Gut microbes modulate (p)ppGpp during a time-restricted feeding regimen

ABSTRACT

Mammals exhibit daily fasting and feeding patterns that produce a fluctuating environment in the gut. By colonizing germfree mice with Bacteroides thetaiotaomicron and examining gene expression through a time-restricted feeding cycle, we demonstrate that this prominent gut commensal exhibits gene expression patterns characteristic of a stringent response and increases ppGpp levels during the host-fasting phase of the feeding regimen. Mutants unable to produce (p)ppGpp fail to produce this transcriptional response, exhibit unrestrained chromosomal replication, and cannot maintain population size during the host-fasting phase. Additionally, B. thetaiotaomicron requires (p)ppGpp to utilize host glycans both in vitro and in vivo, and mutants unable to produce (p)ppGpp display deficiencies in mucus layer colonization during the host-fasting phase. Complete gut microbial communities from mice and humans also increase ppGpp levels in this manner, demonstrating that this response appears to be general across species and conserved across mammalian gut communities. Together, these results identify an intracellular signal that allows gut microbes to coordinate their physiology with a time-restricted feeding regimen of their host.

IMPORTANCE

Mammals do not eat continuously, instead concentrating their feeding to a restricted portion of the day. This behavior presents the mammalian gut microbiota with a fluctuating environment with consequences for host-microbiome interaction, infection risk, immune response, drug metabolism, and other aspects of health. We demonstrate that in mice, gut microbes elevate levels of an intracellular signaling molecule, (p)ppGpp, during the fasting phase of a time-restricted feeding regimen. Disabling this response in a representative human gut commensal species significantly reduces colonization during this host-fasting phase. This response appears to be general across species and conserved across mammalian gut communities, highlighting a pathway that allows healthy gut microbiomes to maintain stability in an unstable environment.

INTRODUCTION

The mammalian gut microbiota comprises one of the densest known microbial communities, reaching levels of 10¹¹–10¹² cells per milligram (1). These microbes compete for limited nutrients in a fluctuating environment, where periodic waves of nutrients enter and exit the digestive tract in accordance with the feeding patterns of the host (2, 3). Both mice and humans do not eat continuously, instead concentrating feeding to a restricted portion of the day (primarily daytime for humans, primarily night for mice) (4–7). Previous research indicates that fasting and feeding cycles affect gut microbiota community composition, metabolite production, and adherence to epithelial cells (8–11). Such outcomes have important implications for human health: for example, host fasting and feeding cycles modulate the production of antimicrobial peptides in a microbiota-dependent manner (12) and fasting enhances colonization resistance to pathogens (13). While these studies indicate that the microbiota senses and responds to host eating (10), how resident gut microbes coordinate their physiology with host feeding and fasting is largely unexplored.

Periodic feeding impacts the range of nutrients available to gut microbes. Gut bacteria possess the ability to utilize many different nutritional sources, including a wide array of complex polysaccharides and fibers. For example, members of the genus Bacteroides (the most prominent genus in the gut of many individuals) typically devote 20% of their genomes to polysaccharide utilization loci (PULs) which are involved in sensing and degrading diverse carbohydrates (14). This allows human gut Bacteroides and their relatives to metabolize a wide variety of food sources, including diet-derived glycans that largely escape host digestive enzymes, and host-derived mucus glycans that line the gut lumen (14, 15).

The bacterial response to starvation and nutrient availability has been most widely explored in laboratory studies of Escherichia coli and Bacillus subtilis. Under laboratory conditions, these species and others adapt to nutrient starvation through the intracellular molecules guanosine-5′−3′-bispyrophosphate (ppGpp) and guanosine pentaphosphate (pppGpp), collectively referred to as (p)ppGpp (16–18). In response to nutrient deprivation and other stresses, bacteria produce (p)ppGpp, which, in turn, leads to a repression of growth and induction of stasis through multiple transcriptional and post-transcriptional mechanisms (16–19). This “stringent response” allows bacteria to allocate resources away from protein production in the absence of necessary resources for growth, in some cases activating alternate pathways. In a classic example, E. coli responds to amino acid starvation in laboratory culture by producing (p)ppGpp, which, in turn, represses expression of tRNAs and other translational machinery and activates amino acid biosynthetic pathways (20, 21). (p)ppGpp⁰ strains, which lack (p)ppGpp biosynthetic enzymes and are unable to activate this response, fail to survive amino acid starvation (22). While these in vitro studies typically use no-carbon conditions or chemical inhibition of tRNA synthetases to induce the stringent response, whether E. coli or other commensal bacteria engage these pathways during routine or extreme nutrient fluctuations in the gut is unexplored.

Commensal gut anaerobes also produce (p)ppGpp (23, 24). In the prominent gut commensal Bacteroides thetaiotaomicron, (p)ppGpp levels are controlled by two enzymes, BT0700 (which encodes a (p)ppGpp synthase domain) and BT3998 (which encodes both synthase and hydrolase domains) (24). As observed in E. coli, B. thetaiotaomicron requires (p)ppGpp to respond to environmental stresses and nutrient limitation, including oxygen exposure and carbon starvation (23, 24). Notably, a (p)ppGpp⁰ strain (that lacks BT0700 and BT3998) exhibits a loss of viability during carbon starvation and a significant fitness defect in gnotobiotic mice despite exhibiting normal exponential growth in rich and minimal medium in vitro (24). When and why this signal to halt growth is important in the competitive gut environment is unclear.

Here, we report that mouse gut microbial communities, and human gut microbiomes transplanted into gnotobiotic mice, respond to brief periods when the host is not eating through a transient wave of (p)ppGpp production. This adaptation is mirrored in the model human gut commensal B. thetaiotaomicron; in this species, (p)ppGpp⁰ strains exhibit a disrupted transcriptional program and significantly drop in population size within 12 h of host fasting. Together, these studies identify a general and conserved mechanism that allows gut microbiomes to align bacterial physiology with a time-restricted feeding regimen of their host.

RESULTS

B. thetaiotaomicron gene expression profiles vary during a timed feeding regimen in monoassociated gnotobiotic mice

To investigate how gut microbes change their physiology to adapt to the diurnal fasting patterns of the host, we colonized germfree mice with B. thetaiotaomicron and maintained these animals on a periodic feeding schedule in which food was available for 12 h during the dark phase, when mice are generally active and eating, and removed for 12 hours during the light phase, when mice are generally resting and not eating (Fig. 1A). This schedule allowed us to simulate the periodic feeding that humans follow (eating during active periods, fasting during resting periods). We conducted RNA sequencing on cecal contents collected at ZT24 (at the end of the feeding period, “fed” state) and ZT12 (at the end of the fasting period, “fasted” state).

Fig 1.

B. thetaiotaomicron alters its transcriptome during a time-restricted feeding regimen in monoassociated gnotobiotic mice. (A) Schematic of the mouse-feeding schedule. (B) Principal component analysis (PCA) of wild-type (WT) B. thetaiotaomicron transcriptomes during fed (blue) and fasted (orange) timepoints. Samples were collected from cecal contents of monoassociated gnotobiotic mice. (C) Volcano plot of WT B. thetaiotaomicron gene expression during fed and fasted timepoints. Gene expression changes were filtered for significance using a false discovery rate (FDR) of <0.05 and a log₂fold change >1 or <−1. Log₂fold change and P-adjusted value thresholds are represented by dotted lines. Genes significantly upregulated in fed or fasting phases are colored in blue or orange, respectively. (D) Gene set enrichment analysis of WT B. thetaiotaomicron gene expression during fed and fasted timepoints. Adjusted normalized enrichment scores are plotted for significant gene sets with a nominal P-value <0.05 and FDR <0.25. “Biosynthesis,” “secondary,” “metabolism,” and “environments” are abbreviated.

Principal component analysis (PCA) of normalized transcriptomic data showed that PC1, which explains ~80% of the observed variance, separates fed-state from fasted-state samples and suggests that these brief fasting and feeding periods affect the transcriptional profile of B. thetaiotomicron (Fig. 1B). Comparison of B. thetaiotaomicron transcriptomes between fed (ZT24) and fasted (ZT12) timepoints identified a subset of genes (348/4,902, representing ~7% of the genome) that were significantly differentially regulated between these conditions (twofold cutoff, P < 0.05 after multiple hypothesis testing correction) (Fig. 1C). To identify pathways upregulated in the fed or fasted state, we conducted gene set enrichment analysis (GSEA) (25) using B. thetaiotaomicron KEGG GO pathways (26) (Fig. 1D). Notably, aminoacyl-tRNA biosynthesis and ribosomal subunit production pathways were the most significantly enriched among the sets of genes upregulated in the fed state and repressed in the fasted state. By contrast, genes upregulated in the fasted state were enriched in glycan degradation pathways and sphingolipid metabolism, both of which include PULs responsible for metabolism of host and diet-derived polysaccharides.

Repression of aminoacyl-tRNA biosynthesis and ribosomal subunit production are both associated with the stringent response in diverse bacteria, including B. thetaiotaomicron (24, 27–29). We reasoned that if repression of aminoacyl-tRNA and ribosomal subunit pathways during the fasted phase was due to increased (p)ppGpp levels, a (p)ppGpp⁰ strain would fail to repress these pathways at this time point. To test this, we colonized germfree mice with a B. thetaiotaomicron (p)ppGpp⁰ strain (24) and conducted RNA-seq on cecal contents collected in the fed and fasted timepoints as above. PCA of wild-type (WT) and (p)ppGpp⁰ transcriptomes from these four groups showed that samples clustered first by B. thetaiotaomicron genotype, then by fed or fasted state (Fig. 2A), suggesting that the ability to produce (p)ppGpp affects gene expression in both states. This concurs with previous studies of gene expression in wild-type and (p)ppGpp⁰ B. thetaiotaomicron strains collected from monocolonized mice at an intermediate time of day (24). Compared to the wild-type strain, the (p)ppGpp⁰ strain differentially regulated nearly twice as many genes (659/4,902) between the fed and fasted time points (Fig. 2B). Many of these changes in gene expression in the (p)ppGpp⁰ strain are distinct from those found in the wild-type strain (Fig. S1A); shared responses are distributed between fed-upregulated and fasting-upregulated genes (Fig. S1B). Notably, GSEA analyses revealed that the (p)ppGpp⁰ strain failed to repress aminoacyl-tRNA biosynthesis and ribosomal subunit production pathways at the fasted time point relative to the fed timepoint, suggesting that these gene expression changes are (p)ppGpp-dependent in the wild-type strain (Fig. 2C). Together, these studies suggest that the (p)ppGpp⁰ strain exhibits a distinct transcriptional profile during both fed and fasted timepoints and that this signaling pathway is required for the stringent response signature exhibited by the wild-type strain during the fasted timepoint.

Fig 2.

B. thetaiotaomicron exhibits (p)ppGpp-dependent gene expression changes during host fed and fasted phases. (A) Principal component analysis plot of WT (triangle) and (p)ppGpp⁰ (circle) B. thetaiotaomicron transcriptomes during fed (blue) and fasted (orange) timepoints. (B) Volcano plot of gene expression of the B. thetaiotaomicron (p)ppGpp⁰ strain during fed and fasted timepoints. Significantly differentially expressed genes (FDR < 0.05 and a log₂fold change >1 or <−1) are colored in blue and orange. (C) Gene set enrichment analysis of B. thetaiotaomicron (p)ppGpp⁰ gene expression during fed and fasted timepoints. Adjusted normalized enrichment scores are plotted for significant gene sets with a nominal P-value <0.05 and FDR <0.25. Gene sets upregulated during the Fed state as in the WT strain are denoted with asterisks. Gene sets upregulated in the fasted or fed states in the WT but not the (p)ppGpp⁰ strain are noted in orange and blue text, respectively. “Biosynthesis,” “secondary,” “metabolism,” and “environments” are abbreviated.

B. thetaiotaomicron increases ppGpp levels during the fasting period of a timed-feeding regimen in mice

We next reasoned that if B. thetaiotaomicron alters its gene expression program in a (p)ppGpp-dependent manner between fed and fasted host states, the levels of this intracellular signal would be modulated accordingly. To test this, we first adapted and validated a ppGpp extraction and HPLC-based quantification method (30, 31) (Fig. S2A) that allowed the detection of this molecule at a lower limit of 2.5 µM (Fig. 3A; Fig. S2B). In cecal contents collected from B. thetaiotaomicron (p)ppGpp⁰-monoassociated gnotobiotic mice, ppGpp was not detected, establishing the specificity of the method in the presence of all other B. thetaiotaomicron and host metabolites in these gut contents (Fig. S2B).

Fig 3.

B. thetaiotaomicron increases ppGpp levels during the fasted phase of a time-restricted feeding regimen. (A) Chromatogram traces of HPLC-based ppGpp quantification. Varying shades of purple represent different concentrations of ppGpp standard. The structure of ppGpp (lower left) and area under the curve values (upper right) are also provided. (B) Cecal ppGpp levels during fed (blue) and fasted (orange) phases in WT B. thetaiotaomicron monoassociated gnotobiotic mice. Significance is indicated with asterisks (*P < 0.05, Mann-Whitney U test).

We next colonized germfree mice with wild-type B. thetaiotaomicron, provided food as described in Fig. 1A, and measured cecal ppGpp concentrations normalized to colony forming units (CFU) at the fed and fasted phase (Fig. 3B). Notably, ppGpp levels were significantly elevated in the fasted phase relative to the fed phase, consistent with the (p)ppGpp-associated changes in gene expression during the fasted timepoint.

B. thetaiotaomicron maintains constant abundance in fed and fasted mice in a (p)ppGpp-dependent manner

In vitro, B. thetaiotaomicron uses (p)ppGpp to maintain viability during carbon starvation; the (p)ppGpp⁰ strain exhibits a significant loss of viability under these conditions (24). To test the hypothesis that (p)ppGpp production in the fasted mouse gut similarly enables B. thetaiotaomicron to maintain its abundance when the host is not eating, we measured bacterial abundance (as determined by CFU/g cecal contents) of B. thetaiotaomicron wild type and (p)ppGpp⁰-monoassociated mice in fed and fasted states. Indeed, despite the lack of food consumption by the host during the fasting period, wild-type B. thetaiotaomicron maintains constant abundance at the fed and fasted phases; furthermore, wild-type and (p)ppGpp⁰ strains do not significantly differ in abundance during the fed state (Fig. 4A; if the mouse with highest colonization in the (p)ppGpp⁰ fed group is removed from the analysis, there remains no significant difference between groups). By contrast, the abundance of the (p)ppGpp⁰ strain drops significantly (by 53%, P < 0.05) during the fasted phase compared to the fed phase; this colonization level is also significantly lower than the wild-type strain at either timepoint (Fig. 4A).

Fig 4.

B. thetaiotaomicron requires (p)ppGpp to maintain abundance and coordinate chromosomal replication during fed and fasted phases. (A) Bacterial density (colony forming units per gram cecal content) during fed (blue) and fasted (orange) phases in gnotobiotic mice monoassociated with B. thetaiotaomicron WT or (p)ppGpp⁰ strains. (B) GRiD values during fed (blue) and fasted (orange) phases in gnotobiotic mice monoassociated with B. thetaiotaomicron WT or (p)ppGpp⁰ strains. GRiD values represent the peak-to-trough ratio of the number of sequencing reads mapping to the B. thetaiotaomicron origin and terminus of replication in each condition. In both panels, significance is indicated with asterisks (*P < 0.05; **P < 0.01; ***P < 0.001, ****P < 0.0001) based on a Kruskal-Wallace test followed by Dunn’s multiple comparison test.

B. thetaiotaomicron requires (p)ppGpp to coordinate chromosomal replication with a timed feeding regimen in mice

One of the hallmark features of (p)ppGpp production in E. coli and B. subtilis is the reduction of chromosomal replication rate, which allows bacteria to coordinate DNA replication with growth (32–34). Consequently, E. coli (p)ppGpp⁰ strains have higher chromosomal replication rates compared to their wild-type counterparts in vitro, regardless of growth rate (32, 34). We reasoned that if (p)ppGpp production modulates B. thetaiotaomicron chromosomal replication in the gut, wild-type bacteria would exhibit reduced chromosomal replication in the gut compared to the (p)ppGpp⁰ strain. To estimate B. thetaiotaomicron chromosomal replication rates, we conducted peak-to-trough analysis, which uses shotgun DNA sequencing to evaluate the relative abundance of DNA mapping to the chromosomal origin of replication compared to the terminus (35, 36). High peak-to-trough ratios reflect the initiation of one or more rounds of chromosomal replication at the origin before the previous round of replication is completed (higher replication rate), whereas lower ratios indicate fewer rounds of replication occurring concurrently (lower replication rate) (35). As expected, peak-to-trough analysis of B. thetaiotaomicron in vitro cultures revealed significantly higher peak-to-trough ratios in exponential phase compared to stationary phase or carbon starvation; the (p)ppGpp⁰ strain exhibited similar behavior (although with slightly but significantly higher peak-to-trough ratios during the stress conditions; Fig. S3). We colonized germfree mice with either wild-type or (p)ppGpp⁰ B. thetaiotaomicron, transferred animals to the restricted feeding schedule as previously described (Fig. 1A), and collected cecal contents at ZT12 and ZT24 for shotgun DNA sequencing and calculation of peak-to-trough ratios. Notably, the (p)ppGpp⁰ strain exhibited significantly elevated peak-to-trough ratios relative to the wild-type strain at both time points (Fig. 4B). These results suggest that B. thetaiotaomicron uses (p)ppGpp to limit chromosomal replication rates in the gut.

B. thetaiotaomicron utilizes host-associated glycans in a (p)ppGpp-dependent manner

In addition to modulating chromosomal replication rate, (p)ppGpp production has widespread impacts on gene expression (Fig. 2B and C). We reasoned that in addition to (p)ppGpp-repressed genes (encoding aminoacyl-tRNAs and ribosomal subunits), genes activated during fasting in a (p)ppGpp-dependent manner could also contribute to the ability of B. thetaiotaomicron to maintain its abundance in the fasted phase and reveal differences in the niche of wild-type and (p)ppGpp⁰ strains when the host is not eating. GSEA analysis highlights “other glycan degradation” and “sphingolipid metabolism” as the pathways enriched among fasting-upregulated genes in the wild-type strain but not in the (p)ppGpp⁰ strain (Fig. 1D and 2C).

Notably, most (29/38, ~75%) of the genes in these pathways are encoded in PULs. To investigate PUL regulation during the fed and fasted phases, we compared the expression of susC and susD genes between the wild-type and (p)ppGpp⁰ strains in fed and fasted states (Fig. 5A). SusC and susD are conserved across PULs, encode the central transport components of these multiprotein machines (37), and are generally expressed in response to their substrate (38). Consistent with published observations that B. thetaiotaomicron upregulates host glycan-targeting PULs in the presence of host glycans if plant polysaccharides are limiting (39), the wild-type strain upregulated multiple PULs predicted to target host glycans during the fasted timepoint (Fig. 5A). By contrast, the (p)ppGpp⁰ strain failed to upregulate many of these host glycan-targeting PULs during this timepoint (Fig. 5A).

Fig 5.

B. thetaiotaomicron utilizes host-associated glycans in a (p)ppGpp-dependent manner. (A) Heatmap of gene expression of susC/D pairs in B. thetaiotaomicron WT or (p)ppGpp⁰ strains in cecal contents of monoassociated gnotobiotic mice during fed or fasted timepoints. PUL number and predicted substrates are from PULDB (40). PULs shown are those with altered expression in the WT strain between fed and fasted mice [susCD log₂(FoldChange) >0.5 or <−0.5]. Rhamnogalacturonan and arabinogalactan are abbreviated as RG and AG, respectively. PULs 14, 82, and 72 also metabolize N-glycans. (B) Growth of B. thetaiotaomicron WT (black) and (p)ppGpp⁰ (green) strains in minimal medium containing 0.1% (wt/vol) polygalacturonic acid (PGA) and 0.8% (wt/vol) porcine mucin O-glycans (PMOG) as the sole carbon source. (C) Growth of WT and (p)ppGpp⁰ strains in minimal medium containing 0.2% PGA and 0.6% PMOG. (D) Growth of WT and (p)ppGpp⁰ strains in minimal medium containing either 0.1% PGA (left) or 0.1% PMOG (right) as the sole carbon source. (E) Bacterial density during the fasted timepoint in cecal lumen (left) or mucus layer (right) in gnotobiotic mice monoassociated with B. thetaiotaomicron WT or (p)ppGpp⁰ strains. Significance is indicated with asterisks (*P < 0.05, Mann-Whitney U test).

In the presence of dietary and mucosal glycans in vitro, B. thetaiotaomicron generally prioritizes plant polysaccharides and upregulates host glycan-targeting PULs once the plant polysaccharides are depleted (39, 41). We measured growth of the wild-type strain in minimal medium containing a representative diet-derived plant polysaccharide (polygalacturonic acid; PGA) in combination with host-derived glycans (porcine mucin O-glycans; PMOG). The wild-type strain exhibits diauxic growth in this substrate combination as expected (39), with the dietary polysaccharide PGA consumed first; altering the PGA:PMOG ratio from 0.1%:0.8% to 0.2%:0.6% (wt/vol) increases growth during the first stage of exponential growth (Fig. 5B and C), and PGA-responsive PULs are upregulated during this stage (Fig. S4). Consistent with prior reports (39), PMOG-responsive PULs are upregulated during the second stage of exponential growth (Fig. S4).

These growth dynamics are profoundly altered in the (p)ppGpp⁰ strain, which grows readily during the first (PGA) exponential phase but does not exhibit a second exponential phase and instead fails to grow once PGA is likely exhausted from the culture medium (Fig. 5B and C; Fig. S5). To further verify that the growth defect of the (p)ppGpp⁰ strain under these conditions results from a lack of ability to grow on PMOG, we cultured each strain on each substrate separately (Fig. 5D). Indeed, the wild-type and (p)ppGpp⁰ strains grew similarly on PGA alone, while the latter exhibited a drastic growth defect in minimal medium containing PMOG as the sole carbon source (Fig. 5D). The wild-type and (p)ppGpp⁰ strains expressed many PGA- and PMOG-responsive PUL genes similarly in these in vitro conditions (one PGA-responsive and one PMOG-responsive PUL gene exhibited statistically significant differences in expression levels although both strains regulated these genes in the same direction), suggesting that the growth defect of the (p)ppGpp⁰ strain in PMOG is likely not due to transcriptional regulation of these genes (Fig. S4B).

We reasoned that if the (p)ppGpp⁰ strain cannot grow efficiently on mucin O-glycans, it would exhibit a colonization defect in the mucus layer in vivo. To test this, we colonized germfree mice with wild-type or (p)ppGpp⁰ B. thetaiotaomicron, provided food as described in Fig. 1A, and measured levels of lumen- and mucus-associated bacteria in the cecum at the fasted time point. While luminal colonization of the (p)ppGpp⁰ strain was ~twofold lower than the wild-type strain (consistent with Fig. 4A), mucosal colonization was ~fourfold lower in the (p)ppGpp⁰ strain compared to the wild-type strain (Fig. 5E), showing that this decrease in abundance during fasting is further exacerbated in the mucus layer.

Complex bacterial communities modulate ppGpp levels during a timed host feeding regimen

To determine if the coordination of ppGpp production with host eating also occurs in complex microbial communities, we placed conventional C57BL/6 mice on the feeding/fasting schedule described in Fig. 1A and measured cecal ppGpp concentrations at fasted and fed states as above. As observed in B. thetaiotaomicron monocolonized mice, these complex communities significantly elevated ppGpp levels in the fasted timepoint (Fig. 6A). Next, we repeated this experiment with gnotobiotic mice carrying human fecal microbiomes; as observed in conventional mice, these human gut communities similarly increased ppGpp levels at the fasted timepoint (Fig. 6B). Together, these results suggest that the ability of B. thetaiotaomicron to coordinate ppGpp production with host eating extends to other species and is conserved across mammalian gut microbiomes.

Fig 6.

Complex microbial communities increase ppGpp levels in fasting mice. (A) Cecal ppGpp levels during fed (blue) and fasted (orange) phases in conventional mice carrying a complete microbiome. (B) Cecal ppGpp levels during fed (blue) and fasted (orange) phases in gnotobiotic mice colonized with complete human fecal microbiomes. Significance is indicated with asterisks (*P < 0.05, Mann-Whitney U test).

DISCUSSION

In the gut, bacteria must compete for limited space and nutrients in a dynamic environment (3). In this report, we demonstrate that both mouse and human gut microbiomes respond to host feeding dynamics through the production of the intracellular second messenger (p)ppGpp. This response is recapitulated by the prominent gut commensal B. thetaiotaomicron, whose transcriptional response to short-term host fasting highlights genes associated with the stringent response. B. thetaiotaomicron requires (p)ppGpp to correctly regulate chromosomal replication, maintain viable cell density through the feeding-fasting cycle, and utilize host-associated glycans when the host is not eating.

Quantification of (p)ppGpp in complex communities and environments such as the gut is challenging due to the inherent instability of these compounds. Measurement of expression of (p)ppGpp-dependent genes (42) can provide an indirect indication of (p)ppGpp levels, although these genes are often regulated by multiple signals (43–45). Thin-layer chromatography (TLC) (46), (p)ppGpp-responsive biosensors (47), and mass spectrometry (30) provide direct approaches for (p)ppGpp quantification. However, the requirement for radiolabeled precursors (for TLC), genetic tractability and suitable reporters (for biosensors), or specialized equipment (for mass spectrometry) complicate the application of these strategies to host-associated environments or natural microbial communities. Here, we adapt an HPLC-based strategy (30, 31) that can be readily applied to individual species and complex communities in the mammalian gut (and potentially other) environments.

The specific signals that trigger ppGpp production during the fasting phase of a timed host feeding regimen in B. thetaiotaomicron and human and mouse gut microbial communities are unknown. Mice generally consume food at night and fast during the day, and it is possible that nutritive (fasting), visual (daytime), and/or circadian cues contribute to the (p)ppGpp response in the gut microbiome. Carbon starvation directly activates (p)ppGpp production in B. thetaiotaomicron and many other bacteria (16, 24, 48); fasting-dependent changes in host physiology, including altered hormones or gut motility, could also affect (p)ppGpp production (49). Furthermore, diurnal feeding establishes oscillations in antimicrobial peptide and MHC Class II expression in the small intestine and IgA secretion in the colon of mice (12, 50–52), which could also provide (p)ppGpp-inducing signals to gut commensal bacteria. Because food consumption is the primary driver of day/night differences in gut antimicrobial peptide production (12, 53) and microbiome community composition (8, 10), we expect that the major signals governing (p)ppGpp production also derive from food consumption. While a day-restricted feeding schedule could potentially be used to further explore these signals, this stressful condition induces arrhythmicity in gene expression and metabolic syndrome-like pathology in mice (54–58), potentially complicating the interpretation of (p)ppGpp measurements in the gut of day-fed mice. Given the global nature of (p)ppGpp signaling in bacteria, it is also likely that this response plays additional roles in the gut environment during both fed and fasting states.

Although (p)ppGpp induction alters gene expression in many bacteria (20–22, 24), wild-type B. thetaiotaomicron changes the expression of far fewer genes between the fed and fasted state (348) compared to the (p)ppGpp⁰ strain (659) (Fig. 1C vs Fig. 2B; Fig. S1). Additionally, these strains alter gene expression between fasted and fed states in different ways: of the 659 differentially expressed genes in the (p)ppGpp⁰ strain, only 101 (~15%) are similarly regulated in the wild-type strain (Fig. S1A). In E. coli, isoleucine starvation in vitro (which triggers (p)ppGpp production in this species) alters the expression of significantly fewer genes in a wild-type strain compared to a (p)ppGpp⁰ strain, with only 20% similarly regulated in both strains (20). In the gut environment, the increased number of differently expressed genes in the B. thetaiotaomicron (p)ppGpp⁰ strain likely reflects compensatory gene regulation in the absence of (p)ppGpp-mediated physiological adaptation. The (p)ppGpp⁰ strain cannot maintain wild-type colonization density during fasting (Fig. 4A), suggesting that these compensatory measures are insufficient to replace the function of (p)ppGpp signaling during host fasting. It should be noted that the wild-type and (p)ppGpp⁰ strains exhibit many transcriptional differences in both the fed and fasted states, and it is likely that (p)ppGpp signaling is required for competitive fitness in both conditions (24).

Notably, B. thetaiotaomicron upregulates gene sets involved in glycan and sphingolipid metabolism in a (p)ppGpp-dependent manner during host fasting (Fig. 1D and 2C). Many of these genes are localized in PULs, and many predicted mucin O-glycan targeting susC/susD pairs are upregulated in the fasted gut by wild-type but not (p)ppGpp⁰ B. thetaiotaomicron (Fig. 5A). In vitro, the (p)ppGpp⁰ strain induces expression of representative PMOG-responsive PULs during a shift from PGA to PMOG, yet fails to grow on PMOG (Fig. 5B and D). This suggests that expression of other genes could be required for growth on PMOG or that (p)ppGpp influences the expression of PMOG utilization machinery at the post-transcriptional level. Indeed, (p)ppGpp directly binds and modulates the function of multiple enzymes across diverse bacterial phyla (28, 59–62).

Given that the (p)ppGpp⁰ strain has the capacity to upregulate representative mucin O-glycan-responsive PULs when exposed to PMOG in vitro, why does this strain fail to induce these genes in vivo? One possibility is that the fraction of the B. thetaiotaomicron population exposed to the inducing signals in the fasted gut is lower in the (p)ppGpp⁰ strain compared to the wild type. Indeed, direct CFU measurements suggest that the mucus-associated population in B. thetaiotaomicron (p)ppGpp⁰-monoassociated mice is reduced by ~75% compared to the mucus-associated population in B. thetaiotaomicron wild-type-monoassociated mice (Fig. 5E). Alternatively, murine O-glycans could provide different (p)ppGpp-dependent signals compared to PMOG, although the wild-type strain induces predicted mucin O-glycan-responsive PULs both during the fasted state in mice and in response to PMOG in vitro (Fig. 5A; Fig. S4B). Taken together, these results suggest that B. thetaiotaomicron could use (p)ppGpp signaling to orchestrate the utilization of multiple carbon sources in the gut. E. coli (p)ppGpp⁰ strains exhibit limited metabolism of a variety of carbon sources in vitro (20), indicating that these connections could extend to other gut microbes.

Most gut bacteria possess enzymes to produce and degrade (p)ppGpp (16, 17). Some bacterial species, including B. thetaiotaomicron and E. coli, have a bifunctional synthase/hydrolase enzyme and a monofunctional synthase enzyme, the latter of which results from gene duplication and loss of function in the hydrolase domain (17, 24). Other bacteria, including many Firmicutes, encode a single bifunctional synthase/hydrolase with or without other small alarmone synthases, which contain only a (p)ppGpp synthase domain (17). The diverse repertoire of (p)ppGpp synthase and hydrolase enzymes across members of the gut microbiome suggests that the signals that regulate (p)ppGpp levels, and the functions under (p)ppGpp control, likely also vary across these taxa. For example, in E. coli, the monofunctional synthase (RelA) is involved in sensing amino acid starvation through direct association with the ribosome (63, 64), while the bifunctional synthase/hydrolase (SpoT) is involved in sensing fatty acid starvation through acyl carrier protein (65, 66) and carbon starvation through Ytfk (48, 67). However, the observation that mouse and human gut communities comprising hundreds of species coordinate ppGpp levels with a time-restricted feeding regimen in mice (Fig. 6A and B) suggests that this condition produces a widespread (p)ppGpp response. This could result from the coordinated activity of most members of these communities or from a subset that respond to these conditions via (p)ppGpp production. Mouse and human communities have limited species overlap (1), favoring the possibility that modulating (p)ppGpp levels under these conditions is a widespread response across community members that is conserved across the gut microbiomes of mammals.

Our results complement previous studies that have examined the role of (p)ppGpp in bacterial pathogens during infection. In these pathogens, (p)ppGpp⁰ mutants generally exhibit reduced virulence and antibiotic tolerance (19, 68–70). While our study focused on commensal members of the microbiome, the same (p)ppGpp-dependent coordination of bacterial physiology with host conditions could also play a role in pathogen colonization. Determining how microbes modulate their physiology in response to host feeding can potentially inform therapeutic approaches to various diseases. For example, insulin resistance (71, 72), drug response, and infection outcome (13, 73, 74) are each impacted by time of day and time of feeding and have also been linked to the gut microbiota. Thus, further understanding of how gut microbes coordinate their physiology with host feeding and other behaviors could reveal new factors that extend our understanding of the role of the microbiome in health and disease.

MATERIALS AND METHODS

Experimental model and subject details

Animal models

Germfree C57BL/6 mice (Taconic) were bred and maintained in the Goodman lab gnotobiotic facility on an autoclaved diet (Purina, #5K67). Mice were maintained in flexible plastic gnotobiotic isolators or sterile isocages (Allentown LLC) on a 12-h light/dark cycle. Animals were housed alone or in same-sex groups. Feeding schedules were ad libitum or restricted (described below). Germfree mice were colonized at 12–21 weeks of age by introduction of bacteria via gastric gavage. Within experiments, mice were mixed sex and age-matched within 2 weeks.

Restricted feeding

For restricted feeding experiments, mice were provided no access to food from 8:00 (denoted as ZT0) to 20:00 (denoted as ZT12) and ad libitum access to food from 20:00 to 8:00 (ZT12-ZT24). Room lighting was coordinated to the fasting period (lights on 8:00 to 20:00 [ZT0-ZT12]; lights off 20:00 to 8:00 [ZT12-ZT24]). At the initiation of each fasting period, bedding was replaced and enrichment removed to prevent consumption by mice; enrichment was returned at the initiation of each feeding period. Restricted feeding was maintained for 4 days before mice were sacrificed at ZT12 or ZT24, and ~100 mg aliquots of cecal material were collected into cryotubes and either immediately snap frozen in liquid nitrogen before storing at −80°C or used directly for CFU plating. Aerobic exposure was minimized during sample collection and was matched between ZT12 and ZT24 timepoints.

For B. thetaiotaomicron monoassociation experiments, germfree mice were colonized during the fasting phase of the first day of restricted feeding with 200 µL of stationary phase cultures (~2 × 10⁷ CFU). For colonization with human fecal samples, germfree mice were colonized 7 days before the first day of restricted feeding with 100 µL of cryopreserved human fecal samples from three separate donors (75). The same number of mice colonized with each donor microbiota was measured at each timepoint. For both humanized and conventional C57BL/6 mice (Taconic Biosciences), mice were acclimatized to isocages for 7 days before the first day of restricted feeding.

All mouse experiments were performed using protocols approved by the Yale University Institutional Animal Care and Use Committee.

Bacterial culture conditions

Bacteroides thetaiotaomicron VPI-5482 tdk (76), designated as the WT strain, and an isogenic ppGpp⁰ strain (either previously constructed (24) or independently generated as described below) were cultured anaerobically in glucose minimal medium [0.5% glucose, 100 mM KH₂PO₄ (pH 7.2), 15 mM NaCl, 8.5 mM (NH4)₂SO₄, 0.5 g L⁻¹ cysteine, 0.2 mM histidine, 1.9 µM hematin, 100 mM MgCl₂, 1.4 mM FeSO4, 50 mM CaCl₂, 1 mg mL⁻¹ vitamin K3, and 5 ng mL⁻¹ vitamin B₁₂] (14), liquid TYG medium [1% tryptone peptone (IBI Scientific), 0.5% yeast extract (Difco), 0.2% glucose, 100 mM potassium phosphate buffer pH 7.2, 0.08 mM MgSO₄*7H₂O, 5 mM NHCO₃, 1.5 mM NaCl, 0.2 mM histidine, 1.9 µM hematin, 1.4 mM FeSO₄, 50 mM CaCl₂, 1 mg mL⁻¹ vitamin K3], or on brain heart infusion (BHI) (Becton Dickinson) agar with 10% horse blood (Quad Five). Anaerobic procedures were performed in an anaerobic chamber (Coy Laboratory Products) containing 20% CO₂, 10% H₂, and 70% N₂.

Generation of (p)ppGpp⁰ strains

Four B. thetaiotaomicron (p)ppGpp⁰ strains (with nonpolar deletions of BT0700 and BT3998) were independently constructed as described (24). Briefly, pExchange_0700 (24) was transferred into B. thetaiotaomicron tdk by conjugation, and second recombination events selected by plating on BHI blood agar containing 200 µg/mL 5-fluoro-2′-deoxyuridine (FUdR). After PCR verification, pExchange_3998 was used to delete BT3998 as above.

Method details

RNA sample preparation

For in vivo RNA-seq measurements, aliquots of ~100 mg of frozen cecal samples were thawed in 10 mL of a 2:1 Bacterial RNAProtect (Qiagen):RNase-free deionized water solution, homogenized by vortexing, incubated at room temperature (RT) for 5 min, vortexed again, and pelleted by centrifugation for 10 min at 7,000g at 4°C. RNA was then extracted using a PowerMicrobiome kit (Qiagen) with on-column DNase treatment. After elution from the PowerMicrobiome kit, RNA was treated with TURBO DNase (ThermoFisher) in a total volume of 100 µL at room temperature for 30 min. RNA was then purified using an RNeasy kit (Qiagen) with an on-column DNase treatment (Qiagen) before being frozen at −80°C for use in RNA-seq. RNA quality was assessed using 2100 Bioanalyzer (Agilent).

For in vitro quantitative reverse transcription (qRT-PCR) preparation, a culture corresponding to ~5 × 10⁸ cells, or 1 mL at OD₆₀₀ = 1, was centrifuged at 7,000 G for 5 min, decanted aerobically, quickly resuspended in 5 mL of Bacterial RNAProtect (Qiagen), and centrifuged at 7,000g for 5 min. Cultures were then decanted, frozen on dry ice, and transferred to −80°C until RNA extraction. Cell pellets were then thawed on ice, resuspended in 200 µL lysis buffer (10 mM Tris, 1 mM ETDA, 5 mg mL⁻¹ lysozyme [Sigma], 12.5 µg mL⁻¹ proteinase K [pH 8.0]), and incubated RT for 10 min. Crude RNA was extracted using an RNeasy Mini kit (Qiagen), treated with TURBO DNase (ThermoFisher) in a total volume of 100 µL at room temperature for 30 min, and processed again through an RNeasy Mini kit. Quality was assessed via gel electrophoresis before cDNA synthesis using Superscript II Reverse Trancriptase (Invitrogen) according to manufacturer’s instructions.

Preparation and sequencing of mRNA libraries

Library preparation and RNA sequencing were performed at the Yale Center for Genome Analysis. rRNA was depleted using NEBNext rRNA Depletion Kit for bacteria (NEB), and RNAseq libraries were prepared using NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (NEB), as recommended by the manufacturer. Sequencing was performed on a NovaSeq6000 S4 flow cell with 1 × 75 bp single-end sequencing to generate ~25–35 million reads per sample.

RNA-seq analysis

Sequences were aligned to the B. thetaiotaomicron VPI-5482 reference genome (NC_004663) using Bowtie2 (77). Reads were counted using featureCounts (78), and differential expression analysis was carried out using DESeq2 (79). PCA plots were generated with DEseq2 data and ggplot (80). Gene set enrichment analysis was performed with GSEA (25), and volcano plots were generated with EnhancedVolcano (81). RNA-seq data are available in NCBI as BioProject PRJNA935381 (see also Table S1).

ppGpp extraction and measurement

ppGpp was extracted as described (30) with modifications. To this end, aliquots of ~100 mg of frozen cecal samples were transferred from −80°C storage to dry ice. Samples were extracted individually by the addition of 1 mL of ice-cold 2 M formic acid along with ~250 mg of acid-washed beads (BioSpec). Samples were then transferred to ice to thaw for 5 min before disruption by bead beating (Biospec) for 2 × 20 s with 1 min on ice between each 20-s treatment. Homogenized samples were then transferred to 15-mL conical tubes (Corning), and 2 mL of ice-cold 2 M formic acid was added to each sample. After 30 min incubation on ice, 50 mM NH₄OAc (pH 4.5) was added to each sample to bring the total volume to 6 mL. One sample per tissue group/time was split into two 3-mL aliquots in 15-mL conical tubes: 20 µL of 200 µm ppGpp or pppGpp standard (Jena Biosciences) was added to one of the aliquots to estimate extraction efficiency. Samples were then centrifuged at 2,400g for 5 min at 4°C or until the material was firmly pelleted at the bottom of the conical tube. Samples were then placed on ice prior to extraction using an Oasis WAX 1 cc Vac Cartridge, 30 mg Sorbent per Cartridge, 60 µm (Waters).

The Oasis WAX 1 cc Vac Cartridge was pre-treated with 1 mL of MeOH followed by 1 mL of 50 mM NH₄OAc (pH 4.5), and the supernatant of the sample solution was loaded 1 mL at a time onto the column; the column was washed with 1 mL of 50 mM NH₄OAc (pH 4.5), followed by 1 mL of MeOH. The nucleotide pool was eluted from the cartridge with 1 mL MeOH/H₂O/NH₄OH (20:70:10) solution into a 15-mL conical tube (Corning) and immediately frozen on dry ice before storing at −80°C for at least 2 h before preparing samples for lyophilization.

To lyophilize the extracted samples, tubes were transferred to dry ice, and caps were removed and replaced with a Kimwipe (KimTech) secured with a rubber band around the neck of the tube. Samples were returned to −80°C for 30 min to ensure the samples were frozen before lyophilization. Frozen samples were quickly transferred on dry ice to a lyophilizer (Labconco), and dry ice was packed around the glass holder to ensure samples remained frozen while reaching pressure. Lyophilized samples were stored at −80°C.

For measurement, lyophilized samples were dissolved in 100 µL of nuclease-free water, filtered through a Qiaquick spin column (Qiagen) by centrifugation for 1 min at 13,000g, and immediately used for ppGpp quantification. After measurement, samples were stored in HPLC vials at −80°C and retained similar ppGpp measurements for at least 3 months.

Chromatography was performed as described (31) with modifications. Specifically, 20 µL of extracted samples and standards were analyzed with an Agilent 1200 Infinity Series HPLC System (equipped with a Sphereclone SAX column, 4.6 × 150 mm, 5 µm) via DAD detection. The mobile phase employed a 2 mL/min isocratic delivery of 0.36 M NH₄H₂PO₄ (pH 3.4, 2.5% [vol/vol] acetonitrile) up to 1 h. The detection wavelength was set to 252 nm. The mobile phase was prepared with analytical grade NH₄H₂PO₄ (>99% pure), H₂O filtered through a Milli-Q system, and HPLC-grade acetonitrile. The pH of the solution was adjusted to 3.4 through the addition of H₃PO₄ (85% aq. solution). Output data were converted into concentrations using a standard curve of ppGpp (Jena Biosciences).

CFU calculations and measurements

B. thetaiotaomicron abundance was monitored by CFU counting on BHI-HK plates under anaerobic conditions. For groups of two, significant differences were determined by Mann-Whitney U test with Benjamini-Hochberg correction; for groups greater than two, significant differences were determined by Kruskall-Wallis followed by Dunn’s multiple correction.

gDNA extraction

For gDNA extraction, ~100 mg of frozen cecal material was added to 500 µL of buffer PB (Qiagen), ~250 µL of acid-washed silica beads, 250 µL of 20% SDS, and 550 µL of phenol:chloroform pH 8.0 premixed with isoamyl alcohol (25:24:1) (VWR) in a 2-mL cryotube at room temperature. Samples were disrupted for 2 min in a Bead Beater (BioSpec), rested on ice for 1 min, and then disrupted again for 2 min. Samples were centrifuged at 8,000g at 4°C for 5 min, and the aqueous layer was extracted and placed on a Qiaquick spin column, centrifuged for 1 min at 13,000g, and then washed twice with 750 µL of PE buffer (Qiagen) before eluting in 100 µL of nuclease-free water (Qiagen).

Whole-genome sequencing library preparation and analysis

Libraries were prepared with the IDT Lotus kit (Integrated DNA Technologies), followed by 6 cycles of amplification, and sequenced on NovaSeq6000 S4 flow cell at 40× coverage with ~15 million reads per genome. Peak-to-trough analysis was performed on WGS sequencing data using GRiD (82) package in R, using annotation from the NCBI taxonomy RefID 226186.

PUL annotation and analyses

PUL numbers in Fig. 5A; Fig. S4B correspond to “literature-derived PULs” from the PULDB database (40). To analyze PUL gene regulation, susC/susD gene pairs were identified using PULDB (40) and changes in expression were calculated as the average log2(fold change) between fed and fasted phases.

B. thetaiotaomicron growth measurements

B. thetaiotaomicron cultures were cultured for 16–20 h in minimal medium with glucose (5 mg/mL), washed twice with 2× minimal medium with no carbon source, back-diluted 1:1,000 (single carbon source experiments) or 1:200 (diauxic shift experiments) in 2× minimal medium without glucose, and aliquoted into 96-well plates (Corning). PMOG was generously provided by E. Martens and N. Pudlo (University of Michigan) and extracted from porcine mucosal glycans (Sigma Aldrich) as described (39). PMOG and PGA (Sigma Aldrich) were used at varying concentrations as indicated. Growth curves were carried out in a microplate reader (BioTek) set at 37°C and housed in an anaerobic chamber. OD₆₀₀ measurements were collected every 15 min with shaking before each measurement. Growth measurements were conducted using both the previously constructed strain (24) and independently generated (p)ppGpp⁰ strains, with similar results.

Quantitative reverse transcription PCR (qRT-PCR)

qRT-PCR was performed as previously described (24), using a CFX96 Realtime system (Biorad) and SYBR FAST universal qPCR master mix (KAPA Biosystems). Relative changes were calculated with ∆∆Cq method, normalizing to the xylanase gene of B. thetaiotaomicron. Primers are described in Table S2.

16S copy quantification

Femto Bacterial DNA Quantification Kit (Zymo) was used for absolute quantification of 16S gene copies in cecal samples. DNA was extracted from cecal samples as described above, and quantification of 16S gene copies was performed according to the manufacturer’s instructions (Zymo). qPCRs were prepared in 96-well plate format and performed on a CFX96 Realtime system (Biorad) with the following conditions: 10 min at 95°C, followed by 40 cycles of 30 s at 95°C, 30 s at 50°C, and 1 min at 72°C, then final extension of 7 min at 72°C. 16S concentrations were determined using a calibration curve of known concentrations of E. coli DNA.

Quantification and statistical analysis

Statistical analysis of results was performed using Prism 9 (GraphPad). All graphs are expressed as means ± SEM, and P-values less than 0.05 were regarded as statistically significant. The statistical difference between any two groups was determined Mann-Whitney U test or unpaired t-test. For multiple groups, the statistical difference was determined using either Kruskal-Wallace test with Dunn correction or the one-way analysis of variance (ANOVA) with post-hoc Tukey test, depending on assumptions of variance.

DATA AVAILABILITY

RNA-seq data have been deposited in NCBI as BioProject PRJNA935381. This paper does not report original code. Any additional information required to reanalyze the data reported in this paper is available from the corresponding author upon request.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/mbio.01907-23.

Fig. S1. mbio.01907-23-s0001.eps.

Limited overlap between gene expression changes in B. thetaiotaomicron WT and (p)ppGpp⁰ strains during fed and fasted phases in mice.

Fig. S2. mbio.01907-23-s0002.eps.

Representative chromatograms used in ppGpp quantification.

Fig. S3. mbio.01907-23-s0003.eps.

B. thetaiotaomicron GRiD values in varying growth conditions in vitro.

Fig. S4. mbio.01907-23-s0004.eps.

Transcriptional responses of B. thetaiotaomicron WT and (p)ppGpp⁰ strains during diauxic growth in PGA and PMOG.

Fig. S5. mbio.01907-23-s0005.eps.

Doubling time of independently constructed B. thetaiotaomicron (p)ppGpp⁰ strains in minimal medium containing PGA and PMOG (0.2%:0.6%).

Supplemental Legends. mbio.01907-23-s0006.docx.

Legends for supplemental figures and tables.

Table S1. mbio.01907-23-s0007.xlsx.

RPKM values of (p)ppGpp⁰ and WT strains isolated from cecal contents of fed and fasted mice.

Table S2. mbio.01907-23-s0008.xlsx.

Primers used for qRT-PCR.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.