Immune deconvolution and temporal mapping identifies stromal targets and developmental intervals for abrogating murine low-grade optic glioma formation.

BACKGROUND: Brain tumor formation and progression are dictated by cooperative interactions between neoplastic and non-neoplastic cells. This stromal dependence is nicely illustrated by tumors arising in the Neurofibromatosis type 1 (NF1) cancer predisposition syndrome, where children develop low-grade optic pathway gliomas (OPGs). Using several authenticated Nf1-OPG murine models, we previously demonstrated that murine Nf1-OPG growth is regulated by T cell function and microglia Ccl5 production, such that their inhibition reduces tumor proliferation in vivo. While these interactions are critical for established Nf1-OPG tumor growth, their importance in tumor formation has not been explored.
METHODS: A combination of bulk and single-cell RNA mouse optic nerve sequencing, immunohistochemistry, T cell assays, and pharmacologic and antibody-mediated inhibition methods were used in these experiments.
RESULTS: We show that T cells and microglia are the main non-neoplastic immune cell populations in both murine and human LGGs. Moreover, we demonstrate that CD8+ T cells, the predominant LGG-infiltrating lymphocyte population, are selectively recruited through increased Ccl2 receptor (Ccr4) expression in CD8+, but not CD4+, T cells, in a NF1/RAS-dependent manner. Finally, we identify the times during gliomagenesis when microglia Ccl5 production (3-6 weeks of age) and Ccl2-mediated T cell infiltration (7-10 weeks of age) occur, such that temporally-restricted Ccl2 or Ccl5 inhibition abrogates tumor formation >3.5 months following the cessation of treatment.
CONCLUSIONS: Collectively, these findings provide proof-of-concept demonstrations that targeting stromal support during early gliomagenesis durably blocks murine LGG formation.

Selective intratumoral CD8+ T cell infiltration reflects increased Ccr4 expression

Optic glioma microglia Ccl5 production precedes Ccl2-mediated T cell infiltration

Temporally-restricted Ccl2 or Ccl5 inhibition durably abrogates gliomagenesis

While prior studies have characterized the tumor microenvironment in pediatric low-grade brain tumors, we leveraged authenticated murine models of neurofibromatosis type 1 (Nf1) low-grade optic glioma to identify tumor-specific stromal cells, as well as the temporal course of microglia expansion, T cell infiltration, and paracrine factor expression during tumorigenesis. By targeting the immune axis during defined periods of tumor development, we blocked Nf1 optic glioma formation. These findings represent a first proof-of-principle demonstration that disrupting paracrine factor circuitry during gliomagenesis is sufficient to abrogate tumor formation, suggesting future more refined chemoprevention strategies for at-risk patient groups.

Low-grade gliomas (LGGs), including pilocytic astrocytomas (PAs), are the most common pediatric brain tumors, and while associated with good overall survival, frequently cause life-long co-morbidities[1]. In contrast to other LGGs, PAs typically have low proliferative indices, lack mitotic figures, and do not progress to malignant glioma. PAs are caused by mutations that increase MAPK/ERK signaling, either from BRAF gene rearrangement (KIAA1549:BRAF fusion) or biallelic inactivation of the Neurofibromatosis type 1 (NF1) tumor suppressor gene[2]. This latter group of PAs typically arises within the optic pathway (optic pathway gliomas; OPGs) of children with NF1, often resulting in vision loss[3].

While both sporadic and NF1-PAs exhibit increased ERK signaling, this is usually amplified in the setting of mitogens supplied by non-neoplastic cells in the tumor microenvironment. As such, human PA cell lines are difficult to grow in vitro in the absence of growth factors[4] and frequently undergo oncogene-induced senescence[5-7], suggesting that the LGG cells are highly dependent upon growth/survival factors elaborated by non-neoplastic stromal cells. In this regard, as 50% of the cells in PAs are monocytes (yolk sac-derived resident brain microglia and bone marrow-derived macrophages) and, to a lesser extent, T cells[2,8]. Importantly, using authenticated murine models of Nf1 optic glioma[9-11], we have previously shown that T cells and microglia contribute to established tumor growth[12,13], culminating in microglial elaboration of Ccl5, a paracrine factor critical for glioma cell expansion[12-14].

To define the role of non-neoplastic cells during tumor development, we took advantage of a murine Nf1-OPG model with high tumor penetrance (>95%)[10] and a well-characterized time course of tumor evolution[15]. In this report, we leveraged bulk RNA sequence deconvolution, single-cell RNA sequencing (scRNAseq), and immunohistochemical analyses to define the non-neoplastic (stromal) cellular composition and its temporal development, which were then used to identify targetable paracrine factors and time windows for tumor prevention. Employing this strategy, we show that inhibition of T cell infiltration and microglial function abrogates Nf1-OPG formation.

Materials and Methods

Mice

Wild type (WT), Nf1flox/flox (equivalent to WT)[9], Nf1+/− [16], Nf1+/R681X (R681X[17]), Nf1+/R1278P (R1278P[11]), Nf1-OPG (Nf1+/− mice with somatic Nf1 gene inactivation in neuroglial progenitors at E15.5), and Nf1flox/R681X; GFAP-Cre (Nf1-R681X-OPG; [17]) mice were housed in a pathogen-free barrier facility with controlled light-dark cycles (12:12 h) and ad libitum access to food and water, and used in accordance with an approved Animal Studies protocol. Nf1flox/flox and WT mice were used interchangeably, as they are identical with respect to all parameters measured.

In vivo Mouse Treatments

Two hundred and fifty micrograms of anti-Ccl5 (R&D Systems MAB478) or anti-IgG2a control antibodies (R&D Systems) were intraperitoneally injected into Nf1-OPG mice daily from three to six weeks of age (WOA). In addition, seven-week-old Nf1-OPG mice were injected intraperitoneally five days per week for three consecutive weeks with either 2 mg/kg anti-Ccl2 or control IgG antibodies (BioXCell, BE0185, and BE0091, respectively). Mice were analyzed at 12 and 24 WOA. PLX33397 (PLX) treatments of Nf1-OPG mice involved replacing normal chow (AIN-76A rodent diet; Research Diet Inc.) with 275 mg/kg PLX3397-containing chow (Free Base) from four to seven WOA. PLX-treated (n = 7) or control chow (n = 8) mouse optic nerves were collected and analyzed at 24 WOA.

RNA Bulk Sequencing, Analysis, and Compilation of CNSortR

Analyses was performed using Partek Flow software, version 9.0[18]. Bulk RNA-seq reads from FAC-sorted CNS cell types[19] were aligned to the Ensembl release 95 v2 top-level assembly with STAR version 2.7.3a[20]. Sequencing performance was assessed for the total number of aligned reads, total number of uniquely aligned reads, and features detected. Gene counts and isoform expression were normalized by quantile normalization to be consistent with the previously-compiled ImmuCC reference file containing gene counts for various immune cell types[21]. Gene-specific analysis was performed using the lognormal with limma-trend shrinkage model[22], where the results were filtered by P-values and fold-changes, to identify the top 20 most significant marker genes for each population. Filtered gene counts were amended to the existing ImmuCC reference file and uploaded into CIBERSORTx[23]. Bulk RNA-seq data from two distinct models of murine Nf1-OPG (GSE102345) were analyzed and deconvoluted using CNSortR in CIBERSORTx (Supplementary Figure S1A). Nf1-OPG, as well as human NF1-PAs (GSE163071), were analyzed using TIMER2.0[24], which includes CIBERSORTx. CCL2 and CCL5 expression in human PA was determined from GSE163071 and GSE44971.

Single Cell RNA Sequencing (scRNAseq)

Nf1-OPG optic nerves were collected from eight 12-week-old mice and pooled. Mice were perfused using D-PBS with calcium, magnesium, glucose, pyruvate (ThermoFisher), and the adult brain dissociation kit (Miltenyi Biotec 130-107-677) was used to prepare single-cell suspensions. Single-cell barcoding, cDNA amplification and Gene Expression Library Construction were then performed using Chromium Single-Cell 3′ Reagent Kits v3, and10x Genomics® was used for sequencing. Nf1-OPG single-cell RNA sequencing data were analyzed in Partek Flow. Sequencing reads were aligned using STAR to the Ensembl annotation model. Cell quality was assessed by total reads, expressed gene count, and mitochondrial read percent. CPM normalization and PCA analysis was conducted followed by graph-based clustering using 20 principal components and the Louvain clustering algorithm. Clustering yielded 12 clusters with differentially expressed genes which were further visualized by UMAP plots generated using Euclidian distance.

Human PA Samples

De-identified human NF1-associated and sporadic PA specimens (tissue microarray) were obtained from Dr. Sonika Dahiya under an approved human studies protocol. Immunohistochemistry was performed on adjacent paraffin sections using antibodies listed in Supplementary Table S1, and detected using the Vectastain ABC kit (Vector Laboratories, Burlingame).

Immunohistochemistry

WT spleens were fixed in 4% paraformaldehyde (PFA) in 0.1M sodium phosphate buffer (pH 7.4) overnight. Following transcardial perfusion with 4% PFA, optic nerves/chiasms were dissected and post-fixed in 4% PFA. Slides were deparaffinized in xylene, subjected to antigen retrieval, and incubated overnight with the appropriate antibodies (Supplementary Table S1), followed by incubation with biotinylated secondary antibodies (1:200; Vector Laboratories, Burlingame, CA). Signals were amplified with the Vectastain ABC kit (PK4000, Vector Laboratories), developed using the Vectastain DAB kit (SK4100, Vector laboratories), and counterstained with hematoxylin. Specimens were scanned using a NanoZoomer 2.0-HT slide scanner (Hamamatsu Photonics K.K., Japan) with a 20x objective, and the percentage of immunopositive cells calculated by manual counting. Optic nerve sections were deparaffinized, hydrated and stained in toluidine blue (1% toluidine blue in 70% ethanol; pH 2.3) for 3 minutes, washed in distilled water, dehydrated, and mounted. Stained optic nerves were imaged for quantification on a Leica ICC50W microscope using LAS EZ software.

Immunofluorescence

Nf1-OPG optic nerves sections were prepared as above and incubated with appropriate primary and Alexa Fluor secondary antibodies (Supplementary Table S1).

ELISA

Active RAS (RAS-GTP; Fisher Scientific, 50673628), Ccl2 (Fisher Scientific, MJE00B), and Ccl5 (Fisher Scientific, MMR00) ELISAs were performed per manufacturers’ instructions.

RNA Extraction, cDNA Production, and qPCR

RNA from mouse optic nerves and spleens was isolated using the NucleoSpin® RNA Plus kit (NC1498148, Takara), reverse transcribed into cDNA using the Applied Biosystems High-Capacity cDNA Reverse Transcription Kit (Thermo- 4374967), and subjected to TaqMan real-time quantitative PCR (RT-qPCR) (Supplementary Table S2). ΔΔCT values were calculated using Gapdh expression as an internal control.

T Cell Isolation

Mouse spleens were homogenized into single-cell suspensions in PBS containing 0.1% BSA and 0.6% Na-citrate, followed by 15 min incubation with 120 Kunitz units of DNase I, and erythrocyte lysis (eBioscience, 00433357). Cells were then filtered through a 30-µM cell strainer. CD8+ and CD4+ T cells were isolated using CD8a (Miltenyi Biotec, 130-104-075) and CD4 (Miltenyi Biotec, 130-117-043) T cell isolation kits, respectively. T cells were grown at 2.5 x 106 cells/ml in RPMI-1640 medium supplemented with 10% FBS and 1% P/S. After 24h, Nf1+/- CD8+ T cells were treated for 24 h with 10 µM Pan-RAS-Inhibitor-IN-1 (IN-1[HY-101295, MedChem Express]) or 10 µM Ccr4 inhibitor (AZD2098, Selleckchem). All experiments were performed with n = 3.

Migration Assays

About 5 × 105 T cells were placed in the upper transwell chambers (Corning, 6.5 mm insert, CLS3398-2EA) containing 200 µl serum-free RPMI 1640 media. 500 µl of chemoattractant media containing Ccl2 (20ng/ml, 479-JE-050/CF-R&D systems) or Ccl12 (25ng/ml, 428-P5-025-R&D systems) was added to the lower chamber[11], and the number of T cells in the lower chambers counted by microscopy at 20x after 6 h.

Proliferation Assay

T cell proliferation was assessed using the BrdU Cell Proliferation ELISA Kit (Ab126556, Abcam) following the manufacturer’s instructions. Briefly, 5 × 105 cells were plated in a 96-well plate and incubated for 72 h prior to the addition of BrdU for 24 h. Cells were then fixed, permeabilized, and subjected to DNA denaturation. Following incubation with anti-BrdU antibodies for 4 h, the cells were washed and incubated with peroxidase-conjugated secondary antibody, and analyzed using a colorimetric assay at 450nm.

Statistical Analyses

Data analyses were performed using GraphPad Prism using unpaired two-tailed Student’s t-tests (differences between two groups) and one-way or two-way analysis of variance (ANOVA) with Dunnett’s or Tukey’s multiple comparisons test, with significance set at p<0.05. Data are presented as mean values ± SEM. All experiments were performed at least thrice with similar results, and statistical analyses verified by a bioinformatician (O.C.).

Results

The Nf1-OPG Microenvironment Contains Mostly T Cells and Microglia

To determine the non-neoplastic cell composition of murine Nf1-OPGs, we analyzed bulk tumor RNAseq data from two models of Nf1-OPG (Nf1-OPG, Figure 1A, left; Nf1-R681X-OPG; Figure 1A, right; GSE102345) at 12 WOA when >95% of the mice harbor tumors. Using TIMER 2.0, we found B cells, plasma cells, dendritic cells, NK cells, and mast cells, in addition to monocytes and T cells (CD8+ and CD4+ cells). Similarly, human PAs (NF1-associated and sporadic) contained B cells, plasma cells, macrophages, T cells, NK cells, dendritic cells, neutrophils, mast cells, and eosinophils (Supplementary Figure S1B).

Figure 1.

The Nf1-OPG microenvironment contains largely T cells and microglia. A, Using TIMER 2.0, Nf1-OPGs and Nf1-R618X-OPGs contain B cells (B), macrophages and monocytes (Mɵ), CD8+ T cells (CD8), natural killer cells (NK), CD4+ T cells (CD4), dendritic cells (DC), neutrophils (N), mast cells (MC) and eosinophils (Eo) (n = 5 each). B, Using CNSortR, Nf1-OPGs and Nf1-R618X-OPGs contained mostly astrocytes (A) and oligodendrocytes (O), but also neurons (Ne), microglia (Mg), macrophages and monocytes, CD8+ T cells, CD4+ T cells, NK cells, plasma cells (PC), B cells, DCs, mast cells, endothelial cells (En), neutrophils, and eosinophils. C, Using scRNAseq, most populations matched those identified using CNSortR. D, No mast cells (toluidine blue), B cells (CD138), dendritic cells (dendritic cell receptor-2), or plasma cells (CD94) were detected in the optic nerves of control (CTL) or Nf1-OPG mice. E, Increased microglia (Iba1+ cells), endothelial cells (VCAM1+ cells), astrocytes (GFAP+ cells), and oligodendrocytes/oligodendrocyte precursor cells (Olig2+ cells), as well as greater CD8+ and CD4+ T cell content, were detected in the optic nerves of Nf1-OPG mice relative to controls (CTL). F, Nearly all Iba1+ cells were Tmem119+ cells (microglia) at all ages examined. Scale bar, 50 µm.

Since TIMER was constructed using RNAseq data from peripheral immune cells, we amended ImmuCC with data from the brainrnaseq.org[19] to provide a better representation of the brain populations in murine Nf1-OPGs (CNSortR v.1.0). With the inclusion of neural cell types, we found predominantly oligodendrocytes and astrocytes (GSE102345, Figure 1B).

Next, scRNAseq of optic nerves from Nf1-OPG mice (n = 8) at 12 WOA identified 12 cell populations (Figure 1C, upper), including endothelial cells, mast cells, neutrophils, B cells, T cells, microglia, oligodendrocytes, and brain cells (astrocytes, neurons). To compare these to those detected by deconvolution analysis, we used the top 20 differentially expressed genes from each of the scRNAseq clusters, and matched them with the genes expressed in the CNSortR populations (Figure 1C). While most populations aligned well, with minor differences, all methods indicated that Nf1-OPGs contain mast cells, plasma cells, B cells, dendritic cells, and natural killer (NK) cells.

To determine whether these newly-identified cells were actually present in Nf1-OPGs, we performed immunostaining of optic nerves from CTL and Nf1-OPG mice at 12 WOA using established antibody markers for each population. Contrary to the bulk RNA-seq deconvolution and scRNAseq results, no mast cells (Toluidine blue-positive), B cells (CD19+), dendritic cells (Dendritic Cell Inhibitory Receptor 2+), plasma cells (CD138+) or NK cells (CD94+) were detected in any of the murine optic nerves (Figure 1D), in normal human brain (n = 26) or in patient PA tissue specimens (n = 67; Supplementary Figure S1C). In contrast, we found that Nf1-OPG mouse tumors contained monocytes (Iba1+ cells), endothelial cells (VCAM1+ cells), astrocytes (GFAP+ cells), and oligodendrocyte precursors/oligodendrocytes (Olig2+ cells), as well as CD8+ and CD4+ T cells (Figure 1E). We also observed that the T cells identified in murine Nf1-OPG tumors at 12 WOA were sometimes, but not always, adjacent to VCAM-1+ blood vessels (Supplementary Figure S1D).

Since RNAseq data from peripheral immune cells may not accurately identify the analogous populations in the brain, we hypothesized that the RNAseq markers defining those populations (BAFF-R, B cells; TPSB2, mast cells) might be expressed by another population. As such, BAFF-R is largely expressed by Iba1+ monocytes (macrophages/microglia) in the Nf1-OPG optic nerves, whereas TPSB2 was mostly found in GFAP+ cells (astrocytes) (Supplementary Figure S1E). Since Iba1 is not a specific marker for microglia, and also identifies other monocyte populations, we sought to determine what proportion of the Iba1+ cells were microglia. Using Tmem119 as a marker of microglia[25], nearly all of the Iba1+ cells were Tmem119+ (Figure 1F).

Murine Nf1-OPG Growth Begins After 6 Weeks of Age

Since the majority of optic nerve myelination occurs postnatally[26], we assessed proliferation in the optic nerves of WT, Nf1+/–, and Nf1-OPG mice at 3, 4.5, 6, 9, and 12 WOA (Figure 2A). While both non-tumor-forming (WT, Nf1+/–) and tumor-forming (Nf1-OPG) mouse optic nerves had high levels of proliferation at 3 WOA, this increased proliferation was reduced in non-tumor-forming mice by 6 WOA. Conversely, tumor-forming (Nf1-OPG) optic nerves increased their proliferation over time. As expected, almost all proliferating cells at 3 and 6 WOA in all mice were Olig2+ oligodendroglial lineage cells (Figure 2B), demonstrating that Nf1-OPG development begins after 6 WOA when normal oligodendrocyte precursor cell proliferation declines to 20% (Figure 2C). At 12 weeks of age, the majority of the proliferating cells are CD133+ and ABCG1+ glioma cells[27,28] (Figure 2C), but not CD3+ T cells or Tmem119+ microglia.

Figure 2.

Optic glioma proliferation begins after 6 weeks of age. A, No difference in proliferation (%Ki67+ cells) in the optic nerves of CTL (n = 6), Nf1+/– (n = 6), and Nf1-OPG (n = 5) were noted at 3, 4.5, or 6 WOA. At 9 and 12 WOA, Nf1-OPG mice have more Ki67+ cells compared to CTL and Nf1+/– mice (P = 0.006 and P < 0.001, respectively). B, At 3 (n = 7 per group) and 6 (n = 6 per group) WOA, most Ki67+ cells were Olig2+ cells in all optic nerves. C, At 12 WOA, the majority of the Ki67+ cells in the murine Nf1-OPGs were Olig2+ cells or CD133+ and ABCG1+ glioma cells.

Microglia and T Cells Exhibit Different Patterns of Temporal Evolution

To define the kinetics of microglia expansion and Ccl5 production, we analyzed optic nerves from control (WT or Nf1flox/flox), Nf1+/–, and Nf1-OPG mice beginning at 3 WOA, before tumors developed[15], as well as at 4.5, 6, and 9 (tumor evolution) and 12 (fully formed tumors) WOA. While Tmem119+ microglia content was similar in all groups at 3 and 4.5 WOA, Tmem119+ microglia increased in Nf1-OPG mice at 6 WOA relative to control (WT) and Nf1+/- mice (Figure 3A). Importantly, Iba1+ (Supplementary Figure S2A) and CD11b+ microglia (Supplementary Figure S2B) exhibit similar patterns to the Tmem119+ microglia profiles (Figure 3B). At 6 WOA, in parallel with increased microglia content, Ccl5 expression increased (Figure 3C) in Nf1-OPG relative to WT mice. Relevant to human LGGs, CCL5 RNA expression was increased in sporadic and NF1-associated pediatric PAs relative to non-neoplastic brain in two independent GEO datasets (Figure 3D–E).

Figure 3.

Nf1-OPG microglia content and Ccl5 production increase by 6 weeks of age. A, Similar Tmem119+ cell content in the optic nerves of all mice were noted at 3 WOA. Nf1-OPG mice (n = 5) have more Tmem119+ cells at 6 and 9 weeks compared to both Nf1+/– and CTL mice (both n = 6, P = 0.0029). At 12 WOA, CTL mice have the lowest density of microglia. Nf1-OPG (P < 0.0001) and Nf1+/– (P = 0.006) optic nerves have more microglia relative to controls. B, No differences in Iba1+, Tmem119+ and CD11b+ cell content were observed between control and Nf1-OPG mice. C, Higher Ccl5 levels were detected in the optic nerves of 6- (P < 0.0001), 9- (P < 0.0001) and 12- (P < 0.0001) week-old Nf1-OPG mice, but not in 3-week-old mice, relative to controls (n = 3 all). Human PAs have increased CCL5 expression by D, RNAseq and E, microarray relative to non-neoplastic brain (CTL). Scale bar, 50 µm.

Next, we analyzed the temporal dynamics of optic nerve T cell infiltration in WT, Nf1+/–, and Nf1-OPG mice at 3, 4.5, 6, 9, and 12 WOA (Figure 4A). CD3+ T cell content begins to rise exclusively in Nf1-OPG mice after 9 WOA. Since we had previously shown that Ccl2 mediates T cell infiltration into the optic nerves of Nf1-OPG mice[11], we quantified Ccl2 expression as a function of tumor evolution, and found that Ccl2 expression increases after 6 WOA in Nf1-OPG mice (Figure 4B). Additionally, CCL2 expression was increased in sporadic and NF1-associated pediatric PAs relative to non-neoplastic brain in two independent GEO datasets (Figure 4C–D).

Figure 4.

Nf1-OPG T cells and Ccl2 levels rise after 6 weeks of age. A, Nf1-OPG (n = 5) CD3+ T cells infiltrate after 9 WOA (P < 0.0001). There was no difference in CD3+ T cells between CTL and Nf1-OPG mice (both n = 6) at 3, 4.5, 6, and 9 WOA. Nf1+/– mice have similar CD3+ T cell content as controls at 3, 4.5, 6, 9, and 12 WOA. B, Ccl2 levels are similar to CTL mice at 3, 4.5, 6 and 9 WOA, but increase by 12 WOA in Nf1-OPG mice compared to controls (n = 3 all, P < 0.0001). Human PAs have increased CCL2 expression by C, RNAseq and D, microarray relative to non-neoplastic brain (CTL). Scale bar, 50 µm.

CD8+T Cell Predominance is Governed by NF1/RAS Regulation of Ccr4 Expression

The majority of the tumor-infiltrating T cells were CD8+ T lymphocytes (Figure 5A), as previously reported in mice[13] and in human NF1-LGGs[8]. In contrast, the number of CD4+ T cells was similar in tumor-bearing (Nf1-OPG) and nontumor-bearing (WT, Nf1+/–) optic nerves from 3 to 12 WOA (Figure 5B). Importantly, the sum of the CD4+ and CD8+ T cells in the optic nerves equaled the total number of CD3+ T cells (Supplementary Figure S3A), suggesting that these are the major T cell populations in murine Nf1-OPGs. No regulatory (Foxp3+), helper (CD25+), or exhausted T (PD-1+ or Lag3+) cells were detected in these tumors at any time point (Supplementary Figure S3B).

Figure 5.

CD8+T cell predominance is governed by NF1/RAS regulation of Ccr4 expression. A, CD8+ T cells content in Nf1-OPG optic nerves (n = 10) is similar to controls (n = 8) at 3, 4.5, 6, and 9 WOA, but increases by 12 WOA (P < 0.0001). Nf1+/– mice (n = 6) have similar numbers of CD8+ T cells as controls from 3 to 12 WOA. B, CD4+ T cell content does not change in Nf1-OPG optic nerves relative to controls at 3, 4.5, 6, 9, and 12 WOA. C, Ccl2 did not increase CD4+ T cell migration in WT or Nf1+/– mice. Ccl12 or medium (control) does not increase CD4+ T cell migration. Ccl2 increased Nf1+/– CD8+ T cell migration relative to WT controls (P < 0.0001). Nf1+/- and WT CD8+ T cells treated with medium (Control) or Ccl12 exhibit similar migration. D, Nf1+/- and WT CD4+ T cells had similar levels of Ccr4 expression. Nf1+/– CD8+ T cells have increased Ccr4 expression relative to their WT counterparts (P = 0.0474). E, Ccr4 inhibition (AZD2098; 15µM) in CD8+ T cells decreases Ccl2-induced, but not Ccl12-induced, migration (P = 0.0063). F, Nf1+/– and WT CD4+ T cells had similar levels of Ras activity. Nf1+/– CD8+ T cells have increased Ras activity relative to their WT counterparts (P = 0.0363) G, Inhibition of Ras activity (10µM IN-1) reduces Ccl2-induced,but not Ccl12-induced, CD8+ T cell migration (P = 0.0104). H, RAS activity inhibition (10µM IN-1) decreases Ccr4 expression in Nf1+/- CD8+ T cells (P = 0.0069). I, Proposed model of Ccl2 signaling through Ccr4 in Nf1+/– CD8+ T cells.

To determine the etiologic basis for the selective recruitment of CD8+ T cells, we examined directed migration of CD4+ and CD8+ T cells to Ccl2 and Ccl12, chemokines secreted by neoplastic cells to mediate T cell migration[11]. While Nf1+/– and WT CD4+ T cell migration to Ccl2 and Ccl12 were similar, Nf1+/– CD8+ T cells had greater migration to Ccl2, but not Ccl12, than their WT counterparts (Figure 5C). One plausible explanation for the increased directed migration of CD8+ T cells to Ccl2 is differential expression of Ccl2 receptors (Ccr2 and Ccr4). Using RT-qPCR, we detected increased Ccr4 expression in Nf1+/- CD8+ T cells relative to WT CD8+, CD4+ T cells, and Nf1+/– CD4+ T cells (Figure 5D), without any change in Ccr2 expression (Supplementary Figure S3C). Similar increases in Ccr4 expression were observed in CD8+ T cells from other Nf1-OPG strains (R681X and R1278P) relative to their WT counterparts (Supplementary Figure S3D). Consistent with Ccr4 as the receptor responsible for increased Nf1+/– CD8+ T cell Ccl2-directed migration, treatment of Nf1+/– CD8+ T cells with a selective Ccr4 inhibitor (AZD2098) reduced the migration of Nf1+/– CD8+ T cells to Ccl2, but not Ccl12 (Figure 5E).

One of the consequences of increased Ccr4 activation in Nf1+/– CD8+ T cells is elevated RAS activity. To this end, WT CD8+ T cells have higher levels of RAS activity than WT CD4+ cells, which is further increased in the setting of an Nf1 mutation (Nf1+/– T cells; Figure 5F). Consistent with NF1/RAS regulation of migration, inhibition of RAS activity (IN-1) decreased CD8+ T cell migration to Ccl2, but not Ccl12, with no effects on CD4+ T cell migration (Figure 5G). Since the increase in Ccr4 expression was larger than the increase in RAS activity detected in CD8+ T cells, we sought to determine whether RAS activation regulates Ccr4 expression. Consistent with a feed-forward loop where Nf1 mutation leads to greater RAS activity and Ccr4 expression, IN-1 treatment reduced Ccr4 expression in Nf1+/– CD8+ T cells (Figure 5H-I). In addition to increased Ccl2-mediated chemotaxis, Nf1+/– CD8+ T cells exhibited greater proliferation than their WT counterparts (Supplementary Figure S3E).

Interrupting T Cell and Microglia Function During Tumor Formation Blocks Nf1-OPG Progression

Leveraging the unique temporal patterns of microglia and T cell infiltration, we performed three proof-of-principle experiments to determine whether blocking these events during the specific time intervals would abrogate glioma progression. Since microglia content and Ccl5 production begin to increase after 3 WOA, we first treated Nf1-OPG mice with the PLX3397 inhibitor from 4-7 WOA to deplete microglia, and analyzed the optic nerves at 6 months of age. While there were fewer microglia and CD3+ T cells (Supplementary Figure S4), optic nerve proliferation (Ki67+ cells) was unchanged, suggesting that depleting microglia was insufficient to block tumor development.

Second, Nf1-OPG mice were treated with anti-Ccl5 antibodies from 3 to 6 WOA, when Ccl5 expression rises, and aged to 12 WOA when Nf1-OPG become evident. Compared to control antibody-treated mice, Nf1-OPG mice treated with anti-Ccl5 antibodies had reduced proliferation, without changes in microglia content (Figure 6A) or optic nerve volume (Figure 6B). Importantly, the lack of tumor proliferation was associated with reduced Ccl2 expression (Figure 6C) and Ccl2-mediated T cell infiltration (Figure 6A). After aging a separate cohort of anti-Ccl5 antibody-treated mice to 6 months of age, the decrease in proliferation and infiltration of T cells persisted (Figure 6D).

Figure 6.

Interrupting microglia or T cell function during tumor evolution blocks Nf1-OPG progression. A, Ccl5 neutralizing antibody (n = 5) treatment (3 to 6 WOA) results in decreased Nf1-OPG proliferation (%Ki67+ cells; P = 0.0012) and CD3+ T cells content relative to anti-IgG2a isotype controls (n = 6, P = 0.0174). No difference in microglia content was found between both groups. B, No difference in the volumes of the optic nerves were observed between anti-Ccl5- and anti-IgG2a-treated mice. C, Nf1-OPG mice treated with anti-Ccl5 neutralizing antibodies had reduced Ccl2 expression (P = 0.0004) relative to isotype control-treated mice (both n = 3). D, At 24 WOA, anti-Ccl5-treated mice (3 to 6 WOA) had decreased proliferation (%Ki67+ cells; P = 0.0002) and CD3+ T cell content (P = 0.0393) compared to anti-IgG-treated mice. E, Anti-Ccl2 neutralizing antibody treatment (n = 8, 7 to 10 WOA) resulted in decreased proliferation (%Ki67+ cells; P = 0.0109) and decreased CD3+ T cell infiltration (P = 0.0065). No difference in microglia content was seen between anti-Ccl2- and anti-IgG-treated (n = 7) mice. F, No differences in optic nerve volumes were observed between both groups. G, Anti-Ccl2 antibody treatment reduced optic nerve Ccl5 levels (P = 0.0022) compared to IgG-treated mice. H, At 24 WOA, anti-Ccl2-treated mice (7 to 10 WOA) had decreased proliferation (%Ki67+ cells; P = 0.0014) and CD3+ T cell content (P = 0.0443) relative to IgG-treated mice.

Third, we treated Nf1-OPG mice with anti-Ccl2 antibodies from 7 to 10 WOA, when Ccl2 expression and Ccl2-mediated CD8+ T cell infiltration increase. Relative to mice receiving control antibodies, Nf1-OPG mice treated with anti-Ccl2 antibodies had reduced tumor proliferation (Ki67+), CD3+ T cell content, and Ccl5 production, with no change in microglia content or optic nerve volume (Figure 6E–G). The reduced proliferation and T cell infiltration seen at 12 WOA remained when a separate cohort of mice were analyzed at 6 months of age (Figure 6H).

Discussion

Stromal dependence is a hallmark of most solid cancers, including brain tumors, where its importance to tumor formation and progression has been mainly explored in malignant gliomas[29-39]. In the current study, we employed Nf1-mutant mouse strains to characterize the LGG immune microenvironment composition, define the temporal sequence of tumor evolution, and use this information to identify therapeutic windows to block glioma progression. Herein, we show that durable inhibition of gliomagenesis is achieved by interrupting Ccl2-mediated T cell infiltration or microglia production of Ccl5. Moreover, we demonstrate that selective CD8+ T cell enrichment in these tumors results from NF1/RAS-regulated Ccr4 expression. Collectively, these findings raise three major points relative to LGG pathobiology.

First, we show that disrupting stromal cell function during critical periods of tumor formation is sufficient to block tumorigenesis. In this manner, we show that anti-Ccl2 antibody-mediated interruption of T cell infiltration and anti-Ccl5 antibody-mediated abrogation of glioma cell microenvironmental support reduces Nf1-OPG proliferation, one key hallmark of these tumors in children. However, microglia content and optic nerve volumes are not restored to WT levels. We interpret this result to indicate that disrupting the stromal axis during tumor evolution blocks Nf1-OPG progression, but does not abrogate glioma initiation. In this manner, the events required for tumor initiation have already occurred, such that these treatments merely block further evolution of the tumor, ameliorating its growth (%Ki67+ cells), but not the microglia infiltration or optic nerve volume. To this end, observations from our laboratory have revealed that Nf1+/- mice have increased optic nerve volumes and microglia content, similar to some children with NF1[40], arguing that these stroma-directed treatments do not completely normalize the NF1-associated optic nerve pathology. In contrast, we have previously shown that Nf1-OPG initiation is dictated by retinal ganglion cell neuronal activity-dependent production of neuroligin-3 (NGLN3)[41]. Taken together, we envision gliomagenesis involves at least three separable epochs: (1) initiation, reflecting the impact of Nf1 mutation on the cells of origin for these tumors[42] or on neuronal activity-regulated Nlgn3 production, (2) propagation, which entails T cell recruitment and elaboration of microglia-produced gliomagens (e.g., Ccl5), and/or (3) maintenance involving some of the same signals and cell types.

Second, prior studies in Nf1-OPG mice[11,13] and NF1 patient gliomas[8] revealed a striking predominance of CD8+ T cells. We previously demonstrated that neoplastic cell-produced Ccl2 (which signals through both Ccr2 and CCr4 receptors) and Ccl12 (which signals through Ccr2 receptor) are responsible for T cell-directed migration[11]. Surprisingly, Nf1 mutation increases RAS activation, proliferation, and Ccl2-mediated migration more dramatically in CD8+ T cells, where it selectively increases Ccr4 RNA expression in a RAS-dependent manner. Moreover, the increased Ccr4 expression is likely responsible for the increased CD8+ T cell content observed in NF1-mutant tumors, as further supported by increased Ccr4 expression in CD8+ T cells from mice harboring different germline Nf1 gene mutations (Supplementary Figure S3D) and inhibition of T cell motility in response to Ccl2. Additional studies are underway to determine why RAS is selectively hyperactivated in Nf1+/– CD8+ T cells, and how this RAS/Ccr4 feedforward loop is maintained.

While CD8+ T cells are increased in the tumors and are critical for glioma proliferation[13], it is not clear whether they are exhausted[43, 44] or dysfunctional[45] CD8+ T cells or a unique population of T cells that promote microglia support of tumor cell growth. Analysis of these lymphocytes reveals that they do not express CD94 (NK cells), PD-1 (exhausted T cells) or Foxp3/CD25 (T helper cells). While further characterization of Nf1-OPG CD8+ T cells is necessary, a population of innate CD8+ T cells that do not recognize classic peptide antigens has been reported to promote tumorigenesis[46].

Third, these findings support the need for further research into stroma-directed therapies. While microglia inhibitors, like PLX3397 (CSF1R targeting), were promising in preclinical studies[47, 48], they only transiently deplete microglia, and had limited efficacy in a phase II clinical trial[49]. Similarly, PLX3397 treatment of Nf1-OPG mice from 4 to 7 WOA had no durable effect on tumor growth when assayed at 25 WOA, despite a persistent reduction in microglia content. In contrast, interrupting the T cell/microglia immune axis resulted in durable effects, with mice exhibiting reduced glioma proliferation >3 months after the cessation of treatment. Taken together, these data reveal critical windows of tumor evolution during which stromal targeting abrogates glioma progression. Moreover, the ability to inhibit the supportive tumor microenvironment for a short period of time is particularly attractive for the treatment of pediatric LGGs that arise in the context of a developing brain.

Click here for additional data file.