Host identity is the dominant factor in the assembly of nematode and tardigrade gut microbiomes in Antarctic Dry Valley streams

Alpha diversity differences among communities

Nematode gut microbiomes were assigned into their respective species categories of E. antarcticus and P. murrayi based on 18S host data that was consistent with morphology (see Methods “Microinvertebrate haplotypes”). In contrast, due to recovery of three undiscernible 18S tardigrade haplotypes, the gut microbiomes were assigned to Tardigrada. Mat bacterial communities were significantly (Tukey’s HSD, P < 2e−16) more diverse than communities of all microinvertebrate gut microbiomes for all four alpha diversity metrics tested (i.e., Richness, Shannon’s Index, Simpson’s Index, and Faith’s Phylogenic Diversity) (Fig. 2, Supplementary Table S1). In contrast to the significance of community type (i.e., mat, E. antarcticus, P. murrayi, Tardigrada) for all alpha diversity metrics (GLM, P < 0.001, χ²(3) > 58.21), there was no effect of mat type (P > 0.65, χ²(1) < 0.21) on bacterial alpha diversity, while stream was significant for Shannon’s and Simpson’s indices (P < 0.01, χ²(3) > 11.47) but not for Richness or Faith’s PD (P > 0.38, χ²(3) < 3.07) (Supplementary Table S1). The interaction between community type and stream was also significant (P < 0.001, χ²(9) > 23.59) for all metrics tested (Supplementary Table S1a). Gut microbiomes of E. antarcticus and P. murrayi were less diverse (Supplementary Table S1b) than those of Tardigrada for Shannon’s, Richness, and Faith’s PD (Tukey HSD, P < 0.05), but not for Simpson’s (Tukey HSD, P > 0.39). Although there was no difference in alpha diversity between nematode species gut microbiomes for Richness, Faith’s PD, and Simpson’s, the Shannon index indicated that gut communities of E. antarcticus were the least diverse (P < 0.05, Fig. 2).

Within microinvertebrates, most eukaryotic reads predictably assigned to the host (89.25% of E. antarcticus, 99.10% for P. murrayi, 99.45% of Tardigrada), however when removed, there was sufficient coverage and sequencing depth for further analysis of non-host eukaryotic communities. Mat eukaryotic communities were more diverse than all non-host eukaryotic gut communities for all metrics (Tukey HSD, Richness, Shannon’s, and Faith’s PD, P < 1.884e−05; Simpson’s P = 0.09) (Supplementary Table S2). Eukaryotic alpha diversity differed between mat types for only Shannon’s (GLM, P = 0.06, χ²(1) = 3.57) but not for Richness, Simpson’s, and Faith’s PD (P > 0.15, χ²(1) < 2.10). Among streams, Richness and Faith’s PD were significantly variable (P = 0.02, χ²(3) > 9.82, Supplementary Table S2a), with Canada Stream being more diverse (Tukey HSD, P < 0.05), but not for Shannon’s (P = 0.22) or Simpson’s indices (P = 0.12) (Supplementary Table S2). Similar to bacterial diversity, the Shannon index indicated that the non-host eukaryotic communities of E. antarcticus were the least diverse, followed by P. murrayi, and then Tardigrada as the most diverse (Tukey HSD, P < 0.05) (Table 2b). For Richness, all gut microbiomes were similarly less diverse than mats, while Faith’s PD showed an overlap of significance for E. antarcticus of the other two microinvertebrate communities although P. murrayi and Tardigrada did separate from each other (Tukey HSD, P < 0.05) (Supplementary Table S2).

Compositional differences among bacterial communities

Black and orange mats represented two dissimilar microbial communities. Examining only mats, although both mat type (PERMANOVA, P < 0.003, F(1) = 3.94) and stream (P < 0.003, F(3) = 4.04) significantly affected the composition of bacterial communities, the communities primarily clustered by mat type (Supplementary Fig. S1a), with only Canada Stream communities separating from those from the other streams (Supplementary Fig. S1b). Consequently, compared to mat type (10%), stream explained the most variation (32%) (Supplementary Table S3a) despite all other streams overlapping in NMDS space (Supplementary Fig. S1b). In contrast, the gut microbiomes of microinvertebrates did not cluster by mat type (Supplementary Fig. S2a) nor by stream, but instead by host identity (e.g., E. antarcticus, P. murrayi, Tardigrada) (Supplementary Fig. S2b). Although all investigated factors significantly affected gut microbiome compositions (PERMANOVA, P < 0.05, F(1–6) > 0.48), mat type and stream explained only 1% and 4% of the microbial community variation, respectively (Supplementary Table S4a). In contrast, host identity played the most dominant role in the assembly of gut microbiomes and explained 14% of total variation. However, most of the variation (72%) remained unexplained. At the genus level of taxonomic resolution, 75% of the taxa within Tardigrada, 78% within P. murrayi, and 87% within E. antarcticus were shared with mats. The remaining proportion of gut taxa was not found within any mats but was in low abundance across all samples. Black and orange mats shared 40% of ASVs, while 27% unique ASVs were assigned to black mats and 33% to orange mats. However, at the genus level, 71% of genera were observed in both mat types with 17% unique genera in black mats and 13% unique in orange mats.

Cyanobacteria, Bacteroidota, and Proteobacteria were the most abundant phyla across all microbiomes comprising 86.40% of the total community composition (Fig. 3a, Supplementary Table S5a). Indicator species analysis confirmed Cyanobacteria, Bacteroidota, and Proteobacteria as significantly indicative phyla of the four microbial community types. Expectedly, Cyanobacteria was the most indicative phylum of the mat communities and although there were six other indicative phyla (Supplementary Table S6), their cumulative relative abundance was low (< 1.2%). Proteobacteria was the sole indicative phylum of the gut microbiomes of E. antarcticus. In contrast, Bacteroidota was the sole indicative phylum of the gut microbiomes of P. murrayi and was also indicative of Tardigrada. Although Patescibacteria was also indicative of Tardigrada, it comprised < 0.1% of all microbiomes and < 0.28% of Tardigrada gut microbiomes. A phylum of predatory bacterium, Bdellovibrionota, was enriched in all gut types (1.02%) vs. mats (0.2%). Due to both their high proportion within communities, and their significance as indicator species, taxa of the three most abundant phyla were selected for further analysis.

There was no effect of mat type in respect to the relative abundance of the three dominant bacterial phyla, but there was a very strong effect of community type (GLMM, P < 1.39e−15, χ²(3) > 72.28, Supplementary Table S7a). Streams significantly affected the abundance of Cyanobacteria and Bacteroidota (P < 0.001, χ²(3) > 14.98), but not of Proteobacteria (P = 0.11, χ²(3) = 5.93). For Cyanobacteria, both mat types were dominated by a similar overall relative abundance (48% and 45% for black and orange, respectively, Fig. 3a, Supplementary Table S5a), but orange mats contained 70.8% more Phormidium, while black mats contained 87.2% more Nostoc (Fig. 3b, Supplementary Table S5b). In comparison to mats, all microinvertebrate gut microbiomes were significantly depleted of Cyanobacteria (Fig. 3b) and when compared to each other contained similar amounts of Cyanobacteria. The contribution of cyanobacterial Nostoc declined from an average 19.2% within both mat types to 0.2% within gut microbiomes. Although cyanobacterial Tychonema was an order of magnitude more abundant than Nostoc within gut microbiomes, its relative abundance was more than twice as abundant in all mats (7.5%) than in the guts (3.4%).

Compared to mats, gut microbiomes of P. murrayi and Tardigrada, but not of E. antarcticus, were similarly enriched in Bacteroidota (Tukey’s HSD, P < 0.05) (Fig. 3c, Supplementary Table S5a). Although gut microbiomes of both P. murrayi and Tardigrada were characterized by similar relative abundance of Bacteroidota at the phylum level, at the genus level Tardigrada microbiome taxa were similar to the mats but at larger relative abundances. In contrast, gut microbiomes of P. murrayi were distinct from the mats and Tardigrada and significantly enriched by the single genus Larkinella making up 37% of its gut community compared to 0.03% of all other microbiome types (Fig. 3c, Supplementary Table S5b). In contrast to Tardigrada, P. murrayi, and mats that all contained similar proteobacterial communities, gut microbiomes of E. antarcticus were significantly enriched (Tukey HSD, P < 0.05) by Proteobacteria and that of the family Pseudomonadaceae in particular (Fig. 3d). Other Proteobacteria of interest within E. antarcticus included taxa in Rickettsiaceae, a family recognized for its intracellular symbionts. Wolbachia, a well-known intracellular bacterium was absent from any microbiome. Comamonadaceae, the most abundant family of Proteobacteria across all communities was almost entirely (97.91%) represented by Polaromonas.

The linear discriminant analysis effect size algorithm (LEfSe) confirmed and further refined these compositional results. LEfSe identified a total of 49 distinctive taxa at different taxonomic ranks specifically affected by the four community types (i.e., mats, E. antarcticus, P. murrayi, Tardigrada) (LDA effect size = 4, P < 0.05) (Fig. 4), but no distinctive taxa when testing for stream (Canada, Bowles Creek, Delta, Von Guerard) or mat type (black or orange). Cyanobacteria (phylum) and Cyanobacteriia (class) were the two most relevant taxa for the mat communities, while Proteobacteria (phylum) and Gammaproteobacteria (subphylum) were the most relevant for gut communities of E. antarcticus (Fig. 4b). Acidobacteriota and Planctomycetota were two other indicative phyla of E. antarcticus but were lower in effect size score. In contrast, there were no characteristic phylum level taxa of P. murrayi nor Tardigrada gut microbiomes. Instead, the most significant taxa of P. murrayi gut microbiomes were Cytophagales (order), Spirosomaceae (family), and Larkinella (genus) (Fig. 4b), all highlighting taxonomic congruence as Cytophagales and Spirosomaceae are the order and family for Larkinella. The five most significant taxa within Tardigrada hosts were all Bacteroidia (class), and not a single taxon was significant above the order level (Fig. 4b). Highlighting the trophic level similarities, indicative taxa for P. murrayi and all but one taxon of Tardigrada were within the same class (i.e., Bacteroidia). However, all indicative taxa at the order, family, and genus level were host specific. Of indicative bacterial taxa for all groups, very little overlap of host types was observed among the entire bacterial tree (Fig. 4a), with only five of the 49 taxa of one host microbiome being nested within another.

Clearly distinct microbiomes within mats and microinvertebrates were also supported by preliminary predictions of their functional characteristics. Using PICRUSt2, functional profiles of mat communities (Supplementary Fig. S3) varied primarily by mat type (PERMANOVA, P = 0.04, F(1) = 2.76), and stream (P = 0.07, F(3) = 1.79) but stream explained more than twice the variation in the model (21.2% vs 10.9%). Similar to bacterial composition data, microinvertebrate gut microbiome functional profiles (Supplementary Fig. S4) statistically varied by both host identity (PERMANOVA, P < 0.01, F(2) = 21.10), mat type (P = 0.04, F(1) = 2.16), and stream (P < 0.01, F(3) = 8.60). However, host identity was the most important factor in the functional profile as it explained more variation (12.9%) than mat type (< 0.1%) or stream (7.2%) did.

Compositional differences among eukaryotic communities

Eukaryotic mat communities were dominated by tardigrades, rotifers, nematodes, and algae (Supplementary Fig. S5). Examining the eukaryotic composition of mats, there was a significant difference between mat types (PERMANOVA, P = 0.03, F(1) = 2.37) and among streams (P < 0.001, F(3) = 6.18), although stream explained a much larger amount of variation (40.7%) than mat type (5.2%) (Supplementary Table S3b). Canada Stream was particularly important for mat communities, making up half of the overall stream explained variation highlighting the uniqueness of mats from this location. Compositional differences of microinvertebrate non-host eukaryotic communities varied by microinvertebrate type. Although the effects of stream, mat type, and host type were all statistically significant to non-host eukaryotic gut communities (P < 0.001, F(1–3) > 1.78), similar to bacterial communities, host identity explained the most variation (3.7%), followed by stream (2.5%), and mat type (0.8%) (Supplementary Table S4b). Indicator species analysis identified Metazoa and Chloroplastida as associated with mat communities (P < 0.001, R² = 0.69 and 0.43 respectively), while all three microinvertebrates shared the same importance of Fungi (P < 0.001, R² = 0.69), at the kingdom level of taxonomic ranking.

Among microinvertebrate gut communities, Fungi comprised the largest (67.3%) component of each microbiome, followed by microbial eukaryotes (including the SAR supergroup), and metazoans (Supplementary Fig. S5). In comparison to mats, all microinvertebrate gut microbiomes were significantly (Tukey HSD, P < 3.54e−07) enriched in Fungi (3.3% vs. 48.9–78.5% respectively) (Supplementary Fig. S5b), resulting in a corresponding change in overall community composition. The relative abundance of Fungi in E. antarcticus was significantly lower (P < 0.01) than in P. murrayi and Tardigrada (48.9%, 76.7%, 78.5%, respectively) (Supplementary Fig. S5b). In contrast to mat types, streams affected the fungal abundance within guts (GLMM, P = 0.03, χ²(3) = 9.00, Supplementary Table S7). Compared to mats, all gut microbiomes were dominated by the subphylum Pezizomycotina and depleted of Blastocladiomycota (Tukey HSD, P < 0.05) (Supplementary Fig. S5b). Eurotiomycetes and Leotiomycetes were the most abundant fungal classes (15.7% and 9.5%), and all fungal communities of microinvertebrate guts contained a high diversity of taxa that were compositionally similar.

Metazoans significantly differed in relative abundance among community types (GLMM, P < 0.01, χ²(3) = 84.74), but there was no influence of mat type or stream (P > 0.47, χ²(1,3) < 2.51, Supplementary Table S7c). Non-host metazoan ASVs were significantly more common in the gut microbiome of E. antarcticus than P. antarcticus and Tardigrada (Tukey HSD, P < 0.001, Supplementary Fig. S5a). E. antarcticus eukaryotic communities were mostly of tardigrade (22.9%) and rotifer origin (5.2%) (Supplementary Fig. S5c). Interestingly, rotifer ASVs were detected in all microinvertebrate guts at similar relative abundances (Tukey HSD, P > 0.32). P. murrayi reads were detected in only a single specimen of E. antarcticus, and in seven out of 94 Tardigrada samples P. murrayi reads were in extremely low abundance (< 0.1%). No reads of E. antarcticus were found in any guts of P. murrayi nor those of tardigrades.

Microinvertebrate haplotypes

As expected, E. antarcticus and P. murrayi host 18S ASV data indicated a single species of each across all samples. However, host 18S ASV data indicated the potential presence of three molecular haplotypes of tardigrades all with equal assignments to known Dry Valley tardigrades (possibly Hypsibiidae Acutuncus antarcticus or Macrobiotidae Richtersius and Paramacrobiotus, or Milnesiidae Milnesium). None of the alpha diversity metrics tested nor the relative abundance of the three dominant bacterial phyla (i.e., Cyanobacteria, Proteobacteria, Bacteroidota) varied among the three haplotypes (GLM, GLMM, P > 0.43, χ²(2) < 4.21) (Supplementary Fig. S6). Although the overall bacterial community composition did statistically vary with stream and haplotype explaining 9.9% and 8.1% variation, respectively, they ordinated with complete overlap in NMDS space (Supplementary Fig. S7). Therefore, three possible host haplotypes were combined for this study due to undiscerned haplotype identity.

Source: Ecology - nature.com