C. difficile intoxicates neurons and pericytes to drive neurogenic inflammation

Abstract

Clostridioides difficile infection (CDI) is a major cause of healthcare-associated gastrointestinal infections^1,2. The exaggerated colonic inflammation caused by C. difficile toxins such as toxin B (TcdB) damages tissues and promotes C. difficile colonization^3-6, but how TcdB causes inflammation is unclear. Here we report that TcdB induces neurogenic inflammation by targeting gut-innervating afferent neurons and pericytes through receptors, including the Frizzled receptors (FZD1, FZD2 and FZD7) in neurons and chondroitin sulfate proteoglycan 4 (CSPG4) in pericytes. TcdB stimulates the secretion of the neuropeptides substance P (SP) and calcitonin gene-related peptide (CGRP) from neurons and pro-inflammatory cytokines from pericytes. Targeted delivery of the TcdB enzymatic domain, through fusion with a detoxified diphtheria toxin, into peptidergic sensory neurons that express exogeneous diphtheria toxin receptor (an approach we term toxogenetics) is sufficient to induce neurogenic inflammation and recapitulates major colonic histopathology associated with CDI. Conversely, mice lacking SP, CGRP or the SP receptor (neurokinin 1 receptor) show reduced pathology in both models of caecal TcdB injection and CDI. Blocking SP or CGRP signalling reduces tissue damage and C. difficile burden in mice infected with a standard C. difficile strain or with hypervirulent strains expressing the TcdB2 variant. Thus, targeting neurogenic inflammation provides a host-oriented therapeutic approach for treating CDI.

Extended Data Fig. 1:. Histopathology score analysis and time course of TcdB-induced inflammation in a mouse cecum injection model.

a, A list of criteria for assessing histopathology scores of colonic tissues. b-e, H&E micrographs illustrating oedema (b, black line indicates oedema extent); epithelial disruption (c); haemorrhage or congestion (d); immune cell infiltration (e). f, Representative histopathology of human CDI colon tissues compared to a normal control, illustrating neutrophilia, oedematous inflammation, and pseudomembranes. g-j, The subcategories for the histopathology scoring in Fig. 1b: oedema (g), immune cell infiltration (h), epithelial disruption (i), and haemorrhage or congestion (j). n = 5, 6, 9, 5 mice. k-l, SP (k) and CGRP (l) levels in colonic explants from the mouse cecal TcdB injection model at the indicated time (min) following TcdB injection. Vehicle was at 120 min. n = 6, 7, 7, 7, 7 mice. m-n, The levels of SP (m) and CGRP (n) in colonic explants from CDI mouse models infected with either C. difficile 630 or a toxin-null strain (630 A- B-) at 23 h (n = 5), 37 h (n= 5), and 48 h (n = 3), normalized to mice treated with antibiotics but gavaged with saline (time point 0, n= 5). g-j, assessed by one-way ANOVA with post hoc Dunnett’s test. k-l, assessed by Kruskall-Wallis test with post-hoc Dunn’s test. Centre line, mean; error bars reflect the standard error of the mean (SEM); n.s., not significant, p values are exact. Scale bar is 50 μm in b, c and 20 μm in d and e. n is described from left to right.

Extended Data Fig. 2:. Histological scoring for cecum injections of TcdB in the indicated mouse models as described in Fig. 1c-d.

a-e, Histopathology total scores and subscores of TcdB injected Tac1 KO mice, n = 7, 8 mice. f-j, Histopathology total scores and subscores of TcdB injected Nk1r KO mice, n = 7, 7 mice. k-o, Histopathology total scores and subscores of TcdB injected Calca KO mice, n = 8, 10 mice. p-t, Histopathology total scores and subscores of TcdB injected Calcb KO mice, n = 3, 5 mice. u, Representative histopathology of Tac1 KO mice injected with TcdB compared to Tac1 KO mice injected with TcdB and SP in the cecal TcdB injection model (120 min incubation). v, Histopathology scores of experiments described in u, n = 3, 4 mice. Assessed by Student’s T-test, two tailed, relative to WT control mice. Centre line: mean; error bars: SEM; n.s., not significant, p values are exact. n is described from left to right.

Extended Data Fig. 3:. Histological scoring for CDI in the indicated KO mouse models as described in Fig. 1e-f.

a-e, Histopathology total scores and subscores of CDI in Tac1 KO mice, n = 11, 20 mice. f-j, Histopathology total scores and subscores of CDI in Nk1r KO mice, n = 7, 10 mice. k-o, Histopathology total scores and subscores of CDI in Calcb KO mice, n = 6, 9, 7 mice. a-j, assessed by Student’s T-test, two tailed, relative to WT control mice. k-o, assessed by one-way ANOVA with post hoc Dunnett’s test. Centre line: mean; error bars: SEM; n.s., not significant, p values are exact. n is described from left to right.

Extended Data Fig. 4:. FZD1/2/7 are expressed in DRG and enteric neurons.

a, Single cell RNA-seq analysis of FZD1, 2, and 7 expression in different cell types within colonic tissues in mice. Size of dot indicates proportion expressing whereas the colour indicates expression level (red high, blue low). Putative excitatory motor neuron (PEMN), putative inhibitory motor neuron (PIMN), putative sensory neuron (PSN), putative sudomotor and vasodilator neurons (PSVN), putative interneuron (PIN). b, FZD1, 2, and 7 expression in DRG neurons that innervate colonic tissues are displayed as violin plots. c, In situ hybridization of spinal cord sections shows enriched expression of FZD1/2/7 in dorsal root ganglia areas. Data are from Allen Brain Atlas. d, Phase-contrast images show that DRG neurons exposed to TcdB (1 pM, 60 min) exhibit no morphological changes. Scale bar is 50 μm. n=3 independent replicates. e, RNAscope in situ hybridization showing localization of Pdgfrb expression (magenta) within Cspg4⁺ cells (cyan, Cspg4-dsRed mice) in colonic tissues. Panels in bottom row show regions in yellow boxed area at higher magnification. Scale bar: 50 μm and lower panel is 20 μm.

Extended Data Fig. 5:. Cspg4⁺ pericytes wrap around blood vessels and form close contact with CGRP-containing afferent neurons.

a, Colon tissues from Cspg4-DsRed mice were harvested and subjected to whole mount immunofluorescence analysis, using an antibody against CD31 (a marker for blood vessel endothelial cells; magenta). DsRed (cyan) signals were detected directly. DsRed-positive cells surround CD31-positive capillaries in colonic crypts. Maximum intensity projection of the whole stack is presented in Figure 2e. Scale bar is 50 μm. b, Experiments were carried out as in panel a, except that lymphatic endothelial cells were labeled with an antibody against Lyve-1 (magenta) and DsRed signals were detected using an antibody against DsRed. DsRed-positive cells are clearly separate from lymphatic endothelial cells. Scale bar is 50 μm. c-f, Experiments were carried out as described in panel a. Enteric glial cells were visualized with GFP fluorescence in examining proteolipid protein 1 (PLP1)-GFP/-Cspg4-dsRed dual reporter mice (panel c). Fibroblasts were labeled using an antibody against platelet-derived growth factor receptor alpha (PGDFRA, magenta, panel d). Myofibroblasts were labeled using an antibody against smooth muscle actin (SMA, magenta, panel e). Mast cells and interstitial cells of Cajal were labeled using an antibody against cKit (magenta, panel f). Scale bar: 10 μm. g, TcdB induced cell-rounding of cultured human brain vascular pericytes. Scale bar: 25 μm h, The indicated TcdB mutants were injected into mouse ears via intradermal injection. Pericytes surrounding ear arterioles were labeled and visualized through DsRed. TcdB and TcdB-FzM induced morphological changes of Cspg4⁺ pericytes surrounding ear arterioles, whereas TcdB-Cspg4M showed no effect. Scale bar, 50 μm. n= 3 mice/group. i, Experiments were carried out as described in panel h, except that WT mice were utilized and pericytes were detected by immunostaining using an antibody against smooth muscle actin (labeling pericytes). TcdB-FzM disrupted pericytes around ear arterioles, whereas TcdB-Cspg4M has no effect. Scale bar: 50 μm. n= 3 mice/group. j, Experiments were carried out as in panel a, except that neuronal processes were labeled using an antibody against Tubb3 (magenta, a marker for neuronal processes), showing that neuronal processes are extending alongside DsRed-positive pericytes. Scale bar: 10 μm. k, Experiments were carried out as in panel a, except that an antibody against CGRP was added to detect CGRP-positive nerve terminals. 3-dimensional reconstruction of images showed that DsRed⁺ pericytes (cyan) surround the vasculature (CD31 endothelial marker; magenta) and contact CGRP-expressing nerve terminals (CGRP; yellow). Scale bar: 10 μm. l, Immunoblot analysis of cell lysates showed expression of CGRP receptors (CALCRL: calcitonin receptor like receptor; and RAMP1: receptor activity modifying protein 1) in primary cultured human brain vascular pericytes. Two human cell lines, HeLa and U87, were analyzed in parallel as controls, which do not express detectable levels of CGRP receptors. Total protein staining with Coomassie blue was used as a control for protein loading. For gel source data, see Supplementary Figure 1. N = 2 replicates.

Extended Data Fig. 6:. AAV-DTR mediates selective expression of DTR in ENS and DRG neurons in Tac1-Cre mice.

a, AAV-DTR was administered to Tac1-Cre mice via tail vein injection. Colon tissues were harvested 3-weeks later and subjected to whole mount immunofluorescence analysis. DRG were harvested and their cryosections were analyzed. HuC/D (magenta) marks neuron soma in colon tissues. An antibody against β3-tubulin (TUBB3, magenta) was utilized to label DRG neurons. DTR-GFP was labeled using an antibody against GFP (cyan). Scale for DRG: 100 μm; for myenteric plexus: 50 μm. Images were collected using a confocal microscope and maximal intensity projection is presented. GFP is directly fused to the C-terminus of DTR, thus cyan color represents DTR expression. n = 4 mice. b, Experiments were carried out as described in upper panel. 5-HT (magenta) labels enterochromaffin cells in colonic tissues, showing no co-expression of DTR-GFP (cyan). Scale bars represent 100 μm (50 μm for right panel, which shows boxed region at higher magnification). c-d, Experiments were carried out as described in Fig. 3e. Colonic tissues (panel c) and DRG (panel d) were isolated and subjected to analysis as described in panel a. WT mice not exposed to AAV-DTR were analyzed in parallel as a negative control (right panels). HuC/D labels myenteric plexus neurons within the proximal colon and TUBB3 labels DRG neurons of a wildtype mouse injected with AAV-DTR. No GFP positive cell bodies or fibers were detected in WT mice. Scale bar is 50 μm. e-h, Sub-scoring of histopathology, related to Fig. 3f, n = 7, 6, 4 mice. One-way ANOVAs were applied with post hoc Tukey’s test as indicated. Centre line: mean; error bars: SEM; n.s., not significant, p values are exact. n is described from left to right.

Extended Data Fig. 7:. Extrinsic afferent neuron-dependent secretion of SP is necessary and sufficient for TcdB mediated pathology.

a-d; Histopathology subscoring of experiments following TcdB administration to RTX treated mice. Related to Fig. 3h, n = 10, 10 mice. e-j, Histopathology of RTX treated/vehicle mice treated with TcdB alone or TcdB and SP for 120 min, demonstrating that SP can restore the effects of TcdB in RTX-treated mice. Representative histopathology in e, and histopathology scores f-j, n = 4, 4, 6, and 6 mice. The scale bar is 50 μm. k, CGRP secretion from RTX treated mice is reduced in the cecal TcdB injection model, and is restored by co-administration of SP with TcdB. n = 2, 4, 6, 4, and 6 mice. l-m, Nav1.8^Cre/+ Rosa-LSL-Tdtomato mice have no evident Tomato expression (magenta) in the ENS (myenteric neurons marked with HuC/D, cyan) with only sparse Tomato⁺ nerve fibers from extrinsic neurons labeled in the colon (l). In contrast, there are many Tomato⁺ neurons in DRG (m). Scale bar: 50 μm. n-q, Histopathology subscores of Nav1.8^Cre/+, Rosa-LSL-DTR mice treated with GTD-iDT related to Fig 3i-j; n = 7, 9 mice. For a-d and n-q, significance was assessed by student’s two-tailed t-test. For f-k, significance was assessed by one-way ANOVA with post hoc Dunnett’s test. Centre line: mean; error bars: SEM. p values are exact. n is described from left to right.

Extended Data Fig. 8:. Inhibiting SP or CGRP signaling reduces severity of C. difficile colitis.

a, CDI models were carried out on WT versus Tac1 KO mice, using either C. difficile 630 Δerm or a toxin-null mutant strain (A-/B-). C. difficile CFUs from faeces were analyzed at 2 days post infection. The toxin-null strain showed reduced colonization on WT mice compared with the standard 630 Δerm, suggesting that the presence of toxins facilitates C. difficile colonization. In contrast, 630 Δerm showed similar levels of CFUs as the toxin-null strain in Tac1 KO mice, suggesting that the toxin’s contribution to C. difficile colonization relies on intact SP signaling. n= 7, 6, 4, 3 mice. b, CDI models were carried out on WT versus Calcb KO mice (using C. difficile 630 Δerm). C. difficile CFUs from faeces showed reduced colonization in Calcb KO mice compared with WT mice, n= 4, 5 mice. c, Representative H&E images for the experiments described in Fig. 4c. Scale bar is 50 μm. d-h, Sub-categories of histopathological scoring of Fig. 4c; n = 15, 12, 11, 3. h-l, Histopathology scoring of mice administered vehicle or aprepitant at the indicated times (0, 30 or 60 min following cecum injection and wound closure); n = 7, 6, 6, 7 mice. Significance was assessed for a – Two-Way ANOVA with post hoc Sidak’s test and by T-test for b. One Way ANOVA with post hoc Dunnett’s test was used for d-l. Centre line: mean; error bars: SEM. p values are exact. n is described from left to right.

Extended Data Fig. 9:. Inhibiting SP or CGRP reduces histopathological scores in cecum injection assays and in CDI models.

a, Representative H&E images for the experiment described in Fig. 4d. Scale bar is 50 μm. b-e, Sub-categories of histopathological scoring of Fig. 4d, n = 13,7, 8, 8 mice. f, C. difficile 630 Δerm fitness (measured by OD600) was not affected by treatment with aprepitant or olcegepant (20 μM) in vitro. Representative of 2 independent experiments (3 wells/group). g-h, CDI experiments (with C. difficile 630) were carried out on WT and Tac1 KO mice (co-housed from weaning), followed by treatment with vehicle or aprepitant. C. difficile colonization (panel g) and histopathological scores (panel h) were shown, n = 7, 7, 6, 7 mice. i-j, CDI experiments (with C. difficile 630) were carried out on WT and Calcb KO mice (littermates, co-housed from birth), followed by treatment with vehicle or olcegepant. C. difficile colonization (panel i) and histopathological scores (panel j) were shown. n = 4, 4, 5, 7 mice. k, TcdA induced oedema in footpad injection assays in mice, which is prevented by pre-treatment with aprepitant. Boiled TcdA solution did not induce any oedema. n = 7,4, 3 mice. Significance was assessed by one-way ANOVA with post hoc Dunnett’s test. Centre line: mean; error bars: SEM. n.s., not significant, p values are exact. n is described from left to right.

Extended Data Fig. 10:. Inhibiting SP or CGRP reduces histopathological scores and colonization of hypervirulent strains that express TcdB2.

a, Representative H&E images for the experiments described in Fig. 4i. Scale bar is 50 μm. b-e, Sub-categories of histopathological scoring of Fig. 4i, n = 3, 9, 7 mice. f, Representative H&E images for the experiment described in Fig. 4j. Scale bar is 50 μm. g-j, Sub-categories of histopathological scoring of Fig. 4j, n = 21, 10, 10, 12, 10, 5 mice. Significance was assessed by one-way ANOVA with post hoc Dunnett’s test. Centre line: mean; error bars: SEM; n.s., not significant; p values are exact. n is described from left to right.

Fig. 1:. SP- and CGRP-mediated neurogenic inflammation in a TcdB cecum injection model and in a CDI mouse model.

a-b, Time course of histopathological changes after injection of TcdB into the cecums of mice. Representative micrographs of Haemotoxylin and Eosin (H&E) stained tissues are shown in a and histopathology scores are shown in b. Scoring criteria and the sub-categories of histopathology scores are shown in Extended Data Fig. 1; n=5, 6, 9, 5 mice. c-d, TcdB was injected into the cecums of wild type (WT) mice or the following KO mice: Tac1 (encoding the precursor for SP), Nk1r (encoding a receptor for SP), Calca (encoding CGRPα), or Calcb (encoding CGRPβ), with representative H&E staining shown in c and histopathology scores shown in d; n=25, 8, 7, 10, 5 mice. e-f, CDI experiments were carried out on WT, Tac1 KO, Nk1r KO, and Calcb KO mice. Colon tissues were harvested and analyzed 48 h later, with representative H&E staining shown in e and histopathology scores shown in f, n=24, 20, 10, 7 mice. Statistical analyses in b (p<0.001 overall), d (p<0.001 overall), and f were one-way ANOVAs with post-hoc Dunnett’s test compared with vehicle-treated mice. P values: n.s., not significant, or exact values shown if <0.05. Centre line indicates mean and error bars reflect the standard error of the mean (SEM). Scale bar: 50 μm. “n” is described from left to right for each graph.

Fig. 2:. TcdB targets neurons via FZD1/2/7 and pericytes via CSPG4.

a, Single cell expression analysis of FZD1, 2 and 7 in human colon tissues (red high, blue low). b, Cultured mouse DRG neurons were exposed to TcdB or TcdB mutants for 1 h. n= 10, 7, 7, 7, 3, 3 culture wells across 3 cultures (each from 5 pooled animals). SP secretion was assessed using ELISA. FzM: TcdB mutant deficient in binding to FZD1/2/7; GtdM: TcdB mutant with mutations in the GTD domain. c, Mouse cecal explants were exposed to TcdB (6 μg, 6 h), TcdB^FzM, or saline, and secretion of SP was measured. n=6,6,5 mice. d, Single cell analysis of CSPG4 expression in human colonic stroma. e-f, Immunostaining of cecum tissues from Cspg4-DsRed mice shows Cspg4⁺ cells (DsRed, cyan) surrounded by CD31⁺ vessels (magenta). e: Whole mount muscularis view, f: cross section. Scale bar: 50μm. g, Cultured human brain pericytes were exposed to TcdB or TcdB mutants for 5 h and cell rounding (%) was assessed. Cspg4M: TcdB mutant deficient in binding to CSPG4; FzCspg4M: a TcdB double mutant deficient in binding to FZD1/2/7 and CSPG4. n= 3 independent cultures. h, TcdB, but not TcdB-Cspg4M, induced secretion of IL-8 from cultured human pericytes. n=4,7,3,4 from 2 independent cultures. i-j, TcdB, but not TcdB-Cspg4M, increased CXCL1 in colonic explants after cecal TcdB injections in WT, but not Cspg4 KO mice. In i n= 11, 13, 8 mice, and in j n= 5, 8, 4, 4, 3 mice. k, Immunostaining of colon tissues from Cspg4-dsRed mice shows extensive overlap and contacts of neurons (Tubb3, magenta) with Cspg4⁺ cells (cyan). DAPI: yellow. Scale bar: 100 μm. i, Immunostaining of colon tissues from Cspg4-dsRed mice shows appositions of CGRP (yellow) to Cspg4⁺ cells (cyan). Blood vessels were labeled with CD31 (magenta). Scale bar: 10 μm. m, CGRP induced IL-8 release from cultured human pericytes. n=4 culture wells (representative of 3 replicates). Statistical analyses in b, c: one way ANOVA (<0.0001 overall) with post-hoc Dunnett’s test compared with vehicle-treated mice, n.s., not significant, p values <0.05 shown. Centre line: mean; error bars: SEM. n is described from left to right.

Fig. 3:. Delivery of TcdB-GTD into peptidergic sensory neurons using the toxogenetic approach is sufficient to induce neurogenic inflammation.

a, Mouse footpad injection assays were carried out for the indicated toxins or reagents. Footpad oedema was measured and shown as ΔThickness. N=4, 3, 8, 4,4,4,4,4 mice. b, Footpad injection assays were carried out on WT, Tac1^−/−, C-kit^W-sh (deficient in mast cells), and Cspg4^−/− mice. N=5, 3, 3, 3 mice. c, Schematic illustration of the toxogenetic approach. PNS: peripheral nervous system. ENS: enteric nervous system. DRG: dorsal root ganglion. d, DTR was expressed in Tac1⁺ peptidergic neurons through intravenous injection of AAV.PHP.s.FLEX.DTR:GFP virus into Tac1-Cre mice. Footpad injection assays were carried out. GTD-ciDT: TcdB-GTD fused to a detoxified diphtheria toxin (iDT); iGTD-ciDT: catalytic inactive form of TcdB-GTD fused to ciDT; n=5, 6, 3, 3 mice. e-f, Intraperitoneal administration of GTD-ciDT induced colonic tissue damage and inflammation in Tac1-Cre mice transduced with AAV.PHP.s.FLEX.DTR:GFP. Representative H&E staining of colonic tissues are shown in e, and histopathology scores are shown in f; n=7, 6, 4 mice. g-h, Ablating TRPV1⁺ sensory neurons using RTX reduced the levels of SP and histopathology scores in the TcdB cecal injection model; n=7, 10, 10 mice for g and n=10, 10 mice for h. i-j Systemic administration of GTD-ciDT induced colonic tissue damage and inflammation in Nav1.8^Cre/+ Rosa-LSL-DTR mice; n=7, 9 mice. Statistical tests in a (p<0.001 overall), b (p<0.001 overall), d (p<0.001), and f (p<0.0001) were one-way ANOVAs. Post-hoc Dunnett’s test compared with TcdB treatment (a) or WT treated with TcdB (b). Post-hoc Tukey’s tests as indicated in d and f. P values are exact or n.s.; Centre line: mean; error bars: SEM. Scale bars are 50 μm. n is described from left to right.

Fig. 4:. Inhibiting SP or CGRP signaling alleviates inflammation and reduces C. difficile colonization for endemic and hypervirulent epidemic strains.

a-b, CDI were carried out (630 Δerm strain), and colony forming units (CFUs) of C. difficile in faeces were assessed and plotted in a; n=5, 5 mice. Microbiome diversity was assessed and presented as Fisher’s alpha diversity in b; n=18, 16, 5, 10 mice. c, Cecum injection assays were carried out with TcdB on mice pretreated with vehicle, aprepitant, olcegepant, or fremanezumab; n=15, 12, 11, 3. d-e, CDI were carried out (630 Δerm) on mice pretreated with vehicle, aprepitant, olcegepant, or fremanezumab; n=13, 7, 8, 8 mice in d; n=3 (d1) and n=6 (d2) in e. f-g, TcdB2 induces oedema in WT and Cspg4^−/− mice in the footpad injection assay; n=3, 4, 4 mice. Aprepitant and olcegepant can prevent oedema induced by TcdB2. h.i. TcdB2: heat inactivated TcdB2, n=5, 5, 4 mice. h-i, CDI with the strain M7404 (TcdA-/TcdB2) of mice pre-treated with vehicle, aprepitant, or olcegepant. Survival is plotted in h, n=31, 28, 21. Histological scores are plotted in i, n=3, 9, 7 mice. j-k, CDI with the hypervirulent strain R20291 of mice pre-treated with the indicated inhibitor (Combination F+A: fremanezumab plus aprepitant; Combination O+A: olcegepant plus aprepitant). Histopathological scores are shown in j, n= 21, 10, 10, 12, 10, 5 mice; CFUs are shown in k, n= 21, 9, 10, 12, 10, 5 mice. Statistical analyses for a, b, and e: two-way ANOVA, variables are time and genotype/drug treatment, post-hoc Sidak’s test with comparisons by day. For c, d, f, g, i, j and k: one-way ANOVA (<0.0001 overall) with post-hoc Dunnett’s test compared with vehicle (c, d, g, i, j and k). For h, drugs were compared by Fisher’s test to vehicle controls. Centre line: mean; error bars: SEM; n.s., not significant; p values are exact. n is described from left to right.