Glial TIM-3 Modulates Immune Responses in the Brain Tumor Microenvironment

T-cell immunoglobulin and mucin domain–containing molecule 3 (TIM-3) is a transmembrane glycoprotein consisting of an immunoglobulin variable domain, a mucin domain, and a cytoplasmic tail with tyrosine phosphorylation motifs (1). TIM-3 was originally proposed as a T helper type 1 (Th1)-specific protein that is selectively expressed in terminally differentiated Th1 cells (2). However, subsequent studies have found that it is expressed on Th17 cells, regulatory T (Treg) cells, and even non-T cells such as dendritic cells (DC), natural killer (NK) cells, monocytes, and macrophages (1, 3). Accumulating evidence suggests that TIM-3 has diverse immune functions, which differ according to the specific cell type and cell status. Notably, TIM-3 appears to play both positive and negative roles in a context-dependent manner (1, 4–6). For example, TIM-3 can inhibit T-cell activation to dampen the responses of CD4⁺ and CD8⁺ T cells and induce peripheral tolerance. In certain cases, however, TIM-3 can participate in activating various innate immune cells, including quiescent macrophages, thereby contributing to eliminate pathologic stimuli (7, 8). A recent study showed that TIM-3 promotes short-term effector T cells during viral infection (9). In addition, it has been reported that dysregulation of TIM-3 expression results in either excessive or inadequate responses in immune cells (1, 7).

Experimental and clinical studies have implicated TIM-3 in a number of diseases, including autoimmune diseases, chronic infections, and ischemia. Alteration and dysregulation of TIM-3 expression have been associated with the onset and severity of the pathological status in patients with multiple sclerosis (MS) and those infected with human immune deficiency virus (HIV) or hepatitis virus (10, 11). TIM-3 has been correlated with high-level expression of IFNγ and TNFα in patients with MS, as well as the distortion of Th1-induced Th2 responses in allergic diseases (10, 12). Moreover, blockade of TIM-3 has been shown to affect the pathological severity of experimental allergic encephalomyelitis (EAE) and the onset of diabetes in nonobese diabetic (NOD) mice (2, 13). Jones and colleagues showed that blockade of TIM-3 signaling restored proliferation and increased cytokine production in HIV-specific T cells (11). In those TIM-3–associated diseases, negative regulation of Th1 or Th17 immunity by TIM-3 is suggested as a possible mechanism leading to pathologic conditions, such as T-cell dysfunction or exhaustion of T cells, distortion of Th2 responses to Th1 signaling, and proinflammatory states of Th1 responses (14). However, we know little regarding the detailed features and roles of TIM-3 in specific immune statuses and distinct diseases, particularly innate immunity-linked pathologic conditions.

In recent years, great deal of research has focused on immune checkpoint molecules and immune surveillance in the context of cancer growth and eradication. This has led to the development of numerous therapeutic strategies, including therapeutic antibodies targeting immune checkpoint molecules (15). TIM-3 has also gained growing attention as a possible target for the immunomodulation of cancer. Studies have shown that TIM-3 expression is associated with several cancers, and TIM-3 has been suggested to play a role in controlling tumor growth (16–18). TIM-3 is highly expressed on CD4⁺ and CD8⁺ tumor-infiltrating T cells from patients with non–small cell lung cancer, and TIM-3 expression on CD4⁺ T cells has been associated with poor clinicopathologic parameters, such as nodal metastasis and advanced cancer (19). Sakuishi and colleagues showed that numbers of intratumoral FOXP3⁺ Tregs express TIM-3 and that TIM-3⁺ Tregs are more immunosuppressive than TIM-3⁻ Tregs (20). In addition, recent research showed that TIM-3 is characteristically expressed on tumor cells in patients with clear cell renal carcinoma and hepatocellular carcinoma (21–23). Considering these multiple facets of TIM-3, it would be useful to clearly define the detailed characteristics of TIM-3 in the tumor microenvironment, as this could help inform new strategies for providing effective clinical approaches against tumors.

Our group has been investigating the characteristics and possible roles of TIM-3 in specific immune cells, particularly in innate immune cells, under pathophysiologic conditions of the brain (24). In this study, we questioned whether TIM-3 could be involved in immune surveillance against brain tumors, and thereby affect tumor immunity in this context. To address this possibility, we used an intracranial mouse brain tumor model system to explore whether and how TIM-3 responds to brain tumor and whether it could play a role in specific immune cells of the brain tumor microenvironment. Here, we provide new information regarding the unique expression and action of TIM-3 in glia, brain-resident immune cells, in the brain tumor microenvironment. Our results suggest that glial TIM-3 may rapidly and actively respond to brain tumor, thereby affecting not only intracellular immune signaling but also intercellular communication with T cells in the brain tumor microenvironment. These findings could facilitate the development of TIM-3–based therapeutics against brain tumors, including strategies for its timely combination with immune checkpoint molecules.

Sprague–Dawley (SD) rats and C57BL/6 were purchased from ORIENT BIO. B6.129-Tlr2^tm1Kir/J, B6(Cg)-Tlr4^tm1.2Karp/J, and B6N.129S-Havcr2^tm1Bmed/J (Tim-3^mut) were purchased from The Jackson Laboratory. Mice carrying HIF1α-floxed alleles (HIF1α^+f/+f) were obtained from Dr. Randall Johnson (University of Cambridge, Cambridge, United Kingdom), and mice lacking HIF1α in myeloid lineage cells were generated by crossbreeding HIF1α^+f/+f mice to LysM-Cre transgenic mice. Mice carrying gene targeting of the Rosa26 locus with a Flag-TIM-3 construct (FSF-TIM-3) were produced by Dr. Lawrence P. Kane (University of Pittsburgh, Pittsburgh, PA; ref. 9). All animals were maintained and bred under SPF conditions in the Association for Assessment and Accreditation of Laboratory Animal Care–accredited National Cancer Center animal facility. All animal procedures were performed according to ARRIVE guidelines and NCC guidelines for the care and use of laboratory animals. The protocol was approved by the Committee on the Ethics of Animal Experiments of the NCC (Permit No.: NCC-11-125). To avoid bias, the animal studies in this study were properly randomized in a blinded manner with respect to the genotypes and treatments.

The murine CT-2A glioma cell line, rat B35 cell line, and human U373MG cell line were purchased from MERCK (#SCC194), ATCC (#CRL-2754), and Sigma (#89081403), respectively. The murine GL26 cell line was kindly provided by Dr. Y.-K. Hong (The Catholic University of Korea, Seoul, Korea). For EGFP-expressing GL26 cells, GL26 cells were transfected with PLL3.7.EF1α producing lentivirus LV-EGFP, and the cells were sorted with gating to 90% purity using a FACSAria (BD Biosciences). All cell lines were used in experiments between passage 2 and 15, regularly confirmed to be free of mycoplasma contamination (most recent date December 2019) and authenticated by STR analysis (NCC Genomic Core).

Primary glial cells were cultured from the cerebral cortices of 1- to 3-day-old SD rats and mice as described in the previous studies (25, 26). Briefly, cortices were triturated into single cells in MEM (Sigma) containing 10% FBS (Hyclone), plated in 75-cm² T-flasks (four hemispheres/flask for mice) and incubated for 2 weeks. The proportion of microglia in the mixed glial cultures was demonstrated to be 30% to 50% by FACS analyses using an anti-CD11b antibody. The microglia were detached from the flasks by mild shaking and applied to a nylon mesh to remove astrocytes and cell clumps. Cells were plated in 24-well plates (5 × 10⁴ cells/well). After removal of the microglia, primary astrocytes were prepared by trypsinization. The cells were demonstrated to be more than 95% authentic microglia and astrocytes because of their characteristic morphology and the presence of the microglia marker CD11b or astrocytes marker GFAP. Conditioned media (CM) from tumor cells and normal astrocytes were prepared by culturing the tumor cells and primary astrocytes for 48 hours. The media were collected from culture dishes and centrifuged at 1,500 rpm for 10 minutes, and then filtered in 0.2 μm syringe filter (Pall corporation). 100% CM were used for this study.

GL26 cells or CT-2A glioma cells were implanted into the right cerebral hemisphere of 8- to 10-week-old female B6 mice as described by Chang and colleagues (27). To generate the mouse brain tumor models, 3 × 10⁵ tumor cells were stereotaxically implanted into the brain using the following coordinates: AP = +0.3 mm; ML = +2 mm; and DV = −3 mm from the bregma (Supplementary Fig. S1B).

Mouse tumor-infiltrating immune cells were isolated using a discontinuous Percoll gradient method. Briefly, animals were anesthetized and transcardially perfused with ice-cold PBS. Brains were removed and sliced at 1-mm thickness using stainless steel brain matrices. Brain slice sections (3 mm) including the injection site were divided into left and right hemispheres. The left and right hemispheres were enzymatically digested with DNase I (100 U/mL) and collagenase (0.1 mg/mL) for 15 minutes at 37°C in serum-free DMEM (Hyclone). After quenching the digestion with the addition of complete media, the cell suspension was obtained by forcing the tissue through 70 μm meshes and separated on a discontinuous 35%/70% Percoll gradient, and lymphocytes were collected from the interface. The collected fractions were washed and centrifuged to remove the Percoll and obtain the microglia. T cells were further distinguished by flow cytometry using fluorochrome-conjugated antibodies against CD45, CD3, CD4, CD8, and CD11b (for mouse).

A 1517-bp fragment of the mouse TIM-3 promoter (−1,517 to +1 relative to the start codon) was PCR amplified from genomic DNA and cloned into the PGL3 basic vector (Promega). Mouse primary mixed glial cells were transfected using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. After transfection, cells were incubated with B35CM or ACM for 6 hours, and reporter gene activity was determined with a luciferase assay system (Promega). β-Galactosidase activity was measured for the normalization of transfection efficiency.

All data are presented as the mean ± SD (n = number of individual samples). All statistical analyses were performed using the GraphPad Prism 5 (GraphPad Software). P < 0.05 was taken as indicating a significant difference.

To gain new insights into the characteristics of TIM-3 in the brain tumor microenvironment, we prepared EGFP-expressing mouse GL26 glioma cells using the PLL.3.7 EF1α lentiviral vector system (28), and implanted the cells stereotactically into the right cerebral hemisphere of C57BL/6 mice (Supplementary Figs. S1A and S1B). At 21 days after implantation, IHC was performed on histologic sections of optimal cutting temperature (OCT)-embedded frozen brain tissue samples from the intracranial brain tumor model. As shown in Fig. 1A and Supplementary Fig. S1C, we observed that TIM-3 was expressed in both green fluorescent GL26 glioma cells and surrounding nontumor cells such as immune cells of tumor-bearing regions. FACS analysis of tumor-infiltrating immune cells from the mouse intracranial model further showed that TIM-3 was expressed in not only CD3⁺CD45⁺ T cells but also in F4/80⁺ or CD11b⁺ innate immune cells of the brain tumor-bearing hemisphere (Supplementary Fig. S2).

Glioma is a type of tumor that arises from glial cells, which serve as brain resident immune cells. Given that TIM-3 expression was detected in CD11b⁺CD45^mid microglia and we know little about the characteristics of TIM-3 in tumor-associated innate immune cells, we compared the basic expression features of TIM-3 in CD11b⁺CD45^mid glia and GL26 glioma cells. We first performed IHC using an antibody against ionized calcium binding adaptor molecule 1 (Iba1, a microglial maker), and confirmed that TIM-3 was expressed on microglia in our intracranial glioma mouse model (Fig. 1B). To define the features of glial TIM-3 in detail, we cultured primary microglia from healthy B6 mice, and then examined the basal level of TIM-3 compared with that in glioma cell lines. The TIM-3 transcript was expressed at a detectable level in primary cultured mixed glia and microglia from B6 mice. The basal level of TIM-3 was considerably lower in GL26 glioma cells than in primary glia at both transcript and cell-surface levels (Supplementary Figs. S3A–S3C). Similar results were observed in rat primary microglia and B35 rat brain tumor cells (Supplementary Figs. S3A, S3D, and S3E). Interestingly, unlike TIM-3, PD-L1 was more highly expressed in glioma than primary glia at the cell surface expression level (Supplementary Fig. S3F). Under the same condition, there were no noticeable differences in the cell surface levels of 4-1BBL, PD-1, and 4-1BB between glia and glioma (Supplementary Figs. S3G–S3I). Collectively, these results indicate that TIM-3 shows a distinctive expression pattern in glia and brain tumor cells, as compared with other representative immune checkpoint molecules.

Glia are brain-resident immune cells that first communicate with pathologic changes in abnormal conditions (29). We thus investigated whether glial cells undergo alteration of TIM-3 expression in response to brain tumor, which would conceivably affect immune responses. Intriguingly and contrary to our expectations, the TIM-3 transcript level was rapidly and significantly reduced both in mouse and rat primary microglia exposed to CM from mouse GL26, rat B35, and human U373 brain tumor cells (Fig. 2A; Supplementary Fig. S4). The TIM-3 protein level was also decreased in CD11b⁺ primary glia exposed to brain tumor CM compared with normal astrocyte CM, and this alteration was found to be time-dependent (Fig. 2B; Supplementary Figs. S5A–S5C). No such alteration was detected when the same cells were exposed to CM derived from normal primary astrocytes, or 5% DMEM (Supplementary Fig. S5D). In addition, luciferase assays showed that TIM-3 promoter activity was reduced by exposure to B35 tumor cells (Fig. 2C). Similar results were obtained from our immunochemical analyses (Fig. 2D and E). To obtain the relative information for the combination therapeutic approach with immune checkpoint molecules, we examined the expression patterns of other immune checkpoint molecules in brain tumor CM-exposed glia, and compared them to the pattern observed for TIM-3. Unlike TIM-3, PD-L1 was significantly upregulated upon exposure to GL26CM and U373CM (Fig. 2F and G). The expression levels of 4-1BB and PD-1 were not changed under the same conditions (Fig. 2H; Supplementary Fig. S4).

To validate these in vitro results, we investigated whether the glial TIM-3 level was reduced by brain tumor in a mouse intracranial brain tumor model. GL26 cells were implanted into C57BL/6 mice, and tumor-infiltrating cells were isolated using Percoll gradient centrifugation, and stained with anti-CD11b and anti-CD45 antibodies. As shown in Fig. 3A, the TIM-3 level of CD11b⁺CD45^mid microglia was significantly lower in cells from the ipsilateral tumor bearing region (red) compared with those from the contralateral region (black). We observed a similar decrease of the TIM-3 level in CD11b⁺CD45^high macrophage cells sorted among the tumor-infiltrating cells (Fig. 3A, right). However, in the same models, TIM-3 levels were significantly increased in tumor-infiltrating CD8⁺ T cells (Fig. 3B; Supplementary Fig. S4). In addition, the TIM-3 expression patterns of mice bearing CT-2A murine glioma cells were similar to those of the GL26 model (Supplementary Fig. S6). To more precisely assess the characteristics of TIM-3 in innate immune cells, we investigated whether the TIM-3 level differs in the CD11b⁺ population sorted among peripheral blood mononuclear cells (PBMC) of tumor-bearing and healthy control mice. Consistent with the results obtained using tumor-infiltrating cells, we found that the proportion of TIM-3^lowCD11b⁺ cells was considerably larger in PBMCs from GL26 tumor-bearing mice compared with control mice (Fig. 3C). However, we did not observe any meaningful difference in the TIM-3 levels of CD4⁺ and CD8⁺ T-cell populations obtained from PBMC of tumor-bearing and control mice (Fig. 3D). Under the same condition, the proportion of PD-L1^high cells was increased in the CD11b⁺ cells of tumor-bearing versus control mice (Fig. 3E). Taken together, these in vitro and in vivo data support the idea that the levels of TIM-3 among CD11b⁺ cells, including microglia, are downregulated in response to brain tumor.

As the above results strongly suggested that glial TIM-3 may play a role in the brain tumor microenvironment, we examined the possible impact of TIM-3 in immune and inflammatory signaling. We used the Tim-3 mutant (Tim-3^mut) mouse, which harbors a homozygous deletion of the TIM-3 cytoplasmic domain and shows intracellular signaling defects (30). Primary microglia were cultured from Tim-3^mut mice and C57BL/6 wild-type (WT) mice, and incubated with GL26CM or normal astrocyte CM for the indicated times. Conventional RT-PCR and real-time RT-PCR showed that the several inflammation-related genes including IL12p35, IL12p40, and IL23p19 were significantly increased in normal WT glia exposed to GL26CM compared with those in ACM-treated controls, but showed a smaller enhancement in Tim-3^mut glia exposed to GL26CM (Fig. 4A; Supplementary Fig. S7A). In the same condition, some molecules, including TNFα, exhibited similar expression levels in WT and Tim-3^mut glia (Fig. 4A). In addition, induction of iNOS expression by GL26CM or Pam₃CSK₄, a known inducer of iNOS, was relatively lower at the RNA and protein levels in Tim-3^mut glia compared with WT glia (Fig. 4B and C).

We also found that the GL26CM-induced enhancement of PD-L1 was less pronounced in TIM-3 signaling-defective CD11b⁺ glia (Fig. 4D). Similar results were observed in Tim-3^mut CD11b⁺ glia treated with Pam₃CSK₄ (Supplementary Fig. S7B). To further confirm the expression of PD-L1 in Tim-3^mut mice, we used FACS analysis to examine the level of PD-L1 in CD11b⁺CD45^mid microglial cells and CD11b⁺CD45^high macrophages sorted from GL26 tumor-bearing mice. As shown in Fig. 4E and Supplementary Fig. S7C, PD-L1 expression was significantly reduced in microglial cells and macrophage cells from GL26 glioma-bearing regions of Tim-3^mut mice compared with those of WT mice (WT = 5, Tim-3^mut = 7).

To further substantiate these findings, we employed primary glia from a flox-stop-flox TIM-3 (FSF-TIM-3) mice, in which TIM-3 expression is driven in a Cre-dependent manner. Primary glial cells cultured from FSF-TIM-3 mice were infected with CMV-Adenovirus (Ad-CMV) or Cre recombinant Adenovirus (Ad-CMV-Cre), and the cells were treated with GL26CM for the indicated times. RT-PCR analysis showed that the transcript levels of the above-listed immune-associated genes were more strongly induced in TIM-3^high glial cells infected with Ad-CMV-Cre compared with control cells infected with Ad-CMV (Fig. 4F; Supplementary Fig. S7D). Similar results were obtained from quantitative real-time PCR (Fig. 4G). Taken together, these results strongly suggest that defects in TIM-3 signaling or high-level expression of TIM-3 influence the expression of several immune-associated genes, including PD-L1.

TLR2 recognizes exogenous and endogenous pathogens, and regulates numerous inflammation-associated genes, including iNOS and PD-L1 (25, 31). Given that the inductions of iNOS and PD-L1 by GL26CM or Pam₃CSK₄ (a representative TLR2 ligand) were meaningfully reduced in Tim-3^mut glia, we questioned whether TLR2 is related to TIM-3–mediated signaling in glia. We thus examined the effect of GL26CM or Pam₃CSK₄ on the level of TLR2 in WT and Tim-3^mut glia. Western blot analysis showed that the level of TLR2 was considerably increased by GL26CM or Pam₃CSK₄, and that this level was markedly lower in Tim-3^mut glia versus WT glia (Fig. 5A and B). However, we did not detect any significant difference in the expression of TLR1 or TLR6 under our experimental conditions. FACS analysis confirmed that the induction of TLR2 by GL26CM or Pam₃CSK₄ was less in CD11b⁺ glia from Tim-3^mut mice compared with WT mice (Fig. 5C and D).

We next used TLR2- and TLR4-knockout (KO) mice to examine whether TLR2 and TIM-3 have a mutual effect in glia of the brain tumor microenvironment. Because we previously found that glial TLR2 promptly responds to and acts against brain tumors (25), we compared the cell surface expression levels of TIM-3 on primary glia from WT, TLR2-KO, and TLR4-KO mice. As shown in Fig. 5E and F, the level of TIM-3 was lower and that of TLR2 was higher in WT glia treated with GL26CM or Pam₃CSK₄ compared with those treated with normal astrocyte CM. Intriguingly, we did not observe this decrease in TIM-3 or increase in TLR2 among TLR2-deficient glia treated with GL26CM or Pam₃CSK₄ (Supplementary Fig. S4). In TLR4-deficient glia, however, treatment with GL26CM or Pam₃CSK₄ reduced TIM-3 and enhanced TLR2 in the same manner seen in WT glia. Under the same condition, the GL26CM- or Pam₃CSK₄-triggered increase of PD-L1 expression was decreased in TLR2-deficient glia compared to WT and TLR4-deficient glia (Fig. 5G). Collectively, these results strongly indicate that TIM-3 is closely associated with TLR2 in glia exposed to tumor CM. These further suggest that the cross-regulation of TIM-3 and TLR2 may affect the immune responses against brain tumor.

Glial cells can respond to pathologic stimuli and activate T cells including CD8⁺ T cells (32–34). To define the function of glial TIM-3 in the brain tumor microenvironment, we used in vitro coculture system to examine whether glial TIM-3 dysfunction could affect the interplay between glia and CD8⁺ T cells. Primary glia were cultured from Tim-3^mut and WT mice, and CD8⁺ T cells were isolated from lymph nodes of WT mice. Primary glia were pretreated with or without GL26CM for 24 hours, and then incubated alone or with CD8⁺ T cells, and the level of IFNγ secreted into the media was examined. When primary WT glia and CD8⁺ T cells were cocultured, IFNγ secretion was markedly increased in the presence of GL26CM. As shown in Fig. 6A, this enhancement was considerably reduced in co-cultures containing Tim-3^mut glia and CD8⁺ T cells. IFNγ secretion was not noticeable in primary glia or CD8⁺ T cells alone.

To further substantiate the above results, we examined the level of IFNγ using primary glia from FSF-TIM-3 mice. Primary glial cells from FSF-TIM-3 mice were infected with Ad-CMV or Ad-CMV-Cre, and the cells were incubated with CD8⁺ T cells in the presence of GL26CM for 24 hours. In accordance with the above results, IFNγ levels were significantly higher in TIM-3^high glial cells infected with Ad-CMV-Cre compared with control cells infected with Ad-CMV (Fig. 6B). These findings suggest that glial TIM-3 may affect the secretion of IFNγ by CD8⁺ T cells in the presence of brain tumor. Collectively, our results indicate that glial TIM-3 functions in both intracellular immune responses and the interplay between glia and CD8⁺ T cells in the brain tumor microenvironment.

Glial TIM-3 expression is regulated by hypoxia in an HIF1-dependent manner (24), and hypoxia is a characteristic of the tumor microenvironment (35). We thus questioned whether the brain tumor CM-induced decrease of the glial TIM-3 level could be affected by hypoxia. For this, primary glia cells were cultured and incubated with GL26CM or ACM control for the indicated times under 20% and 1% O₂ conditions. RT-PCR analysis was performed using two sets of TIM-3 primers that were designed from different sequences. As shown in Fig. 7A, similar to the results obtained under 20% O₂, TIM-3 transcript level was decreased by GL26CM under 1% O₂. To validate this finding, we examined whether TIM-3 expression could be changed by GL26CM in HIF1α-deficient primary glia. Primary glia from LysM-Hif1α^−/− mice were transfected with Cre-expressing adenovirus for depletion of HIF1α, and then incubated with GL26CM or ACM. RT-PCR and FACS analyses showed that TIM-3 expression was decreased by GL26CM in LysM-Hif1α^−/− glia under both 20% O₂ and 1% O₂ (Fig. 7B and C). These results further support that hypoxia do not strongly influence the tumor-dependent decrease of TIM-3 expression in glia. To further assess TIM-3 expression in the hypoxic tumor environment, we examined the level of TIM-3 in CD8⁺ T cells under normoxia and hypoxia. Splenic CD8⁺ T cells were isolated and cultured with ACM or GL26CM in the presence of IL2, anti-CD3, and anti-CD28. As shown in Fig. 7D, TIM-3 levels of CD8⁺ T cells were slightly increased by GL26CM under both normoxia and hypoxia. The GL26CM-triggered enhancement of TIM-3 in CD8⁺ T cells was similar under 20% O₂ and 1% O₂, although TIM-3 level was higher in 1% O₂ than that seen in 20% O₂. Collectively, these results suggest that tumor-dependent alteration of TIM-3 expression may occur irrespective of the oxygen level in the brain tumor microenvironment.

Many recent research efforts have focused on developing therapeutic strategies for improving antitumor immunity and restoring the proper function of immune cells against tumors. Although the initial studies on the disease association of TIM-3 highlighted autoimmune diseases and viral infections, TIM-3 has recently attracted attention as a promising target for strengthening the antitumor activity of immune cells (1). Studies have shown that TIM-3 is expressed not only by tumor-associated immune cells such as CD8⁺ T cells and Treg cells, but also by tumor cells themselves in several tumors (36–38). Furthermore, it has been reported that combined treatment with TIM-3– and PD-1–blocking antibodies was more effective than monotherapy with the PD-1 antibody alone (16). Accumulating evidence suggests that TIM-3 displays diverse characteristics and roles in sometimes conflicting ways, depending on the specific situation (7). Thus, if we hope to develop efficient therapeutic approaches targeting TIM-3 in certain tumors, we must fully understand the detailed functional features of TIM-3 and its mechanisms of action in specific cells that play roles in the tumor microenvironment. In this regard, we explored whether glial TIM-3 responds to brain tumor, and thereby modulates immune responses to combat or help brain tumor (Fig. 7E).

Aggressive brain tumors such as glioblastoma (GBM) generally have a poor prognosis with a high overall mortality rate. The current standard treatment for glioma encompasses surgical resection, radiation, and concomitant and adjuvant chemotherapy with temozolomide (39). Researchers are seeking ways to improve the therapeutic efficacy of such strategies; new innovative clinical trials and experimental studies are ongoing, and immune-based therapy is considered to be promising approach (40). However, brain tumors hide behind the blood–brain barrier, and we know little about their local immune microenvironment. The brain tumor mass is made up of several types of cells, including tumor cells, infiltrating immune cells, and brain-resident immune cells such as microglia (41, 42). In particular, macrophages and microglia have been found to comprise up to 34% of the immune cells that infiltrate into brain tumors (43). To define the precise characteristics of TIM-3 in the brain tumor microenvironment, we first examined the expression patterns of TIM-3 using tissue sections and tumor-infiltrating cells from a mouse intracranial brain tumor model established with EGFP-expressing GL26 glioma cells. IHC and FACS analysis showed that TIM-3 is expressed in both glioma cells and surrounding nontumor cells, including T cells and glial cells. Given that we know relatively little about the function of TIM-3 in the innate immune system, and glia are known to act as frontline immune cells in the brain, we set out to identify the characteristics of TIM-3 in glial cells of the brain tumor microenvironment. Interestingly, we found that the level of glial TIM-3 was markedly reduced by exposure to brain tumor CM. Under the same conditions, the level of PD-L1 was upregulated, and those of 4-1BB, and PD-1 were unchanged in primary glia expose to brain tumor CM. In accordance with our in vitro results, TIM-3 levels were rather lower in tumor-infiltrating CD11b⁺CD45^mid cells from the tumor-bearing hemisphere compared with those from the contralateral hemisphere of an intracranial brain tumor model. Conversely, the TIM-3 levels of tumor-infiltrating CD8⁺ T cells were higher than those of contralateral region. These results suggest that glial TIM-3 may have unique features and functions that could differ from those of TIM-3 expressed on T cells in the brain tumor microenvironment.

The above results raised the possibility that glial TIM-3 might exert unique roles in the brain tumor microenvironment. To address this, we used TIM-3 signaling-deficient mutant mice to explore whether defective glial TIM-3 could affect representative immune responses under brain tumor-exposed conditions. Intriguingly, glial cells from Tim-3^mut mice showed altered expression of IL12p35, IL12p40, IL23p19, PD-L1, and iNOS in comparison with normal glial cells. Our results suggest that TIM-3 may influence the expression of PD-L1 in brain tumor-exposed glia, even though the expression levels of TIM-3 and PD-L1 are differently regulated by brain tumors. More interestingly, TLR2 and TIM-3 were found to reciprocally influence one another's expression levels in tumor-exposed glia. We previously showed that glial TLR2 rapidly responds to brain tumors by exhibiting augmented cell surface expression, upregulating several immune signaling events, and thereby affecting tumor growth (25, 27). Notably, our current results show that the induction of TLR2 expression by GL26 or Pam₃CSK₄ was significantly reduced in glia with TIM-3 signaling defects. Moreover, TIM-3 expression was not decreased by GL26CM or Pam₃CSK₄ in TLR2-deficient glia. Previous reports and our current findings have shown that TLR2 is closely involved in the expression of IL12p35, IL12p40, IL23p19, PD-L1, and iNOS in innate immune cells including glia (44, 45). Considering these results, the interrelation of TLR2 and TIM-3 may be a pathway that leads to the expressional difference of these molecules in glia between Tim-3^mut and WT mice. However, at this time, the mechanism underlying the crosslink between TIM-3 and TLR2 remained unknown. We are currently examining the mechanisms through which TIM-3 interacts with TLR2 in the brain tumor microenvironment.

In contrast to the results supporting the cross-regulation between TIM-3 and TLR2, there was no noteworthy difference in the expression patterns of TIM-3 in WT versus TLR4-deificient glia. There are several reports that TIM-3 and TLR4 may interact in disease-associated conditions, although this appears to depend on the cell types and status. For example, Anderson and colleagues reported that TIM-3 synergized with TLR4 to promote inflammatory responses such as TNFα secretion in innate immune cells from patients with MS (7) and EAE mice. Wang and colleagues reported that TIM-3 inhibited TLR4-triggered apoptosis and proinflammatory responses, whereas TIM-3 blockade exacerbated these responses in decidual stromal cells during pregnancy (46). However, our results show for the first time that TIM-3 and TLR2, but not TLR4, may closely interact and affect each other's expressions in glia when exposed to brain tumor CM or Pam₃CSK₄.

Given that TIM-3 expression has been linked with several diseases, research interest has increasingly focused on the regulation and function of TIM-3 under certain conditions (1). Hypoxia is a feature of inflammation-associated brain diseases, such as tumors and cerebral ischemia (47). We previously showed that TIM-3 is induced in glial cells following hypoxia/ischemia (H/I), and that blockade of TIM-3 significantly reduced infarct size and inflammatory events (e.g., neutrophil infiltration and edema formation) compared that seen in IgG-treated control H/I mice (24). Thus, we herein examined whether hypoxia could affect the reduction of glial TIM-3 by brain tumor using primary glia from LysM-Hif1α mice. However, we did not detect any significant difference in the TIM-3 level of glia under hypoxic and normoxic conditions. Moreover, the TIM-3 expression pattern was similar between normal LysM-Cre glia and HIF1α-deficient LysM-Hif1α glia, and the GL26CM-triggered enhancement of TIM-3 in CD8⁺ T cells was similar under hypoxia and normoxia. These results suggest that the oxygen level alone does not mediate the brain tumor-dependent alteration of TIM-3 in glial cells and CD8⁺ T cells exposed to tumor CM. Considering these results, there may be a factor (or factors) that regulate the differential expression of TIM-3 in glia and CD8⁺ T cells irrespective of oxygen level. Previously, we found that, unlike the expression of TLR2 in glia, there was no significant difference in the TLR2 expression of CD4⁺ and CD8⁺ T cells between intracranial and contralateral hemispheres of tumor-bearing mice at any tested time point (27). In an effort to identify the mechanism underlying the differential regulation of TIM-3 in glia and T cells, we are currently starting to explore possible signaling molecules, including TLR2.

Tumor progression is modulated by interactions among tumor cells and their surrounding immune cells: immune cells cooperate to fight and destroy encountered tumor cells, whereas the tumor cells seek to educate the immune cells to suppress antitumor immunity and/or promote tumor progression (41). In this study, we observed that the tumor CM-triggered increase of IFNγ secretion, which is mediated by an interaction between glia and CD8⁺ T cells, was less pronounced in cocultures containing Tim-3^mut glia and CD8⁺ T cells, compared with those containing WT glia and CD8⁺ T cells. Previously, we reported that glial TLR2 plays an essential role in the antigen presentation system by regulating MHCI expression, and that the TLR2–MHCI axis contributes to the proliferation and activation of CD8⁺ T cells by brain tumor (27). Thus, the previous and current findings suggest that glial TIM-3 seems to interact with TLR2, thereby affecting the interplay between glia and CD8⁺ T cells in the brain tumor microenvironment. We have not yet clearly identified which factor(s) could trigger the interplay leading to glial TIM-3–mediated intracellular and intercellular immune signaling events, nor do we fully understand how glial TIM-3 affects tumor progression. However, our present results suggest that glial TIM-3 may interact with brain tumor cells and other immune cells to affect the brain tumor microenvironment.

Recently, clinicians and researchers have shown great interest in immune checkpoint blockade-based combination therapy. For efficient therapy, it is important to select proper partners that can reduce the tumor burden and increase tumor immunogenicity (48). TIM-3 has been reported to serve as both an inhibitory and activating regulator in a context dependent manner. A change in TIM-3 expression can track with changes in other checkpoint receptors in some—but not all—tumor types. Moreover, TIM-3 blockade has yielded discrepant effects in different types of tumors (16, 49, 50). Thus, we need to obtain clinically relevant information regarding the features and functions of TIM-3 if we hope to appropriately target this factor in specific tumor types. Glia drive early immune responses against tumors, by recruiting T cells and neutrophils into brain tumor sites hiding behind the blood–brain barrier. Considering the function of glial cells, glial TIM-3, with its distinctive expression patterns and functions, could be a target to modulate antitumor immunity in the brain. Our current and ongoing studies will help improve our understanding of the specialized characteristics of glial TIM-3 in the brain tumor microenvironment (Fig. 7E), which could facilitate the development of detailed and systematic approaches for using immune checkpoint molecules to treat brain tumors.

L.P. Kane has ownership interest (including patents) in Lawrence Kane. No potential conflicts of interest were disclosed by the author.

Conception and design: S.S. Kim, E.J. Park

Development of methodology: H.-S. Kim, C.Y. Chang, S.S. Kim, E.J. Park

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L.P. Kane, E.J. Park

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.-S. Kim, C.Y. Chang, H.J. Yoon, K.S. Kim, E.J. Park

Writing, review, and/or revision of the manuscript: H.-S. Kim, H.J. Yoon, S.-J. Lee, L.P. Kane, E.J. Park

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H.-S. Kim, K.S. Kim, H.S. Koh, E.J. Park

Study supervision: E.J. Park

We thank M. Kim (Microscopy Core), S. Jeon (Molecular Imaging Core), and T. Kim (Flow Cytometry Core) for expert assistance and suggestions. We also thank all members of E.J. Park's laboratory for their helpful comments and discussions. This work was supported by the National Cancer Center (NCC-1810230, 1911263 to E.J. Park) and National Research Foundation of Korea (NRF-2017R1A2B2010289 to E.J. Park).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.