MAPK targeting in cancer often fails due to MAPK reactivation. MEK inhibitor (MEKi) monotherapy provides limited clinical benefits but may serve as a foundation for combination therapies. Here, we showed that combining a type II RAF inhibitor (RAFi) with an allosteric MEKi durably prevents and overcomes acquired resistance among cancers with KRAS, NRAS, NF1, BRAFnon-V600, and BRAFV600 mutations. Tumor cell–intrinsically, type II RAFi plus MEKi sequester MEK in RAF complexes, reduce MEK/MEK dimerization, and uncouple MEK from ERK in acquired-resistant tumor subpopulations. Immunologically, this combination expands memory and activated/exhausted CD8+ T cells, and durable tumor regression elicited by this combination requires CD8+ T cells, which can be reinvigorated by anti–PD-L1 therapy. Whereas MEKi reduces dominant intratumoral T-cell clones, type II RAFi cotreatment reverses this effect and promotes T-cell clonotypic expansion. These findings rationalize the clinical development of type II RAFi plus MEKi and their further combination with PD-1/L1-targeted therapy.

Significance:

Type I RAFi + MEKi are indicated only in certain BRAFV600MUT cancers. In contrast, type II RAFi + MEKi are durably active against acquired MEKi resistance across broad cancer indications, which reveals exquisite MAPK addiction. Allosteric modulation of MAPK protein/protein interactions and temporal preservation of intratumoral CD8+ T cells are mechanisms that may be further exploited.

This article is highlighted in the In This Issue feature, p. 521

RAS–RAF–MEK–ERK signaling is hyperactivated in more than 30% to 40% of human cancers (1), including approximately 70% of advanced melanoma driven by BRAFV600, non-BRAFV600 (atypical BRAF), NRAS, and NF1 mutants. Type I RAF inhibitors (RAFi), such as vemurafenib, dabrafenib, and encorafenib, were first developed successfully against advanced BRAFV600MUT melanoma, but acquired resistance is almost universal. As a single agent, allosteric MEK1/2 inhibitors (MEKi), such as cobimetinib, trametinib, binimetinib, and selumetinib, elicit limited clinical response rates in a wide range of malignancies that harbor RAF and RAS mutations, for example BRAFV600MUT and NRASMUT melanoma (2, 3). However, the combination of type I RAFi + MEKi suppresses acquired resistance in tumors driven by BRAFV600 mutants and increases the therapeutic window in patients (4, 5), although acquired resistance caused by MAPK pathway reactivation is still commonplace (6–8). In human cancers where RAS–RAF–MEK–ERK signaling is not hyperactivated by BRAFV600 mutants, a MEKi-based combination that can sustainably block the MAPK pathway and resistance-associated reactivation has not been developed successfully.

A durably effective MEKi-based combination should prevent MAPK reactivation (9, 10) and preserve and/or promote antitumor T-cell immunity. MAPK-targeted therapy combined with immune-checkpoint blockade is undergoing clinical testing (11, 12), and antitumor T-cell immunity has been proposed to contribute to the durability of MAPK-targeted therapy (13). MEKi may be deleterious to tumor antigen–specific T cells (14, 15), which may be ameliorated by pulsatile MEKi dosing (16).

Next-generation MAPK pathway inhibitors may help overcome MEKi resistance commonly caused by MAPK reactivation, which is in turn caused by upregulation of “back-to-back” RAF/RAF and/or “face-to-face” RAF/MEK dimerization. For instance, BRAFV600 mutants, which signal as monomers but form dimers due to resistance-associated alterations, and non-V600, BRAF-activating mutants, which signal as RAS-independent dimers, can both be targeted by a novel dimer breaker, PLX8394 (17). These types of compounds “break” the paradox of type I RAFi, which is BRAF monomer–specific but has the liability of inducing RAF dimers and causing paradoxical MAPK hyperactivation in RAS-activated cancer cells. However, paradox breakers are not effective against CRAF wild-type (CRAFWT) or BRAFWT homodimers or heterodimers, which drive MAPK hyperactivation in the majority of cancers. Type II RAFi (aka dimeric or omni-RAFi) has antitumor activity in RASMUT or BRAFMUT cancers (18–22) but, as a single agent, does not appear to be highly active clinically (23, 24).

To evaluate the clinical potential and mechanisms underlying the combination of type II RAFi and allosteric MEKi, we investigated: (i) durability in preventing as well as overcoming acquired MEKi resistance across potentially MAPK-addicted cancer lineages, (ii) combinatorial mechanistic action consistent with preventing MAPK reactivation, and (iii) in vivo T-cell impacts.

Type II RAFi Combination Prevents and Overcomes Acquired MEKi Resistance

We tested type II RAFi (BGB-283 or RAF-709) at a submicromolar (0.5 μmol/L) concentration and/or allosteric MEKi (trametinib or binimetinib) at a nanomolar (20 nmol/L) concentration against a panel (n = 22) of human melanoma, colorectal carcinoma, pancreatic ductal adenocarcinoma (PDAC), and non–small cell lung carcinoma (NSCLC) cell lines driven by BRAFV600E, NF1−/−, NRASMUT, or KRASMUT in short-term (14-day) clonogenic assays (Fig. 1A; Supplementary Fig. S1A; Supplementary Table S1 for a list of cell lines and in vivo models used in this study). In general, although type II RAFi or MEKi individually were ineffective at preventing short-term growth, type II RAFi + MEKi prevented macroscopic growth over 14 days. Moreover, using a panel of BRAFV600E melanoma sublines (n = 4) with acquired resistance to type I RAFi + MEKi (vemurafenib + selumetinib/AZD-6244; ref. 7), we tested whether switching from type I RAFi to type II RAFi would overcome resistance (Fig. 1B). In one set of cultures, we maintained these sublines with both type I RAFi and MEKi or withdrew one or both inhibitors. As expected, withdrawal of both reduced growth fitness due to drug addiction (7, 25). In two additional sets of cultures, we introduced (or switched to) a type II RAFi (BGB-283 or RAF-709). Importantly, switching from type I to type II RAFi (over 10–15 days) overcame growth of resistant sublines, but this occurred only when an allosteric MEKi was present.

In patient-derived xenograft (PDX) models of PDAC (n = 2, KRASMUT) or of NSCLC (n = 1, BRAFnon-V600MUT; n = 1, KRASMUT), daily treatment with BGB-283 (20 mg/kg, orally) or trametinib (3 mg/kg, orally) alone led to minimal or transient tumor growth deceleration (Fig. 1C and D). However, in all four PDXs, combination treatment achieved durable tumor regression (86–104-day follow-ups), with complete responses (CR) noted in both NSCLC PDXs. Because the combination of type II RAFi + MEKi was highly effective in preventing MEKi resistance in short-term cultures of NRASMUT melanoma cell lines (Fig. 1E; Supplementary Fig. S1A), we also tested the combination against PDXs of NRASMUT melanoma (n = 3). In vivo, responses to daily treatment with BGB-283 (20 mg/kg, orally) or trametinib (3 mg/kg, orally) alone ranged from none to tumor stabilization (without regression; Fig. 1E). In all three NRASMUT melanoma PDXs, the combination achieved highly durable tumor regression in all tumors, resulting in 2/5 to 4/5 CRs (113–141-day follow-ups; Fig. 1E). CRs were confirmed in one PDX model by stopping treatment with both type II RAFi + MEKi at day 125, with no relapse during a 109-day follow-up. Tumors that relapsed in two mice were sensitive to retreatment on day 162 with both inhibitors (secondary CRs; Supplementary Fig. S1B).

Differences of Type II RAFi + MEKi (vs. Type II RAFi + ERKi) in Antagonizing Acquired MEKi Resistance

We then tested whether the durability of type II RAFi + MEKi in preventing acquired MEKi resistance is specific to this combination when compared with type II RAFi + ERKi. To compare the relative combinatorial potencies in preventing or overcoming acquired MEKi resistance, we first identified concentrations of MEKi (trametinib) versus ERKi (SCH772984 or BVD-523) with functionally equivalent impacts on the short-term (10 days) growth of MEKi-naïve, parental NRASMUT melanoma cell lines (M207, M245, and M296) or the short-term (8–11 days) growth of isogenic NRASMUT melanoma sublines with acquired MEKi (trametinib, 0.1 μmol/L) resistance (M207 SDR1, M245 SDR5, M296 SDR3; SDR, single-drug resistant; ref. 25). In MEKi-naïve, parental cells, 0.01 μmol/L trametinib and 0.1 μmol/L SCH772984 were approximately equivalent in suppressing short-term (10 days) growth or short-term (2 hours) pERK levels (Fig. 1F). In isogenic sublines with acquired MEKi resistance, switching from trametinib at 0.1 μmol/L to SCH772984 at 0.1 μmol/L or BVD-523 at 1 μmol/L had no or minimal impact on short-term (8–11 days) growth, as measured by the MTT assay, viable cell counting, or clonogenic assay (Supplementary Fig. S1C and S1D). We also evaluated the temporal impact of single-agent MEKi versus ERKi doses on the MAPK pathway in NRASMUT melanoma sublines with acquired MEKi resistance. We measured pERK and p-p90RSK levels after treatment with MEKi or ERKi for 1 hour or 48 hours, with or without prior treatments with the same inhibitor for 0, 2, 4, or 6 days (Supplementary Fig. S1E–S1G). Within this time span, treatment(s) with trametinib at 0.1 μmol/L, SCH772984 at 0.1 μmol/L, or BVD-523 at 1 μmol/L resulted in similar p-p90RSK levels, which are consistent with similar effective ERK activities and growth rates. However, in all acquired-resistant sublines, the highly similar effective ERK activities over time (as measured by p-p90RSK levels and cellular growth rates) between MEKi versus ERKi treatments contrasted glaringly with pERK levels. Only with ERKi (but not MEKi) treatment did we observe high rebounding levels of pERK 1 hour and 48 hours after the last fresh dose of treatment, regardless of the duration of prior ERKi treatment. This pERK rebound was stronger with BVD-523 and MK-8353 than with SCH772984. Unlike BVD-523 and MK-8353, SCH-772984 is known to inhibit both the kinase activity of ERK and MEK-mediated phosphorylation of the ERK activation loop. Rebound pERK levels in response to BVD-523 (1 μmol/L) and MK-8353 (1 μmol/L) treatment was abundant within 1 hour after treatment but were also readily detectable within 2 days of SCH-772984 at 0.4 μmol/L and within 4 to 6 days of SCH-772984 at 0.1 μmol/L. To our surprise, ERKi-induced pERK rebound levels did not accelerate growth rates of NRASMUT melanoma sublines with acquired MAPKi resistance.

Having established growth rate–equivalent concentrations of MEKi versus ERKi, we compared the relative potencies of combination with type II RAFi in preventing acquired MEKi resistance in drug-naïve cells. In cancer cell lines, acquired resistance to targeted therapy develops through temporal stages: early slow-cycling or drug-tolerant persistence (pseudosenescence) and later fast-cycling or proliferative growth (26). Accordingly, using parental NRASMUT melanoma cell lines (n = 3) in short-term (15 days) or long-term (30 days) cultures (Fig. 1G), we compared the efficacies of type II RAFi (BGB-283 at 0.5 μmol/L) in preventing growth when combined with functionally equivalent doses of MEKi (trametinib) versus ERKi (SCH772984; Fig. 1F). Although BGB-283 + trametinib and BGB-283 + SCH772984 were similarly effective at preventing 15-day growth, BGB-283 + trametinib was far more effective than BGB-283 + SCH772984 at preventing 30-day growth. Because proliferative MEKi-resistant clones are expected to arise with longer-term culture, type II RAFi + MEKi may be more active than type II RAFi + ERKi in overcoming the growth of proliferative MEKi-resistant clones. To test this hypothesis, we compared the relative potencies of type II RAFi + MEKi versus type II RAFi + ERKi in overcoming acquired MAPKi resistance in three melanoma subsets: NRASMUT melanoma sublines (n = 3) with acquired MEKi resistance (Fig. 1H; Supplementary Fig. S1H), BRAFV600MUT melanoma sublines (n = 4) with acquired resistance to type I RAFi + MEKi (Supplementary Fig. S1I and S1J), and NF1−/− melanoma sublines with acquired MEKi resistance (n = 4; Supplementary Fig. S1K and S1L). Using functionally equivalent doses of MEKi (trametinib) versus ERKi (SCH772984, BVD-523, and MK-8353; Fig. 1H; Supplementary Fig. S1C–S1G), we observed consistently (in 11 of 11 acquired MAPKi-resistant melanoma sublines) that BGB-283 + trametinib was more effective than BGB-283 + SCH772984 (or BVD-523, MK-8353) at overcoming the growth of established acquired-resistant clones.

NRASMUT Melanoma Acquires MEKi Resistance via ERK Reactivation

Given the durability of type II RAFi + MEKi in preventing and overcoming MEKi resistance, we tested the hypothesis that NRASMUT melanoma acquires MEKi resistance predominantly through ERK reactivation. We treated NRASMUT PDXs (n = 5) with trametinib (5 mg/kg/day) until acquired-resistant tumors (n = 16) emerged (Fig. 2A). The durability of response was highly variable across distinct and within each model(s). Together with aforementioned NRASMUT melanoma parental (P; n = 3) and isogenic, acquired trametinib-resistant cell lines (n = 5), we extracted genomic DNAs (gDNA) and total RNAs from vehicle-treated tumors and isogenic resistant (R) tumors for whole-exome sequencing (WES; along with patient-matched normal gDNAs) and RNA sequencing (RNA-seq). We then integrated WES/RNA-seq analysis to evaluate the recurrence of gain-of function (GOF) or loss-of-function (LOF) gene-based events of 723 cancer-related genes (COSMIC v.88; ref. 27; Supplementary Table S2). We rank-ordered the recurrence of somatic, resistance-associated alterations based primarily on sample counts and secondarily on patient frequencies (Supplementary Fig. S2A). Among the top 20 GOF genes were overlapping cMET and ERBB2 mRNA upregulation (≥2-fold). However, the most recurrent GOF genes at the genomic level were NRAS [7/21 samples with copy-number gain or mutant allele-specific loss-of-heterozygosity (LOH) events] and RAF1/CRAF (5/21; Supplementary Fig. S2A; Supplementary Tables S3 and S4). We have previously shown (4, 7, 13, 26, 28, 29) that BRAFV600MUT melanoma acquires resistance to type I RAFi + MEKi by omic alterations (e.g., BRAFV600E/K amplification–driven overexpression and RAF-regulated MEK1/2 mutants; refs. 7, 30) that enhance RAF/RAF and/or RAF/MEK interactions. Although RAF1/CRAF genomic alterations are not known to cause MAPKi resistance in BRAFV600MUT melanoma, we observed highly recurrent (and largely mutually exclusive) GOF alterations in NRAS, RAF1/CRAF, BRAF, and MAP2K1/2 (Fig. 2B and C). MAP2K2 harbored resistance-causative somatic mutations (F57L, V64F, and F133L; Supplementary Fig. S2B), as equivalent positions in MAP2K1 (F53, V60, and F129) have been shown to confer MAPKi resistance in BRAFV600MUT melanoma (7). Phylogenetically, resistant tumors or sublines are sometimes distantly related to common ancestral clones (Fig. 2D; Supplementary Fig. S2C), indicating that tumor heterogeneity (preexisting or induced by therapy) contributes to MEKi resistance.

To evaluate the functional roles of NRASMUT and RAF1/CRAF upregulation, we first overexpressed NRASMUT (vs. NRASWT) in the M245 P line to a level similar to NRASMUT upregulation observed in M245 SDR4 or SDR5 (Fig. 2E, left). We also knocked down NRAS overexpression in M245 SDR4 and SDR5 using two independent shRNAs (Fig. 2E, right). We observed in clonogenic assays that NRASMUT overexpression conferred resistance to trametinib relative to M245 P or M245 P overexpressing NRASWT (Fig. 2F, left). We also observed that NRAS knockdown in either M245 SDR4 or SDR5 sensitized these sublines to 0.1 μmol/L of trametinib (Fig. 2F, right). As we have reported (25), MEKi-resistant NRASMUT melanoma sublines are highly addicted to MEKi withdrawal. NRAS knockdown in the single-drug resistant (SDR) sublines abolished MEKi addiction, likely because NRASMUT overexpression drives pERK rebound and hence cell death after MEKi withdrawal. Moreover, M245 SDR3 upregulates CRAF expression; M245 SDR5, both CRAF and BRAF; and M207 SDR1, BRAF (Supplementary Tables S1, S3, and S4). Thus, we knocked down CRAF alone, both CRAF/BRAF, or BRAF alone in the respective SDR sublines using previously validated shRNAs (refs. 7, 29; Fig. 2G). Whereas P lines were highly sensitive to 0.1 μmol/L of trametinib and all three SDR sublines grew even in 1 μmol/L of trametinib, CRAF and/or BRAF knockdown in the SDR sublines conferred sensitivity to 0.1 μmol/L of trametinib in clonogenic assays (Fig. 2H).

Type II RAFi + MEKi Stabilize RAF/MEK and Uncouple MEK/ERK Complexes

We then sought to understand the mechanism underlying the durability of type II RAFi + MEKi in overcoming acquired MEKi resistance. Consistent with NRASMUT, BRAF, and CRAF alterations, we detected enhanced endogenous BRAF–CRAF interaction via coimmunoprecipitation (co-IP) in all SDR sublines (vs. isogenic P lines; Fig. 3A). M296 SDR3, in particular, harbors both subclonal MAP2K1 (MEK1F53L) and MAP2K2 (MEK2F133L) mutations, which disrupt negative regulation of the kinase domain by helix A and result in RAF-dependent ERK activation (30, 31). The RAF dependency of MEK1F53L and MEK2F133L may promote BRAF/CRAF complexes, analogous to allosteric BRAF activation by MEK binding to Kinase Suppressor of RAS (KSR; ref. 32). We then tested how type II RAFi dose-dependently affects growth, BRAF/CRAF interaction, and activation-associated phosphorylation of MEK and ERK in MEKi-treated/resistant cells (vs. MEKi-naïve/nontreated P cells). Type II RAFi (BGB-283, BGB-3245, RAF709, or LXH254) dose-dependently resensitized SDR sublines to MEKi (Fig. 3B; Supplementary Fig. S3A–S3D). Consistently, type II RAFi dose-dependently and sustainably reversed pERK levels or ERK reactivation in SDR sublines in the presence of trametinib (Fig. 3C and D), but, unexpectedly, did not reverse pMEK accumulation (Fig. 3D). Thus, type II RAFi cotreatment of MEKi-treated/resistant sublines led to pERK loss but curiously pMEK persistence.

Type II RAFi RAF709 as a single agent has been reported to induce RAF dimerization (19). In MEKi-naïve/nontreated P cells, type I RAFi (vemurafenib, 0.5 μmol/L) as a single agent induced both BRAF/CRAF interaction and pERK levels. In contrast, type II RAFi (BGB-283 or RAF709) as a single agent at 10 μmol/L induced equivalent levels of BRAF/CRAF interaction but suppressed pERK levels (Fig. 3E; Supplementary Fig. S3E). In MEKi-treated/resistant SDR cells, type I RAFi cotreatment also induced both BRAF/CRAF interaction and pERK levels. Importantly, type II RAFi cotreatment induced BRAF/CRAF interaction (on top of higher preexisting levels compared with P cells) at 0.1 μmol/L, leading to pERK suppression at submicromolar concentrations of type II RAFi. These findings suggest that the effective concentration of type II RAFi to induce signal-incompetent (pERK-suppressed) BRAF/CRAF complexes depends on MEKi cotreatment and/or upregulated levels of BRAF/CRAF complexes (both conditions present in SDR sublines compared with P lines). The upregulated levels of BRAF/CRAF complexes are apparently still capable of phosphorylating MEK (Fig. 3C) but somehow incapable of productively signaling to ERK.

We then investigated whether the high abundance of BRAF/CRAF complexes, which confers MEKi resistance (Fig. 2), also confers sensitivity of SDR sublines to type II RAFi cotreatment. In support of this hypothesis, individual or combined CRAF and/or BRAF knockdown abolished the efficacy of BGB-283 combination treatment (Fig. 3F). Using SDR sublines, we then investigated whether MEKi cotreatment enhances the ability of type II RAFi to induce signal-incompetent BRAF/CRAF complexes, resulting in pERK suppression. In this hypothetical model, type II RAFi + MEKi (i) foster BRAF/CRAF interaction (since BRAF and/or CRAF protein upregulation fosters the combinatorial efficacy of type II RAFi + MEKi; Fig. 3F); (ii) sequester or stabilize MEK/pMEK within the BRAF/CRAF scaffold (which promotes pMEK levels; Fig. 3C); (iii) retard pMEK release from BRAF/CRAF (which would in turn reduce MEK dimerization and thereby reduce the pool of active MEK accessible to ERK); and (iv) retard MEK/ERK binding (which is required for ERK activation). To test this model, we used MEKi-resistant NRASMUT melanoma sublines and measured the effects of single (type II RAFi or MEKi) versus double (type II RAFi + MEKi) inhibitor treatments on the in situ levels of protein/protein complexes within the MAPK pathway by proximity ligation assay (PLA). Consistent with prediction (i) of the model, acute (2-hour) treatment with type II RAFi + MEKi (BGB-283 + trametinib, LXH-254 + trametinib, or RAF-709 + binimetinib) upregulated BRAF/CRAF complexes compared with single-agent treatment (Supplementary Fig. S4A–S4C). After 12 days of treatments with inhibitor(s) (treatment refreshed every two days with additional last dose 12 hours prior to analysis), we observed that type II RAFi alone had little effect on CRAF/MEK or MEK/ERK levels. In contrast, type II RAFi + MEKi significantly induced CRAF/MEK and reduced MEK/ERK levels (Fig. 4A and B), consistent with predictions (ii) and (iv). Induction of CRAF/MEK and reduction of MEK/ERK were also observed as early as 2 hours after treatment with the aforementioned three type II RAFi + MEKi plus two additional combinations (BGB-283 + binimetinib, BGB-283 + cobimetinib; Supplementary Fig. S4A–S4D).

We had shown earlier (Fig. 1H; Supplementary Fig. S1C–S1G) that BGB-283 + trametinib was more effective than BGB-283 + SCH772984 (or BVD-523, MK-8353) at overcoming the growth of established acquired MEKi-resistant melanoma subclones. SCH772984, compared with trametinib, acutely (2 hours) and minimally induced CRAF/MEK and reduced MEK/ERK levels in conjunction with BGB-283 (Supplementary Fig. S4A and S4B). With prolonged (12 days) treatment(s), type II RAFi + ERKi (BGB-283 + SCH772984 or BGB-283 + BVD-523) failed to induce CRAF/MEK or reduce MEK/ERK levels (Fig. 4A and B). Consistently, after 12 days of treatment, BGB-283 cotreatment was able to reduce the p-p90RSK level only in trametinib- but not in SCH772984- or BVD-523–cultured NRASMUT melanoma sublines with acquired resistance (Fig. 4C). As observed previously, single-agent ERKi treatment led to rebound pERK levels (BVD-523 > SCH772984), which curiously failed to increase the growth rate (Supplementary Fig. S1C–S1G). BGB-283 cotreatment with BVD-523 did not reduce the rebounding pERK level and with SCH772984, which unlike BVD-253 inhibits ERK phosphorylation by MEK, did reduce the pERK rebound level but only to a level comparable to that with trametinib alone (Fig. 4C).

To validate our model further, we performed additional PLA assays in NRASMUT melanoma sublines with acquired MEKi resistance. Although BGB-283 + trametinb induced BRAF/CRAF complexes, this combination reduced BRAF or CRAF homo-complexes (Fig. 4D and E). BRAF/MEK complexes, just like CRAF/MEK complexes, were induced by BGB-283 + trametinib. Importantly, BGB-283 + trametinib reduced MEK1 homo-complexes, suggesting that RAF-phosphorylated/sequestered MEK could not be released from BRAF/CRAF scaffolds and thereby could not homodimerize and become activated. We also performed PLA assays in KRASMUT non-melanoma tumor cell lines. First, we derived sublines (HCT116-R, Su86.86-R, and H2122-R) from KRASMUT colorectal carcinoma, PDAC, and NSCLC cell lines (Fig. 1A) that had adapted to increasing doses of trametinib (up to 0.05 μmol/L) with proliferative resistance. In short-term (7 days) clonogenic assays, adding type II RAFi (BGB-283 or RAF-709 at 0.5 μmol/L) to MEKi (trametinib or binimetinb, respectively, at 0.05 μmol/L) strongly reduced clonogenic growth (Supplementary Fig. S4E). With 6 hours of treatment with inhibitor(s), we found that BGB-283 addition to trametinib (vs. BGB-283 alone) induced BRAF/CRAF and CRAF/MEK protein complex levels and reduced MEK/ERK and pERK levels (Supplementary Fig. S4F and S4G) in all three non-melanoma cancer cell lines with acquired MEKi resistance. To validate key mechanistic features of our model in vivo, we retransplanted the NRAS_PDX1 R2 (Fig. 2) acquired trametinib-resistant tumor into mice that were treated daily with trametinib. Trametinib-resistant tumors (≅ 500 mm3) were assigned into three groups: trametinib (5 mg/kg/day, orally), BGB-283 (20 mg/kg/day, orally), or trametinib (5 mg/kg/day, orally) plus BGB-283 (20 mg/kg/day, orally). Switching from trametinib to BGB-283 did not induce tumor regression, whereas the combination of trametinib and BGB-283 rapidly induced tumor regression (CRs in 2 of 5 mice or tumors; Fig. 4F). We then analyzed NRAS_PDX1 R2 tumors early (days 3 and 5) on each of the three treatments for CRAF/BRAF, CRAF/MEK, and MEK/ERK in situ complexes by PLA and for pERK levels by immunofluorescence (IF; Fig. 4G and H). Consistent with cell line observations, type II RAFi + MEKi induced BRAF/CRAF and CRAF/MEK levels concomitant with loss of MEK/ERK and pERK levels. To corroborate PLA findings, we performed co-IP and observed that type II RAFi + MEKi cotreatment preferentially (vs. type II RAFi + ERKi) induced endogenous CRAF/MEK and CRAF/pMEK complexes and reduced MEK/ERK and especially MEK/pERK complexes (Fig. 4I; Supplementary Fig. S4H and S4I).

To predict how RAF binding to BGB-283 and MEK binding to trametinib might enhance RAF/MEK complexes, we modeled the structure of the BRAF/MEK1 dimer bound to these inhibitors and calculated the change in buried solvent accessible surface area (SASA) as a result of dual inhibitor binding (see details in Methods). We superimposed the individual experimental or predicted structures of trametinib-bound MEK1 and BGB-283-bound BRAF (PDB ID 3PP1 and 4R5Y) on a dimer of the BRAF/MEK1 tetramer (PDB 4MNE; Supplementary Fig. S5). It is thought that BRAF and MEK1 form a face-to-face dimer (Supplementary Fig. S5A), with contribution to binding from the activation loop of both kinases (Supplementary Fig. S5B–S5F). When this model is compared with the BRAF and MEK1 conformations in the BRAF/MEK1 tetramer (Fig. 4J), the conformational changes of the activation loop of MEK1 upon trametinib binding and of the P-loop of BRAF upon BGB-283 binding could increase the contact surface between the two proteins and thereby the binding affinity. To estimate the effect of the latter conformational change, we calculated variation of the buried SASA upon binding of BRAF and MEK1, for both the apo conformation of BRAF (PDB ID 4MNE) and its BGB-283–bound conformation. Because the P-loop of BRAF was not resolved in the X-ray crystal structure of the apo form, we modeled it ab initio. For the residues of the BRAF P-loop (465–469) and for the MEK1 residues in the vicinity (residues 73–82 and 97–101), we calculated that the total buried SASA increased from 78 ± 5 Å2 for the apo form of BRAF to 98 Å2 for the BGB-283–bound BRAF, which suggests stronger binding.

Type II RAFi + MEKi Elicit CD8+ T Cell–Mediated Tumor Regression

Beyond tumor cell–intrinsic mechanisms, we evaluated the contribution of CD8 T cells to tumor regression. First, we introduced mutational burden into a syngeneic model of murine NrasQ61R melanoma (called NIL) we recently reported (25) by exposing it to radiation (UV), thereby deriving a subline called NILER1-4 and generating a relative increase of 25.3 mutations per megabase of gDNA. As subcutaneous allografts, NILER1-4 tumors displayed increased durability of MEKi (trametinib 3 mg/kg/day) response (vs. NIL tumors; starting volume at ∼100 mm3; Fig. 5A). However, systemic CD8+ T-cell neutralization abolished this gain in durability, while having no effect on MEKi-resistance development in NIL tumors. These findings suggest that CD8+ T cells suppress acquired MEKi resistance by recognizing neoantigen(s). Using larger (∼200 mm3) NILER tumors, we tested a non–tumor-regressive dose of trametinib (1 mg/kg/day, orally), two doses of BGB-283 (10 or 20 mg/kg/day, orally), and their combinations. Comparable to the studies using PDXs, BGB-283 at these doses did not induce tumor regression. Combination with trametinib at the higher dose (20 mg/kg/day) of BGB-283 induced durable tumor regression beyond 35 days and 7 of 7 CRs on day 52 (with 3 of 7 confirmed CRs after treatment withdrawal on day 52; Fig. 5B). The average weekly body weights of mice in all treatment groups increased commensurate with the Jackson Laboratory reference weights. However, NILER tumor–bearing mice treated with trametinib (1 mg/kg/day) + BGB-283 (20 mg/kg/day) began to lose weight after 40 days of continuous treatments.

We also tested trametinib and BGB-283 at these dosages in syngeneic tumor models of other cancer lineages. In KrasMUT colorectal carcinoma (CT-26) and PDAC (KPC) syngeneic models, neither trametinib (1 mg/kg/day, orally) nor BGB-283 (20 mg/kg/day, orally) alone induced regression of established (∼200 mm3) tumors, but the combination elicited durable tumor regression in both colorectal carcinoma and PDAC models, resulting in 6/8 and 2/8 CRs, respectively, on day 42. Body weight loss was noted only in mice treated with the combination, beginning around day 24 (Fig. 5C and D). To determine whether the addition of BGB-283 can overcome well-established acquired trametinib-resistant tumors (∼450 mm3), we began treating mice bearing ∼200 mm3 CT-26 tumors with higher doses (2 or 3 mg/kg/day, orally) of trametinib to achieve tumor volume stabilization. When tumors resumed growth and reached an average tumor volume of 450 mm3, we assigned mice into three groups: (i) continued trametinib at the prior dosage, (ii) BGB-283 (20 mg/kg/day, orally), or (iii) trametinib + BGB-283. Only in the third group did we observe tumor regression, which was associated with body-weight loss (Fig. 5E and F).

Furthermore, by reducing the trametinib dose 10-fold to 0.1 mg/kg/day (orally), we evaluated the superiority of type II RAFi + MEKi (vs. +ERKi) in vivo, as suggested by cell line studies (Fig. 1), and the contribution of CD8 T cells to tumor regression elicited by type II RAFi + MEKi. Using NILER tumor–bearing mice, we first identified trametinib at 0.1 mg/kg/day (orally) and SCH772984 at 25 mg/kg/day (intraperitoneally) as having similar suboptimal antitumor activities (Fig. 5G). We then compared the relative antitumor activities of trametinib + BGB-283 versus SCH772984 + BGB-283 (vs. trametinb + SCH772984). It was clear that only trametinib + BGB-283 has tumor-regressive activity, which was achieved with only 0.1 mg/kg/day (orally) of trametinib and associated with improved tolerability (Fig. 5H). Moreover, NILER tumor regression (trametinib 0.1 mg/kg/day + BGB-283 20 mg/kg/day, orally) or transient stabilization (trametinib 0.1 mg/kg/day + BGB-283 10 mg/kg/day, orally) was strongly dependent on CD8 T cells, as their systemic neutralization strongly undercut combinatorial efficacies (Fig. 5I). Similar observations were made in CT-26 and KPC models, although the contribution of CD8 T cells to combo-elicited tumor regression was less (Fig. 5C and D). Instead of neutralizing CD8 T cells (Fig. 5I), we invigorated CD8 T cells by treatment of NILER tumor–bearing mice with anti–PD-L1 (Fig. 5J). Anti–PD-L1 treatment by itself has minimal, if any, impact on vehicle-treated tumors. However, triplet treatments incorporating anti–PD-L1 improved the efficacy over that induced by doublet (trametinib + BGB-283) treatments, which is consistent with systemic and/or intratumoral CD8 T cells being an important effector of type II RAFi + MEKi in immune-competent hosts.

Type II RAFi + MEKi Expand T Naïve/T Central Memory and Shrink T Regulatory Cell Compartments Systemically

We evaluated NILER tumor histology and T-cell infiltration levels and patterns, on days 4 and 11, in response to treatment with vehicle, BGB-283 (20 mg/kg/day, orally), trametinib (1 mg/kg/day, orally), or both (Fig. 5B). Vehicle-treated tumors displayed sheets of spindled tumor cells, immature blood vessels, and scattered necrosis (Fig. 6A and B) as well as tumor-infiltrating lymphocytes (TIL; Supplementary Fig. S6A), consistent with findings of CD4+ and CD8+ immunohistochemistry (Fig. 6B). Tumors treated with BGB-283 or trametinib alone, which displayed only growth deceleration but not regression (Fig. 5B), showed small foci of tumor regression on days 4 and 11 and large clefts on day 11 and increased TILs in both the invasive margins and tumor centers (Fig. 6A; Supplementary Fig. S6A). In contrast, durable tumor regression induced by BGB-283 + trametinib was associated with a marked histologic response characterized by large foci of tumor regression and necrosis, tumor cell balloon degeneration, apoptosis, and melanosis (Fig. 6B). Maximal TIL scores were reached by day 4 in both tumor centers/margins, along with extensive infiltration by histiocytes and melanophages. In vehicle- or trametinib-treated tumors, the CD8+:CD4+ ratios were 1 or less. In BGB-283–treated tumors, this ratio was consistently 1. However, by day 11 on BGB-283 + trametinib treatment, this ratio was 2–3 (Supplementary Fig. S6A), consistent with CD8+ T-cell expansion.

To assess systemic T-cell impacts of non–tumor-regressive (single-agent) versus tumor-regressive (combination) versus vehicle treatments (Fig. 5B) in NILER tumor–bearing mice, we performed mass cytometry (CyTOF) on peripheral blood mononuclear cells (PBMC) and dissociated secondary lymphoid organs (tumor-draining lymph nodes and spleens) on days 4 and 11 (n = 3 mice/group). In all T-cell compartments, single and combined inhibitor treatment, short- and long-term, induced CD4+ and CD8+ T cells as percentages of CD45+ cells (Fig. 6C; Supplementary Fig. S6B). We analyzed further T-cell subpopulations by t-distribution stochastic neighbor embedding (t-SNE; Fig. 6D; Supplementary Fig. S6C and S6D). Short- and long-term, single and combined inhibitor treatments tended to induce the proportion of CD4+ and CD8+ T-naïve cells (TN; CD62L+CD44) and T central memory cells (TCM; CD62L+CD44+) but reduce CD8+ T effector memory cells (TEM; CD62LCD44+; Fig. 6E and F; Supplementary Fig. S6E–S6H). This pattern was more consistent on day 4 (during maximal tumor volume reduction by combination treatment) in response to type II RAFi + MEKi, but not to type II RAFi or MEKi alone. Consistently, only among mice cotreated with type II RAFi + MEKi, the proportions of CD8+ TCM and TEM cells were significantly anticorrelated (Fig. 6G; vehicle, R = −0.35, P = 0.14; type II RAFi, R = −0.43, P = 0.078; MEKi, R = −0.058, P = 0.82; type II RAFi + MEKi, R = −0.6, P = 0.0091). Moreover, combination type II RAFi + MEKi treatment reduced the levels of CD4+ T regulatory cells (TREG; CD4+FOXP3+) and their proliferation (Supplementary Fig. S6I and S6J). We also analyzed the profiles of splenic T cells (collected on day 4) after ex vivo cultures for 3 or 4 days with or without anti-CD3 + anti-CD28 stimulation. We found that CD4+ and CD8+ T cells alike from the spleens of mice that were treated in vivo with single-agent or combined inhibitors were more capable of activation (Supplementary Fig. S7A and S7B). Moreover, splenic CD4+ TREG cells from inhibitor-treated mice (especially type II RAFi + MEKi-treated mice on ex vivo day 3) were less capable of ex vivo expansion (Supplementary Fig. S7C). Ex vivo stimulation-induced T-cell proliferation as well as activation and/or effector marker expression was more robust among mice treated with inhibitors (Supplementary Fig. S7D and S7E). Consistent with prior analysis (Fig. 6), CD4+ and CD8+ T-cell compartments from the spleens of inhibitor-treated tumor-bearing mice, with and without ex vivo stimulation, expanded the TN and TCM subpopulations at the expense of the TEM subpopulation (Supplementary Fig. S7F).

Type II RAFi + MEKi Expand Intratumoral CD8+ TEM and Activated T Cells

To assess intratumoral T-cell impacts, we performed CyTOF on dissociated vehicle- versus type II RAFi + MEKi-treated tumors (n = 3 tumors/group) on day 5 and observed that combination treatment increased CD8+ T cells (% of CD45+ cells) by ∼7-fold, based on t-SNE analysis (Fig. 7A; Supplementary Fig. S8A). This finding indicates an even higher increase in the CD8+ T cell-to-tumor cell ratio, given induction of the CD45+ and reduction of the tumor cell compartments elicited by combination therapy. We analyzed further CD8+ T-cell subpopulations by t-SNE (Fig. 7B; Supplementary Fig. S8B). Importantly, type II RAFi + MEKi elevated the levels of intratumoral CD8+ TEM (∼3-fold), T cytotoxic (Tc; ∼3-fold), and T terminally differentiated (TTD; ∼5-fold) cells (% CD45+ cells; Fig. 7C). PD-1+ expression expanded from <0.5% (vehicle) to ∼5% and ∼7% (type II RAFi + MEKi) in, respectively, the CD8+ TEM and TTD subpopulations (Fig. 7D). This was accompanied by an increase in Ki-67 positivity among the CD8+ TTD cells. We corroborated these CyTOF findings with single-cell RNA-seq (scRNA-seq) analysis of combined CD4+ and CD8+ T cells sorted from vehicle- versus type II RAFi + MEKi–treated tumors on day 5 (sorted cells from four independent tumors were combined per condition). T-cell subpopulations visualized by t-SNE were analyzed for differential gene expression (Fig. 7E). Importantly, CD8+ T cells (as a percentage of all CD8+ T cells) that coexpress activation- and exhaustion-associated genes increased from ∼20% to ∼65% in vehicle- to type II RAFi + MEKi–treated tumors (Fig. 7F). An increase in the activated/exhausted CD8+ T cell-to-tumor cell ratio is expected to be even greater, given expansion in both the CD45+ and CD8+ T-cell compartments elicited by combination therapy. Consistently, type II RAFi + MEKi–treated tumors harbored CD8+ T cells with higher activation/exhaustion score and expression of the master regulator of exhaustion, Tox (Fig. 7G and H; refs. 33, 34).

We then evaluated whether intratumoral expansion of activated CD8+ T cells induced by type II RAFi + MEKi is clonal by performing T-cell receptor sequencing (TCR-seq) analysis of vehicle-, BGB-283–, trametinib-, or trametinib + BGB-283–treated NILER1-4 tumors (n = 3/group) on days 4 and 11. By analyzing the CDR3 clonotypes of both TCRα and β chains, we observed that trametinib + BGB-283 increased the number of T-cell clones (Fig. 7I). Moreover, whereas trametinib temporally reduced the sizes of the top or all dominant (>5%) T-cell clones, BGB-283 cotreatment erased and enhanced the sizes of top or dominant clones (Fig. 7J; Supplementary Fig. S8C and S8D). We also calculated the diversity and Gini (clonality) indices, which showed that trametinib increased but trametinib + BGB-283 reduced the diversity of T-cell clonotypes. Accordingly, trametinib reduced, whereas trametinib + BGB-283 maintained, T-cell clonality (Supplementary Fig. S8E and S8F). To evaluate the overlap of T-cell clones across treatment conditions and time points, we calculated the Jaccard indices and overlap coefficients (Fig. 7K; Supplementary Fig. S8G). Importantly, sample-to-sample comparisons among combination-treated tumors generally exhibited the highest percentages of intersecting clones (up to 8%). Group-to-group comparisons consistently showed the highest levels of overlap when combination-treated tumors as a group were compared against any other group, including itself. We observed within the combination treatment group the most significant positive correlation between d4 and d11 TCR clonotype frequencies (Fig. 7L; Supplementary Fig. S8H), suggesting that type II RAFi + MEKi elicits persistent expansion of tumor antigen–specific TCR clones. These findings suggest that type II RAFi + MEKi elicited the most robust tumor-specific T-cell reactivity, as trametinib + BGB-283 cotreatment increased TCR clonality as well as the number of distinct clones and converged T-cell clonotypes. Taken together with prior findings (Figs. 5 and 6), we conclude that type II RAFi + MEKi induces durable tumor regression by recruiting PD-L1 therapy–responsive, tumor antigen–specific CD8+ T cells.

MAPK-targeted therapy elicits clinically meaningful activity in only a handful of BRAFV600MUT cancers. From recent BRAFV600MUT-focused basket trials, it remains unknown whether lack of addiction to the MAPK pathway explains poor clinical efficacy in some BRAFV600MUT cancer histologies. In RASMUT or BRAFnon-V600MUT cancers, addiction to the MAPK pathway has also been cast into doubt by trials testing single-agent MEK or ERK inhibitors. Our study showing the highly durable efficacy of a specific pair of MAPK-targeted agents (type II RAFi + allosteric MEKi) across multiple driver mutations and cancer histologies (BRAFV600, BRAFnon-V600, KRAS, NRAS, and NF1-mutant melanoma and colorectal, pancreatic, and lung cancers) strongly supports MAPK pathway addiction.

The first combinatorial therapy developed successfully against the MAPK pathway in the clinic (for patients with advanced BRAFV600MUT melanoma) consists of type I RAFi plus an allosteric MEKi. However, despite high response rates, acquired resistance is commonplace and frequently due to MAPK pathway reactivation. Our data suggest that a specific combination of type II RAFi + MEKi could help overcome acquired resistance to the current standard of care (type I RAFi + MEKi).

Clinical development of MAPK-targeted combinatorial agents in oncology has centered on direct impacts on tumor cells (vs. immune cells) via inhibition of the activity of kinase(s). Insights from this study shine a light on the importance of indirect impacts on T cells and allosteric dysregulation of protein/protein interactions. This study also provides foundational knowledge for clinical development vis-à-vis identifying susceptible cancer histologies, biomarkers, and rational combinations with immunotherapies.

Previously, the allosteric action of dual MAPK inhibitors has not been examined at the level of protein/protein interactions in the MAPK pathway. Oncogenic RAS (which forms dimers) recruits cytosolic RAF/MEK heterodimers to the cell surface through the RAS-binding domain of RAFs. Face-to-face RAF/MEK heterodimers are then brought together by the back-to-back dimerization of RAFs, which facilitates cis-autophosphorylation of the RAF activation loop. Loosening of the RAF/MEK heterodimers is thought to facilitate the assembly of MEK homodimers on the RAF surface. Phosphorylation of both protomers of MEK permits release of MEK dimers from RAF dimers, MEK/ERK interaction, and ERK phosphorylation. Here we provided evidence that type II RAFi and allosteric MEKi act in concert to stabilize and sequester pMEK in RAF complexes (specifically BRAF/CRAF complexes, which is considered the most active complex among all RAF dimers). This action, which is likely facilitated by the high abundance of RAF/RAF and/or RAF/MEK complexes in MEKi-resistant tumor clones, may therefore be selectively robust against acquired resistance. The end result of MEK sequestration by RAF is reduced MEK dimerization and uncoupling of MEK or pMEK interaction with ERK.

ERKi has been proposed to overcome MEKi resistance (35). Because MEKi or ERKi each can strongly suppress the MAPK pathway in normal cells, untoward toxicities are likely to arise with their combination. Type II RAFi as a single agent does not appear to be highly active in suppressing oncogenic MAPK signaling, which is consistent with low single-agent antitumor activity from early-phase clinical trials. Here, we showed that type II RAFi + MEKi is superior to type II RAFi + ERKi (SCH772984) in preventing and overcoming acquired MEKi resistance. This correlates with the superior ability of type II RAFi + MEKi to physically stabilize RAF/MEK and uncouple MEK/ERK interactions. The synergy derived from this allosteric mechanism may afford MEKi dose reduction and improve the therapeutic index, permitting triplet combination with immune-checkpoint blockade therapy.

We also presented evidence that type II RAFi + MEKi may have favorable T-cell impacts that directly rationalize combination with anti–PD-1/L1 therapy. Prior studies (14, 16) have analyzed the T-cell impacts of MEKi monotherapy. Here, we observed that type II RAFi + MEKi induce systemic levels of CD8+ TCM cells and reduce CD4+ TREG cells. Studies have implicated the importance of CD8+ TCM cells to adoptive immunotherapy or anti–PD-1 therapy (36, 37). Tumor antigen–specific CD8+ TCM (vs. TEM) cells are thought to exhibit more potent in vivo antitumor (i.e., eradication of large established tumors) recall responses (38, 39). Consistent with this, homing to secondary lymphoid tissues (e.g., spleen, lymph nodes) appeared to be required for optimal tumor eradication. That is, highly effective antitumor T cells were those that initially targeted secondary lymphoid tissues rather than peripheral/tumor sites (38). Intratumorally, we observed that type II RAFi + MEKi induced not only sustained immune infiltration but also the relative abundance of CD8+ TEM and TTD cells, their expression of PD-1/Ki-67 expression, and activation/exhaustion genes and signatures. Inside the tumor, CD8+ TEM cells acquire effector functions more rapidly than TCM cells. This superior tumor cytotoxicity in situ is thought to be a key mechanism through which preexisting CD8+ TEM cells mediate secondary or de novo priming of effector T cells in the tumor-draining lymph nodes. Intratumorally, type II RAFi + MEKi also promoted expansion and convergence of T-cell clonotypes, in contrast to MEKi monotherapy. Collectively, these findings rationalize triplet combination trials of type II RAFi + MEKi + anti–PD-1/PD-L1 therapy in RAS/MAPK-hyperactivated cancers.

Cell Lines

All cell lines and drug-resistant sublines were routinely tested for Mycoplasma and profiled and identified by RNA-seq and the GenePrint 10 system (Promega) at periodic intervals during the course of this study for banking and experimental studies. All cell lines were maintained in DMEM-high glucose with 10% heat-inactivated FBS (Omega Scientific) and 2 mmol/L glutamine in humidified, 5% CO2 incubator. To derive resistant sublines, parental human BRAFV600MUT, NRASMUT, or NF1−/− melanoma cells seeded at low density were treated with BRAFi (PLX4032) + MEKi (AZD6244; for BRAFV600MUT lines) or MEKi (trametinib; for NRASMUT or NF1−/− lines) every 2 to 3 days for 12 to 15 weeks, and proliferative colonies were ring-isolated and expanded. To derive resistant polyclonal sublines of KRASMUT PDAC, NSCLC, and colorectal lines, cells seeded at low density were treated with MEKi (trametinib) every 2 to 3 days for 25 days and expanded. The NILER1-4 murine melanoma cell line was derived from NIL by exposure to 1 round of high-dose UVB radiation followed by ring-clonal selection and expansion.

Mice

C57BL/6 and NSG (NOD/SCID gamma) were obtained from the Radiation Oncology breeding colony at University of California, Los Angeles (UCLA). Male or female mice were used at 4 to 6 weeks of age. All animal experiments were conducted according to the guidelines approved by the UCLA Animal Research Committee.

Constructs and Inhibitors

shNRAS, shCRAF, and shBRAF were subcloned into the lentiviral vector pLL3.7 as described (4, 7, 29, 40). NRAS overexpression and all knockdown constructs were packaged into lentiviral particles for infection. Experiments were carried out 3 days after transduction. Inhibitors were obtained from the following sources: PLX4032 (Plexxikon), trametinib in vitro and in vivo (LC Laboratories), BGB-283 and BGB-3245 (Beigene), RAF-709 (Cayman Chemical), LXH-254 (Selleckchem), cobimetinib (Selleckchem), binimetinib (LC Laboratories), and SCH772984 (Chemietek).

Cell Growth Assays

For clonogenic assays, cells were plated at single-cell density in six-well plates. Data presented are representative of at least two independent replicates. Inhibitor(s) and media were replenished every 2 days for the number of days noted. Colonies were fixed in 4% paraformaldehyde (PFA) and stained with 0.05% crystal violet. For temporal measurements of cell growth, cells were plated at single-cell density in 96-well plates and treated with indicated treatments every 2 to 3 days for 8 to 11 days. Cell viability in relative light units was measured using CellTiter-Glo every 2 to 3 days. Experiments were performed in triplicates. For cell counting, cells were plated at single-cell density in six-well plates, and experiments were performed in triplicates. Indicated treatment and media were replenished every 2 to 3 days and, after trypsinization, viable (trypan blue negative) cells were counted on the days noted. Additional plates for the last time point (11 days) were fixed in 4% PFA and stained with 0.05% crystal violet.

Protein Detection

Cells were lysed in IP lysis buffer (Thermo Fisher Scientific) with Halt protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific) for IP and Western blotting. Dynabeads (Thermo Fisher Scientific) were used to immunoprecipitate proteins of interest based on the manufacturer's protocol. For immunofluorescence (IF), tissues were fixed in 4% PFA and sucrose and cryoprotected in optimal cutting temperature (OCT) or in formalin followed by embedding in paraffin [formalin-fixed, paraffin-embedded (FFPE)]. For FFPE tissues, after deparaffinization and rehydration, tissue sections were subjected to heat for antigen retrieval. PFA/OCT sections were not subjected to antigen retrieval. Cell lines were fixed with 4% PFA. IF of both tissue and cell lines was performed with Alexa Fluor–conjugated secondary antibodies (Life Technologies). Nuclei were counterstained by DAPI. Duolink PLA (Sigma-Aldrich) was used to detect in situ protein/protein proximity interactions by following the manufacturer's protocol. Fluorophore signals were captured with a Zeiss microscope (AXIO Imager A1) mounted with a charge-coupled device camera (Retiga EXi QImaging), and the images captured by Image-pro plus 6.0. PLA signals were quantified by counting signals per cell in each field (n = 5 fields per condition). IP, Western blots, IF, IHC, and PLA assays were performed using the following antibodies: CD4 (#183685 from Abcam), CD8 (14-0808-80 from Invitrogen/Thermo Fisher), BRAF (sc-5284 from Santa Cruz), CRAF (#53745, #12552 from Cell Signaling Technology), ERK (#9101, #4696 from Cell Signaling Technology), pERK1/2 (#5726, #4695 from Cell Signaling Technology), pMEK1/2 (#9154 from Cell Signaling Technology), MEK1/2 (#9126 from Cell Signaling Technology, sc-81504 from Santa Cruz), MEK1 (sc-6250 from Santa Cruz, #9146 from Cell Signaling Technology), p-90RSK (#11989 from Cell Signaling Technology), RSK1/2/3 (#9355 from Cell Signaling Technology), NRAS (sc-519, Santa Cruz), and TUBULIN (T9026 from Sigma-Aldrich).

PDX, Syngeneic Mouse Models, Treatments, and Tissue Collection

To develop PDX models, tumor fragments derived from NRASMUT metastatic melanoma, PDAC, or NSCLC, with approval by the local Institutional Review Boards, were transplanted subcutaneously in sex-matched NSG mice (4–6 weeks old). One tumor fragment was implanted in each mouse. Tumors were measured with a caliper every 2 days, and tumor volumes were calculated using the formula (length × width2)/2. Tumors with tumor volumes around 500 mm3 were randomly assigned into experimental groups. For syngeneic melanoma models, C57BL/6 mice were subcutaneously injected on both flanks with either one million NIL or NILER1-4 cells. Once tumors reached a size of 150 to 200 mm3, mice were assigned randomly into experimental groups. αCD8a and isotype control antibodies were administrated intraperitoneally (200 μg/mouse) on day −1, day 0, and then twice a week. Special mice diets (for NSG and C57BL/6) were generated by incorporating trametinib at 1, 3, or 5 mg/kg to facilitate daily drug dosing and to reduce animal stress (Test Diet). BGB-283 (10 or 20 mg/kg/day) was administered to mice via oral gavage and SCH772984 (10 or 25 mg/kg/day) intraperitoneally. Tumors were excised from mice, minced, and digested to single-cell suspensions using the tumor dissociation kit and gentleMACS Octo Dissociator (Miltenyi Biotech). Spleens were manually homogenized, mashed through 45-μm filters into RPMI-1640 supplemented with 10% FBS. Red blood cells in single-cell suspensions were lysed using ACK lysis buffer (Lonza). Intracardiac blood samples were collected in the presence of heparin, and PBMCs were obtained through density centrifugation (1,500 × g at 22°C for 30 minutes) using Lympholyte M (Cedarlane). Draining lymph nodes were collected, and single-cell suspensions were obtained using a spleen dissociation kit according to the manufacturer's protocol (Milteny Biotec).

Histologic Evaluation

Following tissue fixation, paraffin embedding, and sectioning, histologic evaluation was performed by a dermatopathologist (P.O. Scumpia) blinded to the identity of the samples. TIL scoring was performed using a slight modification of standard methodology (41, 42). After identifying the tumor invasive margin and tumor core, lymphocyte counts were performed from three high-power fields. TIL scores were defined as: 1 = 0%–10% of total cells; 2 = 20%–40% of total cells; 3 = 50%–70% of total cells; 4 = 80%–100% of total cells. These counts were confirmed using IHC for CD4 and CD8. The ratio of CD8:CD4 cells was estimated in three distinct tumor regions with the highest TIL infiltration by counting CD4+ and CD8+ cells in the same area.

Ex Vivo T-cell Activation

Splenocytes from NILER1-4 tumor–bearing mice with or without in vivo MAPKi treatment were collected and then seeded at 1 million cells/mL in 24-well plates, with or without anti-CD3 (145-2C11, 2 μg/mL) and anti-CD28 (37.51, 2 μg/mL). After 48 hours, media were refreshed using RPMI-1640 + 7.5% FBS + 0.1% β-mercaptoethanol+ 30 U/mL mIL-2, without antibodies. At 72 and 96 hours, cells were collected and analyzed by mass cytometry.

Mass Cytometry of Murine Tissues

Cells (2 × 106 or fewer) were incubated with 2% of FBS in PBS with 25 μg/mL of 2.4G2 antibody at 4°C for 10 minutes prior to surface staining with an antibody cocktail at 4°C for 30 minutes in a 50-μL volume. Cells were incubated with 2.5 μmol/L 194Pt monoisotopic cisplatin (Fluidigm) at 4°C for 1 minute. Cells were then washed twice with FACS buffer and barcoded using palladium metal barcoding reagents according to the manufacturer's protocol (Fluidigm). Subsequently, fixation and permeabilization were performed using the FOXP3 fix and permeabilization kit according to the manufacturer's protocol (eBioscience). Cells were then stained with an intracellular stain antibody cocktail (FOXP3, Ki-67, granzyme B, T-bet, iNOS, and EOMES) for 30 minutes at room temperature. Cells were then washed twice with FOXP3 permeabilization buffer, twice with FACS buffer, and incubated overnight in 1.6% PFA PBS with 100 nmol/L iridium nucleic acid intercalator (Fluidigm). Cells were then washed twice with PBS with 0.5% BSA, filtered, and washed twice with water with 0.1% BSA prior to analysis. Samples were analyzed using a Helios mass cytometer based on the Helios 6.5.358 acquisition software (Fluidigm).

CyTOF Data Analysis

All the samples were preprocessed by CATALYST, including normalization, debarcoding, and compensation. The normalized fcs files were then uploaded into Cytobank (43), and data were gated to exclude beads and to include only live, single cells. The CD8+ and CD4+ T cells were gated from the CD45+CD3+ populations, and data were downloaded separately into individual files for each sample. We applied Cytofkit (44) to perform the t-SNE analysis separately on the manually gated CD4+ and CD8+ populations from PBMC, spleen, and lymph node samples. We selected 5,000 events/sample to ensure equal representation of cells across samples. For CD4+ T cells, 12 markers, including CD44, CD62L, CD25, CD69, CD366, FOXP3, PD-1, CTLA4, ICOS, EOMES, T-bet, and Ki-67, were used to cluster the cell populations. For CD8+ T cells, CD44, CD62L, CD25, CD69, CD366, granzyme B, PD-1, CTLA4, ICOS, EOMES, T-bet, and Ki-67 were used. We chose 3,000 iterations, perplexity of 30 and theta of 0.5, as the standard t-SNE parameters. Mean intensity values of markers in each cluster were calculated and visualized via heat maps. Cells were assigned to different populations on the basis of the local gradient expression of known markers, e.g., CD44, CD62L, Granzyme B, and FOXP3. Numbers of cells and percentages of different immune cell subsets were calculated for each sample. Ex vivo cultures stained with a reduced antibody panel (CD45, CD4, CD8, CD44, CD62L, FOXP3 ICOS, T-bet, PD-1, CTLA4, and Ki-67) were analyzed using the FlowJo software.

WES and RNA-seq Data and Analysis

Thirty-six PDX tumors and cell lines and matched normal (or surrogate normal) tissue specimens were subjected to WES and RNA-seq. Sequencing was performed using paired-end sequencing with read length of 2 × 150 bps based on either the Illumina HiSeq3000 or the NovaSeq V4 platform. We called single-nucleotide variants (SNV) and small indels as we reported previously (7, 28). Mutations were annotated for coding/noncoding alterations using the stand-alone version of Oncotator (45). Copy numbers were called using the intersection of copy-number calls derived from Sequenza (46) and VarScan2 (47). Recurrent gene alteration events were visualized using the OncoPrint tool, and the MEKi resistance–specific mutations on the MAP2K2 gene were visualized using MutationMapper (48, 49). GOF alterations include known oncogenic missense mutations (from COSMIC v88), copy-number amplification, and/or mRNA overexpression (≥2-fold up). LOF alterations include truncating mutations (nonsense, splice site, frameshift), copy-number loss, and/or mRNA down-expression (≥2-fold down). Missense/in-frame indels of unknown significance were counted as both GOF and LOF events. CNV-related differential gene-expression events were defined as concurrent copy-number gain and mRNA overexpression (at least 1.5-fold of genomic copy-number gain and 2-fold mRNA overexpression). We applied the same cutoffs for copy-number loss (at least 1.5-fold of copy-number loss and 2-fold mRNA down-expression). Paired-end, 2 × 150 bp RNA-seq reads were mapped to the Genome Reference Consortium Human Build 38 (GRCh38) reference genome using HISAT2 (50). Gene-level counts were generated by the htseqcount (51) program, and we took log2 counts per million (CPM) as normalized gene-expression values. We added a pseudo CPM count of 0.1 to avoid taking the log of zero. The phylogenetic analysis was performed using the PHYLIP program as in our previous studies (7, 28).

For the syngeneic murine model NILER1-4, the paired-end reads were aligned to the mouse reference sequence (GRCm38) using the bwa-mem algorithm. The aligned reads were then processed using Picard tools and GATK (version 4.0) for deduplication and base quality recalibration prior to mutation detection. SNVs and indels were identified with MuTect2 using WT C57BL/6J mouse as the germline control. Variants were filtered by FilterMutectCalls using GATK default thresholds. Accumulated mutations in NILER1-4, presumably caused by UV irradiation, were detected as novel variants compared with the parental line NIL.

Generation and Analysis of scRNA-seq Data

To sort TILs, single-cell suspensions of digested tumor were stained for 5 minutes at room temperature with Fc block (anti-CD16/32) and then stained with primary antibodies in staining buffer (PBS with 20% FCS) for 30 minutes. Sorting of live TILs (TER119CD45+CD8+CD4 and TER119CD45+CD4+CD8) was performed on a BD FACSAria II (BD Biosciences). For each treatment condition, TILs from four different tumors were pooled together. Droplet-based 3′ end massively parallel scRNA-seq was performed by encapsulating sorted, live TILs into droplets, and libraries were prepared using Chromium Single Cell 3′ Reagent Kits v3 according to the manufacturer's protocol (10x Genomics). The generated scRNA-seq libraries were sequenced using Illumina NovaSeq.

Alignment to GRCm38 reference genome, barcode, and unique molecular identifier (UMI) counting were performed by using Cell Ranger v2.1.0 (10x Genomics). Seurat package (52) was used for downstream analysis. In brief, cells with fewer than 500 genes detected or greater than 10% mitochondrial RNA content were excluded from analysis. Raw UMI counts were normalized to UMI count per million total counts and log-transformed. Variable genes were detected based on average expression and dispersion for each data set independently. We then use CellCycleScoring function to calculate scores of S and G2–M cell-cycle phases for each cell. Single cells from vehicle- and trametinib + BGB-283–treated samples were integrated into a single assay based on variable genes identified from each sample. We then used the ScaleData function to calculate scaled z-scores of each variable gene in the integrated assay and regress out the effect of number of genes per cell, mitochondrial RNA content, and cell-cycle score differences (between S phase score and G2–M phase score). This scaled data set was then used for principal component analysis (PCA) for cells. Clusters and UMAP plots were generated based on top 30 PCA dimensions. Cell clusters expressing markers of both T cells (i.e., CD3d, CD3e, CD8a, and CD4) and other cell types (i.e., CD19, B cells; KLRCs1, natural killer cells; CSF1R, macrophages; FLT3, dendritic cells) were defined as doublets and excluded from further analysis. T-cell clusters were annotated by identifying differentially expressed marker genes with log-fold change higher than 0.4 using MAST in FindAllMarkers function. Finally, scores of T-cell terminal exhaustion were assigned to each cell by using Seurat's AddModuleScore function based on the gene sets previously reported (34).

Generation and Analysis of TCR-seq Data

Total RNAs were extracted from frozen tissues stored in RNALater using the QIAGEN All Prep DNA/RNA Mini Kit and the Ambion mirVana miRNA Isolation Kit. RNA (600 ng) was used as input to construct libraries with the QIAGEN QIAseq Immune Repertoire RNA Library Kit–T-cell Receptor Panel. Briefly, RNA was reverse-transcribed using a pool of TCR gene–specific primers against the constant region for the T-cell receptor alpha, beta, gamma, and delta genes. The resulting cDNA was then ligated to an oligo containing one side of sample index and UMI. After reaction cleanup, a single primer extension was used to capture the T-cell receptor using a pool of gene-specific primers. The resulting captured sequences were amplified and purified using QIAseq beads. Libraries were then sample indexed on the other side using a unique sample index primer and a universal primer to amplify the library and introduce platform-specific adapter sequences. The dual indexed sample PCR fragment was purified and then quantified for absolute quantification of amplifiable libraries (DNA with adaptors at both ends) in triplicate by real-time qPCR using the QIAGEN QIAseq Library Quant Array Kit. For sequencing, each library was diluted to 4 nmol/L, pooled, and denatured. Denatured library pool (1.2 pmol/L) was run with the QIAseq A Read1 Primer on Illumina NextSeq 500 Mid Output Kit using v2.5 chemistry for 300 cycles with an asymmetrical paired-end 261/41 bp read for the CDR3 region.

Raw reads were analyzed from the QIAGEN GeneGlobe Data Analysis Center (https://www.qiagen.com/us/shop/genes-and-pathways/data-analysis-center-overview-page/), which estimates the abundance of reads of unique CDR3 sequence and generate TCR clonotype calls. Briefly, raw reads were trimmed and randomly down-sampled to control the oversequencing error in UMI and CDR3 sequences. Paired R1 and R2 were then merged into one read with trimmed V regions. Clonotypes were called by IMSEQ (53), which clustered highly similar CDR3 sequences. CDR3 calls that did not have at least one UMI supported by three reads were excluded from downstream analysis. R package tcR (54) was used to perform all the statistical analyses for TCR repertoires, including (i) size of the largest clone, top 3 clones, and large clones with frequency higher than 5%; (ii) diversity estimation using ecological diversity and Gini–Simpson index; (iii) similarities of TCR repertories by calculating the Jaccard index and overlap coefficient between every pair of samples based on their unique alpha or beta chain CDR3 sequences. The Jaccard index was calculated using jaccard.index function in tcR package, and heat maps were generated using R pheatmap package. The comparison of overlap coefficients intra- and across treatment groups was performed using the Wilcoxon rank-sum test.

Molecular Modeling

SASAs were calculated using the analytical module for surface calculation of the CHARMM molecular modeling package (55), with water molecules described as spheres with a radius of 1.4 Å. The value of SASA buried upon the binding of MEK1 to BRAF was computed for the residues belonging to the BRAF P-loop (residues 465–469) and for MEK1 residues situated in the vicinity (i.e., residues 73–82 and 97–101) and by the difference in the SASA of these residues between the complex and the isolated MEK1 and BRAF proteins in the same conformation. The buried SASA was calculated for the apo and BGB-283–bound forms of BRAF to estimate its variation as a result of BGB-283 binding.

The conformation of the MEK1/BRAF heterodimeric complex with the apo form of BRAF was taken from the experimental 3-D structure of the BRAF/MEK1 heterotetramer complex, involving the apo form of BRAF and the G-573/ACP-bound MEK1 (PDB ID 4MNE; ref. 56). Because the P-loop residues were not resolved in this structure, they were modeled using the MODELLER program (57). For this, 5,000 conformations of the P-loop were modeled with DOPE-based loop modeling classes of MODELLER. The 50 top-ranked conformations according to MODELLER were used to calculate the average SASA buried upon the binding of MEK1 to BRAF.

The conformation of the MEK1/BRAF heterodimeric complex with the BGB-283–bound form of BRAF was obtained by superimposing the experimental structure of the complex between BGB-283 and BRAF (PDB ID 4R5Y; ref. 20) on the experimental structure of the MEK1/BRAF heterodimeric complex (PDB ID 4MNE) using UCSF Chimera (58).

Statistical Analysis

No statistical methods were used to predetermine sample size. The paired t test was performed to determine the statistical significance of differences between two variables. All statistical analyses were carried out using R and GraphPad Prism 7.

Data Availability

Raw sequencing files of RNA-seq, scRNA-seq, and TCR-seq data are available at the Gene Expression Omnibus (GEO158610). Mass cytometry data are deposited at FlowRepository (http://flowrepository.org/) using the experiment ID FR-FCM-Z34M. WES data have been made available through the Sequence Read Archive at the accession number PRJNA666070.

S.M. Dubinett reports grants from Johnson & Johnson, other from T-Cure Biosciences, other from Early Diagnosis, Inc, other from LungLife AI, and other from Johnson & Johnson Lung Cancer Initiative outside the submitted work. O. Michielin is an occasional consultant for BMS, Roche, Amgen, MSD, and GSK, and has received honoraria from BMS, Roche, Amgen, MSD, and GSK to participate in advisory boards and to speak at sponsored meetings A. Ribas reports personal fees from Agilent, Bristol-Myers Squibb, 4C Biomed, Apricity, Arcus, Highlight, Compugen, ImaginAb, MapKure, Merus, Rgenix, Lutris, PACT Pharma, Tango, Advaxis, CytomX, Five Prime, RAPT, Isoplexis, and Kite-Gilead, and grants from Agilent, and Bristol-Myers Squibb outside the submitted work. R.S. Lo reports grants from Merck, grants from Pfizer, grants from BMS, grants from OncoSec, personal fees from Array, personal fees from Amgen, and personal fees from Novartis outside the submitted work; and a Material Transfer Agreement with BeiGene, Inc. for supplying BGB compounds for this study. No disclosures were reported by the other authors.

A. Hong: Conceptualization, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. M. Piva: Conceptualization, resources, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. S. Liu: Data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–review and editing. W. Hugo: Formal analysis, investigation, visualization, and methodology. S.H. Lomeli: Resources, formal analysis, investigation, methodology, writing–review and editing. V. Zoete: Formal analysis, visualization, and methodology. C.E. Randolph: Data curation and methodology. Z. Yang: Resources, investigation, and methodology. Y. Wang: Resources, investigation, and methodology. J.J. Lee: Resources and investigation. S.J. Lo: Resources and investigation. L. Sun: Visualization. A. Vega-Crespo: Resources. A.J. Garcia: Methodology. D.B. Shackelford: Resources. S.M. Dubinett: Resources. P.O. Scumpia: Formal analysis, visualization, and methodology. S.D. Byrum: Methodology. A.J. Tackett: Methodology. T.R. Donahue: Resources. O. Michielin: Formal analysis, visualization, and methodology. S.L. Holmen: Resources, writing–review and editing. A. Ribas: Resources. G. Moriceau: Conceptualization, resources, data curation, formal analysis, supervision, validation, investigation, visualization, methodology, writing–review and editing. R.S. Lo: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing.

We thank Lusong Luo (BeiGene Inc.) for providing BGB-283 and BGB-3245 and Robert H. Vonderheide for providing the KPC syngeneic PDAC cell line. We would like to acknowledge the support of the NIH (1P01CA168585 and 1P01CA244118-01A1; to R.S. Lo and A. Ribas). This research was also supported by grants (to R.S. Lo) from the NIH (1R01CA176111A1 and 1R21CA215910-01), Melanoma Research Alliance (MRA; Team Science Award), and V Foundation for Cancer Research (Translational Award). Additional funding was provided by Department of Defense Horizon Award (to A. Hong), MRA Dermatology Fellows Award (to S. Liu), National Cancer Center Postdoctoral Fellowship (to Z. Yang), JCCC Postdoctoral Fellowships (to S. Liu and Z. Yang), JCCC Postdoctoral Seed Grant (to Z. Yang), Dermatology Foundation Career Development Award (to G. Moriceau), NIGMS P20GM121293 (to A.J. Tackett), Huntsman Cancer Foundation (to S.L. Holmen), and Steven C. Gordon Family Foundation (to R.S. Lo). We acknowledge use of the HCI Biorepository and Molecular Pathology (BMP) Shared Resource supported by P30CA042014 awarded to HCI from the NCI. We also thank the Huntsman Cancer Institute (HCI) Preclinical Research Resource for assistance with PDX models. Flow cytometry and CyTOF were performed in the UCLA Jonsson Comprehensive Cancer Center (JCCC) Flow Cytometry Core Facility that is supported by NIH award P30 CA016042, the JCCC, the David Geffen School of Medicine at UCLA, the UCLA Chancellor's Office, and the UCLA Vice Chancellor's Office of Research. We would like to thank the Technology Center for Genomics and Bioinformatics at UCLA for excellent technical support. R.S. Lo and A. Ribas are especially grateful to the Ressler Family Foundation for its long-term support. R.S. Lo dedicates this study to the memory of Waun Ki Hong, M.D.

1.
Samatar
AA
,
Poulikakos
PI
. 
Targeting RAS-ERK signalling in cancer: promises and challenges
.
Nat Rev Drug Discov
2014
;
13
:
928
42
.
2.
Dummer
R
,
Schadendorf
D
,
Ascierto
PA
,
Arance
A
,
Dutriaux
C
,
Di Giacomo
AM
, et al
Binimetinib versus dacarbazine in patients with advanced NRAS-mutant melanoma (NEMO): a multicentre, open-label, randomised, phase 3 trial
.
Lancet Oncol
2017
;
18
:
435
45
.
3.
Flaherty
KT
,
Robert
C
,
Hersey
P
,
Nathan
P
,
Garbe
C
,
Milhem
M
, et al
Improved survival with MEK inhibition in BRAF-mutated melanoma
.
N Engl J Med
2012
;
367
:
107
14
.
4.
Nazarian
R
,
Shi
H
,
Wang
Q
,
Kong
X
,
Koya
RC
,
Lee
H
, et al
Melanomas acquire resistance to B-RAF(V600E) inhibition by RTK or N-RAS upregulation
.
Nature
2010
;
468
:
973
7
.
5.
Su
F
,
Viros
A
,
Milagre
C
,
Trunzer
K
,
Bollag
G
,
Spleiss
O
, et al
RAS mutations in cutaneous squamous-cell carcinomas in patients treated with BRAF inhibitors
.
N Engl J Med
2012
;
366
:
207
15
.
6.
Long
GV
,
Fung
C
,
Menzies
AM
,
Pupo
GM
,
Carlino
MS
,
Hyman
J
, et al
Increased MAPK reactivation in early resistance to dabrafenib/trametinib combination therapy of BRAF-mutant metastatic melanoma
.
Nat Commun
2014
;
5
:
5694
.
7.
Moriceau
G
,
Hugo
W
,
Hong
A
,
Shi
H
,
Kong
X
,
Yu
CC
, et al
Tunable-combinatorial mechanisms of acquired resistance limit the efficacy of BRAF/MEK cotargeting but result in melanoma drug addiction
.
Cancer Cell
2015
;
27
:
240
56
.
8.
Wagle
N
,
Van Allen
EM
,
Treacy
DJ
,
Frederick
DT
,
Cooper
ZA
,
Taylor-Weiner
A
, et al
MAP kinase pathway alterations in BRAF-mutant melanoma patients with acquired resistance to combined RAF/MEK inhibition
.
Cancer Discov
2014
;
4
:
61
8
.
9.
Lamba
S
,
Russo
M
,
Sun
C
,
Lazzari
L
,
Cancelliere
C
,
Grernrum
W
, et al
RAF suppression synergizes with MEK inhibition in KRAS mutant cancer cells
.
Cell Rep
2014
;
8
:
1475
83
.
10.
Lito
P
,
Saborowski
A
,
Yue
J
,
Solomon
M
,
Joseph
E
,
Gadal
S
, et al
Disruption of CRAF-mediated MEK activation is required for effective MEK inhibition in KRAS mutant tumors
.
Cancer Cell
2014
;
25
:
697
710
.
11.
Ascierto
PA
,
Ferrucci
PF
,
Fisher
R
,
Del Vecchio
M
,
Atkinson
V
,
Schmidt
H
, et al
Dabrafenib, trametinib and pembrolizumab or placebo in BRAF-mutant melanoma
.
Nat Med
2019
;
25
:
941
6
.
12.
Ribas
A
,
Lawrence
D
,
Atkinson
V
,
Agarwal
S
,
Miller
WH
 Jr
,
Carlino
MS
, et al
Combined BRAF and MEK inhibition with PD-1 blockade immunotherapy in BRAF-mutant melanoma
.
Nat Med
2019
;
25
:
936
40
.
13.
Hugo
W
,
Shi
H
,
Sun
L
,
Piva
M
,
Song
C
,
Kong
X
, et al
Non-genomic and immune evolution of melanoma acquiring MAPKi resistance
.
Cell
2015
;
162
:
1271
85
.
14.
Ebert
PJ
,
Cheung
J
,
Yang
Y
,
McNamara
E
,
Hong
R
,
Moskalenko
M
, et al
MAP kinase inhibition promotes T cell and anti-tumor activity in combination with PD-L1 checkpoint blockade
.
Immunity
2016
;
44
:
609
21
.
15.
Dushyanthen
S
,
Teo
ZL
,
Caramia
F
,
Savas
P
,
Mintoff
CP
,
Virassamy
B
, et al
Agonist immunotherapy restores T cell function following MEK inhibition improving efficacy in breast cancer
.
Nat Commun
2017
;
8
:
606
.
16.
Choi
H
,
Deng
J
,
Li
S
,
Silk
T
,
Dong
L
,
Brea
EJ
, et al
Pulsatile MEK inhibition improves anti-tumor immunity and T cell function in murine kras mutant lung cancer
.
Cell Rep
2019
;
27
:
806
19
.
17.
Yao
Z
,
Gao
Y
,
Su
W
,
Yaeger
R
,
Tao
J
,
Na
N
, et al
RAF inhibitor PLX8394 selectively disrupts BRAF dimers and RAS-independent BRAF-mutant-driven signaling
.
Nat Med
2019
;
25
:
284
91
.
18.
Peng
SB
,
Henry
JR
,
Kaufman
MD
,
Lu
WP
,
Smith
BD
,
Vogeti
S
, et al
Inhibition of RAF isoforms and active dimers by LY3009120 leads to anti-tumor activities in RAS or BRAF mutant cancers
.
Cancer Cell
2015
;
28
:
384
98
.
19.
Shao
W
,
Mishina
YM
,
Feng
Y
,
Caponigro
G
,
Cooke
VG
,
Rivera
S
, et al
Antitumor properties of RAF709, a highly selective and potent inhibitor of RAF kinase dimers, in tumors driven by mutant RAS or BRAF
.
Cancer Res
2018
;
78
:
1537
48
.
20.
Tang
Z
,
Yuan
X
,
Du
R
,
Cheung
SH
,
Zhang
G
,
Wei
J
, et al
BGB-283, a novel RAF kinase and EGFR inhibitor, displays potent antitumor activity in BRAF-mutated colorectal cancers
.
Mol Cancer Ther
2015
;
14
:
2187
97
.
21.
Vakana
E
,
Pratt
S
,
Blosser
W
,
Dowless
M
,
Simpson
N
,
Yuan
XJ
, et al
LY3009120, a panRAF inhibitor, has significant anti-tumor activity in BRAF and KRAS mutant preclinical models of colorectal cancer
.
Oncotarget
2017
;
8
:
9251
66
.
22.
Ramurthy
S
,
Taft
BR
,
Aversa
RJ
,
Barsanti
PA
,
Burger
MT
,
Lou
Y
, et al
Design and discovery of N-(3-(2-(2-hydroxyethoxy)-6-morpholinopyridin-4-yl)-4-methylphenyl)-2-(trifluorom ethyl)isonicotinamide, a selective, efficacious, and well-tolerated RAF inhibitor targeting RAS mutant cancers: the path to the clinic
.
J Med Chem
2019
;
63
:
2013
27
.
23.
Sullivan
RJ
,
Hollebecque
A
,
Flaherty
KT
,
Shapiro
GI
,
Rodon Ahnert
J
,
Millward
MJ
, et al
A phase 1 study of LY3009120, a pan-RAF inhibitor, in patients with advanced or metastatic cancer
.
Mol Cancer Ther
2019
;
19
:
460
7
.
24.
Yen
I
,
Shanahan
F
,
Merchant
M
,
Orr
C
,
Hunsaker
T
,
Durk
M
, et al
Pharmacological induction of RAS-GTP confers RAF inhibitor sensitivity in KRAS mutant tumors
.
Cancer Cell
2018
;
34
:
611
25
.
25.
Hong
A
,
Moriceau
G
,
Sun
L
,
Lomeli
S
,
Piva
M
,
Damoiseaux
R
, et al
Exploiting drug addiction mechanisms to select against MPAKi-resistant melanoma
.
Cancer Discov
2018
;
8
:
1
20
.
26.
Song
C
,
Piva
M
,
Sun
L
,
Hong
A
,
Moriceau
G
,
Kong
X
, et al
Recurrent tumor cell-intrinsic and -extrinsic alterations during MAPKi-induced melanoma regression and early adaptation
.
Cancer Discov
2017
;
7
:
1248
65
.
27.
Tate
JG
,
Bamford
S
,
Jubb
HC
,
Sondka
Z
,
Beare
DM
,
Bindal
N
, et al
COSMIC: the catalogue of somatic mutations in cancer
.
Nucleic Acids Res
2019
;
47
:
D941
D7
.
28.
Shi
H
,
Hugo
W
,
Kong
X
,
Hong
A
,
Koya
RC
,
Moriceau
G
, et al
Acquired resistance and clonal evolution in melanoma during BRAF inhibitor therapy
.
Cancer Discov
2014
;
4
:
80
93
.
29.
Shi
H
,
Moriceau
G
,
Kong
X
,
Lee
MK
,
Lee
H
,
Koya
RC
, et al
Melanoma whole-exome sequencing identifies (V600E)B-RAF amplification-mediated acquired B-RAF inhibitor resistance
.
Nat Commun
2012
;
3
:
724
.
30.
Gao
Y
,
Chang
MT
,
McKay
D
,
Na
N
,
Zhou
B
,
Yaeger
R
, et al
Allele-specific mechanisms of activation of MEK1 mutants determine their properties
.
Cancer Discov
2018
;
8
:
648
61
.
31.
Yuan
J
,
Ng
WH
,
Tian
Z
,
Yap
J
,
Baccarini
M
,
Chen
Z
, et al
Activating mutations in MEK1 enhance homodimerization and promote tumorigenesis
.
Sci Signal
2018
;
11
:
eaar6795
.
32.
Lavoie
H
,
Sahmi
M
,
Maisonneuve
P
,
Marullo
SA
,
Thevakumaran
N
,
Jin
T
, et al
MEK drives BRAF activation through allosteric control of KSR proteins
.
Nature
2018
;
554
:
549
53
.
33.
Kim
K
,
Park
S
,
Park
SY
,
Kim
G
,
Park
SM
,
Cho
JW
, et al
Single-cell transcriptome analysis reveals TOX as a promoting factor for T cell exhaustion and a predictor for anti-PD-1 responses in human cancer
.
Genome Med
2020
;
12
:
22
.
34.
Yao
C
,
Sun
HW
,
Lacey
NE
,
Ji
Y
,
Moseman
EA
,
Shih
HY
, et al
Single-cell RNA-seq reveals TOX as a key regulator of CD8(+) T cell persistence in chronic infection
.
Nat Immunol
2019
;
20
:
890
901
.
35.
Hatzivassiliou
G
,
Liu
B
,
O'Brien
C
,
Spoerke
JM
,
Hoeflich
KP
,
Haverty
PM
, et al
ERK inhibition overcomes acquired resistance to MEK inhibitors
.
Mol Cancer Ther
2012
;
11
:
1143
54
.
36.
Takeuchi
Y
,
Tanemura
A
,
Tada
Y
,
Katayama
I
,
Kumanogoh
A
,
Nishikawa
H
. 
Clinical response to PD-1 blockade correlates with a sub-fraction of peripheral central memory CD4+ T cells in patients with malignant melanoma
.
Int Immunol
2018
;
30
:
13
22
.
37.
Wu
F
,
Zhang
W
,
Shao
H
,
Bo
H
,
Shen
H
,
Li
J
, et al
Human effector T cells derived from central memory cells rather than CD8(+)T cells modified by tumor-specific TCR gene transfer possess superior traits for adoptive immunotherapy
.
Cancer Lett
2013
;
339
:
195
207
.
38.
Klebanoff
CA
,
Gattinoni
L
,
Torabi-Parizi
P
,
Kerstann
K
,
Cardones
AR
,
Finkelstein
SE
, et al
Central memory self/tumor-reactive CD8+ T cells confer superior antitumor immunity compared with effector memory T cells
.
Proc Natl Acad Sci U S A
2005
;
102
:
9571
6
.
39.
Roberts
AD
,
Ely
KH
,
Woodland
DL
. 
Differential contributions of central and effector memory T cells to recall responses
.
J Exp Med
2005
;
202
:
123
33
.
40.
Shi
H
,
Moriceau
G
,
Kong
X
,
Koya
RC
,
Nazarian
R
,
Pupo
GM
, et al
Preexisting MEK1 exon 3 mutations in V600E/KBRAF melanomas do not confer resistance to BRAF inhibitors
.
Cancer Discov
2012
;
2
:
414
24
.
41.
Hendry
S
,
Salgado
R
,
Gevaert
T
,
Russell
PA
,
John
T
,
Thapa
B
, et al
Assessing tumor-infiltrating lymphocytes in solid tumors: a practical review for pathologists and proposal for a standardized method from the International Immuno-Oncology Biomarkers Working Group: part 2: TILs in melanoma, gastrointestinal tract carcinomas, non-small cell lung carcinoma and mesothelioma, endometrial and ovarian carcinomas, squamous cell carcinoma of the head and neck, genitourinary carcinomas, and primary brain tumors
.
Adv Anat Pathol
2017
;
24
:
311
35
.
42.
Saldanha
G
,
Flatman
K
,
Teo
KW
,
Bamford
M
. 
A novel numerical scoring system for melanoma tumor-infiltrating lymphocytes has better prognostic value than standard scoring
.
Am J Surg Pathol
2017
;
41
:
906
14
.
43.
Kotecha
N
,
Krutzik
PO
,
Irish
JM
. 
Web-based analysis and publication of flow cytometry experiments
.
Curr Protoc Cytom
2010
;
Chapter 10:Unit10 7
.
44.
Chen
H
,
Lau
MC
,
Wong
MT
,
Newell
EW
,
Poidinger
M
,
Chen
J
. 
Cytofkit: a bioconductor package for an integrated mass cytometry data analysis pipeline
.
PLoS Comput Biol
2016
;
12
:
e1005112
.
45.
Ramos
AH
,
Lichtenstein
L
,
Gupta
M
,
Lawrence
MS
,
Pugh
TJ
,
Saksena
G
, et al
Oncotator: cancer variant annotation tool
.
Hum Mutat
2015
;
36
:
E2423
9
.
46.
Favero
F
,
Joshi
T
,
Marquard
AM
,
Birkbak
NJ
,
Krzystanek
M
,
Li
Q
, et al
Sequenza: allele-specific copy number and mutation profiles from tumor sequencing data
.
Ann Oncol
2015
;
26
:
64
70
.
47.
Koboldt
DC
,
Zhang
Q
,
Larson
DE
,
Shen
D
,
McLellan
MD
,
Lin
L
, et al
VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome sequencing
.
Genome Res
2012
;
22
:
568
76
.
48.
Cerami
E
,
Gao
J
,
Dogrusoz
U
,
Gross
BE
,
Sumer
SO
,
Aksoy
BA
, et al
The cBio cancer genomics portal: an open platform for exploring multidimensional cancer genomics data
.
Cancer Discov
2012
;
2
:
401
4
.
49.
Gao
J
,
Aksoy
BA
,
Dogrusoz
U
,
Dresdner
G
,
Gross
B
,
Sumer
SO
, et al
Integrative analysis of complex cancer genomics and clinical profiles using the cBioPortal
.
Sci Signal
2013
;
6
:
pl1
.
50.
Kim
D
,
Langmead
B
,
Salzberg
SL
. 
HISAT: a fast spliced aligner with low memory requirements
.
Nat Methods
2015
;
12
:
357
60
.
51.
Anders
S
,
Pyl
PT
,
Huber
W
. 
HTSeq—a Python framework to work with high-throughput sequencing data
.
Bioinformatics
2015
;
31
:
166
9
.
52.
Butler
A
,
Hoffman
P
,
Smibert
P
,
Papalexi
E
,
Satija
R
. 
Integrating single-cell transcriptomic data across different conditions, technologies, and species
.
Nat Biotechnol
2018
;
36
:
411
20
.
53.
Kuchenbecker
L
,
Nienen
M
,
Hecht
J
,
Neumann
AU
,
Babel
N
,
Reinert
K
, et al
IMSEQ—a fast and error aware approach to immunogenetic sequence analysis
.
Bioinformatics
2015
;
31
:
2963
71
.
54.
Nazarov
VI
,
Pogorelyy
MV
,
Komech
EA
,
Zvyagin
IV
,
Bolotin
DA
,
Shugay
M
, et al
tcR: an R package for T cell receptor repertoire advanced data analysis
.
BMC Bioinformatics
2015
;
16
:
175
.
55.
Brooks
BR
,
Brooks
CL
 3rd
,
Mackerell
AD
 Jr
,
Nilsson
L
,
Petrella
RJ
,
Roux
B
, et al
CHARMM: the biomolecular simulation program
.
J Comput Chem
2009
;
30
:
1545
614
.
56.
Haling
JR
,
Sudhamsu
J
,
Yen
I
,
Sideris
S
,
Sandoval
W
,
Phung
W
, et al
Structure of the BRAF-MEK complex reveals a kinase activity independent role for BRAF in MAPK signaling
.
Cancer Cell
2014
;
26
:
402
13
.
57.
Webb
B
,
Sali
A
. 
Protein structure modeling with MODELLER
.
Methods Mol Biol
2017
;
1654
:
39
54
.
58.
Pettersen
EF
,
Goddard
TD
,
Huang
CC
,
Couch
GS
,
Greenblatt
DM
,
Meng
EC
, et al
UCSF Chimera—a visualization system for exploratory research and analysis
.
J Comput Chem
2004
;
25
:
1605
12
.