Fiber Supplementation Reduces Arthritis Symptoms

Propionate-induced microbial metabolites transfer pro-resolving effects. Prophylactic nutritional fiber or propionate supplementation was shown to be effective in RA and MS animal models (29, 30). To study potential therapeutic effects, we supplemented the drinking water of CIA mice from the peak of disease at 30 days after immunization (dpi) with 150 mM propionate (C3) or high-fiber supplementation. Twenty days after C3 or high-fiber supplementation, CIA mice showed significantly improved clinical signs of arthritis (Figure 1A and Supplemental Figure 1, A–C; supplemental material available online with this article; https://doi.org/10.1172/JCI184697DS1), reduced splenic Th17 cells (Figure 1B and Supplemental Figure 1D), and restored systemic bone density compared with respective nontreated controls (Figure 1C). C3 treatment increased total SCFA concentrations in the intestine (Figure 1D) and reestablished a gut microbial composition similar to the preclinical phase in healthy controls by reducing previously identified arthritis-related species such as from the Akkermansiaceae (31) and Enterobacteriaceae (32) family while increasing the Shannon index over non-C3-treated CIA control mice (Figure 1, E and F). To address whether C3-modulated microbiota contributes to this finding, we established a fecal microbiota transfer (FMT) model in CIA mice (Figure 1G). To that end, FM donor DBA/1 mice were supplemented with C3 in their drinking water for 3 weeks. FM (C3 FM) was harvested and further divided into FM pellet (C3 pellet) and FM supernatant (C3 supernatant) after centrifugation. Similar to therapeutic C3 supplementation in drinking water, C3 FMT, when done at the peak of arthritis, promoted fast resolution compared with transfer of control FM (Figure 1H). When comparing the efficacy of the two treatments from the peak of the disease, the administration of C3 FM resulted in a significant improvement in arthritis scores after only 5 days, whereas the supplementation of C3 in drinking water only led to a significant improvement after 20 days. Within the 3 FMT groups, only C3 supernatant was as effective as complete C3 FM, whereas the C3 pellet fraction showed no statistical clinical improvement (Figure 1H). This finding was reflected in the flow cytometry analyses of spleen cells, which showed that splenic Th17 cells decreased fast after C3 FM treatment as well as after supplementation of C3 in the drinking water (Supplemental Figure 1E). Comparison of 16s rRNA bacterial community profiles after C3 FMT revealed clear differences in the β-diversity 5 days after transfer over FM controls (Figure 1I) and an enrichment in Lactobacillaceae accompanied by lower levels of Lachnospiraceae (Figure 1, J and K). 16s rRNA analysis revealed further enriched Lactobacillaceae in C3 supernatant over the C3 pellet group (Supplemental Figure 1F). The effectiveness of Lactobacillaceae as probiotics, especially for Lactobacillus johnsonii, which was increased in our experiments, has been shown in a variety of inflammatory diseases (33–35).

Propionate-induced microbial metabolites transfer pro-resolving effects. (A) Clinical arthritis score shown as paw thickness (mm) of CIA mice ± 150 mM C3 (n = 4–5) in drinking water starting at 30 dpi. (B) Flow cytometric analysis of IL-17+ CD4+ T cells in the spleen of healthy mice and CIA mice ± C3. (C) quantification of bone mass (BV/TV) and representative μCT images of the trabecular part of tibial bone of healthy mice and CIA mice ± C3, analyzed at 77 dpi. (D) SCFA levels in the cecum content of CIA mice ± C3. (E) 16s rRNA-Seq of the cecum content of healthy mice, CIA mice, and CIA mice ± C3. (F) Alpha-diversity measure of 16s rRNA-Seq data. (G) Experimental layout of FMT experiment. This overview was generated with BioRender. (H) Clinical arthritis score shown as paw thickness (mm) of CIA mice treated with FMT of naive donors, C3-treated donors, pellet of C3-treated donors, or supernatant of C3-treated donors. (I) PCoA plot of 16s rRNA-Seq of mice after control or C3 FMT. (J) Relative abundance of the bacterial families identified by 16s rRNA-Seq. Permutational multivariate analysis of variance (ADONIS) was significant (R2 = 0.27225, P = 0.0167). (K) Relative abundance of most strongly changed Lactobacillaceae strains. (L) Volcano plot of untargeted metabolomics analysis of stool supernatant fraction obtained from control versus C3-treated FMT donors (cutoffs: P < 0.05 and –1 < log2FC < 1). Data are expressed as mean ± SD. Statistical difference was determined by ADONIS (L), 1-way ANOVA (B, C, F), Student’s t test (D), and AUC (A and H). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. CIA, collagen-induced arthritis; C3, propionate; BV, bone volume; TV, tissue volume; FMT, fecal microbiota transfer; PCoA, principal coordinates analysis.

Because C3 supernatant FMT was equally effective as complete C3 FMT, we next performed untargeted metabolic analyses to identify possible effector molecules induced by C3 nutritional supplementation. Volcano plots of C3 supernatant FM identified significant upregulated and downregulated metabolites over FM control supernatant (Figure 1L). Next, to identify responsible effector molecules in C3 supernatant FM, size exclusion chromatography (SEC) was used to separate the supernatant into 5 subfractions with different ranges of molecular size. Strikingly, only C3 supernatant FM fraction number 1, containing only the smallest molecules, showed similar pro-resolving effects as unfractionated C3 supernatant FM (Supplemental Figure 1, G and H). Untargeted metabolomics analysis of the nonfractioned C3 supernatant identified histamine among the most upregulated metabolites, which is also part of the pro-resolving fraction 1, due to its small molecular weight of 111.15 Da (Figure 1K). Another study also showed a tendency for histamine levels to increase in the colon after C3 treatment (36). Taken together, all these observations suggest potent peripheral pro-resolving effector functions of intestinal histamine that is locally increased in the intestine after high-fiber or C3 dietary supplementation in mice.

Resolution of arthritis depends on intestinal histamine and H3R signaling. We then attempted to identify the metabolite responsible for the prompt pro-resolving effect to continue with targeted mechanistic analyses. Therefore, CIA mice (30 dpi) were orally treated with histamine at concentrations as identified in histamine ELISA of stool extracts from the samples from patients with MS or RA (175.2 nM ± 210.8) (Supplemental Figure 2A). Oral treatment with 125 nM histamine at the peak of disease rapidly improved clinical arthritis scores in CIA mice within 5 days of treatment (Figure 2A). Histological analysis of hind paws confirmed this observation, showing reduced inflammatory lesions in the joints of histamine-treated CIA mice (Figure 2B). Multiplex cytokine and chemokine immunoassays of respective sera specifically revealed differences in CCL5 and CCL2 immune cell chemotactic cytokines, whereas other mediators remained unchanged (Figure 2C and Supplemental Figure 2, B–S). Further, β-diversity analysis together with taxonomic profiling of the bacterial community following 16s rRNA amplicon sequencing identified changed intestinal microbiota compositions after oral histamine treatment, again with an increased relative abundance of Lactobacillaceae (37) (Figure 2, D–F). Furthermore, members of this family, such as Lactobacillus reuteri, have been described as histamine producers in the intestine. Histamine produced by Lactobacillus inhibited proinflammatory cytokine production (38). This led us to investigate whether microbial-secreted histamine would have similar pro-resolving effects as orally supplemented histamine on synovial inflammation. Recently, Barcik et al. (39) developed an E. coli BL21 strain that was genetically modified to express the Morganella morganii–derived HDC (E. coli HDC+), which is responsible for catalyzing the decarboxylation of histidine to histamine. Oral transfer of E. coli HDC+ at the peak of disease significantly reduced arthritis in CIA mice compared with E. coli–treated or untreated CIA control mice (Figure 2G and Supplemental Figure 2, T–V).

Resolution depends on intestinal histamine and H3R signaling. (A) Clinical arthritis score shown as paw thickness (mm) of CIA mice ± histamine (n = 12–14) at the peak of disease. (B) Area of inflammation expressed as absolute mm2 per analyzed H&E-stained paw sections of CIA mice ± histamine (n = 5) and example histology images. Scale bars: 500 μm. (C) Serum levels of RANTES/CCL5 and MCP-1/CCL2. (D) PCoA plot of 16s rRNA-Seq of CIA mice ± histamine. Permutational multivariate analysis of variance (ADONIS) was significant (R2 = 0.18944, P = 0.0437). (E) Relative abundance of the bacterial families identified by 16s rRNA-Seq. (F) Relative abundance of most changed Lactobacillaceae strains after histamine treatment. (G) Clinical arthritis score shown as paw thickness (mm) of CIA mice after transfer of PBS (control), E. coli, or HDC positive E. coli at the peak of disease (n = 4–5). (H) Clinical arthritis score shown as paw thickness (mm) of CIA mice after oral transfer of histamine or specific agonists for H1R–H4R. (n = 4–5). (I) t-SNE plot of spectral flow cytometric analysis of the spleen from CIA mice plus H3R agonist RαMH. (J) CD4+ T cells in the synovium. (K) RORγt+ CD4+ T cells in the synovium. (L) RORγt+ CD4+ T cells in the pLN. (M) Neutrophils in the pLN. (N) CD8+ T cells in the pLN. Data are expressed as mean ± SD. Statistical difference was determined by Student’s t test (B, C, and J–N), t test or 1-way ANOVA of AUC (A, G, and H), and ADONIS (D). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. CIA, collagen-induced arthritis; pLN, popliteal lymph node; PCoA, principal coordinates analysis.

Histamine was shown to exert its effects via 4 histamine receptors, H1R–H4R (40). These receptors differ in their affinity for histamine, with H1R and H2R having a lower affinity and H3R and H4R having a higher affinity (41). Oral treatment of CIA mice with selective H1R–H4R agonists at the peak of disease (30 dpi) identified that only the H3R agonist replicated the strong pro-resolving effects of histamine itself (Figure 2H). This pro-resolving effect was independent from the type of H3R agonist used, as both R(-)-alpha-methylhistamine dihydrochloride (RαMH) (42) and immethridine dihydrobromide (43) showed similar pro-resolving effects (Supplemental Figure 3A). To confirm that only a site-specific increase of histamine in the intestine was initiating the resolution of arthritis, we compared i.p. and intrathecal to oral treatment with the H3R agonist in CIA mice (Supplemental Figure 3, A and B). Only oral delivery showed significant pro-resolving effects on arthritis along with increased intestinal length (Supplemental Figure 3C) as an indicator for reduced intestinal inflammation. Moreover, with increasing H3R agonist concentrations, pro-resolving effects disappeared and even exacerbated clinical arthritis scores (Supplemental Figure 3D). Although t-distributed stochastic neighbor embedding (t-SNE) plots of flow cytometry analysis in the spleen did not show differences in cell clustering with H3R agonist treatment (Figure 2I), CD4+ T cells and Th17 cells decreased in the synovial tissue (Figure 2, J and K), and Th17 cells, neutrophils, and CD8+ T cells decreased in the draining popliteal lymph node (pLN) after H3R agonist treatment (Figure 2, L–N). As previously shown by others and us, a key mechanistic contributor to the antiinflammatory effect of C3 is the induction of Tregs (44, 30). However, H3R agonist treatment did not induce Tregs (Supplemental Figure 3E). Therefore, the pro-resolving mechanism of C3-induced intestinal histamine is independent of the classical C3-associated Treg induction. Together, these results demonstrated that low-level histamine concentrations in the intestine promote the resolution of arthritis via local H3R signaling.

H3R agonist stimulates the enteric nervous system in arthritic mice, causing antiinflammatory responses. Our results showed that low-dosage, bacteria-derived histamine induced resolution of inflammation via specific H3R activation in the intestinal tract. H3R is mostly expressed on cells of the nervous system. In the intestine specifically, expression is interestingly limited to cells of the enteric nervous system (ENS) (45) and a very low number of endocrine cells (22). To assess H3R-induced changes in the ENS, we applied an ex vivo gut organ culture system (46) that maintains tissue architecture, yet allows tight experimental control, to perform whole-mount staining on the myenteric plexus after H3R agonist stimulation (Figure 3A). The c-fos gene is induced by a broad range of stimuli and has been commonly used as a reliable marker for neuronal activity (47). Histological MFI quantification in WT ex vivo gut organ cultures after H3R agonist (RαMH) treatment revealed increased cFOS nuclear localization in myenteric neurons positive for βIII-Tubulin (Tuj1) (Figure 3B), consistent with previously published results showing increased excitation of enteric neurons through H3R activation (48).

H3R agonist stimulates the ENS in arthritic mice, causing antiinflammatory responses. (A) Experimental layout of ex vivo intestinal organ culture system. This overview was generated with BioRender. (B) Normalized MFI of c-FOS in βIII-Tubulin+ enteric neurons after RαMH stimulation and example images. Scale bars: 50 μm. (C) Heatmap of the top 50 regulated genes identified by RNA-Seq of intestinal tissue of CIA mice after in vivo RαMH treatment. (D) Gene set enrichment analysis (GSEA) of intestinal RNA-Seq data. (E) Volcano plot of untargeted metabolomics data of serum of CIA mice ± RαMH (cutoffs: P < 0.05 and –1 < log2fc < 1). Data are expressed as mean ± SD. Statistical difference was determined by Student’s t test. ****P < 0.0001.

Bulk RNA-Seq experimental data from intestinal tissues of CIA mice treated with H3R agonist (RαMH) revealed significant differences in gene expression profiles over nontreated CIA mice (Figure 3C). Gene ontology analysis showed that oral H3R agonist treatment of CIA mice induced a prominent inflammation-suppressing phenotype in the intestinal tissue (Figure 3D). The most suppressed genes were associated with adaptive immune response (CD3g, Jchain, Btla, Tnfrsf17), lymphocyte-mediated immunity (Iglc2), and B cell–mediated immunity (IL7R, Igha) (Supplemental Figure 4). These findings were further supported by untargeted metabolomics analysis of serum metabolites (Figure 3E). Metabolites that were increased upon H3R stimulation were uridine and 3-hydroxybutyric acid, which are not only known for their antiinflammatory properties (49, 50) but also for their role in neuroprotection (51, 52) and suppression of microglia responses (53). Taken together, these data suggest that local H3R activation in the intestine induces an antiinflammatory milieu and increases the neuronal activity of enteric neurons.

Microglia depletion impairs pro-resolving effects of histamine. To formally address the contribution of the CNS in histamine-mediated resolution of arthritis, we next characterized cellular changes in the CNS, followed by respective in vivo cell depletion assays. Peripheral inflammation increases p38 phosphorylation in neurons and microglia, especially in the dorsal horn in the spinal cord (54, 55). As demonstrated by Boyle et al., local inhibition of p38 phosphorylation in the spinal cord suppresses arthritis (15). Analysis of the lumbar area (L3–L6) of the spinal cord (Figure 4A) that innervates the paws (56) after oral H3R agonist treatment in CIA mice showed reduced c-fos and p38 phosphorylation (Figure 4, B and C). Spectral flow cytometry multiparameter analysis of spinal cord single cells revealed significant cellular changes after H3R agonist treatment in CIA mice (Figure 4D), most prominent within the CD86+ microglial cell fraction that was restored to the levels observed in healthy control mice (Figure 4E). Further microglial characterization revealed a shift from inflammatory back to TMEM119+ homeostatic microglia (Figure 4F). Interestingly, when we treated microglia, astrocytes, or neurons directly with an H3R agonist in vitro, we did not observe lower levels of inflammatory gene expression or of CCL2 and CCL5 chemokine secretion (Supplemental Figure 5). When applying oral H3R agonist treatment around peak disease in the EAE animal model for MS, where microglial cells were shown to promote inflammation (57, 58), we observed significantly attenuated clinical scores and reduced inflammatory microglial cells in the spinal cord (Supplemental Figure 6), as seen in CIA mice. Next, we investigated the influence of microglia in CIA mice using the colony-stimulating factor 1 receptor inhibitor, PLX5622, to deplete the microglia population (58–60). Effective depletion was confirmed 3 days after the last PLX5622 treatment in CIA mice (Supplemental Figure 6G). After short-term PLX5622 treatment (25–30 dpi) in CIA mice, the prominent H3R agonist–mediated pro-resolving effects were lost when microglia were depleted (Figure 4G). Taken together, these results show that H3R activation in the intestine results in a phenotypic switch from proinflammatory to homeostatic microglial cells in the spinal cord, which are essential for the H3R-induced resolution of arthritis.

Microglia depletion impairs histamine’s pro-resolving effects. (A) Visual representation of the analyzed L3–L6 area of the spinal cord. This overview was generated using BioRender. (B) Quantification of phosphorylated p38 protein expression in the L3–L6 area of the spinal cord normalized on β-actin (Western blot). (C) Quantification of c-Fos+ cells in the spinal cord and example images of immunofluorescence-stained slides. (D) UMAP plot of spinal cord cells isolated from CIA ± RαMH and donut chart (% of live cells) of the different cell clusters analyzed by spectral flow cytometry. (E) CD86+ inflammatory microglia. (F) TMEM119+ homeostatic microglia. (G) Clinical arthritis score shown as paw thickness (mm) of CIA mice with and without microglia depletion (25–30 dpi) ± RαMH at the peak of disease. Data are expressed as mean ± SD. Statistical difference was determined by Student’s t test (B and C), 1-way ANOVA (E and F), and 1-way ANOVA of AUC (G). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Oral H3R agonist treatment influences microglia function in the spinal cord. Our finding that microglia are essential for the intestinal histamine-induced resolution of arthritis prompted us to analyze their transcriptomes in CIA mice after oral H3R agonist treatment. Therefore, CD11b+ spinal cord microglia were isolated from CIA mice after oral H3R treatment and processed for bulk RNA-Seq (Figure 5A and Supplemental Figure 6H). Microglia from H3R agonist–treated CIA mice showed a significantly altered gene expression profile, predominantly characterized by an increase of immune modulatory genes such as Fmr1nb, C4b, Spp1, and CD72 (61–64) (Figure 5B). Gene set enrichment analysis (GSEA) of hallmark genes revealed an overall antiinflammatory phenotype. Inflammatory pathways such as IL-6, JAK/STAT3, and TNF-α signaling were downregulated, whereas antiinflammatory pathways such as oxidative phosphorylation were upregulated (Figure 5C). Further, analysis of the microglial microenvironment by untargeted metabolomics in spinal cords of H3R agonist–treated CIA mice identified a significantly changed secreted metabolite pattern of increased thiamine, acetylcholine, and GABA in lumbar spinal cord tissue supernatants (Figure 5D). Of note, identical spinal cord metabolites were previously linked to attenuated clinical scores in RA (65, 66) by reducing p38 phosphorylation in the CNS (15, 67). Metabolite set enrichment analysis of identified spinal cord metabolites revealed aspartate and purine metabolism as well as arginine biosynthesis among the highest enriched metabolites (Figure 5E). L-arginine was previously shown to inhibit arthritis and associated inflammatory bone loss in mice (68). In addition, patients with RA exhibit lower purine metabolism activity (69), and a shortage of aspartate was shown to fuel synovial inflammation in RA (70). Together, these data identify oral histamine as a pro-resolving regulator of microglial gene expression and spinal cord metabolites.

Oral H3R agonist treatment reverses the proinflammatory environment of the CNS. (A) RNA-Seq experimental layout (generated using BioRender). (B) Heatmap of the top 50 regulated genes in CD11b+ cells of the spinal cord in CIA mice ± RαMH. (C) Gene set enrichment analysis of top 7 upregulated and downregulated hallmark pathways in the spinal cord. (D) Volcano plot of untargeted metabolomics data of spinal cord tissue supernatant of CIA mice ± RαMH (cutoffs were P < 0.05 and –0.3 < log2fc < 0.3). (E) The most upregulated metabolic pathways were identified using MetaboAnalyst.

Intestinal H3R signaling reduces vascular leakage in inflamed paws. Microglia sense neuronal activity and can directly modulate their functions (71). In RA models, sciatic nerve branches were shown to control vascular leakage in arthritic paws (72–74). Therefore, plantar nerve fibers were isolated from the peak of activity of CIA after oral H3R agonist treatment. Sort-purified CD11b– nerve cells were analyzed by bulk RNA-Seq (Figure 6A). The genes most differentially expressed included representatives of several signaling pathways critical for regulating cell and tight junction organization as well as GABA receptor activation and neurovascular coupling, as indicated by enriched pathway analysis (Figure 6B). Further analysis of upstream regulators for neurovascular coupling signaling identified nuclear receptor (NR) Nr4a3 and Mef2c as most significantly upregulated, both described as involved in vascular biology and microglial inflammatory responses (75, 76) (Figure 6C and Supplemental Figure 7, A and B). The smoothelin-like protein 1 (SMTNL1) upstream regulator was most reduced in plantar nerve CD11b– cells, similar to what was found in K/BxN mice after denervation of the sciatic nerve (74) (Figure 6C).

Intestinal H3R signaling reduces vascular leakage in inflamed paws. (A) Heatmap of the top 50 regulated genes in CD11b– cells of the nervus plantaris in CIA mice ± RαMH. (B) Most enriched pathways identified by IPA (QIAGEN). (C) Top 10 upstream regulators of the enriched neurovascular coupling pathway identified by IPA (QIAGEN). (D) Kep (transfer constant) in the paws of CIA mice ± RαMH on third treatment day. (E) Example MRI of T2 STIR, T1 POST KM, T1 POST KM, and map of parameter of CIA mice ± RαMH. (F) Raw signal intensity of DCE measurement over time of CIA mice ± RαMH. (G) Clinical arthritis score shown as paw thickness (mm) of CIA mice with or without QX-314 and bupivacaine-induced nerve blockage ± RαMH (n = 5) at the peak of disease. Data are expressed as mean ± SD. Statistical difference was determined by Student’s t test (D), Student’s t test of AUC (F), and 1-way ANOVA of AUC (G). *P < 0.05, **P < 0.01, ***P < 0.001.

To analyze actual changes in vasoconstriction and vasodilation in CIA paws, in vivo MRI was performed at 28 and 31 dpi, before and after oral H3R agonist treatment, respectively. Reduced Kep values, indicative of reduced vascular leakage, were found in H3R agonist–treated mice (Figure 6D and Supplemental Table 1), along with overall lower raw signal intensity after contrast agent application (Figure 6, E and F). Furthermore, the tendency to lower inflammatory area and paw thickness could be identified by magnetic resonance tomography as soon as the last H3R agonist intervention administration (Supplemental Figure 7, C and D). Also, a reduction of synovial CD4+ T cells was visible 8 days after the last RαMH treatment (Supplemental Figure 7E). To investigate the dependence between the microglia changes in the spinal cord and the vascular leakage in the joints, we used a pharmacological approach to temporarily shut down nerve activity (73). QX-314, a membrane-impermeable lidocaine derivative that must enter a cell to block sodium channel conductance, was locally injected in combination with bupivacaine in the footpad minutes before H3R agonist treatment on the peak activity days of CIA. QX-314 treatment did not directly affect paw thickness in CIA mice, but abrogated the antiinflammatory effects after oral H3R agonist administration (Figure 6G). Taken together, these data suggest that local intestinal H3R activation restores vascular leakage in the inflamed joints by affecting centrally controlled local nerve innervation.

Propionate supplementation increases local histamine levels in patients with RA or MS. To translate our findings to human disease, we utilized samples from 2 human studies, where patients with RA or MS were administered supplementation with the SCFA propionate over the time course of several weeks. In the ProDarMi study (German Clinical Trials Register ID: DRKS00023985), healthy participants were compared with patients with RA, psoriatic arthritis, psoriasis, ankylosing spondylitis, or at risk for RA. The publication of the initial clinical studies showed the beneficial effect of C3 or high-fiber supplementation on disease activity of the patients (77–80). To elucidate whether this effect of propionate was due to an increase in local histamine levels in the gut in the human setting, we analyzed stool samples from these patients at baseline and after 28 days (patients with RA) or 90 days (patients with MS) of C3 treatment. Interestingly, C3 supplementation over the course of 28 and 90 days significantly increased histamine levels in the stool samples of the patients (Figure 7, A and B). To further elucidate the role of microbiota-derived histamine, we reanalyzed RNA-Seq data from the ex vivo gut explant model utilized by Duscha et al. (77), where the gut explants were treated with the microbiota of patients before and after C3 supplementation. Here, we looked at genes that are associated with H1R and H2R activation. Interestingly, H1R- and H2R-associated genes (histamine response network) were downregulated after C3 treatment (Figure 7C), although local histamine levels were significantly increased. This observation supports our hypothesis that this low-level histamine derived from microbiota acts via H3R activation and does not induce allergy or intolerance-associated reactions mediated by H1R or H2R activation. Furthermore, histological analysis of H3R expression in ileal biopsies in patients with RA and healthy controls revealed low levels of H3R expression in the healthy controls, consistent with previously published data (45). In patients with RA, however, H3R expression in ileal tissues was strongly increased (Figure 7, D and E). To further strengthen the translational aspect of our study, we had the unique opportunity to retrospectively analyze FMT samples from the published FLORA trial (ClinicalTrials.gov NCT03058900) (81) within an ongoing collaboration. Our analysis revealed that FMT recipients who received stool samples with high histamine concentrations showed a 100% improvement in clinical response in a small sample group (Figure 7F). Furthermore, we found significant differences in fecal histamine concentrations among FMT recipients, with higher histamine levels associated with clinical improvement (Figure 7G). Taken together, these data indicate that our results in preclinical models may be applicable to human patients.

Propionate supplementation increases local histamine levels in patients with RA or MS. (A) Histamine levels in stool extracts from patients with RA before and after 28 days of propionate supplementation. (B) Histamine levels in stool extracts from patients with MS before and after 90 days of propionate supplementation. (C) Volcano plot of RNA-Seq data from ex vivo colon culture infused with stool samples from patients before and after 90 days of propionate supplementation. Overlay with pathway genes of H1R and H2R activation (GSEA pathway: pos histamine response network). (D) Example images of histology staining of ileum biopsies from healthy controls or patients with RA stained for H3R. (E) Quantification of histological H3R staining of ileum biopsies from patients and healthy controls. (F) Histamine levels in stool processed into fecal microbiota transplants (FMTs) from healthy donors (107). FMTs were given to patients with active arthritis despite ongoing treatment with a steady-state dose of methotrexate. After 26 weeks, patients were categorized as being either treatment failures or treatment responders, according to the European psoriatic arthritis recommendations (108). FMT products of donors 1, 2, 3, and 4 were given to 6, 4, 3, and 2 patients, respectively. (G) Histamine was measured from the stool of patients with psoriatic arthritis at baseline before FMT. Data are expressed as mean ± SD. Statistical difference was determined by paired t test (A and B) and Student’s t test (E). *P < 0.05, ***P < 0.001.

The concept that the nervous system senses environmental stimuli and transmits these signals to immune cells to maintain tissue homeostasis is well-established (82). However, during chronic arthritic inflammation, it has been shown that the nervous system can exert both proinflammatory and antiinflammatory functions. For example, noninvasive electrical vagus nerve stimulation was shown to be effective in attenuating arthritis in CIA mice and patients with RA (83, 84). On the other hand, denervation of the sciatic nerve protected mice from K/BxN serum transfer arthritis, and individuals with paralysis on one side of the body developed arthritis only on the neurologically unaffected contralateral side (74, 85).

So far, irrespective of the clinical effects, the regulation of arthritis by the CNS has only been investigated in a two-dimensional approach between the CNS and the joints. Here, we identified spinal cord microglia as an essential gut-joint interface responsible for transmitting the antiinflammatory pro-resolving effect of microbiota-derived histamine to the joints, thereby extending the neuroimmunomodulatory concept in RA to a third dimension, the gut.