Optimal timing and sequence of combining stereotactic radiosurgery with immune checkpoint inhibitors in treating brain metastases: clinical evidence and mechanistic basis

Metastatic brain tumors occur in approximately 20% of patients with malignant tumors [1]. The existence of brain metastases often indicates poor quality of life and survival with a 2 year survival rate of patients with brain metastases less than 10% in general [2]. The evolving anti-tumor systemic management has improved the survival of patients to a certain degree [3]. However, due to the pharmacokinetic blocking effect of therapeutic agents by the blood–brain barrier (BBB) or the blood-tumor barrier (BTB), and differentiated tumor microenvironment (TME) from the primary tumor, these treatments have rather limited efficacy in suppressing the progression of brain metastases [4]. Still, around one-third of patients with brain metastases die of intracranial progression eventually [4, 5]. Thus, treatments for brain metastasis are yet to be improved.

In the past decade, immune checkpoint inhibitors (ICIs) have primarily reshaped the landscape of anti-tumor treatments for various solid tumors [6]. The mechanism of ICIs is restoring anti-tumor immunity by blocking immune checkpoints, such as cytotoxic T-lymphocyte antigen 4 (CTLA-4), programmed cell death protein 1 (PD-1), and programmed cell death protein ligand 1 (PD-L1) [6]. Also, several meta-analyses support that ICI monotherapy is efficacious for metastatic brain tumors [7, 8]. However, in the included studies of these meta-analyses, only patients with treatment-naïve metastases were enrolled, and the vast majority of prospective studies excluded patients with symptomatic brain metastases [7, 8]. Yet, only 20% of the patients benefited from immunotherapy [7]. The efficacy of ICIs for other patients needs further investigation.

Stereotactic radiosurgery (SRS) has been broadly applied in brain metastases of various cancers [9]. Owing to its modest improvement in patients’ survival and lower chance of causing neurocognitive toxicities compared with whole-brain radiotherapy (WBRT), the American Society for Radiation Oncology recommended SRS as a preferred treatment or an alternative to WBRT for patients with newly diagnosed single or multiple brain metastases who have a good performance status [10]. And the American Society of Clinical Oncology recommended SRS alone for patients with 1 to 4 unresected brain metastases (small cell lung cancer excluded) and patients with 1 to 2 safely resected brain metastases for treating remaining intracranial disease [11]. SRS also has excellent potential to be combined with ICIs because of its trait to temporarily “open up” the BBB/BTB and cause the death of tumor cells which can induce an inflammatory microenvironment with enriched infiltration of antigen-presenting cells (APCs) and cytotoxic T lymphocytes (CTLs) [12, 13]. Although previous meta-analyses denied the clinical advantage of radiotherapy (RT) + ICI compared with ICI monotherapy [7, 14], potential bias may exist due to the retrospective nature of the included studies and baseline differences in patients’ status of the two groups, because only patients with symptomatic brain metastases receive RT + ICI in clinical practice and those who receive ICI monotherapy usually have better baseline performance statuses. As for stage III non-small cell lung cancer (NSCLC), patients who started ICI treatment < 14 days after RT gained significantly better benefits than those who received ICIs ≥ 14 days after RT in a randomized controlled trial [15]. However, the optimal SRS + ICI schedule for brain metastases, especially the sequence of and the interval between SRS and ICI, remains controversial. The understanding of the brain microenvironment after RT may help us guide the development of more effective strategies for brain metastases treatment.

In this review, we aim to summarize recent clinical and preclinical evidence concerning the combination of SRS and ICI in treating brain metastases. From that summary, we aim to provide a perspective on how to sequence and time the two therapies to improve the clinical outcomes of patients with brain metastases.

The BBB can keep out harmful macromolecular substances, playing an important role in maintaining intracranial homeostasis. It consists of tightly connected endothelial cells, mural cells, astrocytes, and basement membranes [16]. The presence of primary or metastatic brain tumors can harm the tight junctions of the BBB and relatively increase the permeability to fuel growth and invasion, and the BBB is often referred to as the BTB in the context [17]. Although the brain was initially considered to be “immune-privileged” due to the isolating capacity of BBB/BTB, newer evidence has shown that brain immunity is just limited rather than “silent” [18]. Intracranial antigens can be recognized by local APCs or translocated to cervical lymph nodes, where the APCs present these antigens to T cells and activate them [19]. Meanwhile, identical antigens released from primary tumors can be recognized and presented to T cells by APCs in adjacent lymph nodes [20]. Activated T cells in the blood, when entering the intracranial metastatic site, get involved in a series of leukocyte-endothelial cell interactions, enabling CTLs to roll and crawl along the vessels to reach a most permissible region for diapedesis [21, 22]. The extravascular CTLs then secrete matrix metalloproteases to decompose dystroglycans in the glia limitans for the final traversal through the BBB/BTB [22, 23]. However, due to chronic tumor antigen exposure, intracranial tumors can send an “off” signal to the CTLs by the binding of immune checkpoint proteins to their ligands. Blocking the interaction between the checkpoints and the ligands by an ICI (anti-PD-L1, anti-PD-1, or anti-CTLA-4) allows the CTLs to regain activity and recover their ability to kill the tumor (Fig. 3A).

The BBB/BTB has a limited permeability and only allows the penetrations of small and medium molecules [16]. Though ICIs have promising efficacy in treating brain metastases of melanoma and NSCLC, they are generally large molecules (146 kDa-149 kDa) that are not likely to cross the normal BBB and may hardly penetrate the BTB [24‐27]. Even so, nivolumab, an anti-PD-1 agent, was detected in cerebrospinal fluid (CSF) of treated melanoma patients with leptomeningeal metastases. In the study, the investigators found that the CSF nivolumab concentration was 35–150 ng/mL, and the CSF/serum ratio of nivolumab concentration was 0.88–1.9% [28, 29]. Van Bussel et al. [29] proposed that the transportation was mediated by the FcRn receptors on the resided macrophages in the epithelial layer of the blood-CSF barrier, which could bind to serum nivolumab, induce endocytosis, and transport and release it to the CSF. However, the efficacy of ICIs on brain metastases is commonly believed to be based on not the direct penetration of ICIs through the BBB/BTB but the relatively autonomous trafficking of CTLs through the BBB/BTB and the moderately disrupted BTB, as mentioned above [18, 21].

Besides the immune barrier effect of the BBB/BTB, brain resident cells, such as microglia and astrocytes, also exhibit immunomodulatory effects in brain microenvironment. Microglia are resident macrophages of the central nervous system originating from the yolk sac and represent the most abundant immune cell population in brain [30]. A recent single-cell analysis has revealed considerable heterogeneity among microglia in the TME, ranging from a homeostatic to reactive phenotype continuum [31]. The homeostatic subpopulation with a ramified morphology possesses phagocytic abilities, which can enhance the therapeutic effect of ICI by presentation of tumor antigens to T cells and leading to activation of CTLs [32]. However, proliferating microglia with amoeboid morphology can inhibit the therapeutic effect of ICIs by contributing to an immunosuppressive TME [32]. Chronic IFN-γ activation in microglia has been linked to an immunosuppressive program, as demonstrated by ex vivo coculture studies. Guldner et al. [33] found that microglia in brain metastases had elevated expression of CXCL10, V-domain immunoglobulin suppressor of T cell activation, and PD-L1, which eventually result in recruitment of malfunctional T cells. Furthermore, cytotoxic T cells' production of granzyme B and IFN-γ is reduced after coculture with proliferating microglia. Proliferating microglia may also promote the depletion of CD8 + T cells. Additionally, activated microglia, along with recruited macrophages, can release a wide array of growth factors and cytokines that support tumor cell proliferation and angiogenesis [34]. Likewise, although the functions of astrocytes in brain tumors vary across subsets, some subgroups of astrocytes exhibit anti-inflammatory properties and may contribute to the immunosuppressive microenvironment [35]. Heiland et al. [36] revealed that astrocytes overexpressed interleukin-10 and transforming growth factor β (TGFβ) in company with microglia or macrophages, and the anti-inflammatory cytokines could further lead to resistance to immunotherapy and radiotherapy [37, 38]. Another subset of astrocytes characterized by high expression levels of the immune checkpoint PD-L1 and activation of the immunomodulatory factor STAT3 was identified in the peritumoral area, where they may potentially act as a barrier against anti-tumor T lymphocytes [39]. Interestingly, the presence of these phosphorylated-STAT3 immunosuppressive astrocytes is induced by tumor and microglia cells [39]. Additionally, reactive astrocytes have been shown to upregulate immunosuppressive and tumor-promoting molecules in microglia and macrophages, thus establishing a positive feedback loop between these cells and TAMs [36]. Hence, the bidirectional crosstalk between astrocytes and microglia plays a crucial role in shaping the immunosuppressive microenvironment in brain tumors.

Moreover, favorable ICI efficacy requires adequate checkpoint expressions and an immune-supportive microenvironment for both primary tumors and brain metastases. Some tumors exhibit primary resistance to ICIs or develop secondary resistance to ICIs during treatment [40]. The mechanisms of ICI resistance are rather knotty and yet to be explored, which mainly include the lack of checkpoint expression, T cell exclusion, impaired interferon signaling, antigen loss, and defective tumor antigen presentation. [41]. Previous studies have revealed a relatively equivalent level of PD-L1 expression between the primary tumor and the paired brain metastases despite noticeable temporal and spatial heterogeneities [42, 43]. This phenomenon suggests that brain metastases originating from ICI-resistant primary tumors are also unlikely to respond to ICIs. However, brain metastases are commonly characterized by much lower immune cell infiltrations and higher proportions of immunosuppressive cells compared with primary tumors, which largely limit the efficacy of ICIs in brain metastases [44]. Our research team investigated transcriptional profiles of 70 brain metastases lesions and 12 samples of paired lung adenocarcinoma and brain metastases, and we found that brain metastases presented an immunosuppressed TME compared with the primary tumor, manifested in inhibition of immune-related pathways, low expression of immune checkpoint, decreased infiltration of CD8 + T cells and cytotoxic lymphocyte, increased proportion of suppressive M2 TAMs [43]. Efforts have been made to transform this “cold tumor” phenotype into a “hot tumor”, including targeting transforming growth factor β, indolamine 2,3-dioxygenase, and tumor-associated macrophages (TAMs), etc. [18].

In summary, ICIs can restore the cytotoxic ability of CTL by inhibiting the interaction between the checkpoint and its ligand. Nonetheless, the therapeutic efficacy of ICIs may be hindered by several factors, including the limited permeability of the blood–brain barrier, primary resistance, as well as the immunosuppressive microenvironment within the brain. To our knowledge, SRS is one of the most promising options to fuel intracranial anti-tumor immunity, and we will discuss it below.

In order to identify clinical studies on the combination of SRS and ICI in treating brain metastases, we searched PubMed using terms of “brain metastasis”, “stereotactic radiosurgery”, and “immunotherapy” or “immune checkpoint” and manually selected retrospective or prospective clinical studies where at least 1 arm involves SRS + ICI as a treatment for patients with brain metastases from any origins. We summarized information from all available studies on the combination of SRS and ICI in brain metastases (Table 1), including the first author, year of publication, NCT registration number (if available), type of study (retrospective/prospective), cancer type, treatments, and sample sizes (patients and lesions) of the intervention arm and the control arm, and intervals. Intervals were defined as the time interval between the first day of SRS and the most adjacent day of ICI infusion in the "concurrent" SRS + ICI treatments. The studies involved 3 classes of primary tumor types, 7 types of intervention/control combinations, and 9 definitions of intervals for “concurrent” SRS + ICI treatments (Fig. 1). In these studies, the definitions for intervals of “concurrent” SRS + ICI were quite different, ranging from 0.1 months to 6 months (Fig. 1). We demonstrated the hazard ratios for the overall survival (OS) in these studies in Fig. 2, and we also mentioned other endpoints below, such as distant brain failure, best objective response (BOR), and intracranial local control, if these data are related to the topic of this review. In short, concurrent SRS + ICI led to better outcomes compared with several controls despite varied definitions of intervals for “concurrent” SRS + ICI (Fig. 2).

Studies aiming to compare concurrent SRS + ICI with non-concurrent SRS + ICI generally showed that concurrent SRS + ICI significantly improved local control of patients with brain metastases compared with non-concurrent administrations. Murphy et al. [45] identified patients with metastatic melanoma who received SRS within 30 days of receiving an ICI infusion (pembrolizumab, nivolumab, and/or ipilimumab) as those who received “concurrent” SRS + ICI. The multivariate analysis showed that concurrent timing of SRS + ICI was an independent predictor of patients’ regional progression-free survival [hazard ratio (HR) = 0.17, P < 0.0001] compared with the non-concurrent schedule [45]. Likewise, in another study by Qian et al. [46], the SRS + ICI combination was identified as “concurrent” if SRS was given within 4 weeks away from the beginning or end of ICIs, and the results demonstrated a sharper trend of reduction in metastatic melanoma lesion volume of the concurrent group compared with that of the non-concurrent group at 1.5 months, 3 months, and 6 months (P < 0.0001). Also, in the study by Skrepnik et al. [47], concurrent SRS + ipilimumab (≤ 30 days) significantly improved the regional brain control (75% vs. 23.5, P = 0.03) and prolonged median CNS progression time (not reached vs. 5.7 months; P = 0.02) for patients with intracranial melanoma metastases compared with the non-concurrent group (> 30 days), though this advantage failed to be embodied in OS. Additionally, Le et al. [48] supported that concurrent SRS + ICI (≤ 30 days) significantly decreased distant brain failure of patients with melanoma and NSCLC brain metastases compared with non-concurrent SRS + ICI or no ICI (HR = 0.15; 95% CI 0.05–0.47, P = 0.0011). Kotecha et al. [49] more specifically defined “concurrent” SRS + ICI as those with an interval ≤ 5 half-life of the ICI and “immediate” SRS + ICI as those with an interval ≤ 1 half-life, and the results revealed that the “immediate” schedule met a superior BOR of − 100% while the “concurrent” group met a BOR of − 67%. Moreover, Yang et al. [50] conducted a meta-analysis including 9 studies to compare concurrent SRS/WBRT + ICI (interval ≤ 1 month) with sequential SRS/WBRT + ICI in NSCLC patients with brain metastases, and the results showed that the concurrent schedules significantly improved intracranial local control (HR = 0.19; 95% CI 0.09–0.42; P < 0.001). However, Cabanie et al. [51] added that the time-lapse between immunotherapy and SRS was not a significant predictor of local control. They reported a 76%, 76%, and 83% 1-year local control rate for patients with an interval of less than 7 days, an interval between 1 and 2 weeks, and an interval of more than 2 weeks, respectively.

In contrast, the extracranial and survival benefits patients with brain metastases receive from “concurrent” SRS + ICI seemed to correlate with the definitions on time intervals. The study by Qian et al. [46] showed that the difference in OS between the concurrent (interval ≤ 4 weeks) and non-concurrent groups was not significant (concurrent vs. non-concurrent, median 19.1 months vs. 9.0 months, P = 0.0691). Nevertheless, Koenig et al. [52] also defined “concurrent” SRS + ICI as those with the interval within 4 weeks in patients with brain metastases of various cancers, but the results showed significantly better OS (multivariable HR = 0.57; 95% CI 0.33–0.99; P = 0.044) and lower extracranial failure rate (multivariable HR = 0.60; 95% CI 0.42–0.87; P = 0.007) compared with the non-concurrent therapy. Likewise, the aforementioned meta-analysis by Yang et al. [50] showed that concurrent SRS/WBRT + ICI (within 4 weeks to 1 month) significantly prolonged OS compared with the sequential administrations of SRS/WBRT and ICIs (HR = 0.39; 95% CI 0.16–0.97; P = 0.043). Narrowing the intervals between SRS + ICI can lead to consistent conclusions. When narrowing the defined interval of “concurrent” SRS + ICI to less than 2 weeks in patients with brain metastases in NSCLC, melanoma, and renal cell cancer, Chen et al. [53] showed an OS benefit of concurrent SRS + ICI compared with non-concurrent SRS + ICI (non-concurrent vs concurrent HR = 2.40, P = 0.006) on multivariate analysis. Moreover, when the definition of “concurrent” SRS + ICI was narrowed to less than 1 week, the research of Scoccianti et al. [54] upheld that the concurrent group had a longer OS, and the time interval between SRS and ICIs had no impact on the toxicity.

Taken together, concurrent SRS + ICI typically leads to better local control of patients with brain metastases than that of nonconcurrent schedules. Whereas it is noteworthy that although the intervention arms of these studies are all defined as “concurrent”, different intervals between SRS and ICI are likely to affect the outcomes of patients with brain metastases. There is a trend that more narrowed definitions of the time intervals result in more favorable survival benefits, suggesting that shorter intervals between SRS and ICIs lead to better clinical outcomes.

The timing and sequence can be important factors in influencing the safety of SRS + ICI. However, contradictory results have been observed in the impact of the interval of SRS + ICI. Retrospective studies with relatively large samples have shown that SRS + ICI does not increase the rates of adverse events in patients with brain metastases compared with SRS alone, and the chance of radiation necrosis is rather low [49, 53, 55, 56]. Chen et al. [53] reported no increase in CNS toxicity or immune-related adverse events in the concurrent SRS + ICI (interval < 2 weeks) group compared with the noncurrent group (30% vs. 32%) in patients with brain metastases from various primary tumors based on a median follow-up of 9.2 months. Kotecha et al. [49] reported similar 12-month cumulative chances of radiation necrosis (3.2% vs. 3.5%) in the patients treated with immediate ICI (interval ± 1 half-life of the ICI) and all patients treated with concurrent or non-concurrent SRS + ICI. Nevertheless, some contradictory results suggested that, compared with SRS alone, SRS + ICI increased the likelihood of symptomatic necrosis within 4 years after SRS [57]. Koenig et al. [52] revealed that concurrent SRS + ICI (interval < 4 weeks) led to a higher risk of adverse radiation events (HR = 4.47, 95% CI 1.57–12.73, P = 0.005) compared with non-concurrent SRS + ICI based on a maximum follow-up of 36 months. Kiess et al. [58] also reported an increased chance of grade 3–4 adverse events in the concurrent group compared with the non-concurrent group based on a maximum follow-up of 50 months. The contradictions can be caused by distinct durations of follow-up of these studies because radiation-related or immune-related adverse events can occur even after 1 year from radiation [57]. For example, delayed radiation-induced vasculitis leukoencephalopathy related to SRS of brain metastases could be observed in 9 to 18 months after treatment [59].

In short, concurrent SRS + ICI may increase the incidences of adverse events compared with nonconcurrent administrations from the perspective of long-term follow-up, while RCTs with a large sample size and long follow-up are needed to draw a final conclusion.

The synergy between SRS and ICIs has been one of the hottest topics in treating brain metastases over years. All of the preclinical and clinical results above showed us a trend that, after 24–72 hours post-SRS, shorter intervals between SRS and ICI indicate more favorable clinical benefits for patients with brain metastases. And single high-dose irradiations appear to cause more potent immunostimulatory effects and more rapid BBB/BTB opening than fractioned low-dose ones. However, metastases from different primary tumors have varied radio-sensitivities and neo-antigen load. Additionally, the optimal sequence and interval may vary with the specific ICI administered, as CTLA-4 and PD-1/PD-L1 antagonists have different mechanisms. Hence, the optimal timing of SRS + ICI addressed in this review remains an active inquiry, which calls for well-designed prospective studies for a reliable answer.

Besides SRS, emerging therapeutic techniques are being developed to overcome the obstacles in the ICI treatment of brain metastases. Nanomedicine can target the BBB/BTB, tumor cells, immunosuppressive cells, APCs, or T cells to boost ICD and the intracranial efficacy of ICIs [98, 99]. A series of local treatments, such as focused ultrasound, tumor-treating fields, and laser therapy, can also modulate the permeability of the BBB/BTB and may have synergy with ICIs [13]. These advances may provide a solid backstop for ICI therapy for brain metastases.