Although radiation has traditionally been considered an immunologically inert process, the recent discovery of immunogenic cell death (ICD), a unique mode of cell death induced by RT or chemotherapy via potent host-mediated antitumor response (28), has suggested otherwise. Cell death occurs differently depending on the identity and maturity of the phagocytic cell, location and manner of phagocytosis, the availability of helper T-lymphocytes, type of death pathway that is triggered, release of immunosuppressive mediators (TGF-β, IL-10), and the immune cells that are exposed to antigens (29). ICD in particular is primarily defined by unique molecular processes, including the translocation of calreticulin (CRT) to the cell surface, ATP release, upregulation of costimulatory molecules, and the extracellular release of high-mobility group protein B1 (HMG-B1), which enhances antigen cross-presentation and secretion of proinflammatory cytokines (28, 30).

Although RT has been speculated to exert immunosuppressive effects via increased TGF-beta expression, M2 macrophage polarization, and T-regulatory (T-reg) cell recruitment, its immunostimulatory effect is beginning to be understood (31–34). Ionizing radiation has been shown to increase translocation and expression of CRT (35) and promote gene transcription of proinflammatory factors via HMG-B1 (36, 37), which are the essential components of ICD, as well as reduce production of immune suppressive cytokines and increase expression of MHC-I and synthesis of novel peptides for cytotoxic T cell recognition (38). Moreover, RT has been shown to promote re-oxygenation and decrease interstitial fluid pressure within the tumor microenvironment, improving immune cell recruitment and infiltration into irradiated tumor (39). Finally, RT induces the release of tumor-associated antigens (TAAs), diversifying the TCR (T-cell receptor) repertoire of infiltrating CTLs and leading to increased efficacy of CTLs (27, 40). Notably, these cancer-specific and stromal-associated responses occur simultaneously and define radiation-induced immunogenicity of the tumor cells. Preclinical studies have clearly suggested that radiation, although conventionally perceived as a local therapy, can potentially exert systemic antitumor effects at least through both cancer cell intrinsic and tumor microenvironmental modulations. These mechanisms are illustrated in Figure 1.

While the brain has traditionally been considered an immunoprivileged organ system, it is now commonly accepted that dysregulation of immune surveillance may be involved in the pathogenesis and progression of multiple CNS malignancies (41). Similar to other types of cancers, primary brain tumors can evade immune system detection through multiple immunosuppressive mechanisms, including upregulation of immune checkpoints. For example, PD-L1 has been shown to be upregulated in glioblastoma cells as well as circulating monocytes and macrophages through oncogenic signaling resulting from PTEN loss and modulation of autocrine–paracrine IL-10 signaling (42–44). Increased PD-1 and PD-L1 expression has also been observed in brain metastases and primary CNS lymphomas (45, 46). These findings warrant further exploration of ICB, in particular PD-1 or PD-L1 blockers, in the pathogenesis and treatment of primary and metastatic CNS malignancies.

Radiation therapy is the pillar of standard therapy for primary and secondary brain tumors. The immunostimulatory effects of RT make it a natural candidate as a synergistic partner with ICB. Indeed, prior preclinical studies have shown survival benefits of radiation with tumor vaccination in glioma mice models (20). In melanoma, RT has been shown to diversify TCR repertoire of infiltrating CTLs and enhance their function via RT-induced release of TAAs (27). This mechanism worked in concert with an anti-CTLA-4 agent leading to inhibition of T-reg cell function and increased CD8⁺ T cell/T-reg ratio. Interestingly, the addition of an anti-PD-L1 agent further prevented T cell exhaustion (40). In murine glioma models, combining radiation concurrently with anti-PD-1 and anti-CTLA-4 or 4-1BB (a costimulatory molecule) agonist yielded improved survival and increased tumor-infiltrating leukocyte (TIL) population, compared with either modality alone (24, 47). The rationale for combining RT with PD-1/PD-L1 blockers is further supported by the observation that local inflammation mediated by RT results in PD-L1 upregulation on cancer cells, macrophages, and dendritic cells (DCs) (48). Moreover, a recent study in murine melanoma brain metastases (MBM) showed that increasing radiation dose with concurrent immunotherapy improved median and long-term survival and prolonged tumor dormancy (49). Finally, while the penetrance of ICB agents into the CNS via the blood–brain barrier (BBB) is not fully known, RT has been shown to increase BBB permeability (50), further facilitating the penetrance of activated antitumor immune cells and possibly the access of ICB agents as well – providing another compelling rationale for combination RT-ICB therapy in the treatment of primary and metastatic CNS tumors (Figure 1).

Clinical evidence supporting the efficacy of combining RT and ICB in the treatment of CNS tumors is primarily garnered from melanoma patients with metastatic disease and is retrospective in nature. Three retrospective single-institution studies have suggested that ipilimumab in combination with RT may be more effective than RT alone in MBM (57–60). Knisely et al. reviewed the outcomes of 77 patients with MBM who received SRS as well as ipilimumab. Patients who received combination therapy demonstrated improved survival compared with those who received SRS alone (58). Moreover, a study by Silk et al. explored the benefits of combination therapy vs. RT alone in 70 MBM patients who received brain radiation (either SRS or WBRT) with ipilimumab. Retrospective analysis demonstrated an OS benefit of 19.9 months with combination therapy vs. 4.0 months for SRS alone, with no associated increase in toxicity with addition of ipilimumab to SRS (59). An interesting study by Kiess et al. investigated the impact of timing in administration of SRS for brain metastases and ICB on patient outcome. Patients who received SRS before or concurrently with ipilimumab appeared to exhibit improved outcomes in OS and distant intracranial control compared to patients who received radiation after immunotherapy (60). Notably, responses to ipilimumab therapy in the brain, as in extracranial disease, can be quite durable with some reported response beyond 4 years (61). Given the retrospective and single institutional nature of these studies, caution needs to be taken regarding interpretation of their results. Several prospective early phase clinical trials at multiple institutions are investigating different combination regimens of RT with immune checkpoint blockers in the setting of brain metastases (Table 1). Results from these studies will one day shed light on the clinical benefit of combining radiation with immunotherapy for the treatment of brain metastases.

For primary CNS tumors, such as glioblastoma, several combination approaches with radiation have been explored clinically (62). For example, immunotherapy in the form of vaccines derived from autologous DCs and pulsed with tumor lysate and various tumor antigens (such as the EGFRvIII vaccine Rindopepimut) have shown some promise in the clinical setting (20). A randomized phase II trial testing ICT-107, a vaccine composed of autologous DCs and tumor antigens, administered with radiation and temozolomide for newly diagnosed glioblastoma, demonstrated benefit in terms of OS and PFS for a subset of HLA-A2 positive patients (63).

Although evidence supporting the combination of ICB and RT for glioblastoma is currently lacking, several clinical trials are underway examining this combination for newly diagnosed and recurrent glioblastoma (Table 1). Two of these are large multi-institutional trials in phases II and III (NCT02667587, NCT02617589) examining OS in patients given nivolumab in combination with radiation (with/without concurrent temozolomide) for newly diagnosed glioblastoma, while another smaller trial is examining similar combinations with pembrolizumab (NCT02530502). A phase II trial sponsored by the Ludwig Institute is investigating OS in monotherapy with the PD-1 inhibitor durvalumab (MEDI4736) in comparison to combination approaches with standard RT or the VEGF inhibitor bevacizumab. Finally, a phase I trial is examining maximum dosage of pembrolizumab when it is used with combination therapy with bevacizumab and a hypofractionated radiation regimen given over 5 days for recurrent glioma. Additional clinical trials investigating combination approaches in the context of primary CNS malignancy will soon be underway.

Glioblastoma is notorious for employing a wide variety of immunosuppressive strategies to disrupt the function of immune cells, lower immunoglobulin levels, and generate a plethora of immunosuppressive processes, involving TGF-beta, IL-10, MHC-I downregulation, T-regs recruitment, and increased expression of PD-L1 (20, 64–68). Development of therapeutics for glioblastoma is complicated by the variability in antigen expression within individual tumors and subtypes. The ability to target the unique signature of tumor-specific mutations and antigens (“neoantigens”) is crucial in effective tumor control, and the expression of these neoantigens has been shown to predict long-term response to ICB in metastatic melanoma and NSCLC (69, 70). Strategies such as biomarker and microRNA (miRNA) development are under investigation to predict and modulate tumor response to immunotherapy and RT.

Along with clinical trials, current translational research efforts are exploring potential predictive biomarkers for response to immunotherapies in glioblastoma (71). Increased PD-L1 expression in circulating monocytes has been identified as a biomarker for tumor-induced immunosuppression and a prognostic factor for poor survival. For example, in one study of glioblastoma patients, the median overall survival for high-expressers of PD-L1 was 18.0 months compared with 44.7 months for low PD-L1 peripheral monocyte expressers (5). Another example is level of tumor EGFRvIII expression as a biomarker for response to the EGFRvIII vaccine in glioblastoma (72). Increased peripheral blood levels of CD33⁺HLA-DR⁻ myeloid-derived suppressor cells (MDSCs), with the potential to induce T-regs and apoptosis in activated lymphocytes as well as secrete immunosuppressive cytokines and oxidizing molecules, have also been reported patients with glioblastoma. RT in combination with ICB has been shown to reduce tumor-associated MDSCs in comparison to single therapy via the activity of CD8⁺ T cells and tumor necrosis factor (TNF) signaling in vivo (48). Moreover, a diminution in the level of circulating T-regs has shown promise as a biomarker for treatment response to immunotherapies (73).

MicroRNAs can act as oncogenes or tumor suppressors through controlling target gene expression and binding of mRNA, and also carry promise as biomarkers (74). In glioblastoma, tumor miR-21 has been a promising candidate as a biomarker in diagnosis and response to chemotherapies (75). In the serum, levels of miR-128, miR-320, and miR-574-3p have been found to be elevated (76, 77). Interestingly, the presence of metastatic disease in the brain can be distinguished from glioblastoma via the identification of the miR-200 family in cerebrospinal fluid (78). miRNA can also serve as targets for therapy. Double-stranded miRNA mimics can reinstate activity of tumor suppressor miRNA, and oncogenic miRNAs can be disenabled by single-stranded antisense nucleotides known as “antimiRs” (79). One example in the CNS is miR-296, which led to reduced tumor volume and angiogenesis in a murine glioma model (80). Another miRNA, miR-221/222, demonstrated reduced tumor volume in murine melanoma (79). A potential target for an miRNA mimic in ICB is miR-138, which has been demonstrated to target both CTLA-4 and PD-1, inhibiting tumor-infiltrating T-regs and subsequently releasing the brake set by these immunosuppressive cells within the glioblastoma microenvironment (81). Whether specific miRNA can be used to predict treatment responses to immunotherapy warrants further investigation.

Newer techniques, such as multi-parameter flow cytometry, time-of-flight mass cytometry (CyTOF) and whole-repertoire T and B receptor amplification, can be used to characterize the immunologic cellular makeup of primary and metastatic brain lesions at baseline and changes in circulatory immune cell profile during treatment (1, 72, 82, 83). RNAseq transcriptome profiling of tumor samples can quantify the presence of markers, tumor antigens, and certain genes. The latter has demonstrated utility in predicting survival in neuroblastoma and treatment response in breast cancer (84). Finally, “imaging genomics” or “radiomic” analysis may be a cost-effective and non-invasive strategy to link the radiophenotype of a tumor it its underlying genomic and proteomic characteristics, via the high-throughput extraction and analysis of large amounts of radiographic imaging for distinctive features, which can then be mapped to certain genomic and proteomic patterns such that predictive and prognostic classifications may be made (85–88). Radiomic analysis may be particularly useful to characterize and distinguish radiation necrosis (RN), a potential late adverse effect of high-dose focal irradiation to the brain occurring more frequently in patients receiving ICB (89), from tumor progression or recurrence and the phenomenon of pseudoprogression. Furthermore, this strategy may be used with traditionally measured biomarkers and miRNA profiles to construct multi-dimensional schemes for prediction of response to combining ICB and RT, development of tailored treatment plans, and assessment of adverse effects of such therapy.