https://orcid.org/0000-0002-2444-9748
Abstract
The intestinal microbiota regulates normal brain physiology and the pathogenesis of several neurological disorders. While prior studies suggested that this operates through immune cells, the underlying mechanisms remain unclear. Leveraging 2 well-characterized murine models of low-grade glioma occurring in the setting of the neurofibromatosis type 1 (NF1) cancer predisposition syndrome, we sought to determine the impact of the gut microbiome on optic glioma progression.
Neurofibromatosis type 1 (Nf1)-mutant mice genetically engineered to develop optic pathway gliomas (Nf1OPG mice) by 3 months of age were reared under germ-free (GF) conditions, treated with specific cocktails of antibiotics, or given fecal matter transplants (FMTs). Intestinal microbial species were identified by 16S genotyping. Neutralizing transforming growth factor-beta (TGFβ) antibodies were delivered systemically, while in vitro experiments used isolated murine microglia and T cells. Single-cell RNA sequencing analysis was performed using established methods.
Nf1OPG mice raised in a GF environment or postnatally treated with vancomycin did not harbor optic gliomas or exhibit OPG-induced retinal nerve fiber layer thinning, which was reversed following conventionally raised mouse FMT or colonization with Bacteroides species. Moreover, this intestinal microbiota-regulated gliomagenesis was mediated by circulating TGFβ, such that systemic TGFβ neutralization reduced Nf1-OPG growth. TGFβ was shown to act on tumor-associated monocytes to induce Ccl3 expression and recruit CD8+ T cells necessary for glioma growth.
Taken together, these findings establish, for the first time, a mechanistic relationship between Bacteroides in the intestinal microbiome and NF1-LGG pathobiology, suggesting both future predictive risk assessment strategies and therapeutic opportunities.
Optic glioma growth is reduced under germ-free conditions or postnatal vancomycin treatment.
The responsible microbe, Bacteroides, regulates glioma growth through TGFβ.
TGFβ induces microglia Ccl3-mediated T cell attraction and glioma growth.
Understanding how the tumor microenvironment controls neoplastic cell growth in the brain, not only reveals potential nonneoplastic cell types for therapeutic targeting, but may also provide a conceptual framework to define how systemic exposures converge on these stromal cells to influence brain tumor biology. In this study, we leverage authenticated murine models of neurofibromatosis type 1 and low-grade (optic) glioma to demonstrate that intestinal Bacteroides regulates optic glioma progression in mice through a circulating paracrine factor (TGFβ). As a secondary messenger, TGFβ induces tumor-associated monocyte attraction of CD8+ T cells to establish a permissive microenvironment for optic glioma growth and tumor-induced retinal pathology. Importantly, these findings suggest potential treatment opportunities, as well as offer new insights into predictive risk factors, for brain tumors in this cancer predisposition syndrome.
Our approach to cancer has been revolutionized by the discovery that tumors are maintained by stromal circuits comprised of paracrine factor-mediated cellular interactions, which dictate tumor formation, progression, and response to therapy.1 These supportive tumor microenvironment (TME) circuits are being elucidated for many solid cancers, including those arising in the central nervous system.2 Within the brain, the most common tumor in children is the low-grade glioma (LGG; World Health Organization grade 1 pilocytic astrocytoma [PA]), often occurring in the setting of the neurofibromatosis type 1 (NF1) cancer predisposition syndrome.3 Histological examination of these tumors reveals that nearly half of the cells are nonneoplastic cells, including tumor-associated monocytes (TAM) and T lymphocytes.4,5 Using authenticated preclinical mouse models of Nf1-LGG that mimic nonlethal pediatric PA of the optic pathway (optic pathway glioma; OPG), prior studies from our laboratory have demonstrated that T cells are critical for mouse Nf1OPG LGG development and progression. As such, systemic depletion of T cells (anti-CD3 or anti-CD8 antibody treatment),6 inhibition of T lymphocyte brain infiltration (anti-VLA4 or anti-Ccl2 treatment6,7), genetic T cell depletion (athymic nu/nu mice8), or silencing of T cell paracrine factor production by neurons (lamotrigine treatment9) all suppress murine Nf1-OPG growth in vivo.
Given that T cells traffic into the brain from the blood or calvarial bone marrow, we hypothesized that their function could be modified by systemic disease or environmental exposures. One such LGG risk factor is asthma, which confers a protective effect in both the general population10 and in children with NF1.11 Since Nf1OPG mice develop LGGs with high penetrance (>95%) and in a predictable temporal pattern (between 6 and 12 weeks of age), we found that asthma induction in Nf1OPG mice blocked tumor formation through T cell expression of decorin and impaired TAM paracrine support of tumor growth.12 These findings raised the intriguing possibility that other extracranial exposures might similarly mediate brain tumor pathogenesis by converging on the immune circuits responsible for controlling glioma growth. One of these potential modifiers is the intestinal microbiota.13
Multiple prior reports have linked the intestinal microbiota to brain pathology. First, antibiotic treatment alters learning and memory in wild-type mice.14 Second, depletion of the intestinal microbiota in a murine model of Alzheimer’s disease decreases τ-mediated neurodegeneration.15 Third, following traumatic brain injury in germ-free (GF) mice, fecal material transplants from conventionally reared (Conv-R) mice restore neurogenesis relative to antibiotic-treated Conv-R mice.16 To explore the potential relationship between the intestinal microbiota and brain tumors, we leveraged Nf1OPG mice to demonstrate that intestinal Bacteroides drives NF1-LGG development and progression by regulating TAM chemoattraction of T cells.
Materials and Methods
Mice
Experiments were performed under an approved Washington University Institutional Animal Care and Use Committee protocol. Nf1OPG (Nf1flox/mut, hGFAP-Cre)17 and Nf1OPG-Y2083X (Nf1flox/Y2083X, hGFAP-Cre)17 mice were maintained on a strict C57BL/6 background using 12-h light/dark cycles in a barrier facility with ad libitum food and water access.
Antibiotic Treatments
Six- or twelve-week-old Nf1OPG mice were administered a cocktail of 0.5 g/L vancomycin, 1 g/L neomycin, and 1 g/L ampicillin (VNA), or 1 g/L vancomycin, neomycin, or ampicillin individually, in sterile distilled drinking water for 6 weeks. Control mice received sterile distilled water.18
Gnotobiotic Mice
Gametes were harvested from 2 12-week-old Nf1flox/mut and Nf1flox/flox; hGFAP-Cre males, as well as 10 12-week-old Nf1flox/mut and Nf1flox/flox; hGFAP-Cre females, for in vitro fertilization. Zygotes were transferred to a GF pseudopregnant dam one day postmating with a vasectomized GF male. F0 pups were born within the flexible film isolator in the Washington University Gnotobiotic Core. The GF status of the embryo recipient females and pups was confirmed and monitored by (a) shotgun metagenomic analysis (Transnetyx), (b) aerobic bacteria, anaerobic bacteria, and fungus cultures, and (c) fecal sample wet mounts (Charles River Labs). The resulting GF progeny were intercrossed to generate GF Nf1OPG (Nf1flox/mut; GFAP-Cre) mice. The corresponding generation of Conv-R mice served as controls.
Fecal Transplantation
Fecal samples collected from 4 to 5 12-week-old Conv-R Nf1OPG mice were pooled, homogenized in sterile reduced phosphate-buffered saline (PBS) (5 pellets/10 mL), and stored at −80°C until fecal microbiota transplantation (FMT). A 200-μL aliquot was administered to sex-matched 9-week-old GF mice via oral gavage. The resulting FMT-GF mice were housed in plastic flexible film gnotobiotic isolators, with 2–5 mice of the same sex per cage. Nf1OPG mice treated with vancomycin from 6 to 9 weeks of age were administered 1 or 2 doses of a 200-μL aliquot at 9 weeks or 9 and 10 weeks of age, respectively.
Bacteroides Treatment
Bacteroides cultures were generated at 37oC in a vinyl anaerobic chamber containing a 70% N2, 20% CO2, and 10% H2 gas mixture. Bacteroides isolates were initially grown on BD Difco Brain Heart Infusion (BHI) agar, supplemented with 10% defibrinated sheep blood, and subsequently in tryptone-yeast extract-glucose broth (10 g/L tryptone, 5 g/L yeast extract, 2 g/L glucose, 500 mg/L cysteine free base, 100 mM pH 7.2 potassium phosphate, 20 mg/L MgSO4-7H2O, 400 mg/L NaHCO3, 80 mg/L NaCl, 100 mg/L resazurin, 0.0008% CaCl2, 0.4 mg/L FeSO4, 1 mg/L menadione, 0.2 mM histidine, and 1.9 mM hematin). Liquid cultures were collected by centrifugation at 21 000 × g, resuspended in PBS, and normalized to an optical density of 2.0. Frozen stocks were prepared by combining 0.5 mL of normalized bacterial cell suspension with 0.5 mL 30% glycerol in PBS for storage at −80°C. One stock of each isolate was plated to obtain accurate CFU counts for each strain. Nf1OPG mice treated with vancomycin from 6 to 9 weeks of age were gavaged with a cocktail (150 µL) of Bacteroides thetaiotaomicron dNLKV9, Bacteroides fragilis (ATCC 25285), Bacteroides ovatus (ATCC 8483), Bacteroides caccae (ATCC 43185), and Bacteroides intestinalis (DSM 17393).19
In Vivo Mouse Treatments
Four-week-old Nf1OPG mice were treated with 200 μg anti-transforming growth factor-beta (TGFβ) neutralizing antibodies (BE0057, Bio X cell) or anti-IgG2a (BE0083, Bio X cell) control antibodies by intraperitoneal injection.12
Blood Collection, ELISA, and Cytokine Array
Blood was collected by retro-orbital bleed using heparinized capillary tubes, followed by serum separation, and centrifugation in EDTA-coated vials. TGFβ ELISA (ab119557, Abcam) and cytokine array (ARY006- R&D systems) were performed as per the manufacturer’s instructions.
Optic Nerve Volume Analysis
Micro-dissected optic nerves were photographed and volumes were quantified using a Leica DFC 3000 G camera and National Institutes of Health (NIH) ImageJ software as previously described.12
Immunohistochemistry and Immunocytochemistry
Mice were euthanized and transcardially perfused with Ringer’s solution, followed by fixation with 4% paraformaldehyde (PFA). Serial 4-μm paraffin sections of the optic nerve were immunostained with appropriate primary and secondary antibodies (Supplemental Table 1) and developed using the Vectastain ABC kit (Vector Laboratories). Hematoxylin and eosin (H&E) staining was performed following the manufacturer’s instructions (StatLab). For immunocytochemistry, eye sections were immunostained with appropriate primary and secondary Alexa Fluor-conjugated antibodies (Supplemental Table 1). Images acquired using Image Studio Lite Version 5.2 and LAS AF Lite 3.2.0 software were analyzed using NIH ImageJ 1.53a software.
Microglia Isolation
Single-cell suspensions of microglia were isolated from mouse brains perfused with D-PBS using the multi-tissue dissociation kit12 and grown in minimal essential medium supplemented with 1 mM l-glutamine, 1 mM sodium pyruvate, 0.6% d-glucose, 1 ng/mL GM-CSF, 100 μg/mL penicillin/streptomycin, and 10% fetal bovine serum (FBS). After 14 days in vitro (DIV), microglia were separated from the astrocyte layer by gentle shaking (200 g, 5 h, and 37°C). Recombinant mouse TGFβ1 or 30 nM MEK inhibitor (PD032590) was added to 5 × 105 microglia for Ccl3 quantification by ELISA and ERK activity by western blot.
Quantitative Reverse Transcription Polymerase Chain Reaction
RNA isolated from 2 mouse optic nerves using the NucleoSpin RNA Plus kit was reverse transcribed with the Applied Biosystems High-Capacity cDNA Reverse Transcription Kit. Real-time quantitative PCR (qPCR) was performed using TaqMan primers (Supplemental Table 2) on a QuantStudio 3 system. ΔΔCT values were calculated using Gapdh as an internal control.
Western Blotting
Forty microgram of protein extracted using RIPA buffer, and quantified by Pierce BCA protein assay, was electrophoretically separated in precast SDS-polyacrylamide gels and transferred onto PVDF membranes, followed by blocking and incubation with primary antibodies (Supplemental Table 1) overnight. Proteins were detected with IRDye-conjugated secondary antibodies in LI-COR Odyssey Imaging system using Image Studio v5.2.
T Cell Isolation
Mouse spleen single-cell suspensions in PBS with 0.1% BSA and 0.6% Na-citrate underwent red blood cell lysis (eBioscience, 00433357)12 and filtering through a 30-µm cell strainer prior to CD8+ T cell isolation using a CD8a T cell isolation kit. 2.5 × 106 CD8+ T cells/mL maintained in RPMI-1640 medium supplemented with 10% FBS and 1% penicillin/streptomycin were activated with 1.25 µg/mL anti-mouse CD3 and 2 µg/mL anti-mouse CD28 antibodies for 2 days. To assay migration, 5 × 105 T cells were placed in the upper transwell chambers containing 500 µL serum-free RPMI-1640 media. 500 µL of media containing Ccl3 (20 ng/mL) was added to the lower chamber,7 and the number of T cells in the lower chambers counted after 24 h on a Leica DMi1inverted microscope at 20×.
16S rDNA qPCR and Illumina Sequencing and Analysis
Phenol:chloroform-extracted DNA from fecal pellets was used for both 16S rRNA SYBR green qPCR for 16S with 515-F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 805-R (5′-GACTACCAGGGTATCTAATCC-3′) primers, as well as for 16S sequencing, as previously described.20 Read quality control and the resolution of amplicon sequence variants were performed with the dada2 R package.21 Amplicon sequence variants not assigned to the kingdom Bacteria were filtered out. The remaining reads were assigned taxonomy using the Ribosomal Database Project (RDP trainset 16/release 11.5) 16S rRNA gene sequence database.22 Ecological analyses were performed using PhyloSeq and additional R packages,23 and differentially abundant taxa between sample groups were identified by performing pairwise comparisons using LEfSe.24 16S sequencing data were uploaded to the European Nucleotide Archive (Accession number PRJEB77572).
Human Tumor Single-Cell Data Processing, Filtering, Cohort Data Integration, and Dimension Reduction
Under an approved Human Studies Protocol, human sporadic PA samples were obtained as dissociated frozen tumor single cells from the St. Louis Children’s Hospital Pediatric Tumor Bank and processed for single-cell RNA sequencing (scRNAseq). Raw sequencing data were processed using the 10X Genomics Cell Ranger pipeline (version 6.1.1) to generate gene count matrices and alignment to the GRCh38 human genome reference prior to analysis using Partek Flow software (version 11.0). Analyses were performed as previously reported using Seurat3 integration, Louvain algorithm graph-based clustering, and t-SNE/UMAP generation employing Euclidean distance.25 Differential analyses were performed using gene-specific analysis (GSA), and the results were filtered for genes with P-values < .05, and fold changes ≤−2 or ≥2.
Mouse Single-Cell Data Processing, Filtering, Cohort Data Integration, and Dimension Reduction
Optic nerves from 10 12-week-old Nf1OPG mice treated with either PBS (vehicle) or vancomycin from 6 to 12 weeks of age were collected and dissociated using the neural dissociation kit followed by debris removal, stained with 7-AAD, and live cells sorted using MoFlo for scRNAseq. Raw sequencing data from vehicle and vancomycin groups were processed through the 10x Genomics Cell Ranger pipeline (version 7.2.0) to generate gene count matrices and subsequently aligned to the mouse reference genome (mm10-2020-A). The scRNAseq data was processed utilizing Partek Flow (version 11.0.24) as previously reported.25 Cell integration was performed using the Seurat3 integration method and integrated expression data was subjected to principal component analysis (PCA). Graph-based clustering using the Louvain algorithm was applied including the top 20 principal components, and UMAP plots were generated employing Euclidean distance on these components. Differential expression analysis was executed using the Hurdle model (MAST) to compare and identify differentially expressed genes between vancomycin and vehicle samples. Results were filtered to only include those genes with the False Discovery Rate (FDR) less than 0.05, and fold changes ≤−2 or ≥2.
Quantification and Statistical Analysis
Data analysis was performed using GraphPad Prism software (version 10). Differences between 2 groups were determined using a 2-tailed Student’s t-test, while multiple comparisons were analyzed using a one-way analysis of variance (ANOVA) with Dunnett’s multiple comparisons test. Statistical significance was defined as P ≤ .05. All experiments were independently repeated at least thrice with at least 3 biological replicates.
Results
Intestinal Microbiota Controls Nf1OPG Progression and Growth
To determine whether bacteria in the intestinal flora regulate Nf1-OPG growth, we performed 3 complementary sets of experiments. First, we raised Nf1OPG mice under GF conditions from birth. At 12 weeks of age, the optic nerves of GF Nf1OPG mice exhibited reduced optic nerve proliferation (%Ki67+ cells; Figure 1A) relative to Conv-R mice. Consistent with reduced tumor progression (rather than initiation).17,26 GF Nf1OPG mice had decreased Olig2+ and Blbp+ cellular content (Figure 1B, C) without any change in volume (Supplementary Figure 1A). It should be noted that Nf1+/− mice (without tumors) have enlarged optic nerve volumes, similar to those seen in Nf1OPG mice and some children with NF1.12 While the optic nerve volumes of GF Nf1OPG mice were indistinguishable from Conv-R Nf1OPG mice, the optic nerves lacked the architectural distortion typical of those observed in tumor-bearing Nf1OPG mice (Supplementary Figure 1B).
Intestinal microbiota control Nf1-OPG growth. Nf1OPG mice were raised in germ-free (GF) facility, while conventionally reared (Conv-R) mice were reared in a specific pathogen-free vivarium. Isolated optic nerves and eyes were analyzed at 12 weeks of age. Immunohistochemical staining and quantification of (A) tumor cell proliferation (%Ki67+ cells), (B) Blbp+ cell content (%Blbp+ cells; n = 5), (C) Olig2+ cell content (%Olig2+ cells; n = 5), and (D) retinal nerve fiber layer (RNFL; n = 6 female Nf1OPG mice) thickness (Smi-32+ fibers) from GF mice relative to Conv-R mice. (D) Dotted lines in photomicrographs outline the RNFL in female Nf1OPG mice, while the dotted line in the graph denotes the average RNFL thickness in WT mice (Supplementary Figure 1D). Male and female Nf1OPG mice received drinking water containing antibiotics (vanco, vancomycin; amp, ampicillin; neo, neomycin; and VNA, all 3 antibiotics) from 6 to 12 weeks of age. Controls received sterile distilled water. (E) Immunohistochemical analysis and quantification of tumor cell proliferation (%Ki67+ cells) in the optic nerves of Nf1OPG mice (Conv-R, n = 5; VNA, n = 9; amp, n = 6; vanco, n = 7; neo, n = 6) and representative images and quantification of (F) Blbp+ cell content (%Blbp+ cells), (G) Olig2+ cell content (%Olig2+ cells) in the optic nerves of Nf1OPG mice (n = 5), and (H) RNFL thickness (Conv-R, n = 5; VNA, n = 4; amp, n = 3; vanco, n = 4; neo, n = 3 female Nf1OPG mice). Dotted lines in the photomicrographs outline the RNFL in female Nf1OPG mice, while the dotted line in the graph denotes the average RNFL thickness in WT mice (Supplementary Figure 1D). All mice were analyzed at 12 weeks of age. Scale bars, A–C and E–G 40 µm; D and H 100 µm. Data are represented as the mean ± SEM. (A–D) Two-tailed student’s t-test. (E–H) One-way ANOVA with Dunnett’s posttest correction. P-values are indicated within each graph.
Importantly, since vision loss is one of the most common clinical sequelae affecting children with NF1-OPG, we also measured retinal nerve fiber layer (RNFL) thickness and retinal ganglion cell (RGC) numbers to obtain similar information as acquired by ocular coherence tomography (OCT) in nonverbal patients.27,28 Since the effect of NF1-OPG on vision loss is sexually dimorphic in mice,29,30 only females were used. While there was no change in RGC numbers (Supplementary Figure 1C), we observed an increase in RNFL thickness (SMI-32+ fibers; Figure 1D) to levels comparable to WT mice (Supplementary Figure 1D). The apparent dissociation between RGC number and RNFL likely reflects the temporal and spatial course of OPG-induced injury, in which axonal damage with the optic nerve and RNFL thinning precede RGC apoptosis and death.28
Second, we treated Conv-R Nf1OPG mice with a cocktail of antibiotics composed of vancomycin, ampicillin, and neomycin (VNA), or each antibiotic separately, from 6 to 12 weeks of age. The fecal microbiota was serially sampled, and animals were euthanized at 12 weeks of age. Like GF Nf1OPG mice, all antibiotic treatments, except ampicillin, reduced Nf1OPG mouse optic nerve proliferation (%Ki67+ cells; Figure 1E). It should be noted that the Ki67+ (proliferating) cells in these nerves at 12 weeks of age are tumor cells, rather than other cell types.1 While there was restoration of normal optic nerve tissue architecture (Supplementary Figure 1E), there was no reduction in optic nerve volume (Supplementary Figure 1F) or RGC number (Supplementary Figure 1G). In contrast, only VNA and vancomycin decreased Olig2+ and Blbp+ cellular content (Figure 1F, G) and restored RNFL thickness (SMI-32+ fibers; Figure 1H) to WT levels (Supplementary Figure 1D). Neomycin, vancomycin, and ampicillin, but not neomycin, caused significant depletions in overall bacterial load as measured by 16S rRNA gene qPCR (Supplementary Figure 5A–D), although the most dramatic reductions were observed with VNA or vancomycin. Comparable results were also observed using a second strain of Nf1OPG mice harboring a patient-derived germline NF1 gene mutation (Y2083X; Supplementary Figure 2), highlighting the generalizability of these findings.
Vancomycin Treatment Suppresses Nf1OPG Tumor Growth
To determine the impact of vancomycin treatment on optic glioma growth after tumors have already formed, 2 different experiments were performed. First, 12-week-old Nf1OPG mice were treated with vancomycin for 6 weeks and analyzed at 18 weeks of age. Treated Nf1OPG mice exhibited reduced optic nerve proliferation (%Ki67+ cells; Figure 2A) and increased RNFL thickness (SMI-32+ fibers; Figure 2B) to WT levels (Supplementary Figure 1D), as well as reduced Blbp+ and Olig2+ cellular content (Figure 2C, D). While there was restoration of tissue architecture (Supplementary Figure 3A), there was no change in optic nerve volume or RGC numbers (Supplementary Figure 3B, C). Second, to determine whether this tumor suppression was durable, Conv-R Nf1OPG mice were treated with vancomycin from 6 to 12 weeks of age and analyzed at 24 weeks of age. Vancomycin-treated Nf1OPG exhibited reduced optic nerve proliferation (%Ki67+ cells; Figure 2E) and increased RNFL thickness (SMI-32+ fibers; Figure 2F) to WT levels (Supplementary Figure 1D), as well as reduced Blbp+ and Olig2+ cellular content (Figure 2G, H) demonstrating a durable effect. While there was restoration of tissue architecture (Supplementary Figure 3D), there was no change in optic nerve volume or RGC number (Supplementary Figure 3E, F). These results establish that the intestinal microbiota is a critical modulator of optic glioma progression and growth, rather than initiation, and demonstrate that tumor suppression is durable.
Male and female Nf1OPG mice receiving drinking water containing vancomycin from 12 to 18 weeks of age were analyzed at 18 weeks of age. Controls received sterile distilled water. (A) Immunohistochemical analysis and quantification of tumor cell proliferation (%Ki67+ cells) in the optic nerves of Nf1OPG mice (n = 5) and (B) retinal nerve fiber layer (RNFL) thickness (n = 5 female Nf1OPG mice). Dotted lines in the photomicrographs outline the RNFL in female Nf1OPG mice, while the dotted line in the graph denotes the average RNFL thickness in WT mice (Supplementary Figure 1D). Quantification and representative images of (C) Blbp+ cell content (%Blbp+ cells; n = 5) and (D) Olig2+ cell content (%Olig2+ cells; n = 5). Male and female Nf1OPG mice receiving drinking water containing vancomycin from 6 to 12 weeks of age were analyzed at 24 weeks of age. Control Nf1OPG mice received sterile distilled water. (E) Immunohistochemical analysis and quantification of tumor cell proliferation (%Ki67+ cells) in the optic nerves of Nf1OPG mice (n = 5) and (F) RNFL thickness (n = 4 female Nf1OPG mice). Dotted lines in the photomicrographs outline the RNFL in female Nf1-OPG mice, while the dotted line in the graph denotes the average RNFL thickness in WT mice (Supplementary Figure 1D). Quantification and representative images of (G) Blbp+ cell content (%Blbp+ cells; n = 5) and (H) Olig2+ cell content (%Olig2+ cells; n = 5). Scale bars, A, C–E, G, and H 40 µm; B and F 100 µm; data are represented as the mean ± SEM. (A–H) Two-tailed student’s t-test. P-values are indicated within each graph.
Fecal Microbiota Transplant Reverses Nf1OPG Tumor Proliferation
To determine whether GF or antibiotic control of Nf1OPG growth is mediated by changes in the gut microbiota, we reared Nf1OPG mice under GF conditions or treated Conv-R Nf1OPG mice with vancomycin from 6 to 9 weeks of age prior to a single oral gavage of fecal material from 12-week-old Conv-R Nf1OPG mice at 9 weeks of age. Consistent with intestinal microbial control of Nf1OPG growth, Conv-R Nf1OPG mouse fecal microbiota transplant (FMT) increased optic nerve proliferation (%Ki67+ cells) (Figure 3A), as well as increased Blbp+ cell content (Figure 3B) with no change in Olig2+ cell content (Figure 3C). Similarly, 6–9 weeks of vancomycin treatment reduced optic nerve proliferation and Blbp+ cell content, which was partially reversed when Nf1OPG mice were colonized with Conv-R fecal microbiota at 9 weeks of age and analyzed at 12 weeks of age (Figure 3D–F). In both treatments, while there was an increase in optic nerve cellularity (Supplementary Figure 4A, B), there was no reduction in optic nerve volume (Supplementary Figure 4C, D) or change in RGC number (Supplementary Figure 4E, F). There was a trend toward reversal of RNFL thickness toward Conv-R mice (Supplementary Figure 4G, H). Given this partial effect, we next sought to determine whether 2 sequential doses of FMT would be more effective. Nf1OPG mice were treated with vancomycin from 6 to 9 weeks of age followed by 2 oral gavages of fecal material from 12-week-old Conv-R Nf1OPG mice at 9 weeks and 10 weeks of age. Two sequential FMT treatments increased optic nerve proliferation (%Ki67+ cells) (Figure 3G), as well as increased Blbp+ cell content (Figure 3H) and Olig2+ cell content (Figure 3I) in Conv-R Nf1OPG mice with vancomycin from 6 to 9 weeks. While there was increased optic nerve cellularity (Supplementary Figure 4I), there was no reduction in optic nerve volume (Supplementary Figure 4J) or change in RGC number (Supplementary Figure 4K). Importantly, there was now a reversal of RNFL thickness toward levels seen in Conv-R mice (Supplementary Figure 4L).
Fecal microbiota transplant from Conv-R mice regulates Nf1-OPG growth. (A) Quantification and representative images of tumor cell proliferation (%Ki67+ cells; n = 6), (B) Blbp+ cell content (%Blbp+ cells; n = 4), and (C) Olig2+ cell content (%Olig2+ cells; n = 4) in the optic nerves of Nf1OPG mice raised in a GF facility and given a single oral gavage of Conv-R mouse fecal material at 9 weeks of age. All mice were analyzed at 12 weeks of age. (D) Immunohistochemistry, quantification, and representative images of tumor cell proliferation (%Ki67 + cells; n = 5), (E) Blbp+ cell content (%Blbp+ cells; n = 4), and (F) Olig2+ cell content (%Olig2+ cells; n = 4) in Nf1OPG mice given drinking water containing vancomycin from 6 to 9 weeks of age followed by a single oral gavage of Conv-R mouse fecal material at 9 weeks of age. Immunohistochemistry, quantification, and representative images of (G) tumor cell proliferation (%Ki67+ cells; n = 5), (H) Blbp+ cell content (%Blbp+ cells; n = 5), and (I) Olig2+ cell content (%Olig2+ cells; n = 5) in the optic nerves of Nf1OPG mice given drinking water containing vancomycin from 6 to 9 weeks of age followed by 2 oral gavages of Conv-R mouse fecal material at 9 weeks and 10 weeks of age. All mice were analyzed at 12 weeks of age. Scale bars, A–I 40 µm; data are represented as the mean ± SEM. (A–I) One-way ANOVA with Dunnett’s posttest correction. P-values are indicated within each graph.
Bacteroides Reverses Antibiotic-Mediated Reduction of Nf1-OPG Tumor Proliferation
To delineate the candidate microbiota taxa responsible for Nf1-OPG growth, we used 16S rRNA gene sequencing to identify taxa preferentially depleted by vancomycin relative to neomycin treatment (Supplementary Figure 5A–D). We performed Linear discriminant analysis Effect Size (LEfSe) analysis to detect differentially abundant taxa (Figure 4A). One of the most consistently depleted groups after vancomycin, relative to neomycin, treatment belonged to the Bacteroidales order (Figure 4A, B). To determine whether taxa in Bacteroidales were involved in intestinal microbiota regulation of tumor growth, Nf1OPG mice were treated with vancomycin from 6 to 9 weeks of age and then gavaged with a single oral dose of a cocktail of Bacteroidales species (B. thetaiotaomicron, B. fragilis, B. ovatus, B. caccae, and B. intestinalis) for analysis at 12 weeks of age. Bacteroides colonization increased mouse optic nerve proliferation (%Ki67+ cells; Figure 4C) to near Conv-R levels, without any change in optic nerve volume (Supplementary Figure 5E). There was a nonstatistically significant trend toward Conv-R levels of RNFL thickness (SMI-32+ fibers; Figure 4D), but no change in RGC number (RBPMS+; Figure 4E). In addition, while Bacteroides colonization did not change Olig2+ and Blbp+ cell content (Supplementary Figure 5F, G), it resulted in a moderate amount of architectural distortion similar to Conv-R Nf1OPG mice (Supplementary Figure 5H).
Mouse intestinal Bacteroides restores the antibiotic-mediated reduction in Nf1-OPG tumor proliferation. (A) LEfSe analysis of 16S rRNA gene sequencing of Nf1OPG mice treated with neomycin or vancomycin identified statistically significant bacterial taxa biomarkers. Green indicates taxa relatively enriched in neomycin-treated mice, while blue indicates taxa enriched in vancomycin-treated mice. (B) Relative abundance of Bacteroidales in fecal samples collected from neomycin- or vancomycin-treated mice. Male and female Nf1OPG mice were treated with drinking water containing vancomycin (vanco) from 6 to 9 weeks of age. At 9 weeks of age, a group of mice received a single oral gavage of Bacteroides cocktail, while controls received sterile phosphate-buffered saline (PBS). Isolated optic nerves and eyes were analyzed at 12 weeks of age. (C) Immunohistochemical analysis and quantification of tumor cell proliferation (%Ki67+ cells) of the optic nerves from Nf1OPG mice (Conv-R, n = 5; VNA, n = 9; amp, n = 6; vanco, n = 7; and neo, n = 6) and (D) RNFL thickness in female Nf1OPG mice (Conv-R, n = 5; vanco, n = 4; and vanco + Bacteroides, n = 4). Dotted lines in the photomicrographs outline the RNFL in female Nf1OPG mice, while the dotted line in the graph denotes the average RNFL thickness in WT mice (Supplementary Figure 1D). (E) Retinal ganglion cell numbers in female Nf1OPG mice (Conv-R, n = 4; vanco, n = 4; and anco + Bacteroides, n = 4). Scale bars, C 40 µm; D 100 µm. (B) Wilcox test. (C–E) One-way ANOVA with Dunnett’s posttest correction. P-values are indicated within each graph.
GF and Antibiotic-Treated Mice Have Reduced Intra-tumoral CD8+ T Cell Content
Previous studies using Nf1OPG mice revealed that the progression and maintenance of murine optic gliomas depend on the presence of CD8+ T cells and tumor-associated monocytes (TAM).6 As such, CD8+ T cells are recruited by glioma tumor cell Ccl2 production, resulting in T cell Ccl4 secretion, which in turn stimulates microglia to produce Ccl5, a growth factor necessary for the formation and growth of Nf1-optic gliomas7 (Figure 5A). To determine whether the intestinal microbiota might regulate T cell function, Nf1OPG mice were raised under GF conditions or treated with vancomycin. Following these exposures, there was reduced optic nerve Ccl4 and Ccl5 expression at 12 weeks of age (Figure 5B, C). Importantly, this reduction was partly reversed following oral gavage with Bacteroides (Supplementary Figure 6A, B). To determine whether reduced T cell Ccl4 and microglial Ccl5 production in Nf1OPG mice resulted from reduced CD8+ T cell or TAM optic glioma function, we performed scRNAseq and immunocytochemistry. We found reduced CD8+ T cell content in the optic nerves of vancomycin-treated, as well as GF, Nf1OPG mice relative to controls (Figure 5D–F), which was reversed following Bacteroides treatment (Supplementary Figure 6C), without any change in optic nerve TAM content (Supplementary Figure 6D). Additionally, there was no change in Ccl2 production, a chemokine made by optic glioma tumor cells previously demonstrated to be important for CD8+ T cell intra-tumoral infiltration (Figure 5G; Supplementary Figure 6E).
Germ-free (GF) and antibiotic-treated mice exhibit reduced intra-tumoral CD8+ T cell content. (A) Schematic representation of the mouse Nf1-OPG immune cancer cell axis. TAM, tumor-associated monocytes (microglia). (B) Ccl4 and (C) Ccl5 RNA expression in the optic nerves of Nf1OPG mice treated with vancomycin or raised in a GF facility. (D) t-SNE visualization plot of optic nerve single-cell RNA sequencing results from Nf1OPG mice treated with vancomycin (vanco) from 6 to 12 weeks of age. Control mice were given sterile distilled water (n = 10 in each group). (E) Violin plot showing changes in the CD8a RNA expression. (F) Immunofluorescence and quantification of CD8+ T cells in the optic nerves of Nf1OPG mice treated with vanco or raised in a GF facility relative to conventionally reared mice (Conv-R, n = 6; vanco, n = 5; and GF, n = 6). (G) Ccl2 RNA expression in the optic nerves of Nf1OPG mice treated with vancomycin or raised in a GF facility relative to conventionally reared (Conv-R) mice (n = 4). (H) Top transcripts reduced in microglia from Nf1OPG mice treated with vancomycin relative to conventionally reared mice. (I) Bubble plot showing Ccl3 gene expression in different cell types in the optic nerves of 3-month-old Nf1OPG mice by single-cell RNA sequencing. Microglia and monocytes constitute tumor-associated monocytes (TAM). (J) Ccl3 RNA expression in the optic nerves of Nf1OPG mice treated with vancomycin or raised in a GF facility (n = 4). Scale bars, F 40 µm. Data are represented as the mean ± SEM. (B, C, F, G, and J) One-way ANOVA with Dunnett’s posttest correction. P-values are indicated within each graph. Abbreviation: R.E., relative expression.
To identify the mechanism by which the intestinal microbiota regulates CD8+ T cell optic glioma infiltration, we leveraged scRNAseq data from Conv-R and vancomycin-treated Nf1OPG mouse optic nerves, as well as from human pediatric LGG tumors (PAs). Following filtering of differentially expressed transcripts (P value ≤ .01, FDR ≤ 0.05, and log fold change ≥ 5), we examined the top 20 differentially regulated pathways in TAM (Figure 5H). We chose to focus on Ccl3 for 3 reasons. First, previous reports demonstrated a critical role for Ccl3 in the regulation of CD8+ T cell migration.31 Second, Ccl3 was the only secreted factor found among the top 20 transcripts. Third, Ccl3 expression was highest in TAM from Nf1OPG mouse optic nerves (Figure 5I) relative to other cell types in the tumor and monocytes (microglia and monocytes) from WT and Nf1+/− mice (Supplementary Figure 6F). Consistent with the hypothesis that TAM Ccl3 might regulate CD8+ T cell migration in response to intestinal microbiota changes, Ccl3 expression was reduced in the optic nerves of GF and vancomycin-treated Nf1OPG mice relative to Conv-R Nf1OPG mice (Figure 5J), which was partially reversed following Bacteroides oral gavage (Supplementary Figure 6G). In addition, CCL3 expression was also increased in a single NF1-PA relative to sporadic (non-NF1; n = 4) PA (Supplementary Figure 6H), where it was enriched in the TAM from this NF1-PA by scRNAseq (Supplementary Figure 6I). Taken together, these findings demonstrate that the intestinal microbiota regulates the ability of CD8+ T cells to migrate into the optic nerve by reducing the expression of Ccl3, a chemokine produced by TAM.
TGFβ Converges on the Nf1-OPG Growth Axis by Increasing Microglia Ccl3 Production
Prior studies implicate inflammatory cytokines in Bacteroides-induced intestinal effects.19 While no differences in serum inflammatory cytokine expression in control versus vancomycin-treated Nf1OPG mice (Supplementary Figure 7A), our scRNAseq results (not shown) and prior literature implicate TGFβ.19 To explore this potential etiologic factor, we first demonstrated that TAMs from both murine Nf1OPG optic gliomas (Supplementary Figure 7B, C) and the one human NF1-PA (Supplementary Figure 7D) express the TGFβ1 receptor (TGFbr1). Second, TGFβ levels were reduced in the serum of vancomycin-treated and GF-raised Nf1OPG mice relative to serum from Conv-R Nf1OPG mice (Figure 6A). Third, TGFβ levels were also reduced in brain of vancomycin-treated mice relative to Conv-R mice (Figure 6B). Fourth, using the same TGFβ concentration found in the serum of Conv-R Nf1OPG mice, TGFβ induced microglia Ccl3 production in vitro (Figure 6C). TGFβ-induced microglia Ccl3 production is dependent on ERK activation, such that pharmacologic ERK inhibition reduced TGFβ-induced Ccl3 production (Figure 6D, E).
TGFβ sustains Nf1-OPG growth by increasing Ccl3 production in microglia. (A) TGFβ ELISA using serum isolated from Nf1OPG mice raised in a conventional facility (Conv-R), treated with vancomycin (vanco), or raised in a germ-free (GF) facility (n = 5). (B) Representative western blot and quantification showing TGFβ protein expression in brains isolated from Nf1OPG mice raised in a conventional facility (Conv-R) and treated with vancomycin (vanco) (n = 3). (C) Nf1+/− microglia were treated with either PBS or TGFβ for 24 h. The conditioned media (CM) was used for Ccl3 ELISA (n = 4). (D) Expression of ERK and phospho-ERK (p-ERK) in Nf1+/− microglia treated with either PBS or TGFβ alone or with 30 nM MEK inhibitor for 24 h. (E) Ccl3 was detected in the CM by ELISA (n = 4). Nf1OPG mice were treated with anti-TGFβ (i.p 200 µg) or anti-IgG isotype control antibodies 3 times a week from 6 to 12 weeks of age. Optic nerve and eyes were analyzed at 12 weeks of age. Immunohistochemical quantification of (F) tumor cell proliferation (%Ki67+ cells; IgG n = 4; αTGFβ n = 5), (G) CD8+ T cells in the optic nerve (n = 5), and (H) retinal nerve fiber layer (RNFL) thickness (Smi-32 + fibers; IgG, n = 3; and αTGFβ, n = 4 female Nf1OPG mice). Dotted lines in the photomicrographs outline the RNFL in female Nf1OPG mice, while the dotted line in the graph denotes the average RNFL thickness in WT mice (Supplementary Figure 1D). (I) Ccl4, (J) Ccl5, and (K) Ccl3 RNA expression in the optic nerves of Nf1OPG mice treated with IgG or αTGFβ antibodies (n = 4). (L) The proposed mechanistic model of intestinal microbiota-mediated control of Nf1-OPG growth. Bacteroides residing in the gut induces intestinal epithelial cell TGFβ release, which enters the systemic circulation to induce Ccl3 expression in microglia and the recruitment of CD8+ T cells to the optic nerve. Scale bars, F and G 40 µm; H 100 µm. Data are represented as the mean ± SEM. (A and E) One-way ANOVA with Dunnett’s posttest correction. (B, C, and F–K) Two-tailed student’s t-test. P-values are indicated within each graph. Abbreviations: F.C., fold change; R.E., relative expression; TAM, tumor-associated monocytes (microglia); TGFBR1, TGF beta receptor 1.
To demonstrate that TGFβ is sufficient to block optic glioma growth, Nf1OPG mice were systemically treated with anti-TGFβ antibodies from 6 to 12 weeks of age. Similar to GF or antibiotic-treated Nf1OPG mice, anti-TGFβ antibody treatment of Nf1OPG mice reduced optic nerve proliferation (%Ki67+ cells; Figure 6F) and CD8+ T cell content (Figure 6G) at 12 weeks of age, as well as restored RNFL thickness (SMI-32+ fibers; Figure 6H) to levels comparable to WT mice (Supplementary Figure 1D). Moreover, TGFβ neutralization reduced Ccl4, Ccl5, and Ccl3 RNA expression (Figure 6I–K) and partly normalized optic nerve tissue architecture (Supplementary Figure 7E), without any change in optic nerve volume (Supplementary Figure 7F), TAM content (Supplementary Figure 7G), or RGC number (Supplementary Figure 7H). Collectively, the findings presented in this report suggest a model in which NF1-LGG risk is modified by intestinal microbiota alterations (Figure 6L), where TGFβ drives CD8+ T cell attraction and immune axis establishment of a microenvironment supportive of tumor growth.
Discussion
The relationship between the intestinal microbiota and human neurological diseases has garnered significant scientific interest, with the demonstration that the depletion of intestinal bacteria has major effects on numerous neurological diseases in mice. Leveraging a well-characterized murine model of LGG affecting children with NF1, we demonstrate that Bacteroides is responsible for regulating TAM chemoattraction of CD8+ T cells into the optic nerve and immune stromal support of glioma progression. This mechanistic dissection raises several key points relevant to the intestinal microbiota–brain axis control of neurological disorders.
While prior studies in experimental models of central nervous system diseases implicated immune system dysfunction, the exact mechanism(s) underlying these immunologic effects was not fully elucidated. In this regard, antibiotic treatment was associated with increased immature microglia32 in wild-type mice, intestinal microbiota depletion in a murine model of Alzheimer’s disease reduced microglia disease phenotypes,15 and fecal matter transplant in GF mice after traumatic brain injury increased microglial activation and decreased T cell infiltration.16 Using the only available authenticated genetically engineered preclinical model of NF1-LGG, we found that the intestinal microbiota is necessary for immune support of tumor progression and growth by regulating TAM production of the Ccl3 chemokine. In this setting, Ccl3 functions as a chemoattractant for T lymphocyte infiltration into the optic glioma. As such, prior studies in human NF1-LGG (PA) demonstrate immune cell infiltration,3 particularly with CD8+ T cells.5 This observed T cell attraction effect in NF1-LGG is supported by studies in other cancers, in which Ccl3 controls CD8+ T cell migration in melanoma metastasis, confers a poor prognosis for esophageal squamous cell carcinoma, and increases colorectal cancer growth.33 The demonstration that Ccl3 functions as another chemokine responsible for mediating T cell tumor infiltration, in addition to Ccl27 produced by Nf1-OPG tumor cells and TAM-elaborated Ccl12 and Cxcl13,25 highlights the diversity of paracrine mechanisms governing immune cell brain trafficking.
Second, we discovered that Bacteroidales content was reduced in the fecal microbiota of vancomycin-treated and GF mice. This observation adds to previous reports implicating Bacteroides, a significant component of the intestinal microbiota, in neurological disorders. While the presence of glioma is associated with increased gut Bacteroides content in both mice34 and humans,35 there are no prior reports describing the mechanistic relationship between the intestinal microbiome and glioma growth. Elevated levels of Bacteroides strains have been associated with Alzheimer’s disease, where B. fragilis metabolites activate microglia and induce AD pathology.36 Studies are currently underway to identify the specific Bacteroides species responsible and employ viral phage approaches to therapeutic Bacteroides ablation.37,38
Third, the finding that immune circuits, formed by CD8+ T cells and TAM, which are essential for Nf1-OPG progression and continued growth, can function as convergence points for brain tumor risk factors offers a new perspective for understanding the gut-brain axis at the circuit level. During brain development and maintenance, the intestinal microbiota regulates the differentiation and function of microglia32,39 and astrocytes,40 which are critical for neural specification, neurotransmission, CNS immune activation, and maintaining blood–brain barrier integrity.41 Additionally, intestinal microbiome alterations can diminish cancer therapy effectiveness, as intestinal bacteria are critical for the efficacy of chemotherapy and immunotherapy in both mice and humans.42,43 Herein, we demonstrate that intestinal Bacteroides regulates Nf1-OPG biology through the elaboration of TGFβ. Prior studies have demonstrated that Bacteroides can induce intestinal inflammation through the production of TGFβ,19 which can regulate regulatory T cells (Tregs), type 17 helper (Th17) cells, innate lymphoid cells (ILCs), and B cells.44 While the exact mechanism by which Bacteroides triggers chronic intestinal inflammation and TGFβ secretion has not been fully elucidated, these bacteria contain esterase gene clusters that regulate the release of ferulic acid, which results in increased TGFβ expression.45 In addition, Bacteroides enterotoxin can induce colonic epithelial cells to release TGFβ,19 which is associated with the increased neutrophil infiltration.46 While TGFβ can signal through ERK or NFκB47 in other contexts, we demonstrate increase that TGFβ-mediated TAM Ccl3 production is ERK-dependent in the setting of Nf1-OPG. The demonstration that TGFβ inhibition was sufficient to reduce tumor growth in Nf1OPG mice by dampening T cell-TAM paracrine (Ccl4, Ccl5) stromal support suggests other opportunities for clinical grade TGFβ inhibitors in brain cancer therapy.48
While we conclusively demonstrate an obligate role for intestinal Bacteroides in murine low-grade optic glioma pathogenesis, further studies will be required to extend these observations to other types of brain tumors in mice and humans. For example, it is tempting to extend these findings to sporadic LGG (harboring the KIAA1549:BRAF genomic rearrangement) based on our prior work demonstrating that KIAA1549:BRAF-expressing neural progenitors attract Ccr2-expressing monocytes through the production of Ccl2 to drive LGG formation.49 However, the lack of robust preclinical models of sporadic LGG limits a more exhaustive exploration of the relationship between TAMs and T cells, which also populate sporadic LGG.50 Similarly, the translation of these studies to children under real-world conditions will necessitate further investigation.
Collectively, we establish a mechanistic relationship between the intestinal microbiota and NF1-LGG growth, raising the intriguing possibility that intestinal bacteria content may represent a predictive risk factor for NF1-LGG development and a future therapeutic target for halting NF1-LGG progression. Moreover, the dissection of the paracrine circuit responsible for gut-brain communication in the setting of brain cancer helps elucidate new actionable opportunities for treating brain tumors.
Funding
This work was funded by grants from the National Institutes of Health [R35NS097211 to D.H.G.], Gilbert Family Foundation [Vision Restoration Initiative to D.H.G.], and Schnuck Markets [to D.H.G.].
Conflict of interest statement. None declared.
Acknowledgments
We thank the Genome Technology Access Center in the Department of Genetics at the Washington University School of Medicine for help with genomic analysis. The Washington University Genome Engineering and iPSC Core Center is subsidized by funding from an NCI Cancer Center Support Grant (P30-CA091842) and by ICTS/CTSA Grant #UL1TR002345 from the National Center for Research Resources (NCRR), a component of the National Institutes of Health (NIH), and NIH Roadmap for Medical Research. The Washington University Gnotobiotic Core generated and provided germ-free mice. The Washington University Ophthalmology Core facility is supported by funding from the National Eye Institute (P30EY002687). Figure illustrations (Figures 5A and 6L) were created using BioRender.com.
Authorship statement
J.C. and D.H.G. designed and analyzed the experiments. J.C., X.Q., T.E., M.O., S.L.B, S.A., D.E.C., L.A.S., and M.T.B, conducted and/or interpreted the experiments. R.M. and X.L. performed the scRNAseq analyses. The manuscript was assembled by J.C. and D.H.G. D.H.G. was responsible for the final production of the manuscript.
Data availability
The single-cell RNA sequencing datasets were deposited in GEO (GSE270264; GSE244433). The 16S rRNA gene sequencing data has been deposited to the European Nucleotide Archive (accession PRJEB77572).
