A melanocyte lineage program confers resistance to MAP kinase pathway inhibition

Abstract

Malignant melanomas harbouring point mutations (Val600Glu) in the serine/threonine-protein kinase BRAF (BRAF(V600E)) depend on RAF-MEK-ERK signalling for tumour cell growth. RAF and MEK inhibitors show remarkable clinical efficacy in BRAF(V600E) melanoma; however, resistance to these agents remains a formidable challenge. Global characterization of resistance mechanisms may inform the development of more effective therapeutic combinations. Here we carried out systematic gain-of-function resistance studies by expressing more than 15,500 genes individually in a BRAF(V600E) melanoma cell line treated with RAF, MEK, ERK or combined RAF-MEK inhibitors. These studies revealed a cyclic-AMP-dependent melanocytic signalling network not previously associated with drug resistance, including G-protein-coupled receptors, adenyl cyclase, protein kinase A and cAMP response element binding protein (CREB). Preliminary analysis of biopsies from BRAF(V600E) melanoma patients revealed that phosphorylated (active) CREB was suppressed by RAF-MEK inhibition but restored in relapsing tumours. Expression of transcription factors activated downstream of MAP kinase and cAMP pathways also conferred resistance, including c-FOS, NR4A1, NR4A2 and MITF. Combined treatment with MAPK-pathway and histone-deacetylase inhibitors suppressed MITF expression and cAMP-mediated resistance. Collectively, these data suggest that oncogenic dysregulation of a melanocyte lineage dependency can cause resistance to RAF-MEK-ERK inhibition, which may be overcome by combining signalling- and chromatin-directed therapeutics.

Figure 1. Near genome-scale functional rescue screens for resistance to RAF, MEK and ERK inhibitors

a, A375 cells transduced with the lentiviral expression library were treated in duplicate (technical) with indicated inhibitors and assayed for viability in the presence of compound alone (x-axis) and viability in compound relative to DMSO (y-axis). Values are presented as a z-score. Genes (n=169) with z-scores ≥ 2.5 (green dashed line) were nominated as candidate resistance genes. b, Summary of protein classes of candidate-genes identified in primary drug resistance screens. Only protein classes containing ≥ 2 genes are shown.

Figure 2. Candidate resistance genes segregate into validating protein classes

a, Area under the curve (AUC) was calculated for MAPK-i drug sensitivity curves in A375 expressing candidate and control genes. Data is presented as a z-score (y-axis), relative to the AUC of all control genes across each MAPK-i. b, Seven BRAF^V600E-malignant melanoma cell lines expressing the indicated candidate/control genes were assayed in technical duplicate for viability following treatment with indicated MAPK-i. Cellular viability is presented as a z-score relative to control genes. Genes with a z-score ≥4 in ≥2 conditions (drug or cell line) are shown. c, The strength of the resistance phenotype for each candidate and control gene across all MAPK-i and cell lines is quantified and presented as a composite rescue score.

Figure 3. A cyclic AMP signaling network mediates resistance to RAF, MEK and ERK inhibitors

a, Average fold change (relative to DMSO) in MAPK-i GI₅₀ or AUC in a panel of BRAF^V600-mutant cell lines treated with vehicle (DMSO), forskolin and IBMX (FSK/I) or dbcAMP and IBMX (cAMP/I). n=8 technical replicates, representative of 3 independent experiments. b, Heat map showing relative cell viability (percent of DMSO) following treatment with forskolin and IBMX (FSK/I) or dbcAMP and IBMX (cAMP/I) in the presence of vehicle (DMSO), the PKA inhibitor H89 and a single dose of indicated MAPK-i. c, Immunoblot analysis of phosphorylated CREB/ATF1(Ser133/Ser63) in lysates from WM266.4 virally transduced with the indicated expression constructs. d, Viability of WM266.4 expressing either LacZ (control) or dominant-negative CREB alleles (CREB^R301L or A-CREB) following treatment with forskolin and IBMX (FSK/I) in the presence indicated MAPK-i. Viability is expressed as a percentage of DMSO. Error bars represent s.d. of mean, n=6 technical replicates, representative of two independent experiments. e, Quantification of pCREB and pATF1 expression following immunoblot analysis of lysates extracted from BRAF^V600-mutant human tumors. Tumors were biopsied pre-initiation of treatment (P, n=5), following 10–14 days of MAPK-inhibitor treatment (on-treatment, O, n=6) or following relapse (R, n=7). MAPK-inhibitor therapy is noted. All available samples were tested and reported. Pre-treatment and on-treatment samples are paired. * p<0.05, 1-tailed T-Test on treatment cohorts, which may not directly inform responses in individual patient samples. f, Immunoblot analysis of lysates from WM266.4 following treatment with forskolin and IBMX (FSK/I) or dbcAMP and IBMX (cAMP/I) in the presence of indicated MAPK-i. Quantification of pCREB relative to CREB is shown.

Figure 4. Candidate resistance genes are transcriptional effectors of the MAPK and cAMP-pathways

a, Schematic outlining the identification of candidate resistance genes endogenously regulated by cAMP. b, Quantification of TBP-normalized mRNA levels using real-time quantitative PCR (relative to DMSO-treatment) following a time course of MEK-i treatment. Error bars represent s.d. of mean, n=3 technical replicates representative of 3 independent experiments. c, Quantification of TBP-normalized mRNA levels using real-time quantitative PCR (relative to DMSO-treatment) following treatment with forskolin and IBMX (FSK/I) for the indicated times in the presence of vehicle (DMSO) or MEK-i. Error bars represent s.d. of mean, n=3 technical replicates representative of 2 independent experiments. d, Immunoblot analysis of lysates from WM266.4 treated with DMSO or MEK-i followed by treatment with panobinostat, vorinostat or entinostat and subsequently stimulated with forskolin and IBMX (FSK/I). e, Cellular viability of WM266.4 treated with the indicated combinations of MAPK-i, HDAC-i and forskolin and IBMX (FSK/I). Cell viability is shown as a percent of DMSO in un-stimulated/non drug-treated cells. Error bars represent s.d. of mean, n=6 technical replicates representative of 2 independent experiments. f, A mechanistic model of GPCR-mediated resistance.

Extended Data Figure 1. A systematic, functional approach to identifying drug resistance genes

Schematic outlining the experimental approach taken to identify membrane-to-nucleus signaling pathways that mediate resistance to MAPK-pathway inhibitors. Resulting data were used to identify gene networks capable of mediating drug resistance.

Extended Data Figure 2. Near genome-scale ORF/cDNA screens identify candidate MAPK-pathway inhibitor resistance genes

a, Histogram of infection efficiency in A375 observed in the primary resistance screens. Percent of total ORFs above and below 65% infection efficiency are noted (red, dashed line). b, Histogram of the z-score of A375 viability in DMSO observed in the primary resistance screen. Total ORFs above, below and within the indicated z-score thresholds are noted. c, Scatter plots and correlation (R) of A375 viability (raw luminescence values) in the primary resistance screens. Colors distinguish viral screening plates. d, Heat map summary of controls and candidate resistance genes identified in primary resistance screens. Protein class and ORF class are indicated (positive control, red; negative control, yellow; experimental ORF, black). Asterisk (*) identifies two genes whose empirical sequence is significantly divergent from its annotated reference sequence.

Extended Data Figure 3. Patterns of drug resistance induced by candidate resistance genes

a, Heat map displaying the percent rescue (viability in drug/viability in DMSO) for each candidate resistance ORF and control ORFs in the presence of log-fold concentrations of the indicated MAPK-pathway inhibitor. These data were used to generate drug sensitivity curves, for which b, the area under the curve was calculated (red dashed lines denote significance thresholds). c, Heat map showing ERK phosphorylation data for all candidate resistance genes and controls in A375. d, Matrix of genes ectopically expressed in A375 (vertical axis) versus treatment condition (horizontal axis). Sensitivity is defined as yielding an area under the curve z-score of <1.96, resistance is defined as z >1.96 (p<0.05). e, Venn diagram showing the overlap of validated resistance genes, grouped by MAPK-pathway inhibitor, in A375. f, Schematic showing the number of validated genes that confer resistance or sensitivity to indicated MAPK-i.

Extended Data Figure 4. Broad validation of candidate resistance genes in a panel of BRAF^V600-mutant melanoma cell lines

a, Drug sensitivity curves for PLX4720, AZD6244 and VRT11E in the panel of 8 BRAF^V600E-mutant malignant melanoma cell lines used for the primary and validation screening experiments (described in Main Fig. 2). Error bars represent s.d. of mean, n=6 technical replicates. b, Western blot analysis following treatment with indicated MAPK-i in the panel of 8 BRAF^V600E-mutant malignant melanoma cell lines used in a. c, Box plot of all candidate and control ORF infection efficiencies in the panel of 8 cell lines used in the validation screening experiments. Center line represents the median value, box defines the 25^th−75^th percentile and whiskers define the 5–95% confidence interval. Outliers are shown as individual data points. d, Summary of the cellular viability (relative to DMSO) of negative and neutral control genes observed in validation screens. Bar graph shows the average viability (relative to that of DMSO treatment) of each cell line when expressing the 59 negative and neutral control genes included in all validation screening experiments. Error bars represent s.d. of mean, each measured in technical duplicates. e, Average composite rescue score of each class of proteins identified among the resistance candidates (relates to Main Fig. 2). Number of genes within each protein class is shown in parentheses. f, ADCY9 was identified as a resistance candidate in the primary resistance screen, but was a DNA failure in our independent prep of candidate virus. Therefore, ADCY9 was not included in the high throughput validation screens, but was included in all subsequent validation work. These data show that ADCY9 is able to confer resistance to all tested MAPK-i to a similar degree as forskolin/IBMX treatment. Error bars represent s.d. of mean, n=6 technical replicates. g, Western blot analysis of the expression of V5-epitope tagged eGFP and ADCY9 in WM266.4.

Extended Data Figure 5. Cyclic AMP induces CREB/ATF1 phosphorylation and induces MAPK-pathway inhibitor resistance

a, Mean fold-change in intracellular cAMP following treatment with forskolin + IBMX (FSK/I) or dbcAMP + IBMX (cAMP/I) using a competitive cAMP ELISA assay (n=2 technical replicates, representative of 2 independent experiments). b, Bar graphs showing the change in the half-maximal inhibitory concentration (GI₅₀) of BRAF^V600E-mutant cell lines treated with escalating doses of indicated MAPK-pathway inhibitor in the presence of vehicle (DMSO), forskolin/IBMX (FSK/IBMX) or dbcAMP/IBMX (cAMP/IBMX). c, Relative cell viability (percent of DMSO) following FSK/IBMX or cAMP/IBMX treatment in the absence of MAPK-pathway inhibitor treatment. Error bars represent s.d. of mean, n=8 technical replicates. Data is representative of 2 independent experiments. d, Number of viable cells treated with the indicated compounds in the presence of vehicle (DMSO) or forskolin and IBMX (FSK/I). Error bars represent s.d. of mean, n=3 technical replicates. e, Immunoblot analysis of WM983b following pre-treatment with the PKA inhibitor H89 and stimulation with forskolin and IBMX. f, Viability of WM266.4 treated with the indicated compounds and doses in the presence of vehicle (DMSO) or forskolin and IBMX (FSK/I). Error bars represent s.d. of mean, n=6 technical replicates.

Extended Data Figure 6. Candidate GPCR/PKA pathway genes induce cAMP and CREB/ATF1 phosphorylation

a, Western blot of BRAF^V600-mutant melanoma cell lines stimulated with forskolin/IBMX or cAMP/IBMX. b, Western blot analysis of WM266.4 treated with AZD6244, followed by stimulation with forskolin and IBMX (FSK/IBMX). c, Western blot analysis of 293T lysates transfected with indicated genes or stimulated with forskolin/IBMX. d, Quantification of immunoblot analyses of 293T transiently transfected with the indicated expression constructs, pre-treated with IBMX (arbitrary units, n=2 biological replicates). e, Mean control or candidate gene-induced cAMP production was measured following transfection of 293T with indicated expression constructs or treatment with forskolin and IBMX (FSK/I). cAMP levels were determined using an immuno-competition assay in the presence (red bars) or absence (black bars) of IBMX (n=2 technical replicates, data is representative of three independent experiments). The green dashed line represents levels of cAMP in negative controls (eGFP, Luciferase, LacZ). f, Western blot analysis of WM266.4 expressing indicated constructs and treated with AZD6244 and/or forskolin/IBMX (FSK/IBMX).

Extended Data Figure 7. CREB activity is regulated in the context of drug treatment in patient biopsies

a, Summary of patient sample characteristics. b, Immunoblot analysis of lysates extracted from BRAF^V600E-mutant human tumors biopsied pre-initiation of treatment (P), following 10–14 days of MAPK-inhibitor treatment (on-treatment, O) or following relapse (R). MAPK-inhibitor therapy is noted (vemurafenib, RAF inhibitor; dabrafenib, RAF inhibitor; tremetinib, MEK inhibitor). c, Comparison of quantified pCREB and pATF1 from b, shown as individual tumors. d, Statistical analysis of pATF1 and pCREB as in c, normalized to Pre-treatment levels. Samples analyzed are restricted to the subset of the biopsies that are patient matched, lesion-matched and treatment-paired *p < 0.0023, by 1-tailed T-test. e, Immunoblot analysis of WM266.4 following treatment with forskolin and IBMX (FSK/I) or dbcAMP and IBMX (cAMP/I) in the presence of vehicle (DMSO) or indicated MAPK-i.

Extended Data Figure 8. Identification of candidate resistance genes that are co-regulated by MAPK- and cAMP/PKA- signaling pathways

a, Candidate and neutral control genes containing cAMP response elements (CREs) were identified using gene sets extracted from MSigDB. Fold enrichment of the percent of CRE-containing genes in candidates over all genes screened for each gene set are noted. b, Matrix of CRE and candidate genes indicates the presence (black box) or absence (white box) of indicated CRE. Composite resistance score for each gene (summarized in Fig. 2c) is noted. Red dashed line indicates a composite resistance score of 50. c, Global endogenous mRNA expression (Log₂ RMA) of candidate and neutral control genes across a panel of melanoma cell lines. Red arrows identify the four genes hypothesized to be regulated by both the MAPK-pathway and the cAMP/PKA/CREB pathway in melanoma: MITF, FOS, NR4A1 and NR4A2. Asterisks identify the subset of cell lines used in for validation and primary screens.

Extended Data Figure 9. cAMP/PKA regulation of MITF mediates resistance to MAPK pathway inhibition

a, Immunoblot analysis and b, quantification in lysates from WM266.4 cells treated as indicated. Arrow indicates the slower migrating, phosphorylated form of MITF, error bars represent s.d. of mean, n=3 biological replicates. c, Western blot analysis of WM266.4 following treatment with AZD6244 and stimulated for the indicated times with forskolin/IBMX. Forskolin/IBMX was washed out of the cells and replenished with normal growth media. Cell lysates were collected at the indicated times. d, Immunoblot analysis of WM266.4 cells following treatment with forskolin/IBMX (FSK/I) for the indicated times in the presence of vehicle (DMSO) or MEK-i. Genes identified in resistance screens are underlined. e, Immunoblot analysis of a panel of BRAF^V600-mutant malignant melanoma cell lines following treatment with AZD6244 in the presence of vehicle (DMSO), forskolin/IBMX (FSK/I) or dbcAMP/IBMX (cAMP/I). f, Immunoblot analysis of WM266.4 cells following treatment with forskolin/IBMX (FSK/I) in the presence of vehicle (DMSO) or indicated MAPK-pathway inhibitor. g, Gene signatures for all candidates and controls were generated in A375 and compared to the signatures of cAMP-stimulating small molecules, including forskolin and its water-soluble derivative, colforsin. Individual genes are grouped as Candidates or Neutral controls, with each gene represented by a vertical line. Genes are ranked by similarity with colforsin, with #1 being the most similar. A subset of the most similar genes is noted. h, Immunoblot analysis of WM266.4 after viral expression of the indicated genes or treatment with forskolin/IBMX (FSK/I) in the presence of vehicle (DMSO) or AZD6244.

Extended Data Figure 10. Inhibition of PKA or MITF impairs cAMP-mediated resistance to MAPK pathway inhibitors

a, Cell viability of WM266.4 expressing a control shRNA (shLuciferase) or shRNAs targeting MITF treated with indicated MAPK-i and concomitant treatment with either DMSO or forskolin/IBMX (FSK/I). Error bars represent s.d. of mean, n=6 technical replicates, data is representative of two independent experiments. b, Western blot analysis of WM266.4 expressing the shRNA-constructs used in a. c, Western blot analysis of WM266.4 treated with AZD6244, followed by pre-treatment with DMSO or H89 and subsequent stimulation with forskolin/IBMX (FSK/I) for the indicated times. d, Immunoblot analysis of lysates extracted from human BRAF^V600E positive melanoma biopsies. Biopsies were obtained prior to treatment (P), on MAPK-inhibitor treatment for 10–14 days (on-treatment, O) or following relapse (R). e, Immunoblot analysis of WM266.4 treated with the indicated concentration of HDAC-inhibitor. f, Immunoblot analysis of SKMEL19 and SKMEL28 in the presence of vehicle (DMSO) or AZD6244, followed by treatment with the indicated HDAC-inhibitor (panobinostat; Pan, vorinostat; Vor) and subsequent stimulation with forskolin/IBMX (FSK/I). g, Drug sensitivity curves of Panobinostat and Vorinostat in WM266.4 expressing LacZ or MITFm. Error bars represent s.d. of mean, n=3 technical replicates.