13 RESEARCH ARTICLE Anomalous δ C in Particulate Organic Carbon at the 10.1029/2019JG005276 Chemoautotrophy Maximum in the Cariaco Basin Special Section: Mary I. Scranton1 , Gordon T. Taylor1 , Robert C. Thunell2,3 , Frank E. Muller‐Karger4, Special Collection to Honor 4,5 6 7 7 the Career of Robert C. Yrene Astor , Peter Swart , Virginia P. Edgcomb , and Maria G. Pachiadaki Thunell 1School of Marine and Atmospheric Sciences, Stony Brook University, Stony Brook, USA, 2Department of Geological Sciences, University of South Carolina, Columbia, SC, USA, 3Deceased on 30 July 2018, 4College of Marine Science, Key Points: University of South Florida, St. Petersburg, FL, USA, 5Fundación La Salle de Ciencias Naturales, EDIMAR, Caracas, • Particulate organic carbon at the O2/ Venezuela, 6Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, FL, USA, 7Woods Hole H S interface in the Cariaco is 2 Oceanographic Institution, Woods Hole, MA, USA occasionally isotopically very heavy 13 (δ CPOC as high as −16‰) 13 • The maximum in δ CPOC corresponds with low C/N ratios of Abstract A chemoautotrophy maximum is present in many anoxic basins at the sulfidic layer's upper particulate organic matter, implying boundary, but the factors controlling this feature are poorly understood. In 13 of 31 cruises to the Cariaco the carbon is relatively fresh Basin, particulate organic carbon (POC) was enriched in 13C(δ13C as high as −16‰) within the • The maximum also corresponds POC with peaks in chemoautotrophy and oxic/sulfidic transition compared to photic zone values (−23 to −26‰). During “heavy” cruises, fluxes of O2 − − RuBisCO form II genes implying and [NO3 +NO2 ] to the oxic/sulfidic interface were significantly lower than during “light” cruises. δ13 that CPOC is affected by Cruises with isotopically heavy POC were more common between 2013 and 2015 when suspended particles chemoautotrophy below the photic zone tended to be nitrogen rich compared to later cruises. Within the chemoautotrophic 13 Supporting Information: layer, nitrogen‐rich particles (molar ratio C/N< 10) were more likely to be C‐enriched than nitrogen‐poor • Supporting Information S1 particles, implying that these inventories were dominated by living cells and fresh detritus rather than • Supporting Information S2 laterally transported or extensively decomposed detritus. During heavy cruises, 13C enrichments persisted to 1,300 m, providing the first evidence of downward transport of chemoautotrophically produced POC. Dissolved inorganic carbon assimilation during heavy cruises (n = 3) was faster and occurred deeper than Correspondence to: M. I. Scranton, during light cruises (n = 2). Metagenomics data from the chemoautotrophic layer during two cruises [email protected] support prevalence of microorganisms carrying RuBisCO form II genes, which encode a carbon fixation enzyme that discriminates less against heavy isotopes than most other carbon fixation enzymes, and Citation: metatranscriptomics data indicate that higher expression of form II RuBisCO genes during the heavy cruises Scranton, M. I., Taylor, G. T., Thunell, at depths where essential reactants coexist are responsible for the isotopically heavier POC. R. C., Muller‐Karger, F. E., Astor, Y., Swart, P., et al. (2020). Anomalous δ13C in particulate organic carbon at the chemoautotrophy maximum in the 1. Introduction Cariaco Basin. Journal of Geophysical Research: Biogeosciences, 125, Carbon isotopic composition of particulate organic matter in aquatic environments is initially determined by e2019JG005276. https://doi.org/ the isotopic composition of the dissolved inorganic carbon (δ13C ) used by autotrophs and by the isotopic 10.1029/2019JG005276 DIC 13 fractionation that takes place when carbon is fixed. DIC in seawater typically has a δ CDIC of around 0 to 13 Received 10 JUN 2019 −1‰ (Hoefs, 2009) and δ CPOC of marine particulate organic matter typically varies between −18 and Accepted 3 JAN 2020 −34‰, depending on location (Goericke & Fry, 1994). While important exceptions exist (see Fry & Accepted article online 30 JAN 2020 Wainright, 1991; Reinfelder et al., 2000), most phytoplankton appear to use the C3 carbon fixation pathway

leading to overall isotopic fractionation factors (εp) of 16 to 25, which result in phytoplankton biomass with δ13C signatures of −19 to −24‰. (Fry, 2006; O'Leary, 1988). 13 One might expect to see declines in δ CPOC at depths where chemoautotrophs fix carbon in stratified mar- ine systems, such as observed in Kyllaren fjord (van Breugel et al., 2005). This results from isotopically light organic matter being preferentially respired back to DIC pool throughout the water column. Thus, DIC and 13 δ CPOC produced by chemoautotrophs would become lighter at depths where significant respiration is occurring. However, the extent of isotopic fractionation by chemoautotrophic microorganisms varies among taxa. For example, ε‐proteobacterial symbionts of hydrothermal vent gastropods and polychaetes, which

chemoautotrophically fixCO2 by the reductive tricarboxylic acid (rTCA) cycle, can be isotopically much hea- δ13 − − ‰ vier ( CBIOMASS = 8to 12 ) than typical photoautotrophic primary producers (Campbell et al., 2003;

Zbinden et al., 2015). In fact, at least six different CO2 fixation pathways are known and have widely varying

©2020. American Geophysical Union. energetic costs (ATP and reductants consumed per pyruvate molecule produced) and carbon isotopic fractio- δ13 − − ‰ All Rights Reserved. nations yielding CBIOMASS ranging from 0.2 to 30 (reviewed in Berg et al., 2010, and Hügler & Sievert,

SCRANTON ET AL. 1of17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

2011). Even among organisms that share the Calvin‐Benson‐Basham (CBB) reductive pentose phosphate

cycle, εp varies depending on which form of the ribulose‐1,5‐bisphosphate carboxylase/oxygenase (RuBisCO) enzyme is used. For example, Robinson and Cavanaugh (1995) determined that endosymbiotic γ‐proteobacteria in the hydrothermal vent vestimintiferan Riftia patchyptila fix carbon with form II δ13 − − ‰ − − ‰ RuBisCO and produce CBIOMASS of 9to 16 compared to 17 to 25 evident in organisms using δ13 form I RuBisCO. Thus, CBIOMASS signatures can provide insights into mechanisms by which DIC is assimilated autotrophically. The widespread importance of chemoautotrophy in environmental redox gradients has become increas- ingly evident in recent years. Significant dark carbon fixation has been reported in the Black Sea (Jørgensen et al., 1991; Yilmaz et al., 2006), in the Cariaco Basin (Taylor et al., 2001; Tuttle & Jannasch, 1973), in the Baltic Sea (Jost et al., 2008), in Mariager fjord (Zopfi et al., 2001), in Namibian shelf waters (Lavik et al., 2009), and in the oxygen minimum zones of the Eastern Tropical South Pacific (Schunck et al., 2013). In most cases, a light scattering (particulate) maximum coincides with elevated microbial bio- mass in the upper portion of the sulfidic zone (Taylor et al., 2001; Yilmaz et al., 2006) where sulfur‐ oxidizing bacteria have been identified (Glaubitz et al., 2010; Jørgensen et al., 1991; Kirkpatrick et al., 2018; Tuttle & Jannasch, 1972; Zopfi et al., 2001). Modeling geochemical balances that are consistent with measured rates of dark carbon fixation has been challenging (Li et al., 2012). However, recent biogeochem- ical flux models yield much better agreement between supply and demand for reactants (Louca, Scranton, et al., 2019, Louca, Astor, et al., 2019). As in many other stratified oxygen‐depleted water columns, chemoautotrophic assemblages across the Cariaco's redox transitional zone are comprised of ammonia‐oxidizing, nitrite‐oxidizing, anaerobic ammonia‐oxidizing (anammox), and reduced sulfur‐oxidizing microorganisms (Cernadas‐Martin et al., 2017; Montes et al., 2013; Suter et al., 2018; Taylor et al., 2001; Wakeham et al., 2012). During the present study's timeframe (2013–2017), ammonia‐oxidizing archaea in marine group I, ammonia‐oxidizing beta‐ and γ‐proteobacteria, and nitrite‐oxidizing Nitrospina (δ‐proteobacteria) have been observed across the oxycline (Cernadas‐Martin et al., 2017; Suter et al., 2018). Peaks in anammox bacteria (Candidatus Scalindua) also have been detected in Cariaco's transitional waters by an array of methods, including lipid biomarkers, fluorescence in situ hybridization, quantitative polymerase chain reaction (PCR), and 16S rRNA sequencing (Wakeham et al., 2012; Rodriguez et al., 2015; Cernadas‐Martin et al., 2017; Suter et al., 2018). Over the entire Cariaco Ocean Time‐Series program (1995–2017), taxa comprising some chemoautotrophic functional groups have changed demonstrably. For example, the dominant S‐oxidizers taxa early in the record were ε‐proteobacteria, which appear to have been replaced by members of the γ‐proteobacteria after May 2009 (Lin et al., 2006; Madrid et al., 2001; Taylor et al., 2018). Putative γ‐proteobacteria S‐oxidizers (GSO) dominating the shallow anoxic and euxinic depths were represented by the SUP05 clade and a single OTU in the Ectothiorhodospiraceae family, closely matching an uncultured Thiorhodospira sp. (Suter et al., 2018). Related GSOs dominate redoxclines of other oxygen‐depleted environments, such as the Black Sea (Fuchsman et al., 2011), Baltic Sea (Glaubitz et al., 2013), shelf water on the Namibian coast (Lavik et al., 2009), Saanich Inlet (Walsh et al., 2009), and the Pacific Ocean OMZs (Stewart et al., 2012). Other likely che- moautotrophic sulfur oxidizers were also present in Cariaco surveys, including the δ‐proteobacteria MG‐B (a.k.a. SAR324), Arctic97B‐4, and SAR202 (Swan et al., 2011). Many of the observed taxa, such as SUP05, − are known or suspected to use NO3 to oxidize reduced S species (Shah et al., 2017; Walsh et al., 2009). Thus, the Cariaco Basin's redoxcline is inhabited by an assortment of chemoautotrophs that link the S, N, and C cycles through multiple pathways. 13 Previous measurements of suspended δ CPOC collected from the Cariaco Basin using bottle casts (Fry et al., 1991) or large volume pumps (Wakeham et al., 2012) ranged from about −24 to −28‰, with lowest values around the oxic/sulfide interface and coinciding with the chemoautotrophy maximum (Wakeham et al., 13 2012). δ CPOC in sediment trap material from 275, 455, 930, and 1,255 m varied from −22.6 to −17.6‰ between 1996 and 1999, with the heaviest values occurring during strong upwelling (Woodworth et al., 2004). Woodworth and colleagues hypothesized a connection between periods of intense upwelling, such as those in 1996 and 1997, and times when isotopically heavy carbon was detected in the sinking fraction.

They found a negative correlation between the concentration of dissolved CO2 and the estimated

SCRANTON ET AL. 2of17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

fractionation factor, εp. They suggested that more intense upwelling (and higher surface pCO2) was asso- ciated with heavier isotopic signatures in sinking particles due to a carbon concentrating mechanism similar to C4 photoautotrophs as demonstrated for cultured diatoms by Reinfelder et al. (2000). This provided an explanation for why lower values were not seen during times of upwelling of isotopically lighter DIC from deep waters. However, this study did not address the possibility that autotrophic organisms at depth using different carbon fixation pathways may contribute significant amounts of 13C‐enriched organic matter to 13 the sinking flux. The present study seeks to examine the relationship between the suspended δ CPOC and dark carbon fixation within the basin.

2. Materials and Methods 2.1. Site Description and Sampling The Cariaco Basin is 1,400‐m deep, located in a seasonal upwelling zone on the continental shelf of Venezuela and isolated from the Caribbean by a shallow sill (~100 m) with two channels extending only to about 150‐m water depth. A general description of carbon cycling in the basin and its relationship to glo- bal carbon cycling can be found in Muller‐Karger, Astor, Benitez‐Nelson, Buck, et al. (2019). The basin has an oxic/anoxic transition zone, currently between ~250 and 300 m. Sampling for this entire study was con- ducted at a single site (10°30′N, 64°40′W), the primary sampling station of the Cariaco Ocean Time Series program (http://www.imars.usf.edu/cariaco), aboard the B/O Hermano Ginés, operated by Estación de Investigaciones Marinas (EDIMAR), Fundación La Salle de Ciencias Naturales located on Margarita Island, Venezuela. Between May 2013 and January 2017, discrete water samples for particulate organic car- 13 bon (POC), particulate nitrogen (PN), DIC, δ CPOC, oxygen, hydrogen sulfide, prokaryoplankton biomass, and nutrients were collected with a SeaBird rosette fitted with 12 TFE‐lined 8‐ or 12‐L Niskin bottles on reg- ular monthly Cariaco Ocean Time Series cruises. Standard sampling depths for monthly cruises were as fol- lows: 2, 7, 15, 25, 35, 55, 75, 100, 130, 160, 200, 250, 270, 300, 350, 400, 500, 750, and 1,302 m (+ 2m) plus an additional depth located to capture the optical backscattering maximum near the oxygen‐sulfide interface. On additional dates, during semiannual biogeochemistry cruises, samples for hydrogen sulfide, particulate sulfur, oxygen, nutrients, and chemoautotrophic rates were also collected at fixed distances below and above (i.e., approximately 0, +10, +20, and +30m) the midwater light scattering maximum, which usually coin- cided with microbial biomass/activity peaks (Taylor et al., 2001). In November 2015, samples also were col- 13 lected on the biogeochemistry cruise for δ CDIC measurements. Details of cruise dates and samples appear in Table 1. Hydrographic profiling utilized a rosette equipped with a SeaBird CTD, a Wetlabs C‐star trans- missometer (660 nm), a Wetlabs ECO chlorophyll fluorometer, and an SBE43 oxygen probe. Sampling and processing protocols for hydrography, pH, total alkalinity (TA), discrete dissolved oxygen, POC, PN, nutrients, and hydrogen sulfide concentrations appear in Astor et al. (2011) and Scranton et al. (2014). Detection limits for oxygen were 1–2 μM and for sulfide were about 1 μM. Particulate elemental sulfur was as analyzed as described by Li et al. (2008). Nutrient data were provided by Drs. K. Fanning and K. Buck, and W. Abbott (Univ. South Florida). Data and protocols for all biogeochemistry parameters are avail- able from Scranton et al. (2019) at https://www.bco‐dmo.org/dataset/3120 and Muller‐Karger, Astor, Scranton, et al. (2019) at https://www.bco‐dmo.org/dataset/3093. CTD data are available from Muller‐ Karger, Astor, Benitez‐Nelson, Scranton, et al. (2019) at https://www.bco‐dmo.org/dataset/3092.

2.2. Isotopic (δ13C) Composition of POC 13 Samples for suspended δ CPOC were collected at the standard Cariaco Ocean Time‐Series depths on 31 cruises between 9 May 2013 and 12 January 2017 (Table 1). Two‐liter samples from each Niskin (one‐liter samples from November 2015 which were collected on the biogeochemistry cruise) were filtered onto 25‐mm precombusted GF/F glass fiber filters and stored in petri dishes. These samples were dried at 60°C for 12 hours in the shore laboratory, returned to University of South Carolina, and analyzed for δ13C using a GV Isoprime isotope ratio mass spectrometer coupled to a Euro elemental analyzer. All samples and stan- dards were combusted at 1020°C. The results are reported in the conventional δ notation relative to standard 13 Vienna Pee Dee Belemnite. Precision based on replicate analysis of standards was < 0.15‰. δ CPOC data are available at https://doi.org/10.6084/m9.figshare.8214470.v1.

SCRANTON ET AL. 3of17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

Table 1 Summary of Cruises With (“Heavy”) and Without (“Light”) Subsurface Enrichments in 13C of the POC Pool Cruise ###_leg Date δ13C‐POC at 7 m (‰) Maximum δ13C‐POCc (‰) Prokaryoplankton biomassd (g C/m2) Additional analysesf

Heavy CAR‐201_2 9 May 2013 −24.15b −16.00 1.21 CAR‐201_3 14 May 2013 Chemo CAR‐203_1 11 July 2013 −24.61 −20.43 1.01 CAR‐204_1 13 August 2013 −23.51 −20.12 1.09 CAR‐205_1 11 September 2013 −23.57 −18.11 0.92 CAR‐206_1 8 October 2013 −24.17 −16.23 0.79 CAR‐209_1 14 January 2014 −22.08 −25.08 0.40 CAR‐210_1 4 February 2014 −19.58 −21.66 0.63 CAR‐211_1 1 April 2014 −21.85 −23.72 1.10 CAR‐212_1 6 May 2014 −20.61b −23.75 CAR‐212_2 8 May 2014 Meta CAR‐212_3 12 May 2014 1.20 Chemo CAR‐213_1 13 June 2014 −22.09 −17.03 1.52 CAR‐214_1 3 September 2014 −23.61 −19.65 0.69 CAR‐215_1 7 October 2014 −23.65 −20.41 1.78 13 CAR‐224_1 14 November 2015 Chemo, δ CDIC CAR‐224_2 16 November 2015 −23.09b −24.13 0.43 mean −22.81 −20.41 0.98 Light CAR‐202_1 11 June 2013 −24.55 n.d. 0.86 CAR‐207_1 7 November 2013 Chemo CAR‐207_3 13 November 2013 −25.18 −25.05 1.32 CAR‐208_1 4 December 2013 −24.95 −25.26 0.45 CAR‐216_2 6 November 2014 Meta CAR‐216_3 10 November 2014 0.71 Chemo CAR‐216_4 12 November 2014 −24.03 −25.30 CAR‐217_1 4 December 2014 −22.72 −25.48 1.24 CAR‐218_1 15 February 2015 −23.26 −24.30 0.51 CAR‐219_1 10 March 2015 −20.65 −22.87 0.51 CAR‐220_1 24 April 2015 −19.75 −24.71 0.75 CAR‐221_1 29 July 2015 −22.90 −25.51 n.d. CAR‐222_2 18 August 2015 −22.91 −24.35 0.89 CAR‐223_1 10 September 2015 −24.04 −24.82 0.86 CAR‐225_1 9 December 2015 −24.97 −25.51 0.64 CAR‐226_1 12 January 2016 −21.31b −25.16 0.56 CAR‐227_1 4 February 2016 −22.05 −24.25 0.56 CAR‐228_1 11 May 2016 −24.31 −25.25 1.12 CAR‐230_1a 20 September 2016 −23.35 −24.79 0.47 CAR‐231_1 21 December 2016 −23.54 −24.64 0.74 CAR‐232_1 12 January 2017 −21.35 −25.35 0.58 mean −23.10 −25.05 0.75 t‐test p‐valuee <0.31 <0.001 <0.035 13 Note. “Heavy” cruises are defined as those with one or more CPOC sample in the 200‐ and 400‐m depth range enriched by at least 2‰ compared to shallower and deeper samples. Abbreviations: DIC: dissolved inorganic carbon; POC: particulate organic carbon. aCAR‐229_1 omitted because samples below 55 m unavailable. bSample from 15 m because 7‐m sample lost. cBetween 200 and 400 m. dProkaryoplankton biomass integrated between 200 and 400 m and based on microscopic enumeration. eResults from one‐tailed t tests comparing 13C‐enriched to unenriched sample sets; p < 0.05 indicates that differences in data sets are greater than that would be expected by chance. fChemo = dark 14C‐DIC assimilation profiled; 13 13 meta = metagenomics and metatranscriptomics sampled; δ CDIC = C‐signature of the pool profiled.

2.3. Total DIC and Its δ13C Niskin bottle subsamples from all time series standard depths were processed to measure pH and TA to per- mit calculation of total DIC. For pH, seawater samples were drawn directly into a 10‐cm spectrophotometer cell and colorimetric analyses were made shipboard within 1 hour of sample collection. TA samples were

collected in 250‐ml opaque glass bottles, poisoned with 50 ml of saturated HgCl2 solution, stored at 4°C, and then analyzed in the shore laboratory. Precision of alkalinity and pH measurements are ±4 μmol/kg and ±0.003 pH units, respectively. Total DIC was calculated as described in Astor et al. (2013).

SCRANTON ET AL. 4of17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

13 Samples for determination of δ CDIC were collected in November 2015 during our biogeochemistry cruise from Niskin bottles using a syringe equipped with a disposable 0.2‐μm pore size filter to remove any parti- culate calcium carbonate. Water was filtered directly into 50‐ml round‐bottomed glass vials that could be sealed with no bubble using a butyl rubber septum and metal crimp. Before applying septum and crimp seal,

200 μl of 250 mM HgCl2 solution were added as a preservative. Samples were run in the lab of Peter Swart using a Gasbench attached to a Thermo Delta V. Precision among triplicate analyses of prepared samples varied between 1 and 32% (relative standard deviation) and averaged 7% (about 0.2‰). Data are presented in Table S1 in the supporting information.

2.4. Prokaryoplankton Biomass Prokaryoplankton (Bacteria + Archaea) were enumerated in 2% buffered formaldehyde preserved samples (10–20 ml) captured on 0.2‐μm polycarbonate membranes and stained with 4',6‐diamidino‐2‐phenylindole as described in Taylor et al. (2001). Losses in cell counts during storage at 4°C were corrected for storage time using an empirical decay curve. For biomass estimates, prokaryotic cell sizes in a subset of all samples (n = 215 samples) were estimated by visually sorting into eight morphology/size classes based on their linear dimensions. Mean biovolumes of each of the eight size classes were computed, and cellular carbon biomass (C) was estimated from biovolume (V) using an allometric carbon to volume extrapolation function (pg C = 0.12 V0.72; Norland, 1993). Total prokaryoplankton biomass (μg C/L) in each sample was computed by multiplying mean cellular biomasses (fg C/cell) for each depth by its cell concentration.

2.5. Chemoautotrophy Dissolved inorganic carbon fixation below the euphotic zone was not a routine monthly time series measure- ment but was measured on biogeochemistry cruises (Table 1; CAR 201_3, 207_1, 212_3, 216_3 and 224_2; cruise#_leg#). These cruise legs were within 2–6 days (mean = 4) of the monthly cruises when pH, TA, oxy- 13 14 − gen, nutrient, POC, PN, and δ CPOC collections were made. Assimilation of C‐HCO3 into acid‐insoluble particles (presumably cells) was measured after 18‐hour dark incubations of triplicate subsamples under simulated in situ conditions as described by Taylor et al. (2001). Total DIC assimilation was calculated from 14C‐activity (μCi) on filters (control‐corrected), specific activity (measured total DIC divided by total μCi added), and an isotope fractionation factor (1.06). Variability among triplicate subsamples ranged between 0 and 140% (relative standard deviation) and averaged 41%, which primarily reflects heterogeneous distribu- tions of microbiota among the subsamples contained in 40‐ml gas‐tight incubation vials.

2.6. Metagenomes and Metatranscriptomes Routine molecular biological analyses were beyond the scope of the CARIACO Ocean Time‐Series pro- gram. However, a related project enabled collection of material for DNA and RNA analyses during CAR 212_2 and CAR 216_2. Water samples collected using Niskin bottles were used for DNA analyses to con- struct metagenomes. Samples were filtered sequentially through EMD Millipore 2.7‐μm glass fiber filters, then through 0.2‐μm Sterivex filters. DNA from the 0.2‐μm Sterivex filters (that have captured prokaryotic cells with cell diameter from 0.2 to 2.7 μm) was extracted according to Frias‐Lopez et al. (2008) and Ganesh et al. (2014) described in detail in Suter et al. (2018). After extraction, DNA was purified with the Genomic DNA Clean and Concentrator‐25 kit (Zymo Research), eluted into 10mM Tris‐HCl, and fro- zen until downstream analysis. An aliquot of the extracted DNA was sent to Georgia Genomics Facility for library preparation and sequencing. The Illumina NextSeq platform was used for paired‐end 2×150 sequencing. The R1 and R2 reads were filtered using Trimmomatic (Bolger et al., 2014), which performs a “sliding window” trimming, removing sequence data once the average quality within the window (eight nucleotides used here) falls below a threshold (set to 12). The length of the trimmed sequences was set to a minimum of 50 nucleotides. The trimmed reads were assembled into contigs using IDBA 1.1.1 (Peng et al., 2010). The assembled metagenomes were annotated using the IMG pipeline (https://img.jgi.doe.gov/cgi‐ bin/mer/main.cgi). RNA samples also were collected to construct metatranscriptomes from several depths chosen to capture anoxic and sulfidic regimes using the modified Microbial Sampler‐In Situ Incubation Device (MS‐SID), a robotic instrument for conducting in situ tracer incubation studies and/or collecting and preserving filtered microbial samples in situ (Edgcomb et al., 2016; Taylor et al., 2015; Taylor & Doherty, 1990). For the May 2014 cruise (CAR 212_2), depths were 295 and 314 m, and for the November 2014 cruise (CAR 216_2),

SCRANTON ET AL. 5of17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

13 13 Figure 1. Vertical profiles of (a) δ C values of particulate organic carbon (δ CPOC), (b) dark dissolved inorganic carbon (DIC) assimilation rates by chemoautotrophs and hydrogen sulfide concentrations, (c) suspended POC concentrations, (d) carbon:nitrogen ratios of suspended particulates,and (e) total DIC (o) and the δ13C of the DIC pool (⋄). Samples for 13 POC, POC/PN, and δ CPOC are plotted for samples from 9 May 2013 (CAR‐201_2) and 12 November 2014 (CAR‐216_4). DIC assimilation rates and H2S were measured on 14 May 2013 (CAR‐201_3) and 10 November 2014 (CAR‐216_3). 13 DIC and δ CDIC were measured in samples collected during CAR‐224 (14 November 2015). The error bars represent 1 SD of triplicate samples. P‐values > 0.05 of analysis of variance (ANOVA) comparing cruises were deemed not significant (n.s.).

the depths were 247 and 267 m; their geochemical conditions are described in Table S2. Water was sequentially filtered through EMD Millipore 2.7‐μm glass fiber filters and then through 0.2‐μm Millipore Express polysulfone membranes and preserved immediately in situ with RNAlater® to preserve total RNA. Upon MS‐SID retrieval, preserved filters were transferred to cryovials with additional RNAlater and stored frozen until extraction. RNA was extracted from the 0.2‐μm polysulfone membranes (that have captured prokaryotic cells with cell diameter from 0.2 to 2.7 μm) using a modification of the mirVana miRNA Isolation kit (Ambion, Life Technologies, Carlsbad, CA, USA) as in Stewart et al. (2012). Briefly, filters were thawed on ice and RNA stabilizing buffer was removed by pipette from cryovials. Cells on filters were lysed by adding lysis buffer and miRNA homogenate additive (Ambion) into the cryovial or cartridge. After vortexing and incubation on ice, lysates were transferred to RNAase‐free tubes and processed via acid–phenol/chloroform extraction according to the kit protocol. The TURBO DNA‐free kit (Ambion, Foster City, CA, USA) was used to remove DNA, and the extracts were purified using the RNeasy MinElute Cleanup Kit (Qiagen, Hilden, Germany). Removal of DNA was confirmed by PCR using as the forward primer (5′‐AYTGGGYDTAAAGNG‐3′) and a mix of reverse primers (5′‐GCCTTGCCAGCCCGCTCAG, TACCRGGGTHTCTAATCC, TACCAGAG TATCTAATTC, CTACDSRGGTMTCTAATC and TACNVGGGTATCTAATCC‐3′ in a 6:1:2:12 ratio, respec- tively), designed to cover most of the Bacteria domain (Cole et al., 2009). cDNA libraries were prepared using the ScriptSeq RNA‐Seq Library Preparation Kit (Illumina). Excess nucleotides and PCR primers were removed from the library using the Agencourt AMPure™ XP (Beckman‐Coulter) kit. The Illumina NextSeq platform was used for paired‐end 2×150 sequencing at the Georgia Genomic Facility. The R1 and R2 reads were filtered using Trimmomatic (Bolger et al., 2014).

SCRANTON ET AL. 6of17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

Figure 2. Temporal and vertical variations in δ13C signatures of suspended particulate organic carbon throughout the observation period (9 May 2013 [CAR‐201] to 12 January 2017 [CAR‐232]).

Genes encoding for RuBisCO were extracted from the metagenomes. In order to detect different forms of the retrieved RuBisCO genes, a phylogenetic tree was constructed. RuBisCO gene sequences from cultured iso- lates available in the NCBI database were aligned with sequences in our assembled metagenomes using ClustalW (Thompson et al., 1994). Tree construction used MEGA7 (Kumar et al., 2016) and the Maximum Likelihood method based on the Le and Gascuel (2008) model. Initial tree(s) for the heuristic search were obtained automatically by applying Neighbor‐Join and BioNJ algorithms to a matrix of pairwise distances estimated using a JTT model, and then selecting the topology with superior log likelihood value. A discrete Gamma distribution was used to model evolutionary rate differences among sites. The rate variation model allowed for some sites to be evolutionarily invariable. The analysis involved 166 amino acid sequences. All positions with less than 95% site coverage were eliminated. Fewer than 5% alignment gaps, missing data, or ambiguous bases were allowed at any position. A total of 312 positions were in the final dataset.

Figure 3. δ13C signatures of suspended particulate organic carbon (POC) observed during cruises in which subsurface 13C‐enrichment were observed (shaded boxes = “heavy cruises”, n = 13; see Table 1) compared to remainder of cruises (open boxes = “light cruises”, n = 18). Data are segregated by layers: (a) epipelagic (7–200 m), (b) redoxcline (200–400 m), and (c) basin interior (400–1,300 m). Analysis of variance (ANOVA) was used to compare data dispersion among “heavy” and “light” cruises in each of the three depth intervals. ANOVA p‐values > 0.05 were deemed not significant (n.s.). Boxes represent the interquartile ranges of all observations (25th–75th percentiles). Internal vertical lines, whiskers, and open circles are medians, 10th to 90th and 5th to 95th percentiles, respectively. The solid and broken lines track the medians of “heavy” to “light” samples, respectively.

SCRANTON ET AL. 7of17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

Based on their placement within the phylogenetic tree, retrieved RuBisCO genes were assigned to either form I or II. Additionally, within both forms several phylogenetically distinct RuBisCO “types” were identified. Metagenomic and metatranscriptomic reads were mapped against these genes using a Burrows‐Wheeler Aligner (Li & Durbin, 2010). The recruited reads were normalized for gene length and library size and con- verted to Reads Per Kilobase of gene per Million reads (RPKM) values. The raw sequencing data were deposited in Sequence Read Archive (SRA).

3. Results and Discussion 13 3.1. Variability in Organic Particulate Matter δ CPOC 13 Water column profiles of δ CPOC isotopic signatures were obtained for the 31 cruises completed between May 2013 and January 2017. Thirteen 13 of these cruises exhibited subsurface maxima in δ CPOC within the oxic/anoxic transition zone between 200 and 400 m (more than 2‰ hea- vier than surface values). An example is shown in Figure 1a (9 May 2013, CAR‐201_2). Cruises when these subsurface 13C enrichments were found are referred to here as “heavy” cruises and the remaining 18 cruises 13 are referred to as “light” (Table 1). δ CPOC of the suspended pool during the light cruises were similar to each other (see example in Figure 1a for on 12 November 2014, CAR‐216_4) or became 13C‐depleted in the upper 400 m of the water column and remained relatively constant (x¯= −25.18 ± 0.73‰) with depth. Isotopically heavy POC was more frequently observed between 2013 and 2015 in samples collected between 200 and 400 m (Figure 2). Differences between heavy and light cruises are best exemplified by exam- Figure 4. Vertical profiles of (a and b) dark dissolved inorganic carbon (DIC) assimilation rates, a measure of chemoautotrophy, and (c and d) ining discrete layers within the epipelagic zone (Figure 3a), the particulate sulfur (symbols + lines), and hydrogen sulfide (lines only) oxygen/sulfide transition zone or redoxcline (Figure 3b), and the euxinic “ ” “ ” 13 concentrations obtained during (a and c) heavy and (b and d) light or sulfidic layers (Figure 3c). Variations in δ CPOC signatures between cruises. The error bars represent 1 SD of triplicate samples. See Table 1 for heavy and light cruises are statistically indistinguishable in the upper cruise dates. 13 150 m or epipelagic zone (median δ CPOC = −24.26 vs −24.44; Figure 3 13 a). Below 150 m, however, δ CPOC values are significantly (p < 0.001) 13 heavier (median δ CPOC = −24.67) and more variable in samples obtained from heavy cruises than light 13 cruises (median δ CPOC = −25.65; Figure 3b). While subtle, this heavier signature was sustained in under- 13 lying waters (median δ CPOC = −24.61 vs −25.28; Figure 3c), suggesting that POC produced in the redox- cline is transported downward. It is often assumed that POC in the water column ultimately derives from in the surface waters or from continental export to coastal waters. However, in the Cariaco Basin, surface water POC typi- 13 cally has a δ CPOC of around −24 to −25‰ and there is little historical evidence for terrestrial POC input 13 (Goni et al., 2009). Thus, the maxima seen in δ CPOC near the oxygen/sulfide transition are intriguing and strongly suggest different sources of POC in the surface euphotic and deeper redox transitional layers. Previous biogeochemical research in the Cariaco Basin examining biomarkers and stable isotopes in sedi- ment trap and high‐volume filtration samples failed to detect signatures of chemoautotrophically produced POC exported to depth (Goni et al., 2009; Wakeham et al., 2012). However, our current observations (Figure 3c) strongly suggest that this is the case. Given that chemoautotrophic bacteria, protistan predators, and their egestate are likely too small to sink, aggregation and particle scavenging are likely mechanisms for their delivery to depth.

3.2. Chemoautotrophy Shipboard measurements of dark DIC assimilation rates were used to define the chemoautotrophic layer and were made within a few days of 3 of the 13 heavy cruises and 2 of the 18 light cruises. Bacterial

SCRANTON ET AL. 8of17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

Figure 5. Temporal and vertical variations in (a) suspended particulate organic carbon (POC; shaded contour plot) and hydrogen sulfide concentrations (white contour lines), (b) suspended particulate nitrogen (PN) concentrations, and (c) molar C/N ratios of suspended organic particulate matter throughout the observation period.

chemoautotrophic production measured by dark DIC assimilation 5 days prior to the CAR‐201_2 and 2 days before the CAR‐216_4 POC collections exhibited pronounced peaks over the same depth interval as 13C enrichments, but depth of peaks differed among variables (Figures 1a–1c). Exact agreement in depth distributions is not necessarily expected because of the temporal separation in measurements. Furthermore, δ13C of the entire POC pool is a longer‐lived signature of activity than “instantaneous” DIC assimilation and both may fluctuate with temporal variations in water mass properties or internal wave propagation. During the heavy cruises, rates of chemoautotrophy were significantly higher and occurred deeper in the water column than rates observed during light cruises (Figures 4a and 4b). Differences in depth distributions

of H2S and particulate sulfur between heavy and light cruises were consistent with those of DIC assimilation (Figures 4c and 4d). Integrated DIC assimilation rates across the transitional zone varied from 35 to 134 − − − − mmol C · m 2 · day 1 during heavy cruises and 13–38 mmol C · m 2 · day 1 during light cruises. Total inte- grated prokaryoplankton biomasses within the oxygen/sulfide transition zone (200–400m) among all heavy cruises were also significantly higher (p < 0.035; t test) than during light cruises (means = 0.98 and 0.75 g C/m2, respectively; Table 1). These observations suggest that when midwater productivity was higher, the dominant chemoautotrophs discriminated less against the heavy carbon isotope producing heavier POC.

13 3.3. Isotopic Composition of Dissolved Inorganic Carbon (δ CDIC) 13 Samples to measure δ CDIC were collected on 14 November 2015 (CAR‐224), as shown in Figure 1e (data in Table S1). Near‐surface values (25–250 m) were about −0.60 ± 0.25‰ and decreased with depth to a

SCRANTON ET AL. 9of17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

Figure 6. As an index of suspended organic particle quality, particulate organic carbon/particulate nitrogen (POC/PN) ratios observed during “heavy” cruises (shaded boxes, n = 13; see Table 1) are compared to “light” cruises (open boxes, n = 18). Analysis of variance (ANOVA) was used to compare data dispersion among “heavy” and “light” cruises in each of the three depth intervals. ANOVA p‐values > 0.05 were deemed not significant (n.s.). The box and whisker parameters are the same as Figure 3.

minimum of −1.77‰ at 300 m just above the particle maximum at the oxygen‐sulfide transition. At depths near the first appearance of sulfide (320‐m depth based on odor, 350‐m depth based on measurements) 13 in November 2015, the values of δ CDIC varied between −0.85 and −1.77‰ near the sulfide interface and averaged −1.64‰ below 350‐m depth. These values are about 1‰ lower than those reported by Deuser 13 (1973). Variations in δ CDIC are clearly too small and in the wrong 13 direction to explain fluctuations in δ CPOC.

3.4. POC, PN, and C/N Ratios in Particles POC concentrations did not significantly differ (p > 0.60 analysis of var- iance [ANOVA]) in the examples of heavy (CAR‐201) and light cruises (CAR‐216). POC typically declined with depth, with small positive excur- sions between 200 and 400 m (Figure 1c). These typical small excursions 13 did not coincide with variations in δ CPOC or chemoautotrophy. POC/PN ratios, however, indicate that suspended particles collected dur- ing the light cruise (CAR‐216) were significantly nitrogen‐depleted (atomic C/N = 6.3–28.3; p < 0.001) at most depths compared to the rela- tively constant and low ratios (atomic C/N = 4.4–9.3) observed in the heavy samples (CAR‐201; Figure 1d). Biological production of organic particles in the euphotic zone is respon- sive to seasonal upwelling of nutrients with biomass maxima typically observed between December and May (Muller‐Karger, Astor, Benitez‐ Figure 7. The tendency of isotopically heavy particles to exhibit low Nelson, et al., 2019). This is reflected in variations in the POC and PN con- particulate organic carbon/particulate nitrogen (POC/PN) ratios. centrations in the upper 75 m of the water column over the entire obser- Data collected between 240 and 350 m from all cruises (n = 31) were subjected to Pearson product moment correlation analysis. The box and vation period (Figures 5a and 5b). The POC pool sizes decreased whisker parameters are the same as Figure 3, except that data are binned in between 75‐ and 200‐m depth. Below 200 m, secondary maxima were increments of 1.0 C/N unit. observed sporadically, apparently uncoupled from the annual upwelling

SCRANTON ET AL. 10 of 17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

Figure 8. Phylogenetic tree of the ribulose‐1,5‐biphosphate carboxylase/oxygenase (RuBisCO) genes retrieved from metagenomic samples collected from 295 and 314 m on 7–9 May 2014 (CAR‐212_2) and from 247 and 267 m on 5–7 November 2014 (CAR216_2). Tree is broadly segregated into RuBisCO form I and RuBisCO form II and more finely into RuBisCO types A‐E. All five types of RuBisCO were represented in the Cariaco metagenomic database. Operational taxonomic units obtained from our Cariaco Basin are designated as “Dxxxxxxxxx” and presented with nearest known prokaryotic autotrophs.

cycle, and often centered near the upper boundary of detectable hydrogen sulfide (Figure 5a). POC concentrations were statistically indistinguishable among all compiled heavy and light cruise samples. However, particulate nitrogen pool sizes were larger in heavy cruise samples below the euphotic zone (typi- cally >80m), resulting in significantly lower C/N ratios (p < 0.001) in the deeper water column (Figure 6). Atomic C/N ratios of particles in the upper 75 m varied between 6 and 28 and were significantly lower (p < 0.007 ANOVA) during heavy cruises (median POC/PN = 8.1) than during light cruises (median POC/PN = 8.9), but only slightly so (Figure 6a). Median POC/PN ratios in samples below the euphotic zone were 7.8 and 10.2 for the heavy and light cruises, respectively (Figures 6b and 6c). C/N ratios of marine par- ticulates often are used as a proxy for organic matter quality, with low values near the Redfield ratio (6.6) indicating fresh biological material and high values indicating decomposed organics. Using this metric, we interpret observed POC/PN ratios to indicate that POC collected throughout the water column during heavy cruises was enriched in fresher material as well as intact cells, consistent with prokaryoplankton bio- mass trends (Table 1). 13 Comparing all POC/PN ratios to their δ CPOC signatures exclusively in the chemoautotrophic layer (240– 350 m) reveals that isotopically heavy samples tend to also be nitrogen enriched (lower POC/PN ratios; 13 Figure 7). POC/PN ratios exceeding 12 consistently have δ CPOC values between −26.76 and −25.42‰, similar to the deepest particulates retrieved from the basin during light cruises (Figures 3c and 6c).

SCRANTON ET AL. 11 of 17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

Figure 9. Bubble plots of the relative abundances of the five types of RuBisCO (a) genes and their (b) transcripts at two depths within the redoxcline during 9 May (“heavy”) and 6 November 2014 (“light”) cruises. While all five RuBisCO gene types and transcripts were present, Type B (form II) known to produce isotopically heavy fixed carbon dominated all gene and transcript libraries. RPKM = reads per kilobase per million mapped reads.

3.5. Metagenome and Metatranscriptome Data An important factor that can control the isotopic composition of POC is the phylogenetic composition of the chemoautotrophic community (Hügler & Sievert, 2011) and their carbon fixation pathways. We constructed metagenomes and metatranscriptomes of microbial assemblages collected from the redox transitional zone at 295 and 314 m in May 2014 (CAR‐212_2) and from 247 and 267 m in November 2014 (CAR216_2; Table S2). Consistent with observations reported in Taylor et al. (2018), genes encoding for the rTCA cycle enzymes (e.g., ATP‐citrate lyase and citryl‐CoA synthetase) used by ε‐proteobacteria and producing isotopi- δ13 − − ‰ cally heavy biomass ( CBIOMASS = 8to 12 ; Campbell et al., 2003; Zbinden et al., 2015) were not detect- able in any of our metagenomes. However, genes for the CBB cycle, present in γ‐proteobacteria, were common. Therefore, we examined the diversity of RuBisCO genes. The RuBisCO genes derived from anoxic and shallow sulfidic depths from both cruises were broadly organized into forms I and II (Figure 8). The form I RuBisCO protein has eight large and eight small subunits and is known to strongly discriminate 13 13 against CO2, yielding light δ CBIOMASS signatures (Robinson et al., 2003). The form II RuBisCO protein is comprised of multiple dimers of a single subunit similar to the large subunit of form I RuBisCO and is 13 13 known to only weakly discriminate against CO2, yielding relatively heavy δ CBIOMASS signatures (Robinson et al., 2003; Tabita, 1999). Within each form, several phylogenetically distinct genes clusters (designated as “types”) were identified (Figure 8). The quantitative importance of each RuBisCO gene type was assessed by comparing the reads per kilobase per million mapped reads (RPKM) in each metagenome and metatranscriptome (Figure 9). Form II, type B RuBisCO gene sequences were far more abundant than any other type in samples from both dates and both depths, followed by type A (Figure 9a). Form I, type E gene sequences were common in all samples at relatively low abundances. Type C (form II) and type D (form I) in the November 2014 (light) metagenomes were more common than in May 2014 (heavy) samples.

SCRANTON ET AL. 12 of 17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

Figure 10. Vertical distributions of (a) dissolved oxygen [O2] and hydrogen sulfide [H2S] (□), (b) combined nitrite − − + + nitrate [NO2 +NO3 =NOx] (o) and ammonium [NH4 ](⋄) concentrations compiled from the redoxcline during “heavy” cruises (black symbols and lines, n = 13; see Table 1) compared to “light” cruises (gray symbols and lines, n = 18). + Group means + 1 SD are presented from CTD O2 sensor and from wet chemistry NOx,NH4 , and H2S analyses of + samples from discrete depths. analysis of variance for O2 compared groups from 70 to 250 m, while those for NOx,NH4 , and H2S compared groups from 250 to 350 m.

Previous observations revealed that GSO dominated the Cariaco Basin's chemoautotrophic layer late (2009– 2016) in the time series, supplanting the dominant ε‐proteobacteria observed between 1996 and 2009 (Lin et al., 2006; Madrid et al., 2001; Rodriguez‐Mora et al., 2015; Suter et al., 2018; Taylor et al., 2018). GSO are known to fix carbon via the CBB cycle, while thioautotrophic ε‐proteobacteria use the rTCA cycle (Hügler & Sievert, 2011). Expression of RuBisCO genes fluctuated more between depths and sampling dates than abundances of the genes in metagenomes. Transcripts of all RuBisCO gene types tended to be more highly represented in the metatranscriptomes of the shallower samples where chemoautotrophy‐supporting 0 − reductants (H2S, S ) and oxidants (O2,NOx ) may coexist (Figures 4c, 4d, and 10b). During the May 2014 0 heavy cruise (CAR‐212), both H2S and S were significantly more abundant at 314 m than they were at 267 m during November 2014 light cruise (CAR‐216; Table S2). We note that NOx,H2S, and sulfur samples were collected 4 days after sampling for metatranscriptomics and thus may not reflect the exact geochemical conditions experienced by microbial communities. Form I RuBisCO genes, the product of which is known to 13 δ13 strongly discriminate against CO2 and to produce light CBIOMASS signatures (Robinson et al., 2003) tended to be more highly expressed in the November 2014 samples (CAR216_2). However, form II 13 RuBisCO type B gene expression (which results in a relatively heavy δ CPOC) dominated the metatranscrip- tomes on both dates. While we have very limited temporal metagenomic coverage, existing data are consis- tent with the hypothesis that 13C‐enriched POC periodically observed across the redoxcline results from carbon fixed predominantly by chemoautotrophs using form II RuBisCO. Chemoautotrophs fixing carbon by form I RuBisCO or other pathways are also likely present in this layer, and their relative contributions 13 to the δ CPOC signature vary with hydrographic conditions. Temporal changes in depth distributions and intensities of DIC assimilation suggest that either the abun- dance or activity of the organisms participating in DIC assimilation and/or the availability of reductants and oxidants were variable. Previous results have shown that chemoautotrophy in the Cariaco appears to

SCRANTON ET AL. 13 of 17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

be primarily fueled by hydrogen sulfide and intermediate oxidation state sulfur species, like thiosulfate and elemental sulfur (Li et al., 2008). DIC assimilation rates are consequently controlled by fluxes of oxidant and reductant to the oxygen/sulfide transition zone (Li et al., 2012). In order to assess whether supply of reactants and hydrographic conditions might promote differences between heavy and light cruises, pooled distributions of chemoautotrophy‐supporting oxidants and reductants during heavy and light cruises were compared (Figure 10). Dissolved oxygen concentrations below 100 m during light cruises were signifi-

cantly higher than during heavy cruises (Figure 10a). Diffusive fluxes of O2 computed from water density profiles and O2 concentration gradients (Figure 10a) revealed that median O2 fluxes between 190 and 250 m during light cruises were about 1.5 times greater than during heavy cruises (p < 0.005, ANOVA). − − Concentrations of the other major oxidants (NO3 +NO2 =NOx) during light cruises also tended to be higher between 250 and 350 m than during heavy cruises (Figure 10b); medians = 0.15 and 0.04 μmol N/kg, respectively (p < 0.05, ANOVA). However, concentrations and diffusive fluxes of the major reductants + (H2S and NH4 ) over this same depth range were statistically indistinguishable (p > 0.05) between heavy and light cruises, so oxidant supply may have been more important than reductant supply in influencing abundances and activities of chemoautotrophic organisms.

4. Conclusions Acknowledgments We thank the captain and crew of the A persistent chemoautotrophy maximum is seen in the Cariaco Basin at the upper boundary of sulfidic B/O Hermano Gines and the staff of conditions. The depth of this feature is similar to the depth at which unusually high values of δ13C have Estación de Investigaciones Marinas, POC Fundación de la Salle de Ciencias been observed in 13 of 31 cruises (~40%) but understanding the microbiology and biogeochemistry that con- Naturales, Margarita Island, Venezuela, trol this feature is hampered by limitations in vertical and temporal sampling resolution as well as analytical fi − − for their eld and laboratory assistance. detection limits. Vertical inventories near, and fluxes of oxidant (O and NO +NO ) to, the We are also indebted to the many stu- 2 3 2 dents, colleagues, and technicians who oxygen/sulfide interface during heavy cruises were lower than during light cruises, which may have favored have participated in this project, in members of the chemoautotrophic assemblage with low carbon isotopic fractionation (εp) factors. Cruises particular, L. Medina Faull for contour δ13 plots, E. Tappa (USC) for POC and with high CPOC values also were more common between 2013 and 2015 when suspended particles below 13 ‐ δ CPOC data measured in Robert the photic zone tended to be nitrogen rich (low POC/PN) compared to later cruises. Within the chemoauto- Thunell's lab, and K. Fanning and K. trophic layer, cruises with nitrogen‐rich particles (POC/PN< 10) were more likely to be 13C‐enriched than Buck and W. Abbott (USF) for nutrient ‐ data. Digna‐Rueda‐Roa, Laura those with nitrogen poor particles, suggesting that these samples were dominated by living cells and fresh 13 Lorenzoni, and Matt Biddle assisted detritus rather than laterally transported or extensively decomposed biogenic debris. The persistent C greatly in getting the data into a format enrichment of POC observed down to 1,300 m during heavy cruises is the first evidence that POC chemoau- ‐ suitable for submission to the BCO fl DMO database. We are also grateful to totrophically produced in the redoxcline actually is transported to the sea oor in this system. two anonymous reviewers for their Metagenomics data from two cruises support prevalence in the chemoautotrophic layer of microorganisms insightful comments. This research was supported by grants from NSF (OCE‐ carrying RuBisCO form II genes. These genes encode for a carbon fixation enzyme that has been shown in 1259110 awarded to M. I. S. and G. T. T.; cultured isolates to discriminate less against heavy isotopes than most other carbon fixation enzymes. ‐ ‐ OCE 1258991 to R. C. T.; OCE 0326268, Metatranscriptomics data confirm that form II RuBisCO genes were more highly expressed during both OCE‐0963028, OCE‐1259043, and OCE‐ 1649626 to F. M. K.; and OCE‐1336082 cruises and at depths where essential reactants coexist. However, our interpretation is based on data from and OCE‐1335436 awarded to V. P. E. cultured microorganisms and whether natural populations fractionate in the same manner is unknown. and G. T. T., respectively), from δ13 Venezuela's FONACIT (2000001702 The patterns in C/N ratios, CPOC, chemoautotrophy, and metagenomics data are all consistent with the and 2011000353 to Y. A.), and a WHOI hypothesis that chemoautotrophic organisms relying on form II RuBisCO to fixCO2 produce isotopically subaward A101259 to M. G. P. heavy POC observed in the redoxcline. Biological and Chemical Oceanography Data Management Office Metadata landing page for the Cariaco Time ser- ies Niskin bottle data is/https://www. References bco‐dmo.org/dataset/3093. For the data Astor, Y. M., Lorenzoni, L. & Scranton, M. I. (Eds.) (2011). Handbook of methods for the analysis of oceanographic parameters at the from our biogeochemistry cruises the Cariaco Time Series Station. Retrieved from http://reef01.marine.usf.edu/sites/default/files/project/cariaco/publications/CARIACO_ BCO‐DMO Metadata landing page is Methods_Manual_v2.pdf https://ww.bco‐dmo.org/dataset/3120 Astor, Y. M., Lorenzoni, L., Thunell, R., Varela, R., Muller‐Karger, F., Troccoli, L., et al. (2013). Interannual variability in sea surface and for the Time series CTD data is temperature and fCO2 changes in the Cariaco Basin. Deep‐Sea Research Part II: Topical Studies in Oceanography, 93,33–43. https://doi. https://www.bco‐dmo.org/dataset/ org/10.1016/j.dsr2.2013.01.002 13 3092. δ CDIC DATA ARE PRESENTED IN Berg, I. A., Kockelkorn, D., Ramos‐Vera, W. H., Say, R. F., Zarzycki, J., Hügler, M., et al. (2010). Autotrophic carbon fixation in archaea. TABLE S1. METAGENOME AND Nature Reviews Microbiology, 8(6), 447–460. https://doi.org/10.1038/nrmicro2365 METATRANSCRIPTOME DATA ARE Bolger, A. M., Lohse, M., & Usadel, B. (2014). Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics, 30(15), AVAILABLE FROM SRA (ACCESSION 2114–2120. https://doi.org/10.1093/bioinformatics/btu170 13 NUMBER PRJNA544741). Δ CPOC DATA Campbell, B. J., Stein, J. L., & Cary, S. C. (2003). Evidence of chemolithoautotrophy in the bacterial community associated with Alvinella ARE AVAILABLE AT HTTPS://DOI.ORG/ pompejana, a hydrothermal vent polychaete. Applied and Environmental Microbiology, 69(9), 5070–5078. https://doi.org/10.1128/ 10.6084/M9.FIGSHARE.8214470.V1. aem.69.9.5070‐5078.2003

SCRANTON ET AL. 14 of 17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

Cernadas‐Martin, S., Suter, E. A., Scranton, M. I., Astor, Y., & Taylor, G. T. (2017). Aerobic and anaerobic ammonium oxidizers in the Cariaco Basin: Distributions of major taxa and nitrogen species across the redoxcline. Aquatic Microbial Ecology, 79,31–48. Cole, J. R., Wang, Q., Cardenas, E., Fish, J., Chai, B., Farris, R. J., et al. (2009). The Ribosomal Database Project: Improved alignments and new tools for rRNA analysis. Nucleic Acids Research, 37(Database issue), D141–D145. https://doi.org/10.1093/nar/gkn879 Deuser, W. G. (1973). Cariaco Trench: Oxidation of organic matter and residence time of anoxic water. Nature, 242(5400), 601–603. https:// doi.org/10.1038/242601b0 Edgcomb, V. P., Taylor, C., Pachiadaki, M. G., Honjo, S., Engstrom, I., & Yakimov, M. (2016). Comparison of Niskin vs. in situ approaches for analysis of gene expression in deep Mediterranean Sea water samples. Deep Sea Research, Part II, 129, 213–222. https://doi.org/ 10.1016/j.dsr2.2014.10.020 Frias‐Lopez, J., Shi, Y., Tyson, G. W., Coleman, M. L., Schuster, S. C., Chisholm, S. W., & Delong, E. F. (2008). Microbial community gene expression in ocean surface waters. Proceedings of the National Academy of Sciences of the United States of America, 105(10), 3805–3810. https://doi.org/10.1073/pnas.0708897105 Fry, B. (2006). Stable Isotope Ecology (p. 308). New York: Springer‐Verlag. https://doi.org/10.1007/0‐387‐33745‐8 Fry, B., Jannasch, H. W., Molyneaux, S. J., Wirsen, C. O., Muramoto, J. A., & King, S. (1991). Stable isotope studies of the carbon, nitrogen and sulfur cycles in the Black Sea and the Cariaco Trench. Deep Sea Research Part A. Oceanographic Research Papers, 38, S1003–S1019. https://doi.org/10.1016/S0198‐0149(10)80021‐4 Fry, B., & Wainright, S. C. (1991). Diatom sources of 13C‐rich carbon in marine food webs. Marine Ecology Progress Series, 76,149–157. https://doi.org/10.3354/meps076149 Fuchsman, C. A., Kirkpatrick, J. B., Brazelton, W. J., Murray, J. W., & Staley, J. T. (2011). Metabolic strategies of freeliving and aggregate‐ associated bacterial communities inferred from biologic and chemical profiles in the Black Sea suboxic zone. FEMS Microbiology Ecology, 78(3), 586–603. https://doi.org/10.1111/j.1574‐6941.2011.01189.x Ganesh, S., Parris, D. J., DeLong, E. F., & Stewart, F. J. (2014). Metagenomic analysis of size‐fractionated picoplankton in a marine oxygen minimum zone. The ISME Journal, 8(1), 187–211. https://doi.org/10.1038/ismej.2013.144 Glaubitz, S., Kiesslich, K., Meeske, C., Labrenz, M., & Jurgens, K. (2013). SUP05 dominates the gammaproteobacterial sulfur oxidizer assemblages in pelagic redoxclines of the central Baltic and Black Seas. Applied and Environmental Microbiology, 79(8), 2767–2776. https://doi.org/10.1128/AEM.03777‐12 Glaubitz, S., Labrenz, M., Jost, G., & Jürgens, K. (2010). Diversity of active chemolithoautotrophic prokaryotes in the sulfidic zone of a Black Sea pelagic redoxcline as determined by rRNA‐based stable isotope probing. FEMS Microbiology Ecology, 74(1), 32–41. https://doi. org/10.1111/j.1574‐6941.2010.00944.x 13 Goericke, R., & Fry, B. (1994). Variations of marine plankton δ C with latitude, temperature, and dissolved CO 2 in the world ocean. Global Biogeochemical Cycles, 8(1), 85–90. https://doi.org/10.1029/93GB03272 Goni, M. A., Aceves, H., Benitez‐Nelson, B., Tappa, E., Thunell, R., Black, D. E., et al. (2009). Oceanographic and climatologic controls on the compositions and fluxes of biogenic materials in the water column and sediments of the Cariaco Basin over the Late Holocene. Deep‐ Sea Research Part I: Oceanographic Research Papers, 56(4), 614–640. https://doi.org/10.1016/j.dsr.2008.11.010 Hoefs, J. (2009). Stable Isotope Geochemistry (6th edition) (pp. 150–151). Berlin: Springer. Hügler, M., & Sievert, S. M. (2011). Beyond the : Autotrophic carbon fixation in the ocean. Annual Review of Marine Science, 3(1), 261–289. https://doi.org/10.1146/annurev‐marine‐120709‐142712 Jørgensen, B. B., Fossing, H., Wirsen, C. O., & Jannasch, H. W. (1991). Sulfide oxidation in the anoxic Black Sea chemocline. Deep‐Sea Research, 38(Suppl. 2), S1083–S1103. https://doi.org/doi.org/10.1016/S0198‐0149(10)80025‐1 Jost, G., Zubkov, M. V., Yakushev, E., Labrenz, M., & Jürgens, K. (2008). High abundance and dark CO 2 fixation of chemolithoautotrophic prokaryotes in anoxic waters of the Baltic Sea. Limnology and Oceanography, 53(1), 14–22. https://doi.org/10.4319/lo.2008.53.1.0014 Kirkpatrick, J. B., Fuchsman, C. A., Yakushev, E., Egorov, A. V., Staley, J. T., & Murray, J. W. (2018). Dark nitrogen fixation: nifH expression in the redoxcline of the Black Sea. Aquatic Microbial Ecology, 82,43–58. Kumar, S., Stecher, G., & Tamura, K. (2016). MEGA7: Molecular Evolutionary Genetics Analysis Version 7.0 for Bigger Datasets. Molecular Biology and Evolution, 33(7), 1870–1874. https://doi.org/10.1093/molbev/msw054 Lavik, G., Stührmann, T., Brüchert, V., Van der Plas, A., Mohrholz, V., Lam, P., et al. (2009). Detoxification of sulphidic African shelf waters by blooming chemolithotrophs. Nature Letters, 457(7229), 581–584. https://doi.org/10.1038/nature07588 Le, S. Q., & Gascuel, O. (2008). An Improved General Amino Acid Replacement Matrix. Molecular Biology and Evolution, 25(7), 1307–1320. https://doi.org/10.1093/molbev/msn067 Li, H., & Durbin, R. (2010). Fast and accurate long‐read alignment with Burrows–Wheeler transform. Bioinformatics, 26(5), 589–595. https://doi.org/10.1093/bioinformatics/btp698 Li, X. N., Taylor, G. T., Astor, Y., & Scranton, M. I. (2008). Relationship of sulfur speciation to hydrographic conditions and chemoauto- trophic production in the Cariaco Basin. Marine Chemistry, 112(1–2), 53–64. https://doi.org/10.1016/j.marchem.2008.06.002 Li, X. N., Taylor, G. T., Astor, Y., Varela, R., & Scranton, M. I. (2012). The conundrum between chemoautotrophic production and reductant and oxidant supply: A case study from the Cariaco Basin. Deep‐Sea Research Part I‐Oceanographic Research Papers, 61,1–10. https://doi. org/10.1016/j.dsr.2011.11.001 Lin, X., Wakeham, S. G., Putnam, I. F., Yrene, M., Scranton, M. I., Chistoserdov, A. Y., & Taylor, G. T. (2006). Comparison of vertical distributions of prokaryotic assemblages in the anoxic Cariaco Basin and Black Sea by use of fluorescence in situ hybridization. Applied and Environmental Microbiology, 72(4), 2679–2690. https://doi.org/10.1128/AEM.72.4.2679 Louca, S., Astor, Y. M., Doebeli, M., Taylor, G. T., & Scranton, M. I. (2019). Microbial metabolite fluxes in a model marine anoxic ecosystem. Geobiology, 17(6), 628–642. https://doi.org/10.1111/gbi.12357 Louca, S., Scranton, M. I., Taylor, G. T., Astor, Y. M., Crowe, S. A., & Doebeli, M. (2019). Circumventing kinetics in biogeochemical models. Proceeding of the National Academy Sciences (USA), 23, 11,329–11,338. https://doi.org/10.1073/pnas.1819883116 Madrid, V. M., Taylor, G. T., Scranton, M. I., & Chistoserdov, A. Y. (2001). Phylogenetic diversity of bacterial and archaeal communities in the anoxic zone of the Cariaco Basin. Applied and Environmental Microbiology, 67(4), 1663–1674. https://doi.org/10.1128/aem.67.4.1663‐ 1674.2001 Montes, E., Altabet, M. A., Muller‐Karger, F. E., Scranton, M. I., Thunell, R. C., Benitez‐Nelson, C., et al. (2013). Biogenic nitrogen gas production at the oxic–anoxic interface in the Cariaco Basin, Venezuela. Biogeosciences, 10, 267–279. Muller‐Karger, F., Astor, Y., Benitez‐Nelson, C., Scranton, M. I., Taylor, G. T., Thunell, R. C., et al. (2019). “Time series composite CTD profiles from R/V Hermano Ginés cruises in the Cariaco Basin from 1995 through 2017 (CARIACO Ocean Time‐Series Program),” 2019‐ 06‐06. https://doi.org/10.1575/1912/bco‐dmo.3092.1, https://hdl.handle.net/1912/24203

SCRANTON ET AL. 15 of 17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

Muller‐Karger, F., Astor, Y., Scranton, M. I., Taylor, G. T., Thunell, R. C., Varela, R., et al. (2019). "Time‐series Niskin‐bottle sample data from R/V Hermano Gines cruises in the Cariaco Basin from 1995 through 2017 (CARIACO Ocean Time‐Series Program)", 2019‐06‐07. https://doi.org/10.1575/1912/bco‐dmo.3093.1, hdl.handle.net/1912/24228 Muller‐Karger, F. E., Astor, Y. M., Benitez‐Nelson, C. R., Buck, K. N., Fanning, K. A., Lorenzoni, L., et al. (2019). The scientific legacy of the CARIACO Ocean Time‐Series Program. Annual Review of Marine Science, 11(1), 413–437. https://doi.org/10.1146/annurev‐marine‐ 010318‐095150 Norland, S. (1993). The relationship between biomass and volume of bacteria. Handbook of Methods in Aquatic Microbial Ecology, 1, 303–308. https://doi.org/10.1177/1350507606060978 O'Leary, M. H. (1988). Carbon isotopes in . Bioscience, 38(5), 328–335. https://doi.org/10.2307/1310735 Peng, Y., Leung, H. C. M., Yiu, S. M., & Chin, F. Y. L. (2010). IDBA—A practical iterative De Bruijn graph De Novo assembler. In B. Berger (Ed.), The 14th Annual International Conference on Research in Computational Molecular Biology (RECOMB 2010) (Vol. 6044, pp. 426–440). Berlin Heidelberg: Springer‐Verlag. Reinfelder, J. R., Kraepiel, A. M. L., & Morel, F. M. M. (2000). Unicellular C4 photosynthesis in a marine diatom. Nature, 407(6807), 996–999. https://doi.org/10.1038/35039612 Robinson, J. J., & Cavanaugh, C. M. (1995). Expression of form I and form II Rubisco in chemoautotrophic symbioses: Implications for the interpretation of stable carbon isotope values. Limnology and Oceanography, 40, 1496–1502. Robinson, J. J., Scott, K. M., Swanson, S. T., O'Leary, M. H., Horken, K., Tabita, F. R., & Cavanaugh, C. M. (2003). Kinetic isotope effect and characterization of form II RubisCO from the chemoautotrophic endosymbionts of the hydrothermal vent tubeworm Riftia pachyptila. Limnology and Oceanography, 48(1), 48–54. https://doi.org/10.4319/lo.2003.48.1.0048 Rodriguez‐Mora, M. J., Scranton, M. I., Taylor, G. T., & Chistoserdov, A. Y. (2015). The dynamics of the bacterial diversity in the redox transition and anoxic zones of the Cariaco Basin assessed by parallel tag sequencing. FEMS Microbiology Ecology, 91(9), 1–13. https://doi. org/10.1093/femsec/fiv088 Schunck, H., Lavik, G., Desai, D. K., Grosskopf, T., Kalvelage, T., Loscher, C. R., et al. (2013). Giant hydrogen sulfide plume in the oxygen minimum zone off Peru supports chemolithoautotrophy. PLoS ONE, 8(8), e68661. https://doi.org/10.1371/journal.pone. 0068661 Scranton, M., Taylor, G., Thunell, R., Benitez‐Nelson, C., Muller‐Karger, F., Fanning, K., et al. (2014). Interannual and subdecadal variability in the nutrient geochemistry of the Cariaco Basin. Oceanography, 27(1), 148–159. https://doi.org/10.5670/oceanog. 2014.18 Scranton, M. I., Taylor, G. T., Muller‐Karger, F., Astor, Y., Varela, R., Fanning, K., et al. (2019). “Biogeochemistry and microbiology from the R/V Hermano Gines cruises in the Cariaco Basin from 1995 to 2015 (CARIACO Ocean Time‐Series Program),” 2019‐06‐07. https:// doi.org/10.1575/1912/bco‐dmo.3120.1, hdl.handle.net/1912/24222 Shah, V., Chang, B. X., & Morris, R. M. (2017). Cultivation of a chemoautotroph from the SUP05 clade of marine bacteria that produces nitrite and consumes ammonium. International Society of Microbial Ecology Journal, 11, 263–271. Stewart, F. J., Ulloa, O., & DeLong, E. F. (2012). Microbial metatranscriptomics in a permanent marine oxygen minimum zone. Environmental Microbiology, 14(1), 23–40. https://doi.org/10.1111/j.1462‐2920.2010.02400.x Suter, E. A., Pachiadaki, M., Taylor, G. T., Astor, Y., & Edgcomb, V. P. (2018). Free‐living chemoautotrophic and particle‐attached het- erotrophic prokaryotes dominate microbial assemblages along a pelagic redox gradient. Environmental Microbiology, 20(2), 693–712. https://doi.org/10.1111/1462‐2920.13997 Swan, B. K., Martinez‐Garcia, M., Preston, C. M., Sczyrba, A., Woyke, T., Lamy, D., et al. (2011). Potential for chemolithoautotrophy among ubiquit ous Bacteria lineages in the dark ocean. Science, 333(6047), 1296–1300. https://doi.org/10.1126/science.1203690 Tabita, F. R. (1999). Microbial ribulose 1,5‐bisphosphate carboxylase/oxygenase: A different perspective. Photosynthesis Research, 60(1), 1–28. https://doi.org/10.1023/A:1006211417981 Taylor, C. D., & Doherty, K. W. (1990). Submersible Incubation Device (SID), autonomous instrumentation for the in situ measurement of primary production and other microbial rate processes. Deep Sea Research Part A. Oceanographic Research Papers, 37(2), 343–358. https://doi.org/10.1016/0198‐0149(90)90132‐F Taylor, C. D., Edgcomb, V. P., Doherty, K. W., Engstrom, I., Shanahan, T., Pachiadaki, M. G., et al. (2015). Fixation filter, device for the rapid in situ preservation of particulate samples. Deep‐Sea Research Part I: Oceanographic Research Papers, 96(January), 69–79. https:// doi.org/10.1016/j.dsr.2014.09.006 Taylor, G. T., Iabichella, M., Ho, T.‐Y., Scranton, M. I., Thunell, R. C., Muller‐Karger, F., & Varela, R. (2001). Chemoautotrophy in the redox transition zone of the Cariaco Basin: A significant midwater source of organic carbon production. Limnology and Oceanography, 46(1), 148–163. https://doi.org/10.4319/lo.2001.46.1.0148 Taylor, G. T., Suter, E. A., Pachiadaki, M. G., Astor, Y., Edgcomb, V. P., & Scranton, M. I. (2018). Temporal shifts in dominant sulfur‐ oxidizing chemoautotrophic populations across the Cariaco Basin's redoxcline. Deep Sea Research Part II: Topical Studies in Oceanography, 156,80–96. https://doi.org/10.1016/j.dsr2.2017.11.016 Thompson, J. D., Higgins, D. G., & Gibson, T. J. (1994). CLUSTAL W: Improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position‐specific gap penalties and weight matrix choice. Nucleic Acids Research, 22(22), 4673–4680. https://doi.org/10.1093/nar/22.22.4673 Tuttle, J. H., & Jannasch, H. W. (1972). Occurrence and types of Thiobacillus‐like bacteria in sea. Limnology and Oceanography, 17(4), 532. Tuttle, J. H., & Jannasch, H. W. (1973). Sulfide and thiosulfate‐oxidizing bacteria in anoxic marine basins. Marine Biology, 20(1), 64–70. https://doi.org/10.1007/bf00387676 van Breugel, Y., Schouten, S., Paetzel, M., Nordeide, R., & Sinninghe Damsté, J. S. (2005). The impact of recycling of organic carbon on the stable carbon isotopic composition of dissolved inorganic carbon in a stratified marine system (Kyllaren fjord, Norway). Organic Geochemistry, 36(8), 1163–1173. https://doi.org/10.1016/J.ORGGEOCHEM.2005.03.003 Wakeham, S. G., Turich, C., Schubotz, F., Podlaska, A., Li, X. N., Varela, R., et al. (2012). Biomarkers, chemistry and microbiology show chemoautotrophy in a multilayer chemocline in the Cariaco Basin. Deep Sea Research Part I: Oceanographic Research Papers, 63, 133–156. https://doi.org/10.1016/j.dsr.2012.01.005 Walsh, D. A., Zaikova, E., Howes, C. G., Song, Y. C., Wright, J. J., Tringe, S. G., et al. (2009). Metagenome of a versatile chemolithoautotroph from expanding oceanic dead zones. Science, 326(5952), 578–582. https://doi.org/10.1126/science.1175309 Woodworth, M., Goñi, M., Tappa, E., Tedesco, K., Thunell, R., Astor, Y., et al. (2004). Oceanographic controls on the carbon isotopic compositions of sinking particles from the Cariaco Basin. Deep‐Sea Research Part I: Oceanographic Research Papers, 51(12), 1955–1974. https://doi.org/10.1016/j.dsr.2004.08.003

SCRANTON ET AL. 16 of 17 Journal of Geophysical Research: Biogeosciences 10.1029/2019JG005276

Yilmaz, A., Çoban‐Yildiz, Y., Telli‐Karakoç, F., & Bologa, A. (2006). Surface and mid‐water sources of organic carbon by photoautotrophic and chemoautotrophic production in the Black Sea. Deep‐Sea Research Part II: Topical Studies in Oceanography, 53(17–19), 1988–2004. https://doi.org/10.1016/j.dsr2.2006.03.015 Zbinden, M., Marqué, L., Gaudron, S. M., Ravaux, J., Léger, N., & Duperron, S. (2015). Epsilonproteobacteria as gill epibionts of the hydrothermal vent gastropod Cyathermia naticoides (North East‐Pacific Rise). Marine Biology, 162(2), 435–448. https://doi.org/10.1007/ s00227‐014‐2591‐7 Zopfi, J., Ferdelman, T. G., Jørgensen, B. B., Teske, A., & Thamdrup, B. (2001). Influence of water column dynamics on sulfide oxidation and other major biogeochemical processes in the chemocline of Mariager Fjord (Denmark). Marine Chemistry, 74(1), 29–51. https://doi. org/10.1016/s0304‐4203(00)00091‐8

SCRANTON ET AL. 17 of 17