Commensal bacteria weaken the intestinal barrier by suppressing epithelial neuropilin-1 and Hedgehog signaling

Abstract

The gut microbiota influences intestinal barrier integrity through mechanisms that are incompletely understood. Here we show that the commensal microbiota weakens the intestinal barrier by suppressing epithelial neuropilin-1 (NRP1) and Hedgehog (Hh) signaling. Microbial colonization of germ-free mice dampens signaling of the intestinal Hh pathway through epithelial Toll-like receptor (TLR)-2, resulting in decreased epithelial NRP1 protein levels. Following activation via TLR2/TLR6, epithelial NRP1, a positive-feedback regulator of Hh signaling, is lysosomally degraded. Conversely, elevated epithelial NRP1 levels in germ-free mice are associated with a strengthened gut barrier. Functionally, intestinal epithelial cell-specific Nrp1 deficiency (Nrp1^ΔIEC) results in decreased Hh pathway activity and a weakened gut barrier. In addition, Nrp1^ΔIEC mice have a reduced density of capillary networks in their small intestinal villus structures. Collectively, our results reveal a role for the commensal microbiota and epithelial NRP1 signaling in the regulation of intestinal barrier function through postnatal control of Hh signaling.

Fig. 1. Hedgehog pathway is modulated by the gut microbiota via epithelial TLR2.

a, Comparative immunoblot analysis of IHH protein levels of small intestinal tissue lysates from GF versus CONV-R C57BL/6J mice, relative to α-actinin (n = 6 versus 6, P = 0.0262). The 45 kDa IHH precursor is detected. Insert shows a representative western blot. b,c, Relative gene expression of Ihh and Gli1 in (b) C57BL/6J GF, CONV-R and CONV-R mice treated with antibiotics (CONV-R + Abx) (Ihh: n = 7 versus 6 versus 7; GF versus CONV-R, P < 0.0001; GF versus CONV-R + Abx, P = 0.0045; CONV-R versus CONV-R + Abx, P = 0.0413; Gli1: n = 8 versus 6 versus 6; GF versus CONV-R, P = 0.0097) or (c) Swiss Webster GF versus CONV-R mice (Ihh: n = 14 versus 6, P < 0.0001; Gli1: n = 6 versus 14, P = 0.0303). NS, not significant. d, qRT–PCR array on pooled concentration-adjusted mRNAs of seven mice per group (n = 7 versus 7), showing differential expression of genes involved in the Hh pathway (see legend) between GF and CONV-D mice (relative to CONV-D). Ihh and Gli1 are highlighted with a black arrow. FC, fold change. TGF, transforming growth factor. e–h, Relative gene expression of Ihh and Gli1 in WT mice versus Tlr2^−/− global knockout mice in CONV-R (e) (Ihh: n = 17 versus 11, P = 0.0025; Gli1: n = 12 versus 11, P = 0.0298) or GF (f) (Ihh: n = 8 versus 9; Gli1: n = 4 versus 7) housing conditions and in Tlr2^ΔIEC CONV-R mice (g,h) in comparison to WT littermates (distal small intestine: Ihh, n = 7 versus 7, P = 0.0035; Gli1: n = 7 versus 6, P < 0.0001. IECs: Ihh, n = 7 versus 7, P = 0.0003). qRT–PCR analyses were performed on the whole tissue (distal small intestine) (b–g), whereas for h analyses were on isolated IECs. i, Relative Ihh expression in MODE-K cells after stimulation with the TLR2 agonist PG (n = 4 versus 4, P = 0.0059). For qRT–PCR assays, L32 was used as a housekeeping gene. In all panels, values were normalized for the mean expression of the control group. Individual values are displayed as dots, while mean ± s.e.m. is shown as a column and error bar (a–c,e–i). Statistical analyses were performed with one-way analysis of variance (ANOVA) and Tukey’s multiple comparison test (b). Unpaired Student’s t-test was used (a,c,e–i). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. j,k, sm-FISH for the Hh downstream targets Gli1 (magenta), Ptch1 (green) and Hhip (white) on distal small intestine sections from GF versus CONV-R (j) and Tlr2^ΔIEC in comparison to WT littermates (k). Gli1 and Ptch1 transcripts are highlighted with color-coded arrowheads. For each group, the experiment was performed on n = 3 mice. For each representative image, two magnifications are shown. Scale bars, 100 μm and 20 μm. DAPI, 4,6-diamidino-2-phenylindole. n represents the number of biological independent mice (a–c,e–h,j), whereas in i it represents the number of independent experiments on cell cultures. Source data

Fig. 2. NRP1 protein levels in the gut epithelium are regulated by the gut microbiota through TLR2-mediated lysosomal degradation.

a, Representative immunofluorescence images of NRP1 expression (green) in the distal small intestine of GF and CONV-R mice. The experiment was performed twice. Cell nuclei are counterstained with DAPI (in blue). Scale bars, 100 μm. b, Relative NRP1 protein levels in the distal small intestine of GF, CONV-R and CONV-D mice (n = 4 versus 5 versus 6; GF versus CONV-D, P = 0.0018; GF versus CONV-R, P = 0.0001). c–e, NRP1 (c) protein (n = 5 versus 4, P = 0.0009) and mRNA (n = 13 versus 9) expression (d) in IECs isolated from GF and CONV-R mice or WT mice (e) versus Tlr2^−/− in CONV-R housing conditions (n = 5 versus 4, P < 0.0001). n represents the number of biological independent mice (b–e). f, TLR2-mediated NRP1 degradation in MODE-K cells by lysosome or proteasome. NRP1 degradation by TLR2 is induced by MALP-2 (TLR2/TLR6 agonist) stimulation. Blocking of lysosomal degradation (left) is achieved by stimulation with bafilomycin A1 in 0.125% (v/v) DMSO (vehicle). Lysosome inhibition prevents TLR2/TLR6-induced NRP1 degradation (MALP-2 + bafilomycin A1) (control versus MALP-2, P < 0.0001; MALP-2 versus MALP-2 + bafilomycin A1, P < 0.0001; vehicle versus MALP-2 + bafilomycin A1, P = 0.0122). Blocking of proteasome (right) is performed by stimulation with epoxomicin in 0.1% (v/v) DMSO (vehicle). Proteasome inhibition does not prevent NRP1 degradation by MALP-2 (MALP-2 + epoxomicin) (control versus MALP-2, P = 0.0328; vehicle versus MALP-2 + epoxomicin, P = 0.0340). g, Inhibition of NRP1 degradation by lysosome shown by flow cytometry, using the same experimental conditions of f (medium versus MALP-2, P = 0.0459; MALP-2 versus MALP-2 + bafilomycin A1, P = 0.0011). Representative histograms are shown (right). For western blot analyses, the number of independent experiments (n) on cell cultures is 4–8, whereas for flow cytometry this was n = 5. GF and CONV-R mice were analyzed on different gels that were processed in parallel (c). In the qPCR assay, L32 was used as the housekeeping gene, whereas in western blot, protein expression is relative to α-actinin or β-actin. Values are normalized for the mean expression of the controls (b–f). Individual values are displayed as dots, whereas mean ± s.e.m. is shown as a column and error bar (b–g). Statistical analyses were performed with one-way ANOVA and Tukey’s multiple comparison test (b,f,g), whereas for c–e, an unpaired Student’s t-test was used. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Source data

Fig. 3. Impairment of gut barrier function by the gut microbiota, gut epithelial NRP1-deficiency and inhibition of Hedgehog signaling.

a, FITC-dextran permeability assay on GF versus CONV-R mice (n = 5 versus 11, P = 0.0006). b, Relative gene expression of claudin-4 (Cldn4), junctional adhesion molecule-A (F11r), occludin (Ocln) and ZO-1 (Tjp1) in IECs from GF versus CONV-R mice (n = 6 versus 6; Cldn4: P = 0.0004; F11r: P = 0.0015; Ocln: P < 0.0001; Tlp1: P < 0.0001). c,d, Relative occludin (c) and ZO-1 (d) protein expression in IECs of GF versus CONV-R mice (occludin: n = 13 versus 19, P = 0.0015; ZO-1: n = 4 versus 9, P = 0.0053). e,f, Relative gene expression of Ihh and Gli1 in the distal small intestine of WT littermates versus Nrp1^ΔIEC mice in CONV-R conditions (e) (Ihh: n = 7 versus 7, P = 0.0033; Gli1: n = 6 versus 7, P = 0.0185) and after antibiotic treatment (Abx) (f) (Ihh: n = 7 versus 6; Gli1: n = 6 versus 6). g, Gut microbiota mean relative abundance on the phylum level as determined by bacterial 16S rRNA gene sequencing of small intestinal tissue samples from Nrp1^ΔIEC mice versus Cre-negative WT littermates. h,i, FITC-dextran permeability assay on WT (Cre-negative) littermates versus Nrp1^ΔIEC in CONV-R status (n = 7 vs 5, P = 0.0074) (h) and after antibiotics treatment (Abx) (n = 8 versus 11) (i). j, Relative gene expression of Cldn4, F11r, Ocln, and Tjp1 in IEC from WT littermates versus Nrp1^ΔIEC mice (n = 7 versus 7; Cldn4: P < 0.0001; F11r: P = 0.0099; Ocln: P = 0.0019; Tlp1: P = 0.0003). k,l, Relative occludin (k) and ZO-1 (l) protein expression in IEC from WT littermates versus Nrp1^ΔIEC mice (n = 7 versus 7; ZO-1: P = 0.0337). m, FITC-dextran permeability assay on vehicle controls versus GDC-0449-treated mice (n = 5 versus 6, P = 0.0012). n, Relative gene expression of Cldn4, F11r, Ocln, and Tjp1 in IEC from controls versus GDC-0449-treated mice (n = 7 versus 7; Cldn4: P = 0.0010; F11r: P = 0.0011; Ocln: P < 0.0001; Tjp1: P = 0.0010). o,p, Relative occludin (o) and ZO-1 (p) protein expression in IEC from controls versus GDC-0449-treated mice (occludin: n = 7 versus 11; ZO-1: n = 7 versus 8). n represents the number of biologically independent mice. For the qPCR assays, L32 was used as the housekeeping gene, whereas in western blots, protein expression is relative to α-actinin. Values were normalized for the mean of the control group. Individual values are displayed as dots, while mean ± s.e.m. is shown as a column and error bar (a–f,h–p). Individual values are not shown (g). For all panels, unpaired Student’s t-test was used. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Source data

Fig. 4. Deficiency of epithelial Nrp1 reduces density of blood capillaries without affecting lacteal length in the distal small intestine.

a, Representative immunofluorescence images of PECAM-1 (CD31) expression (green) in the distal small intestine of mice. The analysis was repeated on n = 5 versus 6 mice. b, Quantification of CD31-positive area per villus structure, indicating vessel density (n = 5 versus 6, P = 0.0013). ROI, region of interest. c, Relative gene expression of Pecam1 in the distal small intestine (n = 6 versus 6, P = 0.0103). d, Representative immunofluorescence images of CD31 (green) and LYVE-1 (red) in the distal small intestine of mice. The analysis was repeated on n = 5 versus 5 mice. e–g, Measurements of villus length (e), lacteal length (f) and lacteal-to-villus ratio (g) (n = 5 versus 5). h,i, Semaphorin 3A (SEMA3A) (h) protein (n = 9 versus 14, P = 0.0189) and mRNA levels (i) in isolated IECs (n = 3 versus 6). For all panels, Nrp1^ΔIEC mice are compared to floxed WT littermates (Cre-negative). n represents the number of biologically independent mice. For immunofluorescence images, cell nuclei were counterstained with To-Pro-3 iodide (blue). Scale bars, 200 μm. For each mouse, the mean measurements of 5–10 villi were taken into account and displayed as a single dot. For qPCR assays, L32 was used as the housekeeping gene, while in western blots, protein expression is relative to α-actinin. Individual values are displayed as dots, while mean ± s.e.m. is shown as a column and error bar (b,c,e–i). Independent samples were analyzed by Student’s t-test, *P < 0.05, **P < 0.01. Source data

Fig. 5. Impact of commensals on Hedgehog signaling, NRP1 and intestinal epithelial permeability.

Conserved molecular patterns such as bacterial ligands from the resident gut microbiota stimulate TLRs on the IECs. In turn, TLR2/TLR6 downregulates NRP1 protein levels in the epithelial compartment through lysosomal degradation. When NRP1 is not degraded, it upregulates IHH, which signals from the epithelium to the mesenchymal compartment. a, In the absence of IHH downstream signaling (shown as exemplification in colonized mice), the transmembrane receptor PTCH1 suppresses the transmembrane protein SMO, resulting in a repressor form of the transcription factor GLI (GLI-R). b, Conversely, after IHH binding to PTCH1 (exemplified in GF mice), repression on SMO is released, yielding the GLI activator (GLI-A) transcription factor and subsequent transcription of GLI targets involved in Hh signaling (GLI targets ‘on’). In GF conditions, this pathway results in upregulation of occludin and ZO-1 protein levels, whereas in CONV-R housing conditions (a), the epithelial gut barrier is impaired. In the scheme, upregulated (↑) proteins are shown in green and downregulated (↓) proteins are in red.

Extended Data Fig. 1. Indian Hedgehog expression in monocolonized and TLR-knockout mouse models; generation and characterization of conditional intestinal epithelial Tlr2 knockout mouse model; bone morphogenetic protein 4 regulation by gut microbiota through TLR2.

(a) Relative mRNA expression of Ihh in the distal small intestine of GF mice compared to ex-GF mice monocolonized with Bacteroides thetaiotaomicron (that is B. theta.) (n = 9 versus 9,  P < 0.0001). (b, c) Relative gene expression of Ihh in the distal small intestine of WT mice versus (b) Tlr4^−/− (CONV-R: n = 5 versus 5; GF, n = 7 versus 6) and (c) Tlr5^−/− (CONV-R: n = 6 versus 6; GF, n = 5 versus 6) global knockout mice in CONV-R and GF housing conditions. (d) Generation of the conditional intestinal epithelial Tlr2-deficient mouse line (Tlr2^ΔIEC). Mice used in the breeding pairs are highlighted by colored squares. P: parental generation; F1, F2, F3: first, second, third filial generation. (e) Relative mRNA expression of Tlr2 in the distal small intestine (n = 7 versus 7, P < 0.0001) or IEC (n = 6 versus 7, P < 0.0001) from WT littermates versus Tlr2^ΔIEC. (f) Relative mRNA expression of Ptch1 and Hhip in the distal small intestine of WT littermates versus Tlr2^ΔIEC mice (n = 7 versus 6, Ptch1: P = 0.0047; Hhip: P < 0.0001). (g-i) Relative mRNA expression of the Hh target bone morphogenetic protein 4 (Bmp4) in the distal small intestine of (g) GF versus CONV-R mice (n = 8 versus 5, P = 0.0013), (h) WT mice versus Tlr2^−/− global knockout mice in CONV-R (n = 7 versus 7, P = 0.0123) and GF (n = 7 versus 7) housing conditions and (i) WT littermates versus Tlr2^ΔIEC mice (n = 7 versus 6, P = 0.0238). For panels a-c, e–i, n represents the number of biological independent mice. For the qPCR assays, L32 was used as the housekeeping gene and the values were normalized for the mean expression of the control group. For panels a-c, e–i, Individual values are shown as dots, while mean ± s.e.m. is shown as a column and error bar (except for panel b, due to bimodal distribution) and unpaired Student’s t test was used. *P < 0.05, **P < 0.01, ****P < 0.0001. Source data

Extended Data Fig. 2. NRP1 and NRP2 expression in the gut of GF, CONV-D and CONV-R mice and MODE-K cell culture model.

(a) Representative immunofluorescence images of NRP1 expression (in red) in the distal small intestine of GF and CONV-D mice. Cell nuclei are counterstained with DAPI (in blue). Scale bar: 50 μm. The measurements were repeated one time. (b) NRP2 expression in the distal small intestine of GF, CONV-R and CONV-D mice (n = 8 versus 10 versus 5). (c) NRP1 expression along the small intestine of GF mice, compared to CONV-R controls (n = 1 versus 1). From the proximal to the distal tract, the small intestine (SI) is divided into 8 equally sized segments. For WB analysis, segments 1, 3, 5, 7 were analyzed. (d) Protein expression of NRP1 after stimulation of MODE-K cells with the TLR2/1 agonist Pam₃CSK₄ (n = 5 versus 5). For panels b-c, n represents the number of biological independent mice, whereas for panel d, n is the number of independent experiments performed on cell cultures. For the western blot assays, protein expression is relative to α-actinin or β-actin and the values were normalized for the mean expression of the control group. For panels b, d, individual values are shown as dots, whereas mean ± s.e.m. is shown as a column and error bar. For panel b, statistical analyses were performed with one-way ANOVA, Tukey’s multiple comparison test. For panel d, Unpaired Student’s t test was used. ns: P > 0.05. (e) Gating strategy for flow cytometry analysis on MODE-K cells. In the FSC-A versus SSC-A panel, single cells are selected. Gating of EPCAM⁺ NRP1⁺ DAPI⁻ singlet specific stain (that is, living IEC expressing NRP1) are based on the isotype control (upper panel). Source data

Extended Data Fig. 3. Characterization of the Nrp1^ΔIEC mouse model, efficacy of NRP1 deficiency, NPR2 expression and microbiome analysis.

(a) Generation of the Nrp1^ΔIEC mouse line. Mice used in the breeding pairs are highlighted by colored squares. P: parental generation; F1, F2, F3: first, second, third filial generation. (b) Genotyping of Villin-Cre (upper panel) and Nrp1 fl/fl (lower panel). (c) NRP1 (n = 5 versus 6, P = 0.0019) and NRP2 (n = 12 versus 11) protein expression in IEC from WT littermates versus Nrp1^ΔIEC mice. (d) Mean relative abundance on the phylum level as determined by bacterial 16 S rRNA gene sequencing of intestinal content samples from Nrp1^ΔIEC mice and floxed WT littermates. (e) Principal coordinate analysis (PCoA) based on the Bray–Curtis index and (f) α-Diversity (Shannon index) of content samples and of intestine samples of 16 S rRNA gene microbiome sequencing from small intestine of Nrp1^ΔIEC mice versus floxed WT littermates. (g) Descriptive statistics (P values) of the bacterial 16 S rRNA gene sequencing analysis on Nrp1^ΔIEC mice versus floxed WT littermates. Differences in Proteobacteria in intestine samples are highlighted by a black box. (h) Relative gene expression of Gli1, Ptch1 and Hhip in the distal small intestine of vehicle controls compared to GDC-0449-treated mice (Gli1: n = 7 versus 8, P < 0.0001; Ptch1: n = 7 versus 7, P = 0.0001; Hhip: n = 7 versus 7, P = 0.0008). (i,j) NRP1 expression in (i) the distal small intestine (n = 7 versus 7) and (j) isolated IEC (n = 7 versus 7) from vehicle controls compared to GDC-0449-treated mice. For panels c, h-j, n represents the number of biological independent mice. For the qPCR assays, L32 was used as the housekeeping gene, whereas in western blot, protein expression is relative to α-actinin or β-actin. In panels c and h-j the values were normalized for the mean expression of the control group. For panels c, h-j, individual values are shown as dots, while mean ± s.e.m. is shown as a column and error bar. For panel d-g, individual values are not shown. For panel c, h-j, unpaired Student’s t test was used. **P < 0.01, ***P < 0.001, ****P < 0.0001. Source data

Extended Data Fig. 4. VEGF-A expression in the small intestine of the Nrp1^ΔIEC mouse line and characterization of blood capillaries and lacteals in the distal small intestine of GDC-0449-treated mice.

(a) VEGF-A ELISA in the distal small intestine (n = 7 versus 7) and IECs (n = 5 versus 8) isolated from WT littermates versus Nrp1^ΔIEC mice. (b) Representative immunofluorescence images of VEGF-A expression (in green) in the distal small intestine of Nrp1^ΔIEC mice versus WT littermates. The stainings were repeated on n = 4 mice. For panels c-i, GDC-0449-treated mice are compared to vehicle controls. (c) Representative immunofluorescence images of PECAM-1 (CD31) expression (in green) in the distal small intestine of mice (n = 12 versus 11). (d) Quantification of CD31-positive area per villus structure (ROI: region of interest), representing vessel density (n = 12 versus 11). (e) Relative mRNA expression of Pecam1 in the distal small intestine (n = 12 versus 11). L32 was used as the housekeeping gene. (f) Representative immunofluorescence images of CD31 (green) and LYVE-1 (red) in the distal small intestine of mice (n = 10 versus 10). Measurements of (g) villus length (n = 10 versus 10), (h) lacteal length (n = 5 versus 5) and (i) the lacteal-to-villus ratio (n = 5 versus 5). For panels a, d-e, g-i, n represents the number of biological independent mice and individual values are shown as dots. Mean ± s.e.m. is shown as a column and error bar. For immunofluorescence images, cell nuclei are counterstained with To-Pro-3 iodide (in blue). Scale bar: 100 μm in panel b and 200 μm in panels c, f. For each mouse, the mean measurements of 5–10 villi were taken into account and displayed as a single dot. Unpaired Student’s t test, ns: P > 0.05. Source data