Ruminococcus gnavus and Limosilactobacillus reuteri Regulate Reg3γ Expression through Multiple Pathways

Animals have evolved numerous defense mechanisms to protect themselves against infectious diseases (1), with expression of antimicrobial peptides (AMPs) representing an evolutionarily conserved immune defense mechanism present in nearly all multicellular organisms (2). These AMPs have bactericidal activity that leads to the rapid death of microbes. Most AMPs are produced by the mucosal epithelium (e.g., intestines, respiratory tract, reproductive tract) to help maintain homeostasis with the commensal microbiota and limit infection by pathogens (3). Given that the intestines represent a major route for pathogen entry and, compared with other anatomic regions, harbor the largest number of commensal microbes, it is not surprising that intestinal epithelial cells (IECs) secrete a broad range of AMPs into the lumen to maintain host–microbiota segregation and protect against enteric infection (4–6).

One of the most widely studied intestinal AMPs is regenerating islet-derived protein 3 γ (Reg3γ), which is produced by multiple intestinal epithelial lineages, including enterocytes and Paneth cells (7). Reg3γ is stored in secretory granules, has bactericidal activity against gram-positive bacteria, and is critical for maintaining spatial segregation between the intestinal epithelium and the microbiota (6, 8, 9). Regulation of Reg3γ expression is complex and involves the integration by IECs of at least two different signals. First, IEC expression of myeloid differentiation primary response protein 88 (Myd88), an adaptor protein downstream of many TLRs, is required (10–12), a finding that suggests TLR signaling is critical. Notably, expression of Myd88 specifically by Paneth cells is sufficient for Reg3γ expression (10), which demonstrates the critical role for this cell type. Second, group 3 innate lymphoid cells (ILC3s) present in the lamina propria must secrete IL-22 (13, 14), which binds to IL-22Rs on the basolateral side of IECs and results in Reg3g expression (15). It is not clear why both Myd88 and IL-22 signaling are required or whether the same environmental signal stimulates both pathways.

Although some AMPs are constitutively expressed in the absence of microbial stimulation, others, including Reg3γ, require microbiota-derived signals for normal levels of expression (8, 10, 11, 16). However, a screen of >50 taxonomically diverse bacteria found that none was able to induce small-intestinal expression of Reg3g in mice (16), which demonstrates this critical function is possessed by a limited number of commensal bacteria. Using microbe–phenotype triangulation, we recently identified Ruminococcus gnavus and Limosilactobacillus reuteri [formerly Lactobacillus reuteri (17)] as potent inducers of Reg3g expression (18). We reasoned these specific Reg3γ-inducing bacteria can be exploited to gain insight into the mechanisms underlying microbiome induction of Reg3g expression. We find R. gnavus and L. reuteri each requires Myd88 and ILC3s for inducing expression of Reg3g. Moreover, these commensal bacteria induce expression of several other intestinal AMPs, with notable differences in their requirement for ILC3s. Intriguingly, R. gnavus and L. reuteri suppress AMP expression in the absence of either Myd88 or ILC3s. Taken together, our results demonstrate that microbial regulation of AMP expression is more complex than previously appreciated.

C57BL/6J (stock number 000664), Ahr^fl/fl (stock number 006203), and Rorc-Cre (stock number 022791) mice were purchased from Jackson laboratories, and Myd88^−/− mice were obtained from A. Moseman (Duke University). Gnotobiotic HMb mice were a gift from D. Kasper (Harvard Medical School) (19). All animals were bred and maintained at Duke University, with gnotobiotic mice maintained in sterile vinyl isolators (Class Biologically Clean, Madison, WI). HMb mice were experimentally manipulated in autoclaved individually ventilated cages with autoclaved food and water. Experiments used sex- and aged-matched (ranging from 5- to 12-wk-old) mice. All procedures were approved by the Duke Institutional Animal Care and Use Committee and were conducted in accordance with National Institutes of Health guidelines.

Mice were administered 10⁷–10⁸ CFUs of Parabacteroides distasonis (ATCC 8503), R. gnavus (ATCC 29149), or L. reuteri (BEI HM-102) by oral gavage. R. gnavus– and L. reuteri–treated mice were euthanized 3 and 5 d after administration of bacteria, respectively, unless specified otherwise.

To determine the colonization kinetics of R. gnavus and L. reuteri, we collected fecal samples at 0, 4, 8, 24, 72, and 120 h after bacterial administration. In addition, we harvested the distal 1.5 cm of ileum from mice treated with either R. gnavus or L. reuteri at 72 or 120 h after bacterial administration, respectively. Total DNA was extracted by homogenizing samples with TRIzol (Invitrogen) and purifying with Quick-DNA MiniPrep kit (Zymo) spin columns. Using organism-specific primers for R. gnavus (forward [F], 5′-CCAATTACGGAAAGCTGGAT-3′; reverse [R], 5′-TCTGCTTTCCATGTATCTTCACA-3′) or L. reuteri (F, 5′-CAGACAATCTTTGATTGTTTAG-3′; R, 5′-GCTTGTTGGTTTGGGCTCTTC-3′), we determined copy numbers for each bacterium by quantitative PCR (qPCR; StepOnePlus; Bio-Rad) and normalized them to the entire microbial abundance using 16S rRNA universal primers 341F (5′-CCTACGGGAGGCAGCAG-3′) and 534R (5′-ATTACCGCGGCTGCTGGCA-3′).

qPCR for AMPs was performed as previously described (18). In brief, the distal 1.5 cm of ileum was collected, frozen immediately in liquid nitrogen, and stored at −80°C until needed. Tissues were homogenized in TRIzol (Invitrogen), and RNA was purified according to the manufacturer’s instructions, with a subsequent additional cleanup step (RNeasy Mini kit; Qiagen). cDNA was prepared with the High-Capacity cDNA Reverse Transcription kit (Applied Biosystems), and qPCR was performed on a StepOnePlus (Bio-Rad) using iTaq Universal SYBR Green Supermix (Bio-Rad). The comparative cycle threshold method was used to quantify transcripts that were normalized with respect to GAPDH. Primer sequences are as follows: GAPDH (F, 5′-CCTCGTCCCGTAGACAAAATG-3′; R, 5′-TCTCCACTTTGCCACTGCAA-3′), Reg3g (F, 5′-TTCCTGTCCTCCATGATCAAAA-3′; R, 5′-CATCCACCTCTGTTGGGTTCA-3′), Reg3b (F, 5′-TACTGCCTTAGACCGTGCTTTCTG-3′; R, 5′-GACATAGGGCAACTTCACCTCACA-3′), Defa5 (F, 5′-TTGGGCTCCTGCTCAACAAT-3′; R, 5′-GACACAGCCTGGTCCTCTTC-3′), and Lyz1 (F, 5′-GAGACCGAAGCACCGACTATG-3′; R, 5′-CGGTTTTGACATTGTGTTCGC-3′).

Ileal tissue was fixed in 10% formalin, embedded in paraffin, and 4-µm-thick sections were stained with H&E. Paneth cells, identified as crypt cells containing eosinophilic granules, were enumerated in 10 crypts per sample. Immunohistochemistry was performed as previously described (20). In brief, 4-µm-thick tissue sections were deparaffinized in xylene followed by alcohol rehydration. Ag retrieval was performed in 1× Borg Decloaker solution (Biocare Medical) at 95°C for 3 min in a pressure cooker (Instant Pot). Ab staining was performed with anti-mouse Reg3γ (1:100 dilution; Invitrogen) followed by a biotin-conjugated secondary (1:500 dilution; Jackson Immunoresearch), and the protein was visualized with Vectastain Elite ABC peroxidase detection kit (Vector Laboratories) followed by incubation with liquid DAB+Substrate (Dako). Image acquisition was performed using a Zeiss Axio Imager microscope, and images were processed using Fiji (ImageJ) software.

Data are represented as mean ± SE (SEM) throughout the figures. Prism 9 (GraphPad Software) was used for all statistical analyses. qPCR analyses were compared using a Mann–Whitney U test, one-way ANOVA, or Kruskal–Wallis test, as appropriate. Bacterial 16S rRNA gene copy numbers were compared using a Wilcoxon test. All data were considered statistically significant for p < 0.05.

We previously demonstrated that a single oral dose of R. gnavus or L. reuteri induces ileal Reg3g expression in gnotobiotic Swiss Webster mice colonized with a human microbiota (18). However, it is not clear how quickly, or for how long, Reg3g induction occurs after treatment with these Reg3γ-inducing organisms. Therefore, we treated SPF C57BL/6 mice with a single oral dose of R. gnavus or L. reuteri and harvested ileal tissue at different time points after treatment. Although R. gnavus induced Reg3g expression 3 and 5 d after treatment, this increased expression was no longer present at 7 d (Fig. 1A). In contrast, induction of Reg3g expression by L. reuteri displayed slower kinetics, with an increase present only on day 5 (Fig. 1B). Given that exogenously administered bacteria often do not persist in the intestinal tract (21–23), we determined the colonization kinetics of R. gnavus and L. reuteri by analyzing the fecal abundance of these bacteria at various time points. Surprisingly, even though these organisms induce Reg3g expression at 3–5 d after treatment, we found both organisms were absent from the fecal microbiota by 24 h after treatment (Fig. 1C, 1D). Moreover, neither R. gnavus nor L. reuteri was detected in stool or the distal ileum at time points when Reg3g expression was first noted to be increased (i.e., day 3 for R. gnavus, day 5 for L. reuteri). Taken together, these results indicate that persistent intestinal colonization of R. gnavus or L. reuteri is not required for Reg3g induction.

To ensure this change in Reg3g expression results in altered protein levels, we performed immunohistochemistry for Reg3γ in ileal samples obtained from mice treated with R. gnavus or L. reuteri. Compared with control animals, treatment with either R. gnavus or L. reuteri resulted in greater Reg3γ protein levels at the crypt base and along the villi (Fig. 2), which reflects Reg3γ in both Paneth cells and enterocytes. These results demonstrate that induction of Reg3g expression by R. gnavus and L. reuteri results in increased levels of protein.

This increase in Reg3γ protein in the crypt base could reflect either an increase in the number of Paneth cells or increased Reg3γ production within each Paneth cell. To discriminate between these possibilities, we enumerated Paneth cells in H&E-stained ileal sections and found no change in Paneth cell numbers in mice treated with either R. gnavus or L. reuteri (Fig. 3). These results demonstrate that the increase in Reg3γ is not a consequence of increased Paneth cell numbers but is more likely caused by increased Reg3γ production within each Paneth cell.

Previous studies have revealed that microbiota-induced expression of Reg3γ requires Myd88 and secretion of IL-22 by ILC3s (6, 13). These prior studies largely focused on the microbiota as a whole; however, given the pleiotropy of the microbiota, it is possible that specific microbes differ in their mechanism of Reg3γ induction. As such, we investigated whether Myd88 and/or ILC3s are required for induction of Reg3g by R. gnavus and L. reuteri. Neither R. gnavus nor L. reuteri was able to induce Reg3g expression in Myd88^−/− mice (Fig. 4A), a finding that indicates Myd88 is required for induction of Reg3g by these organisms. Surprisingly, Myd88^−/− mice treated with either bacterium had suppressed levels of Reg3g expression, which indicates that, in the absence of Myd88 signaling, these bacteria inhibit Reg3g expression.

To determine the role of ILC3s in R. gnavus and L. reuteri stimulation of Reg3g expression, we generated ILC3-deficient mice (ILC3 knockout [KO]) by crossing Ahr^fl/fl mice to Rorc-cre mice as previously described (24). Consistent with previous work demonstrating the role of ILC3s in Reg3g expression (13), ILC3 KO mice have reduced Reg3g expression compared with Ahr^fl/fl littermate controls (Fig. 4B). Neither R. gnavus nor L. reuteri induced Reg3g expression levels in ILC3 KO mice (Fig. 4B), which demonstrates that ILC3s are critical for induction of Reg3g by these organisms. Similar to our results with Myd88^−/− mice, we found R. gnavus and L. reuteri suppressed Reg3g expression in ILC3 KO animals (Fig. 4B). Taken together, our results establish that Myd88 and ILC3s are both required for Reg3g induction by R. gnavus and L. reuteri. In addition, these bacteria suppress Reg3g expression in an Myd88- and ILC3-independent manner.

The intestinal epithelium secretes several AMPs in addition to Reg3γ (7, 25). Because R. gnavus and L. reuteri induce ileal expression of Reg3γ, we investigated whether these organisms also induce expression of other AMPs. We treated mice with either Parabacteroides distasonis (an irrelevant bacterial control), R. gnavus, or L. reuteri and examined ileal expression of Reg3b, Defa5 (α-defensin 5), and Lyz1 (lysozyme). Although both R. gnavus and L. reuteri significantly induce Reg3b expression, only R. gnavus increased expression of Defa5 and Lyz1 (Fig. 5). These results demonstrate that R. gnavus and L. reuteri differ in their ability to induce diverse AMPs.

Given that AMPs differ in their regulation (7, 11), we investigated whether Myd88 and ILC3s are required for R. gnavus– or L. reuteri–mediated induction of these additional AMPs. We quantified ileal expression of Reg3b, Defa5, and Lyz1 in Myd88^−/− and ILC3 KO mice treated with either R. gnavus or L. reuteri (Fig. 5B). R. gnavus and L. reuteri are unable to induce expression of these additional AMPs in Myd88^−/− mice (Fig. 5), a finding that establishes Myd88 is required in this process. ILC3s are required for normal expression of Reg3b and Lyz1 as evidenced by lower expression in ILC3 KO mice compared with Ahr^fl/fl littermate controls; neither R. gnavus nor L. reuteri induces these AMPs in ILC3 KO mice, which establishes that these commensal bacteria induce Reg3b and Lyz1 expression in an ILC3-dependent manner. Interestingly, we observed no difference in Defa5 expression between Ahr^fl/fl and ILC3 KO mice, a finding that indicates ILC3s are not required for Defa5 expression (Fig. 5C); however, R. gnavus treatment no longer induces Defa5 expression in the ILC3 KO mice, establishing that this bacterium that normally induces expression of Defa5 requires ILC3s to do so. Similar to our findings with Reg3γ, R. gnavus and L. reuteri suppressed expression of Reg3b and Lyz1 in a Myd88- and ILC3-independent manner, whereas suppression of Defa5 was observed only in Myd88^−/− mice and not ILC3 KO mice. Interestingly, even though L. reuteri was unable to induce Defa5 or Lyz1 expression in wild-type mice, it can suppress expression of these genes in some genetic backgrounds. Taken together, our findings highlight differences between the AMP-inducing activities of R. gnavus and L. reuteri and demonstrate the complexity underlying microbial regulation of these genes, the specifics of which differ between AMPs.

AMPs play a critical role in protecting the host against infection, and their proper regulation is paramount. Earlier studies demonstrated that induction of Reg3γ, a canonical AMP, requires microbial signals to activate Myd88 and ILC3s (8, 10, 13); however, these studies compared conventional mice with a complex microbiota with germ-free mice. As such, it is possible that some bacteria use a Myd88 pathway, whereas others exploit an ILC3-dependent pathway to induce Reg3γ. In this study, we used R. gnavus and L. reuteri as archetypal Reg3γ-inducing commensal bacteria to clarify the mechanism(s) of Reg3γ induction by specific bacteria. We find neither bacterium can induce Reg3g expression in the absence of either Myd88 or ILC3s, findings that demonstrate individual bacteria require both pathways for AMP expression.

For these studies, we used mice that lacked Myd88 in all cells. Although Myd88 expression by Paneth cells is sufficient for Reg3γ expression (10), we do not know whether R. gnavus and L. reuteri are interacting with Myd88 on Paneth cells, other enterocytes, or immune cells (e.g., dendritic cells) in the lamina propria. We speculate these bacteria signal through Myd88 present on non-Paneth cells (potentially in addition to Paneth cell–associated Myd88) because of the increased levels of Reg3γ we observed in the villi and not just the crypts where Paneth cells are localized. We observed a difference between the induction of Reg3g in the villus, with R. gnavus having an effect in the midvillus and L. reuteri induction occurring at the villus tips. However, this distinction likely reflects the fact that these images were taken at different days after bacterial administration (day 3 for R. gnavus and day 5 for L. reuteri), which corresponds with the idea that enterocytes take ∼5 d to complete their migration to the top of the villus (26). Our data suggest that the induction of Reg3g persists for this entire time and overrides the normal zonation program present in the small intestine (27). Given that all our bacterial treatments were done in the context of mice with a complex microbiota, it is possible the requisite Myd88 or ILC3 signal comes from the endogenous microbiota, with R. gnavus or L. reuteri providing the second signal. Discriminating between these differences may require identifying the relevant bacterial factor that induces Reg3γ and determining whether it signals through Myd88, ILC3s, or both.

Interestingly, in the absence of either Myd88 or ILC3s, treatment of mice with either R. gnavus or L. reuteri led to suppression of Reg3g expression. This finding indicates that these bacteria target at least two different pathways that regulate Reg3g expression in discordant ways. Future work will determine whether this decreased expression results in a decreased level of protein. Given that R. gnavus and L. reuteri also lead to suppression of other AMPs in a Myd88- and ILC3-independent manner, it is likely the same pathway regulates multiple AMPs. One possibility is that R. gnavus and L. reuteri inhibit expression of peroxisome proliferative-activated receptor α, which is known to be negatively regulated by commensal bacteria and increase Reg3g expression (28, 29). Comparing transcriptional profiles in wild-type, Myd88^−/−, and ILC3-deficient animals may help identify this inhibitory pathway.

By comparing the colonization kinetics of R. gnavus and L. reuteri with the kinetics of Reg3g induction, we found that colonization was not required for activity. This delay between bacterial exposure and increased Reg3γ levels indicates that the process of Reg3g induction is relatively slow. Although some work has suggested that commensal organisms must colonize the host to be effective (30, 31), others have similarly shown that transient exposure to bacteria and/or bacterial products can have long-lasting effects (32, 33). With Reg3γ specifically, it has been shown that TLR ligands are sufficient for induction (12). It is likely that the bioactive molecule from R. gnavus and L. reuteri is either present on their surface or secreted into the culture media such that the long-term colonization by the organisms is not required.

In summary, we have detailed the mechanism by which two specific commensal bacteria induce Reg3g expression in the ileum. Moreover, our studies revealed that regulation of Reg3g expression by the microbiota is more complex than previously thought and involves both stimulatory and inhibitory pathways. These findings may aid the development of microbiome-derived treatments that protect against infection by increasing AMP expression.

The authors have no financial conflicts of interest.