Main

A subset of small intestinal myeloid cells, characterized by high expression of CX3CR1 (that is, CX3CR1+ cells), regulates intestinal homeostasis1,2,3,4,5. CX3CR1+ cells and other types of intestinal myeloid cells extrude their dendrites into the lumen to take up antigens1,2,3,4,6,7,8. This dendrite protrusion has previously been shown to be impaired in the absence of the CX3CL1–CX3CR1 axis or epithelial TLR signalling6,7,9. However, because CX3CL1 is produced by the epithelial cells, it remains unclear how dendrites of CX3CR1+ cells extrude into the lumen. Dendrite protrusion was severely decreased in mice that were orally treated with a cocktail of antibiotics7 (Extended Data Fig. 1a). Therefore, we speculated that—in addition to TLR ligands—bacterial metabolites produced in the lumen mediate dendrite protrusion, and thus prepared three fractions (diethyl ether-, chloroform- and methanol-soluble fractions) from the luminal contents of the small intestine. Small intestinal CX3CR1+ cells were treated with these fractions (Fig. 1a). The methanol-soluble fraction induced dendrite extension of CX3CR1+ cells. We then prepared methanol-soluble fractions from specific-pathogen-free (SPF) and germ-free mice and used them for stimulation of CX3CR1+ cells (Fig. 1b). The methanol-soluble fraction prepared from germ-free mice did not induce dendrite extension. CX3CR1+ cells isolated from Myd88−/− mice extended their dendrites in response to the methanol-soluble fraction of SPF mice (Extended Data Fig. 1b). These findings indicate that bacteria-dependent luminal products mediate a TLR-independent dendrite protrusion of intestinal CX3CR1+ cells.

Fig. 1: Extension of CX3CR1+ cell dendrites induced by intestinal contents.
figure 1

a, Intestinal CX3CR1+ cells of Cx3cr1gfp/+ mice were incubated with indicated fractions of luminal contents of SPF mice (n = 6 independent samples per group). b, CX3CR1+ cells were treated with the methanol-soluble fraction of SPF mice or germ-free (GF) mice (n = 6 independent samples per group). The morphology of CX3CR1+ cells was automatically analysed, and the proportion of CX3CR1+ cells with extended dendrites was determined. Each symbol represents the percentage of GFP-positive cells in a single well that have dendrites that are over 5 μm in length, from two biologically independent experiments. Data represent the mean ± s.d. Two-tailed Student’s t-test (a) and one-way ANOVA with Tukey’s comparison (b) were performed for statistical analysis. NS, not significant. Scale bar, 10 μm.

Source data

Full size image

To reveal the mechanisms by which luminal products mediate dendrite protrusion, we tried to identify the host molecule that is responsible for the response to the luminal factors. G-protein-coupled receptors (GPCRs) have previously been shown to respond to several bacterial metabolites, such as short-chain fatty acids10,11,12,13,14. Therefore, we screened the expression of GPCRs in subsets of intestinal myeloid cells, and found that genes for several GPCRs were highly expressed in CX3CR1+ cells (Extended Data Fig. 2a, b). In particular, Gpr31 was most preferentially expressed in CX3CR1+ cells. Among several immune-cell populations, Gpr31 was selectively expressed in CX3CR1+ cells of Peyer’s patches and the small intestinal lamina propria (Extended Data Fig. 2c–j). GPR31 has previously been shown to act as a receptor for 12-(S)-hydroxyeicosatetraenoic acid15; however, an in vivo function of GPR31 remains unclear. Therefore, we analysed mice that lack GPR31 (Gpr31b−/− mice) (Extended Data Fig. 3a–c). The number of intestinal CX3CR1+ cells was unaltered in these mice (Extended Data Fig. 3d); however, dendrite protrusion of CX3CR1+ cells in the small intestine was severely reduced in Gpr31b−/− mice (Fig. 2a, Extended Data Fig. 3e, Supplementary Videos 1–3). We then stimulated Gpr31b−/− CX3CR1+ cells with the methanol-soluble fraction of the small intestinal luminal contexts of SPF (Fig. 2b, Extended Data Fig. 3f). The methanol-soluble fraction did not induce dendrite extension of Gpr31b−/− CX3CR1+ cells. Thus, GPR31 is essential for the dendrite protrusion of intestinal CX3CR1+ cells. CX3CR1+ cells are responsible for the uptake of luminal antigens, including intestinal bacteria6. Therefore, we orally administered non-invasive ΔinvA Salmonella enterica serovar Typhimurium (hereafter, S. Typhimurium) and the uptake of the bacteria by CX3CR1+ cells was analysed (Extended Data Fig. 4a). Wild-type CX3CR1+ cells contained an increased number of the bacteria, but the number of bacteria was markedly decreased in Gpr31b−/− CX3CR1+ cells. When CX3CR1+ cells were incubated with ΔinvA S. Typhimurium in vitro, bacterial uptake was equally observed in wild-type and Gpr31b−/− CX3CR1+ cells, indicating that the capacity of bacterial engulfment was not impaired in Gpr31b−/− CX3CR1+ cells (Extended Data Fig. 4b). Because CX3CR1+ cells have previously been shown to mediate the access of luminal antigens to mesenteric lymph nodes3,6, we next analysed the number of bacteria in the small intestine, mesenteric lymph nodes, spleen and liver (Fig. 2c, Extended Data Fig. 4c). The bacterial load in these tissues was severely reduced in Gpr31b−/− mice. Oral administration of non-pathogenic ΔinvA ΔaroA S. Typhimurium induced antibody responses and increased Salmonella-specific IgG in the serum of wild-type mice (Fig. 2d). However, the antibody response was markedly decreased in Gpr31b−/− mice. We next analysed the effect of non-pathogenic S. Typhimurium pretreatment on survival against invasive S. Typhimurium infection (Fig. 2e). Mice infected with invasive S. Typhimurium died within two weeks of infection. Oral immunization of non-pathogenic S. Typhimurium protected wild-type mice from infection with invasive S. Typhimurium. However, the resistance to invasive S. Typhimurium was notably reduced in Gpr31b−/− mice. Wild-type and Gpr31b−/− mice that were immunized with non-pathogenic S. Typhimurium via an intravenous route showed high resistance to infection with invasive S. Typhimurium (Extended Data Fig. 4d). Thus, consistent with the impaired dendrite protrusion of intestinal CX3CR1+ cells, Gpr31b−/− mice showed a reduced immune response to enteric bacteria.

Fig. 2: GPR31-dependent dendrite protrusion of intestinal CX3CR1+ cells.
figure 2

a, Trans-epithelial dendrite protrusion of CX3CR1+ cells in the ileum. Explants of the distal ileum were stained with CMRA (red), and observed by two-photon microscopy. Arrowheads indicate trans-epithelial dendrites. b, Effects of Gpr31b deficiency on the dendrite extension of CX3CR1+ cells in response to the methanol-soluble fraction of intestinal luminal contents. The dendrite extension of CX3CR1+ cells was observed under a microscope. The experiment was repeated three (a) or two (b) independent times with similar results. c, In vivo uptake of non-invasive S. Typhimurium by GPR31-mediated signalling. Mice were orally administered ΔinvA S. Typhimurium, and bacterial loads in the small intestine were analysed (n = 12 mice). Data were from three biologically independent experiments. CFU, colony-forming units. d, Titres of S. Typhimurium-specific IgG in the serum of mice administered non-pathogenic ΔinvA ΔaroA S. Typhimurium (n = 7 mice). e, Effects of non-pathogenic S. Typhimurium pretreatment on survival against invasive S. Typhimurium infection. Mice were administered with ΔinvA ΔaroA S. Typhimurium for six weeks and the survival after invasive S. Typhimurium (1 × 108 CFU) infection was monitored (n = 8 mice). Two-tailed Mann–Whitney test (c), two-tailed Student’s t-test (d) and log-rank test (e) were performed for statistical analysis. Data represent the mean ± s.d. Scale bar, 10 μm.

Source data

Full size image

We next aimed to purify the particular luminal products that activate GPR31. The stimulation of HEK293 cells that express human GPR31 with the methanol-soluble fraction from the luminal contents of SPF mice increased the concentration of the cytosolic cyclic AMP (cAMP) (Extended Data Fig. 5a). With this assay system, we purified a GPR31-activating product in the methanol-soluble fraction step-by-step using anion-exchange chromatography (for the neutral and basic fraction), gel permeation chromatography (fraction number 5 and fraction number 6) and, finally, hydrophilic interaction liquid chromatography (HILIC) (retention time of 8–10 min) (Extended Data Fig. 5b–f). To obtain information on the molecular masses of the HILIC separation at 8–10-min retention time, fraction number 5 and fraction number 6 from the gel permeation chromatography were analysed by liquid chromatography–mass spectrometry using the HILIC mode (HILIC-LC–MS) (Fig. 3a). Mass analysis at the 8–10-min retention time of fraction number 5 and fraction number 6 indicated the presence of a molecule that showed a strong molecular ion peak at m/z 89.023, which represented C3H5O3. The tandem mass spectrometry (MS/MS) spectra of these fractions were consistent with that of lactic acid (C3H6O3) (Fig. 3b). LC–MS chromatograms (at m/z 89.022–89.026) of fraction number 5 and fraction number 6 exhibited a retention time that was similar to that of lactic acid (Extended Data Fig. 6a). Thus, the purified product contained lactic acid. We therefore stimulated HEK293 cells that express mouse GPR31 and human GPR31 with various chemical compounds, including lactic acid (Fig. 3c, d). Lactic acid and pyruvic acid—the reduction of which by lactate dehydrogenase produces lactic acid, and which is accordingly structurally related to lactic acid—activated both mouse GPR31 and human GPR31. d- and l-lactic acids—as well as dl-lactic acid, which includes equal amounts of d- and l-lactic acids—similarly activated mouse GPR31 (Extended Data Fig. 6b). Therefore, we used dl-lactic acid in the experiments described below. We determined the half maximal effective concentrations (EC50) of lactic acid and pyruvic acid (Fig. 3e–h). Lactic acid activated mouse GPR31 and human GPR31 with EC50 values of 400 μM and 1 μM, respectively, whereas pyruvic acid activated the same two proteins with EC50 values of 110 nM and 3 nM, respectively. Concentrations of d-lactate, l-lactate and pyruvate were severely reduced in the small-intestinal luminal contents of germ-free mice as compared with SPF mice (Fig. 3i–k), which indicates that the production of both compounds in the small intestine is dependent on bacteria. The methanol-soluble fraction isolated from SPF mice that were orally treated with vancomycin and neomycin—but not with gentamicin, ampicillin or metronidazole—did not increase cytosolic cAMP in cells that express mouse GPR31 (Extended Data Fig. 7a). In addition, luminal concentrations of d-lactate, l-lactate and pyruvate, as well as the dendrite protrusion of CX3CR1+ cells, were all markedly reduced in SPF mice that were orally treated with vancomycin or neomycin (Extended Data Fig. 7b–f). These findings indicate that specific bacteria (possibly producing lactic acid and/or pyruvic acid) are responsible for the activation of GPR31. Therefore, we next screened commensal bacteria that produce lactate and pyruvate in vitro, and found that Lactobacillus helveticus secreted lactate and high amounts of pyruvate (Extended Data Fig. 8a–c). Concentrations of d-lactate, l-lactate and pyruvate in the luminal contents of SPF mice that were orally administered L. helveticus were increased (Extended Data Fig. 8d–f). Thus, we identified lactic acid and pyruvic acid as GPR31-activating bacterial metabolites, of which pyruvic acid is a potent activator of GPR31.

Fig. 3: Identification of lactic acid and pyruvic acid as GPR31-activating bacterial metabolites.
figure 3

a, Mass spectra at 8–10-min retention time in HILIC-LC–MS analysis of fraction number 5, fraction number 6 (which activated human GPR31 (hGPR31)) and fraction number 2 (which did not activate human GPR31) are shown. The accurate molecular mass of the major ion peak in fraction number 5 and fraction number 6 was determined as m/z 89.023, which corresponds to C3H5O3. b, Left, the structural formula of lactic acid. Right, MS/MS spectra of the precursor ion at m/z 89.023 detected in fraction number 5, fraction number 6 and lactic acid. The experiment was repeated two independent times with similar results (a, b). c, d, Reactivity of mouse GPR31 (mGPR31) (c, n = 6 independent samples per group) and human GPR31 (d, n = 5 independent samples per group) to dl-lactic acid, pyruvic acid and the indicated chemical compounds. Intracellular cAMP levels in HEK293 Tet-on inducible cells were assessed in the presence or absence of doxycycline (Dox). Data represent two biologically independent experiments. e, f, Dose-dependent reactivity of mouse GPR31 to lactic acid (e) and pyruvic acid (f) (n = 3). g, h, Dose-dependent reactivity of human GPR31 to lactic acid (g) and pyruvic acid (h) (n = 3 independent samples per group). i–k, Concentrations of d-lactate (i), l-lactate (j) and pyruvate (k) in luminal contents prepared from the small intestine of SPF and germ-free mice (n = 6 mice each). Two-tailed Student’s t-test was performed for statistical analysis (c, d, i–k). Data represent the mean ± s.d.

Source data

Full size image

We next analysed whether lactic acid and pyruvic acid induce dendrite protrusion of intestinal CX3CR1+ cells. Both lactic acid and pyruvic acid induced dendrite extension of wild-type CX3CR1+ cells, but not Gpr31b−/− cells, in vitro (Fig. 4a, Extended Data Fig. 9a). We then orally administered lactate and pyruvate to wild-type and Gpr31b−/− mice for 21 consecutive days. Both lactate and pyruvate increased the number of dendrite protrusions of CX3CR1+ cells in wild-type mice, but not in Gpr31b−/− mice (Fig. 4b, Extended Data Fig. 9b, Supplementary Videos 4–9). Increased numbers of dendrite protrusion of CX3CR1+ cells were observed at as early as one day after pyruvate administration and two days after lactate administration (Extended Data Fig. 9c). Oral administration of L. helveticus increased the number of dendrite protrusion of CX3CR1+ cells of wild-type mice, but not of Gpr31b−/− mice (Extended Data Fig. 8g). These findings demonstrate that lactate and pyruvate mediate the dendrite protrusion of intestinal CX3CR1+ cells via GPR31. Previous studies have demonstrated that other subsets of myeloid cells, such as CD103+ cells, extend their dendrites in response to oral Salmonella infection7,8. CD103+ cells isolated from the small intestine did not extend their dendrites in response to lactic acid or pyruvic acid in vitro (Extended Data Fig. 9d, e). We then analysed the dendrite protrusion of intestinal CD103+ cells after Salmonella infection. CD103+ cells of both wild-type and Gpr31b−/− mice extended their dendrites after Salmonella infection (Extended Data Fig. 9f). Thus, dendrite protrusion of intestinal CD103+ cells is induced in a GPR31-, lactate- and pyruvate-independent fashion. We next analysed the effect of lactate and pyruvate on the uptake of S. Typhimurium by CX3CR1+ cells. Mice were orally treated with lactate or pyruvate, and then administered ΔinvA S. Typhimurium (Fig. 4c). The bacterial load at five days after administration was markedly increased in wild-type mice treated with lactate and pyruvate. However, lactate and pyruvate failed to increase bacterial uptake in Gpr31b−/− mice. Consistent with the increased bacterial uptake, treatment with lactate and pyruvate increased serum concentrations of Salmonella-specific IgG in wild-type mice, but not in Gpr31b−/− mice, when they received ΔinvA ΔaroA S. Typhimurium (Fig. 4d). We then analysed the effect of lactate or pyruvate, and non-pathogenic S. Typhimurium, on resistance to invasive S. Typhimurium. Mice were sequentially treated with lactate or pyruvate and ΔinvA ΔaroA S. Typhimurium via an oral route, and then infected with invasive S. Typhimurium (Fig. 4e). The survival rate was enhanced by treatment with lactate and pyruvate in wild-type mice. Furthermore, treatment with lactate and pyruvate markedly decreased the number of S. Typhimurium in the liver and spleen, and suppressed liver damage of wild-type mice (Extended Data Fig. 9g–i). By contrast, Gpr31b−/− mice were highly sensitive to infection with invasive S. Typhimurium compared to wild-type mice, and lactate and pyruvate did not increase the survival rate. These findings demonstrate that the dendrite protrusion of intestinal CX3CR1+ cells, dependent on lactate and/or pyruvate and GPR31, is required for the uptake of luminal antigens that leads to resistance to enteric pathogenic bacteria.

Fig. 4: GPR31-dependent dendrite protrusion of CX3CR1+ cells induced by lactic acid and pyruvic acid.
figure 4

a, Effects of lactic acid and pyruvic acid on the dendrite extension of CX3CR1+ cells. Gpr31b+/+ and Gpr31b−/− CX3CR1+ cells were treated with lactic acid, pyruvic acid or propionic acid (100 μM each), and their morphology was observed. b, Effects of oral administration of lactate or pyruvate on the trans-epithelial dendrite protrusion of CX3CR1+ cells. Trans-epithelial dendrites (arrowheads) in explants of the distal ileum were observed by two-photon microscopy after administration of 100 mM lactate or pyruvate for 3 weeks (green, GFP; red, CMRA). Bottom panels show higher-magnification views of outlined areas in top panels. The experiment was repeated two (a) or three (b) independent times with similar results. c–e, Effects of lactate and pyruvate on S. Typhimurium infection. Mice were orally administered lactate or pyruvate for 3 weeks, and bacterial loads after non-invasive S. Typhimurium infection (c), and production of IgG specific to S. Typhimurium (d) were analysed (c, n = 8 mice; d, n = 19 mice). Mice were pretreated with lactate or pyruvate and non-pathogenic S. Typhimurium, and the survival after invasive S. Typhimurium (1 × 109 CFU) infection was monitored (e) (immunized Gpr31b−/− mice, n = 11; immunized Gpr31b+/+ mice and unimmunized mice; n = 12). Two-tailed Student’s t-test (c, d), and log-rank test (e) were performed for statistical analysis. Data represent the mean ± s.d. Scale bar, 10 μm. NS, not significant.

Source data

Full size image

Previous reports have shown that the CX3CL1–CX3CR1 axis mediates the dendrite protrusion of CX3CR1+ cells6,9. We therefore analysed how GPR31 signalling is involved in the CX3CL1–CX3CR1-dependent dendrite protrusion of CX3CR1+ cells. Intestinal CX3CR1+ cells from Cx3cr1gfp/gfp mice showed defective dendrite extension after treatment with lactic or pyruvic acid (Extended Data Fig. 10a–c). In the intestinal CX3CR1+ cells of Cx3cr1gfp/gfp and Cx3cl1−/− mice, the expression of Gpr31b was severely decreased (Extended Data Fig. 10d). CX3CL1 treatment induced Gpr31b expression in Cx3cl1−/− CX3CR1+ cells (Extended Data Fig. 10d). In addition, in Cx3cl1−/− CX3CR1+ cells that were pretreated with CX3CL1, dendrite extension was induced by lactic acid and pyruvic acid (Extended Data Fig. 10e, f). On the basis of these findings, we propose the following mechanism for the dendrite protrusion of intestinal CX3CR1+ cells. CX3CL1 produced by intestinal epithelial cells mediates GPR31 expression of intestinal CX3CR1+ cells; GPR31 then responds to the bacterial metabolite (lactic and/or pyruvic acid) and the cells protrude their dendrites into the intestinal lumen, thereby taking up luminal antigens for the initiation of immune responses. Compared to intestinal CD103+ cells, CX3CR1+ cells show low motile activity and highly dynamic dendrite protrusion8. Therefore, CX3CR1+ cells predictably recognize CX3CL1 that is secreted from epithelial cells and diffused into the lamina propria. Lactic acid and pyruvic acid are promising compounds that could be used for augmentation of immune responses in various clinical aspects, such as the enhancement of the efficacy of oral vaccines.

Methods

No statistical methods were used to predetermine sample size. The experiments were not randomized and investigators were not blinded to allocation during experiments and outcome assessment.

Antibodies and reagents

PerCP/Cy5.5 anti-CD45 (30-F11), phycoerythrin (PE) anti-CX3CR1 (SA011F11), Alexa Fluor 594 anti-CD103 (2E7), APC anti-CD103 (2E7), PE/Cy7 anti-CD11c (N418), Pacific Blue anti-CD11b (M1/70), PE anti-CCR3 (J073E5), Alexa Fluor 594 anti-CD4 (GK1.5), Alexa Fluor 594 anti-CD8a (53-6.7) monoclonal antibodies (mAbs), and recombinant mouse CX3CL1 were purchased from Biolegend. PE anti-CD103 (M290) was purchased from BD Biosciences. PE anti-MHC class II (M5/114.15.2) and anti-CD16/32 mAb (2.4G2) were purchased from Tonbo Biosciences. dl-lactic acid, d-lactic acid, l-lactic acid, pyruvic acid, 3-isobutyl-1-methylxanthine (IBMX), acetic acid, propionic acid, butyric acid, glycolic acid, l-alanine, 2-hydroxy butyric acid, 3-hydroxy butyric acid, sodium dl-lactate and sodium pyruvate were purchased from Sigma Aldrich. Ki16425 was purchased from Cayman Chemical.

Mice

B57BL/6 mice and ICR mice were purchased from Japan SLC. Myd88−/− mice and Cx3cr1-egfp knock-in mice were generated as previously described16,17. The mice were maintained under SPF conditions. Germ-free (IQI/Jic[Gf] ICR) mice were from CLEA Japan and maintained under germ-free conditions. Age- and sex-matched mice were used for experiments. All animal experiments were conducted in accordance with the guidelines of the Animal Care and Use Committee of Osaka University.

Generation of Gpr31b −/− mice

The Gpr31b gene consists of a single exon. The targeting vector was constructed by replacement of a 1-kb fragment of the exon region containing the coding sequence of Gpr31b with a neomycin-resistance gene cassette, and a gene encoding HSV thymidine kinase was inserted into the genomic fragment for negative selection. The targeting vector was transfected into V6.5 embryonic stem cells, and G418 and ganciclovir double-resistant colonies were selected by PCR and Southern blot analysis. Homologous recombinants were micro-injected into blastocysts of ICR female mice, and heterozygous F1 progeny mice were intercrossed to obtain Gpr31b−/− mice. Gpr31b−/− mice were backcrossed onto C57BL/6 mice for at least six generations.

Generation of Cx3cl1 −/− mice

The targeting vector was assembled using a pBS-vloxP-neo-DTA vector, containing a neomycin-resistance gene cassette flanked by vCre recombinase target sites (vloxP) in a pBluescript SK+ (Stratagene) backbone. A 4-kb 5′-arm genomic fragment including exon 2 flanked by loxP sites, and a 4-kb 3′-arm genomic fragment, were separately cloned into pBS-vloxP-neo-DTA to generate the Cx3cl1 targeting vector. The floxed allele was generated by homologous recombination in the B6J-S1UTR embryonic stem cell line18, according to standard procedures. Then, the loxP-flanked Cx3cl1 exon 2 was deleted by transiently expressed Cre recombinase in the targeted embryonic stem cells to obtain the Cx3cl1 null allele. The Cx3cl1−/− mouse line was generated from the embryonic stem cell line with the Cx3cl1 null allele.

Preparation of organic fractions from intestinal luminal contents

Luminal contents in the small intestine and caecum (approximately 10 g), collected from ICR mice, were homogenized in a fivefold amount (v/w) of ultrapure water by repeated vortexing and sonication, and mixed with an equivalent amount of diethyl ether. The samples were centrifuged at 2,500 r.p.m. for 10 min at 4 °C, and the diethyl ether-soluble and -insoluble fractions were individually collected. The insoluble fraction was further homogenized with a fivefold amount of chloroform, and separated into chloroform-soluble and -insoluble fractions by centrifugation (2,500 r.p.m., 10 min, 4 °C). The insoluble fraction was mixed with a fivefold amount of methanol and subjected to centrifugation (2,500 r.p.m., 10 min, 4 °C), and the supernatant was collected as a methanol-soluble fraction. The methanol-soluble fraction was washed with methyl tert-butyl ether (MTBE) and methanol mixture, and then washed with a MTBE:methanol:acetic acid solution and saturated saline. The organic layer was evaporated to dryness, and dissolved with methanol.

Chromatography

For anion-exchange chromatography, the sample (MTBE-2 in Extended Data Fig. 5b) was applied to a HiTrap Q HP column (GE Healthcare), and eluted with 70% methanol or 1 M NaCl-containing 70% methanol to collect the neutral and basic fraction or the acidic fraction, respectively. The neutral and basic fraction was separated by gel permeation chromatography using a column (bed size 2.5 × 40 cm) packed with Sephadex LH-20 resin (GE Healthcare) with methanol as an eluent, and 50 ml per fraction was collected. The mixture of fraction number 5 and fraction number 6 (see Extended Data Fig. 5b) was separated by HILIC using iHILIC-Fusion column (2.1 × 100 mm, PEEK, particle size: 5 μm, pore size: 100 Å; Hilicon AB) with mobile phase A, 10 mM CH3COONH4 in water, and mobile phase B, CH3CN. Gradient elution was used from 95% B to 60% B for 18 min at a flow rate of 0.2 ml/min.

Liquid chromatography–mass spectrometry

Gel permeation chromatography fractions (number 2, number 5 and number 6) were subjected to LC–MS analysis using a system consisting of a UFLC-XR (Shimadzu) and QExactive HF mass spectrometer (Thermo Fisher Scientific) equipped with iHILIC-Fusion column (Hilicon AB). The analysis was operated in a negative electrospray ionization mode. Molecular species eluted at retention time 8–10 min were analysed by full-scan mas spectrometry with a mass range of 75–1,125 m/z and data-dependent MS/MS acquisition on the 10 most-intense ions using the Orbitrap. The mass spectrometry settings were as follows: resolution for full-scan mass spectrometry 120,000; resolution for MS/MS 15,000; spray voltage −2,000 V; capillary temperature 350 °C; probe heater temperature 300 °C; sheath gas 40 arbitrary units (arb); auxiliary gas 10 arb; sweep gas 0 arb; S-lens RF level 55; higher-energy collisional dissociation (normalized collision energy) fragmentation with stepped collision energy 50, 70.

Assessment of intracellular cAMP levels

Intracellular cAMP levels were assessed as previously described19, with modifications. cDNA fragments encoding Flag-tagged human GPR31 or mouse GPR31 were inserted into a pcDNA5/FRT/TO vector (Thermo Fisher Scientific). Flp-In T-REx-293 cells (Thermo Fisher Scientific) were co-transfected with these constructs and the pOG44 Flp recombinase expression vector. The stable transfectants were treated with 10 μg/ml doxycycline hyclate for 24 h, followed by incubation with 10 μM Ki16425 and 500 μM IBMX for 1 h to inhibit lysophosphatidic-acid-receptor-mediated cAMP responses and phosphodiesterase-mediated cAMP degradation, respectively. The cells were then stimulated with indicated materials for 10 min and lysed with 100 mM HCl. The lysate was centrifuged at 3,000 r.p.m. for 10 min, and the cAMP levels in the supernatant were measured using a cyclic AMP EIA kit (Cayman Chemical), according to the manufacturer’s protocol.

Cell preparation

Myeloid cells in the small intestine and Peyer’s patches were prepared as previously described20 with minor modification. In brief, small intestinal tissues were opened longitudinally and washed extensively with PBS. Small intestinal segments and Peyer’s patches were incubated in PBS containing 10 mM EDTA at 37 °C for 20 min and washed several times with PBS. Small intestinal segments, Peyer’s patches, lymph nodes, spleen and brain were digested with continuous stirring in RPMI 1640/10% FCS with 400 M and l U/ml collagenase D (Roche) and 100 μg/ml DNase I (Sigma Aldrich) at 37 °C for 50 min. Low-density cells were enriched by 17.5% Accudenz solution (Accurate Chemical & Scientific). Myeloid cells in the large intestine and lung were prepared as described21,22.

Assessment of dendrite extension in vitro

Intestinal low-density cells (1 × 105 cells) were incubated in a fibronectin (Sigma Aldrich)-coated 96-well plate at 37 °C for 1 h, followed by stimulation with indicated reagents for 4 h. The cells were fixed with 4% paraformaldehyde for 20 min at 37 °C and stained with DAPI. Cell images were automatically taken at 100 pictures per well using an IN Cell analyser (GE Healthcare). Proportions of GFP and DAPI double-positive cells with dendrites more than 5 μm in length were quantified using the IN Cell analyser algorithm.

Assessment of dendrite extension in vivo

The intestine was collected from anaesthetized mice and segments of the ileum (terminal: 0–5 cm and proximal: 6–10 cm, from the caecum) were immediately everted and gently washed with RPMI 1640. The tissue segments were stained by 125 nM CellTracker Orange CMRA Dye (Thermo Fisher Scientific) for 5 min at 37 °C, washed, and readily immobilized on SecureSeal hybridization chambers (Grace Bio-Labs). Images were acquired using an inverted confocal microscope (FV1000-D; Olympus) with a 60× oil immersion lens or an inverted two-photon microscope (A1R-MP; Nikon) equipped with a 20× water immersion lens (Plan Fluor, N.A. 0.75; Nikon) and Chameleon Vision II Ti: sapphire laser (Coherent) set at 880 nm for two-photon excitation. The typical optimal z-step size was 0.5–1 μm. Serial 40–50-μm z-scan images were collected and reconstructed using Imaris software (Bitmap). The number of trans-epithelial dendrites of 5 μm length (from the basal lamina of the epithelial layer) was quantified. To exclude the possibility of different size of villi and number of CX3CR1+ cells between samples, we normalized these two factors and analysed the number of dendrite protrusions, which produced the same results.

Antibiotic treatment

Mice were administered a combination of four antibiotics, consisting of 1 mg/ml ampicillin (Nacalai Tesque), 1 mg/ml neomycin (Nacalai Tesque), 1 mg/ml metronidazole (Nacalai Tesque) and 500 μg/ml vancomycin (Duchefa Biochemie B.V.) in sterilized drinking water for four weeks. In some experiments, mice were administered each antibiotic or 2 mg/ml gentamicin (Nacalai Tesque).

Quantitative PCR

Total RNA was isolated using Total RNA Miniprep Kit (Sigma Aldrich) and single-stranded cDNA was synthesized by Moloney murine leukaemia virus reverse transcription (Promega) and random primers (Toyobo) after treatment with RQ1 DNase I (Promega). PCR was performed by using GoTaq qPCR Master Mix (Promega) at 95 °C for 10 min, followed by 50 cycles at 95 °C for 15 s, and at 60 °C for 60 s. The PCR primer sets were Gpr31: 5′-AGTCTGACAAACAGCCCAGG-3′, and 5′-CACTAGGCAGGAAGCACAGTC-3′, which detect all Gpr31 isoforms (Gpr31a, Gpr31b and Gpr31 unnamed isoform), Gpr65: 5′-TTGCCAGCCTCC TCAGTCA-3′ and 5′-GGTCGGTGCAAATGGGAA-3′; Gpr114: 5′-GACCGG AACTCATCACTGCTC-3′ and 5′-GGGTCTCCAGAGAGCATAAGCA-3′; Gpr132: 5′-TCAGGACTGGCTTGGGTCA-3′ and 5′-AATGGCACCGTGCTGATGTA-3′; Gpr171: 5′-GAGACGACACAGCCAGGTACA-3′ and 5′-CCGGACAGAAC GTTGAACTG-3′; Gpr183: 5′-CCATTCTGAGCAAACACGGAC-3′ and 5′-CTGTGCTGTGGTGGGCATAG-3′; Gapdh: 5′-CCTCGTCCCG TAGACAAAATG-3′ and 5′-TCTCCACTTTGCCACTGCAA-3′. To distinguish the difference in the expression between Gpr31b and other Gpr31 isoforms, PCR was performed with HiDi polymerase (myPOLS Biotec), which discriminates primers with a mismatch at the 3′-end. The primer sets were; Gpr31b: 5′-CATCTCCTTCTGCAACAGTGGC-3′ and 5′-AGGACACTAGGCAGGAAGCACAG-3′; other Gpr31 isoforms: 5′-CATCT CCTTCTGCAACAGTGGG-3′ and 5′- AGGACACTAGGCAGGAAGCACAG-3′; Gapdh: 5′-GGGTGTGAACCACGAGAAAT-3′ and 5′-CCTTCCACAAT GCCAAAGTT-3′.

Flow cytometry

Cells were pretreated with anti-CD16/32 mAb to block nonspecific binding, and then incubated with indicated mAbs. Flow cytometric analysis was performed using a FACSCanto II flow cytometer (BD Biosciences) with FlowJo software (Tree Star). Indicated cells were isolated using a FACSAria (BD Biosciences). Cell populations were gated as follows; in the blood, Gr-1− monocytes (Gr-1−CX3CR1high), Gr-1+ monocytes (Gr-1+CX3CR1int), eosinophils (CD11bintCCR3+CX3CR1−), neutrophils (CD11bhighCCR3−CX3CR1−). In the small intestine, CD11b− dendritic cells (CD11b−CD11c+CX3CR1−), CD11b+ dendritic cells (CD11b+CD11c+CX3CR1−), CX3CR1+ cells (CD11b+CD11c+CX3CR1+) and eosinophils (CD11b+CD11cintCX3CR1−CCR3+). In mesenteric lymph nodes, CD103+ dendritic cells (CD103+CD11c+), CD103− dendritic cells (CD103−CD11c+) and macrophages (CD11c−CD11b+). The gating strategies are shown in Extended Data Fig. 2c–g.

Measurement of lactate and pyruvate

Luminal contents of the small intestine were homogenized in water and centrifuged at 10,000 r.p.m. for 10 min at 4 °C. The concentrations of d-lactate, l-lactate and pyruvate in the supernatants were measured using a d-lactate colorimetric assay kit, lactate colorimetric assay kit II and pyruvate assay kit (BioVision), respectively, according to the manufacturer’s protocol.

S. Typhimurium strains

The wild-type S. Typhimurium strain SL1344 and the non-invasive ΔinvA strain SB13623 were used. To generate the non-pathogenic ΔinvA ΔaroA strain, a deletion of aroA was constructed using the PCR-based λ Red recombinase system24. A DNA fragment was amplified by PCR with aroA-P1 (5′-CTGTGGGGTTTTTATTTCTGTTTTTTGAGAGTTGAGTTTCgtgtaggctggagctgcttc-3′) and aroA-P2 (5′-GACTCG GC GC GCCAGCCCGTCGACTGGCGCAACAGAAGAC catat gaatatcctccttag-3′) primers using pKD3 as a template. The PCR product was introduced into SL1344 carrying pKD46. Transformants were selected on LB agar plates containing chloramphenicol, and the aroA deletion was confirmed by PCR using aroA-Pcheck (5′-GAGCTGCCCGTCCATCCTCGACTAC-3′) and C1 (5′-TTATACGCAAGGCGACAAGG-3′) primers. To construct the ΔinvA ΔaroA mutant (TH1889), the aroA deletion was transduced by phage P22 into SB136 and verified by PCR.

Bacterial culture

Lactobacillus reuteri (JCM 1112), Lactobacillus gasseri (JCM 1131), Lactobacillus ruminis (JCM 1152), Lactobacillus salivarius (JCM 1231), Lactobacillus johnsonii (JCM 2012), Lactobacillus animalis (JCM 5670), Lactobacillus apodemi (JCM 16172), Lactobacillus acidophilus (JCM 1132), Lactobacillus delbrueckii (JCM 1012), L. helveticus (JCM 1120), Lactobacillus crispatus (JCM 1185), Bacteroides sartorii (JCM 17136) and Bifidobacterium breve (JCM 1192) were purchased from Japan Collection of Microorganisms (JCM) of RIKEN BioResource Research Center Microbe Division. These bacteria were cultured in GAM medium supplemented with 1% glucose in anaerobic conditions for in vitro experiments.

Administration of lactate and pyruvate, and lactate- and pyruvate-producing bacteria

Mice were orally administered with sodium dl-lactate or sodium pyruvate via drinking water for the indicated periods (100 mM each, unless otherwise indicated). The lactate- and pyruvate-producing bacterium L. helveticus was cultured in MRS medium for in vivo administration. The bacteria were resupended in fresh MRS medium and 2 × 109 CFU of bacteria were orally injected into SPF mice for consecutive eight days.

S. Typhimurium infection

To analyse S. Typhimurium uptake by CX3CR1+ cells, mice were orally administered 1 × 109 CFU of non-invasive ΔinvA S. Typhimurium strain in 200 μl PBS. After five days of infection, the small intestines were collected and luminal contents were washed off to remove S. Typhimurium in the lumen. Bacteria titres were quantified by plating homogenized intestinal tissues on 50 μg/ml streptomycin-containing LB agar plates. To assess immunoglobulin production induced by S. Typhimurium infection, mice were pretreated with sodium lactate or sodium pyruvate (100 mM each) for three weeks, and orally administered 1 × 109 CFU of non-pathogenic ΔinvA ΔaroA S. Typhimurium strain in 200 μl PBS three times on alternate days. Serum samples were collected at weekly intervals and used for enzyme-linked immunosorbent assay. To analyse the resistance to invasive S. Typhimurium, mice were orally administered wild-type S. Typhimurium SL1344 strain 6 weeks after the first administration of ΔinvA ΔaroA S. Typhimurium, and their survival rates were monitored. For histological analysis, tissues were fixed with 4% paraformaldehyde and frozen sections were stained with haematoxylin and eosin. The serum aspartate aminotransferase (AST) concentration was measured by the MDH-UV methods (Oriental Yeast).

Detection of Salmonella-specific antibodies

Titres of S. Typhimurium-specific antibodies were determined by enzyme-linked immunosorbent assay. A 96-well plate was coated with heat-killed S. Typhimurium (1 × 107 CFU) and blocked by 5% bovine serum albumin in PBS. The wells were incubated with serum samples for 2 h at room temperature, followed by incubation with HRP-labelled goat anti-mouse IgG (Santa Cruz Biotechnology). After being washed, the wells were incubated with TMB substrate (R&D Systems), and the absorbance was read at 450 nm.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this paper.