1p/19q co-deletion status is associated with distinct tumor-associated macrophage infiltration in IDH mutated lower-grade gliomas

Abstract

Background

Tumor-associated macrophages (TAM)s are critical regulators of glioma progression. As yet, however, TAMs in isocitrate dehydrogenase (IDH) mutated lower-grade gliomas (LGGs) have not been thoroughly investigated. The aim of this study was to determine whether 1p/19q co-deletion status affects the TAM phenotype or its prevalence in IDH mutated LGGs.

Methods

TAMs in IDH mutated LGGs were analyzed using transcriptome data from 230 samples in the TCGA database in combination with transcriptome data from single-cell RNA sequencing of IDH-mutated LGGs. Proteins potentially involved in TAM regulation were examined by immuno-staining in primary LGG samples harboring IDH mutations. Essential signaling pathways regulating TAM phenotypes were investigated in a glioma mouse model using small molecule inhibitors.

Results

Most of the TAMs in IDH-mutated LGGs expressed the M1 activation markers CD86 and TNF, whereas a subset of individual TAMs co-expressed both M1 and M2-related markers. Bioinformatics analysis in combination with immuno-staining of IDH-mutated patient samples revealed higher amounts of TAMs expressing M2-related markers in 1p/19q non-codeletion IDH-mutated LGGs compared to 1p/19q codeletion LGGs. The levels of transforming growth factor beta 1 (TGFβ1) and macrophage colony-stimulating factor (M-CSF) were significantly higher in 1p/19q non-codeletion LGGs than in 1p/19q codeletion LGGs. M-CSF and TGFβ1 signal inhibition decreased tumor growth and modulated the TAM phenotype in a glioma mouse model.

Conclusions

Our data indicate that 1p/19q co-deletion status relates to distinct TAM infiltration in gliomas, which is likely mediated by M-CSF and TGFβ1 signaling. M-CSF and TGFβ1 signaling may play a pivotal role in regulating the TAM phenotype in glioma.

Electronic supplementary material

The online version of this article (10.1007/s13402-020-00561-1) contains supplementary material, which is available to authorized users.

Introduction

Diffuse lower-grade gliomas (LGGs) are infiltrative brain tumors that include World Health Organization (WHO) grade II and III neoplasms [1]. Based on mutations in isocitrate dehydrogenase 1 and 2 (IDH1 and IDH2, hereafter collectively referred to as IDH) and codeletion of chromosomes 1p and 19q, LGGs can be classified into three molecular subgroups: (i) LGGs with IDH mutations and 1p/19q codeletion (IDHmut-codel), (ii) LGGs with IDH mutations without 1p/19q codeletion (IDHmut-noncodel) and (iii) LGGs without IDH mutations and without 1p/19q codeletion (IDHwt) [2]. IDHwt LGGs are associated with the most aggressive clinical behavior and a worst outcome, and are molecularly and clinically distinct from the other two groups, but similar to glioblastomas (WHO grade IV) [2]. IDHmut-noncodel LGGs are diagnosed as astrocytoma, whereas IDHmut-codel LGGs are diagnosed as oligodendroglioma and are associated with the best prognosis [2]. Currently, there is no curative treatment option available and patients with LGG eventually die from relapse [3]. The high recurrence rate of glioma is attributed to both their extreme invasiveness and therapy resistance, mediated by interactions between tumor cells and their microenvironment [4].

The tumor microenvironment, which consists of stromal components, soluble factors and different types of non-malignant cells including tumor-associated macrophages (TAMs), contributes to glioma progression and aggravation [4]. Increasing evidence indicates that TAMs may support glioma growth and progression by enforcing immunosuppression [5] and enhancing proliferation [6, 7], migration [8], invasion and angiogenesis [9, 10]. In addition, TAMs have been shown to regulate glioma stem cell pools and to potentiate resistance to anti-angiogenic therapy [11, 12]. Glioma cells have been found to establish an immune suppressive microenvironment that skews TAMs towards a pro-tumorigenic M2-like phenotype and inhibits the development of an anti-tumorigenic M1-like phenotype [13, 14]. However, TAMs in glioma represent a distinct population that cannot be strictly dichotomized to M1/M2 polarization states [15–17]. Nevertheless, it has become clear that high amounts of TAMs bearing M2-related markers (CD163, CD204) are associated with higher tumor grades and a poor prognosis [13, 18]. Infiltration of CD204⁺ TAMs serves as an independent prognostic marker after adjusting for clinical data and the methylation status of O6-methylguanine-DNA methyltransferase in WHO grade III and IV gliomas [19]. TAM-targeting therapy with a colony stimulating factor 1 receptor (CSF-1R) inhibitor has been found to suppress tumor growth and progression and to lead to substantial survival improvements in a preclinical glioma model [20, 21]. This approach, however, failed to improve the clinical outcome of glioblastoma patients in a phase II clinical trial [22]. Interestingly, a study combining bulk expression profiles from the TCGA LGG dataset and single-cell RNA sequencing profiles from IDH mutated glioma patient samples revealed that the differential transcriptional signature observed between IDHmut-codel and IDHmut-noncodel LGGs in bulk tumors largely constituted microglia/macrophage specific genes [23]. As yet, however, the TAM phenotype in IDH mutated LGGs has not been thoroughly investigated, and potential differences in TAM infiltration between IDHmut-codel and IDHmut-noncodel gliomas have so far not been identified.

Materials and methods

Bioinformatics analysis of immune cell infiltration in LGGs

Processed single cell RNA sequencing (scRNA-seq) counts were obtained from the Gene Expression Omnibus (GEO) database for IDHmut-noncodel LGGs (GSE89567) and IDHmut-codel LGGs (GSE70630), and cell type assignment information from the corresponding studies was utilized [23, 24]. Only the cells from samples with a confirmed IDHmut-noncodel or IDHmut-condel status were selected for downstream analysis. An integrated dataset was processed and 2D tSNE analyses were performed using R Pagoda2 package (https://github.com/hms-dbmi/pagoda2, version: 0.0.0.9003).

Datasets including patient’s clinical information and processed RNA-sequencing data were downloaded from The Cancer Genome Atlas (TCGA) (http://cancergenome.nih.gov) or The Chinese Glioma Genome Atlas (CGGA) (http://cgga.org.cn) databases. The CIBERSORT in silico cytometry method was adopted to analyze different types of immune cell infiltration in LGGs using the TCGA LGG dataset [25, 26]. The CIBERSORT Java application was run with default parameters on the RNA expression data from each sample, and the percentages of M1 and M2 macrophages were predicted.

Patient material and image analysis

Permission for using patient samples was granted by the ethics committee of Shaanxi Normal University. Glioma samples were collected retrospectively at the Tangdu Hospital of the Airforce Military Medical University. Pathological classification and molecular characterization of the glioma samples were performed according to WHO criteria (2016). For sample details, see Table S1.

Immunohistochemistry (IHC) staining of CD68, CD204, M-CSF and TGF-β1 was performed as described previously [27] on 5 μm paraffin sections of the patient samples, including 22 IDH mutated 1p/19q codeletion samples and 24 IDH mutated 1p/19q non-codeletion samples. The antibodies and reagents used in this study are listed in Table S2. The specimens were analyzed using a Nikon Eclipse E800 (10×/0.45 or 20×/0.75) microscope, and the images were captured using NIS software (Nikon). CD68 and CD204 staining was quantified according to positive fractions by Image J. M-CSF and TGF-β1 staining was semi-quantified according to staining intensity on a scale from 0 to 2. All the quantifications were done in a blinded manner.

In vitro stimulation of microglia and monocytic cells

The human acute monocytic leukemia cell line THP-1 and human microglia cell line HMC3 were purchased from the Chinese Academy of Sciences Cell Bank and the ATCC, respectively. THP-1 and HMC3 cells were treated with human TGF-β1 (10 ng/ml, Peprotech), human M-CSF (10 ng/ml, Peprotech) or a combination of TGF-β1 and M-CSF diluted in medium daily for 3 days. Next, the cells were harvested for RNA extraction and cDNA synthesis. Genes associated with a M2-like phenotype were analyzed by qRT-PCR. The primer sequences used are listed in Table S3. Conditioned media were collected from the treated cells, centrifuged to remove cellular material and next subjected to an enzyme-linked immunosorbent assay (ELISA) for human IL10 (R&D Systems).

Tumor cell culture and in vitro proliferation assay

The GL261 murine glioma cell line was a gift from G. Safrany, National Research Institute for Radiobiology and Radiohygiene (NRIRR), Budapest, Hungaria. The cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM; Gibco) supplemented with 10% fetal calf serum (FCS; Sigma). For proliferation assessment, GL261 cells (5 × 10⁴) were seeded in 10 cm dishes and treated with the TGF-β pathway inhibitor galunisertib (1 μM, MedChemExpress) or the CSF1R pathway inhibitor pexidartinib (5 μM, MedChemExpress), diluted in full medium daily. Cells were counted every day for up to 6 days using a cell counter (Beckman).

Orthotopic GL261 glioma model

All animal experiments were conducted in compliance with relevant laws and institutional guidelines, and approved by the Animal Experiment Ethical Committee of Shaanxi Normal University. Six-week-old C57BL/6 female mice were provided by the Experimental Animals Centre, Xi’an Jiaotong University. During injection, mice were anesthetized with 2.5% isoflurane. GL261 cells (1 × 10⁴) in 2 μl Dulbecco’s phosphate-buffered saline (PBS; Gibco) were subsequently injected stereotactically 1 mm anterior to bregma, 1.5 mm from the mid-line, and 2.7 mm below the cranial surface using a microliter syringe (Hamilton). After withdrawal of the needle, the incision was sutured and the mice were placed on a pre-warmed pad until full recovery from anesthesia. Tumor-bearing mice were administered daily with galunisertib (150 mg/kg) or pexidartinib (100 mg/kg), formulated in 0.5% methylcellulose/0.5% Tween80, by oral gavage starting from 5 days after tumor inoculation. 10 days after treatment, tumor-bearing mice were euthanized, after which the brains were harvested for analysis. For survival studies, mice were euthanized when the tumors elicited symptoms such as hunched posture, lethargy, persistent recumbency and/or weight loss of more than 10%.

Immunohistochemistry analysis of TAM infiltration in GL261 tumors

Immunohistochemical staining of CD68 and CD206 was performed, as previously described [28], using 6 μm sections of snap-frozen GL216 tumor tissues embedded in OCT medium (Tissue-Tek Sakura). The antibodies and reagents used are listed in Table S2. The images were acquired using NIS software (Nikon). CD68 and CD206 staining was quantified according to positive fractions by Image J.

RNA isolation, cDNA synthesis and qRT-PCR

Total RNA was extracted using TRIzol (Thermo Fisher Scientific), after which cDNA was synthesized using hexamer primers and SuperScript III transcriptase (Thermo Fisher Scientific), according to the manufacturer’s instructions. Quantitative PCRs were performed using a SYBR Green PCR Master Mix (Thermo Fisher Scientific). Gene expression levels were determined relative to Hprt (HPRT) or Cd68 according to the following formula: Relative expression _{gene X} = 2^-(CT_Hprt/Cd68 – CT _{gene X}). The primer sequences used are listed in Table S3.

List of macrophage markers used in the study

The pan-macrophage marker used was CD68. In addition, we used the M1 macrophage markers CD86 and TNF-α (encoded by TNF), and the M2 macrophage markers CD163, CD204 (encoded by MSR1) and CD206 (encoded by MRC1).

Results

Characterization of TAMs in IDH mutated LGGs

To characterize TAMs in IDH mutated LGGs at a single cell level, we analyzed published datasets of single cell RNA-seq (scRNA-seq) from 3 codeletion and 4 non-codeletion LGG tumors harboring IDH mutations [23, 24]. Three main types of cells including tumor cells, oligodendrocytes and macrophage/microglia cells were identified using scRNA-seq analysis (Fig. 1A). 709 cells were identified as TAMs from the whole-tumor scRNA-seq data, and clearly separated from the bulk of tumor cells (Fig. 1A). Interestingly, t-distributed stochastic neighbor embedding (tSNE) analysis of TAMs revealed distinct characteristics of each patient’s tumor (Fig. 1B). TAMs in gliomas can be derived from either peripheral circulating monocytes or brain resident microglia cells [4, 29, 30]. To investigate the origin of TAMs in LGGs, we generated a macrophage signature score and a microglia signature score for each cell using a previously defined gene signature that could discriminate monocyte-derived TAMs from microglia-derived TAMs in glioblastoma [31]. Plotting of the macrophage signature scores versus the microglia signature scores revealed an anti-correlation pattern at the single cell level (correlation analysis: r = −0.3847; p < 0.05) (Fig. S1A-B). The TAMs from codeletion LGGs and non-codeletion LGGs could not be discriminated based on macrophage and microglia scores, indicating a limited effect of the 1p/19q co-deletion status on the cellular origin of TAMs (Fig. S1A-B).

Fig. 1.

Characterization of TAMs in LGGs. (a) t-distributed stochastic neighbor embedding (t-SNE) plot of all cells in 7 IDH mutated LGGs colored by individual patients and 1p/19q co-deletion status (top-right corner). (b) t-SNE plot of 709 TAMs colored by individual patients (left panel) or by 1p/19q co-deletion status (right panel). (c) Bar plot of CD86 and MRC1 expression in 709 individual TAMs from IDH mutated LGGs. (d) Bar plot of CD86 and CD163 expression in 709 individual TAMs from IDH mutated LGGs. (e) tSNE blot of TAMs colored by positive CD86 or MRC1 expression

Interestingly, we found that most of the TAMs expressed M1 activation markers (e.g. TNF, CD86), regardless of the 1p/19q co-deletion status (Fig. S2A-B). For example, CD86 expression was detected in 81.7% of the TAMs from 1p/19q codeletion LGG tumors and 79.7% of the TAMs from 1p/19q non-codeletion LGG tumors. Expression of TNF and CD86 was hardly detected in tumor cells, indicating that TNF and CD86 expression are specific for TAMs (Fig. S2C, S2D). Moreover, a subset of individual TAMs frequently co-expressed M1 (CD86) and M2 (CD163, MRC1) activation markers (Fig. 1C-D), whereas 44% of the TAMs expressed both the M1 marker CD86 and the M2 marker MRC1 (Fig. 1C, E). These results indicate that most of the TAMs in IDH mutated tumors express M1 activation markers, whereas a subset of TAMs simultaneously co-expresses both M1 and M2 activation markers at the single cell level.

1p/19q co-deletion status correlates with distinct TAM infiltration

To determine putative associations of TAM phenotypes with 1p/19q co-deletion status in IDH mutated LGGs, we evaluated TAMs in 86 1p/19q codeletion LGGs and 144 1p/19q non-codeletion samples in the TCGA LGG dataset using the CIBERSORT in silico cytometry method [25]. We found that the M1 macrophage signatures were similar in the 1p/19q codeletion and non-codeletion LGG subtypes (Fig. 2A). Notably, we found that the M2 macrophage signature was significantly increased in the 1p/19q non-codeletion LGG samples (mean fraction: 0.33%) compared to the 1p/19q codeletion LGG samples (mean fraction: 0.28%) (Student’s t test; p < 0.01) (Fig. 2B). To further substantiate the association of 1p/19q co-deletion status with TAM infiltration, we performed immunostaining of CD68 and CD204 to detect total TAMs and TAMs bearing M2-related markers in 22 IDH mut-codel and 24 IDH mut-noncodel LGG samples (Table S1). We found that there was no difference in total TAM (CD68⁺) infiltration between the two subtypes (positive area: 1.19% in codeletion subtype; 1.28% in non-codeletion subtype; Student’s t test; p = 0.56) (Fig. 2C-D). In agreement with our bioinformatics analysis, quantification of CD204 staining revealed a significantly higher prevalence of TAMs bearing M2-related markers in the 1p/19q non-codeletion LGG subtype (positive area: 0.47%) compared to the 1p/19q codeletion LGG subtype (positive area: 0.26%) (Student’s t test; p < 0.01) (Fig. 2E-F). Taken together, these results indicate that 1p/19q non-codeletion status is associated with an increased polarization of TAMs toward a M2 phenotype.

Fig. 2.

1p/19q co-deletion status is associated with distinct TAM infiltration. Comparison of M1 macrophage (a) and M2 macrophage (b) fractions between 1p/19q codeletion and non-codeletion LGGs in the TCGA LGG dataset. Immune cell fractions were estimated using CIBERSORT algorithm. Immunohistochemical staining and quantification of CD68 (c, d) and CD204 (e, f) in 1p/19q codeletion (n = 22) and 1p/19q non-codeletion (n = 24) IDH mutated LGGs (scale bar = 100 μm)

Increased M-CSF and TGFβ1 expression in 1p/19q non-codeletion LGG samples

To uncover the mechanism underlying the association between 1p/19q co-deletion status and altered TAM phenotype in IDH mutated LGGs, we focused on tumor-derived factors (cytokines, chemokines and growth factors) that are encoded by genes located in either chromosome 1p or chromosome 19q, and which can alter the macrophage phenotype. M-CSF, encoded by CSF1 located in chromosome 1p, and TGF-β1, encoded by TGFB1 located in chromosome 19q, have previously been suggested to polarize macrophages to a M2 phenotype [32–34]. We analyzed the expression levels of CSF1 and TGFB1 in 1p/19q non-codeletion and 1p/19q codeletion LGG samples from the TCGA and CGGA databases, and found that the CSF1 and TGFB1 expression levels, likely due to the respective chromosomal deletions, were significantly lower in the 1p/19q codeletion LGG samples (Fig. 3A-D). This finding was confirmed by M-CSF and TGFβ1 immuno-staining in lower grade glioma sections, showing that the staining scores were significantly higher in the 1p/19q non-codeletion samples than in the 1p/19q codeletion samples (Fig. 3E-H). Moreover, CSF-1R (M-CSF receptor) and TGFBR2 (TGFβ1 receptor) expression was exclusively detected in the TAMs, but not in the oligodendrocytes or tumor cells, in the scRNA-seq analyses (Fig. 3I-J). In contrast to the lower expression of CSF1 in 1p/19q co-deletion LGGs we found that IL34, another ligand for CSF-1R and located in chromosome 16, was upregulated in the 1p/19q codeletion LGGs compared to the non-codeletion LGGs (Fig. S3A-B). Conversely, IL34 expression could hardly be detected at the single tumor cell level compared to the substantial CSF1 and TGFB1 levels observed in the tumor cells (Fig. S3C-E). Taken together, these results indicate that the expression levels of M-CSF and TGFβ1 are significantly decreased in 1p/19q codeletion LGG samples, suggesting a mechanistic connection between 1p/19q co-deletion and altered TAM polarization.

Fig. 3.

Higher M-CSF and TGFβ1 expression in 1p/19q non-codeletion LGG samples. (a-d) RNA levels of CSF1 and TGFB1 in IDH mutated LGG samples from the TCGA or CGGA database with or without 1p/19q co-deletion. Immunohistochemical staining and quantification of M-CSF (e-f) and TGFβ1 (g-h) in 1p/19q codeletion (n = 22) and 1p/19q non-codeletion (n = 24) IDH mutated LGGs (scale bar = 100 μm). t-SNE blot of whole tumor cells in 7 IDH mutated LGGs colored by expression of CSF1R (i) and TGFBR2 (j). RNA levels of IL10 (k), CCL2 (l), CCL22 (m) in THP-1 cells with the indicated treatment detected by qRT-PCR. (n) Concentration of IL-10 in conditioned media of THP-1 cells with the indicated treatments detected by ELISA

M-CSF and TGFβ1 can modulate the monocyte phenotype

Previously, it has been shown that TAMs in glioma can be derived from either peripheral circulating monocytes or brain resident microglial cells [4, 29, 30]. To evaluate the effects of M-CSF and TGFβ1 on determining the macrophage phenotype, we assessed their ability to polarize undifferentiated human monocyte THP-1 cells and human microglia HMC3 cells. To this end, THP-1 and HMC3 cells were stimulated with M-CSF, TGFβ1 or a combination of M-CSF and TGFβ1, after which the expression of IL10, CCL2 and CCL22 was analyzed by qRT-PCR. IL-10, CCL2 and CCL22 are known M2-like cytokines/chemokines that are secreted by TAMs to promote tumor growth [35, 36]. We found that the monocytic (THP-1) cells exposed to TGFβ1 alone did not alter M2-like gene expression (Fig. 3K-M). We also found, however, that M-CSF alone was sufficient to upregulate IL10, CCL2 and CCL22 expression in monocytic cells (Fig. 3K-M). Interestingly, the IL10 and CCL22 expression levels were further increased after combination treatment (Fig. 3K-M). Upregulation of IL-10 expression in THP-1 cells by M-CSF and M-CSF/TGFβ1 combination treatment was validated in conditioned media from treated cells by ELISA (Fig. 3N). In microglia (HMC3) cells, only a slight upregulation of IL10 was observed upon M-CSF treatment (Fig. S4A). The expression of CCL2 and CCL22 was not affected by M-CSF or TGFβ1 in microglia cells (Fig. S4B-C). Taken together, these results indicate that M-CSF and TGFβ1 can modulate monocytic cells, but not microglia cells towards a M2-like phenotype. The concomitantly observed higher levels of M-CSF and TGFβ1 in 1p/19q non-codeletion LGG samples suggests a mechanistic connection between 1p/19q co-deletion and altered TAM polarization.

Inhibition of CSF-1R and TGFβ signaling hampers tumor progression and modulates the TAM phenotype

To determine the potential effect of CSF-1R and TGFβ signaling on TAM regulation during glioma progression, we employed a GL261 syngeneic orthotopic glioma model since no IDH-mutated LGG models are currently available. GL261 tumor-bearing mice were treated with galunisertib, a potent TGFβ receptor kinase inhibitor [37], as well as with pexidartinib, a CSF-1R signaling inhibitor [38]. We found that single pexidartinib treatment significantly improved the survival of tumor-bearing mice, while galunisertib treatment alone did not show such a benefit (Fig. 4A). Notably, an enhanced therapeutic effect was observed when pexidartinib and galunisertib where combined, indicating that inhibition of CSF-1R and TGFβ receptor signaling synergistically abrogates in vivo glioma tumor growth. Of note, in vitro administration of pexidartinib and galunisertib did not decrease the proliferation of GL261 cells (Fig. S5). The observation that glioma tumor growth was suppressed in vivo but not in vitro by CSF-1R or CSF-1R/ TGFβ inhibition suggests that these treatments likely exert their effects through regulating the tumor microenvironment.

Fig. 4.

CSF-1R and TGFβ signaling inhibition hampers tumor progression and modulates TAM phenotype. (a) Survival curve of GL261 tumor-bearing mice treated with galunisertib, pexidartinib or a combination of galunisertib and pexidartinib. Median survival time: control group, 25.5 days; galunisertib group, 27 days; pexidartinib group, 30 days; combination group, 34.5 days (n = 10 mice per group). Immunohistochemical staining and quantification of CD68 (b-c) and CD206 (d-e) in GL261 tumors with the indicated treatments. (f) Proportion of CD206-positive to CD68-positive areas in tumors with the indicated treatments. Quantification of mRNA expression of M2-related genes, including Il10, Arg1, Ccl2, Cd163 and Ccl22 in treated gliomas normalized to Hprt (g) and Cd68 (h) (n = 4 per group). Survival of IDH mutated LGG patients retrieved from the CGGA dataset with low or high CSF1(i) or TGFB1(j) expression levels

Since pexidartinib is a macrophage-targeted drug, we examined total TAM and M2-like TAM infiltration in the treated tumors by CD68 and CD206 immunostaining, respectively. We found that both total TAM infiltration (positive area of CD68 staining: 4.01% in control, 3.52% in galunisertib group, 3.04% in pexidartinib group, 2.88% in combination group; ANOVA with Tukey’s multiple comparisons test; p < 0.01) and M2-like TAM infiltration (positive area of CD206 staining: 1.82% in control, 1.39% in galunisertib group, 1.15% in pexidartinib group, 0.85% in combination group; ANOVA with Tukey’s multiple comparisons test; p < 0.05) were significantly reduced in tumors upon treatment with pexidartinib alone or in combination with galunisertib (Fig. 4B-E). Importantly, the ratio of CD206 to CD68 was significantly reduced in the combination treatment group (ratio of CD206 to CD68: 44.28% in control, 35.34% in galunisertib group, 35.64% in pexidartinib group, 28.77% in combination group; ANOVA with Tukey’s multiple comparisons test; p < 0.05), indicating that CSF-1R and TGFβ inhibition synergistically regulates the TAM phenotype in GL261 glioma tumors (Fig. 4F).

We additionally analyzed the expression of a panel of genes (Il10, Arg1, Ccl2, Cd163 and Ccl22) associated with the M2-like phenotype in macrophages/microglia cells using RNA extracted from total glioma tumor tissues from treated and control mice. We found that Il10, Arg1 and Ccl2 were significantly downregulated in the combination group (Fig. 4G). To rule out the possibility that the reduced M2-related gene expression was attributed to a lower infiltration of TAMs in tumors in the combination group, we normalized the M2-related gene expression levels to that of Cd68. In doing so, we found that the expression of Il10 and Arg1 was still significantly lower in the combination treatment group compared to the control group (Fig. 4H), which indicates that concomitant CSF-1R and TGFβ signaling inhibition regulates the TAM phenotype.

Based on the effective suppression of tumor growth by CSF-1R and TGFβ signaling inhibition, we next set out to assess whether CSF1 and TGFB1 expression correlates with patient survival. We found that the expression of CSF1 and TGFB1 did not correlate with patient survival in the TCGA dataset (Fig. S6A-B). We also analyzed the CGGA dataset which includes more patients (TCGA: n = 224; CGGA: n = 263) as well as more deceased cases (TCGA: n = 42; CGGA: n = 117). Notably, we found that higher CSF1 expression was significantly associated with a poor patient survival in this dataset (Fig. 4I). In addition, we found that patients with a high TGFB1 expression showed a trend towards a poor prognosis (p = 0.0606) (Fig. 4J). However, CSF1 and TGFB1 expression did not correlate to prognosis in a multivariate survival analysis when correcting for 1p/19q co-deletion status, which indicates that CSF1 and TGFB1 do not serve as independent prognostic biomarkers (Table S4).

Discussion

By employing bioinformatics analyses together with immuno-staining of patient samples, we found that the number of TAMs expressing M2-related markers was significantly higher in IDHmut-noncodel LGGs than in IDHmut-codel LGGs. The levels of TGFβ1 and M-CSF were found to be significantly higher in 1p/19q non-codeletion LGGs compared to 1p/19q codeletion LGGs. M-CSF and TGFβ1 stimulation sufficiently up-regulated M2-related gene expression in monocytic cells. Inhibition of M-CSF and/or TGFβ1 signaling decreased tumor growth and skewed the TAM phenotype in a glioma mouse model. Our results indicate that the 1p/19q co-deletion status is associated with a distinct TAM infiltration, which is likely mediated by M-CSF and TGFβ1 signaling.

It has been shown that TAMs in glioblastoma resemble inactivated M0 phenotype macrophages that lack immune activation [16, 17] or mixed M1/M2 phenotype macrophages with an elevated expression of both M1- and M2- related genes [39]. We found that most TAMs in IDH mutated LGGs expressed M1 activation markers (TNF, CD86). Distinction of TAM phenotypes between LGGs and glioblastomas warrants further investigation. Individual TAMs in IDH mutated LGGs can co-express both M1 and M2 activation markers at the single cell level, which supports the notion that TAMs in IDH mutated LGGs have a phenotype distinct from canonical M1 or M2 activation phenotypes. Thus individual TAMs in IDH mutated LGGs should not be simplistically categorized to either M1 or M2 activation phenotype, but instead should be defined in a context-dependent manner. The observation that an increased M2 gene signature and higher numbers of TAMs expressing M2-related markers in 1p/19q non-codeletion LGGs compared to 1p/19q codeletion LGGs is indicative of a distinct phenotype, with enhanced M2-related gene expression by TAMs in 1p/19q non-codeletion LGGs rather than higher numbers of canonical activated M2 TAMs in 1p/19q non-codeletion LGGs.

The observed association of 1p/19q co-deletion status with an altered TAM phenotype in IDH mutant LGGs suggests an effect of tumor cells on their microenvironment. We observed lower levels of M-CSF and TGFβ1 expression in 1p/19q codeletion LGGs, which is consistent with their respective gene deletions in these LGGs. M-CSF is a cytokine that stimulates microglia and macrophages, and monocytes treated with M-CSF have been shown be skewed towards a M2 phenotype in vitro [40, 41]. M-CSF levels have been found to be elevated in glioblastoma, and overexpression of M-CSF in murine glioma cells to promote tumor growth and enhance TAM infiltration [42, 43]. Inhibition of CSF-1R signaling has been shown to block tumor growth and decreases M2-related gene expression in TAMs in glioma animal models [6, 20, 21, 38, 44]. Activation of TGFβ signaling promotes glioma growth through increasing the migration and invasion of tumor cells, and through regulating the tumor microenvironment [45, 46]. TAMs in glioma are derived from both local resident microglial cells and circulating monocytes [4, 30, 31]. We found that exposure of monocytes to M-CSF significantly increased M2-related gene expression, and this effect was further enhanced by an additional TGFβ1 treatment. In contrast, only a subtle effect of M-CSF/TGFβ1 on the induction of M2-related gene expression was detected in microglial cells. The distinct response of monocytes to M-CSF/TGFβ1 treatment suggests that monocyte-derived TAMs may be subject to differential conditioning in 1p/19q codeletion LGGs compared to 1p/19q non-codeletion LGGs, resulting in altered M2-related gene expression.

As there are currently no experimental tumor models available that can correctly recapitulate human IDH-mut glioma, we employed a GL261 mouse model which may be used to efficiently investigate the therapeutic effects of CSF-1R and TGFβ signaling inhibition. We found that co-inhibition of CSF-1R and TGFβ signaling pathways reduced GL261 glioma tumor progression and TAM infiltration and decreased M2-related gene expression, indicating that CSF-1R and TGFβ signaling are potential targets for TAM-targeting glioma therapy.

Taken together, our data indicate that 1p/19q co-deletion status associates with a distinct TAM phenotype with an altered M2-related gene expression. Whether 1p/19q co-deletion status can be used as a biomarker to predict the response to TAM-targeting therapy requires further investigation.

Abbreviations

1p/19q

Chromosome 1p and chromosome 19q

IDH

Isocitrate dehydrogenase

LGG

Lower-grade glioma

M-CSF

Macrophage colony-stimulating factor

scRNA-seq

single cell RNA-sequencing

TAM

Tumor-associated macrophage

TGFβ1

Transforming growth factor beta 1

Authors’ contributions

YZ, YX and LH designed and performed research, collected, analyzed and interpreted data and wrote the manuscript; JT and LC performed research, collected, analyzed and interpreted data and wrote the manuscript; QH and QC performed research and interpreted data; KH performed research; WG and AD interpreted data; LW and LZ designed research, analyzed and interpreted data, wrote the manuscript and supervised the study.

Funding

This work was supported by the National Natural Science Foundation of China (No. 81702489, 81911530166, 81772661, 81402081, 81870978), the “1000 Young Scholars” Program of Shaanxi Province, the National Key R&D Program of China (Grant No. 2018YFC1313003), the Natural Science Foundation of Shaanxi Province (No. 2020JQ-429), Fundamental Research Funds for the Central University (No. GK202003050, GK202003048, GK201603110), the Innovation Training Program for University Students (No. cx2019013, s201910718084) and the Natural Science Foundation of Huaihua city (2020R3118, 2020R3116).

Compliance with ethical standards

Ethics approval and consent to participate

Ethical permit of using patient samples was granted by the ethics committee of Shaanxi Normal University (EP 2017–910,221).

Consent for publication

This manuscript does not contain personal data in any form. All patient identifiers were removed prior to analysis.

Competing interests

The authors declare no conflicts of interest.