ABSTRACT
Large spatial scales and long-term shifts of bacterial community composition (BCC) in the open ocean can often be reliably predicted based on the dynamics of physical-chemical variables. The power of abiotic factors in shaping BCC on shorter time scales in shallow estuarine mixing zones is less clear. We examined the diurnal variation in BCC at different water depths in the spring and fall of 2011 at a station in the Gray's Reef National Marine Sanctuary (GRNMS). This site is located in the transition zone between the estuarine plume and continental shelf waters of the South Atlantic Bight. A total of 234,516 pyrotag sequences of bacterial 16S rRNA genes were recovered; they were taxonomically affiliated with >200 families of 23 bacterial phyla. Nonmetric multidimensional scaling analysis revealed significant differences in BCC between spring and fall samples, likely due to seasonality in the concentrations of dissolved organic carbon and nitrate plus nitrite. Within each diurnal sampling, BCC differed significantly by depth only in the spring and differed significantly between day and night only in the fall. The former variation largely tracked changes in light availability, while the latter was most correlated with concentrations of polyamines and chlorophyll a. Our results suggest that at the GRNMS, a coastal mixing zone, diurnal variation in BCC is attributable to the mixing of local and imported bacterioplankton rather than to bacterial growth in response to environmental changes. Our results also indicate that, like members of the Roseobacter clade, SAR11 bacteria may play an important role in processing dissolved organic material in coastal oceans.
INTRODUCTION
Spatial variability of microbes, i.e., the occurrence of distinct patterns of bacterial community composition (BCC) in geographically isolated habitats or different depth zones, has been well established in marine environments (1–5). Annually recurring patterns and pronounced seasonal variability in BCC have also been studied in a number of marine environments (6–9).
However, BCC variations on shorter time scales, such as diurnal cycles, are relatively understudied, and available reports yield conflicting results. For example, BCC showed little diurnal change in the Western English Channel (10) but varied significantly in coastal California (11) and in the upper mixed layer of the Ligurian Sea (northwest of the Mediterranean Sea) (12). Moreover, some BCC studies are based solely on molecular fingerprinting methods, such as automated ribosomal intergenic spacer analysis (7) and terminal restriction fragment length polymorphism (13), and therefore lack taxonomic resolution. Sequence-based studies are available but often are reported at very broad taxonomic levels (such as phylum/class) (8, 11) or are focused only on a few specific taxa (14). Nonetheless, despite differences in sampling dimensions (spatial or vertical) and analytical methods, microbial ecologists have reached a general consensus that the long-term dynamics of open-ocean BCC are regulated by physical variables, most notably temperature (7, 8). However, whether these environmental factors have similar impacts on BCC in estuarine mixing zones and on short time scales remains unclear.
This study investigated the depth and diurnal dynamics of BCC (i.e., relative abundances of bacterioplankton taxa) at the Gray's Reef National Marine Sanctuary (GRNMS) in two consecutive seasons and examined the potential correlations between BCC and environmental factors. The GRNMS is located ca. 32 km off the coast of Georgia (USA) in the transition zone between the nearshore estuarine plume and the continental shelf waters in the South Atlantic Bight. This coastal site has a long-term data set of physical and water quality variables as a result of the GRNMS monitoring program and the efforts of other researchers (15). The water chemistry at the reef is subject to complex seasonal changes and exhibits intense short-term dynamics due to strong tidal mixing and wind-driven advection (15–17). Past GRNMS resource assessment efforts have documented community structures of algae, coral, sponges, and fish (18), leaving the bacterioplankton unexplored.
MATERIALS AND METHODS
Sample collection and processing.Two sets of water samples were collected on two cruises of the R/V Savannah to the GRNMS (31°24.04′N, 80°51.51′W), one in the spring (20 to 21 April 2011) and one in the fall (5 to 6 October 2011). Water samples were collected every 3 h during a 24-h period (8 or 9 casts in total) using Niskin bottles mounted on a rosette sampling system (Sea-Bird Electronics, Bellevue, WA, USA). Samples taken after sunrise and before the following sunset were labeled as day samples, while samples taken after sunset and before the following sunrise were labeled as night samples. Depth profiles of environmental variables, including temperature (T), salinity (S), and photosynthetically active radiation (PAR), were taken in situ with a conductivity-temperature-depth (CTD) water column profiler (Sea-Bird Electronics, Bellevue, WA, USA) that was mounted on the rosette sampler system. Water samples were collected at three different depths (∼2, 4, and 17 m) in the spring, when a pycnocline was present, and at two depths (∼2 and 17 m) in the fall, when the pycnocline was absent (see Fig. S1 in the supplemental material). The tidal stage at the time of sampling was quantified by transforming the data into tidal angles (19). Briefly, a conversion factor was first calculated by dividing the time interval between two successive high tides by 360°. Tidal angles were then calculated by multiplying this conversion factor by the time interval between a sampling point and the previous high tide. As a result, tidal angles of 0° and 180° corresponded to high and low tides, respectively.
Immediately after collection, 1 liter of water samples was filtered sequentially through 3-μm- and 0.2-μm-pore-size membrane filters (Pall Life Sciences, Ann Arbor, MI, USA). The filtrates were collected in 50-ml sterile conical centrifuge tubes and were stored at −20°C until analysis of dissolved organic carbon (DOC), dissolved nitrogen (DN), nitrate plus nitrite (NOx−), and soluble reactive phosphorus (SRP). Filtrates were also collected in 5-ml amber glass vials and were stored at −80°C for analyses of two labile dissolved organic nitrogen groups: dissolved free amino acids (DFAAs) and polyamines (PAs). Cells that were collected on 0.2-μm-pore-size membrane filters were frozen immediately in liquid nitrogen onboard and were stored at −80°C in the lab until DNA extraction. Particulate material in 500-ml whole-water samples was collected on glass fiber filters (GF/F; Whatman International Ltd., Maidstone, England) for chlorophyll a (Chl a) measurement and were stored in the dark at −20°C before analysis.
Part of the water (2 ml) that passed through the 3-μm-pore-size filters was preserved in 1% (final concentration) freshly prepared paraformaldehyde and was stored at 4°C until cells were enumerated by epifluorescence microscopy.
All glassware and GF/F filters were baked at 500°C for 4 h before use. Triplicate samples were taken for analyses of nutrients and cell counts.
Nutrient analysis.Nutrients were measured according to standard procedures (20). Briefly, DOC and DN concentrations were determined with a total-organic-carbon (TOC)/total-nitrogen (TN) analyzer (TOC-VCPN; Shimadzu Corp., Tokyo, Japan) based on combustion-oxidation/infrared detection and combustion/chemiluminescence detection methods, respectively. Concentrations of NOx− and SRP were determined based on the cadmium reduction method and the molybdenum blue method, respectively, using flow injection protocols (QuikChem 8000 series flow injection analysis [FIA+] system; Lachat Instruments, Loveland, CO, USA).
Concentrations of DFAAs and PAs were measured by using a Prominence 20A high-performance liquid chromatography system (Shimadzu Corp., Tokyo, Japan) equipped with a Phenomenex Gemini-NX C18 column (length, 250 mm; inner diameter [i.d.], 4.6 mm; particle size, 5 μm; Phenomenex, Torrance, CA, USA) and following a protocol developed specifically for seawater samples (21). Chl a was extracted from the GF/F filters with 90% acetone and was measured by spectrophotometry (22).
DNA extraction, PCR, and pyrotag sequencing.DNA was extracted from 0.2-μm-pore-size membrane filters using PowerSoil DNA extraction kits (Mo Bio Laboratories, Inc., Carlsbad, CA, USA). The V4-to-V6 region of 16S rRNA genes was PCR amplified using universal bacterial primers B530F (5′-GT GCC AGC MGC NGG GG-3′) (23) and B1100R (5′-GGG TTN CGN TCG TTG-3′) (24). The forward primers were constructed with an adaptor sequence (CCA TCT CAT CCC TGC GTG TCT CCG ACT CAG) (23) and a 10-bp bar code tag, while the reverse primers were constructed with an adaptor sequence (CCT ATC CCC TGT GTG CCT TGG CAG TCT CAG) only. The PCR program included an initial denaturation at 95°C for 3 min, followed by 30 cycles of 95°C for 30 s, 58°C for 1 min, and 72°C for 1 min, with a final elongation step at 72°C for 5 min. For each sample, 5 separate PCRs (each with 25 μl) were performed; the resulting PCR amplicons (125 μl) were pooled and examined by gel electrophoresis (1% agarose) to verify amplicon length. The amplicons were excised from the gels and were purified first with the QIAquick gel extraction kit (Qiagen, Chatsworth, CA, USA) and then with the Agencourt AMPure XP system (Beckman Coulter, Brea, CA, USA). The quantity of purified PCR amplicons was determined using the Quant-iT PicoGreen dsDNA assay kit (Life Technologies, Carlsbad, CA, USA). Equimolar amounts of PCR amplicons from 13 samples (randomly assigned) were combined and sequenced in one run using a 454 GS Junior system (Roche 454 Life Sciences, Branford, CT, USA) with unidirectional Lib-L chemistry. Three pyrosequencing runs were performed to sequence a total of 39 samples collected in this study.
Taxonomic annotation of 16S rRNA gene pyrotag sequences.Bacterial 16S rRNA gene pyrotag sequences were assigned to their samples of origin based on bar codes. The primer and bar code sequences were then removed from each sequence. Sequences that either had wrong base calls in the primer regions, were shorter than 65 bp, or contained chimeras (as determined by UCHIME) (25) were removed from further analysis. The remaining sequences were clustered into operational taxonomic units (OTUs) by using CD-HIT (26) with a 97% identity cutoff. Singleton OTUs were excluded from the OTU list to prevent overestimation of bacterial diversity (27). The longest sequence within each OTU was selected as a representative and was blasted against the SILVA small-subunit (SSU) database for taxonomic annotation (28). Most sequences were summarized at the family level, except for those affiliated with marine bacterial groups that have no official taxonomic standing. The latter sequences were summarized at the clade level, such as SAR11 and SAR116. For simplicity, this family/clade-level organization of sequences is referred to below as family level. Since the study focused on bacterioplankton, sequences annotated as chloroplasts were excluded from further analyses.
Diversity and statistical analysis.Diversity calculations and statistical analyses were performed using PRIMER, version 5 (Plymouth Marine Laboratory, Plymouth, United Kingdom) (29) unless mentioned otherwise. Family-level diversity indices were calculated using the Shannon index. Family-level rarefaction curves were constructed to infer the library coverage (30) using the vegan package in R (31).
The similarity in BCC (i.e., relative abundances of bacterial taxa) among samples was examined using nonmetric multidimensional scaling (NMDS) analysis, based on a Bray-Curtis matrix that was calculated using normalized and square-root-transformed relative abundances of major bacterial families among samples (29). A hierarchical agglomerative clustering analysis was also performed based on the same matrix using the average group linkage method. The robustness of NDMS grouping patterns was assessed by analysis of similarity (ANOSIM), an analogue of the standard univariate analysis of variance (ANOVA). The ANOSIM index rANOSIM was reported on a scale of 0 to 1. When P was <0.05, the sample groups were reported as either well separated (rANOSIM > 0.75), clearly different but overlapping(0.5 < rANOSIM ≤ 0.75), separated but strongly overlapping (0.25 < rANOSIM ≤ 0.5), or not separable (rANOSIM < 0.25) (29). Similarity percentage (SIMPER) analysis was conducted to identify the contribution of each bacterial family to the observed difference between sample groups.
Redundancy analysis (RDA) (32) was used to explore correlations between changes in BCC and environmental factors, including T, S, PAR, tidal stage (Tide), DOC, DN, NOx−, SRP, DFAAs, PAs, and Chl a, by using the vegan package in the R software package (31). The significance of the correlation values was assessed statistically by a Monte Carlo analysis using 1,000 permutations.
The significance of observed differences in bacterial diversity and composition between sampling depths and times (day or night) was tested using Student's t test (for paired samples), or one-way ANOVA (for multiple samples) within the R software package (33). Differences were deemed significant when P was <0.05. Potential correlations between the relative abundances of individual bacterial taxa and environmental variables were examined by calculating Pearson's product-moment correlation coefficients (r) using the R software package (33). Significant correlations were reported when the P value was <0.05. Bonferroni corrections of P values were used for multiple tests.
Nucleotide sequence accession numbers.The partial 16S rRNA sequences determined in this study were deposited in the NCBI Sequence Read Archive (SRA) under project accession numbers SRR1556928, SRR1602557, and SRR1602558.
RESULTS
General statistics of 16S rRNA gene pyrotag sequences and taxonomic composition of bacterioplankton.A total of 234,516 partial sequences of bacterial 16S rRNA genes with an average read length of 561 bp passed quality control steps. The number of sequences per sample ranged from 1,516 to 17,689 in the spring and from 1,260 to 12,862 in the fall, which corresponded to 170 to 2,859 and 292 to 1,662 OTUs (roughly at the species level), respectively (see Table S1 in the supplemental material). These OTUs were widely distributed across the domain Bacteria, with >200 unique families from 23 phyla in total. Rarefaction curves of sequence libraries based on family-level annotations all reached plateaus (see Fig. S2 in the supplemental material), despite differences in library size (either by sequence or by OTU number). Therefore, the recovered pyrotag sequences were sufficient to represent bacterioplankton diversity in our samples at the family level.
More than 85% of recovered bacterial sequences were affiliated with only 12 families of 5 phyla, namely, the Proteobacteria, Cyanobacteria, Actinobacteria, Deferribacteres, and Bacteroidetes. Each of these 12 families accounted for 2% or more of the total sequences (all spring and fall libraries together) (Fig. 1; Table 1) and was designated a major taxon. Among them, the SAR11 clade (Alphaproteobacteria) and family I Cyanobacteria (93.2% were Synechococcus spp.) were the most abundant, accounting for 30.7% and 13.6% of sequences on average, respectively. Cluster analysis revealed that in both seasons, the patterns of SAR11 distribution differed significantly from those for family I Cyanobacteria and other major taxa (Fig. 1).
Heat map and accompanying cluster analysis of the relative abundances of major GRNMS bacterioplankton taxa at the family level in spring (April 2011) (A) and fall (October 2011) (B) samples. Sample identifiers are based on the sampling season (sp, spring; fa, fall), depth (s, surface; m, mid-depth; b, bottom), and time (in 24-h format). Regular and bold font styles are used to denote day and night samples, respectively.
Average abundances and relative contributions of major taxa to the observed BCC differences among samples from the GRNMS, revealed by SIMPER analysisa
Seasonal variations in BCC and influential environmental factors.Family-level Shannon index (H) values showed that spring communities were more diverse than fall communities (P, <0.05 by t test) (see Table S1 in the supplemental material).
NMDS ordination analysis displayed a clear separation between spring and fall BCCs (Fig. 2). ANOSIM further confirmed the robustness of this separation (rANOSIM = 0.99; P < 0.05) (see Table S2 in the supplemental material). The SAR11 clade and family I Cyanobacteria were the two most abundant taxa in both seasons; the former was overrepresented in spring (average, 37.6% of the sequences) relative to fall (21.5%), and the latter was overrepresented in fall (20.1%) relative to spring (9.6%) (Table 1) (P < 0.05 by t test). SIMPER analysis further revealed that these two taxa contributed the most to the seasonal BCC dissimilarities observed (24.0% and 16.5% of dissimilarity, respectively), followed by Vibrionaceae (11.0% of dissimilarity) and Pseudoalteromonadaceae (9.7% of dissimilarity) (Table 1).
Nonmetric multidimensional scaling ordination based on the relative abundances (expressed as percentages) of major GRNMS bacterioplankton taxa at the family level in spring and fall samples. Sample identifiers are explained in the legend to Fig. 1. Symbols denote different sampling depths in spring (triangles) and fall (squares): filled symbols, surface; shaded symbols, mid-depth; open symbols, bottom.
RDA was used to examine the potential correlation between environmental variables and the seasonality of BCC at the GRNMS (Fig. 3A). The two RDA axes, RDA1 and RDA2, captured 60.7% and 4.3% of the total variance in BCC, respectively. Spring and fall samples were largely separated along RDA1, which broadly extracted the gradients of all environmental variables measured (P < 0.05) except for PAR and Tide, with NOx−, DOC, and T contributing the most.
Bioplot diagrams of RDA of the correlations between changes in BCC and environmental variables at the GRNMS for overall (A), spring (B), and fall (C) samples. Sample identifiers are explained in the legend to Fig. 1. Environmental variables are highlighted. Different shades of gray indicate significantly different BCC sample clusters identified by ANOSIM.
Depth profiles of BCC and influential environmental factors.Bacterial diversity (H values) did not differ significantly at different depths in fall or over all samples (P > 0.05). In the spring, however, bacterial diversity was greater in surface water samples than in either mid-depth or bottom samples (P < 0.05 by t test) (see Table S1 in the supplemental material).
Analysis using NMDS followed by ANOSIM revealed significant differences in BCC between surface samples and mid-depth or bottom samples only in the spring (rANOSIM, 0.52 and 0.77, respectively; P < 0.05), not in the fall (rANOSIM, 0.03; P > 0.05) (Fig. 1 and 2; see also Table S2 in the supplemental material). SIMPER analysis showed that nearly 60% of the difference in BCC in the spring was due to differences in the abundances of SAR11, Rhodobacteraceae, family I Cyanobacteria, and Cytophagia family incertae sedis members (Table 1). Among these taxa, the SAR11 clade and Cytophagia family incertae sedis had higher relative abundances in mid-depth (on average, 42.1% and 6.5% of the sequences, respectively) and bottom (39.5% and 8.4%, respectively) samples than in surface samples (31.6% and 5.3%, respectively) (P < 0.05 by ANOVA). The opposite distribution pattern was observed for Rhodobacteraceae and family I Cyanobacteria, whose relative abundances were greatest in surface water (P < 0.05 by ANOVA).
Consistently, RDA ordination also revealed a clear separation of spring BCCs on the basis of sampling depth (Fig. 3B). This separation was mostly along the RDA1spring axis, which explained 78.2% of the total variance and was significantly correlated with PAR (P < 0.05).
Diurnal dynamics of BCC and influential environmental factors.Shannon index (H) values revealed that in either season, the bacterioplankton diversity of the day samples was not different from that of the night samples (P > 0.05 by t test) (see Table S1 in the supplemental material).
Analyses of BCC using NMDS and ANOSIM, on the other hand, identified significant differences between day and night samples in the fall (rANOSIM = 0.50; P < 0.05) but not in the spring (rANOSIM = 0.00; P > 0.05) (Fig. 1 and 2; see also Table S2 in the supplemental material). SIMPER analysis revealed that >40% of the variance between day and night BCCs in the fall was attributable to shifts in the relative abundances of SAR11 and family I Cyanobacteria members (Table 1). SAR11 members were more abundant during the day (28.9% averaged across all fall samples) than at night (15.5%) (P < 0.05 by t test), while, in contrast, family I Cyanobacteria members were more abundant at night (23.8%) than during the day (15.4%) (P < 0.05 by t test). SIMPER analysis also identified Pseudoalteromonadaceae and Vibrionaceae as important contributors (each accounted for ∼11% of the difference) to the day-night BCC shifts identified in fall samples. However, this result was not supported by the t test, which found that the relative abundances of these taxa in daytime samples were not significantly different from those in nighttime samples in the fall (P > 0.05 by t test).
RDA consistently separated the day and night BCCs in the fall, mostly along the RDA1fall axis (47.9% of the total variance) (Fig. 3C). This separation was significantly correlated with variance in PAR and concentrations of polyamines and Chl a (P < 0.05).
Correlation between major bacterial taxa and environmental factors.As mentioned above, SIMPER analysis followed by ANOVA or t tests revealed that the variations in BCC by time and depth were due largely to changes in the relative abundances of six taxa: the SAR11 clade, family I Cyanobacteria, Cytophagia family incertae sedis, Vibrionaceae, Pseudoalteromonadaceae, and Rhodobacteraceae (Table 1). Pearson product-moment correlation coefficients revealed that these changes were significantly correlated with differences in environmental conditions between spring and fall (Table 2) (P < 0.05 with Bonferroni's correction). The relative abundances of SAR11 and Cytophagia family incertae sedis members were positively correlated with DOC (r = 0.75 and 0.76, respectively) and NOx− (r = 0.52 and 0.74, respectively) and were negatively correlated with T, S, and SRP (r ≤ −0.58). The opposite pattern was found for members of family I Cyanobacteria, Vibrionaceae, and Pseudoalteromonadaceae, whose relative abundances were correlated positively with T, S, and SRP (r ≥ 0.52) but negatively with DOC and NOx− (r ≤ −0.50). The relative abundance of Rhodobacteraceae was not significantly correlated with any environmental variable measured.
Correlations between major bacterial taxa and environmental variables based on Pearson's product-moment correlation coefficient
DISCUSSION
The GRNMS is a live-bottom reef ecosystem located in the transition zone between the coastal/estuarine plume and continental shelf waters in the South Atlantic Bight. In accordance with the complex local hydrology and chemistry, bacterioplankton communities in GRNMS water were highly diverse and showed significant temporal and depth variations in BCC.
Despite sample variability, the two most abundant taxa were consistently the SAR11 clade (Alphaproteobacteria) and family I Cyanobacteria (93.2% were Synechococcus spp.). These two taxa represent numerically and ecologically significant groups of marine heterotrophic and autotrophic bacterioplankton, respectively (34–38). The family Rhodobacteraceae was also identified as a major taxon at the GRNMS, but its relative abundance (7.5% ± 0.4% on average) was much lower than that of the SAR11 clade (21.5% ± 1.7%). The Rhodobacteraceae family contains several genera that are collectively known as the Roseobacter clade. Their relative abundance often exceeds that of members of the SAR11 clade in nutrient-rich coastal regions, likely a result of their stronger ability to process plankton-derived DOC (39, 40). The predominance of SAR11 over Roseobacter clade bacteria observed at the GRNMS has also been found in coastal sites of the Sorcerer II Global Ocean Sampling (41) and in Norwegian coastal waters of the North Sea (42). Moreover, we found a significant correlation between nutrients (both DOC and nitrate concentrations) and the relative abundance of members of the SAR11 clade at the GRNMS (Table 2). This finding suggests that the SAR11 clade may play a more important role in organic substrate turnover in coastal seawaters than previously thought.
The physical-chemical characteristics of seawater at the GRNMS, including T, S, turbidity, and nutrient concentrations, typically show strong temporal and depth variabilities (17; this study). These factors have been repeatedly suggested to be important regulators of BCC in many other marine ecosystems (3, 6, 8, 9). Our data further indicate that the observed temporal (both diurnal and seasonal) and depth variations of BCC at the GRNMS were correlated with mixes of different environmental variables. This suggests that the impacts of individual environmental factors on BCC differ among temporal and spatial scales.
BCC seasonality overwhelmed variations in BCC on either diurnal or depth scales (Fig. 2 and 3). This pattern may be ascribed to the overriding influence of environmental variables in defining the seasonal or long-term BCC dynamics at the GRNMS, as found at other coastal sites (7, 9). Furthermore, we found that variables (i.e., NOx−, DOC, and T) explaining the seasonality of BCC were not correlated with changes in BCC on shorter time scales (day versus night within 24 h) or among depths. Located in a coastal mixing zone, seawater at the GRNMS actively exchanges with ambient environments through tidal advection and turbulent mixing on a daily basis (15). The residence time of GRNMS water has been estimated at 0.2 to 1.2 days (15), which may be too short to allow detectable growth of bacterial taxa. Therefore, the diurnal variation identified among our fall samples (October 2011), when the water column is well mixed, may be due more to the mixing of “new” bacterioplankton assemblages with local taxa than to differential growth of local bacterial populations in response to changes in environmental conditions.
BCC variation with depth was found in the spring (April 2011) (Fig. 1), when the water column was stratified (see Fig. S1 in the supplemental material). This indicates that the density difference between waters above and below the pycnocline has created a barrier to vertical mixing, which allowed the BCC at different depths to diverge along the PAR gradient. The GRNMS is known for its productive benthic community and substantial benthic nutrient fluxes (15, 16). However, in agreement with a previous study (15), no significant vertical gradients of nutrient concentrations were observed in this study, likely because of the relatively large standing stock of nutrients in the water column (15) (see Table S3 in the supplemental material). Therefore, the observed depth variations in BCC with depth are more likely due to the difference between the contents of dissolved organic and inorganic nutrients across the pycnocline. Further studies on the correlation between nutrient content and BCC in the GRNMS are needed to test this hypothesis.
Conclusion.The bacterioplankton community at the GRNMS consisted of a diverse group of taxa that are typically found in coastal marine environments, including Proteobacteria, Cyanobacteria, Actinobacteria, Deferribacteres, and Bacteroidetes. The BCC at the GRNMS showed strong seasonal variations and was correlated with environmental factors that regulate the growth of bacterioplankton, such as T, S, and nutrient concentrations. Diurnal BCC variations at the GRNMS, on the other hand, are suggested to be more impacted by water mixing events (tides and vertical advection). The differences in BCC at different depths likely result from poor vertical mixing and different contents of nutrients between the surface and bottom waters.
ACKNOWLEDGMENTS
We thank the staff working on the R/V Savannah for assisting in sample collection and providing CTD data. We thank C. Clevinger, M. Kelly, and H. Bui for technical help. We also thank two anonymous reviewers for providing comments that helped us improve this paper.
This study was supported by National Science Foundation grants (OCE1029607 to X.M. and OCE 1029742 to J.T.H.) and Kent State University.
FOOTNOTES
- Received 26 August 2014.
- Accepted 14 November 2014.
- Accepted manuscript posted online 21 November 2014.
Supplemental material for this article may be found at http://dx.doi.org/10.1128/AEM.02802-14.
- Copyright © 2015, American Society for Microbiology. All Rights Reserved.
