RNA sequence analysis reveals ITGAL/CD11A as a stromal regulator of murine low-grade glioma growth

Abstract

Background

Emerging insights from numerous laboratories have revealed important roles for nonneoplastic cells in the development and progression of brain tumors. One of these nonneoplastic cellular constituents, glioma-associated microglia (GAM), represents a unique population of brain monocytes within the tumor microenvironment that have been reported to both promote and inhibit glioma proliferation. To elucidate the role of GAM in the setting of low-grade glioma (LGG), we leveraged RNA sequencing meta-analysis, genetically engineered mouse strains, and human biospecimens.

Methods

Publicly available disease-associated microglia (DAM) RNA-seq datasets were used, followed by immunohistochemistry and RNAScope validation. CD11a-deficient mouse microglia were used for in vitro functional studies, while LGG growth in mice was assessed using anti-CD11a neutralizing antibody treatment of Neurofibromatosis type 1 (Nf1) optic glioma mice in vivo.

Results

We identified Itgal/CD11a enrichment in GAM relative to other DAM populations, which was confirmed in several independently generated murine models of Nf1 optic glioma. Moreover, ITGAL/CD11A expression was similarly increased in human LGG (pilocytic astrocytoma) specimens from several different datasets, specifically in microglia from these tumors. Using CD11a-knockout mice, CD11a expression was shown to be critical for murine microglia CX3CL1 receptor (Cx3cr1) expression and CX3CL1-directed motility, as well as glioma mitogen (Ccl5) production. Consistent with an instructive role for CD11a⁺ microglia in stromal control of LGG growth, antibody-mediated CD11a inhibition reduced mouse Nf1 LGG growth in vivo.

Conclusions

Collectively, these findings establish ITGAL/CD11A as a critical microglia regulator of LGG biology relevant to future stroma-targeted brain tumor treatment strategies.

Key Points.

• ITGAL/CD11A is enriched in GAM from both murine and human LGGs

• Itgal/CD11a regulates microglia motility and Ccl5 secretion in vitro

• CD11a inhibition reduces murine Nf1-optic glioma growth in vivo

Importance of the Study.

With the recognition that low-grade brain tumors, specifically gliomas, are heavily dependent on their local tumor microenvironment, recent studies have focused on microglia/macrophages as the major cellular component of LGGs. Prior studies from our laboratory and others have shown that GAM govern glioma growth through interactions with T cells and the subsequent production of glioma growth factors. Using RNA sequencing meta-analyses, we found that Itgal/CD11a expression is enriched in GAM in several murine models of Nf1 optic glioma, as well as in human sporadic and NF1-associated pilocytic astrocytomas. CD11a expression predominates in microglia relative to other nonneoplastic stromal cell types, where it functions to regulate microglia chemokine-directed motility and glioma mitogen (Ccl5) production in vitro. Importantly, antibody-mediated CD11a inhibition reduces murine LGG growth in vivo. Collectively, these proof-of-principle experiments identify ITGAL/CD11A as both a biomarker and therapeutic target in GAM, relevant to future stromal targeting of brain tumors in children.

The critical importance of the tumor microenvironment to brain cancer development and progression has gained considerable traction over the past decade, with numerous studies highlighting the key interactions between nonneoplastic and neoplastic cell populations.^1–6 While this stromal dependency exists for both benign and malignant brain tumors, it is perhaps most acutely appreciated in the setting of low-grade brain cancers, likely recapitulating some of the key interactions necessary for normal brain formation and maintenance. As such, mutations in genes important for brain growth regulation [eg, BRAF, FGFR1 and NTRK (pilocytic astrocytoma (PA)^7,8), tumor necrosis factor receptor signaling (meningioma⁹)] and tissue patterning [eg, Hippo pathway (ependymoma, meningioma¹⁰), frizzled receptor signaling (meningioma¹¹)] predominate in these low-grade tumor histotypes.

In low-grade gliomas (LGGs) of the CNS, microglia have emerged as essential microenvironmental mediators of tumorigenesis and neoplastic cell growth,^12–14 comprising 30%–50% of the cells in these tumors.¹⁵ Using a LGG mouse model of the Neurofibromatosis type 1 (Nf1) cancer predisposition syndrome, we have previously demonstrated that microglia are necessary for glioma formation and continued growth. In these studies, Nf1-mutant mice, genetically engineered to form LGGs of the optic nerve (optic pathway glioma; Nf1-OPG), harbor increased numbers of resident (CD11b⁺, CX3CR1⁺, Iba1⁺, CD45^low, Tmem119⁺) microglia.^3,12,15 The importance of these GAM to tumorigenesis has been revealed using multiple strategies: Pharmacologic or genetic silencing of microglia-directed migration,¹² activation,^2,4,13,15 or growth factor production,² is sufficient to delay optic glioma formation and inhibit established tumor proliferation.

Given the critical role of microglia in Nf1-OPG formation and progression, we leveraged RNA sequencing meta-analysis to identify potential markers of glioma-associated microglia (GAM). Using published RNA sequencing data from microglia (CD11b⁺CD45^low cells) isolated from nervous tissue in healthy animals and those with experimental neurological diseases, we discovered the cell surface molecule CD11A (ITGAL) as a GAM-enriched protein in multiple murine models of Nf1-OPG and in human LGG (PA). Relevant to LGG growth regulation, we demonstrated that murine microglia chemokine-directed migration and Nf1-OPG stromal growth factor (Ccl5) production is decreased in CD11a-deficient mice in vitro, and that treatment of Nf1-OPG mice with neutralizing anti-CD11a antibodies reduced tumor volume and proliferation in vivo. Collectively, these findings establish microglial CD11A as a key stromal regulator of LGG pathobiology.

Materials and Methods

Mouse RNA Sequencing and Analysis

Several microglia-specific RNA expression datasets were used in this study. These included microglia from the spinal cords of SOD1-G93A-mutant mice with experimental ALS (GSE43366), microglia from doxycycline-inducible NEFH-tTa/tetO-208-hTDP43 mice with experimental ALS (GSE109171), microglia from the brains of mice with autism-like phenotypes caused by the induction of maternal allergic asthma (GSE94525), microglia from the brains of Mecp2-mutant mice with autism-like phenotypes (GSE66211), microglia from the optic nerves of mice with Nf1 optic gliomas (GSE65831), microglia from mice with several distinct models of high-grade gliomas (nestin-TVA/Ink4a-Arf-null/RCAS-PDGFB mice; GSE65868 and nestin-TVA/RCAS-PDGFB-shP53 and GL261 mice; GSE86573).

Each dataset was aligned, normalized, and analyzed individually to identify the differentially expressed genes (DEGs) between disease-associated microglia (DAM) and their respective controls in each dataset, and the results for each disease state were combined and compared using Partek Flow software, version 8.0. RNA-seq reads were aligned to the mm10—Ensembl Transcripts release 96 assembly with STAR version 2.6.1d.¹⁶ Gene counts and isoform expression were derived from the assembly output. Sequencing performance was assessed for the total number of aligned reads, total number of uniquely aligned reads, and features detected. Normalization size factors were calculated for all gene counts by counts per million (CPM) to adjust for differences in library size. Gene-specific analyses were then performed using the lognormal with shrinkage model (limma-trend method¹⁷) to analyze for differences between conditions, and the results were filtered for only those genes with P-values ≤ .01 and log fold-changes equal or greater than ± 2. DEG contrasts were run for each diseased state versus their respective controls. The resulting DEG lists for each study were compared, and the common intersections of the glioma datasets minus any overlap from the ALS or ASD datasets were used to generate a preliminary DEG list for GAM (Supplementary Figure 1B).

Microglia from Alzheimer’s disease (AD; 5XFAD) mouse microarray datasets (GSE65067, GSE98823) were individually analyzed, and any genes common to the AD lists and the initial list of glioma-specific genes were excluded, leaving 16 GAM transcripts. Hierarchical clustering was performed using Pearson correlation and average linkage. Principal component analysis was conducted using normalized gene counts (Figure 1C).

Fig. 1.

Elevated Itgal expression characterizes murine GAM. (A) Schematic representation of the bioinformatics analysis pipeline used. (B) Heat map comparing mouse microglia-specific gene expression in the setting of amyotrophic lateral sclerosis (ALS), autism spectrum disorder (ASD), low-grade glioma (LGG), and high-grade glioma (HGG). (C) Principal component analysis of the spatial distribution of the datasets used in the analysis performed in B. (D) Heat map showing the Alzheimer's disease (AD)-specific microglia genes and respective controls (CTL) used to exclude shared genes from the list of DEGs obtained from B. (E) Venn diagrams showing all of the disease-specific microglia transcripts upregulated (left) and downregulated (right), both shared and exclusive to each dataset. (F) Final list of sixteen GAM DEGs. (G) Microglia isolated from Nf1 optic glioma (Nf1-OPG)-bearing nerves exhibit a 4.24-fold increase in Itgal expression relative to microglia from nontumor-bearing Nf1^+/− optic nerves (Nf1^+/− microglia: 15.53 ± 1.03, n = 3; Nf1-OPG microglia, 65.89 ± 16.25, n = 4, P = 2.94 × 10^–3).

Human Samples

Flash-frozen and pulverized PA samples obtained from the St. Louis Children’s Hospital Pediatric Tumor Bank included normal brain (N = 9; 2 males, 7 females; 10.25 ± 2.81 years of age), non-NF1 PA (N = 9; 3 males, 6 females; 10.11 ± 1.76 years of age), and NF1 PA (N = 9; 3 males, 6 females; 10.11 ± 1.91 years of age), which were employed for RT-qPCR using the primers listed in Supplementary Table 1. These de-identified specimens were used under an Institutional Review Board-approved protocol. Human sporadic (non-NF1, N = 44) PA samples were immunostained on a previously prepared tissue microarray,¹⁸ and the percent of CD11A-immunoreactive cells was quantitated and compared to nonneoplastic brain tissue (N = 6). All other RNA sequencing datasets were previously acquired, including CD11B⁺, CD45^high macrophages and CD11B⁺, CD45^low microglia from three sporadic human PAs (GSE65867), as well as additional human PA samples and nonneoplastic brain controls (GSE163071; N = 10 sporadic PAs; N = 10 NF1-PAs, n = 6 nonneoplastic brain samples). CPM normalization and limma-trend method differential analysis were used to identify differences in gene expression.

Mice

CD11a-deficient mice were purchased from The Jackson Laboratory (B6.129S7-Itgaltm1Bll/J, stock number 005257), while other mouse strains have been previously described.^3,19 Mice were maintained on a C57BL/6 background, and used in accordance with an approved Animal Studies Committee protocol. Mice were kept on a 12 h light/dark cycle with full access to food, and were randomly assigned to all experimental groups without bias (containing both male and female mice). The investigators were blinded to sample group allocation and subsequent analysis during all of the experiments.

Mouse Treatments

Nf1-OPG mice were injected intraperitoneally with 150 µg of anti-CD3 neutralizing antibodies (n = 5; Bio X Cell, USA - BE0002) or rat IgG2a (n = 5; isotype control, Bio X Cell - BE0089) every other day, 250 µg of InVivo Mab antimouse LFA-1α antibodies (CD11a, n = 12; Bio X Cell, USA - BE0006) 3 days per week, or 250 µg of InVivo Mab rat IgG2a isotype control antibodies (n = 10; Bio X Cell USA - BE0089) 3 days per week, for a total of 6 weeks, beginning at 6 weeks of age. All mice were perfused at 12 weeks of age, and their optic nerves were collected and processed for immunohistochemistry analysis as previously described.¹⁹ PLX33397 treatment was performed by replacing the normal chow of 21-week-old Nf1-OPG mice with the AIN-76A rodent diet containing 275 mg PLX3397 (Free Base) /kg (Research Diet Inc.) for 4 weeks prior to optic nerve processing at 25 weeks of age.

Optic Nerve Volume Measurements

After dissection of the optic nerves, the tissues were photographed with a scale bar, and volumes calculated as previously reported.¹⁹

Immunohistochemistry

Mice were euthanized, perfused transcardially, and processed for paraffin embedding and sectioning, as previously reported.¹⁹ Serial 4 μm paraffin sections were immunostained with mouse anti-Ki67 (1:200 BDB556003 Fisher Scientific), rabbit anti-Iba1 (1:500, Wako Chemicals 019-1794, Japan) and rabbit anti-CD11a (1:100 ab203336, Abcam) primary antibodies prior to species-specific secondary antibody incubation and development using Vectastain Elite ABC kit (Vector Laboratories).

Optic nerves were scanned on a NanoZoomer 2.0-HT slide scanner (Hamamatsu Photonics K.K.) with a 20x objective or a Nikon eclipse E600 microscope (objective: Plan Fluo 20x 0.5; camera: LEICA EC3) and the percent of positive cells was calculated in the region of interest.

RNAScope

RNAScope® was performed according to manufacturer’s protocol (ACDBio). Briefly, after perfusion with 4% PFA, optic nerves were fixed overnight in 4% PFA, embedded in paraffin, sectioned into 4 µm-thick sections and allowed to dry at room temperature overnight. Following deparaffinization, the tissue was incubated with hydrogen peroxide, antigen retrieval was performed, and the sections were incubated with an antisense-Itgal probe. Sections were subsequently blocked for 1h at RT with 10% goat serum, followed by an overnight incubation with rabbit anti-Iba1 (1:500, Wako Chemicals 019-1794), chicken anti-GFAP (1:1000, ab4674, Abcam), mouse anti-O4 (1:100, MAB345M, Fisher Scientific) or rat anti-CD3 (1:50, ab11089, Abcam) antibodies, and then incubation with Alexa 488 goat anti-rabbit secondary antibody (1:200, A11008 Invitrogen,), Alexa Fluor 488 goat anti-chicken secondary antibody (1:200, A11039, Fisher Scientific), Alexa Fluor 488 goat anti-mouse IgG (H+L) secondary antibody (1:200, A11029, Fisher Scientific), or Alexa Fluor 488 goat anti-rat IgG (H + L) secondary antibody (1:200, A11006, Fisher Scientific), respectively, for 1h at RT followed by RNAScope® DAPI for 30s, and mounted using ProLong® Gold Antifade Reagent (9071S, Cell Signaling Technology). The density of Iba1⁺ and CD3⁺ cells was accessed by normalizing to the total number of DAPI⁺ cells, where Itgal-expressing cell content was determined by normalizing the double-positive cells for each cell marker to the total number of Itgal⁺ cells.

Microglia Isolation

Microglia isolation was performed using the multi-tissue dissociation kit (Miltenyl Biotek - 130-110-201) following established protocols.⁴ The resulting cells were maintained in minimal essential medium supplemented with 1 mM L-glutamine, 1 mM sodium pyruvate, 0.6% D-(+)-glucose, 1 ng/ml GM-CSF, 100 μg/ml P/S, and 10% FBS. After 2 weeks, microglia were separated from the astrocytes by shaking (cultured in medium without GM-CSF 3 days before shaking, 200 g, 5 h, 37°C).

Boyden Chamber Migration Assay

Three thousand microglia were placed into the upper chamber of the 96-well Transwell filter (5 µm; Corning) with 100 µL F12- DMEM without FBS. The lower chamber contained 0.2 mL DMEM with 50 nM CX3CL1 (472-FF-025; R&D System). Microglia on the lower chamber side of the membrane were counted after 5 h. All experiments were repeated three times with at least three independent biological replicates.

RNA Extraction, cDNA Production, RT qPCR

CD11a-deficient and WT mouse brainstems were dissected, homogenized and RNA was isolated using the NucleoSpin® RNA Plus kit (740984.205) as per the manufacturer’s instructions. RNA from cultured microglia was isolated using NucleoSpin RNA Plus XS kit (NC1498148). Total RNA was reverse-transcribed into cDNA using a High-Capacity cDNA Reverse Transcription Kit (Cat no-4374967, Applied Biosystems) as per the manufacturer’s instructions. RT-qPCR was performed using TaqMan Gene Expression primers for Itgal, Ccr2, Ccr4, Tmem119, Cx3cr1, Ccl5, Ccl4 and Gapdh (Supplementary Table 1), and relative gene expression was calculated using the ΔΔCT analysis method following the manufacturer’s instructions (Thermo Fisher Scientific). At least three mice were used for each experiment. Similarly, RNA was extracted from (9 nonneoplastic brains, 9 sporadic PA and 9 NF1-associated PA) for qRT-PCR using TaqMan primers (Supplementary Table 1).

Statistical Analysis

All data were analyzed using GraphPad Prism 8 software. Data between two groups were analyzed by unpaired two-tailed Student’s t-tests. Data for multiple comparisons were analyzed with one-way ANOVA or two-way ANOVA, as per the software instructions. Statistical significance was set at P < .05.

Results

Elevated Itgal Expression Characterizes Murine GAM

Based on numerous studies demonstrating that unique populations of monocytes exist during development and brain disease, we sought to identify genes enriched in GAM. For these studies, we initially leveraged published microglia (FACS-isolated CD11b⁺CD45^low cells) RNA sequencing datasets from experimental mouse models of amyotrophic lateral sclerosis (ALS), autism spectrum disorder (ASD) and glioma normalized to their experimental controls (Figure 1A–C; Supplementary Figure 1A). The DEGs from these datasets were then intersected to generate a list of twenty-three DEGs specifically enriched in GAM (Supplementary Figure 1B). As a secondary filter to increase specificity, we next intersected the initial list of twenty-three GAM DEGs with DEGs generated using two independent microarray datasets of microglia from experimental mouse models of Alzheimer’s disease (Figure 1D). After crossing information on all shared and exclusively expressed genes in each disease model (Figure 1E), we identified sixteen GAM-enriched DEGs (Figure 1F), of which eleven were increased and five were decreased in expression relative to controls.

As we sought to identify a potential cell surface marker for this unique microglial population, we focused on Itgal, which codes for the CD11a cell membrane protein. Moreover, Itgal was the DEG with the highest level of overexpression in GAM relative to control microglia. Importantly, Itgal is expressed at very low levels in normal microglia populations, both in the human and mouse brain (Supplementary Figure 1C–E), according to the brainrnaseq.org database, which was confirmed in the control microglia from the DAM and GAM datasets used in our analysis (Supplementary Figure 2A). Consistent with its enrichment in glioma, Itgal expression was increased >4-fold in microglia isolated from the optic nerves of Nf1-mutant mice with optic glioma (Nf1-OPG) relative to microglia isolated from Nf1-mutant mice lacking optic gliomas (Nf1^+/− mice)² (Figure 1G).

CD11a⁺ Microglia Are Enriched in Murine Nf1-OPGs and Human PA

To confirm that CD11a is expressed in microglia in murine LGGs, we leveraged our prior experience with a series of four independently generated Nf1-mutant mouse strains that develop LGG (Nf1-OPG). For these analyses, we used Nf1^flox/flox (FF) mice that do not develop optic gliomas, as well as Nf1^flox/neo; GFAP-Cre (Nf1-OPG), Nf1^flox/R681X; GFAP-Cre (Nf1-OPG R681X), Nf1^flox/R1278P; GFAP-Cre (Nf1-OPG R1278P), and Nf1^flox/C383X; GFAP-Cre (Nf1-OPG C383X) mice, all of which develop optic gliomas by 12 weeks of age.^3,19 Consistent with the RNA sequencing data, CD11a-immunoreactive cells were detected at low levels in the optic nerves of nontumor-bearing mice (FF). In contrast, 12%–16% of the cells in the optic nerves of the four distinct Nf1-OPG strains were CD11a⁺ (Figure 2A). It is worth noting that Nf1-OPG C383X mice have reduced tumor penetrance,³ which likely accounts for the wide variation in CD11a⁺ cell content in their optic gliomas. In addition, CD11a⁺ cell content in the optic nerves of Nf1-OPG mice increases as a function of age (Supplementary Figure 2A), as previously reported.²⁰

Fig. 2.

CD11a⁺ cells are increased in both murine and human LGGs. (A) CD11a⁺ cells were low or nearly undetectable in Nf1^flox/flox (FF) (nonneoplastic controls; FF: n = 6, 2.12 ± 1.02%), whereas four different Nf1-OPG models harbor 12%–16% CD11a⁺ cells (Nf1-OPG: n = 9, 15.92 ± 1.47%, P < .0001; Nf1-OPG-R681X: n = 4, 15.68 ± 0.55%, P = .0002; Nf1-OPG-R1278P: n = 4, 13.02 ± 1.32%, P = .0024; Nf1-OPG-C383X: n = 8, 10.73 ± 4.24%, P = .0004). (B) ITGAL expression is increased in human pilocytic astrocytoma (PA) relative to nonneoplastic brain tissue (P = .0001; GSE44971), (C) which was confirmed in two independent cohorts of PA tumors, including sporadic (non-NF1 PA, dark blue) and NF1-associated PA (light blue) (nonneoplastic tissue, n = 9, 1.005 ± 0.0931; PA, n = 18, 4.487 ± 1.160; P = .0459) by qRT-PCR (left) and sporadic PA (N = 20) compared to nonneoplastic brain (N = 6) by RNA-sequencing (GSE163071, P = .006; right). (D) Increased percentages of CD11A⁺ cells were observed in PAs (N = 44, 4.246 ± 0.5758) relative to nonneoplastic brain (N = 6, 0.3467 ± 0.3467; P = .0172). Scale bar, 50 µm.

To determine whether ITGAL/CD11A is similarly increased in human LGGs, we performed several experiments. First, we used a publicly available RNA microarray dataset (GSE44971) of human PAs to demonstrate increased ITGAL expression in the tumors²¹ relative to control tissues (Figure 2B). Second, we confirmed elevated ITGAL expression by RT-qPCR using an independent set of PAs, including nine samples from individuals with NF1³ (Figure 2C, left). Third, we performed RNA sequencing on another cohort of NF1-associated and sporadic PA tumors (N = 20), which similarly show increased ITGAL expression in the tumors relative to nonneoplastic brain (N = 6); (GSE163071) (Figure 2C, right). Fourth, we leveraged a tissue microarray containing a fourth independent set of PAs and control tissues to demonstrate increased CD11A⁺ cell content in the tumors relative to controls (Figure 2E).

CD11a Is Enriched in Microglia From Murine and Human LGGs

To ascertain whether the CD11a⁺ cells were microglia, we coupled RNAscope (in situ RNA hybridization) using a probe to detect Itgal in the optic nerves of Nf1-OPG mice with immunolabeling using cell type-specific antibodies (Iba1, microglia; GFAP, astrocytes; O4, oligodendrocytes; CD3, T cells). While Itgal was not exclusively expressed in Nf1-OPG-associated microglia, over 50% of the Itgal⁺ cells in the murine Nf1-OPGs were Iba1⁺ microglia (Figure 3A). The Iba1⁺ cells in these murine gliomas are microglia, rather than macrophages, based on prior FACS analysis,^12,15 Tmem119 expression,³ and double labeling demonstrating that all Iba1⁺ cells are Tmem119⁺ microglia (Supplementary Figure 2B).

Fig. 3.

CD11a is enriched in microglia from both murine and human LGGs. (A) RNAscope® in situ hybridization was used to identify Itgal (red) expression in the various cell populations (green: GFAP, astrocytes; Iba1, microglia; O4, oligodendrocytes; CD3, T lymphocytes) in Nf1-OPG (n = 3). The majority of the cells (52.78 ± 3.21%) expressing Itgal in the Nf1-OPG mice were Iba1⁺ microglia. (B) left: Treatment with PLX3397 decreases the density of Iba1⁺ cells in the optic nerves of Nf1-OPG mice (Control n = 10, 16.12 ± 1.56 % vs PLX3397: n = 6, 3.78 ± 0.41%, P < .0001). right: Treatment with PLX3397 decreases the density of CD11a⁺ cells in the optic nerves of Nf1-OPG mice relative to control mice (controls: n = 10, 19.12 ± 0.8805%; PLX n = 6, 12.92 ± 1.153%, P = .0008. (C) CD11a⁺ cell content in the optic nerves of Nf1-OPG mice is not affected by anti-CD3 antibody treatment (n = 5, 15.61% ± 0.82) related to isotype control-treated mice (n = 5, 15.17 ± 0.4178, P = .2196). (D) Microglia isolated from human sporadic pilocytic astrocytomas have increased ITGAL expression (n = 3, 92.32 ± 20.96 gene counts) relative to whole pilocytic astrocytoma tissue samples (sporadic PA, dark blue; NF1-PA, light blue; n = 20, 7.605 ± 1.67 gene counts, P < .0001) and sporadic PA-associated macrophages (n = 3, 21.46 ± 5.14 gene counts; P = .0304). Scale bar, 50 µm.

Since T cells also express CD11a, we performed three additional experiments to confirm the identity of the CD11a⁺ cells in the tumors. First, we treated Nf1-OPG mice with PLX3397 to partially deplete microglia, resulting in a 4-fold reduction in Iba1⁺ cells. Since ~50% of the CD11a⁺ cells in the tumors are Iba1⁺ cells (Figure 3A), this would result in a ~2-fold reduction in CD11a⁺ cell content, close to the 1.5 fold decrease observed (Figure 3B). Second, we treated Nf1-OPG mice with anti-CD3 antibodies to deplete T cells, but observed no change in CD11a⁺ cell content (Figure 3C). Since T cells account for ~0.007% of the cells in these tumors (Supplementary Figure 2C), complete elimination of T cells, representing ~20% of the CD11a⁺ cells, would have little effect on overall CD11a⁺ cell content, as observed. Third, we analyzed Itgal expression in isolated Nf1-mutant microglia and neoplastic cells (Nf1-deficient optic glioma stem cells; GSE102345) relative to the intact tumors (GSE102345). Whereas no enrichment was detected in the neoplastic cells, there was a ~52-fold enrichment of Itgal expression in the GAM (GSE65868) (Supplementary Figure 2D), far greater than the observed increase in microglia content. Fourth, leveraging another RNAseq dataset (GSE65867) in which microglia (CD11B⁺, CD45^low) and macrophages (CD11B⁺, CD45^high) were isolated from three human sporadic PAs, ITGAL expression was enriched 12-fold in microglia relative to the bulk tumors (GSE163071) or PA-associated macrophages (4-fold) (Figure 3D).

CD11a Controls Microglia Migration and Mitogen Production

To determine the functional relevance of CD11a expression in microglia, we used CD11a knockout (CD11aKO) mice in which the expression of Itgal was ablated (Figure 4A). First, we examined microglia content in the optic nerves of CD11aKO and WT mice. Quantitation of the percent of Iba1⁺ cells (microglia) in these specimens revealed no effect of CD11a loss on microglia density in the optic nerve (Figure 4B). Second, we hypothesized that CD11a loss could alter the expression of genes encoding cell surface proteins involved in microglia homeostasis (Tmem119) or migration (Ccr2, Ccr4, Cx3cr1). Based on prior studies examining the effect of T cell depletion on microglia function, we performed RT-qPCR on RNA from the brainstems of these mice. While Tmem119, Ccr2, and Ccr4 expression were similar in CD11aKO and WT mice, CD11a-deficient mice exhibit reduced Cx3cr1 expression (Figure 4C). Third, to determine whether these gene expression patterns were also observed in brain microglia, RNA was isolated from microglia for RT-qPCR. Similar to the brainstem results, there was reduced Cx3cr1 expression in CD11a-deficient microglia relative to controls, with no changes in Tmem119, Ccr2, and Ccr4 expression (Figure 4D). Fourth, since CX3CR1 is the microglia receptor responsible for mediating directed migration in response to the CX3CL1 chemokine, we used a Boyden chamber migration assay to examine CX3CL1-mediated microglia chemotaxis. Briefly, microglia were cultured for 2 weeks, and added to the upper chamber of a well with a porous membrane, while CX3CL1, the ligand for CX3CR1, was added to the bottom chamber of the same well (Figure 4E, left). After 5 h, the number of WT and CD11aKO microglia that crossed the porous membrane into the bottom well was quantified. Consistent with decreased Cx3cr1 expression, the directed migration of CD11aKO microglia to CX3CL1 was reduced relative to their WT counterparts (Figure 4E, right). Fifth, based on prior work demonstrating that GAM in murine Nf1-OPGs express Ccl5, a stromal mitogen required for the formation and continued growth of these LGGs,² we measured Ccl5 levels in brainstem tissue and microglia isolated from brains of WT and CD11a-deficient mice. Compared to WT mice, CD11a loss in microglia led to reduced Ccl5 expression (Figure 4F). Finally, microglia from CD11aKO mice exhibit impaired phagocytosis relative to their WT counterparts (Supplementary Figure 3A). Taken together, these studies reveal significant cellular and molecular defects in CD11a-deficient microglia germane to Nf1-OPG growth.

Fig. 4.

CD11a controls microglia migration and mitogen production. (A) CD11aKO mice (n = 4) do not express Itgal relative to wild-type (WT) controls (n = 4, P < .0001). (B) The density of Iba1⁺ cells (microglia) in CD11aKO mice (n = 5, 7.93 ± 0.42%) was similar to that seen in WT mice (n = 5, 7.26 ± 0.57%, P = .37). RNA RT-qPCR analysis of (C) total brainstem tissue and (D) isolated brain microglia revealed no differences in the expression of Ccr2, Ccr4 and Tmem119 in CD11aKO relative to WT mice. Reduced expression of Cx3cr1 was observed in CD11aKO (n = 3) relative to WT (n = 3) mice (Brainstem tissue: WT, 1.00 ± 0.04 fold; CD11aKO, 0.67 ± 0.07 fold; P = .0151; microglia: 1.00 ± 0.15; CD11a-KO, 0.34 ± 0.094; P = .02). (E) 3000 microglia (blue) from CD11aKO or WT mice were placed in the upper chamber and 50 nM CX3CL1 (pink) was added to the lower chamber of the transwell. CD11aKO microglia had reduced migration (n = 4, 45.5 ± 3.18%, P = .02) compared to their WT counterparts (n = 4, 62.50 ± 3.38%, right). (F) cDNA RT-qPCR analysis of total brainstem tissue and brain microglia revealed reduced Ccl5 expression in CD11aKO relative to WT mice (brainstem tissue (n = 4 in each group): WT, 1.00 ± 0.04 fold; CD11aKO, 0.67 ± 0.07 fold; P = .0151; microglia (n = 3 in each group): WT, 1.00 ± 0.15 fold; CD11aKO, 0.34 ± 0.094 fold; P = .02). Scale bar, 50 µm.

CD11a Inhibition Reduces Nf1-OPG Microglia Infiltration and Proliferation

Since CD11a deficiency resulted in reduced microglia migration and Ccl5 production, we sought to determine whether antibody-mediated CD11a inhibition would reduce Nf1-OPG growth in vivo. For these experiments, 6-week-old Nf1-OPG mice were treated three times a week with 250 µg of neutralizing anti-CD11a or anti-IgG2a (isotype control) antibodies until 12 weeks of age when >90% of Nf1-OPG mice harbor LGGs (Figure 5A). Relative to anti-IgG antibody-treated Nf1-OPG mice, Nf1-OPG mice receiving anti-CD11a antibodies had reduced density of CD11a⁺ cells and Iba1⁺ microglia (Figure 5B), consistent with the finding that CD11a loss reduces Cx3cr1 expression and CX3CL1-directed microglia chemotaxis (Figure 4E). Since CD11a-deficient microglia have >50% reduction in Ccl5 expression and Ccl5 is essential for Nf1-OPG growth, we examined tumor volume and proliferation (%Ki67⁺ cells) following anti-CD11a antibody treatment, and observed reduced optic nerve volumes and a 5-fold decrease in Nf1-OPG proliferation in anti-CD11a antibody-treated mice relative to IgG isotype antibody-treated controls (Figure 5B). These results demonstrate that CD11a inhibition attenuates Nf1-OPG growth in vivo.

Fig. 5.

CD11a inhibition reduces Nf1-OPG proliferation and microglia infiltration. (A) Nf1-OPG mice were treated thrice weekly with either 250 µg of CD11a neutralizing antibody (n = 12) or an IgG2a isotype control antibody (n = 10) from 6 to 12 weeks of age. (B) Nf1-OPG mice treated with anti-CD11a neutralizing antibodies have fewer CD11a-positive cells (%CD11a⁺ cells) in their optic nerves relative to IgG2a-treated control mice (anti-IgG2a, 15.95 ± 0.6467%; anti-CD11a, 6.163 ± 0.3106%, P < .0001), as well as fewer Iba1⁺ microglia (anti-IgG2a, 14.84 ± 2.327%; anti-CD11a, 6.279 ± 0.4860%; P = 0.0008). Anti-CD11a treated mice also have reduced optic nerve volumes following anti-CD11a antibody treatment (IgG2a, 0.07137 ± 0.005352 mm³; anti-CD11a, 0.05597 ± 0.002759 mm³; P = .00142), as well as a decrease in tumor cells’ proliferation (%Ki67⁺ cells; IgG2a, 3.095 ± 0.5040%; anti-CD11a, 0.5992 ± 0.1371%; P < .0001). Scale bar, 50 µm for CD11a, Iba1 and Ki67 images; 1 mm for volume images.

Discussion

In this report, we employed RNA sequence analysis of various DAM populations to identify ITGAL/CD11A as a potential GAM marker. We show that CD11a was enriched in microglia isolated from Nf1-OPG mice relative to the overall tumor tissue, as well as relative to microglia isolated from nontumor-bearing mice. Similarly, ITGAL/CD11A expression was increased in human PA tumors relative to nonneoplastic brain, where it was enriched in microglia. The importance of CD11a expression to microglia function was further revealed using CD11a-deficient microglia, which exhibited reduced Cx3cr1 expression and impaired CX3CL1 chemokine-directed migration in vitro, as well as reduced Ccl5 expression, a growth factor critical for Nf1-OPG growth. Finally, inhibiting CD11a function using neutralizing antibodies attenuated murine Nf1-OPG proliferation and microglia content in vivo. These exciting observations raise several important points relevant to glioma pathobiology.

First, CD11A was originally identified as the α-subunit of the lymphocyte function-associated antigen-1 (LFA-1), which binds numerous adhesion molecules, including intercellular adhesion molecule-1 (ICAM-1, CD54) and F11R (junctional adhesion molecule-A; JAM-A).²² While not extensively studied in monocytes (macrophages and microglia), LFA-1 mediates integrin-mediated cell adhesion and migration in lymphocytes,²³ as well as modulates receptor tyrosine kinase signaling.²⁴ In the setting of glioma, we hypothesize that CD11a most likely functions at the level of microglia, based on its increased expression in GAM relative to Nf1^+/− nonneoplastic optic nerve microglia, as well as its enrichment in human PA-associated microglia. Additionally, the finding that CD11a-deficient microglia have reduced Cx3cr1 expression, CX3CL1-mediated migration, and Ccl5 expression, coupled with our prior report that Nf1-OPG mice with Cx3cr1 genetic reduction exhibit delayed microglia infiltration and tumor formation,¹² underscores its relevance to LGG biology. Similarly, microglia Ccl5 is critical for glioma formation and growth, as Nf1-OPG stem cells do not form tumors in Ccl5-deficient mice, and Ccl5 neutralizing antibodies dramatically inhibit Nf1-OPG growth in vivo.²

Second, while Itgal is expressed in other nonneoplastic cells (T cells, oligodendroglia, and astrocytes), we suggest that its tumor-relevant function involves microglia, since the greatest reduction in CD11a⁺ cell content was observed following PLX3397-mediated microglia depletion, rather than CD3⁺ T cell depletion. Since T cells communicate with microglia as part of a neuron-immune-glioma cell axis,⁴ it is possible that this circuit is also disrupted at the level of T lymphocytes. As such, Nf1-mutant neurons produce midkine, which induces T cell Ccl4-mediated microglial Nf1-OPG growth. However, CD11a-knockout T cells produce Ccl4 following midkine stimulation (Chatterjee J, unpublished data), and Ccl4 levels in the brainstems of CD11a-KO mice were similar to their WT counterparts (Supplementary Figure 3B).

Third, while CD11a plays a critical role in Nf1-OPG growth, it does not appear to confer a survival advantage in children. This is likely due to the general excellent overall survival in children with sporadic or NF1-associated PA.^25,26 However, a reanalysis of nonredundant TCGA datasets identified a trend towards worse overall survival in patients with grade 3 astrocytomas and high ITGAL expression (Supplementary Figure 3C, P = .149), but not in adult patients with glioblastoma or children with other pediatric cancers, including medulloblastoma, neuroblastoma, Wilm’s tumor, acute lymphoid leukemia, and acute myeloid leukemia (data not shown). Interestingly, another LFA-1 binding partner (JAM-A/F11R) has been reported to function as a sex-specific (female) microglia tumor suppressor in glioblastoma.⁵ Future investigation will be required to define the prognostic value of CD11A expression and the LFA-1/integrin axis in predicting patient survival.

Fourth, in light of the recent report that CD11a might also mark macrophages present in brains of mice with AD,²⁷ it is conceivable that CD11a identifies a population of monocytes with specific functions in the setting of neurological disease. The notion that CD11A identifies a subset of microglia/macrophages is consistent with the idea that brain microglia are composed of numerous subpopulations with specialized capabilities. As such, prior bulk and single cell RNA sequencing studies coupled with cytometry time-of-flight mass spectrometry (CyTOF) analyses have revealed significant levels of microglia heterogeneity in the normal brain with respect to CNS (brainstem, cortex) location,²⁸ sex,^29,30 and developmental stage (PAM).^31,32 Similarly, in the context of disease, multiple distinct populations of monocytes have been identified, including those with increased phagocytic activity in neurodegenerative disorders (DAM),³³ as well as heterogeneity of monocytes (mixtures of peripheral infiltrating macrophages and resident microglia) in glioma. Each of these populations have different, and often opposing, functions in glioma, with both tumor-suppressive and tumor-promoting actions reported in experimental malignant glioma model systems.^6,21,34

This spatial, temporal and functional diversity of microglia in health and disease has additional implications for both understanding the role of microglia in specific disease contexts,³⁵ but also with respect to microglia-directed therapies.³⁶ While additional studies will be required to discern the tumor-specific functions of CD11a⁺ microglia, we propose that CD11a operates to regulate microglia migration and Nf1-OPG growth factor production relevant to the generation of a supportive LGG microenvironment. The success of the anti-CD11a neutralizing antibody treatment highlights the need for future mechanistic dissections of the LFA-1 stromal circuit in LGG essential to defining the role of this unique subset of microglia in brain tumor pathobiology.

Funding

This work was funded by grants from the National Institute of Neurological Disorders and Stroke (1-R35-NS07211-01 Research Program Award to D.H.G.), National Institute of Health (S10-RR0227552 Shared Instrument), and National Eye Institute (Core Grant for Vision Research; P30EY002687).

Conflict of interest statement. D.H.G. has a licensing agreement with the Tuberous Sclerosis Alliance (GFAP-Cre mice). The other authors have no relevant conflicts of interest to disclose.

Authorship statement. D.H.G. worked on the conceptualization. A.D.A.C., J.C., O.C., Sh.S., X.G. worked on the methodology. A.D.A.C., D.H.G. worked on the writing. A.D.A.C. performed the investigation. Sh.S., Su.S., and J.T. performed most of the experiments. O.C. performed the RNA sequencing analysis. S.D. procured the human specimens. D.H.G. was responsible for the resources. D.H.G. was in charge of supervision and funding.