Are Offspring-Heteromorphic Species Hedging Their Bets?

Item Type text; Electronic Dissertation

Authors Scholl, Joshua

Publisher The University of Arizona.

Rights Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction, presentation (such as public display or performance) of protected items is prohibited except with permission of the author.

Download date 24/09/2021 16:33:15

Link to Item http://hdl.handle.net/10150/642172 1

ARE OFFSPRING-HETEROMORPHIC SPECIES HEDGING THEIR

BETS?

Joshua P. Scholl

A Dissertation Submitted to the Faculty of the

DEPARTMENT OF ECOLOGY AND EVOLUTIONARY BIOLOGY

In Partial Fulfillment of the Requirements

For the Degree of

DOCTOR OF PHILOSOPHY

In the Graduate College

THE UNIVERSITY OF ARIZONA

2020

THE UNIVERSITY OF ARIZONA GRADUATE COLLEGE

As members of the Dissertation Committee, we certify that we have read the dissertation prepared by: Joshua Scholl, titled: “Are Offspring Heteromorphic Species Hedging Their Bets?” and recommend that it be accepted as fulfilling the dissertation requirement for the Degree of Doctor of Philosophy.

Jul 10, 2020 Date: D. Lawrence Venable May 28, 2020 Date: Judith Becerra May 29, 2020 Date: Katrina Dlugosch May 29, 2020 Date: Michael Sanderson

Final approval and acceptance of this dissertation is contingent upon the candidate’s submission of the final copies of the dissertation to the Graduate College.

I hereby certify that I have read this dissertation prepared under my direction and recommend that it be accepted as fulfilling the dissertation requirement.

Jul 10, 2020 Date: D. Lawrence Venable Dissertation Committee Chair Department of Ecology & Evolutionary Biology

ACKNOWLEDGEMENTS PAGE

I want to thank Dr. Larry Venable for being an exemplary advisor. His passion for natural history made my dissertation possible and growth as a scientist enjoyable. Additionally, I would like to thank my committee members Dr. Katrina Dlugosch, Dr. Michael Sanderson, Dr. Judith Becerra. Dr. Sanderson provided valuable feedback on the comparative biology component of my dissertation and helped improve my phylogenetic analyses and tree thinking in general. Dr. Becerra was particularly quick to respond to my inquires and chapter drafts and thus a valuable resource in their improvement. Dr. Dlugosch’s impeccable command of evolutionary biology and ecology always provided new and helpful insights across all facets of my graduate work. Dr. Judith Bronstein served as an invaluable mentor without whom this dissertation would not have been possible. Her passion for ecology, mentorship and teaching provided motivation and helped me to persevere. Together, Dr. Bronstein and Dr. Venable served as my most important mentors and role models. I would also like to thank Dr. John Wiens whose passion for herpetology and impeccable knowledge of comparative biology elevated the quality of my graduate teaching experience and dissertation, respectively. Dr. Evelyn Frazier, my undergraduate thesis advisor, introduced me to ecology and continues to offer support and discussion to this day. Dr. Robert Lubarsky has always inspired me to consider mathematical approaches to the worlds many puzzles and has served as an important academic mentor since my freshman undergraduate year. I would like to thank my lab mates, Monica (Xing-Yue) Ge and Ursula Basinger who I am grateful to also call my friends. From jointly working on exciting science projects to hiking the Grand Canyon, I will always be indebted to Monica and Ursula for helping to make my graduate experience an almost entirely positive one. Additionally, I am indebted to the many field and lab undergraduate researchers that made this work possible. Specifically, I would like to thank Kayla Cuestas, Bethany Farrah, and Gabriel Gudenkauf who dedicated countless hours to processing seeds and making the lab and greenhouse a productive and fun environment. My family and close friends provided invaluable emotional support. In particularly my mother Karin Scholl, sisters Chalena and Chiara Scholl and late father Udo Scholl were always supportive of my ecological ventures. My parents encouraged my curiosity of the natural world and instilled in me the discipline and perseverance to pursue my passion professionally. My sisters always provided fun conversations and served as valuable syntax reviewers of this dissertation. I would also like to thank my friends, especially Adam Chen, Ariel Zeiger, Roman Vega, Sid Sirsi, Michiel Pillet, Shannon McWaters, Alyssa Smith, John Fallon, Alison Harrington, Leonardo Calle, Gordon Smith, Anton, and Nick Kortessis for keeping me sane and happy. I am especially thankful to Candle Pfefferle. She helped me collect and process my data and review my manuscripts, award, and job applications. In addition, Candle has been my most valued travel companion, roommate, and a sanctuary from the stresses of the academic world. I also thank my other friends and faculty members in Arizona and Florida, and the undergraduate and graduate students who enriched my life and graduate school in many ways. I would also like to thank my funding sources including the SEEDS Program, National Science Foundation for a Graduate Research Fellowship and Doctoral Dissertation Improvement Grant, Arizona Native Plant Society, University of Arizona Graduate and Professional Student Council, University of Arizona Galileo Circle Scholarship, William G. McGinnies Graduate Scholarship in Arid Lands Studies, and Ecological Society of America Research Grants and Travel Awards without whom this research and its dissemination would not have been possible.

TABLE OF CONTENTS

ABSTRACT………………………………………………………………..………………...…5

I. INTRODUCTION……………………………………………………..…………………...... 8

II. PRESENT STUDY………………………………………...……………….…..…….……16

III. REFERENCES……………………………………………….………………………...... 22

APPENDIX A: Offspring polymorphism and bet hedging: a large-scale phylogenetic analysis...……………………………………………………………….…………….……...... 26

APPENDIX B: Bet hedging and plasticity: an integrated strategy in a variable environment…………………………………..……………………………………………….66

APPENDIX C: Fitness consequences of heteromorphic seed types in a variable environment ...………………………………………………………………………………...123

ABSTRACT

Most organisms have canalized the production of their offspring to one form or shape which is

presumably best adapted for the environment they inhabit. Nevertheless, at least 500 species of

plants, among other organisms, defy this general pattern and instead produce multiple,

morphologically different seeds simultaneously. Researchers have proposed numerous drivers of

this phenomenon. Theoretical work strongly suggests bet hedging as the major underlying mechanism and this has become a widely accepted but empirically untested assumption. At its core, seed heteromorphism can be attributed to only three underlying mechanisms or some

integrated strategy among them: bet hedging, plasticity, and adaptive tracking. In this thesis I

focus on empirically testing the long-standing assumption that seed heteromorphism serves as a

bet hedging strategy.

First, I take a broad perspective in summarizing the occurrence of seed heteromorphism in

southwestern North America. Then I use this dataset to phylogenetically evaluate seed

heteromorphism as a bet-hedging strategy across an unprecedented 96 seed-heteromorphic

species from 51 genera and 9 angiosperm families. In so doing, I evaluate the pervasive, and to

date empirically untested, assumption that the occurrence of seed heteromorphism is spatially

associated with environmental unpredictability as measured, quantitatively, by aridity. This is the

first study to statistically evaluate large-scale evidence for bet hedging among seed- heteromorphic species.

Second, I narrow my focus and use field and experimental work to evaluate seed heteromorphism as a bet hedging strategy across multiple populations in a particularly tractable winter, desert annual species, Pectocarya heterocarpa (I.M. Johnston) I.M. Johnston

(Boraginaceae). This species produces visually distinct and spatially separated seed morphs which allow for the assessment of phenotypic expression in reproductive investment and facilitate experimental manipulation. I couple my field collections with greenhouse studies to evaluate the role of plasticity in the expression of seed heteromorphism. I demonstrate that P. heterocarpa hedges its bets by plastically adjusting its ratio of low risk to high risk seed morphs across the aridity gradient in the direction predicted by bet hedging theory. I also show that seed germination fraction across the aridity gradient follows patterns expected from bet hedging with seeds from more variable sites displaying increased dormancy.

Third, I focus even more locally on just one population of P. heterocarpa to quantify the fitness consequences of its different seed morphs across years, environmental conditions, and germination cohorts. Using a field and greenhouse experiment I evaluate how the natural germination timing of P. heterocarpa seed morphs translate to fitness differences across different rainfall regimes and years. Overall, I find evidence for bet hedging in that seed types never had significant average fitness differences across treatments, years, or germination cohorts, but instead exhibited complex interactions with these factors such that each seed type had highest fitness under different combinations of conditions. My results demonstrate that the different seed morphs of P. heterocarpa, by virtue of these fitness differences, act to reduce variance in fitness both within and across years as predicted by bet hedging theory.

The assumption that offspring heteromorphism serves as a bet hedging strategy is pervasive in the literature despite the lack of rigorous empirical tests. In this dissertation I combine large- and narrow-scale analyses to present a rigorous analysis of bet hedging among offspring- polymorphic species. Overall, this work helps us gain a better understanding of how organisms cope with change, which is crucial for gaining any predictive power for population dynamics amidst global climate change.

I. INTRODUCTION

Most organisms have canalized the production of their offspring to one form or shape which is presumably best adapted for their environment. Nevertheless, some species defy this general pattern and instead produce multiple, morphologically different offspring simultaneously.

Known as offspring polymorphism or offspring heteromorphism, this strategy is observed across the tree of life. Several freshwater fish species are known to produce dimorphic offspring, possibly as a bet-hedging strategy to cope with unpredictable stream sizes (Koops et al. 2003;

Gregersen et al. 2009). Soil mites (Crean & Marshall 2009), grasshoppers (Caesar et al. 2007), and frogs (Lips 2001; Dziminski & Alford 2005) also appear to hedge their bets in unpredictable conditions by producing polymorphic offspring. The strategy is best documented in plants where it is commonly termed seed heteromorphism (Venable 1985b) and defined as the production of seeds by a single individual that differ in morphology and ecology (Venable 1985b; Mandak

1997; Imbert 2002). This reproductive strategy is observed in plants on all continents except

Antarctica (Wang et al. 2010). Its occurrence among plants is estimated to be around 500 species but is probably much higher as it has yet to be explored in many regions of the globe and cases often go unrecognized in regions in which it has been explored (Imbert 2002; Mandak & Pysek

2005; Wang et al. 2010). Furthermore, it can be expressed by either discretely different or continuously varying seed morphs. The two or more seed morphs may also vary in size, dormancy or dispersal, among other ecological characteristics.

Seed heteromorphism is most often qualitatively associated with arid regions (e.g. deserts) and many studies have attributed this to the lack of predictable soil moisture due to stochastic

precipitation patterns (e.g. Venable 1985a, b; Venable et al. 1987; Mandak 1997; Sadeh et al.

2009; Wang et al. 2010). In addition, seed heteromorphism is observed in pioneer or weedy

species that occur outside of deserts (Mandak 1997; Porras & Muñoz 2000; Imbert 2002; Wang

et al. 2010). Despite numerous observations about the association between unpredictable

environments and seed heteromorphism, and a large body of theory that investigates this

relationship, the hypothesis that seed heteromorphism represents a bet-hedging strategy that is adaptive in unpredictably fluctuating environments has yet to be statistically tested as the primary mechanism explaining this widespread phenotype across diverse plant genera. In addition, very few studies have made rigorous assessments of bet hedging in individual seed-

heteromorphic species (Venable 1985a; Venable et al. 1995).

In broad-scale analyses of seed heteromorphic plants, authors have addressed an underlying bet

hedging hypothesis and found some

support for it (Harper 1977; Mandak

1997; Cruz-Mazo et al. 2009; Wang

et al. 2010), however, a study of

194 heteromorphic and

homomorphic species by Imbert

(2002) could not discern an

association between environmental

unpredictability and seed

heteromorphism as expected by a bet hedging hypothesis. Furthermore, no one has analyzed the ecological occurrence of seed

heteromorphism across genera and in a proper phylogenetic framework to ensure sample

independence.

Evaluations of bet hedging among specific seed-heteromorphic species have demonstrated that

different seed morphs have unique ecological consequences which have been shown to translate

to fitness differences (e.g. Venable & Levin 1985a; Baskin et al. 2014). Nevertheless, a

quantitative explanation of the ratios of seed morphs these species produce and how they relate

to environmental variability, such as variability in rainfall, is still nascent (Cheplick & Quinn

1983; Venable 1985a; Sadeh et al. 2009).

Furthermore, the integration between plasticity and bet-hedging, in general, and among seed-

heteromorphic species, is poorly studied (Sadeh et al. 2009; Simons 2014). The few studies that

have investigated it found a large contribution of plasticity among hypothesized bet hedging

strategies suggesting its importance and ubiquitous nature among organisms in variable

environments (Sadeh et al. 2009; Donaldson-Matasci et al. 2013; Gremer et al. 2016). For

example, in investigations of 12 Sonoran Desert winter annual species, Gremer et al. (2014)

found strong evidence for bet hedging with respect to delayed germination. In a follow-up study

Gremer et al. (2016) found that the same species generally, but not always, maximized their

long-term fitness through an integrated strategy of bet hedging and predictive germination

(plasticity). Among seed-heteromorphic species, Sadeh et al. (2009) found strong evidence of a

bet hedging mechanism fine-tuned by plasticity. Interestingly, and somewhat paradoxically, they

found that the degree of plasticity in the adjustment of seed morph ratios was positively

associated with increased environmental variability.

Thus, despite theoretical predictions that seed heteromorphism is a form of bet-hedging (Venable

1985b; Starrfelt & Kokko 2012) and should occur in unpredictable environments, sound empirical evidence of this claim is lacking. Furthermore, the integration of bet hedging with plasticity is poorly studied despite its apparent prevalence in nature (Sadeh et al. 2009; Gremer et

al. 2016).

Bet hedging

Akin to “not putting all of your eggs in one basket”, biological bet hedging is a counterintuitive

strategy wherein organisms spread risk over time by increasing long-term fitness at the expense

of immediate fitness. This allows organisms to buffer against environmental variation through

time and it can

reduce extinction

risk (Cohen 1966;

Seger &

Brockmann 1987;

Philippi & Seger

1989; Starrfelt &

Kokko 2012).

Mathematically, bet

hedging can be

explained in terms

of geometric mean fitness. Population growth is determined by multiplying the yearly population growth rate across

a sequence of years and thus the average population growth rate is given by the geometric mean

of these individual yearly growth rates, i.e., the product of n annual growth rates taken to the nth

( ) root ( 1 , Stearns 2000). An interesting property of the geometric mean is that an annual 𝑛𝑛 𝑛𝑛 𝑖𝑖=1 𝑖𝑖 plant which∏ 𝜆𝜆manages to consistently produce three offspring per year can have higher long term

fitness than one whose offspring number averages above three but varies from year to year

because temporal variance in offspring production directly reduces geometric mean fitness

unless there is a compensating increase in arithmetic mean fitness (Fig. 1).

The key to rigorously testing for a bet-hedging strategy lies in quantifying geometric and arithmetic mean fitness and demonstrating a tradeoff between them (Simons 2011; Starrfelt &

Kokko 2012). For seed-heteromorphic species this translates to quantifying the fitness

consequences of different seed morphs over time and in different environmental conditions. Bet

hedging is demonstrated if different seed morphs are found to have different combinations of

years and environmental conditions in which they do best. Thus, in any given condition there is

always a mismatch between one seed type and the environment but viewed collectively, the

contributions of seed types minimize fitness variance across all environments. This maximizes

geometric mean fitness at the cost of arithmetic mean fitness.

The southwestern region of North America provides a prime opportunity for the large-scale exploration of seed heteromorphism and its function as a bet hedging strategy because it includes a vast matrix of arid, semi-arid and mesic biomes. In particular, the region includes the Sonoran,

Chihuahuan, Great Basin and Mojave deserts. These differ widely in their environmental conditions but share the characteristic that rainfall is generally low, variable and unpredictable

(Davidowitz 2002; Reynolds et al. 2004), prime conditions for the evolution and maintenance of bet hedging strategies (Venable 1985b). To evaluate evidence for an association between seed heteromorphism and environmental unpredictability as expected under a bet hedging hypothesis,

I first compiled an extensive dataset of seed-heteromorphic species in the region by reviewing floras spanning California, Arizona, New Mexico, northern Mexico and surrounding areas. I then analyzed this dataset in a phylogenetic context to statistically evaluate the association between seed heteromorphism and unpredictable habitats.

Moreover, the seed heteromorphic, desert annual species, Pectocarya heterocarpa (I.M.

Johnston; Boraginaceae), is well suited for more detailed within species studies of bet hedging as

the underlying

mechanism of

seed

heteromorphism.

This candidate

species is

favorable in that

it produces fruits

of exactly four heteromorphic seeds, greatly facilitating seed counting and allowing for high sample sizes (Fig.

2). Fruits can be further divided into basal and aerial fruits stemming from cleistogamous (not opening) and chasmogamous (opening) flowers, respectively. The basal and aerial seeds are easily distinguished both visually and spatially, further facilitating seed counts (Fig. 2). The

species is also widespread spanning California, Arizona, New Mexico, and northern Mexico

which facilitates collections across an extensive aridity gradient. Finally, studies of other plants

in the genus Pectocarya (Karron 1989) and preliminary work with P. heterocarpa (Felker,

unpublished data) indicates both self-compatibility and high self-fertilization rates. This is important for propagation of maternal lines for use in experimental work. In addition, we have found no indication of strong seed predation pressure acting on P. heterocarpa. Consequently, variability in water availability is likely the main driver of seed heteromorphism.

Evaluating bet hedging in P. heterocarpa was the focus of my second and third chapter. I evaluated the degree to which bet hedging and plasticity function as an integrated strategy in chapter two. To accomplish this, I collected P. heterocarpa from twelve sites across an aridity gradient spanning southeastern Arizona to southwestern California, to address bet hedging predictions regarding the phenotypic expression of its seed morph ratios with respect to environmental unpredictability (aridity). In this study I also quantitatively assessed germination characteristics across the collection gradient and the role of plasticity in fine-tuning the candidate bet hedging mechanism of P. heterocarpa. In my third chapter, I examined the fitness consequences of different seed types of P. heterocarpa across varying levels of water availability, germination cohorts and years. I conducted this experiment in a natural setting in the field and complemented it with a similar but more controlled experiment in the greenhouse.

My dissertation combines large-scale comparative analyses with individual and population level analyses to provide a comprehensive evaluation of the hypothesis that seed-heteromorphic species are hedging their bets. Ultimately, through this work I strive to understand organismal

15 population dynamics in the face of global climate change and associated increases in environmental aridity and consequently variability.

II. PRESENT STUDY

Below is a summary of the three chapters included in this dissertation and a description of how

they relate to one another. Each chapter includes an abstract, introduction, methods, results,

discussion, and references that are presented in the appendices following this introduction.

CHAPTER I (APPENDIX A): OFFSPRING POLYMORPHISM AND BET HEDGING: A LARGE-SCALE

PHYLOGENETIC ANALYSIS

The assumption that seed heteromorphism serves as a bet hedging strategy in stochastically

varying rainfall environments, such as deserts, is pervasive, yet lacks rigorous, large-scale empirical tests in which shared evolutionary relationships are statistically accounted for. While many fascinating natural history accounts exist of seed heteromorphism and its tendency to be associated with unpredictable environments like deserts (e.g. Zohary 1962), few rigorous empirical tests exist. Furthermore, several reviews do an excellent job of summarizing this bizarre plant reproductive strategy across the globe but fall short in providing summaries without accounting for phylogenetic relationships between species (Mandak 1997; Imbert 2002; Wang et al. 2010). By not accounting for the lack of independence among seed-heteromorphic species

these reviews are unable to statistically evaluate their findings (Felsenstein 1985; Imbert 2002).

The two studies that have incorporated phylogenies in their evaluations of seed heteromorphism

do so for only one genus each (Imbert 2002; Cruz-Mazo et al. 2009). Of these two studies,

Imbert (2002) finds no evidence to support a bet hedging hypothesis in the genus Crepis while

Cruz-Mazo et al. (2009) find strong evidence of the same hypothesis in the genus Scorzoneroides

(Asteraceae). Nonetheless, both studies use only qualitative measures of environmental

17 unpredictability including descriptions such as “desert,” “semi-desert,” “Mediterranean” and

“montane” (Imbert 2002; Cruz-Mazo et al. 2009). In fact, such qualitative approaches to habitat evaluation in the testing of bet hedging among a group of seed-heteromorphic species dominate the literature (Mandak 1997; Imbert 2002; Cruz-Mazo et al. 2009; Wang et al. 2010). Yet, the assumption that such broad biome descriptors are related to increased environmental unpredictability is questionable. Davidowitz (2002) explicitly tests the assumption that variability in precipitation increases from mesic to xeric biomes and finds little empirical support for this generalization.

In my first chapter I conduct the first review of seed heteromorphism in southwestern North

America and leverage the resulting dataset to test the pervasive hypothesis that seed heteromorphism serves as an underlying bet hedging mechanism. This chapter contributes to the scarce but increasingly important literature on empirical assessments of bet hedging. I focus specifically on employing a large, taxonomically diverse dataset on seed heteromorphism, quantitative measures of environmental variability, and evaluating my hypotheses statistically in a phylogenetic context. This work provides a broad scale test of the hypothesis that seed heteromorphism acts as a bet hedging strategy. In the ensuing chapters, I focus my efforts on fine-scale tests of this assumption by focusing on a particular seed-heteromorphic species and evaluating bet hedging predictions regarding its population differentiation in the expression of seed heteromorphism across space (Chapter 2) and fitness consequences across environmental conditions in time (Chapter 3).

CHAPTER II (APPENDIX B): BET HEDGING AND PLASTICITY: AN INTEGRATED

STRATEGY IN A VARIABLE ENVVIRONMENT

In chapter one I found large-scale evidence of an association between seed-heteromorphic species and unpredictable environments as measured by aridity. While this was an important first step, analyses at such large scales can provide only correlational evidence for bet hedging

(Mandak 1997; Simons 2011). Demonstrating a strong link between environmental unpredictability and seed heteromorphism requires detailed analyses within a species and is the aim of my second and third chapter (Venable 1985b). In chapter two I focused on assessing the phenotypic expression of seed heteromorphism in a desert winter annual species, Pectocarya heterocarpa, across populations along an aridity gradient. I also evaluated germination patterns from seeds collected along the transect. In this chapter I also contribute to understanding the increasingly important role of plasticity in fine-tuning bet hedging strategies.

Preliminary work on the life history of P. heterocarpa shows delayed germination of basal relative to aerial seeds (Sarah Felker, unpublished data). Thus, if seed heteromorphism is a bet- hedging strategy, then in the face of increasing environmental variability or increasing probability of death after germination, P. heterocarpa should maximize geometric mean fitness by increasing the ratio of basal to aerial seeds (Cohen 1966; Gremer & Venable 2014). This is because delayed, fractional germination is a risk reducing strategy. I tested this prediction by sampling and assessing seed ratios of P. heterocarpa from twelve sites across an aridity gradient spanning southern California and Arizona. I found the ratio of basal to aerial seeds to increase with increasing rainfall variation as expected. In addition, I found that when the field collected

seeds are grown in a common-garden environment the pattern was reversed, implicating

plasticity. Interestingly, plants from more variable sites showed greater plasticity in their fine-

tuning of their seed ratios.

Regarding germination, I expected plants from more variable rainfall sites to produce offspring

with lower germination fractions (Cohen 1966; Gremer & Venable 2014). As predicted, I found that germination fractions were lowest in more variable sites. I also assessed fitness differences due to originating seed type in the F1 generation but found no significant differences in plant weight, seed set or seed ratios.

Overall, our chapter two results provide strong evidence for bet hedging in a seed-heteromorphic species while highlighting the increasingly recognized importance of the interplay between bet-

hedging and plasticity (Simons 2014; Gremer et al. 2016). While not finding fitness differences

associated with plants arising from different seed types was somewhat unexpected based on past

research in other similar species (McNamara & Quinn 1977; Cheplick 1994), it was not

contradictory to bet hedging predictions. This is because I germinated, transplanted, and grew our species under identical, non-water limited conditions. From bet hedging theory I expect differences in fitness consequences of seed types to operate largely through germination timing within and between years as a result of variability in the amount and timing of rainfall (Donohue

2002; Donohue et al. 2005). In fact, this is what I tested and demonstrated in chapter three.

CHAPTER III (APPENDIX C): FITNESS CONSEQUENCES OF HETEROMORPHIC SEED

TYPES IN A VARIABLE ENVIRONMENT

Chapter two provided important evidence for differentiation in the phenotypic expression of seed heteromorphism in P. heterocarpa across populations as predicted by bet hedging theory (Cohen

1966; Venable 1985b). In addition, we found evidence of plasticity in the adjustment of seed ratios and consistent, predicted differences in germination fractions across populations. In this chapter I extended this work to evaluate evidence that the observed differences in seed morph ratios and their germination characteristics translate to fitness differences across years and environmental conditions. This represents some of the strongest evidence for bet hedging

(Simons 2011). If P. heterocarpa is hedging its bets, then we should find that different seed types exhibit different fitness consequences in different combinations of environmental conditions and years. In other words, we predicted that different seeds should do better in different environments with no seeds being optimal across all environments (Venable 1985b).

To accomplish this, we designed a field experiment in which we planted all three seed morphs of

P. heterocarpa. We then applied three different watering treatments at the level of the seed, one of which was natural rainfall (no supplemental water). We monitored germination and reproduction across three years in this experiment, however in the third year we did not observe any germination. We also repeated a similar experiment in a more controlled greenhouse setting.

To achieve different germination cohorts in the artificial greenhouse setting we germinated seeds in a staggered fashion in growth chambers.

Both the field and greenhouse experiments exhibited different fitness consequences of different seed types in different combinations of germination cohorts, watering treatments and years.

Interestingly, seed type itself was rarely a significant explanatory variable with respect to fitness.

Rather it was the interaction between seed type and watering treatment, germination cohort and year that resulted in differential fitness effects. These results, in conjunction with those of chapters one and two, provide some of the most rigorous evidence of bet hedging documented for seed-heteromorphic species.

III. REFERENCES:

Baskin, J.M., Lu, J.J., Baskin, C.C., Tan, D.Y. & Wang, L. (2014). Diaspore dispersal ability and

degree of dormancy in heteromorphic species of cold deserts of northwest China: a review.

Perspect. Plant Ecol. Evol. Syst., 16, 93–99.

Caesar, S., Ahnesjo, J. & Forsman, A. (2007). Testing the role of coadapted genes versus bet-

hedging for mating strategies in colour polymorphic pygmy grasshoppers. Biol. J. Linn.

Soc., 90, 491–499.

Cheplick, G.P. (1994). Life history evolution in amphicarpic plants. Plant Species Biol., 9, 119–

131.

Cheplick, G.P. & Quinn, J.A. (1983). The shift in aerial subterranean fruit ratio in Amphicarpum

purshii: causes and significance. Oecologia, 57, 374–379.

Cohen, D. (1966). Optimizing reproduction in a randomly varying environment. J. Theor. Biol.,

12, 119–129.

Crean, A.J. & Marshall, D.J. (2009). Coping with environmental uncertainty: dynamic bet

hedging as a maternal effect. Philos. Trans. R. Soc. Lond. B. Biol. Sci., 364, 1087–96.

Cruz-Mazo, G., Buide, M.L., Samuel, R. & Narbona, E. (2009). Molecular phylogeny of

Scorzoneroides (Asteraceae): evolution of heterocarpy and annual habit in unpredictable

environments. Mol. Phylogenet. Evol., 53, 835–847.

Davidowitz, G. (2002). Does precipitation variability increase from mesic to xeric biomes? Glob.

Ecol. Biogeogr., 11, 143–154.

Donaldson-Matasci, M.C., Bergstrom, C.T. & Lachmann, M. (2013). When unreliable cues are

good enough. Am. Nat., 182, 313–327.

Donohue, K. (2002). Germination timing influences natural selection on life‐history characters in

Arabidopsis thaliana. Ecology, 83, 1006–1016.

Donohue, K., Dorn, L., Griffith, C., Kim, E., Aguilera, A., Polisetty, C.R., et al. (2005). The

evolutionary ecology of seed germination of Arabidopsis thaliana: variable natural selection

on germination timing. Evolution (N. Y)., 59, 758–770.

Dziminski, M.A. & Alford, R.A. (2005). Patterns and fitness consequences of intraclutch

variation in egg provisioning in tropical Australian frogs. Oecologia, 146, 98–109.

Felsenstein, J. (1985). Phylogenies and the Comparative Method. Am. Nat., 125, 1–15.

Gregersen, F., Vøllestad, L.A., Olsen, E.M. & Haugen, T.O. (2009). Sibling-size variation in

brown trout Salmo trutta in relation to egg size and stream size. J. Fish Biol., 74, 1259–

1268.

Gremer, J.R., Kimball, S. & Venable, D.L. (2016). Within- and among-year germination in

Sonoran Desert winter annuals: bet hedging and predictive germination in a variable

environment. Ecol. Lett., 1209–1218.

Gremer, J.R. & Venable, D.L. (2014). Bet hedging in desert winter annual plants: optimal

germination strategies in a variable environment. Ecol. Lett., 17, 380–387.

Harper, J.L. (1977). Population Biology of Plants. Academic Press, London.

Imbert, E. (2002). Ecological consequences and ontogeny of seed heteromorphism. Perspect.

Plant Ecol. Evol. Syst., 5, 13–36.

Karron, J.D. (1989). Breeding systems and levels of inbreeding depression in geographically

restricted and widespread species of Astragalus (fabaceae). Am. J. Bot., 76, 331–340.

Koops, M.A., Hutchings, * Jeffrey A. & Adams, B.K. (2003). Environmental predictability and

the cost of imperfect information: influences on offspring size variability. Evol. Ecol. Res.,

5, 29–42.

Lips, K. (2001). Reproductive trade-offs and bet-hedging in Hyla calypsa, a Neotropical

treefrog. Oecologia, 128, 509–518.

Mandak, B. (1997). Seed heteromorphism and the life cycle of plants: a literature review. Preslia

(Prague), 69, 129–159.

Mandak, B. & Pysek, P. (2005). How does seed heteromorphism influence the life history stages

of Atriplex sagittata (Chenopodiaceae)? Flora, 200, 516–526.

McNamara, J. & Quinn, J.A. (1977). Resource-allocation and reproduction in populations of

Amphicarpum purshii (Gramineae). Am. J. Bot., 64, 17–23.

Philippi, T. & Seger, J. (1989). Hedging ones evolutionary bets, revisited. Trends Ecol. Evol., 4,

41–44.

Porras, R. & Muñoz, J.M. (2000). Achene heteromorphism in the cleistogamous species

Centaurea melitensis. Acta Oecologica, 21, 231–243.

Reynolds, J.F., Kemp, P.R., Ogle, K. & Fernandez, R.J. (2004). Modifying the “pulse-reserve”

paradigm for deserts of North America: precipitation pulses, soil water, and plant responses.

Oecologia, 141, 194–210.

Sadeh, A., Guterman, H., Gersani, M. & Ovadia, O. (2009). Plastic bet-hedging in an

amphicarpic annual: an integrated strategy under variable conditions. Evol. Ecol., 23, 373–

388.

Seger, J. & Brockmann, H.J. (1987). What is bet-hedging? Oxford Surv. Evol. Biol., 4, 182–211.

Simons, A.M. (2011). Modes of response to environmental change and the elusive empirical

evidence for bet hedging. Proc. R. Soc. B-Biological Sci., 278, 1601–1609.

Simons, A.M. (2014). Playing smart vs. playing safe: the joint expression of phenotypic

plasticity and potential bet hedging across and within thermal environments. J. Evol. Biol.,

27, 1047–1056.

Starrfelt, J. & Kokko, H. (2012). Bet-hedging - a triple trade-off between means, variances and

correlations. Biol. Rev., 87, 742–755.

Stearns, S. (2000). Daniel Bernoulli (1738): evolution and economics under risk. J. Biosci., 25,

221–228.

Venable, D.L. (1985a). Ecology of achene dimorphism in Heterotheca latifolia: III.

Consequences of varied water availability. J. Ecol., 73, 757–763.

Venable, D.L. (1985b). The evolutionary ecology of seed heteromorphism. Am. Nat., 126, 577–

595.

Venable, D.L., Burquez, A., Corral, G., Morales, E. & Espinosa, F. (1987). The ecology of seed

heteromorphism in Heterosperma pinnatum in Central Mexico. Ecology, 68, 65–76.

Venable, D.L., Dyreson, E. & Morlaes, E. (1995). Population dynamic consequences and

evolution of seed traits of Heterosperma pinnatum (Asteraceae). Am. J. Bot., 82, 410–420.

Venable, D.L. & Levin, D.A. (1985). Ecology of achene dimorphism in Heterotheca latifolia: I.

Achene structure, germination and dispersal. J. Ecol., 73, 133–145.

Wang, L., Dong, M. & Huang, Z.-Y. (2010). Review of research on seed heteromorphism and its

ecological significance. Chinese J. Plant Ecol., 34, 578–590.

Zohary, M. (1962). Plant Life of Palestine. The Ronald Press Company, USA.

APPENDIX A

Title: Offspring polymorphism and bet hedging: a large-scale, phylogenetic analysis

Running title: Seed heteromorphism in North America

Authors and affiliations: Joshua Scholl1*, Leo Calle2, Nick Miller3 and D. Lawrence Venable1

1University of Arizona, Department of Ecology and Evolutionary Biology, Tucson, AZ 85721

2Montana State University, Department of Ecology, Bozeman, MT 59717

3National Ecological Observatory Network, Tucson, AZ 85721

*primary contact: Joshua Scholl, [email protected], 2237 East Beverly Drive, Tucson

Arizona, 85719, USA (phone: 954-815-7731)

Other author email addresses: Larry Venable ([email protected]), Leo Calle

([email protected]), Nick Miller ([email protected])

Published: https://doi.org/10.1111/ele.13522

Keywords: seed heteromorphism, offspring polymorphism, bet hedging, reproductive biology,

life history evolution

Statement of authorship: JPS and DLV designed the study, JPS and NM collected data, LC and

JPS performed the spatial analysis, JPS and DLV performed all other analyses, JPS wrote the

first draft of the manuscript and all authors contributed substantially to revisions

Data accessibility statement: If this manuscript is accepted for publication, the data supporting

the results will be archived in the Dryad public data repository and the data DOI will be included

at the end of the article.

ABSTRACT

Offspring polymorphism is a reproductive strategy where individual organisms simultaneously produce offspring that differ in morphology and ecology. It occurs across the Tree of Life but is

particularly common among plants, where it is termed seed (diaspore) heteromorphism. The

prevalence of this strategy in unpredictably varying environments has resulted in the assumption

that it serves as a bet-hedging mechanism. We found 101 examples of this strategy in

southwestern North America. We provide phylogenetically informed evidence for the hypothesis

that the occurrence of seed heteromorphism increases with increasing environmental variability,

though this pattern was only significant for aridity, one of our two rainfall variability metrics.

We provide a strong test of bet hedging for a large, taxonomically diverse set of seed-

heteromorphic species, lending support to the hypothesis that bet hedging is an important

mechanistic driver for the evolution of seed heteromorphism.

INTRODUCTION

Unpredictable environments are ubiquitous in nature and a mechanistic understanding of how organisms cope with them is becoming increasingly relevant (Simons 2011, Gremer and Venable

2014). Two coping mechanisms are rapid evolution and plasticity, both of which tend to maximize average fitness in variable environments (Bell & Collins 2008). Alternatively, organisms may employ bet-hedging mechanisms by sacrificing average fitness to reduce variance in fitness and maximize long-term population growth rate (Seger & Brockmann 1987;

Starrfelt & Kokko 2012).This allows organisms to buffer against unpredictable environmental variation, such as variable desert rains, and it tends to reduce extinction risk (Cohen 1966;

Slatkin 1974). Putative examples of bet hedging are abundant and include offspring heteromorphism, or the simultaneous production of two or more morphologically distinct offspring by an individual organism. Offspring heteromorphism is often qualitatively associated with unpredictable environmental factors (e.g. rainfall) that could result in the evolution of bet hedging. However, strong quantitative demonstration that bet hedging is the underlying mechanism driving the evolution of strategies like offspring heteromorphism is very rare

(Simons 2011; Gremer & Venable 2014). Here we compiled an extensive dataset of offspring heteromorphic angiosperms and their homomorphic congeners in southwestern North America and combined this with quantitative habitat information on water availability and its variation.

We then analyzed this large-scale dataset in a phylogenetic context to statistically evaluate if the prevalence of offspring heteromorphism is correlated with increasing environmental variability as predicted by bet hedging theory (Cohen 1966; Venable 1985b). This provides an unusually strong, large-scale test of bet hedging as the underlying mechanism driving the evolution of offspring heteromorphism.

Although offspring heteromorphism occurs across the tree of life (e.g. fish (Koops et al. 2003;

Gregersen et al. 2009), soil mites (Crean & Marshall 2009), grasshoppers and frogs (Lips 2001;

Dziminski & Alford 2005)), it is best documented in angiosperms. Termed seed (diaspore)

heteromorphism, it is defined in plants as the production, by a single individual, of seeds or

dispersal units that differ in morphology and ecology (Venable 1985b; Mandak 1997; Imbert

2002). The variation can be discrete or continuous if the extreme morphs are dramatically

different (e.g., Heterosperma pinnatum, Venable et al. 1987). At least 292 seed heteromorphic plant species are known (Wang et al. 2010). The prevalence of this strategy among plants likely is due in large part to their modular structure, which affords them exceptional functional

flexibility (Tomlinson 1982; Lloyd 1984; Venable 1985b; Haukioja 1991).

Past efforts to characterize seed heteromorphism have focused on the conditions driving the

evolution of bet hedging – specifically, unpredictably fluctuating environments. The most

common drivers suggested to date include aridity (Ellner & Shmida 1981; Venable 1985a;

Mandak 1997; Imbert 2002; Wang et al. 2010) and disturbance (e.g. roads and fire; Cheplick &

Quinn 1982; Venable & Levin 1985a). The association of seed heteromorphism with desert or

disturbed habitats has been hypothesized and explored by several reviews (e.g. Mandak 1997;

Imbert 2002; Wang et al. 2010) and individual studies (e.g. Venable et al. 1987; Sadeh et al.

2009). This association has been noted in many early qualitative reports on seed heteromorphism

(e.g. Zohary & Pascher 1937; Ellner & Shmida 1981; Barker 2005).

The benefits of spreading reproductive risk across different seed morphs also should be greater for annual or monocarpic species because they reproduce only once and cannot spread risk across multiple reproductive bouts like perennial polycarpic species (Rees 1994). Consequently,

30 if seed heteromorphism is a bet-hedging strategy, it might be expected to be more prevalent in annual species (Plitmann 1986; Mandak 1997; Imbert 2002).

Previous qualitative analyses have not successfully addressed the question of whether adaptive evolution or shared ancestry is responsible for associations of seed heteromorphism with variable environments or the annual habit. If, for example, the species described in past reviews were all derived from genera that had many seed-homomorphic species, all of which also grew in variable environments, then a phylogenetic analysis would have revealed that the apparent correlation between seed heteromorphism and variable habitats could be explained simply by evolutionary relationships.

Only two studies, both on composite genera, have explored the association of seed heteromorphism with deserts and other unpredictable habitats in a phylogenetic context. Imbert

(2002) explored the association of seed heteromorphism in Crepis with deserts and other unpredictable habitats in a phylogenetic context, but this study lacked statistical evaluation.

Cruz-Mazo et al. (2009) conducted a phylogenetically robust study of 21 species in the genus

Scorzoneroides that demonstrated a statistically significant relationship between seed heteromorphism and unpredictable habitats and annual habit. This study used five, very-closely related seed-heteromorphic species contrasted with 16 seed homomorphic species. Large-scale statistical evidence remains elusive and the broadly assumed conclusion that seed heteromorphism acts as a bet-hedging mechanism is premature.

Plants in North American deserts offer a great opportunity to quantitatively examine evidence supporting the claim that seed heteromorphism evolved as a bet-hedging mechanism. This region encompasses vast areas of desert ecosystems, including the Sonoran, Chihuahuan, Great Basin

and Mojave deserts. These deserts differ widely in temperature and rainfall patterns (UNEP

1992) but share the characteristic that rainfall is generally low, variable and unpredictable

(Davidowitz 2002; Reynolds et al. 2004). Desert floras are also replete with annual plants (~50%

of species in local Sonoran desert floras, Venable & Pake 1999). Thus North American deserts

are likely fertile ground for the evolution of bet hedging strategies (Seger & Brockmann 1987;

Starrfelt & Kokko 2012) and these regions are predicted to contain a high diversity of seed- heteromorphic species.

Our first goal in this analysis was to summarize the occurrence and characteristics of seed- heteromorphic species among all angiosperms in southwestern North America. Specifically, we determined the (1) growth form, (2) weediness and (3) climatic niche for each species of angiosperm that produced heteromorphic seeds and for all congeneric homomorphic species.

Second, we used phylogenetic comparative methods to statistically test the hypothesis that seed heteromorphism is concentrated among weeds or annual plants and that they tend to occur in arid environments or environments with high variability in water availability. Our extensive, diverse list of seed-heteromorphic species and quantitative assessment of environmental unpredictability evaluated within a phylogenetic context make this a robust statistical test of bet-hedging as the long-assumed underlying mechanism favoring the evolution of seed heteromorphism.

METHODS

Definitions

We defined seed heteromorphism as the production, by a single individual, of seeds or dispersal units that differ discretely or extremely in morphology (Venable 1985b; Mandak 1997; Imbert

2002). Thus, we did not include in our definition plants producing continuously varying seeds

without extreme differences (Baskin & Baskin 2014) or cryptic heteromorphism where seeds are morphologically identical but differ in their ecological responses such as germination (Venable

1985b). The former would include too many species with no clear boundaries and the latter are rarely documented and require ecological experimentation to detect (e.g. germination trials).

However, we did include species with continuous variation when seeds at either extreme had widely divergent morphologies. This is the case for many species of Asteraceae, e.g.,

Heterosperma pinnatum (see Fig. 1 in Venable et al. 1995). This is slightly different from the definition of Baskin and Baskin (2014) who nominally exclude continuous variation but in practice include species with continuous but extreme variants such as H. pinnatum as heteromorphic.

In our study we assume that morphological differences among seeds translate to unique ecological consequences. This assumption is robustly supported by studies which have investigated the link between morphology and ecology among seeds of seed-heteromorphic species (e.g. Cruz-Mazo et al. 2009; Ma et al. 2010; Baskin et al. 2014; Zhang et al. 2016). For example, in their meticulous review of 20 seed-heteromorphic species from China, Baskin et al.

(2014) show that all species had morphs which differed in either their dispersal ability or degree of dormancy, and in most cases both. Indeed, we are unaware of any study that challenges the assumption that morphological differences among seeds translate to unique ecological consequences.

Defining and assigning weediness was non-trivial, and despite our use of it as a binary trait we acknowledge its continuous and relative nature (Hanan et al. 2015). We used several sources and indicators of weediness as described in detail in the supplemental materials (Appendix S1 and

database S1). However, our designation of plants as weedy versus non-weedy was largely based

on explicit descriptions of plants as weedy or non-weedy or collection site or habitat as described

in floras and herbaria following Hart (1976) and applied by Hanan (2015). Thus, any plants

described as occurring in disturbed sites (roadsides, railroads, field margins, etc.) or in secondary

or ruderal vegetation were classified as weedy species. We categorized plant life cycle duration

(habit) into annual vs perennial based on descriptions in the floras, but we also consulted

additional databases (listed in supplementary materials) when duration was not clearly indicated

in the floras.

Quantifying environmental unpredictability is difficult. Past studies of seed heteromorphism

have simply relied on general biome descriptions such as “desert,” “semi-desert,”

“Mediterranean” and “montane” (Mandak 1997; Imbert 2002; Cruz-Mazo et al. 2009; Wang et al. 2010). Here we use an aridity index and the coefficient of variation (CV) of precipitation as practical and globally available indicators of variability in water availability. While the CV of precipitation is a direct measure of rainfall variability, aridity is a quantitative way of accounting for many factors which influence water stress for plants including sunlight, rainfall, temperature and elevation making it a nuanced and potentially more relevant measure of water availability

(LeHouérou 1996; Elfaki et al. 2011). Also, previous studies have shown a direct tie between aridity and unpredictability in water availability (Davidowitz 2002; Berg & Hall 2015; Yoon et al. 2015).

Data Sources and Filtering

To determine the prevalence of seed heteromorphism in this region, we reviewed every

angiosperm in five floras (Table S1). We then reviewed the literature regarding all seed-

heteromorphic species we found and used supplemental publications to aid our understanding of

specific species. For example, for the genera Pectocarya and Cryptantha, we found supplemental

publications (Johnston 1925; Hasenstab-Lehman & Simpson 2012) that more thoroughly

described the seed heteromorphism in these groups and allowed us to include several additional

species not listed in the floras (e.g. Guilliams et al. 2013). We also recorded the identity and characteristics of all seed homomorphic species in genera with at least one seed-heteromorphic species.

We reviewed all species names for validity and synonymy using the Taxonomic Name

Resolution Service (TNRS, Boyle et al. 2013). Although we initially recorded all seed homomorphic species from genera that contained at least one seed-heteromorphic species, the nomenclature update resulted in many genera consisting solely of seed homomorphic species.

For the purposes of our analyses we assumed that seed-heteromorphic species with name changes since the original flora publication were still closely related to the seed homomorphic species formerly of the same genus and therefore did not remove them from our data set.

We downloaded species presence data for all these heteromorphic and homomorphic species from two electronic databases, the Southwest Environmental Information Network (“SEINet”

2017) and the Global Biological Information Facility (“GBIF” 2017). Occurrences were limited to the United States and Mexico. To reduce well-documented biases associated with herbarium collections, especially overrepresentation near cities (Rich & Woodruff 1992; Crawford &

Hoagland 2009), we randomly thinned species presence records to a density of one record per 25

km2. Known as spatial filtering, this procedure has been demonstrated to greatly reduce

collection biases (Kramer-Schadt et al. 2013).

We extracted aridity data at a resolution of 30 arcsec (ca. 1km) from shapefiles produced by the

CGIAR-Consortium for Spatial Information Global-Aridity and Global-Potential

Evapotranspiration Database (Zomer et al. 2007, 2008; “CGIAR-CSI” 2017) and CV of precipitation data from TerraClimate ( ca. 4km, Abatzoglou et al. 2018). Aridity was defined as the ratio between mean annual precipitation (MAP) and mean annual potential evapotranspiration (MAE), with lower numbers denoting higher aridity (Middleton and Thomas

1997, “CGIAR-CSI” 2017). CV of precipitation data was calculated on total annual precipitation spanning 1960 to 1990.

Next, for the spatially-filtered set of presence records for a given species, we extracted aridity and CV of precipitation values for each presence point and then calculated their averages for each species. We bootstrapped this analysis 100 times for each species (re-randomizing the location records by filtering each time to generate 100 unique maps) and then calculated the mean and standard deviation of the 100 runs as the final species aridity and CV of precipitation measure. As a caution against statistical artefacts from the spatial filtering, we ran this entire procedure on three other spatial filtering densities (50 km2, 75 km2, and 100 km2) for aridity.

However, since the aridity results were very similar, we only discuss those for 25 km2 and only extracted the CV of precipitation for the 25 km2 scheme. Neither indicator of variability in water

availability was obtained for species with less than three occurrence points or for which all

occurrence points were too localized in space (e.g. all occurring in <25 km2; Database S1). We

provide a detailed figure explaining our spatial analysis and the associated datasets in the

36 supplementary materials (Figure S1). Locations and environmental data for each species were projected, spatially filtered, and analyzed using R Statistical Software (R Core Team 2017).

Phylogenetic Analyses

We used phylogenetic independent contrasts to statistically explore the relationships between seed heteromorphism, life span, aridity, CV of precipitation and weediness. We first pruned a dated molecular phylogeny (hereafter base phylogeny; Zanne et al. 2014), to represent 161 of our

559 seed heteromorphic and seed homomorphic species. Of these, 151 species were directly represented in the base phylogeny, while 34 species were included based on their genus and one based on its family (see supplemental materials for pruned phylogeny and details on assembly).

For the analyses we pruned the phylogenetic tree further to exclude subspecies and varieties, species which lacked climate data, and species which were in the same genus with others in cases where only the genus occurred in the base phylogeny and inclusion resulted in undesirable soft polytomies (see supplementary materials, table S1, for details on assembly). Using these filtering criteria, we retained 127 seed homomorphic species and 24 seed-heteromorphic species for the phylogenetic analyses.

We evaluated life-span, aridity, CV of precipitation and weediness as independent variables predicting seed heteromorphism using BRUNCH (R package Caper v. 0.5.2; Purvis & Rambaut

1995; Orme et al. 2012). BRUNCH performs phylogenetically independent contrasts for models of discrete variables or a combination of discrete and continuous variables. We used a one- sample, one-tailed t-test to determine if the mean of the independent contrasts of the explanatory variable was significantly different than zero. For aridity we expected contrasts to be less than zero which would indicate the expected negative correlation between seed heteromorphism and

37 aridity index (lower index is more arid) while for CV of precipitation we expected contrasts to be greater than zero (higher CV is more variable).

RESULTS

We found 101 species of seed heteromorphic angiosperms spread across 51 genera and 9 families (Table 1; Database S1). Of these, five were intraspecific varieties in the Asteraceae which were excluded from our phylogenetic analyses (Table S1). Seed heteromorphism was concentrated in the Asteraceae and Boraginaceae which contributed 64 and 23 species, respectively. Morphological differences between the seeds of the seed-heteromorphic species can be found in Database S1.

Our phylogenetic analyses revealed significant to marginally significant associations between seed heteromorphism and aridity at the different spatial clustering scales (Table 2). Seed heteromorphism was associated with lower aridity values (i.e. found in more arid habitats; at 25 km2: df = 21, p = 0.05; Fig. 1; Table 2). We also found that the association between seed heteromorphism and CV of precipitation was in the right direction (more heteromorphism in more variable environments) but not significantly so (Fig. 1; Table 2). In the full unpruned data set, 74% of seed-heteromorphic species were annuals compared to just 50% of the seed homomorphic species. However, we found no phylogenetic correlation between seed heteromorphism and the annual growth habit or weediness (Fig. 1; Table 2). While only 29 of our 101 seed heteromorphic taxa were classified as weedy (29%), a lower proportion of seed homomorphic species were classified as weedy (104 out of 458, or 23%).

DISCUSSION

In this study we documented the occurrence of seed heteromorphism in southwestern North

America. Using phylogenetic independent contrasts, we conducted the first large-scale statistical

analysis of the association between seed heteromorphism and aridity and CV of precipitation,

both measures of environmental variability, and thus its possible function as a bet-hedging

mechanism. Our study is the first to incorporate quantitative measures related to environmental

variability and examine species spanning multiple families. We found four major results: first,

there is an association between seed heteromorphism and arid habitats, but not CV of

precipitation. Second, we did not find the hypothesized statistical association of seed

heteromorphism with the annual life cycle. Third, we did not find support for an association

between seed heteromorphism and weediness. Fourth, seed heteromorphism is very common in

southwestern North America compared to other global regions and exhibits a different taxonomic

distribution.

Seed heteromorphism and variability in water availability

Aridity was significantly associated with seed heteromorphism in our phylogenetic analysis at

the lowest spatial thinning scale and marginally significantly at the other scales. Seed

heteromorphism was also associated with higher values of CV of precipitation but this

relationship was not significant. We found this difference in pattern strength to be paradoxical

because much previous work has shown a tight relationship between average precipitation and

variability of precipitation (e.g. R2 = 0.97 in Davidowitz 2002). One possibility is that CV of annual precipitation does not fully capture year-to-year environmental variability as experienced by plants. This could be due partly to statistical reasons (annual precipitation has non-normal,

often skewed distributions and with our sample size of 30 years is influenced by outliers and

decadal trends such as SOI oscillations). But the different result for aridity vs CV of precipitation

could also be due to biological reasons. Environmental variation in precipitation may have a

greater impact on variation in plant fitness at lower precipitation values where extreme stress

occurs. Also, the aridity index includes evapotranspiration which operates in concert with

precipitation in creating plant water stress, hence making it more reflective of environmental

variation in plant fitness. While we do not fully understand why aridity and CV of precipitation

have different significance levels, the trend for both is that heteromorphism is associated with

more variable environments which lends support to the hypothesis that seed heteromorphism

functions as a bet-hedging mechanism to cope with unpredictable soil moisture availability.

Wang et al. (2010) amassed records of 292 seed-heteromorphic species across the globe,

including many Asian species, and reported that 87% occurred in unpredictable sites, qualitatively defined as arid, saline, or highly disturbed. However, they did not break this down further or statistically evaluate it. In an analysis of the African and Madagascan floras, Barker

(2005) reported on 23 seed-heteromorphic species and noted that many of them occur in arid

regions or disturbed high-altitude alpine environments. A connection between seed-

heteromorphic species and arid regions was also recognized by Zohary and Pascher (1937)

regarding the Middle Eastern flora. Such comparisons without phylogenetic analyses have

dominated examinations of seed heteromorphism.

In his analysis of 196 species in the genus Crepis, of which 30 were seed heteromorphic, Imbert

(2002) illustrated the problem associated with not taking species shared evolutionary history into

account. While most seed heteromorphic Crepis occurred in arid regions (21 of 30), only 46 of

166 homomorphic species occurred in arid regions. However, when mapped on a phylogenetic

tree, there was not a clear association between seed heteromorphism and arid habitats. Imbert did

40 not conduct a formal phylogenetic analysis and thus did not evaluate his data statistically but examining the phylogeny allowed him to reject an otherwise strong-looking pattern. Cruz-Mazo et al. (2009) provided the most robust phylogenetic assessment of the link between seed heteromorphism and unpredictable environments. Using 21 species in the genus Scorzoneroides, they found strong support for a link between their five seed-heteromorphic species and unpredictable habitats. However, their sample sizes and taxonomic breadth were very small.

Imbert (2002) argued for the importance of ensuring a large enough sample size and diversity across genera and families. It is very likely that certain genera possess many adaptations to variable environments, with seed heteromorphism being only one of them. In this case a phylogenetic analysis might incorrectly confirm or rule out a causal association between variable habitats and seed heteromorphism due to small sample size.

The problem of small sample size is further illustrated by observations and experiments demonstrating that coping with variability in water availability is not the only explanation for the evolution of seed heteromorphism. Cheplick (2005) suggested that Amphicarpum purshii produces heteromorphic seeds to cope with unpredictable variation in habitat disturbance.

Similarly, Heterotheca subaxillaris var. subaxillaris may have evolved seed heteromorphism in response to colonization and competition dynamics (Baskin & Baskin 1976; Lonard et al. 2011).

Heteromorphic seeds have also been related to variability in predation rates by vertebrates (Cook et al. 1971) and invertebrates (Kistenmacher & Gibson 2016; Honek et al. 2017). In addition, developmental and floral constraints may facilitate the evolution of seed heteromorphism

(Dowling 1933; Zohary 1950; Harper 1977), especially in the Asteraceae (Venable 1985b).

Clearly, there are many environmental factors and ecological processes that may lead to the evolution of seed heteromorphism in individual cases. However, to characterize the strategy and

41 gain some predictive power regarding its evolution and occurrence we must focus on general trends that are evaluated in a phylogenetic context and with sufficient sample sizes. Most seed- heteromorphic species appear to be associated with some type of variable environmental factor.

Our analyses statistically support the notion that aridity is an important factor in explaining the occurrence of seed heteromorphism.

Our results are especially striking when considering that we only analyzed seed-heteromorphic species and their seed homomorphic congeners. Thus, all our seed heteromorphic and seed homomorphic species were very closely related. The niche conservatism hypothesis would suggest that they may have relatively similar niches. Consequently, most of our species likely have some type of mechanism to cope with harsh desert environments. Nevertheless, among this group of largely desert-adapted plants, seed heteromorphic ones were still more strongly associated with aridity and CV of precipitation. These patterns would likely be more striking if one were to repeat this analysis for the entire flora of southwestern North America.

Seed heteromorphism and the annual life cycle

To conduct a statistically robust phylogenetic analysis, we were forced to reduce our set of 559 species to 151. This permitted powerful statistical inference free of spurious biases which found no correlation of seed heteromorphism to the annual life cycle. We nonetheless can complement this analysis by cautiously comparing the results to patterns seen in the proportional representation of heteromorphism across environments using the full raw species data set and lists from other studies. Seed heteromorphism was found in three times as many annuals as perennials while seed homomorphic species were equally likely to be annual or perennial. This agrees with findings from past reviews (Imbert et al. 1997; Mandak 1997; Wang et al. 2010) and is a pattern predicted by bet hedging theory (Venable 1985b). We attribute the lack of a

significant pattern in the phylogenetic analysis to the small proportion of seed-heteromorphic species (24%) that we were able to include in our phylogenetic tree. Also, compared to our complete list, we included proportionately more of the perennial than annual seed-heteromorphic species, reducing the annual to perennial ratio among seed-heteromorphic species from 3:1 to

2:1. Future advances in phylogenetic resolution will permit more powerful analyses of the role of the annual habit in the evolution of seed heteromorphism.

Seed heteromorphism and weediness

In contrast with past qualitative work (Mandak 1997; Imbert 2002; Wang et al. 2010) our analysis did not reveal an association of weediness with seed heteromorphism. However, all these studies lumped ‘weediness’ with variable environments and an annual life form, all being purported to be factors likely involved in the evolution of seed heteromorphism, but without any statistical tests. It is very difficult to define weediness, especially in a binary sense. Also, it is difficult to employ any definition at a large scale given the general lack of ecological data available for most species. For these reasons, we interpret our lack of evidence for a link between seed heteromorphism and weediness cautiously.

Abundance of seed-heteromorphic species in southwestern North America

Seed heteromorphism is abundant in our study region compared to other parts of the world. The most recent comprehensive review for the whole world listed 292 seed-heteromorphic species

(Wang et al. 2010). 86 of our of 101 taxa species are new, bringing the new world total to 378.

Since Wang et al. (2010) did not report on the geographic locations of seed-heteromorphic species, we conducted a quick GBIF search of their species. Of the ones with available

occurrence data, 122 were most frequently associated with Europe, 65 with Asia, 53 with North

America, and 31 with Africa. Thus, when including our species, and accounting for duplicates,

North America has nearly 130 seed-heteromorphic species. These are all no doubt under counts

by varying degrees due to differential research intensity across regions. Ellner and Shmida

(1981) reported about 98 seed-heteromorphic species from the Israeli flora alone. Africa,

Oceania and South America, have large expanses of arid and semi-arid vegetation and likely

have many unreported cases of seed heteromorphism.

Of the 292 global species previously recognized, most (146 species) belong to the Asteraceae

(Wang et al. 2010). In contrast, Boraginaceae ranks eighth in global frequency with only seven species. Similarly, our list is dominated by the Asteraceae with 64 species, however,

Boraginaceae is second with 23 species (Table 1). The high occurrence of Boraginaceae in our

seed heteromorphism list is due almost entirely to two genera, Cryptantha and Pectocarya. Both

occur primarily in arid and semi-arid regions of the Western Hemisphere and are taxonomically

distinguished largely by their seed morphology (Hasenstab-Lehman & Simpson 2012). We

expect that future work on South American species of these genera will be equally detailed

regarding seed morphology and thus contribute more species to the global seed heteromorphism

list.

A complete lack of seed heteromorphic Fabaceae in our list (Table 1) is surprising for two

reasons. First, with about 19,400 species, the family is the third largest in the world and we

might expect at least some of them to have evolved seed heteromorphism (Christenhusz & Byng

2016). Second, it has a high concentration of species in the southwestern United States, including

many in semi-arid and arid regions (e.g. Astragalus; Wojciechowski et al. 1999). Nearly all of

the 22 previously described seed heteromorphic Fabaceae employ the relatively rare amphicarpic

44 seed heteromorphic strategy (Wang et al. 2010). This strategy is rarely reported, perhaps because it is difficult to identify as one seed type is buried underground or because the identification of legumes rarely hinges on diaspore forms which consequently are not meticulously described in floras. This may contribute to why we did not find any heteromorphic Fabaceae species.

Conclusions and future directions

It remains to be seen if the association of seed heteromorphism with aridity or CV of precipitation will hold when examined globally or with more resolved future phylogenies but our analyses support the use of aridity as an important feature characterizing seed-heteromorphic species. Neither weediness nor annual life cycle appears to have any merit as a predictor based on our results. However, we suspect that both factors, and especially the latter, will gain statistical support with advances in phylogenetic resolution.

Defining seed heteromorphism is a challenge. Seed variation exists in all plant species, so what degree of variation must seeds have to merit a search for an adaptive explanation? We have suggested here that it is less about the quantitative degree of difference and more about a consistent discrete difference in the extreme seed morphologies; and most importantly that the differences have significant ecological consequences. In our analysis and in any broad scale analysis it is difficult to assess the ecological consequences of each species’ varying seed morphologies. Thus, our list should be considered a conservative estimate of seed heteromorphism in southwestern North America since we only recorded species that were clearly described as having discrete or extreme, consistent morphological differences.

Seed-heteromorphic species are likely much more common than currently appreciated. Many heteromorphic species remain hidden in floras which have yet to be reviewed by someone interested in seed heteromorphism. Also, plant descriptions typically consider traits relevant to the identification and taxonomic discrimination of species as opposed to ecological features.

Thus, descriptions in floras tend to focus more on the details of flowers and less on seeds. Even careful perusal of floras will not uncover all species without physical examination of actual plants. Furthermore, continents with large arid and semi-arid regions (e.g. South America and

Africa) remain relatively unexplored with respect to seed heteromorphism. We expect this to change dramatically in the future as herbaria and other organizations continue to expand the specimens and floras available digitally.

Offspring polymorphism is a remarkable life history strategy observed across the tree of life that is particularly common among plants. We hope that this paper will provide a foundation for future studies into seed heteromorphism in other understudied regions of the planet. Such studies are especially important because plant reproductive strategies like seed heteromorphism play a critical role in population and community dynamics, relevant to species conservation and evolution in a changing global environment.

ACKNOWLEDGEMENTS

This research was supported by NSF grants DEB-1256792 (LTREB) to DLV, DGE-1746060

(NSF-GRFP) to JPS, DEB-1702050 (NSF-DDIG) to JPS and an Arizona Native Plant Society

Grant to JPS. We thank Ty Taylor and Yue Max Li for their help with data analysis and two anonymous reviewers for their valuable comments.

Data accessibility statement: Database S1 is accessible via Figshare https://doi.org/10.6084/m9.figshare.11907189.v1

REFERENCES

Abatzoglou, J. T., S. Z. Dobrowski, S. A. Parks, and K. C. Hegewisch. 2018. TerraClimate, a

high-resolution global dataset of monthly climate and climatic water balance from 1958–

2015. Scientific data 5:1–12.

Barker, N. P. 2005. A review and survey of basicarpy, geocarpy, and amphicarpy in the African

and Madagascan flora. Annals of the Missouri Botanical Garden 92:445–462.

Baskin, C. C., and J. M. Baskin. 2014. Seeds: ecology, biogeography, and, evolution of

dormancy and germination. 2nd edition. Academic Press, Amsterdam.

Baskin, J. M., and C. C. Baskin. 1976. Germination dimorphism in Heterotheca subaxillaris var.

subaxillaris. Bulletin of the Torrey Botanical Club 103:201–206.

Baskin, J. M., J. J. Lu, C. C. Baskin, D. Y. Tan, and L. Wang. 2014. Diaspore dispersal ability

and degree of dormancy in heteromorphic species of cold deserts of northwest China: a

review. Perspectives in Plant Ecology, Evolution and Systematics 16:93–99.

Bell, G., and S. Collins. 2008. Adaptation, extinction and global change. Evolutionary

Applications 1:3–16.

Berg, N., and A. Hall. 2015. Increased interannual precipitation extremes over California under

climate change. Journal of Climate 28:6324–6334.

Boyle, B., N. Hopkins, Z. Lu, J. A. Raygoza Garay, D. Mozzherin, T. Rees, N. Matasci, M. L.

Narro, W. H. Piel, S. J. Mckay, S. Lowry, C. Freeland, R. K. Peet, and B. J. Enquist. 2013.

The taxonomic name resolution service: an online tool for automated standardization of

plant names. BMC Bioinformatics 14:1471–2105.

CGIAR-CSI. 2017. Consultative Group for International Agricultural Research - Consortium for

Spatial Information. Availabel at: https://cgiarcsi.community/category/data/. Last accessed

08-10-2017.

Cheplick, G. 2005. Biomass partitioning and reproductive allocation in the invasive,

cleistogamous grass Microstegium vimineum: influence of the light environment. The

Journal of the Torrey Botanical Society 132: 214-225.

Cheplick, G. P., and J. A. Quinn. 1982. Amphicarpum purshii and the pessimistic strategy in

amphicarpic annuals with subterranean fruit. Oecologia 52:327–332.

Christenhusz, M. J. M., and J. W. Byng. 2016. The number of known plants species in the world

and its annual increase. Phytotaxa 261:201–217.

Cohen, D. 1966. Optimizing reproduction in a randomly varying environment. Journal of

Theoretical Biology 12:119–129.

Cook, A. D., P. R. Atsatt, and C. A. Simon. 1971. Doves and dove weed: multiple defenses

against avian predation. BioScience 21:277–281.

Crawford, P. H. C., and B. W. Hoagland. 2009. Can herbarium records be used to map alien

species invasion and native species expansion over the past 100 years? Journal of

Biogeography 36:651–661.

Crean, A. J., and D. J. Marshall. 2009. Coping with environmental uncertainty: dynamic bet

hedging as a maternal effect. Philosophical transactions of the Royal Society of London.

Series B, Biological sciences 364:1087–96.

Cruz-Mazo, G., M. L. Buide, R. Samuel, and E. Narbona. 2009. Molecular phylogeny of

Scorzoneroides (Asteraceae): evolution of heterocarpy and annual habit in unpredictable

environments. Molecular Phylogenetics and Evolution 53:835–847.

Davidowitz, G. 2002. Does precipitation variability increase from mesic to xeric biomes? Global

Ecology and Biogeography 11:143–154.

Dowling, R. E. 1933. The reproduction of Plantago coronopus: an example of morphological

and biological seed dimorphism. Annals of Botany 47:861–872.

Dziminski, M. A., and R. A. Alford. 2005. Patterns and fitness consequences of intraclutch

variation in egg provisioning in tropical Australian frogs. Oecologia 146:98–109.

Elfaki, E. A., A. M. Adeeb, S. Saidi, and A. Ickowicz. 2011. Application of remote sensing and

GIS in assessment of rain use efficiency and rangeland degradation in central Butana,

Sudan. International Journal of Water Resources and Arid Environments:261–270.

Ellner, S., and A. Shmida. 1981. Why are adaptations for long-range seed dispersal rare in desert

plants. Oecologia 51:133–144.

GBIF. 2017. Global Biodiversity Information Facility. Available at:

https://www.gbif.org/occurrence/search. Last accessed 08-10-2017.

Gregersen, F., L. A. Vøllestad, E. M. Olsen, and T. O. Haugen. 2009. Sibling-size variation in

brown trout Salmo trutta in relation to egg size and stream size. Journal of Fish Biology

74:1259–1268.

Gremer, J. R., and D. L. Venable. 2014. Bet hedging in desert winter annual plants: optimal

germination strategies in a variable environment. Ecology Letters 17:380–387.

Guilliams, M., B. Veno, M. Simpson, and R. Kelley. 2013. Pectocarya anisocarpa, a new

species of Boraginaceae, and a revised key for the genus in western North America. Aliso

31:1–13.

Hanan, A., A. Maria, H. Vibrans, J. L. Villaseñor, N. I. Cacho, E. Ortiz, G. Gomez, and A.

Viicio. 2015. Use of herbarium data to evaluate weediness in five congeners. AoB Plants 8:

plv144.

Harper, J. L. 1977. Population Biology of Plants. Academic Press, London.

Hart, R. 1976. An index for comparing weediness in plants. Taxon 25:245–247.

Hasenstab-Lehman, K. E., and M. G. Simpson. 2012. Cat’s eyes and popcorn flowers:

phylogenetic systematics of the genus Cryptantha s.l. (Boraginaceae). Systematic Botany

37:738–757.

Haukioja, E. 1991. The influence of grazing on the evolution, morphology and physiology of

plants as modular organisms. Philosophical Transactions of the Royal Society of London.

Series B: Biological Sciences 333:241–247.

Honek, A., Z. Martinkova, S. Koprdova, and P. Saska. 2017. The role of invertebrates in

seedling establishment in a heterocarpic plant, Atriplex sagittata. Annals of Applied Biology

171:264–276.

Imbert, E. 2002. Ecological consequences and ontogeny of seed heteromorphism. Perspectives in

Plant Ecology Evolution and Systematics 5:13–36.

Imbert, E., J. Escarre, and J. Lepart. 1997. Seed heteromorphism in Crepis sancta (Asteraceae):

performance of two morphs in different environments. Oikos 79:325–332.

Johnston, I. M. 1925. The North American species of Cryptantha. The Gray Herbarium of

Harvard University, Cambridge MA.

Kistenmacher, M., and J. P. Gibson. 2016. Bet‐hedging against larval herbivory and seed bank

mortality in the evolution of heterocarpy. American Journal of Botany 103:1383–1395.

Koops, M. A., Jeffrey A. Hutchings, and B. K. Adams. 2003. Environmental predictability and

the cost of imperfect information: influences on offspring size variability. Evolutionary

Ecology Research 5:29–42.

Kramer-Schadt, S., J. Urgen Niedballa, J. D. Pilgrim, B. Schr€ Oder, J. Lindenborn, V.

Reinfelder, M. Stillfried, I. Heckmann, A. K. Scharf, D. M. Augeri, S. M. Cheyne, A. J.

Hearn, J. Ross, D. W. Macdonald, J. Mathai, J. Eaton, A. J. Marshall, G. Semiadi, R.

Rustam, H. Bernard, R. Alfred, H. Samejima, J. W. Duckworth, C. Breitenmoser-Wuersten,

J. L. Belant, H. Hofer, and A. Wilting. 2013. The importance of correcting for sampling

bias in MaxEnt species distribution models. Diversity and Distributions 19:1366–1379.

Larson, D. L. 2003. Native weeds and exotic plants: relationships to disturbance in mixed-grass

prairie. Plant Ecology 169:317–333.

LeHouérou, H. N. 1996. Climate change, drought and desertification. Journal of Arid

Environments 34:133–185.

Lips, K. 2001. Reproductive trade-offs and bet-hedging in Hyla calypsa, a Neotropical treefrog.

Oecologia 128:509–518.

Lloyd, D. G. 1984. Variation strategies of plants in heterogeneous environments. Biological

Journal of the Linnean Society 21:357–385.

Lonard, R. I., F. W. Judd, and R. Stalter. 2011. Biological flora of coastal dunes and wetlands:

Heterotheca subaxillaris (J. de Lamarck) N. Britton & H. Rusby.(Report). Journal of

Coastal Research 27:1052–1058.

Ma, W. B., X. J. Zhao, D. Y. Tan, C. C. Baskin, J. M. Baskin, and J. H. Xue. 2010. Nutlet

dimorphism in individual flowers of two cold desert annual Lappula species

(Boraginaceae): implications for escape by offspring in time and space. Plant Ecology

209:361–374.

Mandak, B. 1997. Seed heteromorphism and the life cycle of plants: a literature review. Preslia

(Prague) 69:129–159.

Middleton, N., and D. Thomas. 1997. World atlas of desertification. Arnold, London.

Orme, C. D. L., R. P. Freckleton, G. H. Thomas, T. Petzoldt, S. A. Fritz, and N. J. B. Isaac.

2012. CAPER: comparative analyses of phylogenetics and evolution in R. Methods in

Ecology and Evolution 3:145–151.

Plitmann, U. 1986. Alternative modes in dispersal strategies, with an emphasis on herbaceous

plants of the Middle East. Proceedings of the Royal Society of Edinburgh Section B:

Biological Sciences 89:193–202.

Purvis, A., and A. Rambaut. 1995. Comparative analysis by independent contrasts (CAIC): an

Apple Macintosh application for analysing comparative data. Bioinformatics 11:247–251.

R Core Team. 2017. R: a language and environment for statistical computing. R Foundation for

Statistical Computing.

Rees, M. 1994. Delayed germination of seeds: a look at the effects of adult longevity, the timing

of reproduction, and population age/stage structure. The American Naturalist 144:43–64.

Reynolds, J. F., P. R. Kemp, K. Ogle, and R. J. Fernandez. 2004. Modifying the “pulse-reserve”

paradigm for deserts of North America: precipitation pulses, soil water, and plant responses.

Oecologia 141:194–210.

Rich, T. C. G., and E. R. Woodruff. 1992. Recording bias in botanical surveys. Watsonia 19:73–

95.

Sadeh, A., H. Guterman, M. Gersani, and O. Ovadia. 2009. Plastic bet-hedging in an

amphicarpic annual: an integrated strategy under variable conditions. Evolutionary Ecology

23:373–388.

Seger, J., and H. J. Brockmann. 1987. What is bet-hedging? Oxford Surveys in Evolutionary

Biology 4:182–211.

SEINet. 2017. Southwestern Environmental Information Network. Available at:

http://swbiodiversity.org/seinet/index.php. Last accessed 08-10-2017.

Simons, A. M. 2011. Modes of response to environmental change and the elusive empirical

evidence for bet hedging. Proceedings of the Royal Society B-Biological Sciences

278:1601–1609.

Slatkin, M. 1974. Hedging one’s evolutionary bets. Nature 250:704–705.

Starrfelt, J., and H. Kokko. 2012. Bet-hedging - a triple trade-off between means, variances and

correlations. Biological Reviews 87:742–755.

Tomlinson, P. B. 1982. Chance and design in the construction of plants. Pages 162–183 Axioms

and principles of plant construction. Springer Netherlands, Dordrecht.

UNEP. 1992. World atlas of desertification. Edward Arnold, London.

Venable, D. L. 1985a. The evolutionary ecology of seed heteromorphism. American Naturalist

126:577–595.

Venable, D. L. 1985b. Ecology of achene dimorphism in Heterotheca latifolia: III.

Consequences of varied water availability. Journal of Ecology 73:757–763.

Venable, D. L., A. Burquez, G. Corral, E. Morales, and F. Espinosa. 1987. The ecology of seed

heteromorphism in Heterosperma pinnatum in Central Mexico. Ecology 68:65–76.

Venable, D. L., E. Dyreson, and E. Morlaes. 1995. Population dynamic consequences and

evolution of seed traits of Heterosperma pinnatum (Asteraceae). American Journal of

Botany 82:410–420.

Venable, D. L., and D. A. Levin. 1985. Ecology of achene dimorphism in Heterotheca latifolia:

I. Achene structure, germination and dispersal. Journal of Ecology 73:133–145.

Venable, D.L. & Pake, C.E. 1999. Population Ecology of SonoranDesert Annual Plants. In

Ecology of Sonoran Desert Plants and Communities. (ed Robichaux, R.H.). The University

of Arizona Press Tucson, Arizona, USA, pp. 115–142.

Wang, L., M. Dong, and Z.-Y. Huang. 2010. Review of research on seed heteromorphism and its

ecological significance. Chinese Journal of Plant Ecology 34:578–590.

Wojciechowski, M. F., M. J. Sanderson, and J. M. Hu. 1999. Evidence on the monophyly of

Astragalus (Fabaceae) and its major subgroups based on nuclear ribosomal DNA ITS and

chloroplast DNA trnL intron data. Systematic Botany 24:409–437.

Yoon, J.-H., S.-Y. S. Wang, R. R. Gillies, B. Kravitz, L. Hipps, and P. J. Rasch. 2015. Increasing

water cycle extremes in California and in relation to ENSO cycle under global warming.

Nature Communications 6:8657.

Zanne, A. E., D. C. Tank, W. K. Cornwell, J. M. Eastman, S. A. Smith, R. G. FitzJohn, D. J.

McGlinn, B. C. O’Meara, A. T. Moles, P. B. Reich, D. L. Royer, D. E. Soltis, P. F. Stevens,

M. Westoby, I. J. Wright, L. Aarssen, R. I. Bertin, A. Calaminus, R. Govaerts, F.

Hemmings, M. R. Leishman, J. Oleksyn, P. S. Soltis, N. G. Swenson, L. Warman, and J. M.

Beaulieu. 2014. Three keys to the radiation of angiosperms into freezing environments.

Nature 506:89–92.

Zhang, R., Y. R. Wang, J. M. Baskin, C. C. Baskin, K. Luo, and X. W. Hu. 2016. Germination

and persistence in soil of the dimorphic diaspores of Atriplex centralasiatica. Seed Science

Research 26:273–283.

Zohary, M. 1950. Evolutionary trends in the fruiting head of Compositae. Evolution 4:103–109.

Zohary, M., and A. Pascher. 1937. Die verbreitungsökologischen Verhältnisse der Pflanzen

Palästinas: Die antitelechorischen Erscheinungen. Verlag von C. Heinrich.

Zomer, R. J., D. A. Bossio, A. Trabucco, L. Yuanjie, D. C. Gupta, and V. P. Singh. 2007. Trees

and Water: Smallholder Agroforestry on Irrigated Lands in Northern India. IWMI Res. Rep.,

122.

Zomer, R. J., A. Trabucco, and D. A. Bossio. 2008. Climate change mitigation: A spatial

analysis of global land suitability for clean development mechanism afforestation and

reforestation. Agriculture, Ecosystems & Environment 126:67–80.

MAIN TEXT FIGURES

Figure 1. Distribution of contrasts obtained from the phylogenetically independent contrasts

analyses. A mean of contrast values (red bar) less than zero for aridity and duration would

indicate that the evolution of seed heteromorphism is correlated with higher aridity (lower aridity

index values) and annual duration (annual coded as 0, perennial as 1). A mean of contrast values

greater than zero for CV of precipitation and weediness would indicate that the evolution of seed

heteromorphism is correlated with higher values of CV of precipitation and weediness (non- weedy coded as 0, weedy as 1). The only significant contrast was aridity, indicated by an asterisk. Contrasts with CV of precipitation also showed the expected pattern (greater than zero)

58 while those with duration and weediness showed no pattern and the opposite of the expected pattern, respectively.

MAIN TEXT TABLES

Table 1. Summary statistics for all species analyzed, by family. Table A shows the breakdown for all seed-heteromorphic species and Table B for all seed homomorphic species. The number of species, genera, annual species, and weedy species present in each family are listed. Five seed heteromorphic varieties of Asteraceae were documented and included in this table; however, they were not included in the phylogenetic analyses. They are listed in the caption for Table S1.

Table 2. Phylogenetic independent contrasts between heteromorphic and homomorphic taxa for the four independent variables. The t-test column indicates the predicted direction of the one- tailed test with respect to zero. Thus, more negative contrasts (less than zero) in the association between aridity index and seed heteromorphism would indicate that seed-heteromorphic species are in more arid environments (lower aridity index is more arid). Similarly, for seed heteromorphism versus duration more negative contrasts (less than zero) would indicate that seed-heteromorphic species tend to be annuals (annuals coded as 0, perennials as 1). For the association between CV of precipitation and seed heteromorphism, more positive contrasts would indicate that seed-heteromorphic species occur in environments with a higher CV of precipitation. Similarly, for seed heteromorphism versus weediness more positive contrasts

(greater than zero) would indicate that seed-heteromorphic species tend to be weedy (non-weedy coded as 0, weedy coded as 1). For aridity, results are shown for all four spatial filtering distances, all of which were relatively similar.

Electronic Supplementary Material

Appendix S1. Databases and definitions used for assigning weediness.

Figure S1. Illustration of the spatial filtering technique employed.

Table S1. Floras used for data collection and number of seed heteromorphic and seed

homomorphic species found in each.

Database S1. Species tree assembly, aridity data, seed heteromorphism data, and data sources.

Database S1 is accessible via Figshare at https://doi.org/10.6084/m9.figshare.11907189.v1

(SEPARATE FILE)

Appendix S1. Databases and definitions used for assigning weediness.

Defining Weediness

Generally, we defined weediness as the degree to which a plant species associates with disturbance, either anthropogenic (Hart 1976; Hanan et al. 2015) or natural (Larson 2003). The

habitats we considered as associated with weedy species were: open woods, forest clearings,

fallow land, railroad beds, banks, roadsides, dumps, quarries, cultivated land, dry river beds,

creeks and streams (Hart 1976; Larson 2003).

Assigning Weediness

Given our large dataset, it was simply not feasible to follow the methods proposed by Hart

(1976), i.e. to review about 100 specimens of each species, document the habitat in which they

were collected and use this information to assign weediness rankings. Thus, we chose to use a

binary weediness scale and assigned weediness based on the general habitat descriptions

provided for each species in the floras we reviewed. In cases where habitat was not reported in

the floras, we used digital herbarium data (e.g. SEINet, CalFlora, EFlora, etc.), including

distributions and descriptions of habitat, to determine whether a species on our list was weedy.

For species that we could not categorize using the above two methods, we employed other

resources including journal articles or genus- and region-specific botanical websites. The source used for each species can be found in database S1. Although not necessarily accurate in absolute terms, this method has been shown to be accurate in relative terms, and thus useful for our analysis (Hanan et al. 2015).

SUPPLEMENTARY FIGURES

Figure S1. Illustration of the spatial filtering technique employed. Step (1) Each bootstrap iteration began with a reduction of occurrence points of a given species to a desired density (one point per 25, 50, 75, or 100 km2). Step (2) Thinned occurrence points were projected onto a raster map of aridity. Step (3) We extracted the raster value (aridity) associated with each occurrence point and took the average value. Step (4) We repeated steps one through three 100 times for each thinning density. Step (5) We averaged all 100 aridity means for each thinning density. Step (6) We repeated this analysis for every species. We repeated this analysis for CV of precipitation at the 25km2 filtering scheme.

SUPPLEMENTARY TABLES

Table S1. Floras used for data collection and number of seed heteromorphic and seed homomorphic species found in each flora, listed as the number of unique species (total species in parentheses). Overall, we found 101 seed heteromorphic taxa and 458 seed homomorphic species. Of these, five were varieties and were not included in our phylogenetic analyses or in this table. The varieties were Bidens pilosa var. pilosa, Galinsoga parviflora var. semicalva,

Lasianthaea fruticosa var. alamosana, Palafoxia arida var. arida, and Pectis multiseta var. ambigua. Note: the totals in each column add up to less than the total number of seed heteromorphic and seed homomorphic species because several species were found in more than one flora.

SUPPLEMENTARY REFERENCES

Baldwin, B.G., Boyd, S., Ertter, B.J., Patterson, R.W., Rosatti, T.J., Wilken, D.H., et al. (2002). Jepson Desert Manual. University of California Press, Berkley. Baldwin, B.G., Goldman, D.H. & Vorobik, L.A. (2012). The Jepson Manual: Vascular Plants of California. University of California Press, Berkley. Felger, R.S. (2000). Flora of the Gran Desierto and Rio Colorado of Northwestern Mexico. University of Arizona Press, Tucson, AZ. Guilliams, M., Veno, B., Simpson, M. & Kelley, R. (2013). Pectocarya anisocarpa, a new species of Boraginaceae, and a revised key for the genus in western North America. Aliso, 31, 1–13. Hanan, A., A. Maria, H. Vibrans, J. L. Villaseñor, N. I. Cacho, E. Ortiz, G. Gomez, and A.

Viicio. 2015. Use of herbarium data to evaluate weediness in five congeners. AoB Plants 8:

plv144.

Hart, R. (1976). An index for comparing weediness in plants. Taxon, 25, 245–247. Johnston, I.M. (1925). The North American species of Cryptantha. Stud. Boraginaceae IV. Kearney, T.H. & Peebles, R.H. (1960). Arizona Flora. Second Ed. with Suppl. by John Thomas Howell, Elizab. McClintock, Collab. University of California Press, Berkley. Larson, D.L. (2003). Native weeds and exotic plants: relationships to disturbance in mixed-grass prairie. Plant Ecol., 169, 317–333. Shreve, F. & Wiggins, I.L. (1964). Vegetation and flora of the Sonoran Desert. Stanford University Press, Stanford.

APPENDIX B

Title: Bet hedging and plasticity: An integrated strategy in a variable environment

Running title: Bet hedging and plasticity

Authors and affiliations: Joshua Scholl1* and D. Lawrence Venable1

1University of Arizona, Department of Ecology and Evolutionary Biology, Tucson, AZ 85721

*primary contact: Joshua Scholl, [email protected]

Other author email addresses: Larry Venable ([email protected])

Keywords: seed heteromorphism, offspring polymorphism, bet hedging, reproductive biology, life history evolution, plasticity, Pectocarya heterocarpa

Statement of authorship: JPS and DLV designed the study, JPS collected data, JPS and DLV performed the analyses, JPS wrote the first draft of the manuscript and DLV contributed substantially to revisions

ABSTRACT

Understanding modes of response to environmental variability is paramount to gaining predictive power regarding organismal population dynamics. Biological bet hedging may be an ecologically

important and ubiquitous mechanism for fitness maximization amidst unpredictable

environmental variability. Additionally, as environmental variability is seldom completely

unpredictable, bet hedging and adaptive plasticity may be combined by organisms. Here we investigate bet hedging in a seed-heteromorphic species, that is, one that produces multiple

morphologically different diaspores simultaneously. We collected specimens across an aridity

gradient in the southwestern United States to test whether the ratios of the different seeds

produced follow expectations from bet hedging theory. Furthermore, we germinated and raised a

subset of the field collected offspring in a common-garden environment to examine germination

differences between populations as well as plasticity associated with the ratio of basal to aerial

seeds produced. Across the aridity gradient we found that, in agreement with bet hedging theory,

the ratio of basal to aerial seeds significantly increases with increasing aridity. However, when

we grew F1 individuals from six extreme populations in a common-garden environment, the seed

ratio pattern was reversed. In addition, the observed reversal in seed ratios was greatest for the

most variable populations suggesting greater plasticity in these populations. An allometric

analysis of this data shows that larger plants produce relatively fewer basal seeds. In addition,

plants in more arid, variable habitats tend to be smaller, and this size difference operating

through the allometry contributes greatly to the among population seed ratio pattern observed in

the field. In a non-water limited greenhouse environment all plants grew much larger and plants

from more variable sites responded more strongly, growing bigger than their counterparts from

68 less variable sites. Regarding germination patterns we found that germination time in the growth chamber was fastest for plants from more variable rainfall sites, though their germination fractions were lowest as expected from bet hedging theory. These results suggest that plants from more variable sites may have the lowest field germination rates, but this may be masked in germination trials with ample water because they are much more responsive to water availability.

Overall, we provide strong evidence for bet hedging exhibited by population differentiation in phenotypic expression.

INTRODUCTION

Natural environments are dynamic and forecasting models predict that this variability, with respect to climate, is increasing globally (e.g. Berg & Hall 2015; Yoon et al. 2015).

Consequently, understanding how organisms respond to environmental variability is becoming an increasingly relevant question in biology. Most basically, organisms can have three adaptive responses to environmental variance besides extinction, namely, adaptive tracking, phenotypic plasticity and bet hedging (Simons 2011). The former two tend to maximize average fitness (Bell

& Collins 2008) while the latter sacrifices average fitness to reduce variance in fitness and maximize long-term fitness (Seger & Brockmann 1987; Starrfelt & Kokko 2012). Bet hedging effectively allows organisms to buffer against unpredictable environmental variation, such as variable desert rains, and reduces the risk of complete reproductive failure (Cohen 1966, Slatkin

1974). While the adaptive advantages are clear, strong quantitative demonstrations of bet hedging as the underlying mechanism of response strategies to variable environments are very rare (Simons 2011; Gremer & Venable 2014) and thus our understanding of its frequency of occurrence is very unclear.

An interplay between bet hedging and plasticity is increasingly recognized as an important and common strategy employed by organisms to maximize their fitness in variable environments with limited predictability (Sadeh et al. 2009; Donaldson-Matasci et al. 2013; Simons 2014;

Gremer et al. 2016). For example, Gremer et al. (2014) demonstrated that 12 Sonoran Desert winter annual species rely heavily on bet hedging with respect to their germination fractions.

When re-analyzing the same data, Gremer et al. (2016) showed that fitness benefits among the

same 12 desert annual species are generally, but not always, maximized by an integrated strategy

of bet hedging and predictive germination (plasticity).

Among the many commonly assumed bet hedging mechanisms is seed heteromorphism or the

production, by one individual, of multiple, morphologically-different offspring simultaneously

(Venable 1985; Mandak 1997; Imbert 2002; Wang et al. 2010). Seed-heteromorphic species are

routinely observed in unpredictable and extreme environments and the different seed morphs

have been shown to translate to unique ecological consequences (e.g. Venable & Levin 1985a;

Baskin et al. 2014). Yet we lack a quantitative explanation of the ratios of seed morphs these

species produce and how they relate to environmental variability, such as variability in rainfall.

Specifically, our understanding of the degree of bet-hedging and plasticity involved in the

allocation to different seed morphs among seed-heteromorphic species is nascent. Here we collect a seed heteromorphic, desert annual species from different populations across an aridity gradient spanning Arizona and California, USA. We combine this with quantitative habitat information on variability in water availability to test if changes in seed morph ratios across the gradient can be explained by environmental variability as predicted by bet hedging theory, plasticity, or an integrated strategy. This provides a uniquely, strong test of the long-standing assumption of bet hedging as the major underlying mechanism of seed heteromorphism.

To accomplish this, we focus on an amphicarpic seed-heteromorphic species, Pectocarya heterocarpa. Such species are ideal for investigations of bet hedging as they produce visually distinct and spatially separated (subterranean and aerial) seeds, which make monitoring seed ratios and germination times easier than for other seed heteromorphic strategies. In amphicarpy,

subterranean seeds are usually several times heavier than their aerial counterparts and usually

exhibit higher germination time (Koller & Roth 1963; Cheplick 1987; Zhang et al. 2015). In such cases subterranean seeds serve to reduce risk and aerial seeds to aid dispersal, respectively

(Koller & Roth 1963). Greater delayed germination of subterranean seeds also is consistent with

the general trend of higher germination times with decreasing dispersal seen across plants

(Imbert 2002). Overall, amphicarpic plants produce larger, faster-growing basal seeds with lower

germination times as compared to their aerial seeds (but see Zhang et al. 2015).

Furthermore, amphicarpic species, including P. heterocarpa usually exhibit allometry in seed ratios with bigger plants typically producing relatively more aerial seeds. Thus, environmental differences that affect plant size will affect seed ratios. This allows for plasticity in seed ratio with respect to environmental factors that affect plant size and sets up the possibility for plasticity to interact with any potential bet hedging mechanism. Specifically, amphicarpic plants start producing their basal seeds before they begin producing aerial seeds. Once aerial seed production is initiated, basal seed production slows down until a point at which remaining resources are invested almost solely in aerial seed production (Cheplick 1987; Sadeh et al. 2009).

This results in an allometric relationship between basal and aerial seeds and suggests that the final basal to aerial seed ratio is highly plastic and driven largely by environmental conditions that determine plant size (Callahan & Waller 2000). For example, favorable conditions (e.g. higher than average rain) will result in plants with very low basal to aerial ratios (McNamara &

Quinn 1977; Cheplick 1987). Likewise, Emex spinosa seeds collected from two sites differing in environmental variability, shifted their seed ratio in response to favorable greenhouse conditions and those from the more variable site did so to a much greater degree (Sadeh et al. 2009).

Throughout this paper we use the term plasticity to refer to seed ratio changes with plant size and

reproductive output, under the assumption that variation in plant size is heavily influenced by

environmental factors including water, temperature, and nutrients. Meanwhile we recognize that

some aspects of plant growth rate and size have a genetic component too.

Preliminary work on the life history of P. heterocarpa shows greater delayed germination and

increased drought resistance in basal versus aerial seeds (Huang et al. 2015). Pilot studies also

have shown that it pursues the typical amphicarpic strategy with respect to allometric

reproduction. In the present study we collected P. heterocarpa, a desert winter annual species,

from different populations across an aridity gradient in southwestern USA to test if the ratio of

slow germinating and low dispersing basal seeds to fast germinating and high dispersing aerial seeds increases with increasing variability in water availability as predicted by bet hedging theory. In so doing we address five questions: 1) Do the ratios of basal to aerial seed morphs

increase with increasing unpredictability in rainfall as predicted by bet hedging theory? 2) Are

the field-observed patterns in ratios of basal to aerial seed morphs maintained in a common environment in the greenhouse? 3) Does the allometry in reproduction differ between sites with

different aridity? 4) Across the gradient, does P. heterocarpa conform to the typical amphicarpic

strategy in which basal seeds have the longest time to germination and lowest germination

fraction followed by aerial seeds? 5) Are germination times and fractions reduced across sites

and within seed morphs based on increasing aridity?

METHODS

Natural History

Pectocarya heterocarpa (I.M. Johnston; Boraginaceae) is a winter, annual plant which

simultaneously produces multiple morphologically distinct nutlets (Fig. 1-2). Cauline nutlets originate from chasmogamous, or opening, flowers, each of which produce a fruit of four nutlets consisting of two abaxial and two adaxial nutlets (Veno 1979). The adaxial nutlets are longer and have a greatly reduced margin as compared to the abaxial ones. In addition, the pedicel is fused to one of the adaxial nutlets (Veno 1979). In general, both adaxial nutlets exhibit stronger attachment than the abaxial nutlets, which very easily detach upon maturation. The conspicuous margin or “wing” surrounding the abaxial nutlets combined with their propensity for detachment suggest that their main role is dispersal and colonization of new patches.

Basal nutlets originate from cleistogamous, or non-opening, flowers. As in cauline flowers, four

basal nutlets are produced per fruit. In contrast to cauline nutlets, all basal nutlets have severely

reduced ornamentation in the form of hairs and margins. Basal nutlets are also very hard to

detach and typically remain attached to the plant until the subsequent growing season. Often,

basal nutlets germinate in a subsequent season, still attached to the woody remnant of the

maternal plant and frequently even grow through its remaining hollow stems. Although basal

fruits have both adaxial and abaxial pairs of nutlets like cauline fruits, the differences between

them are usually very subtle and they are hard to distinguish. Hence, in this paper we distinguish

three different nutlet forms on P. heterocarpa: cauline long nutlets, cauline winged nutlets and

basal nutlets.

P. heterocarpa germinates with the onset of winter rains from October to January with occasional cohorts germinating as late as early March. Depending on the timing and amount of rainfall as well as latitude and temperature, plants flower between February and April (Mulroy &

Rundel 1977) and die in late March or early April. Complete failure to mature and reproduce often occurs when there is not enough rainfall subsequent to initial germination rains. The species is found throughout arid regions of Arizona, southern California, Nevada, Utah, New

Mexico and Northwestern Mexico (Johnston 1939; “SEINet” 2017). Throughout this broad distribution it experiences a range of environments which differ in their total amount of rainfall and aridity (Table 1). For example, Blythe, California lies in the center of its distribution and is characterized by low rainfall amounts and high rainfall variability. In contrast, Portal, Arizona lies more towards the species range edge and may exhibit some of the highest total rainfall and lowest rainfall variability the species encounters anywhere in its range.

Field Plant Harvesting

We selected 12 populations spread across southern Arizona and California encompassing the

Sonoran, Mohave and Chihuahuan Deserts, USA. Each population is within 20 km of a weather station for which climatic data are available from the National Oceanic and Atmospheric

Administration (NOAA). The historical precipitation record depicts significant differences among the 12 locations (Table 1). Mean winter rainfall (October–March) ranged from a low of

54mm (Ocotillo, CA) to 185mm (Portal, AZ) at the most mesic site (Table 1).

Quantifying environmental unpredictability is difficult. Past studies of seed heteromorphism

have simply relied on general biome descriptions such as “desert,” “semi-desert,”

“Mediterranean” and “montane” (Mandak 1997; Imbert 2002; Cruz-Mazo et al. 2009; Wang et al. 2010). Here we use an aridity index as a practical and globally available indicator of variability in water availability. Aridity is a quantitative way of accounting for many factors which influence water stress for plants including sunlight, rainfall, temperature and elevation making it a nuanced and relevant measure of water availability (LeHouérou 1996; Elfaki et al.

2011). Also, previous studies have shown a direct tie between aridity and unpredictability in water availability (Davidowitz 2002; Berg & Hall 2015; Yoon et al. 2015). Weather data required to calculated aridity were downloaded as daily summaries from the National

Oceanographic and Atmospheric Administration (NOAA 2018) for all 12 sites. Data were subset

(to include only winter season), cleaned (checked for missing rows in days or month and missing values in weather metrics) and summarized (averages, etc.) using the R statistical software (R

Core Team 2017). In rare cases where weather station data within a site for a particular year or particular months was missing, a neighboring weather station was used as a substitute. We were careful to select collection sites based on the availability of data from nearby, and similar elevation weather stations. An aridity index was calculated by dividing the precipitation by the

potential evapotranspiration, with lower numbers denoting higher aridity or drier sites (PET;

UNEP 1992). For clarity, we generally refrain from referring to this index throughout the paper

and instead refer to sites as being more or less arid, in the general sense, meaning more or less

dry. Precipitation data came from NOAA and PET was calculated for each month in the winter

growing season (Oct. – Mar.) using the Thornthwaite (1948) equation. Generally, populations

with lower mean winter precipitation had lower aridity values (Table 1).

At each site we established a belt transect within which we used a random number table to

determine how far along and to either side of the center of the transect to collect each of 20 individuals. Plants were collected up to 5m to either side of the center of the belt transect

resulting in a collection area of 600m2 or a 10m by 60m belt transect. Plants were collected upon

death. Due to differences in rainfall across the sites, the collection dates for sites differed. We

carefully cut plants at the top of their roots and placed them in plastic Ziploc containers for

separation into aerial and basal parts. The Ziploc containers captured any aerial seeds that were

shed during handling. Basal seeds are firmly attached to the plant and very difficult to remove

and thus only rarely fell off during handling. Aerial and basal parts were then transferred to

labeled coin envelopes. Any aerial seeds that came lose were also transferred to the aerial parts

envelope.

In the lab we dried the collected plants inside their envelopes in a Fisher Isotemp Drying Oven

(200 Series) at 30C for at least 48 hours and no more than 72 hours. Plants were subsequently

weighed to the nearest mg using an XS105 Dual Range Mettler Toledo scale. Next, we separated

and counted the seeds on all the collected plants. For the aerial parts we counted the adaxial or

long seeds separately from the abaxial or winged seeds. For the basal parts we counted the total

number of seeds. All counted seeds were stored in separate coin envelopes for each seed type

and maternal plant. Although P. heterocarpa produces aerial fruits of four seeds, two of which

are long and two of which are winged, we generally had significantly lower counts of winged

seeds. We attribute this to the ease of detachment of the winged seeds which we have

anecdotally observed detaching even before the whole plant dries out. For this reason, we

assumed that matured winged seeds were equal to matured long seeds and thus doubled our

count of long seeds for our analyses of ratios of basal to aerial seeds. This is supported by our

observations of whole fruit abortions in nature and the greenhouse and only rarely, individual

seed abortions.

Germination Times and Fractions

Seeds collected in part 1 were over-summered for three months at the Desert Laboratory on

Tumamoc Hill in Tucson, Arizona (Adondakis & Venable 2004). Specifically, seeds were placed

in fine-mesh polyorganza bags which were stored inside closed plastic storage bins at our

Tumamoc field site. Experiencing hot summer temperatures is important for natural after-

ripening and dormancy-breaking for Sonoran Desert winter annual species (Adondakis &

Venable 2004; Huang et al. 2015). After over-summering seeds, we conducted germination trials in agar-plated petri dishes in three separate growth chambers using three different temperature settings.

Germination trials for the F1 generation were initiated on January 6th, 2016. We conducted power analyses to determine sample sizes for this and subsequent experiments with respect to germination data based on Huang et al. (2015). We selected three of each of P. heterocarpa’s three seed morphs, from each of 20 plant samples and six populations for a grand total of

1080W, 1080L and 1080B seeds. In some cases, there were not enough seeds from certain individual plants to reach the desired seed numbers, this was particularly true for plants stemming from extremely dry sites. In these cases, seeds from other individuals of the same population were used as a supplement. 360 seeds of each type (W, L, B) were put in each of

three growth chamber temperatures, 12C/16C, 8C/12C and 4C/8C which were set to 12hr light

and 12hr dark with the lower of each temperature being the dark or night temperature. These

temperatures and daylengths are similar to those experienced in the field during the germination

months. Germination was monitored every four hours. After 168 hours or seven days, 75% of all

seeds had germinated so we reduced the monitoring time to once every 12 hours. After 336 hours or 14 days, 79% of seeds had germinated so we reduced it to once every 24 hours. After 1200 hours or 50 days, 80% of all seeds had germinated and we ended the experiment. We repeated this procedure for the second generation of seeds collected from the greenhouse from the F1 plants as shown in figure 3.

Greenhouse Common-garden Experiment

Throughout the F1 and F2 germination trials, germinants were carefully transplanted to individual conetainers, which were then randomly placed in racks of 20 conetainers and immediately placed in the greenhouse. Seedlings transplanted to the greenhouse were all taken from the 8C/12C growth chamber regimen. We attempted to transfer at least 14 seedlings of each seed type from each population but this was not always possible due to a lack of germinants or a lack of enough initial seeds from certain arid populations that had only very small, seed-poor plants. Each conetainer was filled with a 90-grit silica sand and SunGro sunshine mix #3 soil mixture in a 2:3 ratio, respectively. Germinants were transplanted once they achieved a radicle length of 8-15 mm and a hypocotyl length of 5-12 mm. Greenhouse temperatures were set to mimic outdoor growing season conditions for Jan-Mar (Jan=18C/5C, Feb=20C/7C,

Mar=23C/9C; NOAA). In addition, plants received mostly natural light with some shading in the

morning and afternoon due to non-removable greenhouse shutters that shaded some portions of

the bench on which plants were placed. To minimize the influence of this shading we randomly

moved conetainer racks and the plants within the conetainer racks weekly. In the greenhouse,

plants were watered often and liberally such that each plant received ca. 100mL every third day.

Plants were harvested in late March and early April upon death and processed the same way as previously described for field collected plants.

Statistical Analyses

First, we compared the 2014-2015 winter aridity of each population to the 29-year average

aridity of each population (1986-2016, excluding the 2014-2015 season) using ordinary least

squares regression (OLS). Differences of 2014-2015 from the long-term average were evaluated by comparing the resulting regression line to a reference line (1:1 line) using a linear hypothesis

test implemented in the car package (Fox & Weisberg 2018) for the R statistical software (R

Core Team 2017).

For the field collected plants and the first (F1) and second (F2) greenhouse generations, we

evaluated the influence of aridity on the basal to aerial seed ratio, basal seeds produced, and

aerial seeds produced using OLS. For data normalization, all response variables were

transformed using the natural logarithm prior to analyses. For any plants with zero basal seeds,

we changed the basal seeds produced from zero to one to avoid an undefined logarithm.

We evaluated the allometric relation of basal and aerial seed production using OLS to determine

if allometric seed production varied across sites and if this variation could be attributed to aridity.

In this analysis we treated basal seeds set as the response variable and used aerial seeds set and

site of origin as predictor variables (the slope of log basal on log aerial seed types is the allometric constant). For greenhouse plants we added year (for F1 or F2) as an additional predictor variable. We analyzed both the additive effect of predictor variables and their

interactions. Non-significance of the interaction between site of origin and aerial seeds set

indicates that the allometric constant (slope) is invariant across sites. Significance of the additive

effect of site of origin in the model indicates a changing intercept and thus that plants differed in

their basal to aerial ratios for any given plant size.

We evaluated the influence of site or the site’s aridity, seed type and growth chamber

temperature treatment on germination fractions and times using generalized linear mixed models

(GLMMs) with petri dish included as a random effect in all models. Germination time in hours

was transformed using the natural logarithm. Models for germination fraction were fit using

binomial error distributions and logit link functions to test for the effects of predictor variables

on the probability of seed germination. Significance of seed type indicates differences in

germination times or fractions across the different seed types while significance of aridity

indicates differences in germination characteristics between the different sites of origin.

For analyses with multiple variables, various models were constructed using the top-down

technique (Diggle et al. 1998; West et al. 2014). The best model for each response variable was

selected using the Akaike Information Criterion (AIC). All statistical tests were conducted with

R statistical software (R Core Team 2017). The R package nlme (Pinheiro et al. 2019) was used to specify GLMMs. The significance of fixed effects was evaluated with Type II tests using the car package (Fox & Weisberg 2018).

RESULTS

Seed ratios and aridity in the field and in the greenhouse

During the field collection season, winter 2014-2015, the aridity index at each collection site, except four (Red Rock, Wikieup, Barstow, Tumamoc Hill), was lower than its 30-year average, indicating generally dryer than average conditions across the gradient (Fig. 4). When the extreme outlying site, Portal, was not included, the slope of 2014-2015 aridity versus the average aridity did not differ from one. A graph with Portal included is provided in the supplemental materials

(Fig. S1). Hence, while the year of our field collection was slightly drier than average, rankings among sites were not different than average.

For the field collected plants, unpredictability in the environment as measured by aridity explained 18% of the variation in seed ratios when the outlying Portal site was removed and 22% when it was included (Fig. 5A and S2). The ratios showed the expected pattern of increasing basal to aerial seed ratios with increasing unpredictability in the environment (R2 = 0.18, p <<

0.001, slope = -0.16; Fig. 5A). For the F1 and F2 seeds grown in a common-garden environment in the greenhouse, aridity also explained between 14% and 18% of the variation in seed ratios.

However, the pattern of increasing basal to aerial ratios with decreasing aridity index (more arid)

of the site, was reversed (F1: R2 = 0.14, p << 0.001, slope = 0.62; F2: R2 = 0.18, p << 0.001,

slope = 0.57; Fig. 5B-C). Specifically, plants from more unpredictable sites (more arid) increased

their aerial seed production under the relatively wet greenhouse conditions as compared to the

field, much more than their counterparts originating from more predictable (less arid) sites (Fig.

6A-C). Consequently, the relationship between the number of aerial seeds and the aridity index was positive for field collected plants and negative for the F1 and F2 generations. Thus, for field collected plants more aerial seeds were associated with more mesic sites while for the F1 and F2 generations the opposite was true. In contrast, basal seed production, although increased in the greenhouse compared to the field for plants from all sites, did not exhibit a reversed pattern between the field and greenhouse generations. The slopes of basal seeds versus the aridity index for the greenhouse generations, while both positive (more basal seeds in more mesic sites) like that of the field collected plants, were not significantly different from zero (Fig. 6D-F).

Regarding the fitness of different seed types, we did not detect any significant differences in plant weight, or basal, long or total seeds set by plants originating from different seed types for the F1 (all p > 0.5; Fig. S3) or the F2 generations ((all p > 0.3; Fig. S4).

Seed ratio allometry in the field and greenhouse

All field collected plants displayed an allometric relationship in the relative production of basal as compared to aerial seeds (Fig. 7A). The slope, or allometric constant, for log transformed basal vs log transformed aerial seeds across all sites was 0.47 (Table 2, Model 1). Thus, plants with greater seed production have relatively more aerial seeds as compared to basal seeds. This slope, or allometric constant, did not vary significantly with population (p = 0.09; Table 2,

Model 2). The intercept of this allometric relationship did, however, vary with population (p <<

0.0001; Table 2, Model 3), implying that populations differed in their basal to aerial ratios for any given level of seed production. The difference in intercept between sites does not differ with site aridity (p = 0.295, Table 2, Model 4) but is correlated to site differences in mean plant weight, with populations with heavier plants originating from less arid sites and having relatively more aerial seeds (partial R2 = 0.094, p << 0.0001; Table 2, Model 5).

For the six populations grown in the greenhouse for two generations, the F1 and F2 generations displayed a significant overall allometric relationship of basal to aerial seeds (F1: slope = 0.23, p

<< 0.001; F2: slope = 0.58, p << 0.0001; Table 2, Models 6 and 11). Thus, for the F1 generation, basal seeds increased less rapidly with aerial seeds as compared to the field plants (Fig. 7B). For the F2 generation, the opposite was true such that basal seeds increased more rapidly with aerial seeds as compared to the field plants (Fig. 7C). Furthermore, this allometric constant varied significantly with population for the F1 generation (p < 0.001; Table 2, Model 7) but not the F2 generation (p = 0.4; Table 2, Model 12) indicating a varying allometry across collection sites for the F1 generation but a constant allometry across all sites for the F2 generation. For both the F1 and F2 generations, this allometric curve varied significantly with population (F1: p < 0.01; F2: p

= 0.01; Table 2, Models 8 and 13) suggesting that populations in the F1 and F2 generations differed in their basal to aerial ratios for any given common level of seed production. However, for the F2 generation this significance was due to one site, Barstow (Fig. 7C), which when removed results in an insignificant trend (p = 0.4; Table 2, Model 14). For the F1 generation the difference in intercept between sites differs significantly and positively with aridity (F1: partial

R2 = 0.03, p << 0.001; Table 2, Model 9) and is also correlated to site differences in mean plant

weight (partial R2 = 0.24, p << 0.001; Table 2, Model 10). For the F2 generation this trend with

aridity was significant (partial R2 = 0.09, p < 0.001; Table 2, Model 15) but not significant when

Barstow was removed (partial R2 = 0.01, p =0.1; Table 2, Model 17). For the F2 generation the

difference in intercept between sites differed significantly with mean plant weight with Barstow

included (partial R2 = 0.24, p << 0.001; Table 2, Model 16) and with Barstow removed (partial

R2 = 0.17, p << 0.001; Table 2, Model 18).

To summarize, the allometric results for field collected plants and the F1 and F2 generations were similar. Plants that produce more seeds produce relatively fewer basal vs aerial seeds across all populations. Thus, some of the population variation in seed ratio is due to this general allometry interacting with the observation that some populations have lower seed production.

Additionally, there is an effect of mean plant weight differences among populations, but not aridity differences (except for the F1 generation and when Barstow is included for the F2 generation), on seed ratio for plants of a given level of seed production. Thus, plants in arid sites produce fewer seeds which accentuates basal seed production through the general allometry, and even more basal relative to aerial seed production due to an additional effect of plant size on ratio at a given level of seed production.

Germination times and fractions

Average germination time of basal, long and winged seeds differed as expected with the basal seeds having the longest average time to germination, followed by the long seeds, and finally the winged seeds which germinated fastest (Fig. 8, Table 3, seed type main effect, P << 0.0001).

Seeds also germinated faster in hot then warm then cold treatments (temperature main effect, P

<< 0.0001) and the difference in germination time between seed types was greatest in the hot and least in the cold (temperature x seed type interaction P << 0.0001). More arid sites had slightly slower germination for winged seeds, but faster germination for basal and long seeds (seed type

* aridity interaction P << 0.0001).

Winged seeds had the highest germination fractions and basal seeds the lowest (seed type main effect, P << 0.0001; Fig. 9, Table 3). Sites with more arid winter seasons had lower germination fractions for all treatments and seed types (main effect of aridity, p << 0.0001) but the magnitude of the aridity effect varied with seed type with the germination fraction of basal seeds being least sensitive to differences among sites in aridity (P << 0.0001). Temperature and its interactions had no significant effects and were not part of the best AIC model.

DISCUSSION

Seed heteromorphism among angiosperms has long been heralded as a bet-hedging mechanism but sound quantitative, empirical evidence remains scarce. Furthermore, the interplay between bet hedging and plasticity is likely important but remains largely uninvestigated. Here we investigated the change in ratios of different seed morphs of a seed heteromorphic, amphicarpic, annual plant species across an aridity gradient to evaluate evidence for bet hedging. We also raise field plants in a common-garden environment to assess the role of plasticity. While we

found some evidence for bet hedging our results highlight the significant contribution of

plasticity to the assumed bet hedging strategy of seed heteromorphism in P. heterocarpa.

First, we found that unpredictability in winter water availability alone explains 22% of the

variation in seed morph ratio shifts across an aridity gradient with more variable sites producing

more basal relative to aerial seeds as predicted by bet hedging theory. An allometric analysis of

this data shows that larger plants produce relatively fewer basal seeds, plants in more arid,

variable habitats tend to be smaller, and this size difference operating through the allometry

contributes substantially to the among population pattern described above.

Second, when field collected seeds were grown in a non-water limited greenhouse environment,

seed heteromorphism persisted but the pattern of seed morph ratios with unpredictability was

reversed. Specifically, maternal lines originating from more variable sites produced the lowest

basal to aerial ratios. A close examination of this shift in ratio between generations revealed that it came largely from an increased production of aerial seeds while the production of basal seeds remained largely unchanged across populations. The field to greenhouse shift in seed ratios was greatest for the most variable populations suggesting that they may have high plasticity that allows them to respond to rare favorable conditions like those encountered in the greenhouse.

This pattern was counter to bet-hedging predictions and highlights instead the potential role of

plasticity in fine-tuning this species reproductive strategy. This evidence of plasticity is further

supported by the lack of aridity as a significant predictor of the F2 generation’s allometry.

Third, aridity explains differences in germination fractions with more variable sites having lower

germination fractions for each seed type as predicted by bet hedging theory.

Overall, our findings suggest that P. heterocarpa achieves the predicted pattern from bet hedging

theory by producing a greater proportion of basal seeds in more variable, arid environments and

having a lower germination fraction of all seed types in more variable, arid environments. Yet

the adaptive pattern of seed ratio variation appears to be almost entirely due to a shared general allometric pattern of seed production and a pattern of smaller plants with lower seed production being found in more arid sites. It behooves us to ask if this apparently adaptive pattern achieved through plasticity is just a fluke due to conditions in the year we collected our field data or if it is

typical across years. To understand this, we compared aridity of our sites in our year of field

study with the long-term average aridity and found that the 2014-2015 season, while a little drier than average, deviated similarly along the gradient. Thus, it is likely that a pattern of size differences interacting with a fixed allometry will often generate this adaptive pattern. In unusual years when, for example, it might be much wetter than average in dry sites and much drier than average in wet sites, the plastic response will alter the deployment of seed ratio accordingly across sites. But the similarity of the 2014-2015 weather to the long-term average suggests that

the pattern we found will occur frequently. These results highlight the role of plasticity in

allowing this species to respond quickly and dramatically to alter its bet-hedging strategy in response to environmental changes. This affords the plant with a great degree of flexibility with respect to its final ratio of basal to aerial seeds.

Environmental unpredictability and seed morph ratio shifts

Bet hedging theory predicts that increases in environmental variability should result in species spreading risk or adopting a more risk averse strategy (Cohen 1966; Slatkin 1974). In the case of a typical amphicarpic annual plant, such as P. heterocarpa, that produces low dispersing and slow germinating basal seeds and high dispersing and high germinating aerial seeds, this typically translates to increasing the ratio of basal to aerial seeds if the variability lies in water availability (Cheplick 1994). We found this expected pattern in our study for the field collected plants. While only a handful of studies have investigated seed ratio shifts across the landscape in seed-heteromorphic species with respect to a quantitative measure of variability (McNamara &

Quinn 1977; Cheplick & Quinn 1982; Venable et al. 1998), they have found similar patterns. In addition, studies investigating this pattern qualitatively in nature and quantitatively only in controlled greenhouse and natural experiment settings (McNamara & Quinn 1977; Weiss 1980;

Cheplick & Quinn 1982, 1983; Lerner et al. 2008; Sadeh et al. 2009) also find similar patterns to the one we detected in the field.

For example, McNamara and Quinn (1977) investigated shifts in the ratio of aerial to basal seeds across five populations of Amphicarpum purshii Kunth (Poaceae) with respect to a disturbance

(fire) gradient. They found that the ratio of aerial to basal seeds typically decreased with less recent disturbance. That is, plants from sites with recent fires tended to have higher numbers of aerial seeds as compared to those from fire suppressed and thus largely overgrown habitats.

While their observed seed ratio shifts were only significant for one site and their measure of disturbance was more qualitative than quantitative, these results have been supported in subsequent research on the same species (Cheplick & Quinn 1982, 1983). Another study by

Weiss (1980) investigated the properties of seeds of Emex spinosa (Polygonaceae) from five

89 natural populations. Although the populations were very close spatially (ca. within 5km2) and he did not explicitly investigate seed ratios it can be seen from his results (Table 4, Weiss 1980) that the ratios of aerial to basal seeds differed significantly between the sites. Specifically, the number of basal seeds produced between sites was relatively similar, but the number of aerial seeds varied from an average of 16 at one site to an average of 434 at another. He also experimentally manipulated the growing environment of field collected seeds in the greenhouse and showed that limiting nitrogen and increasing density resulted in relatively large reductions in aerial seed production but had little effect on basal seed production. A third study of an amphicarpic species by Lerner et al. (2008) explored two Poaceae species and found contrasting results in an experiment testing seed ratios in response to nutrient stress. They found that their native grass species reduced its basal to aerial seed ratio in response to nutrient reductions while the invasive grass increased its basal to aerial seed ratio as was expected. This appears to highlight different strategies. Reducing the basal to aerial ratio under stress may be an attempt to escape in space or reduce risk across space (i.e. high dispersal of aerial seeds) while increasing the basal to aerial ratio may be an effort to reduce risk in time (i.e. high dormancy of basal seeds).

Studies of other seed heteromorphic but non-amphicarpic species also show similar results to our field collected plants. While these plants do not produce their different seed morphs with such dramatic spatial separation as amphicarpic plants, such plants still produce morphs that combine fast germination with high dispersal and morphs that combine slow germination with low dispersal. In their investigations of Heterosperma pinnatum (Asteraceae) Venable et al. (1987,

1995) found an increase in the ratio of slow germinating and low dispersing seeds to fast

90 germinating and high dispersing seeds which appeared to be correlated with decreasing seasonal precipitation and decreasing canopy coverage. In another study they explicitly tested this prediction quantitatively and found that the percentage of dispersing seeds set was strongly and positively associated with canopy coverage (Venable et al. 1998). In addition, they found that H. pinnatum populations growing in environments with a more reliable onset of the germination season, and thus lower germination risk, produced relatively more fast germinating seeds. These patterns were somewhat correlated as high dispersing seeds of H. pinnatum tend to also be fast germinating. Many of the studies cited above are considered as providing evidence of bet hedging among seed-heteromorphic species, however, none of them grew F1 generations in a common garden to evaluate the contribution of plasticity.

Seed morph ratios in a non-water limited common garden

By growing our plants in a common-garden environment we sought insights into the contribution of genetic and plastic components to the seed ratios observed in the field. While we found that plants from all sites produced both aerial and basal seeds, showing that seed heteromorphism is a fixed trait in P. heterocarpa, we also found that the plants from more variable sites had a greater shift in their aerial to basal ratio as compared to the ratios their maternal plants had in nature. A close look at this shift shows that the degree to which plants adjusted their ratios in the greenhouse is related to the variability of their maternal site. This pattern suggests that the plants from sites with high rainfall variability have greater plasticity in their ability to produce aerial seeds during periods of favorable conditions. Perhaps with fewer opportunities to produce large

quantities of seeds, but also diminished competition in their environment because of its harsh

nature, they have been selected to respond quickly to rare opportunities.

Such plant responses have been demonstrated extensively in Amphicarpum purshii, a pioneer

species in secondary succession (McNamara & Quinn 1977; Cheplick & Quinn 1982, 1983). As succession progresses and vegetation density increases, A. purshii responds by increasing its basal to aerial seed ratio dramatically (Cheplick & Quinn 1983). Its basal seeds show

significantly higher survival rates in the soil seedbank and upon germination amidst competition,

as compared to its aerial seeds. Upon disturbance the species responds by decreasing its basal to

aerial seed ratio to take advantage of available microsites through its smaller, dispersive aerial

seeds. As with P. heterocarpa, the increases and decreases in seed ratios in A. purshii are driven

predominantly by changes in aerial seed production (McNamara & Quinn 1977; Cheplick &

Quinn 1983).

A similar result was observed in another species displaying multiple seed strategies, Emex

spinosa. When plants collected from two different sites were raised in a common garden, plants

from the more variable site had a larger plastic response than those from the less variable site

(Sadeh et al. 2009). Specifically, they more drastically increased their production of aerial seeds

as we saw in our common-garden environment.

Moreover, Martorell and Martinez-Lopez (2014) experimentally explored plasticity in seed

production of H. pinnatum, a seed-heteromorphic species of the semi-arid and arid parts of

Mexico and Southwestern US. When sown along a natural moisture gradient with varying levels

of shade and intraspecific competition, H. pinnatum decreased its proportion of high dormancy

seeds with increasing competition and with increased shading. The increase in production of low

dormancy achenes under shading was interpreted as an adaptive response to increasing water

availability. As water availability was correlated to increased shading and low-dormancy achenes to more-dispersive achenes, it stands to reason that the increase in dispersing seeds with shading

could be an adaptive response to locally occupied safe sites. Interestingly, the plasticity in seed

ratio shifts observed in H. pinnatum is opposite that of A. purshii. A. purshii reduces production of its dispersing aerial seeds in response to shading and increased vegetation density in an apparent attempt to escape in time while H. pinnatum increases it in response to these conditions in an apparent attempt to escape in space.

The seed-heteromorphic species discussed above all displayed high plasticity in seed ratios.

Nevertheless, these species appear to always produce more than one type of seed with plasticity serving as an important mechanism to adaptively modify the ratio. This suggests the importance of seed heteromorphism as a bet hedging mechanism in the face of uncertainty, because if plants could accurately predict which seed type will be best, we would expect them to only produce aerial seeds in favorable conditions, especially by the second generation grown in favorable conditions (e.g. our F2 lines), and only basal seeds in unfavorable conditions.

Allometric Relationships

Our allometric analysis allowed us to investigate differences between populations in their slope and intercept of basal to aerial seed ratios. For the field collected plants we found no significant

93 effect of aridity. For the F1 and F2 generations we found a significant and positive but very small effect of aridity. In addition, the significance of this trend for the F2 generation was entirely due to Barstow which had an unusual allometry. When the Barstow site was removed, so was the effect of aridity. While an allometric relationship in the reproductive rate of amphicarpic plants is well established, we are unaware of studies that have analyzed it to attribute differences in slope and intercept across populations to environmental uncertainty.

Germination timing and fractions across our aridity gradient

Overall, basal, long and winged seeds had consistent and significantly different germination times across treatments and populations (Fig 8). These different germination responses establish the potential for differential fitness outcomes for plants originating from the different seed types, due in large part to the interaction of germination time with growing season length but also with variability in the size and timing of rainfall events (e.g. Venable et al. 1995; Weinig 2000;

Donohue et al. 2005). Furthermore, we found that germination time of basal and long seeds in the growth chamber was, on average, fastest for plants from the more variable rainfall sites in all three growth chamber treatments (Fig. 8). There was no discernible pattern for winged seeds with site variability. The pattern for long and basal seeds disagrees with our hypothesis that plants from more variable sites would have slower germination times corresponding to greater risk spreading within and among growing seasons (Tielbörger et al. 2012; Gremer et al. 2016).

But germination time does not respond to selection per se, instead it is the physiological and morphological mechanisms that control germination responses that evolve. Since increasing rainfall variability is associated with lower overall rainfall (Davidowitz 2002), plants from more

variable sites typically experience less rainfall than those from less variable and more mesic

sites. Consequently, the species from more variable rainfall sites are likely physiologically and

morphologically much more sensitive to water availability. Hence, although field germination

times are probably slower for species from more variable sites, exposing all populations to the

same, high moisture levels inside the petri dishes used for the germination tests results in the

reverse pattern. This type of counter-gradient greenhouse result has been observed in other

winter annuals where a similar conclusion was made based on field germination trials (Clauss

and Venable 2000).

Germination fractions of each seed type showed the expected pattern of decrease in response to increasing variability in water availability as expected from bet hedging theory (Cohen 1966;

Gremer & Venable 2014). While we are unaware of studies showing this among seed-

heteromorphic species, this link between decreasing germination fractions and increasing environmental variability has been demonstrated broadly among non-seed-heteromorphic species

(e.g. Clauss & Venable 2000; Arroyo et al. 2006; Venable 2007; Tielbörger et al. 2012; Gremer

& Venable 2014). Interestingly this pattern occurred among all seed morphs in our study. We

expected it among the basal seeds but less so among the dispersing aerial seeds, especially the

winged aerial seeds. This is because dispersal and dormancy are theoretically expected to trade

off, all else being equal (e.g. Venable & Lawlor 1980; Levin et al. 1984), though because all else

usually is not equal in nature the empirical relationship between the two remains murky (e.g.

Siewert & Tielbörger 2010; De Casas et al. 2015). In any case, both of the significant differences

between the different seed types in germination time and fraction serve as supporting evidence

that seed heteromorphism in P. heterocarpa serves as a bet-hedging strategy (Simons 2011).

We also assessed plant differences due to originating seed type in the F1 generation but found no significant differences in plant weight, seed counts or ratios upon maturity. At least one other study found that seed morphs of a seed-heteromorphic species differ only in their germination characteristics and dispersal (Lenser et al. 2016). In both our study and theirs, the ecological consequences of the different seeds appear to be primarily driven by their emergence time and place, which can have significant fitness impacts (Venable et al. 1995, Donohue et al. 2005, Lu et al. 2016). Like Lenser et al. (2016), our primary focus was not on assessing the differences between plants arising from different morphs, so we did not monitor other earlier post- germination characteristics such as time to first flowering and early relative growth rate. We also did not stress our plants by limiting water or nutrients or expose them to competition. Early germination characteristics have been shown to differ significantly between plants arising from basal and aerial seed morphs and can translate to significant fitness differences under high density or when nutrients or moisture are limited to simulate environmental stress (Cheplick &

Quinn 1982; Venable & Levin 1985b; Lerner et al. 2008; Wang et al. 2012).

Unfortunately, we had real difficulty, with all six sites, in getting our F2 generation of seeds to germinate. Consequently, we were unable to robustly analyze the germination times and fractions as we did for the F1 generation. In another study of a sister taxa of P. heterocarpa, namely P. recurvata, the authors ran into similar germination issues with their F2 generation

(pers com. Monica Ge). This raises interesting questions regarding the germination requirements of our species that warrant further investigation.

Fitness consequences of different seed morphs in the greenhouse

Contrary to other studies, and specifically research on amphicarpic species (Cheplick 1987), we did not find any differences in fitness metrics of plants germinating from different seed types. In part, this may be due to our small sample sizes but it could also be due to there just not being any important differences in plants coming from different seed types in a non-competitive, non-water limited, constant greenhouse environment, especially when the effect of germination timing is removed as it was for our greenhouse experiments (all seeds transplanted at the same time).

Fitness differences among seed types in nature may operate through germination time within and between years as a result of the timing of rainfall in more variable natural environments

(Donohue 2002; Donohue et al. 2005).

Conclusions

Presently, the interplay of bet-hedging and plasticity and the environmental conditions that foster it are relatively poorly characterized empirically as very little research focuses on understanding bet-hedging and even less on how it interacts with plasticity. Our results provide a strong test of bet hedging predictions in an amphicarpic, seed-heteromorphic species and highlight the increasingly recognized importance of the interplay between bet-hedging and plasticity (Sadeh et al. 2009; Simons 2011; Gremer et al. 2016). Clearly there is a genetic component to seed heteromorphism in P. heterocarpa as the species always produces three different types of seeds

(at least across two greenhouse grown generations) and the ratio of basal to aerial seeds is significantly different between populations for both field collected and greenhouse grown plants.

In addition, it appears that the ratio of these different seeds is very strongly adjusted by plasticity

as plants from the field do not exhibit the same ratios as their offspring grown in the greenhouse.

Effectively, the predicted bet hedging pattern of plants producing relatively more slow

germinating and low dispersing seeds in more variable environments, is achieved through

plasticity.

ACKNOWLEDGEMENTS

This research was supported by NSF grants DEB-1256792 (LTREB) to DLV, DGE-1746060

(NSF-GRFP) to JPS, DEB-1702050 (NSF-DDIG) to JPS and a University of Arizona Research

Grant to JPS. We thank Candle Pfefferle, Gabriel Gudenkauf, Aubrey Reynolds, Maddie Seifert,

Kayla Cuestas, Bethany Farrah and Jack Buckman for their assistance with data collection.

Without their tremendous efforts this project would not have been feasible.

REFERENCES

Adondakis, S., and D. L. Venable. 2004. Dormancy and germination in a guild of Sonoran

Desert annuals. Ecology 85:2582–2590.

Arroyo, M. T. K., P. Chacón, and L. A. Cavieres. 2006. Relationship between seed bank

expression, adult longevity and aridity in species of Chaetanthera (Asteraceae) in central

Chile. Annals of Botany 98:591–600.

Baskin, J. M., J. J. Lu, C. C. Baskin, D. Y. Tan, and L. Wang. 2014. Diaspore dispersal ability

and degree of dormancy in heteromorphic species of cold deserts of northwest China: a

review. Perspectives in Plant Ecology, Evolution and Systematics 16:93–99.

Bell, G., and S. Collins. 2008. Adaptation, extinction and global change. Evolutionary

Applications 1:3–16.

Berg, N., and A. Hall. 2015. Increased interannual precipitation extremes over California under

climate change. Journal of Climate 28:6324–6334.

Callahan, H. S., and D. M. Waller. 2000. Phenotypic integration and the plasticity of integration

in an amphicarpic annual. International Journal of Plant Sciences 161:89–98.

De Casas, R. R., K. Donohue, D. L. Venable, and P.-O. Cheptou. 2015. Gene-flow through space

and time: dispersal, dormancy and adaptation to changing environments. Evolutionary

Ecology 29:813–831.

Cheplick, G. P. 1987. The ecology of amphicarpic plants. Trends in Ecology & Evolution 2:97–

101.

Cheplick, G. P. 1994. Life history evolution in amphicarpic plants. Plant Species Biology 9:119–

131.

Cheplick, G. P., and J. A. Quinn. 1982. Amphicarpum purshii and the pessimistic strategy in

amphicarpic annuals with subterranean fruit. Oecologia 52:327–332.

Cheplick, G. P., and J. A. Quinn. 1983. The shift in aerial subterranean fruit ratio in

Amphicarpum purshii: causes and significance. Oecologia 57:374–379.

Clauss, M. J., and D. L. Venable. 2000. Seed germination in desert annuals: An empirical test of

adaptive bet hedging. American Naturalist 155:168–186.

Cohen, D. 1966. Optimizing reproduction in a randomly varying environment. Journal of

Theoretical Biology 12:119–129.

Cruz-Mazo, G., M. L. Buide, R. Samuel, and E. Narbona. 2009. Molecular phylogeny of

Scorzoneroides (Asteraceae): evolution of heterocarpy and annual habit in unpredictable

environments. Molecular Phylogenetics and Evolution 53:835–847.

Davidowitz, G. 2002. Does precipitation variability increase from mesic to xeric biomes? Global

Ecology and Biogeography 11:143–154.

Diggle, P. J., J. A. Tawn, and R. A. Moyeed. 1998. Model‐based geostatistics. Journal of the

Royal Statistical Society: Series C (Applied Statistics) 47:299–350.

Donaldson-Matasci, M. C., C. T. Bergstrom, and M. Lachmann. 2013. When unreliable cues are

good enough. The American Naturalist 182:313–327.

Donohue, K. 2002. Germination timing influences natural selection on life‐history characters in

Arabidopsis thaliana. Ecology 83:1006–1016.

Donohue, K., L. Dorn, C. Griffith, E. Kim, A. Aguilera, C. R. Polisetty, and J. Schmitt. 2005.

The evolutionary ecology of seed germination of Arabidopsis thaliana: variable natural

selection on germination timing. Evolution 59:758–770.

Elfaki, E. A., A. M. Adeeb, S. Saidi, and A. Ickowicz. 2011. Application of remote sensing and

100

GIS in assessment of rain use efficiency and rangeland degradation in central Butana,

Sudan. International Journal of Water Resources and Arid Environments:261–270.

Fox, J., and S. Weisberg. 2018. An R companion to applied regression. Sage publications.

Gremer, J. R., S. Kimball, and D. L. Venable. 2016. Within- and among-year germination in

Sonoran Desert winter annuals: bet hedging and predictive germination in a variable

environment. Ecology Letters:1209–1218.

Gremer, J. R., and D. L. Venable. 2014. Bet hedging in desert winter annual plants: optimal

germination strategies in a variable environment. Ecology Letters 17:380–387.

Huang, Z., S. Liu, K. J. Bradford, T. E. Huxman, and D. L. Venable. 2015. The contribution of

germination functional traits to population dynamics of a desert plant community.

Ecology:150821143437005.

Imbert, E. 2002. Ecological consequences and ontogeny of seed heteromorphism. Perspectives in

Plant Ecology Evolution and Systematics 5:13–36.

Johnston, I. M. 1939. Studies in the Boraginaceae, XIII new or otherwise noteworthy species,

chiefly from Western United States. Journal of the Arnold Arboretum 20:375–402.

Koller, D. O. V, and N. Roth. 1963. Studies on the ecological and physiological significance of

amphicarpy in Gymnarrhena micrantha (Composltae). Amer Jour Bot 51:26–35.

LeHouérou, H. N. 1996. Climate change, drought and desertification. Journal of Arid

Environments 34:133–185.

Lenser, T., K. Graeber, Ö. S. Cevik, N. Adigüzel, A. A. Dönmez, C. Grosche, M. Kettermann, S.

Mayland-Quellhorst, Z. Mérai, and S. Mohammadin. 2016. Developmental control and

plasticity of fruit and seed dimorphism in Aethionema arabicum. Plant Physiology

172:1691–1707.

101

Lerner, P. D., Y. Bai, and E. F. A. Morici. 2008. Does seed heteromorphism have different roles

in the fitness of species with contrasting life history strategies? Botany 86:1404–1415.

Levin, S. A., D. Cohen, and A. Hastings. 1984. Dispersal strategies in patchy environments.

Theoretical Population Biology 26:165–191.

Lu, J.J., Tan, D.Y., Baskin, C.C. & Baskin, J.M. (2016). Effects of germination season on life

history traits and on transgenerational plasticity in seed dormancy in a cold desert annual.

Scientific Reports 6: 25076.

Mandak, B. 1997. Seed heteromorphism and the life cycle of plants: a literature review. Preslia

(Prague) 69:129–159.

Martorell, C., and M. Martinez-Lopez. 2014. Informed dispersal in plants: Heterosperma

pinnatum (Asteraceae) adjusts its dispersal mode to escape from competition and water

stress. Oikos 123:225–231.

McNamara, J., and J. A. Quinn. 1977. Resource-allocation and reproduction in populations of

Amphicarpum purshii (Gramineae). American Journal of Botany 64:17–23.

Mulroy, T. W., and P. W. Rundel. 1977. Annual Plants: Adaptations to Desert Environments.

BioScience 27:109–114.

NOAA. 2018. National Oceanographic and Atmospheric Administration.

http://www.ncdc.noaa.gov/cdo-web/datatools/findstation.

Pinheiro, J., D. Bates, S. DebRoy, and D. Sarkar. 2019. Linear and nonlinear mixed effects

models. R package version 3:57.

R Core Team. 2017. R: a language and environment for statistical computing. R Foundation for

Statistical Computing.

Sadeh, A., H. Guterman, M. Gersani, and O. Ovadia. 2009. Plastic bet-hedging in an

102

amphicarpic annual: an integrated strategy under variable conditions. Evolutionary Ecology

23:373–388.

Seger, J., and H. J. Brockmann. 1987. What is bet-hedging? Oxford Surveys in Evolutionary

Biology 4:182–211.

SEINet. 2017. Southwestern Environmental Information Network. Available at:

http://swbiodiversity.org/seinet/index.php. Last accessed 08-10-2017.

Siewert, W., and K. Tielbörger. 2010. Dispersal-dormancy relationships in annual plants: putting

model predictions to the test. The American Naturalist 176:490–500.

Simons, A. M. 2011. Modes of response to environmental change and the elusive empirical

evidence for bet hedging. Proceedings of the Royal Society B-Biological Sciences

278:1601–1609.

Simons, A. M. 2014. Playing smart vs. playing safe: the joint expression of phenotypic plasticity

and potential bet hedging across and within thermal environments. Journal of Evolutionary

Biology 27:1047–1056.

Slatkin, M. 1974. Hedging one’s evolutionary bets. Nature 250:704–705.

Starrfelt, J., and H. Kokko. 2012. Bet-hedging - a triple trade-off between means, variances and

correlations. Biological Reviews 87:742–755.

Thornthwaite, C. W. 1948. An approach toward a rational classification of climate. Geographical

review 38:55–94.

Tielbörger, K., M. Petruů, and C. Lampei. 2012. Bet-hedging germination in annual plants: a

sound empirical test of the theoretical foundations. Oikos 121:1860–1868.

UNEP. 1992. World atlas of desertification. Edward Arnold, London.

Venable, D. L. 1985. The evolutionary ecology of seed heteromorphism. American Naturalist

103

126:577–595.

Venable, D. L. 2007. Bet hedging in a guild of desert annuals. Ecology 88:1086–1090.

Venable, D. L., A. Burquez, G. Corral, E. Morales, and F. Espinosa. 1987. The ecology of seed

heteromorphism in Heterosperma pinnatum in Central Mexico. Ecology 68:65–76.

Venable, D. L., E. Dyreson, and E. Morlaes. 1995. Population dynamic consequences and

evolution of seed traits of Heterosperma pinnatum (Asteraceae). American Journal of

Botany 82:410–420.

Venable, D. L., E. Dyreson, D. Piñero, and J. X. Becerra. 1998. Seed morphometrics and

adaptive geographic differentiation. Evolution 52:344–354.

Venable, D. L., and L. Lawlor. 1980. Delayed germination and dispersal in desert annuals:

escape in space and time. Oecologia 46:272–282.

Venable, D. L., and D. A. Levin. 1985a. Ecology of achene dimorphism in Heterotheca latifolia:

I. Achene structure, germination and dispersal. Journal of Ecology 73:133–145.

Venable, D. L., and D. A. Levin. 1985b. Ecology of achene dimorphism in Heterotheca latifolia:

II. Demographic variation within populations. Journal of Ecology 73:743–755.

Veno, B. A. 1979. A revision of the genus Pectocarya (Boraginaceae) including reduction to

synonymy of the genus Harpagonella (Boraginaceae). University of California, Los

Angeles.

Wang, L., J. M. Baskin, C. C. Baskin, J. H. C. Cornelissen, M. Dong, and Z. Huang. 2012. Seed

dimorphism, nutrients and salinity differentially affect seed traits of the desert halophyte

Suaeda aralocaspica via multiple maternal effects. BMC plant biology 12:170.

Wang, L., M. Dong, and Z.-Y. Huang. 2010. Review of research on seed heteromorphism and its

ecological significance. Chinese Journal of Plant Ecology 34:578–590.

104

Weinig, C. 2000. Differing selection in alternative competitive environments: shade‐avoidance

responses and germination timing. Evolution 54:124–136.

Weiss, P. W. 1980. Germination, reproduction and interference in the amphicarpic annual Emex

spinosa (L) Campd. Oecologia 45:244–251.

West, B. T., K. B. Welch, and A. T. Galecki. 2014. Linear mixed models: a practical guide using

statistical software. CRC Press.

Yoon, J.-H., S.-Y. S. Wang, R. R. Gillies, B. Kravitz, L. Hipps, and P. J. Rasch. 2015. Increasing

water cycle extremes in California and in relation to ENSO cycle under global warming.

Nature Communications 6:8657.

Zhang, K., J. M. Baskin, C. C. Baskin, X. Yang, and Z. Huang. 2015. Lack of divergence in seed

ecology of two Amphicarpaea (Fabaceae) species disjunct between eastern Asia and eastern

North America. American journal of botany 102:860–869.

105

MAIN TEXT TABLES AND FIGURES

Figure 1. Aerial (chasmogamous) fruit of Pectocarya heterocarpa. Each fruit has four seeds

(nutlets). Of these two have margins and are identical (winged seeds) while the other two have severely reduced margins and are elongated (long seeds). Image courtesy of Michael Simpson

(2017).

106

Figure 2. Electron scanning microscope images of the three main seed (nutlet) types and aerial fruit of Pectocarya heterocarpa. A. Aerial long seed. B. Aerial winged seed. C. Basal seed. D.

Aerial fruit. Image courtesy of Larry Venable.

107

Figure 3. Experimental flow chart. Plants were collected from 6 populations, weighed and their

seeds harvested. Seeds were divided into winged, long and basal types and over-summered in the field. In the ensuing winter they were sown on agar plates in the lab and placed in growth chambers. Germination times and fractions were recorded for each seed type. Seedlings were transplanted to the green house and reared in a common-garden environment. Upon maturation plants were collected and the whole process was repeated for a second generation.

108

Figure 4. Average 29-year (1986-2016, excluding the 2014-2015 winter aridity) winter (Oct-

Mar) aridity patterns across eleven sites compared to 2014-2015 winter aridity. Portal was not included in this graph due to its outlying aridity values. A complete graph with Portal is included in the supplemental materials. The dashed line indicates the 1:1 reference line. Lower aridity values mean drier (more arid) conditions.

109

Figure 5. Basal to aerial seed ratio of P. heterocarpa plants versus average winter season (Oct-

Mar) aridity. A) Results for the field collected plants with the outlying collection site (Portal)

removed (R2 = 0.18, p << 0.001, slope = -0.16). B) Results for the first filial (F1) generation

raised in the greenhouse (R2 = 0.14, p << 0.001, slope = 0.62). C) Results for the second filial

(F2) generation raised in the greenhouse (R2 = 0.18, p << 0.001, slope = 0.57). Field plants

showed a positive relationship with the basal seeds increasing relative to aerial seeds with

increasing unpredictable environmental variability. Greenhouse raised plants showed the opposite pattern. Seed ratios were natural log transformed.

110

Figure 6. Basal and Long seed counts versus average winter season aridity (1986-2016) for the

six collection sites selected for germination and common-garden experiments. A-C. Long seeds

set by field collected plants (A, R2 = 0.29, p << 0.001, slope = 0.26), the F1 common-garden generation (B, R2 = 0.09, p << 0.001, slope = -0.56) and the F2 common-garden generation (C,

R2 = 0.06, p << 0.001, slope = -0.32). D-F. Basal seeds set by field collected plants (D, R2 =

0.16, p << 0.001, slope = 0.11), the F1 common-garden generation (E, R2 < 0.001, p = 0.74,

slope = 0.02) and the F2 common-garden generation (F, R2 = 0.03, p = 0.09, slope = 0.24). All

graphs share the same scale.

111

Figure 7. Seed allometry in the field and in the greenhouse represented as the number of basal seeds set versus the number of aerial seeds set. A. Relationship for all 12 field collection sites. B.

Relationship for F1 generation of 6 populations grown in the greenhouse. C. Relationship for F2 generation of 6 populations grown in the greenhouse. Lines are grouped by originating population and shaded by originating aridity value with dark blue representing the most mesic site and red the driest site. Graphs B and C share the same color for each population while populations in graph A have different colors due to the increased sites. Seeds set were transformed using the natural logarithm. A positive slope is almost universal for all populations, indicating that bigger P. heterocarpa plants tend to have more aerial seeds as compared to basal seeds.

112

Figure 8. Germination time across six of the twelve sites from which P. heterocarpa was collected. Top Row. Pattern graphed as grouped boxplots with categorical x axes. For all sites, basal seeds (yellow) had the slowest average germination time followed by long seeds (blue) and finally winged seeds (red). For all sites, across all treatments, except for Tacna (0.49) in the cold treatment, germination times between basal, long and winged seeds were significantly different

(ANOVA, all sites except Tacna with p < 0.01). Bottom Row. Pattern graphed as grouped scatter plots with continuous x axes. Long and basal regression line slopes (blue and gold, respectively) were positive across all treatments indicating that their germination times were fastest for plants from the most variable sites (lowest aridity values) and decreased with decreasing variability (highest aridity values). In contrast, winged seed regression line slopes

113 were negative across all treatments. The interaction between seed type and aridity was significant across all seeds (p << 0.001).

114

Figure 9. Germination fractions across six of the twelve sites from which P. heterocarpa was collected. Across all sites and treatments (hot, warm and cold), germination fractions decreased with increasing environmental unpredictability (p << 0.0001). Basal seeds tended to have the lowest germination fractions across all sites and treatments followed by long seeds and lastly winged seeds, but this was not significant. There were no significant interactions between aridity, germination chamber treatment or seed type.

115

Table 1. Weather metrics for the 12 plant collection sites across southern Arizona and

California. Average elevation, precipitation (mm), temperature (C) and aridity for a sample of 30 winter seasons (October through March) between 1986 and 2016. Locations are listed in order of increasing aridity index. Lower values on the aridity index indicate drier (more arid) sites. WS =

Weather station.

116

Table 2. Model statistics for the allometry analysis for field, F1 and F2 plants. Each row represents one analysis conducted. The

“Model Number” column serves as a reference row for the results presented in the main text. The data column indicates the dataset

(field collected plants, F1 generation, or F2 generation) used for that row, the “Model” column specifies the full model applied in that row and the “Model Component Evaluated” specifies which part of the model we were interested in, using the base notation y ~ x1 * x2. The “Reason” column indicates that we were either interested in how the slope (allometric relationship) of basal to aerial seeds differs across populations and with aridity or the intercept of this slope and how much variation in intercepts can be explained by originating population, aridity and plant weight. The final columns on the right report the R2, P-value and Slope. For models with multiple explanatory variables slopes are not reported. For models in which we were only interested in the interaction of two predictor variables, we do not report R2 values. R2 values with an asterisk denote partial R2 for the model component evaluated in that line.

117

118

Table 3. Results of the linear mixed models used to evaluate the effects of seed type, originating collection site aridity and

temperature treatment on germination time (natural log transformed) and fraction (arcsin square root transformed) of P. heterocarpa in

the growth chamber. Petri dish was used a random effect for all models. The effect column shows the overall effect of a particular

predictor on the associated germination parameter. The significant interactions column lists the significant interactions from each model. The model R2 column lists the marginal R2 which represents the variance explained by the fixed effects and the condiditonal

R2 which represents the variance explained by the entire model.

119

Supplemental Figures

Figure S1. Average 29-year (1986-2016, excluding the 2014-2015 winter aridity) winter (Oct-

Mar) aridity patterns across eleven sites compared to 2014-2015 winter aridity. The dashed line

indicates the 1:1 reference line. Lower aridity values mean drier (more arid) conditions. Due to

the outlying Portal site, the regression line (sloid black line) is significantly different from the

reference line (p << 0.001) in this graph.

120

Figure S2. Basal to aerial seed ratio of field collected P. heterocarpa plants versus average winter season (Oct-Mar) aridity (R2 = 0.22, p << 0.001) with all 12 collection sites included.

Field plants showed a positive relationship with the basal seeds increasing relative to aerial seeds with increasing unpredictable environmental variability. Seed ratios were natural log transformed.

121

A B C

Figure S3. Average fitness metrics of 2016 greenhouse plants from different originating populations (x-axis) and seed types (colors). A. Average plant weight (mg). B. Average aerial seeds set. C. Average basal seeds set. All y-axes variables were natural log transformed. Error bars represent one standard deviation above and below the mean. Basal seeds are shaded in gold, long seeds in blue and winged seeds in red. No significant patterns were detected.

122

A B C

Figure S4. Average fitness metrics of 2017 greenhouse plants from different originating populations (x-axis) and seed types (colors). A. Average plant weight (mg). B. Average aerial seeds set. C. Average basal seeds set. All y-axes variables were natural log transformed. Error bars represent one standard deviation above and below the mean. Basal seeds are shaded in gold, long seeds in blue and winged seeds in red. No significant patterns were detected.

123

APPENDIX C

Title: Fitness consequences of heteromorphic seed types in a variable environment

Running title: Fitness consequences of heteromorphic seeds

Authors and affiliations: Joshua Scholl1* and D. Lawrence Venable1

1University of Arizona, Department of Ecology and Evolutionary Biology, Tucson, AZ 85721

*primary contact: Joshua Scholl, [email protected]

Other author email addresses: Larry Venable ([email protected])

Keywords: seed heteromorphism, offspring polymorphism, bet hedging, reproductive biology, life history evolution, Pectocarya heterocarpa

124

ABSTRACT

Bet hedging is hypothesized to be one of the main strategy’s species could employ to cope with

environmental change. While proposed bet-hedging traits span the Tree of Life, strong empirical assessments of candidate bet-hedging traits are very rare. Here we rigorously evaluate the long- assumed hypothesis that seed-heteromorphic species are hedging their bets. We do so by conducting both natural-field and greenhouse experiments on an annual plant species across multiple water treatments and germination cohorts within and across years to determine differential fitness due to seed morph. We found that seed types never exhibited differences in average fitness across experiments but instead significantly interact with both the timing of germination and watering treatment. Thus, different seed types do better in different combinations of environmental conditions as required for a bet-hedging explanation.

Importantly, we find this result in both our controlled greenhouse study and our natural field experiment. We discuss these results in the context of the adaptive significance of seed heteromorphisms and fitness in variable environments.

125

INTRODUCTION

Besides extinction, organisms have three ways of coping with environmental change: phenotypic

plasticity, adaptive tracking, and bet hedging (Simons 2011). From the plasticity showcased by

environmental sex determination in reptiles (Janzen & Paukstis 1991) to the adaptive tracking observed in beak sizes of Darwin’s finches in response to changes in food availability (Grant &

Grant 2005), the former two are relatively well understood and abundant empirical examples exist throughout the tree of life (Simons 2011). In contrast, bet hedging is relatively poorly understood and few robust empirical examples exist (Gremer & Venable 2014). Yet, many

species appear to be employing this strategy, from bacteria to birds, and it seems to be

particularly important in the face of unpredictable environmental change (Cohen 1966; Slatkin

1974; Venable 1985b).

Akin to “not putting all of your eggs in one basket,” bet hedging is a counterintuitive strategy

wherein organisms spread risk over time by increasing long-term fitness at the expense of immediate fitness (Slatkin 1974; Seger & Brockmann 1987; Starrfelt & Kokko 2012). This allows organisms to buffer against environmental variation through time to increase fitness and reduce extinction risk. This is accomplished by reducing variance in fitness by following a range of strategies that are each the most successful at different times (e.g. Gremer & Venable 2014).

Bet hedging appears to be particularly common in nature. For example, many annual plants produce seeds with multiple germination years termed “delayed germination” which serve as classic examples of bet hedging (Philippi 1993; Adondakis & Venable 2004; Simons & Johnston

2006; Venable 2007; Tielbörger et al. 2012; Gremer & Venable 2014). Density-independent and

126 dependent models predict decreased germination fractions in response to a decreased probability of a good year (Cohen 1966) and increased variance between years (Ellner 1985a, b, 1987;

Gremer & Venable 2014), respectively. Importantly, selection acts on the variance in seed germination in addition to the mean (Simons 2009).

A particularly interesting candidate for bet hedging is seed heteromorphism. In seed- heteromorphic species each individual produces multiple morphologically different seed types, simultaneously (Mandak 1997; Imbert 2002; Wang et al. 2010). Seed-heteromorphic species are hypothesized to hedge their bets by producing multiple seed types with variable but decoupled success such that the variance in the parent plant’s fitness across years is reduced (Venable

1985b). One way this can be accomplished is for a plant to produce one seed morph that has the highest fitness in wet years and one with the highest fitness in dry years. In this way, one seed type will always be mismatched with the environment but overall, through the fitness returns of both seed types, variance in fitness is reduced and long-term fitness is thereby increased

(Venable 1985b).

While theoretical justification of bet hedging as the underlying mechanism of seed heteromorphism is strong, empirical evidence is weak (Childs et al. 2010; Simons 2011;

Tielbörger et al. 2012). This is due to the difficulty associated with quantifying environmental unpredictability and gathering long-term data on temporal fitness variance resulting from a candidate bet-hedging trait (Simons & Johnston 2003; Simons 2011; Gremer & Venable 2014).

Consequently, many studies have investigated germination, dispersal, growth and reproduction

(Cheplick & Quinn 1982; Imbert et al. 1997; Yang et al. 2015; Zhang et al. 2016; Ma et al.

127

2018) of different seed morphs and shown differences between them. While this is an important step in the process of evaluating seed heteromorphism as a bet hedging strategy, it in and of itself results only in very weak evidence (Simons 2011). Furthermore, these studies have largely been carried out in controlled growth chamber or greenhouse environments (but see Venable 1985a).

A rigorous assessment of bet hedging in seed-heteromorphic species requires examining the fitness consequences of different seed morphs as they relate to germination timing, growth and reproduction across various environmental conditions and years. Furthermore, a rigorous test should demonstrate that seeds differ in their ecological behavior, that this behavior translates to fitness differences in the field and that these differences fluctuate across environmental conditions such that different seeds do best in different environmental conditions. Ideally such a study would be conducted in the field to maximize realism. Furthermore, it should be carried out over multiple years to account for the fitness contribution of seeds that remain dormant during the first year or even first few years.

Here we assess empirical evidence for these assumptions required to link seed heteromorphism to bet hedging. Specifically, we employed a multi-year field experiment to quantify the variable fitness consequences of the three different seed types of a seed-heteromorphic species,

Pectocarya heterocarpa (I.M. Johnston) I.M. Johnston (Boraginaceae). We did so across three years and exposed seeds to nine different environmental conditions or year types. We also repeated a similar study in a more controlled greenhouse environment to evaluate the robustness of our field results. In both environments, we quantified germination characteristics, survival, and reproduction by seed type across multiple environmental conditions.

128

We hypothesized that if P. heterocarpa is employing seed heteromorphism to hedge its bets

against environmental unpredictability then we expect to see seed morphs display varied fitness

consequences that interact with environmental conditions, years and germination cohorts

(within-season germination timing). Specifically, we addressed three questions. First, do germination patterns match those observed in the lab from previous studies (Huang et al. 2015;

Liu et al. 2020; Scholl & Venable 2020) where basal seeds had the lowest germination fractions across seasons and slowest germination timing within seasons followed by long and finally winged seeds? Second, do basal, long, and winged seeds display significantly different fitness consequences? If so, are these fitness differences consistent across environmental conditions, years, and germination cohorts or do they differentially interact with them? Finally, does variation in plant size affect ratios of seed morphs across environmental conditions, years, and germination cohorts?

METHODS

Natural History

Pectocarya heterocarpa (I.M. Johnston; Boraginaceae) is a winter, annual plant which simultaneously produces multiple morphologically distinct single-seeded nutlets. A thorough description of its nutlet morphology and positioning along a mature plant can be found in the introduction to chapter two. Hereafter we employ the more common term of “seed” in favor of the technically correct name “nutlet.” P. heterocarpa produces three morphologically distinct seed types which we refer to as long, winged, and basal seeds. Long and winged seeds are both

129 aerial seeds and grow above ground level. In contrast, basal seeds grow at ground level or are sometimes even set in the soil. Aerial seeds stem from chasmogamous, or opening, flowers, each of which produces a fruit consisting of two long seeds and two winged seeds (Veno 1979). Long seeds are longer and heavier than winged seeds which in turn have webbed bristles, presumably to aid in dispersal. Winged seeds also detach much more readily than long seeds. Basal seeds stem from cleistogamous, or non-opening, flowers. Like their aerial counterparts, they are produced in bunches of four per fruit. Unlike their aerial counterparts, all basal seeds have severely reduced ornamentation in the form of hairs and margins. In addition, they are typically heavier and much more firmly attached to the maternal plant than long seeds. Often, basal seeds remain attached to the plant until the subsequent growing season or longer.

P. heterocarpa completes its life cycle during the winter growing season (Oct – Mar), typically germinating between October and January and, depending on rainfall, flowering between

February and April (Mulroy & Rundel 1977). Plants that survive to reproduction typically perish at the end of the winter rainy season in March or early April. Successful reproduction depends on sufficient rainfall events during the winter growing season. Winter seasons occur in which no P. heterocarpa individuals reproduce. P. heterocarpa has a relatively broad distribution, growing throughout southern and western Arizona, southern and eastern California, Nevada, Utah, New

Mexico and Northern Mexico (Johnston 1939; “SEINet” 2017).

130

Seed Collection

Seeds were collected from Tumamoc Hill in Tucson, Arizona, USA at the end of the 2016-2017

winter growing season (April 30 – May 5, 2017). At the base of Tumamoc Hill, on the northwest

corner of the hill, we established a belt transect within which we used a random number table to

determine how far along and to either side of the center of the transect to collect each of 200

individuals. Plants were collected up to 10m to either side of the center of the belt transect

resulting in a collection area of 2,000m2 or a 20m by 100m belt transect. Notably, the 2016-2017

winter growing season received very little rain resulting in sparse, tiny P. heterocarpa plants

with few seeds overall and especially few aerial seeds (ca. 8/plant). Consequently, the plants we

collected were insufficient to achieve our originally intended sample sizes as a substantial

portion of their aerial seeds were not viable and had to be discarded. All plants were collected

upon death or once they turned brown to ensure full seed maturation. We carefully cut plants at

the top of their roots and placed them in plastic Ziploc containers for separation into aerial and

basal parts. The Ziploc containers captured any seeds that were shed during handling. We

separated plants into aerial and basal parts in the field because it is much easier to tell the

different basal and aerial fruits (cleistogamous and chasmogamous originating, respectively)

apart while the plant is fully intact. Aerial and basal parts were then transferred to labeled coin

envelopes. Any aerials seeds that came lose were also transferred to the aerial parts envelope.

In the lab we separated the aerial and basal seeds from their plant material. We separated aerial

seeds into long and winged seeds but did not divide basal seeds up further as they are all very

similar and lack ornamentation. All seeds were separated from their maternal plant meticulously

131

and individually. Any seeds that did not look viable (typically abnormally small, shriveled seeds, often still green and very fragile) were discarded. Once separated, each seed type was pooled with others of its kind across all collected plants.

From the three pooled seed batches we took subsamples of 20 seeds each and performed a

viability test. This involved cutting or poking through seed coats to check for fleshy embryos,

which were regarded as viable (Pake & Venable 1996; Adondakis & Venable 2004). We chose

this method over tetrazolium chloride staining due to the difficulty associated with manipulating

and perceiving the stain results in the relatively small seeds of P. heterocarpa (Pake & Venable

1996). All evaluated seeds were viable. Viability tests were conducted prior to experimentation

because our experiments were designed such that seeds would be exposed to multiple

germination events (watering) over a relatively long time (months for greenhouse and years for

field). Consequently, assessing viability afterwards would not have allowed us to confidently

discriminate between seeds that were initially unviable and seeds that became unviable during

the experiment. We also attempted to recover ungerminated seeds at the end of our field

experiment to check for viability, however this proved to be extremely difficult.

We stored the seeds in coin envelopes and over-summered them for three months in rain-proof,

plastic storage bins at the Desert laboratory on Tumamoc Hill in Tucson, Arizona. Experiencing

hot summer temperatures is important for natural after-ripening and dormancy-breaking for

Sonoran Desert winter annual species (Adondakis & Venable 2004; Huang et al. 2015). After over-summering, seeds were stored at room temperature in our lab until the initiation of the field and then greenhouse experiment.

132

Field Experiment

We conducted our field experiment at the base of the Desert Laboratory on Tumamoc Hill. We selected a flat, open area on an alluvial plain dominated by Larrea divaricata (creosote bush) to

minimize clearing of natural vegetation. A more detailed description of the site can be found in

Pake and Venable (1996). To prevent germination of non-experimentally placed seeds from the

seedbank, we removed the top 2.5 cm of soil from this area. Viable seeds of winter annual

species rarely occur below this depth (Pake & Venable 1996). Furthermore, winter annual

species often have strong light requirements (e.g. Ma et al. 2018; Wang et al. 2019), and P. heterocarpa is no exception as we observed in a pilot study. Thus, we filled in the layer of sand we had removed with a 90-grit silica sand and SunGro sunshine mix #3 soil mixture in a 2:3 ratio. This further prevented any rare P. heterocarpa seeds buried below 2.5 cm, from compromising the experiment.

On the flat surface we prepared three plots, each 150 by 180 cm, and with a 60cm gap between them. We created three separate plots as opposed to one giant plot for ease of access as we needed to be able to sow seeds and monitor and harvest plants from each plot throughout the

experiment. For each plot we built a wooden framework to support chicken wire which we

placed around and on top of the plot to discourage large granivores (e.g. rodents and birds) and

prevent accidental trampling by larger mammals (e.g. javelinas and humans). The sparse wooden framework consisted of two wooden rectangles, 150 by 180cm, and separated vertically by 40cm with wooden blocks. Thin blocks of wood (1.3cm by 1.3cm thick) were used for the construction

of the entire cage to minimize shading effects. We also wrapped 19-gauge mesh wire around the

133

base of each plot. This wire had tiny 1.3cm by 1.3cm openings to discourage the passage of very

small granivores. The mesh wire was sunk 6 cm below ground and warped outwards to discourage rodents from digging under it (Brown & Munger 1985). The mesh wire was installed to a height of 30cm around the entire plot. We did not continue the mesh wire beyond a height of

30cm or install aluminum flashing above it as in Brown and Munger (1985) to prevent rodents from climbing over it because this would have contributed to significant shading. Pilot studies of this design in the same area showed no rodent or other herbivore damage inside the plot as determined by tracks in the soil.

Within each plot we established a grid of 15 by 12 cells for a total capacity of 180 seeds. Each row and column started 20 cm from the edge of each plot to avoid potential edge-effects

(Yahdjian & Sala 2002). Within the grid, cells were separated from each other by 10cm to

minimize plant interactions. Pilot studies indicated nearly zero germination among buried P.

heterocarpa seeds, even when seeds were only buried to a depth of 0.5cm. Consequently, we

decided to sow seeds by simply dropping them onto the soil. To prevent seeds from moving out

of their designated cell, either by wind or rain, we centered clear plastic tubes, open on both

ends, with a diameter of 7.6cm and a height of 7.5cm, on each cell. Each tube was sunk into the

soil to a depth of approximately 2.5cm to ensure that it remained in place throughout the

experiment.

We sowed 60 of the previously described seeds of each P. heterocarpa seed morph in each of

our three plots by dropping them into their randomly assigned cell tubes. Thus, a total of 180

seeds of each seed type were sowed in this experiment.

134

Each cell within a plot was randomly assigned a watering treatment of either low (no water), medium (7mm per watering) or high (14 mm per watering). No natural rainfall restrictions were assigned to any cells within plots. We powered our irrigation system using a small surface pump

(8000 Series Diaphragm Pump, model 8000-143-136, ShurFlo, Cypress, California, USA). This delivered water at the cell level through Rainbird drip nozzles. On average, we applied water once, approximately every 14 days post sunset to minimize evaporative loss. We watered plots more often during early season germination rain events (Oct-Dec of 2017 and 2018) to simulate a wetter rainfall year more accurately by taking advantage of cloud cover and lower temperatures typically associated with rain events.

We checked each cell within each plot for germination five and ten days after any rainfall period as seeds typically emerge within five to seven days after rainfall events and no later than ten days

(Pake & Venable 1995). We defined a rainfall period as any consecutive or near-consecutive days of rainfall exceeding 5mm. We monitored survival rates by checking germinants just before watering bouts, so approximately every 14 days, throughout the growing season. Plants were collected upon death in late March and early April. Germinants that were not able to establish usually did not accumulate enough biomass to last, even as skeletons, throughout the growing season and were thus not collected at the end of it. All plants that reproduced, and even some that did not were harvested at the close of the winter season. For collection, plants were carefully cut at the top of their roots and placed in small plastic containers where they were separated into aerial and basal parts. The containers served to capture any aerial seeds that were shed during handling (almost exclusively winged seeds). Long and basal seeds are firmly attached to the plant and, in the case of basal seeds especially, very difficult to remove. Thus, these two seed

135 types rarely fell off during handling. Plants were separated into aerial and basal parts in the field because it is much easier to tell the different basal and aerial fruits (cleistogamous and chasmogamous originating, respectively) apart while the plant is fully intact. Separated aerial and basal parts were then transferred to labeled coin envelopes and any aerials seeds that came lose in the plastic container were added.

Coin envelopes and their contents were dried in a Fisher Isotemp Drying Oven (200 Series) at

30C for at least 48 hours and no more than 72 hours. Next, we weighed all plants to the nearest mg using an XS105 Dual Range Mettler Toledo scale before separating and counting the seeds of each plant. Almost all basal seeds and some long seeds are firmly fused to their pedicel. If these seeds detach from the plant a small portion of the pedicel usually remains attached to the seed until germination (and in case of the long seeds, dispersal). Consequently, when separating basal and long seeds from plants we took care to keep these short pedicel portions fused with the seeds to avoid altering any effects they may have on natural germination. Aerial seeds were divided up into long and winged seeds prior to counting while basal seeds were counted in total.

P. heterocarpa produces aerial fruits of four seeds, two of which are long and two of which are winged, however we usually had significantly lower counts of winged seeds. We assumed that this was due largely to the ease of detachment of the winged seeds which we have anecdotally observed detaching even before the whole plant dries out. For this reason, we assumed that matured winged seeds were equal to matured long seeds and thus doubled our count of long seeds to attain our number of aerial seeds.

136

Greenhouse Experiment

We complemented our field experiment with a similar greenhouse experiment. First, seed germination fractions of the field collected P. heterocarpa plants were assessed in a growth chamber. These seeds came from the same batch as those used for the field experiment and were germinated using agar-filled petri dishes and randomized sowing.

In total we used 1,050 seeds of which 480 were basal, 330 were long and 240 were winged. We would have preferred to have 480 seeds of all seed types, but this was not possible due to our lack of aerial seeds, as described in the plant collection section above. We carried out three germination bouts each with 10 petri dishes and 35 seeds per dish such that each germination bout involved 160 basal seeds, 110 long seeds and 80 winged seeds. All germination bouts were carried out in 12C/22C night/day settings with 12 hours of light during the day. The first germination bout was initiated on November 6, 2017, the second on Dec 6, 2017 and the third on

January 6, 2018.

Once they achieved a radicle length of 8-15 mm and a hypocotyl length of 5-12 mm, germinants were carefully transplanted from their petri dishes to individual Cone-Tainer planting pots. These were then randomly placed in racks of 20 conetainers, where each pot had a diameter of 6.3cm and a depth of 25cm, and immediately placed in the greenhouse. For all 3 germination bouts, transplants occurred between 10-16 days after germination. We attempted to transfer all germinants to the greenhouse to maximize our sample sizes across seed types. Each conetainer was filled with the same sand and soil mixture in a 2:3 ratio as used in the field experiment, respectively. Greenhouse temperatures were set to mimic outdoor growing season conditions for

137

Nov-Mar (Nov= 23C/9C, Dec= 18C,6C Jan=18C/6C, Feb=20C/7C, Mar=23C/9C; NOAA) and the greenhouse was free of whitewash and shade screens. After transplanting, we watered plants often and liberally during the first two weeks to maximize establishment. We randomly repositioned conetainer racks and the plants within the conetainer racks weekly.

After allowing the plants two weeks to become established we initiated our watering treatments.

While it was our goal to have approximately 60 plants per seed type, per watering treatment, we were only able to accomplish this fully for basal seeds. Lack of germination among long and winged seeds resulted in relatively low sample sizes for them (Table S1). Watering treatments were assigned randomly across conetainer racks but evenly across seed types. We administered a low, medium, and high-watering treatment. The low treatment consisted of 3 seconds of hose water, the medium of 6 seconds and the high of 9 seconds. Treatments were administered once every four days using a new pesticide sprayer backpack which had never been used to hold chemicals. This frequency resulted in visible differences between the three treatments suggesting its effectiveness and our continued use of it throughout the experiment. Upon completion of the experiment plants were harvested using the same procedure applied to field collections and described above.

In the lab we dried the collected greenhouse plants inside their envelopes in a Fisher Isotemp

Drying Oven (200 Series) at 30C for at least 48 hours and no more than 72 hours. Plants were subsequently weighed to the nearest mg using an XS105 Dual Range Mettler Toledo scale. Next, we separated and counted the seeds on all the collected plants. For the aerial parts we counted both the adaxial or long seeds and the abaxial or winged seeds. For the basal parts we focused on

138

the total number of seeds. All counted seeds were stored in separate coin envelopes for each seed

type and maternal plant. Although P. heterocarpa produces aerial fruits of four seeds, two of

which are long and two of which are winged, we generally had significantly lower counts of

winged seeds. We attribute this to the ease of detachment of the winged seeds which we have

anecdotally observed detaching even before the whole plant dries out. For this reason, we

assumed that matured winged seeds were equal to matured long seeds and thus doubled our count of long seeds for our analyses of ratios of basal to aerial seeds. This is supported by our observations of whole fruit abortions in nature and the greenhouse and only rarely, individual seed abortions.

Statistical Analyses

For the field experiment we evaluated the influence of seed type, water treatment and season on germination fractions and times using generalized linear mixed models (GLMMs) with plot

st included as a random effect in all models. Germination time (days since Oct. 1 , 2017), was transformed using the natural logarithm to improve normality. For the greenhouse experiment we evaluated the influence of seed type and germination bout on germination fraction including petri dish as a random effect. Germination time was not recorded for the greenhouse experiment.

Models for germination fraction were fit using binomial error distributions and logit link

functions to test for the effects of factors on the probability of seed germination. Significance of

seed type indicates differences in germination times or fractions across the different seed types.

139

For the field and greenhouse experiment we evaluated the influence of seed type, germination

cohort, water treatment (and season for the field experiment) on various plant fitness metrics,

including plant weight, total seeds set, aerial seeds set, basal seeds set and basal to aerial seed

ratios, using generalized linear mixed models (GLMMs). Plot or conetainer rack was included as

a random effect in all models for field and greenhouse experiments, respectively. Fitness metrics

were transformed using the natural logarithm. Significance of seed type indicates differences in

fitness outcomes across the different seed types.

Various models were constructed using the top-down technique (Diggle et al. 1998; West et al.

2014). The best model for each combination of response variables that included seed type was

selected using the Akaike Information Criterion (AIC). All statistical tests were conducted in the

R statistical software (R Core Team 2017). The R package nlme (Pinheiro et al. 2019) was used to specify GLMMs. The significance of fixed effects was evaluated with Type III tests using the car R package (Fox & Weisberg 2018). Effect size or partial R2 values were calculated for mixed

models using the MuMin R package (Barton & Barton 2019) which applies the method described

by Nakagawa et al. (Nakagawa & Schielzeth 2013).

RESULTS

Natural Weather Patterns

Rainfall during the 2017-2018 winter growing season was 1.9 standard deviations below the 30-

year average for Tumamoc Hill. In contrast, rainfall during the 2018-2019 season was 2.6

standard deviations above the 30-year average rainfall. Rainfall during the 2019-2020 season

140

was 1.26 standard deviations above the 30-year average rainfall but no germination was observed

during this season.

Germination Characteristics

In our field experiment 199 (37%) seeds germinated of which 119 were harvested and of which

115 (58%) reproduced (Table 1A). Non-harvested plants were plants that germinated but failed to establish and reproduce and withered away prior to our late March and early April collection bouts. Germination was highest during the first year with 32% of seed germinating and 53% of them surviving to reproduction. In the second year only 5% of the remaining seeds germinated, however, 89% of them survived to reproduction.

During the 2017-2018 winter season of the field experiment germination fractions were very similar across watering treatments and seed types such that no significant patterns were observed between seed types (Table 1). In the 2017-2018 season, the non-significant trend was for basal seeds to have the highest germination fraction in the low- and high-water treatments, and long seeds in the medium-water treatment (Fig. 1A). During the 2018-2019 winter season of the field experiment, germination fractions showed more variation across the different water treatments.

The trend was for germination fractions to be highest in the medium-water treatment followed by the low-water treatment (Fig. 1B). Only basal seeds germinated in the high-water treatment.

Basal, long, and winged seeds had very similar times to germination and were not statistically significant though watering treatment and year were (Fig. 1C and D). The best model explained

141

41% of the variation in germination time across the field experiment (Table 3). In 2017-2018

germination time across all seeds decreased with increasing watering treatments (Fig. 1C). In

2018-2019 the same but weaker pattern was observed with germination time decreasing similarly

for all seeds with increasing watering treatments (Fig. 1D). Germination times were faster during

the wetter 2018-2019 field season than the 2017-18 season.

In our greenhouse experiment 362 (48%) seeds germinated (Table 2A). The germination

fractions were similar to those observed in the field. Across all three cohorts, germination

fractions were highest for basal seeds (0.75) followed by winged seeds (0.27) and finally long

seeds (0.23; Table 2A; main effect of seed type, p << 0.001, R2 = 0.36; Table 4). Germination

time was not monitored for seeds germinated in the growth chamber for the greenhouse

experiment. Of the germinants, 342 were successfully transplanted to the greenhouse and of

these 337 (99%) reproduced (Table 2B). Nearly all successful transplants survived to

reproduction.

Fitness Consequences of Different Seed Types in the Field

The best models for the number of basal seeds set in the field and aerial seeds set in the field

included all predictors (water received, germination cohort, seed type and season) with plot as a

random effect (Table 5). Water received throughout the germination and growth period and

germination cohort were the best predictors across all models (Table 5). As expected, increasing

water led to plants that produced more seeds of all types (Fig. 3). In addition, seeds germinating

in earlier germination cohorts produced plants with more seeds of all types (Fig. 3). Also, as

142

natural rainfall was much higher during the 2018-2019 winter growing season, plants set more seeds of all types during this season (Fig. 5B) as compared to 2017-2018 (Fig. 5A). Overall, the fixed effects of these three models explained an average of 70% of the variance while the combined fixed and random effects explained an average of 71.5% of the variance (Table 5).

While the originating seed type was never significant as a main effect regarding the basal and aerial seed set, it interacted significantly with cohort, watering treatment and season, indicating that different seed types had higher fitness in different combinations of conditions (all p < 0.05,

Table 5). The effect size or partial R2 of seed type as an interacting term was 0.04 (p < 0 .05) for

aerial seeds set and 0.02 (p < 0.05) for basal seeds set (Table 5).

The best model for the basal to aerial seed ratios observed in the field experiment included seed

type, water received and season as predictors with plot as a random effect (Table 5, Fig. 3).

Except for seed type, all predictor variables were significant and showed the expected pattern

with increasing watering treatment and the 2018-2019 winter growing season resulting in plants with lower basal to aerial ratios. While seed type was not a significant predictor alone it interacted significantly with water and season (3-way interaction; p = 0.03, partial R2 = 0.02;

Table 5). For this model, the fixed effects explained 53% of the variance while the combined

fixed and random effects explained 55% of the variance.

Fitness Consequences of Different Seed Types in the Greenhouse

For the greenhouse plants the best models for basal and aerial seeds set included seed type, water

received and germination cohort as predictors with conetainer rack as a random effect (Table 6).

143

Water received throughout the germination and growth period was the best predictor across all

models (Fig. 4). Increasing water led to bigger plants that produced more seeds of all types. With

respect to aerial seeds set, seed type was a significant predictor variable (p < 0.001) and

interacted significantly with watering treatment (p = 0.002, partial R2 = 0.12; Table 6). While the originating seed type was not a significant model predictor alone for basal seeds set it was a significant interaction term (p = 0.005, partial R2 = 0.04; Table 6). The effects of originating seed

type on basal and aerial seeds set were complicated. Across germinating cohorts and water

treatments the fitness component ranking of seed types flip-flopped (Table 6; Figure 6). Overall, the fixed effects of the two models regarding basal and aerial seeds set explained an average of

35% of the variance while the combined fixed and random effects explained an average of 47% of the variance (Table 6).

The best model for the basal to aerial seed ratios observed in the greenhouse experiment included seed type, water received and germination cohort as predictors with plot as a random effect

(Table 6, Fig. 4). Of these only seed type (p << 0.001, partial R2 = 0.06) and water received (p

<< 0.001) had significant main effects. The interaction between seed type and all other predictor

variables was also significant. As with the field experiment, increasing water resulted in plants

with lower basal to aerial ratios. In contrast to field experiment results, later germination resulted

in lower basal to aerial ratios. For this model, the fixed effects explained 47% of the variance

while the combined fixed and random effects explained 48% of the variance (Table 6).

Basal and aerial seed allometry in the field and greenhouse

144

Plants from the 2017-2018 and 2018-2019 field seasons displayed an allometric relationship in their relative production of basal as compared to aerial seeds (Fig. 7A and B). The slope, or

allometric constant, for log transformed basal vs log transformed aerial seeds across all watering

treatments and germination cohorts was significant and 0.66 for the 2017-2018 plants and 0.57 for the 2018-2019 plants (both p << 0.001; Table 7, Models 1 and 4). Thus, plants with greater seed production had relatively more aerial seeds as compared to basal seeds. In both seasons, this slope, or allometric constant, did not vary significantly with originating seed type (both p > 0.8;

Table 7, Models 2 and 5; Fig. 7A and B) indicating a constant allometry across all three seed types. The intercept of the allometric relationship also did not vary with seed type (both p > 0.78;

Table 7, Models 3 and 6; Fig. 7A and B), implying that plants originating from different seed types did not differ in their basal to aerial ratios for any given level of seed production.

All greenhouse collected plants displayed an allometric relationship in the relative production of basal as compared to aerial seeds (Fig. 7C). The slope, or allometric constant, for log transformed basal vs log transformed aerial seeds across all watering treatments and germination cohorts was 0.28 (Table 7, Model 7). Thus, plants with greater seed production had relatively more aerial seeds as compared to basal seeds. This slope, or allometric constant, did not vary significantly with originating seed type (p = 0.17; Table 6, Model 8; Fig. 7C) indicating a constant allometry across all three seed types. The intercept of the allometric relationship also did not vary with seed type (p = 0.81; Table 7, Model 9; Fig. 7C), implying that plants originating from different seed types did not differ in their basal to aerial ratios for any given level of seed production.

145

DISCUSSION

Claims of bet hedging span the Tree of Life, however, empirical evidence of this important life history strategy is rare (Childs et al. 2010; Simons 2011). Here we rigorously evaluate evidence for bet hedging in a relatively common and phylogenetically diverse plant reproductive strategy

(Wang et al. 2010; Scholl et al. 2020). We evaluate fitness consequences of seed types of a seed- heteromorphic species across multiple environmental conditions in a controlled greenhouse environment and in nature. Overall, we find evidence for bet hedging. While seed types never had significant differences in average fitness metrics, they exhibited complex interactions with other relevant environmental variables such as water levels and growing season (year). Thus, different seeds performed better under unique combinations of environmental conditions.

Fitness Consequences of Seed Types Depend on Complex Interactions

We found that seed type was never a significant predictor of fitness alone in P. heterocarpa.

Instead seed type interacted significantly with both the timing of germination (germination cohort) and watering treatment (simulated rainfall year) in both the greenhouse and field to predict fitness outcomes. Thus, different seeds performed better in different environmental conditions. This result was similar to findings by Venable (1985a) for Heterotheca latifolia in which fitness of ray and disc seeds interacted significantly with germination timing, density and water availability. H. latifolia displayed clear differences in germination timing and fraction with ray seeds germinating later and to a lower degree than disc seeds. This translated to differences in final plant biomass and fitness. For example, plants originating from ray seeds were typically

146 smaller than their counterparts arising from disc seeds due largely to delayed within-season germination (Venable 1985a). Yet, in drought conditions many of the earlier germinating disc seeds failed to survive to reproduction. These differences are primarily driven by differences in germination behavior. In our study of P. heterocarpa, the only significant interaction we found between germination timing and seed type was for the basal to aerial seed ratio of greenhouse plants, and the magnitude of this effect was relatively small.

Furthermore, P. heterocarpa seed types did not have simple differences consistently related to earlier or later germination or higher or lower water as seen in H. latifolia (Venable 1985a).

Instead, our species exhibited multi-order interactions such as variability in fitness among germination cohorts that depended on water treatment levels and growing season (year). Thus, the relative fitness of seed types displayed complex variation with different combinations of environmental factors. While this result is consistent with bet hedging theory (Venable 1985b) it has not been reported in other studies of seed-heteromorphic species.

Our results for P. heterocarpa were also different from previous work on other amphicarpic species. In particular, studies of amphicarpic plants typically find that basal seeds result in more competitive and bigger plants than aerial seeds (Cheplick 1994; Mandak 1997). Basal seeds of

Gymnarrhena micrantha, produced bigger plants than aerial seeds when grown in a common environment but seedlings from basal seeds grew more slowly initially as compared to aerial seedlings (Koller and Roth 1963). This provides a potential mechanism through which fitness outcomes of different seed types can interact with water availability and germination timing. For example, due to their slow growth basal seeds need to germinate earlier in a season to have time

147

to reach their full potential. In contrast, aerial seeds can germinate later as they require less time

to reach their full potential size. This difference in growth strategy also resulted in basal seeds

being more drought tolerant (Koller & Roth 1963). In another study, Cheplick and Quinn (1982)

showed that Amphicarpum purshii plants arising from basal seeds are more competitive, have

higher survival rates and achieve greater fitness than plants arising from aerial seeds across water

gradients and densities. In addition, they found that aerial originating plants never produce aerial

seeds. The results for A. purshii and G. micrantha are typical for seed heteromorphism in

amphicarpic species.

In comparison to P. heterocarpa, these amphicarpic species have much bigger size differences

between seed types. In a review of amphicarpic plants Cheplick (1994) reported basal to aerial seed weight ratios between 1.5 and 18. Our species is at the lower end of this range with basal seeds that are 2.6 times heavier than winged seeds and two times heavier than long seeds (Sarah

Felker, unpublished data). In contrast, basal seeds of the well-studied Gymnarrhena micrantha are 18 times heavier than aerial seeds (Cheplick 1994) while those of Amphicarpum purshii are 5 times heavier (Cheplick & Quinn 1982). Most studies of amphicarpic species have focused on species with dramatic size differences between basal and aerial seeds (Cheplick 1987, 1994). It is possible that the larger basal seeds of P. heterocarpa result in slightly faster growing seedlings or even slightly larger plants than the aerial seeds but that we were not able to detect this pattern given our relatively small sample sizes and relatively dramatic water treatment differences.

The observed differences caused by environmental interactions with seed type in our study appear relatively small when contrasted with the large effects of water and germination timing.

Nevertheless, these effects could compound over many generations and become large enough to

148

reduce temporal fitness variance and increase long-term fitness, which is the hallmark of a bet- hedging adaptation.

Furthermore, amphicarpic plants are known to exhibit differences in dispersal and competition

(Venable & Levin 1985; Cheplick 1994; Mandak 1997). In amphicarpic plants especially, basal

seeds typically do not disperse at all while aerial seeds disperse quite widely (Cheplick 1987).

Though not evaluated in this study, dispersal differences provide another source of variation in

fitness and further increase seed type by environment interactions in fitness. For example, early

vigor of seedlings from basal seeds may allow them to establish more successfully in

environments with high intraspecific and interspecific competition. This competitive superiority

would be masked in a non-competitive environment such as that in our experiments where the

cost of producing larger seeds would be a fitness detriment.

Germination Characteristics in the Field and Greenhouse Experiment

Germination characteristics of the different P. heterocarpa seed morphs in the field and

greenhouse study did not match previous germination results for this species. Specifically, past

work has shown that basal seeds of P. heterocarpa germinate more slowly and have lower

germination fractions than winged and long seeds (Huang et al. 2015; Liu et al. 2020; Scholl &

Venable 2020). In contrast, in this experiment basal seeds had consistently, albeit not

significantly, higher germination fractions in both the field and greenhouse as compared to the

long and winged seeds. Meanwhile, the long and winged seeds had very similar germination

149 fractions with the winged seeds having slightly higher fractions in the greenhouse and the long seeds having slightly higher fractions in the field.

While studies of this species are rare, especially studies which distinguish between seed types, the few that exist show clear and contrasting trends. Both Huang et al. (2015) and Scholl and

Venable (Scholl & Venable 2020) have found that basal seeds germinate more slowly and have lower germination fractions than either long or winged seeds. This pattern has been shown across various temperature, water and over-summering treatments as well as across different populations of P. heterocarpa in the Southwest (Huang et al. 2015; Scholl & Venable 2020). In addition, Huang et al. (2015) used filter paper to germinate P. heterocarpa seeds as well as numerous pre-germination treatments including testing naturally over-summered seeds and fresh seeds. Scholl and Venable (2020) in turn used agar as a germination medium. Nevertheless, all treatments within and across both studies resulted in the same conclusion, namely that basal seeds have the slowest time to germination and lowest germination fractions. Importantly, the germination behavior observed for basal seeds in the present study mirrors that demonstrated by

Huang et al. (2015) and Scholl (2020). The aerial seeds, however, had significantly lower germination fractions than those measured in previous work.

While we remain puzzled by these results for aerial seeds, we speculate that the rainfall environment experienced by the maternal plant may be a contributing factor. The average winter season (Oct. – Mar.) rainfall for our collection site at Tumamoc Hill is 126mm (NOAA 2018;

Scholl & Venable 2020). Both Scholl and Huang used seeds collected during relatively wet years with 137mm and 226mm of rainfall, respectively. In contrast, the winter season during which we

150 collected seeds for the present study (2016-2017) only received 83mm of rainfall at Tumamoc

Hill, AZ (NOAA 2018). This resulted in very small P. heterocarpa plants that invested sparingly in the production of aerial seeds. Since aerial seeds are set and matured after basal seeds, it stands to reason that maturation was more tenuous for aerial seeds. This could explain why basal seeds behaved relatively normally but aerial seeds exhibited much lower time to germination as well as germination fractions. The differences observed for aerial seeds could have been due to relatively large rates of immaturity among collected seeds or by plastic changes in maternal provisioning to reduce germination fractions. In the former case, thought unlikely, seeds may not have been viable beyond the summer months but were viable just long enough to pass our viability assays that we administered post-collection and prior to over-summering. Perhaps a more plausible explanation is that diminished water availability to the maternal plants resulted in increased dormancy in the aerial seeds. Wang et al. (2012) have shown that the seed heteromorphic Atriplex aucheri of arid China responds to increased water by increasing germination among its two smallest seed types. Dormancy of its largest seed type remained mostly unaffected by water treatment. This could explain why our basal seeds behaved normally while our aerial seeds showed significant decreases in germination fractions.

Lastly, it is also possible that germination behavior is strongly tied to positioning on the maternal plant (Gutterman 2012). Because plants were small and aerial seed set was very low during our collection year, the seeds we acquired came almost entirely from the lowest parts of branches.

We suspect that a combination of seed positioning on the maternal plant and a plastic maternal effect in response to low water contributed to the odd germination characteristics we observed

151 for P. heterocarpa. This highlights the important interaction between bet hedging and plasticity in seed-heteromorphic species.

Seed Allometry

In line with past work on P. heterocarpa (Scholl & Venable 2020) we found that the allometric investment in basal versus aerial seeds was not influenced by the seed type a plant germinated from but rather by environmental conditions in both the greenhouse and the field. These results are corroborated by Sadeh et al. (2009) who investigated seed ratios in Emex spinosa and found them to respond plastically to water, nutrient availability, and competition. However, in species like A. purshii and G. micrantha previous studies show that originating seed type can have predictable consequences for final basal to aerial seed ratios (Koller & Roth 1963; McNamara &

Quinn 1977; Cheplick & Quinn 1982; Cheplick 1994). Additional plasticity in the basal to aerial seed ratio has been demonstrated due to competition (Cheplick & Quinn 1983).

Seed longevity

We carried our field experiment out for three years. Based on studies by Moriuchi et al. (2000) for Pectocarya recurvata, we expected germination of P. heterocarpa seeds in the third year and possibly the fourth. Yet, we observed germination for only two years. This is in agreement with

1-2 year survival of seed types from Salsola ferganica investigated by Ma et al. (2018) but may contradict expectations of seed longevity among annual species of variable environments

(Moriuchi et al. 2000). Furthermore, Moriuchi et al. showed that only a small percentage of

152

seeds in the seed bank were more than two years old (ca. 16 percent). This small percentage

combined with our relatively small sample sizes would have made the potential germinant pool

around 28 seeds per seed type. Combined with the observed decline in germination fraction

between year one and year two and the fact that germination rainfall was relatively high during

both years of the field study, it is not surprising that we did not see any germination beyond the

second year. Nevertheless, rare germination events occurring in the third and possibly fourth

year could offer important contributions to fitness, particularly when the first two years did not

receive adequate rainfall.

Conclusions

We find strong evidence for variation in the fitness of seed types in a seed-heteromorphic species with different combinations of environmental factors that should provide a bet hedging effect.

We demonstrate that seed types of P. heterocarpa do not display average differences in fitness across environmental conditions, germination cohorts and years, but rather complex interactions with these factors. Thus, the marriage of the three seed types in P. heterocarpa likely allows the species to reduce its variance in fitness as compared to a hypothetical species producing only one of the seed types. This provides a strong empirical example of diversified bet hedging.

ACKNOWLEDGEMENTS

This research was supported by NSF grants DEB-1256792 (LTREB) to DLV, DGE-1746060

(NSF-GRFP) to JPS, DEB-1702050 (NSF-DDIG) to JPS and a University of Arizona Research

Grant to JPS. We thank Candle Pfefferle, Gabriel Gudenkauf, Aubrey Reynolds, Maddie Seifert,

153

Kayla Cuestas, Bethany Farrah and Jack Buckman for their assistance with data collection.

Without their tremendous efforts this project would not have been feasible.

154

References

Adondakis, S. & Venable, D.L. (2004). Dormancy and germination in a guild of Sonoran Desert

annuals. Ecology, 85, 2582–2590.

Barton, K. & Barton, M.K. (2019). Package ‘MuMIn.’ R Packag. version, 1.

Brown, J.H. & Munger, J.C. (1985). Experimental manipulation of a desert rodent community:

food addition and species removal. Ecology, 66, 1545–1563.

Cheplick, G.P. (1987). The ecology of amphicarpic plants. Trends Ecol. Evol., 2, 97–101.

Cheplick, G.P. (1994). Life history evolution in amphicarpic plants. Plant Species Biol., 9, 119–

131.

Cheplick, G.P. & Quinn, J.A. (1982). Amphicarpum purshii and the pessimistic strategy in

amphicarpic annuals with subterranean fruit. Oecologia, 52, 327–332.

Cheplick, G.P. & Quinn, J.A. (1983). The shift in aerial subterranean fruit ratio in Amphicarpum

purshii: causes and significance. Oecologia, 57, 374–379.

Childs, D.Z., Metcalf, C.J.E. & Rees, M. (2010). Evolutionary bet-hedging in the real world:

empirical evidence and challenges revealed by plants. Proc. R. Soc. B-Biological Sci., 277,

3055–3064.

Cohen, D. (1966). Optimizing reproduction in a randomly varying environment. J. Theor. Biol.,

12, 119–129.

Diggle, P.J., Tawn, J.A. & Moyeed, R.A. (1998). Model‐based geostatistics. J. R. Stat. Soc. Ser.

C (Applied Stat., 47, 299–350.

155

Ellner, S. (1985a). ESS germination strategies in randomly varying environments. I. Logistic-

type models. Theor. Popul. Biol., 28, 50–79.

Ellner, S. (1985b). ESS germination strategies in randomly varying environments. II. Reciprocal

Yield-Law models. Theor. Popul. Biol., 28, 80–116.

Ellner, S. (1987). Competition and Dormancy: A Reanalysis and Review. Am. Nat., 130, 798–

803.

Fox, J. & Weisberg, S. (2018). An R companion to applied regression. Sage publications.

Grant, P.R. & Grant, B.R. (2005). Darwin’s finches. Curr. Biol., 15, R614–R615.

Gremer, J.R. & Venable, D.L. (2014). Bet hedging in desert winter annual plants: optimal

germination strategies in a variable environment. Ecol. Lett., 17, 380–387.

Gutterman, Y. (2012). Survival strategies of annual desert plants. Springer Science & Business

Media.

Huang, Z., Liu, S., Bradford, K.J., Huxman, T.E. & Venable, D.L. (2015). The contribution of

germination functional traits to population dynamics of a desert plant community. Ecology,

150821143437005.

Imbert, E. (2002). Ecological consequences and ontogeny of seed heteromorphism. Perspect.

Plant Ecol. Evol. Syst., 5, 13–36.

Imbert, E., Escarre, J. & Lepart, J. (1997). Seed heteromorphism in Crepis sancta (Asteraceae):

performance of two morphs in different environments. Oikos, 79, 325–332.

Janzen, F.J. & Paukstis, G.L. (1991). Environmental sex determination in reptiles: ecology,

evolution, and experimental design. Q. Rev. Biol., 66, 149–179.

156

Johnston, I.M. (1939). Studies in the Boraginaceae, XIII new or otherwise noteworthy species,

chiefly from Western United States. J. Arnold Arbor., 20, 375–402.

Koller, D.O. V & Roth, N. (1963). Studies on the ecological and physiological significance of

amphicarpy in Gymnarrhena micrantha (Composltae). Amer Jour Bot, 51, 26–35.

Liu, S., Bradford, K.J., Huang, Z. & Venable, D.L. (2020). Hydrothermal sensitivities of seed

populations underlie fluctuations of dormancy states in an annual plant community.

Ecology, 101, e02958.

Ma, Y., Wang, J., Zhang, J., Zhang, S., Liu, Y. & Lan, H. (2018). Seed Heteromorphism and

Effects of Light and Abiotic Stress on Germination of a Typical Annual Halophyte Salsola

ferganica in Cold Desert. Front. Plant Sci.

Mandak, B. (1997). Seed heteromorphism and the life cycle of plants: a literature review. Preslia

(Prague), 69, 129–159.

McNamara, J. & Quinn, J.A. (1977). Resource-allocation and reproduction in populations of

Amphicarpum purshii (Gramineae). Am. J. Bot., 64, 17–23.

Moriuchi, K.S., Venable, D.L., Pake, C.E. & Lange, T. (2000). Direct measurement of the seed

bank age structure of a Sonoran Desert annual plant. Ecology, 81, 1133–1138.

Mulroy, T.W. & Rundel, P.W. (1977). Annual Plants: Adaptations to Desert Environments.

Bioscience, 27, 109–114.

Nakagawa, S. & Schielzeth, H. (2013). A general and simple method for obtaining R2 from

generalized linear mixed‐effects models. Methods Ecol. Evol., 4, 133–142.

NOAA. (2018). National Oceanographic and Atmospheric Administration. Available at:

157

http://www.ncdc.noaa.gov/cdo-web/datatools/findstation. Last accessed .

Pake, C.E. & Venable, D.L. (1995). Is coexistence of Sonoran desert annuals mediated by

temporal variability reproductive success. Ecology, 76, 246–261.

Pake, C.E. & Venable, D.L. (1996). Seed banks in desert annuals: Implications for persistence

and coexistence in variable environments. Ecology, 77, 1427–1435.

Philippi, T. (1993). Bet-Hedging Germination of Desert Annuals: Beyond the First Year. Am.

Nat., 142, 474–487.

Pinheiro, J., Bates, D., DebRoy, S. & Sarkar, D. (2019). Linear and nonlinear mixed effects

models. R Packag. version, 3, 57.

R Core Team. (2017). R: a language and environment for statistical computing. R Found. Stat.

Comput., R Foundation for Statistical Computing.

Sadeh, A., Guterman, H., Gersani, M. & Ovadia, O. (2009). Plastic bet-hedging in an

amphicarpic annual: an integrated strategy under variable conditions. Evol. Ecol., 23, 373–

388.

Scholl, J.P., Calle, L.M., Miller, N.M. & Venable, D.L. (2020). Offspring polymorphism and bet

hedging: a large-scale, phylogenetic analysis. Ecol. Lett. Press.

Scholl, J.P. & Venable, L.D. (2020). Bet hedging and plasticity: an integrated strategy in a

variable environment. Prep.

Seger, J. & Brockmann, H.J. (1987). What is bet-hedging? Oxford Surv. Evol. Biol., 4, 182–211.

SEINet. 2017. Southwestern Environmental Information Network. Available at:

http://swbiodiversity.org/seinet/index.php. Last accessed 13 March 2018.

158

Simons, A.M. (2009). Fluctuating natural selection accounts for the evolution of diversification

bet hedging. Proc. R. Soc. B-Biological Sci., 276, 1987–1992.

Simons, A.M. (2011). Modes of response to environmental change and the elusive empirical

evidence for bet hedging. Proc. R. Soc. B-Biological Sci., 278, 1601–1609.

Simons, A.M. & Johnston, M.O. (2003). Suboptimal timing of reproduction in Lobelia inflata

may be a conservative bet‐hedging strategy. J. Evol. Biol., 16, 233–243.

Simons, A.M. & Johnston, M.O. (2006). Environmental and genetic sources of diversification in

the timing of seed germination: implications for the evolution of bet hedging. Evolution (N.

Y.) 60, 2280.

Slatkin, M. (1974). Hedging one’s evolutionary bets. Nature, 250, 704–705.

Starrfelt, J. & Kokko, H. (2012). Bet-hedging - a triple trade-off between means, variances and

correlations. Biol. Rev., 87, 742–755.

Tielbörger, K., Petruů, M. & Lampei, C. (2012). Bet-hedging germination in annual plants: a

sound empirical test of the theoretical foundations. Oikos, 121, 1860–1868.

Venable, D.L. (1985a). Ecology of achene dimorphism in Heterotheca latifolia: III.

Consequences of varied water availability. J. Ecol., 73, 757–763.

Venable, D.L. (1985b). The evolutionary ecology of seed heteromorphism. Am. Nat., 126, 577–

595.

Venable, D.L. (2007). Bet hedging in a guild of desert annuals. Ecology, 88, 1086–1090.

Venable, D.L. & Levin, D.A. (1985). Ecology of achene dimorphism in Heterotheca latifolia: I.

Achene structure, germination and dispersal. J. Ecol., 73, 133–145.

159

Veno, B.A. (1979). A revision of the genus Pectocarya (Boraginaceae) including reduction to

synonymy of the genus Harpagonella (Boraginaceae). University of California, Los

Angeles.

Wang, L., Baskin, J.M., Baskin, C.C., Cornelissen, J.H.C., Dong, M. & Huang, Z. (2012). Seed

dimorphism, nutrients and salinity differentially affect seed traits of the desert halophyte

Suaeda aralocaspica via multiple maternal effects. BMC Plant Biol., 12, 170.

Wang, L., Dong, M. & Huang, Z.-Y. (2010). Review of research on seed heteromorphism and its

ecological significance. Chinese J. Plant Ecol., 34, 578–590.

Wang, Y., Li, X., Liu, L., Zhao, J., Song, G. & Zhou, Y. (2019). Dormancy and germination

strategies of a desert winter annual Echinops gmelini Turcz. in a temperate desert of China.

Ecol. Res., 34, 74–84.

West, B.T., Welch, K.B. & Galecki, A.T. (2014). Linear mixed models: a practical guide using

statistical software. CRC Press.

Yahdjian, L. & Sala, O. (2002). A rainout shelter design for intercepting different amounts of

rainfall. Oecologia, 133, 95–101.

Yang, F., Yang, X., Baskin, J.M., Baskin, C.C., Cao, D. & Huang, Z. (2015). Transgenerational

plasticity provides ecological diversity for a seed heteromorphic species in response to

environmental heterogeneity. Perspect. Plant Ecol. Evol. Syst., 17, 201–208.

Zhang, R., Wang, Y.R., Baskin, J.M., Baskin, C.C., Luo, K. & Hu, X.W. (2016). Germination

and persistence in soil of the dimorphic diaspores of Atriplex centralasiatica. Seed Sci. Res.,

26, 273–283.

160

Main Text Figures

Figure 1. Germination time (B) and germination fraction (C) of the different seed types in the field experiment. Mixed effects models showed no significant differences in germination time or fraction between the different seed morphs across or within water treatments or plots (mixed effects model, all p > 0.1). Water was a significant predictor of germination time across all seed types (p << 0.001) as was season (p < 0.001). Seed type explained 4% of the variation in germination fractions in 2018-2019 but was not a significant predictor (p = 0.24).

161

Figure 2. Germination fraction of the different cohorts and their seed types in the greenhouse experiment.

162

163

Figure 3. The effect of originating seed type on two fitness metrics, aerial seeds (A-C) and basal

seeds (D-F), and basal to aerial ratio (G-I) in the field. In each row, a response variable is

graphed for the different seed types against the total water available to the plant during

germination and growth (A, D and G) and the plant’s germination cohort (B, E and H). Graphs

C, F and I show the response variable of a row graphed directly against seed type. 2017-2018

and 2018-2019 data are combined in all graphs. Color legend at the top left applies to all colored

graphs (B = basal seeds, L = long seeds, W= winged seeds). Overall, plants that received more

water and germinated earlier had increased production of aerial and basal seeds. Within this

pattern, aerial and basal seed production increased allometrically with aerial seeds being

produced more rapidly. Thus, basal to aerial ratios decreased with increasing rainfall and earlier germination cohorts.

164

165

Figure 4. The effect of originating seed type on two fitness metrics, aerial seeds (A-C) and basal seeds (D-F), and basal to aerial ratio (G-I) in the greenhouse. In each row a response variable is graphed for the different seed types against the total water available to the plant during germination and growth (A, D and G) and the plant’s germination cohort (B, E and H). Graphs

C, F and I show the response variable of a row graphed directly against seed type. Color legend at the top left applies to all colored graphs (B = basal seeds, L = long seeds, W= winged seeds).

Overall, plants that received more water had increased production of aerial and basal seeds.

Within this pattern, aerial and basal seed production increased allometrically with aerial seeds being produced more rapidly. Thus, basal to aerial ratios decreased with increasing rainfall.

Earlier germination cohorts had lower production of aerial seeds but higher production of basal seeds, except for plants originating from long seeds which produced less basal seeds in earlier germination cohorts. Basal to aerial seed ratios decreased in later germination cohorts.

166

167

Figure 5A. 2017-2018 Mean fitness metrics (natural log transformed) of field plants parsed by watering treatment and germination cohort. Each row represents one water treatment while all the graphs in that row represent the germination cohorts exposed to that watering treatment. For each fitness metric along the x axis, bars represent means which are provided for and shaded by basal (B), long (L) and winged (W) seed originating plants. Fitness metrics: 1. Aerial seeds set

(log), 2. Basal seeds set (log), 3. Basal to aerial seed ratio. Error bars represent one standard deviation. If error bars are missing, then there were not enough plants in that cohort and water treatment combination to calculate a standard deviation. Blank graphs indicate that no germinants survived to reproduction in that water treatment and cohort combination. All fitness metrics were natural log transformed prior to calculation of means and standard deviations.

168

169

Figure 5B. 2018-2019 Mean fitness metrics (natural log transformed) of field plants parsed by watering treatment and germination cohort. Each row represents one water treatment while all the graphs in that row represent the germination cohorts exposed to that watering treatment. For each fitness metric along the x axis, bars represent means which are provided for and shaded by basal (B), long (L) and winged (W) seed originating plants. Fitness metrics: 1. Aerial seeds set

170

171

Figure 6. Mean fitness metrics (natural log transformed) of greenhouse plants parsed by germination cohort and watering treatment. Each row represents one water treatment while all the graphs in that row represent the germination cohorts exposed to that watering treatment. For each fitness metric along the x axis, means are provided for basal (B), long (L) and winged (W) originating plants. Fitness metrics: 1. Aerial seeds set (log), 2. Basal seeds set (log), 3. Basal to aerial seed ratio. Error bars represent one standard deviation. All fitness metrics were natural log transformed prior to summarizing.

172

Figure 7. Seed allometry in the field and in the greenhouse represented as the number of basal

seeds set versus the number of aerial seeds set for the different seed morphs of P. heterocarpa.

Regression lines for graphs are grouped by seed type (B = basal, L = long, W = winged). A.

Relationship for the first field season (2017-2018). B. Relationship for the second field season

(2018-2019). C. Relationship for the plants grown in the greenhouse. Seeds set were transformed using the natural logarithm. Legend at the top right applies to all graphs. A positive slope was observed for all groups, indicating that bigger P. heterocarpa plants tend to have more aerial seeds as compared to basal seeds. This relationship was stronger for field (A-B) as compared to

greenhouse plants (C).

173

Main Text Tables

Table 1. Germinants, germination fractions, plants that reproduced and reproduction fractions across water treatments and years

(rows) in the field experiment. The top row provides summaries for each seed type across all years and treatments. The second and third row further divide this data up by year, and the remaining rows by year and water treatment.

174

Table 2. Germinants, germination fractions, plants that reproduced and reproduction fractions across germination cohorts and water

treatments in the greenhouse. A. Summaries of germination and germination fractions across and within cohorts. The top row provides summaries for each seed type across all germination cohorts while the rows beneath it summarize the data by germination cohort. B.

Summaries of sample sizes for greenhouse plants (successful transplants) across all watering treatments (top row) and within watering

treatments (remaining rows).

175

Table 3. Results of the linear mixed models used to evaluate the effects of seed type on germiantion timing and fraction in the field.

Plot was used as a random effect in all models. For every model the degrees of freedom (df), Chi-sqaured statistic (X2) and p-value

(P) are provided. In addition, the “Effect” column shows the overall effect of the various predictor variables on germiantion. The

“Significant Interactions” column at the end of the table lists any significant interactions in the models. The “Model R2 (Mar., Con.)” column lists the marginal R2 (Mar.) which represents the variance explained by the fixed effects and the condiditonal R2 (Con.) which represents the variance explained by the entire model. Finally, the partial R2 for any seed type interactions are listed in the “Partial R2 for seed type interaction” column. Significant model parameters are bolded.

176

Table 4. Results of the linear mixed models used to evaluate the effects of seed type on germiantion fraction in the greenhouse. Petri dish was used as a random effect in all models. For every model the degrees of freedom (df), Chi-sqaured statistic (X2) and p-value (P) are provided.

In addition, the “Effect” column shows the overall effect of the various predictor variables on germiantion fraction. The “Model R2 (Mar., Con.)” column lists the marginal R2 (Mar.) which represents the variance explained by the fixed effects and the condiditonal R2 (Con.) which represents the variance explained by the entire model. Significant model parameters are bolded.

177

Table 5. Results of the linear mixed models used to evaluate the effects of seed type, germination cohort, watering treatment (total

water received during the germination and growth period, log transformed) and season on various fitness outcomes (all log

transformed) of P. heterocarpa in the field experiment. Plot was used a random effect for all models. For every model the degrees of

freedom (df), Chi-sqaured statistic (X2) and p-value (P) are provided. In addition, the “Effect” column shows the overall effect of the

various predictor variables on the two fitness metrics and basal to aerial ratio. The “Significant Interactions” column at the end of the

table lists any significant interactions in the models while the “Interaction p-values” columns lists their corresponding p-values. The

“Model R2 (Mar., Con.)” column lists the marginal R2 (Mar.) which represents the variance explained by the fixed effects and the

condiditonal R2 (Con.) which represents the variance explained by the entire model. Finally the partial R2 for any seed type

interactions are listed in the “Partial R2 for seed type interaction” column. Significant model parameters are bolded.

178

Table 6. Results of the linear mixed models used to evaluate the effects of seed type, germination cohort and watering treatment (total

water received during the germination and growth period, natural log transformed) on various fitness outcomes (all natural log

transformed) of P. heterocarpa in the greenhouse experiment. Rack was used a random effect for all models. For every model the

degrees of freedom (df), Chi-sqaured statistic (X2) and p-value (P) are provided. In addition, the “Effect” column shows the overall

effect of the various predictor variables on the two fitness metrics and basal to aerial ratio. The “Significant Interactions” column at

the end of the table lists any significant interactions in the models while the “Interaction p-values” columns lists their corresponding p- values. The “Model R2 (Mar., Con.)” column lists the marginal R2 (Mar.) which represents the variance explained by the fixed effects

and the condiditonal R2 (Con.) which represents the variance explained by the entire model. Finally the partial R2 for any seed type

interactions are listed in the “Partial R2 for seed type interaction” column. Significant model parameters are bolded.

179

Table 7. Model statistics for the allometry analysis for plants from the field and greenhouse experiments. Each row represents one analysis conducted. The “Model Number” column serves as a reference row for the results presented in the main text. The data column indicates the dataset (field 2017-2018, field 2018-2019 or greenhouse) used for that row, the “Model” column specifies the full model applied in that row and the “Model Component Evaluated” specifies which part of the model we were interested in, using the base notation y ~ x1 * x2. The “Reason” column indicates that we were either interested in how the slope (allometric relationship) of basal to aerial seeds differs across populations and with aridity or the intercept of this slope. The final columns on the right report the R2, p-values and slopes. Slopes are only reported for the first models. For models in which we were only interested in the interaction of two predictor variables, we do not report R2 values.