Background Although immunotherapy works well in glioblastoma (GBM) pre-clinical mouse models, the therapy has not demonstrated efficacy in humans. To address this anomaly, we developed a novel humanized microbiome (HuM) model to study the response to immunotherapy in a pre-clinical mouse model of GBM. Methods We used five healthy human donors for fecal transplantation of gnotobiotic mice. After the transplanted microbiomes stabilized, the mice were bred to generate five independent humanized mouse lines (HuM1-HuM5). Results Analysis of shotgun metagenomic sequencing data from fecal samples revealed a unique microbiome with significant differences in diversity and microbial composition among HuM1-HuM5 lines. All HuM mouse lines were susceptible to GBM transplantation, and exhibited similar median survival ranging from 19-26 days. Interestingly, we found that HuM lines responded differently to the immune checkpoint inhibitor anti-PD-1. Specifically, we demonstrate that HuM1, HuM4, and HuM5 mice are non-responders to anti-PD-1, while HuM2 and HuM3 mice are responsive to anti-PD-1 and displayed significantly increased survival compared to isotype controls. Bray-Curtis cluster analysis of the five HuM gut microbial communities revealed that responders HuM2 and HuM3 were closely related, and detailed taxonomic comparison analysis revealed that Bacteroides cellulosilyticus was commonly found in HuM2 and HuM3 with high abundances. Conclusions The results of our study establish the utility of humanized microbiome mice as avatars to delineate features of the host interaction with gut microbial communities needed for effective immunotherapy against GBM.

Although the immunotherapy anti-PD-1 works well in glioblastoma (GBM) pre-clinical mouse models, the therapy has not demonstrated a similar efficacy in patient clinical trials. Recent studies have linked the gut microbe composition to tumor growth and response to immunotherapy in some cancers. To date, all GBM pre-clinical studies have been done in mouse models using mouse gut microbiomes. There are significant differences between mouse and human microbial gut compositions, with up to 85% of gut bacteria found in laboratory mice not found in humans. Because it is known that the gut microbe composition can impact the immune system, we hypothesize that the non-responsiveness of GBM patients to immunotherapy may be due to the composition of the gut microbiome. Therefore, we have generated a humanized microbiome mouse model in which mice have been colonized by human donor microbes in their GI tract (two different healthy human donors (HuM1 and HuM2)). In preliminary results, we have found that HuM1 mice are resistant to anti-PD-1, while HuM2 mice are responders to anti-PD-1 in the GL261 syngeneic intracranial model. These mice are genetically identical and only differ in gut microbiome composition. Furthermore, we found that HuM2 mice exhibited a significant increase in cytotoxic CD8+T-cells producing IFN-γ and significant increased CD8+/Treg ratio in the spleen following anti-PD-1 treatment, which was not observed in the HuM1 mice. When testing the efficacy of standard of care temozolomide (TMZ) in our humanized mice, we found that TMZ significantly prolonged survival of both HuM1 and HuM2 mice with intracranial tumors. However, HuM2 mice exhibited superior efficacy (p< 0.001; 57% survival), compared to HuM1 mice (p< 0.01; 0% survival). We are extending these studies to analyze additional humanized microbiome lines as well as GBM patient donor lines to more accurately understand individual responses to tumor growth and responsiveness to therapies.

Although immunotherapy works well in glioblastoma (GBM) pre-clinical mouse models, the therapy has unfortunately not demonstrated efficacy in humans. In melanoma and other cancers, the composition of the gut microbiome has been shown to determine responsiveness or resistance to immune checkpoint inhibitors (anti-PD-1). Most pre-clinical cancer studies have been done in mouse models using mouse gut microbiomes, but there are significant differences between mouse and human microbial gut compositions. To address this inconsistency, we developed a novel humanized microbiome (HuM) model to study the response to immunotherapy in a pre-clinical mouse model of GBM. We used five healthy human donors for fecal transplantation of gnotobiotic mice. After the transplanted microbiomes stabilized, the mice were bred to generate five independent humanized mouse lines (HuM1-HuM5). Analysis of shotgun metagenomic sequencing data from fecal samples revealed a unique microbiome with significant differences in diversity and microbial composition among HuM1-HuM5 lines. Interestingly, we found that the HuM lines responded differently to anti-PD-1. Specifically, we demonstrate that HuM2 and HuM3 mice are responsive to anti-PD-1 and displayed significantly increased survival compared to isotype controls, while HuM1, HuM4, and HuM5 mice are resistant to anti-PD-1. These mice are genetically identical, and only differ in the composition of the gut microbiome. In a correlative experiment, we found that disrupting the responder HuM2 microbiome with antibiotics abrogated the positive response to anti-PD-1, indicating that HuM2 microbiota must be present in the mice to elicit the positive response to anti-PD-1 in the GBM model. The question remains of whether the “responsive” microbial communities in HuM2 and HuM3 can be therapeutically exploited and applicable in other tumor models, or if the “resistant” microbial communities in HuM1, HuM4, and HuM5 can be depleted and/or replaced. Future studies will assess responder microbial transplants as a method of enhancing immunotherapy.

Although immunotherapy works well in glioblastoma (GBM) pre-clinical mouse models, the therapy has not demonstrated efficacy in GBM patients. Since recent studies have linked the gut microbial composition to the success with immunotherapy for other cancers, we utilized a novel humanized microbiome (HuM) model in order to study the response to immunotherapy in a pre-clinical mouse model of GBM. We used five healthy human donors for fecal transplantation of gnotobiotic mice since it is now recognized that microbe strain level differences render individual humans with a unique microbial community composition. After the transplanted microbiomes stabilized, the mice were bred to generate 5 independent humanized mouse lines (humanized microbiome HuM1-HuM5). Analysis of shotgun metagenomic sequencing data from fecal samples revealed a unique microbiome composition with significant differences in diversity and microbial composition among HuM1-HuM5 lines. We next analyzed the growth of intracranial glioma cells in the HuM lines. All HuM mouse lines were susceptible to GBM transplantation, and exhibited similar median survival ranging from 19-26 days. Interestingly, we found that HuM lines responded differently to the immune checkpoint inhibitor anti-PD-1. Specifically, we demonstrate that HuM1, HuM4, and HuM5 mice are non-responders to anti-PD-1 resulting in the death of the mice from the intracranial tumors, while HuM2 and HuM3 mice are responsive to anti-PD-1 and displayed significantly increased survival compared to isotype controls. Bray-Curtis cluster analysis of the 5 HuM gut microbial communities revealed that HuM2 and HuM3 were closely related. Detailed taxonomic comparison analysis at the top 5 across all HuM mouse lines revealed that Bacteroides cellulosilyticus was commonly found between HuM2 and HuM3 with high abundances. The results of our study establish the utility of humanized microbiome mice as avatars to delineate features of the host interaction with gut microbe communities needed for effective immunotherapy against GBM.