Tumor associated microglia/macrophages utilize GPNMB to promote tumor growth and alter immune cell infiltration in glioma

Abstract

Tumor-associated microglia and blood-derived macrophages (TAMs) play a central role in modulating the immune suppressive microenvironment in glioma. Here, we show that GPNMB is predominantly expressed by TAMs in human glioblastoma multiforme and the murine RCAS-PDGFb high grade glioma model. Loss of GPNMB in the in vivo tumor microenvironment results in significantly smaller tumor volumes and generates a pro-inflammatory innate and adaptive immune cell microenvironment. The impact of host-derived GPNMB on tumor growth was confirmed in two distinct murine glioma cell lines in organotypic brain slices from GPNMB-KO and control mice. Using published data bases of human glioma, the elevated levels in TAMs could be confirmed and the GPNMB expression correlated with a poorer survival.

Keywords: CD44; Experimental glioma; GPNMB; Glioblastoma; Macrophage; Microglia; Mouse; RCAS.

Fig. 1

Depletion of host-derived GPNMB impairs glioma formation and proliferation in vivo. A Disease progression score of tumor-bearing WT and KO mice injected with PDGFb-driven RCAS-RFP glioma cells. Animals were sacrificed upon the first animal reaching disease score 3 (day 50). B Representative macroscopic image of a tumor-bearing brain from WT (top) and KO (bottom) mice. The white dotted line marks the tumor in WT and in KO. Scale bars represent 5 mm. C RFP⁺ fluorescence (grey scaled for visibility) labeling of glioma cells of a representative brain slice from a tumor-bearing WT (top) and KO (bottom) mice. The white dotted line marks the whole brain tissue. Scale bars represent 2 mm. D RFP⁺ fluorescence (grey scaled for visibility) of tumor cells at the injection site of a representative KO brain slice. Scale bar represents 500 µm. E Tumor volume of tumor-bearing WT (n = 7) and KO (n = 8) mice. Black dotted line represents the mean value of each group. Statistical analysis was performed using unpaired t-test. F Representative staining of Ki-67 (left) and merged with DAPI (right; Ki-67 = green, DAPI = blue) of brain slices from tumor-bearing WT (n = 6) and KO (n = 7). Scale bar represents 50 µm. Statistical analysis was performed using an unpaired t-test. Error bars represent SD. *p < 0.05

Fig. 2

GPNMB is upregulated in the murine TME and is predominantly expressed by TAMs. A Representative FACS analysis of GPNMB⁺ populations in MG/TAMs (I), MonoLy6c^low (II), MonoLy6c^high (III), neutrophils (IV) and lymphocytes (V) in naïve brain and spleen tissue (top row) and tumor-bearing brain tissue (bottom row). B Quantification of GPNMB expression in non-immune cells (CD45⁻CD11b⁻). Comparison of naïve brain (n = 4) against tumor-bearing (n = 5) brain tissue. GPNMB expression in microglia (MG)/macrophages (MPH)/TAMs (I; CD45⁺CD11b⁺Ly6G⁻Ly6c⁻). Comparison between naïve brain (n = 4), naïve spleen (n = 4) and tumor brain (n = 5). GPNMB expression in monocytes with low Ly6c (II; CD45⁺CD11b⁺Ly6G⁻Ly6c^low), monocytes with high Ly6c (III; CD45⁺CD11b⁺Ly6G⁻Ly6c^high), neutrophils (IV; CD45⁺CD11b⁺Ly6G⁺Ly6c⁺) and lymphocytes (V; CD45⁺CD11b⁻). Comparison of naïve spleen (n = 4) against tumor-bearing (n = 5) brain tissue. Error bars represent SD. Statistical analysis in 3 groups was performed using Ordinary one-way ANOVA with Tukey’s multiple comparisons test and in 2 groups using unpaired t-test. *p < 0.05, **p  < 0.01, ***p < 0.001, ns, not significant. C Representative brain slices from tumor-bearing WT mice stained for IBA1 (top row), GPNMB (middle row) and merge with DAPI (bottom row; IBA1 = green, GPNMB = red, DAPI = blue). Left column: Lower magnifications of the invasive edge (white dotted line = invasive edge; left: brain tissue; right: tumor tissue). Scale bar represents 200 µm. Middle column: Higher magnifications of the invasive edge. Scale bar represents 50 µm. Right column: Representative GPNMB⁺/IBA1⁺ cells in the tumor core. Scale bar represents 20 µm

Fig. 3

GPNMB-KO mice display a higher density and soma size of microglia and accumulate pro-inflammatory TAMs. A Representative staining of IBA1 (left) and merge with DAPI (right; IBA1 = green, DAPI = blue) of brain slices from naïve WT (n = 3) and KO (n = 3). Scale bar represents 100 µm. The inserts show an expanded magnification of an individual microglia. B IBA1⁺ cell density normalized to DAPI in % and C soma volume (in µm³) as obtained from slices as shown in A. Statistical analysis was respectively performed using unpaired t-test. D Representative stainings of IBA1 (left) and merge with DAPI (right; IBA1 = green, DAPI = blue) of brain slices from tumor-bearing WT (n = 7) and KO (n = 7) in three different brain regions. Ipsilateral: Outside of the tumor in the ipsilateral hemisphere. Invasive edge (IE): Border region from tumor to non-tumor brain tissue, defined by the RFP signal and DAPI density. Core: Core tumor tissue area. Scale bar represents 40 µm. The inserts show an expanded magnification of an individual microglia. The bar graph on the right summarizes the percentage of IBA1⁺ cells normalized to the DAPI⁺ cells for these three tissue regions. Statistical analysis was performed using 2-way ANOVA with Sidak´s multiple comparisons test. E Representative staining of MHCII (left), IBA1 (middle) and merge with DAPI (right; MHCII = green, IBA1 = red, DAPI = blue) of brain slices from tumor-bearing WT (n = 7) and KO (n = 6). Scale bar represents 50 µm. Statistical analysis was performed using an unpaired t-test. Error bars represent SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Fig. 4

Loss of host-derived GPNMB promotes a pro-inflammatory tumor immune microenvironment in a murine glioma model. A Representative core region staining of CD3 (left), Ki-67 and merge with DAPI (right; CD3 = red, Ki-67 = green, DAPI = blue) of brain slices from tumor-bearing WT and KO. Scale bar represents 20 µm. The graphs (Top left, core, CD3⁺DAPI⁺: WT n = 7, KO n = 7; Top right, invasive edge (IE), CD3⁺DAPI⁺: WT n = 7, KO n = 7; Bottom left, core, CD3⁺Ki-67⁺DAPI⁺: WT n = 7, KO n = 7; Bottom right, IE, CD3⁺Ki-67⁺DAPI⁺: WT n = 7, KO n = 6) show the summary data from WT and KO mice. B Representative core region staining of CD8 (left) and merge with DAPI (right; CD8 = green, DAPI = blue) of brain slices from tumor-bearing WT and KO. Scale bar represents 40 µm. The graphs (left, core: WT n = 7, KO n = 7; right, IE: WT n = 7, KO n = 7) show the summary data from WT and KO mice. C Representative core region staining of Gzmb (left) and merge with DAPI (right; Gzmb = red, DAPI = blue) of brain slices from tumor-bearing WT and KO. Scale bar represents 40 µm. The graphs (left, core: WT n = 7, KO n = 7; right, IE: WT n = 7, KO n = 7) show the summary data from WT and KO mice. D Representative core region staining of Foxp3 (left) and merge with DAPI (right; Foxp3 = red, DAPI = blue) of brain slices from tumor-bearing WT and KO. Scale bar represents 40 µm. The graphs (left, core: WT n = 7, KO n = 7; right, IE: WT n = 7, KO n = 7) show the summary data from WT and KO mice. Statistical analysis was performed using an unpaired t-test. Error bars represent SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Fig. 5

Loss of host-derived GPNMB impairs tumor growth in OBS cultures injected with glioma cells with low and high intrinsic expression of GPNMB. A RT-qPCR of GPNMB expression relative to cultured murine astrocytes (n = 3) in cultured RCAS-PDGFb cells (n = 3) and the GL261 glioma cell line (n = 3). Statistical analysis was performed using one-way ANOVA with Tukey’s multiple comparisons test. B Western Blot of total GPNMB protein isolated from cultured RCAS-PDGFb, GL261 glioma cells, cultured MG (neonatal) and MG (adult). GAPDH expression serves as a reference. Each column represents an individual batch of culture (n = 3). C OBS from WT and KO animals injected with cultured PDGFb-driven RCAS-RFP glioma cells. Images are shown on the left. Top row: Orthogonal view on RFP immunofluorescence based on z-stack and tile scanning. Bottom row: representative 3D-reconstruction in Imaris based on z-stack and tile scanning. The graph on the right summarizes the volumes (in mm³) of WT (n = 11) and KO (n = 9) tumors as determined from the scans. D OBS injected with cultured mCherry-GL261 glioma cells. Images are shown on the left. Top row: Orthogonal view on mCherry immunofluorescence based on z-stack and tile scanning. Bottom row: representative 3D-reconstruction in Imaris based on z-stack and tile scanning. The graph on the right summarizes the volumes (in mm³) of WT (n = 30) and KO (n = 42) tumors as determined from the scans. The scale bars represent 500 µm. Statistical analysis was performed using unpaired t-test. Error bars represent SD. ****p < 0.0001, ns, not significant

Fig. 6

TAMs are the predominant source of GPNMB in resection tissue from GBM patients. A RT-qPCR of GPNMB expression in CD11b⁺ and CD11b⁻ cells separated from 9 patients with GBMs. Statistical analysis was performed using paired t-test. B Representative staining of patient-derived GBM and non-tumor slices stained for GPNMB (left), IBA1 (middle) and merge with DAPI (right; GPNMB = red, IBA1 = green, DAPI, blue). The image to the right shows a magnified view in the tumor slice. Scale bars, including the magnification, represent 20 µm. C Summary of the percentage of IBA1⁺/GPNMB⁺ cells of the samples from GBM (n = 9) and non-tumor (n = 3). Statistical analysis was performed using unpaired t-test. Error bars represent SD. D Pearson correlation of tumor-associated macrophage/microglia markers (y-axis) with GPNMB (x-axis). Top: CD204 (r = 0.84; p ≤ 0.0001), OPN (r = 0.81; p ≤ 0.0001) and CD68 (r = 0.79; p ≤ 0.0001). Middle: PTPRC/CD45 (r = 0.70; p ≤ 0.0001), CD163 (r = 0.70; p ≤ 0.0001) and CD204 (r = 0.59; p ≤ 0.0001). Bottom: HEXB (r = 0.64; p ≤ 0.0001), TMEM119 (r = 0.37; p ≤ 0.0001) and P2RY12 (r = 0.31; p ≤ 0.0001). Data derived from all primary GBM samples of the CGGA data set (n = 225). E GPNMB gene expression of non-immune cell population (CD45⁻), microglia (MG), macrophages (MPH) and neutrophils in human glioma with IDH wildtype (IDH^wt), IDH mutant (IDH^mut) and brain metastasis (BrM) using the Brain Tumor Immune Micro Environment dataset [18]. Unlabeled statistical analysis were performed in comparison to the non-immune cell population with 2way ANOVA and Tukey's multiple comparisons test. Error bars represent min to max. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. F The Rho value of correlation between uncommitted (M0, left), pro-inflammatory (M1, middle) and anti-inflammatory (M2, right) macrophage infiltration level (based on TIMER algorithm) and GPNMB gene expression

Fig. 7

High GPNMB expression in GBM are negatively prognostic for the disease course and positively correlated for the expression of immune checkpoint markers. A Kaplan–Meier survival curves of GBM patients based on GPNMB expression in the CGGA data set. Left: Cutoff set by median GPNMB expression into GPNMB^high (n = 188, events = 162) and GPNMB^low (n = 189, events = 152). Right: Cutoff by top/low 25% GPNMB expression set into GPNMB^high (top 25%; n = 94, events = 81) and GPNMB^low (low 25%; n = 94, events = 73). B Kaplan–Meier survival estimates of GBM patients based on GPNMB expression in the TCGA data set. Left: Cutoff set by median GPNMB expression into GPNMB^high (n = 258, events = 223) and GPNMB^low (n = 255, events = 212). Right: Cutoff by top/low 25% GPNMB expression set into GPNMB^high (top 25%; n = 128, events = 114) and GPNMB^low (low 25%; n = 128, events = 106). Statistical analysis was performed using Log-rank (Mantel–Cox) test, Gehan-Breslow-Wilcoxon test and Hazard Ratio (Mantel–Haenszel). C Gene expression of immune checkpoint markers (Top: PD-L1, PD-1, CTLA4, TIM3 and BTLA; Bottom: ICOS, ICOSLG, GATA3, CD47 and SIRP-alpha) in GPNMB^high (top 25%; n = 94) and GPNMB^low (low 25%; n = 94) primary GBM. Data from CGGA. D Representative core region staining of PD-1 (left), CD3 (middle) and merge with DAPI (right; PD-1 = yellow, CD3 = red, DAPI = blue) of brain slices from tumor-bearing WT and KO. Scale bar represents 20 µm. The graphs (left, core: WT n = 6, KO n = 7; right, IE: WT n = 6, KO n = 7) show the summary data from WT and KO mice. Statistical analysis was performed using unpaired t-test. Error bars represent SD. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001

Fig. 8

Tissue expression of CD44 is affected by GPNMB in the RCAS-PDGFb model and associated with GPNMB expression in human GBM. A Representative staining of CD44 (left) and merge with DAPI (right; CD44 = red, DAPI = blue) of brain slices from tumor-bearing WT and KO. Scale bar represents 50 µm. The graph on the right shows the summary data of the area of CD44 staining (in % of total area) from WT (n = 7) and KO (n = 6) mice. B Left: Gene expression of CD44 (Cutoff: median) in GPNMB^high (n = 111) and GPNMB^low (n = 111) primary GBM. Data from CGGA. Right: Gene expression of CD44 (Cutoff: median) in GPNMB^high (n = 262) and GPNMB^low (n = 263) primary GBM. Data from TCGA. C Left: Gene expression of CD44 in GPNMB^high (top 25%; n = 55) and GPNMB^low (low 25%; n = 54) primary GBM. Data from CGGA. Right: Gene expression of CD44 in GPNMB^high (top 25%; n = 130) and GPNMB^low (low 25%; n = 131) primary GBM. Data from TCGA. Statistical analysis was performed using unpaired t-test. Error bars represent SD. D Kaplan–Meier survival curves of GBM patients based on GPNMB expression in the CGGA (left) and TCGA (right) data set. Cutoff set by median of CD44 and GPNMB expression into GPNMB^highCD44^high (CGGA: n = 87, events = 77; TCGA: n = 169, events = 149) and GPNMB^lowCD44^low (CGGA: n = 86, events = 66; TCGA: n = 170, events = 137). *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001