Solute samples were filtered using 0.4 µm pore size cellulose acetate syringe filters. Solutes measured were ammonium (NH4+), soluble reactive phosphorus (SRP), total dissolved sulfide (DS = [S2-] + [HS-] + [H2S]), nitrite (NO2-), nitrate (NO3-), and dissolved manganese (DM). Concentrations of NH4+, SRP, DS, and DM were analyzed colorimetrically (Brewer and Spencer 1971; Parsons et al. 1984). Concentrations of NO2- and NO3- were determined by segmented flow analysis after cadmium reduction (Lane et al. 2000).

Dissolved inorganic carbon (DIC = [H2CO3] + [HCO3-] + [CO32-]) was measured with an infrared-based analyzer (Apollo SciTech; Cai and Wang 1998). Standard solutions of sodium carbonate were prepared using distilled water over a range from 1000 to 2000 µmol C L-1 and preserved with mercuric chloride. The standard solution was transferred to autoclaved borosilicate glass tubes fitted with a septum. Standardization was repeated several times prior to analyzing the first sample of each cruise and daily during the transect survey as described in Goyet and Hacker (1992). The standard’s precision was 0.04-0.12 % depending on sampling volume, with higher precision (0.04-0.06 %) when 0.75 ml of sample was analyzed. The calibration was also checked twice with certified reference material (CRM; 2064.11 µmol C L-1; A. Dickson, Scripps Institution of Oceanography). A comparison of standard solution and CRM resulted in a difference of <2 µmol C L-1 (<0.1 %).

Dissolved oxygen to argon ratios were measured with a membrane inlet mass spectrometer (Kana et al. 1994) yielding a precision of <0.05 %. Thermostat temperature-controlled (±0.02 °C of incubation temperature) deionized water was used for standardization and gas ratios were converted to gas concentrations using temperature and salinity-based argon concentrations (Colt 1984).

Cellular CO2 fixation was estimated using radiolabeled NaH14CO3 (14C, PerkinElmer) as described in García-Cantizano et al. (2005). A working 14C stock was prepared using filtered (0.2 µm polycarbonate filter) anoxic seawater before the dilution of 14C to a final concentration of 5 µCi per sample in an anoxic bag. Water samples were collected in triplicate 12 ml vials with silicon septum caps from the top third to sixth depths for photosynthetic fixation experiment, including oxygenic (OxyPP) and anoxygenic (AnoxyPP) photoautotrophic production. Similarly, water samples were collected from the top third to eighth depths for chemolithoautotrophic (DarkPP) experiment. The 14C stock was added into each vial through silicon septa with a gas-tight Hamilton syringe in an anoxic bag. Total four treatments were prepared according to Pedrós-Alió et al. (1993) and incubated for 4 h in an anoxic bag in a temperature-controlled environmental chamber (Thermmax Scientific Products) or an on-deck incubator. The first set of samples was covered with neutral-density screens allowing passage of four photosynthetically active radiation (PAR) levels (6.1, 3.7, 2.2, and 1.3 % of surface irradiance). The second set of triplicate samples was incubated under the same PAR with an addition of 3-(3’, 4’-dichlorophenil)-1,1’-dimethyl urea (DCMU) dissolved in ethanol at a final concentration of 2 µmol L-1 to inhibit photosystem II reaction. We incubated an additional set of samples with the same amount of ethanol in DCMU to check the effect of ethanol. The third set of samples was incubated in dark. Finally, the last set of samples was incubated with formaldehyde (4%, volume:volume) to correct for abiotic reactions. At the end of incubation, samples were immediately filtered onto 0.2 µm polycarbonate membrane filters (Millipore) and dried 24 h. Then, filters were fumed with hydrogen chloride for 30 min and immersed in scintillation cocktail (PerkinElmer) in scintillation vials. Radioactivity, expressed as disintegration per minute (DPM), was counted in a liquid scintillation analyzer (Tri-Carb 3100TR, Packard). The DPM assimilation by OxyPP were calculated by subtracting DPM in illuminated vials with DCMU from DPM in illuminated vials. The DPM assimilation by AnoxyPP was calculated by subtracting DPM in dark vials from DPM in illuminated vials with DCMU. The DPM assimilation by DarkPP were calculated by subtracting DPM in killed samples from DPM in dark vials.

Prokaryotic cells were enumerated by flow cytometry after glutaraldehyde fixed samples were stained with a 1:2000 dilution of Sybr Green overnight at 4ºC in the dark. Samples were diluted 10-fold (lower pycnocline and below) or 20-fold (upper pycnocline and surface) with 0.2µm filtered water, and 20µl bead stock (Becton Dickinson Trucount Controls, Becton Dickinson, Franklin Lakes, New Jersey) was added as an internal reference. Enumeration was performed using a FACSCalibur flow cytometer (Becton Dickinson) using green fluorescence as an incident trigger, with a target range for sample event rates at 100-1000 particles s-1 to prevent incident coincidence. Data was analyzed using CellQuest Pro version 4.0.2 (Becton Dickinson), and populations were gated based upon a forward angle light scatter (FSC), side scatter (SSC, 488 + 5 nm), green fluorescence (FL1, 530 + 15 nm), orange fluorescence (FL2, 585 + 21 nm), and red fluorescence (FL3, >650 nm). FL1 is the fluorescence due to nucleic acids stained with SYBR GreenI, FL2 is the auto-fluorescence of phycoerythrin, and FL3 is the auto-fluorescence of chlorophyll a pigments contained in cells (Troussellier et al., 1995).

Respiration rates were measured with dark bottle incubations. Initial samples were preserved with 2 µl per ml of 50 % saturated mercuric chloride in either 12 ml vials with silicon septum caps or 7 ml test tubes with ground-glass stoppers. Simultaneously, incubations were conducted in triplicate 60 ml biochemical oxygen demand (BOD) bottles covered with opaque plastic bags. Bottles were incubated in either a thermostated bath or a temperature-controlled environmental chamber (±1°C of in situ water temperatures) for 24 h. The linear increase of DIC was examined with water samples collected in different oxic conditions (e.g., r2 ≥ 0.97, p < 0.001). At the end of incubation, samples were siphoned from the BOD bottles to the vials and preserved with mercuric chloride, submerged in water at below-ambient temperatures, and analyzed in <1 wk. Respiration rates were estimated as differences in DIC and dissolved oxygen concentrations between initial and final samples divided by incubation times.

Heterotrophic bacterial production was measured by determining the rate of 3H-leucine incorporation during 1-hour incubations. All incubations were conducted in an anaerobic chamber (Coy) in sealed plastic syringes with the headspace removed. All plastic was stored in an oxygen-free environment 24 hours prior to use. Carbon production was calculated based on leucine incorporation using a ratio of cellular carbon to protein of 0.86, a fraction of leucine in protein of 0.073, and an intracellular leucine isotope dilution of two (Kirchman 1993).


For DNA extraction, each filter was removed from the Sterivex casing by cracking the housing with pliers, sliced on a sterile cutting board, placed in a 2ml microcentrifuge tube, along with the extraction buffer (DEB) in which it had been stored. Samples were extracted using a phenol-chloroform extraction method adapted from Zhou et al. (1996) and Crump et al. (2003). DNA from sediment samples was extracted with the MoBio Power Soil Kit (MoBio Laboratories, Solana Beach, CA, USA).

RNA extractions for 16S amplicon sequencing occurred in an RNA hood to prevent contamination with DNA, and all instruments and surfaces were pretreated with RNase Zap (Ambion) to remove RNase from the area. RNA later was pushed out of the filter with a sterile syringe, and the sample was then rinsed with a sterile solution of 1% Phosphate Buffered Saline (PBS). Each filter casing was cracked and sliced as above, and extracted using the RNeasy Mini Kit (Qiagen) with the Supplementary Protocol for purification of total RNA from bacteria, adapted for extraction from filters as follows: slices of filter were placed in a 2ml microcentrifuge tube with ~100mg of acid-washed glass beads and 1400 µl RLT buffer to cover the surface of the filter, supernatant liquid after bead beating was transferred into two tubes instead of one to accommodate the extra volume, and sample was recombined in the spin column steps. DNA was removed from the RNA extraction using a Turbo-DNase kit (Ambion), followed by Acid -Phenol Chloroform extraction. cDNA was generated using the Superscript II Reverse Transcriptase kit (Invitrogen) with random hexamer primers.

Extracted DNA and RNA was PCR-amplified using primers targeting bacterial 16S ribosomal RNA genes. Each sample was assigned a uniquely barcoded reverse primer and amplified in four replicate 20 μl reactions (Hamady et al. 2008). Primers used for amplification were bacteria-specific primers focusing on the V1-V2 regions, 27F with 454B FLX linker (GCCTTGCCAGCCCGCTCAG TC AGRGTTTGATYMTGGCTCAG) and 338R with 454A linker and unique 8 base pair barcode, denoted by N in primer sequence (GCCTCCCTCGCGCCATCAG NNNNNNN CA TGCWGCCWCCCGTAGGWGT) (Modified from Hamady et al. 2008). Replicate amplicons were combined, quantified, pooled in normalized amounts, purified with MoBio Ultraclean PCR Cleanup Kits (MoBio Laboratories, Solana Beach, CA, USA), and pyrosequenced on a Roche-454 FLX Pyrosequencer at Engencore at the University of South Carolina using titanium chemistry (

RNA samples for metatranscriptomics were retrieved from -80°C storage and extracted using previously described methods (Hewson et al. 2009). For 2011 samples (but not 2010 samples), filters were spiked with 25 ng of internal standard (linearized pGEM-Z transcript [Gifford et al, 2011]) to account for extraction efficiency variation between samples. Briefly, tubes containing filters were amended with 9 ml ZR lysis buffer (Zymo) and 100 µl sterile glass beads, bead beaten (Biospec instruments) for 2 min, and placed on ice, and filters were manually masticated with 10-ml pipettes. Tubes were centrifuged at 3,000 x g for 5 min to remove filter debris before the supernatant was transferred to a clean 15-ml tube. All 8.5 ml of supernatant was run through Zymo Spin Columns II (with repeated centrifugation at 10,000 x g for 30 s and discarding of flowthrough). The filter columns were washed with Zymo wash buffer and then RNA eluted in two rinses of 25 μl, 45°C nuclease-free water. Total RNA extracts were then subjected to treatment with the DNA-free RNA kit (Zymo Research), following the manufacturer’s recommendations. rRNAs were depleted using a previously validated approach (35, 36) employing enzymatic terminator exonuclease degradation (mRNA-ONLY; Epicentre) and bead hybridization capture (MicrobExpress; Ambion). Finally, RNA, which was rRNA depleted, was subject to amplification using the Message-AmpII-Bacteria (Ambion) kit, which employs in vitro transcription to linearly amplify RNAs from nanogram quantities to microgram quantities.

Our treatment of amplified RNA (aRNA) differed from 2010 to 2011 due to improvements in DNA sequencing and method development (Illumina 2 x 200 bp RNAseq). In 2010, aRNA was first converted to double-stranded cDNA using the Superscript III and RNase-dependent DNA polymerase I approach outlined by Hewson et al. (2010). A total of 1 μg double-stranded cDNA from 2010 samples was submitted for 100-bp paired-end Illumina Sequencing (HiSeq 200) at the Cornell University Core Laboratory Center. In 2011, aRNA (~20 μg) was submitted to the Columbia Genome Center laboratory for Illumina RNASeq, employing 100-bp paired-end sequencing.

Our annotation of metatranscriptomes from 2010 included assembly of libraries using the CLC Genomics Workbench 4.0, and annotation of contigs by BLASTx against the RefSeq and SEED databases (MG-RAST). The enormous amount of data generated by HiSeq-Illumina prevented (at the point of metatranscriptome annotation in 2012 and 2013) annotation of these data sets using approaches used previously in the Hewson Lab based on pyrosequencing, for example BLAST-ing files on our internal server, at CAMERA, or annotation of individual paired-end reads using MG-RAST. To account for this, we annotated the assembled sequence data from 2010 and 2011, then recruited library reads to contigs that generated the phylogenetic and functional annotations. This approach allowed quantitative interpretation of metatranscriptomic annotations, with greater confidence than permitted by reads alone. Development of this necessitated re-tooling our handling of large data sets, including the naming convention and storage pedagogy of linked files. Overall, annotation of 2010 and 2011 samples was successful via this approach.

An additional targeted approach used for both data sets was the curation of a respiratory gene database, which includes genes distinct to the expected cascade in electron acceptor use (oxygen → sulfur), and comparison of the library reads against this respiratory gene database. A challenge with formation of this dataset has been identification of genes distinct to each process. For example, genes involved solely in oxic metabolism are not well known compared to those that have flexible metabolic use. In this case, our efforts targeted low-oxygen and high-oxygen cytochromes, instead of solely oxic and anoxic genes. Some pathways have limited description of gene utilization (e.g., Mn reduction), while others (e.g., denitrification) are well described as functional markers of denitrification. Another concern is that some organisms lack the complete pathway necessary for electron acceptor usage (e.g., denitrification). We worked with a specialist in denitrification gene expression, Jim Shapleigh, to ameliorate spurious annotations and to refine the respiratory gene database.