ABL1/2 and DDR1 Drive MEKi Resistance in NRAS-Mutant Melanomas by Stabilizing RAF/MYC/ETS1 and Promoting RAF Homodimerization

Simple Summary NRAS-mutant melanoma is a highly aggressive subtype with few treatment options. Although both BRAF-mutant and NRAS-mutant melanomas have activation of the MEK/ERK pathway, MEK inhibitors (MEKi) are only effective for the BRAF-mutant subtype. The aim of this study was to understand why MEKi are ineffective in NRAS-mutant melanomas with the long-term goal of identifying new treatment regimens. Here, we show that ABL and DDR kinases are critically important for MEKi resistance because they cooperate to promote the stability of key proteins involved in driving melanoma growth and survival. FDA-approved drugs that inhibit ABL1/2 and DDR1 have been used for decades to treat leukemia. We showed that one such inhibitor prevents MEKi resistance from developing in a NRAS-mutant melanoma animal model. Thus, the data in this study provide the rationale for testing the use of drugs targeting ABL1/2 and DDR1 in combination with MEKi for patients with NRAS-mutant melanomas who have failed to respond to immunotherapy. Abstract Melanomas harboring NRAS mutations are a particularly aggressive and deadly subtype. If patients cannot tolerate or the melanomas are insensitive to immune checkpoint blockade, there are no effective 2nd-line treatment options. Drugs targeting the RAF/MEK/ERK pathway, which are used for BRAF-mutant melanomas, do little to increase progression-free survival (PFS). Here, using both loss-of-function and gain-of-function approaches, we show that ABL1/2 and DDR1 are critical nodes during NRAS-mutant melanoma intrinsic and acquired MEK inhibitor (MEKi) resistance. In some acquired resistance cells, ABL1/2 and DDR1 cooperate to stabilize RAF proteins, activate ERK cytoplasmic and nuclear signaling, repress p27/KIP1 expression, and drive RAF homodimerization. In contrast, other acquired resistance cells depend solely on ABL1/2 for their survival, and are sensitive to highly specific allosteric ABL1/2 inhibitors, which prevent β-catenin nuclear localization and destabilize MYC and ETS1 in an ERK-independent manner. Significantly, targeting ABL1/2 and DDR1 with an FDA-approved anti-leukemic drug, reverses intrinsic MEKi resistance, delays acquisition of acquired resistance, and doubles the survival time in a NRAS-mutant mouse model. These data indicate that repurposing FDA-approved drugs targeting ABL1/2 and DDR1 may be a novel and effective strategy for treating patients with treatment-refractory NRAS-driven melanomas.


Introduction
If caught early, most melanomas are cured (98% five-year survival rate); however, the five-year survival rate dramatically decreases (27%) for patients with distant disease (ACS Facts & Figures 2021). Compared with more common BRAF-mutant melanomas (50-70% of cases), the understudied NRAS-mutant subtype (15-25% of cases) is more aggressive, and patients have a lower median overall survival [1]. Low patient survival is due to the high melanoma mitotic index, thicker presentation at diagnosis, high incidence of distant metastases, and fewer treatment options [1]. First-line therapy for patients with metastatic NRAS-mutant melanomas is immune checkpoint blockade; however, only a subset of patients respond (30-60%) [2,3]. Unlike melanomas harboring BRAF mutations, which are highly sensitive to MAPK inhibition (BRAF and MEK inhibitors; BRAFi, MEKi), BRAFi treatment of melanomas harboring wild-type BRAF (including NRAS-mutant) activates MEK/ERK and increases melanoma growth [4]. This phenomenon, termed paradoxical ERK activation, is due to activation of CRAF as a result of heterodimerization with drugbound, wild-type BRAF [4]. NRAS-mutant melanomas are also resistant to MEKi and median progression-free survival (PFS) in clinical trials was limited to 2.8-3.25 months at best [5]. Thus, the PFS gain was too modest to warrant FDA approval [5]. Therefore, for patients who cannot tolerate immune-checkpoint blockade or whose melanomas are resistant to their effects, 2nd-line drugs are limited to cytotoxic agents, which are largely ineffective [1,4].
ABL1/2 (a.k.a. c-ABL, ARG) non-receptor tyrosine kinases are most known for their oncogenic roles in leukemia; however, accumulating evidence over the past two decades indicates that they also play critical roles in solid tumors including melanoma [6,7]. During melanoma progression, ABL1/2 drive proliferation, survival, invasion, and metastasis by phosphorylating substrates including the CRK/CRKL adapter proteins, which results in induction of transcription factors and proteases that degrade the extracellular matrix [6,8]. However, the contribution of ABL1/2 to MEKi resistance in NRAS-mutant melanomas represents a gap in our knowledge. FDA-approved drugs targeting ABL1/2 have been utilized for decades to treat leukemias driven by activated forms of ABL1 (e.g., BCR-ABL) [8]); therefore, repurposing these agents for treatment-refractory melanomas would be an attractive treatment strategy.
The DDR1 receptor tyrosine kinase, which is activated by binding fibrillar collagen I and network-forming collagen IV, regulates adhesion of melanocytes to the basement membrane [9][10][11][12]. Conflicting tumor-suppressing and tumor-promoting roles for DDR1 have been documented [13]. Although a number of FDA-approved drugs target both ABL1/2 and DDR1, due to the similarity of their kinase domains, the relationship between ABL1/2 and DDR1 activation has not been previously explored. Moreover, to date, the contribution of DDR1 to NRAS-driven melanoma growth and MEKi resistance has not been investigated. Here, we demonstrate that DDR1 contributes to ABL1/2 potentiation during MEKi resistance, and ABL1/2 and DDR1 activities are necessary and sufficient to drive MEKi resistance. Moreover, cotargeting ABL1/2 and DDR1 decreases BRAF/CRAF stability, increases RAF heterodimerization, reduces coupling of NRAS and ERK to BRAF/CRAF, prevents survival, induces apoptosis, reverses intrinsic resistance and delays the onset of acquired resistance, in vivo. These exciting data indicate that drugs targeting ABL1/2 and DDR1, many of which are FDA-approved, may be effective in conjunction with MEKi for patients with NRAS-mutant melanomas, a notoriously hard-to-treat class of patients.

Materials and Methods
Supplemental Methods are available online with this article.

Reagents
Details for reagents and software/algorithms are in Table S1.

Cell Lines
SK-MEL-2P/MR and SK-MEL-30P/MR cells were cultured in RPMI+glutamine, HEPES, and 10% FBS, while SK-MEL-147P/MR cells were grown in DMEM with the above additives. Cells were cultured at 37 • C and 5% CO 2 . Resistant lines were generated by culturing parental cells in increasing concentrations of trametinib up to 20 nM and were then maintained in trametinib (20 nM).

Viability Assays
Cells were plated in triplicate or quadruplicate in 96-well plates, drug-treated the following day, and harvested 72-96 h later using CellTiter Glo (see manufacturer's instructions, Table S1). Trametinib was removed from resistant cells two days prior to plating in order to be able to assess the effects of ABL/DDR inhibitors in the absence or presence of trametinib. To measure trypan blue exclusion, cells were plated in 60 mm dishes, drugtreated for 96 h, trypsinized, and the percentage of live cells were scored by diluting cells 1:1 in trypan blue (Biorad, Hercules, CA, USA) and counting on a TC-20 automated cell counter (Biorad, Hercules, CA, USA).
Since no toxicity was noted, nilotinib was increased to 50 mg/kg on d12. Tumors were measured three times weekly with linear calipers, and the volumes were calculated (LxW 2 /2). Mice were euthanized when tumors reached 1500 mm 3 or became ulcerated. The experiments were performed under IACUC protocol #2020-3426 (approval 6 November 2021) in accordance with University and NIH guidelines. These guidelines include: (a) using alternatives to animals, when possible (replacement); (b) using only the number of animals needed to achieve statistical significance (reduction); and (c) refining the experimental conditions to minimize pain and discomfort (refinement).

Statistics
In vitro studies were analyzed with one-way ANOVA (>two comparisons), two sample (comparisons between treatment groups), or one-sample t-tests (comparisons against normalized controls) using the Holm's method for multiple comparisons adjustment for one-and two-sample tests and Bonferroni multiple comparisons test for one-way ANOVA. In vivo studies were analyzed with the linear mixed model (SAS, v9.4) to compare logtransformed tumor volumes for trametinib and trametinib+nilotinib groups. Kaplan-Meier and logrank tests (R, v4.0.0) were used for comparing time to tumor doubling. In all cases, two-tailed p values were reported (p < 0.05 was considered statistically significant).

Targeting ABL1/2 Reverses Intrinsic and Acquired Trametinib Resistance
To examine whether ABL1/2 activation is required for intrinsic and acquired MEKi resistance, we treated cells with the second generation, ATP-competitive ABL1/2 inhibitor, nilotinib. On its own, nilotinib was modestly effective at reducing the viability of parental cells, but was less efficient in matched resistant SK-MEL-2MR and SK-MEL-30MR cell lines ( Figures 1D-F and S1A-C). In contrast, nilotinib alone was effective in reducing the viability of SK-MEL-147MR resistant cells ( Figures 1F and S1C). In combination with trametinib, nilotinib dramatically reduced survival, and reversed intrinsic (SK-MEL-147P), and acquired trametinib resistance (MR lines) as assessed using short-term viability ( Figures 1D-F and S1A-C) and long-term clonogenic ( Figure 1G) assays. Moreover, the loss of viability was permanent as illustrated by induction of apoptosis via caspase-3 and PARP cleavage blots in parental lines and in SK-MEL-2MR and SK-MEL-30MR ( Figure 1H). Interestingly, while nilotinib and nilotinib+trametinib dramatically reduced SK-MEL-147MR viability and percentage of live cells ( Figures 1F and S1D), we did not observe cleaved PARP or caspase-3 in combination-treated cells ( Figure 1H). These data indicate that the mechanism by which nilotinib induces cell death in this cell line does not involve apoptosis but rather is via another form of cell death. Indeed, combination treatment increased LC-3B II, an indicator for the presence of autophagosomes ( Figure 1H) [23].
Peeper and colleagues showed that some melanoma cells treated with BRAF inhibitors become "addicted" to the drugs, such that drug withdrawal induces cell death [24]. Interestingly, SK-MEL-30MR, but not the other resistant cell lines, appeared to be addicted to trametinib, as cells grew much better in trametinib as compared to vehicle ( Figure 1G). Interestingly, a similar effect was also observed with nilotinib indicating that nilotinib and trametinib likely impact the same pathway (e.g., ERK2) to drive drug addiction [24].

ABL1/2 and DDR1 Drive Acquired MEKi Resistance
To identify the nilotinib target(s) that drive acquired resistance, we first examined whether treatment with a highly specific, allosteric ABL1/2 inhibitor (GNF-5; exclusively targets ABL1/2) [8], phenocopies nilotinib's effects. If nilotinib reverses resistance solely by targeting ABL1/2, then GNF-5 should mimic the effects of nilotinib. SK-MEL-2MR and SK-MEL-30MR cells were relatively insensitive to GNF-5 on its own, similar to their response to nilotinib (Figure 2A). In the presence of trametinib, GNF-5 inhibited colony formation but only by ∼ =50%, indicating that it is much less efficient than nilotinib (95-98% reduction; Figure 2A). Similar effects were observed with a second highly specific, allosteric BCR-ABL inhibitor, ABL001/asciminib [21] ( Figure S1E). These data contrast with those obtained with SK-MEL-147MR cells as GNF-5 was extremely efficient (90-95%) at inhibiting colony formation in the presence of trametinib, mimicking nilotinib in this cell line ( Figure 2A). Moreover, GNF-5 was also highly effective in the absence of trametinib in SK-MEL-147MR cells (Figure 2A). Since the allosteric inhibitors are less sensitive than nilotinib, they require higher doses to inhibit ABL1/2; however, the doses utilized here are in-range with those used by other groups [17,18]. Taken together, these data indicate that nilotinib's effects in SK-MEL-2MR and SK-MEL-30MR are likely only partially mediated by ABL1/2 whereas in SK-MEL-147MR, nilotinib likely solely acts by targeting ABL1/2. sis pathways in resistant lines ( Figure S2). Thus, DDR1 is activated in SK-MEL-2MR and SK-MEL-30MR, in the absence of external collagen stimulation, likely due to collagen upregulation and secretion by the melanoma cells themselves. Moreover, activation of ABL1/2 and DDR1 is required for MEKi resistance. Consistent with these conclusions, DDR1 was activated following plating of parental cells on collagen I ( Figure 2F-bottom), which increased cell viability in the presence of trametinib, and cooperated with activated forms of ABL1/2 (PP) [25] to induce trametinib resistance (Figures 2F and S1F).  In order to identify the second nilotinib target involved in trametinib resistance in SK-MEL-2MR and SK-MEL-30MR, we treated the lines with another ATP-competitive ABL1/2 inhibitor (ponatinib) that has a different set of non-ABL targets. Interestingly, ponatinib's effects were identical to nilotinib ( Figure 2B), indicating that targets in common between nilotinib and ponatinib drive resistance. In addition to ABL1/2, nilotinib and ponatinib also inhibit KIT, CSF1R, PDGFR, and DDR1 [8]. To identify which of these four kinases is involved in MEKi resistance in SK-MEL-2MR and SK-MEL-30MR, we utilized specific inhibitors for each molecule. KIT/CSF1R inhibition (PLX3397) had no effect on cell viability, whereas PDGFR (CP673451) blockade reduced viability but did not cooperate with trametinib ( Figure 2C). In contrast, DDR1 inhibition (DDR-IN-1) reversed trametinib resistance and cooperated with GNF-5 (ABL1/2 inhibitor; Figure 2C,D), similar to nilotinib which targets both DDR1 and ABL1/2. Consistent with a role for DDR1 in driving resistance in SK-MEL-2MR and SK-MEL-30MR, DDR1 was activated in resistant cells ( Figure 2E). DDR1 is activated by binding collagen I and IV and stimulates collagen IV synthesis [9,10,12]. Indeed, RNA sequencing (RNA-seq) demonstrated that collagen I and collagen IV subunits were upregulated in acquired resistance cells (SK-MEL-2MR: COL1A1, COL1A2, COL4A2, COL4A5, COL4A6; SK-MEL-30MR: COL1A1, COL4A5; Dataset S1). Moreover, GSEA analysis identified upregulation of numerous collagen synthesis pathways in resistant lines ( Figure S2). Thus, DDR1 is activated in SK-MEL-2MR and SK-MEL-30MR, in the absence of external collagen stimulation, likely due to collagen upregulation and secretion by the melanoma cells themselves. Moreover, activation of ABL1/2 and DDR1 is required for MEKi resistance. Consistent with these conclusions, DDR1 was activated following plating of parental cells on collagen I ( Figure 2F-bottom), which increased cell viability in the presence of trametinib, and cooperated with activated forms of ABL1/2 (PP) [25] to induce trametinib resistance ( Figures 2F and S1F).

SK-MEL-147MR Are Exquisitely Dependent on ABL1/2
In contrast to SK-MEL-2MR and SK-MEL-30MR, which are insensitive to nilotinib or GNF-5 in the absence of trametinib, the survival of SK-MEL-147MR cells was permanently abrogated by GNF-5 and ABL001 in the absence or presence of trametinib (Figure 2Aright,G,H). Importantly, drugs targeting ABL1/2 (nilotinib, GNF-5, ABL001) also reduced the viability of SK-MEL-147P cells that develop trametinib resistance in vivo, demonstrating that the data are clinically relevant ( Figure 2I). To confirm these findings, we attempted to silence ABL1/2 with an shRNA that targets both proteins. Unfortunately, even when using an inducible system (IPTG), we were unable to stably silence ABL1/2 in SK-MEL-147MR as the clones either died or re-expressed ABL1/2. These data are consistent with the notion that SK-MEL-147MR cells are highly dependent on ABL1/2 as even low-level loss of ABL1/2 (leaky inducible system) was not compatible with survival. Importantly, expression of constitutively active forms of ABL1/2 (PP) into parental SK-MEL-147P cells was sufficient to increase survival and promote trametinib resistance, driving cells that had partial intrinsic resistance into an acquired resistance phenotype ( Figure 2J). Thus, acquired MEKi resistance in SK-MEL-147MR cells appears to be driven solely by ABL1/2 unlike SK-MEL-2MR and SK-MEL-30MR, which require cooperation of ABL1/2 and DDR1.

MEKi-Resistant Cells Utilize Diverse Mechanisms to Activate ERK/MYC/ETS1/RSK1 Signaling
To identify the mechanism of MEKi resistance, we first assessed activation of the ERK and AKT pathways, since NRAS activates both pathways in NRAS-mutant melanoma cells [3]. In parental cells, ERK activation and subsequent activation of ERK nuclear (FRA1, MYC) or cytoplasmic (ETS1, RSK1) downstream targets was efficiently inhibited by trametinib. In contrast, in all three acquired resistance cell lines, ERK/MYC/ETS1/RSK1 remained induced/activated in the presence of trametinib ( Figure 4A; compare lanes 1,5 with 2,6). AKT phosphorylation was upregulated in SK-MEL-30MR, whereas in the other two resistant lines, pAKT was decreased ( Figure 4A; lanes 1,2 and 5,6). Thus, ERK/MYC/ETS1/RSK1 activation is important for resistance in all lines, and PI3K/AKT may contribute to SK-MEL-30MR resistance.
BRAF and CRAF are both required to drive ERK activation in mutant NRAS-dependent melanomas [44]; however, the role of the various RAF proteins during resistance is unknown. ERK pathway activation requires both BRAF and CRAF in all three parental NRAS-mutant cell lines as well as in acquired resistant SK-MEL-2MR cells. However, SK-MEL-30MR cells shifted towards relying solely on BRAF, while SK-MEL-147MR shifted towards relying solely on CRAF ( Figure 5C). Interestingly, silencing ARAF either did not impact (SK-MEL-2MR, SK-MEL-147MR) or even increased ERK signaling (parental and SK-MEL-30MR; Figure 5D).
Wild-type RAF isoforms function as homo-and heterodimers, and BRAF/CRAF heterodimers drive ERK signaling downstream of mutant NRAS in MEKi-sensitive melanoma cells [44]. However, we found that RAF heterodimers were difficult to observe in trametinibtreated, acquired MEKi resistance cells (Figures 5E and S6L). Importantly, cotargeting ABL1/2 and DDR1 using nilotinib or GNF-5+DDR-IN-1 dramatically increased heterodimerization in the presence of trametinib ( Figures 5E,F and S6L). Nilotinib also reduced coupling of NRAS to CRAF and ERK to BRAF in both cell lines, and in SK-MEL-2MR, nilotinib reduced NRAS/ARAF binding ( Figures 5E,F and S6L). Thus, reversal of MEKi resistance induced by targeting ABL1/2 and DDR1, results in heterodimerization and degradation of RAF proteins, and reduces coupling of RAF to upstream regulators and downstream targets.   Figure S6L. Representatives of n = 2-3 independent experiments are shown. The uncropped blots are shown in File S1.

Allosteric ABL1/2 Inhibitors Prevent SK-MEL-147MR Survival by Inhibiting MYC and ETS1 Expression and Blocking β-Catenin Nuclear Localization
SK-MEL-147MR are extremely sensitive to highly specific, allosteric ABL inhibitors ( Figure 2G-I) in the absence of trametinib, even though they are less efficient than nilotinib at inhibiting ABL1/2 activity (pCRKL blots; Figure 6A vs. Figure 4A). The reduced efficiency of GNF-5 and ABL001 may be due to strong compensatory upregulation of ABL1 and/or ABL2 protein ( Figure 6A). Interestingly, on their own, GNF-5 and ABL001 had little to no effect on pERK in resistant cells, but effectively inhibited activation/nuclear localization of the transcription factors ETS1, MYC, and β-catenin ( Figure 6A,B, compare lane 4 to 2 and lane 12 to 10). Importantly, expression of exogenous MYC, ETS1, or β-catenin rescued GNF-5/ABL001-mediated reduction in survival ( Figure 6C), indicating that all three transcription factors are required for ABL1/2-driven survival. MYC is transcriptionally regulated by ETS1 and β-catenin, in addition to ERK [7,45]. However, exogenous expression of ETS1 or β-catenin only partially rescued the inhibitory effects of ABL001/GNF-5 on MYC ( Figure S7A), indicating that ABL1/2-mediated regulation of ETS1 and β-catenin is only partially responsible for increased MYC expression. Importantly, GNF-5/ABL001 also decreased the stability of MYC and ETS1, and GNF-5 destabilized β-catenin ( Figure 6D). In contrast, neither drug affected cyclin D stability. Thus, ABL1/2 are critically important for the survival of SK-MEL-147MR cells due to their ability to induce ETS1/β-catenin-mediated induction of MYC and stabilize MYC and ETS1 proteins.
In contrast to their effects on their own, in the presence of trametinib, GNF-5 and ABL001 efficiently inhibited activation of ERK and downstream nuclear and cytoplasmic targets (FRA1/ETS1/RSK1/MYC) ( Figure 6A, compare lane 8 to 6 and 16 to 14), similar to the effects of nilotinib ( Figure 4A). Moreover, expression of constitutively active forms of ABL1/2 in SK-MEL-147P parental cells was sufficient to drive ERK pathway activation in the presence of trametinib, whereas silencing ABL1/2 in resistant SK-MEL-147MR cells reduced pERK/MYC/pETS1/pRSK1 in the presence of trametinib ( Figure S7B,C). Thus, in SK-MEL-147MR cells, allosteric ABL inhibitors act on their own to prevent survival via ERK-independent effects on MYC, ETS1, and β-catenin, whereas ABL inhibitors reverse trametinib resistance via an ERK-dependent mechanism.
Since ABL1/2 activation of MYC is involved in trametinib resistance in all three resistant cell lines, we assessed whether ABL1/2 and MYC mRNA levels correlate in patient samples. We examined RNA-seq data in clinical trial samples from patients harboring NRAS-mutant melanomas who had been treated with a MEKi [46]. Interestingly, ABL2 mRNA correlated with MYC mRNA expression in this dataset ( Figure 6E), indicating that our data have clinical relevance.

Targeting ABL1/2 and DDR1 Reverses Intrinsic/Adaptive Resistance by Preventing Activation of Cytoplasmic but Not Nuclear ERK Targets
In addition to reversing acquired resistance, nilotinib also efficiently reversed partial intrinsic resistance in SK-MEL-147P parental cells ( Figure 1F,G). Interestingly, although trametinib efficiently inhibited MEK/ERK activation at 24 h, MEK/ERK became reactivated at 48-72 h, indicating that the "intrinsic" resistance is actually short-term adaptive resistance ( Figure 7A, compare lanes 3 and 7) [47]. Importantly, treatment with nilotinib efficiently blocked MEK/ERK reactivation at the 48 h timepoint ( Figure 7A, lanes 7 and 8), and induced apoptosis ( Figure 1H, 48 h timepoint is shown). Furthermore, at 72 h, combinationtreated cells were completely dead and could not be blotted. Interestingly, MEK/ERK reactivation at the 48 h timepoint resulted in activation of cytoplasmic (ETS1, RSK1) but not nuclear (FRA1, MYC) ERK downstream targets, which are subsequently inhibited by the addition of nilotinib ( Figure 7A). Thus, nilotinib reverses cytoplasmic and nuclear MEK/ERK signaling during acquired resistance but specifically suppresses cytoplasmic ERK signaling to reverse adaptive resistance. In addition to reversing acquired resistance, nilotinib also efficiently reversed partial intrinsic resistance in SK-MEL-147P parental cells ( Figure 1F,G). Interestingly, although trametinib efficiently inhibited MEK/ERK activation at 24 h, MEK/ERK became reactivated at 48-72 h, indicating that the "intrinsic" resistance is actually short-term adaptive resistance ( Figure 7A, compare lanes 3 and 7) [47]. Importantly, treatment with nilotinib efficiently blocked MEK/ERK reactivation at the 48 h timepoint ( Figure 7A, lanes 7 and 8), and induced apoptosis ( Figure 1H, 48 h timepoint is shown). Furthermore, at 72 h, combination-treated cells were completely dead and could not be blotted. Interestingly, MEK/ERK reactivation at the 48 h timepoint resulted in activation of cytoplasmic (ETS1, RSK1) but not nuclear (FRA1, MYC) ERK downstream targets, which are subsequently inhibited by the addition of nilotinib ( Figure 7A). Thus, nilotinib reverses cytoplasmic and nuclear MEK/ERK signaling during acquired resistance but specifically suppresses cytoplasmic ERK signaling to reverse adaptive resistance.  Figure S8B. The uncropped blots are shown in File S1.

Nilotinib Reverses Intrinsic MEKi Resistance, Delays the Onset of Acquired Resistance, and Prolongs Survival, In Vivo
In patients, MEKi are ineffective due to a combination of intrinsic/adaptive and acquired resistance. To mimic this situation, in vivo, we established SK-MEL-147 parental xenografts. Due to intrinsic/adaptive resistance, tumors were initially relatively resistant to trametinib alone, maintaining a stable size for 20-25 days at which time they acquired complete resistance, and their growth accelerated (Figures 7B and S8A). In contrast, treatment upfront with nilotinib in combination with trametinib reversed intrinsic resistance, induced early tumor regression, delayed the onset of acquired resistance, and doubled the survival time ( Figure 7B,C). Overall, animal weights were not significantly different between trametinib vs. combination-treated mice over the course of the experiment (e.g., p = 0.2 on d35; Figure S8B), suggesting that the combination is well-tolerated. In summary, these exciting data indicate that targeting ABL1/2 and DDR1 in combination with MEKi may be an effective therapeutic regimen for patients with aggressive NRAS-driven melanomas, who have a poor prognosis and limited therapeutic options.

Discussion
NRAS-mutant melanomas are a highly aggressive subtype and there is an unmet need to identify second-line treatment options for patients who fail to respond to immunecheckpoint blockade. In this manuscript, we describe a potential new therapy for patients with this subtype by characterizing the mechanism by which NRAS-mutant melanoma cells resist the effects of MEKi via short-term adaptive/intrinsic resistance and long-term acquired resistance. In some models, long-term acquired MEKi resistance was driven by activation of ABL1/2 and DDR1, which cooperated to drive resistance. This conclusion is based, in part, on loss-of-function studies in acquired resistant cell lines using siRNA and pharmacological approaches which demonstrated that ABL1/2 and DDR1 are both required to drive MEKi resistance. Moreover, we also utilized complementary gain-of-function studies to show that activation of DDR1, induced by plating parental cells on collagen I, together with expression of activated forms of ABL1/2, was sufficient to drive MEKi resistance. Furthermore, we identified the mechanism by demonstrating that ABL1/2 and DDR1 cooperated to activate ERK cytoplasmic and nuclear signaling (MYC/ETS1/RSK1), stabilize RAF and MYC proteins, and promote RAF homodimerization (Figure 8). In contrast, other long-term MEKi acquired resistance cells were highly dependent solely on ABL1/2, which stabilized MYC and ETS1 proteins in an ERK-independent manner, and induced β-catenin nuclear localization to drive melanoma survival ( Figure 8). Finally, we showed that intrinsic/adaptive resistance occurs via a different mechanism as it is driven by reactivation of cytoplasmic but not nuclear ERK targets, which was reversed by the ABL1/2 and DDR1 inhibitor, nilotinib.
Interestingly, we found that CRAF and ERK as well as DDR1 contributed to ABL1/2 activation in NRAS-mutant cells that acquired long-term MEKi resistance. While receptor tyrosine kinases such as EGFR, PDGFR, and IGF1R have previously been implicated in ABL1/2 activation in other cell contexts [28,[48][49][50], this is the first demonstration that DDR1 contributes to ABL1/2 activity. In contrast, DDR1 is likely activated by mRNA upregulation of collagen subunits in resistant cells. Indeed, plating parental cells on collagen activated DDR1 which cooperated with activated forms of ABL1/2 to drive resistance.
Using next generation sequencing, we identified new acquired mutations in GNAI1 (G125E; SK-MEL-30MR) and EPHA4 (N140fs, SK-MEL-2MR), RAF overexpression, and ABL1/2 and DDR1 as critical nodes that stabilized RAF and MYC proteins and destabilized p27/KIP1 while also upregulating MYC mRNA and repressing p27/KIP1 mRNA. ABL1/2 and DDR1 inhibition with either nilotinib or GNF-5+DDR-IN-1 (in the presence of trametinib) increased RAF heterodimerization in resistant lines, and reduced RAF binding to upstream regulators (NRAS), and downstream targets (ERK). Dimerization is required for RAS activation of wild-type RAF, and binding of 14-3-3 to RAF proteins not only retains RAF in an inactive conformation, but also is required for RAF dimerization and activation [51]. Heterodimerization is also enhanced by 14-3-3 proteins whereas ERK phosphorylation of BRAF-T753 promotes heterodimer disassembly [52]. ABL1 and DDR1 bind 14-3-3 proteins in other cell contexts [53,54], and here, we showed that ERK phosphorylation of ABL2 created a putative 14-3-3 binding site. Thus, ABL1/2 and DDR1 may impact heterodimerization and assembly of the NRAS/RAF/MEK/ERK complex by influencing RAF/14-3-3 scaffold interactions. Interestingly, peptides/drugs that bind the RAF dimerization motif not only prevent RAF dimerization but also induce RAF degradation [51], indicating that the effects of ABL1/2 and DDR1 inhibition on RAF heterodimerization and degradation may be linked.   Marais and colleagues demonstrated that some ABL1/2 and DDR1 inhibitors weakly bind RAF (following 3 h treatment), and induce RAF heterodimerization and paradoxical MEK/ERK activation in chronic myelogenous leukemia cells harboring a drug-resistant form of BCR-ABL (T315I) and tumor cells with activated RAS [55]. In contrast, while we showed that nilotinib treatment (24 h) induced RAF heterodimerization in MEKiresistant NRAS-mutant melanoma cells, the result is inhibition of MEK/ERK and decreased coupling of NRAS and/or ERK to RAF rather than MEK/ERK activation. Moreover, since nilotinib's effects are phenocopied by GNF-5+DDR-IN-1, the effects of nilotinib on RAF heterodimerization are likely not drug-dependent but rather are due to ABL1/2 and DDR1 inhibition.
Unlike BRAF and CRAF, silencing ARAF had little impact or even increased pERK activation (SK-MEL-30MR), and nilotinib increased ARAF mRNA expression in SK-MEL-30MR. ARAF was also absent from BRAF or CRAF complexes in SK-MEL-30MR cells unless nilotinib was present ( Figure S6A, left). These data are consistent with published reports indicating a potential tumor suppressive role for ARAF in some cell contexts [56,57].
In contrast to the other two acquired MEKi-resistant cell lines, SK-MEL-147MR cells gained a MEK/ERK pathway mutation (MEK1; F129L). Although previously identified as an activating mutation that drives BRAFi resistance, this mutation was insufficient to drive MEK1/2 activation in SK-MEL-147MR. Phosphorylation of MEK1/2 by upstream RAF kinases and other MAP3Ks can reduce trametinib binding and cooperate with low transforming MEK mutations [42]. Indeed, we also identified a mutation in MAP3K19, a kinase that phosphorylates MAP2Ks including MEK1/2 in SK-MEL-147MR cells (Table S2 and Dataset S2) [58].
Our exciting in vivo data have uncovered a potential new treatment for NRAS-driven melanomas, an understudied and highly aggressive subtype with notoriously poor outcomes. ABL1/2 and DDR1 inhibition in combination with MEKi not only reversed established and intrinsic resistance in vivo, but also significantly delayed acquired resistance from developing in the first place, and substantially increased survival. Since a number of ABL1/2 and DDR1 inhibitors are FDA-approved for treating leukemia, these data may pave the way for testing their clinical efficacy in combination with MEK for treatment-refractory NRAS-driven melanomas.

Conclusions
NRAS-mutant metastatic melanoma is an aggressive and deadly disease with no effective second-line treatments. In this manuscript, we demonstrated that in some resistant lines, ABL1/2 cooperate with DDR1 to drive MEKi acquired resistance by inducing activation of ERK/MYC/ETS1/RSK1, stabilizing RAF and MYC, and promoting RAF homodimerization ( Figure 8). In other resistant lines, MEKi-resistant cells were highly dependent solely on ABL1/2, which promoted survival by stabilizing MYC and ETS1, in an ERK-independent manner, and inducing β-catenin nuclear translocation ( Figure 8). Finally, ABL1/2 and DDR1 are also critical nodes during intrinsic resistance and are required for adaptive reactivation of cytoplasmic ERK targets. Significantly, using a mouse model that mimics intrinsic and acquired MEKi resistance, targeting ABL1/2 and DDR1, with an FDA-approved drug, reversed intrinsic resistance and delayed the onset of acquired MEKi resistance in vivo, thereby enhancing survival. Thus, repurposing inhibitors targeting ABL1/2 and DDR1, many of which are FDA-approved, may be an effective therapeutic approach, in combination with MEKi, for patients with metastatic NRAS-mutant melanomas that fail to respond to immune-checkpoint blockade and thus, have no effective second-line treatment options.