Activity and Resistance of a Brain-Permeable Paradox Breaker BRAF Inhibitor in Melanoma Brain Metastasis

Malignant melanoma is the leading cause of mortality from skin neoplasms with over 57,000 fatal cases in 2020 worldwide (1). Over recent decades, the incidence of melanoma continues to rise (2), further supporting the urgent need to develop more effective treatments. Importantly, a high incidence of brain metastases is typically observed in melanoma patients. Between 7% and 20% of patients present with radiologically detectable brain metastatic lesions at diagnosis, and up to 70% of patients at later stages of disease (3–5). For patients with metastatic melanoma, approved therapeutic options include immune-checkpoint inhibitors (ICI) and combinations of BRAF and MEK inhibitors (BRAFi/MEKi) in BRAF V600E/K mutation–positive tumors. Although these therapies have greatly improved survival rates, not all patients respond to ICI. For targeted therapies, the acquisition of resistance is the major challenge in achieving durable responses (6–8).

Mechanisms of resistance to the three approved BRAFi (vemurafenib, dabrafenib, and encorafenib) in combination with MEKi (cobimetinib, trametinib, and binimetinib) largely occur through the restoration of MAPK signaling, which is predominantly mediated by RAF dimerization (9, 10). Multiple genetic events can enable RAF dimerization including acquired NRAS mutations, receptor tyrosine kinase (RTK) signaling activation, BRAF amplification, and expression of the BRAF splice variant p61 (11–13).

The development of brain metastasis in melanoma patients constitutes an independent negative prognostic factor. Patients with BRAF V600E/K-positive melanoma with brain metastases can respond to available BRAFi/MEKi combinations, but the clinical benefits are notably more short-term compared with metastatic patients with no intracranial metastatic lesions; furthermore, frequent relapse occurs primarily at the brain metastatic site (14).

In brain metastasis, the limited blood–brain barrier (BBB) permeability of approved BRAFi could explain the short duration of the observed responses. Moreover, several authors have described that the brain microenvironment itself can promote activation of alternative signaling and/or drive clone selection to induce resistance to BRAF inhibition (15, 16).

Recently (17), a novel BRAFi, compound 1a (C1a), has shown optimized brain penetration properties without inducing MAPK hyperactivation in BRAF WT settings (paradox breaker). Here, we characterized the activity of this novel agent in the context of melanoma brain metastasis. In particular, we investigated its potential use in BRAFi/MEKi-treated patients, studied its mechanisms of resistance, and examined combination strategies with immune-checkpoint inhibition to drive durable therapeutic responses.

A375 (CRL-1619), YUMM1.7 (CRL-3362), SK-MEL-28 (HTB-72), SK-MEL-5 (HTB-70), and B16-F0 (CRL-6322) were obtained from the ATCC. The cell lines SK-MEL-3 (ACC321), SK-MEL-30 (ACC151), and MEL-JUSO (ACC74) were from the German Collection of Microorganisms and Cell Cultures (DSMZ). All cell lines were maintained in a humidified atmosphere with 5% CO₂ at 37°C.

Cell line identity was confirmed through short-tandem-repeat PCR (performed at Mycrosynth-Switzerland), and the absence of Mycoplasma contamination was verified through the MicoAlert KIT (Lonza) according to the manufacturer's instructions.

The patient-derived lines referred to as patient 1 (Pt1) and patient 2 (Pt2) were maintained as tumorspheres in neurobasal medium (#21103049; Life Technologies) supplemented with B27 (#A3582801; Life Technologies), penicillin/streptomycin (#15140148; Life Technologies), 20 ng/mL EGF (#AF-100-15; PeproTech), and 20 ng/mL FGF-2 (#100-18C-0100; PeproTech).

Encorafenib (HY-15605), dabrafenib (HY-14660A), binimetinib (HY-15202), trametinib (HY-10999), LY3009120 (HY-12558), and cobimetinib (HY-13064A) were purchased from MedChemExpress; C1a was internally synthesized; mouse PD1 antibody (m PD1) clone RMP1-14 (Bio X Cell #BP0146) was administered to mice i.v. through tail-vein injection.

A375 parental cells were transduced with virus particles generated as described elsewhere (18) ORF sequences are reported in Supplementary Materials and Methods. Viral transduction was performed for 24 hours in the presence of 0.8 μg/mL polybrene infection reagent (Millipore; cat. #TR-1003-G), and cells were selected with the addition of 1 μg/mL puromycin (Thermo Fisher; cat. #A1113803).

Homogeneous time-resolved fluorescence (HTRF) assay for the measurement of pERK levels was performed according to the manufacturer's instruction by using the Advanced ERK phospho-T202 /Y204 HTRF assay (Cisbio #64AERPEH) and the total ERK HTRF assay kit (Cisbio #64NRKPEG).

Western blot was performed according to previously reported procedures (17) on lysates generated upon cell incubation with IP Lysis buffer (Pierce; #87788), supplemented with 1× protease/phosphatase inhibitor cocktail (Thermo Fisher Scientific; #78444). The antibodies utilized are listed in Supplementary Materials and Methods.

siRNA transfection was performed using Lipofectamine RNAiMAX Transfection reagent (Thermo Fisher; cat. #13778150) and ON-TARGETplus Smartpool siRNA pool from Horizon Discovery Ltd (L-003460-00-0020).

Animal experiments were approved by and performed according to the guidelines of the Institutional Animal Care Committee of the VHIO Institute in agreement with the European Union and national directives. Female and male NOD scid gamma mice of 4- to 5-week-old were purchased from Charles River. For intracranial models, 5 × 10⁵ A375-LUC were stereotactically inoculated into the corpus striatum of the right brain hemisphere (1 mm anterior and 1.8 mm lateral to the lambda; 2.5 mm intraparenchymal) of mice. For the subcutaneous model, 5 × 10⁶ A375 melanoma cells were injected into one flank. When tumors reached 100 to 300 mm³ (subcutaneous) or bioluminescence signal was detected (intracranial), mice were randomized and drugs were administered daily by oral gavage. Tumor progression was monitored by bioluminescence measurements upon intraperitoneal injection of d-luciferin (Deltaclone #12507), by using the Xenogen IVIS Spectrum apparatus. Bioluminescence images were taken at the same time and exposure to be comparable among experimental groups (Supplementary Fig. S1). Subcutaneous tumor size was measured through a digital caliper.

The allograft Yumm1.7 model was conducted with 4- to 6-week-old C57/BL6 female mice at Labcorp Drug Development Inc. Animal care and studies were conducted in an Association for Assessment and Accreditation of Laboratory Animal Care International–accredited facility in alignment with relevant animal welfare regulatory requirements.

Human melanoma brain metastasis samples were obtained from the Vall d'Hebron University Hospital and Clinic Hospital. The protocol to obtain samples was approved by the Hospital Institutional Review Board (PR(AG)478/2017), and informed consent was obtained for all patients in accordance with ethical guidelines. Tumor fragments from surgical human brain metastasis were implanted subcutaneously in mice flank with Matrigel (Corning; #356237). Upon tumor establishment, mice were randomized and treated as indicated.

For the intracranial model, tumoroids of Pt2 were previously transduced with a luciferase-encoding lentivirus vector. 3 × 10⁵ tumor-derived cells were stereotactically inoculated into the corpus striatum of the right brain hemisphere (1 mm anterior and 1.8 mm lateral to the lambda; 2.5 mm intraparenchymal).

RNA and DNA were coextracted from 1×10⁶ fresh frozen cells (resistant cell lines derived from xenografts), buffy coat, or human brain tissue samples using the Qiagen AllPrep DNA/RNA Mini kit (cat. #80204) according to the manufacturer's instructions. For the patient samples, only DNA was extracted from blood (germline) and tumor tissue. Quality control was performed on a Bioanalyzer (Agilent RNA Nano kit; cat. #5067-1511), and RNA and DNA quantities were determined on a Nanodrop reader.

Whole-exome sequencing was performed using the Human All Exon V6+COSMIC kit (part # 5190-9307) according to the manufacturer's instructions.

Reads were mapped onto the human genome draft GRCh38 using the program BWA (19). SNVs and small indels were called using the GATK suite v.4 with the HaplotypeCaller (20) and MuTect2 (21) modules for germline and somatic variants, respectively. Variants were annotated using the program SnpEff (22) and the GRCh38.86 annotation.

RNA-seq was performed using the Illumina TruSeq Stranded mRNA (cat. #20020595) workflow according to the manufacturer's instructions.

Base calling was performed with BCL to FASTQ file converter bcl2fastq2 version 2.20.0 (Illumina). FASTQ files were quality checked with FastQC version 0.11.5. RNA-seq paired-end reads were mapped onto the human genome (build hg38) with read aligner STAR version 2.7.3a using default mapping parameters (23) and gene fusions were identified using STAR-Fusion v.1.9.0 (24). Differential gene expression was performed using the limma package in R (25) and gene set enrichment analysis (GSEA) using the GSEA program version 4.2.3 (26) and the hallmark gene sets.

RNA was reverse transcribed, amplified by PCR using primers annealing to exon 18 (forward) and exons 10 and 11 (reverse) of BRAF, and the purified products were sequenced by Sanger. BRAF_exon18 (fwd): 5′-GCATCAGAACCCTCCTTGAA-3′, BRAF_ exon10 (rev): 5′-GACTTCCTTTCTCGCTGAGG-3′, BRAF_exon11 (rev): 5′- CTCGAGTCCCGTCTACCAAG-3′. The MKRN-1–BRAF fusion was confirmed through PCR with primers annealing on exon 4 of MKRN1 (fwd) 5′-CCACGGAGATTCTTGTGACAT-3′ and BRAF on Exon 10 (rev) 5′-GCAGACAAACCTGTGGTTGA-3′. The purified PCR product was then subjected to Sanger sequencing.

IHC and flow cytometry for the detection of CD45 and CD3-positive cells were conducted according to standard procedures and are reported in Supplementary Material and Methods.

In vivo data were analyzed using GraphPad Prism 5.0 software. To compare two different groups, we calculated P value (P) using the Mann–Whitney test for nonparametric variables. Survival curves comparison was performed using the log-rank (Mantel–Cox) test. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. For the in vitro experiments statistics, ANOVA analysis has been performed.

Statistic for relapse rate in the Yumm1.7 study was performed as time to relapse using the log-rank test model.

Whole-exome sequencing (WES) and RNA-seq data generated in this study are available in Sequence Read Archive (PRJNA837110). All data presented in this article were generated by the authors and included in the article, in the supplementary section or available on request.

The novel BRAFi C1a was developed to have optimized brain permeability, to avoid MAPK paradoxical activation (paradox breaker) and have selective activity in BRAF V600E/K tumors. We initially aimed to validate and confirm the properties of this novel BRAFi. The treatment of BRAF WT and HCT116 cells with C1a revealed that this agent, unlike dabrafenib and encorafenib, did not promote increased levels of pERK at clinically relevant doses, thus confirming its paradox breaker properties (Fig. 1A). Moreover, it only triggered robust cytotoxic responses in BRAF V600E models within a panel of melanoma cell lines (Fig. 1B). These initial data supported the potent activity of C1a and its differentiated profile on paradoxical MAPK activation in comparison with approved BRAFi.

We then tested C1a in patient-derived tumoroids generated from brain metastases of two patients affected with BRAF V600E-positive melanoma. Both patients manifested progressive disease at the site of brain metastases after prolonged treatment with dabrafenib/trametinib and other lines of therapy (Fig. 1C). Interestingly, sequencing data of these tumors did not reveal, outside of the expected BRAF V600E mutation, acquired genetic alterations known to drive resistance to BRAFi/MEKi (Fig. 1D). Hypothesizing that the limited brain exposure of approved BRAFi/MEKi could represent the predominant causative event responsible for the brain metastatic relapse, we evaluated the impact of BRAF inhibition by measuring pERK levels upon drug treatment and included the BRAF V600E-mutant and BRAFi-sensitive cell line A375 as a reference control.

All BRAFi tested triggered pERK repression in the patient-derived models similarly to the A375 cell line (Fig. 1E). This suggested that the patient-derived models retained the intrinsic cellular dependence on BRAF despite they were derived from brain metastatic lesions progressing on BRAFi/MEKi.

We next examined the in vivo activity of dabrafenib and C1a on subcutaneous xenografts derived from these patients.

Results showed that whereas dabrafenib at a dose of 100 mg/kg effectively triggered tumor growth inhibition in Pt1-derived tumors and tumor stasis in Pt2-derived tumors, C1a promoted complete tumor regression at all tested doses (5–10–20 mg/kg), in both models (Fig. 2A and B). Tumor recurrence to the administered agents was noted only upon drug treatment suspension.

These results indicated that both patient-derived models were sensitive to BRAF inhibition in vivo and that C1a showed higher efficacy when compared with dabrafenib, suggesting that the better pharmacokinetic properties of C1a and/or its paradox breaker properties improved its efficacy.

We then assessed the potency of BRAF inhibition in the intracranial setting. Orthotopic patient-derived xenografts (PDX) models of Pt2 brain metastasis were treated with C1a. A clear antitumor activity associated with tumor regression (Fig. 2C and D) and survival benefits (Fig. 2E) were observed in C1a-treated mice, demonstrating its efficacy in the intracranial setting.

Of note, a higher dose of C1a was utilized in this experiment in order to maximize exposure to this agent. Importantly, as described in previous studies, no sign of intolerability nor neurologic effects was reported with C1a dose levels of up to 180 mg/kg (17).

We next aimed to study the efficacy of C1a in parallel subcutaneous and intracranial A375 xenografts upon relapse to dabrafenib. The A375 model was chosen as representative of a BRAFi-naïve context. Mice were divided into four groups: one vehicle control, one receiving daily oral administration of dabrafenib 100 mg/kg, one receiving oral C1a 10 mg/kg, and another group receiving dabrafenib until evidence of tumor progression, at which point, dabrafenib was suspended and mice were rechallenged with C1a at 10 mg/kg. Results from the subcutaneous xenografts revealed that dabrafenib triggered tumor stasis until day 26 from study initiation followed by slow tumor regrowth, which occurred simultaneously in all treated mice (Fig. 3A). Conversely, treatment with C1a triggered complete tumor regression in all mice with individual tumors asynchronously manifesting treatment escape and resistance acquisition within a treatment window of up to 110 days (Fig. 3A and B). Moreover, in tumors progressing on dabrafenib, the rechallenge with C1a resulted in rapid and complete remission, subsequently followed by a pattern of tumor escape similar to the one observed in tumors treated with C1a alone (Fig. 3A). Survival analysis of this study clearly demonstrated greater survival benefits in mice treated with C1a and in mice receiving C1a treatment upon dabrafenib relapse, when compared with vehicle or dabrafenib (Fig. 3B).

We performed an analogous experiment in the intracranial setting with A375 luciferase-expressing cells. Results indicated that whereas dabrafenib exerted some initial tumor control in brain xenografts (Fig. 3C), minimal survival benefits were observed with this agent. Conversely, C1a treatment promoted marked tumor regression (Fig. 3C and D).

Survival analysis of these brain implanted tumors again demonstrated similar benefits with C1a treatment in both drug-naïve and dabrafenib-pretreated cohorts. However, mice ultimately succumbed due to tumor progression within 38 days from study initiation (Fig. 3E) evidencing differential activity even for the brain-penetrant C1a treatment between tumors in the subcutaneous or the brain setting.

In order to enhance the exposure of the BRAFi in the brain compartment and test an optimal dose regimen, A375 brain xenograft-bearing mice were treated with C1a at 75 mg/kg. Mice treatment was continued until tumor relapse.

Moreover, in this study we aimed to compare our results to the clinically relevant doses of the combinations of BRAFi/MEKi dabrafenib (100 mg/kg)/trametinib (0.25 mg/kg; D/T) and encorafenib (36 mg/kg)/binimetinib (10 mg/kg; E/B).

The D/T- and E/B-treated mice were further divided into two subgroups: one treated with these combinations until relapse, or the second where D/T and E/B treatment were suspended at early signs of disease progression, and mice were rechallenged with C1a at 75 mg/kg.

The E/B and D/T combinations triggered an antitumor response and an evident survival benefit superior to those observed with dabrafenib monotherapy in the previous study (Fig. 4A–D), but manifested disease progression within a maximum of 64 days from the study initiation. Conversely, C1a treatment exerted a potent antitumor response (Fig. 4B–D), resulting in prolonged survival benefits (Fig. 4A).

Moreover, the groups progressing to D/T and E/B, manifested dramatic tumor regression upon rechallenge with 75 mg/kg C1a (Fig. 4E–H for D/T and Fig. 4I–L for E/B), ultimately resulting in prolonged survival benefit lasting up to 120 days from the initiation of the study (Fig. 4A);

Collectively, these data indicate that C1a as monotherapy, at a well-tolerable regimen, is highly effective in achieving prolonged responses in peripheral and metastatic tumors including brain metastasis that progressed on approved BRAFi/MEKi.

We next aimed at exploring mechanisms of tumor relapse to C1a to inform on eventual genetic determinants of resistance to this novel agent by generating a collection of cell lines from explanted tumors that manifested obvious drug escape during in vivo treatment.

The impact of C1a on cell proliferation and clonogenicity on the ex vivo–derived resistant clones from subcutaneous tumors (Fig. 3A) revealed resistance to C1a (up to a concentration of 1 μmol/L) in cells from tumors that escaped C1a treatment (lines #1–#5), and in cells from tumors that were rechallenged with C1a upon dabrafenib relapse (lines #11–#15; Fig. 5A). Moreover, these cells manifested similar cross-resistance in vitro to dabrafenib, but to a lesser extent to PanRAFi LY3009120 (Supplementary Fig. S2; ref. 27).

We then evaluated the impact of C1a on MAPK signaling by evaluating pERK levels through HTRF assay. In resistant clones, sustained pERK levels were maintained despite drug treatment across all lines tested. This was consistent with the evident observed shift of pERK IC₅₀ (Fig. 5B and C). Furthermore, basal hyperactivation of MAPK signaling with pERK levels well above those observed in parental A375 cells were noted in the absence of drug treatment.

We analyzed additional clones from tumors relapsing on C1a at 10 mg/kg or dabrafenib from brain xenografts (Fig. 3E). These cell lines demonstrated sensitivity to C1a or dabrafenib treatment analogous to the parental A375 cells (Supplementary Fig. S3), suggesting that, in these clones, tumor resistance was not associated with ex vivo retained mechanisms, but most likely due to adaptation to insufficient drug exposure in the brain.

As our previous xenograft study with C1a at the optimized dose to achieve effective CNS exposure of 75 mg/kg clearly demonstrated long-lasting responses and second-line activity in D/T and E/B relapsed brain metastasis (Fig. 4), we further explored the resistance associated with this optimized drug regimen.

The additional ex vivo lines from brain xenografts relapsing on C1a treatment at 75 mg/kg (lines #1–75 to #5–75) or that relapsed after C1a rechallenge of D/T (lines #6DT-75 to #8DT-75) and E/B (lines #11-EB75 to #14EB-75) treated mice were established.

Ex vivo resistance to C1a was observed to these ex vivo—derived clones with a shift of C1a activity of ∼ 10-fold compared with parental A375 cells (Fig. 5D).

The subsequent evaluation of the MAPK pathway activity clearly revealed elevated basal pERK levels in all resistant clones and a relevant shift in the C1a concentrations required to trigger pERK repression compared with parental A375 (Fig. 5E and F).

Taken together, these results highlight that MAPK signaling restoration is the predominant drug-escape mechanism occurring with C1a in vivo when administered at relevant regimens to achieve prolonged antitumor response, in both the peripheral and brain metastatic settings.

The analysis of ex vivo–derived lines from tumors demonstrated a highly resistant phenotype consistent with the ability to withstand C1a treatment up to micromolar concentrations. Moreover, an analogous drastic phenotype was also observed in one additional A375-derived line obtained in vitro (A375R1) upon chronic exposure to C1a (Fig. 6A and B).

We then aimed to explore the genetic events occurring in these models that are responsible for such a dramatic ability to resist C1a treatment. The 10 in vivo–derived cell lines presenting resistance to C1a together with the in vitro–derived line and parental A375 cells were subjected to WES to identify genetic drivers of resistance. Although no obvious resistance-causing genetic events were detected in 8 out of the 10 cell lines, BRAF immunoblotting revealed an additional band of 140 KDa reactive with the BRAF antibody. Interestingly, the same additional band was also detected in the A375R1 clone raised to resist C1a in vitro (Fig. 6C and D).

We next performed BRAF knockdown experiments to evaluate whether these cell lines retain BRAF dependency. BRAF ablation was associated with a drastic repression of pERK levels and a loss of cell viability. Importantly, the siRNA-mediated BRAF knockdown also ablated the additional 140 KDa band in all the lines, clearly revealing that a BRAF protein rearrangement had occurred (Fig. 6E and F).

We thus performed RNA-seq of these resistant models, which showed discordantly mapping reads on exons 10, 11,13 and 17, 18 at the BRAF locus (Supplementary Fig. S4) in the 8 ex vivo cell lines and in the in vitro–generated A375R1 model, which were positive for the additional 140 KDa BRAF band.

Similar observations on discordantly mapping reads and an additional 140 KDa band in BRAF WB experiments were previously reported by Kemper and colleagues (28) in models and patient samples manifesting resistance to vemurafenib. The authors demonstrated that the acquisition of a tandem domain duplication of the kinase domain was responsible for resistance to this agent.

We thus investigated whether the BRAF kinase domain duplication occurred in our models through PCR on cDNA with primers flanking exons 18 and exon 10 or 11, respectively (Fig. 7A). PCR products and subsequent Sanger sequencing (Supplementary Fig. S5A) effectively confirmed the existence of the internal tandem duplication in all samples that presented the additional 140 KDa BRAF band.

Of note, the involvement of the activation of the PI3K pathway as a determinant of resistance was ruled out as p-AKT levels remained unaffected or even diminished in these lines (Supplementary Fig. S5B).

Between the two resistant cell lines lacking the additional BRAF band, WES identified an acquired NRAS G13R mutation in one, an event further supported by the elevated RAS-GTP activity in this line (Supplementary Fig. S5C). The other line carried an MKRN1—BRAF fusion that was detected by RNA-seq data analysis and subsequently confirmed by PCR and Sanger sequencing (Supplementary Fig. S5D).

Interestingly, GSEA in the eight clones presenting the kinase domain duplication compared with A375 parental cells showed upregulation of genes relative to the MYC transcriptional program and to the inflammatory response (Supplementary Fig. S6). The latter, in particular, supports further investigation of combination therapies with immune-checkpoint blockade.

To validate if the kinase domain duplication and the MKRN1—BRAF fusion were effective drivers of resistance to C1a, we ectopically stably expressed these ORFs in A375 (Fig. 7B) and SK-MEL-28 cells (Supplementary Fig. S7).

Notably, two potentially alternative forms of kinase domain duplication were included in validation (KDD1 and KDD2) based on the inclusion or not of an alternative spliced exon within the BRAF transcripts (Supplementary Fig. S4).

Results showed that ectopic expression of BRAF KDD1, KDD2, and the MKRN1–BRAF fusion rescued the activity of C1a and dabrafenib on pERK levels and cell viability in A375 cells (Fig. 7C and D) and in the SK-MEL-28 model (Supplementary Fig. S7).

As our previous result evidenced the enrichment of a proinflammatory transcriptional program in C1a-resistant cells, and considering the already established clinical use of ICI in melanoma, we aimed at exploring whether a combinatorial approach with C1a and anti–PD-1 could prolong therapeutic benefits and limit tumor relapse.

We thus established allografts from the BRAF V600E model Yumm1.7 in immunocompetent C57/BL6 mice and, at tumor establishment, treated the mice with C1a at 5 mg/kg daily, mouse anti–PD-1 antibody at 12.5 mg/kg i.v. once a week or the combination of the two agents. The treatment was performed until day 31 from tumor implantation and mice were subsequently monitored, untreated, until day 61 to follow tumor recurrence.

Results (Fig. 8A) illustrated that whereas anti–PD-1 did not produce a meaningful antitumor activity in Yumm1.7-derived tumors, evidence previously documented (29), C1a promoted complete tumor remission in all mice throughout the duration of the treatment. Moreover, during the subsequent untreated monitoring phase, 4 of 10 mice in the C1a-treated group evidenced tumor recurrence, whereas in the C1a/anti–PD-1 combination group, none of the 10 mice displayed tumor recurrence.

These results are indicative that the treatment with C1a contributes to unleash an immune-mediated phenotype that sensitized the tumor to anti–PD-1 in a model otherwise poorly immunogenic and nonresponsive to ICI (29).

To validate this hypothesis, we quantified the immune infiltrate in Yumm1.7 allografts upon treatment with C1a, anti–PD-1, and their combination.

Flow cytometry–based quantification of the immune cell marker CD45 and the T-cell marker CD3 in tumors confirmed a dramatic increase of immune infiltration in the C1a-treated group and the C1a + anti–PD-1 combination (Fig. 8B). Further validation of the dramatic increase of the immune infiltrate generated by C1a treatment was observed through CD45 IHC (Fig. 8C). Furthermore, to explore the impact of the drug treatment on T cells, we assessed the activity of C1a but also of dabrafenib (paradox inducer BRAFi) and trametinib (MEKi) on primary T-cell proliferation. We observed that whereas MEK inhibition drastically impaired T-cell proliferation, the BRAFi dabrafenib enhanced T-cell proliferation (consistent with the paradoxical MAPK activation), and C1a left T-cell proliferation unaffected (Supplementary Fig. S8).

Taken together, our results showed that C1a not only induces tumor regression as monotherapy but also promotes potent antitumor responses when combined with anti–PD-1 preventing disease relapse in our experimental setup.

Here, we investigated the preclinical activity of a novel brain-penetrant, paradox breaker BRAFi in BRAF V600E melanoma brain metastasis and provide evidence to support the clinical testing of this novel agent.

Our studies with patient-derived samples from brain metastases progressing on BRAFi/MEKi showed a similar activity between C1a, dabrafenib, and encorafenib when tested in vitro. Conversely, the evident superiority of C1a compared with dabrafenib was observed in vivo. These data support that clinical relapse to approved BRAFi/MEKi treatment in the brain compartment can occur with adaptation to these drugs in the absence of genetic events driving drug resistance. Corroborating this evidence, no obvious mutations known to drive resistance to BRAFi/MEKi were observed in these patient-derived tumors. Moreover, our results also suggest that suboptimal drug exposure in brain metastases could represent the predominant mechanism responsible for the faster tumor recurrence observed in the clinic to currently available agents.

In line with these data, multiple reports have revealed that bulky brain metastatic lesions exhibit compromised BBB permeability allowing effective local drug perfusion (30, 31). However, after an initial response, BBB restoration at the regressed metastatic site and the spread of tumor cells in the brain parenchyma may constitute a local microenvironmental condition, which limits drug exposure, facilitates adaptation to drug pressure, and allows fast tumor relapse (32, 33). Supporting this hypothesis, a recent clinical study reported responses to encorafenib/binimetinib in brain metastasis previously progressing on dabrafenib/trametinib treatment, thus providing clinical evidence that disease progression devoid of evident genetically driven resistance can occur (34).

Our results on PDX also clearly evidenced dramatic superiority of C1a compared with approved BRAFi as supported by the complete tumor remission achieved with low doses of C1a.

We then further examined whether C1a could present activity in patients progressing on available BRAF/MEKi. We utilized the A375 subcutaneous and intracranial xenograft models and treated them with C1a, first-line or upon relapse to approved targeted therapies. Our initial study of C1a in the A375 model demonstrated that prolonged responses in subcutaneous tumors were triggered at the low dose of 10 mg/kg and prolonged survival benefits were evident even when C1a treatment was performed after relapse to available BRAFi. We next evaluated, for C1a, whether an optimized and well-tolerated regimen, which produces optimal exposure in the CNS, could be efficacious in brain metastases even upon relapse to BRAFi/MEKi. The treatment with C1a at 75 mg/kg triggered long-lasting responses in both drug-naïve and BRAFi/MEKi relapsing mice, indicating that the maximization of the exposure in the brain compartment should be a primary goal to achieve effective responses in patients with melanoma brain metastasis. Furthermore, our data indicate that an optimal C1a regimen could drive superior and prolonged benefits in brain metastasis, even upon relapse on available BRAFi/MEKi, thus supporting the clinical investigation of C1a in patients with brain metastatic lesions progressing to approved BRAFi/MEKi.

The genetic drivers of resistance to approved BRAFi and BRAFi/MEKi predominantly converge toward MAPK reactivation that is enabled by RAF dimer–mediated signaling (9, 35). Because C1a presents paradox breaker properties that could address several RAF dimer–mediated mechanisms of resistance, we aimed at deciphering the resistance mechanisms occurring upon chronic C1a exposure. The identification of these mechanisms can provide predictive biomarkers to exclude patients who might not benefit from C1a, and trigger additional studies on drug combinations to overcome these mechanisms and further prolong antitumor responses. The ex vivo analysis of tumors manifesting relapse on C1a demonstrated that MAPK reactivation occurs even with this novel BRAFi, despite its paradox breaker properties. This evidence aligns well with previously reported findings on the paradox breaker PLX8394 (36).

Importantly, brain tumors relapsing on C1a at 75 mg/kg retained resistance ex vivo, which was associated with MAPK reactivation, further supporting that, with this regimen, effective drug exposure was achieved and resistance developed through genetic determinants.

The profiling of highly resistant clones to C1a uncovered the acquisition of a BRAF kinase domain duplication as a predominant genetic driver of resistance, together with a rare NRAS G13R mutation (37, 38) and a previously described MKRN1–BRAF fusion (39). Follow-up mechanistic studies then confirmed that both the kinase domain duplication and the MKRN1–BRAF fusion could trigger MAPK restoration and resistance to C1a and dabrafenib (but less to PanRAF inhibitors).

Worthy of note is that Kemper and colleagues (28) previously identified the same kinase duplication that we observed driving resistance to C1a in ∼10% of vemurafenib-resistant PDX available in their tumor collection, indicating that a patient population acquiring this genetic event upon relapse to approved BRAFi is expected to present cross-resistance to C1a. It is noteworthy that the analysis of the RNA-seq data in these resistant cells revealed an unpregulation of a proinflammatory transcriptional program, which further prompted us to test whether the combination of C1a with immune-checkpoint inhibition could represent a valuable strategy to limit resistance acquisition.

Using a model refractory to anti–PD-1 treatment, we observed that C1a treatment can drive significant tumor inflammation as revealed by a robust CD45- and CD3-positive infiltrate, and, importantly, the treatment combination C1a plus anti–PD-1 proved to dramatically reduce the onset of relapse when compared with C1a monotherapy.

It is noteworthy that recent clinical data reported clear benefits of the concomitant administration of vemurafenib (BRAFi), cobimetinib (MEKi), and atezolizumab (anti–-PD-L1) albeit in the presence of substantial toxicity (40, 41). Conversely, a second clinical trial examining the combination of only MEKi and anti–PD-L1 failed to show superior clinical benefits over the anti–PD-1 arm but still manifested an increase in toxicity of the combination regimen (42). Together, these clinical data clearly suggest that the potential of omitting the administration of MEKi in drug combinations involving BRAF targeted therapies and immune-checkpoint blockage, an opportunity offered by the paradox breaker features of C1a, could represent a key advantage to maximize therapeutic benefits and improve the safety profile of the therapeutic regimen.

Collectively, our findings support that C1a, by presenting excellent brain-penetration properties, could represent a novel valuable therapeutic opportunity for the treatment of patients with melanoma brain metastasis, even after relapse on approved BRAFi/MEKi. The identified mechanisms of resistance to C1a could potentially inform patient stratification strategies. Moreover, the combination of C1a with immune-checkpoint blockade could further prolong therapeutic benefits offering new hope to patients.

E. Bonfill-Teixidor reports grants from F. Hoffmann-La Roche AG during the conduct of the study and has a patent licensed to F. Hoffmann-La Roche AG. R. Iurlaro reports grants from F. Hoffmann-La Roche AG during the conduct of the study; in addition, R. Iurlaro has a patent licensed to F. Hoffmann-La Roche AG. C. Handl reports personal fees from Hoffmann-La Roche during the conduct of the study and personal fees from Hoffmann-La Roche outside the submitted work. J. Wichmann reports personal fees from F. Hoffmann-La Roche AG during the conduct of the study and personal fees from F. Hoffmann-La Roche AG outside the submitted work; in addition, J. Wichmann has a patent for WO2021/116055 pending and a patent for the use of BRAFi pending. A. Arias reports grants from F. Hoffmann-La Roche AG during the conduct of the study. I. Cuartas reports grants from F. Hoffmann-La Roche AG during the conduct of the study. J. Emmenegger reports personal fees from F. Hoffmann- La Roche outside the submitted work. A. Romagnani reports personal fees from Hoffmann-La Roche during the conduct of the study and personal fees from Hoffmann-La Roche outside the submitted work. T. Lorber reports personal fees from F. Hoffmann-La Roche during the conduct of the study and personal fees from F. Hoffmann-La Roche outside the submitted work. M. Berrera reports personal fees from F. Hoffmann-La Roche AG during the conduct of the study and personal fees from F. Hoffmann-La Roche AG outside the submitted work. C. Godfried Sie reports personal fees from F. Hoffmann-La Roche AG during the conduct of the study and personal fees from F. Hoffman-La Roche AG outside the submitted work. F. Köchl reports personal fees from F. Hoffmann-La Roche AG during the conduct of the study and personal fees from F. Hoffmann-La Roche AG outside the submitted work. J. Eckmann reports personal fees from F. Hoffmann-La Roche during the conduct of the study and personal fees from F. Hoffmann-La Roche outside the submitted work. R. Feddersen reports personal fees from F. Hoffmann-La Roche during the conduct of the study and personal fees from F. Hoffmann-La Roche outside the submitted work. M. Kornacker reports being an employee of F. Hoffmann-La Roche Ltd. and owning shares of F. Hofmann-La Roche Ltd. G. Schnetzler reports personal fees and other support from Roche during the conduct of the study. E. Muñoz-Couselo reports personal fees from BMS, Merck, Sanofi, Pierre Fabre, Novartis, and Roche outside the submitted work. J. Tabernero reports personal fees from Array Biopharma, AstraZeneca, Avvinity, Bayer, Boehringer Ingelheim, Chugai, Daiichi Sankyo, F. Hoffmann-La Roche Ltd, Genentech Inc., HalioDX SAS, Hutchison MediPharma International, Ikena Oncology, Inspirna Inc, IQVIA, Lilly, Menarini, Merck Serono, Merus, MSD, Mirati, Neophore, Novartis, Ona Therapeutics, Orion Biotechnology, Peptomyc, Pfizer, Pierre Fabre, Samsung Bioepis, Sanofi, Seattle Genetics, Scandion Oncology, Servier, Sotio Biotech, Taiho, Tessa Therapeutics, and TheraMyc and personal fees from Imedex, Medscape Education, MJH Life Sciences, PeerView Institute for Medical Education and Physicians Education Resource outside the submitted work. J.R. Bischoff reports other support from Roche outside the submitted work. P. Pettazzoni reports personal fees from F. Hoffmann-La Roche AG during the conduct of the study and personal fees from F. Hoffmann-La Roche AG outside the submitted work; in addition, P. Pettazzoni has a patent for WO2021/116055 pending. J. Seoane reports grants from Hoffmann-La Roche during the conduct of the study and grants from Roche Glycart, AstraZeneca, Northern Biologics, and Mosaic Biomedicals outside the submitted work; in addition, J. Seoane has a patent pending. No disclosures were reported by the other authors.

E. Bonfill-Teixidor: Conceptualization, resources, data curation, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. R. Iurlaro: Conceptualization, data curation, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. C. Handl: Conceptualization, resources, data curation, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. J. Wichmann: Conceptualization, resources, data curation, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. A. Arias: Conceptualization, data curation, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. I. Cuartas: Conceptualization, resources, data curation, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. J. Emmenegger: Conceptualization, resources, data curation, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. A. Romagnani: Conceptualization, resources, data curation, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. L. Mangano: Conceptualization, resources, data curation, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. T. Lorber: Resources, data curation, software, validation, investigation, methodology, writing–original draft. M. Berrera: Resources, data curation, software, investigation, methodology, writing–original draft. C. Godfried Sie: Resources, data curation, software, investigation, methodology. F. Köchl: Resources, data curation, investigation, methodology, writing–original draft, project administration, writing–review and editing. J. Eckmann: Resources, data curation, investigation, methodology, writing–original draft, project administration, writing–review and editing. R. Feddersen: Resources, data curation, investigation, methodology, writing–original draft, project administration, writing–review and editing. M. Kornacker: Resources, writing–original draft, project administration, writing–review and editing. G. Schnetzler: Resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing. M. Cicuendez: Resources. E. Cordero: Resources, data curation, software, supervision, funding acquisition, investigation, methodology, writing–original draft. T.E. Topczewski: Resources, data curation, software, supervision, funding acquisition, validation, investigation, methodology, writing–original draft. A. Ferres-Pijoan: Resources, data curation, software, supervision, funding acquisition, validation, investigation, methodology, writing–original draft. J. Gonzalez: Conceptualization, resources, data curation, software, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. F. Martínez-Ricarte: Conceptualization, resources, data curation, software, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. E. Muñoz-Couselo: Conceptualization, resources, data curation, software, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. J. Tabernero: Conceptualization, resources, data curation, software, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. J.R. Bischoff: Conceptualization, resources, data curation, supervision, funding acquisition, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. P. Pettazzoni: Conceptualization, resources, data curation, supervision, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing. J. Seoane: Conceptualization, resources, data curation, validation, investigation, methodology, writing–original draft, project administration, writing–review and editing.

This work was supported by the Roche Postdoctoral fellowship program, the Fundación Asociación Española contra el Cáncer (AECC), FERO (EDM), Ramón Areces Foundation, Cellex Foundation, BBVA (CAIMI), the ISCIII, and the FIS (PI19/00318). The authors thank Jeannine-Petrig Schaffland, Alessandro Brigo, Erich Küng, David Dejardin, and Vera Griesser for technical assistance (Roche Pharma Research and Early Development pRED).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.