Brain metastases (BMs) are the most common form of intracranial malignant neoplasms in adults and are associated with a dismal prognosis (1–3). Lung cancer is the most frequent primary tumor to develop BM (4) and globally accounts for approximately 40%–50% of all BMs (1, 5). Different risk factors have been associated to lung cancer, namely, cigarette smoking history, radon exposure, and occupational exposure to agents such as arsenic, asbestos, beryllium, cadmium, chromium, coal smoke, diesel fumes, nickel, silica, soot, and uranium. Other risk factors include history of prior malignancies such as lymphoma or head and neck cancers, history of chronic obstructive pulmonary disease, or pulmonary fibrosis (6). Among the different subtypes of lung cancer, it is estimated that up to 22%–57% of patients with non-small cell lung cancer (NSCLC) will develop BMs at some point of the history of their disease (7–10). Overall, 20% of these patients present BM at diagnosis and up to 57% develop BM over the period of the disease (7, 11). The prevalence of BMs in a subset of patients with NSCLC and oncogenic drivers can be higher in some cases and increase throughout the course of their disease. These patients experience different rates of BM at baseline and through the history of their disease, as follows for the most frequent oncogenic drivers: EGFR mutations are 24.4% and 52.9% (12), ALK rearrangements are 23.8% and 72% (12, 13), and ROS1 rearrangements are 19.4% and 34% (13). Regarding the less common genomic alterations, the percentages for baseline BM and throughout the history of the disease are as follows: BRAF V600E mutations are 31% (14) and not reported (possibly due to the low prevalence of BRAF mutations in NSCLC), HER2 mutations are 14% and 28% (15, 16), RET rearrangements are 25% and 46% (17), and KRAS mutations are (G12C and non-G12C) 24% and 35.1% (15, 18). Some patients with oncogenic drivers and BMs benefit from targeted therapies with tyrosine kinase inhibitors, which can surpass the blood–brain barrier (BBB), as evidenced by significant intracranial response rates in different clinical trials (19–22). For patients with small cell lung cancer (SCLC), it is estimated that 10%–20% of patients will present with BM at diagnosis, and up to 50%–80% of patients will develop BMs at some point of the history of their disease (23–26).

Detecting BM at baseline has been associated to a worse prognosis, and developing BM during ongoing therapies reflects tumor progression and resistance to treatment, associated to the disruption of the physiologic BBB.

Current guidelines recommend the use of brain magnetic resonance imaging (MRI) with contrast or, when not possible, brain computed tomography with contrast for patients with NSCLC and SCLC stages II–IV to detect BM (27, 28). Although BMs have a negative impact in survival, in recent years, patients’ outcomes have improved significantly (29, 30). Some of the treatments that have contributed to improving survival of these patients include surgery, stereotactic radiosurgery, whole-brain radiation therapy, prophylactic cranial irradiation, and the addition of better systemic therapies (10, 31–33). To develop effective treatment strategies, it is necessary to understand the genomic and molecular mechanisms leading to BM. The purpose of this review is to provide an integrated synthesis of the epidemiology of BM in lung cancer patients. To describe its pathophysiology and molecular mechanisms, as well as its genomic landscape. Furthermore, we intend to provide a cohesive summary of the clinical evidence for the different treatment strategies, focusing on immunotherapy, and immunotherapy based combinations. Finally, we present future perspectives for the management of BMs.

The mechanisms leading to BMs are complex and different mechanisms are required; among them, local invasion, intravasation, survival in circulation, extravasation, and tissue colonization represent the most remarkable ( Figure 1A ) (34, 35). The most difficult steps in this scenario are extravasation and colonization, since the BBB must be bypassed (35). Under physiological conditions, the BBB regulates the homeostasis of the central nervous system (CNS); it provides the brain with nutrients, limits the transportation of ionic substances and other molecules, and functions as a critical obstacle for drug transportation to the brain (36–38). Key components of the BBB include endothelial cells, pericytes, astrocytes, and the extracellular matrix. Endothelial cells form tight junctions that control transportation and entry of different molecules, allowing a restricted permeability of the BBB (37, 39, 40). These cells secrete growth molecules such as PDGF-β, TGF-β, and Vascular Endothelial Growth Factor (VEGF), which modulate pericytes (40). Pericytes wrap blood vessels through their end feet processes and regulate the vascular function. They are involved in angiogenesis, neovascularization, and antigen cell presentation (under certain conditions), as well as communicating with astrocytes within the neurovascular unit (41–43). Astrocytes are necessary for different metabolic processes in the brain, including metabolization of different substances and production of antioxidants, as well as regulation of signaling pathways. They contribute to the function of the BBB and secrete growth factors targeted to the endothelial cells (40). Another important component of the BBB is the extracellular matrix, composed of the basal lamina that surrounds the BBB (43). Disruption of the extracellular matrix has been associated with an increased permeability of the BBB and as a consequence development of BM (40). In addition to the BBB, other components of the CNS environment include oligodendrocytes and microglia; the latter are the innate macrophage cells of the brain and can exert disrupting effects in the BBB, facilitating the formation of BM (43, 44).

Prior to the development of BM, an adequate microenvironment must be secured for the establishment of the metastatic cells; this pretumoral microenvironment is referred to as the premetastatic niche ( Figures 1A, B ) (45). Different molecules contribute to the formation of the premetastatic niche, as well as to the disruption of the BBB, including VEGF, PLGF, and matrix metalloproteinases, which elicit changes in the tight junctions leading to a disruption of BBB (46–48). Other contributing mechanisms accounting for the conditioning of the premetastatic niche are tumor-secreted exosomes, which are vesicles surrounded by a bilayer of lipids. Exosomes contain nucleic acids, lipids, and proteins, which can regulate integrins and modulate glucose uptake in astrocytes (49, 50). MicroRNAs (miRNAs) are a type of nucleic acids contained in exosomes; these are characterized by being small non-coding RNAs that can alter gene expression. Different miRNAs have been associated to the development of BM, such as miR-378, which is relevant to the promotion of cell migration, invasion, and angiogenesis via upregulation of PRKCA (51–53). A different type of nucleic acids involved in BMs are long non-coding RNAs (lncRNAs), which are RNA sequences longer than 200 nucleotides, which are not translated into functioning proteins (54). The lncRNA MALTA1 has been associated with promoting BMs by inducing epithelial to mesenchymal transition (EMT), whereas the lncRNA HOTAIR contributed to BMs by increasing cell migration and anchorage (55, 56). In addition, contributing factors to the disruption of the BBB can be attributed to endothelial cells, which express different adhesion molecules prompted by tumor cells, facilitating invasion conditions for these tumor cells (57).

Once the BBB has been disrupted, the blood–tumor barrier (BTB) is established ( Figures 1C, D ). The BTB is characterized by alterations in the transportation of molecules, changes in protein expression profiles involving transportation channels, and a heterogeneous permeability to different molecules (44). However, this increased permeability does not necessarily translate into a more favorable anti-cancer drug delivery (44). As a former BBB, the BTB presents similar components, such as endothelial cells, which in the BTB scenario upregulate TNF receptors 1 and 2 and develop alterations in cell adhesion proteins leading to a loss of tight junctions. The loss of endothelial tight junctions increases paracellular permeability and downregulates specific transporters related to influx and efflux of molecules (58). Other cells that contribute to the disruption of the BBB are pericytes, which locate around blood vessels, forming double layers of cells and producing collagen, which leads to a thickening of blood vessels (59). An upregulation in the expression of VEGF in the BTB has been described; however, it does not always translate into an effective formation of blood vessels (60, 61). Other components associated with the establishment of the BTB include the production of inflammatory molecules, membrane proteins, and growth factors, such as claudin-5 (62, 63). Following the disruption of the BBB and the formation of the BTB, metastatic cells can then settle in the brain. Once they are established, tumor cells interact with the innate microenvironment of the brain. The tumor microenvironment is a complex ecosystem composed of tumor cells, as well as non-malignant cell types such as endothelial cells, astrocytes, pericytes, and immune cells (64).

Once, in the brain parenchyma, the tumor cells trigger an inflammatory response led by astrocytes, microglia, and other immune cells (65). Astrocytes are able to exert both pro-tumoral and antitumoral functions (66). Initially astrocytes exert antitumoral effects through the production of nitric oxide and plasminogen activator, as well as creating an astrocytic wall, which separates the metastasis from the brain parenchyma (67–69). Plasminogen activator triggers the production of plasmin, which activates the FasL pathway and deactivates L1CAM, ultimately suppressing BM. This results in a death signal for cancer cells and inhibition of metastatic growth. In response to the deleterious effects exerted by plasmin to the tumor cells, the latter secrete serpins which elicit inhibitory effects against the plasminogen activator and therefore counteract the antitumoral functions of astrocytes (68, 70). A worth noting pro-tumoral effect of astrocytes is the ability to establish gap junctions with tumor cells, facilitating the transfer of cGAMP to astrocytes. Consequently, the STING pathway is activated, leading to the production of cytokines such as IFN-α and TNF, promoting tumor cell growth and chemoresistance. Additionally, these gap junctions hinder the uptake of calcium by tumors, further contributing to tumor growth and chemoresistance (71–73). In addition, reactive astrocytes (RAs) are astrocytes that undergo changes in their molecular, morphological, and functional states as response to pathologic situations in the surrounding environment (74). Tumor cells promote the activation of STAT3 in RA, contributing to the modulation of the immune system, and promote the establishment of a pro-metastatic environment through a decrease in the activation of CD8+ T cells (75). RA may also contribute to tumor proliferation through the secretion of IL-6, TNF- α, and IL-1β (69).

The microglia are the resident macrophages of the CNS and not derived from the bone marrow. However, bone marrow–derived macrophages can reach the CNS when the BBB is disrupted, as a response to disturbances in the CNS. Both of them share lineage and activation markers, and in the context of malignant neoplasms, these cells are referred to as tumor-associated macrophages (TAMs) (76, 77). These macrophages can be subsequently divided according to their functional characteristics; M1-like macrophages are usually described to be proinflammatory and are stimulated by Toll-like receptor ligands, as well as IFN-γ and TNF-α. They are thought to exert tumor-suppressive functions, through the production of IL-1, IL-12, and nitric oxide, among other factors (78, 79). On the other hand, M2-like macrophages are usually considered anti-inflammatory; are activated by IL-4 and IL-13; and produce molecules such as TGF-β, arginase, IL-10, and profibrotic factors. M2-like macrophages are associated with tumor promoting functions through the inhibition of CD8+ T-cell proliferation (78, 79) ( Figure 2 ).

Tumor-infiltrating lymphocytes (TILs) in BMs are usually composed of different subsets of CD8+ T cells and CD4+ T cells, including T-regulatory cells (Tregs) (80). TILs in BM have been associated to peritumoral edema in the brain. Some of the main TILs present in BMs include CD3+, CD8+ cytotoxic TILs, and CD45RO+. A high density of these subpopulations has been associated with a prolonged overall survival (OS) (81). For patients with metastatic SCLC, resection of BM is not a standard of practice; thus, characterization of BM in these patients is not often achieved (82). However, a study in 32 patients with SCLC and resected BM studied the immune microenvironment of these metastases. CD3+ TILs, CD8+ TILs, CD45RO+ TILs, and FOXP3 + TILs were present in over 45% of the studied specimens. PD-L1 expression was positive in 75% of the studied BM, whereas for TILs and TAMs, the percentage of expression was 25% and 28.1%, correspondingly. The expression of PD-L1 on TILs was associated with improved survival (83). In a different study including 49 patients with NSCLC evaluated peripheral blood immune cells, including immunosuppressive monocytes, myeloid-derived suppressor cells, and Tregs. The study found that patients with BM had an increased number of PD-L1+ monocytes, myeloid-derived suppressor cells, and Tregs in peripheral blood, compared to patients who did not have BM and healthy controls. The study also found that patients with elevated numbers of peripheral myeloid PD-L1–expressing cells were associated with a worse progression-free survival and OS (84).

Different studies have evaluated primary NSCLC tumors and paired BM to assess the tumor microenvironment, as well as the expression and mutational profiles in both samples. A study in 39 patients with NSCLC and resected BM performed DNA sequencing of 158 hotspot mutations, TCR-β sequencing analysis, as well as immunohistochemistry staining to assess differences and similarities between the studied specimens. The variant allelic frequency of mutated genes was higher in BM compared to primary tumors, with a median difference of 21.6%. BM showed an inhibition of dendritic cell maturation, Th1 immune response, as well as leukocyte extravasation signaling, and a reduced expression of the adhesion molecule VCAM1. Furthermore, a bioinformatic evaluation of the immune subpopulations based on gene expression detected that the abundance of Th1 or CD8 T genes was lower in BM compared to primary tumors, which was associated with reduced levels of CD8+ T, dendritic cells, and macrophages in BM. The T-cell richness and T-cell densities were significantly lower in BM; however, levels of bone marrow–derived M2-like macrophages were higher in BM. The results from this study suggest that the immune tumor microenvironment of BM presents increased immunosuppressive features compared to the one in primary tumors (85). Another study in 23 patients with matched primary NSCLC and BM identified that PD-L1 expression in tumor cells was higher and CD8+ TILS was lower in BM compared to primary tumors. However, patients categorized as CD8 high stromal TILs in BM had a trend toward a better OS, compared to those with CD8 low stromal TILs (86). In addition, a study involving 43 NSCLC patients evaluated samples of matched primary tumor and BM, identifying higher numbers of neutrophils, CD4+ T cells and dendritic cells, and lower fractions of M1-like macrophages and Tregs in BM. The study also detected a downregulation in the expression of CTLA-4 in BM, as well as lower expression of PD-L1 and CD8A between primary tumors and BM. Interestingly, an analysis between patients with metachronic tumors who received adjuvant treatment and later had a brain relapse, compared to those who were not exposed to prior treatments and later had a brain relapse, showed no differences except for the cell fraction of natural killer cells, which were higher in the group that received adjuvant treatment. Furthermore, this study identified that the immune microenvironment differed between primary tumors and BM in the metachronic group compared to the synchronic group. The results of this study concur with the previous ones in detecting an immunosuppressed tumor microenvironment in BM, compared to the primary tumor (87). A study in 34 patients with NSCLC and BM, who underwent surgical resection or autopsy, compared primary tumors with BM. This study found that only CD204+ cells, corresponding with macrophages, were higher in tumor areas of BM compared to the primary tumor, whereas CD4+ T cells, CD8+ T cells, and CD+ FOXP3 T cells were higher in the primary tumor. The study found that higher densities of CD4+ T cells and CD8+ T cells in both BM and the stromal areas of BM were associated with a higher OS (88).

A study compared the expression of PD-L1 in primary and metastatic lesions; 747 BM samples from patients with NSCLC were included. The study compared the expression of PD-L1 between primary tumors and BM and found a discordance in the expression of PD-L1. BM had a low or negative expression of the biomarker, highlighting the heterogeneity of PD-L1 according to the tumor sample site (89).

Other studies have focused on genomic profiling of NSCLC primary tumors and BM. As such, a study found that non-synonymous tumor mutational burden (TMB) was higher in BM compared to primary tumors; however, the neoantigen load was similar in BM and primary tumors. The study also found a reduced T-cell heterogeneity in BM associated to a immunosuppressive tumor microenvironment (90). A different study profiled BM and primary tumor samples from patients with NSCLC and identified a discrepancy in the mutational landscape between the samples, with a median of 8.3% of shared genetic mutations and higher numbers of somatic mutations in BM. Different pathways were enriched in BM compared to primary tumors, including invasion/metastasis and metabolic associated pathways. The study also found a phylogenetic divergence between BM and primary tumors, indicating parallel progression models for each of the lesions (91). Another study in 11 patients with metastatic lung adenocarcinoma, assessed the mutational landscape of primary tumors and BM. The study found that more unique mutations were present in BM, and identified mutations in FAM129C and ADAMTS present specifically in BM. Alterations in the APOBEC signature were also higher in BM compared to primary tumors (92). A different study performed in samples from 2,309 patients with lung adenocarcinoma, including 238 samples of CNS metastases identified genes associated to the development of BMs. Alterations in the Hippo pathway were associated to a shorter time to developing CNS metastases, and STK11 alterations were more prevalent in CNS metastases (93). Another study evaluated 73 samples of BM from lung adenocarcinoma and compared them against a control population of 503 primary lung adenocarcinoma tumors. The study found different genes that were considered as candidates for the development of BM through the identification of regions with higher amplification frequencies or deletions. These genes included MYC, YAP1, MMP13, and CDKN2A/B (94). A study in 51 patients with lung adenocarcinoma, as well as squamous cell carcinoma, identified that BM had a higher number of somatic copy number alterations. Gene alterations associated with BM driving mechanisms include CDK12, DDR2, ERBB2, and NTRK1 (95). A different study in 54 paired samples from BM and NSCLC patients identified that most driver alterations were present both in primary tumor and BM. Private alterations to one or the other sites were present in 22%–26% of the cases. KRAS mutations were more frequent in primary tumors that developed BM compared to data from The Cancer Genome Atlas (CTGA) (96). The preceding studies highlight that BM are composed of an immunosuppressive tumor environment, which has been associated to a deleterious prognosis. Furthermore, different mutational patterns in primary tumors versus BMs represent the tumor heterogeneity between the two tumor sites and could in part explain the development of BM and the different response to therapy. The most relevant results from these studies are presented in Table 1 .

Different strategies have been used to approach treatment of BM; however, there is not a definitive consensus regarding the sequence of treatment. Neurosurgery should be considered in patients with large tumors that exert a mass effect as well as in patients with minimal intracranial disease. Patients with multiple BM or a single BM could be considered for whole brain radiotherapy (WBRT) or stereotactic radio surgery (SRS) or stereotactic radiotherapy (SRT). However, radiotherapy is not recommended for patients with a Karnofsky Performance Status (KPS) ≤50 or <70 and no systematic therapy options. The use of WBRT, SRS, or SRT can also be considered for patients with a KPS <70 or with comorbidities impeding neurosurgery. Combination of SRS plus WBRT (with our radioprotective strategies) is associated with higher rates of cognitive impairment. Thus, the combination of SRS with WBRT could be used with hippocampal avoidance or administration of memantine to decrease neurological toxicity. In some instances, local treatment can be deferred to avoid potential adverse effects. Situations that might prompt to a deferral for local treatments include asymptomatic BM and a “favorable” location of the BM and systemic treatments with high CNS penetrance and responses (i.e., third-generation EGFR tyrosine kinase inhibitors) (97). Following local treatment strategies, a systemic treatment should be considered, such as immunotherapy, tyrosine kinase inhibitors, chemotherapy, or combination therapies. It is worth noting that administering concurrent radiotherapy to the brain and chemotherapy is not recommended, since it can trigger the development of cognitive impairment, white matter damage, and radionecrosis (98–100).

More recently, in the era of immunotherapy, PD-1 or PD-L1 inhibitors alone or in combination with chemotherapy or anti–CTLA-4 agents have demonstrated activity both NSCLC and SCLC patients, with BM (27, 82, 101–103). The use of Immune checkpoint-inihbitors (ICI) is an attractive strategy to stimulate the immune response of the immunosuppressed tumor microenvironment in BM (104, 105). ICI has demonstrated to be effective in the treatment of BM in other tumors including melanoma, which is coined one of the pioneers of immunotherapy. Clinical trials have demonstrated that patients with asymptomatic BM benefit from the nivolumab–ipilimumab combination, reporting an intracranial benefit of 57%, and an OS at 3 years of 71.9% in this population (106–108). Different clinical trials and retrospective studies have evaluated the effect of ICI alone or in combination with chemotherapy or radiotherapy in BM from lung cancer; these studies are summarized in Table 2 . The current use of immunotherapies is based on regulation from medical agencies, as well as clinical guidelines. Current approvals can be summarized as follows: for tumors with a PD-L1 expression 0%–49% treatment recommendations include platinum-based chemotherapy with anti–PD-1/PD-L1 ± ipilimumab ± bevacizumab (141). Tumors with a PD-L1 expression ≥50% benefit from the prior mentioned regimens or anti–PD-1/PD-L1 agents in monotherapy. For patients with SCLC, the current recommendations are the use of chemoimmunotherapy in the first line of treatment (82).

The development of different therapies for lung cancer and BM can be represented in a timeline, as depicted in Figure 3 (142, 143).

Different biomarkers have been described and are currently being developed to detect and characterize genomic, molecular, or TME alterations in BM, as biomarkers of prognosis or targeted treatment. As such, detection of circulating tumor DNA (ctDNA) in the cerebrospinal fluid (CSF) can be an attractive strategy and less invasive than conventional biopsies. ctDNA allows for the characterization of different genomic alterations within CNS metastases and provides a landscape of tumor heterogeneity (151). High TMB has been associated with response to ICIs, and high TMB has been detected in BM (89). Using TMB as a biomarker of response is an attractive strategy, since it can also be detected in ctDNA from CSF (152). However, there is no clinical evidence that a high TMB in BM correlates to a benefit from ICIs, and the use of ctDNA from CSF in patients with BM is not currently a standard practice.

The most validated biomarker associated with response to ICIs is detection of PD-L1, either in tumor cells, immune cells, or both and to a lesser extent TMB. High expression of both of these biomarkers has been associated with an increased survival in NSCLC patients treated with ICI (153). As previously presented TILs in BM have been associated with an improved survival in patients with NSCLC and SCLC (81, 83) and could potentially serve as biomarkers of benefit. In addition, a different biomarker is a high neutrophil-to-lymphocyte ratio, which has been associated with a worse prognosis in patients being treated with radiosurgery or neurosurgery for BM (154, 155).

Although not particularly associated with BM, an increased representation of different bacteria from the gut microbiota has been detected in the stool of patients who responded to treatment with ICI or ICI-based therapies. Among the most representative bacteria, B longum, E hirae, A muchiniphila, B fragilis, Collinsella aerofaciens, and F prausnitzii stand out as biomarkers associated with response (156–161). Furthermore, clinical trials in melanoma patients have performed fecal microbiota transplantation (FMT) from patients who received ICI and responded to patients with who received the same treatment and did not respond. Patients with primary resistance to ICI who received a FMT from responders had an approximate 20% rate of response following the FMT and continuing treatment with ICI. These results show that manipulation of the gut microbiome through FMT is a strategy useful to overcome resistance to ICI (162, 163).

Regarding future therapeutic strategies, targeting other regulatory checkpoints, such as LAG3, TIGIT, or TIM3, either alone or in combination with anti–PD-1/PD-L1 inhibitors are being explored as means to overcome resistance to current immunotherapies. Many clinical trials with different combinations are currently undergoing. Furthermore, suppression of immune cells of molecules that elicit an immunosuppressive environment could also be a target, such as M2-TAMs, MDSC, and anti-TGFβ treatments. Other strategies are being used in NSCLC patients with BM, such as clinical trials involving vaccines with mRNA tumor antigens pulsed into dendritic cells (NCT02808416) or administration of autologous dendritic cells to the tumor through the intrathecal space (NCT03638765). Results from these studies are expected to provide a new strategy to treat these patients.

BMs represent a devastating event in patients with lung cancer, with alarming numbers regarding its incidence. The development and establishment of BMs is a complex process that needs to be studied from different perspectives, including genetic, epigenetic, immunologic, physiologic, and those associated with the BBB and the brain parenchyma. The cascade of signaling events leading to BMs is complex, and still not completely understood, in part due to a limited access to BMs samples, and the heterogeneity among these tumors. Immunotherapy represents a milestone in the treatment of patients with lung cancer, and it has clearly demonstrated that either alone or in combination with chemotherapy or radiotherapy, patients with BM derive a benefit. However, selecting patients who will benefit from one combination of therapies or another, continues to represent a challenge, and so does the sequence of treatment administration. Although important breakthroughs have been achieved with the arrival of immunotherapy, patients with lung cancer and BMs still present a deleterious diagnosis which needs to be approached by a multidisciplinary team and requires further research.

AR-H performed the literature search and developed the outline as well as the initial draft. EA edited the manuscript, added further content and provided supervision. All authors contributed to the article and approved the submitted version.