Hydrostatic Pressure Regulates Microbial Structuring And Metabolic Functions In The Pelagic Ocean

Background: Microbial-mediated decomposition of particulate organic matter (POM) during its downward transport from the surface to the deep ocean constitutes a critical component of the global ocean carbon cycle. However, the extent to which hydrostatic pressure affects microbial community structuring and metabolic functions is largely underexplored. Results: In this study, we investigated microbial community succession, phylogenetic and functional diversity, and metabolic capabilities during POM decomposition by particle-attached (PAM) and free-living microorganisms (FLM) under increasing hydrostatic pressures. Diatom-originated 13 C-labeled POM was used to incubate surface water microbial communities from the East China Sea (ECS) at pressures of 0.1, 20, and 40 MPa (megapascal). Our results showed that the PAM and FLM communities exhibited contrasting patterns and pressure-dependencies in diversity, richness, and evenness. Microbial assembly was governed predominantly by stochastic processes at low pressure and by deterministic processes at high pressure. Network analysis uncovered the non-randomly structured PAM and FLM communities and clusters of operational taxonomic units (OTUs) that reected different functional and ecological capacities of the subgroups. Metatranscriptomic analysis revealed that gene expression of known metabolic pathways (carbohydrate, amino acid, and energy production) varied greatly with pressure and between PAM and FLM. Furthermore, the FLM communities maintained higher metabolic activities than the PAM communities at high pressures, indicating the apparent difference in resource utilization capacity and ecological functions of PAM and FLM in different pelagic zones of the ocean. Overall, we demonstrated that marine heterotrophic microbial assemblage patterns were non-random; the PAM were crucial in community structuring, whereas the FLM played more important roles in POM decomposition in the deep. Conclusions: Our results provide detailed insights into and increased mechanistic understanding of the structuring and succession of microbial communities and metabolic functions associated with POM degradation in the pelagic ocean. concentration) so as to eliminate other factors (like nutrient dynamics) in affecting microbial succession, we were able to pinpoint the impact of hydrostatic pressure on the metabolic and ecological functions of the PAM and FLM communities. We showed that high pressure led to obviously decreased stochasticity in microbial community assembly, higher volunability of the PAM communities and more resilience of the FLM communities. Likewise, the PAM communities exhibited decreased transcriptomic responses to increasing pressure in carbohydrate, amino acid and energy metabolism, while the FLM communities showed sharply upregulated gene expressions in these same metabolisms, suggesting the ecologically more important roles for the FLM communities in POM decomposition and carbon cycling the deep ocean. Our ndings provide further ecological insights and increased mechanistic understanding on how hydrostatic pressure affected the metabolic specialization and ecological functions of members of the PAM and FLM communities.


Background
Marine microbial physiology and trophic styles have often been linked to life strategies, including microbial motility, gene expression and substrate acquisition, and ecological functions in the ocean. This linkage is increasingly recognized and documented for the ecological dichotomy of heterotrophic prokaryotes in the marine environment, the particle-attached (PA), and free-living (FL) microbes (PAM, FLM). PAM and FLM play important but different roles in the global ocean carbon cycle, particularly in the decomposition of particulate organic carbon (POC) and degradation of dissolved organic carbon (DOC) [1][2][3][4]. It has been shown that PAM and FLM had different morphological, physiological, and genomic characteristics. For instance, PAM cells are often found larger in size and have higher enzymatic and metabolic activities than their FLM counterparts in the surrounding water [5][6][7]. Additionally, some studies showed that the abundance and diversity of marine PAM were often less than those of FLM counterparts [6,8,9]. Other studies revealed that the PAM and FLM communities often differ phylogenetically in various marine environments [10][11][12][13].
However, these previous studies have been limited to alpha-and/or beta-diversity analysis of shallow water microorganisms, and do not determine the associations between PAM and FLM, and how changes in environmental conditions (e.g., temperature, pressure, and composition of organic matter in the water column) affect microbial lifestyles and metabolisms. In this regard, change in hydrostatic pressure is particularly important, as microbes attached to the descending particles would experience the direct effect of signi cantly changed pressure from surface water to the deep ocean [4]. Hydrostatic pressure affects bacterial physiology, metabolic activity, and carbon cycling in the deep ocean [14][15][16]. Previous studies showed that increasing pressure resulted in reduced bacterial cell numbers [17] and bacterial metabolic activity [18]. A recent DNA-stable isotope probing (SIP)-based study showed that only a subset of the microbes, particularly the rare bacterial taxa, were actively involved in POM decomposition and degradation [19].
In this study, we examined microbial interactions, community assembly, and shifting in metabolic capacities of PAM and FLM in POM (i.e., diatom detritus) decomposition and resource utilization under different hydrostatic pressures by combining the sensitive DNA-SIP technique and the informative metatranscriptomic sequencing. We hypothesize that the PAM and FLM communities are assembled via different community assembly processes, and hydrostatic pressure exerts signi cant effects on microbial metabolic capabilities. We further predict that PAM and FLM communities play a different metabolic role in POM decomposition, and their relative contributions varied at different hydrostatic pressures. This study aimed to test this hypothesis and determine how different microbes respond to pressure changes in community dynamics and metabolic functions.

Results
Microbial diversity, richness, and evenness A total of 1,033 OTUs were identi ed from 3,615,779 high-quality sequences from the ECS in-situ surface water (ISW), with 929 and 835 OTUs (731 shared OTUs) for communities of PAM and FLM, respectively.
The Good's coverage ranged from 99.8 to 100%, suggesting that the diversities of the microbial communities were well covered in this study.
Overall, the PAM and FLM communities showed contrasting patterns and pressure dependencies in diversity, richness, and evenness. Microbial diversity and richness showed a decreasing trend with growth pressure, whereas evenness exhibited a more varied pattern. In general, the FLM communities had higher alpha-diversity than the PAM communities at low pressure (0.1 MPa), and comparable values at high pressures (20 and 40 MPa) (Fig. 1). On the other hand, the PAM assemblages had higher evenness than the FLM communities at all pressures. However, the two groups had nearly the same species richness at all three pressures.
In this study, we de ne active communities as those with relative OTU abundance ≥ 1%, i.e., the abundance of an OTU retrieved in the 13 C-heavy DNA fraction minus the corresponding abundance in the 12 C-heavy DNA fraction after incubations (Additional le 1: Figure S1) [19,20]. The active PAM and FLM assemblages exhibited more contrasting differences in the three indices. For instance, compared to PAM, the FLM communities had higher alpha diversity, and the difference became even larger at high pressures. PAM showed much higher evenness than the FLM communities, especially at 20 MPa (Fig. 1). Overall, microbial diversity and species richness decreased precipitously with growth pressure, whereas evenness exhibited an increasing trend.
Venn diagrams showed that the PAM and FLM assemblages shared a high proportion of OTUs at the three pressures, from 63 to 73% for the entire communities (Additional le 1: Figure S4). Interestingly, the proportions of shared OTUs between the active PAM and FLM assemblages were much lower (14 to 27%) than observed in the total communities (Additional le 1: Figure S4). It is also instructive to observe that three distinct clusters were identi ed by NMDS analysis that corresponded to the microbial communities at 0.1, 20, and 40 MPa, irrespective of microbial lifestyles, in both the total and active communities (Fig.   3). This result suggests that the microbial communities were signi cantly in uenced by hydrostatic pressure (P < 0.05) (Fig. 3). However, slight differences were observed between the PAM and FLM communities at all three pressures (P > 0.05; Fig. 3). Microbial community assembly To determine the relative importance of stochastic vs. deterministic ecological processes in microbial community assembly, we calculated the taxonomic normalized stochasticity ratio (NST) [21] for PAM and FLM communities (Table 1 and Fig. 4). In microbial community assembly, deterministic processes mainly include biotic (competition and other biotic interactions) and abiotic factors (environmental ltering) that lead to species sorting; stochastic processes include neutral dispersal (immigration and emigration), drift (random birth and death events), and diversi cation [22]. Table 1 Variations of the normalized stochasticity ratio (NST) for the PAM and FLM microbial communities at 0.1, 20, and 40 MPa, respectively, after the addition of diatom detritus. The microbial composition tended to more stochastic (NST > 50%) or more deterministic (NST < 50%) based on the calculation of NST.  Fig. 4).

Ecological networks of PAM and FLM and their topological features
We sought to determine microbial associations based on network analysis. Six networks were constructed for the PAM and FLM communities. The constructed networks consisted of 274 OTUs, with nodes representing OTUs and links representing correlations (positive or negative) between OTUs (Fig. 5).
The overall topology indexes suggest that the networks were scale-free (R 2 = 0.43-0.81; Additional le 1: Table S2), implying that a few hub OTUs (taxa) in the networks were highly connected while most OTUs had a few connections [23,24]. All networks exhibited small-world features, as indicated by the higher average path distance (GD) and average clustering coe cient (avgCC) compared to the respective randomized networks (Additional le 1: Table S2), i.e., microbes were highly and e ciently connected in the small-world networks. Overall, nodes in both the PAM and FLM networks tended to be more negatively correlated, as indicated by the number of total positive/negative links, 570/620 and 279/602 for PAM and FLM networks, respectively. However, the number of positive links increased with hydrostatic pressure for both the PAM and FLM networks, and the negative links decreased with pressure for the FLM networks, and rst decreased and then increased for the PAM networks (Additional le 1: Table S2).
46 connectors were identi ed in the constructed networks (Additional le 1: Table S2). However, no network hubs were found for any of the constructed networks. Examining the topological properties shows that all three PAM networks had one or more module hubs (Zi ≥ 2.5, Pi < 0.62), whereas only one module hub (OTU930) was found in the FLM networks (0.1 MPa FLM). Furthermore, some of the identi ed module hubs were members of unclassi ed genera (Additional le 1:  Table S2).
Connectors were detected in all PAM and FLM networks, but the distribution and taxa compositions were rather different between the PAM and FLM networks. At pressures of 0.1 and 20 MPa, there were much more connectors in the FLM networks than in the PAM networks, and it is opposite at high pressure (40 MPa). Furthermore, the number of connectors identi ed in the FLM networks (35) was much more than that in the PAM networks (11). Thus, there were more module hubs in the PAM networks and more connectors in the FLM networks, suggesting the different ecological roles of the PAM and FLM communities (Additional le 1: Table S2).

Genetic repertoires and metabolic functions of PAM and FLM
Metatranscriptome sequencing analysis of 13 C-labeled RNA allowed us to detect 26 most abundant active genera (> 1% in the relative abundance of mRNA transcripts) ( Fig. 2b and Additional le 1: Table   S3-S4). In the PAM fraction, the most active microbial taxa were Alteromonas (38%) and Marinomonas (21%) at 0.1 MPa, Pseudoalteromonas (31%) and Alteromonas (8%) at 20 MPa, and Alteromonas (32%) and Tenacibaculum (19%) at 40 MPa. In the FLM fraction, the most active taxa were Amphritea (46%) and Pseudoalteromonas (23%) at 0.1 MPa; Amphritea (32%) and Pseudoalteromonas (27%) at 20 MPa, and Pseudoalteromonas (25%) and Vibrio (25%) at 40 MPa ( Fig. 2b and Additional le 1: Table S4). Thus, the active taxa identi ed based on metatranscriptome analysis are comparable to those detected by DNA-SIP, except for those at 0.  Table S5 (Additional le 1). The PAM and FLM communities differed in their genetic repertoires. First, the genetic machinery for fast growth varied distinctly between the PA and FL microbes. For instance, the FLM communities exhibited steadily increased expressions with pressure in genes related to genetic information processing such as, translation, DNA replication and repair, and ribosome biogenesis (Additional le 1: Figure S6 and Table S5). The PAM fraction, on the other hand, showed a more varied pattern in expressions of the same genes. Second, expressions of genes related with intake of extracellular compounds (i.e., membrane transport, signal transduction, cell motility proteins and the two-component system) showed the similar patterns as the genetic machinery described above, that is, increased expressions at high pressure (40 MPa) for the PAM as well as FLM assemblages (Additional le 1: Figure S6 and Table S5). Finally, both PAM and FLM showed in ated fractions of genes involved in protein folding stability, sorting and degradation, suggesting the global regulation of protein folding and tra cking in environmental adaptation and resource utilization at high pressures (Additional le 1: Figure S6 and Table S5).
Furthermore, the PAM and FLM assemblages differed signi cantly in the common metabolic processes, carbohydrate, amino acid, lipid, nucleotide, and energy metabolism.
Finally, corresponding to the genetic repertoires described above, agellin and aerobic respiration control protein ArcA were found to be highly expressed, mainly by Marinomonas and Pseudoalteromonas (Figs. 6 and 8). Among the cellular processes, we found two proteins, agellar hook-associated protein 2 and agellar basal-body rod protein FlgB, were highly expressed in agellar assembly, and the former was produced mainly by Amphritea and Marinomonas, found in both the PAM and FLM communities (Figs. 6 and 8).

Discussion
Microbial succession and community assembly processes under different pressures Our study revealed that microbial succession and community assembly processes were signi cantly affected by hydrostatic pressure. NMDS analysis illustrated that communities of both PAM and FLM at the same growth pressure clustered together, irrespective of microbial lifestyle (Fig. 3). Metatranscriptome sequencing results con rmed that these clusters were associated with different functional capacities (metabolism, information processing, and directional cellular tra cking), suggesting that these clusters were functionally and ecologically important subgroups at respective hydrostatic pressure (Figs. 2b and  6). Indeed, these subgroups were comprised of different taxa, phylogenetically and functionally, at different pressures (Figs. 2b and 6). Correspondingly, both PAM and FLM communities showed obviously less stochastic assembly, denser between-taxa associations, and higher network complexity under high pressures (20 and 40 MPa) than at low pressure (0.1 MPa), suggesting pressure-induced selection of taxa with speci c functions interacting with the biotic/abiotic environment.
More interestingly, PAM and FAM communities showed different responses to pressures. The PAM groups had consistently lower Shannon diversity than the FLM groups (Fig. 1), consistent with the notion that sinking particles in the ocean are carbon-and nutrient-rich microniches, explored and utilized predominantly by a few well adapted fast growers attached to the particles, the so-called r-strategists [4,27,28].
Our results further revealed that members of the PAM assemblages constituted most of the network module hubs, while members of the FLM communities served as most network connectors (Additional le 1: Table S2). These results indicate that the PAM assemblages played a more important role in structuring the networks and maintaining network stability, while the FLM taxa were mainly "communicators" in information processing and transfer in the networks. Surprisingly, most of the identi ed module hubs (44.4%) and connectors (50.0%) were unclassi ed at genus level and of low abundances, suggesting that taxa of the rare biosphere may play more important roles in structuring bacterial communities and that our knowledge is limited about the potentially important ecological roles of these taxa in the interacting communities of PAM and FLM in the ocean. This nding is similar to that observed in a freshwater lake [29].
Furthermore, the PAM networks had more positive associations than the FLM networks at all pressures, whereas the FLM networks had more negative links than the former at 0.1 and 20 MPa. Positive associations may include cross-feeding, co-aggregation in bio lms, co-colonization, niche overlap, whereas a negative relationship may result from amensalism, predation, and competition [23]. These results re ect the different lifestyles, or dichotomy, of marine microbes, that is, the PA microbes, as copiotrophs in the ocean, prefer the particle-attachment life strategy, for rich resources of carbon and nutrients in the descending particulates in the water column [4,5].
Microbial network topological properties indicate niche differentiation and differed microbial interactions under different pressures [30,31]. Network modules can be considered as niches or microbial functional units [32,33]. The steadily reduced modularity for the PAM networks (Additional le 1: Table S2) suggests decreased PA community segregation, whereas the increased modularity for the FLM networks re ects more niche separation of the FL communities into ner niches and functional units, perhaps around the particulates [4]. This increased compartmentalization would enhance the stability in networks of the FLM communities and increase population diversity [31,34], consistent with our ndings in alpha-diversity of the two groups of communities (Fig. 1). The avgK index indicates network complexity [31]. Compared to FLM networks, the PAM networks had relatively higher average degree (or connectivity) and higher modularity, suggesting greater network vulnerability for the PAM networks, particularly at high pressures.
We postulate that these observed differences in network stability and microbial interactions between PAM and FLM can be attributed to microbial physiology, trophic lifestyles, and life strategies in the ocean. For one, compared to FLM, PAM are supposedly more closely associated between each other in microniches of the particulates [5,35], and therefore, closer, and perhaps more e cient metabolic interactions [27,36,37]. For the other, the lower avgK, larger network sizes and higher microbial diversity of the FLM networks indicate greater network resilience or stability of the FLM communities [31].

Pressure-regulated metabolic activities of PAM and FLM
Hydrostatic pressure dramatically in uenced microbial metabolic activity in POM decomposition and utilization. It is remarkable that expression of genes involved in various metabolic processes such as carbohydrate, amino acid, nucleotide, and energy metabolism were downregulated with pressure for PAM, and upregulated for FLM.
The expression abundance of enzymes in glycolysis/gluconeogenesis decreased with pressure in the PAM fraction while increased with pressure in the FLM fraction. For example, the expressed abundance of phosphoenolpyruvate carboxykinase (ATP) that catalyzes ATP to ADP, and fructose-bisphosphate aldolase, class II that catalyzes the reversible reactions for transferring D-fructose 1,6-bisphosphate to glycerone phosphate and D-glyceraldehyde 3-phosphate in glycolytic/gluconeogenesis [38] decreased with pressure in the PAM fraction. In contrast, the proportion of this enzyme for the FLM fraction increased with pressure. Similar variation patterns of enzyme expressions affected by pressure were observed in the TCA cycle (Fig. 7). Furthermore, FLM expressed higher activity of succinyl-CoA synthetase beta subunit at 40 MPa than PAM in the TCA cycle, the FPKM of the enzyme for FLM at 40 MPa was more than eightfold of that at 0.1 MPa. Meanwhile, the FPKM of this enzyme for FLM at 40 MPa were about 3 times that of PAM at the same pressure ( Fig. 6 and Additional le 1: Table S5). These results suggest that FLM taxa maintained higher transcriptional and enzyme activity in carbohydrate metabolism than PAM at high hydrostatic pressures, and likely play a more important role in carbon cycling in the deep ocean [4]. This functional difference is probably related to microbial physiology and life strategies in the ocean, as discussed above. Our results are consistent with the notion that growth rates and metabolic activities of free-living microorganisms are higher than those of particle-attached microorganisms in the deep ocean [37].
Pressure effect on enzyme activity of amino acid metabolism showed the same variation pattern. For instance, the transcriptional abundance of the alanine dehydrogenase in the FLM fraction increased dramatically from 8 at 0.1 MPa to 1,448 at 40 MPa, while the corresponding values in the PAM fraction decreased from 33 to 25. Also, the FPKM of glycine dehydrogenase and s-adenosylmethionine synthetase decreased for PAM while increased for FLM with increasing pressure. These results indicate that the FLM communities maintain a stronger ability and higher activity than the PAM communities for the utilization of amino acids at high pressures.
F-type H + -transporting ATPase is a functional ATP synthase for ATP production, which is a key multisubunit enzyme for living microorganisms meeting their energy requirements [39]. The seven F-type H +transporting ATPase subunits, a, b, alpha, beta, gamma, delta, and epsilon related to the oxidative phosphorylation (ko00190) in energy metabolism also exhibited the aforementioned variation pattern. The FPKM of these enzymes for the FLM were upregulated sharply with pressure and downregulated with pressure for the PAM assemblage ( Fig. 6 and Additional le 1: Table S5). These observations supported the hypothesis that the FLM communities are more active in carbon and energy metabolic processes at high pressure than the PAM assemblages, and potentially play a more important ecological role in the deep ocean [19,37].
In summary, there was a clear discrepancy in gene expressions for enzymatic activities between PAM and FLM. This differential expression can be attributed to microbial interactions and niche overlap in community assembly, affected by hydrostatic pressure. As shown in Fig. 5 and Additional le 1: Table S2, the modularity of the PAM community networks decreased signi cantly with pressure, so was the total number of modules. However, the average degree, and avgK which represents the frequency of microbial interactions in networks [31], increased substantially with pressure. As modules can be considered as niches or microbial functional units [32,33], decreases in modules and modularity in PAM networks can be interpreted as reduced functional segregation or increased niche overlap, and therefore, enhanced microbial interactions and network complexity, in the PAM communities. The decreased niche differentiation is in accordance with the observed decrease in microbial diversity with pressure. This inference is also consistent with the variations in positive and negative links in PAM and FLM communities (Additional le 1: Table S2). The decreased negative associations with pressure for the FLM networks can be perceived as reduced competition, and therefore, increased utilization of POM.
Generalists or specialists, the evolving ecological roles of PAM and FLM with hydrostatic pressure in the pelagic ocean It is believed that most marine microorganisms are generalists, but with dual lifestyles [40], and therefore, able to grow in either free-living or attached-on-particle mode [41,42]. However, microbes can switch lifestyles leading to species sorting, depending on the environmental conditions [43,44]. Regardless, our results showed that hydrostatic pressure affected microbial life strategies and community assembly, which in turn in uenced microbial metabolic activity and ecological functions. This is clearly demonstrated by the vastly varied metabolic activities of Alcanivorax, Vibrio, and Marinomonas with growth pressure. The metabolic activity of particle-attached Alcanivorax decreased with pressure, while enzyme expression of the free-living Alcanivorax increased with pressure (Additional le 1: Table S6). For example, the FPKM of DNA-directed RNA polymerase subunit beta of PA Alcanivorax decreased with pressure, from 1,122 at 0.1 MPa to 156 FPKM at 40 MPa, whereas that of FL Alcanivorax increased sharply, from 399 at 0.1 MPa to 13,685 FPKM at 40 MPa. Similarly, the expression abundance of glyceraldehyde 3-phosphate dehydrogenase reduced from 267 at 0.1 MPa to 67 FPKM at 40 MPa for PA Alcanivorax, while the corresponding value increased sharply for FLM from 115 at 0.1 MPa to 2,112 FPKM at 40 MPa (Fig. 8). Additionally, it is well known that Alcanivorax can degrade different hydrocarbons [45], but are unable to decompose carbohydrates and amino acids [46]. Our data, however, revealed that Alcanivorax was able to express various enzymes in carbohydrate and amino acid metabolism. Thus, our results suggest that that the FLM communities likely play a more important role than the PAM communities in degrading recalcitrant organic matter in the deep ocean.

Conclusions
By combining DNA-SIP with metatranscriptome sequencing, we provided detailed insights into the community ecology and metabolic functions of particle-attached and free-living microbes in decomposition of diatom detritus with increasing hydrostatic pressure. This approach allowed us to piece together the community assembly processes, interactions between and the vital role served by, the PAM and FLM taxa. By exposing the same microbial communities to the same particulates (in composition and concentration) so as to eliminate other factors (like nutrient dynamics) in affecting microbial succession, we were able to pinpoint the impact of hydrostatic pressure on the metabolic and ecological functions of the PAM and FLM communities. We showed that high pressure led to obviously decreased stochasticity in microbial community assembly, higher volunability of the PAM communities and more resilience of the FLM communities. Likewise, the PAM communities exhibited decreased transcriptomic responses to increasing pressure in carbohydrate, amino acid and energy metabolism, while the FLM communities showed sharply upregulated gene expressions in these same metabolisms, suggesting the ecologically more important roles for the FLM communities in POM decomposition and carbon cycling the deep ocean. Our ndings provide further ecological insights and increased mechanistic understanding on how hydrostatic pressure affected the metabolic specialization and ecological functions of members of the PAM and FLM communities.

Seawater sampling
Seawater samples were collected from water depth of 15 m at the Eastern China Sea (30°39'48''N, 122°29'48''E) in September 2018 (Additional le 1: Figure S7). The corresponding geochemical parameters of the water samples are listed in Additional le 1: Table S7.
Triplicate 3 L of the seawater was ltered sequentially through 3.0 and 0.22 µm pore-size polycarbonate membrane (47 mm; Merck Millipore Ltd.) to obtain PAM and FLM, respectively [11] and the lters were preserved at -80°C. 13 C labeled diatom-derived detritus Diatom species Thalassiosira weiss ogii (strain CCMA-102) was selected for producing 13 C-labeled POM experiment. T. weiss ogii was cultured according to Liu et al. [19] (see details in Additonal le 2: Supplementary method).

DNA/ RNA extraction and SIP ultracentrifugation
After incubation, triplicate 350 mL of the incubation solution from 13 C-labeled or 12 C-control treatments were ltered as described above, and the lters were stored at -80°C for the molecular microbiological analysis.
The lters were rst cut into pieces and then transferred to 2 mL sterilized centrifuge tubes. The DNA and RNA were extracted with RNeasy PowSoil DNA Elution Kit (QIAGEN) and RNeasy PowerSoil Total RNA Kit The DNA-SIP gradient fractionation was conducted by following the method as previously described [19,49] (see details in Additonal le 2: Supplementary method).

Ecological Network Analysis
The ecological networks were constructed based on Molecular Ecological Network Analyses Pipeline (MENAP, http://ieg4.rccc.ou.edu/mena/). Only taxa detected in more than six samples were used for network analysis. Pearson correlation coe cients (r) between any two OTUs were estimated to generate the association matrix. The Random Matrix Theory (RMT) approach was used to determine the threshold of correlation coe cient to construct the network of non-random associations [25,26]. Network properties were calculated using MENAP as previously described [

Microbial community assembly processes
The relative importance of stochastic processes in microbial community assembly was quanti ed using the normalized stochasticity ratio (NST) [21]. The NST in this study was calculated based on Ružička dissimilarity and null model algorithm 'PF' (keep occurrence frequency proportional to observed values and richness the same as observed). The computation was performed using IEG statistical analysis pipeline (http://ieg3.rccc.ou.edu:8080) built on the Galaxy platform. NST can evaluate the community assembly as more deterministic (< 50%) or more stochastic (> 50%), with higher values indicating higher stochasticity.

Metatranscriptome Sequencing and Data Analysis
The removal of ribosomal RNAs (rRNAs) from the total RNA and construction of RNA-based library were performed by using the Ribo-zero rRNA Removal Kit (EpiCentre, WI, USA) and the TruSeq™ RNA Sample The raw sequences were trimmed to obtain high quality pair-end reads, by applying Seqprep (https://github.com/jstjohn/SeqPrep) and Sickle (https://github.com/najoshi/sickle). Open reading frames (ORFs) were predicted with TransGeneScan (http://sourceforge.net/projects/transgenescan/).
The non-redundant gene catalog was generated by using the CD-HIT (http://www.bioinformatics.org/cdhit/) based on the sequences with > 95% sequence identity and > 90% coverage. Reads were mapped to the representative genes with 95% identity and the FPKM (fragments per kilobase of transcript per million reads mapped) were assessed with RSEM (http://deweylab.biostat.wisc.edu/rsem/). Taxonomic annotations were performed by aligning non-redundant gene catalogs against NCBI non-redundant database with e-value of 1e − 5 , by applying BASTP (v2.2.28+; http://blast.ncbi.nlm.nih.gov/Blast.cgi). For functional annotation, KEGG annotation was conducted by aligning non-redundant gene catalogs against KEGG (Kyoto Encyclopedia of Genes and Genomes, http://www.genome.jp/kegg/) database, also by using BLASTP.    Variations of the normalized stochasticity ratio (NST) for the PAM and FLM microbial communities at 0.1, 20, and 40 MPa, respectively, after the addition of diatom detritus. The microbial composition tended to more stochastic (NST >50%) or more deterministic (NST <50%) based on the calculation of NST.