Bifidobacterium can mitigate intestinal immunopathology in the context of CTLA-4 blockade

Significance

The major stumbling block in the use of checkpoint inhibitors for cancer treatment is the severe autoimmunity that often results. In this study, we found the toxicity of a checkpoint blockade antibody can be ameliorated via administration of Bifidobacterium, a widely available probiotic. These results suggest that it may be possible to mitigate the autoimmunity caused by anti–CTLA-4 and perhaps other checkpoint inhibitors by manipulating gut microbiota.

Abstract

Antibodies that attenuate immune tolerance have been used to effectively treat cancer, but they can also trigger severe autoimmunity. To investigate this, we combined anti–CTLA-4 treatment with a standard colitis model to give mice a more severe form of the disease. Pretreatment with an antibiotic, vancomycin, provoked an even more severe, largely fatal form, suggesting that a Gram-positive component of the microbiota had a mitigating effect. We then found that a commonly used probiotic, Bifidobacterium, could largely rescue the mice from immunopathology without an apparent effect on antitumor immunity, and this effect may be dependent on regulatory T cells.

Cancer immunotherapy has focused on harnessing the host immune system to stimulate an antitumor response. Monoclonal antibodies (mAbs) that block immune inhibitory pathways, specifically the CTLA-4 pathway and the PD-1/PD-L1 axis, have been successfully used in the clinic to improve the overall survival of patients with advanced cancers (1–3). However, immune checkpoint blockade has the drawback of frequently producing immune-mediated effects on various organ systems that can lead to autoimmunity, most commonly colitis (4). This checkpoint blockade-associated toxicity can be serious and life threatening and therefore requires prompt and appropriate management to ensure that it is relatively quickly resolved (5), although this can fail to prevent life-long autoimmune syndromes or death.

Recently, studies have begun to shed light on the crucial role of the gut microbiota in antitumor responses induced by checkpoint blockade antibodies (6, 7). These syndromes most frequently affect sites that are exposed to commensal microorganisms (i.e., the skin and gastrointestinal tract). However, there remains a lack of knowledge regarding how the gut microbiota influences this type of toxicity. Here, we show that the gut microbiota of mice can be optimized by the oral administration of appropriate antibiotics followed by supplementation with probiotics. This optimization allows checkpoint blockade to achieve the desired immune response of stimulating antitumor immunity with minimal immunopathology.

To establish a checkpoint blockade-related autoimmune mouse model, mice were tested to determine their response to orally administered dextran sulfate sodium (DSS) with an injection of the anti–CTLA-4 antibody 9D9 or an isotype control. After the administration of 2% or 3% DSS in drinking water for 7 d, mice that received an anti–CTLA-4 antibody showed more severe weight loss compared with mice that received DSS plus the isotype control antibody (Fig. 1 A and B). No significant weight loss was observed in anti–CTLA-4-treated mice in the absence of DSS (Fig. S1). Disease was particularly severe in mice that received 3% DSS plus anti–CTLA-4 antibodies and 40% of the mice died (Fig. 1C). In addition, H&E-stained histological sections from the colons of mice show that the combined treatment showed exacerbated hyperplasia, inflammatory leukocyte infiltration, and ulceration compared with controls, with worse histopathological scores (Fig. 1 D and E). We also observed a therapeutic effect of the anti–CTLA-4 antibody against established B16F10 melanoma in the same mice (Fig. S2) as reported previously (7). Thus, the data from our colitis and tumor models are consistent with clinical observations of patients who received ipilimumab, an antibody against CTLA-4 (8), where colitis is the most frequent problem encountered.

Fig. 1.

Increased susceptibility of DSS-induced colitis in anti–CTLA-4 Ab-receiving mice. (A and B) Percent of initial weight of mice receiving the IgG isotype control (Iso Ctrl) or the αCTLA-4 mAb. Mice were given 2% (A) or 3% DSS (B) for 7 d. There were five mice in each group. The data show mean with SEM analyzed by two-way ANOVA with Sidak’s correction for multiple comparisons. (C) Survival of the mice receiving the IgG isotype control (Iso Ctrl) or the αCTLA-4 mAb with 4% DSS administration. Survival was monitored for 14 d, n = 5 per group. (D) The histological score of mice receiving the IgG isotype control (Iso Ctrl) or the αCTLA-4 mAb with 3% DSS administration, n = 6 per group, means with SEM analyzed by unpaired Student’s t test. (E) Representative colon sections from mice treated with injection of isotype control (Left) or αCTLA-4 mAb (Right) and 3% DSS administration. Colon samples were collected at day 14 (H&E stained). (Scale bar, 50 μm.) *P < 0.05 and **P < 0.01.

A previous study showed that vancomycin worsens histopathological signs of gut inflammation but promotes enhanced antitumor immunity (7) (Fig. S3). Using our DSS plus anti–CTLA-4 colitis model, we assessed the impact of vancomycin on the severity of colitis under CTLA-4 blockade. Mice were pretreated with either vancomycin or a water control for 2 wk before the induction of colitis. In the vancomycin group, mice began to lose body weight after only 4 d of DSS and anti–CTLA-4 treatment (Fig. 2A), showing a much earlier onset of colitis than without antibiotic treatment (9, 10). Consistently, more severe weight loss was observed in mice pretreated with vancomycin than in mice that received the water control (Fig. 2A). By day 10, all mice in the control group remained alive, whereas 80% of the vancomycin plus DSS and anti–CTLA-4 treated mice had died (Fig. 2B). Histopathological scores were also significantly worse for vancomycin-treated mice than for the controls (Fig. 2C). H&E staining of colon sections showed severe immune cell infiltration and complete ulceration in vancomycin-treated mice (Fig. 2D). Similar serum levels of three inflammatory cytokines, KC, IL-6, and CFS3, were also dramatically increased in vancomycin-treated mice (Fig. 2E).

Fig. 2.

Vancomycin augments the immunopathology of CTLA-4 blockade. (A) Percent initial weight of H₂O- or vancomycin-treated mice after injection of the αCTLA-4 mAb and the administration of 3% DSS. In each group, n = 5. ****P < 0.0001; Data show means with SEM analyzed by two-way ANOVA with Sidak’s correction for multiple comparisons. (B) Survival of H₂O- or vancomycin-treated mice after injection of the αCTLA-4 mAb and administration of 3% DSS, with n = 5 per group. (C) Histological score of H₂O- or vancomycin-treated mice after injection of the αCTLA-4 mAb and administration of 3% DSS, n = 5 per group, ****P < 0.0001, unpaired Student’s t test. (D) Representative colon sections of mice treated with H₂O or vancomycin after injection of αCTLA-4 mAb and 6-d administration of 3% DSS (H&E stained). (Scale bar, 100 μm.) (E) Concentrations of KC, IL-6, and CSF3 in the serum of mice treated with H₂O or vancomycin after CTLA-4 blockade and 6-d administration of 3% DSS, n = 5–7 per group. *P < 0.05, unpaired Student’s t test.

Vancomycin is a broad-spectrum antibiotic with activity against Gram-positive bacteria. We hypothesized that we might be able to substitute for this apparent protective effect of the microbiota in our system with one of the most commonly used probiotics, Bifidobacterium (a genus of Gram-positive anaerobic bacteria), which has been suggested as an effective treatment for inflammatory bowel disease (11–16). Indeed, Quantitative PCR (qPCR) analyses targeting Bifidobacterium species showed that the administration of vancomycin decreased the abundance of these species to an undetectable level (Fig. 3A). To test whether this loss of Bifidobacterium strains or others contributed to the severe colitis, we obtained a commercially available mixture of four Bifidobacterium species and administered this mixture to mice via oral gavage before the induction of DSS colitis. This Bifidobacterium treatment resulted in a 10-fold increase in the relative abundance of these bacteria in feces (Fig. 3A).

Fig. 3.

Bifidobacterium ameliorates the immunopathology, but does not affect antitumor immunity of vancomycin and CTLA-4 blockade. (A) The relative abundance of Bifidobacterium was quantified with real-time PCR. This was normalized to total bacteria, with n = 4–5 per group. **P < 0.01, unpaired Student’s t test. (B) Percent initial weight of 2.5% DSS-induced colitis in αCTLA-4 mAb-injected mice treated with H₂O + PBS, vancomycin + PBS, or vancomycin + Bifidobacterium. In each group, n = 5. **P < 0.01, ****P < 0.0001. Data show means with SEM analyzed by two-way ANOVA with Sidak’s correction for multiple comparisons. Histological score (C) and representative colon sections (D) of 2.5% DSS-induced colitis in vancomycin and αCTLA-4 mAb-treated mice oral gavaged with PBS or Bifidobacterium, n = 5 per group, **P < 0.01, unpaired Student’s t test (H&E stained). (Scale bar, 50 μm.) (E) Concentrations of KC, IL-6, and CSF3 in serum from mice oral gavaged with PBS or Bifidobacterium. These mice were also repeatedly injected i.p. with the αCTLA-4 mAb and a 6-d administration of 2.5% DSS, n = 5–7 per group. *P < 0.05; **P < 0.01, unpaired Student’s t test. B16F10 tumor growth kinetics (F) and tumor size at day 18 postimplantation (G) in mice pretreated with vancomycin followed by treatment with PBS or Bifidobacterium by oral gavage. The αCTLA-4 mAb was injected at 3, 6, 10, and 13 d posttumor implantation. n.s., not significant.

In our CTLA-4 blockade condition, Bifidobacterium treatment resulted in substantially less weight loss; the average weight of vancomycin-treated mice improved from 75% of their initial weight in the DSS plus anti–CTLA-4 group to 95% in the Bifidobacterium group on day 8 after DSS administration (Fig. 3B, blue and pink lines). Mice treated with both Bifidobacterium and vancomycin exhibited less weight loss than control mice (Fig. 3B, yellow and pink lines), suggesting that Bifidobacterium treatment ameliorated the immunopathology associated with CTLA-4 blockade by helping to rescue vancomycin-induced gut dysbiosis. Consistent with this assessment, H&E staining of colon sections revealed that Bifidobacterium treatment resulted in a reduced histopathological score with partial restoration of the colon structure and less leukocyte infiltration into gut tissue (Fig. 3 C and D). Bifidobacterium treatment also decreased serum levels of the inflammatory cytokines KC, IL-6, and CFS3 in colitic mice (Fig. 3E).

Importantly, we found that the growth kinetics of established B16F10 melanoma tumors were not affected by the Bifidobacterium treatment (Fig. 3F). Tumors of comparable size were found in Bifidobacterium-treated mice and PBS-treated control mice at day 18 postinoculation (Fig. 3G). These data suggested that Bifidobacterium ameliorated gut immunopathology without compromising the therapeutic efficacy of vancomycin and CTLA-4 blockade against melanoma in this system.

We then investigated the immunologic mechanism underlying the observed amelioration of colitis in Bifidobacterium-treated mice. Interestingly, we found that the protective effect of Bifidobacterium feeding was abrogated in regulatory T cell (T_reg)-depleted mice (Fig. 4A). PBS and Bifidobacterium-treated mice developed equally severe colitis with elevated serum inflammatory cytokine levels after T_reg cell depletion (Fig. 4 B and C). These data indicate that the effects of Bifidobacterium are dependent on the host immune regulatory response. We next tested whether administration of Bifidobacterium increases the frequency of T_reg cells as has been reported in other studies of microbiome influence on these cells (17–22), but we did not see any differences between PBS- and Bifidobacterium-treated mice with respect to CD4⁺ Foxp3⁺ T cells isolated from the spleen or lamina propria (LP) (Fig. S4). Instead, we used RNA sequencing to profile the genome-wide transcriptional pattern of LP T_reg cells isolated from Bifidobacterium-treated mice versus controls. Gene Ontology (GO) analysis of differentially expressed genes identified nine metabolic pathways, including cellular macromolecules, nitrogen compounds, and organic substances (Fig. S5). Thus, it is likely that Bifidobacterium acts primarily via modulating these metabolic functions of T_reg cells (23–25) without either systematically or locally affecting the number of T_reg cells.

Fig. 4.

Immune regulatory function of Bifidobacterium depends on T_reg cells. (A) Percent initial weight of DSS colitis in vancomycin and αCTLA-4 mAb-treated FoxP3-DTR mice after oral gavage with PBS or Bifidobacterium under control or T_reg depletion condition (with DTX injection). Histological score (B), and concentrations of serum KC (C) were measured in DTX-injected mice in A at day 7. For each group, n = 5. Data show means with SEM analyzed by two-way ANOVA with Sidak’s correction for multiple comparisons in A. Unpaired Student’s t test in B and C. DTX, diphtheria toxin; **P < 0.01; n.s., not significant.

In summary, our study demonstrates a role for Bifidobacterial strains in ameliorating the immunopathology associated with CTLA-4 blockade. If the same mechanisms are at work in human beings, this may present a way to reduce or eliminate the autoimmunity that often accompanies checkpoint blockade therapies without diminishing the anticancer responses. This principle could apply to other checkpoint-related immunotherapies, such as antibodies targeting the PD-1/PD-L1 axis. Importantly, recent studies have indicated that Bifidobacterium species promote antitumor response in PD-L1 blockade by augmenting dendritic cell function (6). Thus, it would be valuable to explore the role of Bifidobacterium in the combined blockade of the CTLA-4 and PD-1 pathways. Our work also suggests caution in combining antibiotics with anti–CTLA-4 treatment. Our results indicate a role for T_reg cells in colitis-ameliorating function of Bifidobacterium, and we hope this stimulates further investigations in the molecular mechanism underlying this process. In addition, since a failure in T_reg activity has been implicated in both autoimmunity and allergy generally, the type of enhancement provided by this widely available probiotic that we describe here may be helpful in controlling those diseases as well (26).

Materials and Methods

Mice.

C57BL/6 and Foxp3-DTR mice were purchased from The Jackson Laboratory. For all experiments, 6- to 14-wk-old female mice were used. Mice were maintained in the Research Animal Facility at the Stanford University Department of Comparative Medicine Animal Facility in accordance with guidelines of the National Institutes of Health. Mice experiments were approved by the Stanford University Administrative Panel on Laboratory Animal Care.

Induction of DSS Colitis.

Mice received 2–4% DSS (MP Biomedicals) in their drinking water for 6–7 d. Weight was recorded daily. For gut commensal manipulation, mice were provided with vancomycin (0.5 g/L; Sigma) in the drinking water for at least 14 d. After that, DSS was added to the antibiotic-containing water. Mice were injected with 100 μg of anti–CTLA-4 mAb (clone 9D9) or isotype control twice (started 1–3 d before the DSS administration).

Melanoma Model.

Mice were s.c. injected into the right flank with 1 × 10⁵ B16F10 tumor cells. Mice were injected intraperitoneally (i.p.) with 100 μg of isotype control (clone MPC11) or anti–CTLA-4 mAb (clone 9D9). Mice were injected four times at ∼3-d intervals with the mAb. Tumor size was measured by a caliper until the mandated endpoint (500 mm²). Tumor size was determined as length × width.

Histology.

Freshly isolated colon was fixed in formalin and embedded in paraffin. H&E staining was performed by using a standard protocol. For histological quantitative analysis, five criteria were used to grade each section of intestine: (i) severity of inflammation, (ii) percent of area affected by inflammation, (iii) degree of hyperplasia, (iv) percent of area affected by hyperplastic changes, and (v) ulceration. Section of colon was graded by a pathologist.

Bifidobacterium Administration.

Lyophilized Bifidobacterium species including Bifidobacterium bifidum, Bifidobacterium longum, Bifidobacterium lactis, and Bifidobacterium breve (Seeking Health) were resuspended in PBS. Each mouse was given 300 μL of Bifidobacterium (1 × 10⁹ cfu per mouse) by oral gavage. For DSS colitis, Bifidobacterium was given before DSS administration. For the tumor model, Bifidobacterium was given at 7 and 14 d following tumor inoculation.

Quantification of Bifidobacterium by qPCR Assay.

Fecal samples were collected 4 d after oral administration of Bifidobacterium species. Bacterial DNA was isolated with the Qiagen DNA Stool Mini Kit following the manufacturer’s instructions. Taqman-targeted qPCR was applied using primers and probes specific for Bifidobacterium: Bifid forward (CGGGTGAGTAATGCGTGACC), Bifid reverse (TGATAGGACGCGACCCCA), and probe (6FAM-CTCCTGGAAACGGGTG). The relative abundance of Bifidobacterium was normalized by primers and probe targeting all bacteria: all bacteria forward (CGGTGAATACGTTCCCGG), all bacteria reverse (TACGGCTACCTTGTTACGACTT), and probe (6FAM-CTTGTACACACCGCCCGTC).

Serum Cytokine Measurement.

Blood samples were collected at days 6–7 after DSS colitis induction. After clotting at least 30 min at room temperature, serum was separated with centrifuge (10 min at 1,200 relative centrifugal force). Luminex assay was performed following the product manual at the Stanford Human Immune Monitoring Center.