1 INTRODUCTION

Cutaneous melanoma is the deadliest form of skin cancer, with a propensity to aggressively metastasize to multiple organs.1 One common site of melanoma metastasis development is the brain, with 40–60% of patients with advanced melanoma showing evidence of CNS involvement.2, 3 Left untreated, MBMs progress rapidly, with most patients dying within 3 months.4 As such, MBMs development accounts for 54% of melanoma deaths.5 Melanoma cells disseminate to the brain through the circulatory system, with MBMs most commonly forming in areas of the brain with the highest blood flow. Around 80% of melanoma brain metastases (MBM) are located in the cerebral hemispheres, with a further 5% and 15% of MBM being located in the brain stem and cerebellum, respectively.6 Initial clinical presentation of MBM includes headache, neurological impairment and seizures.7 Risk factors for MBM development are male gender, head or neck as a primary disease site, the presence of visceral or nodal metastases, metastases at three or more visceral sites, high serum lactate dehydrogenase (LDH: which is a marker of higher disease burden) and high Clark's level/Breslow thickness of the primary melanomas.7, 8

For many years, MBM development had a dismal prognosis with no systemic therapies demonstrated to alter the natural history of the disease. Radiation therapy and surgery were the mainstays of treatment. The recent development of effective systemic therapies for advanced melanoma including immune checkpoint inhibitors (anti-PD-1 and anti-CTLA-4) and the BRAF-MEK inhibitor-based targeted therapies has dramatically improved outcomes and overall survival.9-12 These therapies have been also evaluated in patients with MBM, with good rates of response frequently noted.13, 14 Despite these successes, differences have been noted in the duration of therapy responses between intracranial and extracranial metastatic sites. In the case of targeted therapy, rates of response in MBMs, while similar to visceral metastases, seem to be shorter in duration.13 For immunotherapy, the combination of anti-PD-1-CTLA-4 checkpoint inhibitors is associated with response rates of ~50% in patients with asymptomatic brain metastases but yet much lower rates of ~20% in patients with symptomatic MBM.15 It is likely the unique microenvironment of the brain accounts for these differences in response, but this remains poorly characterized. It is further possible that physical constraints, such as poor drug transport into the brain due to the blood-brain barrier, the unique diffusion characteristics of the brain and high interstitial pressure, may also contribute to the reduced effectiveness of these drugs. Strategies to improve durations of systemic therapy response in patients with MBM are urgently needed. In this review, we will briefly describe the mechanisms by which melanoma cells seed to the brain and form metastases. We will then discuss recent data highlighting the role of the unique metabolic and immune environments of the brain in MBM progression before turning our attention to the development of novel therapeutic strategies for MBM and the clinical studies that are currently underway.

2 THE BIOLOGY OF MELANOMA BRAIN METASTASES 2.1 Dissemination of melanoma cells to the brain

The development of distant metastases is a complex, multi-step process that involves the cancer cells leaving the primary tumor, migrating into blood vessels, surviving the high shear stress of the circulation, arresting in microcapillaries, exit from the vasculature and the eventual establishment of new micrometastases16 (Figure 1). The progression of initial micrometastases to macrometastases often requires the cancer cells to co-opt neighbouring host cells, reprogramming them to provide important survival, metabolic and growth signals.17, 18 Although it is technically challenging to visualize the process of melanoma dissemination to the brain, recent advances in 2-photon microscopy and new methods to directly image the brain vasculature through cranial window chambers have shed new light on this process.16 These analyses have shown that the melanoma cells are carried through the circulatory system of the brain in a passive manner until reaching the small microvessels.16 Here, the cells arrest and then remain quiescent for some time before beginning to extravasate. The rate of cancer cell migration into the brain parenchyma is typically much slower than that observed for other visceral organs, perhaps explaining why brain metastases typically develop much later in tumor progression than metastases at other sites. Once extravasated, the melanoma cells adhere to the external (abluminal) side of the endothelial cells and appear to co-opt them.3 Most initial micrometastases appear to grow along the external linings of the blood vessels (Figure 1).

Schematic showing the process of MBM initiation. Melanoma cells leave the primary tumor and enter the circulation. When in the microvessels of the brain, some cells migrate through the BBB and grow along the abluminal side of the endothelial cells. Eventually, some of these cells emerge from quiescence and form micrometastases, a limited number of micrometastases can form macrometastases

One major challenge for melanoma cells establishing new brain metastases is penetrating the protective blood-brain barrier (BBB): the highly selective and tight interface between blood and parenchyma of the brain. The BBB consists of a physical cellular barrier, layers of extracellular matrix (ECM) and drug/solute transporters.19-22 The cellular barrier comprises of endothelial cells linked by multiple tight junctions that prevent passive diffusion of ions and other polar solutes between endothelial cells. The endothelial cells are supported by tightly associated layer of pericytes, along with astrocytes that form perivascular sheath.23 The disruption of BBB integrity is a key step in the development of MBM. This is a multi-step process that involves the dilation of the cerebral microvasculature, the initiation of an inflammatory response, increased secretion of pro-angiogenic factors like vascular endothelial growth factor (VEGF) and loss of tight junction integrity.24 Melanoma cells have multiple strategies to disrupt the BBB. These include physical disruption of the barrier, such as adopting an amoeboid, rounded phenotype with cellular protrusions that serve to push the endothelial cells apart (such as that mediated through EphA2) and increased levels of pro-invasive integrins (such as α3β1, αvβ3 and α4β1).25, 26 Other mechanisms involve upregulation of multiple enzymes that degrade the ECM of the BBB including heparanase (HPSE), which degrades heparan sulphate chains of proteoglycans in endothelial cells and the matrix metalloproteinases (MMP)-2 and 9.22, 27 Other proteases have also been reported to be important for BBB breakdown, including cathepsin-S, which can be either secreted by the tumor cells, or other host cells such as macrophages.28 In this instance, cathepsin-S disrupts the BBB through cleavage of the tight junction protein JAM-B.28 Crosstalk between melanoma cells and host astrocytes may also contribute to MBM development with studies demonstrating the existence of a pro-inflammatory IL-23 driven signalling loop from the astrocytes that increases MMP-2 secretion from the melanoma cells.29 Tumor-derived TGF-β2 was also found to decrease BBB integrity through downregulation of occludin and claudin expression in the endothelial cell barrier30 (Figure 1).

Melanoma cells are derived from the neural crest, and there is some evidence that chemokines or brain-derived ligands may play a role in the homing of melanoma cells to the brain vasculature.16 Chemokine receptors such as CCR7 have been implicated in melanoma metastasis to CCL21-rich lymph nodes, CXCR4/CCXCL12 has been associated with pulmonary metastasis development, and high levels of CCR4 expression have been associated with brain-specific metastasis.31 In xenograft models of melanoma brain metastasis development, increased expression of the endothelin receptor B (EDNRB) was associated with increased development of metastases.32 Melanoma cells also exhibit upregulated levels of p75NTR and TrkC, receptors for nerve growth factor (NGF) and NT-3 which are secreted by astrocytes at the stromal-tumor interface, potentially increasing the homing of the melanoma cells to the brain.32, 33

There has been some suggestion from other cancers (mostly breast and lung) that metastatic cells may acquire additional mutations (in genes such as EGFR/HER2 and the PI3 K/AKT/mTOR pathway) that increase the likelihood of dissemination to the brain.34 A recent analysis of a large cohort of lung adenocarcinoma brain metastases reported an increased frequency in Myc, YAP1, MMP13 and CDKN2A/B mutations compared to case control studies.35 Other recent studies revealed MBMs to have a similar profile of driver mutations to cutaneous melanomas in the TCGA data set.36 Interestingly, analysis of the MBM that emerged as the first site of recurrent disease demonstrated an enrichment of mutant KRAS in up to 8% of cases. The identified KRAS mutations were clonal and concordant with the extracranial tumors, suggesting that these may have been present in the primary melanoma.36 It was suggested that melanoma patients harbouring KRAS mutations in their extracranial tumors should be targeted for increased surveillance for future MBM development.

2.2 The interaction of host cells in MBM progression

The microenvironment of the brain parenchyma is hostile to cancer cells, and there are several challenges that melanoma cells must overcome in order to become established as macrometastases. One of the major protective mechanisms found in the brain are the high levels of proteases, such as plasmin, which serves to activate pro-death signal ligands on astrocytes such as FAS ligand – inducing apoptosis in the invading cancer cells.28 As such, brain-metastatic cancer cells often secrete high levels of protease inhibitors such as the Serpins to neutralize these effects.28, 37 Once these mechanisms are overcome and the MBMs become established in the brain, they are quickly able to recruit host cells to support their growth.37

One of the most important cell types implicated in MBM development are the glia. Glia are a family of non-neuronal cells that provide structural, housekeeping and metabolic support to the neurons of the brain. Multiple types of glia exist including astrocytes (analogous to fibroblasts), microglia (similar in function to macrophages), ependymal cells (involved in the production of cerebrospinal fluid: CSF) and oligodendrocytes (which provide insulation to axons).38 Astrocytes are glia that constitute the major non-neuronal cell type in the brain. They primarily play roles in the maintenance of the brain architecture through regulation of metabolic homeostasis and regulating the formation, maturation and stability of synapses.39 They also are critical for maintenance of the BBB, with 80–99% of the basement membrane of the BBB being composed of astrocyte-foot processes.39 Astrocytes also play a protective role and undergo a hypertrophic/hyperproliferative response following injury and inflammation, a process termed reactive gliosis. Multiple studies have suggested that astrocytes support the growth of brain metastases through the secretion of multiple growth factors, chemokines and cytokines including interleukin (IL)-6, tumor necrosis factor (TNF)-α and IL-1β.37, 40, 41 There is evidence that astrocytes can communicate directly with the tumor cells, inducing them to express increased levels of anti-apoptotic and EMT-associated genes such as BCL2L1 and TWIST as well as ECM degrading metalloproteinases such as MMP-2 and MMP-9.40, 42 There is also evidence of cell-cell communication between cancer cells and astrocytes through membrane-spanning Gap junction proteins, such as Connexin (Cx)-43 and adhesion proteins such as the protocadherins.43 Among these, Cx43 has received special attention as a potential mechanism by which the cancer cells transfer the second messenger cGAMP to astrocytes leading to activation of the STING pathway.43 Activation of STING in astrocytes has been shown to reciprocally increase cancer cell survival through NFκB and STAT1 signalling.41, 43 Suppression of activation astrocytes in experimental models through strategies such as PDGF-B receptor antagonists has been shown to slow the growth of breast cancer brain metastases.

Astrocytes have also been demonstrated to be drivers of drug resistance in brain metastases. Early studies demonstrated astrocytes to protect tumor cells from chemotherapy by sequestering intracellular calcium through increased gap junction communication.41 There is also evidence that astrocytes can protect cancer cells from TRAIL-mediated apoptosis.43 Work in MBM has additionally demonstrated that astrocytes modulate sensitivity to targeted therapy. Multiple studies have now shown that CSF from rats, as well as patients with MBM and leptomeningeal melanoma metastases, (metastases to the leptomeninges: the CSF space that surrounds the brain) can reduce the sensitivity of melanoma cells to the BRAF inhibitor vemurafenib and the BRAF-MEK inhibitor combination.44-46 Similar findings were also reported for melanoma cells cultured in in vitro with astrocyte-conditioned media, with PI3 K inhibitors being found to reverse these protective effects.47 It is likely that CSF contains multiple factors that help shape melanoma behaviour. Recent work from our group demonstrated that CSF from patients with LMM induced a transcriptional programme in melanoma cells associated with PI3 K/AKT, integrin, B-cell activation, S-phase entry, TNFR2, TGF-β and oxidative stress responses.44 Among these, increased TGF-β levels in the CSF correlated with LMM progression and were found to induce BRAF inhibitor resistance in the melanoma cells.44, 48

In addition to the secretion of soluble growth factors, astrocytes have been also shown to epigenetically shape cancer cell behaviour through the release of exosomes containing microRNAs.49 In one prominent recent example of this, it was found that breast cancer cells with expression of the tumor suppressor PTEN frequently lost this when grown as brain metastases. Removal of these cells from the brain led to expression of PTEN being restored. Further analysis showed that the breast cancer cells took up exosomal microRNAs from the brain microenvironment, leading to silencing of PTEN.49 Loss of PTEN in the breast cancer cells was critical for tumor progression through a mechanism involving increased CCL2 expression, myeloid-derived suppressor cell (MDSC) recruitment and the suppression of anti-tumor immunity. Although most data to date have implicated astrocytes in brain metastasis progression, there is also evidence for microglia involvement. Work in breast cancer has demonstrated a role for microglia in increasing brain metastasis growth and survival, in part through activation of WNT signalling.42, 50, 51 Recent single-cell RNA-Seq studies have shown that microglia are highly plastic and can adopt a wide range of transcriptional states.38, 52 It remains to be determined whether brain-metastatic cancer cells reprogramme the microglia to a less immune-competent phenotype that favours tumor progression.

2.3 The metabolic environment of MBM

Among all sites of metastasis, the brain and CNS have a unique metabolic environment that may help shape tumor progression. One key feature of the brain metabolic environment is a reliance upon high levels of glucose oxidation that is tailored to the high energy demands of neuronal cells.53, 54 The fat content of the brain is very high (the highest fat content of any organ), and it is composed of almost 50% lipids.55 The brain environment, particularly the CSF and brain interstitial fluid, is also depleted in many nutrients compared to the surrounding plasma, and it contains very low levels of amino acids.56-58 Taken together, the CNS represents a harsh microenvironment, comprising a unique milieu that successful brain-metastatic cancers must metabolically adapt to. Studies in breast cancer have shown that cells metastasizing to the brain reprogramme their metabolism to a dependency upon glycolysis, the TCA cycle and oxidative phosphorylation.59 Upregulation is also seen in the pentose phosphate system and the glutathione metabolism, which provide protection against the high levels of reactive oxygen species (ROS) encountered.59 RNA-Seq and mass spectrometry-based metabolomic analysis of melanoma metastases from the brain and extracranial sites showed that the MBM were enriched for oxidative phosphorylation (OXPHOS).60 This metabolic dependency was of clinical significance, with small molecule OXPHOS inhibitors being found to prolong survival of mice with intracranial melanoma xenografts.60 Significantly, a switch to OXPHOS has been associated with resistance to BRAF-MEK inhibitor therapy in melanoma, potentially explaining why patients with MBM tend to show shorter progression-free survival (PFS) on BRAF-MEK inhibitor therapy than individuals with disease at extracranial sites.60-62

The fat content of the brain is distinct from that of adipose tissue in that its primary lipid content is polyunsaturated fatty acids. The lipid metabolism of the brain is primarily regulated by the astrocytes.63, 64 Recent work has shown that brain-metastatic melanoma cells take advantage of the fat-rich microenvironment for their growth through the proliferator-activated receptor (PPAR) pathway.65 PPARs are fatty acid-activated transcription factors that are members of the nuclear hormone receptor superfamily. In MBM, the cancer cells co-opt the neighbouring astrocytes to provide arachidonic acid and mead acid, which in turn activate PPARγ signalling in the cancer cells, driving their growth.65 Further studies showed that inhibition of PPAR signalling through specific antagonists retarded the growth of MBM in mice. Additional support for the role of fatty acid metabolism in brain metastasis progression has come from studies in breast cancer that showed brain-metastatic cells to express increased levels of the fatty acid transporter fatty acid-binding protein 7 (FABP7).66 It was found that expression of FABP7 was required for HER2+ breast cancer cells to maintain their glycolytic phenotype and maintain their storage of fat droplets while growing in the brain microenvironment.66

Another recent study has uncovered unique therapeutic vulnerabilities in brain metastases that stem from the depletion of amino acids in the CSF.57, 58, 67, 68 Amino acids are critical for the synthesis of many of a cells’ biomolecules including proteins, lipids and antioxidants. In amino acid-depleted environments, cells can synthesize non-essential amino acids including serine, glycine and aspartate through diversion of glycolytic intermediates into the serine and glycine synthesis pathways. This metabolic reprogramming is driven through increased expression of 3-phosphoglycerate dehydrogenase (PHGDH). It was found that in the amino-depleted environment of the brain, metastatic cancer cells were uniquely dependent upon PHGDH activity for their de novo amino acid synthesis.67 Targeting of PHGDH using specific small molecule inhibitors was found to inhibit the growth of brain metastases (including melanoma, breast and renal cell carcinoma) but not matched tumors at extracranial sites.67 It thus seems that there is a potential to develop brain metastasis-specific therapies that leverage these environment specific metabolic dependencies. Whether or not these strategies can achieve anti-tumor efficacy without impairing brain metabolic function remains to be seen.

2.4 The immune environment of MBM

Despite the belief that the CNS an immune-privileged site, both the brain and the spinal cord are under continual immune surveillance. The primary immune protection of the brain is mediated by the resident microglia which communicate with neurons, astrocytes and other glia.69, 70 Microglia serve as the first line of immune defense of the brain.38 In their inactive state, they express low levels of HLA-DR but can dramatically upregulate their expression of MHC II, CD80, CD86, CD40, CD11a, CD54 and CD58 in response to pathological changes, allowing them to present antigen to T cells.69-72 Microglia also respond to infection and respond in a macrophage-like fashion leading to the release of cytokines (TNF-α, IFN-γ) and the production of many chemokines that attract and recruit immune cells from the circulation.38, 72 In the healthy CNS, low numbers of immune cells from the general circulation also enter the brain and spinal cord, although at a lower rate than that of other peripheral organs. There is evidence that circulating T cells interact with the vascular endothelium of the CNS at 10% of the level than those of other organs.73 Other immune peripheral immune cells such as perivascular macrophages and meningeal dendritic cells also sample the BBB interface allowing foreign antigens to be detected. Despite this capacity to respond to infectious insults, the immune system of the brain is usually quiescent. This damping of baseline immunity is mediated in part by CD200 expressed on neuronal cells interacting with CD200 ligand on the microglia.74 Immune suppression may also be mediated indirectly through the electrical activity of neurons that has the potential to suppress MHC expression on microglia and astrocytes. Expression of neuron-derived neurotrophins such as BDNF, NT-3 and NGF also limits MHC I expression on microglia, impairing their ability to activate CD8+ T cells.75

In cancer, there is evidence that brain-resident immune cells may play a role in the infiltration of tumor cells into the brain. Studies in breast cancer have shown that cells in the primary tumor can induce the accumulation of CD11b+Gr1+ myeloid cells in the CNS, which then, in turn, express the inflammatory chemokines S100A8 and S100A9 which function as chemoattractants.76 These myeloid cells additionally express high levels of the chemokine CCL9, which have been shown to contribute to the growth of melanoma cells and breast cancer cells through a TGF-β dependent mechanism.77

One of the mainstays of therapy for advanced melanoma are immune checkpoint inhibitors that release the brakes on the immune system by targeting cell surface molecules such as CTLA-4 and PD-1 expressed on T cells and myeloid cells.78 One surprising recent finding was the observation that the combination of anti-PD-1 and CTLA-4 was highly effective in ~50% of patients with MBM and that levels of response were similar at cranial and extracranial metastatic sites.14 These intriguing results have led to histopathological studies to more fully define the microenvironment of MBMs. A recent report on 37 paired intracranial and extracranial melanoma metastases demonstrated that the brain metastases had a lower level of T-cell infiltration and microvessel density than those from extracranial sites.79 A second histological study also confirmed the lower level of CD8+ T-cell infiltrate in MBM compared to matched extracranial metastases.80 Although the mechanisms underlying the reduced levels of T cells in MBM are not clear, there is evidence from studies on neuroinflammation that have implicated a role for activated microglia in limiting T-cell infiltration into the brain.81

One intriguing anecdote from clinical studies of immunotherapy responses in MBM has been the observation that patients may respond better in the brain if they have both MBM and extracranial disease – suggesting a requirement for peripheral immune education. This idea has been supported by mouse MBM models which have demonstrated immunotherapy responses in the brain to be dependent upon a prior peripheral expansion of T cells that then trafficked into the brain.82 Studies on the immune environment of matched cranial and extracranial melanoma metastases have shown that samples from MBM have less T-cell receptor (TCR) diversity than the extracranial samples, again suggesting that fewer T-cell clones may make it to the brain.60 These observations are further supported by clinical studies on primary brain tumors (that are confined to the CNS), such as glioblastoma multiforme are so poorly infiltrated with immune cells and respond less well to immunotherapy.83 Melanoma cells have multiple mechanisms to escape immune recognition, and it is likely that these may underlie reduced durations of immunotherapy response in the CNS. Tumor intrinsic β-catenin activation in melanoma cells results in decreased dendritic cell accumulation, thus suppressing T-cell recruitment.84 Correlations have also been noted between depleted PTEN and increased VEGF-A levels in melanoma cells and reduced T-cell recruitment.85 Whether or not these mechanisms of evasions are also involved in immune evasion in MBM requires further exploration.

2.5 Melanoma signalling in the brain microenvironment

Genomic sequencing and signalling studies have shown the majority of cutaneous melanomas to be uniquely addicted to signals through the mitogen-activated protein kinase (MAPK) pathway.86, 87 This signalling, which can arise following the acquisition of activating mutations in BRAF (~50%), NRAS (~15–20%), loss of NF1 or RTK signalling, is critical for the growth, survival and invasive properties of melanoma cells (see88-90 for comprehensive reviews). Analyses of signalling behaviour of melanoma metastases from cranial and extracranial sites using reversed phase protein arrays (RPPA) identified the MBMs to have a unique dependency upon the PI3 K/AKT signalling pathway, with 60% of brain-derived samples having reduced PTEN expression and elevated PI3 K/AKT phosphorylation.91 A follow-up investigation focused upon matched brain and visceral metastases from the same patients confirmed the AKT activation findings, but did not show such a strong correlation with PTEN loss.92 Other studies have, however, shown that loss of PTEN expression (detected in this case by IHC staining) was predictive of eventual MBM development in a cohort of melanoma patients with stage IIIB/C melanoma.93 Further exploration of these clinical findings in transgenic mouse models of spontaneous MBM development confirmed the role of AKT in brain metastasis formation and demonstrated that introduction of myristolated AKT1 (activated AKT1) co-operated with PTEN loss to drive the establishment of CNS metastases.94 Additional work demonstrated that among the 3 isoforms of AKT, the E17 K AKT1 mutant was the most efficient at driving MBM formation in the mice and that this primarily worked through the activation of focal adhesion kinase (FAK) and modulation of integrin signalling.95 The likely therapeutic utility of targeting PI3 K signalling (both alone and in combination with the BRAF inhibitor encorafenib) has been demonstrated in MBM models where human melanoma cells are xenografted into mouse brains.92, 96

Although most studies to date have focused on PI3 K/AKT/mTOR signalling as being important in MBM development/progression, other pathways have been also implicated. Xenograft studies have shown that increased expression of STAT3, which is known to increase cyclin D1 and MMP1 and VEGF expression, enhances MBM formation in mice.97 There is also evidence that increased expression of cell surface receptors on melanoma cells including EDNRB and the p75NTR neurotrophin receptor as being associated with increased levels of MBM development.32, 33

3 TREATMENT OF MELANOMA BRAIN METASTASES

While anti-PD-1 and anti-CTLA4 immunotherapies and MAPK-targeted therapies have been widely used in the treatment of metastatic melanoma, data have only recently become available from single-arm studies looking at the efficacy of these regimens in participants with brain metastases. The COMBI-MB study explored the effectiveness of the BRAF/MEK inhibitors dabrafenib and trametinib in BRAF-V600 melanoma brain metastases.13 Of the 76 participants with asymptomatic BRAF-V600E-mutant melanoma to the brain in the primary cohort without prior local therapy such as radiation, 44 participants (58%; 95% CI 46–69) achieved an intracranial response. However, median progression-free survival (PFS) was only 5.6 months, with 72% of participants progressing with intracranial or intra +extracranial metastases.

The efficacy of anti-PD-1 monotherapy with pembrolizumab in patients with melanoma brain metastases was also investigated in a phase 2 trial.98 In the 23 patients with asymptomatic and untreated brain metastases who received pembrolizumab, six patients (26%) had an intracranial response. Brain metastasis and systemic responses were concordant. The median progression-free and overall survival were 2 and 17 months, respectively. In the Checkmate-204 trial of ipilimumab at 3 mg/kg and nivolumab 1 mg/kg q3 weeks x 4 doses, followed by nivolumab in 100 participants with melanoma brain metastases, a similar intracranial response rate of 54% was observed.14 This rate dropped to 22% in 18 participants with neurological symptoms with or without steroids (of up to 4 mg a day of dexamethasone).15 Median PFS in the asymptomatic cohort had not yet been reached, with 6-month PFS rate of 63%, dropping to 19% in symptomatic participants, with a median PFS of only 1.2 months. Furthermore, there was a grade 3/4 adverse event rate of 55%.

The same combination immunotherapy regimen was also tested in the Australian ABC trial of ipilimumab and nivolumab in participants with melanoma brain metastases.99 In the 35 asymptomatic participants who received the combination and with no prior local therapy, an intracranial response rate of 51% and 24-month intracranial PFS rate of 49% were observed.100 There was a similarly high 54% rate of grade 3/4 adverse events. In the 27 asymptomatic patients on the same study who received nivolumab monotherapy, intracranial response rate was 20%, with a 24-month intracranial PFS rate of 15%; in the 16 patients treated with nivolumab who had symptomatic brain metastases or failed prior local therapy, the intracranial response rate dropped to 6% with a 24-month follow-up. The data to date suggest that responses to targeted therapy are less durable than those seen to immunotherapy. There is also some suggestion that responses to immunotherapy are diminished when used after BRAF-MEK inhibitor failure. The mechanisms underlying these observations are still under exploration but may be related to the transcriptional reprogramming seen following targeted therapy escape, which is frequently associated with decreased MHC and melanoma antigen expression, as well as increased expression of immune checkpoints such as PD-L1.101

There are currently several ongoing trials in patients with melanoma brain metastases with the goal of improving clinical outcomes including in patients with symptomatic disease. An ongoing phase 2 trial is evaluating the combination of bevacizumab (VEGF blocking antibody) with pembrolizumab in patients with brain metastases from either melanoma or lung cancer (NCT02681549). Another trial of bevacizumab plus the anti-PD-L1 antibody atezolizumab in patients with melanoma brain metastases and includes separate cohorts for asymptomatic and symptomatic patients (NCT03175432).

Studies with BRAF/MEK targeted therapy in melanoma brain metastases are also ongoing. Higher doses of BRAF/MEK inhibitors with encorafenib + binimetinib are under investigation, with the study exploring a 600 mg daily dosing of encorafenib (standard dose is 450 mg) with binimetinib in brain metastases (NCT03911869). As suggested by the preclinical efficacy of PI3 K inhibitors in MBM, PI3 K inhibitors are also being investigated, with an ongoing study of buparlisib with encorafenib + binimetinib in BRAF mutant melanoma in a phase 2 trial (NCT02159066).

Two ongoing studies are investigating the combination of BRAF-targeted and immunotherapy (Table 1). The TRICOTEL phase 2 study is exploring the efficacy of the BRAF/MEK inhibitor combination vemurafenib plus cobimetinib with atezolizumab in BRAFV600 mutant melanoma with brain metastases (NCT03625141). The primary outcome of this study is intracranial response rate. The SWOG S2000 study is randomizing patients with melanoma brain metastases to the triplet combination regimen of encorafenib plus binimetinib with nivolumab vs ipilimumab + nivolumab, with a primary endpoint of progression-free survival (either intra or extracranial; NCT04511013). Both of these trials permit patients with symptomatic brain metastases, who may be on corticosteroids up to 8 mg of dexamethasone at time of study entry, to explore the efficacy of a triplet regimen in symptomatic patients. The S2000 trial will also allow for a direct comparison of a triplet regimen to ipilimumab + nivolumab in symptomatic and asymptomatic patients.

TABLE 1. Ongoing clinical trials in melanoma brain metastases Study Regimen Eligibility criteria Primary outcomes NCT02681549

Single-arm phase 2 trial: pembrolizumab with bevacizumab

(melanoma or NSCLC)

N = 53

Asymptomatic, prior radiation/surgery allowed

if untreated target lesion,

no prior immunotherapy,

prior BRAF/MEKi allowed

Intracranial response rate using RECIST 1.1 NCT03175432

Single-arm phase 2 trial:

atezolizumab with bevacizumab

N = 40

Cohort A: asymptomatic,

Cohort B: mildly symptomatic

Prior radiation/surgery allowed

if untreated target lesion, BRAF/MEKi allowed

Intracranial response rate using irRANO NCT03911869

Randomized phase 2 trial:

high dose encorafenib with binimetinib vs standard dose encorafenib with binimetinib

N = 110

Asymptomatic, BRAFV600 mutant,

Cohort 1: received prior local therapy

Cohort 2: no prior local therapy

Intracranial response rate using RECIST 1.1 NCT03625141

Single-arm phase 2 trial:

vemurafenib plus cobimetinib with atezolizumab

N = 120

Symptomatic permitted, BRAFV600 mutant.

Prior BRAF/MEKi not allowed

Intracranial response rate using RECIST 1.1 NCT04511013

Randomized phase 2 trial:

encorafenib + binimetinib + nivolumab vs

ipilimumab + nivolumab

N = 112

Symptomatic permitted, BRAFV600 mutant

Prior systemic