in our present study, we could not differen- + + + citramalate ATP 2 NADH 3CO2) with citramalate tiate between reserve remobilisation and utilisation of recently and/or methylglutamate recycling via 4-hydroxy-4-methyl-2- fixed carbon in respiratory carbon sources. Under occasional (that oxoglutarate aldolase (regenerating 2 pyruvate) (Fig. S7). is, nonchronic) K deficiency and waterlogging, it is likely that Taken as a whole, leaf respiration could be satisfactorily pre- reserves accumulated in the trunk (stem) are of considerable dicted from omics patterns (Figs. 4, S5) that encapsulated many importance to sustain catabolism and therefore CO2 production. features (metabolites or peptides). This shows that sources of Likewise, our cultivation conditions did not allow us to measure CO2 production by leaves were numerous (Krebs cycle, amino root and stem respiration and therefore to compare the effect of acid degradation, oxidative pentose phosphate pathway, alterna- K deficiency on CO2 efflux in heterotrophic organs as compared tive pathways). The balance between these sources may vary with with leaves. These aspects, that are critical for oil palm carbon environmental conditions (O’Leary et al., 2018), and our data balance in the field, will be addressed in a future study using 13C- suggest that both K limitation and waterlogging had a strong labelling. effect on this balance. Acknowledgements Consequences for carbon balance and perspectives The authors warmly thank Cyril Abadie, Illa Tea, Anne-Marie At the leaf level, both low K and waterlogging had a huge effect Schiphorst, Camille Bathellier and members of the RSB Plant on carbon balance because net photosynthesis decreased and Service for their help during oil palm sampling. The support of Rdark increased (Fig. 1). This effect was found to scale up to the the Joint Mass Spectrometry Facility (ANU) is acknowledged. JC plant level, with an alteration of total biomass at low K and under was supported by an Australia Awards PhD Scholarship, and the waterlogging (Fig. 5b). Still, the calculated CUE was conserva- research was supported by the Australian Research Council via a tively low at high K due to a plateauing biomass at 11 and Future Fellowship, under contract FT140100645. The authors 12 months despite a proportionally high photosynthesis input thank S. Cholathan (Siam Elite Palm) for providing oil palm (Fig. 1). The effect of high K was also visible under waterlogging. seeds. This shows that despite a higher C fixation and higher biomass, oil palm saplings lost proportionally higher amounts of carbon in Author contributions respiration at high K. This was likely due to higher growth respiration (the growth rate roughly doubled at high K compared with GT and EL designed the experiments. JC performed palm culti- low K) and maybe, higher maintenance respiration rates (in vation, biomass measurements, and metabolomics analyses. MD gCg 1 C) in heterotrophic organs. Accordingly, 13C-labelling performed proteomics analyses. JC, GT, MD and MZ carried on adult Eucalyptus trees has shown that K availability increases out ’omics data integration and statistics. All authors contributed not only photosynthesis (gross assimilation) but also sink demand to writing the paper. and therefore respiration (Epron et al., 2016). The fact that CRL and Rdark did not show similar changes at low K suggests that ORCID metabolic consequences on CO2-producing pathways in heterotrophic organs were different from those found in leaves. Guillaume Tcherkez https://orcid.org/0000-0002-3339-956X

Ó 2019 The Authors New Phytologist (2019) New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com

116 New 12 Research Phytologist

Hanfrey C, Sommer S, Mayer MJ, Burtin D, Michael AJ. 2001. References Arabidopsis polyamine biosynthesis: absence of ornithine decarboxylase and Abadie C, Blanchet S, Carroll A, Tcherkez G. 2017. Metabolomics analysis of the mechanism of arginine decarboxylase activity. The Plant Journal 27: 551–560. postphotosynthetic effects of gaseous O2 on primary metabolism in illuminated leaves. Functional Plant Biology 44: 929–940. Hartley C. 1988. The oil palm. London, UK: Logman Scientific and Technical. Ahmad P, Ashraf M, Hakeem KR, Azooz MM, Rasool S, Chandna R, Akram Henson I, Harun M, Chang K. 2008. Some observations on the effects of high NA. 2014. Potassium starvation-induced oxidative stress and antioxidant water tables and flooding on oil palm, and a preliminary model of oil palm defense responses in Brassica juncea. Journal of Plant Interactions 9:1–9. water balance and use in the presence of a high water table. Oil Palm Bulletin – Akbar U, Tampulobon F, Amiruddin D, Ollagnier M. 1976. Fertilizer 56:14 22. experimentation on oil palm in North Sumatra. Oleagineux 31: 305–316. Hossain MS, Dietz K-J. 2016. Tuning of redox regulatory mechanisms, reactive Amthor JS. 2000. The McCree–de WitPenning de VriesThornley respiration oxygen species and redox homeostasis under salinity stress. Frontiers in Plant – paradigms: 30 years later. Annals of Botany 86:1–20. Science 7: 548 558. Armengaud P, Sulpice R, Miller AJ, Stitt M, Amtmann A, Gibon Y. 2009. Hussain SS, Ali M, Ahmad M, Siddique KHM. 2011. Polyamines: natural and Multilevel analysis of primary metabolism provides new insights into the role of engineered abiotic and biotic stress tolerance in plants. Biotechnology Advances – potassium nutrition for glycolysis and nitrogen assimilation in Arabidopsis 29: 300 311. roots. Plant Physiology 150: 772–785. Jones LH. 1961. Some effects of potassium deficiency on the metabolism of the – Ashraf MA, Ahmad MSA, Ashraf M, Al-Qurainy F, Ashraf MY. 2011. tomato plant. Canadian Journal of Botany 39: 593 606. Alleviation of waterlogging stress in upland cotton (Gossypium hirsutum L.) by Jones LH. 1966. Carbon-14 studies of intermediary metabolism in potassium- – exogenous application of potassium in soil and as a foliar spray. Crop and deficient tomato plants. Canadian Journal of Botany 44: 297 307. Pasture Science 62:25–38. Jung J-Y, Shin R, Schachtman DP. 2009. Ethylene mediates response and – Azcon-Bieto J, Osmond CB. 1983. Relationship between photosynthesis and tolerance to potassium deprivation in Arabidopsis. Plant Cell 21: 607 621. Kim MJ, Ciani S, Schachtman DP. 2010. A peroxidase contributes to ROS respiration: the effect of carbohydrate status on the rate of CO2 production by respiration in darkened and illuminated wheat leaves. Plant Physiology 71: 574– production during Arabidopsis root response to potassium deficiency. Molecular – 581. Plant 3: 420 427. Baccini A, Goetz S, Walker W, Laporte N, Sun M, Sulla-Menashe D, Hackler J, Lamade E, Setiyo E, Purba A. 1998. Gas exchange and carbon allocation of Beck P, Dubayah R, Friedl M. 2012. Estimated carbon dioxide emissions from oil palm seedlings submitted to waterlogging in interaction with N fertilizer tropical deforestation improved by carbon-density maps. Nature Climate application. International Oil Palm Conference: Commodity of the Past, Change 2: 182–185. Today and the Future. Medan: Indonesian Oil Palm Research Institute. – Besford RT, Maw GA. 1976. Effect of potassium nutrition on some enzymes of 573 584. the tomato plant. Annals of Botany 40: 461–471. Lamade E, Tcherkez G, Darlan NH, Rodrigues RL, Fresneau C, Mauve C, 13 Castro L, Rodriguez M, Radi R. 1994. Aconitase is readily inactivated by Lamothe-Sibold M, Sketriene D, Ghashghaie J. 2016. Natural C peroxynitrite, but not by its precursor, nitric oxide. Journal of Biological distribution in oil palm (Elaeis guineensis Jacq.) and consequences for allocation – Chemistry 269: 29409–29415. pattern. Plant, Cell & Environment 39: 199 212. Chapman GW, Gray HM. 1949. Leaf analysis and the nutrition of oil palm. Langella O, Valot B, Balliau T, Blein-Nicolas M, Bonhomme L, Zivy M. 2017. ! Annals of Botany 13: 415–433. X TandemPipeline: a tool to manage sequence redundancy for protein – Corley R, Tinker P. 2016. The oil palm, 5th edn. Chichester, UK: Wiley- inference and phosphosite identification. Journal of Proteome Research 16: 494 Blackwell. 503. Lloyd SJ, Lauble H, Prasad GS, Stout CD. 1999. The mechanism of aconitase: Cui J, Abadie C, Carroll A, Lamade E, Tcherkez G. 2018. Responses to K deficiency and waterlogging interact via respiratory and nitrogen metabolism. 1.8 A resolution crystal structure of the S642A:citrate complex. Protein Science – Plant, Cell & Environment 42: 647–658. 8: 2655 2662. Dwivedi SK, Kumar S, Bhakta N, Singh SK, Rao KK, Mishra JS, Singh AK. Lynn R, Guynn RW. 1978. Equilibrium constants under physiological 2017. Improvement of submergence tolerance in rice through efficient conditions for the reactions of succinyl coenzyme A synthetase and the application of potassium under submergence-prone rainfed ecology of Indo- hydrolysis of succinyl coenzyme A to coenzyme A and succinate. Journal of – Gangetic Plain. Functional Plant Biology 44: 907–916. Biological Chemistry 253: 2546 2553. Epron D, Cabral OMR, Laclau J-P, Dannoura M, Packer AP, Plain C, Battie- Martin-Tanguy J. 2001. Metabolism and function of polyamines in plants: et al. 13 recent development (new approaches). Plant Growth Regulation 34: 135–148. Laclau P, Moreira MZ, Trivelin PCO, Bouillet J-P 2016. In situ CO2 pulse labelling of field-grown eucalypt trees revealed the effects of potassium Meyer C, Lea US, Provan F, Kaiser WM, Lillo C. 2005. Is nitrate reductase a nutrition and throughfall exclusion on phloem transport of photosynthetic major player in the plant NO (nitric oxide) game? Photosynthesis Research 83: – carbon. Tree Physiology 36:6–21. 181 189. Evans HJ. 1963. Effect of potassium and other univalent cations on activity of Nowak T, Mildvan AS. 1972. Nuclear magnetic resonance studies of the pyruvate kinase in Pisum sativum. Plant Physiology 38: 397–402. function of potassium in the mechanism of pyruvate kinase. Biochemistry 11: – Foster HL. 2002. Assessment of oil palm requirements. In: Fairhurst TH, 2819 2828. H€ardter R, eds. The oil palm, management for large and sustainable yields. Ochs R. 1965. Contribution al’etude de la fumure potassique du palmier a huile. – Singapore City, Singapore: Potash and Phosphate Institute of Canada Oleagineux 20: 433 436. (ESEAP), 382. Ochs R, Olivin J, Quencez P, Hornus P. 1991. Response to potassium fertilizer – Freeman GG. 1967. Studies on potassium nutrition of plants. II. Some effects of on acid sands of tertiary sediments. Oleagineux 46:1 11. potassium deficiency on the organic acids of leaves. Journal of the Science of Okamoto S. 1966. Effect of mineral nutrition on metabolic change induced in – Food and Agriculture 18: 569–576. crop plant roots (III). Soil Science and Plant Nutrition 12:13 17. Gautam P, Lal B, Tripathi R, Shahid M, Baig MJ, Maharana S, Puree C, Nayak Okamoto S. 1967. Effects of potassium nutrition on the glycolysis and the Krebs – AK. 2016. Beneficial effects of potassium application in improving cycle in taro plants. Soil Science and Plant Nutrition 13: 143 150. submergence tolerance of rice (Oryza sativa L.). Environmental and Okamoto S. 1968. The respiration in the roots of broad bean and barley Experimental Botany 128:18–30. under a moderate potassium deficiency. Soil Science and Plant Nutrition – Gupta KJ, Shah JK, Brotman Y, Jahnke K, Willmitzer L, Kaiser WM, Bauwe H, 14: 175 182. Igamberdiev AU. 2012. Inhibition of aconitase by nitric oxide leads to O’Leary BM, Asao S, Harvey Millar A, Atkin OK. 2018. Core principles which induction of the alternative oxidase and to a shift of metabolism towards explain variation in respiration across biological scales. New Phytologist 222: – biosynthesis of amino acids. Journal of Experimental Botany 63: 1773–1784. 670 686.

New Phytologist (2019) Ó 2019 The Authors www.newphytologist.com New Phytologist Ó 2019 New Phytologist Trust

117 New Phytologist Research 13

Ollagnier M, Daniel C, Fallavier P, Ochs R. 1987. The influence of climate and Zin Z, Hamdan A, Singh G. 1993. Palm oil extraction rates in peninsular soil on potassium critical level in oil palm leaf analysis. Oleagineux 42: 446– Malaysia: the effect of fertilizers. National Seminar on Palm Oil Extraction Rate: 449. Problems and Issues. Kuala Lumpur. Ollagnier M, Ochs R. 1977. Effect of fertilizers on composition of lipids produced by perennial tropical oil plants and on their yield. Oleagineux 32: 409–426. Supporting Information Orchard PW, Jessop RS, So HB. 1986. The response of sorghum and sunflower to short-term waterlogging: IV. Water and nutrient uptake effects. Plant and Additional Supporting Information may be found online in the Soil 91:87–100. Supporting Information section at the end of the article. Orchard PW, So HB. 1985. The response of sorghum and sunflower to short- – term waterlogging. Plant and Soil 88: 407 419. Fig. S1 Schematic representation of the experiment carried out Paterson RRM, Lima N. 2018. Climate change affecting oil palm agronomy, and oil palm cultivation increasing climate change, require amelioration. Ecology in this paper. and Evolution 8: 452–461. Phukan UJ, Mishra S, Shukla RK. 2016. Waterlogging and submergence stress: Fig. S2 Elemental composition of oil palm organs. affects and acclimation. Critical Reviews in Biotechnology 36: 956–966. Richards FJ, Coleman RG. 1952. Occurrence of putrescine in potassium- Fig. S3 Metabolomics pattern in leaves under low, medium or deficient barley. Nature 170: 460–462. RSPO. 2017. Impact update. Roundtable on Sustainable Palm Oil. Kuala Lumpur, high K nutrition. Malaysia: RSPO. 13 Shin R, Berg RH, Schachtman DP. 2005. Reactive oxygen species and root hairs Fig. S4 Quantitation of major metabolites by C-NMR in oil in Arabidopsis root response to nitrogen, phosphorus and potassium deficiency. palm leaves. Plant and Cell Physiology 46: 1350–1357. Sung J, Sonn Y, Lee Y, Kang S, Ha S, Krishnan HB, Oh TK. 2015. Compositional changes of selected amino acids, organic acids, and soluble Fig. S5 Multivariate relationship between leaf dark respiration sugars in the xylem sap of N, P, or K-deficient tomato plants. Journal of Plant and protein content. Nutrition and Soil Science 178: 792–797. Takusagawa F, Kamitori S, Markham GD. 1996. Structure and function of S- Fig. S6 Metabolomics pattern in roots under low, medium or adenosylmethionine synthetase: crystal structures of S-adenosylmethionine high K nutrition. synthetase with ADP, BrADP, and PPi at 2.8 A resolution. Biochemistry 35: 2586–2596. Tcherkez G. 2017. Tracking the orchestration of the tricarboxylic acid pathway Fig. S7 Summary of plausible catabolic pathways involved in in plants, 80 years after the discovery of the Krebs cycle. In: Tcherkez G, CO2 production. Ghashghaie J, eds. Plant respiration: metabolic fluxes and carbon balance. – Amsterdam, the Netherlands: Springer, 285 298. Notes S1 Methods used for metabolomics analyses. Teoh K, Chew P. 1988. Potassium in the oil palm ecosystem and some implications to manuring practices. In: Hassan H, eds. International Oil Palm Conference: Progress and Prospects. Kuala Lumpur, Malaysia: Oil Palm Research Notes S2 Methods used for proteomics analyses. Institute of Malaysia, 277–286. Tohiruddin L, Prabowo N, Foster H. 2007. Efficiency of fertiliser use of oil palm Table S1 Nutrient solution composition used in experiments. planted on soils of Northern and Southern Sumatra. Proceedings of the PIPOC 2007 International Palm Oil Congress (Agriculture, Biotechnology & Sustainability), 26-30. Table S2 List of significant proteins corresponding to the statisti- Tortora V, Quijano C, Freeman B, Radi R, Castro L. 2007. Mitochondrial cal analysis of Fig. 2. aconitase reaction with nitric oxide, S-nitrosoglutathione, and peroxynitrite: mechanisms and relative contributions to aconitase inactivation. Free Radical Table S3 List of best metabolites in the relationship with Rdark of – Biology and Medicine 42: 1075 1088. Fig. 4. Verniquet F, Gaillard J, Neuburger M, Douce R. 1991. Rapid inactivation of plant aconitase by hydrogen peroxide. Biochemical Journal 276: 643– 648. Table S4 List of proteins in the specific relationship with Rdark. Wang C, Fan L, Gao H, Wu X, Li J, Lv G, Gong B. 2014. Polyamine biosynthesis and degradation are modulated by exogenous gamma- Please note: Wiley Blackwell are not responsible for the content aminobutyric acid in root-zone hypoxia-stressed melon roots. Plant Physiology – or functionality of any Supporting Information supplied by the and Biochemistry 82:17 26. authors. Any queries (other than missing material) should be Wang Y, Wu W-H. 2013. Potassium transport and signaling in higher plants. Annual Review of Plant Biology 64: 451–476. directed to the New Phytologist Central Office.

Ó 2019 The Authors New Phytologist (2019) New Phytologist Ó 2019 New Phytologist Trust www.newphytologist.com

118 NewPhytologistSupportingInformation

Articletitle:Metabolicresponsestopotassiumavailabilityandwaterloggingreshape respirationandcarbonuseefficiencyinoilpalm Authors:JingCui,MarlèneDavanture,MichelZivy,EmmanuelleLamade,andGuillaume Tcherkez Articleacceptancedate:12February2019

ThefollowingSupportingInformationisavailableforthisarticle:

NotesS1.Methodsusedformetabolomicsanalyses. NotesS2.Methodsusedforproteomicsanalyses. TableS1.Nutrientsolutioncompositionusedinexperiments. TableS2.ListofsignificantproteinscorrespondingtothestatisticalanalysisofFig.2. TableS3.ListofbestmetabolitesintherelationshipwithRdarkofFig.4. TableS4.ListofproteinsinthespecificrelationshipwithRdark. Fig.S1.Schematicrepresentationoftheexperimentcarriedoutinthispaper. Fig.S2.Elementalcompositionofoilpalmorgans. Fig.S3.Metabolomicspatterninleavesunderlow,mediumorhighKnutrition. Fig.S4.Quantitationofmajormetabolitesby13CͲNMRinoilpalmleaves. Fig.S5.Multivariaterelationshipbetweenleafdarkrespirationandproteincontent. Fig.S6.Metabolomicspatterninrootsunderlow,mediumorhighKnutrition. Fig.S7.SummaryofplausiblecatabolicpathwaysinvolvedinCO2production.

119 Supplementary Notes S1. Metabolomics analysis

Five (5) volumes of cold (–20°C) extraction medium (methanol:water 90:10 v:v with 0.01 mg/mL ribitol) were added to grounded plant tissue. Samples were then vortexed and incubated at 60°C for 15 min with 1,400 rpm shaking. After centrifugation for 10 min at 14,800 g, 10 μL of the supernatant was transferred to a glass vial and vacuum-dried. Samples were derivatized with methoxylamine and N-methyl-N-(trimethylsilyl) trifluoroacetamide (MSTFA) in pyridine and 1 μL of derivatized extract was injected into a VF-5 capillary column (30mu250μmu0.25μm). Samples were injected in split and splitless modes. Injector temperature was set at 230°C. Initial GC oven temperature was set at 70°C. 1 min after injection, the GC oven temperature was raised to 325°C at 15°C min-1, and finally kept at 325°C for 3 min. Helium was used as the carrier gas with a constant flow rate of 1 mL min-1. Measurements were done with electron impact ionization (70 eV) in the full scan mode (m/z

40-600). 5 μL of alkane mix (C12, C15, C19, C22, C28, C32, C36 alkanes standard at 28 mg/L) were injected to compute the retention index. Peak integration and identification was done using Metabolome Express (Carroll et al., 2010). Briefly, peak detection was carried out using a slope thresholdௗofௗ200, minimal peak areaௗofௗ1000, minimal peak heightௗofௗ500, minimal peak purity factorௗofௗ2 and minimal peak width of ௗ5 scans. Peaks were identified against a MSRI library using the retention index (RI) and mass-spectral similarity as identification criteria, with matching parameters as follows: RI window tolerance ofௗ±ௗ2 RI units, MST centroid distance of ±ௗ1 RI unit, minimal peak area of 5000, MS qualifier ion ratio error tolerance ofௗ30%, minimum number of correct ratio qualifier ionsௗofௗ2, and maximum average MS ratio errorௗofௗ70%. The primary MSRI library used contained entries derived manually from analyses of authentic standards run under the same GC-MS conditions as well as entries for unidentified peaks that were automatically generated by Metabolome Express while processing the data. Peak areas were normalized to the internal standard (ribitol). As a quality control filter, samples were checked for the presence of a ribitol peak with an area of at least 105 and a deviation from the median internal standard peak area of less than 70%. Metabolomics data were normalized to fresh weight.

Supplementary Notes S2. Proteomics analysis

Protein extraction and in-solution digestion. Plant samples were finely ground in liquid nitrogen using a mortar. Proteins were extracted using the TCA acetone protocol described in (Méchin et al., 2007). Pellets proteins were solubilized in 30 μL per mg of extract of solubilization buffer (6 M urea, 2 M thiourea,10 mM dithiothreitol (DTT), 30 mM Tris-HCl pH 8.8, and 0.1% zwitterionic acid labile surfactant (ZALS I, Proteabio, Morgantown, WV, USA)). Protein concentrations were determined using the PlusOne 2D-Quant Kit (GE Healthcare, Little Chalfont, UK) and adjusted to 0.2 μg μL-1. Digestion was performed in 0.2 mL strip tubes, 20 μg of each sample was alkylated by incubation in darkness for 1 h at room temperature with 20 ȝL of 300 mM iodoacetamide (in 50 mM ammonium bicarbonate). Proteins were then diluted ten times in 880 ȝL of 50 mM ammonium bicarbonate and digested using 3 ȝL trypsin solution (enzyme:substrate ratio of 1:30 w:w). Digestion was stopped with

120 5 ȝL of trifluoroacetic acid. Peptides were desalted using C18 solid phase extraction (SPE) cartridges (strata XL 100 μm ref 8E-S043-TGB, Phenomenex, Torrance, CA, USA).

LC-MS/MS analyses. HPLC was performed on a Nano-HPLC (Eksigent). Buffers A and B were prepared with 0.1% formic acid in water, and with 0.1% formic acid in acetonitrile, respectively. A 4 ȝL-sample of the peptide solution was loaded at 7.5 ȝL min-1 for 1 min on a

Biosphere C18 trap-column (particle size: 5 ȝm, pore size: 12 nm, inner/outer diameters: 360/100 ȝm, length: 20 mm; NanoSeparations) and desalted with 0.1% formic acid, 2% acetonitrile in water. Then peptides were separated on a biosphere C18 column (particle size: 3 ȝm, pore size: 12 nm, inner/outer diameters: 360/75 ȝm, length: 300 mm; NanoSeparations). Peptide separation was achieved at 300 nL min-1 with the following steps: equilibration in 95% buffer A for 9 min, separation with a linear gradient from 5 to 30% buffer B for 75 min, linear gradient from 35 to 95% buffer B for 3 min and regeneration with 95% buffer B for 10 min. Eluted peptides were analyzed on-line with a Q-Exactive Plus mass spectrometer (ThermoFisher Scientific, Courtaboeuf, France) using a nanoelectrospray interface. Ionization (1.5 kV ionization potential) was performed with a glass needle (non- coated capillary silica tips, 360/20-10, New Objective). Peptide ions were analyzed using Xcalibur 4.0.27 with the following datadependent acquisition steps: (1) full MS scan (mass- to-charge ratio (m/ z), 350 to 1,400, in profile mode) with a resolution of 70,000 and (2) MS/MS (isolation window = 1.5 m/z, AGCtarget = 1E5, max Ion Time = 120 ms, collision energy = 27%; profile mode, resolution = 17,500). Step 2 was repeated for the eight major ions detected in step 1. Dynamic exclusion was set to 50 s. Only doubly and triply charged precursor ions were subjected to MS/MS fragmentation.

Protein identification and peptide quantification. Xcalibur raw data were transformed to mzXML open source format and centroided using the msconvert software of the package ProteoWizard 3.0.3706 (Kessner et al., 2008). Protein identification was performed using the X!Tandem Alanine (version 2017.2.1.4; www.thegpm.org) by querying MS/MS data against the GCF_000442705.1_EG5_protein.faa protein library together with a custom contaminant database (trypsin, keratins). The following static and possible modification were accounted for: one missed trypsin cleavage allowed, alkylation of cysteine and oxidation of methionine, respectively. Precursor mass tolerance was set to 10 ppm and fragment ion mass tolerance was 0.02 Da. A refinement search was added with similar parameters except that possible N- terminal acetylation with peptide signal cleavage was included. Identified proteins were filtered and grouped using X!TandemPipeline (C++) 0.2.27 (pappso.inra.fr/bioinfo/xtandempipeline) (Langella et al., 2017), with (1) a minimum of two different peptides required with an E-value smaller than 0.01, (2) a protein E-value (calculated as the product of unique peptide E-values) smaller than 10-4. Using a reverse version of the GCF_000442705.1_EG5_protein.faa database as a decoy, the false discovery rate (FDR) was estimated by X!Tandem to be 0.08% and 0.00% for peptide-spectrum match and protein identification, respectively. Peptide quantitation used the normalized spectral abundance factor (NSAF) (McIlwain et al., 2012).

121 Table S1. Nutrient solution composition used in experiments. KCl concentration was fixed at 4, 1 or 0.2 mM to have high, medium and low K conditions, respectively.

Compound Concentration(mM) KCl 4/1/0.2 NaNO3 4 Ca(NO3)2 4 MgSO4 1.5 NaH2PO4 1.33 FeͲNa(EDTA) 0.11 Ͳ3 MnSO4 210 Ͳ3 ZnSO4 210 Ͳ3 CuSO4 210 H3BO3 0.1 Ͳ3 Na2MoO4 110 Co(NO3)2 410Ͳ4

122 7DEOH 6 /LVW RI VLJQLILFDQW SHSWLGHV FRUUHVSRQGLQJ WR WKH VWDWLVWLFDO DQDO\VLV RI )LJ PDLQ WH[W sorted by pathway. “Effect” means the direction of change upon K deficiency (low K) or waterlogging. “Interaction” means the occurrence of a K u waterlogging interaction using a two-way ANOVA (i.e. with –log(P) > 4.5).

PeptidessignificantfortheKeffect Pathway Effect Interaction pcorr Ͳlog(P) probablegammaͲaminobutyratetransaminase3,mitochondrial Aminoacidmetabolism UP 0.77 6.93 alanineaminotransferase2isoform Aminoacidmetabolism UP Yes 0.63 5.33 isovalerylͲCoAdehydrogenase,mitochondrialisoform Aminoacidmetabolism UP 0.71 5.29 glutaminesynthetasenoduleisozyme Aminoacidmetabolism DOWN Ͳ0.84 5.06 alanineaminotransferase2 Aminoacidmetabolism UP Yes 0.49 4.77 glyoxylate/succinicsemialdehydereductase1 Aminoacidmetabolism DOWN Ͳ0.77 4.57 aspartateaminotransferase,cytoplasmic Aminoacidmetabolism UP 0.82 4.53 5ͲmethyltetrahydropteroyltriglutamateͲͲhomocysteinemethyltransferase1 Aminoacidmetabolism UP 0.80 7.48 28kDaribonucleoprotein,chloroplastic Generegulation DOWN Ͳ0.93 9.90 histoneH4 Generegulation DOWN Ͳ0.84 4.97 23kDajasmonateͲinducedprotein Generegulation UP 0.68 4.79 ACTdomainͲcontainingproteinACR12 Generegulation DOWN Ͳ0.83 4.73 probablemediatorofRNApolymeraseIItranscriptionsubunit37c Generegulation UP 0.80 4.60 chloroplaststemͲloopbindingproteinof41kDab,chloroplasticisoform Generegulation DOWN Ͳ0.68 4.58 DNAͲdamageͲrepair/tolerationproteinDRT102Ͳlike Generegulation UP 0.78 4.52 phosphoglyceratekinase,cytosolic Glycolysis DOWN Ͳ0.90 8.10 triosephosphateisomerase,cytosolic Glycolysis UP 0.87 7.15 triosephosphateisomerase,cytosolicisoform Glycolysis UP 0.86 7.00 fructoseͲbisphosphatealdolase5,cytosolic Glycolysis DOWN Ͳ0.82 6.99 2,3ͲbisphosphoglycerateͲindependentphosphoglyceratemutase Glycolysis UP 0.89 6.08 fructoseͲbisphosphatealdolase5,cytosolicisoform Glycolysis DOWN Ͳ0.83 5.95 enolaseͲlike Glycolysis UP 0.85 5.06 triosephosphateisomerase,cytosolic Glycolysis UP 0.80 4.91 fructoseͲ1,6Ͳbisphosphatase,cytosolic Glycolysis DOWN Ͳ0.79 4.87 pyrophosphateͲͲfructose6Ͳphosphate1Ͳphosphotransferasesubunitalpha Glycolysis UP Yes 0.56 4.61 phospholipaseDalpha1isoform Lipidmetabolism UP 0.88 6.19 patellinͲ3isoform Membranetrafficking UP 0.86 6.37 clathrinheavychain1 Membranetrafficking UP 0.75 6.31 clathrinheavychain1Ͳlike Membranetrafficking UP 0.72 5.63 peptidylͲprolylcisͲtransisomeraseCYP19Ͳ1 Other UP 0.82 5.34 flagellarradialspokeprotein5 Other DOWN Ͳ0.80 5.01 probablecarboxylesterase12 Other UP 0.69 4.76 nascentpolypeptideͲassociatedcomplexsubunitalphaͲlikeprotein4 Other UP 0.80 4.55 dehydrogenase/reductaseSDRfamilymember12 Oxidoreductases, UP Yes 0.45 13.00 general alcoholdehydrogenase1 Oxidoreductases, UP 0.93 7.78 general glyceratedehydrogenaseͲlike Oxidoreductases, DOWN Ͳ0.91 7.64 general nonͲfunctionalNADPHͲdependentcodeinonereductase2 Oxidoreductases, DOWN Ͳ0.84 5.58 general uncharacterizedaldoͲketooxidoreductase(NADP)At1g06690,chloroplastic Oxidoreductases, DOWN Ͳ0.55 5.02 general alcoholdehydrogenaseclassͲ3 Oxidoreductases, UP 0.73 4.73 general glucoseͲ6Ͳphosphate1Ͳdehydrogenase,cytoplasmicisoformisoform Pentosephosphates UP 0.84 5.46 violaxanthindeͲepoxidase,chloroplasticisoform Photosynthesis DOWN Yes Ͳ0.92 13.00 proteinLOWPSIIACCUMULATION3,chloroplasticisoform Photosynthesis DOWN Yes Ͳ0.87 13.00 chlorophyllaͲbbindingproteinCP26,chloroplasticͲlike Photosynthesis DOWN Yes Ͳ0.90 11.52 chlorophyllaͲbbindingproteinCP26,chloroplasticͲlike Photosynthesis DOWN Yes Ͳ0.89 10.41 oxygenͲevolvingenhancerprotein3Ͳ1,chloroplastic Photosynthesis DOWN Yes Ͳ0.77 10.29 ribulosebisphosphatecarboxylasesmallchain,chloroplasticͲlike Photosynthesis DOWN Ͳ0.91 9.96 oxygenͲevolvingenhancerprotein3Ͳ2,chloroplastic Photosynthesis DOWN Ͳ0.91 9.15 chlorophyllaͲbbindingprotein8,chloroplastic Photosynthesis DOWN Ͳ0.94 8.61 serineͲͲglyoxylateaminotransferaseͲlike Photosynthesis DOWN Ͳ0.94 8.34 ribulosebisphosphatecarboxylasesmallchain,chloroplastic Photosynthesis DOWN Ͳ0.88 7.96 chlorophyllaͲbbindingprotein5,chloroplastic Photosynthesis DOWN Yes Ͳ0.79 7.55 glycinedehydrogenase(decarboxylating),mitochondrial Photosynthesis DOWN Ͳ0.85 7.54 probableriboseͲ5Ͳphosphateisomerase3,chloroplastic Photosynthesis DOWN Ͳ0.90 7.51 fructoseͲbisphosphatealdolase,chloroplastic Photosynthesis DOWN Ͳ0.91 6.94 ATPsynthasegammachain1,chloroplastic Photosynthesis DOWN Ͳ0.88 6.90 ribulosebisophosphatecarboxylase(chloroplast) Photosynthesis DOWN Ͳ0.83 6.88 oxygenͲevolvingenhancerprotein2,chloroplasticͲlike Photosynthesis DOWN Yes Ͳ0.71 6.82 photosystemII5kDaprotein,chloroplasticͲlike Photosynthesis DOWN Ͳ0.81 6.79 malatedehydrogenase,glyoxysomalisoform Photosynthesis DOWN Ͳ0.89 6.62 glyceraldehydeͲ3ͲphosphatedehydrogenaseB,chloroplastic Photosynthesis DOWN Ͳ0.89 6.20 peroxisomal(S)Ͳ2ͲhydroxyͲacidoxidaseGLO1Ͳlikeisoform Photosynthesis DOWN Ͳ0.90 6.19 thylakoidlumenal29kDaprotein,chloroplasticisoform Photosynthesis DOWN Ͳ0.82 6.15 ATPsynthaseCF1alphasubunit(chloroplast) Photosynthesis DOWN Ͳ0.87 6.15 photosystemIreactioncentersubunitV,chloroplasticͲlike Photosynthesis DOWN Ͳ0.67 6.14 fructoseͲbisphosphatealdolase,chloroplastic Photosynthesis DOWN Ͳ0.86 6.06 glutamateͲͲglyoxylateaminotransferase2 Photosynthesis DOWN Ͳ0.90 6.05

123 ribulosebisphosphatecarboxylase/oxygenaseactivase2,chloroplasticisoform Photosynthesis DOWN Ͳ0.86 5.88 oxygenͲevolvingenhancerprotein1,chloroplastic Photosynthesis DOWN Yes Ͳ0.66 5.87 chlorophyllaͲbbindingprotein6,chloroplastic Photosynthesis DOWN Ͳ0.83 5.87 photosyntheticNDHsubunitofsubcomplexB2,chloroplastic Photosynthesis DOWN Yes Ͳ0.69 5.84 ribulosebisphosphatecarboxylase/oxygenaseactivase,chloroplasticisoform Photosynthesis DOWN Ͳ0.88 5.62 ribulosebisphosphatecarboxylase/oxygenaseactivase2,chloroplasticisoform Photosynthesis DOWN Ͳ0.85 5.61 ferredoxinͲͲNADPreductase,leafisozyme,chloroplastic Photosynthesis DOWN Ͳ0.82 5.61 probableadenylatekinase5,chloroplasticisoform Photosynthesis DOWN Ͳ0.76 5.58 ATPsynthasesubunitb',chloroplastic Photosynthesis DOWN Ͳ0.82 5.58 glycinedehydrogenase(decarboxylating),mitochondrial Photosynthesis DOWN Ͳ0.79 5.55 ribuloseͲphosphate3Ͳepimerase,chloroplastic Photosynthesis DOWN Ͳ0.82 5.50 serineͲͲglyoxylateaminotransferase Photosynthesis DOWN Ͳ0.86 5.37 thylakoidlumenal17.4kDaprotein,chloroplastic Photosynthesis DOWN Ͳ0.69 5.31 ATPsynthaseCF1betasubunit(chloroplast) Photosynthesis DOWN Ͳ0.83 5.27 chlorophyllaͲbbindingproteinCP2410A,chloroplasticͲlike Photosynthesis DOWN Ͳ0.80 5.26 serinehydroxymethyltransferase,mitochondrial Photosynthesis DOWN Ͳ0.82 5.13 chlorophyllaͲbbindingproteinofLHCIItype1 Photosynthesis DOWN Ͳ0.77 4.94 chlorophyllaͲbbindingproteinofLHCIItype1Ͳlike Photosynthesis DOWN Ͳ0.78 4.94 chlorophyllaͲbbindingproteinofLHCIItype1 Photosynthesis DOWN Ͳ0.77 4.89 peroxisomal(S)Ͳ2ͲhydroxyͲacidoxidase Photosynthesis DOWN Ͳ0.77 4.82 cytochromef(chloroplast) Photosynthesis DOWN Ͳ0.78 4.77 ribulosebisphosphatecarboxylase/oxygenaseactivase,chloroplasticisoform Photosynthesis DOWN Ͳ0.81 4.57 fructoseͲ1,6Ͳbisphosphatase,chloroplasticisoform Photosynthesis DOWN Ͳ0.81 4.53 serinecarboxypeptidaseͲlike50 Proteases UP 0.91 8.46 subtilisinͲlikeproteaseSBT1.7 Proteases UP Yes 0.78 7.92 subtilisinͲlikeproteaseSBT1.9 Proteases UP 0.87 6.77 LOWQUALITYPROTEINsubtilisinͲlikeproteaseSBT1.9 Proteases UP 0.82 6.31 subtilisinͲlikeproteaseSBT1.2 Proteases UP 0.80 5.73 LOWQUALITYPROTEINsubtilisinͲlikeproteaseSBT1.4 Proteases UP Yes 0.61 5.69 probablecytosolicoligopeptidaseA Proteases DOWN Ͳ0.88 5.67 subtilisinͲlikeproteaseSBT1.2 Proteases UP 0.88 5.66 subtilisinͲlikeproteaseSBT1.7 Proteases UP 0.83 5.56 oryzainalphachain Proteases UP 0.84 5.44 oryzainalphachain Proteases UP 0.69 5.14 subtilisinͲlikeproteaseSBT1.7 Proteases UP 0.72 4.98 subtilisinͲlikeproteaseSBT1.2 Proteases UP 0.77 4.83 serinecarboxypeptidaseͲlike Proteases UP 0.85 4.67 subtilisinͲlikeproteaseSBT1.7 Proteases UP 0.82 4.64 glutathionereductase,cytosolic Redox UP Yes 0.75 6.88 lactoylglutathionelyase Redox UP 0.82 6.87 probablenucleoredoxin1Ͳ2 Redox UP Yes 0.67 6.38 superoxidedismutase[CuͲZn]isoform Redox UP 0.82 6.32 superoxidedismutase[CuͲZn] Redox UP 0.86 6.07 probableNAD(P)Hdehydrogenase(quinone)FQR1Ͳlike1 Redox UP 0.68 6.04 peroxidase4 Redox UP 0.88 6.02 superoxidedismutase[CuͲZn] Redox UP 0.86 5.99 2ͲCysperoxiredoxinBAS1,chloroplastic Redox DOWN Ͳ0.71 5.87 thioredoxinreductaseNTRC Redox DOWN Ͳ0.87 5.55 NAD(P)Hdehydrogenase(quinone)FQR1Ͳlikeisoform Redox UP 0.61 5.55 probableNAD(P)Hdehydrogenase(quinone)FQR1Ͳlike1 Redox UP 0.62 5.25 phospholipidhydroperoxideglutathioneperoxidase1,chloroplastic Redox DOWN Ͳ0.76 5.02 peptidemethioninesulfoxidereductaseB5,partial Redox UP 0.83 4.93 cationicperoxidaseSPC4 Redox UP 0.82 4.88 peroxiredoxinͲ2EͲ1,chloroplastic Redox DOWN Ͳ0.77 4.87 proteindisulfideͲisomerase Redox UP 0.82 4.86 ferredoxinͲthioredoxinreductasecatalyticchain,chloroplasticisoform Redox DOWN Ͳ0.74 4.57 aconitatehydratase,cytoplasmic RespirationandTCAP UP 0.89 7.97 succinatedehydrogenase[ubiquinone]flavoproteinsubunit,mitochondrial RespirationandTCAP UP 0.81 7.81 aconitatehydratase,cytoplasmicͲlike RespirationandTCAP UP 0.85 7.25 dihydrolipoyllysineͲresidueacetyltransferasecomponent2ofPDH RespirationandTCAP UP 0.87 6.96 putativeaconitatehydratase,cytoplasmic RespirationandTCAP UP 0.87 6.30 dihydrolipoyllysineͲresidueacetyltransferasecomponent2ofPDH RespirationandTCAP UP 0.83 5.50 probableATPsynthase24kDasubunit,mitochondrial RespirationandTCAP DOWN Ͳ0.73 5.37 probableATPsynthase24kDasubunit,mitochondrial RespirationandTCAP DOWN Ͳ0.75 4.87 40SribosomalproteinS2Ͳ2Ͳlike Ribosomes, translation UP Yes 0.73 6.85 60SribosomalproteinL4 Ribosomes, translation UP 0.83 6.82 elongationfactor1Ͳalpha Ribosomes, translation UP 0.76 5.68 elongationfactor1Ͳalpha Ribosomes, translation UP 0.80 5.28 elongationfactor1Ͳalpha Ribosomes, translation UP 0.80 4.75 polyphenoloxidase,chloroplasticͲlike Secondarymetabolism UP 0.94 10.32 salutaridinereductaseisoform Secondarymetabolism UP 0.85 6.28 dihydropyrimidinaseisoform Secondarymetabolism UP 0.79 6.03 dirigentprotein2Ͳlike Secondarymetabolism UP 0.54 5.59 alphaͲterpineolsynthase,chloroplasticͲlike Secondarymetabolism DOWN Ͳ0.62 5.16 2Ͳalkenalreductase(NADP(+)Ͳdependent)Ͳlike Secondarymetabolism UP 0.45 5.00 tropinonereductasehomologAt5g06060 Secondarymetabolism UP 0.30 4.89 polyphenoloxidase,chloroplasticͲlike Secondarymetabolism UP 0.73 4.76 betainealdehydedehydrogenase2 Secondarymetabolism UP 0.71 4.68

124 nonͲfunctionalNADPHͲdependentcodeinonereductase2 Secondarymetabolism DOWN Ͳ0.74 4.55 heatshockcognate70kDaprotein2 Stressresponse UP 0.85 6.73 heatshock70kDaprotein14Ͳlike Stressresponse UP 0.77 6.41 universalstressproteinPHOS32Ͳlike Stressresponse UP 0.82 5.10 chaperoneproteinClpB3,chloroplasticisoform Stressresponse DOWN Ͳ0.72 4.57 betaͲgalactosidase8Ͳlike Sugarmetabolism UP Yes 0.79 9.98 LOWQUALITYPROTEINbetaͲgalactosidase6Ͳlike Sugarmetabolism UP 0.84 8.84 betaͲgalactosidase15isoform Sugarmetabolism UP Yes 0.61 8.24 alphaͲmannosidaseAt3g26720 Sugarmetabolism UP 0.91 8.07 endoͲ1,3;1,4ͲbetaͲDͲglucanase Sugarmetabolism DOWN Yes Ͳ0.63 7.07 alphaͲ1,4glucanphosphorylaseLisozyme,chloroplastic/amyloplastic Sugarmetabolism UP 0.77 5.96 mannose/glucoseͲspecificlectinisoform Sugarmetabolism UP Yes 0.69 5.88 proteinTIC110,chloroplastic Transporters DOWN Ͳ0.96 8.75 proteintranslocasesubunitSECA1,chloroplastic Transporters DOWN Ͳ0.81 5.93 ABCtransporterGfamilymember7 Transporters UP 0.78 5.01 VͲtypeprotonATPasecatalyticsubunitA Transporters UP 0.80 7.05 VͲtypeprotonATPasesubunitB2 Transporters UP 0.78 4.77 uncharacterizedproteinLOC105049839isoform Unknown DOWN Yes Ͳ0.80 13.00 uncharacterizedproteinLOC105047048 Unknown DOWN Yes Ͳ0.63 13.00 LOWQUALITYPROTEINRNAͲbindingprotein1Ͳlike Unknown UP 0.84 6.97 KHdomainͲcontainingproteinAt4g18375 Unknown UP 0.86 5.99 RNAͲbindingKHdomainͲcontainingproteinPEPPER Unknown UP 0.87 5.50 methylͲCpGͲbindingdomainͲcontainingprotein11 Unknown UP 0.63 5.11 uncharacterizedproteinLOC105034880 Unknown DOWN Ͳ0.82 4.76

Peptidessignificantforwaterloggingeffect Pathway Effect Interaction pcorr Ͳlog(P) alanineaminotransferase2 Aminoacidmetabolism DOWN Yes 0.53 6.18 alanineaminotransferase2isoform Aminoacidmetabolism DOWN Yes 0.44 4.61 5ͲmethyltetrahydropteroyltriglutamateͲͲhomocysteinemethyltransferase1 Aminoacidmetabolism DOWN 0.51 6.12 actinͲ101 Cytoskeleton UP Ͳ0.81 5.46 actinͲ101Ͳlikeisoform Cytoskeleton UP Ͳ0.75 5.33 actinͲ2 Cytoskeleton UP Ͳ0.79 4.79 pyrophosphateͲͲfructose6Ͳphosphate1Ͳphosphotransferasesubunitalpha Glycolysis DOWN Yes 0.52 5.20 uncharacterizedproteinLOC105058713,putativeformamidase Other UP Ͳ0.82 5.87 dehydrogenase/reductaseSDRfamilymember12 Oxidoreductases, UP Yes Ͳ0.17 13.00 general uncharacterizedaldoͲketooxidoreductase(NADP)At1g06690,chloroplastic Oxidoreductases, UP Ͳ0.71 7.58 general uncharacterizedaldoͲketooxidoreductase(NADP)At1g06690,chloroplastic Oxidoreductases, UP Ͳ0.75 5.49 general aldehydedehydrogenasefamily2memberB7,mitochondrial Oxidoreductases, UP Ͳ0.69 4.70 general violaxanthindeͲepoxidase,chloroplasticisoform Photosynthesis UP Yes Ͳ0.10 13.00 proteinLOWPSIIACCUMULATION3,chloroplasticisoform Photosynthesis DOWN Yes 0.14 13.00 photosystemIreactioncentersubunitV,chloroplasticͲlike Photosynthesis UP Ͳ0.59 6.36 fructoseͲ1,6Ͳbisphosphatase,chloroplastic Photosynthesis UP Ͳ0.75 4.88 zeaxanthinepoxidase,chloroplastic Photosynthesis UP Ͳ0.80 4.86 oxygenͲevolvingenhancerprotein2,chloroplasticͲlike Photosynthesis UP Yes Ͳ0.42 4.85 oxygenͲevolvingenhancerprotein3Ͳ1,chloroplastic Photosynthesis UP Yes Ͳ0.24 4.82 LOC105047047,hypotheticalthylakoidmembraneprotein Photosynthesis UP Ͳ0.75 4.67 rhodaneseͲlikedomainͲcontainingprotein4,chloroplastic Photosynthesis UP Ͳ0.70 4.58 subtilisinͲlikeproteaseSBT1.7 Proteases DOWN Yes 0.38 5.02 putativepeptidase,proteinLOC105058230isoform Proteases UP Ͳ0.77 4.75 uncharacterizedproteinLOC105059265,rubredoxinͲlikeprotein Redox DOWN 0.86 5.57 2Ͳoxoglutaratedehydrogenase,componentE1decarboxylating,mitochondrial RespirationandTCAP DOWN 0.67 5.45 ATPsynthasesubunitgamma,mitochondrial RespirationandTCAP DOWN 0.81 5.00 elongationfactorTuA,chloroplastic Ribosomes, translation DOWN 0.81 5.43 ubiquitinͲ40SribosomalproteinS27a Ribosomes, translation UP Ͳ0.82 4.78 elongationfactorTuA(tRNAandGTPbinding),chloroplastic Ribosomes, translation DOWN 0.75 4.61 2Ͳalkenalreductase(NADP(+)Ͳdependent)Ͳlike Secondarymetabolism UP Ͳ0.81 7.21 dirigentprotein2Ͳlike Secondarymetabolism UP Ͳ0.70 5.81 perakinereductaseisoform Secondarymetabolism UP Ͳ0.74 5.68 betaͲgalactosidase15isoform Sugarmetabolism DOWN Yes 0.49 7.99 polygalacturonaseͲlike Sugarmetabolism UP Ͳ0.86 7.11 alphaͲgalactosidase1 Sugarmetabolism DOWN 0.77 5.63 betaͲgalactosidase8Ͳlike Sugarmetabolism DOWN Yes 0.29 5.34 VͲtypeprotonATPasesubunitG Transporters UP Ͳ0.84 4.60 uncharacterizedproteinLOC105049839isoform Unknown UP Yes Ͳ0.04 13.00 uncharacterizedproteinLOC105047048 Unknown DOWN Yes 0.42 13.00 uncharacterizedproteinLOC105037336,WDͲ40propellerfamily Unknown DOWN Yes 0.77 6.58 proteinPHLOEMPROTEIN2ͲLIKEA1 Unknown UP Ͳ0.69 5.43 proteinPHLOEMPROTEIN2ͲLIKEA1 Unknown UP Ͳ0.66 4.96 germinͲlikeprotein2Ͳ4 Unknown UP Ͳ0.80 4.86

125 Table S3. List of best metabolites in the relationship with Rdark (Fig. 4) with variable importance for the projection (VIP) > 1 (most important metabolites). Abbreviations: PPP, pentose phosphate pathway; TCAP, tricarboxylic acid cycle. This table differentiates metabolites specifically related to Rdark (and thus less related to K or waterlogging) and those also strongly related to K and waterlogging (VIP > 1 in the corresponding OPLS in Fig. X). Here, the full analyte name was kept (with the number indicating that it is one of the silylated derivatives of the metabolite of interest). The loading in the relationship pcorr (here equal to the correlation coefficient r with Rdark) is also shown.

VIP pcorr(r) Variation SpecificallyrelatedtoRdark AlsoinvolvedinWL/K Metabolitegroup Pathwayorcomments withRdark groupdiscrimination 2.80Ͳ0.60 Down Uracil Pyrimidine EͲalaninepathway 2.36 0.45 Up Myoinositol Sugar Hexitols 2.22 0.44 Up Cellobiose_2 Sugar Disaccharides 2.21Ͳ0.45 Down Phytol Lipids/terpenoids Chlorophylldegradation 2.20 0.46 Up Catechine Phenylpropanoids 2.16 0.45 Up Gentiobiose_2 Sugar Disaccharides 2.14 0.46 Up EͲsitosterol Lipids/terpenoids Membranesterol 2.11 0.45 Up Ascorbate Sugar Redoxmetabolism 2.07 0.44 Up GlucoseͲ6Ͳphosphate_1 Sugar Glycolysis/photosynthesis 2.05Ͳ0.44 Down EͲalanine Polyamine EͲalaninepathway 2.00 0.40 Up 3Ͳphosphoglycerate Sugar Glycolysis/photosynthesis 1.92Ͳ0.38 Down FructoseͲ6Ͳphosphate Sugar Glycolysis/photosynthesis 1.87 0.39 Up Benzeneethanamine Aminoacid Phedegradation 1.84 0.38 Up Glucosylglycerol Sugar Glycolipidsynthesis 1.79 0.35 Up Xylitol Sugar Pentitols 1.77 0.35 Up Putrescine Polyamine Polyaminesandureacycle 1.72 0.35 Up 3,4,5Ͳtrihydroxybenzoate Phenylpropanoids 1.71 0.35 Up Maltose_1 Sugar Disaccharides 1.63Ͳ0.32 Down Ornithine_1 Polyamine Polyaminesandureacycle 1.60Ͳ0.30 Down GlycerolͲ3Ͳphosphate Sugar Glycolipidsynthesis 1.57 0.36 Up GlucoseͲ6Ͳphosphate_2 Sugar Glycolysis/photosynthesis 1.53Ͳ0.33 Down Serine_1 Aminoacid 1.53 0.34 Up Kaempferol Phenylpropanoids 1.51 0.29 Down Citrulline Polyamine Polyaminesandureacycle 1.47Ͳ0.30 Down Alanine_2 Aminoacids 1.46 0.31 Up Sucrose Sugar Disaccharides 1.46 0.29 Up Glucuronate_2 Sugar Aldaratemetabolism 1.40Ͳ0.31 Down Benzoate Aminoacid Phenylalanine degradation 1.35 0.29 Up AlphaͲtocopherol Lipids/terpenoids Redoxmetabolism 1.35Ͳ0.28 Down OͲacetylͲserine Aminoacid 1.34 0.31 Up Octadecanoate Lipids/terpenoids 1.32 0.26 Up Galactose_2 Sugar 1.31Ͳ0.26 Down 5Ͳaminolevulinate_1 Lipids/terpenoids Chlorophyllsynthesis 1.31Ͳ0.30 Down 3ͲmethylͲ2Ͳoxovalerate_1 Aminoacid Iledegradation 1.29 0.26 Up Fumarate Organicacid TCAP 1.25 0.26 Up Maltose_2 Sugar Disaccharides 1.22 0.25 Up MannoseͲ6Ͳphosphate Sugar 1.17 0.24 Up Raffinose Sugar 1.15Ͳ0.26 Down Threonine_1 Aminoacid 1.14Ͳ0.24 Down Lyxonate Sugar Aldaratemetabolism 1.13Ͳ0.23 Down Glucarate Sugar Aldaratemetabolism 1.13 0.25 Up Glucose_1 Sugar 1.13 0.23 Up Gentiobiose_1 Sugar Disaccharides 1.12 0.24 Up Glycerate Sugar Photorespiration 1.09 0.21 Up Maleate Organicacid TCAP 1.09Ͳ0.23 Down Xylulose Sugar PPP 1.09 0.21 Up TransͲaconitate Organicacid TCAP 1.08 0.23 Up 2Ͳmethylserine Aminoacid Alamethylation 1.06 0.19 Up Sorbitol Sugar Hexitols 1.04Ͳ0.23 Down 2Ͳoxogluconate Sugar Aldaratemetabolism 1.02 0.19 Up Pipecolate Aminoacid Lysdegradation 1.01Ͳ0.19 Down 2Ͳoxogluconate Sugar Aldaratemetabolism 1.01 0.20 Up ParaͲaminobenzoate Aminoacid Trpdegradation

126 Table S4. List of peptides in the specific (i.e. not related to waterlogging or K deficiency) relationship with Rdark. Proteins involved in catabolism (glycolysis, respiration, amino acid degradation) are in red, and correspond to red discs in Fig. S5. Other proteins correspond to black discs in Fig. S5. VIP, variable importance for the OPLS projection.

Peptides Loading VIP Variation pcorr withRdark Peptidesassociatedwithcatabolism: 3ͲhydroxyisobutyrylͲCoAhydrolaseͲlikeprotein3,mitochondrialisoform 0.49 1.24 Up ATPsynthasesubunitd,mitochondrial Ͳ0.40 1.07 Down citratesynthase,mitochondrial Ͳ0.44 1.24 Down 2Ͳoxoglutaratedehydrogenase,dihydrolipoyllysineͲresiduesuccinyltransferaseE2 0.42 1.12 Up component,mitochondrial fructoseͲbisphosphatealdolase1,cytoplasmic 0.42 1.35 Up fructoseͲ1,6Ͳbisphosphatase,cytosolic Ͳ0.54 1.38 Down glucoseͲ6Ͳphosphateisomerase,cytosolic1 0.35 1.05 Up isocitratedehydrogenase[NADP] 0.52 1.55 Up methylcrotonoylͲCoAcarboxylasesubunitalpha,mitochondrialisoform Ͳ0.39 1.04 Down mitochondrialcarnitine/acylcarnitinecarrierͲlikeprotein Ͳ0.47 1.31 Down mitochondrialcarnitine/acylcarnitinecarrierͲlikeprotein Ͳ0.37 1.10 Down mitochondrialoutermembraneproteinporin1 0.38 1.13 Up NADHdehydrogenase[ubiquinone]1alphasubcomplexsubunit6 Ͳ0.54 1.38 Down probableNADHdehydrogenase[ubiquinone]1alphasubcomplexsubunit5, 0.50 1.47 Up mitochondrial putative4ͲhydroxyͲ4ͲmethylͲ2Ͳoxoglutaratealdolase2 0.37 1.03 Up pyruvatedehydrogenaseE1componentsubunitalphaͲ1,mitochondrial 0.48 1.23 Up pyruvatedehydrogenaseE1subunitalphaͲ1,mitochondrialͲlikeisoform 0.41 1.19 Up succinateͲCoAligase[ADPͲforming]subunitalpha,mitochondrial 0.47 1.27 Up Otherpeptides(byreversedalphabeticalorder): zerumbonesynthaseͲlike 0.55 1.64 Up UPF0678fattyacidͲbindingproteinͲlikeproteinAt1g79260 0.42 1.12 Up UPF0603proteinOs05g0401100,chloroplastic Ͳ0.49 1.21 Down universalstressproteinPHOS34 0.40 1.06 Up universalstressproteinPHOS32 0.41 1.18 Up uncharacterizedproteinLOC105060389isoform 0.44 1.06 Up uncharacterizedproteinLOC105060161 Ͳ0.35 1.04 Down uncharacterizedproteinLOC105058674isoform 0.38 1.20 Up uncharacterizedproteinLOC105053842 0.41 1.25 Up uncharacterizedproteinLOC105046447 0.61 1.55 Up uncharacterizedproteinLOC105037888isoform Ͳ0.35 1.05 Down uncharacterizedproteinLOC105035433 0.63 1.65 Up uncharacterizedproteinAt4g14100Ͳlike 0.54 1.48 Up UDPͲglycosyltransferase88A1Ͳlike 0.42 1.06 Up UDPͲglycosyltransferase83A1Ͳlike 0.44 1.12 Up UBP1Ͳassociatedprotein2CͲlike 0.56 1.47 Up ubiquitinreceptorRAD23c 0.59 1.52 Up thylakoidlumenal16.5kDaprotein,chloroplasticisoform Ͳ0.39 1.16 Down thiosulfate/3Ͳmercaptopyruvatesulfurtransferase2 0.62 1.68 Up thiosulfate/3Ͳmercaptopyruvatesulfurtransferase1,mitochondrial 0.44 1.22 Up thioredoxinͲlikeproteinCDSP32,chloroplastic Ͳ0.40 1.10 Down TͲcomplexprotein1subunittheta 0.50 1.34 Up superoxidedismutase[Mn],mitochondrialͲlikeprecursor Ͳ0.37 1.02 Down sucroseͲphosphatase2isoform 0.45 1.20 Up subtilisinͲlikeproteaseSBT1.2 0.52 1.33 Up subtilisinͲlikeproteaseSBT1.2 0.52 1.31 Up serinecarboxypeptidaseͲlike27 0.45 1.31 Up serinecarboxypeptidaseͲlike27 0.40 1.01 Up ruBisCOlargesubunitͲbindingproteinsubunitalphaisoform Ͳ0.35 1.06 Down RNAͲbindingproteinCP33,chloroplastic 0.48 1.47 Up ribosomalproteinL14(chloroplast) Ͳ0.46 1.16 Down putrescinehydroxycinnamoyltransferase1Ͳlike 0.48 1.31 Up psbPdomainͲcontainingprotein6,chloroplastic Ͳ0.45 1.07 Down proteinͲribulosamine3Ͳkinase,chloroplastic 0.50 1.41 Up proteinTIC62,chloroplastic Ͳ0.53 1.40 Down proteinTHYLAKOIDFORMATION1,chloroplastic Ͳ0.59 1.59 Down proteinplastidtranscriptionallyactive16,chloroplastic Ͳ0.42 1.05 Down proteinLURPͲoneͲrelated15 0.42 1.29 Up proteinHHL1,chloroplasticisoform Ͳ0.47 1.31 Down proteinFLXͲlike2isoform 0.50 1.41 Up proteinCROWDEDNUCLEI1 0.47 1.17 Up proteasomesubunitbetatypeͲ2ͲB 0.38 1.07 Up proteasomesubunitbetatypeͲ1 0.42 1.07 Up 127 proteasomesubunitalphatypeͲ4 0.39 1.01 Up probableplastidͲlipidͲassociatedprotein11isoform Ͳ0.48 1.21 Down probableNͲacetylͲgammaͲglutamylͲphosphatereductase,chloroplasticisoform 0.40 1.15 Up probablemembraneͲassociated30kDaprotein,chloroplastic Ͳ0.42 1.07 Down probablemediatorofRNApolymeraseIItranscriptionsubunit37c 0.42 1.13 Up probablelipoxygenase8,chloroplastic Ͳ0.59 1.64 Down probablelactoylglutathionelyase,chloroplastic Ͳ0.39 1.01 Down probableinactivelinolenatehydroperoxidelyase Ͳ0.70 1.85 Down polypyrimidinetractͲbindingproteinhomolog3Ͳlikeisoform 0.38 1.02 Up polyadenylateͲbindingproteinRBP45Ͳlike 0.51 1.30 Up plastocyaninB'/B'' Ͳ0.64 1.66 Down photosystemIIrepairproteinPSB27ͲH1,chloroplastic Ͳ0.41 1.07 Down photosystemIIphosphoprotein(chloroplast) Ͳ0.61 1.61 Down photosystemIIcytochromeb559alphasubunit(chloroplast) Ͳ0.43 1.16 Down photosystemII10kDapolypeptide,chloroplastic Ͳ0.62 1.58 Down photosystemIreactioncentersubunitXI,chloroplasticͲlike Ͳ0.41 1.05 Down photosystemIreactioncentersubunitpsaK,chloroplastic Ͳ0.50 1.28 Down photosystemIP700apoproteinA2(chloroplast) Ͳ0.46 1.26 Down peptidylͲprolylcisͲtransisomeraseFKBP12isoform 0.43 1.26 Up patellinͲ3 0.40 1.17 Up patellinͲ3 0.50 1.41 Up oxygenͲevolvingenhancerprotein1,chloroplastic Ͳ0.48 1.25 Down oligouridylateͲbindingprotein1Bisoform 0.55 1.45 Up nucleosidediphosphatekinaseB 0.49 1.42 Up nucleosidediphosphatekinase2,chloroplastic 0.52 1.58 Up nonͲspecificlipidͲtransferprotein2Ͳlike Ͳ0.41 1.21 Down nonͲspecificlipidͲtransferprotein1Ͳlike Ͳ0.61 1.54 Down nitrogenregulatoryproteinPͲIIhomologisoform 0.38 1.12 Up multipleorganellarRNAeditingfactor9,chloroplastic Ͳ0.44 1.15 Down monodehydroascorbatereductase 0.54 1.59 Up malatedehydrogenase,chloroplastic 0.48 1.20 Up malatedehydrogenase[NADP]1,chloroplasticisoform Ͳ0.53 1.38 Down magnesiumprotoporphyrinIXmethyltransferase,chloroplastic Ͳ0.43 1.20 Down LOWQUALITYPROTEINuncharacterizedproteinLOC105041010 0.38 1.10 Up LOWQUALITYPROTEINphotosystemIreactioncentersubunitIII,chloroplastic Ͳ0.41 1.09 Down LOWQUALITYPROTEINbetaͲhexosaminidase1 0.53 1.46 Up LOWQUALITYPROTEINalanineͲͲtRNAligaseͲlike 0.44 1.08 Up longchainacylͲCoAsynthetase4 0.37 1.04 Up leukotrieneAͲ4hydrolasehomologisoform 0.45 1.05 Up leucineͲrichrepeatextensinͲlikeprotein2 0.41 1.03 Up leucineaminopeptidase1 0.43 1.27 Up lachrymatoryͲfactorsynthaseͲlike Ͳ0.44 1.18 Down Keratin,typeIcytoskeletal16(CytokeratinͲ16)Homosapiens 0.36 1.15 Up hypersensitiveͲinducedresponseprotein1 0.53 1.43 Up histoneH2BͲlike 0.63 1.59 Up histoneH1Ͳlike 0.50 1.33 Up histoneH1Ͳlike 0.48 1.25 Up highmobilitygroupBprotein1 0.53 1.39 Up hexokinaseͲ2 0.49 1.34 Up heterogeneousnuclearribonucleoprotein1Ͳlike 0.46 1.21 Up heatshockprotein81Ͳ1 0.38 1.05 Up heatshock70kDaprotein 0.40 1.02 Up granuleͲboundstarchsynthase2,chloroplastic/amyloplastic Ͳ0.50 1.38 Down glyoxysomalfattyacidbetaͲoxidationmultifunctionalproteinMFPͲa 0.39 1.08 Up glycineͲrichRNAͲbindingproteinGRP2A 0.40 1.16 Up glycineͲrichRNAͲbindingprotein2,mitochondrial 0.40 1.11 Up glutathionesynthetase,chloroplasticisoform 0.38 1.10 Up glutamateͲͲglyoxylateaminotransferase2 Ͳ0.52 1.36 Down glucoseͲ1Ͳphosphateadenylyltransferasesmallsubunit,chloroplastic/amyloplasticͲlike Ͳ0.42 1.05 Down glucoseͲ1Ͳphosphateadenylyltransferaselargesubunit1,chloroplasticͲlike Ͳ0.44 1.17 Down glucanendoͲ1,3ͲbetaͲglucosidaseͲlike 0.45 1.17 Up germinͲlikeprotein8Ͳ14 0.53 1.30 Up GDSLesterase/lipaseAt3g26430 Ͳ0.48 1.25 Down gammaͲinterferonͲinduciblelysosomalthiolreductase 0.41 1.10 Up fruitproteinpKIWI502 Ͳ0.46 1.18 Down fructoseͲbisphosphatealdolase1,chloroplastic 0.44 1.24 Up fructoseͲbisphosphatealdolase1,chloroplastic 0.52 1.32 Up FGGYcarbohydratekinasedomainͲcontainingproteinisoform Ͳ0.38 1.08 Down ferredoxinͲdependentglutamatesynthase,chloroplasticisoform Ͳ0.46 1.16 Down ferredoxinͲdependentglutamatesynthase,chloroplastic Ͳ0.51 1.37 Down farupstreamelementͲbindingprotein3Ͳlike 0.49 1.40 Up eukaryotictranslationinitiationfactor3subunitDͲlike 0.47 1.24 Up eukaryotictranslationinitiationfactor3subunitA 0.47 1.21 Up endoplasminhomolog 0.42 1.19 Up elongationfactorTu,mitochondrialͲlike 0.55 1.47 Up elongationfactorTu,mitochondrial 0.53 1.42 Up elongationfactorTu,mitochondrial 0.56 1.50 Up DNArepairRAD52Ͳlikeprotein2,chloroplastic 0.41 1.14 Up 128 DEADͲboxATPͲdependentRNAhelicase56isoform 0.45 1.23 Up cytosolicendoͲbetaͲNͲacetylglucosaminidase1 0.63 1.78 Up cytochromeb6/fcomplexsubunitIV(chloroplast) Ͳ0.37 1.06 Down cysteinesynthase Ͳ0.45 1.13 Down coppertransportproteinAT 0.48 1.16 Up cinnamoylͲCoAreductase1 0.38 1.12 Up chorismatesynthase2,chloroplastic 0.49 1.30 Up chorismatesynthase2,chloroplastic 0.51 1.34 Up chloroplasticlipocalinisoform Ͳ0.49 1.34 Down chloroplaststemͲloopbindingproteinof41kDaa,chloroplastic Ͳ0.53 1.34 Down chlorophyllaͲbbindingproteinCP2410A,chloroplasticͲlike Ͳ0.44 1.02 Down chlorophyllaͲbbindingprotein4,chloroplastic Ͳ0.41 1.12 Down chitinase6Ͳlike 0.46 1.26 Up chalconeͲͲflavononeisomerase 0.50 1.41 Up celldivisioncycleprotein48homolog 0.42 1.17 Up celldivisioncycleprotein48homolog 0.43 1.18 Up celldivisioncycleprotein48homolog 0.50 1.35 Up calnexinhomolog 0.46 1.28 Up biotincarboxylase2,chloroplastic 0.43 1.02 Up auxinͲbindingproteinABP20Ͳlike 0.62 1.53 Up ATPͲdependentClpproteaseproteolyticsubunit6,chloroplastic Ͳ0.40 1.24 Down ATPͲdependentClpproteaseproteolyticsubunit5,chloroplastic Ͳ0.39 1.02 Down ATPͲdependentClpproteaseATPͲbindingsubunitClpAhomologCD4A,chloroplasticͲlike Ͳ0.59 1.54 Down ATPsynthaseCF1epsilonsubunit(chloroplast) Ͳ0.43 1.16 Down ATͲhookmotifnuclearͲlocalizedprotein9 0.40 1.14 Up ATͲhookmotifnuclearͲlocalizedprotein10 0.40 1.11 Up asparagineͲͲtRNAligase,cytoplasmic3isoform 0.46 1.09 Up asparagineͲͲtRNAligase,cytoplasmic1 0.41 1.01 Up apoptoticchromatincondensationinducerinthenucleusisoform 0.56 1.54 Up aminomethyltransferase,mitochondrial Ͳ0.59 1.47 Down amidase1Ͳlike Ͳ0.43 1.03 Down alphaͲxylosidase1 Ͳ0.44 1.05 Down alphaͲmannosidaseͲlikeisoform 0.38 1.02 Up alphaͲmannosidaseAt3g26720Ͳlikeisoform 0.36 1.15 Up alanineͲͲtRNAligaseͲlike 0.51 1.31 Up adenosinekinase2Ͳlike 0.50 1.27 Up acidphosphatase1 0.37 1.09 Up 60SribosomalproteinL7Ͳ2 0.36 1.01 Up 60SribosomalproteinL3isoform 0.36 1.03 Up 60SacidicribosomalproteinP0 0.37 1.12 Up 50SribosomalproteinL12,chloroplasticͲlike Ͳ0.43 1.18 Down 4ͲhydroxyͲ3ͲmethylbutͲ2ͲenͲ1Ͳyldiphosphatesynthase(ferredoxin),chloroplastic 0.41 1.05 Up 40SribosomalproteinS20Ͳ1 0.53 1.43 Up 40SribosomalproteinS10Ͳ1 0.41 1.05 Up 3Ͳisopropylmalatedehydrogenase2,chloroplastic 0.48 1.33 Up 30SribosomalproteinS1,chloroplastic Ͳ0.51 1.28 Down 28kDaribonucleoprotein,chloroplasticͲlike Ͳ0.45 1.22 Down 26SproteasomenonͲATPaseregulatorysubunit4homologisoform 0.42 1.20 Up 26SproteasomenonͲATPaseregulatorysubunit14homologisoform 0.61 1.52 Up 26SproteasomenonͲATPaseregulatorysubunit1homologA 0.46 1.08 Up 26Sproteaseregulatorysubunit7A 0.40 1.08 Up 14Ͳ3Ͳ3Ͳlikeproteinisoform 0.47 1.19 Up 14Ͳ3Ͳ3Ͳlikeprotein 0.53 1.39 Up

129 Fig. S1. Schematic representation of the experiment carried out in this paper. Note that the total duration of the experiment was 12 months and 3 weeks, and that there were 7 sampling campaigns, representing 140 trees overall. The experiment was organized in such a way that the effect of K conditions alone (monofactorial experiment) was examined before the onset of waterlogging, such that the age of oil palm trees was not exactly the same between the monofactorial experiment and the crossed effect experiment (maximal difference of 3 months, between first and last sampling dates). The color code used throughout the paper is recalled at the bottom of this scheme.

130 35 0.120 aaaa abcd (a) (d) ) 30 abac ) Ͳ 1 Ͳ 1 0.100 g g abac 25 0.080 (mg (mg 20 abab 0.060 15 content abcd content 0.040 10 abbc c abaab abbc 0.020 Element 5 Element abca 0 0.000 Ca K Mg P S Na Fe Zn Cu Mn

35 0.06 aaaa abcd (b) (e) ) 30 Ͳ 1 ) 0.05 1 g Ͳ g 25 0.04 (mg (mg 20 abca abcd 0.03 15 abcac content abcd abbc content abcd 0.02 10 aabc 0.01 Element 5 abcb Element 0 0 Ca K Mg P S Na Fe Zn Cu Mn

80 abcb (c) 1.6 abab (f) 70 1.4 ) 1 Ͳ

g 60 1.2 (mg/g) 50 1 (mg 40 0.8 content 30 0.6 content abab abcc 20 0.4 abcb Element abcd 10 abcb 0.2 Element abbb abcc aaaa 0 0 Ca K Mg P S Na Fe Zn Cu Mn

Fig. S2. Elemental composition of oil palm organs (at sampling date 7) in the K u waterlogging crossed experiment (, low K; , high K; , low K + waterlogging; , high K + waterlogging): leaves (a, d), stem (b, e) and roots (c, f). this figure presents macronutrients on left (a, b, c) and micronutrients on right (d, e, f). Elemental contents were determined using IC-OES. Note that K deficiency is compensated for by elevated calcium (Ca) and magnesium (Mg) in leaves and roots, and also sodium (Na) and iron (Fe) in roots. This figure shows meanrSD (n = 5). For each chemical element, letters stand for statistical classes (P < 0.05) when comparing conditions.

131 Fig. S3. GC-MS metabolomics pattern in oil palm leaves under low, medium or high K nutrition, before the onset of waterlogging. Here, “resupply” means cultivation at low K (0.2 mM) for 9 months and then at 4 mM (high K). (a) OPLS analysis showing that K cultivation conditions could be discriminated easily (qualitative Y variable), and high K and resupply could not be differentiated (single spot of blue points)); (b) hierarchical clustering showing that high K and resupply conditions are indistinguishable; (c) loading plot of the OPLS to show most important discriminating metabolites for each condition; (d) heat map of significant metabolites (one-way ANOVA, threshold P-value of 0.05 with Bonferroni correction, i.e. 310-4; blue cells: decrease; red cells: increase), showing three groups: metabolites less abundant under extreme conditions (both low and high K) mostly after 11 months growth (top), metabolites increased under low K (middle), and metabolites decreased under low K (bottom). The -9 OPLS model was siignificant (PCV-ANOVA < 10 ) and explained most of the variance in the data (R² = 0.972; Q² = 0.864). Note that the effect of K is not simply linear and thus an OPLS model with K conditions as a quantitative Y variable would be much less robust than the model presented here (K conditions as qualitative Y variable). In each case (age and K level), n = 5.

132 Fig. S4. Quantitation of major metabolites by 13C-NMR in oil palm leaves (at natural 13C abbundance), under low (green) or high (blue) K conditions. Main panel: overview of the full spectrum. Insets: magnification of the 23-60 ppm region (a) and the 23.6-24.4 ppm region (b) (shadowed in (a)). Note the larger amount of ornithine (o) under high K and its decrease at low K while putrescine (u) increases. Citrulline and spermidine (common peaks i) are visible under both K conditions. CDTA (black arrows) was used to have a good peak definition (resolution) at high chemical shift (COOH region). Also note the considerable increase in citrate (c) and malate (m) at low K (but the very low content in fumarate, hardly visible at 136 ppm, pink arrow). Legend: a, aspartate; c, citrate; g, glutamate; i, citrulline + spermidine (could not be resolved due to very close chemical shifts); m, malate; o, ornithine; p, phenylpropanoid; s, chlorogenate-like compound; t, succinate; u, putrescine. NMR acquisition was performed on perchloric extracts neutralized to pH 7 (with KOH), with a proton-decoupled pullse program (zgig) using a Waltz16 sequence, accumulated over 10,000 scans. Here, the two spectras are displayed with the same height of the internal standard maleate (M) which represented 125 μmol added to each sample.

133 Fig. S5. Multivariate relationship between leaf dark respiration and protein content: (a) observed versus predicted relationship; (b) volcano plot showing the most important discriminating proteins (peptides): grey, proteins unrelated (VIP < 1) or also related to waterlogging/K discrimination (Fig. 2 in main text); red and black, proteins related to respiration (VIP > 1) and not to waterlogging or K (VIP < 1 in Fig. 2) with enzymes involved in catabolism (red) and other proteins (black). The OPLS model generated used Rdark as a quantitative variable. The statistical model was highly significant (PCV-ANOVA < 10-6) and predictive (R² = 0.968, Q² = 0.789). Proteins related to respiration (black and red) are listed in Table S4.

134 Fig. S6. GC-MS metabolomics pattern in oil palm roots under low, medium or high K nutrition. Here, “resupply” means cultivation at low K (0.2 mM) for 9 months and then at 4 mM (high K) from 10 or 11 months. (a) OPLS analysis showing that K cultivation conditions could be discriminated easily (qualitative Y variable), and high K and resupply could not be differentiated (single spot of blue points); (b) hierarchical clustering showing that high K and resupply conditions are indistinguishable; (c) loading plot of the OPLS to show most important discriminating metabolites for each condition; (d) heat map of significant metabolites (one-way ANOVA, threshold P-value of 0.05 with Bonferroni correction, i.e. 310- 4), showing two groups: metabolites less or more abundant under extreme conditions (both low and high K) mostly after 11 months as well as metabolites decreased under low K (bottom); metabolites increased -9 under low K (top),. The OPLS model was significant (PCV-ANOVA < 410 ) and explained most of the variance in the data (R² = 0.960; Q² = 0.834). Note that tthe effect of K is not simply linear and thus an OPLS model with K conditions as a quantitative Y variable would be much less robust than the model presented here (K conditions as qualitative Y variable). In each case (age and K leevel), n = 5 except at 10 and 11 months after resupply (n = 10). In (d), note the specific behaviour of Į-tocopherol which is elevated under both low and high K at 9 months. Further note the strong similarity with Fig. S3 (with putrescine as the low K biomarker) except at medium K were the best leaf biomarker was ornithine while it was glucarate in roots.

135 glycolysis Ͳ HCO3 PEPC PEP ADP ATP pyruvate MH2OGA

CO2

CO2 acetylͲCoA citrate ME ATP oxaloacetate CO2 aconitate itaconate citramalate malate ADP+Pi Krebscycle CO isocitrate 2 aspartate parapyruvate CO2 2Glu fumarate 2OG 2OG arginine ADP+Pi methylͲglutamate CO2 succinylͲCoA CO2 succinate putrescine ATP

Fig. S7. Summary of plausible catabolic pathways involved in CO2 production, showing steps affected by low K or waterlogging (green), putrescine synthesis (pink) and the involvement of 4-methyl- 4-hydroxy-2-oxoglutarate aldolase (MH2OGA), phosphoenolpyruvate (PEPC) + malic enzyme (ME) to regenerate pyruvate when pyruvate kinase activity is limiting. Further details on the polyamine/ȕ-alanine pathway are given in Fig. 3 (main text). References Carroll AJ, Badger MR, Harvey Millar A. 2010. The MetabolomeExpress Project: enabling web-based processing, analysis and transparent dissemination of GC/MS metabolomics datasets. BMC Bioinformatics 11: 376. Kessner D, Chambers M, Burke R, Agusand D, Mallick P. 2008. ProteoWizard: open source software for rapid proteomics tools development. Bioinformatics 24: 2534-2536. Langella O, Valot B, Balliau T, Blein-Nicolas M, Bonhomme L, Zivy M. 2017. X!TandemPipeline: a tool to manage sequence redundancy for protein inference and phosphosite identification. Journal of Proteome Research 16: 494-503. McIlwain S, Mathews M, Bereman MS, Rubel EW, MacCoss MJ, Noble WS. 2012. Estimating relative abundances of proteins from shotgun proteomics data. BMC bioinformatics 13: 308. Méchin V, Damerval C, Zivy M 2007. Total protein extraction with TCA-acetone. In: Thiellement H, Zivy M, Damerval C, Méchin V eds. Plant proteomics: methods and protocols. Totowa: NJ Humana Press, 1-8.

136 Chapter 5 - General discussion and perspectives

In the present thesis, emphasis has been given to metabolic aspects of potassium nutrition and waterlogging. For the two species of interest (sunflower, oil palm), a specific discussion has been provided in associated papers attached to this manuscript. However, as expected, features common to the two species have been found. In particular, the impact of K deficiency and waterlogging on nitrogen metabolism and plant N fluxes as well as respiration in darkness were observed in both sunflower and oil palm. Also, specific metabolites have been observed under K deficiency (putrescine) in both species, and this deserves further examination of their metabolic roles. The present discussion elaborates on this, taking advantage of preliminary results obtained from analyses of natural 15N abundance. Then limitations and perspectives of the present work are discussed. The specific discussion on the roles of putrescine is attached as an appendix that has been published as a paper to Plant Cell and Environment.

5.1 Comparison of 15N/14N fluxes in sunflower and oil palm under varying K availability and waterlogging (this section has been extended and published in New Phytologist during this thesis been reviewed, see more details in Appendix 1)

In both species (oil palm, sunflower), the present work has shown that K nutrition and waterlogging led to changes not only in amino acid content (as well as other nitrogenous compounds such as polyamines) but also in elemental N content in leaves. In sunflower, detailed analysis of leaf nitrates and N demand for growth has been done and we found that presumably, waterlogging and K deficiency inhibited nitrate transport and thus changed the balance between nitrate influx and nitrate assimilation in leaves. However, at the time of writing the paper, only leaves were analysed and further work was required on other organs and also oil palm. 14N /15N (and 12C/13C) natural abundance of various N or C-containing metabolites provide an useful tool in understanding the metabolic flux between different organs in plants (Yoneyama et al., 2003). In fact, the preference one isotope over the other(s) of transport processes or/and enzyme reactions leads to isotope enrichment or depletion, a process referred to as isotope fractionation (Tcherkez and Hodges, 2007). In practice, there is an isotope fractionation during nitrate assimilation so that plant organic matter is naturally 15N-depleted compared to source nitrate. This comes from the 14v/15v isotope effect (ratio of rates) associated with nitrate transport through

137 membranes in root hairs (probably about 1.003) and reduction by nitrate reductase (1.016). Here, I determined the natural 15N abundance in N-containing fractions including nitrate, total soluble fraction, proteins in different organs (root, stem and leaves) in both sunflower and oil palm. The results –still preliminary– are briefly described here so as to show how they may bring new perspectives for future work.

Figures 1 and 2 show plots that incorporate all data obtained on leaves, with the apparent fractionation Δ15N (i.e., δ15N of raw organic matter relative to source nitrate) plotted against mean-centered N elemental content. Figure 1 also replots data already shown in Chapter 3 on sunflower, plus data obtained at medium K conditions, and K resupply after temporary deficiency. Clearly, low K leads to a decline in %N in sunflower leaves, and waterlogging exaggerates this effect (Figure 1). Medium K conditions and resupply conditions co-localize with high K conditions in this graph, showing that nutrient conditions are already K-sufficient at medium K. However, in oil palm, %N at high K is higher than that at low K regardless of waterlogging (Figure 2), and there is an apparent ranking of conditions nearly forming a straight line. In that sense, ∆15N displays a similar pattern in sunflower and oil palm. That is, waterlogging and low K lead to a high fractionation (higher depletion in 15N) and a low N content. At first glance, this is not surprising since the fractionation is usually larger as N assimilation is low, simply because a low assimilation flux allows substantial fractionation (Tcherkez and Hodges, 2007).

138 7.0

6.0

5.0

4.0 N

15 3.0 ∆ 2.0

1.0

0.0 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 -1.0 D LK HK LK+WL HK+WL MK LK+HK resupply Figure 1. Relationship between apparent fractionation (total organic matter) 15 15 15 15 and N content in sunflower leaves. ∆ N= (δ Nleaf − δ Nsource nutrition)/ (δ Nsource nutrition + 1); D is the relative N content (mean-centered) calculated as (%Nleaf − %Naverage)/SD. All data compiled from plant grown at low K (LK), medium K (MK), high K (HK), with or without waterlogging (WL).

12.0

10.0

8.0 N

15 6.0 ∆

4.0

2.0

0.0 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 D LK HK LK+WL HK+WL

Figure 2. Relationship between apparent fractionation (total organic matter) 15 15 15 15 and N content in oil palm leaves. ∆ N= (δ Nleaf − δ Nsource nutrition)/ (δ Nsource nutrition + 1); D is the relative N content (mean-centered) calculated as (%Nleaf − %Naverage)/SD. All data compiled from plant grown at low K (LK), high K (HK), with or without waterlogging (WL).

139 Of course, Figures 1 and 2 are not fully representative because total N in tissues includes nitrate, and thus both the Δ15N and D values are averages of organic and inorganic nitrogen. This thus makes difficult deciphering whether (i) low K availability and waterlogging led to modifications of nitrogen metabolism and nitrate cycling, and (ii) there were differences between sunflower and oil palm. To clarify these aspects, Figure 3 and 4 summarizes data obtained so far in nitrates, and also proteins. Proteins are here used as compounds reflecting the δ15N of amino acids and thus primarily assimilated N (while the total soluble fraction would also contain nitrate and other nitrogenous compounds that can be synthesized with isotope fractionation). As discussed before in the published paper, in sunflower, leaves contain less nitrate at high K, showing the larger consumption rate (assimilation) compared to influx via the xylem than at low K. This interpretation is reinforced by the fact that in roots, nitrate is not substantially lower at high K while it is lower in stems. By contrast, waterlogging caused a decline in nitrate content in all organs in particular at low K, suggesting an inhibition of nitrate absorption, consistent with the literature (further discussed in the Introduction of the paper). However, it is striking to see that the δ15N in leaf nitrate is roughly constant across sampling time and K/waterlogging conditions (Figure 3b1). This is an unexpected result, not only because nitrate utilisation and transport (and the nitrate pool size) have changed, but also because nitrate is progressively 15N-depleted under low K and waterlogging conditions in both root and stem (Figure 3b2, 3b3). However, these differences between leaf and other organs do not hold in oil palm (Figure 4). In effect, in particular under high K conditions, δ15N in leaf nitrate increased over time (Figure 4b1). Also, root nitrate content in oil palm decline with time while no change was observed in sunflower (Figure 4a3). Taken as a whole, both the nitrate transfer from roots to leaves and assimilation have different dynamics in sunflower and in oil palm. In other words, the coordination between decreased metabolic capacity (and biomass production) in leaves and lower supply from roots under K deficiency does not involve the same flux adjustments in the two species. As a matter of fact, the δ15N value in leaf nitrate remains nearly constant in both species at low K while leaf nitrate pool size is not.

Also, of particular interest is the isotope difference in nitrate between leaves and roots, in particular in sunflower. The preference of 14N over 15N by nitrate reductase is so that nitrate in root tissue should be less enriched than in leaves, because nitrate molecules left behind (that are not assimilated) by roots are transported to stem and leaves where

140 non-assimilated nitrate is further 15N-enriched due to nitrate assimilation (Tcherkez, 2010). But here, the δ15N value in nitrate roots is much higher that that in leaves, in both species (Figure 3-4). In other words, leaf nitrates are much less 15N-enriched than expected. There are two possible explanations (similar to what has been proposed to explain the lower-than-expected 18O-enrichment in leaf water, Barbour, 2007):

• First, nitrate may be compartmentalized in root tissues after having been absorbed. One pool would be of small size and metabolically active, used for export to stems and nitrate assimilation, while the other one would be a storage and would exchange slowly with the first pool. While it is hard to imagine how this compartmentalization takes place since nitrate reductase is a cytosolic protein (Tischner, 2000), it is possible that vacuolar nitrate plays the role of the second, slow turned-over pool. Still, the big storage pool could accumulate 15N-enriched nitrate while the metabolically active pool would be renewed (turned-over) too quickly to show significant enrichment.

• Second, a backflow of nitrate from leaves to the xylem. Since leaf nitrate pool is progressively 15N enriched, this creates a gradient of 15N that is compensated for by the retro-diffusion of 15N-nitrate against the xylemian convection flux (Péclet effect). Alternatively, there could be a “false” Péclet effect whereby leaf nitrate simply feeds the phloem and can thus go back to stem and roots.

Of course, another possibility is that nitrate transporters responsible for loading - 15 NO3 into the xylem fractionates, leading to a depletion in N-nitrate in the xylem and thereby in leaves. Although possible, such a fractionation should be small (in the per mil range or less) and thus cannot explain numerically our results (in sunflower, leaves are up to 35‰ less enriched than expected). In favor of the first hypothesis is the fact that in sunflower roots, proteins (resulting from N assimilation) are always depleted by much more than 16‰, showing that the bulk nitrate pool is likely not the source of nitrate used for assimilation.

Clearly, further work is required to sort out the origin of the δ15N pattern, and to compute fluxes. In fact, modelling using the N demand for biomass production, the size of nitrate pools and δ15N values combined with a set of differential equations can now be carried out to solve for flux values (nitrate circulation fluxes, assimilation fluxes). This exercise is beyond the scope of the present thesis and will be addressed in the future.

141 Nevertheless, it should be recognized that our extensive data on δ15N that includes nitrate (and represent more than 1,500 extractions and isotopic analyses) are very interesting because it provides for the first time a precise picture of leaf nitrate δ15N dynamics.

142

Figure 3. Time course of δ15N and nitrate in sunflower leaves, stems and roots. Nitrate content (a1: leaf; a2: stem; a3: root). δ15N in nitrate (b1: leaf; b2: stem; b3: root). δ15N in proteins (c1: leaf; c2: stem; c3: root). Time is shown on the x axis in weeks after the onset of waterlogging. Mean ± SD (n = 5)

143

Figure 4. Time course of δ15N and nitrate in oil palm leaves, stems and roots. Nitrate amount (a1: leaf; a2: stem; a3: root). δ15N in nitrate (b1: leaf; b2: stem; b3: root). δ15N in proteins (c1: leaf; c2: stem; c3: root). Time is shown on the x axis in days after the onset of waterlogging. Mean ± SD (n =5).

144 5.2 The control of respiration by K availability and thus the impact on carbon use efficiency: summary

Our results have shown that low K conditions decrease net CO2 assimilation (A) and stomatal conductance (gn), but increase considerably leaf respiration in darkness (Rdark) in both sunflower and oil palm. However, waterlogging effects seem to be different between oil palm and sunflower. In oil palm, waterlogging causes a decrease in A and gn but an increase in Rdark regardless of K availability. In sunflower, generally the opposite trend of A and gn was observed, while Rdark increased as well in low K and waterlogging conditions. The effect of K deficiency on respiration rate in darkness has been known for a long time (Okamoto, 1967; Terry and Ulrich, 1973; Azcón-Bieto and Osmond, 1983) but its precise origin is still not very well known. That is, the metabolic origin of the higher CO2 production has never been addressed directly. Here, I summarize potential explanations, with pathways involved and changes in the abundance of several mitochondrial proteins, as revealed by our omics data.

As explained in the Introduction, low K (but also waterlogging) inhibit glycolysis and Krebs cycle steps, shown in green in Figure 5. Pyruvate kinase and succinate thiokinase are inhibited by low K availability since K+ is a cofactor for the reaction. Aconitase is inhibited, probably by both K deficiency and waterlogging, due to oxidative stress and NO generation. In particular, NO and H2O2 has been found to inhibit the enzyme (Verniquet et al., 1991; Gupta et al., 2012). Inhibition of these three steps have pervading consequences and it seems plants use alternative pathways to produce carbon skeletons and energy to survive under such adverse environments (O'Leary et al., 2019).

These alternative pathways may contribute to the enhanced CO2 evolution.

145

Figure 5. Summary of plausible catabolic pathways involved in CO2 production. Glu, glutamate; 2OG, 2 oxoglutarato; ME, malic enzyme; PEP, phosphoenolpyruvate; PEPC, PEP carboxylase.

• When pyruvate kinase becomes limiting, pyruvate may be synthesized via malate, which can indeed be converted to pyruvate via the activity of malic enzyme (ME). A

net CO2 production would be expected, if this reaction is not coupled to bicarbonate fixation by PEPC (oxaloacetate production then reduced to malate). In my study, PEPC and ME protein abundance was not significantly affected by K availability (Chapter 4 Figure 2). In addition, in oil palm, we found an alternative pathway for pyruvate generation, via parapyruvate aldolase (MH2OGA in Figure 5). However, it is worth noting that the abundance is not necessary reflecting enzymatic activities as many enzymes are controlled posttranslationally or via effectors.

146 • Multivariate analysis of metabolite profiles suggested the importance of sugars (including glucose-6-phosphate) of the oxidative pentose phosphate pathway in

contributing to Rdark. Glucose-6-phosphate, which can be converted into pentitols, was positively related to respiration in the OPLS analysis, while glucarate and 2- oxogluconate were negatively correlated with respiration (Figure 6). This strongly suggests that the decarboxylation in the oxidative pentose phosphate pathway via

gluconate, which produce CO2, is favored over oxidation to 2-oxogluconate.

Figure 6. Key metabolites of the oxidative pentose phosphate pathway found to be significant in my study.

147 • The production of putrescine may also contribute to increase Rdark. Induction of the production of putrescine by low K conditions has been reported for decades (Jones, 1961, 1966; Okamoto, 1966; Freeman, 1967; Okamoto, 1967, 1968; Besford and Maw, 1976; Armengaud et al., 2009; Hussain et al., 2011; Jwakyung et al., 2015). The

production of putrescine is in principle associated with CO2 production regardless of the pathway (from ornithine or agmatine) (Figure 7). An enzyme of putrescine

metabolism, hydroxycinnamoyl transferase, was found to be associated with Rdark suggesting the importance of this pathway. Also, in quantitative terms (absolute quantitation in moles), we could detect putrescine in significant amount in oil palm using 13C-NMR, meaning that putrescine synthesis is a small but not negligible source

of CO2 production.

Figure 7. Simplified metabolic pathway of putrescine synthesis and utilization. (a) Chemical structure of putrescine. Note that it contains two N atoms and four C atoms, that all come from glutamate. (b) Pathways showing the direct route starting from glutamate via ornithine (black), putrescine synthesis via arginine (gray) and other polyamines synthesis (blue). Cofactors and other compounds involved in reactions are shown in green or light turquoise. The alternative use of N- acetylornithine as an acetyl donor is shown in dashed green. The recycling of fumarate via the Krebs cycle and aspartate synthesis, and the recycling of ammonium by carbamoyl phosphate synthase are shown in dotted light turquoise. Abbreviations: 2OG, 2-oxoglutarate; ADC, arginine decarboxylase (chloroplastic); CP, carbamoyl phosphate; NAG, N-acetyl glutamate; NAGSA, N-acetyl glutamate semialdehyde; ODC, ornithine decarboxylase (cytosolic); P-NAG, phospho-N- acetyl glutamate; SAE, S-adenosyl methioninamine; SAM, S-adenosyl methionine; SMTA, S-methyl thioadenosine.

148 • Recycling aconitate to citramalate could also contribute to increase Rdark. Convertion

of aconitate to citramalate plays important role in the C5 branched pathway when there is a limiting step in the Krebs cycle, in particular if the conversion of aconitate to isocitrate (by aconitase), and the conversion of succinyl-CoA to succinate (by succinyl-CoA thiokinase) is slow because of low K (and waterlogging). On average, this alternative pathway converts one succinate for two aconitate consumed, the other aconitate molecule being converted to citramalate (Figure 8). Overall, since it requires

extra ATP to generate itaconyl-CoA and citramalate, it leads to a higher CO2 release per ATP generation.

Figure 8. Citramalate biosynthetic pathway (overall equation at bottom). 2OG, 2-oxoglutarate.

Although they are well-supported statistically by our omics analyses, these metabolic reasons for the increased CO2 production would require a precise quantitative assessment so as to know whether their contributions are significant.

149 5.3 Distinct effect of waterlogging on leaf respiration in oil palm and sunflower

Waterlogging altered the dark respiration of both oil palm and sunflower in different K conditions, but in opposite directions. In sunflower, waterlogging suppressed the low K caused increase of dark respiration, while no effect under high K condition (Chapter 3 Figure 1). However, in oil palm, waterlogging increase dark respiration under both low K and high K condition (Chapter 4 Figure 1). The synthesis of putrescine contributes to

CO2 evolution for both species. The waterlogging caused putrescine profile different together with its effect on other metabolic pathways may explain the difference effect of waterlogging on leaf respiration of the two plant species.

In sunflower, the leaf respiration rate correlated to putrescine and citramalate (Chapter 3 Figure S5). Low K induced putrescine and citramalate which causes the increase evolution of CO2. Waterlogging decreased putrescine content (Chapter 3 Figure S4) which then can explain why it caused suppression of leaf respiration.

Similarly, in oil palm, the increase in putrescine caused by low K in principle associated with CO2 production. Different to waterlogging effect on sunflower, waterlogging did not change the putrescine content in both low K and high K condition (Chapter 4 Figure 3). The proteomic data suggests that direct impact of waterlogging on glycolysis and the Krebs cycle (altered pyruvate kinase, aconitase, and succinate thiokinase activity) and alternative pathways can lead to proportionally higher CO2 release per generated ATP.

150 5.4 The role of putrescine (This section has been extended and published during this thesis been reviewed. See published version in Appendix 2)

Consistent with numerous previous reports, the putrescine content increased dramatically under low K availability, in both sunflower and oil palm. However, although putrescine has been found to be accumulated under K deficiency more than 65 years ago, the specific role of this molecule under K deficiency is unknown. So far, too many putative roles have been suggested, and as summarized by Alcázar and co-workers (2006), “the precise role(s) of polyamine metabolism remain ill-defined”. This is why in this thesis, I provide a brief synthesis of the roles played by putrescine under K deficiency and propose the hypothesis that putrescine is accumulated mostly to avoid the cellular effects of Ca2+ increase (in particular on mitochondria), which is indeed more abundant in tissues under K deficiency (Chapter 3 Figure 2 and Chapter 4 Figure S2). This hypothesis is extremely well supported by the literature, including in mammals, where the role of polyamines in the mitigation of Ca2+ effects has been studied a lot in the past decade. In the attached appendix, the review paper (under revision) is provided.

151 5.5 Rationales for the selected K concentration for treatments and their appropriateness in agricultural K fertilization practice

K deficiency has a far-reaching influence on agriculture. In practice, it is reported that many agricultural lands throughout the world are K-deficient; so is the case of 75% of paddy soils of China and 65% of the wheat belt of Southern Australia (Zörb et al, 2014). The vast majority of K in soil is not readily available to support plant growth because it is not in the form of exchangeable K but remains insoluble in the mineral phase. It is typically the case of oil palm (Eleais guineensis), which has an even higher K elemental content compared to N. Major oil palm cultivated regions include Indonesia and Malaysia, as well as South American countries (Brazil, Peru, and Columbia) and Congo are often nutrient-poor, have a low potassium-to-calcium ratio, or have a low potassium exchange capability. It is believed that the concentration of soil potassium is generally less than 0.3 mM around the root zone (Mengel and Kirkby, 1982), so that most plants may easily face low-K stress during growth. In other words, K deficiency naturally occurs in the field when soil quality, climate or fertilization has a negative effect on soil K availability. I, thereby, selected 0.2 mM K as low K treatment.

Comparing to oil palm growing in medium and high K condition, plants grown in low K condition displayed decreased photosynthesis, stomatal conductance and biomass (Chapter 4 Figure 1 and 5), suggesting plants were stressed by low K. In addition, chlorophyll content in low K treated plants is lower than these in medium and high K reflected by SPAD reading (Figure 9). These data indicate that under 0.2 mM K treatment, oil palm is stressed by K deficiency.

The proteomic data of oil palm showed that both low and high K conditions were associated with common features, such as catalase, lipoxygenase, erythronolactone (ascorbate derivative), all related to oxidative stress metabolism. Consistently, antioxidants such as threonate (ascorbate derivative) increased at high K, and α- tocopherol (antioxidant) under both low and high K (Chapter 4, Fig. S6). These data indicate that 4 mM K, like low K (0.2 mM K) causes stresses to oil palm suggesting that 4 mM K applied here is too high for oil palm. In field, K fertilizer is usually only applied locally to the root and only applied at the beginning of the season. This may cause the oil palm plants suffering from high K stress at the beginning of season.

152

70 60 50 40 30

SPAD reading SPAD 20 10 0 0.2 1 4 K level (mM)

Figure 9. Leaf relative chlorophyll content of 11-month oil palm sapling determined by SPAD.

153 5.6 Limitations of this thesis

In this thesis, the consequences of K limitation and waterlogging (and their interaction) on metabolism were investigated. Important conclusions could be drawn about e.g. nitrogen assimilation and CO2 production, including the probable involvement of alternative pathways. However, there are important limitations:

• The source of respiratory carbon (e.g. reserve remobilisation and utilisation of recently fixed carbon) could not be differentiated in the studies of this thesis, not only in terms of metabolic origin (contribution of different pathways), but also at the plant scale. Oil palm accumulates reserves in the trunk (stem), which may be important to sustain catabolism. Under occasional (that is, non-chronic) stress, such as K deficiency or/and waterlogging, there could be some exchange between plant parts (from trunk reserves to leaves, for example), thereby changing the carbon source used for leaf respiration. In the field, it has been shown that under water stress, oil palm can indeed remobilizes non-soluble carbohydrates from the trunk (Legros et al., 2009).

• In this thesis, Rdark was only determined in leaves, not in heterotrophic organs (e.g.

stem, roots); therefore the total CO2 efflux, and thus the carbon balance, could not be derived precisely. Our estimate of carbon use efficiency in oil palm based on biomass and gross photosynthesis suggests that respiration by heterotrophic organs is an important part of the carbon budget, but more experimental data are needed. • GC-MS profiling the metabolites used here did not give direct access to absolute amounts (in moles) since it was used for semi-quantitative analysis (which is adapted to comparison between samples). A more quantitative view of metabolites would be required to compare with rates that have been measured (respiration, photosynthesis). • It is proposed here that accumulation of putrescine under low K availability is to avoid the side effects of Ca2+ on mitochondrial activity, rather than just compensate for the lack of positive charges due to K+ scarcity. However, experimental data is needed to support this hypothesis.

154 5.7 Future directions

To address the limitations mentioned above and also to better understand the metabolic patterns we have found in this work, several perspectives can be envisaged:

• To understand the contribution of root (and stem) in plant CO2 efflux and thus carbon use efficiency, a hydroponic system to grow oil palm with low K or high K could be carried out. • To differentiate between reserve remobilisation and utilisation of recently fixed carbon as respiratory carbon sources, and also to trace metabolic pathways, we could use 13C labelling; also, labelling with 15N-nitrate to follow fluxes of nitrogen under K deficiency and/or waterlogging would be useful. • Also, it would be interesting to have access to sap composition (both phloem and xylem) to know both the metabolic/elemental composition under low K and waterlogging, but also the isotope composition (δ15N) of exchanged nitrate. • In order to address the potential role of putrescine in avoiding side effects of Ca2+: 1) Testing whether Arabidopsis mutants that contains a reduced amount in putrescine (such as adc2-1 mutant whose endogenous putrescine is decreased to about 25% of that in wild-type (Zhu et al., 2016)) is more sensitive to low K treatment than wild-type. 2) Testing whether Arabidopsis mutants with reduced K uptake (such as athak5 akt1 double mutant whose growth is impaired at low K condition (Pyo et al., 2010)) accumulate putrescine or not. 3) Testing whether exogenously applying putrescine can enhance plants tolerance to low potassium. 4) Testing whether reducing Ca2+ in the culture medium (nutrient solution) can alleviate the accumulation of putrescine under low K conditions.

155 References

Alcazar, R., Marco, F., Cuevas, J.C., Patron, M., Ferrando, A., Carrasco, P., Tiburcio, A.F., and Altabella, T. (2006). Involvement of polyamines in plant response to abiotic stress. Biotechnology letters 28, 1867-1876. Armengaud, P., Sulpice, R., Miller, A.J., Stitt, M., Amtmann, A., and Gibon, Y. (2009). Multilevel analysis of primary metabolism provides new insights into the role of potassium nutrition for glycolysis and nitrogen assimilation in Arabidopsis roots. Plant physiology 150, 772-785. Azcón-Bieto, J., and Osmond, C.B. (1983). Relationship between Photosynthesis and Respiration. The Effect of Carbohydrate Status on the Rate of CO2 Production by Respiration in Darkened and Illuminated Wheat Leaves 71, 574-581. Barbour, M.M. (2007) Stable oxygen isotope composition of plant tissue: A review. Functional Plant Biology, 34, 83-94. Besford, R.T., and Maw, G.A. (1976). Effect of Potassium Nutrition on Some Enzymes of the Tomato Plant. Annals of Botany 40, 461-471. Freeman, G.G. (1967). Studies on potassium nutrition of plants. II. Some effects of potassium deficiency on the organic acids of leaves. Journal of the Science of Food and Agriculture 18, 569-576. Gupta, K.J., Shah, J.K., Brotman, Y., Jahnke, K., Willmitzer, L., Kaiser, W.M., Bauwe, H., and Igamberdiev, A.U. (2012). Inhibition of aconitase by nitric oxide leads to induction of the alternative oxidase and to a shift of metabolism towards biosynthesis of amino acids. Journal of Experimental Botany 63, 1773- 1784. Hussain, S.S., Ali, M., Ahmad, M., and Siddique, K.H.M. (2011). Polyamines: Natural and engineered abiotic and biotic stress tolerance in plants. Biotechnology Advances 29, 300-311. Legros, S., Mialet-Serra, I., Clement-Vidal, A., Caliman, J. P., Siregar, F. A., Fabre, D., and Dingkuhn, M. (2009). Role of transitory carbon reserves during adjustment to climate variability and source-sink imbalances in oil palm (Elaeis guineensis). Tree Physiol. 29, 1199-1211. Jones, L.H. (1961). Some effects of potassium deficiency on the metabolism of the tomato plant. Canadian Journal of Botany 39, 593-606. Jones, L.H. (1966). Carbon-14 studies of intermediary metabolism in potassium- deficient tomato plants. Canadian Journal of Botany 44, 297-307. Jwakyung, S., Yeonkyu, S., Yejin, L., Seongsoo, K., Sangkeun, H., B., K.H., and Taek-Keun, O. (2015). Compositional changes of selected amino acids, organic acids, and soluble sugars in the xylem sap of N, P, or K-deficient tomato plants. Journal of Plant Nutrition and Soil Science 178, 792-797. Mengel, K. and Kirkby, E. A. (1982) in Principles of Plant Nutrition 3rd edn, Vol. 1 (eds Kosegarten, H. et al.) Ch. 10, 655 International Potash Institute, Worblaufen- Bern, Switzerland. O'Leary, B.M., Asao, S., Millar, A.H., and Atkin, Owen K. (2019). Core principles which explain variation in respiration across biological scales. New Phytologist 222, 670-686. Okamoto, S. (1966). Effect of mineral nutrition on metabolic change induced in crop plant roots (III). Soil Science and Plant Nutrition 12, 13-17. Okamoto, S. (1967). Effects of potassium nutrition on the glycolysis and the krebs cycle in taro plants. Soil Science and Plant Nutrition 13, 143-150.

156 Okamoto, S. (1968). The respiration in the roots of broad bean and barley under a moderate potassium deficiency. Soil Science and Plant Nutrition 14, 175-182. Pyo, Y.J., Gierth, M., Schroeder, J.I., and Cho, M.H. (2010). High-affinity K+ transport in Arabidopsis: AtHAK5 and AKT1 are vital for seedling establishment and postgermination growth under low-potassium conditions. Plant Physiol 153, 863-875. Tcherkez, G. (2010). Natural 15N/14N isotope composition in C3 leaves: are enzymatic isotope effects informative for predicting the 15N-abundance in key metabolites? Functional Plant Biology 38, 1-12. Tcherkez, G., and Hodges, M. (2007). How stable isotopes may help to elucidate primary nitrogen metabolism and its interaction with (photo)respiration in C3 leaves. Journal of Experimental Botany 59, 1685-1693. Terry, N., and Ulrich, A. (1973). Effects of Potassium Deficiency on the Photosynthesis and Respiration of Leaves of Sugar Beet. Plant Physiology 51, 783-786. Tischner, R. (2000). Nitrate uptake and reduction in higher and lower plants. Plant, Cell & Environment 23, 1005-1024. Verniquet, F., Gaillard, J., Neuburger, M., and Douce, R. (1991). Rapid inactivation of plant aconitase by hydrogen peroxide. Biochemical Journal 276, 643-648. Yoneyama, T., Ito, O., and Engelaar, W.M.H.G. (2003). Uptake, metabolism and distribution of nitrogen in crop plants traced by enriched and natural 15N: Progress over the last 30 years. Phytochemistry Reviews 2, 121-132. Zhu, X.F., Wang, B., Song, W.F., Zheng, S.J., and Shen, R.F. (2016). Putrescine Alleviates Iron Deficiency via NO-Dependent Reutilization of Root Cell-Wall Fe in Arabidopsis. Plant Physiology 170, 558-567. Zörb, C., Senbayram, M. and Peiter, E. (2014). Potassium in agriculture--status and perspectives. J Plant Physiol. 171,656-669.

157

Appendix 1

δ15N values in plants is determined by both nitrate assimilation and circulation

158 Research d15N values in plants are determined by both nitrate assimilation and circulation

Jing Cui1, Emmanuelle Lamade2, Francßois Fourel3 and Guillaume Tcherkez1 1Research School of Biology, ANU Joint College of Science, Australian National University, Canberra, ACT 2601, Australia; 2UPR34 Performance des systemes de culture des plantes perennes, Departement PERSYST, Centre de Cooperation Internationale en Recherche Agronomique pour le Developpement (CIRAD), Montpellier 34398, France; 3UMR CNRS 5023 LEHNA, Universite Claude Bernard Lyon 1, 3 rue Rapha€el Dubois, Villeurbanne 69622, France

Summary Author for correspondence: Nitrogen (N) assimilation is associated with 14N/15N fractionation such that plant tissues are Guillaume Tcherkez generally 15N-depleted compared to source nitrate. In addition to nitrate concentration, the Tel: +61 2 6125 0381 d15N value in plants is also influenced by isotopic heterogeneity amongst organs and metabo- Email: [email protected] lites. However, our current understanding of d15N values in nitrate is limited by the relatively Received: 27 November 2019 small number of compound-specific data. Accepted: 31 January 2020 We extensively measured d15N in nitrate at different time points, in sunflower and oil palm grown at fixed nitrate concentration, with nitrate circulation being varied using potassium (K) New Phytologist (2020) conditions and waterlogging. doi: 10.1111/nph.16480 There were strong interorgan d15N differences for contrasting situations between the two species, and a high 15N-enrichment in root nitrate. Modelling shows that this 15N-enrichment Key words: fractionation, modelling, nitrate, can be explained by nitrate circulation and compartmentalisation whereby despite a numeri- nitrogen isotope, oil palm, potassium, cally small flux value, the backflow of nitrate to roots via the phloem can lead to a c. 30& dif- sunflower, waterlogging. ference between leaves and roots. Accordingly, waterlogging and low K conditions, which down-regulate sap circulation, cause a decrease in the leaf-to-root isotopic difference. Our study thus suggests that plant d15N can be used as a natural tracer of N fluxes between organs and highlights the potential importance of d15N of circulating phloem nitrate.

(Tcherkez, 2011). As a result, specific analysis of plant nitroge- Introduction nous compounds can provide information on metabolic fluxes Plant nitrogen (N) acquisition from nitrate fractionates between and nitrate circulation between plant organs. For example, it has isotopes so that plant total organic matter is in general naturally been shown that isotope effects and metabolic fluxes in most 15 N-depleted compared to source N. When source nitrate is not important reactions (including photorespiratory NH3 production limiting, plant matter can be depleted by several per mil (&) and recycling) can explain the d15N values in individual amino (H€ogberg, 1997; Evans, 2001). Groundwater nitrate isotope acids of rapeseed leaves (Gauthier et al., 2013). composition (natural 15N abundance, d15N) is generally between It is believed that after having been absorbed, nonassimilated 5 and +5& (Kendall & Aravena, 2000), and therefore the nitrate is 15N-enriched due to nitrate reduction in roots, and d15N value in plant matter is usually within 13 and +5&, therefore nitrate molecules transferred to shoots via the xylem are depending on species, plant organs and environmental conditions naturally 15N-enriched (Evans, 2001; Tcherkez & Hodges, (for an example, see Handley et al., 1997). The fractionation 2007). This should in principle lead to a 15N-difference between comes from the 14v/15v isotope effect associated with nitrate dif- plant organs, with leaves being 15N-enriched compared to roots. fusion and transport via channels and transporters (presumably Of course, partitioning of nitrate reduction between roots and c. 1.003) and nitrate reductase (c. 1.016) (Ledgard et al., 1985; shoots depends on species, environmental conditions, develop- Evans, 2001). The net fractionation associated with nitrate utili- mental stage and nitrate concentration and therefore the d15N sation nevertheless depends on nitrate content, with low nitrate value of plant matter can vary accordingly. In addition, the day/ concentration in the soil solution leading to low isotope fraction- night cycle of transpiration and nitrate reduction activity in leaves ation (Kolb & Evans, 2003; Stock & Evans, 2006). In addition, leads to significant changes in d15N values not only in leaves but d15N values vary between plant organs, metabolites (Werner & also in xylem and phloem sap. In fact, in Ricinus communis Schmidt, 2002) and N-atom positions (Sacks & Brenna, 2005) (where phloem sap can be relatively easily collected), it has been due to nitrate redistribution within the plant, and isotope effects shown that there is a diel cycle of d15N values, with a 15N-enrich- associated with enzymatic reactions in anabolism such as glu- ment of several per mil in phloem at midday compared to the tamine synthetase, glycine decarboxylase or transaminases middle of the night, suggesting a progressive enrichment in

Ó 2020 The Authors New Phytologist (2020) 1 New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com

159 New 2 Research Phytologist

exported N-assimilates as leaves consume nitrate in the light by far the major cation in phloem sap and low K conditions are (Peuke et al., 2013). detrimental to phloem-mediated export by leaves (Hartt, 1969; Taken as a whole, N isotope composition of plant compounds Mengel & Haeder, 1977; Doman & Geiger, 1979; Cakmak is dictated by key mechanisms of N metabolism and allocation et al., 1994; Epron et al., 2015). Low K not only affects phloem and, thus, d15N values could be exploited to obtain information sap circulation but also leads to an increase in amino acid export on metabolic fluxes. However, this requires the use of models from roots to shoots. Analyses of xylem composition have indeed that account for nitrate redistribution, N partitioning and shown an increase in glutamine and c-aminobutyrate content, branching points at isotopically sensitive steps. This problem has suggesting an increase in the root-to-shoot ratio of nitrate reduc- been addressed previously with kinetic models (Yoneyama et al., tion (Sung et al., 2015). 1987; Robinson et al., 1998; Comstock, 2001; Johnson & Berry, The present study represents a compilation of more than 2500 2013). Although models can explain the d15N values with fitted isotopic analyses including 600 analyses of d15N values in nitrate, flux parameters, they have three potential problems: first, the using samples collected over one year (the time required to grow number of parameters required, up to nine in Robinson et al. oil palm saplings in the glasshouse). Our objective was to address (1998); second, they are implemented in the metabolic steady- the following questions: Is leaf nitrate systematically 15N-en- state (except for pools formed by accumulation); and third, riched compared to absorbed nitrate and root nitrate? Do K limi- experimental data required for model validation and testing (in tation and waterlogging effectively affect plant nitrate d15N particular d15N values in nitrate) are very limited. As a result, the values? Can we predict d15N values using a simple, nonsteady- utilisation of d15N values in species relying on soil inorganic N in state model and does modelling provide information on nitrate plant physiology remains limited. By contrast, d15N utilisation circulation? Our data show that plant nitrate is considerably has been implemented in other species for a long time, typically enriched compared to source nitrate from the nutrient solution 15 to estimate the proportion of N derived from atmospheric N2 in and that, surprisingly, root nitrate was more N-enriched than legumes (Mariotti et al., 1983). leaf nitrate. Modelling suggests that this surprising result comes Here, we tackle this issue using, at the same time: experimental not only from root nitrate reduction preferentially consuming determination of nitrate isotope composition and content in sev- 14N-nitrate and the ability of roots to accumulate nitrate, but also eral organs at different developmental stages; measurements of from the backflow of 15N-enriched nitrate (although at a very biomass and elemental N content; and a deliberately simple small rate) via the phloem. model of nitrate distribution and utilisation. To do so, we used two species (sunflower and oil palm) followed over an extended Materials and Methods growth period, cultivated under low or high potassium (K) availability, with or without waterlogging, using a nitrate-based nutri- Plant material and sampling ent solution (we thus avoided complications associated with other N sources (ammonium or urea), such as isotope fractiona- Plant cultivation conditions were as in Cui et al. (2019a,b). tion in ammonium acid/base protonation or in urea cleavage). Briefly, sunflower (Helianthus annuus L., var. XRQ) seed pro- We chose these two species because they usually grow in regions vided by the INRA Toulouse (France) and germinated oil palm where K deficiency and waterlogging can occur simultaneously. seeds (Elaeis guineensis Jacq., Dura 9 Pisifera) obtained from In addition, their metabolic response to both K limitation and Siam Elite Palm Co. Ltd (Krabi, Thailand) were sown directly in waterlogging (including N metabolism) has been described sand (washed with distilled water) in the glasshouse, using 7-l recently (Cui et al., 2019a, b), and they are large enough to col- pots. Growth conditions were: 12 h : 12 h, 25°C:18°C pho- lect sufficient sample weight for nitrate purification. Manipulat- toperiod and air temperature and 70/60% relative humidity, ing potassium availability and waterlogging allowed us to change day : night for sunflower; and natural photoperiod (from 13.5 N assimilation and nitrate partitioning, while the availability and h : 10.5 h at emergence (February) to 14.5 h : 9.5 h at the end of d15N of source nitrate were kept constant. Waterlogging affects the experiment (January), with a minimum of 9.5 h : 14.5 h in + ° ° ion (such as K and NO3 ) uptake efficiency in roots (Demid- June), 30 C:24 C air temperature and 70% : 60% relative chik, 2014). In fact, oxygen shortage in roots leads to a deficiency humidity day : night for oil palm. Transpiration was measured in ATP generation by respiration as well as a change in trans- using a portable open system Li-Cor 6400 XT device. Sampling porter activity (such as an inhibition by reactive oxygen species) time points are detailed in Supporting Information Fig. S1. + such that both nitrate and K efflux from root cells increase Plants were subdivided into stems (+ apical bud), roots and leaves (Sharma et al., 2010; Barrett-Lennard & Shabala, 2013; Shabala which were frozen in liquid N2 and freeze-dried, weighed, and & Pottosin, 2014; Zeng et al., 2014). Under waterlogging, it is ground for further utilisation. believed that the rate of nitrate reduction in roots does not decrease at least in the short- to mid-term so that there is a gen- Nutrient conditions and waterlogging eral decrease in nitrate concentration in tissues, as typically seen in sunflower (Aguera€ et al., 1990). The simultaneous decrease in The nutrient solution composition is given in Table S1. The + nitrate and water circulation from roots to shoots and the amount of K was varied by changing the amount of KCl. Two + increase in root K efflux tend to induce a potassium deficiency K availability conditions were used here: ‘low K’ (0.2 mM) and + and thus a decline in sap circulation (Phukan et al., 2016). K is ‘high K’ (4 mM). The amount of nitrate was 12 mM in the

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust

160 New Phytologist Research 3 nutrient solution and was kept constant throughout the experi- deliberately kept as simple as possible to minimise the number of ments. Under nonwaterlogging (control) conditions, 300 ml variables, avoid lengthy parameterisation and therefore allow (sunflower) or 500 ml (oil palm) nutrient solution was provided facile numerical resolution. In practice, the plant is considered to to each pot every day. Excess liquid was allowed to drain freely contain one mesophyll nitrate pool (‘leaves’) which is fed by the from the pot, so that on average 250 ml was retained in pots each xylem, and can transfer some nitrate to the phloem for nitrate day. Waterlogging was performed by filling the pot with the redistribution to (stem and) roots. Here, ‘xylem’ and ‘phloem’ nutrient solution. Some nutrient solution was provided every day are two terms used in a broad sense, that is, including tissues that to compensate for evaporation and transpiration (200 ml d 1 inherit nitrate from vasculature (xylem vessels) and sieve tubes pot). Thus, waterlogging led to a progressive hypoxia rather than (parenchyma, sink organs), and therefore to avoid confusion with abrupt anoxia. We used a relatively high nitrate concentration pure sap-conducting xylem and phloem cells (or saps themselves), + + (12 mM) under our conditions to avoid changes in the d15Nof they are thereafter referred to as xylem and phloem . The model source nitrate: in fact, renewal of nutrient solution (3.6 and includes the possibility to have some transfer of nitrate from 6 mmol nitrate per day per pot) compensated for the consump- xylem to phloem (transfer rate B). Roots are assumed to comprise tion by plants, so that source nitrate was kept isotopically con- two nitrate pools (R1 and R2, respectively): one of small size kept stant (d15N variation during 24 h before the addition of nutrient in the steady-state and used as a transfer compartment from soil solution of the next day would not have exceeded 0.05&). The to xylem sap, and another one that can be of varying size and d15N value of source nitrate (nutrient solution) was +3.5&. used as the source of nitrate for root N assimilation. These two root pools can exchange nitrate (via back and forth fluxes, with rates w and A + w). Isotope analysis and %N At this stage, at least three aspects deserve comments: first, we In total, 100 mg of powdered freeze-dried material was first recognise that the transfer of nitrate from xylem to phloem extracted with hexane to obtain the total lipid fraction (supernatant (transfer rate B) has been found experimentally to be very small that contained chlorophylls). The pellet was extracted with MilliQ (see references in the Discussion section). As such, having a small water. After centrifugation, the supernatant was collected and value of B will be a way to check the relevance of our computa- heated at 100°C for 5 min. The heat-precipitated protein pellet tions. Second, the model is not in the steady-state. That is, both was collected and the deproteinated supernatant was subdivided nitrate d15N and pool sizes varied with time (during growth) and into two halves. One half was freeze-dried and used for total sol- thus incorporating differential equations was necessary. Third, a uble N isotope analysis. The other half was used for nitrate extrac- backflow of nitrate from leaves to sink organs and ultimately to tion as barium nitrate with barium iodide, after Huber et al. roots is considered. The reasons for including a phloem-driven (2011). Plant samples were weighed in tin capsules. d15N analysis flux of nitrate is further explained in the Discussion. As will was carried out using a Pyrocube Elemental Analyser (EA; Elemen- become apparent below, this flux (although quantitatively small) tar France, Lyon France) connected on line (via an open split) in is an important determinant of nitrate d15N because it can carry continuous flow mode to an Isoprime 100 isotope ratio mass spec- 15N-enriched nitrate. trometer (IRMS; Elementar UK, Cheadle, UK), after Fourel et al. The equations used here are based on the general expression of (2014). For combustion in the EA, the quartz tube was filled with derivatives of isotope ratios with respect to time. The derivative tungsten oxide and the reduction tube was filled with copper. of the 15N/14N isotope ratio R in a given nitrate pool follows the Combustion gases were separated through the ‘purge and trap’ relationship below that comes from 15N mass balance: module from the EA. The amount of N was measured via a ther- d15 15 mal conductivity detector. N precision for replicate analyses d Q ¼ dQR ¼ dR þ dQ ¼ Fin Fout & Q R Rin R Eqn 1 was 0.1 . Data were calibrated against international reference dt dt dt dt ain aout material IAEA-600, IAEA-CH6 and IAEA-N2. In addition, all sample batches included standards (glycine, +0.66&;cysteine, where 15Q is the total 15N content and Q is the total 14N content, +7.78&; previously calibrated against IAEA standards glutamic which is very close to the total pool size (because 15N natural acid USGS-40 and caffeine IAEA-600) every 10 samples. When abundance is very small). Fin and Fout are influx and efflux, and < μ 14 15 14 15 the sample mass was occasionally low (typically 0.08 mol N per Rin the isotope ratio of influx. N/ N isotope effects ( v/ v) d15 a sample), the N value was corrected using the response curve of are denoted as . Note that Fin may be different from Fout if the the d15N value to sample mass using standard glycine and cysteine. nitrate pool of interest is not in the steady-state because by defini- d15 d = – N values are expressed in notation, that is deviation in parts tion dQ/dt Fin Fout, rearranging Eqn 1 gives Eqn 2: & d15 = per thousand ( ) relative to atmospheric N2 (air): N (Rsam- 15 14 – Rin 1 ple/Rstandard 1) where R denotes the N: N ratio. Fin a R FoutR a 1 dR in out ¼ Eqn 2 dt Q Modelling and assumptions The model considered here assumes that the plant can be subdi- The set of equations is given in Notes S1. When considered together as a matrix, the differential system can be simply written vided into pools that exchange nitrate via xylem and phloem = (Fig. 1; symbols are defined in Table 1). The model is as dY/dt MY where Y is the vector of isotope ratios and M the

Ó 2020 The Authors New Phytologist (2020) New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com

161 New 4 Research Phytologist

Assimilation Circulation of r N assimilates l al Leaves

T – rx – ax – B bn

rp Assimilation

B Stem + + a Xylem Phloem x ap Assimilation rx

T bn – rp – ap + B

i A+w w Roots rr Circulation of e Assimilation N assimilates

Fig. 1 Model used to compute d15N values in nitrate. For simplicity, plants are assumed to comprise three compartments: leaf mesophyll (abbreviated ‘leaves’), stem and roots. Pools that can vary in size (nonsteady-state) are indicated with dashed grey borders. The only pool that is of ﬁxed size is the

nitrate transfer pool (R1) in the root. That pool is fed by external nitrate (influx i) and nitrate can also escape (efflux e). The stem is assumed to be + composed of two pools: one associated with xylem tissues (including xylem parenchyma; referred to as xylem ) and one associated with phloem, phloem + parenchyma and sink (developing) organs such as apical bud (referred to as phloem ). Rates of nitrate reduction (assimilation) are denoted as r while accumulation (build-up) rates are denoted as a. See also Table 1 for symbol definitions. Assimilates can circulate across the plant (amino acid or amide exchange between organs) and this is symbolised with blue arrows. This has no impact on nitrate in this model because it does not appear in equations. Note that this model is simple enough to rely on three main variables, once nitrate reduction capacities (r values) and accumulation rates (a values) are

known: the rate of exchange between root pools (A), nitrate backﬂow (bn) and xylem-to-phloem nitrate transfer (B).

matrix of (time-dependent) coefficients in equations. This equa- is the value broadly accepted for the fractionation by nitrate tion cannot be solved directly because coefficients of M are not reductase (for a review, see Tcherkez & Farquhar, 2006). The constant (pool sizes) and furthermore M could be nondiagonalis- actual value of the fractionation was also estimated here using iso- able (decomposed into singular values). Therefore, we solved topic mass-balance in leaves. Leaves were chosen because they are numerically by decomposing time windows (referred to as ‘steps’ an important site of nitrate reduction and redistribution of N in Fig. S1) into n time increments. For any incremental time assimilates (at the mature stage, they do not import reduced N to index k in [0;n], the isotope ratio at k + 1 is given by a large extent in both species) and thus the d15N of their organic = + c∙ c Rk+1 Rk (dR/dt)k where is the time unit conversion factor compounds is minimally impacted by the contribution of other between unit of fluxes F (μmol per plant d 1) and time incre- organs. Leaf samples were fractionated into their four main N- ments. n was chosen to be sufficiently large to avoid divergence containing components (proteins, lipids, nitrate and total soluble (artefactual instability due to too large dR/dt values). In practice, fraction), the d15N of which was determined. Therefore, we time increments represented 3.6 and 3.36 min for sunflower and have: oil palm, respectively. Further details on parameterisation are d D provided in Notes S2 (as well as Figs S2, S3 and Table S2). nitrates r dNassimilates ¼ 1 þ Dr ¼ dlipidsxlipids þ dproteinsxproteins þ dTSFxTSF Eqn 3 Verification of the isotope effect ar D The fractionation associated with nitrate reduction (assimilation) where r is the fractionation associated with reduction = a is a critical component of the model and it is fixed at 16&, which ( r 1) and x are mole fractions of N represented by the

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust

162 New Phytologist Research 5

¼ þ þ þ Table 1 Symbols used in the main text and equations. rtot r1 rr rx rp X %N 3 %N 3 m 10 n 0 m 10 n Symbol Definition Units ¼ 14 t 14 t leaf;stem;root t 0 t Fluxes Eqn 4 A Net influx of nitrate from root pool R1 to R2 μmol N per plant d 1 Similarly, the rate of change in nitrate pool size between time t al Variation in nitrate content in leaves μmol N per 0 plant d 1 and t was calculated, for each organ, as: + ap Variation in nitrate content in phloem μmol N per 1 plant d ðm nÞ 0 ðm nÞ a ¼ t t Eqn 5 ar Variation in nitrate content in roots μmol N per 0 t t plant d 1 + ax Variation in nitrate content in xylem μmol N per 1 This allowed us to compute the total net nitrate demand plant d B Transfer of nitrate from xylem to phloem μmol N per (i e): plant d 1 X X + bn Backflow of nitrate from leaves to phloem μmol N per ¼ þ ¼ þ i e r a rtot atot Eqn 6 plant d 1

e Efﬂux of nitrate from root pool R1 back to soil μmol N per plant d 1 Therefore, gross nitrate inﬂux was given by:

i Inﬂux of nitrate from soil to root pool R1 μmol N per plant d 1 r þ a i ¼ tot tot Eqn 7 Q Pool size μmol per 1 e=i plant

rl Flux of nitrate reduction in leaves μmol N per plant d 1 using the value of e/i chosen for parameterisation (Table S2). + rp Flux of nitrate reduction in phloem μmol N per Note that the value of e/i had a very limited impact on calcula- plant d 1 tions of d15N values simply because the isotopic contribution of μ rr Flux of nitrate reduction in roots mol N per efflux (isotope effect of 1.003) was very small compared to the 1 plant d 15 + range of N-enrichment found here (up to 55&). rx Flux of nitrate reduction in xylem μmol N per plant d 1 T Flux of nitrate export from roots μmol N per plant d 1 Results w Flux of nitrate exchange between R1 and R2 μmol N per 1 plant d d15 Isotopes N in leaf organic matter a 14 15 d v/ v isotope effect associated with nitrate Dimensionless The N isotope composition of leaf raw material was within a transport through membranes (influx and & efflux). Fixed at 1.003. rather narrow range, roughly between 8 and 0 , that is, 14 15 15 ar v/ v isotope effect associated with nitrate Dimensionless N-depleted compared to source nitrate in the nutrient solu- reduction. Fixed at 1.016. tion (+3.5&) (Fig. S4). The isotope composition was then 15 14 15 d N N/ N isotope composition with respect to air & expressed relative to source nitrate (D15N = d15N 15 14 source R Isotope ratio ( N/ N) Dimensionless – d15 nitrate Nleaf) as the apparent N isotope fractionation, and plotted against %N expressed as the mean centred value (Fig. 2; see Fig. S5 for %N source data). Nearly all D15N values were positive (only one value obtained at high K in sun- + + = compounds of interest (xlipids xproteins xTSF 1). Eqn 3 flower was negative) showing that, overall, N acquisition in leaf 15 neglects other components that have a small contribution to the tissues generally fractionated against N. In both species, there 15 total N budget (volatilisable ammonia etc.). In addition, ‘TSF’ was a significant effect of K availability: leaves were more N- refers to the nitrate-corrected total soluble fraction as the raw depleted (and less N-rich) at low K, suggesting that nitrate used 15 total soluble fraction also contains nitrate. Eqn 3 can thus be used for N assimilation in leaves was N-depleted or more abun- D to compute r. dant (thereby allowing larger fractionation) than at high K. Waterlogging caused a 15N-depletion, abolished the isotopic difference between high and low K, and led to a lower %N in Determination of total N assimilation and nitrate pool size leaf tissue. This suggests that waterlogging was detrimental to change nitrate assimilation regardless of K availability, and changed μ 1 15 Total N assimilation (denoted as rtot,in mol N per plant d ) nitrate circulation so that leaf nitrate was N-depleted. Such a 15 was calculated as a time average between two time points t and t0 N-depletion was also visible in raw material of sunflower stem (two sampling dates) using dry weight (m, in mg), the %N in and roots, and oil palm roots (Fig. S4). total plant matter and a correction for the quantity of nitrate (n There were differences between leaf fractions (proteins, lipids, in μmol mg 1). That is: total soluble fraction), the nitrate-corrected soluble fraction being

Ó 2020 The Authors New Phytologist (2020) New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com

163 New 6 Research Phytologist

generally 15N-depleted by a few per mil (Fig. S6). When calcu- considerably. Waterlogging generally caused a decrease in d15N lated by mass-balance, the isotope fractionation associated with values, and this was particularly visible at high K in sunflower. nitrate assimilation in leaves was 16.8& on average (Fig. 2, where d15N changes were accompanied by changes in nitrate content. fitting a Gaussian distribution gave a value of 17.8&). This value In fact, the nitrate content (μmol g 1) showed a progressive is very close to the measured isotope fractionation by nitrate decline (particularly pronounced in oil palm roots, Fig. 3l), reductase (16&) (Ledgard et al., 1985; Tcherkez & Farquhar, except in sunflower roots where it remained fairly constant 2006). While tissue fractions fell within a relatively narrow range (Fig. 3f). It is worth noting that roots represented plant organs in sunflower leaves, there were important differences in other where nitrate was most concentrated (except at low K in sun- organs of both species, with lipids being generally 15N-depleted flower, Fig. 3e). compared to proteins, by up to 30& (Fig. S6). In sunflower, the nitrate-corrected soluble fraction appeared to be 15N-depleted in d15N modelling roots, suggesting that it contained organic compounds the synthesis of which was associated with a large isotope fractionation. Biomass and nitrate content data were incorporated into a (nonsteady-state) model to assess whether observed d15N value can be reproduced satisfactorily, using observed N incorpora- Nitrate content and d15N value tion rates computed from biomass N increment and changes Nitrate was extracted and purified from all tissues and its d15N in nitrate pools in each organs (see the Materials and Methods value was measured (Fig. 3). As expected, nitrate was found to section for details). Taken as a whole, the performance of be 15N-enriched (up to 55& in sunflower stems) as a conse- modelling was good, with a R2 between predicted and quence of the isotope fractionation by N assimilation (so that observed values of 0.95 (Fig. 4a). Modelling was not very sen- consumed nitrate molecules were 15N-depleted and nitrate left sitive to parameterisation and led to an average error in d15N behind was 15N-enriched). In sunflower, there was a 15N-enrich- of < 1& (Fig. S2). In addition, the predicted rate of nitrate ment gradient from leaves to roots (Fig. 3a–c) regardless of K transport via the xylem, T, was used to estimate leaf transpira- and waterlogging conditions. d15N values were not highly tion using leaf total surface area, and a fixed xylem concentra- affected by K/waterlogging conditions except at high K where tion of 3 mM (although we recognise that nitrate transport stem and root nitrate was very 15N-enriched. In general, d15N may be disconnected from transpiration under certain circum- values in oil palm were lower, except at high K where it reached stances (White, 2012)). The agreement between observed and nearly 30& 7 wk after the onset of our experiment (Fig. 3g–i). predicted values was good under nonwaterlogged conditions Taken as a whole, with the exception of the last sampling at high (Fig. 4b). Unsurprisingly, there was no good agreement K in oil palm, root nitrate was always 15N-enriched compared to between observed and predicted values of transpiration under other organs. waterlogging conditions (Fig. 4b, grey symbols), demonstrating The d15N value in nitrate changed with time with a progres- the disconnection between water flow and nitrate circulation sive 15N-depletion, except at high K where it increased when roots were waterlogged.

(a) (b) (c) 17.8 7.3 12.0 16 N N 16.8 15 15 ∆ 6.3 ∆ 14 10.0

5.3 12 8.0 4.3 10

6.0 8 3.3

Frequency (%) 6 2.3 4.0

4 1.3 2.0 2 0.3 D D 0.0 0 –3 –2 –1 0 1 2 3 –2.5 –1.5 –0.5 0.5 1.5 2.5 –0.7 0 5 10 15 20 25 30 35 Fractionation (‰) LK HK LK+WL HK+WL MK LK+HK resupply LK HK LK+WL HK+WL

Fig. 2 D15N values in leaf raw material in sunflower (a) and oil palm (b), with K and waterlogging conditions indicated with colours. In sunflower, additional measurements with medium K and low K + K resupply are also shown in green (squares and triangles, respectively). The net fractionation D15N (calculated using source nitrate as the source material, d15N =+3.5&) is plotted against mean-centred nitrogen elemental content (%N), denoted as D. (c) Distribution of the isotope fractionation associated with nitrate assimilation, computed from purified protein, lipids and total soluble fraction from leaves. The average is shown with a blue arrow (16.8&) and the Gaussian distribution with its average (17.8&) are shown in red. d15N values in leaf components (lipids, proteins, soluble fraction) are shown in Supporting Information Fig. S5. Statistical significance (P < 0.05) is shown with asterisks: waterlogging effect at high K (*), waterlogging effect at both high and low K (**), and K effect under nonwaterlogged conditions (***).

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust

164 New Phytologist Research 7

SUNFLOWER OIL PALM (a) Leaves (b) Stem (c) Roots K (g) Leaves (h) Stem (i) Roots W K 60 60 60 I 40 W 40 40 K I High K K K I 35 35 W K 35 50 Low K 50 50 K W W I I K High K + WL 30 30 I 30 )‰(etartinniN Low K + WL K W 40 40 K 40 W K K W 25 25 25 I I I W K 30 30 30 20 W 20 20 K W W 15 W 15 I 15 20 20 20 N in nitrates (‰) 15 δ 15 K δ 10 10 I 10 10 10 10 5 5 5

0 0 0 0 0 0 0 5 10 15 051015051015 024680246802468 Time (d) Time (d) Time (d) Time (w) Time (w) Time (w) (d) (e) (f) (j) (k) (l) 180 180 180 30 30 30 K )WD 160 160 W 160 I K 25 25 25 140 140 W K 140 DW) –1 I W –1 I K o 120 K 120 120 20 20 20 W mg W μ I ( 100 100 100

tnetnoceta K W W I 15 K 15 15 80 80 80 W I W K K 60 60 60 10 10 W K 10 I W rti I 40 I 40 40 K K I Nl 5 5 5

Nitrates content (μmol g W 20 20 20 W 0 0 0 0 0 0 0 5 10 15 0 5 10 15 051015 024680246802468 Time (d) Time (d) Time (d) Time (w) Time (w) Time (w) Fig. 3 Nitrate content and isotope composition (d15N) in leaves, stems and roots of sunﬂower (a–f) and oil palm (g–l). Same colour code as in Fig. 2. The 15 – 1 d N value is shown in per mil (&) on the top row (a–c; g–i). The nitrate content (μmol NO3 g DW) is shown on the bottom row (d–f; j–l). Values are mean SD (n = 5). Statistical signiﬁcance (P < 0.01) is shown with upper-case letters (K, potassium effect; W, waterlogging effect; I, potassium 9 waterlogging effect) at each time point.

Nevertheless, modelled d15N values were not close to observed (1999)) suggest a 15N-enrichment. Such an enrichment can be values (up to 10& difference) at high K in sunflower (Fig. 4a, large and, as a result, leaf nitrate content affects the d15N value in red frames). This shows the limitation of the model used here, leaf total soluble fraction (Gauthier et al., 2013). An important which was based on a minimal number of adjustable parameters. finding of the present study is the 15N-enrichment, sometimes Possible explanations are further detailed in the Discussion sec- considerable, in root nitrate compared to other tissues, including tion. Still, as expected, the model generated low values for all leaves. This was visible in both sunflower and oil palm. This three exchange rates: nitrate transfer from xylem to phloem (B), result is surprising because of the relatively higher d15N value in nitrate backflow from leaves to roots (bn) and exchange between leaf organic matter (at least compared to roots, Fig. S3). This rapid and slow nitrate pools in roots (w) (Fig. 4c). Also, waterlog- lower-than-expected isotope composition in leaf nitrate (and ging caused a decrease in modelled nitrate circulation (both T higher-than-expected in roots) could come from several potential 15 and bn) in sunflower. In oil palm, the effect of waterlogging on T mechanisms: (1) a Peclet effect whereby mesophyll N-nitrate or bn was more variable. In both species, modelling suggested that would behave as leaf water during transpiration, that is, would waterlogging caused a transient increased remobilisation of 15N- back-diffuse to xylem vessels against the convection flow (Bar- enriched nitrate from the root storage pool (w) to feed nitrate cir- bour et al., 2000, 2004); (2) the occurrence of two nitrate pools + culation. As a result, the predicted d15N composition in xylem in roots, including a nitrate storage pool that can accumulate increased under waterlogging, and this effect was mostly visible 15N-enriched nitrate; (3) a flux of nitrate through the phloem in sunflower where stem nitrate content changed (Fig. S6). from leaves back to roots; and (4) a metabolic process that con- It is worth noting that the modelled d15N value of nitrate in sumes 15N-depleted nitrate in roots, thereby enriching nitrate left + + + xylem and phloem differed substantially, with phloem nitrate behind. Amongst these, assumption (1) is unlikely because the + – being much more 15N-enriched than xylem (for an example, see diffusion coefficient of nitrate in water, 1.3 9 10 5 cm2 s 1 (Yeh Fig. S7). This phenomenon is not surprising because nitrate & Wills, 1970), is not large enough to compensate for the high reduction in leaves fractionates against 15N and therefore nitrate water flux via the xylem. In other words, the Peclet number, exported by leaves is 15N-enriched. As a result, nitrate carried by which quantifies the diffusion-to-convection ratio would be too + phloem back to roots is 15N-enriched, thereby explaining the small to explain a significant change in the d15N value of leaf relatively high d15N values observed in root nitrate. nitrate. Assumptions (2), (3) and (4) are plausible. In fact, roots were a site of nitrate accumulation, with very high nitrate content up to Discussion 80 μmol g 1 DW in sunflower (Fig. 3), as observed previously (Alves et al., 2016). The presence of two nitrate pools in roots 15N-enrichment in nitrate has been demonstrated – although the biochemical nature of The d15N in nitrate extracted from plant tissue is relatively these two pools (cytosolic vs. vacuolar, apoplastic vs. symplastic, poorly documented, but available data (e.g. data used for mod- and so on) remains uncertain (Ferrari et al., 1973; MacKown elling in Robinson et al. (1998) or data in Yoneyama & Tanaka et al., 1983; Siddiq et al., 1991; Devienne et al., 1994; Miller &

Ó 2020 The Authors New Phytologist (2020) New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com

165 New 8 Research Phytologist

60 (a) Low K (c) Sunflower Oil palm High K 20

) 1.4 LowHigh KK + + WL WL Low K

50 –1 HighLow KK ++ WLWL 1.2 High K 15 Low K + WL Linear reg. (sunflower) 1.0 High K + WL 40 1 : 1 line 0.8 Linear reg. (oil palm) 10

N 0.6 51 0.4 5

devresbO 30 0.2

T (mmol per plant d 0.0 0 20 0123 01234

) 0.6 0.12 –1 10 0.10 0.5 0.08 0.4 0 0.06 0.3 0.04 0.2

0102030405060 0.02 0.1

15 B (mmol per plant d 0.00 Modelled N 0.0 (b) 0123 01234

) 0.30 4

10 –1 Sunflower, no WL 0.25 Sunflower, with WL 3 ) 0.20 1– Oil palm, no WL s

2– Oil palm, with WL 0.15 2 8 1 : 1 line mlomm(noitaripsnartdevresbO 0.10 1 0.05 (mmol per plant d n

6 b 0.00 0 0123 01234

) 0.8 0.5

4 –1 0.4 0.6 0.3 2 0.4 0.2 0.2 0.1

0 w (mmol per plant d 0.0 0.0 0246810 0123 01234 Predicted transpiration (mmol m–2 s–1) Growth step Growth step Fig. 4 Output of d15N modelling in nitrate. (a) Correlation between modelled and observed d15N values (in &). Red frames highlight points that deviate from the 1 : 1 line. Sunflower and oil palm symbols are discs and squares, respectively. With all points taken together, R2 = 0.95. Each point is the average values of n = 5 replicates (all organs plotted together). (b) Relationship between predicted transpiration (assuming that xylem nitrate concentration is 3 mM) and observed transpiration across all K conditions, with (grey) and without (back) waterlogging. The determination coefficient (R2) is 0.72 without + waterlogging (P < 0.01) while there is no significant correlation under waterlogging. (c) Outputs of the model showing nitrate transport via xylem (T), + + xylem -to-phloem transfer (B), nitrate backflow from leaves (bn), and nitrate exchange between root pools (w). Fluxes are given in mmol nitrate per plant d 1. Note the peak associated with w upon waterlogging onset.

Smith, 1996). It is generally believed that the flux of nitrate via relatively small flux of nitrate via the phloem is possible. Here, + the phloem is very small as nitrate is present at very low concen- d15N modelling suggests that it is indeed the case, with phloem tration if at all detectable in several species (especially legumes), carrying 15N-enriched nitrate (by up to 30& compared to + and 15N-nitrate labelling via the xylem suggests that shoot nitrate xylem ). The flux of nitrate export from leaves generated by the reduction is so efficient that only trace amounts of nitrate circu- model (bn) was rather small under normal conditions (no water- late via the phloem (Pate et al., 1975). However, nitrate is logging), between 7% (sunflower) and 3% (oil palm) of the detectable in phloem sap of other species (such as castor bean and nitrate flux via the xylem (Fig. 4). This order of magnitude seems yucca) and extensive measurements in castor bean (R. communis) to match experimental data on palm trees: in adult palm trees, have shown that it is c. 0.6 mM (Peuke, 2009). In palm trees, xylem water transport represents c. 600 l d 1 and thus c. 2 mol – 1 nitrate in palm wine obtained from phloem exudates has been NO3 d (assuming a concentration of 3 mM) while phloem found to be c. 20–30 μgml 1, that is 0.4 mM (Ojimelukwe, transport through the stem represents c. 50 l d 1 (Van Die & – 1 2001; Chandrasekhar et al., 2012). Interestingly, nitrate trans- Tammes, 1975) and thus 0.02 mol NO3 d (at 0.4 mM), that porters that are present in phloem cells and essential to redis- is, 1% of the xylem flux. Interestingly, under waterlogging, bn tribute leaf nitrate to sink organs have been found in Arabidopsis was found to decrease in sunflower, suggesting an inhibition of (Fan et al., 2009; Wang & Tsay, 2011). Thus, in principle, a phloem circulation as in Ricinus (Peuke et al., 2015) or forest

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust

166 New Phytologist Research 9 trees (Kreuzwieser & Rennenberg, 2014), but increased consider- In addition, our calculations do not take into account the contri- ably in oil palm (suggesting an increase in leaf nitrate remobilisa- bution of veins in leaves, or the loss of organic matter in roots by tion). exudation or fine root turnover. In that sense, our model is rather Root nitrate is further 15N-enriched by nitrate reduction in similar to the structure in basic elements of circulation described root metabolism: although small (shoots prevail in whole-plant previously (Pate, 1975). In doing so, the model was deliberately nitrate reduction in the two species of interest), nitrate assimila- kept simple to avoid the use of many parameters that are not tion participates in enriching root nitrate by several per mil. It is accessible experimentally, but could reproduce observed d15N worth noting that in sunflower roots, the d15N value of proteins data (Fig. 4) by adjusting nitrate fluxes. As expected, the direct was similar to that in leaves (Fig. S6a,c) and this shows that roots xylem-to-phloem nitrate transfer (B) was small (Van Bel, 1990) used nitrate of similar isotopic composition to that in leaves and/ and was thus in agreement with experiments on legumes (Pate, + or inherited amino acids produced in leaves. In oil palm, the 1975). Modelled values of nitrate flux in xylem decreased as d15N value of root proteins was higher than that in leaves by sev- expected under low K (see Mengel & Radomir (1973) in sun- eral per mil (except at week 7 under low K + waterlogging) flower) and led to a consistent order of magnitude for leaf tran- (Fig. S6d,f) showing that root metabolism did not only inherit spiration, using a fixed value of nitrate concentration in xylem amino acids from leaves but also assimilated nitrate that was more sap of 3 mM. Of course, this value is debatable but 3 mM is real- 15N-enriched than leaf nitrate itself. istic. In castor bean, xylem sap contains on average 7 mM nitrate We recognise that our study did not allow the purification and (Peuke, 2009). In sunflower, measurements are scarce but sam- isotopic analysis of nitrate from xylem and phloem sap. The vol- ples of sap obtained with decapitation or a pressure chamber sug- ume of sap required would have been far too large (typically 0.1– gest a value of between 0.9 and 7.3 mM (Mengel & Radomir, 1 ml xylem sap would have been necessary; even more for phloem 1973; Udayakumar et al., 1981; Gollan et al., 1992). To our sap because nitrate concentration is low) and thus this was techni- knowledge, in oil palm, nitrate concentration in xylem sap has cally not feasible. Nevertheless, our dataset includes isotopic analy- never been measured. sis of stems. In sunflower, the d15N value of stem nitrate was Of particular significance is the low performance of modelling generally intermediate between leaves and roots (Fig. 3b). This is in sunflower at high K (Fig. 4a, red frames): with the present unsurprising because stems should reflect a mixture of 15N-de- model, it was not possible to reproduce perfectly the high 15N- pleted xylem nitrate (from root nitrate absorption) and 15N-en- enrichment in roots (with a difference of up to 10&). There are riched phloem nitrate (inherited from leaf nitrate export), and several potential causes for this discrepancy: root turnover and such a mixture must be isotopically lighter than nitrate accumu- exudation, nitrate loss (efflux) from the storage pool or an under- lated in roots. In oil palm, stem nitrate was generally less 15N-en- estimation of leaf nitrate d15N. In fact, fine root turnover and riched than in leaves (Fig. 3h). Nitrate concentration was also very exudation would represent a loss of organic matter that would low in oil palm stems (Fig. 3k), suggesting that there was much less have to be taken into account in rr (nitrate reduction rate in opportunity for phloem nitrate to accumulate (meaning that, in roots). That is, not accounting for this loss would have led to an 15 mathematical terms, ap was a small number) than in sunflower. underestimation of rr and thereby of the N-enrichment in root Interestingly, isotopic analyses in phloem and xylem sap of nitrate. Similarly, not accounting for potential nitrate efflux back Ricinus are consistent with phloem carrying 15N-enriched nitrate. to soil from the storage pool would affect the d15N value because In fact, during the day (when leaf nitrate reduction occurs), efflux fractionates between isotopes (by 3&, Table 1). Leaf phloem d15N was quite close to that in xylem (within 0 and 4&) nitrate could also be slightly underestimated because leaves con- while leaf soluble fraction was 15N-depleted (c. 2& on average tain xylem vessels in veins, and xylem nitrate is isotopically at midday) (Peuke et al., 2013), suggesting that amino acids car- lighter. We have proposed previously that this effect might cause ried by phloem sap could not explain satisfactorily the isotopic sig- an underestimation of d15N in leaf nitrate by several per mil, and nature of phloem sap and thus other N-containing compounds in thus affects the isotope composition of nitrate going back to roots phloem must cause a 15N-enrichment. For example, if nitrate rep- via the phloem (Cui et al., 2019a). resented only 3% of phloem sap N, it could explain the results We also recognise that our study used relatively high nitrate con- obtained by Peuke et al. (2013) with a d15N value of 100&.Itis centration (12 mM) and this allowed us to have isotopically constant worth noting that the diurnal cycle of N transport via xylem and conditions (the d15N value of source nitrate remained constant). phloem is probably pronounced in oil palm because leaf %N (total However, our modelling would remain valid at lower nitrate con- N) varied by 0.05 units (35 μmol N g 1 DW) between day and centration. Should nitrate concentration not be constant, it would night (Scheidecker & Prevot, 1954). only require the addition of a differential equation for medium nitrate to account for changes in content and d15N. Model validity Nitrate circulation under waterlogging and low K Key parameters associated with nitrate circulation could not be measured experimentally and thus we carried out modelling. Of There was an effect of waterlogging and K conditions on nitrate course, our model was crude in that it could not capture all the d15N and concentration. Waterlogging caused a general decrease subtleties of nitrate redistribution and utilisation. That is, we in nitrate content in all tissues (Fig. 3), consistent with an inhibi- used three compartments, one of them (roots) having two pools. tion of nitrate absorption (influx). It also led to a progressive

Ó 2020 The Authors New Phytologist (2020) New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com

167 New 10 Research Phytologist

decline in d15N in nitrate in particular in roots, compared to con- 2002). It is worth noting that the nitrate-corrected soluble frac- trol conditions, resulting from lower nitrate reduction rate tion exhibited a rather variable d15N value, similar to that of pro- (slower biomass N accumulation) and slower nitrate circulation teins, intermediate, or as 15N-depleted as in lipids. This reflects coupled with transient remobilisation of root nitrate (Fig. 4). the fact that the composition of the soluble fraction in N-con- The former effect leads to a lower 15N-enrichment in noncon- taining compounds (amino acids, polyamines, nucleotides, etc.) sumed nitrate molecules and the latter leads to smaller flow of varied and these different compounds probably had distinct d15N 15 + N-enriched nitrate from phloem (lower bn) and allows the values (not analysed here). consumption of nitrate from the root storage pool (w > 0). Such a situation was particularly visible in sunflower (Fig. 4) and is Perspectives consistent with the general depressing effect of waterlogging on N metabolism in roots in both species but also on leaf-exported Our results provide an overview of nitrate d15N is plant organs N assimilates such as asparagine (Cui et al., 2019a,b). and clearly show that nitrate is considerably 15N-enriched in Low K conditions also had a considerable effect on d15N val- roots (Fig. 3) in the two species of interest, sunflower and oil ues, with much less 15N-enriched nitrate (except in sunflower palm. That is, root nitrate appeared to be much more enriched + leaves where leaf nitrate did not change much) (Fig. 3). K is the than would be expected if nitrate were simply transported to and cornerstone of phloem function and a common counter-cation accumulated in leaves, therefore suggesting that 15N-enriched for nitrate in xylem, so an effect on nitrate circulation was nitrate can be transported back to roots from leaves via the expected. In R. communis, potassium limitation increased nitrate phloem (d15N values summarised in Fig. S8). It does not imply concentration in both xylem and phloem (Peuke et al., 2002). that the concentration or flux of nitrate in the phloem has to be + + Here, in sunflower, both modelled xylem and phloem nitrate high. Here, modelling shows that even small (a few per cent of transport rates were much lower at low K. However, this was not xylem nitrate flux), the phloem nitrate flux has a considerable reflected by the nitrate content, which was not lower but higher. impact on root d15N. Of course, this effect probably varies widely This simply shows that plant nitrate absorption was less affected between species, which differ in their ability to carry nitrate via than nitrate circulation or utilisation by metabolism. That is, the the phloem. For example, past measurements have shown that N demand to form organic matter was proportionally lower, as nitrate is usually undetectable in phloem sap of legumes. Still, demonstrated previously (Cui et al., 2019a). In oil palm, nitrate our findings are highly significant because they suggest that the concentration was lower under low K, suggesting an effect on natural abundance in nitrate can be used as a tracer of plant sap nitrate absorption by roots and, furthermore, nitrate concentra- circulation and support the view that nitrate remobilisation from tion correlated with the d15N, in particular in stems (Fig. 3h). leaves is an important aspect of plant N budget. We nevertheless Thus, in contrast to sunflower, low K conditions inhibited not recognise that more data would be required on the isotope com- only nitrate transport but also nitrate acquisition and thus the position of nitrate purified from saps but this is at present too ability to store (15N-enriched) nitrate. The strong effect of potas- challenging technically in particular for phloem, because of low sium limitation on N acquisition is well known in this species, nitrate concentrations and sap volumes that can be sampled (of with many agronomical trials showing the strict dependence of N the order of microlitres (Dinant & Kehr, 2013; Palmer et al., absorption on K (Chapman & Gray, 1949; Ollagnier & Ochs, 2014) and thus maximally 0.3 μmol nitrate per sample, while 1973; Chan, 1981), and the strong decrease in elemental N con- EA-IRMS would require minimally 1–2 μmol nitrate per sam- tent (%N) in all organs (data not shown). ple). Conversely, N circulation in the plant also involves organic molecules highly concentrated in phloem sap, and some information could be drawn from compound-specific analysis. This will Isotope fractionation in N assimilation be addressed in a future study. There were significant differences between tissue fractions, with 15 15 nitrate being the most N-enriched while lipids were N-de- Acknowledgements pleted, except in leaves. Accounting for the d15N of the different fractions allowed us to estimate the fractionation associated with We thank Cyril Abadie, Illa Tea, Anne-Marie Schiphorst, nitrate reduction, which was, as expected, near 16& (Fig. 2) Camille Bathellier and members of the RSB Plant Service for (Ledgard et al., 1985; Tcherkez & Farquhar, 2006). In leaves, the their help during oil palm sampling. The support of the Joint lipid fraction, enriched in chlorophylls, is isotopically similar to Mass Spectrometry Facility (ANU) is acknowledged. JC was sup- proteins (Fig. S4), and this has been found previously in rapeseed ported by an Australia Awards PhD Scholarship, and the research leaves (Gauthier et al., 2013). This simply comes from the fact was supported by the Australian Research Council via a Future that N atoms in chlorophylls come from glutamate (via d- Fellowship, under contract FT140100645. We thank S. Cho- aminolevulinate) with little isotope fractionation. By contrast, N- lathan (Siam Elite Palm) for providing oil palm seeds. containing lipids mostly originate from serine (e.g. ethanolamine 15 and choline moieties of phospholipids), which is generally N- Author contributions depleted compared to other amino acids (Tcherkez, 2011; Gau- thier et al., 2013). In addition, there may be some isotopic frac- GT and EL designed the experiments. JC performed palm culti- tionation during polar lipid synthesis (Werner & Schmidt, vation, biomass measurements and isotopic analyses. FF

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust

168 New Phytologist Research 11 performed extra isotopic analyses. All authors contributed to Fourel F, Martineau F, Seris M, Lecuyer C. 2014. Simultaneous N, C, S stable writing the paper. isotope analyses using a new purge and trap elemental analyzer and an isotope ratio mass spectrometer. Rapid Communications in Mass Spectrometry 28: 2587– 2594. ORCID Gauthier PP, Lamothe M, Mahe A, Molero G, Nogues S, Hodges M, Tcherkez G. 2013. Metabolic origin of d15N values in nitrogenous compounds from Guillaume Tcherkez https://orcid.org/0000-0002-3339-956X Brassica napus L. leaves. Plant, Cell & Environment 36: 128–137. Gollan T, Schurr U, Schulze E-D. 1992. Stomatal response to drying soil in relation to changes in the xylem sap composition of Helianthus annuus. I. The concentration of cations, anions, amino acids in, and pH of, the xylem sap. References Plant, Cell & Environment 15: 551–559. Aguera€ E, de la Haba P, Fontes AG, Maldonado JM. 1990. Nitrate and nitrite Handley L, Robinson D, Forster B, Ellis R, Scrimgeour C, Gordon D, Nevo E, uptake and reduction by intact sunflower plants. Planta 182: 149–154. Raven J. 1997. Shoot d15N correlates with genotype and salt stress in barley. Alves LS, Torres CV, Fernandes MS, Marques dos Santos A, De Souza SR. Planta 201: 100–102. 2016. Soluble fractions and kinetics parameters of nitrate and ammonium Hartt CE. 1969. Effect of potassium deficiency upon translocation of 14Cin uptake in sunflower (Neon Hybrid). Revista Ciência Agronômica 47:13–21. attached blades and entire plants of sugarcane. Plant Physiology 44: 1461–1469. Barbour MM, Roden JS, Farquhar GD, Ehleringer JR. 2004. Expressing leaf H€ogberg P. 1997. 15N natural abundance in soil–plant systems. New Phytologist water and cellulose oxygen isotope ratios as enrichment above source water 137: 179–203. reveals evidence of a Peclet effect. Oecologia 138: 426–435. Huber B, Bernasconi Stefano M, Luster J, Pannatier-Graf E. 2011. A new Barbour MM, Schurr U, Henry BK, Wong SC, Farquhar GD. 2000. Variation isolation procedure of nitrate from freshwater for nitrogen and oxygen isotope in the oxygen isotope ratio of phloem sap sucrose from castor bean. Evidence in analysis. Rapid Communications in Mass Spectrometry 25: 3056–3062. support of the Peclet effect. Plant Physiology 123: 671–680. Johnson JE, Berry JA. 2013. The influence of leaf-atmosphere NH3 (g) exchange Barrett-Lennard EG, Shabala SN. 2013. The waterlogging/salinity interaction in on the isotopic composition of nitrogen in plants and the atmosphere. Plant, + higher plants revisited – focusing on the hypoxia-induced disturbance to K Cell & Environment 36: 1783–1801. homeostasis. Functional Plant Biology 40: 872–882. Kendall C, Aravena R. 2000. Nitrate isotopes in groundwater systems. In: Cook Cakmak I, Hengeler C, Marschner H. 1994. Changes in phloem export of P, Herczeg A, eds. Environmental tracers in subsurface hydrology. New York, NY, sucrose in leaves in response to phosphorus, potassium and magnesium USA: Springer, 261–297. deficiency in bean plants. Journal of Experimental Botany 45: 1251–1257. Kolb K, Evans R. 2003. Influence of nitrogen source and concentration on Chan K. 1981. Nitrogen requirements of oil palms in Malaysia: fifty years of nitrogen isotopic discrimination in two barley genotypes (Hordeum vulgare L.). experiments conducted by Gutheries. In: Pushparajah E, Chew P, eds. The oil Plant, Cell & Environment 26: 1431–1440. palm in agriculture in the eighties. Kuala-Lumpur, Malaysia: The Incorporated Kreuzwieser J, Rennenberg H. 2014. Molecular and physiological responses of Society of Planters, 119–141. trees to waterlogging stress. Plant, Cell & Environment 37: 2245–2259. Chandrasekhar K, Sreevanin S, Seshapani P, Pramodhakumari J. 2012. A review Ledgard S, Woo K, Bergersen F. 1985. Isotopic fractionation during reduction of on palm wine. International Journal of Research in Biological Sciences 2:33–38. nitrate and nitrite by extracts of spinach leaves. Functional Plant Biology 12: Chapman GW, Gray HM. 1949. Leaf analysis and the nutrition of oil palm. 631–640. Annals of Botany 13: 415–433. MacKown C, Jackson W, Volk R. 1983. Partitioning of previously-accumulated Comstock J. 2001. Steady-state isotopic fractionation in branched pathways using nitrate to translocation, reduction, and efflux in corn roots. Planta 157: – – – plant uptake of NO3 as an example. Planta 214: 220 234. 8 14. Cui J, Abadie C, Carroll A, Lamade E, Tcherkez G. 2019a. Responses to K Mariotti A, Mariotti F, Amarger N. 1983. Use of natural 15N abundance in the deficiency and waterlogging interact via respiratory and nitrogen metabolism. measurement of symbiotic fixation. In: Nuclear techniques in improving pasture Plant, Cell & Environment 42: 647–658. management. Vienna, Austria: IAEA, 61–77. Cui J, Davanture M, Zivy M, Lamade E, Tcherkez G. 2019b. Metabolic Mengel K, Haeder H-E. 1977. Effect of potassium supply on the rate of phloem responses to potassium availability and waterlogging reshape respiration and sap exudation and the composition of phloem sap of Ricinus communis. Plant carbon use efficiency in oil palm. New Phytologist 223: 310–322. Physiology 59: 282–284. + Demidchik V. 2014. Mechanisms and physiological roles of K efflux from root Mengel K, Radomir S. 1973. Effect of potassium supply on the acropetal cells. Journal of Plant Physiology 171: 696–707. transport of water, inorganic ions and amino acids in young decapitated Devienne F, Mary B, Lamaze T. 1994. Nitrate transport in intact wheat roots: II. sunflower plants (Helianthus annuus). Physiologia Plantarum 28: 232–236. Long-term effects of NO3 concentration in the nutrient solution on NO3 Miller AJ, Smith SJ. 1996. Nitrate transport and compartmentation in cereal unidirectional fluxes and distribution within the tissues. Journal of Experimental root cells. Journal of Experimental Botany 47: 843–854. Botany 45: 677–684. Ojimelukwe P. 2001. Effect of preservation with Saccoglottis gabonensis on the Dinant S, Kehr J. 2013. Sampling and analysis of phloem sap. In: Maathuis FJM, biochemistry and sensory attributes of fermenting palm wine. Journal of Food ed. Plant mineral nutrients: methods and protocols. Totowa, NJ, USA: Humana Biochemistry 25: 411–424. Press, 185–194. Ollagnier M, Ochs R. 1973. Interaction entre l’azote et le potassium dans la Doman DC, Geiger DR. 1979. Effect of exogenously supplied foliar potassium nutrition des oleagineux tropicaux. Oleagineux 28: 493–507. on phloem loading in Beta vulgaris L. Plant Physiology 64: 528–533. Palmer LJ, Dias DA, Boughton B, Roessner U, Graham RD, Stangoulis JCR. Epron D, Cabral OMR, Laclau J-P, Dannoura M, Packer AP, Plain C, Battie- 2014. Metabolite profiling of wheat (Triticum aestivum L.) phloem exudate. et al 13 – Laclau P, Moreira MZ, Trivelin PCO, Bouillet J-P . 2015. In situ CO2 Plant Methods 10:27 32. pulse labelling of field-grown eucalypt trees revealed the effects of potassium Pate JS. 1975. Exchange of solutes between phloem and xylem and circulation in nutrition and throughfall exclusion on phloem transport of photosynthetic the whole plant. In: Zimmermann MH, Milburn JA, eds. Transport in plants I: carbon. Tree Physiology 36:6–21. Phloem transport. Berlin, Germany: Springer, 451–473. Evans RD. 2001. Physiological mechanisms influencing plant nitrogen isotope Pate JS, Sharkey PJ, Lewis OAM. 1975. Xylem to phloem transfer of solutes in composition. Trends in Plant Science 6: 121–126. fruiting shoots of legumes, studied by a phloem bleeding technique. Planta Fan S-C, Lin C-S, Hsu P-K, Lin S-H, Tsay Y-F. 2009. The Arabidopsis nitrate 122:11–26. transporter NRT1.7, expressed in phloem, is responsible for source-to-sink Peuke AD. 2009. Correlations in concentrations, xylem and phloem flows, and remobilization of nitrate. The Plant Cell 21: 2750–2761. partitioning of elements and ions in intact plants. A summary and statistical re- Ferrari TE, Yoder OC, Filner P. 1973. Anaerobic nitrite production by plant evaluation of modelling experiments in Ricinus communis. Journal of cells and tissues: evidence for two nitrate pools. Plant Physiology 51: 423–431. Experimental Botany 61: 635–655.

Ó 2020 The Authors New Phytologist (2020) New Phytologist Ó 2020 New Phytologist Trust www.newphytologist.com

169 New 12 Research Phytologist

Peuke AD, Gessler A, Tcherkez G. 2013. Experimental evidence for diel d15N- Yeh H, Wills G. 1970. Diffusion coefficient of sodium nitrate in aqueous patterns in different tissues, xylem and phloem saps of castor bean (Ricinus solution at 25°C as a function of concentration from 0.1 to 1.0 M. Journal of communis L.). Plant, Cell & Environment 36: 2219–2228. Chemical and Engineering Data 15: 187–189. Peuke AD, Gessler A, Trumbore S, Windt CW, Homan N, Gerkema E, Van As Yoneyama T, Tanaka F. 1999. Natural abundance of 15N in nitrate, ureides, and H. 2015. Phloem flow and sugar transport in Ricinus communis L. is inhibited amino acids from plant tissues. Soil Science and Plant Nutrition 45: 751–755. under anoxic conditions of shoot or roots. Plant, Cell & Environment 38: 433– Yoneyama T, Yamada H, Yamagata M, Kouchi H. 1987. Nitrate reduction and 447. partitioning of nitrogen in komatsuna (Brassica campestris L. var. rapa) plants: Peuke AD, Jeschke WD, Hartung W. 2002. Flows of elements, ions and abscisic compartmental analysis in combination with 15N tracer experiments. Plant and acid in Ricinus communis and site of nitrate reduction under potassium Cell Physiology 28: 679–696. limitation. Journal of Experimental Botany 53: 241–250. Zeng F, Konnerup D, Shabala L, Zhou M, Colmer Timothy D, Zhang G, Phukan UJ, Mishra S, Shukla RK. 2016. Waterlogging and submergence stress: Shabala S. 2014. Linking oxygen availability with membrane potential + affects and acclimation. Critical Reviews in Biotechnology 36: 956–966. maintenance and K retention of barley roots: implications for waterlogging Robinson D, Handley L, Scrimgeour C. 1998. A theory for 15N/14N stress tolerance. Plant Cell and Environment 37: 2325–2338. fractionation in nitrate-grown vascular plants. Planta 205: 397–406. Sacks GL, Brenna JT. 2005. 15N/14N position-specific isotopic analyses of polynitrogenous amino acids. Analytical Chemistry 77: 1013–1019. Supporting Information Scheidecker D, Prevot P. 1954. Nutrition minerale du palmier a huile a Pobe (Dahomey). Oleagineux 9:13–19. Additional Supporting Information may be found online in the Shabala S, Pottosin I. 2014. Regulation of potassium transport in plants under Supporting Information section at the end of the article. hostile conditions: implications for abiotic and biotic stress tolerance. – Physiologia Plantarum 151: 257 279. Fig. S1 Scheme of the experimental design. Sharma PK, Sharma SK, Choi IY. 2010. Individual and combined effects of waterlogging and alkalinity on yield of wheat (Triticum aestivum L.) imposed at three critical stages. Physiology and Molecular Biology of Plants 16: 317–320. Fig. S2 Sensitivity of the model to parameterisation. Siddiq MY, Glass ADM, Ruth TJ. 1991. Studies of the uptake of nitrate in barley: III. Compartmentation of NO3 . Journal of Experimental Botany 42: Fig. S3 Shoot : root ratio in both species over the course of the – 1455 1463. experiment. Stock WD, Evans JR. 2006. Effects of water availability, nitrogen supply and atmospheric CO2 concentrations on plant nitrogen natural abundance values. d15 Functional Plant Biology 33: 219–227. Fig. S4 N values in total organic matter. Sung J, Sonn Y, Lee Y, Kang S, Ha S, Krishnan HB, Oh TK. 2015. Compositional changes of selected amino acids, organic acids, and soluble Fig. S5 N elemental content in total organic matter. sugars in the xylem sap of N, P, or K-deficient tomato plants. Journal of Plant Nutrition and Soil Science 178: 792–797. 15 15 14 Fig. S6 d N values in tissue fractions (proteins, lipids, soluble Tcherkez G. 2011. Natural N/ N isotope composition in C3 leaves: are enzymatic isotope effects informative for predicting the 15N-abundance in key fraction). metabolites? Functional Plant Biology 38:1–12. Tcherkez G, Farquhar GD. 2006. Isotopic fractionation by plant nitrate Fig. S7 Example of d15N modelling under low K and waterlog- – reductase, twenty years later. Functional Plant Biology 33: 531 537. ging. Tcherkez G, Hodges M. 2007. How stable isotopes may help to elucidate primary nitrogen metabolism and its interaction with (photo) respiration in C3 d15 leaves. Journal of Experimental Botany 59: 1685–1693. Fig. S8 Graphical summary of modelled N values. Udayakumar M, Devendra R, Sreenivasa-Reddy V, Krishna-Sastry K. 1981. Nitrate availability under low irradiance and its effect on nitrate reductase Notes S1 Equations. activity. New Phytologist 88: 289–297. Van Bel A. 1990. Xylem-phloem exchange via the rays: the undervalued route of Notes S2 Model parameterisation. transport. Journal of Experimental Botany 41: 631–644. Van Die J, Tammes P. 1975. Phloem exudation from monocotyledonous axes. In: Zimmermann MH , Milburn JA, eds. Transport in Plants I. Berlin, Table S1 Nutrient solution composition. Germany: Springer, 196–222. Wang Y-Y, Tsay Y-F. 2011. Arabidopsis nitrate transporter NRT1.9 is important Table S2 Model parameterisation. in phloem nitrate transport. The Plant Cell 23: 1945–1957. Werner RA, Schmidt H-L. 2002. The in vivo nitrogen isotope discrimination among organic plant compounds. Phytochemistry 61: 465–484. Please note: Wiley Blackwell are not responsible for the content White P. 2012. Long distance transport in the xylem and phloem. In: Marschner or functionality of any Supporting Information supplied by the P, ed. Mineral nutrition of higher plants. Berlin, Germany: Academic Press authors. Any queries (other than missing material) should be Elsevier, 49–70. directed to the New Phytologist Central Office.

New Phytologist (2020) Ó 2020 The Authors www.newphytologist.com New Phytologist Ó 2020 New Phytologist Trust

170 New Phytologist Supplementary Material

Article title: δ15N values in plants is determined by both nitrate assimilation and circulation Authors: Jing Cui, Emmanuelle Lamade, François Fourel and Guillaume Tcherkez

Table S1. Nutrient solution composition.

Table S2. Model parameterization.

Notes S1. Equations.

Notes S2. Model parameterization.

Fig. S1. Scheme of the experimental design.

Fig. S2. Sensitivity of the model to parameterization.

Fig. S3. Shoot-root ratio in both species over the course of the experiment.

Fig. S4. δ15N values in total organic matter.

Fig. S5. N elemental content in total organic matter.

Fig. S6. δ15N values in tissue fractions (proteins, lipids, soluble fraction).

Fig. S7. Example of δ15N modelling under low K and waterlogging.

Fig. S8. Graphical summary of modelled δ15N values.

Supplementary references

171 Table S1. Nutrient solution composition used in experiments. KCl concentration was fixed at 4 or 0.2 mM to have high and low K conditions, respectively.

Compound Concentration (mM) KCl 4/0.2 NaNO3 4 Ca(NO3)2 4 MgSO4 1.5 NaH2PO4 1.33 Fe-Na (EDTA) 0.11 -3 MnSO4 2⋅10 ZnSO4 2⋅10-3 CuSO4 2⋅10-3 H3BO3 0.1 Na2MoO4 1⋅10-3 Co(NO3)2 4⋅10-4

Table S2. Parameterization used for modelling δ15N values in nitrate. Units and definitions are given in Table 1 (main text). rshoot and rtot refers to total shoot (stem+ leaves) and whole-plant (stem + leaves + roots) N assimilation, respectively. (a) this value was changed to 0.65 under waterlogging (transient inhibition of shoot N assimilation under waterlogging). (b) this value was changed to 0.5 under high K + waterlogging. (c) this value was adjusted for better fitting, to 0.25 (high K + waterlogging, step 0), 0 (high K, step 2), 0.1 (low K, step 4) and 0.9 (low K + waterlogging, step 4). (d) the proportion of root nitrate reduction (with respect to whole plant nitrate reduction) in oil palm was allowed to vary to account for possible changes suggested by the shoot:root ratio (Fig. S3) or caused by waterlogging. In practice, this value varied between 0.05 (high K, steps 3 and 4) and 1.0 (step 1 under waterlogging at high K).

Oil palm Sunflower Parameter Step 0 Step 1 Step 2 Step 3 Step 4 Step 0 Step 1 Step 2 Step 3 ap/(ax+ap) 0.9 0.9 0.9 0.9 0.9 0 0 0 0 e/i 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 0.1 c c c c c (rp+rx)/rshoot 0.5 0.5 0.5 0.5 0.5 0.3 0.3 0.3 0.3 d d d d d a rr/rtot 0.06 0.06 0.06 0.06 b b b b b rx/(rp+rx) 0 0 0 0 0 0.5 0.5 0.5 0.5

172 Supplementary Notes S1. Equations

Pool sizes:

dQR1 Root pool R1: = 0 (root pool R1 is assumed to be in the steady-state) (S1) dt Root pool R2: def dQR2 ==++ar() Aw( bn −−+−−=++−++ rp a p B) wrr() Ab n B() r p a pr r (S2) dt dQ Xylem+: X = a (S3) dt x dQ Phloem+: P = a (S4) dt p dQ Leaves: L = a (S5) dt l In Eqs. (S1-5), it should be noted that a values are not constant over time. They have a different value at the different growth steps (see parameterization in Table S2).

Isotope ratios: Derivatives with respect to time are as follows (general expression of Eq. 2):

dR 1 Rin R =Fin − Fout −−() FFRin out dt Q ααin out Applied to fluxes and pools considered here (depicted in Fig. 1, main text), we have:

Root nitrate pool R1 (pool size in the steady-state thus dQR1/dt = 0):

dRR 1 iR e 1 =i +wR − R ⋅ T +++ A w dt Q ααRR21 Rd1 d  dRR 11iR  1 =i + − − ⋅+ −1 (S6) wR()RR21 R R R 1 i e dt Q αα Rd1 d Root nitrate pool R2:

dRR 1 r 2 r =(branp −−+⋅++⋅−+ p BR) P() AwRR w ⋅−−−++−⋅ RR( branp p BArRr) R dt Q 12α 2 Rr2  dR 11 R2 = −−+⋅ − ++⋅ − +1 − ⋅ (bnp r a p B) () R P R R2() Aw() RR12 R R r r RR2 dt Q α Rr2  (S7) Xylem+ nitrate pool: dR 11 X =TR ⋅−+−⋅ R r1 R (S8) ()RXx1  X dt QXrα Phloem+ nitrate pool: dR 11 P = ⋅−+⋅−+1 −⋅ (S9) BR()()X R P b nLP R R r p RP dt QPrα

173 Leaf nitrate pool: dR 11 L = −− − ⋅ − +1 − ⋅ (S10) (Trxx a B)() R X R L r l RL dt QLrα Because of the steady-state hypothesis on the root pool R1, we have T = i – e – A. Also, if we take into account the imbalance of the leaf nitrate pool al, we have: T – ax – rx – B = bn + rl + al. Note that A can be > 0 or < 0, with the condition A + w ≥ 0, that is, A ≥ – w. Therefore, A can be negative number. To perform computations where only positive numbers have to be solved numerically, we rather solved for w, B and bn, from which T and A could be calculated.

174 Supplementary Notes S2. Model parameterization

From Fig. 1 it can be seen that since the transfer pool in roots is in the steady-state, we have T = i + w – (e + A + w) = i – e – A. Numerical solving allowed the determination of w, bn and B by optimization (minimization of sum of squares between observed and modelled δ15N and pool sizes). The value of i – e was determined with the total demand computed from biomass and %N (increase in total N per plant per time unit). In each organ, nitrate quantitation and biomass was used to calculate pool size variation dQ/dt, referred to as accumulation fluxes al and ar in leaves and roots, respectively, and ax + ap in stems. Since whole stems were analyzed, ax and ap were not measured individually and thus assumptions had to be made on nitrate accumulation partitioning between xylem+ and phloem+. Different scenarios have been tested and the impact on computations is very small (addressed below, and Fig. S2).

Also, r values (nitrate reduction rates) cannot be obtained directly. While the increase in organic N can be easily measured in each organ from biomass, %N and the subtraction of the contribution of nitrate to total N content, N assimilates can be exchanged between organs (typically, exported by leaves). That is, what can be measured is total N assimilation rtot = rl + rx + rp + rr. Since sunflower N assimilation mostly (~94%) comes from shoot nitrate reductase activity (Andrews, 1986), the contribution of different organs to N assimilation was chosen as 0.06 (root), 0.28 (stem) and 0.66 (leaves). These contributions may change with conditions such as K availability (Förster & Dieter Jeschke, 1993); however, changing these contributions does not have a huge impact on computed values of w, bn and B (Fig. S2). In oil palm, to our knowledge, there is no direct information on shoot-root N assimilation (nitrate reduction) ratio. Using an oil palm root EST library, it has been found that gene(s) encoding nitrate reductase are also expressed in roots (Ho et al., 2007). Leaf nitrate reductase activity has been measured in this species - -1 -1 in the field and is, minimally, of 75 µmol NO3 h g DW (if nitrate cellular concentration is at the Km value) (Eschbach, 1982), suggesting that nitrate is mostly assimilated in leaves. However, in oil palm, conditions had a significant impact on the shoot:root ratio (while it was rather invariant in sunflower), suggesting that the proportion of N assimilation by roots (vs. shoots) may change (Fig. S3). Parameterization is given is Table S2. Upon modelling, the minimum sum of squares was obtained for a proportion of root nitrate reduction (rr/rtot) of 0.3. As will become apparent below, the prevalence of shoot (leaf) nitrate assimilation is also reflected by root proteins which were not highly 15N- enriched but rather similar to that in leaves, demonstrating that amino acids used by root metabolism mostly came from leaf nitrate reduction.

175

Sunflower Onset of waterlogging 1 2 3 4 Growth phases and sampling: Step 0 (15 d) Step 1 (2 d) Step 2 (5 d) Step 3 (7 d) Time increments for computations (1 increment = 3.6 min): 6,000 800 2,000 2,000

Oil palm Onset of waterlogging 1 2 3 4 5 Growth phases and sampling: Step 0 (11 months) Step 1 (1 w) Step 2 (1 w) Step 3 (1 w) Step 4 (4 w) Time increments for computations (1 increment = 3.36 min): 132,000 2,800 2,800 2,800 11,200

Colour code: Low K Low K + waterlogging High K High K + waterlogging

Fig. S1. Scheme of the experimental design and conditions. Plants were first grown under low (0.2 mM) or high (4 mM) K conditions for15 days (sunflower) and 11 months (oil palm), the first sampling was done, and then waterlogging was started for one half of the plants (the other half being under non-waterlogged conditions). For computations (numerical solving of differential equations), small time steps (time increments of 3.6 and 3.36 minutes) were used to avoid artefactual divergence of derivatives. The colour code used throughout the paper is recalled at the bottom.

176 0.7 1.4 0.6 1.2 0.5 1.0 w 0.4 0.8 0.3 0.6

b Error(‰) n Error(‰) 0.2 0.4 T 1 2 3 1 2 3 4 B 0.7 6

0.6 5 ) ) -1 -1 0.5 d d 4 -1 -1 0.4 3 0.3

2 0.2 Value (mmol plant (mmol Value Value (mmol plant (mmol Value

1 0.1

0.0 0 1 2 3 1 2 3 4 Scenario Scenario Fig. S2. Sensitivity of modelling with respect to parameterization: example at low K (last growth step) in sunflower (left) and oil palm (right). Note that nitrate exchange rate between root pools w is always found to be zero, and in most cases in oil palm, xylem- to-phloem transfer rate B is found to be zero. Best parameterization is found to be scenario 1 and 4 in sunflower and oil palm, respectively. Parameterization scenarios are as follows:

Sunflower Oil palm Scenario: 1 2 3 1 2 3 4 ap/(ax+ap) 0 0 0.5 0.9 0.9 0.9 0.9 rx/(rp+rx) 0.5 0.5 0.5 0.5 0.5 0 0 rr/rtot 0.06 0.06 0.06 0.06 0.06 0.06 0.3 rstem/rshoot 0.3 0.1 0.3 0.3 0.1 0.3 0.1

177 (a) Sunflower (b) Oil palm 20

8 ) ) 15 -1 -1 6

10 4 Shoot : root ratio (g g (g ratio root : Shoot Shoot : root ratio (g g (g ratio root : Shoot 2 5

0 0

0 2 4 6 8 10 12 14 0 1 2 3 4 5 6 7 Time (d) Time (w)

Fig. S3. Shoot-root ratio (g DW g-1 DW) in sunflower (a) and oil palm saplings (b). Data presented in this figure are mean±SD (n = 5). Time is in days (sunflower) and weeks (oil palm) after the onset of our experiment, that is, t = 0 corresponds to 15 days and 11 months after sowing, respectively. Note the opposite effect of waterlogging in sunflower (small decrease in shoot:root ratio at 14 d) and oil palm (general increase of shoot:root ratio from 1 week onwards). Colors are as in Fig. S1 (, low K; , high K; , low K + waterlogging; , high K + waterlogging).

178 (a) Sunflower leaves (b) Sunflower stem (c) Sunflower roots 2 15 2 LK HK 10 1 LK + WL 1 HK + WL

5 0 0

0 N

15 -1 -1  -5

-2 -2 -10

-3 -3 -15

-4 -20 -4 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 Time (d) Time (d) Time (d)

(d) Oil palm leaves (e) Oil palm stem (f) Oil palm roots 0 0 0

-2 -2 -2

-4 -4 -4

-6 -6 -6 N 15

 -8 -8 -8

-10 -10 -10

-12 -12 -12

-14 -14 -14

0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 Time (w) Time (w) Time (w)

Fig. S4. δ15N value in raw organic matter in leaves, stems and roots in sunflower (a-c) and oil palm (d-f). Mean±SD (n = 5). Note the difference in scale (y axis) in panel (b) compared to (a) and (c). Colors are as in Fig. S1 (, low K; , high K; , low K + waterlogging; , high K + waterlogging).

179 SUNFLOWER (a) leaves (b) stem (c) roots

5 5

4 4

% N % 3 3

2 2

1 1

LK HK LK HK LK HK LK+WLHK+WL LK+WLHK+WL LK+WLHK+WL

OIL PALM (d) leaves (e) stem (f) roots

5 5

4 4

% N % 3 3

2 2

1 1

LK LK HK +WL HK LK HK LK+WLHK LK+WLHK+WL LK+WLHK+WL

Fig. S5. Nitrogen elemental content in leaves, stems and roots in sunflower (a-c) and oil palm (d-f). The elemental content in given in %N (% dry weight). The horizontal red line indicates the mean value. Each data point is the average and SD of n = 5 replicates. The different shades of blue and green stand for the different sampling time points. LK, low K; HK, high K; WL, waterlogging.

180

(a) Sunflower leaves (b) Sunflower stem (c) Sunflower roots 30 30 30 LK, lipids 20 20 20 LK, SF LK, proteins 10 10 10 HK, lipids HK, SF 0 0 0 HK, proteins

N(‰) LK+WL, lipids

15 -10 -10 -10 LK+WL, SF  LK+WL, proteins -20 -20 -20 HK+WL, lipids -30 -30 -30 HK+WL, SF HK+WL, proteins -40 -40 -40 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 Time (d) Time (d) Time (d)

(d) Oil palm leaves (e) Oil palm stem (f) Oil palm roots 20 20 20

10 10 10

0 0 0 -10 N(‰)

15 -10 -10  -20

-20 -20 -30

-30 -40 -30 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 Time (w) Time (w) Time (w)

Fig. S6. Nitrogen isotope composition in tissue fractions (lipids, total soluble fraction, proteins) of different organs: leaves (a,d), stems (b,e) and roots (c,f), in sunflower (a-c) and oil palm (d-f). Total soluble fraction is corrected for nitrate (that is, the isotopic contribution of N from nitrate is subtracted by mass-balance). In (a), the thick black horizontal line stands for the average of leaf nitrate, which did not vary much in sunflower leaves. Data are mean±SD (n = 5). Time starts at the first sampling (15 d after sowing in sunflower, 11 months after sowing in oil palm). Note the generally low δ15N value in lipids (dashed-dotted) in stems and roots due to isotope fractionation during lipid synthesis while lipids are not 15N-depleted in leaves. Proteins appear to be rather close to source nitrate (3.5‰) in particular in sunflower, regardless of K and waterlogging conditions. Waterlogging caused a clear 15N-depletion in several fractions (lipids in sunflower stem and oil palm roots, soluble fraction in oil palm leaves). Colors are as in Fig. S1 (, low K; , high K; , low K + waterlogging; , high K + waterlogging).

181 (a)

(b)

Fig. S7. δ15N modelling and comparison with observed average values under low K conditions + waterlogging as an example, in sunflower (a) and oil palm (b). The onset of waterlogging is indicated with a back arrow (WL). Time is in days after sowing. Note that despite the good agreement between modelled and observed data, leaves appear to be generally more enriched than predicted in sunflower, perhaps reflecting the fact that the δ15N value of total leaf nitrate does not only represent mesophyll since leaves have veins (with xylem and phloem) which contain nitrate. Note the considerable difference 15 between the two root pools (the transfer pool R1 being N-depleted compared with the 15 accumulation pool R2) and between xylem and phloem nitrate (phloem nitrate being N- + enriched). In oil palm, xylem and root R1 are isotopically similar and lines overlap. The temporarily strong effect of waterlogging is easily visible with the transient 15N- enrichment in xylem nitrate, in particular in sunflower.

182

(a) Sunflower (b) Oil palm

17.3 15.9 31.0 Leaf 15.1 Leaf 19.5 0.9 18.9 7.2

2.9 16.6 0.8 15.9 14.8 27.4 3.0 25.9 + + Xylem+ Phloem+ Xylem Phloem 3.4 12.3 0.4 17.5 2.9 12.6 3.1 10.6

Root (R1) Root (R2) Root (R1) Root (R2) 0.8 18.9 0.8 31.4 0.8 27.9 0.8 38.9 0.4 12.9 1.3 21.9 0.9 20.0 0.8 25.9

Fig. S8. Summary of average δ15N values in plant compartments generated by modelling, in sunflower (a) and oil palm (b), showing the strong δ15N difference between xylem+ and phloem+. Colors are as in Fig. S1 (, low K; , high K; , low K + waterlogging; , high K + waterlogging).

Supplementary references

Andrews M. 1986. The partitioning of nitrate assimilation between root and shoot of higher plants. Plant Cell and Environment 9: 511-519. Eschbach J. 1982. Composante biochimique de la production du palmier à huile. Oléagineux 37: 159-168. Förster JC, Dieter Jeschke W. 1993. Effects of potassium withdrawal on nitrate transport and on the contribution of the root to nitrate reduction in the whole Plant. Journal of Plant Physiology 141: 322-328. Ho C-L, Kwan Y-Y, Choi M-C, Tee S-S, Ng W-H, Lim K-A, Lee Y-P, Ooi S-E, Lee W-W, Tee J-M. 2007. Analysis and functional annotation of expressed sequence tags (ESTs) from multiple tissues of oil palm (Elaeis guineensis Jacq.). BMC genomics 8: 381-391.

183

Appendix 2

What is the role of putrescine accumulated under potassium deficiency?

184 Received: 15 January 2020 Revised: 23 January 2020 Accepted: 25 January 2020 DOI: 10.1111/pce.13740

INVITED REVIEW

What is the role of putrescine accumulated under potassium deficiency?

Jing Cui1 | Igor Pottosin2 | Emmanuelle Lamade3 | Guillaume Tcherkez1

1Research School of Biology, ANU Joint College of Sciences, Australian National Abstract University, Canberra, Australian Capital Biomarker metabolites are of increasing interest in crops since they open avenues for Territory, Australia precision agriculture, whereby nutritional needs and stresses can be monitored opti- 2Biomedical Centre, University of Colima, + Colima, Mexico mally. Putrescine has the potential to be a useful biomarker to reveal potassium (K ) 3UPR34 Performance des systèmes de culture deficiency. In fact, although this diamine has also been observed to increase during des plantes pérennes, Département PERSYST, Centre de Coopération Internationale en other stresses such as drought, cold or heavy metals, respective changes are compa- Recherche Agronomique pour le rably low. Due to its multifaceted biochemical properties, several roles for putrescine Développement (CIRAD), Montpellier, France under K+ deficiency have been suggested, such as cation balance, antioxidant, reac- Correspondence tive oxygen species mediated signalling, osmolyte or pH regulator. However, the spe- Guillaume Tcherkez, Research School of + Biology, ANU Joint College of Sciences, cific association of putrescine build-up with low K availability in plants remains Australian National University, Canberra 2601, poorly understood, and possible regulatory roles must be consistent with putrescine ACT, Australia. Email: [email protected] concentration found in plant tissues. We hypothesize that the massive increase of putrescine upon K+ starvation plays an adaptive role. A distinction of putrescine function from that of other polyamines (spermine, spermidine) may be based either on its specificity or (which is probably more relevant under K+ deficiency) on a very high attainable concentration of putrescine, which far exceeds those for spermidine and spermine. putrescine and its catabolites appear to possess a strong potential in controlling cellular K+ and Ca2+, and mitochondria and chloroplasts bioenergetics under K+ stress.

KEYWORDS deficiency, ion balance, polyamines, potassium, putrescine

1 | INTRODUCTION Biomarker metabolites that are tractable using metabolomics are of potential importance in crop management, not only to follow devel- Putrescine, spermine and spermidine are dominant polyamine species, opmental stages, but also to monitor disease progression, nutritional naturally found in all organisms. It is now more than 65 years since needs or abiotic stresses (for a recent review, see [Alexandersson, putrescine was found to accumulate under K+ deficiency in plants Jacobson, Vivier, Weckwerth, & Andreasson, 2014]). Here, putrescine (Coleman & Hegartv, 1957; Coleman & Richards, 1956; Richards & is an interesting candidate to detect K+ deficiency situations, as Coleman, 1952). In fact, when K+ availability is low or very low in the suggested back in the 80s (Smith, 1984). Leaf metabolic biomarkers nutrient solution or in soil, putrescine accumulates in several parts of would be extremely useful to adjust cropping practices and in particu- the plant, particularly in leaves, to levels that can be up to 150 times lar, K+ fertilization. In effect, the simple measurement of K+ levels in higher than the normal content under K+-sufficient conditions. As leaves can be insufficient to characterize the ion status of crops and such, putrescine is one of the first metabolic biomarkers that has been thus to detect K+ deficiency. This is typically the case in oil palm discovered in the history of plant physiology. (Elaeis guineensis, a high K+-demanding species) where variations in

185 2 CUI ET AL. leaf potassium elemental content are relatively small even though K+ what enigmatic. There are several reasons to explain this limita- availability may vary widely. tion. First, several biochemical roles are in principle possible Putrescine is synthesized from ornithine, either via the direct (described below). Second, putrescine (as other polyamines) has route involving ornithine decarboxylase (ODC) or the indirect route been found to accumulate (although to a lower extent and less that involves arginine decarboxylase (ADC) (Figure 1; Slocum, systematically) under stress conditions other than K+ deficiency, 2005). Metabolomics studies on Arabidopsis have suggested that suggesting it is part of a more general stress response (Table 1). putrescine and ornithine are positively correlated with growth Third, putrescine accumulation is metabolically ‘expensive’ (Meyer et al., 2007) and ADC activity is essential for root growth because it requires ATP, redox power (NADPH) and assimilated (Watson, Emory, Piatak, & Malmberg, 1998). In tobacco, the ODC nitrogen (that might be limiting under K+ deficiency because of pathway is believed to be related to growth and proliferation, altered nitrate circulation). Using the direct route, the overall whereas the ADC pathway seems to be associated with morpho- equation gives: genesis and stress response (Masgrau et al., 1997). However, the ! putrescine biosynthetic pathway depends on the plant species. For 2 glutamate + ATP + NADPH putrescine + CO2 + 2OG + NADP + ADP + Pi example, Arabidopsis lacks ODC and thus synthesize putrescine from arginine only (Hanfrey, Sommer, Mayer, Burtin, & Michael, where 2OG stands for 2-oxoglutarate. The overall equation with the 2001). Most other species have both enzymes, with varying propor- indirect route is even more expensive in terms of consumed ATP, as tions of the biochemical route used. For example, in oil palm there is follows (assuming that fumarate is recycled via NAD-dependent a quantitative decrease in ornithine with no appearance of interme- malate dehydrogenase and that carbamoyl phosphate is synthesized diates (like arginine) when putrescine accumulates (Figure S1a), de novo): suggesting that the direct route is used. Similarly, in sunflower leaves, both ornithine and putrescine accumulate under K+ defi- 2 glutamate + 4 ATP + NADPH + NAD ! ðÞ: ciency (Figure S1b) and putrescine generally anticorrelates with putrescine + CO2 + 2OG + NADP + NADH + 4 ADP + Pi arginine (Figure S2), suggesting a competition between arginine and putrescine synthesis from ornithine and therefore, the direct route Considering such an energy requirement, the function of putres- of putrescine biosynthesis. In response to low K+ conditions, cine should be of considerable importance. In this short review, we Poaceae generally synthesize putrescine via ADC (see e.g., [Young & briefly describe possible roles of putrescine, and summarize data that Galston, 1984]). help defining most likely and specific roles of putrescine under K+ Despite its widespread accumulation (Table 1), the precise deficiency. role played by putrescine under K+ deficiency remains some-

(b)

(a)

FIGURE 1 Simplified metabolic pathway of putrescine synthesis and utilization. (a) Chemical structure of putrescine. Note that it contains two N atoms and four C atoms, that all come from glutamate. (b) Pathways showing the direct route starting from glutamate via ornithine (black), putrescine synthesis via arginine (grey) and other polyamines synthesis (blue). Cofactors and other compounds involved in reactions are shown in green or light turquoise. The alternative use of N-acetylornithine as an acetyl donor is shown in dashed green. The recycling of fumarate via the Krebs cycle and aspartate synthesis, and the recycling of ammonium by carbamoyl phosphate synthase are shown in dotted light turquoise. Abbreviations: 2OG, 2-oxoglutarate; ADC, arginine decarboxylase (chloroplastic); CP, carbamoyl phosphate; NAG, N-acetyl glutamate; NAGSA, N- acetyl glutamate semialdehyde; ODC, ornithine decarboxylase (cytosolic); P-NAG, phospho-N-acetyl glutamate; SAE, S-adenosyl methioninamine; SAM, S-adenosyl methionine; SMTA, S-methyl thioadenosine

186 PUTRESCINE ROLES UNDER K+ DEFICIENCY 3

TABLE 1 Summary and list of abiotic stress situations where putrescine quantity varies in plants

Tabulated summary

Stress Do other polyamine accumulate? Change in putrescine content K deficiency No in most cases ×3–×150 Osmotic shock Variable ×2–×14 Drought No, except in drought-tolerant plants? ≈×2 Salinity (NaCl) Yes in most cases Generally decreases Other stresses Generally yes if mineral nutrition also ×2–×10 impacted (heavy metals, N, etc.)

Full table (with references):

Do other polyamines Observed fold change in Stress Species and tissue accumulate? putrescine References K deficiency Various Unknown Unknown (based on Coleman and Richards (1956); Richards colorimetric assays at and Coleman (1952); Smith and that time) Richards (1962) Various No (except in radish) Up to 8 Basso and Smith (1974) Arabidopsis No 5 Watson and Malmberg (1996) Pea Yes (spermidine, Up to 27 (depending on Klein, Priebe, and Jäger (1979) + slightly) NH4 nutrition) Blackcurrant leaves Unknown Very high (undetectable Murty, Smith, and Bould (1971) at high K) Tobacco leaves Unknown Up to 11 (radioactivity Yoshida (1969) upon isotopic arginine feeding) Grapevine leaves No 5.5 Adams, Franke, and Christensen (1990) Lucerne Unknown Up to 150 Smith, Lauren, Cornforth, and Agnew (1982) Scots pine needles No Up to 100 Sarjala and Kaunisto (1993) Lemna species Unknown ≈10 or 100 (depends on Tachimoto, Fukutomi, Matsushiro, species) Kobayashi, and Takahashi (1992) Scots pine seedlings No change (roots) or Up to 9 Sarjala (1996) decline (needles) Tomato leaves No 5? Corey and Barker (1989) Poplar roots and leaves No (spermine declines) 25 (leaves), 80 (roots) Houman, Godbold, Majcherczyk, Shasheng, and Hüttermann (1991) Banana vitroplants Slightly (spermidine) Up to 30 Zaidan, Broetto, de Oliveira, Gallo, and Crocomo (1999) Barley leaves Slightly (agmatine) Up to 30 Sinclair (1969) Gentiana shoots Transiently 56 Takahashi, Imamura, Miyagi, and (spermidine) Uchimiya (2012) Tomato roots No 13 Sung et al. (2015) Oil palm leaves No 10 Cui, Davanture, Zivy, Lamade, and Tcherkez (2019) Sunflower leaves No 35 Cui, Abadie, Carroll, Lamade, and Tcherkez (2019) Rice cells No (decrease) 3 Sung, Liu, and Kao (1994) Sesame leaves Unknown (citrulline 9 Crocomo and Basso (1974) and ornithine also increase) (Continues)

187 4 CUI ET AL.

TABLE 1 (Continued)

Full table (with references):

Do other polyamines Observed fold change in Stress Species and tissue accumulate? putrescine References Osmotic shock Sorbitol Cereal leaves No 2 to 10 Flores and Galston (1982) Rice leaves Unknown 2.5 Chen and Kao (1993) Oat leaves No (slight decrease) Up to 4 (no change if Turner and Stewart (1988) turgor maintained) Arabidopsis leaves Yes (spermine) Up to 3 Feirer, Hocking, and Woods (1998) Wheat leaves Yes (spermidine) Up to 3 Erdei, Trivedi, Takeda, and Matsumoto (1990) Polyethylene Tobacco leaves Unknown 2 Kotakis et al. (2014) glycol Potato cultured cells No 14 (only insoluble Scaramagli, Biondi, Leone, Grillo, and conjugated putrescine) Torrigiani (2000) Various Oat leaves No Up to 5 Flores and Galston (1984) osmotica Mannitol Wheat Yes (cadaverine, Up to 3 (in leaves) Foster and Walters (1991) spermine) Drought/water deficit Barley leaves No (spermidine 2 Turner and Steward (1986) decreases) Rice leaves No 1.5 Capell, Bassie, and Christou (2004) Arabidopsis No Does not change Alcázar et al. (2010) Transient increase in 1.7 (transient increase) Alcázar et al. (2011) spermidine and then declines. Resurrection plant Yes Up to 3 Alcázar et al. (2011) Salt stress (NaCl) Soybean leaves Yes (spermine). Decreases Su and Bai (2008) Spermidine decreases. Olive tree roots Yes ≈1.3 Tattini, Heimler, Traversi, and Pieroni (1993) Soybean roots Yes Decreases Zhang, Xu, Hu, Mao, and Gong (2014) Arabidopsis flowers Yes (spermidine) Decreases Tassoni, Franceschetti, and Bagni (2008) Tomato leaves Yes (spermine). No Decreases Aziz, Martin-Tanguy, and Larher (1998) change in spermidine. Rice seedling roots Unknown Decreases Lin and Kao (2002) Sunflower xylem sap Yes (spermidine) Up to 2.5 Friedman, Levin, and Altman (1986) Arabidopsis Yes (both) Does not change Alet et al. (2012) Rice shoots Variable Up to 1.5 Katiyar and Dubey (1990) Various seedlings Yes Decreases Zapata, Ma, Pretel, Amorós, and Botella (2004) Sunflower seedlings No (decrease) Decreases Benavides, Aizencang, and Tomaro (1997) Arabidopsis Yes (spermidine). Decreases Bagni et al. (2006); Naka et al. (2010) Spermine decreases. (Continues)

188 PUTRESCINE ROLES UNDER K+ DEFICIENCY 5

TABLE 1 (Continued)

Full table (with references):

Do other polyamines Observed fold change in Stress Species and tissue accumulate? putrescine References Rice seedlings Yes Up to 3.5 Basu and Ghosh (1991)

Yes (spermidine). Very 2 Basu, Maitra, and Ghosh (1988) small change in spermine. Sunflower shoots Yes (spermine). Decreases or does not Mutlu and Bozcuk (2007) Spermidine change (depends on decreases. variety) Wheat leaves Yes Does not change Erdei et al. (1990) Arabidopsis Yes (spermine). No 2 Urano et al. (2004) change in spermidine. Various No (decrease) Decrease Priebe and Jäger (1978) Mung bean Yes (spermidine). Up to 4 (decrease in Friedman, Altman, and Levin (1989) Spermine content roots) not measured. Other stresses Magnesium Various No (except in radish) Up to 7.3 Basso and Smith (1974) deficiency Phosphate Rice cells No (decrease) ≈2 Shih and Kao (1996) deprivation (along with K+) Heavy metals: Aluminium Rice roots No (tend to decline) 3 Wang and Kao (2006) (Al3+) Cadmium Oat and bean leaves Spermine increases, Up to 10 Weinstein, Kaur-Sawhney, Rajam, (Cd2+) spermidine does not Wettlaufer, and Galston (1986) change Soybean nodules and roots Yes (spermine) 2.5 (nodules), 1.5 (roots) Balestrasse, Gallego, Benavides, and Tomaro (2005) Sunflower shoots Yes 2.7 Groppa, Ianuzzo, Tomaro, and Benavides (2007) Chromium Barley and rape seedlings No Up to 10 Hauschild (1993) (Cr3+,Cr6+) Copper (Cu2+) Rice leaves Unknown Up to 4 Lin and Kao (1999) Sunflower shoots Yes 1.6 Groppa et al. (2007) Anoxia/ Cereal seedlings Slightly (but numerical Up to 2 Reggiani, Giussani, and Bertani (1990) hypoxia/ data not reported) submergence Rice coleoptile Slightly 2 to 14 Reggiani, Hochkoeppler, and Bertani (1989); Reggiani, Zaina, and Bertani (1992) Scirpus shoots No (decrease) 6 Lee, Shieh, and Chou (1996) Cold Arabidopsis seedlings Spermidine stays Up to 5 Cuevas et al. (2008) constant, spermine decreases Diverse fruits Unknown, or decrease Up to 2.5 Escribano, Aguado, Reguera, and Merodio (1996); McDonald and Kushad (1986) Cucumber seedlings Yes (spermidine) Does not change Wang (1987) (Continues)

189 6 CUI ET AL.

TABLE 1 (Continued)

Full table (with references):

Do other polyamines Observed fold change in Stress Species and tissue accumulate? putrescine References Boron Tobacco leaves and roots Yes Up to 2 (leaves) and Camacho-Cristóbal, Maldonado, and deficiency 5 (roots) González-Fontes, 2005) Change from Tomato No ≈3 Feng and Barker, 1993) nitrate to Pepper and wheat leaves Yes (pepper), no Up to 20 Houdusse, Garnica, Zamarreño, Yvin, + NH4 (wheat) and García-Mina (2008) Mechanical Rapeseed leaves No 2 Cowley and Walters (2005) wounding Bananas No Up to 5 Yoza, Takeda, Sekiya, Nogata, and Ohta (1996) Low pH (< 5) Oat and pea leaves Unknown 2 to 8 Young and Galston (1983)

Note: When putrescine decreases or does not change rather than increase, it is mentioned in italics. When the reference cited also include mutants, data tabulated here only refer to wild-type plants.

2 | IS PUTRESCINE A VERSATILE spermidine appear in significant metabolites). In fact, the biosynthe- BIOMARKER OF K DEFICIENCY? sis of spermine and spermidine requires S-adenosyl methioninamine (SAE, Figure 1), which is produced from S-adenosyl methionine Putrescine accumulates under K+ deficiency up to the 1–10 mM (SAM) decarboxylation. SAM synthetase requires K+ as a cofactor range, with an increase up to by two orders of magnitude as com- (Takusagawa, Kamitori, & Markham, 1996) and therefore its activity pared to its level at optimal K+ (Table 1). For instance, in oil palm is probably very limited under K+ deficiency, thereby impacting not putrescine concentration is ca. 60 μM at high K+ and 1.8 mM at low only polyamines, but also all cellular reactions that use SAM as a K+ (i.e., ≈ 7 μmol g DW−1; Figure S1; Cui, Davanture, et al., 2019). methyl donor. Also, in plants, it is remarkable that putrescine is not Conversely, high (>10 mM) external K+ causes a decrease in putres- an effector of SAM decarboxylase activity (contrary to its mamma- cine content, which is converted to ‘higher polyamines’ (this term lian counterpart; Bennett, Ekstrom, Pegg, & Ealick, 2002) thereby refers to higher molecular weight polyamines synthesized from allowing putrescine accumulation without stimulation of SAE (and putrescine, such as spermine and spermidine; Aurisano, Bertani, thus spermine and spermidine) production. Phosphate deprivation Mattana, & Reggiani, 1993; Reggiani, Aurisano, Mattana, & Bertani, has also been reported to trigger putrescine build-up (Knobloch & 1993) and/or putrescine extrusion (Tamai, Shimada, Sugimoto, Berlin, 1981; Shih & Kao, 1996). However, in (Shih & Kao, 1996) Shiraishi, & Oji, 2000). Thus, putrescine metabolism is sensitive to phosphate abstraction from the medium seems to have been done external K+, but the underlying mechanism is still unknown. It might by withdrawing potassium phosphate from the nutrient solution, be speculated that the increase in putrescine content at low K+ is cau- meaning that the build-up of putrescine was in fact coupled to K+ + sed by the stimulation of ammonium assimilation (see high NH4 con- deficiency. ditions in Table 1), which has indeed been observed in Arabidopsis (Armengaud et al., 2009). Regardless of the underlying metabolic cause for its accumulation, putrescine seems to be a good low-K+ bio- 3 | ONE MOLECULE, TOO MANY ROLES? marker in the bio-statistical sense since its increase is highly significant (order of magnitude of the p-value far below that of many other Plant polyamines have been studied for a long time and quite under- metabolites changed by low K+) and it has a very high weight (loading) standably, the literature on polyamines in plant physiology is now con- in multivariate analyses. Therefore, it might be used as an index for K+ siderable. Taken as a whole, polyamines are believed to be of availability (Cui, Abadie, et al., 2019; Cui, Davanture, et al., 2019). importance under stressful conditions and to play a signalling role dur- Nevertheless, putrescine also may accumulate under other condi- ing plant development (Alcázar et al., 2006; Galston & Sawhney, tions, such as low pH, anoxia, heavy metals, low Mg2+, cold or osmotic 1990; Tiburcio, Altabella, Bitrián, & Alcázar, 2014). Historically, stress. In half of cases, putrescine has been found to decrease under putrescine has been suggested to play a role of (a) a cation to substi- salt stress (Table 1) and to confer no specific advantage for NaCl toler- tute K+, (b) an antioxidant and/or a ROS-mediated signal (via oxida- ance when applied exogenously (Ndayiragije & Lutts, 2006). Poly- tion), (c) an osmolyte under salt or osmotic stress, (d) a root-shoot amines other than putrescine (spermine, spermidine) may also transport molecule (either as a nitrogen-containing metabolite or a accumulate under K+ deficiency although not to the same extent and cation), and (e) a cryoprotectant at low temperature. However, ion- can even decrease (for an example in Arabidopsis, see [Watson & omics analyses have shown that when compared to other cations, Malmberg, 1996]; see also Figure S2 where neither spermine nor putrescine represents a small pool (<5%) of positive charges

190 PUTRESCINE ROLES UNDER K+ DEFICIENCY 7

(a) (b)

FIGURE 2 Leaf cation balance under normal or low potassium availability in oil palm (a) and sunflower (b). In each panel, the inset shows the sum of cations, also in μmol positive charges g−1 DW (dry weight). Abbreviations: Min, other minor cations (Zn2+,Cu2+,Mn2+ and H+ calculated assuming a pH value of 7); Put, putrescine (carrying two positive charges); Orn, ornithine (carrying one positive charge). From source data in Cui, Abadie, et al. (2019); Cui, Davanture, et al. (2019). Asterisks stand for a significant K-availability effect (in sunflower, there is a significant increase in putrescine although it remains very small in terms of positive charge load)

(Figure 2), so its role in charge balance is minor and the same is true sunflower (Cui, Abadie, et al., 2019; Cui, Davanture, et al., 2019). for its role in osmoprotection. The role of antioxidant, although widely Under K+ deficiency, there is also an increase in the difference supported experimentally, seems to depend on concentration and between Ca2+ and the sum Mg2+ +K+ (of about 0.4 mmol positive conditions, since there are examples where polyamine addition may charges g−1 DW in Figure 2). In general, there is a well-supported neg- trigger oxidative stress (Mohapatra, Minocha, Long, & Minocha, 2009) ative relationship between K+ and Ca2+, which has been documented and polyamine catabolism is indeed an important source of hydrogen for nearly 50 years in herbaceous crops (such as sunflower, rapeseed, peroxide and other ROS species, especially under stress conditions tobacco, or wheat). This is here examplified in oil palm, cultivated (Moschou, Paschalidis, & Roubelakis-Angelakis, 2008; Pottosin, under varying K+ fertilization (Figure S3). Similarly, in other species Velarde-Buendía, Bose, Zepeda-Jazo, et al., 2014; Wang et al., 2019). such as castor bean, K+ deficiency causes an increase in Ca2+ and Mg2+, In the next sections, we focus on roles of putrescine, as compared to and a slight decrease in Na+ in leaf lamina, but conversely a consider- higher polyamines, in the regulation of K+ acquisition and re-distribu- able increase in Na+ with little change in Ca2+ and Mg2+ in petioles and tion, Ca2+ signalling, and chloroplast and mitochondrion functions. phloem sap, leading to an excess of positive charges (Peuke, Jeschke, & Hartung, 2002). In grape, low K+ is compensated for by Ca2+ and Mg2+ in leaves and by Na+ in fruits also suggesting that phloem sap carries 4 | PUTRESCINE AND REGULATION OF more Na+ (Ruhl, 1989). While these effects reflect the antagonism CATION TRANSPORT AND BALANCE between K+,Na+ and Mg2+ absorption and exchange (Diem & Godbold, 1993; Jakobsen, 1993), they also show that K+ deficiency is associated + 4.1 | Consequences of K deficiency for ion with more positive charges in the phloem, and thus that putrescine is composition unlikely to play the role as a cation to substitute K+ in sap. However, when K+ deficiency is compensated for by K+-substitution with Na+ or K+ deficiency is not associated with a general decrease, but actually Rb+, putrescine accumulates less, suggesting that there is a link with leads to a significant increase in cation load (Figure 2, insets). That is, cations (Richards & Coleman, 1952; Smith, 1984). Quite remarkably, if quite counter-intuitively, K+ deficiency implies an extra demand in K+-deficiency is accompanied by low Ca2+ provision, putrescine accu- negative charges to reach electro-neutrality, which is met by accumu- mulation is also lower (Coleman & Richards, 1956; Richards & Coleman, lated organic and amino acids (Armengaud et al., 2009). The excess of 1952). These observations suggest that putrescine is mostly a response positive charges mostly comes from the considerable increase in Ca2+ to a disequilibrium in cation composition, in which Ca2+ would be over- (up to twofold increase) and Mg2+ (more than twofold) in oil palm and represented. Mg2+ deficiency also leads to a modest putrescine

191 8 CUI ET AL. accumulation (Table 1), probably because it changes the cation balance either putrescine, spermine or spermidine (Janicka-Russak, KabaŁa, in favour of Ca2+, but to a lower extent than K+ deficiency (due to the MŁodzinska, & KŁobus, 2010). In that case, the inhibition was caused naturally lower Mg2+ content compared to K+; for example, see by a decrease in the expression for an H+-ATPase isoform and not by a Figure 2). direct (physical) interaction affecting ATPase catalysis. In rice coleoptiles, direct stimulation of plasma membrane H+-pumps by all polyamines at millimolar (mM) concentration has been reported, while only + 4.2 | Regulation of H -ATPases by putrescine putrescine may reach such a concentration in physiological situations (Reggiani et al., 1992). In maize roots, plasma membrane H+-pumping is Rather than acting as a charge-balancing cation, putrescine appears to rapidly stimulated by putrescine (in the elongation zone) and depressed regulate the cation balance (summarized in Figure 3). Lowering exter- by spermine (in the maturation zone) (Pandolfi, Pottosin, Cuin, Man- nal K+ concentration causes a rapid (within minutes) membrane hyper- cuso, & Shabala, 2010). Similarly, spermine at high concentrations polarization, which stimulates K+ uptake via inward-rectifying AKT1 suppresses, whereas putrescine has no direct effect, on H+-pumping channels (Chérel, Lefoulon, Boeglin, & Sentenac, 2013; Wang & Wu, in plasma membrane vesicles isolated from pea roots (Pottosin, 2013). When K+ starvation lasts, however, membrane depolarization Velarde-Buendía, Bose, Fuglsang, et al., 2014). This contrasted effect may occur, which correlates with a marked decrease in cytosolic K+ of putrescine and other polyamines on H+-ATPases could originate concentration (Armengaud et al., 2009). To drive K+ uptake, the activ- from difference in competing with Mg2+ for ATP-binding and/or ity of root K+/H+ symporter (mainly via HAK5) energized by plasma ATPase phosphorylation. In fact, putrescine does not bind to ATP, membrane H+-ATPase is critical (Wang & Wu, 2013). Potassium ions but spermine does (Igarashi et al., 1989) while Mg-ATP (and not free uncouple ATP hydrolysis from the H+ extrusion by plasma membrane ATP) acts as a substrate for H+-ATPases. In intact roots, both poly- H+-ATPase (Buch-Pedersen, Rudashevskaya, Berner, Venema, & amines induced Ca2+-pumping, which in turn stimulated H+- Palmgren, 2006). Thus, at low cytosolic K+, ATP/H+ coupling is proba- pumping, most likely via a decrease of H+-ATPase protein phosphor- bly better and H+ extrusion is stimulated, thereby favouring ion ylation by a Ca2+-dependent kinase (see Pottosin, Velarde-Buendía, uptake (Chérel et al., 2013; Wang & Wu, 2013). Then do polyamines Bose, Fuglsang, et al., 2014; and references therein). Thus, putres- and putrescine in particular, influence plasma membrane H+-ATPase? cine stimulates H+-pumping whereas spermine stimulates ATPase at The answer to this question appears to be species- and tissue-depen- low concentration and suppresses H+-pumping at high concentra- dent. Suppression of both plasma membrane and vacuolar H+-ATPase tion. Taken as a whole, putrescine seems to favour H+-pumping activity was observed in cucumber roots pretreated for 24 hr with across the plasma membrane unlike higher polyamines (spermine).

FIGURE 3 Summary of possible roles of putrescine on cellular cation balance under K+ deficiency. Two main roles are highlighted here, via ions channels (orange, left) and H+-ATPases (green, right). See main text for further details. Abbreviations: DAO, diamine oxidase; GABA, γ-aminobutyrate; ROS, reactive oxygen species

192 PUTRESCINE ROLES UNDER K+ DEFICIENCY 9

+ 4.3 | Putrescine, ROS and K transport 1996). In the initial phase, the vacuole will indeed compensate for the decrease in cytosolic K+ by K+-release via selective (TPK) and non- Externally applied polyamines at relatively high (0.5–1mM)concentra- selective monovalent cation FV channels, both marginally sensitive to tion inhibit both inward and outward rectifying K+-selective currents in putrescine at the sub-millimolar range (Brüggemann, Pottosin, & roots (Pottosin, 2015; Zhao, Song, He, & Zhu, 2007), whereas internal Schönknecht, 1998; Dobrovinskaya, Muniz, & Pottosin, 1999; polyamines at 1 mM halved the current mediated by KAT1 in guard Hamamoto et al., 2008). Under very strong K+ deprivation, the elec- cells (Liu, Fu, Bei, & Luan, 2000). It is not very likely, therefore, that trochemical gradient for K+ becomes vacuole-directed (Walker et al., these effects have a huge significance for K+ absorption and retention. 1996). Thus, to minimize passive vacuolar K+ re-uptake, it is certainly On the other hand, a combination of polyamines with oxidative stress crucial to reduce K+-transport by K+-permeable channels. When induces a substantial K+ loss from roots. ROS are produced via the oxi- putrescine reaches millimolar concentration, K+ transport not only dation of putrescine and other polyamines by intrinsic apoplast diamine via SV channels, but also via FV channels will be suppressed and polyamine oxidases (DAO and PAO, respectively) (DiTomaso, (Brüggemann et al., 1998; Dobrovinskaya, Muniz, & Pottosin, 1999). Shaff, & Kochian, 1989; Zepeda-Jazo et al., 2011; Velarde-Buendía, Shabala, Cvikrova, Dobrovinskaya, & Pottosin, 2012; Pottosin, Velarde- Buendía, Bose, Zepeda-Jazo, et al., 2014). The occurrence of DAO and 5 | ROLES OF PUTRESCINE IN PAO is variable, with DAO being more abundant in Dicots and PAO in CHLOROPLASTS Monocots like Poaceae (Moschou et al., 2008). The loss of K+,especially in specialized zones like the root apex, is not necessarily harmful Possible roles of putrescine on chloroplast metabolism are summa- despite oxidative stress. Instead, low intracellular K+ may be sensed and rized in Figure 4. Subcellular fractionation followed by metabolomics induces a metabolic switch to defence responses (Shabala, 2017). analysis has shown that about 40% of cellular putrescine is present in Another product of putrescine catabolism, GABA, has recently been chloroplasts in Arabidopsis leaves (Krueger et al., 2011), perhaps shown to improve K+ retention in Arabidopsis roots by a stimulation of reflecting the activity of chloroplastic ADC (Borrell et al., 1995; plasma membrane H+-ATPase activity, a decrease of stress-induced Bortolotti et al., 2004). Stress-induced stimulation of ADC (Alcázar ROS production and a decrease in the expression of outward-rectifying et al., 2010) might further increase putrescine accumulation in chloro- K+ channel, GORK (Su et al., 2019). plasts. In chloroplasts, polyamines are believed to regulate different aspects of photosynthesis, with reported differences in action between putrescine and other polyamines. Exogenous putrescine 2+ 4.4 | Putrescine and Ca homeostasis decreases non-photochemical quenching (NPQ) and increases photochemical yield (Ioannidis, Sfichi, & Kotzabasis, 2006). Yet, these results Overall,thecationloadaswellastotalCa2+ increase under K+ deficiency have been obtained under non-physiological conditions, with a low- (e.g., Figure 2 and Figure S3). Free cytosolic Ca2+ may be kept low by salt medium, to minimize the interference with other cations (such as (a) efficient Ca2+ extrusion while as mentioned above, there is a stimula- Mg2+) and therefore, are perhaps not so informative. On the other tion of plasma membrane Ca2+ pumps by polyamines; and (b) vacuolar hand, with more physiological saline buffers, all polyamines stimulate Ca2+ sequestration. The latter is especially important, bearing in mind the photophosphorylation at low concentrations, whereas spermidine and observed increase in total Ca2+. In fact, in plant cells, total cellular Ca2+ spermine but not putrescine act as strong uncouplers at high concen- mostly reflects vacuolar Ca2+.Ca2+ accumulates in vacuoles via CAX- tration (>1 mM for spermidine and >0.1 mM for spermine). That is, mediated H+/Ca2+ antiport, fuelled by the trans-tonoplast H+ gradient. only putrescine induces a relatively high and stable stimulation of ATP To ensure efficient vacuolar Ca2+ retention, channel-mediated Ca2+ loss production in chloroplasts (Ioannidis & Kotzabasis, 2007). from the vacuole to the cytosol must be negligible. SV/TPC1 channels Putrescine is a weak base (pKa 10.8) thus its uncharged form are the major routes of vacuolar Ca2+ release (Pottosin & Schönknecht, coexist, albeit at a relatively small fraction (0.04%), with the charged 2007). Consequently, relative expression of TPC1 and CAX is crucial for species at pH 7.4. Light induces stromal alkalization and thylakoid vacuolar Ca2+ accumulation (Gilliham, Athman, Tyerman, & Conn, 2011). lumen acidification and this proton gradient can be damped by trans- Importantly, ionic currents via SV channels are efficiently suppressed by port of uncharged putrescine across the thylakoid membrane. This polyamines in their physiological range of concentrations. Albeit this does not affect the electrical potential difference across the thylakoid effect is charge-dependent, with putrescine having the lowest affinity membrane (ΔΨ) but dissipates ΔpH and reduces lumen acidification, (Dobrovinskaya, Muñiz, & Pottosin, 1999), it could be compensated for optimizing photosynthesis under stress conditions where high ΔpH by a very high putrescine concentration under K+ deprivation. values lead to NPQ (Ioannidis, Cruz, Kotzabasis, & Kramer, 2012). Under K+ deficiency, the decrease in K+ can be compensated for by an increase in Mg2+ (Figure 2). Mg2+ is a charge-balancing cation that + 4.5 | Putrescine and vacuole-cytosol K balance can dissipate ΔΨ and facilitate ΔpH built-up across the thylakoid membrane via Mg2+-permeable channels that are present in thylakoid Under K+ deficiency, maintenance of relatively high cytosolic K+ is membranes (Pottosin & Schönknecht, 1996). Thus, putrescine can achieved at the expense of the vacuolar K+ (Walker, Leigh, & Miller, have a role of Mg2+ antagonist, whereby it prevents excessive energy

193 10 CUI ET AL.

FIGURE 4 Summary of possible roles of putrescine on organelles under K+ deficiency. Putrescine has a general positive effect on ATP synthesis in both mitochondria and chloroplasts via a number of mechanisms, including mitigation of mitochondrial permeability transition (MPT) and non-photochemical quenching (NPQ), respectively. Abbreviations: NDHs, NAD(P)H dehydrogenases; TCAP, tricarboxylic acid pathway

dissipation and decreased photosynthesis, which may be due to the 6.1 | Putrescine and mitochondrial metabolism excessive lumen acidification even at relatively low light (see [Davis, Rutherford, & Kramer, 2017], for further details). It has also been Under stress conditions, putrescine causes a stimulation of the tricar- demonstrated that putrescine up-regulates the expression of ATP- boxylic acid pathway (TCAP) and thus facilitates mitochondrial ATP synthase and exerts a general protective effect on the photosynthetic production (Zhong et al., 2016). So far, this effect has been demon- membrane and in particular PSII structure (Shu et al., 2015). strated for salt stress, when putrescine was supplied exogenously. This still needs to be tested under K+ deficiency, based on large amounts of putrescine accumulated naturally. However, metabolomics + 6 | ROLES OF PUTRESCINE IN analyses have suggested that the increased CO2 release under K MITOCHONDRIA deficiency is not associated with a higher ATP production but rather reflects lower efficiency of the TCAP when K+ is limiting enzymatic Putrescine is synthesized outside mitochondria but can be taken up activity (Cui, Abadie, et al., 2019). Also, it should be noted that mito- by them. It is likely exchanged between the cytosol and the mitochon- chondrial carbonic anhydrase, which might play an important role in drial matrix via a basic amino acid transporter which is able to carry anaplerosis (conversion of catabolic CO2 into bicarbonate), is inhibited arginine, citrulline and ornithine (Hoyos et al., 2003; Palmieri et al., with a high affinity (low Ki) by spermine and spermidine, while putres- 2006). In animal cells, mitochondrial putrescine uptake has a low affin- cine has no effect (Carta et al., 2010). ity (K0.5 ≈ 1–4 mM) but a high capacity driven by electrical gradient, Interestingly, tobacco mitochondrial complex I mutants, which that is, the high negative potential of the mitochondrial matrix (Dalla have a slow growth phenotype, show a significant increase in putres- Via, Di Noto, & Toninello, 1999; Toninello, Dalla Via, Siliprandi, & cine, along with related compounds such as GABA (Lothier, De Garlid, 1992). Similarly, in plants, polyamine accumulation in mito- Paepe, & Tcherkez, 2019). At physiologically attainable K+, higher chondria depends on membrane potential, but its regulation differs polyamines inhibit mitochondrial membrane-bound F0F1-ATPase in somewhat from that in animals (Pistocchi, Antognoni, Bagni, & Vigna (Peter, Pinheiro, & Lima, 1981), which may be partly caused by Zannoni, 1990) and associated molecular mechanisms remain the fact that higher polyamines (but not putrescine) are able to dis- unknown (Fujita & Shinozaki, 2015). Polyamines have diverse effects place Mg2+ from Mg-ATP complexes (Igarashi et al., 1989). That is, + + in mitochondria, typically on metabolism, electron transport and the putrescine can activate mitochondrial F0F1-ATPases even at low K /Na permeability transition (summarized in Figure 4). (in contrast to spermine and spermidine, the action of which decreases

194 PUTRESCINE ROLES UNDER K+ DEFICIENCY 11 at low K+/Na+) (Peter et al., 1981) thereby allowing ATP production In fact, MPT is stimulated by the increase in Ca2+ via ROS genera- despite low K+ concentration encountered under potassium deficiency. tion while polyamines have been found to mitigate ROS generation In addition, enzymatic transglutaminase covalent binding of putrescine and inhibit MPT in both plants and animals (Arpagaus et al., 2002; to mitochondrial membrane proteins is associated with higher F0F1- Tabor, 1960; Toninello, Salvi, & Mondov, 2004). Unlike spermine, ATPase activity and tolerance to osmotic stress (Liu & Zhang, 2004; putrescine has been shown to be inefficient on cytochrome c release Votyakova, Wallace, Dunbar, & Wilson, 1999). Putrescine, albeit with a at up to 1 mM in mitochondria isolated from rat heart (Stefanelli et al., 100 times lower affinity compared to higher polyamines (yet with 2000). The intermediate of putrescine synthesis, agmatine (Figure 1), 2+ K0.5 = 0.3 mM), stimulates the activity of the mitochondrial membrane inhibits Ca -mediated MPT in Mammals (Battaglia et al., 2010). Con- ATP/ADP exchanger (Krämer, Mayr, Heberger, & Tsompanidou, 1986). versely, in yeast, spermine stimulates Ca2+ uptake by mitochondria, This activation may become significant under K+ deficiency, when thereby favouring MPT (Votyakova, Bazhenova, & putrescine reaches millimolar levels. Zvjagilskaya, 1993). Polyamines at a physiological concentration (0.1 mM) lead to a reduction of ΔΨ by 30 and 50%, with putrescine and spermine, 6.2 | Putrescine and mitochondrial membrane respectively; this differential effect of putrescine and spermine has permeability been found to correlate with substrate preference of mitochondrial amine oxidase (Maccarrone et al., 2001) but whether this effect is Polyamines can have an impact on mitochondrial transmembrane effectively mediated by amine oxidase is not known. In plant mito- potential (ΔΨ), perhaps mediated by their effect on mitochondrial chondria under low cytosolic cation load (low K+), putrescine slightly + mito ATP-sensitive K channels ( KATP). Both the molecular identity of stimulates external NAD(P)H dehydrogenases while at high cation mito KATP and their structural similarity with plasma membrane KATP load, it has little effect; this is in contrast with spermidine and spe- channels (which are abundant in animal tissues but absent in plants) rmine, which stimulate NAD(P)H dehydrogenases activity consider- are still a matter of debate (Szabo & Zoratti, 2014; Trono, Laus, ably at low cation load (and inhibit dehydrogenases activity at high Soccio, Alfarano, & Pastore, 2015). Under the assumption that cation load; Phelps & McDonald, 1990; Rugolo, Antognoni, Flamigni, & mito + + KATP are structurally similar to K inward rectifiers (as animal Zannoni, 1991; Sjölin & Møller, 1991). Therefore, when K concentra- + plasma membrane KATP channels are), the K current through the tion is low, spermine and spermidine tend to increase the electron channel pore would be modulated in a voltage-dependent manner by pressure on the mETC and promotes ROS generation, while this effect cytosolic polyamines. In Mammals, spermine, spermidine and putres- does not take place with putrescine. cine can regulate the K+ efflux upon depolarization (Aguilar-Bryan & Surprisingly, although polyamines can inhibit MPT at relatively 2+ Bryan, 1999). Unlike their animal counterparts, plant KATP are not sen- high concentration, they may also favour Ca accumulation in the sitive to Mg2+ (Pastore, Stoppelli, Di Fonzo, & Passarella, 1999) but to mitochondrial matrix, which normally acts as a MPT inducer (reviewed our knowledge, the effect of polyamines has not been documented in [Toninello et al., 2004]). Thus, under K+ deficiency, high putrescine yet. Mitochondrial depolarization by K+ influx is believed to reduce concentration with higher Ca2+ load (MPT promoter) and high Mg2+ ROS production in plants under stress (Trono et al., 2015) and, vice (MPT opposer) may either stimulate or down-regulate MPT, versa, hyperpolarization is associated with excessive electron pressure depending on whether the change in mitochondrial Ca2+ predomi- in the mitochondrial electron transfer chain (mETC) and higher ROS nates over Mg2+ change, ROS limitation and electron pressure mitiga- production. For example, under osmotic stress, a ROS-mediated acti- tion. Alternatively, one might speculate that a brief MPT event may + 2+ vation of KATP has been found in wheat (Trono et al., 2015). Thus, have a protective role, releasing excess ROS and Ca from the matrix mito activation of plant KATP could in principle be efficient to regulate and restoring normal mitochondrial ATP production. However, the mitochondrial activity, since it not only decreases ΔΨ, but also release of ROS and Ca2+ may become self-propagative, causing Ca2 impedes ROS generation. +-induced Ca2+ release and ROS-induced ROS release (Zorov, The effect of polyamines and in particular putrescine on mito- Juhaszova, & Sollott, 2014) and ultimately cell death. It is thus more chondria can also be linked to the control of mitochondrial permeabil- likely that putrescine accumulation under K+ deficiency is beneficial ity transition (MPT), which is a massive increase in permeability of the due to its combination of physiological effects, that is, simultaneous inner mitochondrial membrane, with a collapse of ΔΨ and release of limitation of Ca2+ release in the cytosol (see Section 4.4) and down- pro-apoptotic factors (cytochrome c). In effect, MPT with properties regulation of MPT. similar to those found in animal MPT, such as activation by Ca2+ over- load and ROS, and inhibition by Mg2+ and low pH, has been reported in plants and shown to promote programmed cell death (Arpagaus, 7 | SIDE EFFECTS OF PUTRESCINE Rawyler, & Braendle, 2002; Fortes, Castilho, Catisti, Carnieri, & Ver- cesi, 2001; Lin, Wang, & Wang, 2005; Scott & Logan, 2008; Tiwari, The beneficial effects of putrescine in particular on cation balance Belenghi, & Levine, 2002). Potentially, polyamines can have an action (see above) probably explain why the addition of exogenous putres- on MPT via electron pressure on mETC, Ca2+ concentration, and ROS cine or the production of endogenous putrescine in transgenics has generation. often been described as being advantageous to improve stress

195 12 CUI ET AL. tolerance and mitigate oxidative stress (Ndayiragije & Lutts, 2006; positive effects of ADC overexpression (increase in tolerance to Öztürk & Demir, 2003; Verma & Mishra, 2005). However, over- drought, cold, or salinity in Arabidopsis, or rice [Alcázar et al., 2010; expression of ADC2 in Arabidopsis induces dwarfism and late Wang, Zhang, Liu, & Li, 2011]), toxic effects of putrescine over- flowering (Alcázar, García-Martínez, Cuevas, Tiburcio, & Altabella, production have been observed (see above). It is possible that delete- 2005). Also, overexpression of oat ADC in tobacco leads to short rious effects were caused by enhanced DAO activity and excessive internodes, thin stems and leaves, leaf chlorosis and necrosis, and ROS production. Therefore, one might hypothesize that engineering reduces root growth (Masgrau et al., 1997), which mimics to some plants with simultaneous overexpression of ADC and knock-down of extent the symptoms of some stresses like K+ deficiency or osmotic DAO could be beneficial. In the field, the putrescine content in crops + shock. Conversely, inhibiting putrescine synthesis using D-arginine could be used as a component of the metabolomics signature of K under phosphorus deficiency appears to be beneficial for total bio- nutrition or a marker to detect K+-responsive varieties, because it mass in cultured rice cells (Shih & Kao, 1996). It should be recognized reflects several processes (described above) triggered by intracellular that adding putrescine or boosting putrescine synthesis changes nitro- K+ scarcity. In the near future, it might then be amongst biomarkers gen metabolism and promotes putrescine recycling. In fact, putrescine used by precision agriculture. is believed to be easily recycled via diamine oxidase to GABA (Shelp et al., 2012) and importantly, putrescine oxidation can be a source of ACKNOWLEDGMENTS ROS (see above), signalling a stress response and leading to changes G.T. thanks the financial support of the Région Pays de la Loire and in gene expression (Gupta, Sengupta, Chakraborty, & Gupta, 2016; Angers Loire Métropole via the Connect Talent grant Isoseed. J.C. was Minocha, Majumdar, & Minocha, 2014). Putrescine can thus be occa- supported by an Australia Awards PhD Scholarship. sionally detrimental in terms of oxidative stress or net photosynthesis (Mohapatra et al., 2009; Pál et al., 2018). Whenever the pro-oxidant ORCID effect predominates over the anti-oxidant function of putrescine, the Guillaume Tcherkez https://orcid.org/0000-0002-3339-956X suppression of arginine formation and ADC activity (along with a decrease in putrescine and concomitant decrease of ROS production) REFERENCES may be beneficiary for plant performance under stress (e.g., the Adams, D. O., Franke, K. E., & Christensen, L. P. (1990). Elevated putrescine levels in grapevine leaves that display symptoms of potassium decrease in putrescine synthesis by metasilicic acid (H2SiO3) applica- deficiency. American Journal of Enology and Viticulture, 41, 121–125. tion can alleviate some effects of K+ deficiency (Chen et al., 2016)). Aguilar-Bryan, L., & Bryan, J. (1999). Molecular biology of adenosine + However, such a situation nevertheless seems unlikely under K defi- triphosphate-sensitive potassium channels. Endocrine Reviews, 20, ciency since putrescine accumulates to very high levels, certainly 101–135. reflecting an adaptive trait of plant metabolism. Alcázar, R., Bitrián, M., Bartels, D., Koncz, C., Altabella, T., & Tiburcio, A. F. (2011). Polyamine metabolic canalization in response to drought stress in Arabidopsis and the resurrection plant Craterostigma plantagineum. Plant Signaling and Behavior, 6, 243–250. 8 | CONCLUSIONS AND PERSPECTIVES Alcázar, R., García-Martínez, J. L., Cuevas, J. C., Tiburcio, A. F., & Altabella, T. (2005). Overexpression of ADC2 in Arabidopsis induces Putrescine has specific biochemical properties that differ from other dwarfism and late-flowering through GA deficiency. The Plant Journal, 43, 425–436. polyamines and this probably explains why K+ deficiency appears to Alcázar, R., Marco, F., Cuevas, J. C., Patron, M., Ferrando, A., Carrasco, P., be closely associated with putrescine rather than spermine or … Altabella, T. (2006). Involvement of polyamines in plant response to spermidine. Putrescine accumulation under K+ deficiency is perhaps abiotic stress. Biotechnology Letters, 28, 1867–1876. … advantageous via its concerted action on several cellular processes Alcázar, R., Planas, J., Saxena, T., Zarza, X., Bortolotti, C., Cuevas, J., Altabella, T. (2010). Putrescine accumulation confers drought toler- including cation balance, ultimately down-regulating MPT. To better ance in transgenic Arabidopsis plants over-expressing the homologous understand stress responses where putrescine is involved, a differ- arginine decarboxylase 2 gene. Plant Physiology and Biochemistry, 48, ence should be made between endogenous, natural putrescine pro- 547–552. duction under K+ deficiency and artificial putrescine provision. To Alet, A. I., Sánchez, D. H., Cuevas, J. C., Marina, M., Carrasco, P., Altabella, T., … Ruiz, O. A. (2012). New insights into the role of spe- definitely appreciate the adaptive role of putrescine under K+ defi- rmine in Arabidopsis thaliana under long-term salt stress. Plant Science, ciency, it will be necessary to use plant lines with altered putrescine 182,94–100. content such as ADC overexpression or knock-out lines and at the Alexandersson, E., Jacobson, D., Vivier, M. A., Weckwerth, W., & same time, verify putrescine subcellular distribution, measure both K+ Andreasson, E. (2014). Field-omics—Understanding large-scale molecular data from field crops. Frontiers in Plant Science, 5, 286. and Ca2+ content, and monitor mitochondrial activity (ATP synthesis, Armengaud, P., Sulpice, R., Miller, A. J., Stitt, M., Amtmann, A., & Gibon, Y. transmembrane potential and ROS production). Also, a possible venue (2009). Multilevel analysis of primary metabolism provides new would be to examine further the roles of putrescine in chloroplasts (its insights into the role of potassium nutrition for glycolysis and nitrogen major site of production via the ADC pathway) and in particular, to assimilation in Arabidopsis roots. Plant Physiology, 150, 772–785. Arpagaus, S., Rawyler, A., & Braendle, R. (2002). Occurrence and charac- check its effect on ion and pH homeostasis, electrochemical gradient teristics of the mitochondrial permeability transition in plants. Journal across the thylakoid membrane and ultimately optimization of photo- of Biological Chemistry, 277, 1780–1787. synthesis. It should be kept in mind that aside from examples of

196 PUTRESCINE ROLES UNDER K+ DEFICIENCY 13

Aurisano, N., Bertani, A., Mattana, M., & Reggiani, R. (1993). Abscisic acid Coleman, R., & Hegartv, M. (1957). Metabolism of DL-ornithine-2-14Cin induced stress-like polyamine pattern in wheat seedlings, and its normal and potassium-deficient barley. Nature, 179, 376–377. reversal by potassium ions. Physiologia Plantarum, 89, 687–692. Coleman, R., & Richards, F. (1956). Physiological studies in plant nutrition: Aziz, A., Martin-Tanguy, J., & Larher, F. (1998). Stress-induced changes in XVIII. Some aspects of nitrogen metabolism in barley and other plants polyamine and tyramine levels can regulate proline accumulation in in relation to potassium deficiency. Annals of Botany, 20, 393–409. tomato leaf discs treated with sodium chloride. Physiologia Plantarum, Corey, K., & Barker, A. (1989). Ethylene evolution and polyamine accumu- 104, 195–202. lation by tomato subjected to interactive stresses of ammonium toxic- Bagni, N., Ruiz-Carrasco, K., Franceschetti, M., Fornalè, S., ity and potassium deficiency. Journal of the American Society for Fornasiero, R. B., & Tassoni, A. (2006). Polyamine metabolism and bio- Horticultural Science, 114, 651–655. synthetic gene expression in Arabidopsis thaliana under salt stress. Cowley, T., & Walters, D. R. (2005). Local and systemic changes in arginine Plant Physiology and Biochemistry, 44, 776–786. decarboxylase activity, putrescine levels and putrescine catabolism in Balestrasse, K. B., Gallego, S. M., Benavides, M. P., & Tomaro, M. L. (2005). wounded oilseed rape. New Phytologist, 165, 807–811. Polyamines and proline are affected by cadmium stress in nodules and Crocomo, O., & Basso, L. (1974). Accumulation of putrescine and related roots of soybean plants. Plant and Soil, 270, 343–353. amino acids in potassium deficient Sesamum. Phytochemistry, 13, Basso, L. C., & Smith, T. A. (1974). Effect of mineral deficiency on amine 2659–2665. formation in higher plants. Phytochemistry, 13, 875–883. Cuevas, J. C., López-Cobollo, R., Alcázar, R., Zarza, X., Koncz, C., Basu, R., & Ghosh, B. (1991). Polyamines in various rice (Oryza sativa) Altabella, T., … Ferrando, A. (2008). Putrescine is involved in Ara- genotypes with respect to sodium chloride salinity. Physiologia Pla- bidopsis freezing tolerance and cold acclimation by regulating abscisic ntarum, 82, 575–581. acid levels in response to low temperature. Plant Physiology, 148, Basu, R., Maitra, N., & Ghosh, B. (1988). Salinity results in polyamine accu- 1094–1105. mulation in early rice (Oryza sativa L.) seedlings. Functional Plant Biol- Cui, J., Abadie, C., Carroll, A., Lamade, E., & Tcherkez, G. (2019). Responses ogy, 15, 777–786. to K deficiency and waterlogging interact via respiratory and nitrogen Battaglia, V., Grancara, S., Satriano, J., Saccoccio, S., Agostinelli, E., & metabolism. Plant Cell and Environment, 42, 647–658. Toninello, A. (2010). Agmatine prevents the Ca2+-dependent induction Cui, J., Davanture, M., Zivy, M., Lamade, E., & Tcherkez, G. (2019). Meta- of permeability transition in rat brain mitochondria. Amino Acids, 38, bolic responses to potassium availability and waterlogging reshape res- 431–437. piration and carbon use efficiency in oil palm. New Phytologist, 223, Benavides, M. P., Aizencang, G., & Tomaro, M. L. (1997). Polyamines in 310–322. Helianthus annuus L. during germination under salt stress. Journal of Dalla Via, L., Di Noto, V., & Toninello, A. (1999). Binding of spermidine and Plant Growth Regulation, 16, 205–211. putrescine to energized liver mitochondria. Archives of Biochemistry Bennett, E. M., Ekstrom, J. L., Pegg, A. E., & Ealick, S. E. (2002). Monomeric and Biophysics, 365, 231–238. S-adenosylmethionine decarboxylase from plants provides an alterna- Davis, G. A., Rutherford, A. W., & Kramer, D. M. (2017). Hacking the thyla- tive to putrescine stimulation. Biochemistry, 41, 14509–14517. koid proton motive force for improved photosynthesis: Modulating Borrell, A., Culianez-Macia, F. A., Altabella, T., Besford, R. T., Flores, D., & ion flux rates that control proton motive force partitioning into Δψ Tiburcio, A. F. (1995). Arginine decarboxylase is localized in chloro- and ΔpH. Philosophical Transactions of the Royal Society B: Biological plasts. Plant Physiology, 109, 771–776. Sciences, 372, 20160381. Bortolotti, C., Cordeiro, A., Alcázar, R., Borrell, A., Culiañez-Macià, F. A., Diem, B., & Godbold, D. (1993). Potassium, calcium and magnesium antag- Tiburcio, A. F., & Altabella, T. (2004). Localization of arginine decarbox- onism in clones of Populus trichocarpa. Plant and Soil, 155, 411–414. ylase in tobacco plants. Physiologia Plantarum, 120,84–92. DiTomaso, J. M., Shaff, J. E., & Kochian, L. V. (1989). Putrescine-induced Brüggemann, L. I., Pottosin, I. I., & Schönknecht, G. (1998). Cytoplasmic wounding and its effects on membrane integrity and ion transport pro- polyamines block the fast-activating vacuolar cation channel. The Plant cesses in roots of intact corn seedlings. Plant Physiology, 90, 988–995. Journal, 16, 101–105. Dobrovinskaya, O., Muniz, J., & Pottosin, I. (1999). Inhibition of vacuolar Buch-Pedersen, M. J., Rudashevskaya, E. L., Berner, T. S., Venema, K., & ion channels by polyamines. The Journal of Membrane Biology, 167, Palmgren, M. G. (2006). Potassium as an intrinsic uncoupler of the 127–140. plasma membrane H+-ATPase. Journal of Biological Chemistry, 281, Dobrovinskaya, O., Muñiz, J., & Pottosin, I. I. (1999). Asymmetric block of 38285–38292. the plant vacuolar Ca2+-permeable channel by organic cations. Camacho-Cristóbal, J. J., Maldonado, J. M., & González-Fontes, A. (2005). European Biophysics Journal, 28, 552–563. Boron deficiency increases putrescine levels in tobacco plants. Journal Erdei, L., Trivedi, S., Takeda, K., & Matsumoto, H. (1990). Effects of of Plant Physiology, 162, 921–928. osmotic and salt stresses on the accumulation of polyamines in leaf Capell, T., Bassie, L., & Christou, P. (2004). Modulation of the polyamine segments from wheat varieties differing in salt and drought tolerance. biosynthetic pathway in transgenic rice confers tolerance to drought Journal of Plant Physiology, 137, 165–168. stress. Proceedings of the National Academy of Sciences of the United Escribano, M. I., Aguado, P., Reguera, R. M., & Merodio, C. (1996). Conju- States of America, 101, 9909–9914. gated polyamine levels and putrescine synthesis in cherimoya fruit Carta, F., Temperini, C., Innocenti, A., Scozzafava, A., Kaila, K., & during storage at different temperatures. Journal of Plant Physiology, Supuran, C. T. (2010). Polyamines inhibit carbonic anhydrases by 147, 736–742. anchoring to the zinc-coordinated water molecule. Journal of Medicinal Feirer, R. P., Hocking, K. L., & Woods, P. J. (1998). Involvement of arginine Chemistry, 53, 5511–5522. decarboxylase in the response of Arabidopsis thaliana to osmotic Chen, C. T., & Kao, C. H. (1993). Osmotic stress and water stress have stress. Journal of Plant Physiology, 153, 733–738. opposite effects on putrescine and proline production in excised rice Feng, J., & Barker, A. V. (1993). Polyamine concentration and ethylene leaves. Plant Growth Regulation, 13, 197–202. evolution in tomato plants under nutritional stress. HortScience, 28, Chen, D., Cao, B., Qi, L., Yin, L., Wang, S., & Deng, X. (2016). Silicon- 109–110. moderated K-deficiency-induced leaf chlorosis by decreasing putres- Flores, H. E., & Galston, A. W. (1982). Polyamines and plant stress: Activa- cine accumulation in sorghum. Annals of Botany, 118, 305–315. tion of putrescine biosynthesis by osmotic shock. Science, 217, Chérel, I., Lefoulon, C., Boeglin, M., & Sentenac, H. (2013). Molecular 1259–1261. mechanisms involved in plant adaptation to low K+ availability. Journal Flores, H. E., & Galston, A. W. (1984). Osmotic stress-induced polyamine of Experimental Botany, 65, 833–848. accumulation in cereal leaves. Plant Physiology, 75, 102–109.

197 14 CUI ET AL.

Fortes, F., Castilho, R. F., Catisti, R., Carnieri, E. G. S., & Vercesi, A. E. Jakobsen, S. T. (1993). Interaction between plant nutrients: III. Antagonism (2001). Ca2+ induces a cyclosporin A-insensitive permeability transi- between potassium, magnesium and calcium. Acta Agriculturae tion pore in isolated potato tuber mitochondria mediated by reactive Scandinavica, Section B—Soil & Plant Science, 43,1–5. oxygen species. Journal of Bioenergetics and Biomembranes, 33,43–51. Janicka-Russak, M., KabaŁa, K., MŁodzinska, E., & KŁobus, G. (2010). The Foster, S. A., & Walters, D. R. (1991). Polyamine concentrations and argi- role of polyamines in the regulation of the plasma membrane and the nine decarboxylase activity in wheat exposed to osmotic stress. tonoplast proton pumps under salt stress. Journal of Plant Physiology, Physiologia Plantarum, 82, 185–190. 167, 261–269. Friedman, R., Altman, A., & Levin, N. (1989). The effect of salt stress on Katiyar, S., & Dubey, R. (1990). Changes in polyamine titer in rice seedlings polyamine biosynthesis and content in mung bean plants and in halo- following NaCl salinity stress. Journal of Agronomy and Crop Science, phytes. Physiologia Plantarum, 76, 295–302. 165,19–27. Friedman, R., Levin, N., & Altman, A. (1986). Presence and identification of Klein, H., Priebe, A., & Jäger, H.-J. (1979). Putrescine and spermidine in polyamines in xylem and phloem exudates of plants. Plant Physiology, peas: Effects of nitrogen source and potassium supply. Physiologia Pla- 82, 1154–1157. ntarum, 45, 497–499. Fujita, M., & Shinozaki, K. (2015). Polyamine transport systems in plants. In Knobloch, K. H., & Berlin, J. (1981). Phosphate mediated regulation of T. Kusano & H. Suzuki (Eds.), Polyamines (pp. 179–185). Tokyo: cinnamoyl putrescine biosynthesis in cell suspension cultures of Nicoti- Springer. ana tabacum. Planta Medica, 42, 167–172. Galston, A. W., & Sawhney, R. K. (1990). Polyamines in plant physiology. Kotakis, C., Theodoropoulou, E., Tassis, K., Oustamanolakis, C., Plant Physiology, 94, 406–410. Ioannidis, N. E., & Kotzabasis, K. (2014). Putrescine, a fast-acting Gilliham, M., Athman, A., Tyerman, S., & Conn, S. (2011). Cell-specific com- switch for tolerance against osmotic stress. Journal of Plant Physiology, partmentation of mineral nutrients is an essential mechanism for opti- 171,48–51. mal plant productivity - another role for TPC1? Plant Signaling and Krämer, R., Mayr, U., Heberger, C., & Tsompanidou, S. (1986). Activation Behavior, 6, 16656–16661. of the ADP/ATP carrier from mitochondria by cationic effectors. Groppa, M. D., Ianuzzo, M. P., Tomaro, M. L., & Benavides, M. P. (2007). Biochimica et Biophysica Acta (BBA)-Biomembranes, 855, 201–210. Polyamine metabolism in sunflower plants under long-term cadmium Krueger, S., Giavalisco, P., Krall, L., Steinhauser, M.-C., Büssis, D., or copper stress. Amino Acids, 32, 265–275. Usadel, B., … Steinhauser, D. (2011). A topological map of the com- Gupta, K., Sengupta, A., Chakraborty, M., & Gupta, B. (2016). Hydrogen partmentalized Arabidopsis thaliana leaf metabolome. PLoS One, 6, peroxide and polyamines act as double edged swords in plant abiotic e17806. stress responses. Frontiers in Plant Science, 7, 01343. Lee, T.-M., Shieh, Y.-J., & Chou, C.-H. (1996). Role of putrescine in enhanc- Hamamoto, S., Marui, J., Matsuoka, K., Higashi, K., Igarashi, K., ing shoot elongation in Scirpus mucronatus under submergence. Nakagawa, T., … Nakanishi, Y. (2008). Characterization of a tobacco Physiologia Plantarum, 96, 419–424. TPK-type K+ channel as a novel tonoplast K+ channel using yeast Lin, C., & Kao, C. H. (1999). Excess copper induces an accumulation of tonoplasts. Journal of Biological Chemistry, 283, 1911–1920. putrescine in rice leaves. Botanical Bulletin Academia Sinica, 40, Hanfrey, C., Sommer, S., Mayer, M. J., Burtin, D., & Michael, A. J. (2001). 213–218. Arabidopsis polyamine biosynthesis: Absence of ornithine decarboxyl- Lin, C. C., & Kao, C. H. (2002). NaCl-induced changes in putrescine content ase and the mechanism of arginine decarboxylase activity. The Plant and diamine oxidase activity in roots of rice seedlings. Biologia Pla- Journal, 27, 551–560. ntarum, 45, 633–636. Hauschild, M. Z. (1993). Putrescine (1,4-diaminobutane) as an indicator of Lin, J., Wang, Y., & Wang, G. (2005). Salt stress-induced programmed cell pollution-induced stress in higher plants: Barley and rape stressed with death via Ca2+-mediated mitochondrial permeability transition in Cr(III) or Cr(VI). Ecotoxicology and Environmental Safety, 26, 228–247. tobacco protoplasts. Plant Growth Regulation, 45, 243–250. Houdusse, F., Garnica, M., Zamarreño, A. M., Yvin, J. C., & García-Mina, J. Liu, J., & Zhang, Y.-Y. (2004). Relationship between ATPase activity and (2008). Possible mechanism of the nitrate action regulating free- conjugated polyamines in mitochondrial membrane from wheat seed- putrescine accumulation in ammonium fed plants. Plant Science, 175, ling roots under osmotic stress. Journal of Environmental Sciences, 16, 731–739. 712–716. Houman, F., Godbold, D. L., Majcherczyk, A., Shasheng, W., & Liu, K., Fu, H., Bei, Q., & Luan, S. (2000). Inward potassium channel in Hüttermann, A. (1991). Polyamines in leaves and roots of Populus guard cells as a target for polyamine regulation of stomatal move- maximowiczii grown in differing levels of potassium and phosphorus. ments. Plant Physiology, 124, 1315–1326. Canadian Journal of Forest Research, 21, 1748–1751. Lothier, J., De Paepe, R., & Tcherkez, G. (2019). Mitochondrial complex I

Hoyos, M. E., Palmieri, L., Wertin, T., Arrigoni, R., Polacco, J. C., & dysfunction increases CO2 efflux and reconfigures metabolic fluxes of Palmieri, F. (2003). Identification of a mitochondrial transporter for day respiration in tobacco leaves. New Phytologist, 221, 750–763. basic amino acids in Arabidopsis thaliana by functional reconstitution Maccarrone, M., Bari, M., Battista, N., Di Rienzo, M., Falciglia, K., & Finazzi into liposomes and complementation in yeast. The Plant Journal, 33, Agro, A. (2001). Oxidation products of polyamines induce mitochon- 1027–1035. drial uncoupling and cytochrome c release. FEBS letters, 507(1), 30–34. Igarashi, K., Kashiwagi, K., Kobayashi, H., Ohnishi, R., Kakegawa, T., Masgrau, C., Altabella, T., Farrás, R., Flores, D., Thompson, A. J., Nagasu, A., & Hirose, S. (1989). Effect of polyamines on mitochondrial Besford, R. T., & Tiburcio, A. F. (1997). Inducible overexpression of oat

F1-ATPase catalyzed reactions. The Journal of Biochemistry, 106, arginine decarboxylase in transgenic tobacco plants. The Plant Journal, 294–298. 11, 465–473. Ioannidis, N. E., Cruz, J. A., Kotzabasis, K., & Kramer, D. M. (2012). Evi- McDonald, R. E., & Kushad, M. M. (1986). Accumulation of putrescine dur- dence that putrescine modulates the higher plant photosynthetic pro- ing chilling injury of fruits. Plant Physiology, 82, 324–326. ton circuit. PLoS One, 7, e29864. Meyer, R. C., Steinfath, M., Lisec, J., Becher, M., Witucka-Wall, H., Ioannidis, N. E., & Kotzabasis, K. (2007). Effects of polyamines on the func- Törjék, O., et al. (2007). The metabolic signature related to high plant tionality of photosynthetic membrane in vivo and in vitro. Biochimica growth rate in Arabidopsis thaliana. Proceedings of the National Acad- et Biophysica Acta (BBA)-Bioenergetics, 1767, 1372–1382. emy of Sciences, 104, 4759–4764. Ioannidis, N. E., Sfichi, L., & Kotzabasis, K. (2006). Putrescine stimulates Minocha, R., Majumdar, R., & Minocha, S. C. (2014). Polyamines and abi- chemiosmotic ATP synthesis. Biochimica et Biophysica Acta (BBA)-Bio- otic stress in plants: A complex relationship. Frontiers in Plant Science, energetics, 1757, 821–828. 5, 00175.

198 PUTRESCINE ROLES UNDER K+ DEFICIENCY 15

Mohapatra, S., Minocha, R., Long, S., & Minocha, S. C. (2009). Putrescine Priebe, A., & Jäger, H. J. (1978). Effect of NaCl on the levels of putrescine overproduction negatively impacts the oxidative state of poplar cells and related polyamines in plants differing in salt tolerance. Plant Sci- in culture. Plant Physiology and Biochemistry, 47, 262–271. ence Letters, 12, 365–369. Moschou, P. N., Paschalidis, K. A., & Roubelakis-Angelakis, K. A. (2008). Reggiani, R., Aurisano, N., Mattana, M., & Bertani, A. (1993). Influence of Plant polyamine catabolism: The state of the art. Plant Signaling and K+ ions on polyamine level in wheat seedlings. Journal of Plant Physiol- Behavior, 3, 1061–1066. ogy, 141, 136–140. Murty, K. S., Smith, T. A., & Bould, C. (1971). The relation between the Reggiani, R., Giussani, P., & Bertani, A. (1990). Relationship between the putrescine content and potassium status of black currant leaves. accumulation of putrescine and the tolerance to oxygen-deficit stress Annals of Botany, 35, 687–695. in Gramineae seedlings. Plant and Cell Physiology, 31, 489–494. Mutlu, F., & Bozcuk, S. (2007). Relationship between salt stress and levels Reggiani, R., Hochkoeppler, A., & Bertani, A. (1989). Polyamines in rice of free and bound polyamines in sunflower plants. Plant Biosystems, seedlings under oxygen-deficit stress. Plant Physiology, 91, 141,31–39. 1197–1201. Naka, Y., Watanabe, K., Sagor, G., Niitsu, M., Pillai, M. A., Kusano, T., & Reggiani, R., Zaina, S., & Bertani, A. (1992). Plasmalemma ATPase in rice Takahashi, Y. (2010). Quantitative analysis of plant polyamines includ- coleoptiles; stimulation by putrescine and polyamines. Phytochemistry, ing thermospermine during growth and salinity stress. Plant Physiology 31, 417–419. and Biochemistry, 48, 527–533. Richards, F., & Coleman, R. (1952). Occurrence of putrescine in potassium- Ndayiragije, A., & Lutts, S. (2006). Do exogenous polyamines have an deficient barley. Nature, 170, 460–462. impact on the response of a salt-sensitive rice cultivar to NaCl? Journal Rugolo, M., Antognoni, F., Flamigni, A., & Zannoni, D. (1991). Effects of poly- of Plant Physiology, 163, 506–516. amines on the oxidation of exogenous NADH by Jerusalem artichoke Öztürk, L., & Demir, Y. (2003). Effects of putrescine and ethephon on (Helianthus tuberosus) mitochondria. Plant Physiology, 95,157–163. some oxidative stress enzyme activities and proline content in salt Ruhl, E. (1989). Effect of potassium and nitrogen supply on the distribution stressed spinach leaves. Plant Growth Regulation, 40,89–95. of minerals and organic acids and the composition of grape juice of Pál, M., Tajti, J., Szalai, G., Peeva, V., Végh, B., & Janda, T. (2018). Interac- sultana vines. Australian Journal of Experimental Agriculture, 29, tion of polyamines, abscisic acid and proline under osmotic stress in 133–137. the leaves of wheat plants. Scientific Reports, 8, 12839. Sarjala, T. (1996). Growth, potassium and polyamine concentrations of Palmieri, L., Todd, C. D., Arrigoni, R., Hoyos, M. E., Santoro, A., scots pine seedlings in relation to potassium availability under con- Polacco, J. C., & Palmieri, F. (2006). Arabidopsis mitochondria have two trolled growth conditions. Journal of Plant Physiology, 147, 593–598. basic amino acid transporters with partially overlapping specificities Sarjala, T., & Kaunisto, S. (1993). Needle polyamine concentrations and and differential expression in seedling development. Biochimica et Bio- potassium nutrition in scots pine. Tree Physiology, 13,87–96. physica Acta (BBA)—Bioenergetics, 1757, 1277–1283. Scaramagli, S., Biondi, S., Leone, A., Grillo, S., & Torrigiani, P. (2000). Accli- Pandolfi, C., Pottosin, I., Cuin, T., Mancuso, S., & Shabala, S. (2010). Speci- mation to low water potential in potato cell suspension cultures leads ficity of polyamine effects on NaCl-induced ion flux kinetics and salt to changes in putrescine metabolism. Plant Physiology and Biochemis- stress amelioration in plants. Plant and Cell Physiology, 51, 422–434. try, 38, 345–351. Pastore, D., Stoppelli, M. C., Di Fonzo, N., & Passarella, S. (1999). The exis- Scott, I., & Logan, D. C. (2008). Mitochondrial morphology transition is an tence of the K+ channel in plant mitochondria. Journal of Biological early indicator of subsequent cell death in Arabidopsis. New Phytologist, Chemistry, 274, 26683–26690. 177,90–101. Peter, H. W., Pinheiro, M. R., & Lima, M. S. (1981). Regulation of the Shabala, S. (2017). Signalling by potassium: Another second messenger to F1-ATPase from mitochondria of Vigna sinensis (L.) Savi cv. Pitiuba by add to the list? Journal of Experimental Botany, 68, 4003–4007. spermine, spermidine, putrescine, Mg2+,Na+, and K+. Canadian Journal Shelp, B. J., Bozzo, G. G., Trobacher, C. P., Zarei, A., Deyman, K. L., & of Biochemistry, 59,60–66. Brikis, C. J. (2012). Hypothesis/review: Contribution of putrescine to Peuke, A. D., Jeschke, W. D., & Hartung, W. (2002). Flows of elements, ions 4-aminobutyrate (GABA) production in response to abiotic stress. and abscisic acid in Ricinus communis and site of nitrate reduction under Plant Science, 193, 130–135. potassium limitation. Journal of Experimental Botany, 53,241–250. Shih, C. Y., & Kao, C. H. (1996). Growth inhibition in suspension-cultured Phelps, D. C., & McDonald, R. E. (1990). Inhibition of electron transport rice cells under phosphate deprivation is mediated through putrescine activities in mitochondria from avocado and pepper fruit by naturally accumulation. Plant Physiology, 111, 721–724. occurring polyamines. Physiologia Plantarum, 78,15–21. Shu, S., Yuan, Y., Chen, J., Sun, J., Zhang, W., Tang, Y., … Guo, S. (2015). Pistocchi, R., Antognoni, F., Bagni, N., & Zannoni, D. (1990). Spermidine uptake The role of putrescine in the regulation of proteins and fatty acids of by mitochondria of Helianthus tuberosus. Plant Physiology, 92,690–695. thylakoid membranes under salt stress. Scientific reports, 5, 14390. Pottosin, I. (2015). Polyamine action on plant ion channels and pumps. In Sinclair, C. (1969). The level and distribution of amines in barley as T. Kusano & H. Suzuki (Eds.), Polyamines (pp. 229–241). Tokyo: affected by potassium nutrition, arginine level, temperature fluctuation Springer. and mildew infection. Plant and Soil, 30, 423–438. Pottosin, I., & Schönknecht, G. (1996). Ion channel permeable for divalent Sjölin, A., & Møller, I. (1991). The effect of polyamines and other cations and monovalent cations in native spinach thylakoid membranes. The on NADH oxidation on the inner surface of the inner mitochondrial Journal of Membrane Biology, 152, 223–233. membrane. Plant Physiology and Biochemistry, 29, 607–613. Pottosin, I., & Schönknecht, G. (2007). Vacuolar calcium channels. Journal Slocum, R. D. (2005). Genes, enzymes and regulation of arginine biosyn- of Experimental Botany, 58, 1559–1569. thesis in plants. Plant Physiology and Biochemistry, 43, 729–745. Pottosin, I., Velarde-Buendía, A. M., Bose, J., Fuglsang, A. T., & Shabala, S. Smith, G. S., Lauren, D. R., Cornforth, I. S., & Agnew, M. P. (1982). Evalua- (2014). Polyamines cause plasma membrane depolarization, activate tion of putrescine as a biochemical indicator of the potassium require- Ca2+-, and modulate H+-ATPase pump activity in pea roots. Journal of ments of lucerne. New Phytologist, 91, 419–428. Experimental Botany, 65, 2463–2472. Smith, T. (1984). Putrescine and inorganic ions. In B. Timmermann & Pottosin, I., Velarde-Buendía, A. M., Bose, J., Zepeda-Jazo, I., Shabala, S., & L. F. C S (Eds.), Phytochemical adaptations to stress (pp. 7–54). Boston: Dobrovinskaya, O. (2014). Cross-talk between reactive oxygen species Springer. and polyamines in regulation of ion transport across the plasma mem- Smith, T. A., & Richards, F. J. (1962). The biosynthesis of putrescine in brane: Implications for plant adaptive responses. Journal of Experimen- higher plants and its relation to potassium nutrition. The Biochemical tal Botany, 65, 1271–1283. Journal, 84, 292–294.

199 16 CUI ET AL.

Stefanelli, C., Maddalena, Z., Bonavita, F., Flamigni, F., Zambonin, L., Landi, L., Turner, L. B., & Stewart, G. R. (1988). Factors affecting polyamine accumu- … Caldarera, C. M. (2000). Polyamines directly induce release of cyto- lation in barley (Hordeum vulgare L.) leaf sections during osmotic stress. chrome c from heart mitochondria. Biochemical Journal, 347,875–880. Journal of Experimental Botany, 39, 311–316. Su, G. X., & Bai, X. (2008). Contribution of putrescine degradation to pro- Urano, K., Yoshiba, Y., Nanjo, T., Ito, T., Yamaguchi-Shinozaki, K., & line accumulation in soybean leaves under salinity. Biologia Plantarum, Shinozaki, K. (2004). Arabidopsis stress-inducible gene for arginine 52, 796–801. decarboxylase AtADC2 is required for accumulation of putrescine in Su, N., Wu, Q., Chen, J., Shabala, L., Mithöfer, A., Wang, H., … Shabala, S. salt tolerance. Biochemical and Biophysical Research Communications, (2019). GABA operates upstream of H+-ATPase and improves salinity 313, 369–375. tolerance in Arabidopsis by enabling cytosolic K+ retention and Na+ Velarde-Buendía, A. M., Shabala, S., Cvikrova, M., Dobrovinskaya, O., & exclusion. Journal of Experimental Botany, 70, 6349–6361. Pottosin, I. (2012). Salt-sensitive and salt-tolerant barley varieties dif- Sung, H.-I., Liu, L.-F., & Kao, C. H. (1994). Putrescine accumulation is asso- fer in the extent of potentiation of the ROS-induced K+ efflux by poly- ciated with growth inhibition in suspension-cultured rice cells under amines. Plant Physiology and Biochemistry, 61,18–23. potassium deficiency. Plant and Cell Physiology, 35, 313–316. Verma, S., & Mishra, S. N. (2005). Putrescine alleviation of growth in salt Sung, J., Lee, S., Lee, Y., Ha, S., Song, B., Kim, T., … Krishnan, H. B. (2015). stressed Brassica juncea by inducing antioxidative defense system. Metabolomic profiling from leaves and roots of tomato (Solanum Journal of Plant Physiology, 162, 669–677. lycopersicum L.) plants grown under nitrogen, phosphorus or Votyakova, T. V., Bazhenova, E. N., & Zvjagilskaya, R. A. (1993). Yeast potassium-deficient condition. Plant Science, 241,55–64. mitochondrial calcium uptake: Regulation by polyamines and magne- Szabo, I., & Zoratti, M. (2014). Mitochondrial channels: Ion fluxes and sium ions. Journal of Bioenergetics and Biomembranes, 25, 569–574. more. Physiological Reviews, 94, 519–608. Votyakova, T. V., Wallace, H., Dunbar, B., & Wilson, S. B. (1999). The cova- Tabor, C. W. (1960). The stabilizing effect of spermine and related amines lent attachment of polyamines to proteins in plant mitochondria. on mitochondria and protoplasts. Biochemical and Biophysical Research European Journal of Biochemistry, 260, 250–257. Communications, 2, 117–120. Walker, D. J., Leigh, R. A., & Miller, A. J. (1996). Potassium homeostasis in Tachimoto, M., Fukutomi, M., Matsushiro, H., Kobayashi, M., & vacuolate plant cells. Proceedings of the National Academy of Sciences, Takahashi, E. (1992). Role of putrescine in Lemna plants under potas- 93, 10510–10514. sium deficiency. Soil Science and Plant Nutrition, 38, 307–313. Wang, B.-Q., Zhang, Q.-F., Liu, J.-H., & Li, G.-H. (2011). Overexpression of Takahashi, H., Imamura, T., Miyagi, A., & Uchimiya, H. (2012). Comparative PtADC confers enhanced dehydration and drought tolerance in trans- metabolomics of developmental alterations caused by mineral deficiency genic tobacco and tomato: Effect on ROS elimination. Biochemical and during in vitro culture of Gentiana triflora. Metabolomics, 8,154–163. Biophysical Research Communications, 413,10–16. Takusagawa, F., Kamitori, S., & Markham, G. D. (1996). Structure and func- Wang, C. Y. (1987). Changes of polyamines and ethylene in cucumber tion of S-adenosylmethionine synthetase: Crystal structures of S- seedlings in response to chilling stress. Physiologia Plantarum, 69, adenosylmethionine synthetase with ADP, BrADP, and PPi at 2.8 Å 253–257. Resolution. Biochemistry, 35, 2586–2596. Wang, J.-W., & Kao, C. H. (2006). Aluminum-inhibited root growth of rice Tamai, T., Shimada, Y., Sugimoto, T., Shiraishi, N., & Oji, Y. (2000). Potas- seedlings is mediated through putrescine accumulation. Plant and Soil, sium stimulates the efflux of putrescine in roots of barley seedlings. 288, 373–381. Journal of Plant Physiology, 157, 619–626. Wang, Y., & Wu, W.-H. (2013). Potassium transport and signaling in higher Tassoni, A., Franceschetti, M., & Bagni, N. (2008). Polyamines and salt plants. Annual Review of Plant Biology, 64, 451–476. stress response and tolerance in Arabidopsis thaliana flowers. Plant Wang, Z., Wang, Y., Shi, J., Zheng, Q., Gao, L., Wang, Q., & Zuo, J. (2019). Physiology and Biochemistry, 46, 607–613. Effects of putrescine on the postharvest physiology characteristics in Tattini, M., Heimler, D., Traversi, M. L., & Pieroni, A. (1993). Polyamine cowpea. Food Science and Nutrition, 7, 395–403. analysis in salt stressed plants of olive (Olea europaea L.). Journal of Watson, M. B., Emory, K. K., Piatak, R. M., & Malmberg, R. L. (1998). Argi- Horticultural Science, 68, 613–617. nine decarboxylase (polyamine synthesis) mutants of Arabidopsis Tiburcio, A. F., Altabella, T., Bitrián, M., & Alcázar, R. (2014). The roles of thaliana exhibit altered root growth. The Plant Journal, 13, 231–239. polyamines during the lifespan of plants: From development to stress. Watson, M. B., & Malmberg, R. L. (1996). Regulation of Arabidopsis thaliana Planta, 240,1–18. (L.) Heynh arginine decarboxylase by potassium deficiency stress. Plant Tiwari, B. S., Belenghi, B., & Levine, A. (2002). Oxidative stress increased Physiology, 111, 1077–1083. respiration and generation of reactive oxygen species, resulting in ATP Weinstein, L. H., Kaur-Sawhney, R., Rajam, M. V., Wettlaufer, S. H., & depletion, opening of mitochondrial permeability transition, and Galston, A. W. (1986). Cadmium-induced accumulation of putrescine programmed cell death. Plant Physiology, 128, 1271–1281. in oat and bean leaves. Plant Physiology, 82, 641–645. Toninello, A., Dalla Via, L., Siliprandi, D., & Garlid, K. D. (1992). Evidence Yoshida, D. (1969). Formation of putrescine from ornithine and arginine in that spermine, spermidine, and putrescine are transported electropho- tobacco plants. Plant and Cell Physiology, 10, 393–397. retically in mitochondria by a specific polyamine uniporter. Journal of Young, N. D., & Galston, A. W. (1983). Putrescine and acid stress: Induc- Biological Chemistry, 267, 18393–18397. tion of arginine decarboxylase activity and putrescine accumulation by Toninello, A., Salvi, M., & Mondov, B. (2004). Interaction of biologically low pH. Plant Physiology, 71, 767–771. active amines with mitochondria and their roles in the mitochondrial Young, N. D., & Galston, A. W. (1984). Physiological control of arginine mediated pathway of apoptosis. Current Medicinal Chemistry, 11, decarboxylase activity in K-deficient oat shoots. Plant Physiology, 76, 2349–2374. 331–335. Trono, D., Laus, M. N., Soccio, M., Alfarano, M., & Pastore, D. (2015). Mod- Yoza, K.-I., Takeda, Y., Sekiya, K., Nogata, Y., & Ohta, H. (1996). Putrescine ulation of potassium channel activity in the balance of ROS and ATP accumulation in wounded green banana fruit. Phytochemistry, 42, production by durum wheat mitochondria—An amazing defense tool 331–334. against hyperosmotic stress. Frontiers in Plant Science, 6, 1072. Zaidan, H. A., Broetto, F., de Oliveira, E. T., Gallo, L. A., & Crocomo, O. J. Turner, L. B., & Steward, G. R. (1986). The effect of water stress upon (1999). Influence of potassium nutrition and the nitrate/ammonium polyamine levels in barley (Hordeum vulgare L.) leaves. Journal of Experi- ratio on the putrescine and spermidine contents in banana vitroplants. mental Botany, 37, 170–177. Journal of Plant Nutrition, 22, 1123–1140.

200 PUTRESCINE ROLES UNDER K+ DEFICIENCY 17

Zapata, P. J., Ma, S., Pretel, M. T., Amorós, A., & Botella, M. A. (2004). Poly- Zorov, D. B., Juhaszova, M., & Sollott, S. J. (2014). Mitochondrial reactive amines and ethylene changes during germination of different plant oxygen species (ROS) and ROS-induced ROS release. Physiological species under salinity. Plant Science, 167, 781–788. Reviews, 94, 909–950. Zepeda-Jazo, I., Velarde-Buendía, A. M., Enríquez-Figueroa, R., Bose, J., Shabala, S., Muñiz-Murguía, J., & Pottosin, I. I. (2011). Polyamines interact with hydroxyl radicals in activating Ca2+ and K+ transport across the SUPPORTING INFORMATION – root epidermal plasma membranes. Plant Physiology, 157,2167 2180. Additional supporting information may be found online in the Zhang, G.-W., Xu, S.-C., Hu, Q.-Z., Mao, W.-H., & Gong, Y.-M. (2014). Supporting Information section at the end of this article. Putrescine plays a positive role in salt-tolerance mechanisms by reducing oxidative damage in roots of vegetable soybean. Journal of Integra- tive Agriculture, 13, 349–357. Zhao, F., Song, C.-P., He, J., & Zhu, H. (2007). Polyamines improve K+/Na+ How to cite this article: Cui J, Pottosin I, Lamade E, homeostasis in barley seedlings by regulating root ion channel activi- Tcherkez G. What is the role of putrescine accumulated under ties. Plant Physiology, 145, 1061–1072. potassium deficiency? Plant Cell Environ. 2020;1–17. https:// Zhong, M., Yuan, Y., Shu, S., Sun, J., Guo, S., Yuan, R., & Tang, Y. (2016). doi.org/10.1111/pce.13740 Effects of exogenous putrescine on glycolysis and Krebs cycle metabolism in cucumber leaves subjected to salt stress. Plant Growth Regula- tion, 79, 319–330.

201 Supplementary Material

Figure S1: 13C-NMR spectra (at natural abundance) of the 20-60 ppm region showing the appearance of putrescine

Figure S2: Metabolites significantly affected (P < 0.01) by K availability

Figure S3: Ca-K relationship in oil palm vegetative organs

202 Putrescine Ornithine Citrulline CDTA CDTA-Ca O O O NH 1 2 5 3 5 3 2’ 2 2 NH NH OH 2 2NH OH 2 NH 4 4 2 NH 1’ NH2 2

2+2’ 1+1’

m c

4 cg LOW K g g a s

2 3 5 4 HIGH K HIGH

m 4+2+2’

c 4 5

sg LOW K a cg g HIGH K HIGH

Fig. S1. 13C-NMR spectra (at natural abundance) of the 20-60 ppm region showing the appearance of putrescine under low K conditions in oil palm (a) or sunflower (b) leaves. Peaks associated with putrescine, ornithine and citrulline are labelled with green, black and pink arrows, respectively, with the no. of C atoms recalled (IUPAC numbering, recalled on top in the grey frame). Ion coordination to improve NMR conditions was done with CDTA (grey arrowheads). Abbreviations: a, alanine; c, citrate; cg, chlorogenate; g, glutamate; m, malate; s, succinate. Note that (i) putrescine accumulates at low K at the expense of ornithine in oil palm while ornithine and putrescine increase simultaneously in sunflower; (ii) the concurrent accumulation of organic acids (malate, citrate) under low K; and (iii) putrescine represents a large metabolic pool in oil palm but not in sunflower. Redrawn from Cui et al. (2019a, 2019b).

203 (a) OIL PALM

Month 9 Month 10 Month 11

K0.2 K1 K4 K0.2 K1 K4 R K0.2 K1 K4 R

–

(b) SUNFLOWER

Week 2 Week 4 K0.2 K1 K4 K0.2 K1 K4 R + ±

–

Fig. S2. Metabolites significantly affected (P < 0.01) by K availability in oil palm (a) and sunflower (b) leaves, presented as a heatmap and a hierarchical clustering (on left). Plants were cultivated either in low, medium and high K (i.e., 0.2, 1 or 4 mM KCl in nutrient solution composition: referred to as K0.2, K1 and K4) or cultivated at low K and then resupplied with high K (denoted as R). Metabolites increased at low K are framed in red, while those decreased under low K are framed in blue. In (b), metabolites with a more complicated pattern (with a strong effect of developmental stage) are framed in green. Putrescine is labelled with a black arrowhead. Redrawn from Cui et al. (2019a, 2019b).

204 14 14

12 12 10

10 8 6

DW) 8 5 6 7 8 9 10 11 -1

6 Ca (mg g Ca(mg 4

0 0 10 20 30 40 50 K (mg g-1 DW)

Fig. S3. Relationship between elemental Ca and K content in oil palm, with all vegetative organs represented (green, leaflets; light grey, rachis; dark grey, petiole, dark yellow, upper trunk section; red, middle trunk section; black, bottom trunk section). Inset, magnification of the relationship in leaflets, with a linear regression (Ca = 15.1 – 0.71 K, r = 0.59) with 95% confidence interval (dotted). Symbols stand for four different progenies. Here, 7-year old oil palm were used in a K-fertilization trial in Sumatra (Indonesia). Total sample number n = 440. At the plant scale, there is a clear antagonism between Ca and K, with rachis and petiole being more variable. At the leaflet scale, the antagonism generates a steep negative slope (inset). From Lamade et al. (open archives of the Isopalm initiative, http://agritrop.cirad.fr/564411/).

205

Appendix 3

The metabolomics signature is a reliable tool to assess K nutrition in oil palm saplings

206 Brief communication

The metabolomics signature is a reliable tool to assess K nutrition in oil palm saplings

Jing Cui1, Juan Manuel Chao de la Barca2, Emmanuelle Lamade3 & Guillaume Tcherkez1*

1. Research School of Biology, ANU Joint College of Science, Australian National University, 2601 Canberra, ACT, Australia. 2. Unité Mitovasc, UMR Inserm 1083, UMR CNRS 6015, Bâtiment IRIS 2, Université d’Angers, 3 rue Roger Amsler, 49055 Angers cedex 2, France. 3. UPR34 Performance des systèmes de culture des plantes pérennes, Département PERSYST, Centre de Coopération Internationale en Recherche Agronomique pour le Développement (CIRAD), 34398 Montpellier, France. *Contact author to whom correspondence should be addressed: [email protected]

Number of figures: 1 Word count: Keywords: oil palm, potassium, diagnostic, machine learning, metabolomics Twitter: @IsoSeed ORCID account: G. Tcherkez: https://orcid.org/0000-0002-3339-956X

207 Main text

Oil palm (Elaeis guineensis Jacq.) is the major oil-producing crop in the world, with a global annual production of about 75 Mt. Low potassium (K) availability is a major concern on tropical soils where oil palm is cultivated since they are often naturally poor in exchangeable cations such as K+. In addition, oil palm growth is highly K-demanding and fruit bunch harvesting removes substantial amounts of K from oil palm agrosystems. For example, typical fruit harvesting of 30 tons FFB (fresh fruit bunches) ha-1 y-1 represents a loss of up to 160 kg K ha-1 y-1, that is, 75% of the K fertilization input (reviewed in (Corley and Tinker, 2016)). Oil palm plantations are thus heavily fertilized with K (typically using potassium chloride, KCl) up to 200 kg K ha-1 y-1 (≈1.5 kg K tree- 1 y-1), leading to an annual cost of about $1 billion at the global scale. However, the efficacy of applied K depends on leaching, the efficiency of K absorption by roots, K allocation within the tree and the response of yield to K availability in the variety (cross) of interest. Quite understandably, intense efforts have been devoted for decades to improve fertilization strategies and monitor oil palm K requirement accurately. Up to now, K fertilisation is mostly based on either the leaf diagnostic which relies on critical %K in leaflets or rachis (French method, (Caliman et al., 1994)) or the knowledge of the response coefficient of yield to applied K using differential decomposition (Foster’s method, (Foster, 2003)). However, both techniques rely on prior knowledge of oil palm response to K and thus typical quantities of fertiliser (KCl) listed in reference tables. Therefore, the major problem of these empirical methods is that they requires systematic agronomical trials to document the response to K in the cross of interest, under the climatic conditions and soil type considered. That is, it may require data from a large number of reliable multifactorial fertilization experiments. As such, it is very tedious and in principle lacks a mechanistic understanding of how K availability modulates growth, fruit production and ultimately, yield. To overcome this problem, the description of metabolic effects of K availability offers an excellent perspective, not only to understand how K nutrition controls oil palm physiology but also to find metabolic signatures that can be implemented to determine when palm trees require fertilization to optimize growth and yield. Recently, the potential of putrescine as a biomarker of K deficiency in relation to K, Mg and Ca balance has been described (Cui et al., 2020b). Here, we examined whether the metabolic signature can be used to provide direct indications on the K status and how it relates to growth potential, so as to appreciate whether oil palm requires K fertilization. To do so, we used oil palm saplings (aged ≈1 y) of a commercial cross Deli  LaMé, grown in a greenhouse nursery under controlled K conditions and sampled at 10, 11 and 12.75 months (Fig. 1a). Oil palm saplings were cultivated under low (0.2 mM KCl in nutrient solution), medium (1 mM), high (4 mM), with or without waterlogging from the eleventh month. In the present context, waterlogging is particularly interesting since it inhibits sap circulation and thus alters K nutrition (Cui et al., 2020a). Some low-K saplings were also subjected to K resupply (at 4 mM) for one or two weeks. In doing so, we thus had situations where K provision was continuous and situations where it was perturbed. That is, our dataset gathers oil palm saplings representative of different life itineraries. We carried out leaflet and rachis GC-MS metabolomics and ICP-OES ionomics to quantify metabolites and elements, respectively (data from (Cui et al., 2019) and deposited on Metabolome Express (Carroll et al., 2010) under accession reference 2018 oil palm K). We then used the dataset to perform machine-learning multivariate statistics based on orthogonal projection on latent structures (OPLS).

208 We first looked at the relationship between leaflet K elemental content (in mg g-1 DW) and absolute growth rate (g DW month-1) (Fig. 1b). Here, the vegetative growth rate was the objective response variable that was looked at because the present work deals with saplings, not adult trees. Despite a generally positive relationship, there was no simple correlation since as expected, it depended on both age (the older the higher the growth rate) and growth conditions (e.g., lower growth rate under waterlogging). Data points seemed to be comprised between sigmoid envelope curves, reflecting either a simple Michaelis-Menten dependence of growth rate with potassium (curvature α = 1) or a cooperative behaviour of internal K pools (α = 5). From this point, multivariate statistics could be used with two methods, using the K content as either an input (predicting) variable X or an objective (response) variable Y. The former is useful to appreciate the importance of the role of K (amongst other metabolites and elements) for growth. The latter (in which elemental contents are withdrawn from the dataset to keep only metabolites) is the way to go to check whether leaflet metabolome is a good predictor of K and growth rate, and thus has some potential for utilization in nutritional monitoring. The first method is illustrated in Fig. 1c with a volcano plot that shows the weight of leaflet compounds in defining the component aligned with growth (and independent of age) generated by the OPLS (y axis), against the P-value obtained by classical univariate statistics using linear multiple regression (x axis). This representation is very convenient to locate most important biomarkers of growth at the right and left extremities of the volcano plot. Here, we confirm that K (red arrow) is amongst the major determinants driving growth in addition to carbohydrates (such as glycerol 3-phosphate or fructose 6- phosphate). By contrast, other cations (Ca, Mg) are anti-correlated with growth, reflecting the well-known antagonism between K and other cations. The output of the second method is shown in Fig. 1d as a biplot that represents how growth and K content relate to components and what metabolites can explain them. The superimposition of samples facilitates the understanding of their position in the multidimensional metabolic space. The first component (x axis) was defined by both growth rate and K, with age (time) and metabolites like sucrose (reflecting the photosynthetic input) being important variables. By contrast, the y axis was solely defined by K (the growth rate appeared to have nearly no coordinate on the y axis). This agrees with the fact that the K content co-varies with growth but it also determined by other mechanisms independent of growth and age (Fig. 1b). Such mechanisms manifest themselves by an increase in some metabolites (glycerol 3-phosphate and 2-oxoallonate for example) and a decrease in other (such as putrescine, asparatate and myoinositol) when the K content increases. This response is consistent with the documented metabolic effects of K availability on metabolic pathways in oil palm (Cui et al., 2019; Mirande- Ney et al., 2020). Taken as a whole, the statistical model was highly explicative (R² = 0.97), robust (the cross-validated R², denoted as Q², was 0.93) and highly significant (P = 10–25 for K). When waterlogging or “resupply” samples were discarded from the dataset to build the model, the metabolic signature still appeared to be highly predictive, with an average error of 0.2-0.3 mg g-1 DW only for the K content and only about 2-8 g DW month-1 in growth rate (not shown). Extra samples from 11.5 month-old palms cultivated under low or high K, with or without waterlogging, were then used to test further the model and check whether it could locate samples in the correct region of the growth-K space. In fact, all samples appeared in the proper region. In addition, when a decision line was drawn half-way between low and high K, the model could identify the correct K status in 100% of cases (Fig. 1e). Also, the metabolomics signature allowed proper classification of samples based on OPLS components, as only four out of 65 (i.e. 6%)

209 samples were wrongly classified (red arrows), due to a confusion between two similar K- sufficient situations: resupply and high K oil palms (which are both under 4 mM KCl at the time of sampling). The present work shown for leaflets in Fig. 1a-f has also been carried out in rachis samples. Under our conditions, there was a less strong response of rachis K content to K availability since under high K, the rachis K content seemed to plateau (Fig. 1g). Therefore, the multivariate analysis gave similar results but was overall slightly less predictive (not shown). Of course, we recognize that the present metabolomics tool was implemented in saplings grown in greenhouse nursery and as such, we used vegetative growth rate as an objective variable to be assessed against. Future studies are warranted to test the performance of this tool in adult tress grown in plantation, with yield as an objective response variable. Still, our study shows that leaflet metabolome appears to be an excellent tool to monitor K nutrition in oil palm saplings, allowing proper diagnostic in the growth-K space, which is essential to take fertilization decisions.

Acknowledgements

The authors warmly thank Cyril Abadie, Illa Tea, Anne-Marie Schiphorst, Camille Bathellier and members of the RSB Plant Service for their help during oil palm sampling. The support of the Joint Mass Spectrometry Facility (ANU) is acknowledged. J. Cui was supported by an Australia Awards PhD Scholarship. The authors thank S. Cholathan (Siam Elite Palm) for providing oil palm seeds.

Conflict of interest

The authors declare no conflict of interest.

References

Caliman, J.-P., Daniel, C. and Tailliez, B. (1994) La nutrition minérale du palmier à huile. Plantations, recherche, développement 1, 36-54. Carroll, A.J., Badger, M.R. and Harvey Millar, A. (2010) The MetabolomeExpress Project: enabling webbased processing, analysis and transparent dissemination of GC/MS metabolomics datasets. BMC Bioinformatics 11, 376. Corley, R.H.V. and Tinker, P.B. (2016) The Oil Palm, Fifth Edition. West Sussex, UK:John Wiley and Sons. Cui, J., Davanture, M., Zivy, M., Lamade, E. and Tcherkez, G. (2019) Metabolic responses to potassium availability and waterlogging reshape respiration and carbon use efficiency in oil palm. New Phytologist 223, 310-322. Cui, J., Lamade, E., Fourel, F. and Tcherkez, G. (2020a) δ15N values in plants is determined by both nitrate assimilation and circulation. New Phytologist, In press. Cui, J., Pottosin, I., Lamade, E. and Tcherkez, G. (2020b) What is the role of putrescine accumulated under potassium deficiency? Plant Cell and Environment, In press. Foster, H.L. (2003) Assessment of oil palm requirements. In: The oil palm, management for large and sustainable yields (Fairhurst, T.H. and Härdter, R. eds), pp. 231-257. Singapore: Potash and Phosphate Institute of Canada (ESEAP). Mirande-Ney, C., Tcherkez, G., Balliau, T., Zivy, M., Gilard, F., Cui, J., Ghashghaie, J. and Lamade, E. (2020) Metabolic leaf responses to potassium availability in oil palm (Elaeis guineensis Jacq.) trees grown in the field. Environmental and Experimental Botany 104062, In press.

210 (c)

211 Fig. 1. Machine-learning analysis of the relationship between potassium nutrition and leaflet metabolome in oil palm saplings Deli  LaMé. (a) Adopted experimental design and workflow. R1 and R2 refer to early and late K resupply, respectively. (b) Relationship between growth rate and potassium elemental content, with differentiation of age (symbol size) and K conditions (colors). (c) Volcano plot (–log(P-value) of univariate analysis against loadings in OPLS multivariate analysis, implemented with Simca®) showing best determinants of growth rate independent of tree age. P-values are associated with the relation with growth rate, using multiple linear regression of the form lm(Metabolite ~ Growth rate + Age) implemented in R. The dashed red line represent the Bonferroni significance threshold (P = 10–3.54). (d) biplot showing the location of K, growth rate and samples in the metabolic multidimensional space. (e) Predicted growth rate and K content in extra samples, and their position with respect to the halfway (medium K) decision line. Samples used to draw the figure in panel (b) are recalled with empty symbols. (f) Hierarchical classification of samples using components (axes) of the multivariate analysis. (g) Relationship between leaflet and rachis K content. Data from 11.5 month-old palms (measured rachis K versus predicted leaflet K) are shown with empty symbols. In both (b) and (g), blue lines stand for sigmoids with curvature (α) as indicated. Abbrevations: 2O(G)A, 2-oxoallonate; αToc, α-tocopherol; Asp, aspartate; BEA, benzene ethanamine; βSit, β-sitosterol; Cell, cellobiose (appears two times in the volcano plot since it generates two analytes during derivatization prior to GC-MS analyses); F6P, fructose 6-phosphate; G3P, glycerol 3-phosphate; Ita, itaconate; Myo, myoinositol; PGA, 3-phosphoglycerate; Put, putrescine; Pyr, pyroglutamate; Suc, sucrose; Succ, succinate; Tym, tyramine; Xyl, xylose.

212

Appendix 4

Potassium deficiency reconfigures sugar export and induces dopamine accumulation in oil palm leaves

213 Potassium deficiency reconfigures sugar export and induces dopamine accumulation in oil palm leaves

Jing Cui1, Emmanuelle Lamade2 and Guillaume Tcherkez1*

1. Research School of Biology, ANU Joint College of Sciences, Australian National University, 2601 Canberra ACT, Australia 2. UPR34 Performance des systèmes de culture des plantes pérennes, Département PERSYST, Centre de Coopération Internationale en Recherche Agronomique pour le Développement (CIRAD), 34398 Montpellier, France.

*Contact author to whom correspondence should be addressed: [email protected]

Short title: Impact of K deficiency on major metabolites

Word count (w/o abstract and acknowledgements): 3,323

Number of figures: 5

Keywords: dopamine, sugars, potassium, deficiency, oil palm

Twitter account: @IsoSeed

ORCID account: G. Tcherkez: https://orcid.org/0000-0002-3339-956X

214 Abstract. Metabolic effects of potassium (K) deficiency have been described for nearly 70 years in many cultivated species. It has been shown that K deficiency stimulates putrescine synthesis, inhibits glycolysis, and slows nitrate and sugar circulation via xylem and phloem saps, respectively. However, specific effects of low K availability on sugar composition, sugar export rate and its relationship with other leaf metabolites are not very well documented. Having such pieces of information is nevertheless essential to identify reliable biomarker metabolites and metabolic signatures to monitor K fertilization in crops. This is particularly important in oil-producing crop species such as oil palm (Elaeis guineensis), which is strongly K-demanding and involves high sugar dependence for fruit formation because of low carbon use efficiency in lipid synthesis. Here, we used metabolic analyses, measured sugar export rates with 13C isotopic labeling and 13C-NMR spectroscopy, and thereby examined the effects of K availability on both leaflet and rachis sugar metabolism in oil palm saplings. We show that low K leads to a modification of sugar composition mostly in rachis and decreased sucrose and hexose export rates from leaflets. As a result, leaflets contained more starch and induced alternative pathways such as raffinose synthesis, although metabolites of the raffinose pathway remained quantitatively minor. The alteration of glycolysis by low K was compensated for by an increase in alternative sugar phosphate utilization by tyrosine metabolism, resulting in considerable amounts of catecholamines (tyramine and dopamine).

215 Introduction

Low potassium (K) availability is a major concern for many crop species that are grown in naturally K-poor areas such as oil palm, Elaeis guineensis Jacq. Also, oil palm growth is highly K-demanding and furthermore, fruit bunch harvesting removes substantial amounts of K from oil palm agrosystems. For example, typical fruit harvesting of 30 tons FFB (fresh fruit bunches) ha-1 y-1 represents a loss of up to 160 kg K ha-1 y-1, that is, 75% of the K fertilization input (for a review, see (Corley & Tinker, 2016)). Oil palm plantations are thus heavily fertilized with K (typically using potassium chloride, KCl) at about 200 kg K ha-1 y-1 (≈1.5 kg K tree-1 y-1) (Heffer, 2009), leading to an annual cost of about $1 billion at the global scale. Quite understandably, intense efforts are now being devoted by oil palm industry to improve fertilization strategies and find solutions from precision agriculture to monitor K needs accurately. Describing metabolic effects of K availability is thus of prime importance not only to understand how K nutrition controls metabolic pathways but also to find metabolic signatures that can be implemented (in addition to the leaf diagnostic based on nutrient elemental contents) to determine when palm trees are K-deficient and require fertilization. For example, the potential of putrescine as a biomarker of K availability in relation to K, Mg and Ca balance has been described recently (Cui et al., 2020). However, many metabolites other than putrescine are affected by K availability. The effect of K deficiency in plants has been studied for a long time and includes an inhibition of pyruvate synthesis as well as an accumulation of organic acids and polyamines (Jones, 1961; Jones, 1966; Okamoto, 1966; Freeman, 1967; Okamoto, 1967; Okamoto, 1968; Besford & Maw, 1976; Armengaud et al., 2009; Hussain et al., 2011; Jwakyung et al., 2015). More recently, metabolomics analyses have been carried out to examine concurrent changes in several metabolic pathways. In K-deficient Arabidopsis rosettes, there is a reorchestration of amino acid and organic acid synthesis (in favour of neutral amino acids), an altered glycolytic metabolism, enhancement of the malate shunt (malate production via phosphoenolpyruvate carboxylase and malate dehydrogenase and then malate cleaveage by the malic enzyme) to circumvent the inhibition of pyruvate kinase (Armengaud et al., 2009) and stimulation of secondary metabolism to produce glucosinolates (Troufflard et al., 2010). In K-deficient sunflower leaves, there is also a considerable change in organic acids coming from the C5 branched pathway (citramalate) likely caused by the inhibition of succinate thiokinase (which uses K+ as a cofactor) (Cui et al., 2019a). K deficiency also affects metabolite exchanges between plant organs. In fact, there is an increase in amino acid export from roots to shoots and analyses of xylem composition have indeed shown an increase in glutamine and γ-aminobutyrate content (Jwakyung et al., 2015). Quite critically, potassium availability has a strong effect on leaf carboxydrate export and phloem loading. In fact, isotopic labelling with 14C or 13C have demonstrated a stimulation of isotope translocation velocity by K (foliar application or fertilization) in several crops and trees (Hartt, 1969; Doman & Geiger, 1979; Thompson & Dale, 1981; Geiger & Conti, 1983; Cakmak et al., 1994; Epron et al., 2015; Martineau et al., 2017). In castorbean (R. communis), low K leads to a lower phloem sap exsudation rate and lower osmotic pressure, as well as lower content in hexose 6-phosphate and raffinose but no change in sucrose (which reporesents more than 90% of sugars in phloem sap) (Haeder, 1977; Mengel & Haeder, 1977). By contrast, in phloem sap collected from willow bark strips with aphids, applied potassium increases sucrose concentration (but not K+ concentration) (Peel & Rogers, 1982). In rice, the apparent sugar transport capacity of phloem calculated from bleeding rate (at the neck internode) and sugar concentration

216 decreased by about 20% when no potassium was added to the soil (Ling et al., 2018). In cotton, using phloem sap collected by bleeding in EDTA solution, low K conditions were found to cause a decrease in sugar export rate (expressed in µg sugars g-1 leaf FW h-1), and an increase or a decrease in amino acids export rate depending on the plant line (Wang et al., 2012). Taken as a whole, potassium appears to be essential for proper phloem function and export ability of source leaves. It is believed that K+ ions play a key role in (i) phloem unloading by increasing or maintaining the transmembrane electrical gradient of phloem cells despite the efflux of sucrose and influx of H+; and (ii) phloem loading by controlling transmembrane gradient created by H+ pumping via K+ influx (for a recent review, see (Dreyer et al., 2017)). While most past studies examining the metabolic impact of K deficiency have been carried out in herbaceous plants, specific effects of K availability on oil palm metabolism have only been elucidated recently. In young oil palm trees, suspending K fertilization for 12-24 months has been found to change bunch fruit set composition (fruit number vs. weight balance) and alter amino acid recycling and lipid synthesis in mesocarp tissue, but final oil composition at maturity remains unchanged (Mirande-Ney et al., 2019). In oil palm saplings grown in the greenhouse, low K conditions cause a decline in photosynthesis, fructose 6-phosphate content and an increase in organic acids, accompanied by a dramatic increase in dark respiration in leaflets (Cui et al., 2019b). The increase in respiratory CO2 production has been found to correlate to higher content in pentitols (polyols in C5), maltose, sucrose and glucose 6-phosphate and lower content in glucarate and 2-oxogluconate, suggesting that sugar export is slowed down and instead, sugars are channelled to the oxidative pentose phosphate pathway (Cui et al., 2019b). Also, in adult oil palm trees in plantation, K availability has been found to correlate positively to the glucose-to-starch ratio across several organs (Lamade et al., 2014). Nevertheless, to our knowledge, effects of K availability on starch synthesis and sugar export from leaves have never been examined precisely. In addition, despite some metabolites have been shown to increase or decrease with high statistical significance under low K, their quantitative impact on metabolism is unknown. In other words, metabolomics have provided information on relative metabolic changes but it remains difficult to interpret metabolic data in terms of fluxes. This is an important limitation because absolute quantities (e.g. in mol g-1) do matter to understand carbon allocation and partitioning and thus anticipate how metabolic changes impact on sugar circulation and therefore fruit bunch production. In fact, oil palm is an oil-producing crop species and fruit development is highly sensitive to sugar provision because of the low carbon use efficiency in lipid synthesis (Lamade et al., 2016) and concurrent use of sugars to produce new leaves and carbon storage in trunk as suggested by pruning experiments (Calvez, 1976; Henson, 2002; Legros et al., 2009a; Legros et al., 2009b) and the relationship between trunk soluble sugars and yield (Henson et al., 1999). Here, we conducted metabolic analyses using both 1H-NMR and mass spectrometry (GC-MS metabolomics) in oil palm saplings grown under low or high K conditions in order to quantify metabolic changes and appreciate potential modifications of sugar composition, both in leaflets and leaf rachis. In addition, we performed isotopic 13 13 labelling with CO2 followed by C-NMR analyses in both leaves and rachis, in order to examine carbon redistribution in sugars and determine sugar export rates. Our results show that at low K, the major change in sugar composition is an increase in glucose content in rachis, despite lower export rate and higher leaflet starch content. Furthermore, the quantitatively most important changes amongst non-sugar metabolites are in tyramine and dopamine, suggesting an unprecedented role of catecholamine metabolism under low K availability.

217 Material and methods

Plant material and growth conditions Germinated oil palm seeds (Dura  Pisifera) were obtained from Siam Elite Palm Co. Ltd. (Krabi, Thailand). Plants were sown directly in sand (washed with distilled water) in the greenhouse, using 7-L pots. K-containing nutrient solution (4 mM KCl) was provided for one month until emergence, and then controlled nutrient conditions at varying K concentration were used. Plants were cultivated in the Research School of Biology plant facility at the Australian National University in Canberra (Australia). Growth conditions were: natural photoperiod (from 13.5/10.5 h at emergence (February) to 14.5/9.5 h at the end of the experiment (January), with a minimum of 9.5/14.5 h in June), 30/24°C air temperature, 70/60% relative humidity day/night.

Nutrient conditions sampling The nutrient solution was composed on purpose for this experiment, whereby the amount + of K was varied by changing the amount of KCl. It consisted of NaNO3 (4 mM), Ca(NO3)2 (4 mM), MgSO4 (1.5 mM), NaH2PO4 (1.3 mM) and a mixture of microelements (Na-Fe-EDTA 0.11 mM, MnSO4 2 µM, ZnSO4 2 µM, CuSO4 2 µM, H3BO4 100 µM, Na2MoO4 1 µM, Co(NO3)2 0.4 µM). Three K availability conditions were used: “low K” (0.2 mM), “medium K” (1 mM), and “high K” (4 mM) up to 11 months after sowing. At 11 months, since plant size was already substantial, plants at “medium K” were removed to clear the greenhouse compartment and provide more space to low and high K plants (this is referred to as “thinning” thereafter). Data presented here show results obtained upon seven sampling dates: 9 months, 10 month, 11 months, and then 11 months plus 1, 2, 3 or 7 weeks after thinning. Upon sampling, plants were measured for size, fresh weight, and dissected and kept in liquid nitrogen until further use. Metabolomics was performed on freeze-dried material.

Proteomics Protein data reported here are re-tabulated from our previous work in (Cui et al., 2019b) where we used a factorial experiment with K and waterlogging. Here we report data only related to the K effect, samples from waterlogged palms having been discarded. Briefly, analyses used proteins extracted using TCA-acetone. After digestion, peptides were analysed by LC-MS/MS (nano-HPLC coupled to mass spectrometry via an electrospray interface). Protein identification, filtering and grouping was carried out using the X!Tandem pipeline, after (Langella et al., 2017).

Metabolomics analyzes Gas chromatography coupled to mass spectrometry (GC-MS) analyses were carried out as in (Cui et al., 2019a; Cui et al., 2019b). We used methanol:water extracts derivatized with methoxylamine and N-methyl-N-(trimethylsilyl) trifluoroacetamide (MSTFA) in pyridine. Ribitol was used as an internal standard. 1H-NMR analyzes were performed on 20 mg dry weight extracted with 1 mL phosphate buffer solution (50 mM). After centrifugation, 530 µL supernatant were mixed with 50 µL D2O and 20 µL TMSP (sodium [2,2,3,3-d4] 3-(trimethylsilyl) propionate in D2O; internal standard) and poured in a 5-mm tube. NMR acquisitions were done at 298 K with excitation sculpting water suppression (pulse program zgesgp) on a Bruker Advance 700 MHz spectrometer. GC- MS and 1H-NMR data were normalized to dry weight. Unless otherwise stated, five replicates were analyzed for all conditions. Supervised multivariate analysis of metabolomics data was carried out by orthogonal projection on latent structure (OPLS)

218 with Simca 13 (Umetrics), using K level and developmental time as predicted qualitative Y variables and metabolites as predicting X variables. The absence of statistical outliers was first checked using a principal component analysis (PCA) to verify that no data point was outside the 99% confidence Hostelling region. The goodness of the OPLS model was appreciated using the determination coefficient R² and the predictive power was quantified by the cross-validated determination coefficient, Q². The significance of the statistical OPLS model was tested using a ² comparison with a random model (average  random error), and the associated P-value (PCV-ANOVA) is reported. A permutation test was also performed to check the reliability of the OPLS model, that is, to verify that at maximal permutation (similarity of permuted dataset tending to zero), Q² was always negative. Best discriminating metabolites (biomarkers) were identified using volcano plots whereby the logarithm of the P-value obtained in univariate analysis (two- way ANOVA) was plotted against the rescaled loading (pcorr) obtained in the OPLS. In such a representation, best biomarkers have both maximal –log(P) and pcorr values.

13C labelling and 13C-NMR analysis Labelling was performed in a gas-exchange system consisting of a purpose-built chamber coupled to the Licor 6400-XT (Licor Biosciences, USA), with soft walls that could be cut very quickly and thus allowed instant sampling with liquid nitrogen as described previously (Tcherkez et al., 2012). Two leaflets (at point B of the leaf) with a total surface area of about 70 cm² were placed in the leaf chamber. Gas-exchange conditions were 500 -2 -1 µmol m s PPFD, 400 ppm CO2, 21% O2, 80% relative humidity and 21-23°C air 13 13 temperature. Isotopic labelling was performed using CO2 (Sigma-Aldrich, 99% C) for 4 h (for dark-adapted samples, followed by 2 h darkness) after having reached steady photosynthesis using ordinary CO2 (natural abundance) for ca. 60 min. Two series of 13 experiments were done: with CO2, and with natural CO2. Performing experiments with 13 natural CO2 was strictly required for % C calculations using NMR data. Leaflets and rachis (rachis section to which leaflets were attached) were both sampled. Samples were extracted with perchloric acid and maleate (internal standard) in liquid nitrogen. The extract was neutralized, frozen-dried and resuspended in water, 50 µL D2O were added and the sample was poured in a 5-mm tube. NMR analyzes were performed at 298 K using proton-decoupled (decoupling sequence waltz16) carbon pulse program (zgig) as in (Abadie et al., 2017). To get a good signal-to-noise ratio, 12,500 scans were done, representing about 9 h analysis per sample. NMR data presented in the paper are meanSD of n = 4 replicates.

Starch extraction and analysis Non-soluble starch (amylose + non-soluble amylopectin) was extracted as in (Tcherkez et al., 2003) using HCl-catalyzed gelatinization followed by flocculation in cold methanol. After having discarded the supernatant, starch pellets were frozen-dried and weighed in tin capsules and the absolute 13C content was measured by elemental analysis- isotope ratio mass spectrometry, EA-IRMS (Carlo Erba-Isoprime, Elementar). Soluble amylopectin (phytoglycogen) was visible by NMR (C-1 atoms of glycosyl monomers at a chemical shift of about 100 ppm) and this signal was used for calculations.

Calculation of sugar export rate The export rate was determined with two methods: isotopic mass-balance, and isotopologue abundance fitting. Mass-balance calculations used the difference between assimilated 13C (13A, known from photosynthesis measured by gas exchange monitoring during labelling) and 13C found in sugars (soluble sugars, 13S, and starch, 13U). 13U was

219 obtained from the sum of insoluble (from IRMS analysis) and soluble (from NMR analysis) starch collected after the light period. Of course, this difference did not only represent exported 13C but also 13C in other minor compounds (such as amino acids or organic acids synthesized). However, the contribution of minor compounds was quantitatively small (this is apparent in NMR spectra shown below). Therefore, the quantity of 13C exported from leaflets (13Q) was given by (in µmol 13C): 13QASU 13 13 13 (1) This 13C amount was converted to a flux in µmol m-2 s-1 as: 13Q 13E  (2) T where σ is leaflet surface area and T is the time spent in the light. 13E does not represent the export rate because exported sugars were not completely labelled (i.e. were not at 100% 13C), thus the export rate was computed as: 13E E  (3) p where < p > is the average %13C in sugars. For the export rate in the dark, a similar calculation was done, replacing, in equation (1), 13A by 13U and 13U by the 13C amount 13 represented by starch after 4 h light + 2 h darkness ( Udark). Isotopologue abundance-based calculations exploited the appearance of multi- labelled molecular species of sugars both in leaflets and rachis. That is, the majority of 13C-labelled sugar molecules had several 13C atoms simultaneously (showing the rapid turn-over of sugar active pools). This was readily visible in NMR spectra were leaflet sugars signal were in the form of multiplets coming from double 13C-13C interactions, that is, the presence of two neighbor 13C for each observed 13C atom (this is further explained in Fig. S1). Decomposing the signal of multiplets as in (Abadie & Tcherkez, 2019) was used to (i) calculate the proportion of labelled neighbor 13C atoms and thus of mono-, bi- and tri-labelled molecules for each C-atom position of interest and (ii) calculate the fraction of metabolically inactive pool (that is, non or slowly turned-over and thus not affected by labelling; referred to as “static” thereafter). Fully labelled isotopic species could thus be used as tracers for which ordinary differential equations could be applied. To do so, we used the simple model depicted in Fig. S2 where sugars are made of a dynamic pool and a static pool. For each active pool of interest, if the metabolite active pool size Q is constant, we have: dx x  x P F (4) dt Q where x is the proportion of fully labelled molecules in the pool of interest, xP is the proportion of fully labelled molecules in the precursor of this pool, and F is the flux (turn- over rate). For simplicity, we assumed that x in chloroplastic phosphorylated intermediates (such as triose phosphates) was unity. In fact, Calvin cycle intermediates are present at low concentration (small pool size) and thus they are turned-over within seconds (see for example Fig. S3a). By contrast, cytosolic hexose phosphates are certainly less rapid to be labelled because of the sucrose futile cycle that can lead to sucrose cleavage and hexose recycling. For the dark period, the value of xstarch used (source sugar for metabolism and export) was calculated using the difference in 13C-enrichment of starch (insoluble and soluble) just before the dark period (end of light) and just after the dark period (overall 13C fraction in starch that disappeared in darkness). The values of Q were determined from samples obtained with CO2 at natural abundance (except for hexose phosphates, the pool size of which was assumed to be small, at 1 µmol g-1 FW). Equations were then solved numerically with Excel for flux values so that after 14,400

220 (light) or 7,200 (dark) seconds, modelled x values matched observed x values both in leaflets and rachis. An illustration of kinetic profile obtained therefrom is provided in Figs. SXb-c. In our model, to avoid under-determination, we assumed that glucose and fructose were exported at the same rate (Efg). Therefore, modelling had two outputs: sucrose export rate (Es) and fructose-glucose export rate (Efg). Other assumptions are given in the legend of Fig. S3b-c.

221 Results

Major compounds in oil palm leaflet and rachis Quantitatively major leaflet metabolites were analysed by 1H-NMR and a typical profile is presented in Figs. 1 and S4-5. Apart from gallic acid, leaflets contain homovanillic acid, sucrose, glucose and fructose (Fig. 1) and in quantitatively less important amounts, arginine, malate, alanine, glutamate and its derivatives (Fig. S4-5). Amongst sugars, sucrose prevailed, being at least two times more abundant than glucose or fructose. In fact, when NMR signals were used to determine quantities expressed in nmol g-1 dry weight (DW), the sucrose content was found to be about 200 nmol g-1 while hexoses were about 50-80 nmol g-1 (Fig. 2a). By contrast, the rachis contained a lot of glucose, up to 1150 nmol g-1 thus much more than sucrose (Fig. 2b) (as already found in (Lamade et al., 2009; Lamade et al., 2016)). When K availability varied, leaflet sugar composition was not significantly altered, unlike organic acids (increase in malate and citrate, Fig. S4) and nitrogenous compounds: arginine (decreased), dopamine and tyramine (increased) (Fig. 2a, S4-5). Since the appearance of catecholamines (dopamine and tyramine) is rather unusual, the spectrum of authentic standards was compared with leaflet extracts to locate specific triplets near 2.9 and 3.2 ppm, and multiplets in the 6.7-7.2 ppm region and thus confirm the presence of catecholamines (Fig. S6). In rachis, the sugar composition changed a lot under low K, with a 60% increase in glucose and reciprocally, a decrease in sucrose and fructose (Fig. 2b). As in leaves, amino acids such as arginine and asparatate + asparagine decreased considerably, while malate increased (Figs. 2b, S7). Catecholamines were also visible in rachis at low K although in much lower amount: in fact, they were hardly visible in the 2.9-3.2 ppm region of the NMR spectrum and detectable with a very small signal at high chemical shift (Fig. S8). It should be noted that in both leaflets and rachis, NMR signals of putrescine are hardly visible (Fig. S5). In fact, despite the well-known increase in putrescine under low K conditions (Cui et al., 2020), here putrescine amount appeared to be small compared to major metabolites and furthermore, a fraction of putrescine was present in its acetyl-conjugated form (Fig. S5). Similarly, raffinose and myoinositol appeared to be quantitatively minor compounds, since they were hardly visible in NMR spectra (see the comparison with standards in Fig. S8).

13C labelling in metabolites 13 13 As expected, when leaflets were labelled with CO2, sugars appeared to be highly C- enriched as shown by 13C-NMR spectra (Fig. 3). There was a clear difference between high and low K conditions with much less 13C being recovered at low K (Fig. 3b). The NMR signal of each C-atom in sugars was made of a multiplet (Fig. 3c-d) showing that multiply labelled molecules prevailed in the sugar metabolic pool. In fact, at each C-atom position, the presence of neighbor 13C atoms led to the formation of a multiplet instead of a singlet (see Fig. S1 for further details). The fact that leaflet sugars were also present as multiply labelled molecules under low K (Fig. 3d) demonstrates that the turn-over of photosynthetic intermediates at the origin of sugars was also very fast under low K conditions despite lower photosynthesis rates. In rachis, 13C labelling was much lower and there were little difference in major peaks of the 13C-NMR spectra between labelled and unlabelled samples (Fig. S9). This was unsurprising considering that sugars in rachis originate not only from labelled leaflets but also from all other leaflets (unlabelled) still attached to the leaf, so that labelled leaflet surface area represented ≈5% only of total leaflet surface area. By contrast, there was a clear difference in minor peaks showing the contribution of multiply labelled sugar species. In other words, rachis sugar were made

222 of a big pool of molecules at 13C natural abundance plus a small pool of highly 13C- enriched sugars.

13C allocation and sugar export NMR and IRMS data were used to compute 13C amounts and compare with photosynthetic 13C input (Fig. 4). In agreement with the rather low starch content in leaflets (Fig. 2), 13C allocation to starch was small, less than 12% of net fixed photosynthetic 13C (Fig. 4a). Soluble sugars represented by far the most important fraction of 13C recovered in leaflets. When expressed in %13C, NMR data indicate that sucrose was always the most enriched sugar, nearly 100% labelled under low K in the light (Fig. 4b). Glucose and fructose were much less 13C-enriched than sucrose in the light but not in the dark. At low K, the 13C-difference between sucrose and hexoses was observed in dark. In other words, the turn-over of sucrose in the light is the most impacted flux under low K conditions. As expected, the isotopic enrichment in rachis was small, just above natural abundance (Fig. 4b). It should nevertheless be kept in mind that because of the big pool size of rachis sugars (Fig. 2), even a small 13C-enrichment represents a considerable 13C amount. Also in rachis, starch represented an extremely small 13C amount so that its isotope composition was different from natural abundance by 4‰ only on the δ13C scale (data not shown). Isotopic data (Fig. 4b) were then used with pool sizes (Fig. 4c) to calculate the rate of export using differential equations, or using mass-balance. In agreement with %13C data, at high K, sucrose export was by far the prevalent export flux (about 90% of total export rate) in the light and was nearly suppressed in darkness (Fig. 4d). At low K, sucrose export remained the prevalent flux while export rates decreased nearly two-fold in the light. In darkness, export rates were found to be extremely small.

Metabolomics analysis Since K conditions have mid- to long-term effects, we carried out a metabolic analysis at different time points during palm growth using GC-MS as a high-throughput technique. This represented a set of 100 samples, from which we could analyse the effect of developmental time and K conditions. In other words, this allowed us to extract metabolites that were robustly associated with K conditions. We detected and identified 163 metabolites. The multivariate statistical analysis allowed excellent discrimination between samples (R²=0.945 and Q²=0.916), with a highly significant K and time effect -21 -30 (PCV-ANOVA = 8∙10 and < 10 respectively). The combination of multivariate and univariate analysis shows that metabolites best explain the difference between low and high K samples are putrescine, organic acids, dopamine, myoinositol and galactitol (increased at low K) and glucuronate, 2-oxoallonate, glycerol 3-phosphate and fructose 6-phosphate (decreased at low K) (Fig. 5a). The analysis of the covariation pattern by hierarchical clustering shows that metabolites significantly affected by K formed four groups (Fig. 5b): metabolites with a low K-high K difference all across time points (glucose, glucuronate, glycerol 3-phosphate, cluster 1), metabolites that tended to be more abundant early on (time effect) with a larger K effect later on (amino acids and organic acids, cluster 2), metabolites that were more variable with time (diverse metabolic classes and includes myoinositol, cluster 3) and metabolites that tend to be more abundant later on with larger low K/high K difference also later on (comprising putrescine, galactinol, dopamine, tyrosine, raffinose and some organic acids cluster 4). When examined separately (Fig. 5c), it is clear that sucrose increased with time but showed no effect at all of K conditions. By contrast raffinose and galactinol increased considerably after 11 months, that is, just after thinning, and increased at low K. Dopamine and putrescine were

223 also increased by thinning and were always more abundant under low K across all time points. The same analysis was conducted in rachis (Fig. S10). Sample types could be easily discriminated (R²=0.944 and Q²=0.900) with a highly significant effect of K and -20 -30 time in multivariate analysis (PCV-ANOVA = 7∙10 and <10 , respectively). As in leaflets, best discriminating metabolites included myoinositol and catecholamines (increased at low K) and 2-oxoallonate (decreased at low K) (Fig. S10a). Also, the effect of low K on amino acids was stronger in rachis than leaflets (cluster 1, Fig. S10b), suggesting an inhibition of the export of nitrogenous compounds from leaflets. Like in leaves, there was a significant increase in myoinositol, raffinose, galactitol and galactinol (Figs. S10b-c), suggesting that the entire raffinose pathway was stimulated in both organs. It is worth noting that in both organs, α-tocopherol increased at low K and belonged to the same cluster as dopamine, while these two metabolites are products of tyrosine metabolism.

224 Discussion

Re-orchestration of carbon export from leaflets

K availability impacted not only on sugar synthesis by leaflet photosynthesis (Fig. 4a, (Cui et al., 2019b)) but also sugar export, in agreement with results obtained in other species (see Introduction). Here, we have determined export using mass-balance and 13C isotopologues abundance in both leaf and rachis and at low K, we found a strong reduction in total sugar export rate of about 50% in the light and 90% (or more) in darkness (Fig. 4d). The strong effect of low K observed in darkness not only came from a slower starch turn-over, but also from the fact that in darkness, leaflets exported more hexoses than sucrose (while sucrose cycling within leaflets is so that the isotopic enrichment is similar in sucrose, glucose and fructose in darkness, Fig. 4b). Parenthetically, the small value of sucrose export rate in darkness came from the fact that the 13C-enrichment in source leaflet sucrose was high at the end of the light period and showed a strong decline in darkness, while the most enriched sugar in darkness in rachis was fructose. It is presently difficult to know whether the export pattern changed in darkness under low K conditions because export rates were very small. It is also possible that once loaded in the rachis, glucose movement in rachis sieve tubes was slowed down as suggested by the huge glucose accumulation under low K (Fig. 2, 4c). This would be in agreement with the observed correlation between elemental K content and glucose concentration across trees and organs (Lamade et al., 2014). Overall, the effect of low K was probably primarily linked to K+ scarcity impeding transmembrane energization (K+/H+ gradient) required for phloem loading. We nevertheless recognize that with the present data, it is difficult to know whether the long-distance sugar species circulated to other organs (petiole and stem) is glucose, fructose or sucrose, or a mixture of them, and whether this can change with K availability. Here, our 13C-data and calculated export rates suggest that it is a mixture of sucrose and hexoses where sucrose prevails (Fig. 4d). Under high K conditions, the pool of glucose in rachis decreased in the dark compared to the light while the glucose + fructose export rate increased and the sucrose export rate decreased, suggesting that the glucose pool in the rachis was formed by sucrose cleavage. 14C-labelling experiments and petiole sap exudation with EDTA showed that sucrose was by far the most labelled compound (Houngbossa & Bonnemain, 1985) suggesting that sucrose is the main circulating sugar. However, the low 14C activity observed in glucose could have been the result of the isotopic dilution by the glucose pool of the rachis. Transcriptomics have shown that in fruits, invertase is highly expressed during maturation (Dussert et al., 2013), suggesting that the long-distance transported sugar is sucrose. Still, obtaining a definitive answer would require a strong 13C labelling and then tracing of labelled sugar species in different organs. Low K conditions also affected metabolic pathways involved in sugar export (summarized in Fig. S11). In fact, metabolomics analyses show that there was a significant increase in raffinose and galactinol under low K in both leaflets and rachis (Figs. 5, S10). The enzymatic mechanism responsible for raffinose and galactinol synthesis involves galactinol synthesis from UDP-galactose and myoinositol via inositol galactosyl transferase (IGT) and recombination of galactinol and sucrose into raffinose, regenerating myoinositol, via galactinol sucrose galactosyl transferase (GSGT). Interestingly proteomics have shown that low K was not associated with a general increase in GSGT and IGT content but a decline in GSGT 2 and no significant effect on IGT (Table 1). The oil palm genome comprises 8 genes encoding GSGT (Sanusi et al.,

225 2018) thus presumably the decline in GSGT 2 did not lead to a tangible decline in raffinose synthesis. It should also be noted that GSTS is a reversible enzyme, is inhibited by myoinositol for raffinose synthesis (with a high Ki of about 4.5 mM, far from concentrations observed here, Figs. 1, S8) and can accept lactose (instead of sucrose) as a galactosyl acceptor (Peterbauer et al., 2002; Li et al., 2007). We also show that under low K, the sucrose content in leaflets did not change significantly (Fig. 2,5c) although there was an increase in sucrose synthetizing proteins, sucrose phosphate synthase 2 and sucrose phosphatase (Table 1). Also, myoinositol, galactinol and raffinose increased (Figs. 2,5). This suggests that despite the lower GSGT 2 content, the flux of raffinose synthesis was higher, thereby consuming sucrose and regenerating myoinositol. Interestingly, there was an increase in β-galactosidases (Table 1). This in turn indicates that raffinose metabolism was associated with galactose regeneration via β-galactosidases and thus cleavage of lactose, which can in turn be reformed from glucose and UDP- galactose by β-galactosyltransferase (lactose synthase; the oil palm genome comprises 22 sequences annotated β-galactosyltransferase). Galactose was not amongst significant metabolites (P = 0.09 for the K effect; P = 0.02 for the interaction K  time) but galactitol (polyol synthesized from galactose) was amongst the best biomarkers of low K conditions (Fig. 5). Taken as a whole, the raffinose pathway was re-orchestrated and this represents a response to the lower utilization of sucrose, glucose and fructose by phloem loading. The physiological advantage of raffinose utilization is that it does not require coupling to H+ transport and thus associated ATP consumption, unlike sucrose or hexose loading (De Schepper et al., 2013). In addition to polyols (myoinositol, galactitol, glycerol), sugars were also converted to derivatives of the oxidative pentose phosphate pathway (such as xylitol, xylonolactone, erythritol) and this was at the expense of (i) alternative utilization of UDP- glucose by aldarate metabolism (glucuronate being the best biomarker that declines under low K) (Fig. 5) and (ii) glycolysis as shown by the strong decrease in fructose 6-phosphate and fructose 1,6-bisphosphatase and conversely, the increase in PPi fructose 6-phosphate phosphotransferase (PFP) (Fig. 5, Table 1).

Alternative PEP utilization by dopamine and tyramine production

In addition to the typical build-up of organic acids coming from tricarboxylic and C5 branched acid pathway (summarized in Fig. S11) (Cui et al., 2019a; Cui et al., 2019b), a striking feature of the present metabolic analysis was the accumulation of dopamine (and tyramine) under low K, to levels comparable to sugars at 11 months (Figs. 1-2, S5). Dopamine is usually present in tiny amounts in plants (like all catecholamines, as opposed to animals) and up to now, significant dopamine amounts have only been found in banana (Waalkes et al., 1958; Kulma & Szopa, 2007). Here, dopamine was amongst the best biomarkers of low K conditions (Fig. 2,5a), although its content varied considerably with time (Fig. 5c). Dopamine production involves the shikimate pathway (synthesis of aromatic amino acids) and here, shikimate, phenylalanine and tyrosine were also amongst significant metabolites (the biochemical pathway involving dopamine is recalled in Fig. S11). Proteomics analyses have further shown that a peptide associated with polyphenol oxidase (required for dopamine synthesis) is amongst the most highly significant features (P < 10-10). The biological significance of dopamine accumulation is not clear. However, it should be noted that dopamine synthesis has several metabolic advantages. First, tyrosine synthesis does not consume NADPH and thus does not compete with photosynthesis electron utilization in the chloroplast. In fact, tyrosine involves chorismate synthesis

226 which consumes NADPH (2 phosphoenolpyruvate + erythrose 4-phosphate + NADPH + ATP  chorismate + 3 Pi + NADP + ADP) and chorismate conversion to tyrosine, which produces NADPH (chorismate + glutamate + NADP  tyrosine + CO2 + 2-oxoglutarate + NADPH). Dopamine production also does not require NADPH nor ATP (tyrosine + O2  CO2 + H2O + dopamine). Second, tyrosine synthesis consumes erythrose 4-phosphate that can be synthesized by the cytosolic pentose phosphate pathway (see above) in addition to the Calvin cyle in the chloroplast. Third, tyrosine synthesis consumes phosphoenolpyruvate (PEP) which cannot be consumed by pyruvate kinase (since this enzyme strictly depends on K+ as a cofactor). Fourth, dopamine catabolism ultimately regenerates fumarate and acetyl-CoA. The oil palm genome contains the key enzymes required for dopamine and tyramine catabolism, including fumarylacetoacetase (Fig. S12). Our analyses also have suggested that homovanillic acid appears amongst metabolites at low K (Fig. 1, S4). Taken as a whole, the “dopamine shunt” (synthesis + degradation) allows PEP recycling and compensate for the difficulty to synthesize pyruvate and thus acetyl-CoA due to the inhibition of pyruvate kinase. Nevertheless, this is associated with an extra loss of CO2, and this decreases the efficiency of catabolism. Considering the considerable amount of dopamine involved, this probably contributes to the much higher respiration rates observed at low K.

Perspectives

Apart from the metabolic advantage of the dopamine shunt, the specific roles of dopamine and tyramine as accumulated molecules remain unknown. It has been shown that (exogenous) dopamine can be beneficial for salt or drought tolerance (Zhou et al., 2007; Li et al., 2015; Liang et al., 2018; Jiao et al., 2019). Also, a link has been found between heterologous dopamine receptor expression and sugar accumulation in potato, suggesting that dopamine could have a signaling role to regulate sucrose metabolism (Skirycz et al., 2005). Like putrescine (Cui et al., 2020), dopamine could also play a role in ion 2+ homeostasis via its action on ion (Ca ) channels (as found in Mammals), H2O2 scavenger (as suggested by (Kulma & Szopa, 2007)), or in organelle homeostasis to regulate non- photochemical quenching (in chloroplasts) and mitochondrial permeability transition (in mitochondria). It is worth noting that oil palm possesses genes also involved in alkaloid synthesis (thebaine family) and two genes encoding a 3,4-DOPA dioxygenase extradiol (shown in the dopamine pathway in Fig. S12) (Sanusi et al., 2018), normally encountered in betalain-producing species (Caryophyllales) (Tanaka et al., 2008; Khan & Giridhar, 2015) but also encountered in other species, like tobacco where it is involved in the response to pathogens (Bahramnejad et al., 2010), and mung bean in the response to H2O2 (Li et al., 2017). The committed step of alkaloid production, tyrosine decarboxylase, has been found to be induced by specific nutritional conditions such as high sugars and low phosphate or low nitrogen (for a review, see (Facchini et al., 2000)). Presumably, oil palm metabolism may involve the production of alkaloids perhaps playing a role in stress response such as low K. Tyramine has been found to be amongst significant metabolites under pathogen infection causing oil palm fatal yellowing (Rodrigues-Neto et al., 2018). Interestingly, a protein involved in alkaloid production (salutaridine reductase) is significantly more abundant at low K (P < 10-6) (Cui et al., 2019b). It is also striking that the other documented species that contains relatively high contents in dopamine is banana, also a tropical Monocot tree species. Future studies are warranted to determine whether there is indeed a link between dopamine, taxonomy or region of origin.

227 Acknowledgements

The authors warmly thank Cyril Abadie, Illa Tea, Anne-Marie Schiphorst, Camille Bathellier and members of the RSB Plant Service for their help during oil palm sampling. The support of the Joint Mass Spectrometry Facility (ANU) is acknowledged. J. Cui was supported by an Australia Awards PhD Scholarship, and the research was supported by the Australian Research Council via a Future Fellowship, under contract FT140100645. The authors thank S. Cholathan (Siam Elite Palm) for providing oil palm seeds.

Supplementary information

Supplementary figure S1. Decomposition of the 13C NMR signal. Supplementary figure S2. Metabolic model used for export calculation. Supplementary figure S3. Example of modelled labelling kinetics. Supplementary figure S4. 1H-NMR spectra of leaflet extracts. Supplementary figure S5. Detailed analysis of the 1.5-3.5 ppm region in leaflets. Supplementary figure S6. NMR identification of dopamine and tyramine. Supplementary figure S7. 1H-NMR spectra of rachis extracts. Supplementary figure S8. NMR identification of raffinose and myoinositol. Supplementary figure S9. Overview of 13C spectra in rachis. Supplementary figure S10. Metabolomics analysis of rachis. Supplementary figure S11. Summary of pathways affected by low K. Supplementary figure S12. Dopamine metabolism.

References

Abadie C, Lothier J, Boex-Fontvieille E, Carroll A, Tcherkez G. 2017. Direct assessment of the metabolic origin of carbon atoms in glutamate from illuminated leaves using 13C-NMR. New Phytologist 216: 1079-1089. Abadie C, Tcherkez G. 2019. In vivo phosphoenolpyruvate carboxylase activity is controlled by CO2 and O2 mole fractions and represents a major flux at high photorespiration rates. New Phytologist 221: 1843-1852. Armengaud P, Sulpice R, Miller AJ, Stitt M, Amtmann A, Gibon Y. 2009. Multilevel Analysis of Primary Metabolism Provides New Insights into the Role of Potassium Nutrition for Glycolysis and Nitrogen Assimilation in Arabidopsis Roots. Plant Physiology 150: 772-785. Bahramnejad B, Erickson LR, Goodwin PH. 2010. Induction of expression and increased susceptibility due to silencing a 4,5-DOPA dioxygenase extradiol-like gene of Nicotiana benthamiana in the interaction with the hemibiotrophic pathogens, Colletotrichum destructivum, Colletotrichum orbiculare or Pseudomonas syringae pv. tabaci. Plant Science 178: 147-157. Besford RT, Maw GA. 1976. Effect of Potassium Nutrition on Some Enzymes of the Tomato Plant. Annals of Botany 40: 461-471. Cakmak I, Hengeler C, Marschner H. 1994. Changes in phloem export of sucrose in leaves in response to phosphorus, potassium and magnesium deficiency in bean plants. Journal of Experimental Botany 45: 1251-1257. Calvez C. 1976. The influence on oil palm yield of frond pruning to different levels. Oleagineux 31: 53-58.

228 Corley RHV, Tinker PB. 2016. The Oil Palm, Fifth Edition. West Sussex, UK: John Wiley and Sons. Cui J, Abadie C, Carroll A, Lamade E, Tcherkez G. 2019a. Responses to K deficiency and waterlogging interact via respiratory and nitrogen metabolism. Plant Cell and Environment 42: 647-658. Cui J, Davanture M, Zivy M, Lamade E, Tcherkez G. 2019b. Metabolic responses to potassium availability and waterlogging reshape respiration and carbon use efficiency in oil palm. New Phytologist 223: 310-322. Cui J, Pottosin I, Lamade E, Tcherkez G. 2020. What is the role of putrescine accumulated under potassium deficiency? Plant Cell and Environment: In press. De Schepper V, De Swaef T, Bauweraerts I, Steppe K. 2013. Phloem transport: a review of mechanisms and controls. Journal of Experimental Botany 64: 4839- 4850. Doman DC, Geiger DR. 1979. Effect of Exogenously Supplied Foliar Potassium on Phloem Loading in Beta vulgaris L. Plant Physiology 64: 528-533. Dreyer I, Gomez-Porras JL, Riedelsberger J. 2017. The potassium battery: a mobile energy source for transport processes in plant vascular tissues. New Phytologist 216: 1049-1053. Dussert S, Guerin C, Andersson M, Joët T, Tranbarger TJ, Pizot M, Sarah G, Omore A, Durand-Gasselin T, Morcillo F. 2013. Comparative transcriptome analysis of three oil palm fruit and seed tissues that differ in oil content and fatty acid composition. Plant Physiology 162: 1337-1358. Epron D, Cabral OMR, Laclau J-P, Dannoura M, Packer AP, Plain C, Battie-Laclau 13 P, Moreira MZ, Trivelin PCO, Bouillet J-P, et al. 2015. In situ CO2 pulse labelling of field-grown eucalypt trees revealed the effects of potassium nutrition and throughfall exclusion on phloem transport of photosynthetic carbon. Tree Physiology 36: 6-21. Facchini PJ, Huber-Allanach KL, Tari LW. 2000. Plant aromatic L-amino acid decarboxylases: evolution, biochemistry, regulation, and metabolic engineering applications. Phytochemistry 54: 121-138. Freeman GG. 1967. Studies on potassium nutrition of plants. II.—Some effects of potassium deficiency on the organic acids of leaves. Journal of the Science of Food and Agriculture 18: 569-576. Geiger DR, Conti TR. 1983. Relation of increased potassium nutrition to photosynthesis and translocation of carbon. Plant Physiology 71: 141-146. Haeder H-E 1977. Effects of potassium on phloem loading and transport.In Cooke G. Fertilizer Use and Production of Carbohydrates and Lipids. York, UK: International Potash Institute. 115-121. Hartt CE. 1969. Effect of Potassium Deficiency Upon Translocation of 14C in Attached Blades and Entire Plants of Sugarcane. Plant Physiology 44: 1461-1466. Heffer P. 2009. Assessment of fertilizer use by crop at the global level. International Fertilizer Industry Association. Henson IE. 2002. Oil palm pruning and relationships between leaf area and yield-a review of previous experiments. The Planter 78: 351-362. Henson IE, Chang KC, Siti Nor Aishah M, Chai SH, Hasnuddin MY, Zakaria A. 1999. The oil palm trunk as a carbohydrate reserve. Journal of Oil Palm Research 11: 98-113. Houngbossa S, Bonnemain JL. 1985. Nature des nutriments organiques exportés par la feuille de palmier à huile (Elaeis guineensis Jacq.) : effet de l'acide éthylène

229 diamine tétraacétique sur l'exsudation. Bulletin de la Société Botanique de France 132: 15-23. Hussain SS, Ali M, Ahmad M, Siddique KHM. 2011. Polyamines: Natural and engineered abiotic and biotic stress tolerance in plants. Biotechnology Advances 29: 300-311. Jiao X, Li Y, Zhang X, Liu C, Liang W, Li C, Ma F, Li C. 2019. Exogenous dopamine application promotes alkali tolerance of apple seedlings. Plants 8: Article 580. Jones LH. 1961. Some effects of potassium deficiency on the metabolism of the tomato plant. Canadian Journal of Botany 39: 593-606. Jones LH. 1966. Carbon-14 studies of intermediary metabolism in potassium-deficient tomato plants. Canadian Journal of Botany 44: 297-307. Jwakyung S, Yeonkyu S, Yejin L, Seongsoo K, Sangkeun H, B. KH, Taek-Keun O. 2015. Compositional changes of selected amino acids, organic acids, and soluble sugars in the xylem sap of N, P, or K-deficient tomato plants. Journal of Plant Nutrition and Soil Science 178: 792-797. Khan MI, Giridhar P. 2015. Plant betalains: chemistry and biochemistry. Phytochemistry 117: 267-295. Kulma A, Szopa J. 2007. Catecholamines are active compounds in plants. Plant Science 172: 433-440. Lamade E, Ollivier J, Rozier-Abouab T, Gérardeaux E 2014. Occurrence of potassium location in oil palm tissues with reserve sugars: consequences for oil palm K status determination. 4th International Oil Palm Conference. Bali, Indonesia. 1-16. Lamade E, Setiyo IE, Girard S, Ghashghaie J. 2009. Changes in 13C/12C of oil palm leaves to understand carbon use during their passage from heterotrophy to autotrophy. Rapid Communications in Mass Spectrometry 23: 2586-2596. Lamade E, Tcherkez G, Darlan NH, Rodrigues RL, Fresneau C, Mauve C, Lamothe-Sibold M, Sketriené D, Ghashghaie J. 2016. Natural 13C distribution in oil palm (Elaeis guineensis Jacq.) and consequences for allocation pattern. Plant Cell and Environment 39: 199-212. Langella O, Valot B, Balliau T, Blein-Nicolas M, Bonhomme L, Zivy M. 2017. X! TandemPipeline: a tool to manage sequence redundancy for protein inference and phosphosite identification. Journal of proteome research 16: 494-503. Legros S, Mialet-Serra I, Caliman J-P, Siregar FA, Clément-Vidal A, Fabre D, Dingkuhn M. 2009a. Phenology, growth and physiological adjustments of oil palm (Elaeis guineensis) to sink limitation induced by fruit pruning. Annals of Botany 104: 1183-1194. Legros S, Mialet-Serra I, Clément-Vidal A, Caliman J-P, Siregar FA, Fabre D, Dingkuhn M. 2009b. Role of transitory carbon reserves during adjustment to climate variability and source–sink imbalances in oil palm (Elaeis guineensis). Tree Physiology 29: 1199-1211. Li C, Sun X, Chang C, Jia D, Wei Z, Li C, Ma F. 2015. Dopamine alleviates salt- induced stress in Malus hupehensis. Physiologia plantarum 153: 584-602. Li S-W, Leng Y, Shi R-F. 2017. Transcriptomic profiling provides molecular insights into hydrogen peroxide-induced adventitious rooting in mung bean seedlings. BMC Genomics 18: Article 188. Li S, Li T, Kim W-D, Kitaoka M, Yoshida S, Nakajima M, Kobayashi H. 2007. Characterization of raffinose synthase from rice (Oryza sativa L. var. Nipponbare). Biotechnology Letters 29: 635-640.

230 Liang B, Gao T, Zhao Q, Ma C, Chen Q, Wei Z, Li C, Li C, Ma F. 2018. Effects of exogenous dopamine on the uptake, transport, and resorption of apple ionome under moderate drought. Frontiers in plant science 9: Article 755. Ling L, Jiang Y, Meng JJ, Cai LM, Cao GC. 2018. Phloem transport capacity of transgenic rice T1c-19 (Cry1C*) under several potassium fertilizer levels. PloS one 13: Article e0195058. Martineau E, Domec J-C, Bosc A, Dannoura M, Gibon Y, Bénard C, Jordan-Meille L. 2017. The role of potassium on maize leaf carbon exportation under drought condition. Acta Physiologiae Plantarum 39: 219-224. Mengel K, Haeder H-E. 1977. Effect of Potassium Supply on the Rate of Phloem Sap Exudation and the Composition of Phloem Sap of Ricinus communis. Plant Physiology 59: 282-287. Mirande-Ney C, Tcherkez G, Gilard F, Ghashghaie J, Lamade E. 2019. Effects of potassium fertilization on oil palm fruit metabolism and mesocarp lipid accumulation. Journal of agricultural and food chemistry 67: 9432-9440. Okamoto S. 1966. Effect of mineral nutrition on metabolic change induced in crop plant roots (III). Soil Science and Plant Nutrition 12: 13-17. Okamoto S. 1967. Effects of potassium nutrition on the glycolysis and the krebs cycle in taro plants. Soil Science and Plant Nutrition 13: 143-150. Okamoto S. 1968. The respiration in the roots of broad bean and barley under a moderate potassium deficiency. Soil Science and Plant Nutrition 14: 175-182. Peel AJ, Rogers S. 1982. Stimulation of sugar loading into sieve elements of willow by potassium and sodium salts. Planta 154: 94-96. Peterbauer T, Mach L, Mucha J, Richter A. 2002. Functional expression of a cDNA encoding pea (Pisum sativum L.) raffinose synthase, partial purification of the enzyme from maturing seeds, and steady-state kinetic analysis of raffinose synthesis. Planta 215: 839-846. Rodrigues-Neto JC, Correia MV, Souto AL, Ribeiro JAdA, Vieira LR, Souza MT, Rodrigues CM, Abdelnur PV. 2018. Metabolic fingerprinting analysis of oil palm reveals a set of differentially expressed metabolites in fatal yellowing symptomatic and non-symptomatic plants. Metabolomics 14: Article 142. Sanusi NSNM, Rosli R, Halim MAA, Chan K-L, Nagappan J, Azizi N, Amiruddin N, Tatarinova TV, Low E-TL. 2018. PalmXplore: oil palm gene database. Database 18: Article 95. Skirycz A, Świędrych A, Szopa J. 2005. Expression of human dopamine receptor in potato (Solanum tuberosum) results in altered tuber carbon metabolism. BMC Plant Biology 5: Article 1. Tanaka Y, Sasaki N, Ohmiya A. 2008. Biosynthesis of plant pigments: anthocyanins, betalains and carotenoids. The Plant Journal 54: 733-749. Tcherkez G, Mahe A, Guerard F, Boex-Fontvieille E, Gout E, Lamothe M, Barbour MM, Bligny R. 2012. Short-term effects of CO2 and O2 on citrate metabolism in illuminated leaves. Plant Cell and Environment 35: 2208-2220. Tcherkez G, Nogués S, Bleton J, Cornic G, Badeck F, Ghashghaie J. 2003. Metabolic origin of carbon isotope composition of leaf dark-respired CO2 in French bean. Plant Physiology 131: 237-244. Thompson RG, Dale JE. 1981. Export of 14C- and 11C-labelled assimilate from wheat and maize leaves: effects of parachloromercurobenzylsulphonic acid and fusicoccin and of potassium deficiency. Canadian Journal of Botany 59: 2439- 2444.

231 Troufflard S, Mullen W, Larson TR, Graham IA, Crozier A, Amtmann A, Armengaud P. 2010. Potassium deficiency induces the biosynthesis of oxylipins and glucosinolates in Arabidopsis thaliana. BMC Plant Biology 10: Article 172. Waalkes TP, Sjoerdsma A, Creveling CR, Weissbach H, Udenfriend S. 1958. Serotonin, norepinephrine, and related compounds in bananas. Science 127: 648- 650. Wang N, Hua H, Egrinya Eneji A, Li Z, Duan L, Tian X. 2012. Genotypic variations in photosynthetic and physiological adjustment to potassium deficiency in cotton (Gossypium hirsutum). Journal of Photochemistry and Photobiology B: Biology 110: 1-8. Zhou S, Wei S, Boone B, Levy S. 2007. Microarray analysis of genes affected by salt stress in tomato. African Journal of Environmental Science and Technology 1: 14- 26.

232

Fig. 1. 1H-NMR identification of major sugars in oil palm leaves at 11.5 months at high K: (a) full NMR spectrum (leaflet) highlighting the sugar region in grey; (b) magnification of the 3.2-4.3 ppm region; (c) comparison between leaflet (blade, blue) and rachis (red) for the same amount of dry weight per sample. Abbreviations: 6PG,R, 6-phosphogluconate/D-ribose regions; Arg, arginine; F, D-fructose; g, galactose; G, D-glucose; GA, gallic acid; Suc, sucrose; VA, probable homovanillic acid (also with OCH3 group of O-methylglucose at 3.35 ppm). Note the overlapping of signals, for example between glucose and arginine at 3.25 ppm, sucrose and arginine at about 3.77 ppm, or sugars and vanillic acid at 3.9 ppm. A detailed analysis of the 1.5- 3.5 region is provided in Fig. S5. In (c), note the much larger abundance in glucose, and much lower abundance in vanillic acid in rachis. In both leaflet and rachis, other sugars such as galactinol and raffinose are at low concentration and are not readily visible by 1H-NMR.

233 (a) (b) 250 1200 High K Low K * 1000 200

800 DW) DW) -1 150 -1

600

100 * * * 400 Content (µmol g (µmol Content Content (µmol g (µmol Content

50 200 * * * * * * 0 0 * e e e e e e h e e e x e e h os os os in in in rc s s s s at in rc cr c ct m m gin ta cro co cto A al in ta u lu ru pa yra r S u lu ru M rg S S G F Do T A S G F A

Fig. 2. Content in major compounds of leaflets (a) and rachis (b) in old oil palm saplings at 11.5 months. Absolute quantitation of major compounds including sugars by 1H-NMR spectroscopy and amylose (insoluble starch) by gelatinization-flocculation (starch content given in hexose equivalents). ‘Asx’ stands for aspartate + asparagine, which have similar chemical shifts. Asterisks stand for significant differences between high K and low K (P<0.01). 1H-spectra are further illustrated in Figs. S4-7. This figure does not show gallic acid, a major compound visible by NMR (Fig. 1), as it is not significantly affected by K conditions.

234

Fig. 3. 13C-NMR spectrum of leaflets in oil palm saplings at 11.5 months, after 4 h 13 photosynthesis using CO2 at natural abundance (blue) or CO2 (red) or 4 h photosynthesis with 13 CO2 and then 2 h darkness (green), using oil palm cultivated under high K (a,c) or low K (b,d) conditions. The figure comprises full spectra (a,b) as well as insets corresponding to the 96-104 ppm region to show signal splitting by 13C-13C couplings. Abbreviations: G1β, C-1 of β-D- glucose; F2α and F2β, C-2 of α- and β-D-fructose, respectively; G, gallic acid; M, maleic acid (internal reference for quantitation); SF2, C-2 of the fructosyl moiety of sucrose; SS, soluble amylopectin (phytoglycogen).

235

(a) (b) (d) 180 2.0 Total assimilated Sucrose (glucosyl) (c) Sucrose Glucose + Fructose 160 Starch 100 Sucrose (fructosyl) 800 1.8 Soluble sugars Glucose Total Other Fructose 1.6 Calculated from mass-balance 140 80 600 DW) ) ) -1 -2 -1 1.4

120 400 s

60 -2 Content 1.2

100 g (µmol C 200

13 40 1.0

80 % 0 20 0.8 * 60 * C amount (mmol m (mmol C amount 0.6 Rate (µmol HE m 13 40 2 0.4

20 0.2 * 0 0 0.0 k High K Low K ight ark High K light High K dark Low K light Low K dark Light f Dar Light Light f af D eaf L L Lea Le K Lea achis achis h K igh K ig K R H Low H Low K igh K R ow K Rachis Dark H Low High K Rachis LDark

Fig. 4. 13C allocation pattern in leaves of oil palm saplings at 11.5 months. (a) Total 13C amount fixed by net photosynthesis and 13C amount in leaflet starch, soluble sugars (sucrose + glucose + fructose) and difference between assimilated 13C and 13C in starch and sugars (referred to as “others”). Amounts 13 -2 13 13 reported here are in mmol C m recovered after 4 h labelling in the light with CO2 (99% C), corrected for natural abundance. For each, there was a significant difference between low K and high K (P<0.01). (b) 13C-percentage in soluble sugars just after 4 h labelling (“light”) or 4 h light labelling + 2 h darkness (“dark”). The red horizontal dashed line represents natural abundance (1.1%). Glucosyl and fructosyl moieties of sucrose are differentiated with dotted-black and closed-black filling. (c) Total pool sizes determined by 1H-NMR in the light and in the dark. (d) Export rate from leaflets to rachis calculated using modelling (see Material and methods) for sucrose (black) and glucose + fructose (grey). The sum (total) is shown in dark grey. Total export rate was also calculated from isotopic mass-balance (“others” in panel (a) and average %13C from panel (b)). In (d), all values are in µmol hexose equivalents per m² per second. Asterisks stand for significant difference between low and high K conditions.

236

Fig. 5. Metabolomics pattern of oil palm leaflets: effect of K availability and developmental time. (a) Volcano plot combining univariate (–log(P-value)) and multivariate (loading pcorr) analyses to show most discriminating metabolites between low K and high K conditions. Metabolites that are significant (insignificant) for a K  time interaction affect appear in dark (light) blue. In this representation, pcorr is positive if metabolites increase under K deficiency (low K). The red dash-dotted line stands for the Bonferroni threshold. (b) Heatmap and hierarchical clustering (Pearson correlation) showing the relationship between metabolites significant (P<0.01) for the K effect (using a two-way ANOVA). In this representation, red (blue) stands for an increase (decrease) in the metabolite of interest. Most important groups (clusters) are framed and numbered (in black). (c) Time course of sucrose, raffinose (continuous) and galactinol (dashed), dopamine, and putrescine under low K (open symbols) and high K (closed symbols). The dash-dotted red line stands for the cultivation transition (thinning). Data shown are meanSD (n = 5). Sucrose is not significant for the effect of K availability but is significant for the effect of time. The four other metabolites appear in panel (b). Abbreviations: 2OL, metabolite of the 2- oxoglyconate family (probably 2-oxoallonate); Cit, citrate; CA, cis-aconitate; DAL, δ- aminolevulinate; F6P, fructose 6-phosphate; Fum, fumarate; G3P, glycerol-3-phosphate; Gol, galactitol; Ict, isocitrate; Mae, maleate; Mal, malate; Myo, myoinositol; Put, putrescine; TA, trans-aconitate.

237 Table 1. List of proteins involved in sugar metabolism that changed significantly in leaflets with K availability at 12 months (re-tabulated from (Cui et al., 2019b)). Some proteins appear several times in this list since several unique peptides of the same protein were found in our proteomics analysis. The Bonferroni threshold appears with a transverse dashed line. Here, the maximum loading (absolute value) to compare with is unity. The loading value is negative if the protein of interest decreases under low K.

Name Loading in –log (P) in Effect of low K OPLS analysis ANOVA Beta-galactosidase 8-like 0.79 9.98 UP Low quality protein (beta-galactosidase 6-like) 0.84 8.84 UP Beta-galactosidase 15 0.61 8.24 UP Alpha-mannosidase (At3g26720 homologue) 0.91 8.07 UP Triosephosphate isomerase, cytosolic 0.87 7.15 UP Endo-1,3;1,4-beta-D-glucanase -0.63 7.07 DOWN Triosephosphate isomerase, cytosolic 0.86 7.00 UP Fructose-bisphosphate aldolase 5, cytosolic -0.82 6.99 DOWN Alpha-1,4 glucan phosphorylase L isozyme, 0.77 5.96 UP chloroplastic/amyloplastic Fructose-bisphosphate aldolase 5, cytosolic -0.83 5.95 DOWN Mannose/glucose-specific lectin 0.69 5.88 UP Glucose-6-phosphate 1-dehydrogenase, 0.84 5.46 UP cytoplasmic Triosephosphate isomerase, cytosolic 0.80 4.91 UP Fructose-1,6-bisphosphatase, cytosolic -0.79 4.87 DOWN Pyrophosphate-fructose 6-phosphate 1- 0.56 4.61 UP phosphotransferase, subunit alpha ∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙Bonferroni threshold∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙ Granule-bound starch synthase 2, -0.70 4.50 DOWN chloroplastic/amyloplastic Nucleoside diphosphate kinase B 0.44 4.24 UP Fructose-bisphosphate aldolase 1, cytoplasmic 0.79 3.99 UP Alpha-mannosidase (At3g26720 homologue) 0.69 3.83 UP Galactinol-sucrose galactosyltransferase 2 -0.61 3.70 DOWN Xylose isomerase 0.74 3.56 UP Phosphoglucomutase 2, cytoplasmic 0.73 3.19 UP Sucrose-phosphate synthase 2 0.19 2.99 UP UDP-glycosyltransferase 83A1-like 0.43 2.99 UP UDP-glycosyltransferase 83A1-like 0.41 2.99 UP Glucan endo-1,3-beta-glucosidase 0.63 2.96 UP Glucose-6-phosphate isomerase, cytosolic 0.60 2.64 UP Alpha-mannosidase-like 0.66 2.55 UP Fructose-bisphosphate aldolase 1, cytoplasmic 0.59 2.53 UP Sucrose-phosphatase 2 0.42 2.15 UP Phloem protein 2-like A1 (sugar-binding phloem -0.37 2.07 DOWN lectin) ∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙Threshold P = 0.01∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙∙

238 Supplementary material

Fig. S1. Decomposition of the 13C signal that includes spin-spin interactions: example of the C-atom no. 2 of the fructosyl moiety of sucrose. Because of interactions with two neighbor 13C atoms, the signal does not form a singlet but a quintuplet, whereby integrals represent the abundance of isotopologues (mono-, bi- and tri-labelled). The central peak corresponds to mono- 13 labelled and natural abundance molecules ( C1) plus half the signal of tri-labelled isotopologues. The contribution of natural abundance (at 1.1% 13C) reflects non-labelled molecules, which come from incomplete turn-over and/or the presence of a “static”, isotopically non-active pool. Note 13 that in the rachis, labelled molecules are mostly tri-labelled ( C3), and the contribution of the static pool is very large. Should this pattern be the result of incomplete turn-over, bi-labelled molecules would have been more abundant than tri-labelled molecules (commencing turn-over) or alternatively, mono-labelled molecules would have been much less abundant than bi-labelled molecules (end of turn-over). Here, the fact that bi-labelled molecules are the smallest contributor to the signal means that observed sucrose is a mixture of a small pool of highly labelled (mostly tri-labelled) molecules, and a large pool of unlabeled molecules (static pool). The static pool in rachis simply comes from the contribution of other (non-labelled) leaflets, which export sugars at natural 13C abundance. In leaflets, nearly all molecules are tri-labelled and the contribution of mono-labelled molecules is very small, showing that the sucrose pool is at least 96% active (turned-over), therefore with a very small static pool. Numbers indicated here are proportions of total signal.

239 (a) Light

A Phosphorylated A/6 –S (A/6 –S)/2 13 Fructose 6‐phosphate Glucose 6‐phosphate CO2 intermediates

S (A/6 –S)/2 + C C Sucrose Sucrose C Starch (active) (inactive) Futile sucrose Glucose (active) C + Efg cycle

Glucose (inactive)

Fructose (active)

LEAFLET Fructose (inactive)

E E E RACHIS s fg fg Sucrose (active) Fructose (active) Glucose (active) Phloem flow to petiole Sucrose (inactive) Fructose (inactive) Glucose (inactive)

Phloem flow from other leaflets

(b) Darkness

Fructose 6‐phosphate (D + R)/2 Glucose 6‐phosphate R D (D –R)/2 + C Respiration C Sucrose Sucrose C Starch (active) (inactive) Futile sucrose

C + Efg Glucose (active) cycle

Glucose (inactive)

Fructose (active)

LEAFLET Fructose (inactive)

E E E RACHIS s fg fg Sucrose (active) Fructose (active) Glucose (active) Phloem flow to petiole Sucrose (inactive) Fructose (inactive) Glucose (inactive)

Phloem flow from other leaflets Fig. S2. Simplified metabolic pathway used to compute export rates from isotopic data. In the light (a), hexose phosphates are produced from “intermediates” such as triose phosphates, which are assumed to have a probability (frequency) of being fully 13C-labelled (i.e., all C-atom position labelled) of 1 (this is further apparent in Fig. S3 with the example of ribulose 1,5- 13 bisphosphate which is in its C5 form within 200 seconds only). A futile cycle around sucrose is assumed to take place, with a flux rate C. The export rate of fructose and glucose are assumed to be equal (Efg). Here, it is assumed that the system is in the steady state for pool sizes such that net hexose production A/6 – S equals total hexose export 2  (Es + Efg), that is, (A/6 – S)/2 = Es + Efg. In darkness (b), hexose phosphates come from starch degradation (with rate D). An additional flux of fructose 6-phosphate consumption by dark respiration (R) is considered (day respiration is accounted for in the light since A is a net flux). In the steady state, we have Es + Efg = (D – R)/2. The rate of dark respiration was from (Cui et al., 2019b) (converted in µmol hexoses g-1 FW). Note that futile sucrose cycling improves homogenization between sugars (sucrose, glucose and fructose) but slows down turn-over and thus 13C appearance in pools, since it dilutes isotopically hexoses using non-labelled sugars. In this figure, “active” and “inactive” refer to the ability/inability to be labelled. Calculations done here neglect the slow exchange of molecules between active and inactive pools.

240 Fig. S3. Example of labelling kinetics with a simple kinetic model. (a) Modelled ribulose-1,5-bisphosphate (RuBP) isotopologue frequency. Note that due to the relatively small RuBP pool ( 30 µmol m , i.e., 0.17 µmol g FW) in leaves, the C5 species (fully labelled) is the prevalent form after 200 s. ≈ -2 -1 13 Here, it is assumed that RuBP atoms are renewed statistically, regardless of the pathway represented by the Calvin cycle. (b) and (c) Frequency of fully labelled sugar species ( C12-sucrose and C6-hexoses) with time (in seconds) in leaflets (b) and rachis (c) ‘active’ pools, using pool sizes, photosynthesis 13 13 and allocation to starch determined experimentally, and the model depicted in Fig. S2. This example corresponds to high K conditions in the light. Here the model was parameterized with C (sucrose futile cycle activity) equal to zero.

241 Fig S4. Comparison of NMR spectra in leaflet extracts from oil palm cultivated at high (blue) and low (red) K. The difference between spectra appears in green. Abbreviations: Arg, arginine; C, citrate; D, dopamine; M, malate; Min, minaline (pyrrolidine 2-carboxylic acid); PM, 2- pyrrolidine methanol (prolinol); S, sugars ; T, tyramine; VA, homovanillic acid. Note the very large content in gallic acid (at 7.05 ppm), also shown in Fig. 1 (main text).

242 Fig. S5. Detailed analysis of the 1.5-3.5 ppm region of the 1H-NMR spectrum, showing changes in secondary metabolites (tyramine, dopamine, N-acetylputrescine, prolinol) in leaves under low K conditions. Top, annotated spectrum; Bottom, comparison with standards (same colors as in bullet points). Note the relatively low amount of free putrescine under low K, partly because it is conjugated to other compounds (for example in the form of N-acetylputrescine, NAP).

243 Fig. S6. NMR identification of dopamine and tyramine, showing a leaf sample (blue) and authentic standards of tyramine (green) and dopamine (red). Chemical groups (carbon no.) are indicated with numbers.

244

Fig S7. Comparison of NMR spectra in rachis extracts from oil palm cultivated at high (blue) and low (red) K. The difference between spectra appears in green. Inset: magnification of the 7.0 ppm region showing the appearance of tyramine and dopamine at low K (black arrowheads) and spectra of authentic standards (orange, green). Abbreviations: Arg, arginine; Asp, aspartate + asparagine; C, citrate; G, glucose; GA, gallic acid; M, malate; PHB, probable parahydroxybenzoate or gentisic acid derivative; PM, pyrrolidine methanol (prolinol); S, sucrose.

245

Fig S8. Comparison of the 1H-NMR spectrum of leaflet extracts with raffinose and myoinositol standards. Peaks that can be possibly attributed to raffinose (green) and myoinositol (red) are labelled with arrows. Note that (i) such peaks are very small, and (ii) some uncertainty remains since several peaks do not show up (question marks). This clearly demonstrate that raffinose and myoinositol are quantitatively minor compounds hardly visible by NMR.

246

13 13 Fig. S9. Overview of C-spectra in oil palm rachis after labelling with CO2: control at 13 natural abundance (gas exchange for 4 h with natural CO2 at 1.1% C; blue), 4 h labelling with 13 13 CO2 (red) and 4 h labelling with CO2 and then 2 h in the dark (green). Overview of full spectra under high K (a) and low K (b) conditions. Detailed view of the C-1 and C-6 sugar region (60-64 ppm) under high K (c) and low K (d) conditions.. 13C-13C couplings are shown with black arrowheads. C-atom position identification: F, fructose; G, glucose; SF, fructosyl moiety of sucrose; SG, glucosyl moiety of sucrose. Numbers refer to atom no. Note the virtually invisible difference between unlabeled and labelled situations (a-b), showing that the contribution of current photosynthates to rachis sugar pools is small and thus that the export rate is likely rather low. However, leaflet-exported labelled sugars are visible thanks to 13C-13C couplings, in particular in sucrose at high K.

247

Fig. S10. Metabolomics pattern of oil palm rachis: effect of K availability and developmental time. (a) Volcano plot combining univariate (–log(P-value)) and multivariate (loading pcorr) analyses to show most discriminating metabolites between low K and high K conditions. Metabolites that are significant (insignificant) for a K  time interaction affect appear in dark (light) blue. In this representation, pcorr is positive if metabolites increase under K deficiency (low K). The red dash-dotted line stands for the Bonferroni threshold. (b) Heatmap and hierarchical clustering (Pearson correlation) showing the relationship between metabolites significant (P<0.01) for the K effect (using a two-way ANOVA). In this representation, red (blue) stands for an increase (decrease) in the metabolite of interest. Most important groups (clusters) are framed and numbered (in black). (c) Time course of sucrose, raffinose (continuous) and galactinol (dashed), dopamine, and putrescine under low K (open symbols) and high K (closed symbols). The dash-dotted red line stands for the cultivation transition (thinning). Data shown are meanSD (n = 5). Sucrose and galactinol are not significant for the effect of K availability. The three other metabolites appear in panel (b). Abbreviations: 2OL, metabolite of the 2-oxoglyconate family (probably 2-oxoallonate); 2OU, 2-oxogluconate; Bsit, β-sitosterol; DHB, dihydroxybutanoate; Fum, fumarate; GABA, γ-aminobutyrate; Gla, glycerate; Gly, glycine; Gol, glycerol; Mae, maleate; Myo, myoinositol; Orn, ornithine; PGA, 3-phosphoglycerate; Put, putrescine; Pyro, pyroglutamate; Ser, serine; Shk, shikimate; Succ, succinate; Tha, threonate.

248

Raffinose Galactinol Galactitol Gal export Glc 2‐oxoglyconate Lac Pentoses UDP‐Gal 6‐phosphogluconate OPPP Arg Myoinositol Pentitols Putrescine Sucrose Hexose phosphates CCB cycle Orn G6P, F6P Glycerol‐3‐phosphate Hexose Triose phosphates Triose phosphates E4P phosphates Glu Pro P2C Prolinol G6P, F6P glycolysis shikimate‐ arogenate pathway PK Acetyl‐CoA Pyruvate PEP PEP

Aco? Parapyruvate Tyrosine Tyrosine Tricarboxylic STK Starch acid pathway Citramalate Citrate C5 branched L‐DOPA 4‐hydroxyphenylpyruvate 2OG acid pathway Tyramine Malate Fumarate Dopamine α‐tocopherol Homovanillate Homogentisate

CHLOROPLAST Acetoacetate

Fig. S11. Tentative summary of alternative metabolic pathways induced by low K in oil palm leaves. In red, metabolites that are more abundant at low K, in blue, metabolites less abundant at low K. Five most important effects of low K are shown here: increase in polyamines metabolism (putrescine), alteration of sugar export leading to starch, raffinose and myoinositol accumulation; reconfiguration of the oxidative pentose phosphate pathway; changes in catabolism alleviating the inhibition of pyruvate kinase and Krebs cycle enzymes (aconitase, succinate thiokinase); and stimulation of tyrosine metabolism, which consumes PEP. Abbreviations: Aco, aconitase; Arg, arginine; L-DOPA, L-3,4-dihydroxyphenylalanine; F6P, fructose 6-phosphate; G6P, glucose 6- phosphate; Gal, galactose; Glu, glutamate; Lac, lactose; Orn, ornithine; P2C, pyrrolidine-2- carboxylate; PEP, phosphoenolpyruvate; Pi, inorganic phosphate; PK, pyruvate kinase; Pro, proline; STK, succinate thiokinase. Pathways that are certainly up-regulated (as shown by protein and metabolite levels) appear in red.

249

O OH OH OH Tyrosine O NH2 OH Polyphenol/catechol oxidase like L‐DOPA p5.00_sc00052_p0001 NH2 Tyrosine/DOPA decarboxylase 2 like OH p5.00_sc00102_p0003 p5.00_sc00154_p0001 p5.00_sc02648_p0002 p5.00_sc00154_p0003 p5.00_sc02783_p0001 Tyrosine decarboxylase 1 like p5.00_sc11175_p0001 p5.00_sc00231_p0009 p5.00_sc11267_p0001

NH2 Tyramine OH Tyrosine/DOPA decarboxylase 2 like p5.00_sc00154_p0001 Polyphenol/catechol oxidase like p5.00_sc00154_p0003 p5.00_sc00052_p0001 p5.00_sc00154_p0005 p5.00_sc00102_p0003 p5.00_sc00175_p0025 Dopamine β‐oxygenase p5.00_sc02648_p0002 Tyrosine decarboxylase 1 like p5.00_sc00137_p0012 p5.00_sc02783_p0001 p5.00_sc00231_p0009 p5.00_sc00363_p0003 p5.00_sc11175_p0001 p5.00_sc00621_p0004 p5.00_sc11267_p0001 Norcoclaurine synthase like p5.00_sc00154_p0052 NH2 p5.00_sc00696_p0005 Norcoclaurine 6‐O‐methyltransferase like OH Dopamine p5.00_sc00004_p0313 p5.00_sc00004_p0314 OH Coclaurine N‐methyltransferase like p5.00_sc00001_p0037 Salutaridine reductase like Homogentisate dioxygenase like 3,4 DOPA dioxygenase extradiol‐like p5.00_sc00187_p0029 p5.00_sc00001_p0545 p5.00_sc00051_p0111 p5.00_sc00187_p0031 p5.00_sc00213_p0025 p5.00_sc00092_p0039 Acylpyruvase p5.00_sc00009_p0063 Fumarylacetoacetase p5.00_sc00096_p0075 Betalamic acid derivatives Thebaine family alkaloids Aldehyde dehydrogenase family (many genes) ?

Catabolism to organic acids

Fig. S12. Simplified pathways of dopamine utilization by metabolism, showing gene ID (accession) numbers associated with enzymes in the oil palm genome (gene IDs retrieved from the oil palm genome database on-line palmxplore.mpob.gov.my). O-methyltransferases (catabolism and alkaloid synthesis) are not mentioned here (many genes) for simplificy. 3,4 DOPA dioxygenase extradiol is a typical enzyme of the betalain pathway (synthesis of betalamic acid) and since oil palm does not produce betalains, it is more likely that it acts as an intradiol dioxygenase, maybe forming pyrimidine alkaloids (question mark). This figure mentions all of the genes IDs that could be retrieved for the enzyme activities of interest, but note that some of them located in the same scaffold and physically very close on the chromosome might correspond to the same gene in reality (for example, tyrosine/DOPA decarboxylase 2 like with three very close sequences p5.00_s00154_p0001, p5.00_s00154_p0003, p5.00_s00154_p0005) (Sanusi et al., 2018).

250