Rapid transcriptional and metabolic adaptation of intestinal microbes to host immune activation

Summary

The gut microbiota produces metabolites that regulate host immunity, thereby impacting disease resistance and susceptibility. The extent to which commensal bacteria reciprocally respond to immune activation, however, remains largely unexplored. Herein we colonized mice with four anaerobic symbionts and show that acute immune responses result in dramatic transcriptional reprogramming of these commensals with minimal changes in their relative abundance. Transcriptomic changes include induction of stress response mediators and down-regulation of carbohydrate-degrading factors such as polysaccharide-utilization loci (PUL). Flagellin and anti-CD3 antibody, two distinct immune stimuli, induced similar transcriptional profiles, suggesting that commensal bacteria detect common effectors or activate shared pathways when facing different host responses. Immune activation altered the intestinal metabolome within 6 hours, decreasing luminal short chain fatty acid and increasing aromatic metabolite concentrations. Thus, intestinal bacteria, prior to detectable shifts in community composition, respond to acute host immune activation by rapidly changing gene transcription and immunomodulatory metabolite production.

Graphical Abstract

Becattini et al. show that activation of the host immune system rapidly alters gene transcription in symbiotic bacteria inhabiting the intestinal lumen, with minimal changes in microbiota composition. Transcriptional changes in resident bacteria alter production of SCFAs and other microbe-derived metabolites, with potential consequences for host immunity and health.

Introduction

Resident bacterial communities of the mammalian intestinal tract interact with the host and regulate its physiology by producing a wide range of metabolic products (Belkaid and Hand, 2014). This symbiosis has been investigated by in vivo transfer of defined bacterial consortia into mice, which has provided mechanistic insights into microbiota-host interactions (Atarashi et al., 2013, Ivanov et al., 2009, Fischbach and Sonnenburg, 2011). Although interactions between host and microbiota are bidirectional, most studies have focused on the impact of bacterial strains or more complex microbial populations and their metabolites on the development and function of host cells (Geva-Zatorsky et al., 2017). In contrast, the influence of the host on the function of resident bacterial communities has received less attention.

Increasing recognition that intestinal microbes provide resistance against intestinal pathogens and contribute to inflammatory diseases, and the effectiveness of fecal transplantation for the treatment of some infections, have led to growing interest in the development and eventual administration of well-characterized commensal bacterial strains for therapeutic purposes (Reeves et al., 2012, Lawley et al., 2012, Becattini et al., 2017, Buffie et al., 2015, Caballero et al., 2017, Kim et al., 2019, Brugiroux et al., 2016). Other studies have demonstrated that commensal bacteria can contribute to enhanced responses to cancer immunotherapy (Fessler et al., 2019). A critical question, however, is whether commensal bacteria function similarly in the wide range of circumstances encountered in clinical settings that often include inflammation and immune activation. In this regard, we have very little data on the impact of immune activation on gene regulation and metabolism of bacteria inhabiting the intestine.

Studies of IBD patients (Schirmer et al., 2018) and experiments with mouse models of colitis (Ilott et al., 2016, Patwa et al., 2011, Lengfelder et al., 2019) have demonstrated that the meta-transcriptome of a complex microbiota is impacted by chronic inflammation. Microbe-derived products resulting from these transcriptional changes can impact the host by modulating inflammation and tissue damage (Ocvirk et al., 2015, Lengfelder et al., 2019). Sustained inflammation, however, alters intestinal physiology via a combination of factors, and promotes changes in microbiota composition which reshape the overall functional profile of the community (Morgan et al., 2012).

In contrast to chronic inflammation, acute inflammation is typically transient, with resolution following the disappearance of the inflammatory stimulus. In these settings, it is unclear whether the microbiota responds to host immune activation by altering gene expression and metabolism, or if it undergoes compositional shifts. Furthermore, given the range of different systemic inflammatory responses that can be generated by the host, it is unknown whether commensal bacteria calibrate their transcriptional and metabolic output to the nature and strength of the underlying immune stimulus.

To address these questions, we utilized a reductionist approach in which mice were reconstituted with a consortium of 4 commensal bacterial strains consisting of two species belonging to the Bacteroidetes phylum and two belonging to the Firmicutes phylum, that had previously been shown to reduce gut colonization by VRE (Caballero et al., 2017). Acute activation of the innate and adaptive immune systems, by systemic administration of flagellin or anti-CD3 antibody, respectively, demonstrated in vivo transcriptomic changes in these commensal bacteria within hours, including up-regulation of genes involved in stress-responses and down-regulation of genes involved in carbohydrate utilization, in the absence of compositional shifts. Metabolomic analyses following immune activation revealed decreased short-chain fatty acids (SCFAs) and increased amino acid and aromatic compound concentrations. Thus, the metatranscriptome of the microbiota and its associated metabolic products can change dramatically during acute inflammation, regardless of alterations in its composition, demonstrating that a given microbiota can exist in multiple functional states depending on the host’s immune status.

Results

Flagellin induces rapid transcriptional changes in the microbiota, without altering its composition

To investigate bacterial adaptation to immune responses in the intestine, we reconstituted mice with a consortium of 4 intestinal obligate anaerobic bacterial strains (CBBP) including 2 Firmicutes (Blautia producta, hereafter BP, and Clostridium bolteae, CB) and 2 Bacteroidetes (Bacteroides sartorii, BS, and Parabacteroides distasonis, PD) that confer resistance against VRE in a mouse model via secretion of a lantibiotic by BP (Kim et al., 2019, Caballero et al., 2017).

We administered CBBP to antibiotic-treated mice and found that the four bacterial strains stably colonized the intestine and equilibrated to similar densities, intermingling without noticeable segregation in the cecal lumen (Supplementary Figure 1A, Figure 1A). Baseline transcriptional profiling of the 4 bacterial strains revealed some distinguishing features, such as a prevalence of genes involved in branched-chain amino acid metabolism and genes of the acetogenic Wood-Ljunghdal pathway (‘carbohydrate’) in BP, and polysaccharide utilization locus (PUL) genes in BS and PD (‘membrane transport’) (Figure 1B and Supplementary Figure 1B). Highly transcribed genes in CBBP strains overlapped minimally, suggesting that each species prioritizes distinct metabolic functions at steady state (Supplementary Figure 1C). Thus, CBBP represents a model commensal community to study how representative anaerobic bacterial strains with distinct transcriptional profiles adapt to host immune responses in the gut lumen.

Figure 1. Flagellin treatment modulates transcriptional activity of the microbiota without affecting community structure.

A) Mice were treated with ampicillin and reconstituted with the CBBP consortium. 14 days later ceca were harvested and sectioned, and species-specific FISH staining was performed. Shown is 1 representative image with single as well as merged channels (n=3, 2 independent experiments performed). B) RNA was extracted from cecal contents of animals reconstituted as described in A) and sequenced. Shown is the functional annotation performed on the resulting reads using the Rast SEED Subsystem, at the Class level (n=3) (right), as well as on the respective genomes (left). C) CBBP-reconstituted mice were injected i.p. with flagellin (2.5 μg). Quantitative PCR was carried out for the depicted genes on cecal tissue harvested at the indicated times following treatment (n=7 from 2 independent experiments, shown are means ± SEM; multiple Wilcoxon tests vs baseline, *p<0.05, **p<0.01, ***p<0.001). D) Relative proportion of CBBP members in cecal content as quantified by 16s rRNA gene sequencing at the depicted time points following flagellin treatment (n=3, bars represent individual mice; shown is one representative of 2 experiments). E) Volcano plots showing genes significantly up- and down-regulated (in red and blue, respectively) by the depicted CBBP members, 6h post flagellin treatment. Numbers in corners indicate significantly differentially expressed genes (DEG, FDR < 0.05, log 2-fold change > 1) (n=3). F) Heatmaps depicting changes in the top variant DEG for each CBBP member at 0h (untreated), 6h and 24h following flagellin treatment (n=3). Shown are all DEG that could be annotated based on the SEED Subsystem, Class level. G) Gene set enrichment analysis was performed on gene expression data shown in Figure 1 F. Shown are SEED Subsystem Subclasses that were significantly enriched (p<0.05).

We first tested how CBBP responds to host innate immune activation by systemic administration of flagellin, which engages a well-characterized TLR5/IL-23/IL-22 intestinal circuitry, leading to epithelial production of the anti-microbial lectin RegIII-γ (Kinnebrew et al., 2012). Consistent with previous findings, intraperitoneal injection of flagellin induced rapid and robust up-regulation of IL-22 as well as of other inflammatory cytokines and antimicrobial factors in the cecum, including IFN-γ, TNF-α, iNOS and RegIII-γ (Figure 1C).

Surprisingly, in spite of its strong immuno-stimulatory effects, flagellin injection did not alter the relative proportions of CBBP strains (Figure 1D) nor their RNA/DNA ratio (Supplementary Figure 1D), suggesting these bacteria may be agnostic to immune activation. RNAseq analysis of the 4 strains, however, revealed profoundly altered transcriptional profiles as early as 6h post treatment, with hundreds of genes differentially expressed as compared to baseline (Figure 1E). The transcriptional profiles of the 4 strains largely returned to baseline 24h post injection, with a fraction of genes remaining differentially expressed at this later time point (Figure 1F and Supplementary Figure 1E). Gene set enrichment analysis (GSEA) of individual strains revealed distinct transcriptional signatures. Broadly up-regulated genes included protein-folding factors (chaperones), oxidative species-scavenging molecules and stress-response mediators, while down-regulated genes were mostly encoding metabolic enzymes, in particular those involved in sugar catabolism and amino acid synthesis (Figures 1F and 1G). Importantly, transcripts derived almost exclusively from CBBP members, in line with the efficient reconstitution shown by 16s rRNA gene sequencing (Figure 1D and data not shown). Taken together, these data indicate that intestinal bacteria rapidly respond to host immune activation with tailored transcriptional programs that are activated in the absence of overt dysbiosis, i.e. changes in community structure.

Shared and unique transcriptional adaptations of commensal bacteria to distinct immune stimuli

To determine the extent to which the observed transcriptional response is specific to the immune response induced by flagellin, an innate immune stimulus, we investigated the transcriptional response of CBBP in mice administered the anti-CD3 antibody 145–2C11, which activates the adaptive immune system by stimulating T cells and profoundly impacts intestinal tissues (Esplugues et al., 2011, Huber et al., 2011, Merger et al., 2002, Musch et al., 2002, Zhou et al., 2004). Injection of anti-CD3 up-regulated inflammatory cytokines and anti-microbial factors in cecal tissue and produced neutrophil accumulation in the lamina propria and egress into the gut lumen (Supplementary Figure 2A, B). Anti-CD3 treatment did not alter bacterial density in the cecum, as assessed by quantitative PCR or direct plating (Supplementary Figure 2C, D) or change the microbiota’s composition (Figure 2B) or relative transcriptional rate of the 4 commensal strains (Supplementary Figure 2E). Nevertheless, anti-CD3 induced dramatic transcriptional changes in all CBBP members within the first 6h, with the transcriptome largely returning to baseline by 24h post-treatment (Figure 2 C–D). GSEA revealed al very similar modulation pattern of bacterial pathways to that observed in flagellin-treated mice (Supplementary Figure 2F). Time point-matched, vehicle-injected controls confirmed modulation of signature genes following anti-CD3 treatment, suggesting that neither trauma of IP injection nor circadian oscillations contributed to transcriptional shifts (Supplementary Figure 2G, H).

Figure 2. Effects of in vivo anti-CD3 antibody treatment on a stable intestinal microbial consortium.

A) Mice were reconstituted with CBBP for 14 days and then injected i.p. with anti-CD3 (a-CD3) antibody (100 μg). Quantitative PCR was carried out for the depicted genes on cecal tissue harvested at the indicated time points following treatment (n=3, shown are means ± SEM; multiple t-tests vs baseline, *p<0.05, **p<0.01, ***p<0.001; similar results were obtained in at least 4 other experiments using mouse strains with different microbiota). B) Relative proportion of CBBP members in cecal content at the depicted time points following treatment, as quantified via 16s rRNA gene sequencing (n=3, bars represent individual mice; shown is one representative of 4 experiments). C) Volcano plots showing genes significantly up- and down-regulated (in red and blue, respectively) by the depicted CBBP members, 6h post anti-CD3 antibody treatment of the mice. Numbers in corners indicate significantly differentially-expressed genes (DEG) (n=3). D) Heatmaps depicting significant DEG for each CBBP member at 0h (untreated), 6h and 24h following anti-CD3 antibody treatment (n=3). Data shown in this figure derive from the same experiment shown in Figure 1, untreated animals (0h time point) are in common.

A cross-sectional analysis of the flagellin/anti-CD3 dataset revealed that several genes were selectively induced by only one or the other stimulus (Supplementary Table 1), likely the result of differences in the respective cytokine profiles (Figures 1C and 2A). Genes involved in the utilization of N-acetylneuraminate (sialic acid) and xylose, possibly reflecting a distinct intestinal carbohydrate profile driven by high levels of IL-22, which promotes fucosylation of epithelial cells (Pickard et al., 2014), were selectively induced by flagellin; genes uniquely up-regulated by anti-CD3 included arginase and enzymes involved in inositol phosphate metabolism.

Although stimulus-specific signatures could be identified, principal component analysis of differentially expressed genes (DEGs) revealed preferential co-clustering of samples 6 hours following anti-CD3 or flagellin administration, suggesting substantial overlap in transcriptional changes occurring in each CBBP member responding to the two stimuli (Figure 3A and Supplementary Figure 3A).

Figure 3. Transcriptional adaptation of commensal microbes to distinct immune stimuli.

A) CBBP-reconstituted mice were treated with either anti-CD3 (a-CD3) antibody or flagellin, and sacrificed 6h or 24h post-treatment to perform RNAseq on luminal RNA. Shown is a Principal Component Analysis on rlog transformed counts of all differentially expressed genes (see also Supplementary Figure 3A for PCA on all genes). Polygons represent different time points (n=3 per time point/treatment. Data shown are from the experiment depicted in Figure 1 D–F and Figure 2 B–D). B) Correlation between differential expression of genes in CBBP members following mouse treatment with anti-CD3 antibody or flagellin. Each dot represents a gene, color code indicates significance of the comparison 6h vs baseline (for each treatment). Pearson’s correlation coefficient and p-value are indicated at the top (n=3 per group). C) Volcano plots highlighting genes encoding for the depicted molecules 6h post treatment. Genes of interest are highlighted and color-coded in yellow if present but not-significantly modulated, and in blue if significantly modulated (n=3 per group).

In fact, bacterial gene transcription correlated between flagellin and anti-CD3 -treated mice, with the majority of highly differentially expressed genes significantly up- or down-regulated by both treatments (Figure 3B). Some genes that were significantly impacted by only one treatment nevertheless displayed similar regulation trends across both datasets (Supplementary Table 1).

Interestingly, while some transcriptional responses to flagellin or anti-CD3 treatment were shared by all CBBP members, others were strain-specific. For instance, transcription of chaperone (Hsp20, ClpB) and thioredoxin encoding genes were shared, while catalase production was more prominent in BS and ferric iron uptake was restricted to CB (Figure 3C and Supplementary Figure 3B).

Polysaccharide utilization loci (PUL) were strongly down-regulated in BS and, to a lesser extent, in PD (Figure 3C). Degradation of complex carbohydrates in the intestinal lumen is a critical step for short-chain fatty acid production, in particular butyrate, which dampens host inflammation and enhances colonization resistance against some pathogens (Furusawa et al., 2013, Smith et al., 2013, Arpaia et al., 2013, Sorbara et al., 2018). Of note, transcription of genes belonging to the Lan operon in BP, which encodes a lantibiotic that enhances resistance to VRE (Kim et al., 2019), was not altered by immune activation, suggesting that CBBP would likely provide resistance regardless of the host’s state of immune activation (Supplementary Figure 3C).

Our results demonstrate that in response to immune activation, intestinal bacteria rapidly activate a wide range of genetic modules involving distinct molecular pathways that may contribute to bacterial persistence in the gut.

Microbiota composition impacts bacterial adaptation to immune responses

Intestinal microbiota compositions vary between individuals and can fluctuate over time. The response of a given commensal bacterial strain is likely to be influenced by its surrounding microbial community. To address the impact of microbiota composition on the transcriptional response of bacterial strains to immune activation, we compared bacterial responses induced by anti-CD3 treatment in ex germ-free (GF) mice reconstituted with either CBBP (4mix) or a 3mix (CBP) from which BS was removed. BS was chosen because of its high expression of catalases, which might modify responses of the other CBBP strains. Confirming our previous results, anti-CD3 treatment did not alter relative proportions, density, transcription or replication rates of CBBP/CBP strains (Figure 4A, Supplementary Figure 4A–C). Immune activation induced rapid transcriptional changes in all strains in mice colonized with either CBBP or CBP. However, we detected a higher number of differentially expressed genes in CBP versus CBBP colonized mice, suggesting that the presence of BS impacts the responses of CBP (Figure 4B). The absence of BS most dramatically impacted the transcriptome of PD, the other member of the Bacteroidetes phylum (Figure 4B). PCA analysis revealed that both consortium composition and immune activation were shaping the transcriptional profile of intestinal bacteria, and although anti-CD3 treatment seemed to imprint more prominent changes and segregate samples along the first PCA axis, some genes were differentially regulated by composition at baseline (Figure 4C and Supplementary Figures 4D,E).

Figure 4. Impact of community composition on transcriptional adaptation of commensals to immune response.

A) Germ-Free (GF) mice were reconstituted with CBBP (4mix) or a consortium lacking BS (3mix) for 14 days, then administered anti-CD3 antibody (a-CD3) and sacrificed 6h post-treatment. Shown is fecal microbiota composition as assessed by 16s rRNA gene sequencing, prior to treatment (0h) and 6h after treatment. Shown is also composition of the input (n=3 per group, 1 pellet not obtained for untreated mouse #10). B) Venn diagrams depicting the numbers of differentially regulated genes (padj<0.05, absolute log2 fold-change>1) by CB, BP and PD 6h post anti-CD3 antibody treatment, in the absence or presence of BS (3-mix and 4-mix, respectively). Gray circles represent genes differentially expressed only in 3-mix, blue circles represent genes differentially expressed only in 4-mix, and purple intersections represent genes modulated in both contexts. C) Principal Component Analysis of rlog-transformed counts for all genes that were significantly differentially expressed in at least one comparison. Polygons represent different groups (consortium~time point) and colors represent time point (n=3 per group). D) Expression levels of manganese catalase genes in CB, BP and PD, 6h post anti-CD3 treatment in the presence or absence of BS (4mix vs 3mix). Shown are the adjusted pvalue and log2 fold change in expression of the genes (6h vs 0h) across the two mixes (n=3). E) Correlation plots showing log2 fold change values for significantly DEG between 6h post anti-CD3 antibody administration and baseline in the depicted consortia. Each dot corresponds to a gene, color code is: blue = DE only in 4-mix; black = DE only in 3-mix; purple = DE in both. Pearson’s correlation coefficient and p-value are shown on top (n=3 per group). F) Representative volcano plot DEG for BP in the context of 3mix or 4mix, 6h post anti-CD3 antibody treatment. Genes DE in both consortia are shown in purple, genes that are DE in only one consortium are shown in blue. Upper panels: DEG in common are in front; lower panels: unique DEG are in front. Some of the most up-/down-regulated genes are highlighted by colored filling; top panels: common 3-mix/4-mix DEG are highlighted; bottom panels: common 3-mix/4-mix DEG are highlighted.

In particular, expression of genes involved in nucleoside and nucleotide metabolism (CB), fucose metabolism (BP) and biotin synthesis (BP, PD) was enhanced in the absence of BS, suggesting that BS might provide these molecules or metabolites (Supplementary Table 2).

We next investigated the impact of the absence of BS in the adaptation of the remaining three members to immune activation.

Mn-catalases, which scavenge H₂O_2, are encoded by the genomes of each of the CBBP strains, but they are the most highly up-regulated by BS upon anti-CD3 treatment (Figure 3C). In the absence of BS, anti-CD3 treatment resulted in up-regulation of MNC-encoding genes in BP, CB and PD (Figure 4D). MNCs were previously shown to function in trans to protect other catalase-negative bacteria from the effect of oxidative species (H₂O₂) (Rochat et al., 2006). Co-culture experiments showed that resistance of BP, CB or PD to H₂O₂ challenge was increased in the presence of an E. coli strain expressing BS-derived MNCs (Supplementary Table 3), suggesting that BS could protect neighboring bacteria during inflammation. Thus, microbial adaptation to host immune responses appears to be a coordinated process among community members.

Comparison of transcriptional shifts occurring in individual bacteria within CBP vs CBBP upon immune activation revealed many overlapping DEGs, and similar trends for some DEGs that were significant in only one consortium (Figure 4E) while other genes were bound by community composition. For instance, following a-CD3 treatment, BP up-regulated arginase, ClpB and Hsp20 chaperones, and oleate hydratase, while down-regulating beta-fructofuranosidase and ABC sucrose transporters in the presence and in the absence of BS. In contrast, the long-chain-fatty-acid-CoA ligase was only up-regulated when BP was exposed to immune activation within CBBP, while the MerR-family transcriptional activator and a hydrolase of the beta-lactamase family where only up-regulated in the CBP consortium (Figure 4F and Supplementary Table 4).

Overall, our results demonstrate microbiota compositional impacts on gene transcription in gut symbiotes and that bacterial species impact each other at baseline and following immune activation. However, at least under the two conditions investigated in these experiments, many transcriptomic changes were unaffected by the presence or absence of a prevalent community member.

CBBP genomes and adaptation modules are structured into operons

Bacterial genes are largely organized into operons (Conway et al., 2014, Zheng et al., 2002). While genomic organization into operons is well characterized for common laboratory strains of bacteria, little is known about the occurrence and structure of operons in obligate anaerobes of the intestine, in part because precise determination of operons requires high-quality transcriptomic data sets, which, for many commensals, are in short supply. We decided to leverage the RNAseq datasets obtained from CBBP-reconstituted ex-GF mice, to elucidate operon organization in the CBBP strains and address the regulation of multi-gene operons following host immune activation. Operon prediction was carried out using Rockhopper (McClure et al., 2013). Approximately 75% of genes constituting each CBBP strain genome are organized into operons (Figure 5A), with a mean size of 3 genes, and a median of 3 (CB and BP) or 2 (BS and PD) genes (Figure 5B), in agreement with previous results (Conway et al., 2014, Zheng et al., 2002). While most operons included few genes, a small number reached sizes of 10+ genes (up to 40) (Figure 5C and Supplementary Table 5). Importantly, over 50% of operons carried combinations of genes of known and unknown function (Figure 5D), a feature that could be exploited to bundle genes of unknown function into functional groups since co-transcribed genes within an operon are usually contributing to a common metabolic output. Analysis of the BP Lan operon organization revealed that this structure is actually constituted by 4 distinct transcriptional units (LanP/LanF-E-G-I-K1-R1/LanA1–5/LanR2-K2-B-T-C), in agreement with its previously reported wide transcriptional range (Kim et al., 2019). A validation run carried out on the dataset obtained from 3-mix-reconstituted mice, confirmed the overall number of operons in BP, CB and PD and validated their composition, although for a subset of operons different boundaries were predicted, mainly resulting from splitting of larger gene clusters, or joining of smaller operons (Supplementary Table 5).

Figure 5. Operon structure in CBBP genomes and adaptation modules.

A) RNA sequencing reads from experiment shown in Figure 4 (GF mice, 4-mix) were used to identify operons in the CBBP genomes using a previously published approach (McClure et al., 2013). Bar graphs depict percentage of genes organized into operons for each CBBP member; numbers indicate absolute number of operons. B) Boxplot depicting operon size in CBBP members. Middle line indicates median, box ranges indicate 25 and 75 percentile, whiskers indicate extremes. Dots represent mean values. C) Distribution of operon sizes in CBBP members. Dots indicate presence of an operon with the corresponding size. D) Percentage of operons containing no, all or some unclassified genes, annotated as hypothetical protein by PATRIC. E) Operon organization of top up and down-regulated genes in BS, following anti-CD3 treatment. Bars represent the mean log2-fold change ± SD for genes significantly modulated across 2 independent experiments (n=3 for each). Alternating colors (red, gray) delimitate distinct operons, i.e. color switch corresponds to end of an operon (see Supplementary Figure 5 for a complete depiction of all operons fulfilling the above selection criteria in CBBP members).

Operon visualization for DEGs induced by immune activation across our experiments revealed coordinated transcription of genes and provided additional insights into adaptation to host immune activation (Figure 5E and Supplementary Figure 5).

In CB we observed upregulation of an operon containing two hypothetical Hsp20-family chaperones, an operon containing genes involved in ribose transport and one encompassing four genes encoding ferric iron transport factors. Of note, iron uptake by commensals can play a role in promoting inflammation and delaying tissue repair in a mouse model of colitis (Bessman et al., 2020) and in our system anti-CD3 strongly induced lipocalin-2, an iron-sequestering molecule released in the gut lumen by epithelial cells (Figure 2A).

In BP, transcription of 2 operons involved in C4-dicarboxylate transport, hydrogenase maturation factors (HypA and HypB) and formate dehydrogenase, that enhance fitness of bacteria in the inflamed intestine, was upregulated (Nguyen et al., 2020, Hughes et al., 2017).

In BS, at least 7 operons involved in degradation of dietary complex carbohydrates were down-regulated, demonstrating concerted reduction of this metabolic activity. Two highly up-regulated manganese catalases were included in operons with genes encoding additional hypothetical proteins and/or a transglycosylase-associated protein. Glutamate and aspartate decarboxylases, highly induced by anti-CD3 treatment, were also encoded within operons whose function might be related to increased amino acid metabolism or stress mitigation via consumption of protons to increase intracellular pH (Richard and Foster, 2004) (Figure 5E and Supplementary Figure 5).

In PD, we identified at least one operon containing 2 hypothetical proteins (peg.952, peg.953) carrying domains of unknown function (DUF) 2807 and 4097 found in proteins involved in adhesion of bacterial cells to surfaces (superfamily of trimeric auto-transporter adhesins, TAAs), that are important virulence factors in Gram-negative pathogens, and one sigma factor (ECF) involved in response to envelope stress (Heimann, 2002) (Supplementary Figure 5).

Thus, commensal microbes modify and coordinate transcription of complex functional modules in response to host immune activation.

Immune activation rapidly alters concentrations of microbiota-derived metabolites in the gut lumen

Commensal bacteria of the intestine affect host physiology through the production of numerous biomolecules.

When cultured in medium, CBBP members produce a broad range of physiologically relevant metabolites (Figure 6A).

Figure 6. Host immune activation promotes rapid metabolic shifts in the gut lumen.

A) A) Metabolic profile of CBBP members cultured for 48h in BHI+. Heatmap shows fold-enrichment/depletion of the respective metabolite with reference to the plain medium (shown are average values from 2 independent experiments). B) Antibiotic-treated mice were reconstituted with CBBP (4mix) for 14 days and then treated with anti-CD3 antibody (a-CD3). Cecal contents were collected 6h and 24h post-treatment (or in untreated mice, 0h) and subject to metabolomics analysis for the depicted compounds. Shown are the results for a selection of relevant metabolites) (data are cumulative from 5 independent experiments for the 0h and 6h time points, 2 independent experiments for the 24h time point, for a total of 14, 15 and 6 mice, respectively). C) Naïve SPF mice were treated with anti-CD3 antibody. Cecal contents were collected 6h and 24h post-treatment and subjected to metabolomics analysis for the depicted compounds. Shown are the results for the metabolites also shown in (B) (n=5 per group; Wilcoxon test using 0h as a reference test, *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001)

To investigate whether the transcriptomic changes in commensal bacteria caused by immune activation impact the concentrations of metabolites in the gut lumen, we performed metabolomic analyses on cecal contents from CBBP-reconstituted, anti-CD3 treated mice. We detected a significant decrease in acetate, the most abundant SCFA in the intestinal lumen (Figure 6B, Supplementary Figure 6A), which is consistent with down-regulated expression of carbohydrate-degrading genes in Bacteroidetes. Reduced acetate might result from lowered production by BS and PD or diminished cross-feeding to BP and CB, given that all four species are acetate producers (Figure 6A). Production of propionate, which is a major fermentation product of BS and PD but not CB or BP (Figure 6A), was less impacted by anti-CD3, although still significantly diminished. Butyrate levels in CBBP colonized mice were very low because none of the CBBP strains produce butyrate (Figure 6B and Supplementary Figure 11B). GC-MS assays revealed increased in amino acid concentrations in cecal samples, possibly resulting from increased protein leakage caused by anti-CD3-induced damage to the ileum or enhanced protein degradation, consistent with down-regulation of genes involved in amino acid synthesis during acute inflammation. In particular, concentrations of valine, leucine, lysine, threonine, proline, tryptophan, phenylalanine were increased 6 and 24 h after immune activation. In addition, concentrations of branched-SCFA derived from amino acid metabolism (isovalerate, 2-methylbutyrate) (Figure 6B and Supplementary Figure 6A) were increased, consistent with the ability of BS, PD and BP to produce these metabolites in vitro (Figure 6A). Anti-CD3 treatment led to increased production of phenylacetate, a metabolite produced via phenylalanine catabolism by PD and CB in vitro. Concentration of palmitate was also increased by anti-CD3 treatment, providing a substrate for long-chain fatty acid CoA-ligase, whose expression was significantly induced in BP. Palmitic acid is the precursor for palmitoleic acid, an antimicrobial molecule produced by epithelial cells, which bacteria detoxify using oleate hydratase, a gene product that is also up-regulated in BP (Supplementary Figure 5) (Subramanian et al., 2019). Several of the above-described metabolic changes were observed in CBBP-reconstituted mice treated with flagellin (Supplementary Figure 6B), including decreases in SCFA concentrations and augmented AA and palmitate availability. These results suggest that intestinal commensals, upon host immune activation, switch from carbohydrate- to protein/amino acid-based metabolism.

To determine whether metabolic changes detected in CBBP-colonized mice are generalizable to mice colonized with a complex microbiota, we administered anti-CD3 antibody to naïve WT mice and analyzed cecal metabolites 6h later. We detected metabolite changes in WT mice that largely mirrored those observed than in CBBP-reconstituted animals and were of higher magnitude, likely resulting from the greater density of the complex microbiota. Consistent with our observations in CBBP-colonized mice, anti-CD3 treatment of WT mice led to reduced acetate and butyrate levels, while propionate concentrations were not impacted. The levels of isovalerate, 2-methylbutyrate, isobutyrate (produced from valine), derivatives of tyrosine (phenol, p-cresol), phenylacetate and aminated SCFAs (Figure 6C and Supplementary Figure 6C) were increased upon anti-CD3 treatment. Luminal concentrations of palmitic acid were also increased upon anti-CD3 treatment in naïve mice (Figure 6B,C). Overall, these changes support the hypothesis that immune activation promotes a shift from carbohydrate- to protein-metabolism in resident symbiotic bacteria. The intestinal metabolic profile resulting from anti-CD3 treatment shares similarities to that encountered in chronic diseases associated with inflammation, in particular inflammatory bowel disease (Vernocchi et al., 2016, Franzosa et al., 2018, Lloyd-Price et al., 2019).

Our results demonstrate that intestinal bacteria rapidly respond to host immune activation at the transcriptional level and activate defensive pathways to minimize immune-mediated challenges, in the process reshaping the metabolic milieu of the intestinal lumen. Given the symbiosis between resident bacteria and their mammalian hosts (Rakoff-Nahoum et al., 2004), it is tempting to speculate that commensal bacteria have evolved strategies to protect not only themselves but also their host. Determining how the activation and inactivation of different microbial functions impacts the host’s ability to limit and recover from inflammatory diseases remains an important challenge.

Discussion

The gut microbiota has a profound impact on the composition and function of the host immune system (Geva-Zatorsky et al., 2017) and, more broadly, on the health status of the host through the production of metabolites that affect tissue physiology (Rakoff-Nahoum et al., 2004, Furusawa et al., 2013, Smith et al., 2013, Arpaia et al., 2013), or, conversely, promote disease (Henke et al., 2019, Scher et al., 2013). Our studies demonstrate that the functions of the microbiota, at the transcriptomic and metabolomic levels, are modulated by the activation state of the host’s immune system. Previous studies using longitudinal meta-transcriptional analyses of a complex microbiota from healthy subjects and patients with inflammatory bowel disease (IBD) (Schirmer et al., 2018) demonstrated shifts in the gene expression during chronic inflammation. We extend these findings by demonstrating that transient activation of immune responses with prototypical immunostimulatory agents (flagellin, anti-CD3 antibody) perturbs microbial functions on a remarkably short time scale. In particular, our analyses identified multiple key genes that are modulated within 6h of immune stimulation by a stable consortium of 4 representative obligate anaerobic commensals, 2 Clostridiales and 2 Bacteroidetes (CBBP).

By performing experiments in CBBP-reconstituted mice as well as in an SPF mouse strain bearing a more complex microbiota we showed profound alterations in the intestinal metabolome within 6 hours of immune activation that occur in the absence of changes to gut microbiota composition. Thus, intestinal microbes modify their functional status during acute immune activation of the host, leading to a distinct metabolic milieu in the gut lumen.

In this study, we describe two main axes along which commensal bacterial responses to host immune activation unfold. One axis is up-regulation of stress-related molecules, such as ROS-scavenging enzymes, chaperones, multidrug-export complexes, adhesins, iron uptake mediators and the other is dramatic metabolic restructuring characterized by reduction in metabolism of complex and simple carbohydrates and augmented utilization of amino acids, C4-compounds, and nucleotides. Several differentially modulated microbial functions following flagellin or anti-CD3 administration have been implicated in bacterial adaptation to prolonged inflammation. For example, metatranscriptome analyses in colitis mouse models demonstrated up-regulation of stress-response pathways, reduced polysaccharide utilization and fermentation to butyrate (Ilott et al., 2016, Schwab et al., 2014), although in these studies changes in RNA levels were strongly impacted by shifts in microbial species’ relative abundance, as commonly reported in studies involving complex microbial communities (Franzosa et al., 2014). In a different mouse model of colitis (Patwa et al., 2011), E. coli up-regulated transcription of chaperones, glutamate decarboxylases (involved in acidic stress response) and glycerol-3-phosphate dehydrogenase, which can be involved in osmotic stress responses (Yang et al., 2007, Albertyn et al., 1994). Colitogenic E. faecalis colonization of IL-10−/− mice for 16 weeks reduced bacterial sugar metabolism and up-regulated expression of genes involved in utilization of alternative carbon sources (Lengfelder et al., 2019).

Our results indicate that many of the transcriptional responses of the CBBP strains are conserved, regardless of whether activation of the immune response is initiated by flagellin or anti-CD3 administration, or whether BS is present or absent. Flagellin and anti-CD3 induce similar levels of TNF-α and iNOS in the cecum, which may account for the overlapping commensal transcriptional responses to these two stimuli. On the other hand, flagellin induces higher levels of IL-22 and anti-CD3 promotes stronger IFN-γ up-regulation, which may account for our finding that some DEGs responded uniquely to one stimulus, as is the case for genes involved in sialic-acid import and processing, which were modulated by flagellin but not by anti-CD3.

Along similar lines, expression of some genes was influenced by presence or absence of BS. For example, genes encoding for manganese-catalases were regulated cooperatively, as removal of BS, which expresses high levels of a functional MNC, resulted in up-regulation of analogous MNC genes in the remaining members of the CBP consortium following immune activation, suggesting a division of labor (Tsoi et al., 2018) among CBBP strains. However, elimination of BS from the consortium did not restrict survival and proliferation of the remaining CBP strains (not shown). Noteworthy, catalase production is a prototypical function supporting the ‘Black Queen Hypothesis’ (BQH), which postulates that within bacterial communities, costly functions that are leaky and produce a benefit to multiple species tend to be lost, at the gene level, so that only one or few generous ‘helper’ members maintain them, with an overall benefit in terms of stability of the community (Morris et al., 2012). We propose that understanding how bacterial functions are regulated in real-time via transcriptional modulation by community members, might uncover BQH-compatible strategies employed on a short time-scale to preserve the community.

By clustering CBBP genes based on their chromosomal proximity and expression levels across experimental conditions in vivo (McClure et al., 2013), we have generated a detailed map of operons in these anaerobes’ genomes, and identified polycistronic modules that are likely to be important contributors to the adaptation of commensals to host immune activation, providing an overview of the responses of gut microbes in inflammatory scenarios.

The transcriptional changes in CBBP strains following immune activation were associated with alterations in the intestinal metabolome, alterations that were even more apparent in naïve SPF mice. Immune activation reduced the concentration of short chain fatty acid, particularly acetate and also, in mice harboring a diverse microbiota, butyrate. SCFAs impact mucosal T cell differentiation and can dampen intestinal and systemic inflammation, and their concentration is reduced in patients with inflammatory bowel disease (Marchesi et al., 2007, Franzosa et al., 2018, Lloyd-Price et al., 2019, Bjerrum et al., 2015). Our finding that acute inflammation rapidly reduces SCFA levels in the intestine without altering the community composition suggests that metabolic changes occurring in patients with inflammatory diseases are attributable, at least in part, to metabolic changes occurring within specific members of the commensal community rather than exclusively changes in the microbial community. Following anti-CD3 treatment, we detected other IBD-associated signatures, such as increased tyrosine metabolism, with augmented levels of p-cresol and phenol (Walton et al., 2013), phenylalanine metabolism, (Jansson et al., 2009), as well as higher levels of amino acids, branched AA and branched SCFA (Marchesi et al., 2007, Franzosa et al., 2018, Lloyd-Price et al., 2019, Bjerrum et al., 2015). BSCFA are also enhanced in subjects on protein rich diets (Russell et al., 2011, David et al., 2014) and in patients with hypercholesterolemia (Granado-Serrano et al., 2019), fitting with a potential reprogramming of bacterial metabolism from carbohydrate-degradation to protein-degradation shortly after immune activation. Some of the compounds accumulating in the gut lumen following acute immune activation can have deleterious effects on host health. p-cresol and phenol are genotoxic agents that have been linked to colorectal cancer (Al Hinai et al., 2019, Boutwell and Bosch, 1959, Andriamihaja et al., 2015). Phenylacetate, which is produced in high quantities by commensal microbes from morbidly obese subjects, can promote BCAA metabolism as well as liver steatosis (Hoyles et al., 2018). The long-term health repercussions of repeated or prolonged immune activation and exposure to potentially toxic concentrations of noxious metabolites remain largely undefined.

In conclusion, our work demonstrates that the intestinal microbiota can rapidly adopt distinct functional states following immune activation. The microbiota’s response to the host’s inflammatory state reshapes the intestinal metabolome by redirecting metabolic activities towards reduced carbohydrate metabolism and SCFA production and increased amino acid metabolism with associated production of potentially noxious compounds. We speculate that such transcriptional and metabolic rearrangements might have an impact on host health as well as on the unfolding immune response.

Star Methods

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for reagents or resources may be directed to and will be fulfilled by lead contact Eric G. Pamer (egpamer@uchicago.edu).

Materials Availability

All materials generated in this study will be made available upon reasonable request.

Data and Code Availability

The sequencing data generated during this study have been deposited in the NCBI SRA database under the BioProject ID: PRJNA672114. Metabolomics data, bash and R codes utilized for this work are also available in the GitHub repository: https://github.com/simobeca/CBBP-transcriptomics.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Mouse husbandry

All experiments using wild-type mice were performed with C57BL/6 female mice that were 6–9 weeks old; mice were purchased from Jackson Laboratories.

Germ-free mice were bred in-house in germ-free isolators.

All mice were housed in sterile, autoclaved cages with irradiated food and acidified, autoclaved water. Mouse handling and weekly cage changes were performed by investigators wearing sterile gowns, masks and gloves in a sterile biosafety hood. All animals were maintained in a specific-pathogen-free facility at Memorial Sloan Kettering Cancer Center Animal Resource Center.

For experiments involving different mouse genotypes, mice were co-housed so that each cage would contain each strain, for at least two weeks, prior to proceeding with antibiotic treatment.

Experiments were performed in compliance with Memorial Sloan-Kettering Cancer Center institutional guidelines and approved by the institution’s Institutional Animal Care and Use Committee (protocol # MSK IACUC 00–05-066).

CBBP Bacterial strains and mouse reconstitution

CBBP bacterial strains (Clostridium bolteae, Blautia producta, Bacteroides sartorii, Parabacteroides distasonis) were initially isolated from a mouse colony as described elsewhere (Caballero et al., 2017). Strains were separately maintained in glycerol stock at −80°C. To generate the inoculum to be administered to mice, bacteria were thawed and 50 μl of stock were spread onto pre-reduced Columbia blood agar plates (BD) at 37°C in anaerobic conditions (Coy Laboratory Products) for 48h. Bacterial lawns were collected by scraping and resuspended in pre-reduced PBS to an OD of ~2, equal amounts of suspensions were mixed for the 4 bacteria, and 200 μl of mix were administered to ampicillin-treated mice by oral gavage.

Mice were administered ampicillin in drinking water (0.5 g/l) starting for 4 days prior to reconstitution with CBBP (with the exception of experiments carried on GF mice, for which amipicillin was not used).

METHOD DETAILS

Immune stimulation

CBBP-reconstituted or SPF WT mice were injected i.p. with either anti-CD3 antibody antibody (145–2C11, 100 μg, BioXCell) or ultrapure flagellin (FLA-ST, 2.5 μg, InvivoGen) resuspended in 200 μl of PBS and sacrificed at the indicated time points.

RT PCR (cytokines)

Cecum wall was harvested upon sacrifice, cut longitudinally, emptied, rinsed in PBS, and stored in RNAlater (Millipore) at −20°C.

RNA was isolated from cecum tissue using mechanical homogenization and TRIzol isolation (Invitrogen) according to the manufacturer’s instructions. cDNA was generated using the QuantiTect reverse transcriptase kit (QIAGEN). RT-PCR was performed on cDNA using TaqMan primers and probes in combination with TaqMan PCR Master Mix (ABI), reactions were run on a RT-PCR system (StepOne Plus, Applied Biosystems). Gene expression was normalized to expression of Hprt in untreated mice.

Quantification 16S copy number density by rtPCR.

DNA extracted from intestinal content samples (faeces) was subjected to RT-PCR of 16S rRNA genes using 0.2 mM concentrations of the broad-range bacterial 16S primers 563F (5’-AYTGGGYDTAAAGNG-3’) and 926Rb (5’-CCGTCAATTYHTTTRAGT-3’).

Standard curves were generated by serial dilution of a linearized vector (Invitrogen TOPO pcr2.1 TA vector^AMP amplified into DH5α bacteria) containing the V4 and V5 regions from the E.coli rRNA gene (6 serial dilutions starting from 100 × 10⁶ copies/μl). DyNAmo SYBR Green qPCR Master Mix (Fisher# F-410L) was used to set up the reactions. The cycling conditions were as follows: 95°C for 10 min, followed by 40 cycles of 95°C for 30 s, 52°C for 30s, and 72°C for 1min.

Quantitative PCR (CBBP relative quantification)

DNA extracted from fecal pellets or cecal content was subjected to quantitative PCR using the following custom CBBP member-specific primers.

BS1 (B. sartorii): fw: 5’-ATCCAACCTGCCTTATACTCCC-3’; rev: 5’-TACCGAAGTTCTTTAATCACGAGA-3’; LanB1 (B. producta): fw: 5’-ACTGCCGTTTTTACCTGTGG-3’; rev: 5’-CCATGGGAACTCAGACCACT-3’; CB4 (C. bolteae): fw: 5’-CATCCTTTTGACCGGCGT-3’; rev: 5’-GATTTGCTCCACATCACTGTC-3’; PD3 (P. distasonis): fw: 5’-GGGATGAAGGTTCTATGGATCG-3’; rev: 5’- CGCTGACTTAAAAGGCCGC-3’.

Standard curves were generated by amplification of DNA extracted from pure cultures (48h on Columbia Blood Agar plates, BD) of the respective bacteria (six serial 1:10 dilutions, from 3.3 ng/μl, 3 μl, ~ 10 ng total DNA). A control consisting of equal amounts DNA from the 4 members was always added to confirm accuracy of the quantification. DyNAmo SYBR Green qPCR Master Mix (Fisher# F-410L) was used to set up the reactions. Reactions were run on a StepOne Plus RT PCR system (Applied Biosystems). The cycling conditions were as follows: 95°C for 10 min, followed by 40 cycles of 95°C for 30 s, 52°C for 30s, and 72°C for 1min.

DNA extraction.

DNA was extracted using a phenol–chloroform extraction technique with mechanical disruption (bead beating). In brief, a frozen aliquot of approximately 100 mg per sample was suspended, while frozen, in a solution containing 500 μl of extraction buffer (200 mM Tris, pH 8.0; 200 mM NaCl; and 20 mM EDTA), 210 μl of 20% SDS, 500 μl of phenol:chloroform:isoamyl alcohol (25:24:1), and 500 μl of 0.1-mm-diameter zirconia/silica beads (BioSpec Products). Microbial cells were lysed by mechanical disruption with a bead beater (BioSpec Products) for 2 min, followed by two rounds of phenol:chloroform:isoamyl alcohol extraction. After extraction, DNA was precipitated in ethanol, resuspended in 200 μl of TE buffer with RNase (100 mg/ml), and further purified with QIAamp mini spin columns (Qiagen).

Whole-genome sequencing, assembly and annotation.

An overnight culture grown from a single colony in culture broth was DNA extracted and sequenced using the Illumina MiSeq platform. Purified DNA was sheared using a Covaris ultrasonicator and prepared for sequencing with a Kapa library preparation kit with Illumina TruSeq adaptors to create 300 × 300 bp nonoverlapping paired-end reads. Raw sequence reads were filtered (Phred score ≥ 30, 4 bp sliding window) using Trimmomatic (v.0.36) (Bolger et al., 2014). Trimmed reads were assembled into contigs and annotated with putative open reading frames using the assembly and annotation services in PATRIC40 (v.3.5.25) (Brettin et al., 2015, Wattam et al., 2016).

Microbial composition by 16S rRNA gene sequencing.

Universal bacterial primers—563F (5′-nnnnnnnn-NNNNNNNNNNNN-AYTGGGYDTAAA-GNG-3′) and 926R (5′-nnnnnnnn-NNNNNNNNNNNN-CCGTCAATTYHT- TTRAGT-3), in which ‘N’ represents unique 12-base-pair Golay barcodes and ‘n’ represents additional nucleotides to offset the sequencing of the primers—were used to PCR-amplify the V4–V5 hypervariable region of the 16S rRNA gene. The V4–V5 amplicons were purified, quantified, and pooled at equimiolar concentrations before ligating Illumina barcodes and adaptors using the Illumina TruSeq Sample Preparation protocol. The completed library was sequenced using the MiSeq Illumina platform35. Paired-end reads were merged and demultiplexed. The UPARSE pipeline36 was used for error filtering using the maximum expected error (Emax = 1)37, clustering sequences into operational taxonomic units (OTUs) of 97% distance-based similarity and identifying and removing potential chimeric sequences using both de novo and reference-based methods. Singleton sequences were removed before clustering. A custom Python script incorporating nucleo- tide BLAST, with NCBI RefSeq (Tatusova et al., 2013) as reference training set, was used to perform taxonomic assignment to the species level (E ≤ 1 × 10−10) using representa\ve sequences from each OTU.

RNA extraction.

Samples shown in Figures 1–3 were extracted using an acidic phenol–chloroform protocol. In brief, approximately 100 mg per sample was suspended in 700 μl of RNAlater (ThermoFisher). The suspension was homogenized using a sterile RNase-free spatula and incubated at 4 °C overnight. Samples were pelleted by centrifugation at 13,000g for 10 min and resuspended in 200 μl of RNA extraction buffer supplemented with proteinase K (1 mg ml−1) that was heat-activated at 50 °C for 10 min. Samples were incubated at room temperature for 10 min and vortexed every 2 min. Then, 300 μl of Qiagen RLT Plus Buffer (Qiagen) with β-mercaptoethanol (1%) was added to each sample, vortexed and incubated for 5 min at room temperature. Samples were then transferred to a sterile bead beating tube with 500 μl of 0.1 mm glass beads and 500 μl of acidic phenol:chloroform:isoamyl. Mechanical lysis was performed by bead beating the samples for 3 min (BioSpec Products), followed by one round of acidic phenol–chloroform extraction and one round of chloro- form extraction. RNA was precipitated with 50 μl of 3 M ammonium acetate and 500 μl of 100% isopropanol and incubated at −20 °C overnight. RNA was pelleted by centrifugation at 13,000g for 20 min at 4 °C and washed with 450 μl of 70% ethanol. Ethanol wash was repeated, and the pellet was allowed to air dry at room temperature for 5 min. The pellet was then dissolved in 50 μl of RNase-free water.

Subsequent samples were extracted using the RNeasy PowerMicrobiome kit (Qiagen), with addition of 100 μl phenol–chloroform–isoamyl alcohol (pH 6.5–8.0) as per manufacturer’s instructions.

RNA samples were purified using RNAClean XP (Agencourt), DNA contaminants were removed using TURBO DNA-Free kit (Life Technologies), and ribosomal RNA removed using Ribo-Zero rRNA Removal Kit (Illumina). Following ribosomal RNA depletion, RNAClean XP purification was repeated.

RNA sequencing and analysis.

RNA sample libraries were prepared using the TruSeq Stranded mRNA protocol (Illumina) and sequenced using the Illumina Miseq (samples shown in figures 1–2, 150 bp x 150 bp) or Hiseq (all other samples, 150 bp x 150 bp or 100 × 100 bp) platform (Illumina). Raw sequence reads were filtered using Trimmomatic (v.0.36) (Bolger et al., 2014), sequentially aligned to the mouse genome to remove mouse-derived sequences and then to each of the 4 CBBP genomes using bowtie2 (v.2.3.4.1) (Langmead and Salzberg, 2012). BAM files obtained were analyzed using R (v.3.2.1). Reads were assigned to genes using featureCounts (v.1.6.1) (Liao et al., 2013), and analyzed using ‘DeSeq2’ (v.1.14.0) (Love et al., 2014). GSEA was performed using the package ‘clusterProfiler’ (Yu et al., 2012).

Fluorescence in situ hybridization.

Intestinal tissues with luminal contents were carefully excised and fixed in freshly made nonaqueous Methacarn solution (60% methanol, 30% chloroform and 10% glacial acetic acid) as previously described for 6 h at 4 °C. Tissues were washed in 70% ethanol, processed with Leica ASP6025 processor (Leica Microsystems) and paraffin-embedded by standard techniques. Subsequently, 5-μm sections were baked at 56 °C for 1 h before staining. Tissue sections were deparaffinized with xylene (twice, 10 min each) and rehydrated through an ethanol gradient (95%, 10 min; 90%, 10 min) to water. Sections were incubated at 50 °C for 3 h with custom probes specific to:

BS: [Alexa488]- TACCGAAGTTCTTTAATCACGAGA -[Alexa488])

BP: [Alexa546]- TATAAGACTCAATCCGAAGAGATCAT -[Alexa546]

CB: [Alexa594]- GATTTGCTCCACATCACTGTC -[Alexa594]

PD: [Alexa647]- CAGCGATGAATCTTTAGCAAATATCC -[Alexa647]

Probes were diluted to 5 ng μl−1 in 0.9 M NaCl, 20 mM Tris-HCl at pH 7.2 and 0.1% SDS before use. Sections were later washed twice in 0.9 M NaCl, 20 mM Tris-HCl at pH 7.2 (wash buffer) for 10 min and counterstained with Hoechst (1:3,000 in wash buffer) for nuclear staining.

Image acquisition was performed with a Leica TCS SP5-II upright confocal microscope using a 63x oil immersion lens. Fiji (ImageJ) software was used to modify colors.

qPCR to confirm modulation of selected signature bacterial genes.

RNA was extracted from cecal content as indicated above, cDNA was generated using the QuantiTect reverse transcriptase kit (QIAGEN). Quantitative PCR was carried out using Sybr Green (Applied biosystems) and dedicated primers for the target genes and housekeeping genes selected from our RNAseq datasets based on low variance in expression in spite of host immune activation.

Generation of BS-MNC+ E. coli strain.

Chemically competent BL21(DE3) E. coli cells were transformed with a pETDuet-1 plasmid (Novagen) encoding for two BS genes annotated as Mn-catalase (peg. 1799, peg. 4964) under the regulation of an IPTG-inducible T7 promoter. The plasmid was constructed using the InFusion kit (Takara Bio) according to manifacturer’s instructions, via linearization of the plasmid (primers: linearize_F1: ccgctgagcaataactagc; linearize_R1:tctacgccggacgcat) and ligation with a 1537-bp gene block containing both MNC genes, spaced with the appropriate sequence of plasmid origin (and amplified with primers: gblock_F1: cacgatgcgtccggcgta; gblock_R1: tgctagttattgctcagcggt), so to perform a 1-step reaction. Amplification of the linearized plasmid/block was performed using the HiFi Premix CloneAmp (Takara Bio). Transformed clones were selected by plating on LB agar supplemented with 100 ug/ml of ampicillin. Function of the gene-encoded enzyme(s) was confirmed via challenge with H₂O₂ (50 mM, 15’) 1h following induction of transcription with 1 mM IPTG. Plating confirmed a ~100 fold increase in survival in MNC⁺ E. coli (not shown).

H₂O₂ survival assay.

BP, CB, PD as well as MNC+ E. coli and WT E. coli were cultured on Columbia blood agar plates (BD) for 24h and then inoculated in BHI+ medium (brain heart infusion broth supplemented with yeast extract (5 g/l) and l-cysteine (1 g/l)) in anaerobic conditions. Overnight cultures of BP, CB and PD, respectively, were mixed with an equal volume of overnight culture of either MNC+ E. coli or WT E. coli, doubled in volume with fresh medium and co-cultured anaerobically for 4h in the presence of IPTG 1 mM. Cultures were then challenged with 500 μM H₂O₂ for 1h, and viability of CB, BP, PD was assessed through plating on BHI+ plates supplemented with 100 μg/ml nalidixic acid, to eliminate E. coli.

Metabolomics Analysis

Targeted metabolomics were preformed using a modified version of a previously reported protocol (Haak et al., 2018). Frozen cecal contents were thawed and resuspended at 100 mg to 1 mL ice cold 80% methanol in 2mL screw cap vials containing 2.8 mm ceramic beads (Omni International). D3-acetate, D5-propionate, D7-butyrate and D6-succinic acid (Cambridge Isotope Laboratories) were included with the 80% methanol as internal standards. Samples were homogenized for 2×15 seconds using a bead beater (Biospec products). After homogenization, samples were centrifuged at 20 000 g at 4°C for 20 minutes. 100 mL of the supernatant was transferred to 2 mL glass mass-spectrometry (MS) vials (Agilent Technologies). 100mL of borate buffer (Thermo Scientific) at pH 10, 400mL of acetonitrile containing 150 mM pentafluorobenzyl bromide (Sigma Aldrich), and 400 mL of n-hexane (Sigma Aldrich) were added. Samples were then incubated at 65°C for 60 minutes with shaking at 600 rpm. After incubation, both undiluted and 1:10 dilutions of the n-hexane layer were transferred to fresh MS vials. Samples were then analyzed by GC-MS (8890 GC system and 5977B MSD-Agilent Technologies) running in negative chemical ionization mode with methane as the reagent gas. MS data was first analyzed in Mass Hunter Quantitative Analysis (Agilent Technologies) to give peak areas for the compounds analyzed. Peak areas were normalized to internal standard peak areas, and expressed as a concentration (for acetate, propionate, butyrate) by generating a calibration curve, or as a relative change compared to the average peak area in the control mouse group.

QUANTIFICATION AND STATISTICAL ANALYSIS

Cumulative data are shown as mean ± SEM or mean ± SD. Statistical tests used include Student’s t-test and Wilcoxon test, Pearson’s correlation. Unless otherwise stated, pvalues and FDR<0.05 were considered significant. Detailed information on the n of biological replicates and statistical tests used can be found in figure legends. Statistical analyses were performed using R (v.3.2.1) or GraphPad Prism (v.9.0.0).

Supplementary Material

3

Supplementary Table 1 (related to Figure 3). Shared and unique transcriptional adaptation signatures induced by flagellin or anti-CD3 in CBBP bacteria. Mice were reconstituted with CBBP for 14 days, then administered either anti-CD3 antibody or flagellin, and sacrificed 6h post-treatment to perform RNAseq on luminal RNA. Shown are genes for each bacterial strain that were significantly modulated by only one or both treatments (padj <0.05 and abs(log2FC) > 1).

4

Supplementary Table 2 (related to Figure 4). Comparison of baseline transcriptional differences in 3-mix vs 4-mix. GF mice were reconstituted with CBBP (4-mix) or a modified consortium from which BS had been withdrawn (only CB, BP and PD, referred to as 3-mix). Shown are genes resulting significantly modulated at baseline (comparison: 3mix-0h vs 4 mix-0h).

5

Supplementary Table 4 (related to Figure 4). Comparison of immunity-modulated bacterial genes in 3-mix and 4-mix. GF mice were reconstituted with CBBP (4-mix) or a modified consortium from which BS had been withdrawn (only CB, BP and PD, referred to as 3-mix). Mice were injected i.p. with anti-CD3 antibody and sacrificed after 6h to perform RNAseq on cecal content. Shown are genes resulting significantly modulated by bacteria in the context of either 3-mix, 4-mix or both, when comparing transcriptional levels at 6h post anti-CD3 injection and untreated mice (0h).

6

Supplementary Table 5 (Related to Figure 5). Operon prediction in CBBP (4mix) and CBP (3mix) dataset. GF mice were reconstituted with wither CBBP or a 3-mix lacking BS and injected anti-CD3 i.p. The resulting RNAseq datasets were utilized to predict operon structure (see STAR Methods). Reported is the operon composition predicted from the 2 datasets.

7

Supplementary Table 6. List of primers and probes utilized for this work.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
CD45 BV605	BioLegend	103139
CD11b PETR	ThermoFIsher	RM2817
Ly6G FITC	BD	551460
InVivo Mab anti-CD3e antibody clone 145-2C11	BioXCell	BE0001-1
Bacterial and Virus Strains
Blautia producta	Mouse isolate (https://doi.org/10.1016/j.chom.2017.04.002)	SAMN14679094
Bacteroides sartorii	Mouse isolate (https://doi.org/10.1016/j.chom.2017.04.002)	SAMN14754793
Clostridium bolteae	Mouse isolate (https://doi.org/10.1016/j.chom.2017.04.002)	SAMN14653269
Parabacteroides distasonis	Mouse isolate (https://doi.org/10.1016/j.chom.2017.04.002)	SAMN14645711
E. coli BL21 (DE3)	This manuscript	C600003
E. coli BL21 (DE3) p(BS)MNC	ThermoFisher	NA
Biological Samples
Chemicals, Peptides, and Recombinant Proteins
Ampicillin	ThermoFisher	69-52-3
Ultrapure flagellin from S. typhimurium	InvivoGen	tlrl-epstfla-5
Critical Commercial Assays
DyNAmo SYBR Green qPCR Master Mix	ThermoFisher	F-410L
QuantiTect Reverse Transcription Kit	Qiagen	205311
RNeasy PowerMicrobiome Kit	Qiagen	26000-50
TaqMan Universal PCR Master Mix	Applied Biosystems	4305719
SYBR™ Green PCR Master Mix	Applied Biosystems	4364344
Taqman Probe for Reg3g	ThermoFisher	Mm00441127_m1
Taqman Probe for Ifng	ThermoFisher	Mm01168134_m1
Taqman Probe for Tnfa	ThermoFisher	Mm00443258_m1
Taqman Probe for Il22	ThermoFisher	Mm00444241_m1
Taqman Probe for Lcn2	ThermoFisher	Mm01324470_m1
Taqman Probe for Nos2	ThermoFisher	Mm00440502_m1
Taqman Probe for Hprt	ThermoFisher	Mm00446968_m1
InFusion HD Cloning Kit	Takara	638910
Deposited Data
Shotgun, 16s sequencing and transcriptomics reads	This manuscript	BioProject # PRJNA672114
R and bash codes	This manuscript	https://github.com/simobeca/CBBP-transcriptomics
Experimental Models: Cell Lines
Experimental Models: Organisms/Strains
Mouse: C57BL/6J	The Jackson Laboratory	JAX: 000664; RRID: IMSR_JAX:000664
Oligonucleotides
For primers and probes see Supplementary Table 6	This Study	NA
Recombinant DNA
Software and Algorithms
Trimmomatic	https://doi.org/10.1093/bioinformatics/btu170	NA
Bowtie2	https://doi.org/10.1038/nmeth.1923	NA
Deseq2	https://doi.org/10.1186/s13059-014-0550-8	NA
ClusterProfiler	https://doi.org/10.1089/omi.2011.0118	NA
Rsubread (FeatureCounts)	https://doi.org/10.1093/nar/gkz114	NA
Rockhopper	https://doi.org/10.1016/j.ymeth.2019.03.026	NA
GraphPad Prism v. 9.0.0	GraphPad Software	NA
MacVector v.15.5.0	MacVector	NA
SnapGene v 5.2	GSL Biotech LLC	NA
Muscle	https://doi.org/10.1186/1471-2105-5-113	NA
ggplot2	Wickham H (2016). ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York. ISBN 978-3-319242774, https://ggplot2.tidyverse.org.	NA
dplyr	Hadley Wickham, Romain François, Lionel Henry and Kirill Müller (2020). dplyr: A Grammar of Data Manipulation. R package version 1.0.2. https://CRAN.R-project.org/package=dplyr	NA
ggpubr	Alboukadel Kassambara (2020). ggpubr: ‘ggplot2’ Based Publication Ready Plots. R package version 0.4.0. https://CRAN.R-project.org/package=ggpubr	NA
R v 3.2.1	R Core Team (2018). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available online at https://www.R-project.org/.	NA
Other

Highlights.

• Acute immune activation alters intestinal microbiota gene transcription within hours

• Microbial transcriptional signatures include stress and metabolism-associated genes

• Microbiota composition and the type of immune stimulus impact bacterial responses

• The intestinal metabolome is rapidly reshaped by immune activation of the host OC Blurb