The estuarine microbiome networks are comprised of highly interconnected taxa, formed clustered topologies, and thus contained ‘small-world’ properties . Among the interconnected taxa, the abundant microorganisms contribute greatly to the biomass, nutrient cycling, and primary production [11, 78, 79]. However, the most abundant microbial groups are not those taxa that hold the network together, such as “hub species” or “gatekeepers” at both ASV and family levels (Fig. 1 and S2). Thus, abundant taxa may not be necessarily critical to the prokaryotic microbial network structure or their stability (Table 1 and S2). In this study, the hub species and gatekeepers are relatively low in abundance or belong to rare species (Table S2), but they play fundamental roles in network persistence and contribute greatly to the stability and resilience of these taxa-taxa networks . Recent studies have increasingly emphasized the ecological importance of the rare biosphere, because rare taxa can include more metabolically active microorganisms than abundant taxa (as measured by RNA to DNA ratios), and they may be keystone species in regulating the functioning of aquatic environments [81, 82]. In addition, the rare microbes have been shown to fulfill essential functions associated with nutrient cycling, and may enhance functionality of the abundant microbes . For example, it has been demonstrated that rare species offer the required gene pool to catalyze complex degradation processes of organic compounds, and some pollutants are often degraded by species falling below the detection limit in pristine samples .
Our results further corroborated the significance of “hub species” and “gatekeepers” via stability testing with removal processes. The responsiveness of network fragmentation to removal of nodes with highest betweenness centrality provide important insights into the susceptibility of prokaryotic microbial networks to disturbance . Our results suggest that the loss of those potential “gatekeepers” contributes disproportionately to network fragmentation, which essentially agrees with earlier reports on food webs and mutualistic networks showing high fragility/susceptibility upon selective removal of taxa [33, 70, 85]. Sequencing data allowed us identify that potential “gatekeepers” were affiliated with Actinobacteria (Sporichthyaceae, Microbacteriaceae, CL500-29 marine group and PeM15), Betaproteobacteria (Polynucleobacter), Alphaproteobacteria (Sphingorhabdus), Bacteroidetes (Cryomorphaceae), Planctomycetes (Pirellula), and Gammaproteobacteria (Legionella). These taxa had highest betweenness centrality values (937-1943) and were consistently present in the major component of the co-occurrence networks. The loss of “gatekeepers” may adversely affect the robustness and resistance of microbiome structure and associations [33, 70, 85]. Similarly, strongly interacting species (i.e., “hub species”) are important to CB microbial communities, and they were shown to be able to steer microbiome ecosystems towards certain community types . Additionally, as part of the microbial “seed bank”, rare taxa (including these high degree/betweenness centrality but low abundance species) can potentially drive ecosystem responses to environmental changes and become dominant under favorable conditions , therefore providing a mechanism for community persistence and stability .
Similar to microbiome structures [14, 54], microbial co-occurrence networks showed strong temporal and spatial patterns with distinct property and stability (Fig. 2-5, Fig S2-S4 and Table 1). Our results showed that consistent patterns of microbial networks were observed at different taxonomic levels. These differences are likely due to strong gradients in seasonality, salinity, nutrient availability, and other causal environmental factors [11, 18, 89, 90]. CB, the largest estuary in northeast America, experience typical seasonal changes and constant freshwater/oceanic water input and exchange. The dynamic estuarine gradients lead to large variations in bulk bacterial production and biomass [19, 20], and community composition [11, 22], and subsequently affecting the property and stability of microbial networks (this study, Fig. 2-6 and Table 1; [52, 91, 92]).
Warm season microbial networks revealed high number of edges, high average degree, hign graph density, low number of components and clusters, low fragmentation, high modularity, and high number of negative links. Stronger correlations between node-normalized degree and betweenness centrality and less variability of fragmentation after removing the potential 10 “gatekeepers” were also observed in the networks for summer and autumn (Fig. 4 and Table 1). These results indicated that microbial networks were more stable during warm seasons compared to those in winter and spring. Microbial co-occurrence affected by temporal variations were also found in Alaska Beaufort Sea coast , the San Pedro Ocean Time-series station , and the Lake Mendota . The higher number of nodes in the co-occurrence networks of warm season agrees with the previously reported high microbial diversity in the Bay [14, 54]. High biodiversity is able to promote co-occurrence between microbial communities , and these biotic associations including competition, are commonly thought to increase co-occurrence in microbial networks as they refer to common resources and environmental conditions [30, 95, 96]. In addition, increase in grazer richness has a positive effect on the bacterial richness and evenness , stemming primarily from the widespread distribution of resources and then resulting in ecological niche complementarity [98–100]. Finally, this stability is also due to long residence time and relatively high growth rate (and mortality) of different microbial populations [20, 90], which allows estuarine bacteria overwhelm those allochthonous populations from marine or freshwater . Therefore, stable prokaryotic microbial networks in warm season in the Bay may arise from a better balance of the microbial associations (e.g., mutual, competitive and prey) and nutrient availability during summer and autumn than winter and spring [11, 20, 102]. In contrast, microbial associations in cold seasons contained high fragmentation and it may be reinforced by stronger nutrient limitations and lower growth rates (due to low temperature) .
Compared to the upper Bay, the network in the lower Bay was more connected with more negative correlations (Fig. 3, Table 1). High percentage of negative links in the networks could stabilize co-oscillation within microbial communities and increase network stability . A strong association between node-normalized degree and betweenness centrality and the less variability of fragmentation after removing potential “gatekeeper” nodes were also observed in the network of lower Bay (Fig. 4, 5). Our results suggested that the stability of prokaryotic microbial network was higher in the lower Bay compared to the upper Bay. The stability difference between upper and lower Bay could be due to the disturbance and interference from freshwater vs. oceanic water. Similar environmental variations were also found in many other estuaries, and spatial variation also could affect bacterial associations, including Ems estuary , Hangzhou Bay [43, 53] and Pearl River Estuary . River input, land runoff, suspended particles/turbidity, and accompanied nutrients availability are all pulsating with strong seasonal/inter-annual variations and uncertain patterns in the upper Bay. Conversely, salty water intrusion from the North Atlantic Ocean can be relatively consistent, and we thus hypothesize that less temporal disturbances from the ocean contribute to higher stability of microbial networks in lower Bay compared to the upper Bay.
Estuarine microbiomes continuously experience environmental perturbations including pulse, press and environmental stochasticity . This study was not originally designed and intended to investigate community recovery after defined perturbations such as pulse (e.g., floods) or press (lasting disturbances such as climate changes), therefore the results are more reflecting the ecological stability of microbiome correlations under the perturbation of environmental stochasticity . Ecological stability is a multidimensional concept that captures different aspects of the dynamics of the ecosystem and its response to perturbations . Pimm (1984) summarized five components of ecological stability that are in common use: asymptotic stability, variability, persistence, resistance, and resilience . In this study, multiple metrics in network topology including the number of components, modularity, negative correlations, fragmentation as well as the correspondence between node-normalized degree and betweenness centrality are used to characterize the asymptotic stability and variability, both of which indicate higher network stability in warm season than cold seasons, and lower Bay than upper Bay. However, due to the high correlations between metrics used, it is difficult to propose a single best way of selecting metric(s) in evaluating the independent stability components. For instance, negative correlations between metrics would suggest trade-offs, although they seem to be rare exceptions in complex trophic communities . Further, the choice of the metrics always depends on the system studied and practical constraints, therefore measuring more stability metrics with deep and hierarchical analyses and/or explained variance analyses can help making informed choices and improve our understanding of the full-picture of microbial stability in the Bay for the future studies.
Due to the continuous environmental perturbations, the estuarine microbiomes and their distribution are a comprehensive output of the bio-associations between microbial populations and the response to the surrounding environmental gradients. Our results show that shifts of environmental factors (including temperature, salinity, TSS, nitrogen, phosphorus, and PC) have strong effect on microbial community structure and networks. The shifts within the spatiotemporal variations seem to favor strong co-occurrence patterns (lower fragmentation and higher stability) of prokaryotic microbial associations, implying elevated biotic co-occurrence or species sorting strongly mediated by the local environment [96, 109]. Physiologic predisposition and nutritional tolerance of microbiomes tend to maintain stable communities inter-annually during certain seasons [11, 25, 110]. Focusing on the identified prokaryotic microbial networks and their responses to environmental variations could provide us valuable insights into the microbiome adaptation and habitat partitioning/preferences in the Bay across seasonal changes and spatial gradients [30, 67, 111]. In addition, co-occurrence networks reveal critical information on co-oscillation between microbial taxa and also the stability of these involved communities in the Bay [30, 36, 111]. Therefore, changes in estuarine microbial networks resulting from disturbances provide a potential to examine the legacy effect on estuarine microbiome population structure, ecological function (e.g., contribution to primary productivity, food webs and economical sustainability) and its vulnerability to future disturbances, such as anthropogenic influence (e.g., eutrophication, contamination and damming) and climate change (e.g., drought and flood) [11, 31, 111].
To the best of our knowledge, this is the first systematic and thorough study of the microbiome co-occurrence associations, their stability, and the corresponding environmental factors in an estuary with long residence time, the Chesapeake Bay. There are some limiations and caveates in this study. For instance, most environmental parameters in the current study were not from the on-site samples that collected during the cruises. Instead, we used the data from the long term Chesapeake Bay Program’s (CBP) Water Quality Database (https://www.chesapeakebay.net/what/downloads/cbp_water_quality_database_1984_present). Plenty of previous studies and our own observations have shown clear and repeatable spatiotemporal patterns in the Bay [112–114]. These stable patterns are reoccurring annually and spatially, and therefore the CBP data closest to our sampling sites and dates were used in our analysis. In addition to the environmental gradients included in this study, many other abiotic or biotic associations also play critical roles in microbiome composition and distribution. For example, grazing and viral lysis are important factors that may affect the associations and stability of estuarine microorganisms [115–117], which deserve future investigations. Currently, we are analyzing the structure of microeukaryotic communities in the Bay, which may provide further information on primary producers and their interactions with other estuarine microbiomes. All in all, examining the natural spatiotemporal changes and stability of microbial networks as well as their interactions with other organisms is essential to strengthen the understanding of estuarine ecosystem processes.