Converting biology into clinical benefit: lessons learned from BRAF inhibitors

Abstract

The identification and pharmacological targeting of activating BRAF mutations in melanoma has led to significant improvements in patient outcomes. This perspective paper illustrates the lessons learned from the study of BRAF mutations and the development of BRAF inhibitors. The relevance of these lessons to the development of future targeted therapies is highlighted.

Practice points.

• Mutation of the BRAF gene is the most common oncogenic event identified to date in cutaneous melanomas. The clinical experience with targeted therapies in metastatic melanoma patients with a BRAF ^V600 mutation provides an important proof-of-concept of the potential benefit of personalized therapeutic approaches in this disease and key insights into the factors that are critical to effective therapeutic evaluation and development.

• Different mutations in BRAF and other oncogenes can have different functional and clinical significance.

• The optimal development and evaluation of new targeted therapies should utilize appropriately selected patient populations.

• Pharmacodynamic evaluation of target inhibition is critical to the evaluation of new targeted therapy strategies and experimental agents.

• Targeted therapy resistance is pervasive and can be due to multiple mechanisms, many of which can be heterogeneous within individual patients and even individual tumors.

• Resistance can be mediated by genetic factors, epigenetic changes and the influence of the tumor microenvironment. Understanding and overcoming resistance will likely require integrated analysis and targeting of these disparate factors.

Broad sequencing studies have demonstrated that cutaneous melanomas harbor more mutations than virtually all other cancer types. The overwhelming majority of these mutations appear to be nonfunctional, ‘passenger’ mutations that are induced nonspecifically by ultraviolet radiation. However, most cutaneous melanomas are also characterized by one or more recurrent mutations in oncogenes in key cell signaling cascades. To date, the most prevalent and clinically significant events are point mutations that affect the V600 position of the BRAF protein kinase (BRAF^V600). This landmark discovery in 2001 triggered the era of personalized therapy in melanoma. Clinical testing for BRAF ^V600 mutations is now the standard of care for patients with metastatic melanoma as four different therapeutic regimens have been approved by the US FDA specifically for patients with mutations affecting this site. These treatments provide clinical benefit even in patients with rapidly growing and widely metastatic disease, providing a powerful proof of concept of the positive impact of melanoma molecular analysis. Over time, a number of insights have been made that have improved our understanding of the biology and treatment of melanomas with BRAF mutations. At a very basic level, it is now clear that different mutations in the BRAF gene are not functionally or clinically equivalent. The failure of early targeted therapy trials in melanoma has subsequently provided insights to facilitate the appropriate and efficient evaluation of future experimental agents. Finally, the study of the pervasive resistance to BRAF inhibitors has demonstrated the adaptability and heterogeneity of this disease, as well as the interconnections between melanomas and the tumor microenvironment (TME).

Many preclinical and clinical investigations are underway to identify new targeted therapy approaches for melanoma. The likelihood of the successful translation of these efforts into clinically beneficial treatments for patients will be increased by utilizing insights gained by the development of targeted therapy strategies for BRAF ^V600 mutant melanoma.

The discovery & biology of BRAF mutations

RAS was one of the first oncogenes identified in human malignancy, and RAS-induced hyperactivation of the RAS–RAF–MAPK signaling pathway is one of the most common genetic events in cancer [1]. Activating mutations in NRAS were identified in melanoma in the 1970s, providing initial evidence for a role of this pathway. However, the true significance of this pathway in melanoma was revealed by a systematic screen for mutations in the RAF protein kinases in cancer [2]. This initial study identified mutations in the gene encoding the BRAF serine-threonine kinase in over half of included melanoma cell lines and clinical specimens, as well at a lower prevalence in colon, lung and ovarian cancer specimens (3–18%). Larger follow-up studies have confirmed that BRAF mutations occur in approximately 45% of cutaneous melanoma [3]. Studies that included melanomas of different clinicopathological subtypes have demonstrated that BRAF mutations are much less frequent in acral and mucosal melanomas (18 and 6% respectively), and that they are essentially absent in uveal melanoma [3]. Interestingly, BRAF mutations have been detected in up to 80% of benign nevi, supporting a key role in early melanomagenesis [4].

The mutations that have been detected in BRAF in cancer overwhelmingly affect exons 15 (>95%) and 11 (˜5%). Almost 95% of the reported mutations result in substitutions of the valine residue at position 600 (V600) in the BRAF protein, which is within the activation section of the kinase domain. In melanoma, approximately 80% of mutations at this site are V600E, in which glutamic acid is substituted for valine. V600K mutations comprise approximately 20% and V600D, V600R and V600M are all rare but well-described mutations. There are also rare mutations in the adjacent residues [3]. The V600E mutation results in an amino acid substitution at position 600 in BRAF, from a valine to a glutamic acid. This mutation occurs within the activation segment of the kinase domain. BRAF ^V600 substitutions markedly increase the catalytic activity of the BRAF protein, leading to constitutive activation and phosphorylation of MEK and ERK in the RAS–RAF–MAPK signaling cascade [2]. The V600E mutation increases the kinase activity of BRAF more than 400-fold in vitro [5]. In contrast, the catalytic activity of BRAF proteins with non-V600 mutations is extremely variable. Some of these mutations increase BRAF activity similarly to V600 substitutions, some only weakly activate the protein, and others actually decrease catalytic activity compared with the wild-type protein. However, even mutations that do not increase BRAF catalytic activity appear to activate the downstream pathway by changing the structure of BRAF to promote dimerization with other RAF isoforms (i.e., CRAF) [5].

Interestingly, BRAF ^V600 and non-V600 mutations also demonstrate different molecular associations. For example, in treatment-naive melanomas, BRAF ^V600 and NRAS mutations are essentially mutually exclusive [6]. However, BRAF non-V600 mutations are often detected in melanomas with activating NRAS mutations [7]. Further, as will be described later, there appear to be several significant clinical differences in melanoma patients with V600E and V600K mutations. Thus, different mutations in the BRAF gene may have very different consequences and clinical impact (Table 1). This supports that other candidate oncogenes in melanoma with mutations at multiple sites may require detailed functional, molecular and clinical characterization to fully understand their significance.

Table 1. . Lessons learned from BRAF biology.

Finding	Implication
Functional differences between mutations at different amino acids in BRAF	Need to functionally characterize specific mutations in candidate oncogenes
Variability in outcomes among patients with different BRAFV600 substitutions	Need to distinguish between prognostic and predictive markers
Metastatic site-specific molecular features	Potential need for site-specific characterization and therapeutic approaches

Clinical targeting of BRAF in melanoma

Sorafenib (Nexavar^®, Bay43-9006) is a multikinase inhibitor that is FDA approved for the treatment of metastatic renal cell carcinoma, hepatocellular carcinoma and differentiated thyroid carcinoma. Originally developed as a CRAF (RAF-1) inhibitor, sorafenib was also found to suppress both wild-type BRAF and BRAF^V600E proteins in vitro and additionally inhibited VEGFR2/3, PDGFRβ, c-KIT and FLT [8]. This promiscuity was initially felt to be a strength given inhibition of the RAS-RAF pathway at multiple levels. In xenograft BRAF ^V600E melanoma models, sorafenib slowed tumor growth [9]. Despite these results, 0 of 34 metastatic melanoma patients responded to sorafenib treatment in a Phase II clinical trial [10]. While this study was not restricted to patients with a specific genotype, subsequent analyses demonstrated that patients with and without BRAF ^V600 mutations had similar clinical outcomes. Sorafenib was also investigated in melanoma patients in combination with chemotherapy with similarly disappointing results in a Phase III trial of over 800 patients after a promising Phase I/II study [11–14]. While sorafenib did not benefit patients with BRAF ^V600 mutations, there was a trend toward improved clinical response from the addition of sorafenib to chemotherapy in patients with NRAS mutations [15]. This may be due sorafenib’s more potent CRAF inhibition, which is the RAF isoform utilized preferentially in cells with a NRAS mutation [16].

The negative sorafenib trials led to questioning of the clinical significance and therapeutic value of BRAF ^V600 inhibition in melanoma. However, the ability to answer this question was critically limited by the lack of sample collection to determine if sorafenib actually inhibited BRAF^V600 at clinically tolerated doses. A subsequent Phase II study of 36 metastatic melanoma patients demonstrated that the expression of cyclin D1 (CCND1), which is regulated by the MAPK pathway, was not significantly affected by sorafenib treatment [17]. An additional analysis of a Phase I trial of sorafenib and the mTOR inhibitor temsirolimus showed that the combination achieved minimal inhibition of downstream ERK activation. This is critical as inhibition of phosphorylated ERK proteins (P-ERK), has been subsequently validated to be a robust biomarker of MAPK pathway inhibition. This lack of P-ERK inhibition was in contrast to the P-ERK inhibition seen in preclinical studies [18]. The issue lay in selectivity and potency – the IC₅₀ required for MAPK pathway inhibition appears to be too toxic to be safely achieved in patients [19]. While sorafenib is a multikinase inhibitor, its clinically beneficial effects may be predominantly driven by VEGF inhibition. The diseases in which sorafenib provides clinical benefit are classical VEGF-driven cancers (hepatocellular carcinoma and renal cell carcinoma). In melanoma, an association between VEGFR2 expression and clinical response was observed in one trial [20]. Side effects driven by the inhibition of non-BRAF kinases to which sorafenib binds with higher affinity likely limited dose-escalation to a drug concentration sufficient for BRAF^V600 inhibition. The failure of sorafenib highlights the importance of the identification and validation of biomarkers of target inhibition, and the incorporation of pharmacodynamics in early clinical trials. This information is particularly critical in ‘negative’ trials to distinguish if they represent the failure of a target/strategy or simply a pharmacologic failure that has not appropriately tested the clinical impact of oncogene inhibition (Table 2).

Table 2. . Lessons learned from BRAF inhibitor development.

Finding	Implication
Correlation of pharmacodynamic effects (P-ERK inhibition) with clinical benefit	Need for validated biomarkers and incorporation of pharmacodynamics in clinical evaluation of new agents
Limited bioavailability, and lack of clinical benefit, for initial formulation of vemurafenib	Need for pharmacokinetic analysis in clinical evaluation of new agents
Identification of concurrent NRAS mutations in BRAF-mutant tumors progressing on BRAF inhibitors	Targeted therapies can result in molecular features that are extremely different than treatment-naive tumors
Resistance due to RTK activation in the absence of gene mutation or amplification	Need for multiplatform analysis, not simply DNA sequencing
Resistance induced by growth factors produced by tumor adjacent or infiltrating nontransformed cells	Need for interrogation of the TME
Heterogeneity of resistance mechanisms between different tumors in individual patients	Need for noninvasive approaches to characterize resistance mechanisms in toto
Variability of response among different cancer types with the same mutation	Need for testing of novel agents in individual cancers
Improvement in outcomes with combination therapy upfront	Need for testing of rational combinations in first-line setting
Variability of toxicities among therapies in same class	Need for multiple agents within same class to tailor to patients

RTK: Receptor tyrosine kinase; TME: Tumor microenvironment.

After these initial failures, the efficacy of second-generation, mutation-selective BRAF^V600 inhibitors has decisively established the importance of targeting BRAF^V600 in melanoma. To date, two agents (vemurafenib and dabrafenib) have gained regulatory approval, while other agents have shown promising results in early phase studies. These agents exhibit high selectivity for BRAF over other kinases and, more critically, also have greater affinity for BRAF^V600 than wild-type BRAF [21]. This is in contrast to sorafenib which binds almost equally well to wild-type BRAF and BRAF^V600. Consistent with this in silico specificity, in preclinical studies the mutant-selective BRAF inhibitors markedly reduce the growth of human melanoma cell lines with IC₅₀s several log concentrations lower than the doses required for lines with wild-type BRAF [21]. While sorafenib led predominantly to slower growth in vivo, vemurafenib treatment produces tumor regressions [22]. This distinction is important as it translates clinically into the difference between slowing progressive disease and clinical response.

Despite promising preclinical results, in the initial portion of the Phase I trial of vemurafenib in BRAF ^V600 metastatic melanoma patients, there were no clinical responses [23]. However, pharmacokinetic studies demonstrated poor bioavailability, with serum levels failing to achieve the concentrations required for tumor regression in preclinical models despite dose escalation. The trial was halted and the drug was reformulated. After this change, linear, dose-dependent increases in serum drug levels were observed – as were clinical responses. In the expansion cohort, the unconfirmed response rate was 81%. By contrast, none of the five patients enrolled in the Phase I trial with melanomas without a BRAF ^V600 mutation responded; in fact, four of the five patients had significant tumor progression observed on initial restaging.

A number of patients that participated in the Phase I study of vemurafenib gave permission for tumor biopsies to be collected prior to treatment, and after 1–2 weeks of therapy. Pharmacodynamic analysis showed that the levels of P-ERK inhibition achieved varied significantly and that the degree of P-ERK inhibition was correlated with response, with a nearly linear relationship demonstrated between P-ERK inhibition and tumor shrinkage [24]. The correlation of P-ERK inhibition with clinical benefit established this test as a powerful biomarker for the clinical evaluation of other BRAF inhibitors. The development of vemurafenib highlights the shared importance of target identification, drug selection, and pharmacokinetics and pharmacodynamics in the successful development of targeted therapies (Table 2). Ultimately, this trial also provided an emphatic proof-of-concept that inhibition of activated oncogenes can provide significant clinical benefit in metastatic melanoma, and specific proof that BRAF^V600 is an important therapeutic target in this disease.

Dabrafenib was the second BRAF^V600-selective inhibitor to enter into clinical trials. Phase I and subsequent clinical trials demonstrated that overall it achieved very similar overall response rates (ORR) and progression-free survival (PFS) as vemurafenib and dabrafenib was approved for the treatment of BRAF ^V600 melanoma in May 2013 [25–27]. However, the clinical development of dabrafenib differed in several ways from that of vemurafenib. First, while the clinical testing of vemurafenib almost exclusively included patients with V600E mutations, patients with other V600 substitutions, particularly V600K, were included in early phase clinical trials with dabrafenib. While these studies demonstrated that non-V600E mutations do respond to dabrafenib, the response rates and PFS were significantly lower in patients with V600K mutations versus V600E mutations [25]. This was unexpected given that dabrafenib binds and inhibits all prevalent BRAF V600 substitutions equally in vitro. Retrospective analyses have subsequently demonstrated that melanoma patients with a V600K mutation are characterized by a more aggressive disease phenotype than patients with a BRAF ^V600E mutation. Patients with V600K mutations have a significantly shorter time to development of metastatic disease and shorter overall survival (OS) from time of stage IV diagnosis even among patients not treated with MAPK pathway inhibitors [28,29]. V600K mutations are also associated with distinct clinical phenotypes, including older age at diagnosis and occurrence in the head and neck region, in other words, areas of chronic sun damage [28,29]. Thus, the presence of a V600K mutation is more likely a negative prognostic factor, or a marker of distinct disease biology, rather than a negative predictive factor for BRAF inhibitors (Table 1).

Another difference between the vemurafenib and dabrafenib trials was the inclusion of patients with asymptomatic, untreated brain metastases in the dabrafenib trials. In the dabrafenib Phase I dose expansion cohort, ten patients with brain metastases were included, nine of whom had shrinkage of their brain metastases and four of whom had complete resolution of all brain lesions [25]. The subsequent Phase II BREAK-MB study was the largest prospective study to date for metastatic melanoma patients with active brain metastases, enrolling 172 patients with BRAF ^V600E or BRAF ^V600K melanoma and either untreated or progressive brain metastases [30]. Despite preclinical studies predicting lack of penetration of an intact blood–brain barrier, ORR, PFS and OS clinical trial outcomes support that dabrafenib had significant benefit for these patients. While this is likely due to disruption of the blood–brain barrier by brain metastases, current clinical trials are also evaluating potential CNS penetration and role of several metabolites of dabrafenib. Subsequent studies and case reports have demonstrated that vemurafenib also achieves significant clinical benefit in this population [31–33]. These results support the inclusion of patients with brain metastases in the evaluation of safety and efficacy of new agents, particularly as effective treatments for brain metastases are desperately needed.

Molecular effects & drug toxicities

One of the most common and unexpected toxicities of the mutant-selective BRAF inhibitors was the development of cutaneous squamous cell carcinomas (SCCs). SCCs and keratoacanthomas were detected in approximately 20–30% of patients in the trials of vemurafenib, and approximately 10% of patients treated with dabrafenib [25,34]. These lesions require surgical treatment, but fortunately to date have not been reported to metastasize. BRAF inhibitor treatment can be resumed safely after they have been removed.

The likely cause of these lesions is an unexpected paradoxical effect of the mutant-selective BRAF inhibitors on the MAPK pathway. Treatment of cancer cell lines without a BRAF ^V600E mutation with selective BRAF inhibitors unexpectedly led to accelerated tumor growth in vitro and in vivo, particularly in models with RAS mutations [35–37]. In BRAF wild-type cells, BRAF inhibitor binding induces a conformational change in the BRAF protein that promotes heterodimer formation with other RAF isoforms, and the heterodimers then activate MAPK pathway signaling. This paradoxical effect of the BRAF inhibitors potentially contributed to the rapid progression observed in the majority of patients with BRAF wild-type melanoma enrolled in the Phase I trials of the selective BRAF inhibitors [23,25].

SCCs and keratoacanthomas from patients treated with BRAF inhibitors have a high frequency of activating RAS mutations, and in preclinical models the growth of chemically induced premalignant skin lesion were accelerated by treatment with BRAF inhibitors [38]. This SCC-promoting effect was blocked by concurrent treatment with MEK inhibitors, providing functional evidence that the lesions were caused by paradoxical MAPK pathway activation [38]. Similarly, when BRAF inhibitors and MEK inhibitors are given concurrently to patients, the rate of cutaneous SCCs is also dramatically less, dropping from 15–20 to 1–5% [39,40].

Notably, while cutaneous toxicity is decreased with combined BRAF and MEK inhibition, other side effects are increased, most notably pyrexia (˜70% incidence with dabrafenib + trametinib combination therapy). The incidence of fever is markedly increased with dabrafenib-based combinations compared with vemurafenib-based, while other toxicities are more frequent with vemurafenib [41,42]. The differences in side effect profiles clinically is used by physicians to select or adapt treatments based on patients’ tolerance, and illustrates why it can be useful to have more than one agent available for important targets (Table 2).

Resistance to BRAF inhibitors

By the time of publication of the Phase I/II trial of vemurafenib in August of 2010 [23], the Phase III BRIM-3 trial was 4 months from completion, enrolling 657 patients from January to December 2010 in the fastest Phase III trial ever completed in oncology [34]. Patients with BRAF ^V600-mutant melanoma were randomized to vemurafenib or dacarbazine. PFS was improved from 1.6 months in the dacarbazine group to 5.3 months in the vemurafenib group, and OS at 6 months improved from 64 to 84% (hazard ratio [HR]: 0.37; p < 0.001). Vemurafenib was FDA approved for the treatment of metastatic BRAF^V600E melanoma in August 2011, just 3 years after the first published preclinical studies of the agent. Dabrafenib was subsequently approved for the treatment of BRAF ^V600 melanoma in May 2013 based on improvement in PFS compared with dacarbazine of 5.1 versus 2.7 months (HR: 0.33; p < 0.001) in a pivotal Phase III trial [27].

While the approval of these agents was a key advance in the treatment of BRAF ^V600 melanoma, the clinical benefit that can be achieved with BRAF inhibitor monotherapy is critically limited by resistance. Although almost all patients with BRAF^V600 mutations have some degree of response to BRAF inhibitors, responses are highly variable and are usually short lived. Approximately 10% of patients have progressive disease (PD) as their best response, reflecting the presence of rare de novo/primary resistance. Among the patients that do respond, there are very few complete responses (CRs; ˜5%), and there is significant variation in degree of tumor shrinkage, reflecting an underlying heterogeneity. Furthermore, the average response to BRAF inhibitors lasts 5–7 months, at which point the tumors begin growing again. This change in sensitivity is referred to as acquired, secondary or induced resistance.

Multiple mechanisms of resistance to BRAF inhibitors have now been identified, with distinct therapeutic implications. Resistance mechanisms can be broadly characterized as those that cause reactivation of MAPK pathway signaling or the activation of parallel prosurvival pathways [43,44]. Interestingly, loss of the activating BRAF ^V600 mutation has not been described as a mechanism of acquired resistance. There have furthermore been no secondary BRAF mutations identified, which is in contrast to resistance to other targeted therapies such as imatinib in chronic myelogenous leukemia (CML) and gastrointestinal stromal tumors (GIST) [45,46]. BRAF gene amplification has been observed as a fairly common cause of acquired resistance, detected in approximately 20% of clinical samples at the time of progression [47,48]. Importantly, in cell lines, simply increasing the dose of BRAF inhibitor overcomes this resistance mechanism [47]. This is in contrast to another BRAF modification, alternative splicing, which produces a truncated protein that is able to dimerize and activate the MAPK pathway even in the presence of high doses of selective BRAF inhibitors [49]. Beyond alterations in BRAF, mutations in other components of the MAPK pathway can result in pathway reactivation. The most frequent events are NRAS mutations, which are observed in approximately 20% of tumor samples at time of progression on BRAF inhibitors [48,50]. This was a surprising discovery as BRAF ^V600 and NRAS mutations are essentially mutually exclusive in BRAF inhibitor-naive melanomas. Thus, the selective pressure of effective targeted therapies can actually generate a molecular landscape that is distinct from treatment-naive disease. In addition to NRAS, multiple mutations in MEK1 and MEK2 have been reported in melanomas after they have developed resistance to BRAF inhibitors [48,51]. Some MEK1 mutations have been reported in pretreatment samples from patients who achieved good clinical responses to BRAF inhibitors, while others have only been observed in resistant lesions [52,53]. Subsequent molecular characterization supports that, similar to BRAF, different mutations in the MEK genes can have different effects on the protein’s activity and on sensitivity to BRAF inhibitors.

Activation of the parallel PI3K/AKT pathway is another common mechanism of resistance. Oncogenic activation of the PI3K/AKT pathway is one of the most frequent events in cancer. One of the most potent mechanisms of PI3K activation in melanoma is loss of the negative pathway regulator, PTEN. PTEN loss occurs both in melanomas with a concurrent BRAF ^V600 mutation and tumors with wild-type BRAF and NRAS genes at a prevalence of 20–30%, but it is very rare in tumors with NRAS mutations [54]. The complementation between BRAF ^V600 and PTEN loss is supported functionally by characterization of a genetically engineered mouse model of BRAF ^V600 melanoma. Expression of the BRAF ^V600 mutation alone results in hyperproliferative melanocytic lesions but not invasive melanomas. However, concurrent BRAF ^V600 mutation and PTEN loss invariably results in invasive melanomas that spontaneously metastasize [55]. In vitro, loss of PTEN in BRAF^V600 cell lines causes only a mild decrease in the growth inhibitory effect of BRAF and MEK inhibitors, but it markedly decreases the degree of apoptosis induced by these agents. In early phase trials of dabrafenib, decreased PTEN gene copy number was associated with a trend for reduced PFS compared with patients with an intact PTEN gene (18 vs 32 weeks; p = 0.059) [56]. Patients in the Phase II trial of vemurafenib who failed to achieve clinical responses also had lower PTEN expression by immunohistochemistry (IHC) than responders. Loss of PTEN has also been reported as a mechanism of acquired resistance, albeit a rare one [48,56–57].

The PI3K/AKT pathway can also be activated by receptor tyrosine kinases (RTKs) on the surface of melanoma cells. Overexpression and activation of various RTKs such as IGF1R, PDGFR, ERBB3 (HER3) and EGFR (HER1) has been demonstrated in both cell lines and patients with acquired resistance to BRAF inhibitors [50,58–63]. Increased IGF1R expression has also been found to mediate de novo resistance in vitro [59,64]. Notably, in these studies the RTKs were not mutated or amplified, and the mechanism underlying their upregulation and activation remains unclear. In addition to tumor intrinsic changes, activation of RTKs and the PI3K/AKT pathway can also be induced by the tumor microenvironment. Two different studies independently demonstrated that the production of HGF, the ligand for the c-MET receptor, by nontransformed cells in the tumor microenvironment can induce resistance to anticancer therapies in cancer cells, including melanoma. Critically, this was not only demonstrated in vitro in melanoma cell lines treated with BRAF inhibitors, but analysis of clinical specimens also demonstrated that melanoma patients with evidence of HGF expression in the TME had inferior clinical responses with BRAF inhibitor therapy [65,66]. Notably, HGF did not rescue MAPK pathway activation in the setting of BRAF inhibition, but instead caused activation of PI3K/AKT signaling. This paracrine resistance mechanism could be overcome by blocking HGF, c-MET or PI3K/AKT pathway signaling. In addition to providing additional support to the rationale for the targeting of the PI3K/AKT pathway, these findings point to the importance of studying the potential role of the TME in resistance to new therapies (Table 2).

Whole exome sequencing studies have identified distinct molecular alterations in one or both of these two core cell signaling pathways, MAPK and PI3K, in approximately 75% of clinical samples of tumors progressing on BRAF inhibitors [48]. However, in approximately 25% of patients the mechanism of resistance could not be identified by whole exome sequencing which is limited to detecting changes in the genome. As per the discovery of HGF, it is possible that these tumors may be resistant due to extrinsic factors in the TME. Alternatively, as is observed with RTKs, molecular changes that cause resistance may be induced by epigenetic changes, which are not detected by DNA sequencing. Other cellular processes and pathways not traditionally studied in the analysis of oncogenes, such as tumor metabolism, have also recently been implicated as compensatory survival signals and mediators of resistance [67,68]. Importantly, functional testing in preclinical models suggests that modifiers of these processes, such as histone deacetylase inhibitors and metabolic agents, can overcome resistance to BRAF inhibitors [67,69]. Together, these findings strongly support the need for integrated, multiplatform analysis of clinical samples as opposed to DNA sequencing alone, in the development of future agents and therapeutic strategies.

A further challenge to understanding and overcoming BRAF inhibitor resistance is the molecular heterogeneity of melanoma. Tissue-based studies have demonstrated both spatial and temporal heterogeneity of distinct resistance-associated molecular lesions [48,70]. Pressure from BRAF inhibition drives clonal evolution and even subclonal evolution with distinct resistant mechanisms within single tumors [48,71]. There is also evidence of molecular lesions clustering with specific sites of metastases. Specifically, brain metastases, compared with extracranial metastases within the same patient, are much more likely to demonstrate activation of the PI3K/AKT pathway, a distinction that may help elucidate the pathogenesis of brain metastases as well as offer an avenue for therapeutic interventions [72,73]. This heterogeneity presents a major challenge to targeting and overcoming resistance mechanisms, given that a single biopsy may not reflect the molecular landscape of the patient’s cancer. Empiric combination regimens that target the major branch points of the most prevalent resistance mechanisms is one rational approach, especially given the ‘long tail’ of low frequency resistance mechanisms [74]. Alternatively, approaches that would allow for real-time detection of a multitude of resistance mechanisms are under investigation, including molecular imaging and the analysis of circulating tumor cells or DNA [75,76].

Overcoming resistance to BRAF inhibitors

Given that reactivation of the MAPK pathway is the most common mechanism of resistance to BRAF inhibitors, dual inhibition of the MAPK pathway was a natural strategy to prioritize for clinical testing. The MEK inhibitor trametinib was approved by the FDA as a single agent in 2013 based on superiority compared with chemotherapy [40]; however, the greater value of MEK inhibition in melanoma is in combination with BRAF inhibition. The combination of dabrafenib and trametinib was approved in January 2014 for BRAF inhibitor-naive metastatic melanoma patients with BRAF ^V600 mutations based on a randomized Phase II study which showed an improvement in ORR from 54% with single-agent dabrafenib to 76% with the combination, and an improvement PFS from 5.6 to 10.5 months. A subsequent Phase III trial confirmed the superiority of this combination approach with an improvement in 1-year OS from 65 to 72% (HR: 0.69) compared with vemurafenib alone [39]. A Phase III study of vemurafenib and the MEK inhibitor cobimetinib published concurrently reported similar improvements in outcomes with an ORR of 68% and PFS of 9.9 months [42].

While these studies have demonstrated that combined BRAF and MEK inhibition is very effective in BRAF inhibitor-naive patients, disappointingly these combinations have much less activity in patients who have progressed on single-agent BRAF inhibitor therapy with ORR approximately 10–15% [41,77]. While dual MAPK pathway inhibition delays the emergence of resistance by blocking MAPK pathway reactivation, salvage combination may be less effective at reversing established MAPK pathway reactivation. Alternatively, the presence of parallel prosurvival mechanisms such as PI3K activation may be a negative predictor of response to salvage combination. Ongoing work investigating mechanisms of resistance and predictors of clinical response to this approach should help elucidate rational clinical strategies. However, this finding does raise the question of whether it is most appropriate to test new strategies in BRAF-mutant patients after they have progressed on existing MAPK pathway inhibitors, or if it is better to do combinatorial testing in BRAF inhibitor-naive patients. Presumably similar questions may arise as effective therapies are identified for other oncogenic targets in this disease.

To date, understanding of predictors of response and mechanisms of resistance to combination BRAF and MEK inhibitor has been limited due to lack of pharmacodynamic analysis. One study that examined five patients with acquired resistance to combined dabrafenib and trametinib unexpectedly showed that one patient had BRAF amplification and one had a BRAF splice variant, both of which would have been predicted to be overcome by BRAF and MEK inhibitor combination therapy [78].

In those tumors in which resistance is driven by activation of the PI3K/AKT pathway, combined MAPK and PI3K pathway inhibition is a rational approach. In cell lines, this combination induces synergistic cell killing and overcomes resistance [64]. The PI3K pathway can be targeted at multiple levels, and PI3K, AKT and mTOR inhibitors have all been explored. However, thus far, clinical development of these agents in BRAF ^V600-mutant melanoma has been challenging. Whereas BRAF inhibition selectively targets a mutated oncogene, PI3K pathway inhibition may be difficult to attain at clinically tolerated doses given that this pathway is critically important in nonmalignant cells as well. One strategy under exploration is isoform-specific PI3K inhibitors [79]. Development of a pharmacodynamic biomarker that correlates with pathway inhibition (such as P-ERK in MAPK pathway inhibition) is critical. In addition, there are multiple feedback loops within the PI3K/AKT pathway that allow for escape via compensatory signaling [79,80]. For example, mTORC1 inhibitors have shown little activity in melanoma, and this may be driven by compensatory hyperactivation of AKT due to an mTORC1-mediated negative feedback loop [81,82]. Dual mTORC1/2 inhibitors may be more effective as mTORC2 also blocks AKT activation and thus may abrogate this negative feedback [59]. Interestingly, dual mTORC1/2 inhibitors have recently been shown to have activity against the subset of BRAF melanoma with high oxidative phosphorylation-driven resistance [68]. Thus, while MAPK and PI3K pathway inhibition represents a rational combinatorial approach, there remain a number of key challenges including patient selection, pharmacodynamic biomarker development to assess for pathway inhibition within a narrow therapeutic index and combination approaches to overcome feedback loops (Table 2) [79].

Another possible strategy for combating resistance supported by preclinical studies is via intermittent drug dosing. Continuous BRAF inhibition rapidly led to clonal selection of drug-resistant tumor cells which remained dependent on MAPK pathway signaling via increased BRAF expression in an in vivo model [83]. In this model, resistant tumors grew more rapidly in the presence of BRAF inhibitor, whereas drug withdrawal actually led to tumor regression. This is probably due to markedly (toxically) elevated MAPK pathway activation from unencumbered signaling by increased BRAF proteins causing growth inhibition. In essence, tumors adjust BRAF expression epigenetically to optimize the degree of MAPK signaling, and perturbations in this established homeostasis disrupt the oncogenic network. In these xenograft models an intermittent dosing schedule forestalls the emergence of resistance and prolongs survival [83]. This strategy is being investigated in ongoing and planned clinical trials.

Improving response to BRAF inhibition: beyond targeted therapy

The other class of therapeutic agents that has revolutionized the treatment of melanoma is immunotherapy. Three new antibodies which ‘derepress’ the immune system have gained FDA approval in recent years, the anti-CTLA4 antibody ipilimumab (2011) and the anti-PD1 antibodies pembrolizumab (2014) and nivolumab (2014). While response rates are lower than that achieved by dual BRAF and MEK inhibition (˜10% for ipilimumab and ˜40% for pembrolizumab and nivolumab), the durability of responses to immunotherapies is often very prolonged [84–86]. It is possible that combining immunotherapies with BRAF inhibition could lead to higher rates of durable clinical responses. In preclinical models, there is in fact synergy between immunotherapy and BRAF inhibitors [87,88]. The mechanism underlying this synergy seems to be multifactorial. BRAF inhibition causes increased expression of antigens on tumor cell surfaces. This improves immune recognition of the tumor cells, resulting in an increase in tumor infiltrating lymphocytes [87,88]. Other data support that BRAF inhibitors can decrease the expression of inhibitory cytokines and other tumor-mediated effects on the tumor microenvironment that dampen the antitumor immune response [89–91]. Together, these findings support the importance of evaluating the immune effects of future experimental targeted therapies.

Unfortunately, the first clinical trial of combined targeted and immune therapies, which tested concurrent vemurafenib and ipilimumab, was stopped early due to an unexpectedly high rate of hepatotoxicity [92]. Interestingly, and consistent again with differences in the toxicities of the different BRAF inhibitors, the combination of ipilimumab and dabrafenib was tolerated well without any increased incidence of hepatotoxicity over that expected with ipilimumab alone [93]. Multiple trials combining targeted and immune therapies are now ongoing. In addition to testing the safety and efficacy of combining different agents together, upcoming trials will need to address the optimal dosing and sequencing of these agents.

Conclusion

The identification and successful targeting of the BRAF oncogene has dramatically improved the outcomes of patients with BRAF ^V600 metastatic melanoma. This story represents a compelling example of the power and potential of personalized, targeted cancer therapy. Lessons learned through the development of BRAF inhibitors should be used to facilitate the development of future targeted therapies (Table 2). The failure of early attempts at BRAF inhibition highlights the importance of validated biomarkers of target inhibition, an issue relevant to the development of PI3K inhibitors. We have learned that different mutations in BRAF or in genes secondarily mutated in BRAF inhibitor-resistant melanoma such as MEK are not equivalent and that functional characterization is necessary. Finally, we have gained important understandings of the pervasive and heterogeneous nature of targeted therapy resistance, and the necessity to incorporate integrated, multi-platform analysis of clinical samples in order to understand and overcome it.

A number of challenges remain to optimize the treatment of BRAF ^V600 melanoma. The heterogeneity of this disease, manifested by both variable responses and the wide diversity of resistance mechanisms, is a major hurdle. A major limitation to overcoming these challenges is the burden of acquisition of tissues for analyses. The development of noninvasive blood-based molecular biomarkers is thus a high priority across the spectrum of personalized targeted therapy, from molecular subtyping to guide therapy, monitoring response and predicting relapse, to eventually identifying and forestalling emerging molecular changes that mediate resistance.

There are also lessons to be learned from the development of targeted therapies in other malignancies. Particularly relevant is the unexpected failure of BRAF inhibitors in BRAF ^V600-mutant colon cancer, where the response rates with BRAF inhibitors are on the order of 5% [94]. A major mechanism of resistance in BRAF ^V600-mutant colon cancer is feedback activation of the EGFR, which can be overcome in preclinical models with combination of BRAF and EGFR inhibition [95,96]. Histology-dependent molecular milieus and associated variance in clinical responsiveness should serve as cautionary tales in the design and interpretation of histology-independent, mutation-dependent targeted therapy basket trials.

While the clinical testing of BRAF inhibitors in melanoma has focused on patients with distant metastases, these agents may also achieve clinical benefit in earlier stages of this disease. The efficacy of HER2/Neu inhibitors in breast cancer and c-KIT inhibitors in gastrointestinal stromal tumors were both initially demonstrated in patients with distant metastatic disease. However, these agents are now also FDA approved in the neoadjuvant/adjuvant setting for patients with clinically localized or regional disease. Notably, while HER2/Neu inhibitors delay but do not prevent treatment resistance in stage IV disease, they increase the rate of durable disease control in earlier stage patients. Thus, while BRAF inhibitors do not appear to achieve cures in patients with distant metastases, they may be able to do this in the adjuvant setting. This approach is currently being tested in stage III BRAF ^V600 melanoma, a space in which the need for new therapeutic approaches is tremendous. The extremely high response rate of combined inhibition of BRAF and MEK supports the rationale for neoadjuvant approaches, which may have the added benefit of reducing the extent and morbidity of definitive surgical treatment.

The treatment and prognosis of metastatic melanoma has changed dramatically over the last 5 years through the development and approval of effective new immunotherapies and targeted therapies. Despite these advances, clinical responses remain variable and cures rare. In order to further improve outcomes in this aggressive disease, it is important to thoughtfully incorporate lessons from the successes, as well as early failures, of the BRAF inhibitors.

Future perspective

The treatment of patients with metastatic melanoma has changed dramatically over the last 5 years. Over this time, four targeted therapies and three immunotherapies have gained regulatory approval for patients with stage IV melanoma. Over the next decade there will be many efforts to build upon this tremendous progress. There is a strong rationale to determine if combining targeted therapies and immune therapies will be more beneficial than treatment with either modality alone. In addition, clinical trials have already been initiated to determine the safety and efficacy of the new agents in patients with high-risk regional disease, and approaches will likely also be explored in even earlier-stage patients. Ultimately, each of these clinical efforts will be optimized by an improved understanding of clinical and molecular markers that correlate with the aggressiveness of disease, and the responsiveness and resistance to targeted and immune therapies. Further, based on the experience with the BRAF inhibitors, it will be critical to integrate the planned analysis of progressing tumors in clinical investigations to rapidly identify and prioritize new strategies that are likely to be more effective. The ultimate goal of these efforts will be the identification of personalized approaches for patient management that span not only the full continuum of melanoma, but also the many therapies that are available to combat it.