Skip to main content
mSphere logoLink to mSphere
. 2022 Jun 1;7(3):e00918-21. doi: 10.1128/msphere.00918-21

Biological Microbial Interactions from Cooccurrence Networks in a High Mountain Lacustrine District

Vicente J Ontiveros a,b,#, Rüdiger Ortiz-Álvarez a,#, José A Capitán b,c, Albert Barberán d, David Alonso b, Emilio O Casamayor a,
Editor: Michael J Imperialee
PMCID: PMC9241510  PMID: 35642514

ABSTRACT

A fundamental question in biology is why some species tend to occur together in the same locations, while others are never observed coexisting. This question becomes particularly relevant for microorganisms thriving in the highly diluted waters of high mountain lakes, where biotic interactions might be required to make the most of an extreme environment. We studied a high-throughput gene data set of alpine lakes (>220 Pyrenean lakes) with cooccurrence network analysis to infer potential biotic interactions, using the combination of a probabilistic method for determining significant cooccurrences and coexclusions between pairs of species and a conceptual framework for classifying the nature of the observed cooccurrences and coexclusions. This computational approach (i) determined and quantified the importance of environmental variables and spatial distribution and (ii) defined potential interacting microbial assemblages. We determined the properties and relationships between these assemblages by examining node properties at the taxonomic level, indicating associations with their potential habitat sources (i.e., aquatic versus terrestrial) and their functional strategies (i.e., parasitic versus mixotrophic). Environmental variables explained fewer pairs in bacteria than in microbial eukaryotes for the alpine data set, with pH alone explaining the highest proportion of bacterial pairs. Nutrient composition was also relevant for explaining association pairs, particularly in microeukaryotes. We identified a reduced subset of pairs with the highest probability of species interactions (“interacting guilds”) that significantly reached higher occupancies and lower mean relative abundances in agreement with the carrying capacity hypothesis. The interacting bacterial guilds could be more related to habitat and microdispersal processes (i.e., aquatic versus soil microbes), whereas for microeukaryotes trophic roles (osmotrophs, mixotrophs, and parasitics) could potentially play a major role. Overall, our approach may add helpful information to guide further efforts for a mechanistic understanding of microbial interactions in situ.

IMPORTANCE A fundamental question in biology is why some species tend to occur together in the same locations, while others are never observed to coexist. This question becomes particularly relevant for microorganisms thriving in the highly diluted waters of high mountain lakes, in which biotic interactions might be required to make the most of an extreme environment. Microbial metacommunities are too often only studied in terms of their environmental niches and geographic barriers since they show inherent difficulties to quantify biological interactions and their role as drivers of ecosystem functioning. Our study highlights that telling apart potential interactions from both environmental and geographic niches may help for the initial characterization of organisms with similar ecologies in a large scope of ecosystems, even when information about actual interactions is partial and limited. The multilayered statistical approach carried out here offers the possibility of going beyond taxonomy to understand microbiological behavior in situ.

KEYWORDS: cell-cell interaction, microbes, networks

INTRODUCTION

The occurrence of specific microbes in natural environments is associated with multiple processes. At the spatial scale, geographic distance may create an effect of dispersal limitation (1), but most importantly, the local environment exerts strong filters facilitating the settlement of microbes with similar environmental preferences (2). A good example is the effect of pH on freshwater ecosystems, where a regional pool of microbes distributed along the hydrographical network is filtered according to water alkalinity, among other environmental reasons (3, 4). Although the environment has been extensively studied under the lens of community ecology as a driver of community assembly (4, 5), environmental factors or even dispersal limitation can be incomplete predictors of joint microbe occurrence (6). There is often a lack of understanding about the direct interrelationships among microbes—i.e., biological and ecological interactions—and how these are crucial in regulating the appearance or abundance of species in complex communities (7, 8). Unfortunately, there is a lack of knowledge about the myriad of interactions (9, 10) or ecological roles (9, 11, 12) of many community members. Microbial ecology has traditionally studied microbial interactions by analyzing small sets of species in controlled environments such as in cocultures (13) and biofilms (14), as well as population-level responses through quorum sensing (15) or particular interactions such as competition (7). For example, this issue was easily observed in microeukaryotes under a microscope, spotting the parasitic behavior of aquatic fungi toward phytoplankton (16) or the predatory effect of zooplankton toward phytoplankton (17). Still, there are potentially many specific and unspecific interactions that have not yet been detected but that may be particularly relevant in organizing biological communities. So, when direct observation of interactions is difficult, ecologists may rely on computational methods to infer potential interactions. For example, signals of competitive interactions have been suggested in the case of macroorganisms (18).

Recently, amplicon sequencing through high-throughput technologies has allowed microbial ecologists to study multiple species thriving in their natural communities. Through the inference of cooccurrences and coexclusions, many studies have attempted to unveil why some species tend to occur together in the same locations, while others are never observed sharing the same ecological space (1921). These associations and network properties linked to the association’s structure have been traditionally hard to interpret, since occurrences based on spatial associations do not solely account for biotic interactions and usually blend patterns of interactions and environment (22, 23). To decipher the causes acting on associations (cooccurrences and coexclusions) and infer potential biotic interactions, the study of microbial associations should be conducted within a defined metacommunity (24), i.e., a set of local communities, linked by dispersal, of potentially interacting species (25). A good model system would be one where local communities have some degree of isolation to display specific environmental filters but still are connected enough to be part of the same metacommunity. For this reason, in this study, we used a high mountain lacustrine district (Pyrenees, Spain) with 224 sampled lakes (local communities) comprising a regional metacommunity. In lacustrine districts, each lake acts as a local community, isolated by land but closely influenced by regional processes such as atmospheric deposition (26), catchment processes (27), and water flows (3), all mobilizing the regional species pool (28). Although interactions occur anywhere and can be extremely varied, in habitats with a strong nutrient limitation, such as in the highly diluted waters of high mountain lakes, food-web components and functional interactions are the microbial engine responsible for nutrient mobilization (2932).

The purpose of this study was 2-fold. First, to quantify the drivers of species associations in a high-altitude lacustrine microbial metacommunity, and second, to define potential microbial interaction units and explore their properties. To do so, we identified microbial associations caused by both environmental niches and dispersal limitation, interpreting the unexplained associations as potential biotic interactions. This approach allowed us to extract several layers of information, starting by describing the environmental variables that were primarily linked to taxon associations and ranking the influence of these factors to explain cooccurrence and coexclusion pairs. Once the importance of explanatory variables on association pairs had been quantified, we examined specifically the pairs interpreted as species interactions. By integrating pairs into a network analysis, we sought to describe different groups of species interactions on the basis of network properties and modules. We classified taxa into clusters of more-positive associations and less-negative associations to find groups of species based on interactions (“interacting guilds”) and keystone taxa organizing each group. Targeting these keystones and their guilds in future studies adds a powerful new layer of information based on interactions to the traditional approaches based on taxonomy or functional components in tag-sequencing studies.

For this purpose, we applied the combination of a probabilistic method for estimating significant cooccurrences/exclusions and a conceptual framework for filtering out associations potentially linked to environmental and/or spatial factors.

RESULTS AND DISCUSSION

In the present study, we applied a combination of tools to tell apart biotic interactions from associations driven by niche differentiation or geographic dispersal limitation, using a large amplicon high-throughput sequencing data set. We first inferred significant associations by combining a probabilistic method, followed by variance analysis, to separate the species pairs that can be explained by environmental variables or geographic coordinates. This computational method allowed us to rank the influence of environmental factors and to identify the pairs interpreted as species interactions. By integrating these pairs into a network analysis, we described different groups of species interactions on the basis of network properties and modules, as well as keystone taxa organizing each group. Targeting these unseen keystones and their guilds added a new view based on interactions to the traditional approaches mostly based on taxonomy or functional components in tag-sequencing studies.

Quantifying environmental niches and isolating potential biotic interactions.

The applied approach identified microbial associations caused by both environmental niches and dispersal limitation, interpreting the unexplained associations as potential biotic interactions. The probabilistic method for the calculation of co-occurrences retrieved a total of 52,143 significant cooccurrences and 25,675 coexclusions, of 661 bacterial and 265 eukaryotic zero-radius operational taxonomic units (zOTUs) with 427,812 potential combinations. These pairs harbor significant associations primarily reflecting environmental (analysis of variance [ANOVA] test: 83.34% of pairs) and dispersal (multivariate ANOVA [MANOVA] test: 4.14% of unique pairs) processes. Indeed, cooccurrences have been often used before to show the ecological clusters based on environmentally driven edges (33). Hence, we quantified the environmentally driven edges by sorting the environmental variables based on the unique proportion of cooccurrences and coexclusions explained for each variable. The analysis showed distinct rankings for bacteria and microbial eukaryotes (Fig. 1). In general, environmental variables explained fewer pairs in bacteria than in microbial eukaryotes, with pH alone explaining the highest proportion of bacterial pairs, and leaving 8,202 unexplained pairs by any variable (Fig. 1a and c). Dispersal limitation alone, stood out in explaining ~10% of bacterial coexclusions but did not explain a high proportion of the cooccurrences (~2%). In Eukarya, cooccurrences were mostly explained by NO3 and SO42+ (Fig. 1b) and coexclusions by pH (Fig. 1d). Interestingly, most eukaryotic pairs were explained by environmental and geographic variables, leaving only 212 pairs unexplained. With regard to the unexplained pairs, there are examples in the literature where these precise associations have been explored to search for ecological interactions at the community and regional levels (34, 35), which highlights the potential of the approach. In total, our analyses left 9,749 pairs unexplained (12.52% of the total, 7,017 cooccurring, 2,732 coexclusions), which may be interpreted as potential biotic interactions.

FIG 1.

FIG 1

Proportions of cooccurrence and coexclusion pairs (Bacteria-Bacteria [a and c] and Eukarya-Eukarya [b and d]) estimated using the probabilistic approach, explained by the environment and the geography after conducting ANOVA and MANOVA tests, respectively. Environmental variables were ranked based on the cumulative proportion of links not previously explained by any other environmental variables (black line).

Explanatory variables of associated pairs according to their taxonomic rank.

When combining explanatory variables with taxonomy there was an apparent equilibrium at high taxonomic ranks (i.e., class), with taxa similarly explained by the different variables (Fig. 2). However, there were higher differences between bacterial groups at the genus level (see Fig. S1 and Table S1 in the supplemental material). We did not narrow down to the genus level in eukaryotes because of classification limitations based on the short ribosomal region used (V9, in the 18S-rRNA) (36). In all the cases, we used a chi-square test of independence to analyze the contingency table of the absolute associations between explanatory variables and microbial taxa to explore significant associations (P < 0.0001). The pH was the variable explaining more associations both in Bacteria and in Eukarya and promoted the networks with the highest clustering coefficients. It has been shown that pH is one of the strongest predictors of microbial communities in natural ecosystems such as soils (33, 37), and freshwater ecosystems (3, 4). However, there were no differences regarding how pH explained associations for each high-rank taxon, except for a lower proportion of coexclusions involving Betaproteobacteria. In addition, narrowing down at bacterial genus level (see Table S1), there were PAM k-medoids clusters with more pairs (K2) explained by pH, or with less pairs (K3), showing in turn an increase of pairs explained by nutrient variables (DOC, NO3, and total phosphorous [TP]). Indeed, nutrients were the second most important group of variables explaining association pairs, particularly in Eukarya.

FIG 2.

FIG 2

Proportions of variables (environment, dispersal, or potential species interaction) explaining cooccurrences and coexclusions across taxonomic groups. A single pair constitutes two nodes with the same or different taxonomy; hence, a pair could contribute to two different taxa. The bar plot displays the dominant bacterial classes and ecologically relevant eukaryotic groups, and the number of pairs by taxa subject to an explanatory variable. The asterisks (*) indicate the highest contributors to the significant association between bar and explanatory variables according to a chi-square test.

FIG S1

(a) Bar plot displaying the proportions of variables (environment, dispersal, or potential species interaction) that explain cooccurrences and coexclusions across taxonomic bacterial genera, according to the results obtained using the method of Blois et al. (22). A single pair constitutes two nodes with the same or different taxonomy; hence, a pair could contribute to two different taxa. The bar plot displays the groups of genera according to a partitioning around medoids clustering, relying on the proportion of explanatory variables over the pairs where each genus was participating. (b) Contribution in percentage of the Pearson residuals to the total chi-square value of cooccurrences (left) and coexclusions (right). Download FIG S1, PDF file, 0.2 MB (188.2KB, pdf) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

TABLE S1

Bacterial genera grouped according to PAM k-medoids clustering. Download Table S1, DOCX file, 0.01 MB (13.7KB, docx) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

Although small differences were observed in Bacteria, it is worth to mention that in Flavobacteria cooccurrences, dissolved organic carbon (DOC) explained a higher proportion of pairs, agreeing with previous findings where flavobacterium-like populations were favored during periods of high heterotrophic activity with a high availability of resources (38). Sulfate was also an important explanatory variable of association pairs, particularly in eukaryotic cooccurrences, but scarcely in their coexclusions. Despite not explaining a lot of pairs in Bacteria, there were more pairs explained by sulfate when the pair involved Armatimonadia or the K7 cluster, than in other taxa.

Other variables that to a lower extent may form niches are water-renewal time (RWT), explaining more cyanobacterial pairs, SRP, explaining more Dictyochophyceae and Ciliophora associations, and the K4 PAM k-medoids bacterial cluster, and DRSi, explaining more Cryptomonadales associations. Altitude, as a proxy of temperature and UV radiation, seemed to explain more pairs involving Alphaproteobacteria and Acidimicrobia than in other taxa.

Assuming that we covered the most important environmental variables, as we had shown in previous studies using the same data set (4, 39), we argue that the remaining pairs could be interpreted as potential positive and negative biotic interactions, which are differently quantified and balanced depending on the taxonomic affiliation (see Fig. S2). Although we cannot rule out unmeasured environmental variables or other factors explaining some pairs, the method selected a reduced subset of pairs with the highest probability of species interactions. In fact, microbes that potentially interact, significantly reached higher occupancies than those that do not interact (Bacteria: P < 0.031, Eukarya: P < 0.015) (Fig. 3, right panel). Most importantly, there were significant abundance differences between those microbes that potentially interact (with lower mean relative abundances) and those that do not interact (Wilcox test, Bacteria: P < 0.001, Eukarya: P < 0.013). Indeed, this result has previously been observed in protist experiments (40) and has been explained by the carrying capacity hypothesis (41), in which the maximal abundance is given by the system resources and its idiosyncrasies. Also, microbes that cooperate with many other microbes need to share the available resources while limiting their abundances. This trend reflects the previous observation that niche-expanding positive nontrophic interactions (i.e., facilitation, mutualism, or commensalism) (42) seem to be overrepresented in the pool of inferred potential interactions (43).

FIG 3.

FIG 3

Mean regional relative abundances and regional occupancy between interacting zOTUs or noninteracting zOTUs in bacterial and microbial eukaryotes.

FIG S2

Summary of estimated interactions by taxonomic groups in Bacteria (a) and Eukarya (b), and interdomain (c). Raw counts, without weights. (a) In terms of taxonomic groups, in Bacteria, members of Burkholderiales (Betaproteobacteria), Sphingobacteriales, Flavobacterales, and Cytophagales (Bacteroidetes), Caulobacterales (Alphaproteobacteria), and Frankiales (Actinobacteria) held the most interactions. (b) Except the orders Pseudomonadales (Gammaproteobacteria) and Oligoflexales (Deltaproteobacteria), all orders had more positive than negative potential interactions. In the case of Eukarya, most potential interactions were quantified in the dominant group Chrysophyceae, followed by Chytridiomycota, Dictyochophyceae, Cercozoa, and Dinoflagellata. Only Cryptomycota had more negative than positive potential interactions, and Tracheophyta only was involved in negative pairs. (c) Bacterial relationships with eukaryotes were concentrated in Chrysophyceae and then in Dinoflagellata, Chytridiomycota, and Cercozoa. Interestingly, 9 groups (of 17) had more negative than positive interactions: Tracheophyta, Chlorophyceae, Bicosoecida, Cryptomycota, Dikarya, Cryptomonadales, Dictyochophyceae, Peronosporomycetes, and Cryptomycota. Download FIG S2, PDF file, 0.1 MB (86.8KB, pdf) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

Overall, after separating the cooccurrences and coexclusions that could not be explained by overlapping environmental niches or geographic dispersal, the resulting networks were closer to an ecological network rather than a simple associational network, integrating multiple types and effects of ecological interactions, direct and indirect, taking place across spatial scales from regional to local. In these networks, the internal ecological structure should harbor subgroups of organisms based on an ecological criterion. Although there is some ecological coherence in high-rank bacterial taxa (44), defining alternative groups with a common ecology rather than those with a common phylogeny (45) may be more relevant in terms of biological interactions.

We arranged positive and negative potential biological interactions into domain networks that were subsequently divided in three modules both in bacteria (B1, B2, and B3, respectively, spinglass modularity = 0.31) and in eukaryotes (E1, E2, and E3, respectively, spinglass modularity = 0.53). Using modularity to “separate” subcommunities within a metacommunity network is a common strategy (4648), often assumed to divide a metacommunity network into theoretical community types that would be naturally found in the local samples. These modules hold structural characteristics that can be objectively measured such as diameter, clustering coefficient, average path length, and the average hub value of their nodes (Table 1), weighted degree distributions (see Fig. S3), between-module arrangements (see Fig. S4), or properties quantified at the local community level such as richness and abundance per lake, among others (see Fig. S5). These modules promote within-module cooccurrences, and intermodule coexclusions. These groups can be considered “interacting guilds,” according to a very broad definition: “a group of species that are similar in somewhat that is ecologically relevant, or might be” (45). Since there is not a single definition of “guild,” the criterion we use here to characterize guilds was to separate groups of microorganisms that potentially interact more than expected by chance.

TABLE 1.

Properties of network modulesa

Module No. of nodes Diam Clustering Avg path length Mean Hub value (SD)
B1 268 26.93 0.10 2.81 0.01 (0.02)
B2 186 23.10 0.19 2.54 0.05 (0.09)
B3 189 18.35 0.35 2.13 0.18 (0.19)
E1 54 15.23 0.38 1.63 0.01 (0.02)
E2 52 42.67 0.04 3.38 0.02 (0.13)
E3 26 68.46 0.12 5.54 0.00 (0.00)
a

In eukaryotes, properties exclude five unconnected pairs. The number of edges and the distribution within and between modules is presented in detail in Fig. S4. Additional module metrics are available in Fig. S5.

FIG S3

Density distribution of node weighted degree, displayed for the modules in the network of bacteria (a) and in the network of eukaryotes (b). The mean value of nodes from each module was drawn with a dashed line. Download FIG S3, PDF file, 0.1 MB (90.9KB, pdf) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

FIG S4

Number of edges between and within modules of the Bacteria network (B1, B2, and B3) and the Eukarya network (E1, E2, and E3) and interdomain relationships based on domain-specific modules. Download FIG S4, PDF file, 0.1 MB (68.3KB, pdf) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

FIG S5

Local community quantification of network metrics: module richness/lake (number of zOTUs within a module that are present in each individual sample), module completeness/lake (proportion of nodes from each module that is present in each individual sample, regarding the total number of nodes of each module), module abundance/lake (the total relative abundance of the nodes corresponding to each module), and total node weight/lake (sum of node weighted degree by each module and each lake). Download FIG S5, PDF file, 0.1 MB (97.5KB, pdf) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

Understanding interaction guild components and keystone taxa.

A look in detail for Bacteria showed different properties in the three guilds (Fig. 4a and b; see also Fig. S3 and S5): (i) B1, a low-abundance module with a degree distribution skewed toward negative values, dominated by Alpha-, Beta-, and Gammaproteobacteria, from soil, sediment, and biofilm habitats; (ii) B3, a high-abundance module with a degree distribution skewed toward positive values, dominated by Actinobacteria, Sphingobacteria, and Opitutae, from aquatic habitats; and (iii) B2, with intermediate characteristics between B1 and B3. Interestingly, according to the environmental ontology annotations of their nodes, the interacting bacterial guilds may be related to habitat and microdispersal processes (i.e., aquatic versus soil microbes). This phenomenon has been recognized as an important structuring factor in freshwater ecosystems (49). In the case of Actinobacteria from module B3, it could be a good example of a potential keystone taxa that deserves further investigations. Actinobacteria have actinorhodopsins (50, 51), which may function as energy generation for the trophic chain in ultraoligotrophic alpine lakes. Hypothetically, Actinobacteria may act in these systems as a central community component where other organisms may arrange around to form an interacting guild.

FIG 4.

FIG 4

Bar plots showing the relative abundances of bacterial taxonomic groups (a) and bacterial associated EnvO terms (b) and the relative abundances of eukaryotic taxonomic groups (c) and nutrition strategies (d) for the different biotic-driven modules (“interacting guilds”) found.

In the case of Eukarya, the interacting guilds were split into three categories, as follows: (i) E1, high-abundance mixotrophs with high total weights and an extended degree distribution; (ii) E2, low-abundance parasites and osmotrophs such as Peronosporomycetes, Chytridiomycota, and Diatomea, exhibiting a peaked degree distribution skewed toward negative degrees; and (iii) E3, showing the lowest abundance and total node weights, characterized by Cercozoa and Chrysophyceae, and displaying a degree distribution skewed toward positive values (Fig. 4c and D; see also Fig. S3 and S5). Each of these guilds has taxa whose ecologies are related to their interactions, and module E2 (fungal parasites) might be particularly relevant. The role of fungi in freshwater ecosystems has been highlighted as a modulator of dynamics by releasing nutrients through the degradation of organic matter, although their roles are still being resolved (9, 16). Given that module E2 has most negative interactions with modules E1 and E3 (phytoplankton) but also curiously harbors positive interactions with Diatomea, we wonder whether diatoms can properly manage nonlethal highly specific infections by parasitic/saprotrophic fungi in high mountain lakes, a question meriting further study. Overall, these results strongly suggest that the unveiled bacterial and eukaryotic interacting guilds found have differences based on ecological categories.

Limitations and opportunities of the approach.

We carried out an approach that certainly has the potential to tease apart the influence of the environment and dispersal limitation on community assembly in microbial systems but that also has some limitations that must be acknowledged to avoid interpretations based on false premises. Methodologically, we dealt with presence-absence data and not real abundances. This limitation may obscure the interaction-abundance relationships, but presence-absence has the advantage that it may detect nonlinear relationships that pure correlation methods cannot detect. In addition, there might be other ecological processes that drive joint species occurrence, such as particular dispersal mechanisms taking place in freshwater environments, without actually interacting. For instance, habitat dispersal at the catchment scale in the bacterial component (4, 49) is not easy to isolate, and may add noise to the filtered potential interactions. Also, we might have missed relevant environmental variables that could explain some of the species pairing. However, it seems unlikely that new variables could explain a significant portion of the potential interactions, since additional variables incurred in decreasing percentages of previously unexplained pairs (Fig. 1) and the selected variables already represent the environmental variation at a regional scale. Despite these uncertainties, unexplained pairs are good candidates to correspond to actual biotic interactions and may add helpful information to guide further efforts for a mechanistic understanding of microbial interactions in situ. By narrowing down potential biological interactions with association networks there is the possibility of discovering new interactions that were missed by traditional methods. In our study, we highlight the association networks with which taxa are most likely to establish interactions.

Concluding remarks.

We unveiled here potential keystone taxa for the defined interacting units and showed that even phylogenetically related organisms behave very differently in terms of abundance and interaction strength according to their assigned role. Understanding the interaction of organisms within these metacommunities is the basis for unveiling the processes behind ecological organization. Microbial metacommunities are too often only studied in terms of their environmental niches and geographic barriers due to inherent difficulties that make it harder to quantify biological interactions and their role as drivers of ecosystem functioning. Our study highlights that distinguishing potential interactions in both environmental and geographic niches may help in the initial characterization of organisms with similar ecologies in a large scope of ecosystems, even when information about actual interactions is partial and limited. The multilayered statistical approach carried out here offers the possibility of going beyond taxonomy to understand microbiological behavior in situ.

MATERIALS AND METHODS

Data sets and metadata.

The metacommunity was composed of lake surface plankton samples that were taken during summer of 2011 along the Pyrenees mountain range (4). Most of these lakes lie within ultraoligotrophic/oligotrophic ranges (52), where nitrate-phosphate imbalances caused by atmospheric depositions (53) may alter significantly the phytoplanktonic behavior and trophic status (54). However, pH is often the strongest segregator of community types (5557). DNA was extracted, amplified and sequenced with standard methodology as previously described (4, 58), using a subset of 224 lake samples where both Bacteria and Eukarya were successfully sequenced. Briefly, high-speed multiplexed rRNA gene sequencing with the Illumina MiSeq System was carried out using the primers 515f and 806r (59) for the bacterial 16S V4 region and the primers 1391f and EukBr for the eukaryal 18S V9 region (60). Illumina paired-end raw data were processed and quality filtered using UPARSE (61), and zero-radius OTUs (zOTUs) were recovered with the UNOISE algorithm (62). To avoid a sequencing depth bias and to homogenize the presence/absence probabilities of zOTUs, we first rarefied both bacterial and eukaryotic communities to 5,000 sequences per sample. Then, we kept these zOTUs with a minimum occurrence of 10% (24 lakes). Each zOTU was annotated with its mean relative abundance in the data set, its dominant Environmental Ontology (EnvO) affinity in the case of bacteria through the SeqEnv pipeline (63), and its primary nutrient assimilation strategy in the case of eukaryotes (39), respectively. Taxonomy was then inferred using the SILVA 128 database through the SILVA-NGS online portal (64). Raw data are fully available under BioProject PRJNA413654.

In addition, we specifically selected environmental variables for the 224 samples that broadly represent the data set: pH, DOC, nitrates (NO3), total phosphorus (TP), soluble reactive phosphorus, dissolved reactive silica, sulfate (SO42–), water renewal time, and altitude as a joint proxy of temperature and UV radiation, as previously reported (4).

Inferring potential interactions.

We used a modified version of the framework of Blois et al. (22) that assesses causes of species associations controlling for environmental and spatial factors. The method identifies aggregation, randomness, or segregation of pairs of species across a landscape and then estimates whether the deviations from randomness can be attributed to environmental filtering and/or dispersal limitation. While the original framework uses C scores to quantify the significance of species cooccurrences and relies on matrix randomizations, we opted for Veech’s probabilistic method (65) to carry out this task. This probabilistic method assumes that the probability of a species occurrence at a site is equal to its observed frequency among the set of sites considered, allowing for the calculation of the probability of cooccurrence between two species at a given number of sites under the assumption of independent distribution of the species. Therefore, the model produces an expected distribution of cooccurrences that can be used to assign significance to the actual observed cooccurrences. This method is analytically exact, and fast, and does not imply prior interactive structure. Also, it is based on presence-absence, which avoids problems derived from the compositional nature of microbial amplicon data, such as spurious correlations, low sensitivity due to relative frequencies, and alterations in its covariance structure (66, 67).

Initially, we calculated the probability that each pair of species cooccur more (aggregated pair) or less (segregated pair) than expected, given their occurrences. Then, for species pairs that deviated from the expectations (i.e., that were not random), we tried to identify the cause of these deviations, whether they are produced by environmental or spatial factors. Environmental factors were tested by one-way ANOVA of each environmental variable comparing allotypic sites (sites with only species A versus only species B) in segregated pairs or, in the case of aggregated pairs, sites with both species against empty sites. In the case of spatial factors, the null hypothesis tested with a one-way MANOVA of the coordinates of the sites is that the different types of sites are not spatially segregated. If this hypothesis is rejected, the cause of the observed pattern can be imputed to some form of dispersal limitation, at least partially. In summary, each pair of zOTUs can be assigned to one of the following results: random pairs, segregated pairs due to negative interactions or environmental filtering and/or dispersal limitation, and aggregated pairs due to positive interactions or environmental filtering and/or dispersal limitation (22).

Network and node properties estimation.

We used the inferred positive and negative potential interactions to build a metacommunity network, where nodes (zOTUs) are linked with weighted links. These weights can be positive (cooccurrences) or negative (coexclusions), and their value was the logarithm of the cooccurrence/coexclusion P value obtained when testing the null hypothesis that cooccurrences/coexclusions emerge by chance (68). In practice, this means that weights were proportional to the order-of-magnitude difference between probabilities and the significance level.

We described network structure using standard statistical descriptors for complex networks implemented in iGraph (69). We analyzed network properties such as “diameter” (shortest-path length between the most distant nodes in the network), “average shortest-path length” (average number of steps it takes to get from one node of the network to another), and “clustering coefficient” defined as the average transitivity (the clustering of a node is defined as the ratio of existing links connecting its neighbors among each other to the maximum possible number of such links between neighbors; the network clustering coefficient is the average over the network of each node’s clustering). Modularity was calculated using the spinglass algorithm (70), which allows the definition of community clusters based on both positive and negative edges. The index reveals how significant network community clusters are, optimizing network partitions by including more positive edges within the modules and more negative edges between modules. Finally, we extracted the modules assigned by the spinglass algorithm for downstream analyses. The total number of edges connected to a given node is defined as “node degree.” By classifying zOTUs into network modules, we aimed at finding taxonomic differences in and between the modules and to assess keystone groups based on weighted node degree. When including the sum of the connection strength (weight) of every edge in the network, we refer to the node degree as the “weighted node degree” (71). The nodes that have a significantly larger degree, in comparison to other nodes in the network, are interpreted as hubs or keystones. These are nodes that can be interpreted as facilitator species, highly competitive species, or species subject to a high biotic pressure (i.e., predation).

Statistics.

We used a chi-square test of independence to analyze contingency tables of the contribution of explanatory variables to taxonomic groups and also the contribution of taxonomy, EnvO (72), or nutrient assimilation strategies to different modules. We quantified the contribution in the percentage of the residuals to the total chi-square statistic to indicate particular associations that stand out, not having an effect on the differential interaction numbers included in the categories. To study the explanatory variables and their association with taxonomy at the genus level, we first grouped genera according to a clustering partition around k-medoids (PAM-K) and optimum average silhouette to estimate the optimum number of clusters, as implemented in the R package fpc (73, 74).

Data availability.

The whole gene sequence data sets were deposited to the NCBI Sequence Read Archive and are available through BioProject PRJNA413654.

TEXT S1

Defining potential keystone taxa and guild components. Download Text S1, DOCX file, 0.01 MB (13.6KB, docx) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

ACKNOWLEDGMENTS

We thank the technicians and students who participated in the field sampling expeditions, L. Camarero for ancillary data, and the facilities and logistics of the Center for High Mountain Research (CRAM-UB).

Funding was provided by grant INTERACTOMA RTI2018-101205-B-I00 from the Spanish Agency of Research (AEI-MICINN) and European funding (ERDF) to E.O.C. R.O.-A. and V.J.O. were funded by the Spanish FPI PhD scholarships program (to E.O.C. and D.A., respectively).

Contributor Information

Emilio O. Casamayor, Email: casamayor@ceab.csic.es.

Michael J. Imperiale, University of Michigan—Ann Arbor

REFERENCES

  • 1.Barberán A, Casamayor EO, Fierer N. 2014. The microbial contribution to macroecology. Front Microbiol 5:203. doi: 10.3389/fmicb.2014.00203. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Vellend M. 2010. Conceptual synthesis in community ecology. Q Rev Biol 85:183–206. doi: 10.1086/652373. [DOI] [PubMed] [Google Scholar]
  • 3.Niño-García JP, Ruiz-González C, del Giorgio PA. 2016. Interactions between hydrology and water chemistry shape bacterioplankton biogeography across boreal freshwater networks. ISME J 10:1755–1712. doi: 10.1038/ismej.2015.226. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Ortiz-Álvarez R, Cáliz J, Camarero L, Casamayor EO. 2020. Regional community assembly drivers and microbial environmental sources shaping bacterioplankton in an alpine lacustrine district (Pyrenees, Spain). Environ Microbiol 22:297–309. doi: 10.1111/1462-2920.14848. [DOI] [PubMed] [Google Scholar]
  • 5.Román S, Ortiz-Álvarez R, Romano C, Casamayor EO, Martin D. 2019. Microbial community structure and functionality in the deep sea floor: evaluating the causes of spatial heterogeneity in a submarine canyon system (NW Mediterranean, Spain). Front Mar Sci 6:1–18. [Google Scholar]
  • 6.Lima-Mendez G, Faust K, Henry N, Decelle J, Colin S, Carcillo F, Chaffron S, Ignacio-Espinosa JC, Roux S, Vincent F, Bittner L, Darzi Y, Wang J, Audic S, Berline L, Bontempi G, Cabello AM, Coppola L, Cornejo-Castillo FM, D’Ovidio F, De Meester L, Ferrera I, Garet-Delmas MJ, Guidi L, Lara E, Pesant S, Royo-Llonch M, Salazar G, Sánchez P, Sebastian M, Souffreau C, Dimier C, Picheral M, Searson S, Kandels-Lewis S, Gorsky G, Not F, Ogata H, Speich S, Stemmann L, Weissenbach J, Wincker P, Acinas SG, Sunagawa S, Bork P, Sullivan MB, Karsenti E, Bowler C, De Vargas C, Raes J, Tara Oceans Coordinators , et al. 2015. Determinants of community structure in the global plankton interactome. Science 348:1262073. doi: 10.1126/science.1262073. [DOI] [PubMed] [Google Scholar]
  • 7.Hibbing ME, Fuqua C, Parsek MR, Peterson SB. 2010. Bacterial competition: surviving and thriving in the microbial jungle. Nat Rev Microbiol 8:15–25. doi: 10.1038/nrmicro2259. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Ghoul M, Mitri S. 2016. The ecology and evolution of microbial competition. Trends Microbiol 24:833–845. doi: 10.1016/j.tim.2016.06.011. [DOI] [PubMed] [Google Scholar]
  • 9.Grossart HP, Wurzbacher C, James TY, Kagami M. 2016. Discovery of dark matter fungi in aquatic ecosystems demands a reappraisal of the phylogeny and ecology of zoosporic fungi. Fungal Ecol 19:28–38. doi: 10.1016/j.funeco.2015.06.004. [DOI] [Google Scholar]
  • 10.García-Callejas D, Molowny-Horas R, Araújo MB. 2018. Multiple interactions networks: towards more realistic descriptions of the web of life. Oikos 127:5–16. doi: 10.1111/oik.04428. [DOI] [Google Scholar]
  • 11.Hug LA, Baker BJ, Anantharaman K, Brown CT, Probst AJ, Castelle CJ, Butterfield CN, Hernsdorf AW, Amano Y, Ise K, Suzuki Y, Dudek N, Relman DA, Finstad KM, Amundson R, Thomas BC, Banfield JF. 2016. A new view of the tree of life. Nat Microbiol 1:16048. doi: 10.1038/nmicrobiol.2016.48. [DOI] [PubMed] [Google Scholar]
  • 12.Castelle CJ, Wrighton KC, Thomas BC, Hug LA, Brown CT, Wilkins MJ, Frischkorn KR, Tringe SG, Singh A, Markillie LM, Taylor RC, Williams KH, Banfield JF. 2015. Genomic expansion of domain archaea highlights roles for organisms from new phyla in anaerobic carbon cycling. Curr Biol 25:690–701. doi: 10.1016/j.cub.2015.01.014. [DOI] [PubMed] [Google Scholar]
  • 13.Christensen BB, Haagensen JAJ, Heydorn A, Molin S. 2002. Metabolic commensalism and competition in a two-species microbial consortium. Appl Environ Microbiol 68:2495–2502. doi: 10.1128/AEM.68.5.2495-2502.2002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Li YH, Tian XL. 2016. Quorum sensing and bacterial social interactions in biofilms: bacterial cooperation and competition. Stress Environ Regul Gene Expr Adapt Bact 2:1197–1205. [Google Scholar]
  • 15.Keller L, Surette MG. 2006. Communication in bacteria: an ecological and evolutionary perspective. Nat Rev Microbiol 4:249–258. doi: 10.1038/nrmicro1383. [DOI] [PubMed] [Google Scholar]
  • 16.Frenken T, Alacid E, Berger SA, Bourne EC, Gerphagnon M, Grossart H-P, Gsell AS, Ibelings BW, Kagami M, Küpper FC, Letcher PM, Loyau A, Miki T, Nejstgaard JC, Rasconi S, Reñé A, Rohrlack T, Rojas-Jimenez K, Schmeller DS, Scholz B, Seto K, Sime-Ngando T, Sukenik A, Van de Waal DB, Van den Wyngaert S, Van Donk E, Wolinska J, Wurzbacher C, Agha R. 2017. Integrating chytrid fungal parasites into plankton ecology: research gaps and needs. Environ Microbiol 19:3802–3822. doi: 10.1111/1462-2920.13827. [DOI] [PubMed] [Google Scholar]
  • 17.Catalan J, Curtis CJ, Kernan M. 2009. Remote European mountain lake ecosystems: regionalization and ecological status. Freshw Biol 54:2419–2432. doi: 10.1111/j.1365-2427.2009.02326.x. [DOI] [Google Scholar]
  • 18.Capitán JA, Cuenda S, Ordóñez A, Alonso D. 2021. A signal of competitive dominance in mid-latitude herbaceous plant communities. R Soc Open Sci 8:201361. doi: 10.1098/rsos.201361. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Barberán A, Bates ST, Casamayor EO, Fierer N. 2012. Using network analysis to explore co-occurrence patterns in soil microbial communities. ISME J 6:343–351. doi: 10.1038/ismej.2011.119. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Milici M, Deng ZL, Tomasch J, Decelle J, Wos-Oxley ML, Wang H, Jáuregui R, Plumeier I, Giebel HA, Badewien TH, Wurst M, Pieper DH, Simon M, Wagner-Döbler I. 2016. Co-occurrence analysis of microbial taxa in the Atlantic ocean reveals high connectivity in the free-living bacterioplankton. Front Microbiol 7:1–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Eiler A, Heinrich F, Bertilsson S. 2012. Coherent dynamics and association networks among lake bacterioplankton taxa. ISME J 6:330–342. doi: 10.1038/ismej.2011.113. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Blois JL, Gotelli NJ, Behrensmeyer AK, Faith JT, Lyons SK, Williams JW, Amatangelo KL, Bercovici A, Du A, Eronen JT, Graves GR, Jud N, Labandeira C, Looy CV, McGill B, Patterson D, Potts R, Riddle B, Terry R, Tóth A, Villaseñor A, Wing S. 2014. A framework for evaluating the influence of climate, dispersal limitation, and biotic interactions using fossil pollen associations across the late quaternary. Ecography doi: 10.1111/ecog.00779. [DOI] [Google Scholar]
  • 23.Goberna M, Montesinos-Navarro A, Valiente-Banuet A, Colin Y, Gómez-Fernández A, Donat S, Navarro-Cano JA, Verdú M. 2019. Incorporating phylogenetic metrics to microbial co-occurrence networks based on amplicon sequences to discern community assembly processes. Mol Ecol Resour 19:1552–1564. doi: 10.1111/1755-0998.13079. [DOI] [PubMed] [Google Scholar]
  • 24.Leibold MA, Holyoak M, Mouquet N, Amarasekare P, Chase JM, Hoopes MF, Holt RD, Shurin JB, Law R, Tilman D, Loreau M, Gonzalez A. 2004. The metacommunity concept: a framework for multi-scale community ecology. Ecol Lett 7:601–613. doi: 10.1111/j.1461-0248.2004.00608.x. [DOI] [Google Scholar]
  • 25.Wilson DS. 1992. Complex interactions in metacommunities, with implications for biodiversity and higher levels of selection. Ecology 73:1984–2000. doi: 10.2307/1941449. [DOI] [Google Scholar]
  • 26.Hervas A, Casamayor EO. 2009. High similarity between bacterioneuston and airborne bacterial community compositions in a high mountain lake area. FEMS Microbiol Ecol 67:219–228. doi: 10.1111/j.1574-6941.2008.00617.x. [DOI] [PubMed] [Google Scholar]
  • 27.Felip M, Camarero L, Catalan J. 1999. Temporal changes of microbial assemblages in the ice and snow cover of a high mountain lake. Limnol Oceanogr 44:973–987. doi: 10.4319/lo.1999.44.4.0973. [DOI] [Google Scholar]
  • 28.Barberán A, Casamayor EO. 2014. A phylogenetic perspective on species diversity, β-diversity and biogeography for the microbial world. Mol Ecol 23:5868–5876. doi: 10.1111/mec.12971. [DOI] [PubMed] [Google Scholar]
  • 29.Buchaca T, Catalan J. 2007. Factors influencing the variability of pigments in the surface sediments of mountain lakes. Freshwater Biol 52:1365–1379. doi: 10.1111/j.1365-2427.2007.01774.x. [DOI] [Google Scholar]
  • 30.Auguet J-C, Casamayor EO. 2013. Partitioning of Thaumarchaeota populations along environmental gradients in high mountain lakes. FEMS Microbiol Ecol 84:154–164. doi: 10.1111/1574-6941.12047. [DOI] [PubMed] [Google Scholar]
  • 31.Llorens-Marès T, Catalan J, Casamayor EO. 2020. Taxonomy and functional interactions in upper and bottom waters of an oligotrophic high-mountain deep lake (Redon, Pyrenees) unveiled by microbial metagenomics. Sci Total Environ 707:135929. doi: 10.1016/j.scitotenv.2019.135929. [DOI] [PubMed] [Google Scholar]
  • 32.Catalan J, Camarero L, Felip M, Pla S, Ventura M, Buchaca T, Bartumeus F, de Mendoza G, Miró A. 2006. High mountain lakes: extreme habitats and witnesses of environmental changes. Limnetica 25:551–584. doi: 10.23818/limn.25.38. [DOI] [Google Scholar]
  • 33.Delgado-Baquerizo M, Oliverio AM, Brewer TE, Benavent-González A, Eldridge DJ, Bardgett RD, Maestre FT, Singh BK, Fierer N. 2018. A global atlas of the dominant bacteria found in soil. Science 359:320–325. doi: 10.1126/science.aap9516. [DOI] [PubMed] [Google Scholar]
  • 34.Araújo MB, Rozenfeld A. 2014. The geographic scaling of biotic interactions. Ecography (Cop) 37:406–415. doi: 10.1111/j.1600-0587.2013.00643.x. [DOI] [Google Scholar]
  • 35.Morales-Castilla I, Matias MG, Gravel D, Araújo MB. 2015. Inferring biotic interactions from proxies. Trends Ecol Evol 30:347–356. doi: 10.1016/j.tree.2015.03.014. [DOI] [PubMed] [Google Scholar]
  • 36.Albaina A, Aguirre M, Abad D, Santos M, Estonba A. 2016. 18S rRNA V9 metabarcoding for diet characterization: a critical evaluation with two sympatric zooplanktivorous fish species. Ecol Evol 6:1809–1824. doi: 10.1002/ece3.1986. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Bahram M, Hildebrand F, Forslund SK, Anderson JL, Soudzilovskaia NA, Bodegom PM, Bengtsson-Palme J, Anslan S, Coelho LP, Harend H, Huerta-Cepas J, Medema MH, Maltz MR, Mundra S, Olsson PA, Pent M, Põlme S, Sunagawa S, Ryberg M, Tedersoo L, Bork P. 2018. Structure and function of the global topsoil microbiome. Nature 560:233–237. doi: 10.1038/s41586-018-0386-6. [DOI] [PubMed] [Google Scholar]
  • 38.Zeder M, Peter S, Shabarova T, Pernthaler J. 2009. A small population of planktonic Flavobacteria with disproportionally high growth during the spring phytoplankton bloom in a prealpine lake. Environ Microbiol 11:2676–2686. doi: 10.1111/j.1462-2920.2009.01994.x. [DOI] [PubMed] [Google Scholar]
  • 39.Ortiz-Álvarez R, Triadó-Margarit X, Camarero L, Casamayor EO, Catalan J. 2018. High planktonic diversity in mountain lakes contains similar contributions of autotrophic, heterotrophic, and parasitic eukaryotic life forms. Sci Rep 8:4457. doi: 10.1038/s41598-018-22835-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Holt AR, Warren PH, Gaston KJ. 2002. The importance of biotic interactions in abundance – occupancy relationships. J Anim Ecol 71:846–854. doi: 10.1046/j.1365-2656.2002.00650.x. [DOI] [Google Scholar]
  • 41.Nee S, Gregory RD, May RM. 1991. Core and satellite species: theory and artifacts. Oikos 62:83–87. doi: 10.2307/3545450. [DOI] [Google Scholar]
  • 42.Freilich MA, Wieters E, Broitman BR, Marquet PA, Navarrete SA. 2018. Species co-occurrence networks: can they reveal trophic and non-trophic interactions in ecological communities? Ecology 99:690–699. doi: 10.1002/ecy.2142. [DOI] [PubMed] [Google Scholar]
  • 43.Stachowicz JJ. 2001. Mutualism, facilitation, and the structure of ecological communities. Bioscience 51:235.2.0.CO;2]. doi: 10.1641/0006-3568(2001)051[0235:MFATSO]2.0.CO;2. [DOI] [Google Scholar]
  • 44.Philippot L, Andersson SGE, Battin TJ, Prosser JI, Schimel JP, Whitman WB, Hallin S. 2010. The ecological coherence of high bacterial taxonomic ranks. Nat Rev Microbiol 8:523–529. doi: 10.1038/nrmicro2367. [DOI] [PubMed] [Google Scholar]
  • 45.Wilson JB. 1999. Guilds, functional types and ecological groups. Oikos 86:507. doi: 10.2307/3546655. [DOI] [Google Scholar]
  • 46.Girvan M, Newman MEJ. 2002. Community structure in social and biological networks. Proc Natl Acad Sci USA 99:7821–7826. doi: 10.1073/pnas.122653799. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Ding T, Schloss PD. 2014. Dynamics and associations of microbial community types across the human body. Nature 509:357–360. doi: 10.1038/nature13178. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Williams RJ, Howe A, Hofmockel KS. 2014. Demonstrating microbial co-occurrence pattern analyses within and between ecosystems. Front Microbiol 5:358. doi: 10.3389/fmicb.2014.00358. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Ruiz-González C, Niño-García JP, del Giorgio PA. 2015. Terrestrial origin of bacterial communities in complex boreal freshwater networks. Ecol Lett 18:1198–1206. doi: 10.1111/ele.12499. [DOI] [PubMed] [Google Scholar]
  • 50.Sharma AK, Sommerfeld K, Bullerjahn GS, Matteson AR, Wilhelm SW, Jezbera J, Brandt U, Doolittle WF, Hahn MW. 2009. Actinorhodopsin genes discovered in diverse freshwater habitats and among cultivated freshwater Actinobacteria. ISME J 3:726–737. doi: 10.1038/ismej.2009.13. [DOI] [PubMed] [Google Scholar]
  • 51.Shade A, Kent AD, Jones SE, Newton RJ, Triplett EW, McMahon KD. 2007. Interannual dynamics and phenology of bacterial communities in a eutrophic lake. Limnol Oceanogr 52:487–494. doi: 10.4319/lo.2007.52.2.0487. [DOI] [Google Scholar]
  • 52.Camarero L, Rogora M, Mosello R, Anderson NJ, Barbieri A, Botev IVAN, Kernan M, Kopáček JIŘÍ, Korhola ATTE, Lotter AF, Muri G, Postolache C, Stuchlík E, Thies H, Wright RF. 2009. Regionalization of chemical variability in European mountain lakes. Freshw Biol 54:2452–2469. doi: 10.1111/j.1365-2427.2009.02296.x. [DOI] [Google Scholar]
  • 53.Camarero L, Catalan J. 2012. Atmospheric phosphorus deposition may cause lakes to revert from phosphorus limitation back to nitrogen limitation. Nat Commun 3:1118. doi: 10.1038/ncomms2125. [DOI] [PubMed] [Google Scholar]
  • 54.Giménez-Grau P, Felip M, Zufiaurre A, Pla-Rabès S, Camarero L, Catalan J. 2020. Homeostasis and nonlinear shift in the stoichiometry of P-limited planktonic communities. Ecosphere 11. doi: 10.1002/ecs2.3249. [DOI] [Google Scholar]
  • 55.Marchetto A, Rogora M, Boggero A, Musazzi S, Lami A, Lotter AF, Tolotti M, Thies H, Psenner R, Massaferro J, Barbieri A. 2009. Response of Alpine lakes to major environmental gradients, as detected through planktonic, benthic, and sedimentary assemblages. Adv Limnol 62:419–440. doi: 10.1127/advlim/62/2009/419. [DOI] [Google Scholar]
  • 56.Grossmann L, Jensen M, Pandey RV, Jost S, Bass D, Psenner R, Boenigk J. 2016. Molecular investigation of protistan diversity along an elevation transect of alpine lakes. Aquat Microb Ecol 78:25–37. doi: 10.3354/ame01798. [DOI] [Google Scholar]
  • 57.Catalan J, Barbieri MG, Bartumeus F, Bitušík P, Botev IVAN, Brancelj A, Cogălniceanu DAN, Manca M, Marchetto ALDO, Ognjanova-Rumenova N, Pla S, Rieradevall M, Sorvari S, Štefková E, Stuchlík E, Ventura MARC. 2009. Ecological thresholds in European alpine lakes. Freshw Biol 54:2494–2517. doi: 10.1111/j.1365-2427.2009.02286.x. [DOI] [Google Scholar]
  • 58.Ortiz-Alvarez R, Casamayor EO. 2016. High occurrence of Pacearchaeota and Woesearchaeota (Archaea superphylum DPANN) in the surface waters of oligotrophic high-altitude lakes. Environ Microbiol Rep 8:210–217. doi: 10.1111/1758-2229.12370. [DOI] [PubMed] [Google Scholar]
  • 59.Caporaso JG, Lauber CL, Walters WA, Berg-Lyons D, Lozupone CA, Turnbaugh PJ, Fierer N, Knight R. 2011. Global patterns of 16S rRNA diversity at a depth of millions of sequences per sample. Proc Natl Acad Sci USA 108:4516–4522. doi: 10.1073/pnas.1000080107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Amaral-Zettler LA, McCliment EA, Ducklow HW, Huse SM. 2009. A method for studying protistan diversity using massively parallel sequencing of V9 hypervariable regions of small-subunit ribosomal RNA genes. PLoS One 4:e6372. doi: 10.1371/journal.pone.0006372. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Edgar RC. 2013. UPARSE: highly accurate OTU sequences from microbial amplicon reads. Nat Methods 10:996–998. doi: 10.1038/nmeth.2604. [DOI] [PubMed] [Google Scholar]
  • 62.Edgar RC. 2016. UNOISE2: improved error-correction for Illumina 16S and ITS amplicon sequencing. bioRxiv https://www.biorxiv.org/content/10.1101/081257v1.
  • 63.Sinclair L, Ijaz UZ, Jensen L, Coolen MJ, Gubry-Rangin C, Chroňáková A, Oulas A, Pavloudi C, Schnetzer J, Weimann A, Ijaz A, Eiler A, Quince C, Pafilis E. 2016. Seqenv: linking sequences to environments through text mining. PeerJ 4:e2690. doi: 10.7717/peerj.2690. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 64.Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, Peplies J, Glöckner FO. 2013. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res 41:D590–D596. doi: 10.1093/nar/gks1219. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 65.Veech JA. 2013. A probabilistic model for analyzing species co-occurrence. Glob Ecol Biogeogr 22:252–260. doi: 10.1111/j.1466-8238.2012.00789.x. [DOI] [Google Scholar]
  • 66.Röttjers L, Faust K. 2018. From hairballs to hypotheses: biological insights from microbial networks. FEMS Microbiol Rev 42:761–780. doi: 10.1093/femsre/fuy030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 67.Berry D, Widder S. 2014. Deciphering microbial interactions and detecting keystone species with co-occurrence networks. Front Microbiol 5:219. doi: 10.3389/fmicb.2014.00219. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 68.Martinez-Romo J, Araujo L, Borge-Holthoefer J, Arenas A, Capitán JA, Cuesta JA. 2011. Disentangling categorical relationships through a graph of co-occurrences. Phys Rev E Stat Nonlin Soft Matter Phys 84:046108. doi: 10.1103/PhysRevE.84.046108. [DOI] [PubMed] [Google Scholar]
  • 69.Csardi G. 2008. Large-scale network analysis. Physics (College. Park. Md) 101:603. [Google Scholar]
  • 70.Traag VA, Bruggeman J. 2009. Community detection in networks with positive and negative links. Phys Rev E Stat Nonlin Soft Matter Phys 80:036115. doi: 10.1103/PhysRevE.80.036115. [DOI] [PubMed] [Google Scholar]
  • 71.Gilarranz LJ, Pastor JM, Galeano J. 2012. The architecture of weighted mutualistic networks. Oikos 121:1154–1162. doi: 10.1111/j.1600-0706.2011.19592.x. [DOI] [Google Scholar]
  • 72.Buttigieg PL, Morrison N, Smith B, Mungall CJ, Lewis SE, ENVO Consortium . 2013. The environment ontology: contextualizing biological and biomedical entities. J Biomed Semantics 4:43. doi: 10.1186/2041-1480-4-43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 73.Henning C. 2014. fpc: flexible procedures for clustering. https://CRAN.R-project.org/package=fpc.
  • 74.Ortiz-Álvarez R, De Los Ríos A, Fernández-Mendoza F, Torralba-Burrial A, Pérez-Ortega S. 2015. Ecological specialization of two photobiont-specific maritime cyanolichen species of the genus Lichina. PLoS One 10:e0132718. doi: 10.1371/journal.pone.0132718. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

FIG S1

(a) Bar plot displaying the proportions of variables (environment, dispersal, or potential species interaction) that explain cooccurrences and coexclusions across taxonomic bacterial genera, according to the results obtained using the method of Blois et al. (22). A single pair constitutes two nodes with the same or different taxonomy; hence, a pair could contribute to two different taxa. The bar plot displays the groups of genera according to a partitioning around medoids clustering, relying on the proportion of explanatory variables over the pairs where each genus was participating. (b) Contribution in percentage of the Pearson residuals to the total chi-square value of cooccurrences (left) and coexclusions (right). Download FIG S1, PDF file, 0.2 MB (188.2KB, pdf) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

TABLE S1

Bacterial genera grouped according to PAM k-medoids clustering. Download Table S1, DOCX file, 0.01 MB (13.7KB, docx) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

FIG S2

Summary of estimated interactions by taxonomic groups in Bacteria (a) and Eukarya (b), and interdomain (c). Raw counts, without weights. (a) In terms of taxonomic groups, in Bacteria, members of Burkholderiales (Betaproteobacteria), Sphingobacteriales, Flavobacterales, and Cytophagales (Bacteroidetes), Caulobacterales (Alphaproteobacteria), and Frankiales (Actinobacteria) held the most interactions. (b) Except the orders Pseudomonadales (Gammaproteobacteria) and Oligoflexales (Deltaproteobacteria), all orders had more positive than negative potential interactions. In the case of Eukarya, most potential interactions were quantified in the dominant group Chrysophyceae, followed by Chytridiomycota, Dictyochophyceae, Cercozoa, and Dinoflagellata. Only Cryptomycota had more negative than positive potential interactions, and Tracheophyta only was involved in negative pairs. (c) Bacterial relationships with eukaryotes were concentrated in Chrysophyceae and then in Dinoflagellata, Chytridiomycota, and Cercozoa. Interestingly, 9 groups (of 17) had more negative than positive interactions: Tracheophyta, Chlorophyceae, Bicosoecida, Cryptomycota, Dikarya, Cryptomonadales, Dictyochophyceae, Peronosporomycetes, and Cryptomycota. Download FIG S2, PDF file, 0.1 MB (86.8KB, pdf) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

FIG S3

Density distribution of node weighted degree, displayed for the modules in the network of bacteria (a) and in the network of eukaryotes (b). The mean value of nodes from each module was drawn with a dashed line. Download FIG S3, PDF file, 0.1 MB (90.9KB, pdf) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

FIG S4

Number of edges between and within modules of the Bacteria network (B1, B2, and B3) and the Eukarya network (E1, E2, and E3) and interdomain relationships based on domain-specific modules. Download FIG S4, PDF file, 0.1 MB (68.3KB, pdf) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

FIG S5

Local community quantification of network metrics: module richness/lake (number of zOTUs within a module that are present in each individual sample), module completeness/lake (proportion of nodes from each module that is present in each individual sample, regarding the total number of nodes of each module), module abundance/lake (the total relative abundance of the nodes corresponding to each module), and total node weight/lake (sum of node weighted degree by each module and each lake). Download FIG S5, PDF file, 0.1 MB (97.5KB, pdf) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

TEXT S1

Defining potential keystone taxa and guild components. Download Text S1, DOCX file, 0.01 MB (13.6KB, docx) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.

Data Availability Statement

The whole gene sequence data sets were deposited to the NCBI Sequence Read Archive and are available through BioProject PRJNA413654.

TEXT S1

Defining potential keystone taxa and guild components. Download Text S1, DOCX file, 0.01 MB (13.6KB, docx) .

Copyright © 2022 Ontiveros et al.

This content is distributed under the terms of the Creative Commons Attribution 4.0 International license.


Articles from mSphere are provided here courtesy of American Society for Microbiology (ASM)

RESOURCES