Background
Gliomas, characterized by immune evasive hallmarks, are the major primary tumors in the central nervous system (CNS) [
1]. The immune microenvironment of glioma is a complex neuroinflammatory network that involves both positive and negative immune regulation [
2]. T cells, the main executors in the anti-tumor immune response, are suppressed by various mechanisms at the tumor site [
3‐
6], among which PD-1/PD-L1 axis-mediated functional inhibition plays a key role. PD-L1 is an immune inhibitory receptor ligand expressed on many types of cancer cells, such as melanomas, lymphomas, lung cancers, prostate cancers, and gliomas [
7]. By binding to its receptor PD-1 expressed on the surface of activated T cells, PD-L1 leads to T cell dysfunction and apoptosis [
8,
9]. This facilitates the immunosuppressive microenvironment and tumor progression. Previously, studies have revealed that PD-L1 upregulation depended on IFN-γ-secreting CD8
+ lymphocytes [
10]. IFN-γ binds with receptor and subsequently activates JAK/STAT signaling pathway, which leads to the downstream expression and activation of IRF-1, further inducing PD-L1 expression on tumor cells [
11]. However, the driving factors of PD-L1 expression on various cells in the glioma microenvironment remain to be investigated.
Emerging evidence implies that PD-1/PD-L1 is a promising target to reverse the immune evasion of glioma [
12]. Nduom et al. [
13] measured PD-L1 expression in 94 patients and found that PD-L1 was a negative prognostic indicator for glioblastoma (GBM). Wang et al. [
14] analyzed 976 glioma samples with transcriptome data and concluded that PD-L1 expression was positively correlated with the WHO classification of glioma. While abundant clinical studies on anti-PD-1/PD-L1 antibody specific to gliomas are in progress, the results remain unclear. Based on completed clinical trials of anti-PD-1/PD-L1 therapy targeting other tumors, nevertheless, screening for appropriate patients is crucial for favorable prognosis [
15]. Although PD-L1 immunohistochemistry (IHC) has been approved by the FDA as the only predictive companion test for cancer immunotherapy such as pembrolizumab in non-small cell lung cancer patients, supplementary clinical indicators are urgently needed considering the high false negative rate [
16]. Until now, biomarkers identifying possible responsive glioma patients have not been defined.
In this study, we investigated the distribution of T cells and PD-L1 expression on murine orthotopic glioma model and validated the results in human glioma samples from databases of the Cancer Genome Atlas (TCGA) and the Ivy Glioblastoma Atlas Project. We found that the distribution of PD-L1 in glioma coincides with morphologically apoptotic T cells and that IFN-γ induced PD-L1 expression on primary cultured microglia, bone marrow-derived macrophages (BMDM), and GL261 tumor cells, suggesting IFN-γ derived from tumor-infiltrating T cells may be the lead to induced PD-L1 expression in the microenvironment. We also found that, apart from the tumor cells previously reported, activated microglia and peripheral-derived macrophages in the microenvironment also present significant upregulation of PD-L1. Considering the importance of IFN-γ in inducing PD-L1 in the glioma microenvironment, it is assumed as a supplementary indicator to predict the expression of PD-L1. However, traditional IHC or RNA seq methods are insufficient to accurately measure the IFN-γ level in tumor samples. Here, we proposed IFN-γ score, aggregated from the expressions of seven IFN-γ-induced genes, as an ancillary marker in screening for appropriate glioma patients.
Methods
Mice
C57BL/6 mice (6–8 weeks) were purchased from Shanghai Slac Laboratory Animal Co., Ltd. (Shanghai, China). Mice were maintained under the specific pathogen-free condition and housed in the Animal Facility of Fudan University (Shanghai, China) according to the Guidelines for the Care and Use of Laboratory Animals (No. 55 issued by the Ministry of Health, People’s Republic of China, on January 25, 1998), as administered by the Institutional Animal Care and Use Committee (IACUC) of Fudan University.
GL261 murine glioma model
GL261 murine glioma cell line was kindly provided by Dr. Liangfu Zhou (Huashan Hospital, Shanghai, China). GL261 was cultured in DMEM/F12 (Thermo Fisher, USA) supplemented with 10% heat-inactivated FBS (Thermo Fisher, USA), 2 mM glutamine (Thermo Fisher, USA), 100 U/ml penicillin (Thermo Fisher, USA), and 100 μg/ml streptomycin (Thermo Fisher, USA). Cells were maintained in the incubator at 37 °C in a humidified 5%CO2/95% atmosphere with routine checks for mycoplasma contamination every 3 months. For tumor inoculation, anesthetized mice were immobilized and mounted onto a stereotactic head holder in the flat-skull position. The skin of the skull was dissected in the midline by a scalpel. The skull was carefully drilled with a 20-gauge needle tip (ML + 2.0; RC + 1.0 mm). Then, a microliter Hamilton syringe was inserted to a depth of 3 mm and retracted to a depth of 2.5 mm from the dural surface. Five microliters (2 × 104 cells/μl) of cell suspension or PBS was slowly injected in 2 min. The needle was then slowly taken out from the injection canal, and the skin was sutured. Terminal stage of GL261 murine glioma model was defined by agonal symptoms such as poor grooming, lethargy, weight loss, or seizures.
Primary adult microglia culture
Microglia were prepared from 6- to 8-week-old mice as described previously [
17]. Briefly, the brains were dissected with the cerebella and olfactory bulbs taken off. The tissue was triturated mechanically and washed with PBS by centrifuging for 7 min at 500
g, 4 °C. The supernatant was discarded, and pellets were re-suspended in 37% Percoll. Percoll gradients (70%/37%/30%/0%) were prepared and centrifuged for 5 min at 500
g, 18 °C (low acceleration, brake off). Mononuclear cells were collected at 70%/37% Percoll interface. Microglia were enriched by CD11b microbeads (BD Bioscience, USA) according to the manufacturer’s specification and harvested for purity check and further tests. Isolated microglia were plated onto 24-well plates (1 × 10
5 cells per well) and cultured in basic medium with additional 5 ng/ml recombination TGF-β1 (Miltenyi, Germany) and 10 ng/ml Recombinant Mouse M-CSF Protein (R&D, USA). Half of the medium was changed every 3 days, for a total of 10–14 days.
For T cell co-culture assay, adult microglia were plated onto 96-well plates at a density of 1 × 105 cells per well. Half of the medium was changed every 3 days, for a total of 7 days. On day 8, microglia were treated with or without 20% GCM for 24 h. The CD4+ T Cell Isolation Kit (Miltenyi, Germany) was used for purification of CD4+ T cells from the spleen of OT II mice. CD4+ T cells were stained with CFSE dye (Invitrogen, USA) following the manufacturer’s instructions. The microglia were washed with PBS for three times and then co-cultured with CD4+ T cells (4 × 105 cells per well) for 4 days supplied with 0.1 μM OVA323–339 peptides (Sigma-Aldrich, USA). After co-culture, both microglia and T cells were determined by flow cytometric analysis.
Immunofluorescence
For immunofluorescence, sections were thawed and dried at room temperature and rinsed in PBS. For fixation, cells were washed with PBS and followed by 4% PFA for 5 min. Samples were permeabilized with 0.25% Triton X-100 for 15 min and blocked in blocking buffer containing 10% donkey serum for 2 h at room temperature or overnight at 4 °C. Then, samples were incubated with indicated primary antibodies (Additional file
1: Table S2) overnight at 4 °C. Samples were then washed with PBS and incubated with the appropriate fluorophore-conjugated secondary antibodies, namely Alexafloure-488, 594 (Thermo Fisher, USA) and Cy3 (JacksonImmunoResearch Laboratory, USA), at a dilution of 1:500 in 1% BSA for 1 h at room temperature. 4′, 6-Diamidino-2-phenylindole (DAPI) was used as a counterstain. Images were acquired by a fluorescence microscope Olympus IX73 (Olympus, Japan). Appropriate gain and black level settings were determined by control tissues stained with secondary antibodies. Analyses of images were performed using ImageJ software (NIH, USA). Quantitative analysis was performed with ImageJ to determine the T cell counts and mean intensity of PD-L1. For CD4
+ cell or CD8
+ cell counts, data were collected from five random fields for each region per mouse,
n = 4. For mean intensity of PD-L1, data were collected from at least three and up to seven random fields for each region per mouse,
n = 5.
FACS analyses
For fluorescence-activated cell sorting (FACS) analysis of brain tumor-infiltrating immune cells, mice were euthanized at the defined endpoint. Mononuclear cells in the brains were isolated as previously described and stained afterward with the respective antibodies for FACS analysis. For flow cytometry, cells were counted and incubated with Fc blocker (eBiosciences, USA) for 30 min, followed by another 30-min incubation with conjugated antibodies for extracellular markers. For intracellular cytokine detection, cells were stimulated in vitro with Cell Stimulation Cocktail (eBiosciences, USA) for 5 h at 37 °C before FACS analysis. After stimulation, cells were stained for surface markers and cytokines with Intracellular Fixation and Permeabilization Buffer Set (eBiosciences, USA). All antibodies used for these experiments were listed in Additional file
1: Table S2. Proper isotype controls and compensation controls were performed in parallel. BD Biosciences Canto II (BD Biosciences, USA) was used for flow cytometry. FlowJo software (Tree Star, USA) was used for FACS data analysis.
Quantitative real-time PCR
Total RNA was isolated with RNAiso (Takara, Japan) following the manufacturer’s protocol and reversely transcribed using PrimeScript™ RT reagent Kit with gDNA Eraser (Perfect Real Time) (Takara, Japan). Gene expression was detected using SYBR® Premix Ex TaqTM II (Tli RNaseH Plus) Kit (Takara, Japan). All RT-PCR amplifications were performed in triplicates in a 20-μl reaction volume with the indicated primer pairs. Primer sequences were listed in Additional file
2: Table S1. RT-PCRs were performed using 7500 Fast Real-Time PCR System (Applied Biosystems, USA). The amount of target mRNA was normalized to the expression level of β-actin generated from the same sample and subsequently to controls. Relative expression was calculated as 2
−ΔCt.
IFN-γ score calculation and clinical data analysis
Firstly, 34 genes were sorted out by filtering genes from GO term: response to interferon-gamma (accession GO: 0034341, organism: Homo sapiens) with genes that were positively correlated with PD-L1 expression (p < 0.05; r > 0.5) from the TCGA lower grade glioma (LGG)/GBM datasets. Then, further crossing 34 genes with 840 genes that were positively correlated with PD-L1 expression (p < 0.05; r > 0.3) from the Ivy Glioblastoma Atlas Project, 7 genes were eventually sorted out, namely GBP5, ICAM1, CAMK2D, IRF1, SOCS3, CD44, and CCL2. Combining the relative expression levels of the sorted seven genes, IFN-γ score was calculated as a substitute indicator for IFN-γ level. Myeloid cell-related genes (CD14, CD33, CD36, CD68, CX3CR1, ENG, ITGAL, and ITGAM) were used to calculate myeloid cell score. T cell-related genes (CD2, CD3D, CD3E, CD3G, CD4, CD8A, CD8B, CD28, CCR7, and IL2RA) were used to calculate T cell score. The median value of IFN-γ score was used as the cutoff to divide patients with high IFN-γ score and patients with low IFN-γ score.
Data presentation and statistical analysis
GraphPad Prism 6.0 (GraphPad Software Inc., USA) was used for all data analysis. Parametric data were presented as mean ± standard error of the mean (SEM). Differences between two groups were analyzed using Student’s unpaired t test. Analysis of variance (ANOVA) was used to compare multiple groups, and Pearson’s correlation coefficient was used to analyze the correlation of the expression levels of genes. Statistical significance was determined at p < 0.05 in all cases.
Discussion
Our study identified the distribution of PD-L1 in gliomas and that, apart from tumor cells in the tumor microenvironment, significantly increased PD-L1 expression was also spotted on activated microglia and peripheral-derived myeloid cells. Besides, we provided some evidence that IFN-γ played an important role in inducing the expression of PD-L1 in gliomas. IFN-γ score, aggregated from expression of IFN-γ downstream genes as a substitute for the abundance of IFN-γ, is expected to serve as an auxiliary prognostic indicator for screening potential PD-1/PD-L1 antibody drug-applicable glioma patients.
Previous studies have focused on the mechanisms of PD-L1 expression in tumor cells, which include tumor endogenous proto-oncogenic signal, such as abnormal PI3K/Akt signaling pathway [
21], and adaptive immune resistance, specifically the magnified negative feedback of the immune system that originally prevents over-activated immune cells from damaging the tissue [
22,
23]. In gliomas, the latter mechanism may play a greater role in the expression of PD-L1 in the microenvironment. T cells are activated in the local region of tumor and thus secrete IFN-γ [
24‐
26], which can subsequently induce upregulation of PD-L1 in tumor cells and immune cells in the microenvironment [
11,
27], thereby inhibiting tumor eradication led by T cells. Notably, IFN-γ in the tumor microenvironment comes not only from T cells but also from NK cells. The vicious effect of this negative feedback may be more pronounced in the CNS. The microglia, astrocytes, neurons, and epithelial cells in the CNS can be induced by IFN-γ and upregulate PD-L1 expression [
28‐
30], which may exacerbate T cell dysfunction and apoptosis in gliomas. Such IFN-γ/PD-L1 axis-mediated immune suppression that also exists in the normal tissue inadvertently promotes glioma immune escape.
In the GL261 glioma model, we found that the expression of PD-1 was elevated in both CD4
+ and CD8
+ T cells. Previous studies have revealed that PD-L1 expressed by tumor cells suppresses cytotoxic activity of tumor-infiltrating CTLs [
31‐
33]. In addition, we found that glioma-infiltrated antigen-presenting cells (microglia and peripheral-derived macrophages) overexpressed PD-L1. The abovementioned suggests the importance of PD-1/PD-L1 axis on the functional inhibition of CD4
+ cells in glioma. It is indicated that the adaptive immune resistance not only occurs as the inhibition of CTL-mediated tumoricidal activity, but also as the activation of CD4
+ helper T cells in tumors, thus fundamentally affecting the entire tumor-immune microenvironment networks and disrupting the formation of anti-tumor immune microenvironments.
In the study, we proposed IFN-γ score as a complementary predictive biomarker, which is aggregated from the expression of seven selected genes. IRF1, directly participating in the regulation of PD-L1 expression, is essential in the constitutive and IFN-γ-induced expression of PD-L1 in various cancer cells [
34‐
36]. A recent study confirmed that PD-L1 expression in melanoma cells is mainly regulated by the IFN-γ receptor signaling pathway which subsequently converged to the binding of IRF1 with the PD-L1 promoter [
11]. Upregulation of GBP5 has been recognized in colon cancer [
37]. In gastric cancer, positive correlation between immune cell infiltration and stromal and epithelial GBP5 expression has been reported [
38]. A previous study demonstrated the role of GBP protein in the formation of inflammasome complex as well as its anti-inflammatory and autoimmunity-controlling effect [
39], yet the specific function of the GBP protein family remains to be discovered. SOCS3, promoted by IFN-γ downstream genes STAT1 or STAT3 [
40,
41], is known as a negative regulator of cytokine signaling and participant in control of CNS immunity [
42,
43]. CCL2 is the key chemokine that recruits myeloid-derived cells to the tumor microenvironment in glioma [
44]. According to previous studies, peripherally derived myeloid cells usually perform immunosuppressive functions in gliomas [
45‐
47]. Besides, the expression levels of all seven genes were negatively correlated with the survival of glioma patients. All the evidence indicated that the expression of PD-L1 and other immune inhibitory mechanisms in the glioma microenvironment might serve as negative feedback mechanisms that followed, rather than preceded, T cell activation and IFN-γ secretion.