Gut-associated functions are favored during microbiome assembly across a major part of C. elegans life

Abstract

The microbiome expresses a variety of functions that influence host biology. The range of functions depends on the microbiome's composition, which can change during the host's lifetime due to neutral assembly processes, host-mediated selection, and environmental conditions. To date, the exact dynamics of microbiome assembly, the underlying determinants, and the effects on host-associated functions remain poorly understood. Here, we used the nematode Caenorhabditis elegans and a defined community of fully sequenced, naturally associated bacteria to study microbiome dynamics and functions across a major part of the worm's lifetime of hosts under controlled experimental conditions. Bacterial community composition initially shows strongly declining levels of stochasticity, which increases during later time points, suggesting selective effects in younger animals as opposed to more random processes in older animals. The adult microbiome is enriched in genera Ochrobactrum and Enterobacter compared to the direct substrate and a host-free control environment. Using pathway analysis, metabolic, and ecological modeling, we further find that the lifetime assembly dynamics increase competitive strategies and gut-associated functions in the host-associated microbiome, indicating that the colonizing bacteria benefit the worm. Overall, our study introduces a framework for studying microbiome assembly dynamics based on stochastic, ecological, and metabolic models, yielding new insights into the processes that determine host-associated microbiome composition and function.

Importance: The microbiome plays a crucial role in host biology. Its functions depend on the microbiome composition that can change during a host's lifetime. To date, the dynamics of microbiome assembly and the resulting functions still need to be better understood. This study introduces a new approach to characterize the functional consequences of microbiome assembly by modeling both the relevance of stochastic processes and metabolic characteristics of microbial community changes. The approach was applied to experimental time-series data obtained for the microbiome of the nematode Caenorhabditis elegans across the major part of its lifetime. Stochastic processes played a minor role, whereas beneficial bacteria as well as gut-associated functions enriched in hosts. This indicates that the host might actively shape the composition of its microbiome. Overall, this study provides a framework for studying microbiome assembly dynamics and yields new insights into C. elegans microbiome functions.

Keywords: Caenorhabditis elegans; community dynamics; functional ecology; microbial communities; microbial ecology; microbiome; systems biology.

Fig 1

Genomes and inferred metabolic model similarities: (A) Rooted phylogenetic tree in circular format based on assembled genome similarity (alignment by GtdbTk [45] and tree by IQ-TREE [46], midpoint rooting), (B) Multiple correspondence analysis of reactions that were present in the inferred metabolic models. The colors indicate the different genera of the CeMbio43 community. Strain codes are as in Table 1.

Fig 2

Microbial abundances and diversity across sample types and time. (A) Relative mean abundances per sample type and time point. Only taxa with a maximal relative abundance ≥0.05 are shown. Low abundance CeMbio43 taxa [group: “cembio (low)”] and unidentified taxa (group: “other”) are represented jointly. (B) Differential heat tree of species abundances. A taxonomic tree is shown in gray. Differentially abundant species between three conditions were compared: microbial samples from C. elegans (host), from the plate environment directly associated with the worms (substrate), and from plate environments that were always maintained without worms (controls). The colored trees indicate the comparison: substrate vs host (top left), control vs host (top right). Differentially abundant species were colored brown (enriched in substrate or controls) or turquoise (enriched in hosts). Differential abundances were inferred by DESeq2. Taxa with adjusted P-value of ≤0.05 are shown. The thickness of branches corresponds to mean abundance. (C) α-Diversity of filtered and agglomerated microbial samples on taxon level (Shannon diversity) across time. (D) Ordination of microbial samples shown in a PCoA plot using Aitchison distance. Time is indicated by increasing red color intensities. (E and F) Differences in β-diversity (Aitchison, weighted Unifrac distance) compared between sample types and across time. For comparing host-associated samples, pairs from the same replicate were available, whereas for comparing host with control or substrate with control, pairs were randomly associated (100 repetitions).

Fig 3

Influence of stochastic effects on microbial community dynamics. (A) Best-fit values of the dispersal parameter m of the neutral model across time and the different sample types. (B and C) Taxonomic normalized stochasticity index across time and sample types. Red dashed line (tNST = 0.5) indicates the transition of the influence of stochastic (tNST > 0.5) to deterministic (tNST < 0.5) processes in community assembly. Significant differences in tNST were observed between sample types (B) and across specific adjacent time points (C).

Fig 4

Differentially abundant functions in host compared to substrate and control microbiome. Differential abundance was assessed by DESeq2 for different functional subsystems (along vertical axis) and different comparisons of the sample types (the two panels). For example, for the comparison “substrate vs host,” positive values (log2 fold change) indicate that the host-associated microbiome showed an increased presence of the functions of the respective subsystem, while negative values a comparative decrease of the functional subsystem in the host samples. No differentially abundant functions were found for substrate compared to control microbiome. The following subsystems were considered: cazyme, carboydrate-active enzymes; exchange, carbon/energy source and metabolic byproducts; gut, gut-microbiome gene clusters; interactions, prevalence of ecological interactions based on pairwise metabolic predictions; medium, growth medium components; metabolism, MetaCyc pathways; uast, life-history strategies; virulence, virulence and resistance genes.

Fig 5

Enrichment of gut microbiome functions. Gene clusters were predicted by gutSMASH and combined with microbial abundance data. We found gut functions dominantly enriched in host samples. No differences were found for controls vs substrate. TPP, thiamine pyrophosphate; AA, amino acids; GR, glycyl radical amino acid metabolism; OD, oxidative decarboxylation; PFOR, pyruvate ferredoxin oxidoreductase; energy-converting ferredoxin, NAD+ reductase complex (Rnf).

Fig 6

Contribution of microbial taxa to functions. The attribution of individual taxa to predicted functions was determined by regression analysis. We used a penalized regression model to estimate how well the taxon abundances can predict functional abundances. For each taxon, all coefficients from regression models summarized by subsystem are shown in boxplots. The subsystem “interactions” is not shown because of low variance. Strain codes are as in Table 1; colors indicate different genera as in Fig. 1.