Purpose:

The immunosuppressive tumor microenvironment present in the majority of diffuse glioma limits therapeutic response to immunotherapy. As the determinants of the glioma-associated immune response are relatively poorly understood, the study of glioma with more robust tumor-associated immune responses may be particularly useful to identify novel immunomodulatory factors that can promote T-cell effector function in glioma.

Experimental Design:

We used multiplex immune-profiling, proteomic profiling, and gene expression analysis to define the tumor-associated immune response in two molecular subtypes of glioma and identify factors that may modulate this response. We then used patient-derived glioma cultures and an immunocompetent murine model for malignant glioma to analyze the ability of tumor-intrinsic factors to promote a CD8+ T-cell response.

Results:

As compared with isocitrate dehydrogenase (IDH)-mutant astrocytoma, MAPK-activated pleomorphic xanthoastrocytoma (PXA) harbored increased numbers of activated cytotoxic CD8+ T cells and Iba1+ microglia/macrophages, increased MHC class I expression, enrichment of genes associated with antigen presentation and processing, and increased tumor cell secretion of the chemokine CXCL14. CXCL14 promoted activated CD8+ T-cell chemotaxis in vitro, recruited tumor-infiltrating CD8+ T cells in vivo, and prolonged overall survival in a cytotoxic T-cell–dependent manner. The immunomodulatory molecule B7-H3 was also highly expressed in PXA.

Conclusions:

We identify the MAPK-activated lower grade astrocytoma PXA as having an immune-rich tumor microenvironment and suggest this tumor may be particularly vulnerable to immunotherapeutic modulation. We also identify CXCL14 as an important determinant of the glioma-associated immune microenvironment, sufficient to promote an antitumor CD8+ T-cell response.

This article is featured in Highlights of This Issue, p. 2717

Translational Relevance

The majority of diffuse glioma exhibit a lymphocyte-depleted phenotype. Improved immunotherapeutic strategies require elucidation of the determinants of an effective antitumor cytotoxic T-cell response in brain. By studying an MAPK-activated subset of astrocytic glioma, pleomorphic xanthoastrocytoma (PXA), we identify determinants of a robust cytotoxic T-cell response. Tumor cell expression of chemokine CXCL14 was sufficient to promote cytotoxic T-cell chemotaxis, and in an immunocompetent murine model for malignant glioma increase the glioma-associated CD8+ T-cell response and promote overall survival. Consistent with CXCL14 driving an antitumor cytotoxic T-cell response, the prolonged survival conferred by CXCL14 was dependent on CD8+ T cells. These studies elucidate a novel mechanism of immune regulation in the brain and suggest CXCL14 modulation is a potential therapeutic strategy to suppress tumor growth. In addition, they identify a subset of glioma with an immune-rich tumor microenvironment that may be particularly vulnerable to immunotherapeutic modulation.

Diffuse glioma are characterized as lymphocyte-depleted and studies suggest they share an immunosuppressive tumor microenvironment characterized by T-cell dysfunction and abundant immunosuppressive myeloid cells (1–3). Yet, not all glioma share this immunophenotype and even within diffuse glioma, the immune microenvironment can vary (4–6). A mechanistic understanding of the determinants of the glioma immune microenvironment across distinct glioma types is critical to develop improved immunotherapeutic strategies for brain tumor patients and identify potential biomarkers for optimal therapeutic stratification.

The glioma immune microenvironment is shaped at multiple levels by tumor-cell–extrinsic and tumor-cell–intrinsic factors (7). Compared with other peripheral organs, the brain is often considered immunologically “distinct” (8) and certain aspects of the immune microenvironment, including myeloid cell populations, may contribute to an immunosuppressive microenvironment (2, 7, 9, 10). Supporting the important role for tumor-cell–intrinsic factors, tumors driven by different epigenetic or genetic alterations can exhibit quantitative and qualitative differences in their tumor-associated immune response (9, 11). In pediatric high-grade glioma, mutations in BRAF and NF1 have been associated with increased CD8+ tumor-infiltrating lymphocytes (TIL; ref. 6). Even within isocitrate dehydrogenase (IDH)—wild-type (WT) glioblastoma (GBM), there are differences in the brain tumor immune microenvironment across tumors (4, 12). In IDH-mutant astrocytoma several mechanisms may contribute to tumor immune-evasion, including reduced tumor cell expression of IFNγ-inducible chemokines, including CXC-chemokine ligand 10 (CXCL10; ref. 1).

To date there has been limited investigation of the determinants of the immune microenvironment in diverse glioma subtypes. We hypothesize that mechanisms supporting a glioma-associated CD8+ T-cell response may be relevant across brain tumor subtypes. Thus, in this study, we identify an MAPK-activated lower grade astrocytoma, pleomorphic xanthoastrocytoma (PXA), as a glioma subtype with a more robust CD8+ T-cell response and we use multiplex immune profiling, spatial profiling, and gene expression analysis to elucidate the tumor immune microenvironment. Then, using human tumor tissues, patient-derived cell lines, and an immunocompetent murine model for glioma, we investigate the tumor-intrinsic determinants of this robust glioma-associated CD8+ T-cell response and demonstrate the importance of tumor secreted factors.

Patient samples and inclusion criteria

A total of 89 World Health Organization (WHO) 2016 grade II to IV astrocytoma cases, of which 27 were PXA (mean age 28.7 years), 32 were IDH-mutant astrocytoma (mean age 39.4 years), and 30 were IDH-WT GBM (mean age 55.1 years), were identified from records in the Brain Tumor Center Biorepository and the Division of Neuropathology, Department of Pathology, at University of California, San Francisco (UCSF; San Francisco, CA). PXA is genetically characterized by the combination of CDKN2A biallelic inactivation and oncogenic RAF kinase signaling, most commonly the activating BRAF p.V600E (c.1799T > A) mutation (13, 14). Less common MAPK alterations identified, include BRAF insertion/deletion mutations (15), fusions involving BRAF or RAF1 (16), and mutation of NF1 (13, 17). Molecular classification was performed using the UCSF500 targeted next-generation sequencing panel and/or immunostaining for IDH1 R132H, mutant protein, p16, BRAF V600E mutant protein, ATRX, and p53 (18). The samples included in each analysis are specified in the text and figure legend (for additional details please see Supplemental Methods). Nonneoplastic adult brain samples were used as a control. The assays performed on each case are summarized in Supplementary Table S1. The ethics approval number for the use of deidentified human biospecimens is 10–01318. These studies were in accordance with the ethical standards of the institutional research committee and with the 1964 Helsinki declaration and its later amendments.

Multiplex immunofluorescence and immunostaining

Protein expression, including assessment of tumor-associated immune infiltration, was evaluated using multiplex immunofluorescence (mIFA; for most assays) or IHC (for select mouse studies) on formalin-fixed, paraffin-embedded (FFPE) slides using a Discovery XT autostainer (Ventana Medical Systems) with appropriate controls Cases analyzed by mIFA are listed in Supplementary Table S1, methods described in Supplemental Methods, and all antibodies detailed in Supplementary Table S2.

Image acquisition and analysis

5 × 5 tiled images of immunostained slides were acquired at 200× magnification using a Zeiss Spinning Disc confocal & TIRF (Zeiss). For digital quantification FIJI software was used (Supplemental Methods).

Digital spatial profiling of protein expression

46 immunologically relevant proteins (plus three normalization controls and three negative isotype controls; Supplementary Table S3) were quantified using the NanoString Inc. GeoMX platform (ref. 19; Supplemental Methods). Slides were also stained with fluorescently labeled antibodies directed against CD3 (Cyan,), Iba1 (red), CD34 (yellow), and DAP1 nuclear staining (blue) which served as visualization markers. Regions of interest (ROI) was selected as geometric shapes in T-cell–abundant regions.

Nanostring PanCancer immune profiling

Information on cases selected for Nanostring profiling and more detailed methods are detailed in Supplementary Table S1 and Supplemental Methods. The multiplex gene expression analysis using nCounter PanCancer immune profiling panel (n = 730 genes involved in the immune response and 40 housekeeping genes; NanoString Technologies) was performed at the computer-assisted tomography (CAT) facility, UCSF. Raw counts expression was analyzed using the nSolver Analysis Software (NanoString Technologies RRID:SCR_02171). The R package “limma” (version 3.29.0) was used for differential gene expression analysis. Multiple comparison testing using the Benjamini–Hochberg method was used to control the FDR. Genes with FDR less than 0.1 were reported. Two-by-two comparisons were performed and differentially expressed genes (DEG) were selected using expression levels P ≤ 0.05.

Analysis of RNA sequencing data from Children's Brain Tumor Network

Mutational calls and normalized RNA sequencing (RNA-seq) data from pediatric low-grade (WHO grade I–II) and high-grade astrocytoma (WHO grade III-IV), from the Children's Brain Tumor Network (CBTN) was downloaded via the Gabriella Miller Kids First Data Resource Portal. Data from a total of 314 tumors were divided into three cohorts based on molecular features: RAF-altered astrocytoma (n = 38), defined by mutations or fusions involving RAF1 or BRAF; RAF WT astrocytoma (n = 164), as defined by an absence of RAF1/BRAF alterations; and tumors most likely to be WHO grade I pilocytic astrocytoma (20), as defined by the presence of KIAA1549-BRAF fusion, FAM131-BRAF fusion, or RNF130-BRAF fusions (n = 112). T-effector score was calculated as the geometric mean of the transcripts per kilobase million (TPM) of GZMA, GZMB, PRF1, IFNγ, EOMES, and CD8A (21). The cytolytic score (CYT) was calculated in a similar manner using the geometric mean of the TPM of GZMA and PRF1 as previously described (22). Data analysis and visualization were performed with R version 4.0.0.

RNAscope

FFPE sections were evaluated by RNAscope chromogenic in situ hybridization (CISH) assay for the expression of CXCL14 using Advanced Cell Diagnostics (ACD; RRID:SCR_012481) probes specific for CXCL14 (425299-C2). The RNA Probe PPIB (313909) and dapB (312039) were used as positive and negative control probes, respectively.

Cell culture, condition media, and treatments

Patient-derived PXA cell lines, PXA1 (SF12183) and PXA2 (SF12570), were established from biopsy tissue from recurrent anaplastic PXA, WHO grade III, and maintained under adherent conditions in DMEM media. Denver Brain Tumor Research Group 05 (DBTRG-05MG) (RRID: CVCL_1169) and the immortalized human fetal astrocytes line SVG-p12 (RRID:CVCL_3797) were purchased from ATCC. IDH Mt cells are immortalized normal human astrocytes expressing E6/E7 oncoprotein, catalytic subunit of telomerase, and heterozygous IDH1 R132H mutation (23). Tumor-prone murine progenitor cells (mTPC) containing EGFRvIII were generated in the laboratory and cultured under nonadherent conditions (24). All the established cell lines and primary cell line used were cultured for less than 30 and seven passages, respectively, were maintained at 37°C in a humidified 5% CO2 incubator (Supplementary Table S4), authenticated by analysis of short tandem repeats (STR), and confirmed Mycoplasma negative by PCR twice per year. Glioma cell condition media (CM) was isolated on day 3 when cells were 80% to 90% confluent and centrifuged at 1,000 rpm at 4°C for 4 minutes to remove debris (Supplemental Methods). Cells were treated for 24 hours with recombinant human CXCL14 (Peprotech, catalog no. 300–50) or IFNγ (Peprotech, catalog no. 500-M90). Inhibitor AMD3100 was used at 10 μmol/L (Millipore Sigma, catalog no. A5602) for 15 minutes. siRNA knockdown of CXCL14 was performed using siRNA ON-TARGETplus siRNA SMARTpools targeting CXCL14 (siCXCL14, #SO2860488G) or control siScramble (siScr, #D-001810–10–05; Dharmacon). Cells were analyzed 48 hours from initial transfection. Primer sequences used for qRT-PCR are listed in Supplementary Table S5.

Flow cytometry and ELISA

Single cell suspensions were incubated with Fc Block (BD Biosciences), stained with antibodies (Supplementary Table S2) according to standard protocols, and run on Attune instrument (BD Biosciences). Samples were evaluated with appropriate isotype controls and analyzed using FlowJo software (Tree Star). The CXCL14 ELISA Kit (R&D Systems; catalog no. DY866) was used according to the manufacturer's protocol and absorbance obtained at 450 nm (Epoch Spectrophotometer, Biotech). A standard curve was generated using recombinant ligand and CXCL14 concentration was interpolated.

Isolation of human CD8+ T cells and in vitro transmigration assay

Human CD8+ T cells were isolated using EasySep direct human CD8+ T-cell isolation kit (Stem Cells; catalog no. 19663) and cultured in ImmunoCult TMXF T Cell Expansion Medium (Stem Cells; catalog no. 10981), Human Recombinant IL2 (Stem cells, catalog no. 78220), and ImmunoCult Human CD3/CD28 T Cell Activator (Stem cell, catalog no. 10970). Cells were incubated at 37°C and 5% CO2 for 3 days before plating in the top of Transwell inserts (5-μm pore size; Costar; Supplemental Methods).

Generation of cell lines transduced with CXCL14

mTPCs containing EGFRvIII were transduced with either Lenti open reading frame (ORF) particles, CXCL14-GFP–tagged (mTPCs-CXCL14; Origene technology, catalog no. MR222808L4V), or with pLenti-C-mGFP-P2A (mTPCs-control; Origene Technology, catalog no. PS100093V) according to the manufacturer's protocol. The transduced cells were expanded, cultured as described previously using minimal essential media under nonadherent conditions, and GFP-positive cells were selected by cell sorting (24).

Intracranial injection of transduced mTPCs

All experiments were performed in compliance with institutional guidelines and regulations following approval from the appropriate institutional review board [UCSF Office of Research Institutional Animal Care and Use Committee (IACUC), #AN185514-01]. FVB mice (male, 4 weeks, weight ranged from 20–21 g) were obtained from Charles River (Charles River Laboratories, RRID: IMSR_CRL:207) and nude mice (Foxn1nu, female, 4 weeks, weight ranged from 19–20 g) were obtained from the Jackson Laboratory (The Jackson Laboratory, RRID: IMSR_JAX:007850) and were maintained in accordance with the IACUC guidelines. Intracranial allografts were performed as described previously (ref. 24; Supplemental Methods).

CD8 T immune-cell depletion in vivo

Mice implanted with CXCL14-mTPC received intraperitoneal administration of anti-CD8α (clone GK 2.43) or control isotype anti-IgG (mIgG2b, clone LTF-2) beginning 4 days prior to tumor cell implantation and continuing twice a week until day 60. In parallel, mice without implanted tumor were treated with either anti-CD8α or anti-IgG and splenocytes were isolated on days 2, 23, and 44 postimplantation to verify CD8α cell depletion by flow cytometry (Supplementary Tables S2 and S6).

Statistical analyses

For in vivo studies, 5 to 10 mice were used per treatment group. Data analyzed using GraphPad Prism 9.00 software. Data evaluated by one-way ANOVA, two-sided, unpaired Mann–Whitney U test, and Student t test as stated in figure legends. Survival graphs were calculated using Kaplan–Meier analysis. A P < 0.05 was defined as statistically significant.

Data availability statement

The human sequence data generated in this study are not publicly available due to patient privacy requirements but are available upon reasonable request from the corresponding author. Other data generated in this study are available within the article and its supplementary data files. Some of the data analyzed in this study were obtained from Children's Brain Tumor Network (CBTN) and was downloaded via the Gabriella Miller Kids First Data Resource Portal (https://portal.kidsfirstdrc.org/login).

Quantitative differences in the tumor-associated immune response in molecularly distinct astrocytoma subsets

TILs play an important role in the tumor microenvironment by influencing the growth and progression of cancer cells (25). Given previous reports suggesting that some RAF-altered glioma, predicted to be MAPK-activated, exhibit an increased CD8+ T-cell response (6), we investigated CD3+ TILs in three molecular glioma subsets with astrocytic morphology, as follows: (i) PXA, characterized by RAF alterations, most commonly BRAF p.V600E mutation, and concomitant CDKN2A homozygous deletion, WHO grade II and III (n = 22); (ii) IDH-mutant astrocytoma WHO grade II and III (IDH-mt, n = 29); and (iii) GBM, IDH-WT, WHO grade IV (n = 30; Fig. 1). PXAs harbored a significantly higher mean percentage of CD3+ T cells (8.1%) as compared with IDH-mt astrocytoma (3.9%, P < 0.01) and IDH-WT GBM (2.1%, P < 0.0001; Fig. 1A). In PXA, 23.7% of CD3+ T cells were cytotoxic T cells (dual CD8+CD3+), a 12-fold increase relative to IDH-mt astrocytoma (P < 0.05; Fig. 1B and D; Supplementary Fig. S1A). In contrast, few CD3+ T cells were regulatory T cells (dual FOXP3+CD3+; mean 1.7% vs. 28.9% in IDH-mt astrocytoma, P < 0.05; Fig. 1C; Supplementary Fig. S1B). In PXA, CD3+ T cells were located in perivascular regions as well as diffusely throughout the tumor (Fig. 1E). Cytotoxic activity of T cells requires an activated effector T-cell state, including expression of granzymes, including granzyme B (GZMB), a serine protease secreted by activated cytotoxic effector T cells. In PXA, 22.1.% of CD8+ T cells coexpressed GZMB while in IDH-mt astrocytoma less than 1% of CD8+ T cells coexpressed GZMB (P < 0.05, Fig. 1F and G).

Figure 1.

Quantitative analysis of immune infiltrates in molecularly distinct subsets of astrocytoma. A, CD3+ T cells as a percentage of total cells in PXA (n = 22), IDH-mt (n = 29), and GBM, IDH-WT (GBM, n = 30). Dot plot showing CD3+ T cells double-positive for CD8 (B) or FOXP3 (C) as a percentage of total CD3+ cells in PXA (n = 22) and IDH-mt (n = 27). D, Representative images of dual CD3 (green) and CD8 (violet) immunostaining in PXA. Dotted lines denote magnified region at lower right. Arrowhead highlights CD3+CD8+ T cell. E, Representative images demonstrating regions of PXA with predominant perivascular (arrows) and parenchymal (arrowheads) CD3+ T cells (brown) and associated Iba1+ microglia/macrophages (purple). F, Dot plot showing GZMB+CD8+ cells as a percentage of total CD8+ cells in PXA (n = 22) and IDH-mt (n = 5). G, Representative images demonstrating GZMB+ cells (yellow), CD8+ T cells (violet), and dual positive cells (white) in PXA. Arrowhead highlights GZMB+CD8+ T cell. T-effector cell score (H) and cytolytic cell score (I) from gene expression analysis of pediatric low-grade and high-grade astrocytoma comparing RAF-altered astrocytoma (n = 38) and RAF-WT astrocytoma (n = 164). Data from CBTN downloaded via Kids First Data Resource Portal. J and K, Dot plot showing Iba1+ and CD163+ cells as a percentage of total cells in PXA (n = 23) and IDH-mt (n = 27). L, Representative images of multiplex immunostaining showing Iba1+ (green), CD163+ (violet), and dual positive (white), cells in PXA and IDH-mt. Arrowhead denotes Iba1+CD163+ cells. In (A–C), (F), and (H–K) each dot represents a value from a single patient, black lines represent the mean, and error bars indicate the SD. Nuclei are stained with DAPI (blue). Significance evaluated by one-way ANOVA and two-sided, unpaired Mann–Whitney U test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; NS, P > 0.05. Scale bar: 20 μm. RAF-alt, RAF-altered.

Figure 1.

Quantitative analysis of immune infiltrates in molecularly distinct subsets of astrocytoma. A, CD3+ T cells as a percentage of total cells in PXA (n = 22), IDH-mt (n = 29), and GBM, IDH-WT (GBM, n = 30). Dot plot showing CD3+ T cells double-positive for CD8 (B) or FOXP3 (C) as a percentage of total CD3+ cells in PXA (n = 22) and IDH-mt (n = 27). D, Representative images of dual CD3 (green) and CD8 (violet) immunostaining in PXA. Dotted lines denote magnified region at lower right. Arrowhead highlights CD3+CD8+ T cell. E, Representative images demonstrating regions of PXA with predominant perivascular (arrows) and parenchymal (arrowheads) CD3+ T cells (brown) and associated Iba1+ microglia/macrophages (purple). F, Dot plot showing GZMB+CD8+ cells as a percentage of total CD8+ cells in PXA (n = 22) and IDH-mt (n = 5). G, Representative images demonstrating GZMB+ cells (yellow), CD8+ T cells (violet), and dual positive cells (white) in PXA. Arrowhead highlights GZMB+CD8+ T cell. T-effector cell score (H) and cytolytic cell score (I) from gene expression analysis of pediatric low-grade and high-grade astrocytoma comparing RAF-altered astrocytoma (n = 38) and RAF-WT astrocytoma (n = 164). Data from CBTN downloaded via Kids First Data Resource Portal. J and K, Dot plot showing Iba1+ and CD163+ cells as a percentage of total cells in PXA (n = 23) and IDH-mt (n = 27). L, Representative images of multiplex immunostaining showing Iba1+ (green), CD163+ (violet), and dual positive (white), cells in PXA and IDH-mt. Arrowhead denotes Iba1+CD163+ cells. In (A–C), (F), and (H–K) each dot represents a value from a single patient, black lines represent the mean, and error bars indicate the SD. Nuclei are stained with DAPI (blue). Significance evaluated by one-way ANOVA and two-sided, unpaired Mann–Whitney U test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; NS, P > 0.05. Scale bar: 20 μm. RAF-alt, RAF-altered.

Close modal

Alterations in RAS/RAF/MAPK pathway signaling are relatively common in pediatric and young adult low grade glioma (26). Given the relatively high cytotoxic T-effector TIL population in PXA, we investigated whether this might be a broader phenotype across RAF-altered glioma. Using available genomic and transcriptomic data from 314 pediatric low- and high-grade astrocytoma, WHO grade I to IV, from the CBTN we identified three cohorts of tumors (defined in Methods): RAF-altered astrocytoma (n = 38), RAF-WT astrocytoma (n = 164), and pilocytic astrocytoma (n = 112). RAF-altered astrocytoma had an increased T-cell score (P < 0.001) and an increased cytolytic T-cell score (P < 0.0001) relative to RAF-WT astrocytoma (Fig. 1H and I) and pilocytic astrocytoma (Supplementary Fig. S1C and S1D).

Myeloid cells help regulate the tumor-associated immune response in glioma (9, 10). In PXA (n = 23), there were increased numbers of tumor-associated microglial/macrophages (TAM), as defined by Iba1 expression, relative to IDH-mt astrocytoma (n = 27; mean 32.2% vs. 23.3%, respectively; P < 0.001; Fig. 1J–L). TAMs can exhibit different activation phenotypes. The proportion of CD204- (macrophage scavenger receptor 1) expressing cells was comparable in PXA and IDH-mt astrocytoma [15.9% vs. 15.4%, respectively; P = 0.9, not significant (NS); Supplementary Fig. S1E). In contrast, the proportion of CD163 cells was increased in PXA (3.7% vs. 1.4%, respectively; P < 0.05), although the total population of CD163 cells was low in both tumor types (Fig. 1K and L). Taken together, these data suggest that PXA have a more robust cytotoxic TIL response and TAM response as compared with IDH-mt astrocytoma.

High-plex proteomic profiling of the immune response in PXA and identification of potential therapeutic targets

To better characterize the tumor-associated immune response in PXA, we performed high-plex spatial profiling of 46 immune-response–related proteins in a cohort of PXA (n = 9) and IDH-mt astrocytoma (n = 8). Analysis of CD3+ T-cell rich regions revealed both quantitative and qualitative differences in the tumor-associated immune response (Fig. 2A). PXA demonstrated an increase in overall CD45+ immune cells, CD3+ T cells, CD8+ T cells, HLA-DR antigen-presenting cells, and CD20+ B cells (P < 0.05 for all). In contrast, CD56+ natural killer (NK) cells were not significant but trended lower in PXA and CD11c+ myeloid cells also did not demonstrate a difference (Fig. 2B).

Figure 2.

High-plex spatial proteomic profiling of immune response in PXA and IDH-mt astrocytoma. A, Representative images of PXA (left) and IDH-mt astrocytoma (right) analyzed by digital spatial profiling. CD3+ T-cell–rich ROIs were identified by multiplex immunostaining for CD3 (cyan), Iba1 (red), CD34 (yellow), and nuclei (blue) in PXA (n = 9) and IDH-mt (n = 8). B, Bar graph showing mean immune cell protein marker expression level for PXA (red) and IDH-mt (blue). C–E, Box plots showing protein expression of T lymphocyte co-stimulatory (CD127, CD27, and CD25) and activation (GZMA, GZMB) molecules in PXA (red) and IDH-mt (blue). F and G, Box plots showing protein expression of immunomodulatory molecules in PXA (red) and IDH-mt (blue). H, Cell surface expression of B7-H3 on patient-derived PXA cell lines (PXA1, PXA2), DBTRG, and NHA-IDH1 mutant cell line (IDH-mt) as determined by flow cytometry expressed as mean fluorescent intensity (MFI; n = 3 replicates per line). Data shown are mean | $ \pm \ $ |SEM or box plots denoting 25%, median, 75%, and whiskers represent the minimum and maximum. All experiments analyzed by two-sided, unpaired Mann–Whitney U test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; NS, P > 0.05. Scale bar: 20 μm.

Figure 2.

High-plex spatial proteomic profiling of immune response in PXA and IDH-mt astrocytoma. A, Representative images of PXA (left) and IDH-mt astrocytoma (right) analyzed by digital spatial profiling. CD3+ T-cell–rich ROIs were identified by multiplex immunostaining for CD3 (cyan), Iba1 (red), CD34 (yellow), and nuclei (blue) in PXA (n = 9) and IDH-mt (n = 8). B, Bar graph showing mean immune cell protein marker expression level for PXA (red) and IDH-mt (blue). C–E, Box plots showing protein expression of T lymphocyte co-stimulatory (CD127, CD27, and CD25) and activation (GZMA, GZMB) molecules in PXA (red) and IDH-mt (blue). F and G, Box plots showing protein expression of immunomodulatory molecules in PXA (red) and IDH-mt (blue). H, Cell surface expression of B7-H3 on patient-derived PXA cell lines (PXA1, PXA2), DBTRG, and NHA-IDH1 mutant cell line (IDH-mt) as determined by flow cytometry expressed as mean fluorescent intensity (MFI; n = 3 replicates per line). Data shown are mean | $ \pm \ $ |SEM or box plots denoting 25%, median, 75%, and whiskers represent the minimum and maximum. All experiments analyzed by two-sided, unpaired Mann–Whitney U test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001; NS, P > 0.05. Scale bar: 20 μm.

Close modal

T-lymphocyte activation and immune function are regulated by co-stimulatory molecules such as CD25, CD27, and CD127. CD27 is required for the T-cell expansion and contributes to the accumulation of effector T cells, CD25 (IL2) promotes the differentiation of T cells into effector cells and is transiently upregulated upon activation of T-effectors, and CD127 helps mediate T-cell homeostasis (27). Each of these co-stimulatory molecules was significantly increased in PXA relative to IDH-mt astrocytoma (P < 0.05, Fig. 2C). In addition, cytotoxic T-cell activation markers, granzyme A (GZMA) and GZMB, were increased in PXA relative to IDH-mt astrocytoma (P < 0.01 and 0.05, respectively; Fig. 2D and E), providing further support to the activated T-cell phenotype.

Immune checkpoints are also crucial for controlling effector T-cell function. Increased expression of inhibitory immune checkpoints is common across diverse tumor types, yet the exact mechanism of immune evasion appears to differ. Neither programmed death 1 (PD-1) nor programmed death ligand 1 (PD-L1) expression levels were significantly different in PXA and IDH-mt astrocytoma (Fig. 2F). However, PXA had robust expression of B7 homolog 3 (B7-H3), and this was significantly higher than in IDH-mt astrocytoma (Fig. 2G). Protein expression of CTLA-4 and OX40 L were also increased and expression of GITR was slightly decreased in PXA relative to IDH-mt astrocytoma (Supplementary Fig. S2A). The expression of B7-H3 on the cell surface is critical for its immunomodulatory functions and for it to be targeted therapeutically. Therefore, we investigated the cell surface expression of B7-H3 protein by flow cytometry in three RAF-altered glioma lines, two primary patient-derived PXA cell lines, PXA1 and PXA2, and DBTRG (Supplementary Table S4). All three lines had robust cell surface B7-H3 expression relative to the IDH-mutant expressing normal human astrocytes (NHA; IDH-mt; Fig. 2H). These data suggest PXA has a relatively immune-rich microenvironment and identify B7H3 as a potential immunotherapeutic target in PXA.

Enrichment of immune-related genes including antigen processing and presentation in PXA

To identify factors that may promote an immune-rich microenvironment in PXA, we performed transcriptional analysis of immunologically relevant genes in PXA (n = 21), IDH-mt astrocytoma (n = 6), and nonneoplastic brain (NB; n = 3) using Nanostring. Differential gene expression analysis identified 151 differentially expressed genes between PXA and IDH-mt astrocytoma (P < 0.05) the majority of which (70%; 105/151) were upregulated in PXA consistent with a more immune-rich microenvironment. The list of the top 20 upregulated gene in PXA is shown in Supplementary Table S7. Indeed, several inflammatory immune response pathways were enriched based on Gene Ontology (GO) enrichment analysis with the antigen processing and presentation pathway as most significantly enriched (Fig. 3A).

Figure 3.

Enrichment of immune-related genes including antigen processing and presentation in PXA. A, Top 10 enriched biological processes from GO enrichment analysis of differentially expressed genes between PXA and IDH-mt astrocytoma. B, Dot plot showing high expression of MHC-I genes (HLA-A, -B and -C) in PXA (red, n = 21) compared with IDH-mt (blue, n = 6). C, Heatmap showing upregulated MHC-I genes in PXA (n = 21, red) compared with NB (n = 3, green); P < 0.01 for all, Student t test. Fold change ≤ 0.5 represents downregulated (yellow) and ≥ 1.5 represents upregulated (blue). D and E, Representative images and quantification of HLA-ABC (green) expression in PXA (n = 9) and IDH-mt (n = 7). Arrow highlights HLA-ABC–positive cell. F, Representative image of multiplex immunostaining highlighting HLA-ABC (green) expression on BRAFV600E-mutant tumor cells (violet) denoted by arrowhead in PXA. G, Surface expression of HLA-ABC assessed by flow cytometry expressed as MFI (n = 3 replicates per line) in NHA-IDH1–mt cell lines, DBTRG, PXA1, and PXA2. In (B) each dot represents a value from a single patient, black lines represent mean | $\ \pm \ $ |SD. In (E) and (G) box plots are mean + SEM. Results are data from at least technical triplicates and are representative of experiments performed in biologic triplicate. Significance obtained using ANOVA, two-sided (G), unpaired Mann–Whitney U test (E). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Nuclei are stained with DAPI (blue). Scale bar: D, 50 μm; and F, 20 μm.

Figure 3.

Enrichment of immune-related genes including antigen processing and presentation in PXA. A, Top 10 enriched biological processes from GO enrichment analysis of differentially expressed genes between PXA and IDH-mt astrocytoma. B, Dot plot showing high expression of MHC-I genes (HLA-A, -B and -C) in PXA (red, n = 21) compared with IDH-mt (blue, n = 6). C, Heatmap showing upregulated MHC-I genes in PXA (n = 21, red) compared with NB (n = 3, green); P < 0.01 for all, Student t test. Fold change ≤ 0.5 represents downregulated (yellow) and ≥ 1.5 represents upregulated (blue). D and E, Representative images and quantification of HLA-ABC (green) expression in PXA (n = 9) and IDH-mt (n = 7). Arrow highlights HLA-ABC–positive cell. F, Representative image of multiplex immunostaining highlighting HLA-ABC (green) expression on BRAFV600E-mutant tumor cells (violet) denoted by arrowhead in PXA. G, Surface expression of HLA-ABC assessed by flow cytometry expressed as MFI (n = 3 replicates per line) in NHA-IDH1–mt cell lines, DBTRG, PXA1, and PXA2. In (B) each dot represents a value from a single patient, black lines represent mean | $\ \pm \ $ |SD. In (E) and (G) box plots are mean + SEM. Results are data from at least technical triplicates and are representative of experiments performed in biologic triplicate. Significance obtained using ANOVA, two-sided (G), unpaired Mann–Whitney U test (E). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Nuclei are stained with DAPI (blue). Scale bar: D, 50 μm; and F, 20 μm.

Close modal

MHC class I (MHC-I) expression is essential for antigen presentation and processing to CD8+ T cells. In PXA, the expression of MHC-I molecules, human leukocyte antigen (HLA)-A, -B, and -C, were upregulated relative to IDH-mt astrocytoma (P < 0.01, for all; Fig. 3B) and NB (P < 0.01, for all; Fig. 3C). In dense PXA tumor regions, most cells expressed HLA-ABC protein, mean 57.8% of total cells, and the percentage was increased relative to IDH-mt astrocytoma, mean 21% (P < 0.05; Fig. 3D and E). Dual immunostaining demonstrated HLA-ABC expression on BRAF V600E-mutant tumor cells (Fig. 3F), and cell surface expression of HLA-ABC was robust in all three RAF-altered glioma lines (Fig. 3G). Taken together, these data suggest that an increased antigen processing and presentation pathway, including elevated MHC-I expression, may help promote a robust tumor-associated immune response in PXA.

CXCL14 is upregulated and secreted by tumor cells and patient-derived PXA cell lines

Chemokines are chemotactic cytokines known to help regulate immune cell trafficking and influence the TIL response (28). Differential gene expression analysis revealed an upregulation of the chemokine (C-X-C motif) ligand 14 (CXCL14) in PXA relative to IDH-mt astrocytoma (P < 0.0001; Fig. 4A). The mean normalized mRNA count in PXA was 20-fold higher relative to that in IDH-mt astrocytoma (P < 0.001; Fig. 4B). Comparing CXCL14 expression across pediatric low- and high-grade astrocytoma, WHO grade II to IV, RAF altered astrocytoma had increased CXCL14 expression relative to RAF-WT astrocytoma (P < 0.0001; Fig. 4C) and pilocytic astrocytoma (Supplementary Fig. S3A).

Figure 4.

Increased CXCL14 expression in PXA tumors and patient-derived cell lines. A, Volcano plot showing differentially expressed genes between PXA and IDH-mt astrocytoma highlighting CXCL14 (blue box). B, Dot plot showing CXCL14 normalized mRNA count in PXA (red, n = 21) versus IDH-mt (blue, n = 6). C, Increased expression of CXCL14 in RAF-altered astrocytoma (n = 38) versus RAF-WT astrocytoma (n = 164) in a cohort of pediatric low-grade and high-grade astrocytoma. Data from CBTN downloaded via Kids First Data Resource Portal. D, Representative images from serial sections showing BRAF V600E-mutant tumor cells (brown, arrow) corresponding to regions of abundant CXCL14 mRNA-expressing cells (red, arrow). Dual immunostaining for Iba1 (green, arrowhead) highlights lack of CXCL14 expression in microglia/macrophages. E, Representative image of BRAFV600E (red), CXCL14 (yellow), and nuclei (blue) in BRAF V600E-mutant PXA. F, Expression of CXCL14 in DBTRG and patient-derived PXA lines PXA1 and PXA2 as compared with NHA-IDH1–mt cell line by qRT-PCR. G, Secreted CXCL14 protein in cell-free CM from DBTRG, PXA1, PXA2, and NHA-IDH1–mutant cell lines measured by ELISA. H, Flow cytometric analysis of cell surface HLA-ABC expression on IDH-mt after addition of CXCL14 ligand (200 ng/mL, 24 hours). In (B) and (C) each dot represents a value from a single patient; black lines represent the mean | $ \pm \ $ |SD. In (D) red dots correspond to single RNA molecules. In F to H, data shown in box plots are mean + SEM. Results are data from at least technical triplicates and are representative of experiments performed in biologic triplicate. Significance obtained using ANOVA and two-sided, unpaired Mann–Whitney U test. **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Scale bar: D, 100 μmol/L; E, 20 μm. Ctrl, control.

Figure 4.

Increased CXCL14 expression in PXA tumors and patient-derived cell lines. A, Volcano plot showing differentially expressed genes between PXA and IDH-mt astrocytoma highlighting CXCL14 (blue box). B, Dot plot showing CXCL14 normalized mRNA count in PXA (red, n = 21) versus IDH-mt (blue, n = 6). C, Increased expression of CXCL14 in RAF-altered astrocytoma (n = 38) versus RAF-WT astrocytoma (n = 164) in a cohort of pediatric low-grade and high-grade astrocytoma. Data from CBTN downloaded via Kids First Data Resource Portal. D, Representative images from serial sections showing BRAF V600E-mutant tumor cells (brown, arrow) corresponding to regions of abundant CXCL14 mRNA-expressing cells (red, arrow). Dual immunostaining for Iba1 (green, arrowhead) highlights lack of CXCL14 expression in microglia/macrophages. E, Representative image of BRAFV600E (red), CXCL14 (yellow), and nuclei (blue) in BRAF V600E-mutant PXA. F, Expression of CXCL14 in DBTRG and patient-derived PXA lines PXA1 and PXA2 as compared with NHA-IDH1–mt cell line by qRT-PCR. G, Secreted CXCL14 protein in cell-free CM from DBTRG, PXA1, PXA2, and NHA-IDH1–mutant cell lines measured by ELISA. H, Flow cytometric analysis of cell surface HLA-ABC expression on IDH-mt after addition of CXCL14 ligand (200 ng/mL, 24 hours). In (B) and (C) each dot represents a value from a single patient; black lines represent the mean | $ \pm \ $ |SD. In (D) red dots correspond to single RNA molecules. In F to H, data shown in box plots are mean + SEM. Results are data from at least technical triplicates and are representative of experiments performed in biologic triplicate. Significance obtained using ANOVA and two-sided, unpaired Mann–Whitney U test. **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. Scale bar: D, 100 μmol/L; E, 20 μm. Ctrl, control.

Close modal

In glioma, potential cell sources for CXCL14 include both immune cells and tumor cells (29). In BRAF V600E-mutant tumor dense regions, dual RNAscope revealed robust expression of CXCL14 transcript in Iba1-negative regions (Fig. 4D). Furthermore, dual immunostaining demonstrated CXCL14 protein expression in BRAF V600E-mutant tumor cells (Fig. 4E). In addition, all three RAF-altered glioma lines, PXA1, PXA2, and DBTRG, had high expression of CXCL14 transcript (Fig. 4F) and ELISA demonstrated robust secretion of CXCL14 protein in PXA1 (285.5 pg/mlL), PXA2 (250 pg/mL), and DBTRG (196 pg/mL) cells (Fig. 4G).

In addition to demonstrated roles in immune-cell modulation, CXCL14 has also been implicated in the regulation of MHC-I expression (30). Given the increased MHC-I expression in PXA (Fig. 3C), we investigated whether CXCL14 could alter cell surface expression of HLA-ABC. Addition of exogenous CXCL14 increased cell surface HLA-ABC expression on IDH-mt cells relative to control (Fig. 4H; Supplementary Fig. S3B, P < 0.01). These data suggest CXCL14 is upregulated in RAF-altered tumor cells and may promote increased cell surface MHC-I.

Secreted CXCL14 promotes activated CD8 T-cell chemotaxis in vitro

Our analysis suggested that PXAs harbor abundant CD8+ cytotoxic T cells (Fig. 1BG). To investigate whether factors secreted from tumor cells may promote chemotaxis of activated CD8+ T cells, we performed transwell migration assays using cell-free CM from the three RAF-altered glioma lines, PXA1, PXA2, and DBTRG. CM from all three lines induced increased chemotaxis of activated CD8+ T cells relative to media alone (Fig. 5A). Given that CXCL14 has been implicated in promoting immune cell recruitment in neoplastic and nonneoplastic settings (30, 31), we first demonstrated CXCL14 was sufficient to promote directional migration of activated CD8+ T cells in a dose-dependent manner (Fig. 5B). Next, we knocked down CXCL14 in DBTRG cells (Fig. 5C) and observed a reduction in CM-induced chemotaxis of activated CD8+ T cells to background control levels (Fig. 5D).

Figure 5.

Secreted factors from tumor cells promote activated CD8 T-cell chemotaxis in vitro. A, Relative chemotaxis of activated CD8 T cells toward cell-free CM derived from DBTRG, PXA1, and PXA2 as compared with control media by transwell migration assay. B, Chemotaxis of activated CD8+ T cells toward increasing concentration of CXCL14 ligand (25–100 ng/mL) by transwell migration assay. C, CXCL14 ELISA of cell-free CM after transduction of DBTRG with four pooled siRNA targeting CXCL14 gene (siCXCL14) versus siScr. D, Chemotaxis of activated CD8+ T cells toward cell-free CM from siCXCL14 and siScr transduced DBTRG. E, CXCL14-induced chemotaxis of activated CD8+ T cells following inhibition of CXCR4 (AMD3100) compared with control. Results are data from at least technical triplicates and are representative of experiments performed in biologic triplicate. Mean + SEM. Significance was obtained using one-way ANOVA and Student t test. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; NS, P > 0.05. Ctrl, control.

Figure 5.

Secreted factors from tumor cells promote activated CD8 T-cell chemotaxis in vitro. A, Relative chemotaxis of activated CD8 T cells toward cell-free CM derived from DBTRG, PXA1, and PXA2 as compared with control media by transwell migration assay. B, Chemotaxis of activated CD8+ T cells toward increasing concentration of CXCL14 ligand (25–100 ng/mL) by transwell migration assay. C, CXCL14 ELISA of cell-free CM after transduction of DBTRG with four pooled siRNA targeting CXCL14 gene (siCXCL14) versus siScr. D, Chemotaxis of activated CD8+ T cells toward cell-free CM from siCXCL14 and siScr transduced DBTRG. E, CXCL14-induced chemotaxis of activated CD8+ T cells following inhibition of CXCR4 (AMD3100) compared with control. Results are data from at least technical triplicates and are representative of experiments performed in biologic triplicate. Mean + SEM. Significance was obtained using one-way ANOVA and Student t test. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; NS, P > 0.05. Ctrl, control.

Close modal

CXCR4 is one of potentially several receptors that CXCL14 can use or modulate (32, 33). In PXA, CXCR4 expression was 5-fold higher relative to IDH-mutant astrocytoma (P < 0.001; Supplementary Fig. S4A) and activated CD8+ T cells demonstrated robust expression of CXCR4 (Supplementary Fig. S4B). Inhibition of CXCR4 with AMD3100 (34) blocked CXCL14-induced chemotaxis of activated CD8+ T cells in vitro (Fig. 5E). These data demonstrate that secreted CXCL14 can promote activated CD8+ T cell chemotaxis in a CXCR4-dependent manner, further supporting a potential role for CXCL14 in the tumor-associated immune response.

PXA also demonstrated increased TAMs relative to IDH-mt astrocytoma (Fig. 1J–L). As CXCL14 has previously been implicated in myeloid cell migration (35), we also evaluated the ability of secreted factors from tumor cells, including CXCL14, to influence macrophage chemotaxis. Cell-free CM from PXA1, PXA2, and DBTRG, promoted chemotaxis of the macrophage cell line MV4–11 in transwell migration assays relative to media alone (Supplementary Fig. S4C). We also confirmed that exogenous CXCL14 was sufficient to promote the directional migration of MV4–11 cells in a dose-dependent manner (Supplementary Fig. S4D).

Secreted CXCL14 promotes a tumor-associated CD8+ T-cell response and prolongs survival

IDH-mutant diffuse glioma are characterized as lymphocyte depleted (1–3), and median CXCL14 expression is lower than that in normal brain (Supplementary Fig. S5A). To investigate whether CXCL14 might be sufficient to promote a CD8+ T-cell response in diffuse glioma in vivo, we engineered mTPC from an immunocompetent murine model of malignant glioma (25) to express CXCL14 (CXCL14-mTPC; Fig. 6). CXCL14-mTPC had increased expression of CXCL14 mRNA (Supplementary Fig. S5B) compared with control mTPCs and secreted CXCL14 protein at similar levels to human patient-derived PXA cell lines (282.8 pg/mL; Fig. 6A and Fig. 4G). Murine TPCs were then implanted into the striatum of immunocompetent mice and brains were examined at study end point. Histologic and immunophenotypic analysis revealed a striking increase in the CD3+ and CD8+ T-cell infiltrates in tumors generated from CXCL14 as compared with control-mTPCs (Fig. 6B and C; Supplementary Fig. S5C). In contrast, the number of Iba1+ microglia/macrophages was similar (Fig. 6D; Supplementary Fig. S5D). There was no increased immune infiltrate outside of the tumor region (Supplementary Fig. S5E). In vitro, the CXCR4 inhibitor AMD3100 inhibited CXCL14-induced chemotaxis of activated CD8+ T cells (Fig. 5E). In vivo, treatment with the CXCR4 inhibitor abrogated the CXCL14-mediated CD8+ T-cell response (Supplementary Fig. S5F).

Figure 6.

Secreted CXCL14 promotes an antitumor CD8+ T-cell response in vivo. A, Secreted CXCL14 in cell-free CM from mTPCs expressing CXCL14 or GFP-control. B, Quantification of CD8+ T cells as percentage of total cells in CXCL14 (n = 8 mice) and control tumors (n = 5 mice). C, Representative images demonstrating CD3+ (green), CD8+ (violet), and dual CD3+CD8+ (white, arrow) T cells in CXCL14 and control tumors. D, Quantification of Iba1+ TAMs per tumor area in CXCL14 (n = 7 mice) relative to control tumors (n = 4 mice). E, Kaplan–Meier survival analysis. Mice with intracerebral allograft of CXCL14 mTPC have prolonged survival (median survival of 64.8 days) relative to control mTPC (median survival of 35.8 days; P < 0.003, n = 8, and n = 5, respectively). F, Quantification of proliferating cells as denoted by Ki-67 per tumor area in CXCL14 (n = 7 mice) relative to control tumors (n = 4 mice). G, Schema for tumor implantation in nude mice (n = 5 per group) and FVB mice (n = 10 per group) with biweekly anti-CD8α treatment to deplete CD8+ T cells (blue arrows). Corresponding isotype antibody used as control. Nude mice were implanted with either CXCL14 mTPC or control mTPC and FVB mice were implanted with CXCL14 mTPC. Kaplan–Meier survival analysis (bottom). H, Correlation between CXCL14 gene expression and CD8A gene expression in pediatric astrocytoma (CBTN; Spearman correlation coefficient, r = 0.45, n = 202, P < 0.0001). Bar graphs represent mean + SEM. Significance was obtained using Student t test. Figure 6G (top FVB and nude mouse) was created with BioRender.com. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; NS, P > 0.05. Scale bar: 100 μm. Ctrl, control.

Figure 6.

Secreted CXCL14 promotes an antitumor CD8+ T-cell response in vivo. A, Secreted CXCL14 in cell-free CM from mTPCs expressing CXCL14 or GFP-control. B, Quantification of CD8+ T cells as percentage of total cells in CXCL14 (n = 8 mice) and control tumors (n = 5 mice). C, Representative images demonstrating CD3+ (green), CD8+ (violet), and dual CD3+CD8+ (white, arrow) T cells in CXCL14 and control tumors. D, Quantification of Iba1+ TAMs per tumor area in CXCL14 (n = 7 mice) relative to control tumors (n = 4 mice). E, Kaplan–Meier survival analysis. Mice with intracerebral allograft of CXCL14 mTPC have prolonged survival (median survival of 64.8 days) relative to control mTPC (median survival of 35.8 days; P < 0.003, n = 8, and n = 5, respectively). F, Quantification of proliferating cells as denoted by Ki-67 per tumor area in CXCL14 (n = 7 mice) relative to control tumors (n = 4 mice). G, Schema for tumor implantation in nude mice (n = 5 per group) and FVB mice (n = 10 per group) with biweekly anti-CD8α treatment to deplete CD8+ T cells (blue arrows). Corresponding isotype antibody used as control. Nude mice were implanted with either CXCL14 mTPC or control mTPC and FVB mice were implanted with CXCL14 mTPC. Kaplan–Meier survival analysis (bottom). H, Correlation between CXCL14 gene expression and CD8A gene expression in pediatric astrocytoma (CBTN; Spearman correlation coefficient, r = 0.45, n = 202, P < 0.0001). Bar graphs represent mean + SEM. Significance was obtained using Student t test. Figure 6G (top FVB and nude mouse) was created with BioRender.com. *, P < 0.05; **, P < 0.01; ****, P < 0.0001; NS, P > 0.05. Scale bar: 100 μm. Ctrl, control.

Close modal

CXCL14 expression had no influence on tumor cell proliferation in vivo (Fig. 6F; Supplementary Fig. S5G) or cell growth in vitro, as demonstrated by similar doubling times for CXCL14-mTPC and control mTPC (29.28 hours and 27.26 hours, respectively; P = 0.4). Yet, overall survival was significantly prolonged in mice harboring CXCL14-expressing tumors relative to control (Fig. 6E). To evaluate whether T lymphocytes, specifically CD8+ cytotoxic T cells, were required for the survival difference between CXCL14-expressing and control-mTPC two approaches were taken. As demonstrated in Fig. 6G, there was no CXCL14-mediated survival advantage in nude mice lacking mature T cells or in otherwise immunocompetent mice with CD8+ T-cell depletion (Supplementary Fig. S5H). CD8+ T cells were required for the prolonged survival mediated by CXCL14 (anti-CD8, n = 10) versus anti-IgG control (n = 10, P = 0.0035). While median CXCL14 expression is low in several tumor subtypes, variation across tumors is evident. Using gene expression data, CXCL14 expression was correlated with increased CD8A expression in both pediatric and adult astrocytoma [r = 0.45, P < 0.0001, CBTN n = 202; r = 0.60, P < 0.0001, The Cancer Genome Atlas (TCGA) n = 310, respectively; Fig. 6H; Supplementary Fig. S5I]. Together these data suggest CXCL14 expression in tumor cells can promote an antitumor cytotoxic CD8+ T-cell response.

Improved immunotherapeutic strategies for patients with brain tumor requires a mechanistic understanding of the determinants of the tumor-associated immune response. Using a RAF-driven glioma subtype with a robust immune response, we identify tumor-intrinsic factors that modulate the tumor microenvironment and promote a cytotoxic CD8+ T-cell response, including increased expression of MHC-I and secretion of CXCL14. Highlighting the role for CXCL14 in the tumor microenvironment, increased expression of CXCL14 in an immunocompetent murine model for malignant glioma resulted in an increased tumor-associated immune response, including increased CD8+ T cells, and conferred prolonged survival in a CD8 T-cell–dependent manner. Interestingly, CXCL14 expression was correlated with increased CD8 T-cell expression in both pediatric and adult astrocytic tumors. Modulation of the tumor microenvironment may be an important immunotherapeutic strategy to promote T-cell effector function in diverse brain tumors.

Integrating data from multiplex immune profiling, spatial proteomics, and gene expression analysis we demonstrate a robust immune microenvironment in PXA, including increased cytotoxic CD8+ T cells and TAMs. These data are consistent with several studies suggesting that molecularly distinct brain tumor subtypes harbor unique immune microenvironments (1, 4–6, 12, 36), including those suggesting that some subsets of MAPK-activated brain tumors have a pronounced CD8+ T-cell response (6, 36). While our study focused on the tumor-associated immune response in PXA, analysis of RNA-seq data from pediatric astrocytoma suggests a larger subset of RAF-altered astrocytic glioma may have an increased cytotoxic TIL response and increased CXCL14 expression. These tumors may be more vulnerable to immunotherapeutic strategies. In support of this, activating MAPK mutations have been associated with GBM response to immunotherapy (37, 38). Additional determinants beyond MAPK activation, however, are likely as pilocytic astrocytoma with activating MAPK mutations did not show increased cytolytic T-cell score or CXCL14 expression.

Recent studies have highlighted the importance of the tumor secretome in shaping many aspects of the microenvironment in diverse tumor types (39, 40). In particular, chemokines produced by tumor cells and immune cells help regulate the tumor-associated immune response (28). In our analysis, despite the relatively robust tumor-associated immune response, many of the cytokines commonly implicated in lymphocyte chemotaxis or as mediators of a proinflammatory response demonstrated either no significant differences or reduced expression in PXA relative to IDH-mt astrocytoma. The increased expression of the CXC motif chemokine 14 (CXCL14, BRAK, or MIP-2G) in PXA was a striking exception. CXCL14 is a chemokine implicated in both homeostatic immune functions, particularly in squamous epithelium, and host-tumor immune interactions (29, 41). In several cancers CXCL14 expression is altered, with increased expression reported in tumor cells in papillary thyroid carcinoma (42) and decreased tumor cell expression associated with disease progression in human papillomavirus (HPV)-associated cervical and head and neck carcinoma (30, 31). Manipulation of CXCL14 expression in other systems has revealed roles for the chemokine in the recruitment of immune cells, including CD8 cytotoxic T cells, dendritic cells, and NK cells (30, 43). We demonstrated that CXCL14 is expressed and secreted by tumor cells and promotes chemotaxis of activated CD8+ T cells in a dose-dependent manner. Conversely, CXCL14 knockdown blocked condition media-induced chemotaxis of activated CD8+ T cells. While several factors may promote the immune-rich microenvironment in PXA, our data suggest CXCL14 is an important factor.

The diverse roles reported for CXCL14 in tumor progression may reflect a combination of factors including the cell source of CXCL14, whether the chemokine functions predominantly in an autocrine or paracrine manner, and receptor usage. While CXCL14 is considered an orphan chemokine, some CXCL14-dependent functions may be mediated by CXCR4, the receptor for CXCL12 (32). Recent studies suggest CXCL14 may have both CXCL12-dependent and CXCL12-independent roles in CXCR4 signaling (33). In PXA, both CXCL14 and CXCR4 expression was increased relative to IDH-mt astrocytoma and inhibition of CXCR4 blocked CXCL14-mediated chemotaxis of activated CD8+ T cells. In our murine model for malignant glioma, inhibition of CXCR4 blocked the CXCL14-dependent increase in CD8+ T cells in vivo. Taken together these data suggest the CXCL14-mediated CD8+ T-cell response is mediated largely through CXCR4, although contributions from other receptors cannot be excluded.

In PXA, MHC-I molecules, HLA-ABC, were upregulated. Patient-derived PXA lines also had increased cell surface HLA-ABC expression relative to IDH-mt cells. MHC-I mediated antigen presentation is crucial for CD8 effector T-lymphocyte responses and MHC-I expression correlates with the presence of tumor infiltrating CD8+ T cells in several cancers (44). Loss or downregulation of MHC-I is considered a mechanism of T-lymphocyte evasion and is associated with tumor progression (45, 46). While the precise mechanisms that promote MHC-I expression in PXA are an area of ongoing study, CXCL14 may be a factor, as exogenous CXCL14-induced increased cell-surface HLA-ABC expression on IDH-mt cells. A similar increase in MHC-I was observed in HPV-associated head and neck carcinoma upon reexpression of CXCL14 (30). Given the importance of MHC-I–mediated antigen presentation for CD8 effector function, robust expression of MHC-I in PXA may promote the CD8+ cytolytic T-cell response.

As CXCL14 has been implicated in regulating brain immune development (47), we hypothesized that tumor cell expression of CXCL14 may promote a more robust tumor-associated immune response. Using an immunocompetent murine model for malignant glioma, CXCL14 expression was sufficient to promote an increased tumor-associated CD8+ T-cell response. CXCL14 also conferred prolonged survival in a CD8+ T-cell–dependent manner.

In conclusion, our study demonstrates the relatively immune-rich microenvironment of PXA and identifies CXCL14 as an important factor in promoting an antitumor cytotoxic T-cell response in the brain tumor microenvironment. As recurrence and progression is not uncommon in PXA, immunotherapeutic strategies may be important to consider. In this regard, we identified robust B7-H3 expression in PXA and CAR T cells targeting B7-H3 have been proposed as a potentially promising therapy in subgroups of adult and pediatric patients with glioma (48, 49). More broadly, our data elucidate a novel mechanism of immune regulation in the brain and suggest CXCL14 modulation may be a potential therapeutic strategy to promote an antitumor cytotoxic T-cell response.

No disclosures were reported.

A. Kumar: Conceptualization, formal analysis, investigation, methodology, writing–original draft, project administration, writing–review and editing. E. Mohamed: Data curation, methodology. S. Tong: Resources, data curation, methodology. K. Chen: Resources. J. Mukherjee: Methodology. Y. Lim: Resources, methodology. C.M. Wong: Data curation, methodology. Z. Boosalis: Resources, data curation, methodology. A. Shai: Resources, data curation, methodology. R.O. Pieper: Resources, software, methodology. N. Gupta: Resources, methodology. A. Perry: Resources, investigation, writing–review and editing. A.W. Bollen: Resources, visualization, methodology. A.M. Molinaro: Resources, software, methodology. D.A. Solomon: Resources, methodology. J.T.C. Shieh: Conceptualization, resources, software, supervision, validation, project administration, writing–review and editing. J.J. Phillips: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, writing–original draft, project administration, writing–review and editing.

This study was supported by UCSF Brain Tumor SPORE Developmental Research Project (NIH/NCI 5P50CA097257-18 to A.M. Molinaro, J.J. Phillips; T32 CA151022 to E. Mohamed; NIH/NCI U01 CA168878 to J.J. Phillips), Panattoni Family Research Program (to J.J. Phillips), and UCSF Loglio collective (to J.J. Phillips). We also thank the T.J. Martell Foundation, the Gerson and Barbara Bakar Philanthropic Fund, and the Sence Foundation for support. Resources were provided by UCSF Brain Tumor SPORE Biorepository (NIH/NCI 5P50CA097257-18 to J.J. Phillips), UCSF Helen Diller Family Comprehensive Cancer Center Laboratory for Cell Analysis (P30CA082103), and data were provided by the Children's Brain Tumor Network. We thank Dr. Claudia Petritsch for her scientific input and Sonia Patel for her technical input. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. The authors declare that they have no conflicts of interest with the contents of this article.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Kohanbash
G
,
Carrera
DA
,
Shrivastav
S
,
Ahn
BJ
,
Jahan
N
,
Mazor
T
, et al
.
Isocitrate dehydrogenase mutations suppress STAT1 and CD8+ T cell accumulation in gliomas
.
J Clin Invest
2017
;
127
:
1425
37
.
2.
Müller
S
,
Kohanbash
G
,
Liu
SJ
,
Alvarado
B
,
Carrera
D
,
Bhaduri
A
, et al
.
Single-cell profiling of human gliomas reveals macrophage ontogeny as a basis for regional differences in macrophage activation in the tumor microenvironment
.
Genome Biol
2017
;
18
:
234
.
3.
Woroniecka
KI
,
Rhodin
KE
,
Chongsathidkiet
P
,
Keith
KA
,
Fecci
PE
.
T-cell dysfunction in glioblastoma: applying a new framework
.
Clin Cancer Res
2018
;
24
:
3792
802
.
4.
Wang
Q
,
Hu
B
,
Hu
X
,
Kim
H
,
Squatrito
M
,
Scarpace
L
, et al
.
Tumor evolution of glioma-intrinsic gene expression subtypes associates with immunological changes in the microenvironment
.
Cancer Cell
2017
;
32
:
42
56
.
5.
Petralia
F
,
Tignor
N
,
Reva
B
,
Koptyra
M
,
Chowdhury
S
,
Rykunov
D
, et al
.
Integrated proteogenomic characterization across major histological types of pediatric brain cancer
.
Cell
2020
;
183
:
1962
85
.
6.
Mackay
A
,
Burford
A
,
Molinari
V
,
Jones
DTW
,
Izquierdo
E
,
Brouwer-Visser
J
, et al
.
Molecular, pathological, radiological, and immune profiling of non-brainstem pediatric high-grade glioma from the HERBY Phase II randomized trial
.
Cancer Cell
.
2018
;
33
:
829
42.e5
.
7.
Sampson
JH
,
Gunn
MD
,
Fecci
PE
,
Ashley
DM
.
Brain immunology and immunotherapy in brain tumours
.
Nat Rev Cancer
2020
;
20
:
12
25
.
8.
Kipnis
J
,
Filiano
AJ
.
The central nervous system: privileged by immune connections
.
Nat Rev Immunol
2018
;
18
:
83
4
.
9.
Klemm
F
,
Maas
RR
,
Bowman
RL
,
Kornete
M
,
Soukup
K
,
Nassiri
S
, et al
.
Interrogation of the microenvironmental landscape in brain tumors reveals disease-specific alterations of immune cells
.
Cell
2020
;
181
:
1643
60
.
10.
Wei
J
,
Chen
P
,
Gupta
P
,
Ott
M
,
Zamler
D
,
Kassab
C
, et al
.
Immune biology of glioma-associated macrophages and microglia: functional and therapeutic implications
.
Neuro Oncol
2020
;
22
:
180
94
.
11.
Lieberman
NAP
,
DeGolier
K
,
Kovar
HM
,
Davis
A
,
Hoglund
V
,
Stevens
J
, et al
.
Characterization of the immune microenvironment of diffuse intrinsic pontine glioma: implications for development of immunotherapy
.
Neuro Oncol
2019
;
21
:
83
94
.
12.
Dejaegher
J
,
Solie
L
,
Hunin
Z
,
Sciot
R
,
Capper
D
,
Siewert
C
, et al
.
DNA methylation based glioblastoma subclassification is related to tumoral T-cell infiltration and patient survival
.
Neuro Oncol
2021
;
23
:
240
50
.
13.
Phillips
JJ
,
Gong
H
,
Chen
K
,
Joseph
NM
,
van Ziffle
J
,
Bastian
BC
, et al
.
The genetic landscape of anaplastic pleomorphic xanthoastrocytoma
.
Brain Pathol
2019
;
29
:
85
96
.
14.
Schindler
G
,
Capper
D
,
Meyer
J
,
Janzarik
W
,
Omran
H
,
Herold-Mende
C
, et al
.
Analysis of BRAF V600E mutation in 1,320 nervous system tumors reveals high mutation frequencies in pleomorphic xanthoastrocytoma, ganglioglioma and extra-cerebellar pilocytic astrocytoma
.
Acta Neuropathol
2011
;
121
:
397
405
.
15.
Pratt
D
,
Camelo-Piragua
S
,
McFadden
K
,
Leung
D
,
Mody
R
,
Chinnaiyan
A
, et al
.
BRAF activating mutations involving the β3-αC loop in V600E-negative anaplastic pleomorphic xanthoastrocytoma
.
Acta Neuropathol Commun
2018
;
6
:
24
.
16.
Phillips
JJ
,
Gong
H
,
Chen
K
,
Joseph
NM
,
van Ziffle
J
,
Jin
L-W
, et al
.
Activating NRF1-BRAF and ATG7-RAF1 fusions in anaplastic pleomorphic xanthoastrocytoma without BRAF p.V600E mutation
.
Acta Neuropathol
2016
;
132
:
757
60
.
17.
Vaubel
R
,
Zschernack
V
,
Tran
QT
,
Jenkins
S
,
Caron
A
,
Milosevic
D
, et al
.
Biology and grading of pleomorphic xanthoastrocytoma—what have we learned about it?
Brain Pathol
2021
;
31
:
20
32
.
18.
Kline
CN
,
Joseph
NM
,
Grenert
JP
,
van Ziffle
J
,
Talevich
E
,
Onodera
C
, et al
.
Targeted next-generation sequencing of pediatric neuro-oncology patients improves diagnosis, identifies pathogenic germline mutations, and directs targeted therapy
.
Neuro Oncol
2017
;
19
:
699
709
.
19.
Amaria
RN
,
Reddy
SM
,
Tawbi
HA
,
Davies
MA
,
Ross
MI
,
Glitza
IC
, et al
.
Neoadjuvant immune checkpoint blockade in high-risk resectable melanoma
.
Nat Med
2018
;
24
:
1649
54
.
20.
Collins
VP
,
Jones
DTW
,
Giannini
C
.
Pilocytic astrocytoma: pathology, molecular mechanisms and markers
.
Acta Neuropathol
2015
;
129
:
775
88
.
21.
Bolen
CR
,
McCord
R
,
Huet
S
,
Frampton
GM
,
Bourgon
R
,
Jardin
F
, et al
.
Mutation load and an effector T-cell gene signature may distinguish immunologically distinct and clinically relevant lymphoma subsets
.
Blood Adv
2017
;
1
:
1884
90
.
22.
Rooney
MS
,
Shukla
SA
,
Wu
CJ
,
Getz
G
,
Hacohen
N
.
Molecular and genetic properties of tumors associated with local immune cytolytic activity
.
Cell
2015
;
160
:
48
61
.
23.
Chaumeil
MM
,
Larson
PEZ
,
Yoshihara
HAI
,
Danforth
OM
,
Vigneron
DB
,
Nelson
SJ
, et al
.
Non-invasive in vivo assessment of IDH1 mutational status in glioma
.
Nat Commun
2013
;
4
:
2429
.
24.
Ohkawa
Y
,
Wade
A
,
Lindberg
OR
,
Chen
KY
,
Tran
VM
,
Brown
SJ
, et al
.
Heparan sulfate synthesized by Ext1 regulates receptor tyrosine kinase signaling and promotes resistance to EGFR inhibitors in GBM
.
Mol Cancer Res
2021
;
19
:
150
61
.
25.
Lohr
J
,
Ratliff
T
,
Huppertz
A
,
Ge
Y
,
Dictus
C
,
Ahmadi
R
, et al
.
Effector T-cell infiltration positively impacts survival of glioblastoma patients and is impaired by tumor-derived TGF
-.
Clin Cancer Res
2011
;
17
:
4296
308
.
26.
Ryall
S
,
Zapotocky
M
,
Fukuoka
K
,
Nobre
L
,
Guerreiro Stucklin
A
,
Bennett
J
, et al
.
Integrated molecular and clinical analysis of 1,000 pediatric low-grade gliomas
.
Cancer Cell
2020
;
37
:
569
83
.
27.
Liu
W
,
Putnam
AL
,
Xu-yu
Z
,
Szot
GL
,
Lee
MR
,
Zhu
S
, et al
.
CD127 expression inversely correlates with FoxP3 and suppressive function of human CD4+ T reg cells
.
J Exp Med
2006
;
203
:
1701
11
.
28.
Nagarsheth
N
,
Wicha
MS
,
Zou
W
.
Chemokines in the cancer microenvironment and their relevance in cancer immunotherapy
.
Nat Rev Immunol
2017
;
17
:
559
72
.
29.
Frederick
MJ
,
Henderson
Y
,
Xu
X
,
Deavers
MT
,
Sahin
AA
,
Wu
H
, et al
.
In vivo expression of the novel CXC chemokine BRAK in normal and cancerous human tissue
.
Am J Pathol
2000
;
156
:
1937
50
.
30.
Westrich
JA
,
Vermeer
DW
,
Silva
A
,
Bonney
S
,
Berger
JN
,
Cicchini
L
, et al
.
CXCL14 suppresses human papillomavirus-associated head and neck cancer through antigen-specific CD8+ T-cell responses by upregulating MHC-I expression
.
Oncogene
2019
;
38
:
7166
80
.
31.
Cicchini
L
,
Westrich
JA
,
Xu
T
,
Vermeer
DW
,
Berger
JN
,
Clambey
ET
, et al
.
Suppression of antitumor immune responses by human papillomavirus through epigenetic downregulation of CXCL14
.
mBio
2016
;
7
:
e00270
16
.
32.
Collins
PJ
,
McCully
ML
,
Martínez-Muñoz
L
,
Santiago
C
,
Wheeldon
J
,
Caucheteux
S
, et al
.
Epithelial chemokine CXCL14 synergizes with CXCL12 via allosteric modulation of CXCR4
.
FASEB J
2017
;
31
:
3084
97
.
33.
Tanegashima
K
,
Tsuji
K
,
Suzuki
K
,
Shigenaga
A
,
Otaka
A
,
Hara
T
.
Dimeric peptides of the C-terminal region of CXCL14 function as CXCL12 inhibitors
.
FEBS Lett
2013
;
587
:
3770
5
.
34.
Fricker
SP
,
Anastassov
V
,
Cox
J
,
Darkes
MC
,
Grujic
O
,
Idzan
SR
, et al
.
Characterization of the molecular pharmacology of AMD3100: a specific antagonist of the G-protein coupled chemokine receptor, CXCR4
.
Biochem Pharmacol
2006
;
72
:
588
96
.
35.
Augsten
M
,
Hägglöf
C
,
Olsson
E
,
Stolz
C
,
Tsagozis
P
,
Levchenko
T
, et al
.
CXCL14 is an autocrine growth factor for fibroblasts and acts as a multi-modal stimulator of prostate tumor growth
.
Proc Natl Acad Sci U S A
2009
;
106
:
3414
9
.
36.
Tang
K
,
Kurland
D
,
Vasudevaraja
V
,
Serrano
J
,
Delorenzo
M
,
Radmanesh
A
, et al
.
Exploring DNA methylation for prognosis and analyzing the tumor microenvironment in pleomorphic xanthoastrocytoma
.
J Neuropathol Exp Neurol
2020
;
79
:
880
90
.
37.
Zhao
J
,
Chen
AX
,
Gartrell
RD
,
Silverman
AM
,
Aparicio
L
,
Chu
T
, et al
.
Immune and genomic correlates of response to anti-PD-1 immunotherapy in glioblastoma
.
Nat Med
2019
;
25
:
462
9
.
38.
Arrieta
VA
,
Chen
AX
,
Kane
JR
,
Kang
SJ
,
Kassab
C
,
Dmello
C
, et al
.
ERK1/2 phosphorylation predicts survival following anti-PD-1 immunotherapy in recurrent glioblastoma
.
Nat Cancer
2021
;
2
:
1372
86
.
39.
Georgouli
M
,
Herraiz
C
,
Crosas-Molist
E
,
Fanshawe
B
,
Maiques
O
,
Perdrix
A
, et al
.
Regional activation of myosin II in cancer cells drives tumor progression via a secretory cross-talk with the immune microenvironment
.
Cell
2019
;
176
:
757
74
.
40.
Venkatesh
HS
,
Johung
TB
,
Caretti
V
,
Noll
A
,
Tang
Y
,
Nagaraja
S
, et al
.
Neuronal activity promotes glioma growth through neuroligin-3 secretion
.
Cell
2015
;
161
:
803
16
.
41.
Hromas
R
,
Broxmeyer
HE
,
Kim
C
,
Nakshatri
H
,
Christopherson
K
,
Azam
M
, et al
.
Cloning of BRAK, a novel divergent CXC chemokine preferentially expressed in normal versus malignant cells
.
Biochem Biophys Res Commun
1999
;
255
:
703
6
.
42.
Oler
G
,
Camacho
CP
,
Hojaij
FC
,
Michaluart
P
,
Riggins
GJ
,
Cerutti
JM
.
Gene expression profiling of papillary thyroid carcinoma identifies transcripts correlated with BRAF mutational status and lymph node metastasis
.
Clin Cancer Res
2008
;
14
:
4735
42
.
43.
Shurin
GV
,
Ferris
R
,
Tourkova
IL
,
Perez
L
,
Lokshin
A
,
Balkir
L
, et al
.
Loss of new chemokine CXCL14 in tumor tissue is associated with low infiltration by dendritic cells (DC), while restoration of human CXCL14 expression in tumor cells causes attraction of DC both in vitro and in vivo
.
J Immunol
2005
;
174
:
5490
8
.
44.
Perea
F
,
Bernal
M
,
Sánchez-Palencia
A
,
Carretero
J
,
Torres
C
,
Bayarri
C
, et al
.
The absence of HLA class I expression in non-small cell lung cancer correlates with the tumor tissue structure and the pattern of T cell infiltration: HLA class I expression in NSCLC
.
Int J Cancer
2017
;
140
:
888
99
.
45.
Cordon-Cardo
C
,
Fuks
Z
,
Drobnjak
M
,
Moreno
C
,
Eisenbach
L
,
Feldman
M
.
Expression of HLA-A,B,C antigens on primary and metastatic tumor cell populations of human carcinomas
.
Cancer Res
1991
;
51
:
6372
80
.
46.
McGranahan
N
,
Rosenthal
R
,
Hiley
CT
,
Rowan
AJ
,
Watkins
TBK
,
Wilson
GA
, et al
.
Allele-specific HLA loss and immune escape in lung cancer evolution
.
Cell
2017
;
171
:
1259
71
.
47.
Li
Z
,
Li
Y
,
Jiao
J
.
Neural progenitor cells mediated by H2A.Z.2 regulate microglial development via Cxcl14 in the embryonic brain
.
Proc Natl Acad Sci U S A
2019
;
116
:
24122
32
.
48.
Majzner
RG
,
Theruvath
JL
,
Nellan
A
,
Heitzeneder
S
,
Cui
Y
,
Mount
CW
, et al
.
CAR T cells targeting B7-H3, a pan-cancer antigen, demonstrate potent preclinical activity against pediatric solid tumors and brain tumors
.
Clin Cancer Res
2019
;
25
:
2560
74
.
49.
Haydar
D
,
Houke
H
,
Chiang
J
,
Yi
Z
,
Odé
Z
,
Caldwell
K
, et al
.
Cell-surface antigen profiling of pediatric brain tumors: B7-H3 is consistently expressed and can be targeted via local or systemic CAR T-cell delivery
.
Neuro Oncol
2021
;
23
:
999
1011
.

Supplementary data