The aim of this study was to elucidate the community structure and nifH mRNA expression of diazotrophs associated with plankton with a particle size > 100μm in northern South China Sea. Our results showed that the active PADs communities were taxonomically distinct from the total communities. Trichodesmium predominated among the PADs of the euphotic zone, with approximately equal abundances in the DNA and RNA libraries. Differences in the nifH gene expression levels of the different phylogenetic groups of diazotrophs may be related to the life forms and symbiotic strategies of these organisms.
Active communities are more diverse and divergent than total communities
In determinations of the microbial communities in a particular environment, higher community diversity in DNA- than in RNA-derived libraries is typical and has been demonstrated in several 16S rRNA-based studies [33, 48-51]. Here we directly tracked nifH-defined communities, because nifH gene may provide better insight into the diversity and composition of the communities that are capable of nitrogen fixation in an environment than those extrapolated from16S rRNA data. However, our nifH sequence analysis revealed a higher alpha-diversity in the active (based on RNA) than in the total (based on DNA) communities of PADs (Fig. 2). The evenness of the total PADs communities was also lower than that of the metabolically active communities, as indicated by the very small number of highly dominant taxa. The nifH gene-based OTU abundances from the DNA analysis do not always predict which diazotrophic OTU is actively expressing nifH assessed in the RNA analysis. The former one is presumably considered a proxy for the dominant diazotrophs in a given sample, whereas the latter offers a representation of metabolically active diazotrophs with regard to nitrogen fixation processes. A high alpha diversity in a given community could be a result of high evenness. As such, a higher evenness obtained from nifH RNA-based OTU analysis accounts for closer number among individuals that were actively expressing nifH, relative to the communities comprising a few highly dominant diazotrophs detected in nifH DNA based approach. This is supported by the prevalence of Trichodesium in all DNA-derived communities and the decline in its abundance in the RNA-derived communities, whereas some taxa that were rare in the total communities were of moderate abundance in the active communities (Fig. 5). Thus, although the low-abundance groups represented a small proportion of the total PADs community biomass, they contributed a disproportionately high amount of the activity, presumably due to active transcription of nitrogenase mRNA under suitable growth conditions.
Previous studies have confirmed the dissimilarities between total and active diazotrophic communities [52, 53], but the variability within communities of total and active diazotrophshas not been quantitatively investigated. Given that changes occur faster in RNA than in DNA , we hypothesized a more pronounced variability in the active diazotrophic communities (RNA) across stations than in the total communities (DNA). Indeed, variations in the community structures of PADs among sampling sites were observed and were independent of the presence or activity of the cells. However, in accordance with our hypothesis, this variation was significantly higher in the active fraction of the diazotrophic communities than in the total communities (Fig. 3b). Moreover, the differences between the active communities at different sites were mainly caused by the replacement of taxa, rather than changes in the abundance of common taxa (Fig. 5), suggesting that different selective pressures determine the turnover in membership of metabolically active diazotrophs in the studied ecosystems. Accordingly, active diazotrophic assemblages may better reflect the response of PADs to environmental alterations.
A varying level of metabolic activity among rare taxa
Given that some taxa tend be to conditionally rare , investigations of membership turnover in rare but active populations of natural communities may shed light on microbial diversity and succession. In our study, rare but active PADs groups included heterocystous cyanobacteria, unicellular cyanobacteria, Deltaproteobacteria and Gammaproteobacteria (Figs.5 and 7). Members of the heterocystous cyanobacterial genus Richelia were the most transcriptionally active group. The heterocystous cyanobacteria Richelia has been extensively reported as a diatom symbiote, including the investigations in the South China Sea [34, 35], and was identified to actively transfer nitrogen to their diatom hosts . The association of diazotrophic unicellular cyanobacteria with phytoplankton has been the subject of recent studies [11, 27, 56], including a report describing a high nitrogen fixation rate and rapid nitrogen transfer from symbiotic unicellular cyanobacteria to their hosts, evidenced by N-isotope-based nano-SIMS technology . The highly active transcription of nitrogenase mRNA by intracellular symbiotic Richelia and unicellular cyanobacteria detected in our study is in agreement with previous studies reporting high nitrogen fixation rates by those diazotrophs [11, 15]. Intracellular symbiotic strategies and nutrient exchange mechanisms may facilitate a high efficiency of nitrogen fixation by Richelia as well as unicellular cyanobacteria.
High nitrogenase mRNA transcription level was also detected in Deltaproteobacteria and Gammaproteobacteria, previously reported to be dominant diazotrophic groups [35, 57]. In our study, although Deltaproteobacteria was present at low abundance, its level of nifH mRNA transcription activity was high. This group was detected in copepods in previous studies [24, 26] whereas in the case of diazotrophic Gammaproteobacteria little is known about its symbiotic relationships. Fluorescence in situ hybridization studies of diazotrophic Gammaproteobacteria may lead to the identification of its host and thus provide insights into the ecological importance of this PADs group.
The inactive diazotrophic groups
Nitrogenase is a multi-component enzyme with several different forms, although the nifH gene encodes a highly conserved subunit . However, the nifH gene also shares a similarity with nifH-like genes in non-diazotrophic organisms, such that many nifH sequences retrieved from environmental clones are in fact pseudogenes [58, 59]. In this study, nifH sequences belonging to cluster III were the second most abundant group in the nifH DNA libraries, but few transcripts were detected in the nifH RNA libraries for all stations (Figs 5 and 6). Considering that the cluster III nifH sequences obtained in this study were not transcribed, whether they were pseudogenes or indeed belonged to inactive groups was unclear. The reference sequences that clustered closely with the sequences of cluster III (Fig. 4) originated from zooplankton and the termite gut [24, 60]. In addition, the nitrogenase activity of PADs was shown to be much higher in starved copepods than in full-gut copepods . It was therefore hypothesized that, the presence of sufficient bio-available nitrogen in the zooplankton gut down-regulates nitrogenase gene expression . Similarly, the cluster III diazotrophs found in our study might be those inactive members associated with zooplankton, which would imply that zooplankton-associated heterotrophic diazotrophs are suppliers of extra nitrogen for their hosts, especially under conditions of starvation.
Trichodesmium dominate the PADs
The results of microscopy-based cell counting showed a relatively lower Trichodesmium abundance (1.29×103 –8.22×103 trichomes m−2 in the euphotic waters, Fig. S2) in the South China Sea than previously reported [34, 35, 61]. The differences may have been due to the use of different sampling methods, as the net-tow (100-μm mesh) collection of plankton in our study would have excluded free-living or smaller aggregates of Trichodesmium cells, both of which are captured in previous studies using smaller pore-size filters (0.2-, 10- or 20.μm) [34, 35, 61]. Nevertheless, the sampling method used in our study was chosen to ensure collection of the PADs fraction while excluding free-living diazotrophs, as that different patterns for PADs and whole diazotrophic communities were expected.
Trichodesmium is one of the dominant diazotrophs in tropical and subtropical oceans [5, 6, 62]. In the South China Sea, heterotrophic Proteobacteria was shown to dominate the diazotrophic communities [34, 35]. However, consistent with the results of Farnelid and colleagues, who reported the predominance and enrichment of cyanobacteria, including Trichodesmium, in diazotrophic communities associated with sinking particles , we found that Trichodesmium accounted for nearly all of the PADs (~98%) in our samples. It seems that autotrophic cyanobacterial diazotrophs such as Trichodesmium tend to dominate in large particle-size fraction of the diazotrophic community inhabiting the euphotic-zone, while heterotrophic diazotrophs tend to be free-living. Alternatively, the disproportional enrichment of Trichodesmium in the >100 μm size fraction could be resulted from colonization of their cells. Thus, our collection method revealed a difference in the community structure of bulk diazotrophs vs. PADs associated with large particles.
We also found that Trichodesmium accounted for the majority (87%) of the nitrogenase mRNA transcripts of PADs (Fig. 5). However, an analysis of the nifH RNA/DNA ratio showed that the nifH gene expression level of Trichodesmium was lower than that of rare but active groups (e.g., heterocystous cyanobacteria) (Fig. 7). In a previous isotope-based investigation, more than half of the Trichodesmium cells were incapable of fixing nitrogen . Therefore, changes in the abundance of diazotrophic groups, determined using nifH DNA sequences, may not reflect changes in the nitrogen fixation activity of the respective members, whose ability to fix nitrogen varies. However, newly developed isotope techniques, especially NanoSIMS, allow the nitrogen fixation rate of specific groups to be investigated at the single cell level [11, 63].