Vagal stimulation ameliorates murine colitis by regulating SUMOylation

Abstract

Inflammatory bowel diseases (IBDs) are chronic debilitating conditions without cure that shorten the lifespans of seven million patients worldwide by nearly 10%, the etiologies of which are unknown. Here, we found that increased autonomic parasympathetic dysfunction resulted in increased IBD susceptibility and mortality in mouse models of disease. Conversely, vagal stimulation restored neuromodulation and ameliorated colitis by inhibiting the post-translational modification SUMOylation through a mechanism independent of the canonical interleukin-10/α7 nicotinic cholinergic vagal pathway. Colonic biopsies from patients with IBDs and mouse models showed an increase in small ubiquitin-like modifier (SUMO)2 and SUMO3 during active disease. In global genetic knockout mouse models, deletion of Sumo3 protected against development of colitis and delayed onset of disease, whereas deletion of Sumo1 halted the progression of colitis. Bone marrow transplant from Sumo1-KO but not Sumo3-KO mice into wild-type mice conferred protection against development of colitis. Electric stimulation of the cervical vagus nerve before the induction of colitis inhibited SUMOylation and delayed the onset of colitis in Sumo1-KO mice, and resulted in milder symptoms in Sumo3-KO mice. Treatment with TAK-981, a first-in-class inhibitor of the SUMO-activating enzyme, ameliorated disease in three murine models of IBD and reduced intestinal permeability and bacterial translocation in a severe model of the disease, suggesting the potential to reduce progression to sepsis. These results revealed a pathway of vagal neuromodulation that reprograms endogenous stress-adaptive responses through inhibition of SUMOylation and suggest SUMOylation as a therapeutic target for IBD.

Introduction

Inflammatory bowel diseases (IBDs) are idiopathic, chronic, debilitating conditions affecting nearly 7 million people worldwide. The two major IBDs include ulcerative colitis (UC) with continuous inflammation through the colon, and Crohn's disease (CD) with patchy inflammation in the whole digestive tract from the mouth to the anus. The Centers for Disease Control and Prevention estimates that IBD affects nearly 4 million Americans, with epidemiological studies reporting a 50% increase in IBD prevalence in the last 15 years (1, 2). There is no cure for IBD, and IBDs cause inflammation, colon perforation, increased risk of both sepsis and cancer, and can shorten patients’ lifespans by around 10% (3-5). Current anti-inflammatory therapies reduce the need for surgery, but patients lose responsiveness over time and experience side effects and relapses that impact their quality of life and health (6-8). Therefore, there is a need for alternative therapies to treat these diseases.

Psychological stress is a main factor in the six leading causes of death (9). A typical example is the implications of stress in IBDs, especially ulcerative colitis, which was originally regarded as a psychosomatic disorder (10, 11). Stress directly impacts immune homeostasis through the modulation of the nervous system. Previous studies suggested that stress may cause autonomic dysfunction affecting the parasympathetic system, and deficient vagal tone is increasingly recognized as a physiologic marker for stress (12-14). We reported that vagal and non-cholinergic sciatic nerve stimulations curb systemic inflammation and improve organ function in experimental sepsis (15-17). Bioelectronic medicine is an emerging field based on nerve stimulation to control organ function and restore immune homeostasis in infectious and inflammatory disorders. Several pilot studies of nerve stimulation in infectious and inflammatory disorders such as polyneuropathies (NCT04053127), heart failure (NCT03425422), rheumatoid arthritis, and traumatic brain injury (NCT02974959) are underway, and some have provided promising results (18, 19). Two pilot studies have reported that vagal stimulation can improve clinical outcomes in patients with CD (20-22). One study showed that one-year vagal stimulation treatment improved endoscopic scores in half of the 16 patients with CD who participated (22), and the other study reported remission in six out of nine patients with CD, but the other three patients experienced worsened of their disease (ClinicalTrials.gov Identifier: NCT01569503) (20, 21). These results highlight the need for mechanistic studies to better understand the reasons for patient non-response or worsening and to improve treatment efficacy while avoiding surgical nerve manipulation.

Although the pathogenesis of IBDs is currently unknown, we and others have reported SUMOylation as an emerging stress-adaptive response to safeguard intestinal integrity, a tissue particularly challenged by its bacterial environment (23-25). SUMOylation refers to the covalent, reversible conjugation of target proteins to small ubiquitin-like modifier (SUMO) proteins as a post-translational modifier to regulate protein interactions and functions (26). SUMOylation is critical for survival, and total inhibition of SUMOylation by genetic ablation of Ubc9 (ubiquitin-conjugating enzyme 9 that functions as the sole E2-conjugating enzyme required for SUMOylation) causes early embryonic death (27). SUMOylation is highly responsive to endogenous and environmental stressors and coordinates critical signaling pathways and transcriptional factors to reprogram immune cells at multiple levels (28, 29). A previous study reported reduced Ubc9 abundance in colonic endoscopic biopsies of patients with IBD, and specific Ubc9 knockdown in human epithelial cells inhibited master kinase Akt1 SUMOylation and resulted in increased inflammation (24). However, SUMOylation is a complex mechanism with three main SUMO isoforms expressed in multiple cell types that play different roles in IBD (24, 28-32). Here, we analyze the autonomic dysfunction induced by stress, its effects on acute colitis induced by dextran sulfate sodium (DSS), and whether electric stimulation of the cervical vagus nerve before the induction of colitis could improve colitis progression by regulating SUMOylation and reprogramming immune responses.

Results

Vagal stimulation ameliorates stress-induced autonomic dysfunction and reduces colitis-related mortality in mice.

Restraint stress induced by restraining the mice 6 hours daily for 21 days increased susceptibility to DSS colitis induction following the restrain period and turned colitis lethal with all mice dead by day 5 after the challenge (Fig. 1A). There were no differences in body weights between the groups, however stressed mice experienced more severe disease as measured by the multivariable colitis disease activity index (DAI) (Fig. 1B), mainly related to intestinal bleeding and stool consistency measures. Given that death prevented statistical group comparison at day 7 post-DSS challenge, we indicated the time of death and monitored DAI in control mice to show the disease progression time course (Fig. 1C). To determine whether DSS concentration influences the mice susceptibility to colitis we have tested 3% and 5% concentrations in non-stressed mice. We found no difference between DSS 3% and 5% in terms of stool consistency, intestinal bleeding, body weight loss, or DAI, nor did we see any difference with colon weight and length, organ adhesion, intestinal permeability, and colon cytokines between the two groups (Fig.S1, A to H). Then, we analyzed the three major networks of neuromodulation in stressed and control mice; hypothalamic-pituitary-adrenal (HPA), sympathetic, and cardio-vagal networks, at day 5 post-DSS challenge before the timepoint when the stressed mice are expected to die. We found no significant differences in plasma corticosterone concentrations (a marker of HPA axis) or norepinephrine (a marker for sympathetic adrenal axis). However, there was a 3-fold increase in the low frequency/high frequency (LF/HF) ratio, a measure of heart rate variability (HRV) and parasympathetic tone through cardio-vagal axis (Fig.1D). Because high LF/HF may reflect a parasympathetic deficiency, we analyzed whether electrical stimulation of the parasympathetic vagus nerve could compensate for these effects and improve colitis outcome(33). Acute electric stimulation of the cervical vagus nerve (5v, 50Hz, 100ms) for 15 minutes decreased the LF/HF ratio without affecting corticosterone or epinephrine concentrations compared to sham mice which were subjected the same surgical procedure without vagal stimulation (Fig.1E). A control group of non-stressed mice was tested parallel to the sham group, and we didn’t observe any significant difference between the sham and control groups. Vagal stimulation could rescue stressed mice from DSS colitis induced lethality and significantly improve DAI score and intestinal bleeding without significant change of body weight loss (Fig.1, F to I). These results are consistent with the literature showing that stool consistency and intestinal bleeding/permeability are the first symptoms as reflected in early DAI scores before body weight changes are noticeable normally after day 5 of DSS treatment (34). Vagal stimulation also rescued normal mice (without stress) and prevented colitis-induced intestinal bleeding, DAI, body weight loss, and intestinal permeability (Fig.1, F to J). Vagal stimulation preserved macroscopic and microscopic intestinal integrity (Fig.1, K to M), reduced inflammatory cytokines interleukin 6 (IL-6) and tumor necrosis factor alpha (TNF-α) and increased the anti-inflammatory cytokine interleukin 10 (IL-10) (Fig.1N). Because these results suggested that vagal stimulation may improve colitis by inducing IL-10, we studied colitis and vagal stimulation in IL10-KO mice. IL10-KO mice were highly susceptible to DSS colitis, with early progression of DAI (Fig. 2A) and intestinal bleeding within the first 24 hours (Fig. S2, A and B). Wild-type and IL10-KO mice had similar body weight loss (Fig. 2B), colon permeability, colon length, organ adhesion, and inflammatory cytokine concentrations at euthanasia on day 7 (Fig. 2, C and D; Fig. S2C). Vagal stimulation protected IL10-KO mice from onset of colitis until after the first 4 days and reduced body weight loss, intestinal permeability, stool consistency, and intestinal bleeding with non-significant alteration of IL-6 or TNF-α concentrations by euthanasia at day 7 (Fig. 2, E to H; Fig. S2, D to F). These results suggested that vagal protection against colitis was largely independent of IL-10.

Fig. 1. Vagal stimulation improves outcomes in a mouse model of colitis.

(A) Kaplan Meier survival curve and time course of (B) body weight and (C) disease activity index (DAI), (D) serum concentrations of corticosterone (Cort) or norepinephrine (NE), and heart rate variability (HRV measured as the Low/High-Frequency ratio) in 3%DSS mice with or without stress or (E) stress and non-stress mice with sham or vagal stimulation (VS). Stress mice were subjected to 21 days of restraint then received either sham or VS followed by either clean water (control) or 5 days of DSS treatment. (E) Kaplan Meier survival curve and timecourse of (G) body weight, (H) DAI, and (I) intestinal bleeding in non-stress (dashed lines) and stress (solid lines, St) with and without sham or VS. in stress DSS mice with sham or vagal stimulation. (J) Intestinal permeability, colon weights, (K) representative macroscopic colons and group colon lengths, (L) colonic H&E (20x, scale 200μM), (M) histological score, and (N) colon IL-6, TNF-α, and IL-10 concentrations at day 7 in control and DSS mice with sham or vagal stimulation. Data are representative of experiments that were replicated on different dates. For survival analysis mice were followed for 14days, *P<0.05 vs Sham, #P<0.05 vs St-Sham DSS (n=10 from two repeated experiments, Long Rank test). *P<0.05 vs Control (n=5, Two-way ANOVA with Bonferroni’s post-hoc test). *P<0.05 vs Control, #P<0.05 vs sham DSS (n=5, One-way ANOVA with Bonferroni’s post-hoc test). Graphs represent the mean ± SEM of experiments repeated twice on different dates.

Fig. 2. Vagal regulation of colitis in mice is IL-10-independent and mediated by the spleen.

Time course of (A) DAI, (B) body weight, (C) intestinal permeability, and (D) colon IL-6 and TNF-α concentrations in (a to d) IL10-KO (IL10^KO) mice and their wild-type (WT) littermates at day 7 after 5-day DSS challenge. Time course of (E) DAI, (F) body weight, (G) intestinal permeability, and (H) colon IL-6 and TNF-α concentrations in (E to H) IL10^KO at day 7 after receiving sham or vagal stimulation (VS) immediately followed by 5-day DSS challenge. Time course of (I) DAI, (J) body weight, (K) intestinal permeability, and (L) colon IL-6 and TNF-α concentrations in (I to L) splenectomized (SPX) or sham mice at day 7 after 5-day DSS challenge. Mice were splenectomized 5 days before the DSS, sham, or VS treatment. Time course of (M) DAI, (N) body weight, (O) intestinal permeability, and (P) colon IL-6 and TNF-α concentrations in (M to P) splenectomized (SPX) at day 7 after receiving sham or vagal stimulation (VS) immediately followed by 5-day DSS challenge. *P<0.05 vs Control, #P<0.05 vs DSS (n=5, One-way ANOVA with Bonferroni’s post-hoc test). Graphs represent the mean ± SEM of experiments repeated twice on different dates.

Splenectomy blocks vagal protection against colitis in mice.

The mechanism of vagal regulation of colitis is unknown because most of the colon lacks efferent vagal innervation. We have previously reported that the spleen is a critical link between the nervous and immune systems that regulates innate immune responses to bacterial challenge in endotoxemia (16, 35, 36), and so we postulated that the spleen may also contribute to vagal regulation of colitis. Surgical splenectomy followed by initiation of colitis by administration of DSS 5 days later resulted in little change to early stool consistency, intestinal bleeding, colon length, and organ adhesion, but improved DAI scores after day 5 (Fig. 2I, fig. S3, A to C). Splenectomy before onset of colitis led to improved colon integrity, body weight, and intestinal permeability, and reduced colon IL-6 and TNF-α concentration, without affecting IL-10 concentrations (Fig. 2, J to L). Moreover, splenectomy blocked the protective effects of vagal stimulation in DAI score, colon length, permeability, cytokine concentrations, stool consistency, intestinal bleeding, colon length, and organ adhesion (Fig. 2, M to P; fig. S3, D to F).

Vagal stimulation is partially dependent on α7-nicotinic acetylcholine receptors (α7nAChRs) in controlling colitis.

We and others have reported that vagal stimulation controls inflammation in endotoxemia by inhibiting macrophages via α7-nicotinic acetylcholine receptors (α7nAChRs) (17, 37-39). We therefore analyzed whether α7nAChRs contribute to the vagal regulation of colitis. Chrna7-KO colitis mice had lower DAI scores and experienced less body weight loss and intestinal permeability with no difference in macroscopic images, colon length, and weight compared to wild-type colitis mice, similar to that seen in splenectomized mice (Fig. 3A; Fig. 3, B to E). In addition, there were no differences in stool consistency, intestinal bleeding, and organ adhesion between Chrna7-KO and wild-type colitis mice (Fig. S4, A to C) Chrna7-KO colitis mice had lower IL-6 but similar TNF-α and IL-10 concentrations in the colon as wild-type mice with colitis (Fig. 3F). Vagal stimulation improved DAI scores and stool consistency after day 4 in Chrna7-KO mice (Fig. 3G; Fig. S4D). Although we did not detect a significant effect on body weight loss, intestinal permeability, intestinal bleeding, organ adhesion, macroscopic images, colon weight, and colon length, vagal stimulation reduced IL-6 and TNF-α but not IL-10 concentrations in the colons of Chrna7-KO colitis mice on day 7 (Fig. 3, H to L; fig. S4, E and F). These results suggest that vagal stimulation may induce two protective mechanisms, one dependent on α7nAChRs to induce IL-10 and partially inhibit TNF-α, and a second α7nAChR-independent mechanism to inhibit IL-6 and TNF-α.

Fig. 3. Vagal protection against colitis is partially dependent on α7nAChR.

Time course of (A) DAI, (B) body weight, and (C) representative colons, (D) group colon lengths and weight, (E) intestinal permeability, and (F) colon immune cytokines in (A to F) Chrna7-KO (α7^KO) or wild-type (WT) littermates at day 7 after 5-day DSS challenge. Time course of (G) DAI, (H) body weight, and (I) representative colons, (J) group colon lengths and weight, (K) intestinal permeability, and (L) colon immune cytokines in (G to L) Chrna7-KO (α7^KO) mice at day 7 after receiving sham or vagal stimulation (VS) immediately followed by 5-day DSS challenge. *P<0.05 vs Control, #P<0.05 vs WT or DSS (n=4, One-way ANOVA with Bonferroni’s post-hoc test). Graphs represent the mean ± SEM of experiments repeated twice on different dates.

Acetylcholine is the principal neurotransmitter of the vagus nerve and signals through nicotinic receptors. Given the known protective effects of nicotine on ulcerative colitis (40, 41), we assessed whether nicotine could replicate vagal neuromodulation of colitis. Treatment with nicotine was less effective than vagal stimulation and did not improve DAI scores during the first 4 days of treatment, but did reduce intestinal bleeding and body weight loss on days 6 and 7, although effects were minor (Fig. 4, A to C). By euthanasia on day 7, nicotine substantially reduced intestinal permeability and IL-6 and TNF-α concentrations, without affecting IL-10 (Fig. 4, D and E). Next, we analyzed whether the effects of nicotine were mediated by α7nAChR using Chrna7-KO mice. Nicotine did not improve DAI scores, intestinal bleeding, body weights, colon length, or intestinal permeability in Chrna7-KO mice (Fig. 4, F to I). However, nicotine did inhibit IL-6 and TNF-α concentrations in the colons of Chrna7-KO mice (Fig. 4J). These results suggest that nicotine can protect reduce inflammatory cytokines in colitis through a mechanism independent of α7nAChRs.

Fig. 4. Nicotine improves colitis outcomes in mice through α7nAChRs.

Time course of (A) DAI, (B) intestinal bleeding, and (C) body weight, and (D) colon length, intestinal permeability, and (E) colon immune cytokines in (A to E) wild-type mice at day 7 after 5-day DSS challenge with 3-day nicotine treatment or vehicle. Timecourse of (F) DAI, (G) intestinal bleeding, and (H) body weight, and (I) colon length, intestinal permeability, and (J) colon immune cytokines in (F to J) Chrna7-KO (α7^KO) mice at day 7 after 5-day DSS challenge with 3-day nicotine treatment or vehicle. *P<0.05 vs Control, #P<0.05 vs DSS (n=5, One-way ANOVA with Bonferroni’s post-hoc test). Graphs represent the mean ± SEM of experiments repeated twice on different dates.

SUMOylation is a dynamic regulatory process in human IBD and vagal neuromodulation.

We recently reported that SUMOylation helps orchestrate immune responses to bacterial infection (23, 30, 42). Thus, we first analyzed Ubc9 transcripts in our mRNA databank of human biopsy samples where Single-cell sequencing data are available under GEO accession GSE164985.(43-45). There was no statistically significant difference between the Ubc9 transcripts from 91 patients with CD and 29 non-IBD controls (p=0.34, Fig. 5A). There are three SUMO isoforms, termed SUMO1, SUMO2, and SUMO3, with SUMO2 and SUMO3 referred to together as SUMO2/3 due to their homology. We collected 18 human samples from intestinal biopsies and surgical resections from the ascending colons of 9 healthy controls without IBD, 2 UC, and 7 CD patients to analyze SUMOylation by Western blot and immunohistochemistry. Western blotting did not reveal major changes in SUMO1 but did show a strong increase in SUMO2/3 abundance in samples from patients with CD as shown by the high molecular “smear” representative of polySUMOylation of multiple proteins as compared to non-IBD patients (Fig. 5B). Immunohistochemistry analyses of human biopsies also demonstrated stronger SUMO2/3 staining in samples from patients with CD (Fig. 5C). Western blots of colon samples from WT mice with DSS-induced colitis showed similar patterns of SUMOylation without major changes in SUMO1 but about 2-fold increases in SUMO2/3 smear ≈ 250KDa (Fig. 5D). Immunohistochemistry analyses also demonstrated a stronger and broader distribution of SUMO2/3 in the colon of colitis mice (Fig. 5E). Vagal stimulation decreased SUMO1 by nearly 30% and SUMO2/3 by 70% compared to sham colitis mice (Fig. 5, D and E).

Fig. 5. SUMOylation in human biopsy and vagal regulation of colitis.

(A) DEseq normalized counts for Ubc9 expression in human biopsy samples from 29 patients without IBD (NIBD) and 91 patients with CD (p=0.34). Single-cell sequencing data are available under GEO accession GSE164985. (B) Western blots and (C) representative IHC for SUMO1 (red, center column) and SUMO2/3 (green, right column) of colon biopsy from NIBD (top row), and patients with UC (middle row), and CD (bottom row). Merged images are shown in the left column (nuclei are stained blue with DAPI). Data shows representative results of 7 biopsies from 18 human samples analyzed. Brackets in (B) point to the typical high molecular weight smears induced by poly SUMOylation. (D) Western blots, densitometric analyses, and (E) representative IHC for SUMO1 and SUMO2/3 of colon samples from control or 3% DSS mice with sham or vagal stimulation. White scale bar is equivalent to 250 μm. While arrows point to intestinal crypt epithelium. Western blots represent experiments repeated at least twice on different days and graphs show mean ± SEM of fold of induction (FI). *P<0.05 vs Control, #P<0.05 vs DSS (n=5/group, two-way ANOVA with Tukey’s post hoc test).

SUMO isoforms regulate different stages of colitis in mice.

We next analyzed the specific roles of different SUMO isoforms in colitis and vagal neuromodulation by using global SUMO1-KO and SUMO3-KO mice (46, 47). SUMO2-KO mice were not analyzed because embryos die during early development (46, 47). SUMO1-KO DAI scores plateaued on day 3, and mice were thereafter resistant to the progression of colitis (Fig. 6A; Fig.S5, A to C). By contrast, SUMO3-KO mice were resistant to colitis during the first 5 days (Fig. 6A; Fig.S5, A to C). Both SUMO1-KO and SUMO3-KO mice were more resistant to intestinal permeability and had lower IL-6 and TNF-α colon concentrations on day 7 than WT mice (Fig. 6, B to D). Vagal stimulation protected SUMO1-KO mice against colitis through day 4 as shown by the lower DAI, body weight loss, however with similar colon length, intestinal permeability, colon IL-6 and TNF-α but higher IL-10 compared to sham SUMO1-KO mice (Fig. 6, E to H) Similarly, vagal stimulation improved stool consistency, and intestinal bleeding through day 4, and reduced colon shortening and organ adhesion at day 7 in SUMO1-KO mice (Fig.S5, D to F). By contrast, vagal stimulation improved DAI scores in SUMO3-KO mice after day 5 (Fig. 6i; fig.S5, G to I). Vagal stimulation prevented intestinal permeability in SUMO1-KO and SUMO3-KO mice by 65% and 35%, respectively (Fig. 6, G and K). By day 7, vagal stimulation increased IL-10 concentrations in both SUMO1-KO and SUMO3-KO mice by nearly 4-fold (Fig. 6, H and J). The protection of SUMO1-KO mice after day 3 correlates with the timing of immune cell influx into the mouse intestine during DSS-induced colitis (48, 49), and prompted us to ask whether the protection induced by SUMO1 depletion may be due to immune cells. We therefore performed bone marrow transplants from wild-type, SUMO1-KO, and SUMO3-KO mice into irradiated wild-type mice. Mice received lethal dose of xray irradiation followed by bone marrow rescue 5 hours later and were monitored for 4 weeks until complete donor chimerism was seen, then we started DSS treatment. The transfer of wild-type and SUMO3-KO bone marrow resulted in increased DAI scores, body weight loss, intestinal permeability, and immune responses in irradiated wild-type mice (Fig. 6, M to P). However, the transfer of SUMO1-KObone marrow significantly protected irradiated wild-type mice against increases in DAI, body weight loss, intestinal permeability, and inflammatory responses while increasing colon IL-10 concentrations by nearly 3-fold. These results revealed the differences between SUMO1 and SUMO3 in the pathogenesis and progression of colitis and suggested that vagal stimulation protects against colitis by inhibiting SUMOylation.

Fig. 6. Bone marrow transfer from Sumo1-KO mice confers protection against colitis.

(A) DAI, (B) representative colons, (C) group colon lengths and intestinal permeability, and (D) immune cytokines in (A to D), Sumo1-KO, Sumo3-KO, and their wild-type littermate controls at day 7 after 5-day DSS challenge. (E) DAI, (F) body weight, (G) group colon lengths and intestinal permeability, and (H) immune cytokines in (E to H), Sumo1-KO, at day 7 after receiving sham or vagal stimulation (VS) immediately followed by 5-day DSS challenge. (I) DAI, (J) body weight, (K) group colon lengths and intestinal permeability, and (L) immune cytokines in (I to L), Sumo3-KO, at day 7 after receiving sham or vagal stimulation (VS) immediately followed by 5-day DSS challenge. (M) DAI, (N) body weight, (O) group colon lengths and intestinal permeability, and (P) immune cytokines in (M to O), irradiated wild-type DSS (Ir.WT+DSS) mice transferred with wild-type, Sumo1-KO, or Sumo3-KO bone marrow and received 5-day DSS challenge starting at day 28 after bone marrow transplantation and assessed at day 7 after DSS challenge. *P<0.05 vs Control or WT, #P<0.05 vs DSS (n=4, One-way ANOVA with Bonferroni’s post-hoc test). Graphs represent the mean ± SEM of experiments repeated twice on different dates.

Inhibition of SUMOylation with TAK-981 rescued mice from established colitis.

Because vagal stimulation is an invasive surgical procedure, we next tested whether pharmacological inhibition of SUMOylation mimics vagal stimulation and protects against colitis. We used TAK-981 (Subasumstat), a first-in-class inhibitor of the transfer of SUMO to Ubc9 (43-45, 50). First, we confirmed that TAK-981 inhibits both SUMO1 and SUMO2/3 in the colon of DSS mice (fig. S6, A to D). The efficacy of TAK-981 treatment was also confirmed by western blotting against negative control samples from knockout mice to ensure successful inhibition of SUMOylation (Fig. S6, A to F). Because previous studies reported that bacterial endotoxin induces SUMOylation(23), we also confirmed that TAK981 inhibits SUMOylation in the spleens of endotoxemic mice (Fig. S6, C and D). The specificity of the staining was confirmed in extracts from SUMO1-KO and SUMO3-KO mice. Mice were treated with TAK-981 (7.5 mg/kg) 12 hours before the start of DSS challenge then daily for 4 days. TAK-981 protected against colitis, improved DAI scores, preserved the macroscopic and histological integrity of the colon, prevented body weight loss, and almost reduced intestinal permeability by nearly 90% (Fig. 7, A to E). TAK981 reduced colon IL-6 and TNF-α concentrations by 80% and 65%, respectively without affecting IL-10 concentrations (Fig. 7F).

Fig. 7. Inhibition of SUMOylation with TAK-981 improves colitis outcomes in mice.

(A) DAI, (B) body weight, (C) representative colons and group colon lengths, (D) H&E of colonic tissues (10x, 20x, scale 200uM), (E) histological score, intestinal permeability, and (F) colon immune cytokines in control or colitis mice treated with vehicle or TAK-981 at day 7 after 5-day DSS challenge. Mice were treated with TAK-981 (7.5mg/kg) 12 hours before starting DSS challenge and then daily for 4 days. *P<0.05 vs Control, #P<0.05 vs DSS (n=5, One-way ANOVA with Bonferroni’s post-hoc test).

We next analyzed whether TAK-981 treatment can rescue mice from established colitis. TAK-981 treatment, starting three days after the DSS challenge, ameliorated colitis and improved DAI scores by reducing intestinal bleeding and body weight loss (Fig. 8, A to C). TAK-981 also improved DAI scores, intestinal bleeding, and body weights in experimental CD induced by TNBS (2,4,6- trinitrobenzene sulfonic acid, Fig. 8, D to F). These results prompted us to hypothesize that TAK-981 might preserve intestinal integrity and inhibit colitis progression to sepsis, a potentially lethal complication in patients with IBD. Thus, we used TAK-981 to treat mice with oxazolone-induced colitis, an experimental model of severe ulcerative colitis associated with mucosal damage that can progress to sepsis (51, 52). TAK-981 treatment started before oxazolone challenge improved DAI scores, reduced intestinal bleeding and body weight loss, and improved intestinal integrity (Fig. 8, G to K). TAK-981 treatment also reduced intestinal permeability and peritoneal bacterial leakage by nearly 90% (Fig. 8L) and IL-6 and TNF-α by nearly 70% without substantially affecting IL-10 (Fig. 8M). Together, these results demonstrated that inhibition of SUMOylation with TAK-981 during colitis might prove advantageous for treating patients with IBD.

Fig. 8. TAK-981 treatment ameliorates disease in three mouse models of IBD.

(A) DAI, (B) intestinal bleeding, (C) body weight at day 7 of 5-day DSS challenge in mice treated with vehicle or TAK-981 (rTAK) started at day 3 post-DSS and daily for 4 days., (D) DAI, (E) intestinal bleeding, (F) body weight at day 7 of TNBS (150mg/kg) recall instillation in mice treated with vehicle or TAK-981 started 12h before TNBS challenge and daily for 4 days. (G) DAI, (H) intestinal bleeding, (I) body weight, (J) representative macroscopic images, (K) microscopic images, (L) intestinal permeability and peritoneal lavage bacteria, (m) colon inflammatory cytokines at day 7 of oxazolone (1%) (OXA) recall instillation in mice treated with vehicle or TAK-981 started 12h before oxazolone challenge and daily for 4 days. *P<0.05 vs Vehicle (n=4, Two-way ANOVA with Bonferroni’s post-hoc test). *P<0.05 vs Control, #P<0.05 vs OXA (n=4, One-way ANOVA with Bonferroni’s post-hoc test).

Discussion

Epidemiological studies have suggested an association between stress and IBD relapses, regardless of whether stress is consciously perceived (53). Our results showed that stress results in lethal colitis within 5 days after the DSS challenge. We selected stress induced by physical restraint, without constraint or pain, to avoid circadian or metabolic dysfunction induced by other models such as sleep or food deprivation (54, 55). Stress induced a higher LF/HF ratio in HRV frequency analyses, suggesting parasympathetic dysfunction. We previously reported that vagal dysfunction contributes to inflammatory disorders and vagal stimulation can curb inflammatory responses to bacterial endotoxin(35, 36). Considering the previous studies reporting that stress can affect immune responses in mice via C1 neurons (56), it is possible that the abrogation of stress-induced immune dysfunction is one mechanism of vagal anti-inflammatory effects. The mechanism of vagal regulation of colitis is unknown and likely complex because the colon lacks efferent vagal innervations. Thus, previous studies focused on experimental CD induced by TNBS in the small intestine rather than colitis. Two recent studies reported that prophylactic chronic vagal stimulation can improve outcomes in rats with TNBS-induced colitis (57, 58). These studies are limited by the variability of colon inflammation in IBD experimental models using haptenizing agents like TNBS or oxazolone (58). -Despite its invasiveness, vagal stimulation in IBD mouse models is a very efficient tool to study local and systemic and immune regulation and define new molecular targets for future drug discovery.

Our results concur with clinical studies showing that therapies increasing cardiovagal activity, such as mind-body interventions, alleviate IBD symptoms, but their clinical use is hampered by limited efficacy. Vagal stimulation protected mice from colitis for the first 5 days after DSS exposure. Stool consistency and intestinal bleeding and permeability are generally the first signs of mucosal barrier injury before body weight changes are noticeable around day 5 of DSS treatment. These results prompted us to analyze the local and systemic vagal mechanisms of neuromodulation mediated by the spleen.

The vagal induction of IL-10 concurs with the high susceptibility of IL10-KO mice to colitis. IL-10-deficient mice have normal lymphocyte development and antibody responses but are anemic and develop chronic enterocolitis (59). Vagal stimulation protected IL10-KO mice from colitis, revealing a IL-10-independent mechanism. We have previously reported that the spleen is a critical link between the nervous and immune systems in endotoxemia (16, 60, 61). Vagal stimulation activates the ‘cholinergic anti-inflammatory pathway,’ curbing unfettered inflammation in experimental sepsis by inhibiting the production of inflammatory cytokines in splenic macrophages via α7nAChRs (60, 62-64). We now report that splenectomy prevented vagal regulation of colitis. These results were also consistent with recent clinical studies showing splenomegaly (65) with splenic abscesses as an early manifestation of CD (66), and dysfunctional hypersplenism with higher pitted erythrocyte counts in patients with relapsed UC (67). These results suggested that the spleen may be a therapeutic target for neuromodulation and bioelectronic strategies. We also noted that vagal stimulation protected Chrna7-KO mice against DSS colitis. Because vagal stimulation failed to control systemic inflammation in Chrna7-KO mice in endotoxemia, this suggests a mechanism of vagal neuromodulation in colitis independent of the canonical vagal IL-10/α7nAChR vagal pathway. Future physiologic and cellular studies will be required to delineate this complex vagal-immune neuronal network, as well as the specific cells regulated in the spleen and colon.

SUMOylation is emerging as a stress-adaptive response to safeguard intestinal integrity a tissue particularly challenged by its bacterial environment (23, 24, 28-32). A previous study reported lower Ubc9 expression in colonic biopsy samples from patients with UC and CD (n = 22/group) (24). We did not find a significant difference in Ubc9 expression between samples from 91 patients with CD and 29 non-IBD controls (p=0.34). The reason for this lack of difference may be due to patients’ selection with wide range of ages and disease activity. Western blot and immunohistochemistry revealed an increase of SUMO2/3 but not SUMO1 in human samples and murine models. SUMO1-KO mice were resistant to colitis progression, whereas SUMO3-KO mice were resistant to the early stages of colitis, and bone marrow transplants from SUMO1-KO but not SUMO3-KO mice, conferred protection to irradiated wild-type mice against colitis, demonstrating differences between the contributions of the SUMO1 and SUMO3 isoforms to colitis. The typical high molecular weight smears observed in Western blots on samples frompatients with CD and mouse models reflect SUMOylation of multiple proteins. Indeed, SUMOylation controls multiple pathways and can target several proteins within the same pathway, such asp65, inhibitor of nuclear factor kappa B (IκBα), and Centrobin-Centrosomal Protein 4.1-associated Protein (CPAP) in the Nuclear factor kappa B (NF-κβ) pathway (68-70). Future in vivo and in vitro studies will be required to define the specific targets of SUMO1 and SUMO3 in splenic and colonic cells during colitis.

Our study has several limitations. First, it is unclear which part of the vagus nerve plays a dominant role in IBD regulation. We have recently used an afferent vagal stimulation model to demonstrate the role of vagal sensory regulation in insulin secretion (71). Although this model was successful in improving short-term metabolic readouts, it involves cervical vagotomy, which shortens mice survival, and implementation in colitis models would be technically difficult. Future studies of afferent vagal neuromodulation in colitis models with the help of optogenetics may be warranted. Another limitation of our study is the use of global SUMO KO models, which prevents delineation of the specific effector compartments of SUMO regulation in IBD. Although we started to tackle this problem by using bone marrow transplant chimeras, the availability of conditional knockouts will help us to further elucidate the precise compartments in future studies. Another limitation of this study is the use of male mice only to limit variability in results. In addition, although we have shown the importance of cell trafficking for specific SUMO1 inhibition, it is still unclear whether this is mediated through decreased inflammatory cell influx or increased anti-inflammatory cell influx. Future flow cytometric studies are warranted to further understand the role of trafficking immune cells in conveying the anti-inflammatory properties of SUMO inhibition in IBD. Also, the mechanism of local protective effects of SUMO3 inhibition in the gut are not yet fully understood.

Our results showed that SUMOylation is regulated by neuromodulation and that vagal stimulation can control SUMOylation in murine models of colitis. We then tested whether pharmacological inhibition of SUMOylation mimicked vagal stimulation and protected against colitis. Treatment with the SUMOylation inhibitor TAK-981 improved both DSS colitis and TNBS CD in mice. TAK-981 treatment also reduced intestinal permeability and bacterial leakage in oxazolone-induced colitis, an experimental model of severe ulcerative colitis associated with mucosal damage that can progress to sepsis (51, 52). in conclusion, these results show that modulation of SUMOylation through vagal stimulation or treatment with a SUMOylation inhibitor can ameliorate colitis in mouse models, suggesting these approaches as potential therapies for patients with IBD.

Material and Methods

Study Design.

The aim of this study was to identify the role of vagal neuromodulation in murine colitis and to assess its pharmacologic translation via SUMO inhibition. Sample sizes were determined according to our previous studies and preliminary results. Sample sizes are included in the figure legends. Investigators were blinded to mice treatment groups during DAI score assessment and histological assessments. Mice were randomized for control, vagal stimulation, or drug treatment.

Chemicals and reagents.

Dextran sulfate sodium salt (DSS, MW 35-60KDa; MP Biomedicals) was dissolved in autoclaved deionized water (34, 72, 73). 2,4,6-trinitrobenzene sulfonic acid (TNBS, Sigma-Aldrich) was used at a single dose of 150mg/kg (150μL, pr). 4-Ethoxymethylene-2-phenyl-2-oxazolin-5-one (Oxazolone, Sigma-Aldrich) was used at 1% w/v in 50% ethanol (150μL, pr) 7 days after epicutaneous sensitization with 3% w/v Oxazolone in ethanol. Nicotine was dissolved in autoclaved drinking water at 20 μg/mL (74). Mice were treated with nicotine 3 days before and daily for 5 days after DSS treatment as reported (74). TAK-981 (ChemieTek)(43-45, 50) was dissolved in DMSO (stock concentration 50 mg/mL, stored at −20°C) and diluted in pharmaceutical-grade corn oil (Sigma-Aldrich) at 7.5 mg/kg, and administered s.c. 12 h before the start of colitis induction, then daily for 4 days as we reported (23).

Animal experiments.

Animal procedures were approved by the Institutional Animal Care & Use Committee of Duke University. 6-8-week-old (~25 ± 5 g) male C57BL/6; male IL10-KO mice (59), and Chrna7-KO mice were obtained from The Jackson Laboratory. Sumo1-KO, Sumo3-KO, and wild-type littermate mice were produced in our lab as previously described (46, 47). Genotyping was performed by PCR using tail genomic DNA and the Extract-N-Amp Tissue PCR kit (Sigma-Aldrich)(46, 47). Mice were maintained on 12h light-dark cycle, with free access to food and water (ad libitum), randomized into experimental groups, and outcome assessors were blinded to group treatments.

Restraint stress.

Stress was induced by restraining the mice for 6 hours daily for 3 weeks in well-ventilated polypropylene tubes without physical compression or pain described (54, 55).

Heart rate variability (HRV).

Three stainless steel electrodes were implanted in the right and left pectoral muscles, and the abdominal wall using cardiac pacing wires (A & E Medical), and tunneled under the skin. The electrodes were exteriorized from the mouse nape and connected to MP150 research system (Biopac) for lead II electrocardiogram recording. Mice were allowed to recover for 5 days before recording lead II EKG. AcqKnowledge software was used for HRV frequency domain analyses based on the consecutive RR intervals of the 15 min EKG recordings. Low and high-frequency domains were calculated, and the ratio was used as an indicator of the sympathetic-vagal balance as described (57).

Experimental IBD.

Dextran Sulfate Sodium (DSS) colitis:

In the initial experiments (fig. S1, A to H), we used 3 and 5% DSS in autoclaved drinking water for 5 days as described in the literature (34). Thereafter, all experiments were performed with 3% DSS. For survival analysis, animals treated with DSS 3% were followed for 14 days to rule out delayed deaths.

TNBS (2,4,6-trinitrobenzene sulfonic acid) colitis:

Mice were fasted overnight and then given 150μL of TNBS (150mg/Kg) in 50% ethanol injected intra-rectally using polyurethane P20 tube (Becton Dickinson) advanced gently 4 cm (from the anal verge) into the mouse rectum as previously described.(75, 76) The mice were held vertically for one minute after injection to make sure that TNBS was retained in the rectum.

Oxazolone (4-Ethoxymethylene-2-phenyl-2-oxazolin-5-one) colitis:

A 2x2 cm area of the abdomen was shaved, and mice were pre-sensitized with the percutaneous application of 150 μL of 3% oxazolone w/v in 100% ethanol as previously described.(77) Seven days later, mice received 150 μL of 1% oxazolone w/v in 50% ethanol intrarectally, then held vertically for one minute to ensure the retainment of oxazolone in the rectum.

In the three models of colitis, body weight, stool consistency, blood in stools by Hemoccult test (Hemoccult Sensa, Beckman Coulter), and water consumption were recorded daily during the experiment to calculate Disease Activity Index (DAI) score.(78) Intestinal permeability was analyzed with oral gavage of FITC dextran (0.6mg/g BW in PBS) 3 hours before terminal blood collection.(79) Mice were euthanized on the indicated day, blood was collected by cardiac puncture and intestine was collected for further analysis.

Disease activity index (DAI).

DAI was recorded daily assessing the preclinical parameters as described (78). Weight loss was scored (compared to baseline weight): 0, no loss or weight gain; 1, 5-10%; 2, 10-15%; 3, 15-20%; and 4, >20% loss from baseline weight. Stool consistency was scored: 0, well-formed stools; 1, sticky stools; 2, soft stools; 3, semi-solid loose stools; and 4, watery diarrhea. Rectal bleeding was scored: 0, no bleeding; 2, occult blood (by Hemoccult sensa, Beckman Coulter, CA); 3, bloody stools; and 4, gross rectal bleeding. Mice were immediately euthanized if they lost more than 20% of their baseline body weight.

Vagal stimulation.

Mice were anesthetized by 3-5% isoflurane, intubated, and connected to a ventilator to allow 45-55 mL/ min airflow with 30% oxygen and 1-2% maintenance isoflurane as we described.(80) EKG leads were attached to the right and left forelimbs and hindlimbs to monitor heart rate. The front of the neck was shaved, and sterilized (70% ethanol and povidone-iodine) and a 2 cm longitudinal midline incision was made. Submandibular salivary glands were dissected and retracted laterally to expose the carotid artery and right cervical vagus nerve. The vagus nerve was carefully dissected from the carotid artery and connected to platinum-iridium microelectrodes. Biopac MP150 system was used to deliver continuous electrical stimulation (5V, 50 Hz, 100ms) for 15 minutes. Heart rate was monitored during the procedure to ensure effective vagal stimulation. After the stimulation, tissues were returned to anatomical positions, and the neck was closed with a 6/0 suture. Mice were monitored for 30 minutes in an incubation chamber to ensure full recovery before returning them to the cage. Sham mice were intubated, a longitudinal neck incision was performed, and the glands were retracted without touching the vagus nerve to avoid mechanical stimulation. Vagal nerve stimulation was performed immediately before DSS treatment.

Splenectomy.

Mice were anesthetized with isoflurane 5% and subjected to an abdominal incision on the epigastrium and mesogastrium. The spleen was exposed by gentle retraction of the stomach to the side as we described (81). The three main branches of the splenic artery were stabilized with nylon thread, ligated, and cut. The spleen was removed, and the abdominal wall was sutured with catgut and the skin with nylon thread. Mice were splenectomized 5 days before the experimental procedure.

Peritoneal lavage and bacterial load analysis.

Peritoneal lavage was performed as we described.(82) Six hours before euthanasia, mice were anesthetized with 5% isoflurane, the abdominal wall was disinfected with 70% ethanol, and 5 mL of sterile pyrogen-free PBS was injected i.p. using a 25G needle. Then, the abdomen was massaged for one minute, and 1 mL of peritoneal fluid was aspirated using a sterile syringe. Peritoneal aspirates were centrifuged at 2,000g for 20 min at 4°C and supernatants were frozen at −20°C for further cytokines analysis. Peritoneal lavage bacterial load was analyzed as previously described.(83) Freshly collected peritoneal aspirates were cultured in blood agar using a sterile inoculation loop and agar plates were incubated at 37°C for 24 hours. Bacterial colonies (CFU/mL of peritoneal aspirate) were counted using an ImageJ Fiji software cell analyzer.(84)

Statistical analyses.

Statistical analyses were performed with GraphPad Prism Software (GraphPad Software, San Diego, CA). The sample size was calculated based on our preliminary results using G*Power3.1 and ANOVA (Fixed effects, omnibus, one-way), assuming effect size= 0.25, α=0.05, power (1-β prob Type II error) =0.8. Data are expressed as mean±SEM and the figures represent experiments repeated twice on different days. The student’s t-test or Mann-Whitney U test was used to compare two experimental groups. Three or more groups were analyzed with parametric one-way ANOVA with multiple pair-wise comparisons. Normality and homogeneity of variance were confirmed with Kolmogorov-Smirnov analyses. Pair comparisons in nonparametric ANOVA tests were adjusted post-hoc with the Tukey test (in equal sample sizes) or Bonferroni’s for multiple hypothesis testing. The time courses and pair-wise comparisons were analyzed with the two-way ANOVA for repeated measures. Statistical significance was established at p≤.05 with a 2-sided alpha. Figures show representative data of experiments that were replicated on different dates.