Host and microbiome jointly contribute to environmental adaptation

Most animals and plants have associated microorganisms, collectively referred to as their microbiomes, which can provide essential functions. Given their importance, host-associated microbiomes have the potential to contribute substantially to adaptation of the host-microbiome assemblage (the “metaorganism”). Microbiomes may be especially important for rapid adaptation to novel environments because microbiomes can change more rapidly than host genomes. However, it is not well understood how hosts and microbiomes jointly contribute to metaorganism adaptation. We developed a model system with which to disentangle the contributions of hosts and microbiomes to metaorganism adaptation. We established replicate mesocosms containing the nematode Caenorhabditis elegans co-cultured with microorganisms in a novel complex environment (laboratory compost). After approximately 30 nematode generations (100 days), we harvested worm populations and associated microbiomes, and subjected them to a common garden experiment designed to unravel the impacts of microbiome composition and host genetics on metaorganism adaptation. We observed that adaptation took different trajectories in different mesocosm lines, with some increasing in fitness and others decreasing, and that interactions between host and microbiome played an important role in these contrasting evolutionary paths. We chose two exemplary mesocosms (one with a fitness increase and one with a decrease) for detailed study. For each example, we identified specific changes in both microbiome composition (for both bacteria and fungi) and nematode gene expression associated with each change in fitness. Our study provides experimental evidence that adaptation to a novel environment can be jointly influenced by host and microbiome.


Nematode strains
We used the C. elegans population A0 (1) which was repeatedly used in various evolution experiments (1)(2)(3)(4).Unless otherwise stated, the A0 population was maintained on nematode growth medium (NGM) plates with Escherichia coli strain OP50 and synchronized by bleaching following standard procedures (5).

Bacterial strains
To prepare the CeMbio43 community, bacteria were thawed from frozen stocks, individual bacteria were cultured in tryptic soy broth (TSB) in deep well microtiter plates for 42 h on a circular shaker at 20 °C, then mixed in equal cell numbers, and the bacterial mixture adjusted to OD60010 in PBS.

Mesocosm experiment
Lab compost was prepared by placing 100 g autoclaved compost soil in a sterile box.Each box was closed with a lid with three 3 cm holes closed with a sterile foam plug to allow aeration.Potatoes, apples, kohlrabi, and carrots including leaves were washed with water, chopped, mixed, and approximately 100 g of the plant material was added to each box and mixed with the soil in the box.One ml of freshly prepared CeMbio43 in OD6005 was added directly, and another 1 ml and 1.3 ml were added after 24 h and 48 h, respectively.Compost and bacteria were mixed and moistened with 1 to 2 ml of sterile water.
Subsequently, approximately 3200 C. elegans A0 in different stages were added to the compost.These worms were previously maintained on NGM plates inoculated with OP50 and repeatedly washed with M9-buffer prior to introduction into the mesocosm boxes.The boxes were stored on a table in a separate room at room temperature (seasonal fluctuations between 17 °C and 25 °C).Fresh plant material was added every other week and each compost mixed every week using a sterile wooden spatula.The position of the boxes on the table was shuffled weekly.In order to mimic a natural compost environment, the boxes were not maintained under strictly sterile conditions; therefore, microorganisms (in addition to the initial CeMbio43 inoculum) were regularly introduced to the environment (e.g. via fresh plant material).
After 100 days, we sampled nematodes and bacteria from each box.To collect the microbial communities, a randomly chosen subsample of each compost was placed in a 9-cm petri dish, followed by addition of 25 ml of M9-buffer with 0.025% (v/v) of Triton X100 (M9-T).We transferred 300 µl of the mixture to a 2 ml microtube containing approximately ten sterile 1 mm zirconia beads, followed by homogenization of the material using a bead ruptor (Bead Ruptor 96, Omni International, Kennesaw, Georgia, USA) for 3 min at 30 Hz, and conservation of microbes in 10% (v/v) dimethyl sulfoxide (DMSO) at -80 °C.To obtain microbe-free nematodes, we transferred 150 µl of worm-containing buffer from the same compost subsample onto peptone-free NGM (PFM) plates.The resulting worm population was bleached to remove any bacteria, following standard protocols (5), and the surviving eggs were kept overnight in M9-buffer on a shaker at 20 °C and the hatched L1 larvae frozen in a final concentration on 15% (w/v) glycerol in S-buffer at -80 °C.

Common garden experiment and assessment of nematode popula7on growth rate
We performed two common garden experiment to assess the influence of the host and the microbiome on environmental adaptation.The first common garden experiment included host and microbes from six mesocosms lines (or boxes), while the second common garden experiment was focused on hosts and microbes from mesocosm boxes 1 and 2. Both common garden experiments were performed under compost conditions, while the second was additionally repeated on agar plates.
Laboratory compost consisted of approximately 10 g water-washed, grated potatoes, apples, kohlrabi, and carrots including greens placed in sterile 60 ml containers.Each laboratory compost was inoculated with 1 ml of CeMbio43 (i.e., initial microbiome) or one of the microbial communities isolated at day 100 from the mesocosms (i.e., final microbiomes), always standardized to an OD60010.The containers were covered with an air-permeable film and then left for three days at room temperature, and stirred daily.
The NGM agar plates were inoculated with the same CeMbio43 or final (day-100) microbial communities 24 h before the experiment.At the start of the common garden experiment, we added 100 synchronized C. elegans at the fourth larval stage from the initial or final host populations in M9-T to the compost or NGM plate replicates.For the second experiment, worm populations were prepared to be devoid of males, thus consisting of hermaphrodites only, in order to enhance comparability of the different C. elegans populations.These male-free populations were frozen in aliquots for later usage.They were rechecked for males after thawing and the subsequent initiation of worm cultures with individual L4 nematodes, followed by an assessment of the populations after 5 and 7 days, revealing and confirming the complete absence of males (Supplementary Table S1.14).Compost and plates were stored at 20 °C and, for the second common garden experiment, the compost was mixed daily by gentle shaking.After five or four days for compost or plates respectively, the worms were collected.
For the analysis of the plates, we collected worms in 5 ml M9-T, centrifuged for 1 min at 500 rpm, the supernatant was removed, and the worms washed four more times in fresh M9-T.The washed worm pellet was split in two microtubes.One microtube was stored at -20 °C until population growth rate, worm length, and worm area were determined.The other tube was stored at -80 °C.
For the compost, nematodes and microbial communities were isolated from a randomly chosen subsample following a standardized protocol.For this, the compost was mixed using a sterile wooden spatula and approximately 0.6 g compost was transferred onto a 9 cm petri dish.15 ml sterile M9-T were distributed over the sample by gently circling the Petri dish five times.We then collected worms in 3 x 500 µl buffer from a distance of 1 cm, 2 cm, and 3 cm around the undissolved compost sample and combined the samples into one microtube.Three technical replicates (three Petri dishes containing a 0.6 g compost subsample) were taken from each compost to compensate for variations in worm counts within a compost sample.Worm samples were frozen in 1.5 ml M9-T at -20 °C for later phenotypic analysis.
Population growth rate was determined by first counting worms in 1 µl to 200 µl (depending on worm density) of the frozen sample, repeated a total of three times, and calculation of the average.Offspring per worm was calculated by dividing the count result by the 100 worms initially added to each replicate.
The counted worms were extrapolated to the total frozen buffer and, for the compost experiment, to the total compost weight.Length and area of adult worms were determined for randomly selected individuals using pictures taken with a Leica stereomicroscope (Leica Microsystems GmbH, Wetzlar, Germany) and image analysis with Image J (version 2.3.0).
The two common garden experiments were performed at different Nme points using only minor changes in the experimental protocol (e.g., daily mixing of compost material in the second common garden experiment versus no mixing of compost material in the first common garden experiment), possibly leading to some quanNtaNve differences in the measured traits.Importantly, for each common garden experiment, all relevant treatment groups were always assessed in parallel, under idenNcal condiNons, using a randomized distribuNon of samples (to minimize the influence of uncontrollable gradients in the incubators) and using neutral codes for the treatment groups (to minimize any observer bias), thereby ensuring comparability of the relevant treatments per experiment.The population growth rates, worm length, and area of different treatment groups were compared with a Wilcoxon rank sum test and Bonferroni correction for multiple comparisons or with an ANOVA.All statistical calculations were performed with R studio software (version 2022.07.2) and can be found in Supplementary Tables S1.3, S1.5, S1.7, S1.8, S1.10, S1.12, and S1.13.Graphs were produced with R Studio and edited with Inkscape (version 1.1).

16S and ITS amplicon sequencing for microbiome analysis of common garden experiment
The compost was mixed with a sterile, wooden spatula and a compost (i.e., substrate) sample was collected and frozen at -20 °C until further use.Worms were collected from the compost sample following a previously published protocol (6).Briefly, a compost sample was covered with M9-T and emerging worms were collected using a pipette.The worms were transferred to sterile M9-T in a 3 cm Petri dish and then to a sterile microtube and kept for at least 2 min in 10 mM Tetramisole to stop ingestion and excretion of bacteria.Worms were washed another four times in fresh M9-T and frozen in 300 µl of M9-T in a 2-ml microtube in liquid nitrogen, and finally stored at -80 °C.
For isolation of bacterial DNA from nematodes, we added five to ten sterile 1 mm zirconia beads to the thawed worm samples.After crushing the samples for 3 minutes at 30 Hz, each sample was transferred to a 1.5 ml microtube and centrifuged at 8000 rpm for 3 minutes.All but 100 µl of the supernatant was removed and the pellet was resuspended in the remaining 100 µl.Using a tissue kit (Macherey-Nagel, Düren, Germany), DNA was isolated according to the manufacturer's instructions and stored at -20 °C.This protocol follows our previous methods to characterize the microbiome of C. elegans and we did not observe an apparent bias in the identification of the included bacteria, for example during analysis of initial inocula (7,8).Importantly, even if there was a bias, then it should have affected all treatment groups of the common garden experiment in a similar way.Therefore, the variaNon observed between the treatment groups of a common garden experiment should sNll be valid and informaNve.
For DNA isolation from compost, we shredded approx.100 µl of a compost sample with 5-10 sterile 1 mm zirconia beads and 300 µl RNase-free water in a 2 ml microtube for 3 min at 30 Hz.The tubes were briefly centrifuged to spin down larger substrate particles.DNA was isolated from 100 µl compost supernatant using a modified Cetyl Trimethyl Ammonium Bromide-(CTAB) based protocol (9,10).

Microbiome data analysis
Raw amplicon sequencing reads were processed using the QIIME 2 v2022.2microbiome bioinformatics platform (12).Primer and adapter sequences were removed with cutadapt (13).For the 16S amplicons, reads were filtered based on quality scores and forward and reverse reads joined with vsearch ( 14), and amplicon sequence variants (ASVs) were resolved with deblur (15).Taxonomy was assigned to the 16S ASVs with a naïve Bayes classifier pre-trained on the Silva 138 SSU database (16)(17)(18).For the ITS amplicons, ASVs were resolved from the forward reads with the DADA2 pipeline which removes low quality and chimeric sequences (19).Taxonomy was assigned to the ITS ASVs with a naïve Bayes classifier pre-trained on the UNITE v9.0 database (16,17,20,21).The raw reads are available from the NCBI BioProject database under the BioProject ID PRJNA954426.
All statistical analyses of the microbiome data were conducted in R Studio (22,23).Potential contaminant ASVs for both the 16S and ITS datasets were identified with the decontam package using the prevalence method, which uses a statistical approach to determine the ASVs that are overrepresented in negative controls relative to experimental samples, and removed (24).We then assessed differences in microbial community composition in three main sets of analyses, which addressed the following questions: To conduct these analyses, we first quantified dissimilarity in microbiome composition between samples with robust Aitchison distance (25) and visualized these relationships with ordinations of principal coordinates analyses.We partitioned variation in Aitchison distance among worm type, microbiome type, sample type, and their interactions and tested for significance with permutational multivariate analysis of variance (PerMANOVA) (26,27).To determine which amplicon sequence variants differed in relative abundance between pairs of treatment combinations, we used the ALDex2 R package (v 1.31.0) and considered ASVs with an effect size > 1 or < -1 to be differently abundant (28)(29)(30).

RNAseq for transcriptome analysis of C. elegans popula7ons
Approximately 1000 worms were added to the compost.After 24 h, a compost subsample was placed in a 9-cm Petri dish and covered with M9-T.The emerging worms were transferred three times to fresh M9-T in a 3 cm Petri dish in as little liquid as possible to remove bacteria adhering to the worms.At least 50 worms were frozen in 800 µl TRIzol Reagent (Thermo Fisher Scientific, Waltham, MA, United States) in a 2-ml microtube in liquid nitrogen.Worms in TRIzol were thawed five times at 45 °C and refrozen in liquid nitrogen to break up the worm cuticle.Total RNA was isolated using a Direct-zol RNA MicroPrep Kit (Zymo Research, Irvine, CA, United States) following manufacturer's instructions and stored at -80 °C.The transcriptome was analyzed for five replicates.RNA libraries were prepared for sequencing using the Illumina stranded total RNA kit with RiboZero Plus (Illumina, San Diego, USA, catalog number 20040529) according to the manufacturer's protocol.Libraries were sequenced on an Illumina NovaSeq 6000 with paired-end strategy and read length of 100 bp (Illumina, San Diego, USA, NovaSeq 6000 SP Reagent Kit v1.5 (200 cycles), catalog number 20040719).The raw data is available from the ENA database under the accession numbers ERR11455018-ERR11455037.

Transcriptome data analysis
The obtained sequence data was first processed, checked for quality, and filtered.In detail, we removed ribosomal and transfer RNA sequences and, thereafter, contaminant reads with the program BBSplit of the BBTools package v38.45 (31).Read quality trimming was performed with the BBTools package v38.45 (31), including removal of duplicate reads, adapter sequences, low-entropy reads, and trimming of bases with quality scores < 10.Reads with invalid or ambiguous bases and reads with a length < 50 base pairs (bp) were discarded.Only reads surviving quality trimming as pairs entered downstream analysis.Read quality recalibration and error correction was performed with the BBTools package v38.45 (31).The filtered reads were mapped with STAR v2.7.10b (32) to the reference genome of C. elegans strain N2, release WS286, retrieved from WormBase (33).We then assessed differential gene expression in three main sets of analyses, which addressed the following three main questions: All three sets of analyses included the following steps.At the beginning, we explored variation in gene expression across the considered treatments using a principal component analysis (PCA) with R package ggplot2 v3.4.0 (34) in R v4.2.1 (22).Thereafter, we inferred differential gene expression, always relative to the combination of the ancestral C. elegans population with the reference microbiome, using R package DESeq2 v1.36.0 (35).We only included genes in the analysis if they (i) have an existing Entrez gene symbol, (ii) are coding, (iii) are functional (i.e., not a pseudogene), and (iv) have ≥ 10 counts in ≥ 5 samples.Gene count distribution was modeled using a negative binomial generalized linear model.Statistical significance of the treatment effect on individual genes was determined with the Wald test (36).Effect size is expressed as log2(fold-change) of size-factor-normalized gene counts.Probabilities were adjusted using the false discovery rate (37).For the sake of clarity, only genes with a false discovery rate ≤ 0.1 are shown.
Genes with absolute log2(fold-change) ≥ 1 and FDR ≤ 0.05 or 0.01, respectively, in at least one contrast, were subjected to k-means clustering.The optimal number of clusters was determined based on Akaike's Information Criterion.The process was repeated 100 times to account for random effects, and the median optimal number of clusters was taken as the final result.Gene clusters of contrasts with FDR ≤ 0.01 were further condensed by performing a second round of k-means clustering using the cluster medians of the first round as input.The C Index (38) was used to determine the optimal number of second-round clusters.We used the results of the second round of k-means clustering for subsequent enrichment analyses.
For enrichment analyses, we focused on individual gene clusters that showed the respective transcriptome signatures needed to address the three above questions (e.g., gene clusters indicating a difference in transcriptome response between the Box 1 and Box 2 worm populations to address question (i)).Two types of enrichment analyses were performed.On the one hand, we used the Database for Annotation, Visualization, and Integrated Discovery (DAVID; (39)) and studied enrichment according to Gene Ontology (GO) (40), including only categories with a probability ≤ 0.05 (after FDR correction).Cluster importance of a category was defined as Importance, following: where Hij is the number of genes in cluster i with a hit in GO category j, Ni is the total number of genes in cluster i, and N is the overall number of genes in all clusters.The rationale of this approach is to normalize the proportion of hits within each cluster by relative cluster size to avoid bias due to the latter.Cluster importance of GO categories is visualized as heatmaps.
On the other hand, we performed an enrichment analysis with the C. elegans-specific gene expression database WormExp (41), which contains approx.3000 published gene expression data sets (i.e., gene sets) for C. elegans under diverse conditions, allowing a more taxon-specific inquiry of enriched expression categories.Probabilities of enriched gene sets were adjusted using FDR, and the relationship of enriched gene sets was evaluated using hierarchical clustering, visualized via heatmaps.The conditions, under which the differentially expressed gene sets were obtained, yield a strong indication of their function.For example, the gene sets, which exhibit differential expression upon exposure to low doses of the toxic heavy metal cadmium, are likely involved in the stress response to cadmium.There is quite some overlap among the available transcriptome data sets, both in the considered conditions as well as the resulting gene sets, as also illustrated by the performed hierarchical clustering of enriched gene sets (Fig. S9C, S10C, S11C).Therefore, to enhance accessibility of the often complex results, we combined gene sets with such overlaps and, at the same time, with related functions or measured under related conditions, under a common header in the description of results.For example, overlapping differentially expressed gene sets obtained for different mutants, which all exhibit an increased lifespan, are summarized under the header "Lifespan".Similarly, overlapping differentially expressed gene sets, which were observed in comparisons between different natural C. elegans strains, are summarized under the header "Strain variation".We still provide information on the specific gene sets involved, in order to permit a fine-tuned and taxon-specific exploration of the enriched gene sets and the underlying functions, which is indeed a unique feature of the C. elegans-specific WormExp gene expression database (41)(42)(43).
(i) How do Box 1 and Box 2 substrate microbiomes differ in taxonomic composition, and to what degree are substrate microbiome members selected by worms?To address this question, we compared microbiome composition among substrates to which initial worms and either the initial microbiome (including the CeMbio43 bacterial community), final Box 1 microbiome, or final Box 2 microbiome were added and used differences in microbiome composition between these substrates and their respective worms as an indication of selection.(ii) Which microbiome members are associated with the increase in fitness observed for Box 1 worms colonized by the corresponding Box 1 microbiome?To address this question, we compared microbiomes between initial worms and final Box 1 worms exposed to the final Box 1 microbiome and compared these worms to their respective substrates.(iii) Which members of the final Box 2 microbiome are associated with low worm fitness?Here we used differences between the substrate and worm microbiomes for both final Box 2 and initial worms exposed to the final Box 2 microbiome as an indication of selection by the worm and compared the selected microbiomes of final Box 2 and initial worms exposed to the final Box 2 microbiomes.
(i) To what extent do the final Box 1 and final Box 2 populations differ from each and the ancestral C. elegans population and thus indicate genetic evolution?To address this question, we compared variation in gene expression among the three worm populations upon exposure to the same reference microbiomes (including the CeMbio43 bacterial community) in compost, thus ensuring identical growth conditions for the three populations.Any gene expression variaNon observed under otherwise idenNcal treatment condiNons are most likely due to geneNc differences of the worm populaNons.(ii) Which gene expression changes characterize the final Box 1 worms colonized by their coexisting Box 1-day 100 microbiome and thus underlie the high population growth rate and thus the high values for a relevant fitness component expressed by this assemblage in the Box 1 common garden experiment?For this, we identified and characterized the gene expression signature that is unique to the final Box 1 worm -final Box 1 microbiome combination in comparison to the remaining three host-microbiome treatments of the Box 1 common garden experiment.(iii) Which gene expression changes characterize the low population growth rate of worms exposed to the final Box 2 microbiome?To address this question, we compared gene expression changes in the two nematode populations exposed to the final Box 2 microbiome to those exposed to the reference microbiome.

Figure S1 :
Figure S1: Mesocosm experiment in laboratory compost.(A) A mesocosm experiment was performed in boxes containing laboratory compost.(B) Fresh produce and compost soil were added to the boxes at the beginning of the experiment, followed by the addiNon of 43 naNve microbiota bacteria (CeMbio43) and a geneNcally diverse worm populaNon.The laboratory compost was supplemented with fresh produce every two weeks.(C) A †er two weeks and (D) ten weeks, new microbes were visible and the compost showed signs of decomposiNon.On day 100, worms and microbes were harvested from (E) Box 1 and (F) Box 2.

Figure S2 :
Figure S2: Host and microbiome can jointly influence proxies for nematode fitness in the novel compost environment.Results of common garden experiments, in which worm length and worm area was measured for C. elegans populaNons isolated from mesocosms at day 100 (final) and iniNal worms (iniNal) in the presence of final mesocosm microbiomes (final) or iniNal microbiomes including the CeMbio43 bacterial community (iniNal).Worm length (A, B) and area (C, D) was measured for worms from (A, C) compost and (B, D) plates.Worm length and worm area are shown in arbitrary units (AU) for final Box 1 (red boxes), final Box 2 (blue boxes), and iniNal worm populaNons (white boxes) in the presence of final Box 1 (red dots), final Box 2 (blue dots), or iniNal microbiomes (gray dots).Results are summarized as boxplots with the median as a thick horizontal line, the interquarNle range as box, the whiskers as verNcal lines, and each replicate depicted by a dot or symbol.Significant differences are indicated with different le•ers.n = 5.The variaNon in worm length and area on plates and that for Box 2 in compost are consistent with the variaNon in populaNon growth rate measured under the same condiNons.The variaNon in length and area for Box 1 in compost vary from the corresponding results for populaNon growth rate.(E-H) Overall, both measures of worm body size show a significant posiNve correlaNon (assessed via Spearman rank correlaNon) with worm offspring numbers in compost (E, G) as well as on plates (F, H), supporNng the suggesNon that both are related and that worm size is a meaningful proxy for worm fitness.Lines are predicted from a linear model.Shaded areas indicate the 95% confidence interval.

Figure S3 :
Figure S3: Microbiome treatments resulted in differences in both compost and nematode microbiomes.Pairwise differenNal abundance analyses of worm/substrate bacterial microbiomes (final Box 1, final Box 2 or iniNal inoculum including the CeMbio43 bacterial community).Each point represents an ASV and ASVs are grouped by genus with class and order listed.Colors and shapes indicate treatment and sample type of differenNally abundant ASVs.

Figure S4 :
Figure S4: Microbiome treatments resulted in differences in both compost and nematode microbiomes.Pairwise differenNal abundance analyses of worm/substrate fungal microbiomes (final Box 1, final Box 2 or iniNal inoculum including the CeMbio43 bacterial community).Each point represents an ASV and ASVs are grouped by genus with class and family listed.Colors and shapes indicate treatment and sample type of differenNally abundant ASVs.

Figure S5 :Figure S6 :
Figure S5: Differences in microbiome composi7on were associated with increased fitness in nematodes from the Box 1 mesocosm.Pairwise differenNal abundance analyses of bacterial microbiomes from iniNal or final Box 1 worms/substrates exposed to final Box 1 inoculum or iniNal inoculum including the CeMbio43 bacterial community.Each point represents an ASV and ASVs are grouped by genus with class and order listed.Colors and shapes indicate treatment and sample type of differenNally abundant ASVs.

Figure S8 :Figure S9 :
Figure S8: Differences in microbiome composi7on were associated with decreased fitness in nematodes from the Box 2 mesocosm.Pairwise differenNal abundance analyses of fungal microbiomes from iniNal or final Box 2 worms/substrates exposed to final Box 2 inoculum or iniNal inoculum.Each point represents an ASV and ASVs are grouped by genus with class and family listed.Colors and shapes indicate treatment and sample type of differenNally abundant ASVs.

Figure S10 :Figure S11 :
Figure S10: Differen7al gene expression in the adapted Box 1 C. elegans popula7ons.Transcriptome data analysis for the comparison of all possible host-microbiome combinaNons for the Box 1 common garden experiment.IniNal or final Box 1 worms were combined with either the iniNal microbes (including the CeMbio43 bacterial community) or the final Box 1 microbes.(A) Enriched gene ontology (GO) terms of differenNally expressed genes.GO enrichment analysis was performed by DAVID.(B) General variaNon in gene expression was explored with a principal component analysis, whereby the panel shows the spread of sample variaNon along the third and fourth principal components (PC3, PC4).Symbols indicate final Box 1 (triangle) or iniNal (circle) worm origin; colors indicate final Box 1 (red) or iniNal (gray) microbiome origin.n = 5. (C) shows the results of the focused enrichment analysis of cluster 3 with the C. eleganstailored WormExp database and visualizaNon of differenNal expression using heatmaps, whereby the heatmaps always show the gene overlap in percent.DescripNon on the right gives terms of the gene funcNons and fold change a †er FDR correcNon.

Differences in microbiome composi7on were associated with decreased fitness in nematodes from the Box 2 mesocosm
. Pairwise differenNal abundance analyses of bacterial microbiomes from iniNal or final Box 2 worms/substrates exposed to final Box 2 inoculum or iniNal inoculum including the CeMbio43 bacterial community.Each point represents an ASV and ASVs are grouped by genus with class and order listed.Colors and shapes indicate treatment and sample type of differenNally abundant ASVs.