Genotype-selective combination therapies for melanoma identified by high-throughput drug screening

Abstract

Resistance and partial responses to targeted monotherapy are major obstacles in cancer treatment. Systematic approaches to identify efficacious drug combinations for cancer are not well established, especially in the context of genotype. To address this, we have tested pairwise combinations of an array of small-molecule inhibitors on early-passage melanoma cultures using combinatorial drug screening. Results reveal several inhibitor combinations effective for melanomas with activating RAS or BRAF mutations, including mutant BRAF melanomas with intrinsic or acquired resistance to vemurafenib. Inhibition of both EGF receptor and AKT sensitized treatment-resistant BRAF mutant melanoma cultures to vemurafenib. Melanomas with RAS mutations were more resistant to combination therapies relative to BRAF mutants, but were sensitive to combinations of statins and cyclin-dependent kinase inhibitors in vitro and in vivo. These results show the use of combinatorial drug screening for discovering unique treatment regimens that overcome resistance phenotypes of mutant BRAF- and RAS-driven melanomas.

Significance: We have used drug combinatorial screening to identify effective combinations for mutant BRAF melanomas, including those resistant to vemurafenib, and mutant RAS melanomas that are resistant to many therapies. Mechanisms governing the interactions of the drug combinations are proposed, and in vivo xenografts show the enhanced benefit and tolerability of a mutant RAS -selective combination, which is currently lacking in the clinic.

Fig. 1. Single Agent High-Throughput Screening

(A) Unsupervised clustering heatmap of efficacy, or maximum growth inhibition, for each single agent (columns) per melanoma cell line (rows) relative to DMSO controls. Cell lines named in red have BRAF mutations; green, RAS mutations; blue, wildtype RAS and BRAF (WT). (B) Second round of unsupervised clustering on drugs from Cluster 3 in (A) in order to survey for genotype-associated bias. Varying levels of efficacy are seen across melanomas. BRAF V600E-targeting drugs vemurafenib, PLX4720, and GDC0879, marked in red, were effective in many mutant BRAF melanomas, yet less effective in others (yellow box). Black boxes surrounding compound names are those with shared targets that grouped together when clustered. (C) Concentration-effect curves fitted to median growth inhibition values for vemurafenib in the mutant BRAF, mutant RAS, and WT genotypic groups. Bracket indicates the concentration range of significant differences in drug potency; p<0.0001, Kruskal-Wallis test. (D) Comparison of vemurafenib concentrations required to induce 50% growth inhibition (GI₅₀) in mutant BRAF melanoma lines. Red datapoints indicate cell lines with primary resistance to vemurafenib and extant PTEN. Green datapoints mark sensitive lines with PTEN homozygous deletion. (E) PTEN and RB1 protein status confirming no association with vemurafenib-resistance in these lines. (F) Concentration-effect curves of median growth inhibition values for simvastatin in the three genotypic groups. A trend in higher potency of the drug in mutant RAS lines is observed (2–10µM).

Fig. 2. Genotype-Specific Combinatorial Drug Sensitivities in Melanoma

(A) Unsupervised clustering of maximum growth inhibition achieved out of the nine concentration combinations tested for each cell line (columns) for each unique drug pair (rows). Partitioning of mutant BRAF cultures (Cluster 1) and mutant RAS or WT melanoma cultures (Cluster 2) is readily apparent. (B) Frequency of agents that in pairwise combinations elicit a GI₂₅ or higher effect level and ≥15% growth inhibition in the mutant BRAF group relative to the other two groups. (C) As in (B) but for mutant RAS melanomas.

Fig. 3. Efficacious Drug Combinations for RAS Mutant and Vemurafenib-Resistant BRAF Mutant Melanomas

(A) Drug interaction signatures for each cell line representing all combinatorial data compiled in a 40 by 40 drug matrix. Inset at bottom: Magnified view of the nine concentration combinations for a representative drug pair within its drug interaction matrix, to indicate scale. Yellow bar is the Bliss independent sum of growth inhibition for single agents; red indicates synergy, and green, antagonism (see also Fig. S3A-C). (B) Compilation of the total number of drug pair synergies, additivisms, and antagonisms. P-values calculated by Chi-square test. P<0.05 for BRAF versus RAS for number of synergies and RAS versus BRAF or WT for number of antagonisms. (C) Unsupervised hierarchical clustering of average synergy values obtained out of the nine concentration combinations tested for each drug pair (rows) for each melanoma line (columns), and (D) Average antagonisms of the nine combinations. Only drug pairs showing significant differences are shown. (E) Frequencies of drugs appearing in mutant BRAF-selective combinations yielding significantly higher relative average synergies. Lapatinib demonstrated the largest number of synergies specific to mutant BRAF melanomas. (F) Maximal percent growth inhibition of mutant BRAF lines for vemurafenib combined with other agents at maximal concentrations tested (see Table S5). YUKSI (red), the line most intrinsically resistant to vemurafenib, was also less sensitive to these combinations. The second-most vemurafenib-resistant line 501Mel (black) used in cHTS was also less sensitive to many combinations. The most vemurafenib-sensitive line YULAC (green) was most sensitive to combinations with vemurafenib in almost all cases. Yellow highlight marks vemurafenib combinations that demonstrated increased growth inhibition in the vemurafenib-resistant lines. (G) Drug combinations excluding vemurafenib, which showed highest selectivity towards mutant BRAF melanomas, were with lapatinib and AKT inhibitor MK-2206. YUKSI and 501Mel were also sensitive to these combinations. Bars indicate means. Indicated p-values calculated by Kruskal-Wallis ANOVA. Post-hoc Fishers t-test showed p<0.05 for the mutant BRAF group versus the other groups. (H) Drug combinations with the highest efficacy and selectivity towards RAS mutant melanomas included simvastatin with flavopiridol.

Fig. 4. Confirmation of Synergy and Cytotoxicity of Genotype-Selective Drug Pairs

(A) Formal assessment of synergy of lapatinib or bosutinib with MK-2206, and simvastatin with flavopiridol or 17-DMAG on representative BRAF and RAS mutant melanomas, by Chou-Talalay isobologram analysis. Data are normalized, with connecting line at X and Y = 1 corresponding to the line of additivity. Datapoints falling below line are synergistic, along or near the line are additive, and above the line are antagonistic (see Key). Data represent averages for three separate experiments. Combination indices and growth inhibition values can be found in Table S9. (B) Flow cytometry of Annexin-V and propidium iodide viability markers after lapatinib (2.5µM) and MK-2206 (5.0µM) treatment alone or in combination, or simvastatin (2.5µM) and flavopiridol (0.1µM) alone or in combination with representative lines YUMAC (BRAF mutant) or YUGASP (NRAS mutant). Numbers in the lower right quadrant correspond to early apoptotic cells, while numbers in the upper right and left quadrants correspond to late apoptotic or necrotic cells, respectively. Flow for vehicle-only controls and other combinations are shown in Fig. S3F. (C) Percent viability for combinations tested by flow cytometry, as in (B). N=3 for all experiments. Error bars, mean ± SD. P-values calculated by Student’s T-test, asterisks represent P<0.05. (D) Evaluation of drug-class effects. Representative concentration-response curves of single agents or dual-agent combinations demonstrating enhanced growth inhibition in mutant BRAF cells using the EGFR inhibitor gefitinib with MK-2206 (top row) or atorvastatin/Lipitor with flavopiridol in mutant RAS cells (bottom row) in relation to the predicted Bliss independence model. See also, Fig. S4. For statin combinations with flavopiridol, top and bottom x-axes represent concentrations used for each drug with those same concentrations used in combinations. Same concentration combinations were used for gefitinib and MK-2206.

Fig. 5. Combined Targeting of EGFR and AKT Re-Establishes Vulnerability to Vemurafenib

(A) Representative 2-D clonogenic (i), soft agar (ii), and flow cytometry (iii) assays on the YUKSI primary vemurafenib-resistant mutant BRAF line which also shows marked resistance to lapatinib and MK-2206 used alone or combined. D=DMSO vehicle; L=lapatinib (1.5µM); M=MK-2206 (1.5µM). (B) Reproduced experiments as described in (A) for primary resistant lines YUKSI and YUKOLI, as well as acquired resistant line YULAC-R. Normalized clonogenic fraction refers to two-dimensional clonogenic and soft agar assays, while normalized survival fraction refers to flow cytometry analyses. Lapatinib and MK-2206 both used at 1.5µM. (C) Immunoblotting to assess target engagement of EGFR and AKT by lapatinib and MK-2206, alone or combined. Increase of p-EGFR can be seen upon AKT inhibition by MK-2206. (D) 2-D clonogenic (i), soft agar (ii), and flow cytometry (iii) assays on YUKSI showing enhanced susceptibility to vemurafenib upon lapatinib and MK-2206 combination at 1.5µM each. V=vemurafenib (10µM); other designations same as in (A). (E) Reproduced experiments as in (D) for vemurafenib-resistant lines YUKSI, YUKOLI, and YULAC-R. (F) Molecular consequences of vemurafenib, lapatinib, and MK-2206 as single, dual, or triple-agent combination on downstream activity indicators of MAPK and PI3K pathway activity, including p-ERK and p-AKT. Primary vemurafenib-resistant YUKSI cells were treated with drugs for 24 hours before protein extraction. Concentrations of drugs in µM are shown in parentheses after drug letter designations. (G) p-ERK and p-AKT activity in the in vitro-selected vemurafenib-resistant YULAC-R line after 1 or 24 hour drug treatments. High levels of cleaved PARP could be detected clearly by 24 hours in the triple-agent combination in association with complete loss of p-ERK and p-AKT. (H) Long-term 2-D clonogenic assay on the parental YULAC line that was sensitive to vemurafenib. Cells were plated at 2×10³ cells/well and treated every other day with fresh drug for 2 weeks, followed by 4 weeks of recovery. Only the triple-agent combination completely prevented the emergence of colonies.

Fig. 6. Simvastatin Sensitizes Mutant NRAS Cells to Flavopiridol In Vitro and In Vivo

(A) Immunofluorescence on YUGASP mutant NRAS cells showing depletion of membrane-associated NRAS after simvastatin treatment at 5µM for 24 hours, or 1µM for 72 hours, with fresh drug added daily. (B) NRAS knockdown by siRNA nearly depletes p-ERK protein and attenuates active AKT. Addition of flavopiridol at 0.1µM during the final 48 hours of siRNA treatment completely abrogated p-ERK and resulted in increased levels of pro-apoptotic BIM and cleaved PARP relative to siRNA or flavopiridol treatment alone. EL, L, and S mark the three BIM isoforms; F=full length PARP, C=cleaved PARP. (C) Flow cytometry for viability confirmation in YUGASP mutant NRAS cells after NRAS knockdown with or without flavopiridol (FLAV) treatment at 0.1µM. SCR=scrambled siRNA control. (D) Western blot for p-ERK and p-AKT levels after simvastatin, flavopiridol, or both in mutant NRAS cells. The dual-agent combination reduced p-ERK and p-AKT levels more effectively than single agents. (E) YUGASP cells were treated with simvastatin at 1µM, flavopiridol at 0.1µM, or the combination. Viability (left panels) and clonogenicity (right panels) assessed by flow cytometry or 2-D clonogenic assay, respectively. (F) Tumor weights of xenografted tumors grown in immunocompromised mice and treated with simvastatin, flavopiridol, both, or mock treated with drug vehicle for two weeks. Kruskal-Wallis one-way ANOVA, p<0.0001; Asterisks over brackets indicate significant differences between two treatment groups, **p<0.001, ***p<0.0001, by Dunn’s multiple comparison test. (G) Tumor volume over time plot for those treatments as in (F). Asterisks over endpoint tumor volumes indicate p<0.001 by one-way ANOVA followed by Bonferroni’s multiple comparison test. All pairwise group comparisons demonstrated significant differences with the exception of mock treated versus simvastatin alone.