Marine archaeal dynamics and interactions with the microbial community over 5 years from surface to seafloor

Abstract

Marine archaea are critical contributors to global carbon and nitrogen redox cycles, but their temporal variability and microbial associations across the water column are poorly known. We evaluated seasonal variability of free living (0.2–1 μm size fraction) Thaumarchaea Marine Group I (MGI) and Euryarchaea Marine Group II (MGII) communities and their associations with the microbial community from surface to seafloor (890 m) over 5 years by 16S rRNA V4-V5 gene sequencing. MGI and MGII communities demonstrated distinct compositions at different depths, and seasonality at all depths. Microbial association networks at 150 m, 500 m and 890 m, revealed diverse assemblages of MGI (presumed ammonia oxidizers) and Nitrospina taxa (presumed dominant nitrite oxidizers, completing the nitrification process), suggesting distinct MGI-Nitrospina OTUs are responsible for nitrification at different depths and seasons, and depth- related and seasonal variability in nitrification could be affected by alternating MGI-Nitrospina assemblages. MGII taxa also showed distinct correlations to possibly heterotrophic bacteria, most commonly to members of Marine Group A, Chloroflexi, Marine Group B, and SAR86. Thus, both MGI and MGII likely have dynamic associations with bacteria based on similarities in activity or other interactions that select for distinct microbial assemblages over time. The importance of MGII taxa as members of the heterotrophic community previously reported for photic zone appears to apply throughout the water column.

Introduction

Several studies have shown Thaumarchaea Marine Group I (MGI) and Euryarchaea Marine Group II (MGII), to be numerically important marine microorganisms (Massana et al., 1998; Murray et al., 1999; Karner et al., 2001; Herndl et al., 2005; Teira et al., 2006b; Amano-Sato et al., 2013; Needham and Fuhrman, 2016). Additionally, the ability of all cultured MGI to oxidize ammonia to nitrite, and the global abundance of related archaeal ammonia monooxygenase genes has established the importance of these archaea in global biogeochemical cycles (Francis et al., 2005; Könneke et al., 2005; Beman et al., 2008; Galand et al., 2009; Newell et al., 2011; Beman et al., 2012; Tolar et al., 2013; Qin et al., 2014; Santoro et al., 2015). Although studies have indicated surface MGII taxa are heterotrophic and have demonstrated particle attachment and dissolved protein utilization, less is known about the MGII communities below the photic zone (Alderkamp et al., 2006; Iverson et al., 2012; Martin-Cuadrado et al., 2015; Orsi et al., 2015, 2016).

Time-series studies of archaeal communities have demonstrated temporal and seasonal variability at some sites (Massana et al., 1998; Murray et al., 1998, 1999; Pernthaler et al., 2002; Wuchter et al., 2006; Levipan et al., 2007; Beman et al., 2010; Galand et al., 2010; Beman et al., 2011; Robidart et al., 2012). Until recently, studies tended to use methods with relatively low phylogenetic resolution (e.g. DNA fingerprinting techniques, oligonucleotide probes, and quantitative PCR) that may not accurately represent the amount of diversity within an archaeal phylogenetic group, limiting our understanding of temporal changes in archaeal community structure. More recently, Hugoni et al., 2013 used pyrosequencing of 16S rDNA and rRNA to demonstrate dynamics of abundant and rare archaeal taxa, as well as changes in active community members over 3 years in the surface waters (3 m) of the Bay of Banyuls-sur-Mer in France. Although this study provided thoughtful insights into the dynamics of archaeal communities in surface waters, it does not provide information on the archaeal dynamics below the euphotic zone. Studies at the San Pedro Ocean Time-series (SPOT) station, ~10 km off the coast of Los Angeles, California, have shown that microbial communities, as well as the associations between the microbial members, have distinct temporal patterns at different depths (Beman et al., 2010, 2011; Steele et al., 2011; Chow et al., 2013; Cram et al., 2015). However, these were limited to analysis of archaeal communities through qPCR of certain 16S rRNA and Thaumarchaeal ammonia monooxygenase (amoA) genes. Although this has provided an understanding of how MGI and MGII gene abundances change over time and depth, qPCR analyses cannot provide a comprehensive analysis of changes within archaeal communities.

This study evaluates the long-term (~5 years, sampled approx. monthly) temporal changes in archaeal community at SPOT across the water column (0–890 m, encompassing euphotic and aphotic zones) by 16S rRNA gene analysis via ‘universal’ primers that also amplify bacteria. This allows comparisons of the temporal variability of the MGI and MGII in the context of the entire prokaryotic community. Concurrent analysis of the bacterial community and environmental measurements can provide greater insight into the controls on archaeal diversity and abundance across the water column, and how those relationships change over time and depth. We found many strong archaeal-bacterial associations, some apparently connected with nitrification, and others associated with similarities in metabolic activity; however, since little is known about the activity of the majority of marine microbes, most associations are of unknown function.

Materials and methods

Sample collection and DNA extraction

Seawater samples were collected at the USC Microbial Observatory at the San Pedro Ocean Time-series (SPOT) station in the San Pedro Channel (33.55°N, 118.4°W). Samples of 10 L were collected at 5 m (or by bucket at 0 m when 5 m samples were not obtainable) and deep chlorophyll maximum layer (DCM) and 20 L at 150 m, 500 m, and 890 m as previously described (Fuhrman et al., 2006; Beman et al., 2011; Chow et al., 2013; Cram et al., 2015). Samples analyzed in this study were taken at a nearly monthly basis when possible between February 2009–December 2013 resulting in samples from 54 dates at 5 m, 48 dates from the DCM and 890 m, 49 dates from 150 m, and 46 dates from 500 m (Supplementary Table S1).

Seawater was filtered sequentially through an 80 μm nylon mesh, 1μm glass fiber A/E filter (Pall, Port Washington, NY), and 0.22 μm Durapore filters (ED Millipore, Billerica, MA); only DNA from the 0.22 μm filter was analyzed. DNA was extracted by SDS lysis and purified by phenol-chloroform as previously described (Fuhrman et al., 1988).

Environmental measurements

An in situ sensor (Sea-Bird Electronics, Bellevue, WA, USA) measured depth, temperature, and salinity and an in situ oxygen electrode measured oxygen concentration (Sea-Bird, model 13; note the electrode oxygen concentrations were linearly related to Winkler values at other times in this time series, R²=0.93, Cram et al., 2015). The mixed layer depth (MLD) was calculated from the derived density values (SigmaTheta) as the depth at which SigmaTheta was 0.125 kg/m³ greater than at 10 m (Levitus, 1983).

Nitrite (NO₂⁻), nitrate (NO₃⁻), and phosphate (PO₄³⁻) were measured by colorimetric methods (Parsons et al., 1984), from samples stored at −20 °C. Total prokaryotic cell abundance and viral particle abundance was measured by epi-fluorescence microscopy with SYBR Green (Noble and Fuhrman, 1998; Patel et al., 2007). Bacterial productivity was measured by [³H]thymidine (Thy) for DNA synthesis and [³H]leucine (Leu) incorporation for protein synthesis (Fuhrman et al., 2006); both methods were used at 5 m, DCM, and 150 m, and as leucine is more sensitive and may be incorporated by more organisms, it was used for all 5 depths. Cell turnover time (days) calculated as [cell abundance]/ (cell production estimated by [³H]leucine or [³H]thymidine incorporation).

Satellite and meteorological data were gathered and processed as described previously (Cram et al., 2015; see also Supplementary Table S2). Additionally, absorbance due to detritus and gelbstoff (light absorbing part of dissolved organic matter, ABSDetGel) and absorbance due to phytoplankton (ABSPhyto) at 443nm was downloaded from the ocean color data site (Maritorena et al., 2002; oceancolor.gsfc.nasa.gov) and a 3x3 grid around SPOT was extracted with the SeaDAS program (Fu et al., 1998). See Supplementary Information for measurements and abbreviations used in this study (Supplementary Table S2).

Sequencing and data processing

The 16S rRNA gene was partially amplified and sequenced as described previously, using universal V4-V5 primers 515F-C (5′- GTGCCAGCMGCCGCGGTAA) and 926R (5′- CCGYCAATTYMTTTRAGTTT) (Parada et al., 2015). This study does not use the modified 515F-Y primer described in Parada et al., 2015 as the majority of these samples had been sequenced prior to development of the primer modification. Even though using the altered 515F primer appears to slightly increase the relative abundance of the MGI taxa, the modification does not alter the archaeal community composition of samples at SPOT, so we decided to proceed without the modified primer. Extensive clade-by-clade coverage of these primers are reported by Walters et al., 2015. Further details on primer coverage and construction of custom primers are given in Supplementary Information. Cycling conditions included a 3-min heating step at 95 °C followed by 25 cycles of 95 °C for 45s, 50 °C for 45s, 68 °C for 90s, and a final extension at 68 °C for 5-min. DNA was sequenced on a MiSeq (Illumina) using the 2 × 250 bp or 2 × 300 bp paired-end chemistry. Technical replicates of several samples (marked with an asterisk in Supplementary Table S1) were also sequenced, where DNA from the same sample extract was PCR amplified and sequenced independently. Additionally, blanks (no DNA template) and mock communities (as described in Parada et al., 2015) were also sequenced to evaluate contamination and error rates per sequencing run. Briefly, paired-end sequences were merged with USEARCH v7 (Edgar, 2010). Merged sequences were demultiplexed and quality filtered at a minimum quality score of 33 across a 50 bp sliding window in QIIME (Caporaso et al., 2010). Additional details are given in the Supplementary Information. Sequences were submitted to the EMBL database under bioproject PRJEB12267. Representative OTU sequences and a legend of identifiers for each OTU used in this study are available via FigShare (https://dx.doi.org/10.6084/m9.figshare.3474074.v1)

Sequences were further processed in mothur v1.34.4, generally following the MiSeq standard operating procedure, first accessed on May 14, 2015 (Schloss et al., 2009; Kozich et al., 2013). Chimeras were detected using UCHIME (Edgar et al., 2011) then removed. Remaining sequences were clustered de novo at 99% similarity using the average neighbor algorithm after pre-clustering sequences at a 3-base similarity to reduce the influence of error sequences as verified by examining mock community error rates (Schloss et al., 2009; Kozich et al., 2013). Taxonomy was assigned using the default classifier in mothur and 80% confidence cutoff against the Silva v119 reference database (Wang et al., 2007; Quast et al., 2013). The representative sequence for OTUs classified as ‘Chloroplast’ were also compared to the GenBank (Benson et al., 2005) non-redundant nucleotide database with blastn to obtain further identification when available. Operational taxonomic units (OTUs) with less than 6 total sequences across all samples were discarded, this removed extremely rare OTUs only. We chose this cutoff as a compromise to reduce error OTUs while still retaining some extremely rare real OTUs. All blanks and mock communities were then removed prior to further analyses, as well as SPOT samples with less than 10,000 sequences (only one sample was discarded with this criterion, final range 10,740-191,281, average=46953.9, median=44005.5, n=316). Relative differences in sequence number between replicates did not appear to affect community distributions (see Supplementary Information). The relative abundance of each OTU was calculated as the percent proportion of all sequences in a sample (% Relative Abundance). To aid in the interpretation of the data, the relative abundances of technical replicates were averaged and used in all figures and analyses.

Statistical analyses

Canonical Correspondence Analysis (CCA) ordination plots were used to evaluate differences in MGI or MGII communities at all depths using the ‘cca’ and ‘ordistep’ functions of the ‘vegan’ package in R (R Core Team, 2015; Oksanen et al., 2016). OTU abundances were normalized and the log of each abundance value was taken to reduce the effects of outliers. Environmental variables (from Supplementary Table S2) were included in the CCA analyses to identify parameters that may shape the MGI and MGII communities. Additional details are provided in the Supplementary Information. Environmental measurements were transformed to reduce the effect of unit differences (Chow et al., 2014). MGI and MGII community variability was also evaluated by non-metric multidimensional scaling (NMDS, Supplementary Information) to examine the distribution of communities without constraints.

Differences in community composition between samples were evaluated separately for MGI and MGII communities, by calculating the Bray-Curtis dissimilarity between pairs of samples in QIIME or with the ‘vegan’ package in R (Bray and Curtis, 1957; Caporaso et al., 2010; R Core Team, 2015).

Seasonality was determined as statistically significant correlations (5000 permutations) between the Bray-Curtis dissimilarities of MGI or MGII communities and the difference (e.g., distance) in days between each sample by Mantel Tests (calculated similarly as in Cram et al., 2015; see also Supplementary Information). The same method was used to determine seasonality of picophytoplankton communities (comprised of all chloroplast and cyanobacterial OTUs in the 0.2–1μm size fraction). Mantel tests were performed with the ‘mantel’ function in the ‘vegan’ (Oksanen et al., 2016) package in R (R Core Team, 2015) using the spearman correlation method. Mantel tests on Bray-Curtis dissimilaritiy matrices of MGI or MGII and the picophytoplankton or Nitrospina communities were also used to identify correlations between communities.

Mann-Whitney tests were used to determine if mean Bray-Curtis dissimilarities between communities sampled at opposite times of the year were greater than dissimilarities between communities from similar times of the year. This method evaluates if communities from any January month are more similar to each other than to communities from any July month, regardless of which year the community was sampled. Mann-Whitney tests were performed with the ‘wilcox.test’ function in the ‘stats’ package in R (R Core Team, 2015).

Bray-Curtis similarity values between all pairs of MGI or MGII communities at each depth were plotted against time-lag (months) as box and whisker plots using the ‘boxplot’ function in R (R Core Team, 2015). Mann-Whitney tests were used to evaluate statistical significance of observed troughs and peaks in similarity between samples 6 months or 12 months apart (on untransformed time lags).

Additional details of statistical analyses are provided in the Supplementary Information. Examples of all code used in R are provided through FigShare (https://doi.org/10.6084/m9.figshare.4668169.v2).

Correlation networks were calculated with local similarity (LS) analysis (eLSA program) and visualized in Cytoscape v.2.8.3 (Cline et al., 2007; Xia et al., 2011). As previously described, we determined LS correlations (proportional to Pearson’s correlation coefficients) between variables using a linear interpolation for missing values and a delay up to 1 month (Ruan et al., 2006; Steele et al., 2011; Xia et al., 2011; Chow et al., 2014). Only OTUs or environmental variables occurring and measured in >37 months at a given depth were included in the LSA calculations. P-values were calculated through theoretical approximation followed by 2000 permutation tests (Xia et al., 2013; Chow et al., 2014; Xia et al., 2015). Only correlations with LS values >|0.6| with associated P-values<0.01, q-values (false discovery rate; (Storey, 2002)) <0.01 and alignment length >48 were visualized.

Results

Community composition across the water column

MGI and MGII taxa were found throughout the 900 m water column at SPOT and though the summed relative abundance of MGI generally increased with depth (from an average of ~0.15% at 5 m to an average of 13% at 890 m), the MGII summed relative abundance varied much less (average 3–5%) across depths (Figures 1a and b). Furthermore, several 99% OTUs were dominant members at ‘adjacent’ depths (i.e., MGI 570 and 840 or MGII 248 and 2869), but the dominant OTUs were typically different at each depth (Supplementary Figures S1–S2). MGI and MGII OTUs found more abundant at different depths were typically phylogenetically distinct from each other, though this relationship was most obvious for the MGII (Supplementary Figures S1–S2). Utilizing 99% OTUs allowed differentiation of ecologically distinct MGI and MGII taxa that had different temporal and spatial patterns, despite ⩾97% sequence similarity (Supplementary Figures S3–S5).

Figure 1.

Five-year time-series of the summed relative abundances of (a) MGI and (b) MGII taxa. Filled circles indicate the depth and month a sample was taken, and the weighted-average was used to interpolate between depths and months.

CCA analysis of the MGI and MGII community composition across depths showed that MGI and MGII aphotic communities tended to separate into 3 general clusters for 150 m, 500 m, and 890 m (Figure 2). Although the MGII euphotic (5 m and DCM) communities formed a single distinct cluster, the MGI euphotic samples often occurred outside what might be considered their own general cluster, with composition sometimes resembling that of other depths (Figure 2b). This variability may partly result from CCA sensitivity to rare OTUs, as the 5 m and DCM MGI communities are much less abundant than at other depths (Figure 1 and Supplementary Figure S6a). Some samples fall outside their central depth-associated clusters (i.e., a 500 m sample in Figures 2a and b among the 150 m clusters), but these are few, and were possibly mislabeled prior to sequencing. NMDS plots of MGI and MGII communities also demonstrated clustering by depth (Supplementary Figure S7). The separation in 500 m and 890 m MGI communities is less pronounced than observed in the CCA plots; however, comparison of a 2D-plot and an 80degree rotated 3D-plot of the NMDS ordination, does reveal distinct clusters of these communities (Supplementary Figures S7A and S7C).

Figure 2.

Canonical Correspondence Analysis (CCA) of (a) Thaumarchaea Marine Group I, or (b) Euryarchaea Marine Group II communities from all samples, show distinct clustering of deeper (150 m-890 m) MGI and MGII communities. MGII surface communities formed a combined euphotic cluster, while the low abundance of MGI communities in surface samples likely allowed for greater variability in rarer OTUs causing a more distributed ordination of these samples. The environmental variables that best explained the variability and ordination of the samples are given as vectors in the ordination plots. Note that a few samples appear to have been mislabeled prior to sequencing and cluster with other depths. MLD=Mixed Layer Depth; NO₃⁻=nitrate concentration; Prim_Prod= surface satellite monthly average primary productivity measurement; PAR_SAT= surface satellite photosynthetically active radiation measurement.

Several environmental parameters related to MGI and MGII community ordination (Figure 2). Salinity (P=0.00010, q=0.0025), nitrate concentration (NO₃⁻) (P=0.014, q=0.017), and the mixed layer depth (MLD) (P=0.022, q=0.022) correlated to MGI communities below the euphotic zone, while temperature (P=0.0020, q=0.0033) most strongly correlated with MGI communities of shallower samples (Figure 2a). Surface-water satellite measurements of photosynthetically active radiation (PAR; P=0.017, q=0.017) most strongly associated with the MGII surface cluster, while satellite measurements of primary productivity in surface waters (Prim_Prod; P=0.0070, q=0.011) most significantly correlated to the ordination of MGII communities from 890 m (Figure 2b). Similar to MGI, temperature (P=0.0030, q=0.0075) was correlated to surface MGII communities and salinity (P=0.0090, q=0.011) correlated to the distribution of 890 m MGII communities. No other environmental measurements (of variables in Supplementary Table S2) were significantly correlated to ordination of MGI or MGII communities. Time-series profiles of environmental measurements included in statistical analyses are provided in Supplementary Figures S8–S16.

Archaeal temporal dynamics

Both communities exhibited statistically significant seasonality at every depth sampled (5 m, DCM, 150 m, 500 m, and 890 m; Table 1). Mantel tests of correlation revealed that Bray-Curtis dissimilarity of MGI or MGII communities increased as time between samples approached 6 months at all depths (Table 1). Mann-Whitney tests demonstrated statistically greater mean Bray-Curtis dissimilarities between any pair of samples taken at opposite times of the year (e.g., January vs June) as compared to samples taken at similar times of the year for MGI and MGII communities at all depths (Table 1).

Table 1. Archaeal communities at all depths demonstrated predictable temporal patterns despite differences in the average monthly variability within the MGI or MGII communities.

Taxa	Depth	Mantel ρ (Seasonal)	P-value	Mann-Whitney P-value: time of year
MGI	5 m	0.15±3.1e-2	<0.001*	<0.001*
	DCM	0.093±3.3e-2	<0.001*	3.8e-2*
	150 m	0.26±4.3e-2	<0.001*	<0.001*
	500 m	0.18±3.9e-2	<0.001*	<0.001*
	890 m	0.074±4.2e-2	7.2e-3*	2.2e-2*
MGII	5 m	0.20±4.3e-2	<0.001*	<0.001*
	DCM	0.12±5.4e-2	3.7e-3*	<0.001*
	150 m	0.20±4.0e-2	<0.001*	<0.001*
	500 m	0.26±4.0e-2	<0.001*	<0.001*
	890 m	0.11±4.3e-2	1.6e-3*	8.0e-3*

DCM=Deep Chlorophyll Maximum, Mantel R (ρ) value is given±the 95% confidence intervals. P-value given is based on 5000 permutation tests, *P<0.05.

Mantel Tests of the MGI and MGII communities at each depth show that Bray-Curtis dissimilarity between samples and differences in the time of year the samples were taken are statistically correlated, indicative of seasonality. Mann-Whitney tests between the Bray-Curtis dissimilarity of all pairs of samples show that communities sampled during similar seasons, regardless of year, are more similar to each other than samples taken during an opposite season.

Box and whisker plots of Bray-Curtis similarity between samples and time-lag (in months) between samples (Figure 3) shows apparent decreases in similarity at near 6-month intervals and increases in similarity at 12-month intervals, possibly indicating seasonal patterns (Figure 3). Mann-Whitney tests demonstrated mean Bray-Curtis similarities of MGI communities sampled 12 months apart were statistically greater than similarities between communities sampled 6 months apart at 5 m (P=7.66e-11), DCM (P=0.0132), and 500 m (P=1.603e-3). MGII communities only followed this trend at 500 m (P=4.045e-5). These analyses, however, only evaluate communities sampled exactly 6months or 12months from each other, which may not reflect the exact period of seasonality of these communities or may be more sensitive to outliers than the Mantel and Mann-Whitney tests described above.

Figure 3.

Bray-Curtis Similarities between all pairs of samples, plotted against the time-lag between the samples, show temporal variability and possible seasonality in (a–e) MGI, and (f–j) MGII communities. MGII communities in the euphotic zone (5 m, DCM) were more different (i.e., lower average similarities) across the time-series than communities deeper in the water column (150 m, 500 m, 890 m). Peaks in similarity near 12, 24, 36 month intervals (i.e., same seasons) and troughs near 6, 18, 30 month intervals (i.e., opposite seasons) suggests seasonality in the communities at several depths, most evident in 5 m Thaumarchaea. The average Bray-Curtis Similarity and standard error of the mean (SEM) between technical replicate communities of MGI or MGII for each depth are plotted on right, note due to extremely low abundance of MGI in surface (5 m and DCM) the replication is worse. Average whole community replication was 82.5%±9.25e-2 SEM.

Euphotic communities varied considerably more than deeper ones, i.e., average Bray-Curtis Similarities of MGI and MGII communities generally increased with depth (Figure 3). At 5 m and DCM both groups consistently had community similarities <50% between all possible pairs of samples, though there were peaks in similarity every ~12months (Figure 3a, b, f and g). Average similarities between samples at depths below 150 m were typically >60–70% (Figures 3c–e and h–j). We also noted that technical replicates (as controls) of the MGI communities showed relatively high variation in surface samples (Figures 3a and b), probably because these comprise a much smaller proportion of the total microbial community there (Supplementary Figure S6a), so the abundances (and hence community composition) of these taxa are more sensitive to stochastic variations. Note that similarities between samples at depths of 150 m, 500 m, and 890 m, often fell within the range of technical replicates, suggesting largely stable communities on average (Figures 3c–e and h–j).

Permutational multivariate analysis of variance (PERMANOVA) showed statistically significant (P<0.05) seasonal groupings for MGI and MGII at most depths, with the exception of MGI at 500 m (Supplementary Table S3). However, generally low R² values showed only weak predictability of groups by season. We analyzed MGI and MGII communities by similarity percentages (SIMPER) to identify OTUs that may influence seasonal community shifts. There appeared to be some association between phylogeny of MGI OTUs and seasonal contribution to community structure at 5 m and 150 m where seasonal variation in the OTUs was most obvious. For example, at 5 m and 150 m MGI OTUs 464 and 495 typically had fall-winter peaks (Supplementary Figure S16) and fell into the shallow marine group (note that the representative sequence of MGI_464 matches Candidatus Nitrosomarinus catalina SPOT01 and two strains of Nitrosopumilus maritimus, Supplementary Figures S1 and S3). Meanwhile, OTUs 570, 840, 30, and 1283 had spring-summer peaks and clustered into the deep marine group (Supplementary Figures S1 and S16). The phylogeny of MGII OTUs demonstrated a less obvious association to seasonal contribution. For instance, at 5 m MGII_89 and MGII_4267 had similar summer peaks but were from different clusters, while most OTUs at 500 m clustered into marine group b despite obvious differences in seasonal peaks (Supplementary Figures S2 and S17). Furthermore, MGII_198 had distinct seasonal patterns at 5 m and DCM, suggesting that 16S phylogeny alone may not predict temporal variability of MGII at SPOT.

Thaumarchaea marine group I correlation networks

eLSA networks revealed MGI OTUs correlated with many microbial OTUs at all depths and select OTUs were correlated to some environmental variables (Supplementary Figure S18). Several MGI OTUs were highly connected to other variables (microbial OTUs and environmental parameters) in these networks at all depths (Supplementary Table S4). MGI OTUs had common associations with SAR11 and Marine Group B (SAR324) OTUs across all depths (Supplementary Figure S18). Networks of 150 m to 890 m, also showed many correlations between MGI OTUs and Marine Group A (SAR406), SAR86, Nitrospina, and Verrucomicrobia OTUs (Supplementary Figures S18C–E). However, individual MGI OTUs did not correlate to the same microbial OTUs or environmental variables. For example, at 5 m, only MGI_464 had correlations to environmental parameters, with a positive correlation to oxygen saturation and negative correlations to Temperature, Salinity and day length (Supplementary Figure S18A). At 890 m only one MGI OTU negatively correlated to viral abundance (Supplementary Figure S18E), while several OTUs correlated to density (SigmaTheta), temperature, oxygen saturation, and surface particulate organic carbon (SAT_Surface_POC). Furthermore, only some MGI OTUs at the DCM, 500 m and 890 m correlated to picophytoplankton OTUs (Supplementary Figure S18). Among the MGI to microbial OTU correlations at 150 m, 500 m, and 890 m, we found different MGI OTUs commonly correlated to distinct Nitrospina OTUs (presumably nitrite oxidizing bacteria) and investigated these specific associations further.

Thaumarchaea Marine Group I and Nitrospina assemblages

Time-series plots of MGI and Nitrospina OTUs and eLSA networks at 150 m indicate different MGI-Nitrospina assemblages may be responsible for nitrification at different times of the year (Figure 4). For example, there were distinct correlations between MGI and Nitrospina OTUs and a time-series plot of the distinctly correlated MGI_464-Nitrospina_1255 and MGI_1283-Nitrospina_299 OTUs shows that these two MGI-Nitrospina groups commonly had mirroring, or negatively correlated, temporal patterns (Figure 4b). Similar patterns were observed between other correlated MGI and Nitrospina OTUs at 150 m (Supplementary Figure S19). Mantel tests also indicated a strong correlation between the MGI and Nitrospina communities at 150 m (Mantel ρ=0.71±0.12 95% confidence interval (C.I.), P<0.001). Deeper in the water column, seasonal patterns were not obvious, though MGI and Nitrospina OTUs still formed distinct correlations (Supplementary Figure S20) and changes in MGI communities were correlated to changes in Nitrospina communities (500 m Mantel ρ=0.60±0.16 95% C.I., P<0.001; 890 m Mantel ρ=0.43±0.11 95% C.I., P<0.001; Supplementary Figure S21).

Figure 4.

(a) Microbial association network shows strong correlations between particular MGI and Nitrospina 99% OTUs at 150 m, suggesting that the community structure of nitrifiers changes over time. Triangles are MGI OTUs and circles are Nitrospina OTUs. Boxes enclose MGI and Nitrospina OTUs plotted in B). Only correlations (Local Similarity values; LS) to Nitrospina OTUs with LS>|0.6| (P<0.01, false discovery rate q <0.01) are shown (i.e., Nitrospina are ‘hubs’) and the arrow indicates Nitrospina_1012 lags behind MGI_50 by 1month. The size of the symbols reflect average relative abundance of each OTU, and width of the lines indicate correlation strength. (b) Time-series line graphs show two pairs of correlated MGI and Nitrospina OTUs over 5 years, part of the data underlying the network in A. Additional examples are given in Supplementary Figure S19. Note MGI_464 matches SPOT01 (Ahlgren et al., 2017) cultured from this location.

We examined our data for presence of any other known ammonia oxidizing organisms and found Nitrosococcus and Nitrosomonas taxa (ammonia oxidizing bacteria), but at very low relative abundance (maximum average abundance of 0.0395% at 150 m). Preliminary analysis of the dataset with different clustering and taxonomic assignment methods, yielded at most 2 OTUs of nitrite oxidizing Nitrospira, with a maximum abundance less than 0.007%, but no OTUs were assigned as Nitrospira with the methods used in this study. Therefore, the abundance and consistent occurrence of MGI and Nitrospina OTUs suggest these organisms were the dominant nitrifying organisms in these samples.

Euryarchaea Marine Group II correlation networks

Correlation networks show differences in the bacterial, archaeal, and environmental parameters associated with different MGII OTUs at each depth (Supplementary Figure S22). To narrow the scope of this study, we primarily focused on correlations to the two most abundant MGII OTUs at each depth (Figure 5). At 5 m, these correlations were limited to other MGII OTUs (Figure 5a), despite many correlations to other microbes and environmental variables, including picophytoplankton OTUs, by less abundant MGII OTUs at this depth (Supplementary Figure S22A). Notably, at the DCM, MGII_89 was positively correlated to two cyanobacteria OTUs and to a chloroplast associated with the picoeukaryotic green alga, Bathycoccus prasinos. Only MGII_89 at DCM and MGII_2869 at 890 m were correlated to environmental parameters, interestingly, at 890 m, MGII_2869 was correlated to surface satellite measurements of particulate organic carbon (POC, Figures 5b and e).

Figure 5.

Networks show MGII 99% OTUs at (a) 5m, (b) Deep Chlorophyll Maximum (DCM), (c) 150m, (d) 500m, and (e) 890m are correlated to many distinct OTUs and environmental parameters. The most abundant MGII OTUs are given in red font. Letters in superscripts correspond to Supplementary Table S4. Boxes enclose related OTUs that share the main labels shown. Lines connecting symbols represent correlations (LS) either greater than 0.6 (solid lines) or less than -0.6 (dashed lines) between OTUs, and arrows indicate OTUs that lag behind connecting OTU by 1 month. The size of the symbols reflect average relative abundance of each OTU, and width of the lines indicate correlation strength.

Several phylogenetic groups correlated to MGII OTUs at all depths. OTUs from the SAR86, SAR406, SAR324, and Chloroflexi (SAR202), correlated to the two most abundant MGII OTUs at all depths except 5 m, though some SAR86 and SAR406 were negatively correlated to MGII OTUs (Figure 5). These abundant MGII OTUs also demonstrated common correlations to other Archaea, including the poorly studied MGIII. We found that some OTUs remained associated with this OTU across depths (SAR324_838, MGII_56, and MGIII_57; Figures 5c–e). Additionally, these abundant MGII OTUs often had more network connections than the large majority of other taxa, with MGII_2869 at 890 m having more correlations than 97% of all other nodes at 890 m (Supplementary Table S4).

MGI and MGII associations with picophytoplankton communities

MGI and MGII correlations to distinct picophytoplankton OTUs and phytoplankton derived parameters (i.e., POC, PAR, Prim_Prod) suggest a relationship between these communities. Picophytoplankton communities were detected at all depths sampled (Supplementary Figure S23) and demonstrated seasonality at every depth, as well as depth stratification, despite low abundance (Supplementary Table S5). Although picophytoplankton in the aphotic zone are likely inactive, changes in their community composition were correlated to MGI and MGII communities. Mantel tests of Bray-Curtis dissimilarity matrices of MGI or MGII and the picophytoplankton communities indicated MGI were statistically correlated to picophytoplankton at 150 m, 500 m and 890 m, while MGII was correlated to picophytoplankton at all depths (Supplementary Table S6). Surface (5 m and DCM) picophytoplankton communities were correlated to MGII communities at all depths except 150 m, with only the 890 m MGI communities correlated to DCM picophytoplankton.

Discussion

Community stratification across the water column

Previous studies indicated phylogenetic stratification of MGI and MGII between photic and aphotic environments (Hallam et al., 2006; Santoro et al., 2010; Auguet et al., 2012; Iverson et al., 2012; Luo et al., 2014; Martin-Cuadrado et al., 2015). However, this study demonstrates further partitioning of these communities in the aphotic depths at SPOT. To identify if archaeal communities demonstrate similar stratification globally, requires higher resolution deep vertical profiles at other geographical locations, as well as time-series sampling of these profiles to identify ephemeral, regularly repeating, or constant partitioning of archaeal communities between depths.

Similarities between 5 m and DCM archaeal communities (Figure 2), akin to what was found for bacteria by Chow et al., 2013, probably reflects biological and chemical similarities between these depths and suggests possible exchanges between communities during winter mixing. Cram et al., 2015 noted the mixed layer depth deepens significantly from an average of 10–20 m in March-November to 30–40 m in December-February (see also Supplementary Figure S12A showing reoccurring overlap between the depths of the mixed layer and the DCM). Therefore, clustering of the photic communities could be due to mixing of photic waters at SPOT.

Differences in community composition between depths of the aphotic zone may suggest niche partitioning within MGI and MGII, selecting for ecotypes within the communities better equipped to take advantage of environmental conditions found at each depth. Our results show that the extent and patterns of separation of the communities is different between MGI and MGII (Figure 2). For example, MGI surface communities show high variation, but MGII communities form a much tighter euphotic cluster. Furthermore, MGII aphotic communities separate more distinctly from surface communities than the MGI. We observed that the phylogenetic relationship to depth stratification appeared stronger for the MGII than for MGI (Supplementary Figures S1 and S2). This may indicate stronger selective pressure for distinct MGII taxa at different depths than for the MGI. The niches MGI communities occupy may be more similar throughout the water column, while depth-varying factors controlling MGII communities may more strongly influence segregation between depths (i.e., myriad heterotrophic substrates could vary more in composition and concentrations between depths than do the fewer substrates used by ammonia oxidizers). Furthermore, most studies have focused only on surface dwelling MGII (Iverson et al., 2012; Hugoni et al., 2013; Orsi et al., 2015; Martin-Cuadrado et al., 2015), while our results indicate that these are likely genetically distinct from MGII found below the photic zone. Salazar et al., 2015 demonstrated that the majority of bathypelagic MGII taxa might not be particle associated unlike the surface MGII, indicating that aphotic MGII may indeed carry out their heterotrophy differently from the surface-dwelling MGII.

MGI and MGII seasonality

The MGI and MGII communities at SPOT appear to have predictable seasonal dynamics throughout the water column (Table 1 and Supplementary Table S3). Although, the strength of this seasonality, as measured by the Mantel statistic (Table 1) and R² of the PERMANOVA tests (Supplementary Table S3) varies between depths, the congruency between tests suggests that the seasonality observed at all depths is not coincidental. Seasonal variability in bacterial and myoviral community composition has been shown in the surface and at 890 m at SPOT (Chow and Fuhrman, 2012; Chow et al., 2013; Cram et al., 2015). However, the archaeal-focused data presented here, suggests that archaeal communities at SPOT are seasonal at all depths sampled. The extent to which this water-column wide seasonality applies to archaea at other locations requires further attention.

Similar to what was found in other studies (i.e., Mincer et al., 2007; Beman et al., 2011; Hugoni et al., 2013), though low in abundance, only the 5 m MGI archaea had clear peaks in abundance during SPOT winter months (January and February, Supplementary Figure S6A). Like previous studies, our results suggest that mixed-layer depth is related to MGI community composition (Figure 2a). Additionally, temperature, salinity, and day length, which have winter lows (Supplementary Figures S11–S12), were negatively correlated to the abundant MGI_464 at 5 m (Supplementary Figure S8A), indicating a positive relationship to California winter months. We suggest winter mixing of the photic zone at SPOT (Supplementary Figure S12) brings MGI to the surface where they persist until the water stratifies, and populations are reduced by processes like competition with phytoplankton for ammonia and possibly light inhibition of growth (Yool et al., 2007; Qin et al., 2014; Baer et al., 2014). Previous studies have also shown seasonality of surface MGII communities (Murray et al., 1999; Hugoni et al., 2013; Galand et al., 2015; Needham and Fuhrman, 2016). As MGII taxa are presumed heterotrophs (Iverson et al., 2012; Orsi et al., 2015, 2016; Martin-Cuadrado et al., 2015), shifts in the surface community compositions are likely due to changes in the composition and availability of organic substrates, and/or top-down controls associated with phytoplankton blooms.

We observed potential associations between MGI phylotypes and seasons,, but not for the MGII. This is in contrast to the findings of Galand et al., 2010 and Hugoni et al., 2013, which found MGII Group B to be dominant in winter and MGII Group A dominant in summer months in surface waters of the Mediterranean Sea. However, the shorter lengths of the PCR amplicons or the region amplified in our study may not allow the same phylogenetic distinctions as found in those studies, or could indicate differences in the controls affecting MGII seasonality at these locations.

Potential influence of surface derived organic matter on community structure

Analysis of the community ordinations and correlations of MGI and MGII communities to picophytoplankton (Supplementary Figures S18 and Supplementary Table S6), indicates a possible relationship between surface productivity and archaeal community composition throughout the water column. While we do not have direct measurements of dissolved or particulate organic matter for our samples, we did find that MGI and MGII aphotic communities were correlated to picophytoplankton communities found at different depths (Supplementary Table S6). Although the phytoplankton observed are limited to the 0.2–1 μm size fraction, these were detected at all depths and displayed seasonality and differences in community composition across the water column (Supplementary Table S5 and Supplementary Figure S23). The picophytoplankton observed at depth are likely inactive or sloughed off larger particles during filtration, so that seasonality and depth stratification may reflect inputs of distinct surface derived substrates at different depths influencing archaeal composition throughout the water column. The ordination of 5 m and 890 m MGI communities (Figure 2a), the correlation of the MGII communities to surface primary productivity and PAR, and the correlation of multiple MGI OTUs and MGII_2869 at 890 m to surface POC (Supplementary Figures S18 and Figure 5e) further suggests phytoplankton derived particle flux affects MGI and MGII communities throughout the water column, but most strongly at 5 m and 890 m.

Recently, Cram et al., 2015 also reported significant seasonality of 890 m bacterial communities at SPOT, from analysis of 10-years of ARISA-based community fingerprints (Fisher and Triplett, 1999), attributed primarily to seasonal changes in particle flux from the surface. The authors noted some bacterial taxa at intermediate depths were seasonal, and suggested materials released from sinking particles probably influenced those taxa. This same flux of organic matter may be responsible for seasonality in the archaeal communities we observed. Decomposition of sinking particulate matter could release substrates for MGI ammonia oxidation or potential heterotrophy/mixotrophy (Ouverney and Fuhrman, 2000; Teira et al., 2006a; Qin et al., 2014) as well as organic substrates for stimulating MGII growth as observed by Orsi et al., 2016, leading to the apparent seasonality. Thus seasonal and other temporal changes in surface ocean biota that alter the flux and composition of particles, as shown for SPOT by Collins et al., 2011, may induce the shifts observed in the community composition of MGI and MGII archaea. Differences in the substrate utilization and rate of activity of distinct microbial members can therefore have important implications to biogeochemical cycles if the community shifts cause changes in the rates of remineralization and flux of particles reaching the sediments. Further analysis on the composition of organic matter and larger phytoplankton is needed to thoroughly investigate the relationship between the archaea and this surface derived organic matter.

MGI and MGII in correlation networks

We assessed which archaeal and bacterial OTUs were correlated to identify possible associations among microbial community members. The correlation networks suggest that some MGI and MGII taxa are among the most highly interconnected of any prokaryotes throughout the water column (Figure 5 and Supplementary Table S4). Importantly, the networks show that different MGI and MGII OTUs correlated to distinct microbial members even within a depth, suggestive of niche partitioning within these communities. This reveals the importance of high phylogenetic resolution analysis (e.g., the 99% OTUs we used here, see Supplementary Figures S2–S4) that enables observation of ecologically distinct members of the archaeal communities. Additionally, incorporating environmental measurements showed salinity, temperature, oxygen concentrations, and density likely influence specific taxa (Supplementary Figures S18 and S22), but show that the majority of MGI and MGII are mostly correlated to other microbes as seen previously for microbial communities at SPOT and in the Western English Channel (Gilbert et al., 2011; Steele et al., 2011).

Co-occurrence of MGI and Nitrospina

Correlations and time-series plots of MGI and Nitrospina OTUs (reported to be the presumed marine midwater organisms that complete the nitrification process by oxidizing nitrite to nitrate, Mincer et al., 2007) indicate that different assemblages of MGI and Nitrospina may be responsible for the complete nitrification process at different times of the year. Although, the relative abundance of some Nitrospina OTUs was low (Figure 4 and Supplementary Figure S19), repeating seasonal patterns and correlations to MGI OTUs that persisted over five years, supports the existence of a relationship between individual MGI and Nitrospina OTUs. The very low relative abundance of ammonia-oxidizing bacteria and no detection of other known nitrite oxidizing organisms, suggests these MGI and Nitrospina are effectively the main nitrifiers at SPOT. The reasons for the apparent temporal succession of distinct MGI-Nitrospina assemblages are not clear, whether it is coincidence from external driving forces or a result of specific organism-organism interactions including transfer of nitrite is unknown, suggesting the need for cultures and/or experiments to investigate the controls selecting for these apparently distinct seasonal assemblages. Because the enrichment culture SPOT01 (Ahlgren et al., 2017) has 100% similarity to MGI_464 (Supplementary Figures S1 and S3a), future experiments with it could help explore the MGI_464/Nitrospina relationship observed in the 150 m network (Figure 4). We note that both SPOT01 and N. maritimus match over the sequenced region (although not to each other over the entire 16S rRNA), but because SPOT01 was isolated from SPOT, while N. maritimus was isolated from an aquarium sediment, it is probably more representative.

Correlations between MGI and Nitrospina found in our study are in contrast to previous findings at SPOT, where connections between total MGI archaea and nitrite oxidizing bacteria were not significant (Beman et al., 2010). However, in that study the MGI were measured as a single variable by qPCR, using the total 16S rRNA or amoA gene copy number. Therefore, shifts in the dominant MGI OTU over time could not be captured. Nonetheless, the temporal variability in the identity of nitrifiers observed in this study could reflect seasonal fluctuations in nitrification rates. If these rate changes exist, it may suggest distinct contributions to carbon and nitrogen cycles during different times of the year or during other events that change the microbial community structure. Tolar et al., 2016, for example, found different ammonia oxidation rates between summer and winter in the Antarctic coastal surface waters corresponded to differences in the dominant Thaumarchaeal amoA gene. This suggests that changes in the dominant MGI taxa may in turn affect nitrification potentially through differences in the activity of those MGI or by the effect a specific MGI phylotype may have on the community of nitrite oxidizing bacteria. The changes in the dominant MGI taxa result from changes in substrate availability, removal rates through grazing or viral lysis, or changes in other controls that select for different MGI phylotypes or alter the activity of different members within the MGI community. However, the study by Tolar et al., 2016 focused on only one year, therefore it is unknown if those rate changes occur annually. Seasonal measurements of nitrification rates at SPOT and other geographic locations over multiple years are therefore necessary to evaluate if seasonality in the microbial community structure observed at SPOT is reflected in changes to nitrification rates. These results will allow us to understand if environmental changes that affect the composition of marine nitrifiers will result in rate changes to an important pathway in the nitrogen cycle.

Correlation networks may indicate activity shapes MGII communities

Correlation networks reveal potential associations between MGII and other heterotrophs throughout the water column (Figure 5). These networks show MGII OTUs are often associated with SAR406, SAR324, SAR86, and SAR202 taxa at all depths, suggesting strong direct or indirect relationships among these phylogenetic groups. The SAR406 have been shown to be heterotrophic and potentially capable of sulfur based energy metabolism (Rinke et al., 2013; Wright et al., 2014). Additionally, whole genome amplification has revealed that SAR324 taxa are likely heterotrophic and particle associated (Swan et al., 2011), similar to the anticipated life-style of MGII taxa. Through metagenomics and single-cell analyses, Dupont et al., 2012 demonstrated that SAR86 taxa, appear to be specialized in lipid and polysaccharide degradation. Less is known about the activity of SAR202, though they have been shown to take up amino acids (Varela et al., 2008). Beman et al., 2011 also found correlations between MGII 16S gene abundance and SAR11, SAR86 and Bacteroidetes OTUs, indicating consistent findings of correlations between MGII and heterotrophic bacteria. The large number of associations between MGII and these known heterotrophic bacteria likely indicate activity-based relationships among these taxa, perhaps driven by organism-specific interactions, similar responses to nutrient inputs or other as-yet-unknown details of the degradation and consumption of particulate and dissolved organic matter.

Correlations between MGII and various picophytoplankton were also found throughout the water column at SPOT. At photic depths, several MGII OTUs were correlated to various Prochlorococcus, chloroplasts (including one OTU classified as belonging to Bathycoccus prasinos, Figure 5b) and other unclassified cyanobacteria OTUs (Supplementary Figures S22A–B). These correlations may indicate potential sources of organic substrates promoting MGII heterotrophic growth. Indeed, Orsi et al., 2015 demonstrated that MGII abundance was stimulated by the presence of picoeukaryotic phytoplankton, and similar MGII communities could utilize proteins from these phytoplankton (Orsi et al., 2016). A recent report about microbial associations in surface waters at SPOT showed MGII were at times highly abundant in both <1μm and >1μm size fractions, where one MGII OTU bloomed sharply (to 30% of the entire prokaryote community) in concert with the haptophyte Phaeocystis globosa, possibly indicating a physical association (Needham and Fuhrman, 2016). Additionally, metagenomic studies demonstrate that assembled genomes of MGII taxa indicate heterotrophic lifestyles despite phylogenetic differences (Iverson et al., 2012; Martin-Cuadrado et al., 2015). These studies and our results suggest MGII organisms may be important contributors to the degradation and respiration of phytoplankton biomass throughout the water column, but that the microbial relationships or inputs of organic substrates selects for which MGII taxa may occur at any given time or depth.

Conclusions

This study provides evidence that Thaumarchaea Marine Group I and Euryarchaea Marine Group II at SPOT are stratified and often seasonal in composition. Numerous temporal correlations within microbial communities suggest that archaeal community seasonality reflects temporal variability of microbial assemblages, though this is superimposed over additional variability that appears stochastic. These changes in community structure may therefore alter the rates or efficiencies with which the microbial community drives biogeochemical cycles. Understanding what selects for different assemblages may therefore be key in predicting changes to these cycles in response to regional and global environmental changes.