Infection trains the host for microbiota-enhanced resistance to pathogens

Apollo Stacy; Vinicius Andrade-Oliveira; John A McCulloch; Benedikt Hild; Ji Hoon Oh; P Juliana Perez-Chaparro; Choon K Sim; Ai Ing Lim; Verena M Link; Michel Enamorado; Giorgio Trinchieri; Julia A Segre; Barbara Rehermann; Yasmine Belkaid

doi:10.1016/j.cell.2020.12.011

. Author manuscript; available in PMC: 2022 Feb 4.

Published in final edited form as: Cell. 2021 Jan 15;184(3):615–627.e17. doi: 10.1016/j.cell.2020.12.011

Infection trains the host for microbiota-enhanced resistance to pathogens

Apollo Stacy ^1,^2,^3,^*, Vinicius Andrade-Oliveira ^1,⁴, John A McCulloch ⁵, Benedikt Hild ⁶, Ji Hoon Oh ⁶, P Juliana Perez-Chaparro ², Choon K Sim ^7,⁸, Ai Ing Lim ¹, Verena M Link ¹, Michel Enamorado ¹, Giorgio Trinchieri ⁵, Julia A Segre ⁷, Barbara Rehermann ⁶, Yasmine Belkaid ^1,^2,^9,^*

PMCID: PMC8786454 NIHMSID: NIHMS1654488 PMID: 33453153

Summary

The microbiota shields the host against infections in a process known as colonization resistance. How infections themselves shape this fundamental process remains largely unknown. Here, we show that gut microbiota from previously infected hosts display enhanced resistance to infection. This long-term functional remodeling is associated with altered bile acid metabolism leading to the expansion of taxa that utilize the sulfonic acid taurine. Notably, supplying exogenous taurine alone is sufficient to induce this alteration in microbiota function and enhance resistance. Mechanistically, taurine potentiates the microbiota’s production of sulfide, an inhibitor of cellular respiration, which is key to host invasion by numerous pathogens. As such, pharmaceutical sequestration of sulfide perturbs the microbiota’s composition and promotes pathogen invasion. Together, this work reveals a process by which the host, triggered by infection, can deploy taurine as a nutrient to nourish and train the microbiota, promoting its resistance to subsequent infection.

Graphical Abstract

graphic file with name nihms-1654488-f0001.jpg

Introduction

The microbiota of host metaorganisms presents a formidable barrier to invasion by pathogens. Highlighting this, collateral damage to the microbiota by antibiotic therapy is often followed by the expansion of highly intractable antibiotic-resistant pathogens. Globally, antibiotic-resistant pathogens kill over 700,000 people per year, and over the next several decades, are expected to continue to represent a major cause of worldwide mortality (United Nations, 2019). For this reason, it is imperative to develop strategies to sustain or improve host antimicrobial defenses. One such arm of metaorganism immunity is colonization resistance, or the ability of the microbiota to resist pathogen colonization. Colonization resistance comprises via numerous inhibitory mechanisms that can be either direct or indirect/host-mediated (Kim et al., 2017). The factors controlling colonization resistance, as well as the multiple mechanisms associated with this fundamental function, still remain to be fully explored.

A defining feature underlying the evolutionary success of the immune system is its ability to develop both innate and adaptive memory against pathogens, promoting more rapid and robust host responses to subsequent infections. Though memory was once exclusively attributed to the adaptive immune system, it is now appreciated that the innate immune system can also be trained to fully mature and attain enhanced function (Netea et al., 2020). Moreover, maturation of the adaptive and innate immune systems results not only through infections but, beginning in infancy, through numerous other exposures such as diet and the microbiota (Macpherson et al., 2017). Whether host memory of previous encounters reciprocally shapes the microbiota remains unclear.

The microbiota like the immune system is highly dynamic. Notably, age-associated shifts in the microbiota strongly influence the microbiota’s functional capacity and host fitness (Stewart et al., 2018). A key example is that the immature microbiota of neonates is highly susceptible to pathogen invasion (Singer et al., 2019). Factors that modulate the development and maintenance of the microbiota include cross-talk with the immune system (McLoughlin et al., 2016), antibiotic exposure (Roubaud-Baudron et al., 2019), and defined dietary components (Wotzka et al., 2019). Increasingly appreciated is also the role of pathogens in remodeling the microbiota (Kamdar et al., 2016; Molloy et al., 2013). For instance, by eliciting an overly exuberant immune response, pathogens can gain access to inflammatory byproducts, such as elevated oxygen in normally anaerobic niches, that enable utilization of non-fermentable, strictly respiratory substrates for growth (Lopez et al., 2016).

Due to the prevalence of infections throughout evolution, we hypothesize that host metaorganisms have adapted strategies to harness infection-elicited microbiota for their own benefit and greater defense against subsequent infection. A corollary of this hypothesis is that regular exposure to pathogens may be beneficial for the microbiota to develop optimal antimicrobial function. In industrialized societies, however, improved hygiene and sanitation is thought to have reduced exposure to certain classes of pathogens, contributing to a recent precipitous decline in microbiota diversity and beneficial functions (Sonnenburg and Sonnenburg, 2019; Ramanan et al., 2016). Aligning with this, studies with the microbiota of wild and pet-shop mice (Beura et al., 2016; Rosshart et al., 2017, 2019), which inevitably encounter more infections than their laboratory counterparts, have demonstrated that mice with a history of infection exhibit altered immunity (Reese et al., 2016) and heightened resistance to perturbations of their gut microbiota (Rosshart et al., 2019).

Here we examined the possibility that prior infections could enhance metaorganism colonization resistance and that dissecting the underlying molecular mechanism would lead to the identification of novel microbiota-directed therapeutics. Our results uncover that the host, triggered by transient infection, deploys the sulfonic acid taurine as a nutrient to promote long-term resistance of the microbiota to subsequent infection.

Results

Infection-trained microbiota enhance colonization resistance.

To assess the impact of infection on the ability of the microbiota to promote colonization resistance, we first sought a model pathogen whose colonization of the host is dictated primarily by the microbiota. We therefore chose the hospital-associated pathogen Klebsiella pneumoniae (Kpn) (CDC, 2019), specifically a derivative of a carbapenem-resistant clinical isolate (Ramage et al., 2017). Though extraintestinal infections with Kpn can be life-threatening, otherwise healthy individuals can harbor Kpn at low levels in their gut microbiota, and hospitalized patients colonized with Kpn are at much greater risk for subsequent systemic infection (Conlan et al., 2016; Gorrie et al., 2017). In mice, following oral infection with a high dose, Kpn can be transiently detected in the lumen of the large intestine before it is no longer detectable in the stool by 2 days post-infection (Figures 1A–B). In contrast, mice orally pre-treated with streptomycin, a broad-spectrum antibiotic, sustain a high fecal burden of Kpn (Figure 1B). These results support the idea that the microbiota plays a dominant role in controlling Kpn colonization, making this pathogen suitable for our studies of microbiota-dependent colonization resistance.

To examine how prior infection impacts resistance to Kpn, we employed 2 separate mouse models: (1) the offspring of a specific-pathogen-free (SPF) mouse colony whose germ-free (GF) founders received the microbiota of wild mice (wildR model) (Rosshart et al., 2017) and (2) SPF mice previously infected with an attenuated strain (∆yopM) of the food-borne pathogen Yersinia pseudotuberculosis (post-∆yopM model) (Han et al., 2017) (Figure 1C). The first model is particularly appropriate for addressing our question since, in their natural environment, wild mice are exposed to far more infections than standard laboratory mice. We acknowledge, however, that many factors besides prior infection could also contribute to the altered microbiota of wild mice. Hence, we also turned to the more defined post-∆yopM model. A key advantage of this model is that ∆yopM locally expands and induces gut inflammation but, unlike the Y. pseudotuberculosis (Yptb) wild type, is completely cleared within a few days post-infection (Han et al., 2017). As such, this model allowed us to assess the impact of a transient infectious process on colonization resistance, independently of pathogen persistence.

Following oral infection, wildR mice displayed enhanced resistance to Kpn as compared to control SPF mice (Figure 1D). Because wildR mice were generated by transferring the microbiota of wild mice into GF mice (Figure 1C), these results support the idea that the enhanced resistance of wildR mice was primarily mediated by the microbiota. Furthermore, we also observed enhanced resistance to Kpn in mice previously infected with ∆yopM (Figure 1E). A time course of resistance to Kpn following ∆yopM infection showed that enhanced resistance is detectable only after 4 weeks post-∆yopM infection and persists to at least 15 weeks post-∆yopM infection (Figures 1E and S1A).

To distinguish the role of the microbiota versus host responses in ∆yopM-enhanced resistance, we first considered the possibility that prior infection may accelerate intestinal expulsion. However, intestinal transit time was not impacted following ∆yopM infection (Figure S1B). Next, we broadly surveyed both innate and adaptive immune responses in the lamina propria of the large intestines (Figures S2A–C), the primary site of Kpn colonization (Figure 1A). The most significant immune alteration that we identified was that, in mice at 4 weeks post-∆yopM infection, CD8⁺ T cells exhibit a heightened potential to produce the cytokine interferon (IFN)-γ (Figures S2C–D). Thus, ∆yopM-enhanced resistance could result from altered mucosal immune responses and/or changes to the microbiota. While both phenomena could play a role, we decided to focus our exploration on the microbiota.

To this end, we conventionalized GF mice with fecal microbiota from post-∆yopM mice or control naïve mice (Figure 1C). As previously reported (Han et al., 2017), we could not detect ∆yopM in the transferred microbiota (Figure S1C). Using this approach, we found that the microbiota from post-∆yopM mice was sufficient to confer enhanced resistance to Kpn (Figure 1F). Notably, resistance resulting from microbiota transfer occurred in the absence of enhanced CD8⁺ T cell responses (Figure S2E). These findings illustrate that a single mild infection is sufficient to remodel the microbiota in a way that enhances resistance to Kpn.

We next analyzed the community compositions of the post-∆yopM and wildR microbiota, with the goal of identifying commonly enriched taxa that may enhance colonization resistance. Thus, we examined the 16S rRNA gene profiles of the wildR microbiota (F2 generation from a previously published study (Rosshart et al., 2017)) and the post-∆yopM microbiota. Unlike the post-∆yopM microbiota (which was compared to naïve SPF microbiota as a control), the wildR microbiota exhibited higher alpha diversity compared to its control (microbiota of F2 labR mice, or mice whose GF founders received the microbiota of laboratory mice (Rosshart et al., 2017)) (Figure S1D). On the other hand, both wildR and post-∆yopM microbiota profiles clustered separately from their controls according to beta diversity analyses (Figures 1G–H), and this separation persisted to the F10 (contemporary) generation of wildR mice and was maintained after transfer of the post-∆yopM microbiota into GF mice (Figures S1E–F).

While we did not discern consistent major shifts in highly abundant phyla (Figures S1G–H), a normally minor phylum, Proteobacteria, was the only phylum to be enriched in both the post-∆yopM and wildR microbiota (both F2 and F10 generations) (Figure S1H and Table S1), expanding about 100-fold from 0.01% to 1% following infection with ∆yopM and in wildR mice compared to control mice (Figure 1I). Several members of the Proteobacteria phylum, especially those within the class Gammaproteobacteria, are known to expand in the context of infection and inflammation (Molloy et al., 2013). However, in our settings, the only Proteobacteria classes to be significantly increased in both the post-∆yopM and wildR microbiota (F2 and F10), as compared to their control microbiota, were the Deltaproteobacteria and the Betaproteobacteria (Figures 1J and S1I). Furthermore, following transfer of the post-∆yopM microbiota into GF mice, the only Proteobacteria species to remain enriched was a deltaproteobacterium (Figure S1J and Table S1). Thus, microbiota profiles associated with enhanced resistance to Kpn are enriched for Deltaproteobacteria, supporting the idea that these bacteria may promote colonization resistance.

Infection-trained microbiota are enriched for taurine-utilizing taxa.

Due to the enrichment of Proteobacteria in both the post-∆yopM and wildR microbiota, we hypothesized that they may harbor common community functions. To probe this possibility, we examined the shotgun metagenomes of the wildR (F2 generation) and post-∆yopM microbiota. Relative to its control metagenome, the post-∆yopM metagenome was enriched for 46 functions, while the wildR metagenome was enriched for 379 functions (Enzyme Commission numbers (McDonald et al., 2009)) (Figure 2A and Table S2). Of these enriched functions, 18 were shared (Figure 2B), and 6 of these 18 could be assigned to 2 metabolic pathways: sulfur metabolism and menaquinone biosynthesis (Figures 2C–D).

Figure 2. — (A) Fold changes in the median abundance of functions (EC numbers with log₂ fold change > 0.5, P < 0.05) in the shotgun metagenome of post-*∆yopM* mice at >15 weeks post-*∆yopM* infection (compared to naïve SPF mice) vs. F2 wildR mice (compared to F2 labR mice). Functions differentially abundant in only post-*∆yopM* mice are black diamonds; only wildR, gray circles; enriched in both post-*∆yopM* and wildR, blue; differentially abundant in both post-*∆yopM* and wildR, red.

(B) Number of functions co-enriched in the post-*∆yopM* and F2 wildR metagenomes.

(C) Fold change in the median abundance of functions enriched in the metagenome of post-*∆yopM* vs. naïve SPF mice (n = 11) and F2 wildR vs. F2 labR mice (n = 10). Post-*∆yopM* bars show mean ± SEM of 3 paired cages.

(D) Pathways enriched in the post-*∆yopM* metagenome. Pie charts represent the relative contribution of taxa to a function. A, *mqnA* (same for E, X, B, and C); MQ, menaquinone; other, aggregate of rare taxa.

(E) Model for *Deltaproteobacteria* energy generation. MQ, oxidized menaquinone; MQH₂, reduced menaquinone.

(F) Abundance of taurine in the cecal contents of control and post-*∆yopM* mice at >15 weeks post-*∆yopM* infection (3 pooled experiments, n = 13–14, fold increase = 4.7). a.u., arbitrary units.

(G) The microbiota converts primary (1°) bile acids to taurine and secondary (2°) bile acids.

(H) Abundance of total taurine-conjugated bile acids in the cecal contents of control and post-*∆yopM* mice at >15 weeks post-*∆yopM* infection (3 pooled experiments, n = 12–14, fold increase = 1.6).

(I) Growth yield (measured by absorbance at 600 nm) of *B. wadsworthia* in rich media + vehicle, taurine, or sulfate. Bars show mean ± SEM (1 experiment, n = 4).

(J) *Kpn* CFU in the feces of ex-GF mice, conventionalized with SPF microbiota ± pre-engraftment with *B. wadsworthia* (Bw), at 1 day post-*Kpn* infection (2 pooled experiments, n = 10, fold decrease = 5.3). The dotted line indicates the detection limit.

(K) Model: Elevated taurine post-infection enriches for gut taxa that utilize taurine.

In box and whisker plots, lines connect the medians of paired cages. n.s., not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.01 by Mann-Whitney (A, C, F, H-J) or hypergeometric (B) test. See also Figure S3 and Tables S2–S3.

In agreement with the expansion of Deltaproteobacteria in the post-∆yopM and wildR microbiota (Figure 1J), the primary source of the sulfur metabolism pathway was a deltaproteobacterium (Desulfovibrio genus) in the post-∆yopM metagenome (blue wedges, Figure 2D) and a mixture of Deltaproteobacteria in the wildR metagenome (blue wedges, Figure S3A). A prevalent characteristic of Deltaproteobacteria is the ability to gain energy from anaerobic respiration with sulfur-containing compounds such as taurine or sulfate (Peck et al., 2019) (Figure 2E). Specifically, these bacteria first convert taurine or sulfate to sulfite via taurine-pyruvate aminotransferase (encoded by tpa) or sulfate adenylyltransferase (sat), respectively. Then, they reduce sulfite to sulfide via dissimilatory sulfite reductase (dsr) using electrons donated by menaquinone electron shuttles (Barton et al., 2017) (Figure 2E). To distinguish whether taurine and/or sulfate drives the Deltaproteobacteria expansion in the post-∆yopM and wildR microbiota, we assessed the abundance of tpa and sat in their metagenomes. While tpa was significantly higher in both metagenomes, sat showed a modest increase in only the wildR metagenome, suggesting that taurine rather than sulfate contributes to the expansion of Deltaproteobacteria (Figure 2C).

Because metagenomics can only give insight into the microbiota’s functional potential, we next set out to more directly determine if taurine is indeed elevated in the gut following infection. To this end, we profiled the abundance of 186 defined metabolites using gas chromatography time-of-flight mass spectrometry and in total identified 13 metabolites that were more abundant in the cecal contents of post-∆yopM mice than those of control mice (Figures 2F, S3B, and Table S3). Five of these 13 metabolites were the substrate, precursors, or byproducts of xanthine oxidase (metabolites in bold in Figure S3B), a host antimicrobial effector known to be induced by enteric pathogens (Martin et al., 2004). Notably and confirming our metagenomics-based prediction, taurine was the most significantly increased metabolite in post-∆yopM mice (Figures 2F and S3B).

In the gut, a dominant source of taurine is bile acids (Wahlström et al., 2016). These host detergents are synthesized and conjugated to taurine in the liver, stored in the gallbladder, and released into the gut to aid fat/oil digestion (Figure 2G). Once in the gut, bile acids are chemically transformed by specific members of the microbiota. These transformations include deconjugation, which releases taurine from primary (host-synthesized) bile acids, making taurine bioavailable to the entire microbiota, and dehydroxylation into secondary (microbiota-derived) bile acids (Wahlström et al., 2016) (Figure 2G). The latter activity is encoded by the bai operon, and a gene within this operon (baiE) was jointly enriched in the post-∆yopM and wildR metagenomes, suggesting heightened bile acid-directed activity in the gut microbiota of these mice (Figure 2C).

Because bile acids are synthesized in the liver, we assessed this organ for broad changes post-infection. Post-∆yopM mice, however, did not exhibit gross signs of liver damage based on histology (Figures S3D–E). On the other hand, whole-tissue RNA-seq of post-∆yopM livers revealed differential expression of 894 genes. The 792 upregulated genes were functionally enriched for pathways (Gene Ontology terms (Ashburner et al., 2000)) related to inflammation, cytokine production, and leukocyte activity (Figure S3F and Table S4). In line with this, post-∆yopM liver-derived γδ, CD4⁺, and CD8⁺ T cells exhibited heightened production of various pro-inflammatory cytokines (Figure S3G). Although these cytokines, in particular IL-17, are associated with bile acid-induced inflammation (Li et al., 2017), only 3 genes (Vdr, Slc10a6, Slco1a1) mediating bile acid signaling/transport were differentially expressed in post-∆yopM livers, suggesting that most ∆yopM-induced changes to bile acid metabolism may occur post-transcriptionally or outside the liver. Supporting this, post-∆yopM mice sustained enlarged gallbladders (Figure S3H) and higher intestinal levels of total bile acids (Figure S3I). Thus, a transient infection can induce long-term heightened immunity within the liver, the organ responsible for synthesizing bile acids, along with alterations to the gallbladder, the organ responsible for storing bile acids. Of interest, wildR mice also sustained higher intestinal levels of total bile acids (Figure S3J) but lower levels of taurine, a potential result of enhanced consumption by the microbiota (Figure S3C).

Because bile acids are highly heterogeneous, we next profiled the abundance of 23 defined bile acids using liquid chromatography-mass spectrometry. Total taurine-conjugated bile acids as well as individual primary and secondary bile acids were elevated in the cecal contents of post-∆yopM mice (Figures 2H, S3K, and Table S3). Collectively, these results propose that even a transient infectious encounter can have long-term consequences on host bile acid metabolism, resulting in a sustained increase in intestinal taurine, a phenomenon that may lead to the expansion of taurine-utilizing Deltaproteobacteria (Figure 2K).

To assess whether taurine-utilizing Deltaproteobacteria contribute to colonization resistance, we chose the deltaproteobacterium Bilophila wadsworthia ATCC 49260 as a model. As predicted for the Deltaproteobacteria in the post-∆yopM and wildR metagenomes (Figure 2C), this strain flourishes on taurine, but not sulfate (Figure 2I), in a menaquinone-dependent manner (Figure S3L). To preferentially enrich the gut microbiota with B. wadsworthia, we colonized GF mice with mono-cultures of B. wadsworthia prior to conventionalizing these mice with complex SPF microbiota. This procedure resulted in a nearly 8-fold increase in B. wadsworthia (Figure S3M and Table S1) and was sufficient to enhance resistance to Kpn (Figure 2J). Together, these results support the idea that the expansion of taurine-utilizing Deltaproteobacteria can directly contribute to colonization resistance (Figure 2K).

Taurine-trained microbiota enhances colonization resistance.

Our results thus far support the idea that elevated taurine may by itself promote colonization resistance. To test this, we placed mice onto taurine-supplemented drinking water for 2–3 weeks prior to infection with Kpn. (Figure 3A). Fecal measurements showed that this procedure was able to slightly increase the available taurine in the distal colon without altering total bile acids (Figure 3B). Notably, even a slight increase in colonic taurine levels was sufficient to increase resistance to Kpn (Figure 3C). Suggesting a role for the gut microbiota, taurine supplementation did not enhance intestinal expulsion or resistance to Kpn airway infection (Figures S4A–C).

Figure 3. — (A) Scheme for taurine treatment. Mice were placed onto drinking water (vehicle) ± 200 mM taurine for 2–3 weeks prior to *Kpn* infection.

(B) Abundance of taurine and total bile acids in the feces of vehicle and taurine-treated mice (1 experiment with 2 paired cages, n = 10, taurine fold increase = 2.1).

(C) *Kpn* CFU in the feces of vehicle and taurine-treated mice at 1 day post-*Kpn* infection (8 pooled experiments, n = 58–59). Lines connect the means of paired cages (fold decrease = 4.5; for medians of paired cages, fold decrease = 2.1).

(D) *Kpn* CFU in the feces of ex-GF mice, conventionalized with the microbiota of vehicle or taurine-treated mice, at 1 day post-*Kpn* infection (2 pooled experiments, n = 12, fold decrease = 11).

(E) *C. rodentium* CFU in the feces of vehicle and taurine-treated mice at 5 days post-*C. rodentium* infection (5 pooled experiments, n = 39–40, fold decrease = 241).

(F) Ten most differentially abundant species that were significant in at least 1 of 3 paired groups of vehicle and taurine-treated mice (experiment A, B, and/or C). Species are named according to their lowest taxonomic classification. Order, o; family, f; genus, g; *Lachno.*, *Lachnospiraceae*; *Rumino., Ruminococcaceae*.

(G) Median abundance of functions (EC numbers) in the shotgun metagenome of taurine vs. vehicle-treated mice. Significant functions (log₂ fold change > 0.5, P < 0.05) are red (2 pooled experiments, n = 9). *dsr*, dissimilatory sulfite reductase (fold increase = 8.3); ppm, parts per million (read counts assigned to a function per million total read counts).

(H) Pie chart representing the relative contribution of species to *dsr*. Species are named according to their lowest taxonomic classification (order, o; genus, g). Figure 1. Infection-trained microbiota enhance colonization resistance

(I) Hydrogen sulfide released by the gut microbiota of vehicle and taurine-treated mice during *ex vivo* culture ± taurine (1 experiment with 2 paired cages, n = 9–10, fold increase = 1.6). Data are representative of 2 experiments.

(J) Model: Taurine enriches for gut taxa that generate sulfide.

In box and whisker plots, lines connect the medians of paired cages, unless otherwise indicated. Dotted lines in C, D, and E indicate the detection limit. n.s., not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001 by Mann-Whitney test (B-G, I). See also Figure S4 and Tables S1–S2.

Besides serving as a microbial nutrient, taurine has been shown to stimulate host antimicrobial defenses (Levy et al., 2015). To distinguish whether the microbiota or host primarily mediates taurine-enhanced resistance, we first assessed immune cell composition in the large intestinal lamina propria and epithelium via flow cytometry and RNA-seq, respectively. These approaches revealed that taurine alone elicits neither innate nor adaptive immune responses in the lamina propria and only modestly alters gene expression in purified epithelial cells (all log₂ fold changes < 0.5, Figures S4D–F and Table S4). Interestingly, nearly half (10/21) of the upregulated genes encoded ribosomal proteins, suggesting that taurine primarily stimulates epithelial cell growth and translation rather than antimicrobial defense. Next, we assessed whether taurine can enhance resistance to Kpn in mice devoid of microbiota (Figure S4G). However, taurine treatment of GF mice did not alter protection (Figure S4H). While less powered than our SPF experiments (Figure 3C), our GF experiments nevertheless suggest that, under certain settings, the microbiota may be required for taurine-enhanced resistance. To confirm this, we transferred the fecal microbiota of taurine or vehicle-treated mice into GF mice. Further supporting a role for the microbiota, the taurine-trained microbiota promoted greater resistance to Kpn (Figure 3D).

To determine if taurine can enhance resistance to other pathogens, we utilized Citrobacter rodentium, a murine model for pathogenic E. coli (Figure S4I). C. rodentium primarily colonizes the colon, where one of its virulence strategies is to induce the hyperplasia of colonic crypts. Because epithelial colonocytes represent a major source of oxygen in the colon lumen, crypt hyperplasia promotes C. rodentium aerobic respiration and luminal proliferation (Lopez et al., 2016). Beginning at day 3 post-infection, taurine induced a 1 to 3 log decrease in C. rodentium fecal burden (Figures 3E and S4J), and was also able to mitigate crypt hyperplasia (Figures S4K–L). Thus, these results further demonstrate taurine’s ability to enhance colonization resistance.

We next assessed the impact of taurine on gut microbial ecology. Because pathways for metabolizing taurine are taxonomically widespread (Cook and Denger, 2006), we predicted that diverse taxa would have the potential to benefit and expand in response to taurine. In line with this, 16S profiling revealed that taurine remodeling of the gut microbiota is highly variable at the taxonomic level (Figure 3F and Table S1). This remodeling, however, was conserved at the functional level, as revealed by shotgun metagenomics, where one of the largest increases was in the gene encoding dissimilatory sulfite reductase (dsr) (Figure 3G and Table S2). The dsr gene product catalyzes the final step in taurine metabolism (Figure 2E), and consistent with taurine shaping the post-∆yopM and wildR microbiota, both of their metagenomes were also enriched for dsr (Figure 2C). In the taurine-trained metagenome, the primary sources of dsr were Clostridiales (Figure 3H), and accordingly, many Clostridiales also expanded following taurine treatment (Figure 3F). Of note, these taxa were distinct from the primary source of dsr (Deltaproteobacteria) in the post-∆yopM and wildR microbiota (Figures 2D and S3A). Thus, both taurine supplementation and prior infection enrich for members of the microbiota that functionally share the capacity for metabolizing taurine via dsr.

A major byproduct of taurine metabolism, resulting directly from dsr activity, is the poisonous gas hydrogen sulfide (Barton et al., 2017). Based on our metagenomics data (Figure 3G), we predicted that the taurine-trained microbiota would produce more sulfide than vehicle-treated microbiota. To test this, we utilized an ex vivo culture-based approach, allowing us to modulate the microbiota’s growth environment. Confirming our metagenomics-based prediction, the microbiota of taurine-treated mice was highly responsive to taurine addition ex vivo, producing significantly more sulfide than the microbiota of vehicle-treated mice (Figure 3I). Together, these data support the idea that training of the microbiota with taurine, a bile acid-derived metabolite that is elevated by pathogenic exposures, potentiates the microbiota’s ability to produce sulfide (Figure 3J).

Taurine-derived sulfide inhibits pathogen respiration.

At high concentrations, hydrogen sulfide can inhibit the activity of cytochrome oxidases, the terminal enzymes in aerobic electron transport chains (Forte et al., 2016). Because aerobic respiration is exploited by pathogens to access non-fermentable growth substrates (Lopez et al., 2016) and because taurine potentiates sulfide production by the microbiota (Figure 3I), we proposed the following model for taurine-enhanced colonization resistance: at low levels of taurine, oxygen enables pathogens to access non-fermentable substrates, whereas at high levels of taurine (such as in post-∆yopM or taurine-supplemented mice), taurine-derived sulfide restricts pathogens from these substrates (Figure 4A).

To test our model, we initially employed a high-throughput approach, transposon-seq (Tn-seq) (Chao et al., 2016), to screen Kpn Tn mutants with altered fitness in vivo. To perform Tn-seq, we passaged an input pool of 12,000 Kpn Tn mutants through control and post-∆yopM mice (Ramage et al., 2017). With this experimental design, if a particular gene preferentially promoted fitness in 1 of the 2 mouse groups, mutants in that gene would be preferentially depleted from that group’s output pool (Figure S5A). To identify such mutants, we amplified and deep-sequenced Kpn genomic DNA adjacent to Tn insertions, allowing us to determine both their genomic location and abundance in the output pools. In total, we identified 125 and 79 genes that differentially promote Kpn fitness in control and post-∆yopM mice, respectively (Table S5).

Based on our model (Figure 4A), we first queried our Tn-seq results for Kpn genes that mediate anaerobic respiration. In bacteria, respiration can be performed not only with oxygen but also anaerobic electron acceptors, such as nitrate. While defined enteric pathogens have been demonstrated to exploit anaerobic respiration to colonize the gut (Faber et al., 2017), our model proposes that taurine-derived sulfide specifically inhibits aerobic respiration. Accordingly, Tn-seq confirmed the inability of genes that mediate anaerobic respiration to differentially contribute to Kpn fitness in either control or post-∆yopM mice (Figure S5B).

We next queried our Tn-seq results for Kpn genes that mediate growth on non-fermentable substrates. Based on our model, we focused on substrates, such as formate (Lopez et al., 2016), that gut pathogens have been reported to exploit in vivo via aerobic respiration (Figure S5B). Among the genes that we examined, most striking was that 4 of the 21 genes in the 1,2-propanediol utilization (pdu) locus promoted Kpn fitness in control rather than post-∆yopM mice (enrichment P = 0.003, hypergeometric test) (Figures 4B and S5B). 1,2-propanediol is a byproduct of fermentation by the microbiota, but being strictly respiratory, it itself can only be utilized in the presence of an electron acceptor (Faber et al., 2017). While anaerobic electron acceptors could fuel Kpn growth on 1,2-propanediol, we found that Kpn grew maximally on 1,2-propanediol with oxygen (Figure 4C).

Based on this finding, we next queried our Tn-seq results for terminal oxidases that potentially support Kpn growth on 1,2-propanediol in vivo. Indeed, Tn-seq identified a putative oxidase that, like the pdu locus, contributed to Kpn fitness in control mice (Figure S5B). The gene encoding this oxidase was homologous (>68% identical in amino acid sequence) to the cyxB-encoded subunit of the Salmonella cytochrome oxidase bd-II (Rivera-Chávez et al., 2016). To confirm that Kpn cyxB promotes 1,2-propanediol utilization, we competed the wild type (WT) against a cyxB^- mutant in minimal media with 1,2-propanediol as the sole carbon source. At the level of oxygen in air (21%), the cyxB^- mutant exhibited fitness comparable to the WT (Figure 4D). On the other hand, at levels of oxygen comparable to those found in the gut (<1%) (Rivera-Chávez et al., 2016), the Kpn WT robustly outcompeted the mutant (Figure 4D). Collectively, these data support the idea that Kpn utilizes 1,2-propanediol in vivo by scavenging oxygen via cytochrome oxidase bd-II.

As a further test of our model (Figure 4A), we next assessed whether sulfide could directly block 1,2-propanediol utilization. Indeed, >250 µM sulfide potently inhibited the ability of Kpn to grow on 1,2-propanediol with oxygen (Figure 4E). In contrast, precursors to sulfide, the taurine-conjugated bile acid taurocholate and taurine, were not inhibitory, suggesting that conversion of these precursors to sulfide is necessary for inhibition (Figure 4E). Another prediction of our model is that under low-taurine/sulfide conditions, when respiration is possible, Kpn should experience a growth advantage and thus be able to outcompete a Kpn respiration mutant. In contrast, under high-taurine/sulfide conditions, when respiration is limited, Kpn should not experience a growth advantage and thus be more equal in fitness to a Kpn respiration mutant (Figure 4A). To test these predictions, we competed the Kpn WT and cyxB^- respiration mutant in mice treated with the vehicle or taurine. As we predicted, in vehicle-treated mice, the WT outcompeted the respiration mutant, whereas in taurine-treated mice, the WT and mutant exhibited comparable fitness (Figures 4F and S5E). Similarly, in vehicle-treated mice, the WT outcompeted a 1,2-propanediol utilization mutant (pduQ^-), whereas in taurine-treated mice, the WT and mutant again exhibited comparable fitness (Figures 4F, S5C, and S5E). Further supporting our model, a Kpn mutant (acrE^-) capable of respiring with 1,2-propanediol could not be outcompeted by the WT in either vehicle or taurine-treated mice (Figures S5C–E).

We next tested our model in the context of C. rodentium infection since, like Kpn, C. rodentium exploits aerobic respiration to colonize the gut (Lopez et al., 2016). Specifically, C. rodentium leverages the cytochrome oxidase bd-I, encoded by cydAB, to access non-fermentable substrates such as formate (Lopez et al., 2016) (Figure S5F). As predicted, >500 µM sulfide blocks the ability of C. rodentium to grow on formate with oxygen (Figure S5G). Furthermore, in vehicle-treated mice, the C. rodentium WT ultimately outcompeted a cydAB^- respiration mutant, whereas similar to our observations for Kpn, the competitive advantage of the WT was impeded in taurine-treated mice (Figure S5H). Together, these findings support an overarching model where taurine-driven sulfide production enhances colonization resistance, specifically by inhibiting the ability of pathogens to aerobically respire and thus access strictly respiratory substrates (Figure 4G).

Sulfide sequestration unleashes pathogen respiration.

Because the microbiota produces a baseline level of sulfide (Figure 3I), we next wondered whether homeostatic sulfide itself shapes gut microbial ecology and colonization resistance. To test this, we placed mice onto drinking water containing the sulfide sequestrant bismuth subsalicylate, a compound commonly consumed in over-the-counter remedies for upset stomach (Bierer, 1990) (Figure 5A). Bismuth sequesters sulfide by forming an insoluble precipitate (Bierer, 1990), and indeed, we observed that the gut microbiota releases significantly less sulfide after bismuth treatment (Figure 5B).

This observation led us to test the following model: in the absence of bismuth, homeostatic sulfide restricts respirers from accessing oxygen, whereas in the presence of bismuth, sulfide sequestration increases oxygen availability, promoting invasion by respirers such as Kpn (Figure 5C). Furthermore, since respirers are endogenous to the microbiota, we predicted that sulfide sequestration would promote their expansion as well. Confirming our prediction, bismuth rapidly altered the microbiota by inducing a pronounced expansion of a Bacilli taxon (Figures 5D–E and Table S1). Isolation of this taxon revealed it to be Enterococcus faecalis, a frequent cause of hospital-associated infections (CDC, 2019) (Figure 5F). Unlike many other Enterococcus species, E. faecalis can gain energy from aerobic respiration (Ramsey et al., 2014). In line with this, whole-genome sequencing confirmed the presence of a cytochrome oxidase (bd-I) in bismuth-induced E. faecalis (Table S5). Besides E. faecalis, the potentially pathogenic gammaproteobacterium Escherichia coli also expanded after sulfide sequestration (Figures 5E–F). Because E. coli is capable of respiring with oxygen, its co-expansion with E. faecalis supports our model that homeostatic sulfide shapes gut ecology by subduing low-abundance, potentially pathogenic respirers.

According to our proposed model, sulfide should restrict not only endogenous but also invading oxygen respirers (Figure 5C). To test this, we assessed the impact of sulfide sequestration on colonization resistance. As predicted by our model, we found that bismuth rendered mice highly susceptible to colonization by Kpn, allowing for a nearly 5 log increase in Kpn fecal burden (Figure 5G). Thus, our results support the idea that, by sequestering sulfide, bismuth may promote aerobic respiration by potentially pathogenic, resident members of the microbiota and invading pathogens. A confounding factor to consider when interpreting these studies is that, in addition to sequestering sulfide, bismuth has antimicrobial properties (Bierer, 1990). Nevertheless, these results complement our overarching model that, by increasing the microbiota’s production of sulfide, taurine restricts pathogen respiration and thus enhances colonization resistance (Figure 5H).

Discussion

We are entering a post-antibiotics “dark age” where once curable maladies could become life-threatening due to the spread of antibiotic resistance (United Nations, 2019). Moving forward, one possible path is to leverage the microbiota, as exemplified by fecal microbiota transplantation (Pamer, 2014). However, an obstacle to microbiota-based therapies is the microbiota’s vast inter-individual heterogeneity (Suez and Elinav, 2017). As the majority of beneficial host-microbe interactions are mediated by widely conserved metabolites, therapeutically co-opting metabolites, rather than their microbial targets/producers, may result in wider success (Suez and Elinav, 2017).

In the present work, we discovered that transient infection establishes long-term “metaorganism memory.” Unlike traditional immunological memory, metaorganism memory hinges upon interdependent host and microbiota functions. Following infection, the host amplifies its production/transport of bile acids, and in turn, the microbiota converts a bile acid-derived metabolite, taurine, to the antimicrobial sulfide. Thus, only together do these functions bolster metaorganism resistance to subsequent infection. Because the microbiota function, the metabolism of taurine, is widely distributed (Cook and Denger, 2006), it overcomes the obstacle of inter-individual heterogeneity in microbiota composition. Furthermore, the target of sulfide, aerobic respiration, is a widely (though not universally (Neumann-Schaal et al., 2019)) distributed virulence factor.

While our work focused on how infection-elicited taurine functionally remodels the microbiota, our findings do not fully exclude a role for the immune system. Indeed, the cellular transport of taurine is critical for the establishment of immunological memory (Kaesler et al., 2012), and besides taurine, our data suggest gut infection also elicits the innate immune effector xanthine oxidase (Martin et al., 2004) and the microbial metabolite menaquinone, a precursor of which induces antimicrobial peptides (Fukumoto et al., 2014). Thus, the immune system may work in concert with the microbiota to promote colonization resistance following infection.

Our work suggests that bile acid-derived taurine is beneficial for colonization resistance, but if truly the case, it raises the question of why bile acids are not always maintained at high levels. The most likely answer is that excessive bile acids are associated with various inflammatory disorders and cancers (Jia et al., 2018). Despite this, bile acids are upregulated by enteric infections, most likely because they are antimicrobial and aid pathogen clearance (Uribe et al., 2016). An outstanding question is what infection-induced signaling events lead to host upregulation of bile acids. While this question is made particularly challenging by the numerous receptors and pathways that control bile acid metabolism (Jia et al., 2018), potential candidates are the farnesoid X receptor and the hormone fibroblast growth factor 15 (Romain et al., 2013; Uribe et al., 2016).

Together, our work highlights the key contribution of taurine and sulfide to the trainable resistance of the gut microbiota to infection. By probing the consequence of pathogenic exposures on colonization resistance, we discovered that host metaorganisms adapt to infections by nourishing their gut microbiota with taurine, a strategy that serves to boost the microbiota’s production of sulfide and ensure its heightened resistance against future pathogen invasions.

STAR Methods

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Yasmine Belkaid (ybelkaid@niaid.nih.gov).

Materials Availability

The C. rodentium cydAB^- mutant and the E. faecalis strains isolated in this study are available upon request.

Data and Code Availability

The 16S, shotgun metagenomics, RNA-seq, and Tn-seq datasets generated during this study are available at the Sequence Read Archive (PRJNA666931). The 16S dataset for mice previously infected with the Y. pseudotuberculosis WT (PRJNA276477) and the 16S and metagenomics datasets for generation F2 wildR mice (PRJNA390686) are also available at the Sequence Read Archive. An unpublished in-house package, JAMS, used to analyze metagenomics and whole-genome sequencing data is available at https://github.com/johnmcculloch/JAMS_BW.

Experimental Model and Subject Details

Mice

Conventional SPF C57BL/6 and congenic CD45.1 mice (used interchangeably) were acquired from Taconic Biosciences (only barriers positive for segmented filamentous bacteria) and maintained at NIAID or NIDDK animal facilities. WildR SPF C57BL/6 mice were created, bred, and maintained at an NIDDK animal facility (Rosshart et al., 2017), then transferred entirely to Taconic Biosciences. The F10 wildR mice in this study were analyzed after wildR mice had been re-imported from Taconic Biosciences to NIDDK. GF C57BL/6 mice were bred and maintained at the NIAID Microbiome Program Gnotobiotic Animal Facility. All facilities were accredited by the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International, and all animal procedures were performed in accordance with the Guide for the Care and Use of Laboratory Animals under animal study proposals approved by the NIAID or NIDDK Animal Care and Use Committees. Sex- and age-matched mice between 6–20 weeks of age were used for all experiments. Female and male mice were used for experiments with wildR and GF mice, while only female mice were used for experiments involving only conventional mice.

Bacteria

The Y. pseudotuberculosis, Kpn, C. rodentium, and B. wadsworthia WT strains were 32777, MKP103 (a KPNIH1 derivative with the KPC-3 carbapenemase-encoding gene deleted (Ramage et al., 2017)), ICC169 (an ICC168 derivative resistant to nalidixic acid), and ATCC 49260. The Kpn cyxB^-, pduQ^-, and acrE^- Tn mutants were purified from the MKP103 Tn mutant library (Ramage et al., 2017) and confirmed via PCR. Y. pseudotuberculosis ∆yopM (Han et al., 2017) was routinely revived on LB agar and cultured in 2x YT broth at 25°C with shaking at 200–250 rpm. Kpn and C. rodentium WT and mutant strains were routinely revived on LB agar and cultured in LB broth (10 g/l tryptone, 5 g/l yeast extract, 8 g/l NaCl) at 37°C with shaking at 200 rpm. B. wadsworthia was routinely revived on Wilkins-Chalgren agar + 50 mM taurine and cultured in Wilkins-Chalgren broth + 50 mM taurine at 37°C without shaking in an anaerobic chamber (Whitley Workstation A85).

Method Details

Intestinal infection of mice with gut pathogens

Overnight (12–14 hour) Y. pseudotuberculosis ∆yopM cultures were resuspended in PBS, adjusted to an absorbance at 600 nm (A₆₀₀) of 0.1 in PBS, and gavaged into mice fasted overnight for 13–15 hours (~2 x 10⁷ CFU in 200 µl/mouse). Overnight Kpn cultures were resuspended in LB, adjusted to an A₆₀₀ of 1 in LB, and gavaged into overnight-fasted mice (~2 x 10⁸ CFU in 200 µl/mouse). Overnight C. rodentium cultures were resuspended in PBS, adjusted to an A₆₀₀ of 5 or 25, and gavaged into non-fasted mice (~1 or 5 x 10⁹ CFU in 200 µl/mouse). For C. rodentium experiments, mouse cages were not changed from 3–5 days before infection until 5 days after infection. For in vivo competition experiments, the same procedures as described above were followed, except that the WT and mutant were mixed together at a 1:1 ratio prior to gavage. When pelleting cultures, all bacteria were centrifuged at 3–4,000 x g for 2–5 minutes at room temperature. All absorbances were measured with a spectrophotometer (Biochrom).

Treatment of mice with taurine, bismuth, and streptomycin

To treat mice with taurine, mice were placed onto non-acidified tap water + 200 mM taurine for 2–3 weeks prior to infection, during which fresh water was provided every 3–5 days. At the same as when mice were first placed onto taurine water, mice were exchanged between paired control and treatment cages, the intent being to reduce microbiota variability between cages (generally, 2 mice in a control cage of 5 mice were exchanged with 2 mice in a treatment cage of 5 mice). Starting the evening (12–14 hours) prior to Kpn infection, mice were placed onto non-acidified tap water lacking taurine, to avoid stimulation of Kpn growth by taurine. To treat mice with bismuth, starting the evening prior to infection, mice were placed onto non-acidified tap water + 20 mM bismuth subsalicylate and maintained on bismuth water throughout the course of infection. To treat mice with streptomycin, mice were gavaged with 20 mg streptomycin (200 µl of a freshly prepared 100 mg/ml solution) 1 day prior to infection.

Collection and processing of gut samples to determine CFU

Fecal pellets (colon lumen samples) were collected directly from mice or by placing mice into paper cups. Cecum lumen samples were collected by squeezing out the contents of extracted ceca using forceps. Mucosal samples were collected by cutting extracted tissues (colons, ceca, or ilea) open longitudinally, thoroughly rinsing tissues in Petri dishes with PBS, scraping the interior surface of rinsed tissues using closed forceps, and collecting the scrapings (mucosa). The PBS used to rinse ileal tissues was transferred to 50-ml conical tubes and centrifuged at 4,000 rpm (2,880 x g) for 5 minutes at room temperature to collect the pellets (ileum lumen samples). All gut samples were transferred or collected directly into pre-weighed 1.5-ml tubes, weighed after collection, homogenized in 500 µl PBS using a pipette tip, and vortexed for 5 seconds. After allowing large particles to settle, top fractions were collected to determine CFU.

Determination of CFU

CFU were determined for both gut samples and in vitro growth assays as described using a multichannel pipette (Hung et al., 2009). Briefly, samples were serially diluted in 96-well plates with PBS, and 10 µl of each dilution was spotted 5 times onto LB agar (1 plate per sample, 50 µl per dilution). For in vitro growth assays, LB without antibiotics was used to determine total CFU; LB + 175 µg/ml chloramphenicol, Kpn Tn mutant CFU; LB + 100 µg/ml kanamycin, C. rodentium cydAB^- CFU. For gut samples, LB + 50 µg/ml ampicillin + 100 µg/ml neomycin + 100 µg/ml vancomycin was used to determine total Kpn CFU; the same media + 175 µg/ml chloramphenicol, Kpn Tn mutant CFU; LB + 50 µg/ml nalidixic acid, total C. rodentium CFU (only large colonies); the same media + 100 µg/ml kanamycin, C. rodentium cydAB^- CFU. In competition assays, samples with fewer than 10 total CFU were excluded from final results.

Preparation of NCE medium

No carbon E (NCE) minimal medium (Faber et al., 2017) for in vitro growth assays (below) was prepared from stock solutions in diH₂O: 1) 10x salts (at 1x, 29 mM KH₂PO₄, 28.3 mM K₂HPO₄, 16.7 mM NaNH₄HPO₄), 2) 1000x MgSO₄ (at 1x, 1 mM), and 3) 1000x trace metals (at 1x, 0.3 µM CaCl₂, 3 µM CoCl₂, 0.1 µM CuSO₄, 2 µM MnSO₄, 0.2 µM Na₂MoO₄, 0.2 µM Na₂SeO₃, 0.1 µM NiSO₄, 0.1 µM ZnSO₄).

Kpn and C. rodentium in vitro growth assays

All overnight cultures for growth assays were grown without shaking in an anaerobic chamber, diluted 1:250 into test media that was pre-reduced overnight in the anaerobic chamber, and further grown for 23–25 hours in either the anaerobic chamber, a hypoxic chamber (Baker Ruskinn PhO₂x Box) with 0.8% oxygen, an anaerobic jar with a microaerophilic sachet (6–12% oxygen), or in air (21% oxygen) prior to determining CFU. To assess electron acceptor usage, overnight cultures in LB were diluted into NCE + 5 mM 1,2-propanediol + 50 mM electron acceptor or vehicle (water). For in vitro competition experiments, WT and mutant overnight cultures in NCE + 0.2% yeast extract + 50 mM carbon source (1,2-propanediol or formate) were mixed together at a 1:1 ratio and diluted into NCE + 5 mM carbon source. For inhibition assays, overnight cultures in NCE + 0.2% yeast extract + 50 mM carbon source (1,2-propanediol or formate) were diluted into NCE + 5 mM carbon source + 5 mM taurocholate, 5 mM taurine, sodium hydrosulfide (0.125–1 mM for Kpn or 1–5 mM for C. rodentium), or vehicle (water). Sodium hydrosulfide was freshly prepared for all experiments.

B. wadsworthia in vitro growth assays

To assess sulfur source usage, B. wadsworthia was grown overnight in pre-reduced Wilkins-Chalgren broth + 50 mM taurine without shaking in an anaerobic chamber, diluted 1:500 into pre-reduced Wilkins-Chalgren broth + 50 mM sulfur source (taurine or magnesium sulfate) or vehicle (water), and further grown overnight in the anaerobic chamber prior to measuring A₆₀₀. To assess growth stimulation by DHNA, overnight cultures in NCE (with 1 mM MgCl₂ rather than MgSO₄) + 50 mM taurine + 50 mM formate were diluted 1:500 into the same medium + 276 µg/ml ferrous ammonium sulfate + 200 µg/ml DHNA or vehicle (ethanol) and further grown in the anaerobic chamber.

Construction of C. rodentium cydAB mutant

To delete cydAB from C. rodentium ICC169, pCAL60, a plasmid previously constructed to delete cydAB from C. rodentium DBS100, was used (Lopez et al., 2016). This plasmid contains 1) a kanamycin resistance cassette flanked by the regions up and downstream of cydAB, 2) a chloramphenicol resistance cassette on the backbone, and 3) to allow for blue/white screening, the Salmonella Typhimurium phoN gene on the backbone. To conjugate pCAL60 from E. coli S17–1 λ-pir (donor) into C. rodentium ICC169 (recipient), overnight cultures of the recipient (diluted 1:10 into fresh LB) and donor (not diluted) were resuspended in fresh LB, mixed 1:1, and spotted onto polycarbonate membranes (100 µl/membrane) placed onto LB agar. After 4–6 hours, conjugations were resuspended in 1 ml LB and struck onto LB agar with 50 µg/ml nalidixic acid (to select against E. coli S17–1 λ-pir), 100 µg/ml kanamycin (to select for single crossover events), and 40 µg/ml X-phos (to screen for double crossover events). White colonies were screened for sensitivity to 15 µg/ml chloramphenicol, and sensitive colonies were confirmed for cydAB deletion via PCR.

Lung infection of mice with Kpn

For lung infections, Kpn ATCC 43816 was used since this strain is highly virulent in lung infection models (Broberg et al., 2014). Kpn was inoculated directly from a freezer stock into 50 ml Nutrient broth and cultured for 7 hours at 37°C with shaking at 200 rpm. Mice anesthetized with isoflurane were intratracheally inoculated with 3 x 10³ CFU in 80 µl. After 36 hours, lungs were harvested into pre-weighed tubes with PBS (1 ml/lung) and 1.5 mm zirconium beads, weighed after collection, homogenized at 5,000 rpm for 2 20-second cycles using a Precellys 24 (Bertin Instruments), and plated onto MacConkey agar to determine CFU.

Preservation of microbiota

Fecal samples from post-∆yopM mice were frozen at −80°C prior to being processed, while samples from taurine-treated mice were processed shortly after being collected. Samples were transferred into an anaerobic chamber, homogenized in pre-reduced PBS (~20 µl/mg feces) using a pipette tip, and vortexed for 5 seconds. After allowing large particles to settle, top fractions were diluted 1:5 into pre-reduced PBS, mixed 1:1 with pre-reduced 20% glycerol, aliquoted, and stored at −80°C. Prior to gavage into GF mice, aliquots were thawed in the anaerobic chamber and mixed 1:1 with pre-reduced PBS.

Conventionalization of GF mice

GF mice were transferred using aseptic technique into sterile cages with acidified drinking water, immediately gavaged with 200 µl microbiota (revived in 5% glycerol, as described in the previous section), housed under SPF conditions, and gavaged twice more with microbiota on consecutive days. To pre-engraft GF mice with B. wadsworthia, mice were 1) aseptically transferred to SPF cages, 2) immediately gavaged with 200 µl of concentrated B. wadsworthia (an overnight 25-ml culture resuspended in 1–2 ml Wilkins-Chalgren broth), 3) gavaged the following day again with B. wadsworthia, 4) hours after the second B. wadsworthia gavage, gavaged with 200 µl SPF microbiota (freshly prepared by homogenizing fecal pellets in pre-reduced PBS in an anaerobic chamber), 5) gavaged the following day again with freshly prepared SPF microbiota. After conventionalization, mice were rested for 1.5–2 weeks prior to infection, during which cages were only changed if necessary (e.g. if water bottles leaked). Starting the evening before infection, mice were placed onto non-acidified tap water.

16S rRNA gene profiling

DNA was extracted from fecal pellets using phenol:chloroform (Goodman et al., 2011) or the MagAttract PowerMicrobiome DNA/RNA Kit. 16S V4 rDNA libraries were prepared using the Phusion High-Fidelity PCR Master Mix with primers 515F and 806R (Caporaso et al., 2012), and sequenced on an Illumina MiSeq instrument by the NIAID Microbiome Program core facility, with the exception of fecal pellets from F10 wildR mice, which were processed and sequenced by the NCI Cancer and Inflammation Program Microbiome and Genetics Core.

16S data were analyzed using QIIME 2 (Bolyen et al., 2019). Briefly, reads were denoised using the q2-dada2 plugin (exact parameters were forward reads truncated to 180 bases with 20 bases trimmed from the 5’ end, reverse reads truncated to 150 bases with 20 bases trimmed from the 5’ end, and bases truncated with quality score below 10). Using the q2-feature-table plugin, samples within an experiment were rarefied to the depth of the sample with the fewest reads (>19,000 reads for all experiments), and sequence variants were filtered if within an experiment they did not occur in at least 3 samples. Rooted phylogenetic trees were generated using the align-to-tree-mafft-fasttree method in the q2-phylogeny plugin. Alpha and beta diversity analyses were performed using the core-metrics-phylogenetic method in the q2-diveristy plugin, with a sampling depth of 5,000 reads. Taxonomy was assigned to sequence variants using the q2-feature-classifier plugin and a naïve Bayes classifier pre-trained on the SILVA 132 515F/806R 99% OTUs (Yilmaz et al., 2014) provided as a QIIME 2 data resource. A pseudocount was applied prior to calculating differences in the relative abundance of taxonomic groups. In parallel, the same workflow was used to re-analyze published 16S data for fecal pellets from post-Yptb WT mice (Fonseca et al., 2015) and published 16S data for ileocecal material from F2 labR and F2 wildR mice (Rosshart et al., 2017).

Shotgun metagenomics

Metagenomics libraries were generated using the Illumina DNA Prep kit, with 500 ng purified DNA as template, and sequenced on an Illumina NextSeq instrument (average depth per sample, 3.6 Gb; range, 2.7–4.5 Gb) by the NIAID Microbiome Program core facility.

Data were analyzed using an in-house package (https://github.com/johnmcculloch/JAMS_BW) as described (Rosshart et al., 2019). In parallel, the same package was used to re-analyze published metagenomics data for ileocecal material from F2 labR and F2 wildR mice (Rosshart et al., 2017). Briefly, sequencing reads were processed using Trimmomatic (Bolger et al., 2014); host-derived reads were filtered using Bowtie 2 (Langmead and Salzberg, 2012); microbiota-derived reads were assembled into contigs using Megahit (Li et al., 2015); genes within contigs were predicted and annotated with EC numbers using Prokka (Seemann, 2014); the relative abundance of EC numbers was determined by aligning microbiota-derived reads back to contigs using Bowtie 2; contigs and unassembled reads were taxonomically classified by k-mer analysis using Kraken 2 (Wood et al., 2019) and, as reference, a custom-built database of all genome sequences in Genbank. The relative contribution of taxa to individual EC numbers was determined by tallying the read coverage of each single gene with a particular EC number and obtaining the taxonomic information for that single gene from the Kraken 2 identification of the exact contig in which the gene was predicted ab initio.

Of note, for many of the functions in Figure 2, more than one species of Desulfovibrio was often identified as contributing to each function. However, based on 16S analysis, all post-∆yopM samples were determined to contain only a single species of Desulfovibrio. Thus, the metagenomic identification of multiple species of Desulfovibrio was deemed an artifact due to the relatively low depth of Desulfovibrio-derived sequencing reads. Because of this, the multiple species of Desulfovibrio identified as contributing to each function in Figure 2 were aggregated into a single unidentified Desulfovibrio species.

Profiling of metabolites by mass spectrometry

Cecal contents collected onto dry ice were submitted to the West Coast Metabolomics Center at the University of California, Davis. The relative abundance of 449 metabolites (186 defined, 263 undefined) was quantified using gas chromatography time-of-flight mass spectrometry (GC-TOF-MS). The absolute abundance of 23 bile acids was quantified using liquid chromatography-mass spectrometry (LC-MS).

Measurement of fecal taurine and total bile acids

Fecal taurine and total bile acids were measured using colorimetric assay kits (Cell Biolabs) according to the manufacturer’s instructions, except that 1) all reagent volumes were halved, and 2a) fecal samples were collected into pre-weighed tubes with PBS (10 µl/mg feces) and 0.1 mm zirconia/silica beads, 2b) homogenized at 5,000 rpm for 3 20-second cycles using a Precellys 24 (Bertin Instruments), and 2c) centrifuged at 10,000 x g for 2 minutes to collect supernatants. Fecal pellets for measuring taurine and total bile acids were collected in the morning.

Measurement of sulfide production by ex vivo microbiota cultures

Sulfide production by ex vivo microbiota cultures was measured using the OxiSelect Free Hydrogen Sulfide Assay Kit. Briefly, freshly collected fecal samples were transferred into an anaerobic chamber, homogenized in pre-reduced PBS (~20 µl/mg feces) using a pipette tip, and vortexed for 5 seconds. To establish ex vivo cultures, the top fraction of fecal homogenates was diluted 1:1 with pre-reduced PBS, and 225 µl was transferred into non-adjacent wells of a 96-well plate containing 25 µl vehicle (0.5 M NaCl) or 25 µl 0.5 M taurine and 0.5 M sodium formate. The areas of the 96-well plate lid directly above sample wells were each coated with 20–25 µl of the Lid Coating Mix, according to the manufacturer’s instructions, and after placing the lid over the cultures, it was removed at 15–20 minute intervals from the anaerobic chamber to measure absorbance at 405 nm on a plate reader (BioTek Synergy HTX). Standard curves were generated using freshly prepared sodium hydrosulfide and measured in the same plates as ex vivo cultures.

Tn-seq

The Kpn MKP103 Tn mutant library (Ramage et al., 2017) arrayed into 96-well plates was revived overnight in LB at 37°C without shaking, combined into a single pool, mixed 1:1 with 50% glycerol, and stored as aliquots at −80°C. Thawed aliquots were grown overnight (12–14 hours) and prepared for gavage as described above for routine Kpn infections. DNA extracted from fecal samples collected at 8 hours post-infection was used as template in the 2-PCR method for generating Tn-seq sequencing libraries, according to a published protocol (Stacy et al., 2016), except with primers designed for the Kpn Tn mutant library. Briefly, DNA was C-tailed using terminal deoxynucleotidyl transferase, PCR-amplified with a biotinylated primer targeting the Tn sequence and a primer targeting the C-tail (PCR-1), purified with streptavidin-coated magnetic beads, further PCR-amplified to append forward and reverse Illumina adapter sequences (PCR-2), and sequenced on an Illumina NextSeq instrument by the NCI Cancer and Inflammation Program Microbiome and Genetics Core.

Tn-seq data were analyzed largely as described (Stacy et al., 2016). 1) Reads were processed as follows using Cutadapt (Martin, 2011): a) the exact Tn sequence ATGTGTATAAGAGACAG was trimmed from 5’ ends (this quality-control sequence should be amplified by the second Tn-specific forward primer immediately prior to Kpn DNA), b) C-tails and low-quality bases (Q < 20) were trimmed from 3’ ends, and c) reads were trimmed to 20–25 bases in length. 2) Processed reads were mapped to the KPNIH1 genome using Bowtie 2 (default end-to-end alignment) (Langmead and Salzberg, 2012). 3) In Unix, a) low-quality alignments (Q < 39) were filtered, b) the number of unique Tn per sample was tallied (in doing so, the 9 base duplication of the Tn5 target sequence (Goryshin et al., 1998) was trimmed from alignments mapping to the negative strand), c) to achieve a uniform density of 5,000–6,000 Tn per sample, samples with < 5,000 Tn were concatenated, and samples with > 6,000 Tn were rarefied (alignments were discarded), and d) the relative abundance per unique Tn was tallied for each sample. 4) The relative Tn abundance per gene was tallied for each sample using Rsubread (Liao et al., 2019). 5) Samples were normalized for differences in sequencing depth using DESeq2 (Love et al., 2014), and differential Tn abundance per gene was assessed using the DESeq2 Wald test. The KPNIH1 genome was annotated with K numbers using the KEGG Automatic Annotation Server (Moriya et al., 2007).

Processing of large intestines and livers for flow cytometry

To isolate cells from large intestinal lamina propria, large intestines (LI, colons with ceca) were extracted with minimal attached fat into complete medium (RPMI 1640 supplemented with 1 mM pyruvate, 0.1 mM nonessential amino acids, 2 mM glutamine, 20 mM HEPES, 100 U/ml penicillin, 0.1 mg/ml streptomycin, and 50 µM β-mercaptoethanol) + 3% fetal bovine serum (FBS) on ice. LI were cut open longitudinally, thoroughly rinsed in PBS to remove feces, cut into 1–2 cm pieces, and incubated in 10 ml complete medium + 3% FBS + 5 mM EDTA + 1 mM DTT for 20 minutes at 37°C with stirring. To remove epithelial cells, LI were next vigorously shaken in 10–15 ml complete medium + 2 mM EDTA in 50-ml conical tubes for 15–20 seconds and strained. After performing this procedure 3 times (each time with fresh media), LI were minced into fine pieces with scissors and digested in 10 ml complete medium + 0.1 mg/ml Liberase TL + 0.5 mg/ml DNase I for 25–30 minutes at 37°C with stirring. Digestions were stopped with 10 ml ice-cold complete medium + 3% FBS, passed through 70-µm filters, centrifuged at 1500 rpm for 5 minutes, resuspended in 4 ml 37.5% Percoll, centrifuged at 1800 rpm for 5 minutes, and resuspended in 300 µl complete medium + 3% FBS.

To isolate cells from livers, livers were disassociated through 70-µm filters with 4–6 ml complete medium, centrifuged at 1500 rpm for 5 minutes, resuspended in 10 ml 37.5% Percoll, centrifuged at 1800 rpm for 5 minutes, and resuspended in 300 µl complete medium + 3% FBS.

Flow cytometry

To assess myeloid cells, 100 µl of 300 µl single-cell suspensions were stained on ice for 15 minutes in PBS with fluorochrome-conjugated antibodies (against CD45 diluted 1:300; TCRβ 1:400; TCRγδ, 1:400; B220, 1:400; NK1.1, 1:400; CD11b, 1:200; Ly-6G, 1:500; Siglec-F, 1:300; CD11c, 1:200; CD64, 1:200; MHC-II, 1:1000; CD103, 1:200; CD24, 1:500; Ly-6C, 1:400; CCR2, 1:100), fixable live/dead stain, Fc Block, and counting beads (to quantify cell numbers) and fixed on ice for at least 1 hour using the Fixation/Permeabilization Solution Kit.

To assess T cells, 200 µl of 300 µl single-cell suspensions were restimulated in vitro in complete medium + 10% FBS + 50 ng/ml PMA (phorbol 12-myristate 13-acetate) + 1 mM ionomycin + 1 µl/ml GolgiPlug for 2.5 hours at 37°C, stained on ice for 15 minutes in PBS with fluorochrome-conjugated antibodies against surface targets (CD45 diluted 1:500; CD90.2, 1:400; TCRβ, 1:200; CD4, 1:200; CD8β, 1:400), fixable live/dead stain, Fc Block, and counting beads (to quantify cell numbers), fixed on ice for at least 1 hour using the Foxp3/Transcription Factor Staining Buffer Set, and stained at room temperature for at least 30 minutes with antibodies against intracellular targets (Foxp3 diluted 1:100; IFN-γ, 1:100; IL-17A, 1:100; IL-22, 1:100; TCRβ, 1:200; CD4, 1:200). Cells were acquired on a BD LSRFortessa running FACSDiva software, and data were analyzed using FlowJo.

Processing of large intestines for epithelial cell sorting

Extracted large intestines (LI, colons with ceca) were cut open longitudinally, thoroughly rinsed in gut wash buffer (HBSS supplemented with 1 mM pyruvate, 1 mM nonessential amino acids, 2 mM glutamine, 20 mM HEPES, 100 U/ml penicillin, and 100 mg/ml streptomycin) + 2% FBS to remove feces, and cut into 1–2 cm pieces. Subsequently, LI were incubated in 10 ml gut wash buffer + 1 mM DTT for 7–8 minutes in a 37°C water bath, vigorously shaken, and strained to discard DTT-containing fractions. After performing this procedure 2 times, LI were incubated in 10 ml gut wash buffer + 5 mM EDTA for 7–8 minutes in a 37°C water bath, vigorously shaken, and strained to collect EDTA-containing fractions. After performing this procedure 2 times, the pooled EDTA-containing fractions were passed through 100-µm filters, centrifuged at 1,000 rpm for 5 minutes, washed with DMEM, and digested in 10 ml DMEM + 0.1 U/ml Dispase II + 0.1 mg/ml DNase I for 8 minutes in a 37°C water bath. Digestions were stopped with DMEM + 2.5% FBS, centrifuged at 1,000 rpm for 5 minutes, and washed with gut wash buffer. Cells were stained on ice for 30 minutes in gut wash buffer + 5% FBS with fluorochrome-conjugated antibodies (against EpCAM diluted 1:500; CD45, 1:1000; CD31, 1:200) and DAPI, and epithelial cells were sorted directly into Buffer RLT using a Sony MA900 (sorting chip size, 100 µm).

Liver and gut epithelial cell RNA-seq

Per liver, small pieces (0.5–1 cm in width) from the edge of each lobe were collected together into 1 ml RNAlater, incubated overnight at room temperature, and stored at −20°C. RNA was extracted using the RNeasy Fibrous Tissue Mini Kit according to the manufacturer’s instructions, with the exceptions that 1a) samples were transferred to tubes with 1 ml Buffer RLT and 1.5 mm zirconium beads and 1b) homogenized at 5,000 rpm for 3 20-second cycles using a Precellys 24 (Bertin Instruments), and 2) only 50 µl of the homogenate was mixed with 250 µl fresh Buffer RLT + β-mercaptoethanol (143 mM) and used to extract RNA. For sorted gut epithelial cells, RNA was extracted using the RNeasy Plus Micro Kit. The integrity of extracted RNA was assessed using an Agilent TapeStation. RNA-seq libraries were prepared using the Universal Plus mRNA-Seq Library Preparation Kit (for liver RNA) or the Ovation SoLo RNA-Seq Library Preparation Kit (for epithelial cell RNA) and sequenced on an Illumina NextSeq instrument by the NIH Center for Human Immunology core facility.

RNA-seq data were mapped to the mm10 mouse genome using STAR (default parameters) (Dobin et al., 2013). Samples with clonality greater than 2.5 as assessed by HOMER (Heinz et al., 2010) were excluded. Gene expression was assessed using HOMER’s analyzeRepeats.pl with the options rna, mm10, -count exons, -condenseGenes, and differential gene expression was assessed using DESeq2 (Love et al., 2014). The enrichment of GO terms was assessed using Metascape (Zhou et al., 2019). Ribosomal protein and antimicrobial peptide annotations were obtained from the Gene Ontology Browser of the Mouse Genome Informatics website (Bult et al., 2019).

Measurement of gallbladder weight

Gallbladders were carefully extracted into PBS, and any attached fat was removed in a Petri dish using forceps and a scalpel. Prior to being weighed, gallbladders were blotted onto absorbent wipes to remove PBS. Gallbladders that leaked either during extraction or fat removal were excluded from final results.

Liver and colon histology

Liver samples (part of the largest lobe) and colon samples (distal 0.5–1 cm without feces) were fixed in 1 ml 10% neutral buffered formalin and submitted to Histoserv, Inc. for paraffin sectioning and H&E staining. Sections were imaged on a Leica SCN400 slide scanner and analyzed in a blinded manner, liver sections by a trained pathologist and colon sections using Fiji (Schindelin et al., 2012). The crypt length per mouse was assessed by averaging the length of 20 crypts per colon section.

Measurement of intestinal transit time

As previously described (Dey et al., 2015), non-fasted mice were gavaged with 200 µl carmine solution (6% (w/v) carmine in a 0.5% (w/v) methyl cellulose solution), and feces was collected every 30 minutes to assess for the presence of carmine, by streaking the collected feces onto a white paper towel. The time from gavage to the initial appearance of carmine in feces was recorded as the intestinal transit time.

E. faecalis isolation, identification, and whole-genome sequencing

The feces of 2 non-co-housed bismuth-treated mice was homogenized in PBS and struck onto bile esculin azide agar. Four colonies (2 colonies per mouse) were re-struck onto the same media and submitted to the NIAID Research Technologies Branch for species-level identification using a Bruker MALDI Biotyper. All 4 colonies were identified as E. faecalis. Whole-genome sequencing libraries were prepared using the Illumina DNA Prep kit and sequenced on an Illumina MiSeq instrument. Data were analyzed using the same in-house package as for metagenomics data, with SPAdes as the genome assembler (Bankevich et al., 2012).

Quantification and Statistical Analysis

The statistical details of experiments can be found in the figure legends. Reported n are the total samples (mice or cultures) per group; fold changes are the medians of pooled experiments; n.s., not significant; †, P < 0.1; *, P < 0.05; **, P < 0.01; ***, P < 0.001. Unless otherwise stated, P values were not corrected for multiple testing. Mann-Whitney tests (all unpaired and two-tailed) were performed in Prism or R; Grubbs’ and ROUT outlier tests were performed in Prism; Benjamini-Hochberg corrections, hypergeometric tests, Kruskal-Wallis tests and PERMANOVA (QIIME 2 package), and Wald tests (DESeq2 package) were performed in R; Student’s t tests (all unpaired and two-tailed) were performed in Prism.

Supplementary Material

Figure S1. Infection-trained microbiota enhance colonization resistance, Related to Figure 1

(A) Kpn CFU in the feces of control and post-∆yopM mice at 1, 2, 4, and >15 weeks post-∆yopM infection. The 1 and 2 week time points represent single experiments (n = 5–8). The 4 and >15 week time points are reproduced from Figure 1E. Lines connect the medians of paired cages.

(B) Time for a gavaged dye to transit through the gut of control and post-∆yopM mice at >15 weeks post-∆yopM infection (1 experiment with 2 paired cages, n = 10).

(C) Identity of isolates cultured (on MacConkey agar at 25°C) from the microbiota of mice at >15 weeks post-∆yopM infection. The dotted line indicates the detection limit (1 in 14 or 7% of isolates).

(D) Number of unique 16S sequences identified in the gut microbiota of post-∆yopM, F2 wildR, and respective control mice as a function of read depth. Lines represent mean ± SEM (n = 11–20, with 10 iterations averaged per n). (E-F) PCoA plots of unweighted UniFrac distances between the 16S profiles of (E) control and F10 wildR mice (n = 7–12) and (F) ex-GF mice conventionalized with control or post-∆yopM microbiota (n = 8–9). Percentages represent the variance explained by each PC.

(G) Relative abundance of phyla in the gut microbiota of control and post-∆yopM mice at >15 weeks post-∆yopM infection. Data are representative of 4 experiments.

(H) Fold change in the median abundance of phyla in the gut microbiota of post-Yptb (circles, Yptb ∆yopM; triangles, Yptb WT; n = 23–27), F2 wildR (n = 11–12), and F10 wildR (n = 7–12) vs. respective control mice. Post-Yptb bars show mean ± SEM of 5 paired cages. Only fold changes > 2 were considered significant. n.d., not detected.

(I) Fold change in the median abundance of Proteobacteria classes (indicated by Greek letters) in the gut microbiota of F10 wildR vs. control mice (n = 7–12).

(J) Fold change in the median abundance of Proteobacteria species (classes indicated by Greek letters) in the gut microbiota of ex-GF mice conventionalized with post-∆yopM vs. control microbiota. Bars show mean ± SEM of 2 paired cages (n = 8–9). Colors indicate engraftment (gray, negative; red, positive in post-∆yopM vs. control). Only Proteobacteria species with P < 0.05 are displayed.

The control for post-Yptb (∆yopM and WT) and F10 wildR mice was naïve SPF mice. The control for F2 wildR mice was F2 labR mice. n.s., not significant; †, P < 0.1; *, P < 0.05; **, P < 0.01; ***, P < 0.001 by Mann-Whitney test (A, B, J; in H-I, with Benjamini-Hochberg correction), Kruskal-Wallis test (D), or PERMANOVA (E-F).

NIHMS1654488-supplement-1.pdf^{(220.7KB, pdf)}

Figure S2. Gut immunity of post-∆yopM mice, Related to Figure 1

(A-B) Gating strategy for (A) innate immune and (B) T cells. DC, dendritic cell; moDC, monocyte-derived DC; Tregs, regulatory T cells; IFN, interferon; IL, interleukin.

(C) Numbers of innate immune cells, total T cells, and cytokine-producing T cells (restimulated with PMA/ionomycin) in the large intestinal (cecum and colon) lamina propria of control and post-∆yopM mice at 4–5 weeks post-∆yopM infection. Bars show median ± 95% confidence intervals (n = 5). Data either represent 1 experiment (innate immune cells) or are representative of 2 experiments (T cells).

(D-E) IFNγ-producing CD8⁺ T cells in (D) SPF control and post-∆yopM mice and (E) ex-GF mice conventionalized with control or post-∆yopM microbiota. Left: Numbers in representative contour plots indicate mean frequency ± SEM. Right: Data either (D) are representative of 2 experiments or (E) represent 1 experiment with 2 paired cages (n = 7).

n.s., not significant; †, P < 0.1; **, P < 0.01 by Student’s t test (C-E).

NIHMS1654488-supplement-2.pdf^{(1.2MB, pdf)}

Figure S3. Infection-trained microbiota are enriched for taurine-utilizing taxa, Related to Figure 2

(A) Pathways enriched in the F2 wildR shotgun metagenome. Pie charts represent the relative contribution of taxa to a function. Abbreviations same as in Figure 2D.

(B) Fold change in the median abundance of metabolites enriched in the cecal contents of post-∆yopM vs. control mice at >15 weeks post-∆yopM infection. Bars show mean ± SEM of 3 paired cages (n = 13–14). Metabolites in bold are related to xanthine oxidase.

(C) Abundance of taurine in the cecal contents of F10 wildR mice (1 experiment with 2 paired cages, n = 7, fold decrease = 2).

(D-E) (D) Representative micrographs and (E) histological score of liver sections from control and post-∆yopM mice at 4 weeks post-∆yopM infection. Scale bar, 100 µm.

(F) Top 10 most enriched functions (GO terms) among genes upregulated (log₂ fold change > 0.5, P < 0.1 by Wald test with Benjamini-Hochberg correction) in the livers of post-∆yopM vs. control mice at 4–5 weeks post-∆yopM infection.

(G) Numbers of innate immune cells, total T cells, and cytokine-producing T cells (restimulated with PMA/ionomycin) in the livers of control and post-∆yopM mice at 4–5 weeks post-∆yopM infection. Bars show median ± 95% confidence intervals (n = 5). Data either represent 1 experiment (innate immune cells) or are representative of 2 experiments (T cells). Gating strategies and abbreviations same as in Figure S2.

(H) Weights of gallbladders from control and post-∆yopM mice at >4 weeks post-∆yopM infection (4 pooled experiments, n = 12–16, fold increase = 1.4).

(I) Abundance of total bile acids in the feces of control and post-∆yopM mice at >15 weeks post-∆yopM infection (3 pooled experiments, n = 12–13, fold increase = 1.5).

(J) Abundance of total bile acids in the feces of control and F10 wildR mice (1 experiment with 2 paired cages, n = 7, fold increase = 9.3).

(K) Fold change in the median abundance of individual bile acids enriched in the cecal contents of post-∆yopM vs. control mice at >15 weeks post-∆yopM infection. Bars show mean ± SEM of 3 paired cages (n = 12–14). 1°, primary bile acids; 2°, secondary bile acids; TωMCA; tauro-ω-muricholic acid; GUDCA, glycoursodeoxycholic acid; CA, cholic acid; CDCA, chenodeoxycholic acid; TCDCA, taurochenodeoxycholic acid; βMCA, β-muricholic acid.

(L) B. wadsworthia growth (indicated by black color) in minimal media + taurine ± DHNA + ferrous iron (reacts with sulfide to form a black precipitate).

(M) Relative abundance of B. wadsworthia (Bw) in the gut microbiota of ex-GF mice conventionalized with the microbiota of SPF mice or, prior to conventionalization, pre-engrafted with Bw (2 pooled experiments, n = 10, fold increase = 7.5). The dotted line indicates the detection limit.

The control for post-∆yopM and F10 wildR mice was naïve SPF mice. The control for F2 wildR mice was F2 labR mice. In box and whisker plots, lines connect the medians of paired cages. n.s., not significant; †, P < 0.1; *, P < 0.05; **, P < 0.01 by Mann-Whitney (B, C, H-K, M); hypergeometric (F), or Student’s t test (G). See also Table S4.

NIHMS1654488-supplement-3.pdf^{(911.8KB, pdf)}

Figure S4. Taurine-trained microbiota enhances colonization resistance, Related to Figure 3

(A) Time for a gavaged dye to transit through the gut of vehicle and taurine-treated mice (1 experiment with 2 paired cages, n = 10).

(B) Mice were maintained on taurine after intratracheal inoculation with Kpn.

(C) Kpn CFU in the lungs of vehicle and taurine-treated mice at 36 hours post-intratracheal inoculation with Kpn (2 pooled experiments, n = 9–10).

(D) Numbers of innate immune cells, total T cells, and cytokine-producing T cells (restimulated with PMA/ionomycin) in the large intestinal (cecum and colon) lamina propria of vehicle and taurine-treated mice. Bars show median ± 95% confidence intervals (n = 5, all P > 0.14). Data are representative of 2 experiments. Gating strategies and abbreviations same as in Figure S2.

(E) Pre-sort: Gating strategy to sort epithelial cells from the large intestines of vehicle and taurine-treated mice. Post-sort: Representative contour plots depicting the purity of sorted epithelial cells. Both pre and post-sort are gated on singlets. Numbers represent population frequency.

(F) Differentially expressed genes in large intestinal epithelial cells sorted from taurine vs. vehicle-treated mice (2 pooled experiments, n = 6–7). Horizontal dotted line, corrected P = 0.1; vertical dotted lines, log₂ fold change = 0.5. Points in the red and gray areas represent genes up and down-regulated in taurine, respectively. Only genes with P < 0.3 are displayed. The annotated antimicrobial peptide (Rps19) has P = 0.102.

(G) GF mice were maintained on taurine after Kpn infection.

(H) Kpn CFU in the feces of vehicle and taurine-treated GF mice (2 pooled experiments, n = 13–15).

(I) SPF mice were maintained on taurine after C. rodentium infection.

(J) C. rodentium CFU in the feces of vehicle and taurine-treated mice (5 pooled experiments, n = 39–40).

(K-L) (K) Transverse colonic sections and (L) quantification of colonic crypt lengths of vehicle and taurine-treated mice at 14 days post-C. rodentium infection (2 pooled experiments, n = 15, fold decrease = 1.5). Scale bar, 100 µm.

In box and whisker plots, lines connect the medians of paired cages. n.s., not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001 by Mann-Whitney (A, C, J, H, L), Student’s t (D), or Wald test (F, with Benjamini-Hochberg correction). See also Table S4.

NIHMS1654488-supplement-4.pdf^{(790.3KB, pdf)}

Figure S5. Taurine-derived sulfide inhibits pathogen respiration, Related to Figures 4 and 5

(A) A pool of Kpn transposon mutants was passaged through control and post-∆yopM mice. Genes in which mutants were preferentially depleted (X) from the output of control or post-∆yopM mice contributed to fitness in that condition.

(B) Kpn genes that differentially contribute to fitness in post-∆yopM vs. control mice. Highlighted genes mediate respiration with anaerobic electron acceptors (left panel, circles), utilization of non-fermentable substrates (middle panel, squares), or respiration with oxygen (right panel, triangles). Data represent 2 pooled experiments (n = 6–7 samples per group). Horizontal dotted line, P = 0.05; vertical dotted lines, log₂ fold change = 1. Points in the gray and red areas represent genes that contribute to fitness in control and post-∆yopM mice, respectively.

(D) The CI of Kpn WT:acrE^- in the feces of vehicle and taurine-treated mice at 1 day post-Kpn infection (2 pooled experiments, n = 11–12). The dotted line indicates equal fitness between WT and mutant.

(E) Paired CFU of the Kpn WT (circles) and the cyxB^-, pduQ^-, or acrE^- mutant (triangles) in the feces of vehicle and taurine-treated mice at 1 day post-Kpn infection. Lines connect WT and mutant CFU within individual mice (represented as CI in Figure 4G and panel C). The dotted line indicates the detection limit.

(F) The CI of the C. rodentium WT and cydAB^- mutant after growth on formate ± oxygen (2 pooled experiments, n = 6).

(G) Fold increase in C. rodentium CFU after growth on formate ± sodium hydrosulfide (NaHS) (3 pooled experiments, n = 8).

(H) The CI of C. rodentium WT:cydAB^- in the feces of vehicle and taurine-treated mice (2 pooled experiments, n = 10). The CI for mice with <10 detectable CFU were excluded. Outliers (black points) were identified using the ROUT method (Q = 0.1%) and excluded when determining significance. Lines connect the medians of paired cages. The dotted line indicates equal fitness between WT and mutant. Bars show mean ± SEM. n.s., not significant; †, P < 0.1; *, P < 0.05; **, P < 0.01; ***, P < 0.001 by Mann-Whitney (C, D, F-H), Wald (B), or Wilcoxon matched-pairs signed-rank test (D).

NIHMS1654488-supplement-5.pdf^{(1.1MB, pdf)}

Table S1. Analyzed 16S sequencing data, Related to Figures 1, 3, 5, S1, and S3

Relative abundance (%) of taxa (phyla, order, or species, as indicated in tab titles) in the gut microbiota of individual post-Yptb (∆yopM or WT) mice (tabs: Yptb phyla and Yptb species), GF mice conventionalized with post-∆yopM microbiota (tab: GF_yopM species), wildR F2 mice (tabs: wildR_F2 phyla and wildR_F2 species), wildR F10 mice (tabs: wildR_F10 phyla and wildR_F10 species), conventionalized GF mice pre-engrafted with Bilophila wadsworthia (tab: GF_Bilophila species), taurine-treated mice (tab: taurine species), bismuth-treated mice (tabs: bismuth order and bismuth species), and respective control mice. Colors: yellow, P < 0.05 (Mann-Whitney test, adjusted with the Benjamini-Hochberg method, except for taurine mice); green and red, fold change in median abundance (log2) > 1 and < −1, respectively. Unique feature IDs were generated by QIIME 2.

NIHMS1654488-supplement-6.xlsx^{(1.2MB, xlsx)}

Table S2. Analyzed shotgun metagenomics data, Related to Figures 2, 3, and S3

Relative abundance (ppm with a pseudocount) of EC numbers in the gut metagenomes of individual post-∆yopM (tab: yopM), F2 wildR (tab: wildR), taurine-treated (tab: taurine), and respective control mice, and relative contribution in ppm of individual last known taxa (LKT) to EC numbers in the gut metagenomes of post-∆yopM (tab: yopM pie), F2 wildR (tab: wildR pie), taurine-treated (tab: taurine pie), and respective control mice. Colors: yellow, P < 0.05 (Mann-Whitney test); green and red, fold change in median abundance (log2) > 0.5 and < −0.5, respectively.

NIHMS1654488-supplement-7.xlsx^{(1.1MB, xlsx)}

Table S3. Analyzed mass spectrometry data, Related to Figures 2 and S3

Abundance of primary metabolites (in arbitrary units, tab: primary metabolites) and bile acids (in pg/mg, tab: bile acids) in the cecal contents of individual control and post-∆yopM mice. Colors: yellow, P < 0.1 (Mann-Whitney test); green and red, fold change in median abundance (log2) > 0.5 and < −0.5, for primary metabolites, or > 0 and < 0, for bile acids. C numbers are KEGG compound database identifiers. In Figure S3, the experiment C value for β-muricholic acid was excluded due to being an outlier.

NIHMS1654488-supplement-8.xlsx^{(156.3KB, xlsx)}

Table S4. Analyzed RNA-seq data, Related to Figures S3 and S4

Normalized gene expression (transcripts per million) in the livers of individual control and post-∆yopM mice (tab: liver) and in the large intestinal epithelial cells of individual vehicle and taurine-treated mice (tab: gut). Colors: yellow, P < 0.1 (Wald test, adjusted with the Benjamini-Hochberg method); green and red, fold change (log2) > 0.5 and < −0.5, for livers, or > 0 and < 0, for gut epithelial cells.

NIHMS1654488-supplement-9.xlsx^{(8.2MB, xlsx)}

Table S5. Analyzed Tn-seq and whole-genome sequencing data, Related to Figures 4 and S5

Normalized Kpn Tn counts per gene in the output pools of individual control and post-∆yopM mice infected with the Kpn Tn mutant pool (tab: Tn-seq), and genome annotations of E. faecalis strains isolated from bismuth-treated mice (tabs: A1, A2, B1). Colors: yellow, P < 0.05 (Wald test); green and red, fold change (log2) > 1 and < −1, respectively. The following samples were pooled in silico due to bottlenecking (< 5,000 unique Tn insertions): control 4 and 5 (from experiment A); control 6, 7, 8, and 10 (from experiment B); post-∆yopM 6, 7, 8, and 10 (from experiment B). NA values are DESeq2-filtered outliers. K numbers are KEGG Orthology database identifiers. Genes plotted in Figure S5 are indicated by the last column.

NIHMS1654488-supplement-10.xlsx^{(1.2MB, xlsx)}

Prior intestinal infection increases the gut microbiota’s resistance to subsequent infection
Infection induces host production of taurine and the expansion of taurine-utilizing gut taxa
The gut microbiota converts taurine to sulfide, inhibiting the ability of pathogens to respire
Sulfide sequestration unleashes an expansion of endogenous respirers in the microbiota

Acknowledgements

This work was supported by the Intramural Research Programs of NIAID (1ZIAAI001115–11 to Y.B.), NIDDK (1ZIADK054508–22 to B.R.), NCI Center of Cancer Research (1ZIABC011153–11 to G.T.), and NHGRI (1ZIAHG000180–19 to J.A.S.), the NIH Director’s Challenge Innovation Award Program (Y.B. and B.R.), the DDIR Innovation Award Program (Y.B. and B.R.), the NIGMS Postdoctoral Research Associate Training (PRAT) Program (1FI2GM128736–01 to A.S.), the NIH Office of Dietary Supplements Research Scholars Program (A.S.), the Pew Latin American Fellows Program (V.A.-O.), the Deutsche Forschungsgemeinschaft (HI 2088/1–1 to B.H.), the Human Frontier Science Program (LT000191/2018 to A.I.L.), and the Cancer Research Institute Irvington Postdoctoral Fellowship Program (M.E.). We thank Xingyao He, Shreni Mistry, and Naisarg Modi (NIAID Microbiome Program); Nicolas Bouladoux, Ejae Lewis, and Dana Trageser-Cesler (NIAID Microbiome Program Gnotobiotic Animal Facility); Kelly Paglia (West Coast Metabolomics Center); Shurjo Sen (NCI Cancer and Inflammation Program); Galina Koroleva (NIH Center for Human Immunology); Kim Webber (Belkaid laboratory), Mirian Krystel De Siqueira (now UCLA), and Austin Hall (now UW-Madison) for technical assistance; Margery Smelkinson (NIAID Biological Imaging Section); Michael A. Eckhaus (NIH Veterinary Resources Program) for scoring liver sections; Krystyna Szymczyk (NIAID Research Technologies Branch); Gad Frankel (Imperial College London) for providing C. rodentium ICC169; Andreas Bäumler (UC Davis) for providing pCAL60; the NIAID and NIDDK animal facilities; the NIH HPC Biowulf cluster (http://hpc.nih.gov); and Les Laboratoires Servier (https://smart.servier.com/) for medical art. We also thank Ivan Vujkovic-Cvijin and other members of the Belkaid laboratory for providing additional expertise and support.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Declaration of Interests

NIDDK licensed wildR mice to Taconic Biosciences.

References

Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, Davis AP, Dolinski K, Dwight SS, Eppig JT, et al. (2000). Gene Ontology: Tool for the unification of biology. Nat. Genet 25, 25–29. [DOI] [PMC free article] [PubMed] [Google Scholar]
Bankevich A, Nurk S, Antipov D, Gurevich AA, Dvorkin M, Kulikov AS, Lesin VM, Nikolenko SI, Pham S, Prjibelski AD, et al. (2012). SPAdes: A New Genome Assembly Algorithm and Its Applications to Single-Cell Sequencing. J. Comput. Biol 19, 455–477. [DOI] [PMC free article] [PubMed] [Google Scholar]
Barton LL, Ritz NL, Fauque GD, and Lin HC (2017). Sulfur cycling and the intestinal microbiome. Dig. Dis. Sci 62, 2241–2257. [DOI] [PubMed] [Google Scholar]
Beura LK, Hamilton SE, Bi K, Schenkel JM, Odumade OA, Casey KA, Thompson EA, Fraser KA, Rosato PC, Filali-Mouhim A, et al. (2016). Normalizing the environment recapitulates adult human immune traits in laboratory mice. Nature 532, 512–516. [DOI] [PMC free article] [PubMed] [Google Scholar]
Bierer DW (1990). Bismuth subsalicylate: History, chemistry, and safety. Rev. Infect. Dis 12, S3–S8. [DOI] [PubMed] [Google Scholar]
Bolger AM, Lohse M, and Usadel B (2014). Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120. [DOI] [PMC free article] [PubMed] [Google Scholar]
Bolyen E, Rideout JR, Dillon MR, Bokulich NA, Abnet CC, Al-Ghalith GA, Alexander H, Alm EJ, Arumugam M, Asnicar F, et al. (2019). Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nat. Biotechnol 37, 852–857. [DOI] [PMC free article] [PubMed] [Google Scholar]
Broberg CA, Wu W, Cavalcoli JD, Miller VL, and Bachman MA (2014). Complete Genome Sequence of Klebsiella pneumoniae Strain ATCC 43816 KPPR1, a Rifampin-Resistant Mutant Commonly Used in Animal, Genetic, and Molecular Biology Studies. Genome Announcements 2, e00924–14. [DOI] [PMC free article] [PubMed] [Google Scholar]
Bult CJ, Blake JA, Smith CL, Kadin JA, Richardson JE, the Mouse Genome Database Group. (2019). Mouse Genome Database (MGD) 2019. Nucleic Acids Res 47, D801–D806. [DOI] [PMC free article] [PubMed] [Google Scholar]
Caporaso JG, Lauber CL, Walters WA, Berg-Lyons D, Huntley J, Fierer N, Owens SM, Betley J, Fraser L, Bauer M, et al. (2012). Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J 6, 1621–1624. [DOI] [PMC free article] [PubMed] [Google Scholar]
Centers for Disease Control and Prevention (2019). Antibiotic Resistance Threats in the United States, 2019
Chao MC, Abel S, Davis BM, and Waldor MK (2016). The design and analysis of transposon insertion sequencing experiments. Nat. Rev. Microbiol 14, 119–128. [DOI] [PMC free article] [PubMed] [Google Scholar]
Conlan S, Park M, Deming C, Thomas PJ, Young AC, Coleman H, Sison C, Weingarten RA, Lau AF, Dekker JP, et al. (2016). Plasmid dynamics in KPC-positive Klebsiella pneumoniae during long-term patient colonization. mBio 7, e00742–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
Cook AM, and Denger K (2006). Metabolism of taurine in microorganisms: A primer in molecular biodiversity? In Taurine 6, Advances in Experimental Medicine and Biology, vol. 583, Oja SS, Saransaari P, eds. (Boston, MA: Springer; ), pp. 3–13. [DOI] [PubMed] [Google Scholar]
Dey N, Wagner VE, Blanton LV, Cheng J, Fontana L, Haque R, Ahmed T, and Gordon JI (2015). Regulators of Gut Motility Revealed by a Gnotobiotic Model of Diet-Microbiome Interactions Related to Travel. Cell 163, 95–107. [DOI] [PMC free article] [PubMed] [Google Scholar]
Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, Batut P, Chaisson M, Gingeras TR (2013). STAR: Ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
Faber F, Thiennimitr P, Spiga L, Byndloss MX, Litvak Y, Lawhon S, Andrews-Polymenis HL, Winter SE, and Bäumler AJ (2017). Respiration of microbiota-derived 1,2-propanediol drives Salmonella expansion during colitis. PLoS Pathog 13, e1006129. [DOI] [PMC free article] [PubMed] [Google Scholar]
da Fonseca DM, Hand TW, Han SJ, Gerner MY, Zaretsky AG, Byrd AL, Harrison OJ, Ortiz AM, Quinones M, Trinchieri G, et al. (2015). Microbiota-dependent sequelae of acute infection compromise tissue-specific immunity. Cell 163, 354–366. [DOI] [PMC free article] [PubMed] [Google Scholar]
Forte E, Borisov VB, Falabella M, Colaço HG, Tinajero-Trejo M, Poole RK, Vicente JB, Sarti P, and Giuffre A (2016). The terminal oxidase cytochrome bd promotes sulfide-resistant bacterial respiration and growth. Sci. Rep 6, 1–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
Fukumoto S, Toshimitsu T, Matsuoka S, Maruyama A, Oh-oka K, Takamura T, Nakamura Y, Ishimaru K, Fujii-Kuriyama Y, Ikegami S, et al. (2014). Identification of a probiotic bacteria-derived activator of the aryl hydrocarbon receptor that inhibits colitis. Immunol. Cell Biol 92, 460–465. [DOI] [PubMed] [Google Scholar]
Goodman AL, Wu M, and Gordon JI (2011). Identifying microbial fitness determinants by insertion sequencing using genome-wide transposon mutant libraries. Nat. Protoc 6, 1969–1980. [DOI] [PMC free article] [PubMed] [Google Scholar]
Gorrie CL, Mirčeta M, Wick RR, Edwards DJ, Thomson NR, Strugnell RA, Pratt NF, Garlick JS, Watson KM, Pilcher DV, et al. (2017). Gastrointestinal carriage is a major reservoir of Klebsiella pneumoniae infection in intensive care patients. Clin. Infect. Dis 65, 208–215. [DOI] [PMC free article] [PubMed] [Google Scholar]
Goryshin IY, Miller JA, Kil YV, Lanzov VA, and Reznikoff WS (1998). Tn5/IS50 target recognition. Proc. Natl. Acad. Sci. U. S. A 95, 10716–10721. [DOI] [PMC free article] [PubMed] [Google Scholar]
Han SJ, Glatman Zaretsky A, Andrade-Oliveira V, Collins N, Dzutsev A, Shaik J, Morais da Fonseca D, Harrison OJ, Tamoutounour S, Byrd AL, et al. (2017). White adipose tissue is a reservoir for memory T cells and promotes protective memory responses to infection. Immunity 47, 1154–1168.e6. [DOI] [PMC free article] [PubMed] [Google Scholar]
Heinz S, Benner C, Spann N, Bertolino E, Lin YC, Laslo P, Cheng JX, Murre C, Singh M, and Glass CK (2010). Simple Combinations of Lineage-Determining Transcription Factors Prime cis-Regulatory Elements Required for Macrophage and B Cell Identities. Mol. Cell 38, 576–589. [DOI] [PMC free article] [PubMed] [Google Scholar]
Hung C-S, Dodson KW, and Hultgren SJ (2009). A murine model of urinary tract infection. Nat. Protoc 4, 1230–1243. [DOI] [PMC free article] [PubMed] [Google Scholar]
Jia W, Xie G, and Jia W (2018). Bile acid–microbiota crosstalk in gastrointestinal inflammation and carcinogenesis. Nat. Rev. Gastroenterol. Hepatol 15, 111–128. [DOI] [PMC free article] [PubMed] [Google Scholar]
Kaesler S, Sobiesiak M, Kneilling M, Volz T, Kempf WE, Lang PA, Lang KS, Wieder T, Heller-Stilb B, Warskulat U, et al. (2012). Effective T-cell recall responses require the taurine transporter Taut. Eur. J. Immunol 42, 831–841. [DOI] [PubMed] [Google Scholar]
Kamdar K, Khakpour S, Chen J, Leone V, Brulc J, Mangatu T, Antonopoulos DA, Chang EB, Kahn SA, Kirschner BS, et al. (2016). Genetic and Metabolic Signals during Acute Enteric Bacterial Infection Alter the Microbiota and Drive Progression to Chronic Inflammatory Disease. Cell Host Microbe 19, 21–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
Kim S, Covington A, and Pamer AG (2017). The intestinal microbiota: Antibiotics, colonization resistance, and enteric pathogens. Immunol. Rev 279, 90–105. [DOI] [PMC free article] [PubMed] [Google Scholar]
Langmead B and Salzberg SL (2012). Fast gapped-read alignment with Bowtie 2. Nat. Methods 9, 357–359. [DOI] [PMC free article] [PubMed] [Google Scholar]
Levy M, Thaiss CA, Zeevi D, Dohnalová L, Zilberman-Schapira G, Mahdi JA, David E, Savidor A, Korem T, Herzig Y, et al. (2015). Microbiota-Modulated Metabolites Shape the Intestinal Microenvironment by Regulating NLRP6 Inflammasome Signaling. Cell 163, 1428–1443. [DOI] [PMC free article] [PubMed] [Google Scholar]
Li D, Liu CM, Luo R, Sadakane K, and Lam TW (2015). MEGAHIT: An ultra-fast single-node solution for large and complex metagenomics assembly via succinct de Bruijn graph. Bioinformatics 31, 1674–1676. [DOI] [PubMed] [Google Scholar]
Li M, Cai S-Y, and Boyer JL (2017). Mechanisms of bile acid mediated inflammation in the liver. Mol. Aspects Med 56, 45–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
Liao Y, Smyth GK, and Shi W (2019). The R package Rsubread is easier, faster, cheaper and better for alignment and quantification of RNA sequencing reads. Nucleic Acids Res 47, e47. [DOI] [PMC free article] [PubMed] [Google Scholar]
Lopez CA, Miller BM, Rivera-Chávez F, Velazquez EM, Byndloss MX, Chávez-Arroyo A, Lokken KL, Tsolis RM, Winter SE, and Bäumler AJ (2016). Virulence factors enhance Citrobacter rodentium expansion through aerobic respiration. Science 353, 1249–1253. [DOI] [PMC free article] [PubMed] [Google Scholar]
Love MI, Huber W, and Anders S (2014). Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol 15, 550. [DOI] [PMC free article] [PubMed] [Google Scholar]
Macpherson AJ, De Agüero MG, and Ganal-Vonarburg SC (2017). How nutrition and the maternal microbiota shape the neonatal immune system. Nat. Rev. Immunol 17, 508–517. [DOI] [PubMed] [Google Scholar]
Martin HM, Hancock JT, Salisbury V, and Harrison R (2004). Role of xanthine oxidoreductase as an antimicrobial agent. Infect. Immun 72, 4933–4939. [DOI] [PMC free article] [PubMed] [Google Scholar]
Martin M (2011). Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet.journal 17, 10–12. [Google Scholar]
McDonald AG, Boyce S, and Tipton KF (2009). ExplorEnz: The primary source of the IUBMB enzyme list. Nucleic Acids Res 37, D593–D597. [DOI] [PMC free article] [PubMed] [Google Scholar]
McLoughlin K, Schluter J, Rakoff-Nahoum S, Smith AL, and Foster KR (2016). Host Selection of Microbiota via Differential Adhesion. Cell Host Microbe 19, 550–559. [DOI] [PubMed] [Google Scholar]
Molloy MJ, Grainger JR, Bouladoux N, Hand TW, Koo LY, Naik S, Quinones M, Dzutsev AK, Gao JL, Trinchieri G, et al. (2013). Intraluminal containment of commensal outgrowth in the gut during infection-induced dysbiosis. Cell Host Microbe 14, 318–328. [DOI] [PMC free article] [PubMed] [Google Scholar]
Moriya Y, Itoh M, Okuda S, Yoshizawa AC, Kanehisa M (2007). KAAS: an automatic genome annotation and pathway reconstruction server. Nucleic Acids Res 35, W182–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
Netea MG, Domínguez-Andrés J, Barreiro LB, Chavakis T, Divangahi M, Fuchs E, Joosten LAB, van der Meer JWM, Mhlanga MM, Mulder WJM, et al. (2020). Defining trained immunity and its role in health and disease. Nat. Rev. Immunol 20, 375–388. [DOI] [PMC free article] [PubMed] [Google Scholar]
Neumann-Schaal M, Jahn D, and Schmidt-Hohagen K (2019). Metabolism the difficile way: The key to the success of the pathogen Clostridioides difficile. Front. Microbiol 10, 219. [DOI] [PMC free article] [PubMed] [Google Scholar]
Pamer EG (2014). Fecal microbiota transplantation: Effectiveness, complexities, and lingering concerns. Mucosal Immunol 7, 210–214. [DOI] [PubMed] [Google Scholar]
Peck SC, Denger K, Burrichter A, Irwin SM, Balskus EP, and Schleheck D (2019). A glycyl radical enzyme enables hydrogen sulfide production by the human intestinal bacterium Bilophila wadsworthia. Proc. Natl. Acad. Sci. U. S. A 116, 3171–3176. [DOI] [PMC free article] [PubMed] [Google Scholar]
Ramage B, Erolin R, Held K, Gasper J, Weiss E, Brittnacher M, Gallagher L, and Manoil C (2017). Comprehensive arrayed transposon mutant library of Klebsiella pneumoniae outbreak strain KPNIH1. J. Bacteriol 199, e00352–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
Ramanan D, Bowcutt R, Lee SC, Tang MS, Kurtz ZD, Ding Y, Honda K, Gause WC, Blaser MJ, Bonneau RA, et al. (2016). Helminth infection promotes colonization resistance via type 2 immunity. Science 352, 608–612. [DOI] [PMC free article] [PubMed] [Google Scholar]
Ramsey M, Hartke A, and Huycke M (2014). The Physiology and Metabolism of Enterococci. In Enterococci: From Commensals to Leading Causes of Drug Resistant Infection, Gilmore MS, ed. (Boston, MA: Massachusetts Eye and Ear Infirmary; ). [PubMed] [Google Scholar]
Reese TA, Bi K, Kambal A, Filali-Mouhim A, Beura LK, Bürger MC, Pulendran B, Sekaly RP, Jameson SC, Masopust D, et al. (2016). Sequential Infection with Common Pathogens Promotes Human-like Immune Gene Expression and Altered Vaccine Response. Cell Host Microbe 19, 713–719. [DOI] [PMC free article] [PubMed] [Google Scholar]
Rivera-Chávez F, Zhang LF, Faber F, Lopez CA, Byndloss MX, Olsan EE, Xu G, Velazquez EM, Lebrilla CB, Winter SE, et al. (2016). Depletion of Butyrate-Producing Clostridia from the Gut Microbiota Drives an Aerobic Luminal Expansion of Salmonella. Cell Host Microbe 19, 443–454. [DOI] [PMC free article] [PubMed] [Google Scholar]
Romain G, Tremblay S, Arena ET, Antunes LCM, Covey S, Chow MT, Finlay BB, and Menendez A (2013). Enterohepatic bacterial infections dysregulate the FGF15-FGFR4 endocrine axis. BMC Microbiol 13, 1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
Rosshart SP, Vassallo BG, Angeletti D, Hutchinson DS, Morgan AP, Takeda K, Hickman HD, McCulloch JA, Badger JH, Ajami NJ, et al. (2017). Wild Mouse Gut Microbiota Promotes Host Fitness and Improves Disease Resistance. Cell 171, 1015–1028.e13. [DOI] [PMC free article] [PubMed] [Google Scholar]
Rosshart SP, Herz J, Vassallo BG, Hunter A, Wall MK, Badger JH, McCulloch JA, Anastasakis DG, Sarshad AA, Leonardi I, et al. (2019). Laboratory mice born to wild mice have natural microbiota and model human immune responses. Science 365, eaaw4361. [DOI] [PMC free article] [PubMed] [Google Scholar]
Roubaud-Baudron C, Ruiz VE, Swan AM, Vallance BA, Ozkul C, Pei Z, Li J, Battaglia TW, Perez-Perez GI, and Blaser MJ (2019). Long-term effects of early-life antibiotic exposure on resistance to subsequent bacterial infection. mBio 10, e02820–19. [DOI] [PMC free article] [PubMed] [Google Scholar]
Schindelin J, Arganda-Carreras I, Frise E, Kaynig V, Longair M, Pietzsch T, Preibisch S, Rueden C, Saalfeld S, Schmid B, et al. (2012). Fiji: An open-source platform for biological-image analysis. Nat. Methods 9, 676–682. [DOI] [PMC free article] [PubMed] [Google Scholar]
Singer JR, Blosser EG, Zindl CL, Silberger DJ, Conlan S, Laufer VA, DiToro D, Deming C, Kumar R, Morrow CD, et al. (2019). Preventing dysbiosis of the neonatal mouse intestinal microbiome protects against late-onset sepsis. Nat. Med 25, 1772–1782. [DOI] [PMC free article] [PubMed] [Google Scholar]
Sonnenburg ED, and Sonnenburg JL (2019). The ancestral and industrialized gut microbiota and implications for human health. Nat. Rev. Microbiol 17, 383–390. [DOI] [PubMed] [Google Scholar]
Stacy A, Fleming D, Lamont RJ, Rumbaugh KP, and Whiteley M (2016). A commensal bacterium promotes virulence of an opportunistic pathogen via cross-respiration. mBio 7, e00782–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
Stewart CJ, Ajami NJ, O’Brien JL, Hutchinson DS, Smith DP, Wong MC, Ross MC, Lloyd RE, Doddapaneni HV, Metcalf GA, et al. (2018). Temporal development of the gut microbiome in early childhood from the TEDDY study. Nature 562, 583–588. [DOI] [PMC free article] [PubMed] [Google Scholar]
Suez J, and Elinav E (2017). The path towards microbiome-based metabolite treatment. Nat. Microbiol 2, 1–5. [DOI] [PubMed] [Google Scholar]
Torsten S (2014). Prokka: Rapid prokaryotic genome annotation. Bioinformatics 30, 2068–2069. [DOI] [PubMed] [Google Scholar]
United Nations Interagency Coordination Group on Antimicrobial Resistance (2019). No Time to Wait: Securing the future from drug-resistant infections
Uribe JH, Collado-Romero M, Zaldívar-López S, Arce C, Bautista R, Carvajal A, Cirera S, Claros MG, and Garrido JJ (2016). Transcriptional analysis of porcine intestinal mucosa infected with Salmonella Typhimurium revealed a massive inflammatory response and disruption of bile acid absorption in ileum. Vet. Res 47, 1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
Wahlström A, Sayin SI, Marschall HU, and Bäckhed F (2016). Intestinal Crosstalk between Bile Acids and Microbiota and Its Impact on Host Metabolism. Cell Metab 24, 41–50. [DOI] [PubMed] [Google Scholar]
Wood DE, Lu J, and Langmead B (2019). Improved metagenomic analysis with Kraken 2. Genome Biol 20, 257. [DOI] [PMC free article] [PubMed] [Google Scholar]
Wotzka SY, Kreuzer M, Maier L, Arnoldini M, Nguyen BD, Brachmann AO, Berthold DL, Zünd M, Hausmann A, Bakkeren E, et al. (2019). Escherichia coli limits Salmonella Typhimurium infections after diet shifts and fat-mediated microbiota perturbation in mice. Nat. Microbiol 4, 2164–2174. [DOI] [PMC free article] [PubMed] [Google Scholar]
Yilmaz P, Parfrey LW, Yarza P, Gerken J, Pruesse E, Quast C, Schweer T, Peplies J, Ludwig W, and Glöckner FO (2014). The SILVA and “All-species Living Tree Project (LTP)” taxonomic frameworks. Nucleic Acids Res 42, D643–D648. [DOI] [PMC free article] [PubMed] [Google Scholar]
Zhou Y, Zhou B, Pache L, Chang M, Khodabakhshi AH, Tanaseichuk O, Benner C, and Chanda SK (2019). Metascape provides a biologist-oriented resource for the analysis of systems-level datasets. Nat. Commun 10, 1523. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

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

Supplementary Materials