The nitrogen (N) cycle is complex and N availability can regulate primary production and, in turn, determine where and how new biomass is formed, stored, and exported from the euphotic zone of aquatic systems (Gruber, 2008). Understanding the dynamics of N cycling, including the factors that control individual processes and the organisms involved, has been a major goal of microbial ecologists and biogeochemists for decades. The information presently available has been used to develop global models aimed at further understanding how these dynamics affect community structure and the cycling of other key elements, particularly carbon (C) (Galloway et al., 2008). Ultimately an overarching goal is resolving the extent primary production is driven by new vs. regenerated sources of N so this can be used to further refine these models (Falkowski et al., 1998).

Subsets of microbes contribute to both autotrophic and heterotrophic processes in aquatic freshwater and marine systems. The canonical view of this is that photoautotrophic protists and cyanobacteria are important primary producers while heterotrophic microbes participate primarily in the degradation and remineralization of this newly formed organic material (Kirchman, 1994). Extensive work has shown that diverse clades of microorganisms, e.g., SAR11 and Prochlorococcus, may transition between subgroups of phylogenetically similar organisms that possess distinct physiological characteristics (i.e., ecotypes) based on fluctuations in their physical and chemical environment (Rocap et al., 2003; Johnson et al., 2006; Carlson et al., 2009; Galand et al., 2009, 2010; Vergin et al., 2013). Shifts in the predominance of these specific ecotypes may affect the patterns and regulation of key environmental processes (Rocap et al., 2003; Johnson et al., 2006; Carlson et al., 2009; Galand et al., 2009, 2010; Vergin et al., 2013), e.g., the production and degradation of organic material. However, the dynamics of these interactions are not well understood reinforcing the importance of characterizing the roles of different ecotypes in the environment in order to constrain their effect.

The identification and characterization of key players in microbial communities were, until recently, largely restricted to the correlation of molecular biological observations (e.g., the presence of functional genes or transcripts) with biogeochemical or physiological measurements (Zehr et al., 2001; Montoya et al., 2004; Lam et al., 2009) and the cultivation of isolates. However, most microorganisms in the environment resist current culturing techniques while those isolated are often minor parts of the community that are selected for based on the isolation method (Staley and Konopka, 1985). Strain-specific genetic and functional differences can also impede the disentanglement of an individual’s niche and how a specific ecotype may fit into a complex ecosystem. Traditional bulk assays may only access the most active and abundant organisms during the time of sampling, potentially missing important but relatively rare or ephemeral taxa (Sogin et al., 2006). Such information can provide a foundation for future work, but can be misleading if not interpreted cautiously since this may only supply a superficial glimpse into the potential metabolic diversity and potential of an ecosystem.

Stable isotope probing (SIP) has gained popularity in recent years due to its ability to provide direct evidence of substrate assimilation by specific taxa, thereby revealing their functional roles within a community (Neufeld et al., 2007a; Buckley et al., 2008; Addison et al., 2010; Nelson and Carlson, 2012). Actively growing organisms capable of assimilating substrates enriched with ¹³C or ¹⁵N incorporate these heavy isotopes into newly formed nucleic acids, increasing the density of these molecules relative to those same organisms in controls grown on ¹²C or ¹⁴N-substrates (Neufeld et al., 2007b). The nucleic acids are then separated by density gradient ultracentrifugation and a variety of techniques can then be employed to both detect this enrichment and identify the organisms assimilating the substrate of interest. Uptake of ¹⁵N-substrates can result in a maximum density increase of only ~50% of that achievable with ¹³C-substrates (Buckley et al., 2007a). An additional complication in interpretation of N SIP data is the range of DNA GC-content (Buckley et al., 2007b) found within a sample, making density shifts, particularly in individual species, difficult to differentiate. Offsetting many of these problems, Tag-SIP (16S rRNA gene amplicon or ‘tag’ sequencing, combined with SIP) increases the resolving power of traditional SIP by examining the densities of individual OTU’s DNA to more easily identify and compare shifts (Figures 1 and 2), similar to a recent study (Connelly et al., 2014).

Tag-SIP was applied to coastal surface waters in order to investigate our hypothesis that a population’s capacity to assimilate different N sources, particularly on the subgroup level, is much more diverse than is currently assumed, ultimately leading to an incomplete understanding of community structure, functional capacity, and metabolic activity. We provide evidence of intraclade heterogeneity with respect to both ammonium and nitrate assimilation, shedding light on the activity of individual organisms and the roles they may play in the cycling of nutrients in situ (Fuller et al., 2005; Follows et al., 2007).

DNA recovered from natural populations of prokaryotic assemblages was used to assess a community’s capacity to incorporate a particular isotopically enriched N-substrate. This was accomplished through the identification of a relative increase in density between DNA bands of OTUs recovered from control and isotopically treated samples using the Tag-SIP methodology. In order to establish the density of each OTU, the peak of the DNA band had to be contained in at least three fractions to be assessed, while single point peaks were ignored. In the rare occasion where no definitive peak fitting the criteria was observed, the overall distribution of the bands was compared. Previous methods using similar criteria have shown this general approach to be valid (Buckley et al., 2007a,b; Wawrik et al., 2009; Nelson and Carlson, 2012; Wawrik et al., 2012; Connelly et al., 2014). However, a limit of detection needed to be tested before enrichment could be properly identified in our study.

Based on the results of the LNT, a limit of detection was conservatively set at 0.003 g ml^-1. Additionally, DNA repair can increase the overall incorporation of label into DNA without any subsequent cell divisions (Alberts et al., 2002) and taxa with higher metabolic rates could potentially carry out more than a single cell division during these incubations. This potentially increases the probability that some OTUs within the LNT were labeled, further bolstering this as a conservative cutoff for enrichment. Therefore, we are confident that any shifts in DNA equal to or greater than this limit of detection were due to the active assimilation of labeled substrate making this value more sensitive than those reported previously [>30%, (Murrell and Whiteley, 2011; Connelly et al., 2014]. This cutoff was then applied to the ammonium and nitrate treatments to evaluate these communities’ ability to assimilate different N sources.

Not all OTUs displayed evidence of ammonium assimilation during the incubation (Table 2) although it is generally assumed that most microbes can carry out this process. Thus, this heterogeneity with respect to ammonium assimilation may actually tell us more about the activity of these OTUs than their functional capacity. This is because SIP is not just testing capacity but also activity in general. Many factor can play into the activity and labeling of individual microbes in situ including life history and the environmental conditions, both of which can greatly affect their physiological state (Blazewicz et al., 2013) leading up to and during the incubation period. This can be compounded by taxon diversity where OTUs may possess distinct genetic or physiological machinery that may differ in transcription rates, kinetic effects caused by substrate concentration, or mechanism of regulation (Vergin et al., 2007; Berges and Mulholland, 2008; Tripp et al., 2008). Dormant or dead cells will have little chance of ammonium assimilation, but their DNA can still be recovered and assessed. Conversely, active utilization of the substrate can occur without simultaneous or subsequent DNA turnover. Both situations would lead to a negative result; SIP labeling only occurs after the substrate is taken into the cell and incorporated into DNA at a fast enough rate to significantly increase the density of their DNA above the limit of detection over the incubation period. An isotopically labeled OTU must almost certainly be active in order for this labeling to occur. SIP therefore may be more sensitive in identifying the most active fraction of a mixed community than more standard methods, e.g., rRNA analyses, since there is not always a direct correlation between activity and rRNA concentration (Gausing, 1977; Blazewicz et al., 2013). The OTUs labeled within the ammonium treatment likely represent the most active members of the microbial population during the time of sampling (Table 2).

A diverse array of heterotrophic clades demonstrated significant evidence of nitrate utilization. These results directly complement previous work (Allen et al., 2001; Cai and Jiao, 2008) which had uncovered nasA genes (assimilatory nitrate-reductase) within diverse groups of heterotrophic bacterial clades distributed among contrasting oceanic regimes, and bacterial uptake of nitrate accounted for up to 40% of the total nitrate uptake at five stations in the Barents Sea (Allen et al., 2002). In more recent investigations, culture dependent and independent analyses of diverse marine environments suggested that previously surveyed nasA genes had been underestimated (Jiang et al., 2015) while Wawrik et al. (2012) provided evidence of nitrate assimilation by heterotrophs despite the inability to recover transcripts of nasA genes. These results further demonstrate that heterotrophic organisms can and are actively competing with the phototrophic community for this oxidized N source.

Initial genomic surveys and culture work reported Prochlorococcus’ inability to assimilate nitrate (Moore et al., 2002), but more recent studies have indicated otherwise (Casey et al., 2007; Martiny et al., 2009; Treibergs et al., 2014). Our direct evidence suggests, by way of the presence of prominent, shifted secondary peaks (Figure 2C; Supplementary Figure S2) that assimilation of nitrate by Prochlorococcus occurs in situ. Secondary peaks are not uncommon in SIP (Buckley et al., 2007a; Wawrik et al., 2009) and likely point toward the presence of subpopulations within an individual OTU with different levels of activity or nutrient pathways, which has been proposed previously for this clade (Martiny et al., 2009; Biller et al., 2014; Kashtan et al., 2014). Both OTUs recovered showed enrichment with similar banding patterns providing potential validation of nitrate utilization in this study (Supplementary Figure S2). Phylogenetic analysis reveals that our Prochlorococcus OTUs fall within the high-light (HL) ecotype (Supplementary Figure S3) and are most similar to the MED 4 strain (100 and 99.7%).

Our results are supported by the work of Martiny et al. (2009) who showed a genomic island, or irregular regions of the genome often used to identify possible sites of lateral gene transfer (LGT), containing Synechococcus-like genes involved in nitrite and nitrate acquisition. These genes were also found to code functional proteins and their distribution was shown to be widespread in marine surface waters. Furthermore, a recent study uncovered genes involved in nitrate assimilation within three distinct Prochlorococcus lineages, as well as the growth of isolates solely on nitrate, including members of the HL ecotype (Berube et al., 2015a). Berube et al. (2015a) only recovered these genes in HLII, while our study suggests nitrate utilization in members of HLI but no members of any other subclade were observed or analyzed within the top one hundred OTUs. This would be expected as our study site contains colder waters with higher nutrients where HLI would be likely to dominate (Johnson et al., 2006) compared to the much warmer waters where most of their strains were isolated (Berube et al., 2015a). They suggested that genes for nitrate assimilation may have been gained and lost multiple times as the Prochlorococcus clade has evolved but also had evidence that these genes may be susceptible to LGT (Berube et al., 2015a). Berube et al. (2015a) speculate that this could facilitate the introduction of this ability into other lineages.

These studies, in combination with our evidence of in situ nitrate utilization, confirm that this cyanobacterial clade is much more physiologically and genetically diverse than originally perceived (Moore et al., 2002; Kettler et al., 2007). Prochlorococcus is the most abundant phototroph on the planet (Johnson et al., 2006; Flombaum et al., 2013), and so any changes in the understanding of this clade’s physiology or ecological niche will improve our current perspective of its role in the microbial ecology of the ocean (Rocap et al., 2003; Giovannoni and Stingl, 2005; Johnson et al., 2006; Martiny et al., 2009). Our understanding of N cycling and new production in oligotrophic area, such as the tropical and subtropical open oceans, will likely be affected most since Prochlorococcus can account for up to ~45% of the photosynthetic biomass and carry out a significant fraction of net primary production (~10–50%) in these regions (Goericke and Welschmeyer, 1993; Liu et al., 1997). Maximum abundances are observed when surface waters become stratified and nutrients are depleted, increasing the competition for nitrate. A recent study carried out in the Atlantic and Pacific attempted to examined this by quantifying nitrate assimilation genes within the HLII and found them to be 20–50% of the population under stratified conditions (Berube et al., 2015b). Further work is required to understand the spatial and temporal variations of these nitrate utilizing ecotypes in order to better assess and appreciate the role Prochlorococcus plays in the marine community as well as in both N and C cycling.

Evidence of nitrate assimilation within the SAR11 clade appears to follow phylogenetic lines as all OTUs that did not show evidence of nitrate assimilation were found within subclade II, while those with enrichment were from subclade I (Figure 3). This suggests potential niche partitioning among ecotypes, wherein subpopulations’ distinct phenotypic variations are selected for when certain environmental conditions develop. The generally higher-nutrient coastal waters investigated in this study may be more likely to harbor taxa with the genetic capacity to respond to N sources across a range of oxidation states. Our nitrate amendment may have enriched for this particular ecotype, facilitating its detection. Ecotypes are not uncommon within the SAR11 clade and previous work has suggested it has evolved into at least a dozen specialized ecotypes who’s relative abundances often appear to correlate with fluctuations in environmental parameters (Carlson et al., 2009; Vergin et al., 2013). Many of these subclades harbor hypervariable regions that have been implicated in the acquisition of novel genetic material, enabling strain-specific metabolisms, e.g., sulfur metabolism unique to strain HTCC9565 of SAR11 (Grote et al., 2012). A recent study investigating single amplified genomes (SAG) off the coast of Mexico uncovered the genetic potential to reduce nitrate within several subclades of SAR11 taken from the oxygen minimum zone (Tsementzi et al., 2016). Though assimilatory pathways were not address directly in their study, genes involved in both the uptake and reduction of nitrate were present and actively transcribed (Tsementzi et al., 2016). This result in combination with our direct evidence further bolsters the potential that an assimilatory pathway is present and active within the SAR11 clade. Tag-SIP may facilitate a fine scale look at the functional capacity of individuals, analogous to SAG, through the use of targeted N amendments and fractional gradient sequencing.

Nitrate assimilation may have only recently arisen within the SAR11 lineage when subclades I and II diverged (Figure 3). Acquisition of this capability may have contributed to the divergence. Alternatively, microdiversity may have independently developed multiple times. The physiological state or life history of the community at the time of sampling may have also played a role in these apparent intraclade differences, potentially making these results more of a reflection of increased activity than lack of capacity, as mentioned previously. These differences in activity or capacity could potentially help explain some of the aforementioned variations in abundance and distribution patterns often associated with this clade (Carlson et al., 2009; Vergin et al., 2013). Therefore, a complete understanding of the dynamics of SAR11’s metabolic potential is essential to fully understanding its role in organic C degradation and sequestration in the oceans, particularly in the vast expanses of the oligotrophic gyres where nutrients are low and SAR11 abundances are high (Morris et al., 2002; Carlson et al., 2009; Brown et al., 2012; Vergin et al., 2013).

Similar heterogeneity was observed within many other major bacterial clades including Roseobacter, SAR86, and Flavobacteriaceae (Table 2), which we see as additional evidence of phenotypic variation of closely related organisms and potentially ecotypes exhibiting different metabolic strategies, and/or activities, allowing them to occupy discrete environmental niches. This result was not only restricted to the bacterial domain but was also found within the archaeal clade MG-II. Previous work surveying 16S rRNA genes similarly uncovered potential evidence of functional differences among these subclades (Galand et al., 2009, 2010). However, no cultured representatives exist from MG-II, so there is little known about their modes of nitrogen metabolism, although the ability to both transport and utilize nitrate through dissimilatory and assimilatory pathways has been found widespread throughout the Archaeal domain (Cabello et al., 2004). A single recently closed genome representing multiple strains of MG-II did not appear to contain any known genes associated with nitrate utilization (Iverson et al., 2012) and no cultured representatives currently exists. Since ours is the first dataset of this nature for MG-II, which often dominates archaeal gene surveys in the marine water column (DeLong et al., 2006; Galand et al., 2009), a reconsideration of the role of archaea in the uptake of N is indicated, just as we have had to revise our conceptual model to include a role of Archaea in ammonium oxidation (Wuchter et al., 2006).

It is important to acknowledge the possibility that some of our nitrate results may be affected by cross-feeding, where uptake of labeled nitrate by one organism is then metabolized to another form, e.g., ammonium, released, and assimilated by another organism yielding a false positive. We argue that nitrate uptake, metabolism, and release of sufficient substrate by one organism, in close enough proximity to another organism that is able to take it up and enrich their DNA by the ~20% required for detection, within a 24 h incubation, is a highly unlikely, particularly given the dilution of the enriched substrate by ambient pools of unlabeled nitrogen. A highly sensitive direct visualization technique, i.e., nanoSIMS, showed that, despite being attached directly to an actively fixing nitrogen fixer (diazotroph), epibiotic (i.e., attached) heterotrophic bacteria were only weakly enriched secondarily with released ¹⁵N, and cells that were <10 μm away from the diazotroph showed no observable enrichment (Behrens et al., 2008).

Though rates were not measured directly in this study, measurements in a previous investigation within the Southern California Bight (SCB) can be used to further demonstrate the unlikelihood of cross-feeding here. Bronk and Ward (2005) reported nitrate uptake rates of ~0.01 μmol N L^-1 h^-1 that would equate to an ~20% enrichment of the ~1.25 μM N PON. This enriched PON could theoretically produce ~0.009 μM of ¹⁵N-DON and ~0.13 μM of ¹⁵N-NH₄⁺ during this incubation with release rates of ~0.002 and 0.028 μmol N L^-1 h^-1, respectively (Bronk and Ward, 2005). If these were the sole ¹⁵N sources being released, this would make up a small fraction, i.e., ~8%, of the total N taken up through ammonium assimilation based on the rates measured in the SCB study [~0.072 μmol N L^-1 h^-1; Bronk and Ward, 2005] and this enrichment would be undetectable by Tag-SIP. Furthermore, the relatively high concentration of unlabeled ambient ammonium (1.19 μM, Table 1) and DON [~5–7 μM; Bronk and Ward, 2005] in these waters would dilute any labeled ammonium or DON released, additionally decreasing the possibility of measurable cross-feeding. Therefore, it is likely that only the most metabolically active organisms with consistent access to the primary labeled substrate would be distinguishable from the rest of the community with our method. Finally, in all treatments, the percentages of DNA labeling found in organisms that are thought to be more active, e.g., phototrophs, were similar to those found in the presumably less active heterotrophic community (data not shown), possibly reflecting use of the same labeled source. Therefore cross-feeding was considered highly unlikely to have any major effects on the results.

If any cross-feeding was occurring, it would probably come from highly active organisms living in very close association with primary utilizers. Their metabolic activities would have to be tightly coupled in order to accumulate enough secondary label within these 24 h incubations. This type of cross-feeding would still be quite informative, delineating an effective and rapid trophic cascade. Such results would potentially be evidence of some type of microbial association, e.g., symbiosis, where nutrients were rapidly transferred directly in at least one direction, though the mechanism would remain unclear. Alternatively, no specific associations could be involved but the taxa involved would have to be specifically geared toward scavenging newly produced N quickly and efficiently. These populations may continuously be maintaining significant amounts of resource-acquisition machinery primed and ready or investing heavily in growth machinery to facilitate rapid growth when fluxes of regenerated N occur, similar to the optimization model of phytoplankton (Klausmeier et al., 2004). These tightly coupled associations in combination with the rapid and efficient exploitation of freshly released N could potentially play a role in the global dominance seen in heterotrophic clades like SAR11 and MG-II, permitting them to preferentially access to exudates.

This dominant nature associated with some of the major clades, e.g., MG-II, SAR 11, and Flavobacteriaceae, may be further explained by the functional heterogeneity we observed, where any one member of a subpopulation may be more suited under a particular set of environmental conditions, allowing the population to weather a range of variables, e.g., changes in light or nutrients. This may also shed light on the long standing ‘paradox of the plankton,’ which is the often observed high diversity of organisms under conditions of apparent resource limitation. Each of these fluctuations in the physical and chemical environment may select for a different population or subpopulation, maintain diversity that would otherwise be filtered out by competitive exclusion (Hutchinson, 1961). Alternatively, some ecotypes may have the ability to assimilate the substrate, but their growth rate may be too low to resolve incorporation in the length of our incubations. These differences in activity may also play a role in ecotype selection.

Our results suggest that this functional heterogeneity may be more widespread in the microbial community than previously appreciated. Resolving the extent of functional heterogeneity in the environment will likely be difficult with traditional methods, as detection usually requires long-term surveys to identify the highly structured and rhythmic patterns of variation over space and time. Tag-SIP may provide a much more tractable approach to identifying ecotypes, requiring only a single sample, as well as having the added advantage of providing direct information on their function.

Our work provides insight into the uptake of a substrate by a cell, and then furthers this insight by demonstrating that the organism is metabolically active and assimilating this substrate into biomass, i.e., DNA enrichment, something very difficult to otherwise evaluate. This enabled us to investigate some prevailing perceptions, particularly in regard to heterotrophic prokaryotes’ influence (e.g., SAR11) on nitrate dynamics as well as the potential frequency of intraclade heterogeneity. Our data suggest that this phenomenon is likely much more ubiquitous than previously understood and requires greater examination. We also uncovered what is potentially the first direct evidence of in situ nitrate assimilation by several important marine taxa including SAR11 and MG-II. As the sophistication of global biogeochemical cycling models increases with the inclusion of specific functional groups and ecotypes (Follows et al., 2007; Follows and Dutkiewicz, 2011), the effect of addition of these dominant marine clades to nitrogen cycling sub-models could be substantial. Tag-SIP holds great promise in gaining a robust knowledge of the specific roles of important marine groups in a range of biogeochemical processes. However, further work is required to expand its use and to corroborate these findings in order to better understand how they fit into our overall perspective of global nutrient cycling.

MM and DC designed the research; MM conducted the research; MM analyzed and synthesized the data; and MM and DC wrote the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.