The gut microbial metabolites short-chain fatty acids restrain tissue bacterial load, chronic inflammation, and associated cancer in the colon

Abstract

The intestinal immune system is regulated by microbes and their metabolites. The roles of gut microbial metabolites in regulating intestinal inflammation and tumorigenesis are incompletely understood. We systematically studied the roles of short-chain fatty acids (SCFAs) and their receptors (GPR43 or GPR41) in regulating tissue bacterial load, acute versus chronic inflammatory responses, and intestinal cancer development. SCFA receptor-, particularly GPR43-, deficient mice were defective in mounting appropriate acute immune responses to promote barrier immunity, and developed uncontrolled chronic inflammatory responses following epithelial damage. Further, intestinal carcinogenesis was increased in GPR43-deficient mice. Dietary fiber and SCFA administration suppressed intestinal inflammation and cancer in both GPR43-dependent and independent manners. The beneficial effect of GPR43 was not mediated by altered microbiota but by host tissue cells and hematopoietic cells to a lesser degree. We found that inability to suppress commensal bacterial invasion into the colonic tissue is associated with the increased chronic Th17-driven inflammation and carcinogenesis in the intestine of GPR43-deficient mice. In sum, our results reveal the beneficial function of the SCFA-GPR43 axis in suppressing bacterial invasion and associated chronic inflammation and carcinogenesis in the colon.

Introduction

The intestine harbors a myriad of commensal microorganisms, which benefit the host in many different ways [1, 2]. Digestion resistant-dietary materials reach the colon for microbial fermentation. A major group of dietary materials processed by microbial fermentation includes dietary fibers (DF) and resistant starch. They are fermented to produce short-chain fatty acids (SCFAs), which play diverse roles from fueling colonocytes to regulating cells of the immune system, energy storage/metabolism, and gut barrier function [3–14].

SCFAs exert their functions in the host through both SCFA receptor-dependent and independent mechanisms [3]. SCFA receptors include G-protein-coupled receptors (GPCRs) such as GPR43, GPR41, GPR109A and Olfr78. GPR43 and GPR41 are activated by all major SCFAs such as acetate (C2), propionate (C3), and butyrate (C4) [15, 16]. GPR109A is activated by C4 along with a non-SCFA ligand, niacin [17]. Olfr78 is activated by C2 and C3, but not C4 [18]. All of these receptors are 7-transmembrane spanning GPCRs and activate heterotrimeric G proteins leading to subsequent activation of many proteins, including MAPK/ERK proteins [13, 16, 19, 20]. GPR43, GPR41 and GPR109A are differentially expressed by tissue cells, including gut epithelial cells, and certain hematopoietic cell types such as macrophages and dendritic cells [13, 15, 16]. Olfr78 is expressed in kidney endothelial cells [18].

Inflammation is a major risk factor for cancer development in the digestive tract [21–24]. Colon cancer is increased in patients with inflammatory bowel diseases, such as ulcerative colitis and Crohn’s disease. Similarly, celiac disease and Crohn’s disease are risk factors for small intestinal cancer. It has been documented that high SCFA levels and DF intake decrease colon cancer development [25]. GPR43 expression is decreased in tumor cells and this receptor functions to suppress colon cancer development [26–28]. Employing acute or chronic DSS-induced colitis model, two groups found that GPR43 has heterogeneous effects on dextran sodium sulfate (DSS)-induced inflammatory responses [29, 30]. GPR43 and GPR41 are required for mounting acute inflammatory responses following gut barrier breach [13, 29], which is thought to be important for barrier repair and decrease tissue bacterial load. A potential mechanism for the anti-inflammatory and anti-cancer effect of GPR43 is microbiota dysbiosis in GPR43 deficiency. This idea is supported, in part, by the reports that GPR43 functions to increase beneficial Bifidobacterium species but suppress inflammatory Helicobacter hepaticus and Prevotellaceae [27] and that cohousing of Gpr43^−/− mice with wild type (WT) littermates suppressed acute 2% DSS-induced colitis [28]. However, it is at odd with a report that DF suppress inflammation-associated colon cancer development in a manner independent of the microbiota change but dependent on another SCFA receptor GPR109A [31]. In this regard, the cohousing or microbiota effect on chronic-inflammation-associated colon cancer development is unknown. Moreover, the differential effects of the SCFA-SCFA receptor axis on acute versus chronic responses in the context of barrier function and microbial invasion should be studied in the context of cancer development to better understand the working mechanisms of DF, SCFAs and their receptors.

In this study, we addressed two critical issues regarding the role of the SCFA-SCFA receptor axis in regulating inflammation and cancer. First, we systematically determined the impact of SCFAs and their receptors, with the focus on GPR43, on tissue bacterial load and acute versus chronic inflammatory responses and how this relates to colon cancer development. Second, we interrogated whether the function of GPR43 is dependent on microbiota difference between WT and Gpr43^−/− mice employing cohoused mice. Our work demonstrates the importance of the SCFA receptor-dependent barrier immunity in regulating bacterial invasion, subsequent inflammatory responses and colon cancer development in a microbiota change-independent manner.

Results

SCFA receptors promote acute but suppress chronic inflammatory responses

To systematically understand the roles of SCFA receptors in regulating inflammation, we compared the inflammatory responses of wild type (WT) and SCFA receptor (Gpr43 and Gpr41)-deficient mice induced with DSS administered in acute and chronic inflammatory regimens. In response to an acute inflammation-inducing DSS treatment regimen (3.5% of DSS for 5 days and euthanized at day 9), SCFA receptor-deficient mice were hypo-responsive in mounting inflammatory responses at this early time point. Compared to WT mice, they lost less weight, had longer colons, failed to increase the frequencies of gut effector T cells (Th1 and Th17) and neutrophils, and did not up-regulate the mRNA expression of inflammatory cytokines and chemokines such as Il18, Il22, Il6, and Cxcl1 (Figure 1). Increased frequencies of Tregs were detected in Gpr43^−/− mice. In general, Gpr43^−/− mice were more defective in mounting the acute inflammatory response than Gpr41^−/− mice.

FIGURE 1. Acute DSS-induced inflammatory responses in WT and SCFA receptor-deficient mice.

(A) Body weight change (%) and stool score during DSS-induced acute inflammation. (B) Colon length of WT, Gpr41^−/− and Gpr43^−/− mice after 3.5% DSS treatment. (C) Frequency of IFN-γ⁺, IL-17A⁺, and FoxP3⁺ cells among intestinal lamina propria (LP) CD4⁺ T cells of WT, Gpr41^−/− and Gpr43^−/− mice. (D) Frequency of Gr-1⁺CD11b⁺ cells in colonic LP cells of WT, Gpr41^−/− and Gpr43^−/− mice. (E) Expression of inflammatory cytokine and chemokine genes in colonic tissues of WT, Gpr41^−/− and Gpr43^−/− mice. mRNA expression was measured by qRT-PCR and normalized to β-actin (Actb). Mice received 3.5% DSS in drinking water for 5 days before they were fed regular water prior to sacrifice at day 9. Representative and pooled data obtained (mean ± SEM, n=8–10) from 2 independent experiments are shown. *Significant differences (P < 0.05) by two-way ANOVA followed by Bonferroni’s multiple comparison (A) and by one-way ANOVA followed by Tukey’s Multiple Comparison Test (B-E) from DSS-treated WT mice.

In response to a chronic inflammation-inducing DSS treatment regimen (3 cycles of 1.5% DSS for 6 weeks), Gpr43^−/− mice suffered less initially but more in the late phase based on weight change and stool score (Figure 2A). However, Gpr41^−/− mice were not different from WT mice in the late phase. At termination following the chronic DSS treatment, Gpr43^−/− mice had signs of more severe inflammation based on colon length and frequencies of inflammatory T cells (Th1 and Th17 cells) and neutrophils (Figure 2B-D). In general, the expression of inflammatory cytokines, such as Il18, Il22 and Il6, was higher in SCFA receptor-deficient mice (Figure 2E). Overall, these results indicate that the SCFA receptor GPR43 suppresses chronic inflammatory responses.

FIGURE 2. Chronic DSS-induced inflammatory responses in WT and SCFA receptor-deficient mice.

(A) Body weight change (%) and stool score during DSS-induced chronic inflammation. (B) Colon length of WT, Gpr41^−/− and Gpr43^−/− mice after 3 cycles of 1.5% DSS treatment. (C) Frequency of Gr-1⁺CD11b⁺ cells in the colonic LP cells of WT, Gpr41^−/− and Gpr43^−/− mice. (D) Frequency of IFN-γ⁺, IL-17A⁺, and FoxP3⁺ cells among LP CD4⁺ T cells of WT, Gpr41^−/− and Gpr43^−/− mice. (E) Expression of inflammatory cytokine and chemokine genes in colon tissues isolated from WT, Gpr41^−/− and Gpr43^−/− mice. mRNA expression was measured. Mice received 3 cycles of 1.5% DSS for 7 days followed by regular water for 7 days. Representative and pooled data (mean ± SEM, n=8–10) from 2 independent experiments are shown. *Significant differences (P < 0.05, by two-way ANOVA followed by Bonferroni’s multiple comparison (A) and by one-way ANOVA followed by Tukey’s Multiple Comparison Test (B-E) from DSS-treated WT mice.

GPR43 functions to suppress Th17-driven inflammatory response and intestinal carcinogenesis in a microbiota change-independent manner

Inflammation induces production of cytokines and genotoxic reactive oxygen and nitrogen species, which cause mutations and epithelial damages. These events compromise gut barrier function and cause bacterial invasion, further increasing inflammatory responses leading to epithelial transformation [32–35]. Thus, we determined the impact of SCFA receptor deficiency on azoxymethane (AOM)/DSS-induced carcinogenesis. We found that Gpr43^−/− mice initially suffered less from AOM/DSS-induced weight loss, and diarrhea the first 3–4 weeks but more in the late phase (Figure 3A, S Figure 1). The effect of GPR41 deficiency on cancer development was not as clear as in GPR43 deficiency, which mirrors the weak effect of GPR41 deficiency on chronic inflammation (S Figure 1 and Figure 2). Because of its clear function in suppressing DSS-induced chronic inflammation and AOM/DSS-induced cancer, we focused the remaining study on GPR43. At 10 weeks post-AOM treatment, Gpr43^−/− mice had greater numbers of polyps and advanced (bigger) adenomas often with complete loss of crypts, medullary inflammatory cells (I) and necrotic abscesses (N) (Figure 3B and C). Further, they had enlarged spleen and mesenteric lymph nodes (MLN, Figure 3D), which indicate active systemic and intestinal inflammatory responses respectively.

FIGURE 3. AOM/DSS-induced colon cancer development in WT versus GPR43-deficient mice.

(A) Changes in body weight (%), survival rate (%), and stool scores of WT and Gpr43^−/− mice during AOM/DSS treatment. (B) Tumor burden (adenoma number and size) in the colon of WT and Gpr43^−/− mice. (C) Histological assessment of colon tissues of WT and Gpr43^−/− mice. Paraffin-embedded tissue sections (6 μm) were stained with hematoxylin and eosin. Medullary inflammatory cells (I) and necrotic abscesses (N) are marked in the picture. (D) Enlargement of spleen and MLN following AOM/DSS treatment. The broken line indicates average weight of the spleen of untreated age-matching control WT mice. Mice were treated with AOM (10 mg/Kg of body weight, i.p.) and 3 cycles of 1% DSS in drinking water and sacrificed at day 70 post AOM injection. Representative and pooled data (mean ± SEM, n=8–12) from 3 independent experiments are shown. *Significant differences (P < 0.05, by two-way ANOVA followed by Bonferroni’s multiple comparison for weight change and stool score (A) and by Mann Whitney U test (2-tailed) for single data set comparisons of two groups of interests (B, D).

It has been documented that GPR43-deficient mice have altered microbiota [27, 28] and, therefore, it is possible that the protective effect of GPR43 is mediated by the altered microbiota. To determine this, we induced AOM/DSS-induced colon cancer in co-housed mice and mice with shared bedding, both of which were effective in creating the same shared microbiota profile based on 16S rRNA gene q-PCR of selected bacterial groups such as Enterococcus, Bacteroidetes, Lactobacillus, E. coli, segmented filamentous bacteria, and Clostridia (S. Figure 2). Gpr43^−/− mice maintained the same increased susceptibility to AOM/DSS-induced colon cancer even after the microbiota-equalization procedures (Figure 4A, B). In the colon, the frequencies of Th17 cells and neutrophils were increased whereas there was no change in those of Th1 and FoxP3⁺ Tregs (Figure 4C). In addition, increased organ weight and frequencies of Th1, Th17 and Treg cells were detected in the spleen and MLN, indicating elevated systemic and intestinal inflammation (Figure 4D,E). We also examined the immune cell profile at an early stage (16–17 days after AOM injection) in co-housed mice. We observed longer colon, somewhat decreased numbers of CD4⁺ T cells and increased frequencies of Th17 cells in the colon of Gpr43^−/− mice at this early stage of carcinogenesis (S Figure 3). Thus, sustained high Th17-driven inflammatory response is the characteristic phenotype of Gpr43^−/− mice during the colon cancer development. Importantly, this phenotype was detected in the absence of microbial changes due to GPR43 deficiency.

FIGURE 4. AOM/DSS-induced colon cancer development and T cell responses in co-housed GPR43-deficient mice.

(A) Gross colon images of co-housed WT and Gpr43^−/− mice. (B) Tumor formation (polyp number and size) in the colon of co-housed WT and Gpr43^−/− mice. (C) Effector T cells, Tregs and neutrophils in the colon of co-housed mice. (D) Spleen weight and T helper cell subset frequencies in the spleen of co-housed mice. (E) MLN weight and effector T cells and Tregs in the MLN of co-housed mice. In panel D and E, the broken lines indicate average weight of the indicated organs of untreated age-matching WT mice. To equalize microbiota between WT and Gpr43^−/− mice, female mice were co-housed at least one week prior to and until termination of the experiments. Alternatively, male mice were housed on mixed bedding (20% of used bedding equally from WT and Gpr43^−/− mice and 80% new bedding; changed twice a week). The data from these two experiments were similar and all data were combined. Mice were treated with AOM (10 mg/Kg of body weight, i.p.) and 3 cycles of 1% DSS in drinking water and sacrificed at day 62–63 post AOM injection. Representative and pooled data (mean ± SEM, n=8–12) from 3 independent experiments are shown. *Significant differences (P < 0.05, 2-tailed Mann Whitney U tests) from AOM/DSS-treated WT mice.

Next, we employed another intestinal cancer model (APC^Min/+ mice) to verify the protective role of GPR43. This model develops spontaneous adenomas in both the small intestine and colon. Gpr43^−/− APC^Min/+ mice developed more tumors in the small intestine and colon than APC^Min/+ mice, with greater numbers of adenomas (Figure 5A). Histological analysis revealed disordered epithelial cancer continuous from intestinal crypts in the intestine of Gpr43^−/− APC^Min/+ mice (Figure 5B). The spleen and MLN of Gpr43^−/− APC^Min/+ mice were abnormally enlarged (Figure 5C).

FIGURE 5. Intestinal tumor formation in WT versus Gpr43^−/− Apc^Min/+ mice.

(A) Tumor formation in the small intestine and colon Apc^Min/+ and Apc^Min/+Gpr43^−/− mice. (B) Histological assessment of intestine tissues. (C) Enlargement of spleen and MLN in Apc^Min/+ and Apc^Min/+Gpr43^−/− mice. Apc^Min/+ and Apc^Min/+Gpr43^−/− mice were sacrificed at 4 months of age. In panel C, the broken lines indicate average weight of the indicated organs of untreated age-matching WT mice. Representative and pooled data (mean ± SEM; n=6) from 2 experiments are shown. *Significant differences from Apc^Min/+ mice (P < 0.05, 2-tailed Mann-Whitney U-test).

Gpr43 is expressed by both tissue cells and hematopoietic cells, including gut epithelial cells, enteroendocrine L cells, adipocytes, eosinophils, basophils, neutrophils, monocytes, dendritic cells, mucosal mast cells, and myometrium endothelium [3]. To determine the relative importance of tissue versus hematopoietic cells in mediating the protective function of GPR43 on colon cancer, we performed a bone marrow (BM) transplantation study. We induced colon cancer with AOM/DSS in these mice and measured weight change, survival, tumor burden, and lymphoid tissue enlargement. While differences between groups in not all the readouts were statistically significant, there was a trend that the “Gpr43^−/− (BM donor)→Gpr43^−/− (recipient)” group was most susceptible, followed by the “WT→Gpr43^−/−” group, and then by the “Gpr43^−/− →WT” group (Figure 6). Thus, both cell types may mediate the protective effect of GPR43, but it was the non-hematopoietic tissue cells that play a major role.

FIGURE 6. Comparison of non-hematopoietic and hematopoietic GPR43 in suppressing colon cancer.

(A) Weight change (%) and survival rate (%) of mice during AOM/DSS treatment. (B) Gross colon images and tumor formation in the colon. (C) Enlargement of spleen and MLN. BM from WT or Gpr43^−/− mice (10 million cells) were transferred into lethally irradiated WT or Gpr43^−/− mice, and the chimera mice were treated with AOM/DSS for induction of colon cancer. Chimeras were treated with AOM (10 mg/Kg of body weight, i.p.) and 3 cycles of 1% DSS in drinking water and sacrificed at day 70 post-AOM injection. Representative and pooled data (mean ± SEM, n=7–8) from 2 independent experiments are shown. *Significant differences (P < 0.05, for weight change by two-way ANOVA followed by with Bonferroni’s multiple comparison (A) and by one-way ANOVA followed by Tukey’s Multiple Comparison Test (B, C) from “WT → WT” group. ** Significant differences from “Gpr43^−/− mice → WT” group.

GPR43 partially mediates the protective effect of DF and SCFAs

DF and SCFAs have SCFA receptor-independent functions. For example, SCFAs function as HDAC inhibitors or metabolic regulators [36], which do not require SCFA receptors. To determine if the protective effect of DF on AOM/DSS-induced colon cancer is GPR43-dependent, WT and Gpr43^−/− mice were fed low-fiber diet (LFD, 0% soluble DF: 1:1= pectin and inulin), medium-fiber diet (MFD, 5% soluble DF), or high-fiber diet (HFD,15% soluble DF) and were examined for their susceptibility to AOM/DSS-induced carcinogenesis. As demonstrated in the literature, DF suppresses colon cancer development in WT mice in a dose-dependent manner (Figure 7). DF also suppressed carcinogenesis in Gpr43^−/− mice but this effect was smaller than that in WT. This indicates that DF can suppress carcinogenesis in both Gpr43-dependent and independent manners.

FIGURE 7. Effect of DF on AOM/DSS-induced colon cancer development in WT and Gpr43^−/− mice.

(A) Colon tissues of WT or Gpr43^−/− mice fed different levels of DF (LFD, MFD, and HFD) and treated with AOM/DSS. (B) Weight change (%) of WT and Gpr43^−/− mice during AOM/DSS treatment. (C) Tumor formation in the colon of WT and Gpr43^−/− mice. Mice were treated by AOM (10 mg/Kg of body weight, i.p.) and 3 cycles of 1% DSS in drinking water and sacrificed at day 70 post AOM injection. Representative and pooled data (mean ± SEM, n=7–9) from 2 independent experiments are shown. *Significant differences (P < 0.05) from LFD-fed mice by two-way ANOVA followed by with Bonferroni’s multiple comparison (B) and by one-way ANOVA followed by Tukey’s Multiple Comparison Test and 2-tailed Mann Whitney U test (C).

We next studied if GPR43 mediates the protective effect of C3. C3 was chosen because it is a high affinity ligand for GPR43 and an efficient inhibitor of HDACs [15, 16, 37]. The increased AOM/DSS-induced cancer development, along with spleen enlargement, in LFD-fed WT mice was suppressed by C3 administered in drinking water (Figure 8A and B). C3 decreased cancer formation and spleen enlargement in WT mice and in Gpr43^−/− mice to a lesser degree (Figure 8C and D). Interestingly, C3 and DF (i.e. MFD) had additive suppressive effects on cancer development and spleen enlargement (Figure 8C-E). Overall these results indicate that DF and C3 suppress cancer development in Gpr43-dependent and independent manners.

FIGURE 8. Effect of propionate on AOM/DSS-induced colon cancer development in WT and Gpr43^−/− mice.

(A) Tumor formation in the colon of WT mice fed LFD alone or LFD and C3 (80 mM). (B) Enlargement of spleen following AOM/DSS treatment. (C) Tumor formation in the colon of mice fed indicated diet and drinking water. (D) Histological assessment of colon tissues. Paraffin-embedded tissue sections (6 μm) were stained with hematoxylin and eosin. (E) Spleen enlargement. Mice were fed indicated diet and drinking water and treated with AOM (10 mg/Kg of body weight, i.p.) and 3 cycles of 1% DSS in drinking water and sacrificed at day 70 post AOM injection. Representative and pooled data (mean ± SEM, n=5–7) from 2 independent experiments are shown. *Significant differences (P < 0.05) from AOM/DSS-treated LFD-fed mice by 2-tailed Mann Whitney U test (A, B, E) and one-way ANOVA followed by Tukey’s Multiple Comparison Test (C, E).

Defective barrier function and increased tissue bacteria contribute to intestinal carcinogenesis in Gpr43^−/− mice

Defective regulation of gut permeability in inflammatory conditions has been reported for Gpr43^−/− mice [13]. We found that the expression of ocln (the gene for Occludin) at mRNA level was greatly decreased in the colon of Gpr43^−/− mice after 3 cycles of 1.5% DSS treatment (Figure 9A). C3 increased the expression of ocln and tjp1 (the gene for ZO-1) in the colon of WT mice but failed to increase their expression in Gpr43^−/− mice. Consistently, gut permeability was higher in Gpr43^−/− than WT mice following the chronic DSS treatment (Figure 9B). We performed fluorescent in situ hybridization (FISH) for eubacterial 16S RNA gene and quantitative (q)-PCR analysis to detect bacteria in colon tissues. The abnormal gut permeability in Gpr43^−/− mice was supported by increased levels of eubacteria in colon tissues (Figure 9C). We found that many more bacteria were detected within the colon tissues of Gpr43^−/− compared to WT mice during chronic inflammation. Importantly, C3 feeding decreased tissue bacteria (Figure 9C). We also examined tissue bacteria during AOM/DSS-induced carcinogenesis in the colon. Tissue bacterial load was not different in co-housed WT and Gpr43^−/− mice at an early stage but later it was decreased in WT but remained elevated in Gpr43^−/− mice (Figure 9D), indicating defective control of tissue bacterial load at the chronic stage in GPR43 deficiency.

FIGURE 9. Effects of propionate and GPR43 on gut barrier function and tissue bacterial load during inflammation and colon cancer development in WT versus Gpr43^−/− mice.

In panel A-C, WT and Gpr43^−/− mice were fed regular rodent chow and C3 (80 mM) in drinking water and treated with 3 cycles of 1.5% DSS and regular water. (A) Expression of tight junction protein genes (ocln and tjp1) at mRNA level in colon tissues of WT and Gpr43^−/− mice after DSS treatment. qRT-PCR was performed. (B) Gut permeability in WT and Gpr43^−/− mice after DSS treatment. Shown is FITC-dextran concentration in blood. (C) Tissue-associated bacteria in the colon following DSS treatment. Confocal microscopy and q-PCR was performed. For q-PCR, DNA isolated from colon tissues was PCR-amplified with primers for total eubacteria and the mouse β-actin gene. Relative levels of eubacterial levels normalized by β-actin gene signals and then by control WT levels are shown. (D) Tissue bacteria load in co-housed mice following AOM/DSS-induced colon cancer. Mice were treated with AOM and 3 cycles of 1% DSS in drinking water and sacrificed at day 62–63 post AOM injection. (E) Tumor formation in the colon from WT and Gpr43^−/− mice with or without antibiotics (Abx) treatment. Representative and pooled data (mean ± SEM, n=7–9 mice) from 2 experiments are shown. One-way ANOVA followed by Tukey’s Multiple Comparison Test was used. In panels A-C, significant differences between indicated pairs (*) using 2-tailed Mann-Whitney U-test. Mann-Whitney U-test (2-tailed) was used to compare two groups of interest (D). In panel E, significant differences from WT Abx-treated mice (*) or from non-Abx-treated groups (**) by one-way ANOVA followed by Tukey’s Multiple Comparison Test.

To understand the role of the increased tissue bacteria in Gpr43^−/− mice, mice were treated with antibiotics during intestinal carcinogenesis. Antibiotics suppressed AOM/DSS-induced colon cancer in Gpr43^−/− mice (Figure 9E). Antibiotics suppressed the numbers of bigger tumors (>1 mm) in Gpr43^−/− mice than that in WT mice. Antibiotics also significantly decreased the enlargement of spleen in Gpr43^−/− mice (not shown). Overall, these results indicate that increased colon cancer formation in Gpr43^−/− mice is due in part to high levels of tissue bacteria.

Discussion

We studied, in depth, the impact of DF, SCFAs, and SCFA receptors on tissue bacterial load, intestinal inflammation and carcinogenesis. Based on early in vitro data with cell lines, it has been proposed that SCFAs inhibit HDACs to suppress cancer cells [38–40]. However, this mechanism alone does not fully explain the SCFA effect on carcinogenesis. A mounting body of evidence, including the data presented in this report, suggests that the protective effect of SCFAs is, in part, mediated by SCFA receptors [27, 31], which function in a HDAC-independent manner. We found that SCFAs and SCFA receptors promote barrier immunity and suppress bacterial invasion that can induce chronic inflammatory responses. Importantly, SCFAs, exemplified by C3, and GPR43 have suppressive effects on chronic inflammatory responses and intestinal carcinogenesis.

We studied both GPR43 and GPR41 but focused our efforts on GPR43 because the anti-cancer effect of GPR43 was greater than that of GPR41. The two GPCRs are receptors for C2, C3 and C4 but their affinity to these SCFAs are different [16]. While both receptors are expressed by intestinal epithelial cells, the distribution of the two in the body are very different [3, 37]. For example, GPR43 is expressed by neutrophils, whereas GPR41 is expressed by adipocytes. Also, the two may be expressed by different cell types in the intestine. These differences in receptor expression are likely to account for the observed differences in SCFA receptor functions. Our BM transplantation experiment revealed that non-BM-derived tissue cells play a major role in mediating the GPR43-dependent suppression of intestinal inflammation and carcinogenesis, while hematopoietic cells play a relatively minor role. The most likely cells to mediate the GPR43 function are intestinal epithelial cells, which highly express GPR43 and are effectively stimulated by SCFAs for barrier immunity [13]. This finding is in line with a previous report that non-hematopoietic cells are the major mediator of the GPR109A effect [31]. However, a minor but significant role of hematopoietic cells was still observed in both studies.

SCFAs promote the expression of intestinal tight junction proteins in the steady state and boost anti-microbial responses during infection or following gut barrier breach [14, 41]. T cell responses in the intestine reflect the inflammatory status regulated by SCFAs and their receptors. Particularly, Th17 cells, which contribute to barrier immunity, undergo expansion during an acute phase of immune responses but contraction as inflammatory responses subside. Numbers of Th17 cells remain increased in Gpr43^−/− mice along with chronic inflammation. This effect somewhat mirrors that of GPR109A [31]. These results identify important roles of SCFAs and their receptors in mounting innate and adaptive immune responses to suppress bacterial invasion across the barrier. This finding clarifies the rather unexpected attenuated inflammation observed by Sina et al at an acute phase of DSS-induced inflammation [29]. It has been recently reported that GPR43 and GPR109A promote epithelial inflammasome responses [28]. Our results together with these reports emphasize the defective acute inflammatory response and associated-gut permeability change in Gpr43^−/− mice following DSS or AOM/DSS treatment, observed in this study

Sustained high bacterial burden in colon tissues can cause chronic inflammatory responses. For example, inflammatory cytokines and chemokines, such as IL-18, IL-22, IL-6 and CXCL1, which promote effective barrier immunity to ward off invading bacteria [42–44], were not normally induced at the early phase in Gpr43^−/− mice. During prolonged treatments with DSS or AOM/DSS, Gpr43^−/− mice suffer less than WT mice in the first few weeks but become more susceptible afterward. This indicates that SCFA receptors function to coordinately regulate acute and chronic inflammatory responses. The function of SCFA receptors in regulating chronic inflammation is closely related to their function in suppressing intestinal cancer. In APC^min/+ mice, tumors are formed more in the small, than large, intestine, and thus this model provides additional insights into SCFA-receptor-regulated cancer outside the colon. The phenotype of this model indicates that the SCFA-GPR43 axis has anti-cancer effects in both the small and large intestines.

Inflammation is frequently linked to carcinogenesis. Our results indicate that DF and SCFAs, as represented by C3 in this study, suppress intestinal cancer in GPR43-dependent and independent manners. Our findings are in line with others’ findings that DF, SCFAs and SCFA receptors exert suppressive effects on colon cancer [27, 31, 45–48] and provide a potential mechanism for the suppressive function. Our findings underscore the important role of the SCFA- SCFA receptor axis in suppressing bacterial invasion, resulting chronic inflammation, and inflammation-associated intestinal carcinogenesis. Colon cancer development is closely associated with gut microbiota changes [49–51]. DF changes the gut microbiota. For example, DF enriches Bacteroidaceae and Bifidobacteriaceae and potentially decrease pathogenic bacteria [52]. Firmicutes, particularly those of the Erysipelotrichaceae family, are increased in DF deficiency. Moreover, GPR43 functions to increase beneficial bacterial species as mentioned in the introduction [27]. In this study, we observed that co-housing with WT-littermates failed to suppress elevated chronic Th17 response, bacterial invasion and associated colon cancer development in Gpr43^−/− mice in response to AOM/DSS-treatment. This indicates that control of microbiota community by GPR43 does not have a dominant role on colon cancer development, which underlines the importance of the SCFA effect on host cells rather microbiota in exerting their protective effect. Our findings provide a new insight into the protective effect of the SCFA-GPR43 axis in regulating colon cancer development. We conclude that the SCFA-GPR43 axis functions to boost barrier immune responses against microbiota and this response limits persistent bacterial invasion, chronic inflammatory responses and colon cancer development.

Materials and Methods

Mice, diet and SCFA administration

All animal experiments were approved by the Purdue Animal Care and Use Committee (PACUC). CD45.1 C57BL/6 mice and Apc^Min/+ mice were from the Jackson Laboratory (Bar Harbor, ME). Gpr43^−/− mice were purchased from Deltagen (San Mateo, CA), and Gpr41^−/− mice were obtained from Dr. M Yanagisawa (UT Southwestern Medical Center at Dallas). All mice were on C57BL/6 background. Gpr43^−/− mice were crossed with Apc^Min/+ mice to derive Apc^Min/+Gpr43^−/− mice. The mice were kept at the Purdue Life Science Animal Facility for at least 12 months prior to the study. Mice were fed regular rodent chow (Harlan 2018S Global 18% Protein Rodent Diet) or special DF diets. The special diets containing different amounts of DF (0% of pectin and inulin at 1:1 ratio for LFD, 5% for MFD, and 15% for HFD) were custom-ordered from Envigo. Mice were fed special diets from 3 weeks of age and/or drinking water containing C3 (80 mM, pH 7.4). Mice were treated with DSS or AOM/DSS from 5 or 6 weeks of age for most mice or 11–12 weeks of age for BM chimera mice. When indicated, mice were treated with antibiotics (ampicillin, neomycin, and metronidazole at 1 g/L and vancomycin at 0.5 g/L). When indicated (Figure 4, 9D, and supplementary Figure 2), female mice were co-housed for 1–2 weeks before and during the induction of AOM/DSS-induced colon cancer. Alternatively, WT and Gpr43^−/− mice were housed in cages inoculated with mixed soiled bedding from WT and Gpr43^−/− mice (20% old mixed and 80% new bedding; changed every 3–4 days). The data obtained with the two methods were similar and therefore were combined.

DSS-induced acute and chronic intestinal inflammation

For acute intestinal inflammation, mice were on 3.5% DSS in drinking water for 5 days and then on plain water before they were euthanized on day 9. For chronic intestinal inflammation, mice were treated with 1.5% DSS in drinking water for 7 days and then with regular water for 7 days. This cycle was repeated three times. Intestinal tissues were examined by flow cytometry for T cells and neutrophils and by qRT-PCR for expression of inflammatory cytokines (IL-6, IL-18, IL-22, and CXCL-1) as described previously [13, 53]. Primers used for this study were previously described [13].

Tight junction proteins, tissue bacterial burden, and cecal bacterial 16S rRNA gene PCR analysis

Mice were assessed for expression of tight junction protein genes (ocln and tjp1), gut permeability, and tissue bacteria 42 days after the start of DSS treatment. The expression of ocln and tjp1was determined by qRT-PCR with the primers: ACC ACT ATG AAA CAG ACT ACA CG (Ocln-F), GGA GAA CTA CAC GCT ATT AAA CG (Ocln-R), GAA GAT GAA GAT GAG GAT GGT CA (tjp1-F), ACC AGT GAT TAG TCC TGT CTT TGT (tjp1-R). Tissue bacterial levels were determined by q-PCR of a common eubacterial 16S rRNA gene sequences and normalized by genomic β-actin levels as previously described [13]. Cecal bacterial 16S rRNA gene PCR analysis for cohoused animals was performed as described previously [13, 54].

Gut permeability

For assessing gut permeability, mice were fasted overnight and orally gavaged with fluorescein isothiocyanate (FITC)–conjugated dextran (0.2 mg/g of body weight, mean molecular weight of 3000–5000; Sigma- Aldrich, St Louis, MO) using a round-tip feeding needle. Mice were sacrificed 3 h later, and the FITC-dextran concentration in the plasma was determined with a fluorescent microplate reader (excitation at 485 nm and emission at 535 nm; Synergy HT; BioTek, Winooski, VT).

AOM/DSS and APC^Min/+-induced models of intestinal cancer

Mice were treated i.p. with azoxymethane (AOM, 10 mg/kg of body weight) and rested for 7 days. The mice were then administered with 1% DSS (36–50 kDa) in drinking water for 7 days and rested on regular water for 7 days. This cycle was repeated 3 times. Mice were monitored daily for changes in body weight, signs of illness, and stool score. Mice were sacrificed 62–70 days post-AOM injection and were assessed for polyp size and number and organ (spleen and MLN) weight. Colon tissues were paraffin-embedded and tissue sections were stained with hematoxylin and eosin. Apc^Min/+ mice and Apc^Min/+Gpr43^−/− mice were similarly assessed for tumors in the small and large bowels and organ weight at 4 months of age.

Cell isolation and flow cytometry

For isolation of intestinal LP cells, intestinal tissues were harvested into cold PBS and trimmed to remove mesenteries, fat, and Peyer’s patches. Intestinal tissues were cut into ~2 cm pieces. The epithelial layer was extracted five times with 10 mM EDTA in HBSS buffer at room temperature for 15 min, and then the remaining tissues were digested twice at 37°C for 1 hour with 1.5 mg/ml collagenase (type 3, Worthington). Collected cells were stained with antibodies to Gr-1 (RB6–8C5), CD11b (M1/70) and CD45 (30-F11) for neutrophils or to CD4 (RM4–5) for T cells. For intracellular staining of IL-17 and IFN-γ, cells were activated in RPMI 1640 (10% fetal bovine serum) with phorbol myristate acetate (50 ng/ml), ionomycin (1 mM), and monensin (2mM; Sigma- Aldrich) for 4 h. Cells were then fixed, permeabilized, and stained with antibodies to mIL-17A (TC11–18H10.1) or IFNγ (XMG1.2). For FoxP3 staining, cells were fixed, permeabilized, and then stained with FoxP3 (FJK-16S) according to the manufacturer’s protocol (eBioscience).

Bone marrow transplantation

To prepare host mice, WT or Gpr43^−/− mice were irradiated at 500 rads, and this was repeated after 4 h. Host mice were injected i.v. with 1 × 10⁷ BM cells from WT or Gpr43^−/− mice. The CD45.1 and CD45.2 congenic system was used to distinguish donor and recipient cells. Six weeks later, mice were checked for hematopoietic reconstitution and were treated with AOM and DSS. The marrow reconstitution efficiency in these animals was >95% (spleen).

Confocal microscopy and histological analysis

Colon tissues, harvested from WT or Gpr43^−/− mice subjected to chronic DSS treatment, were frozen in Tissue-Tek OCT compound (Sakura) and were cut into 8-µm sections. The sections were fixed in cold acetone and stained with CY3-labled EU338 (5′-GCT GCC TCC CGT AGG AGT-3, Eurofins) and Pan-cytokeratin. Confocal images were collected with a SP5 II laser scanning microscopic system (Leica). For histological analysis, sections (6 μm) of paraffin-embedded intestinal tissues were stained with hematoxylin and eosin. Histological images were taken at 50× magnification.

Statistical analysis

Two-way ANOVA followed by Bonferroni’s multiple comparison was used for repeated measure data (weight change and/or stool score in Figure 1A, 2A, 3A, 6A, 7B). We used one-way ANOVA followed by Tukey’s Multiple Comparison Test for differences among multiple groups (Figure 1B-E, 2B-E, 6B-C, 7C, 8C, 8E, 9A-C, E). Additionally, Mann-Whitney U-test (2-tailed) was used to compare two groups of interest (Figure 3B, 3D, 4B-E, 5A, 5C, 7C, 8A-B, 8E, 9D). P values smaller than or equal to 0.05 were considered significant. All error bars indicate SEM. Prism (Graphpad, La Jolla, CA) was used for statistical analyses.

Abbreviations:

Abx

antibiotics

BM

bone marrow

AOM

azoxymethane

C2

acetate

C3

propionate

C4

butyrate

DF

dietary fiber

DSS

dextran sodium sulfate

HFD

high fiber diet

GPCR

G-protein-coupled receptors

LFD

low fiber diet

LP

lamina propria

MFD

medium fiber diet

SCFAs

short-chain fatty acids

q-PCR

quantitative PCR