Half of advanced human melanomas are driven by mutant BRAF and dependent on MAPK signaling. Interestingly, the results of three independent genetic screens highlight a dependency of BRAF-mutant melanoma cell lines on BRAF and ERK2, but not ERK1. ERK2 is expressed higher in melanoma compared with other cancer types and higher than ERK1 within melanoma. However, ERK1 and ERK2 are similarly required in primary human melanocytes transformed with mutant BRAF and are expressed at a similar, lower amount compared with established cancer cell lines. ERK1 can compensate for ERK2 loss as seen by expression of ERK1 rescuing the proliferation arrest mediated by ERK2 loss (both by shRNA or inhibition by an ERK inhibitor). ERK2 knockdown, as opposed to ERK1 knockdown, led to more robust suppression of MAPK signaling as seen by RNA-sequencing, qRT-PCR, and Western blot analysis. In addition, treatment with MAPK pathway inhibitors led to gene expression changes that closely resembled those seen upon knockdown of ERK2 but not ERK1. Together, these data demonstrate that ERK2 drives BRAF-mutant melanoma gene expression and proliferation as a function of its higher expression compared with ERK1. Selective inhibition of ERK2 for the treatment of melanomas may spare the toxicity associated with pan-ERK inhibition in normal tissues.

Implications:

BRAF-mutant melanomas overexpress and depend on ERK2 but not ERK1, suggesting that ERK2-selective inhibition may be toxicity sparing.

This article is featured in Highlights of This Issue, p. 933

The MAPK pathway consisting of the RAF–MEK–ERK kinase cascade is dysregulated in a large portion of human cancers resulting in uncontrolled cell growth, proliferation, and resistance to apoptosis (1). Half of advanced melanomas are driven by activating mutations in the kinase BRAF, most commonly a substitution at Val600 (2). Targeting the MAPK pathway is a clinically validated strategy for the treatment of BRAFV600-mutant (BRAFV600mut) cancers including melanoma, as inhibitors of RAF and MEK extend survival in these patients and ERK inhibitors are now in clinical trials (3, 4).

The ERK1 and ERK2 kinases share 84% amino acid identity and all known phosphorylation substrates. However, distinct requirements for one kinase or the other have been reported in several cellular and organismal contexts (5). In some of these cases, expression differences underlie differential dependency on a particular ERK isoform, while in other cases, notably those involving kinase-independent ERK functions, ERK1 and ERK2 are functionally distinct (6, 7). Thus, the degree to which ERK1 and ERK2 are redundant or unique depends greatly on a number of factors including cell type.

In addition to the kinase domain, ERK1 and ERK2 interact with substrates, scaffolds, and regulators through two distinct docking domains: the D-recruitment site (DRS) and the F-recruitment site (FRS) located on opposite faces of the protein (8). The DRS interacts with a hydrophobic-rich motif, termed the D-site motif, of the form (R/K)2–3-X1–6-φ-X-φ, where φ is a hydrophobic amino acid residue. The FRS binds to an F-site motif of the form FXFP. Binding to one or both docking domains is required for some ERK substrates, and interference with protein-protein interactions at either docking site has been explored as a strategy for allosteric inhibition (9, 10).

To elucidate the relative contributions of ERK1 and ERK2 in BRAFV600mut melanoma, we analyzed data from three genetic dependency screens performed in large panels of cancer cell lines. Despite the high homology between ERK1 and ERK2, the vast majority of melanoma cell lines were dependent on ERK2 but not ERK1. We validated these findings with genetic and pharmacologic loss-of-function and rescue experiments and found that overexpressed ERK1 could compensate for ERK2 suppression, indicating functional redundancy between these kinases in BRAFV600mut melanoma. Furthermore, both ERK1 and ERK2 required the FRS but not the DRS to mediate viability. We analyzed transcriptomic and proteomic data sets across large cancer cell line panels and found that ERK2 is relatively overexpressed in melanoma cells compared with ERK1 and compared with other cancer types, and ERK2 dependency correlates with mRNA or protein expression across all cell types. Finally, we analyzed signaling changes upon ERK1 and ERK2 suppression in two ERK2-dependent BRAFV600mut cell lines and found that ERK2 is the primary isoform driving MAPK signaling and transcriptional programing in these cells.

Cell lines and drug treatments

HEK293T viral packaging cells were obtained from ATCC. Normal human epithelial melanocytes from adults (NHEM) were obtained from Lonza. The human melanoma cell lines are part of the Cancer Cell Line Encyclopedia (11) and were authenticated by SNP profiling. All cell lines were tested for Mycoplasma infection on December 7, 2020. The following is the list of cell lines, their RRID, and date of SNP testing: A375, CVCL_0132, January 13, 2014; MEL-HO, CVCL_1402, July 19, 2016; SK-MEL-5, CVCL_0527, August 24, 2011; UACC62, CVCL_1780, July 19, 2016; WM1799, CVCL_A341, October 3, 2017; and WM266–4, CVCL_2765, July 19, 2016. All experiments were begun within 2 weeks of thawing cells.

Cells were maintained in Gibco (Thermo Fisher Scientific) RPMI or DMEM or in Lonza Eagle minimum essential medium supplemented with 10% Seradigm TET-free FBS (Avantor), except for NHEM cells, which were maintained in Melanocyte Growth Medium-4 BulletKit supplemented with Endothelin-3 (Lonza). Cells were incubated at 37°C in 5% CO2 atmosphere.

Cells were treated with 100 ng/mL doxycycline (Millipore Sigma) or 1 μmol/L Shield-1 (AOBIOUS) as indicated. Encorafenib, trametinib, MEK-162, ulixertinib, and ERKi-A1 (ERK1/2 inhibitor example A1 from patent WO 2009105500) were synthesized at Novartis and suspended in DMSO.

Stably infected cells were selected in 5 μg/mL puromycin (Corning), 500 μg/mL G418 (Corning), 5 μg/mL blasticidin (Gibco), or 100 μg/mL hygromycin (Gibco) and maintained in 0.2× selection media.

shRNA target sequences

shNT (nontargeting) control: CAACAAGATGAAGAGCACCAA

shERK1: GCCACCTCTCTCCTTTGCTGA

shERK2: TATCCATTCAGCTAACGTTCT

shERK1-alt (alternative): CGTGCTCCACCGAGATCTAAA

shERK2-alt (alternative): CCCATATCTGGAGCAGTATTA

sgRNA target sequences

sgCDKN2A-02: ACCCGTGCACGACGCTGCCC

Plasmids, transfection, and viral transduction

DNA oligomers encoding shRNA sequences were cloned into the Tet-pLKO-puro lentiviral vector (Addgene; plasmid #21915). The pQCXIN-UbC vector was created by replacing the CMV promoter in the pQCXIN retroviral expression vector (Takara) with the Ubiquitin C promoter. Silent mutations in ERK1 and ERK2 CDS to confer resistance to the shRNA sequences and an N-terminal sequence encoding the DD tag from FKBP12 (Takara) were introduced using the QuikChange Site-Directed Mutagenesis Kit (Stratagene). TERT was subcloned from pcDNA-DEST40-TERT (Thermo Fisher Scientific) into pDONR221 using Gateway BP Clonase (Thermo Fisher Scientific) followed by subcloning into the pNGx-LV-v017 lentiviral vector with a hygromycin resistance cassette using Gateway LR Clonase (Thermo Fisher Scientific). The pXP1512-BRAF-V600E lentiviral plasmid was described previously (12). Plasmids were validated by Sanger sequencing (GENEWIZ).

Retrovirus was produced by cotransfecting HEK293T cells with pCMV-Gag-Pol retroviral packaging vector (Cell Biolabs), VSV-G envelope vector (Addgene; plasmid #8454), and pQCXIN-UbC retroviral expression vector using TransIT-293 transfection reagent (Mirus) diluted in Gibco OptiMEM–reduced serum medium (Thermo Fisher Scientific) according to the manufacturer's directions. Lentivirus was similarly produced using the pCMV-dR8.2 dvpr (Addgene; plasmid #8455) lentiviral packing vector. Virus-containing conditioned media was harvested after 48 hours and passed through 0.45-μm cellulose acetate filters (Corning). Target cancer cell lines were spinfected at 1,500 rpm for one hour with virus-containing conditioned media and 10 μg/mL polybrene (Millipore Sigma). Selection media was added 24 hours after infection, and media were changed to 0.2× selection media when all mock-infected control cells were dead.

NHEM-T cells were created by transducing NHEM cells with pNGx-LV-v017-TERT virus followed by hygromycin selection. NHEM-TC cells were created through transient transfection of NHEM-T cells with the NLS-SpCas9-NLS-T2A-mCherry plasmid described previously (13) containing the sgCDKN2A-02 sequence using the ViaFect (Promega) transfection reagent diluted in Gibco OptiMEM (Thermo Fisher Scientific) according to the manufacturer's directions. Cells were collected 24 hours after transfection, washed, stained with Life Technologies Near-IR Live/Dead Cell Stain (Thermo Fisher Scientific), and sorted using a FACSAria Fusion flow cytometer (BD Biosciences) to select high mCherry-expressing cells into a 96-well plate at one cell per well. A single-cell clone of NHEM-TC was confirmed for CDKN2A knockout (CDKN2Ako) by Western blot for the p16 protein and used for IncuCyte confluency assays (see below).

Proliferation assay

Cells stably expressing doxyxyline-inducible shRNA were pretreated with or without doxycycline for three days. On the fourth day after doxycycline addition (day 0), 103 cells/well (A375, SK-MEL-5, MEL-HO, and UACC-62) or 5 × 103 cells/well (WM266–4 and WM1799) were seeded in n = 6 replicate wells each of eight replicate black-walled 96-well plates. Each day, one replicate plate was removed and CellTiter-Glo viability reagent (Promega) was added according to the manufacturer's directions. Fluorescence was read using an EnVision 2104 plate reader (PerkinElmer). Data were normalized to the day 0 plate values and final day values of treated and untreated controls were compared by one-way ANOVA with Tukey multiple comparisons test using Prism 8 (GraphPad Software).

IncuCyte confluency assay

NHEM-TC cells were transduced with BRAF as described above, and after 24 hours, cells were plated in blasticidin selection media in 6-well plates at 105 cells/well. Confluency was calculated from n = 25 images per well taken at 10× magnification every 12 hours. Media were changed twice weekly to refresh nutrients and doxycycline. The cell lines without doxycycline were expanded and maintained for 30 days and retested in the proliferation assay.

Colony formation assay

Cells were plated in 6-well plates at 105 cells/well. Media were replaced twice weekly. After two weeks, the plates were washed with PBS, fixed for 5 minutes with cold 1% formalin in PBS, washed with PBS again, and stained with 1.5% Crystal violet in 20% methanol solution for 15 minutes at room temperature. Plates were then washed repeatedly with deionized H2O until the waste water was clear and dried overnight.

ERKi-A1 dose–response curves

Cells were seeded at 103 cells/well in a black-walled 96-well plate. The next day, the ERKi-A1 was serially diluted in DMSO in a nine-point half-log dilution series with a final top concentration of 10 μmol/L. To half the wells, Shield-1 was added to a final concentration of 1 μmol/L. Three days following drug treatment, cell viability was measured via CellTiter-Glo assay (Promega) according to the manufacturer's directions, giving n = 3 or 4 replicate wells for each drug concentration and Shield1 condition. Data were normalized to the average values of the DMSO without Shield-1 wells and plotted in Prism 8 (GraphPad Software).

Western blot analysis and antibodies

Cells were washed twice with PBS and then lysed with Pierce RIPA buffer (Thermo Fisher Scientific) containing cOmplete Mini EDTA-free Protease Inhibitor Cocktail (Roche) and Phosphatase Inhibitor Cocktails 2 and 3 (Millipore Sigma). Lysates were cleared by centrifugation and stored at -80°C. Protein concentration was measured using DC Protein Assay (Bio-Rad) using a standard curve of Pierce BSA (Thermo Fisher Scientific). Samples containing 20 μg of lysate were prepared with Invitrogen NuPAGE LDS Sample Buffer and Invitrogen NuPage Sample Reducing Agent (Thermo Fisher Scientific) and electrophoresed on Invitrogen NuPAGE Bis-Tris 4% to 12% precast polyacrylamide gels (Thermo Fisher Scientific), and transferred to nitrocellulose membranes using the Invitrogen iBlot transfer stacks and dry transfer apparatus (Thermo Fisher Scientific). Blots were blocked with 5% nonfat milk in TBS-T, and the indicated proteins were detected using the following primary antibodies obtained from MilliporeSigma or Cell Signaling Technology diluted in 5% BSA in TBS-T followed by horseradish peroxidase (HRP)-conjugated secondary antibody targeting IgG from the appropriate species: BRAF (Millipore Sigma HPA001328 RRID:AB_1078296); EGR1 (Cell Signaling Technology, 4153S RRID:AB_2097038); ERK1/2 (Cell Signaling Technology, 4696S RRID:AB_390780); P-ERK1/2 (phospho-T185/Y187 of ERK2; Cell Signaling Technology, 9101S RRID:AB_331646); FosB/FosB2 (Cell Signaling Technology, 2251S RRID:AB_2106903); FRA1 (Cell Signaling Technology, 5281S RRID:AB_10557418); P-FRA1 (phospho-S265; Cell Signaling Technology, 5841S RRID:AB_10835210); p16INK4A (Cell Signaling Technology, 92803S RRID:AB_2750891); RSK1/2/3 (Cell Signaling Technology, 9355S RRID:AB_659900); P-RSK (phospho-T356;Cell Signaling Technology, 8753S RRID: AB_2783561); Vinculin (Millipore Sigma V9131 RRID: AB_477629); anti-mouse IgG, HRP-linked goat Ab (CST 7076P2 RRID: AB_330924); and anti-rabbit IgG, HRP-linked goat Ab (CST 7074P2 RRID: AB_2099233).

Densitometry analysis of Western blot bands corresponding to ERK1 and ERK2 was performed using ImageJ 1.52n (NIH, Rockville, MD) and normalized to GST-tagged recombinant protein bands of known amount after adjusting for the mass of the GST tag.

mRNA expression analyses

Cells were treated with doxycycline (100 ng/mL) for four (A375) or seven (WM266–4) days or with encorafenib (50 nmol/L), trametinib (3 nmol/L), MEK-162 (300 nmol/L), ERKi-A1 (200 nmol/L), or Shield1 (1 μmol/L) for 24 hours. RNA was extracted using the RNeasy Plus Total RNA extraction kit (Qiagen) according to the manufacturer's directions. RNA quantity and quality was measured by UV spectrophotometry using a DS-11 FX+ spectrophotometer (DeNovix) for qRT-PCR experiments or a NanoDrop ND-2000 spectrophotometer (Thermo Fisher Scientific) for RNA sequencing (RNA-seq) and microarray experiments. RNA quality was assessed using the RNA ScreenTape Assay using a 4200 TapeStation system (Agilent Technologies) for RNA-seq experiments or the Bioanalyzer 2100 system (Agilent Technologies) for microarray experiments.

RNA-seq was performed as described previously (12). Gene-set enrichment analysis was performed using GeneGo MetaCore (Thomson Reuters). RNA-seq data has been deposited as an NCBI SRA BioProject under accession number PRJNA704901.

Triplicate RNA samples were submitted to the Novartis microarray core facility and analyzed on Affymetrix HGU133Plus2 microarray platform according to the manufacturer's instructions. The raw signal was normalized using the affy package (http://bioconductor.org) for R 3.5.3 (R Foundation) with RMA method and HGU133Plus2_Hs_ENTREZG custom CDF for target gene definition (14). Top 20% most variable probesets were used for calculated hierarchical clustering of samples and genes via UPGMA method with default Euclidean distance and average value agglomeration method using Spotfire 10.3.3 (TIBCO Software). RNA microarray data have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO series accession number GSE57721.

Reverse transcription was performed on 2 μg of RNA using an Applied Biosystems High-Capacity RNA-to-cDNA kit (Thermo Fisher Scientific) according to the manufacturer's instructions. Quantitative real-time PCR was performed with TaqMan Fast Advanced Master Mix and the indicated TaqMan FAM-labeled probes duplexed with VIC-labeled GUSB probes to assess expression of the appropriate genes (Thermo Fisher Scientific). Relative expression was determined using the ΔΔCt method using GUSB as housekeeping control. The gene and TaqMan Gene Expression Assays are as follows: GUSB, Hs00939627_m1; MAPK1 (ERK2), Hs01046830_m1; MAPK3 (ERK1), Hs00385075_m1; AREG, Hs00950669_m1; COL21A1, Hs00229402_ m1; DUSP5, Hs00244839_m1; EGR1, Hs00152928_m1; FOSB, Hs00171851_m1; NTRK2, Hs00178811_m1; PDE2A, Hs00159935_ m1; POSTN, Hs01566750_m1; SIRPB1, Hs02565816_s1.

BRAFV600mut melanoma cell lines are more dependent on ERK2 than ERK1

Genetic screens performed in large panels of human cancer cell lines have been successful in identifying novel targets and uncovering biological interactions (15). To determine the requirement of MAPK pathway members in BRAFV600mut melanomas, we analyzed three large-scale dependency screens, namely Project DRIVE: a genome-wide shRNA dependency screen performed in 398 human cancer cell lines (16), and two datasets from Project ACHILLES: (i) a genome-wide siRNA dependency screen performed in 501 cancer cell lines (17, 18), and (ii) AVANA, a genome-wide CRISPR screen performed in 341 cancer cell lines (19, 20). These three screens were performed independently in cell lines that are now all included in the Cancer Cell Line Encyclopedia (CCLE), and the data are publically available on the Broad Institute DepMap hub (depmap.org/portal).

There is significant overlap in the cell lines tested in the three screens, but each utilized cell lines not tested in the other two. Of the 99 BRAF-mutant (BRAFmut) lines tested in at least one screen, 65 (66%) are mutated at BRAFV600. Of the 34 cell lines with non-V600 BRAF mutations, only one is melanoma (SK-MEL-30, which is also NRASQ61L mutant). Finally, only three lines are of unknown BRAFmut status, none of which were melanoma.

We first interrogated the effects of BRAF suppression in melanoma cell lines. As expected, the majority of melanoma cell lines are BRAFV600mut (37/46, 80%) and are highly enriched in the sensitive portion of the BRAF dependency plot in all three screens (Fig. 1A–C; Supplementary Table S1). Similar results are seen when the data are segregated by BRAF mutation status, either by BRAFV600mut specifically or by all BRAF mutations; Supplementary Fig. S1A–S1C; Supplementary Table S1). Restricting this analysis to only BRAFV600mut melanoma lines, all cell lines but one display dependency on BRAF expression in at least one of the three screens (Supplementary Table S2; 36/37, 97%). To evaluate whether melanoma cell lines are more dependent on one ERK isoform over the other, we similarly analyzed the dependency plots for MAPK1, the gene encoding ERK2, and MAPK3, the gene encoding ERK1. Interestingly, while melanoma lines cluster in the sensitive portion of the MAPK1 dependency plot in all three dependency screens, this pattern is noticeably absent in the MAPK3 dependency plots (Fig. 1A–C; Supplementary Table S1). Marking cell lines by BRAFV600mut status also revealed dependency on ERK2 but not ERK1, and similar results are seen analyzing BRAFmut status (Supplementary Fig. S1A–S1C; Supplementary Table S1). The majority of BRAFV600mut melanoma are dependent on MAPK1 in at least one of the three screens, while few are similarly dependent on MAPK3 (Supplementary Table S2; MAPK1: 28/37, 76%; MAPK3: 4/37, 11%). Finally, in all three screens, both BRAF and MAPK1 exhibit the highest codependency on the other among all genes tested (Pearson coefficient = 0.67 for DRIVE, 0.49 for ACHILLES siRNA, and 0.67 for AVANA; P < 0.0001 for each screen). However, no correlation is observed between BRAF and MAPK3 dependency (Pearson correlation coefficients < 0.1; P > 0.05 for each screen). Together, these three independently performed genomic dependency screens reveal that the vast majority of BRAFV600mut melanoma cell lines require ERK2, but not ERK1.

Figure 1.

Melanoma cancer cell lines depend on BRAF and ERK2, but not ERK1. A, ATARiS waterfall plots from the Project DRIVE shRNA screen (16) for dependency on BRAF, MAPK1 (ERK2), and MAPK3 (ERK1), colored by cancer type (melanoma or nonmelanoma). B, DEMETER2 waterfall plots from the Project ACHILLES shRNA screen (17, 18) for dependency on BRAF, MAPK1 (ERK2), and MAPK3 (ERK1), colored by cancer type (melanoma or nonmelanoma). C, CERES waterfall plots from the Project AVANA CRISPR/Cas9 screen (19, 20) for dependency on BRAF, MAPK1 (ERK2), and MAPK3 (ERK1), colored by cancer type (melanoma or nonmelanoma). The dashed lines represent dependency threshold (-0.5 for DRIVE and AVANA; -2.5 for ACHILLES). Diagonal hatch marks indicates abbreviated axes. **, P < 0.0001 for enrichment in sensitive portion for melanoma versus nonmelanoma cancer types by Fisher exact test.

Figure 1.

Melanoma cancer cell lines depend on BRAF and ERK2, but not ERK1. A, ATARiS waterfall plots from the Project DRIVE shRNA screen (16) for dependency on BRAF, MAPK1 (ERK2), and MAPK3 (ERK1), colored by cancer type (melanoma or nonmelanoma). B, DEMETER2 waterfall plots from the Project ACHILLES shRNA screen (17, 18) for dependency on BRAF, MAPK1 (ERK2), and MAPK3 (ERK1), colored by cancer type (melanoma or nonmelanoma). C, CERES waterfall plots from the Project AVANA CRISPR/Cas9 screen (19, 20) for dependency on BRAF, MAPK1 (ERK2), and MAPK3 (ERK1), colored by cancer type (melanoma or nonmelanoma). The dashed lines represent dependency threshold (-0.5 for DRIVE and AVANA; -2.5 for ACHILLES). Diagonal hatch marks indicates abbreviated axes. **, P < 0.0001 for enrichment in sensitive portion for melanoma versus nonmelanoma cancer types by Fisher exact test.

Close modal

Although the majority of BRAFV600mut melanoma cell lines are dependent upon ERK2 but not ERK1, we selected cell lines predicted by Project DRIVE to be dependent on ERK2 only (MEL-HO, WM266–4, and WM1799), both ERK1 and ERK2 (A375), or neither isoform (SK-MEL-5 and UACC-62) for further characterization. To assess the effects of loss of ERK1 or ERK2 in these cell lines we used doxycycline-inducible shRNA against ERK1 or ERK2 (shERK1 and shERK2, respectively) and a nontargeting shRNA (shNT) control. Maximum suppression of ERK1 and ERK2 was observed after four days in A375 and WM266–4 cells (Supplementary Fig. S2A), consistent with the reported half-life of these proteins of greater than 48 hours (21). Therefore, cells were pretreated with doxycycline for three days before being tested in a seven-day proliferation assay. In A375 cells, ERK2 knockdown effectively blocked proliferation, whereas ERK1 knockdown led to only a modest reduction in proliferation rate (Fig. 2A). Induction of a shNT had no significant effect on proliferation. Similar results were seen when alternate hairpins targeting ERK1 and ERK2 (shERK1-alt and shERK2-alt, respectively) were used to exclude off-target effects of the shRNA sequences (Fig. 2B). Colony formation was likewise inhibited by shERK2 but not shERK1 in A375 cells (Supplementary Fig. S2B). Similar results were observed in WM266–4 cells where knockdown of ERK2, but not ERK1, blocked proliferation (Fig. 2C).

Figure 2.

ERK2 knockdown suppresses proliferation of BRAFV600mut melanoma cell lines. A–I, Proliferation and Western blot analysis of the indicated human BRAFV600mut melanoma cell line (A–G, I) or NHEM-TCB cells (H) stably expressing doxycycline (Dox)-inducible shRNA against a non-targeting sequence (shNT), ERK1 (shERK1), or ERK2 (shERK2) or an alternate shRNA sequence (alt) against ERK1 or ERK2 to confirm specificity. I, Proliferation of A375 cells stably expressing doxycycline-inducible shERK2 and empty vector (e.v.) or exogenous shRNA-resistant ERK1 or ERK2 (E1SR and E2SR, respectively). Points and error bars represent the mean ± SEM of 6 (A–H) or 10 (I) replicate wells. *, P < 0.05 and **, P < 0.0001 by one-way ANOVA of final day fold proliferation with Tukey multiple comparisons test. Representative of at least two independent replicate experiments.

Figure 2.

ERK2 knockdown suppresses proliferation of BRAFV600mut melanoma cell lines. A–I, Proliferation and Western blot analysis of the indicated human BRAFV600mut melanoma cell line (A–G, I) or NHEM-TCB cells (H) stably expressing doxycycline (Dox)-inducible shRNA against a non-targeting sequence (shNT), ERK1 (shERK1), or ERK2 (shERK2) or an alternate shRNA sequence (alt) against ERK1 or ERK2 to confirm specificity. I, Proliferation of A375 cells stably expressing doxycycline-inducible shERK2 and empty vector (e.v.) or exogenous shRNA-resistant ERK1 or ERK2 (E1SR and E2SR, respectively). Points and error bars represent the mean ± SEM of 6 (A–H) or 10 (I) replicate wells. *, P < 0.05 and **, P < 0.0001 by one-way ANOVA of final day fold proliferation with Tukey multiple comparisons test. Representative of at least two independent replicate experiments.

Close modal

In contrast to the results in A375 and WM266–4 cells, knockdown of either ERK1 or ERK2 in WM1799 and MEL-HO cells was sufficient to significantly inhibit proliferation (Fig. 2D and E). In SK-MEL-5 and UACC62 cells, knockdown of neither ERK1 nor ERK2 was sufficient to completely block proliferation, but shERK2 had a greater effect than shERK1 (Fig. 2F and G). Thus, using doxycycline-inducible shRNAs, we validated the observation that BRAFV600mut melanoma cells generally depend more on ERK2 than ERK1. We also characterized six BRAFV600mut cell lines as being dependent on ERK2 but not ERK1 (A375 and WM266–4), on both ERK1 and ERK2 (WM1799 and MEL-HO), or on neither isoform alone (SK-MEL-5 and UACC62). Although DRIVE data predicted A375 to be ERK1-dependent and MEL-HO and WM1799 to be ERK1-independent, this discrepancy may be due to differences in the time course of the experiments or competition effects from the pooled shRNA screen.

Transformation of primary human melanocytes by BRAFV600E requires both ERK1 and ERK2

We next asked whether the differences in dependency on ERK1 and ERK2 are a product of the melanocyte cell lineage or an adaptive trait of BRAFV600mut melanoma cells. To address this question, we immortalized primary human melanocytes through expression of the reverse transcriptase subunit of human telomerase (TERT) followed by CRISPR-mediated knockout of CDKN2A (Supplementary Fig. S2C), a tumor suppressor gene encoding the p16INK4A and p14ARF proteins commonly lost or mutated in melanoma (22, 23). We then stably infected these cells with lentivirus encoding doxycyline-inducible shNT, shERK1, or shERK2 followed by infection with lentivirus encoding human BRAFV600E. These transformed melanocytes, termed NHEM-TCB (for Normal Human Epithelial Melanocytes with TERT, CDKN2Ako, and BRAFV600E), were maintained in selection media and allowed to adapt to oncogenic signaling for at least 30 days prior to doxycycline treatment. Induction of shERK1 or shERK2 similarly reduced proliferation whereas shNT had no significant effect (Fig. 2H). In addition, the ability of these transformed melanocytes to form colonies from sparsely seeded cells was similarly impaired by either shERK1 or shERK2, but not shNT (Supplementary Fig. S2B), indicating that BRAFV600E-transformed melanocytes depend on both ERK1 and ERK2. Similar results were seen when these transformed melanocytes were treated with doxycycline either four days prior to infection or seven days after infection with lentivirus expressing BRAFV600E in an imaging-based confluency assay (Supplementary Fig. S2D). These results suggest that ERK2 dependency may be an acquired trait in advanced BRAF-driven melanoma, as both isoforms were required for proliferation in melanomagenesis.

ERK1 can functionally compensate for ERK2 in A375 cells

To assess whether ERK1 and ERK2 have distinct functions in A375 cells that could account for the difference in dependency on these two isoforms, we rescued knockdown of ERK2 with exogenous expression of shRNA-resistant ERK1 or ERK2 (ERK1SR or ERK2SR, respectively). As reported above, induction of shERK2 with doxycycline lead to a significant proliferation deficit in A375 cells in a seven-day growth assay (Fig. 2I). As expected, rescue of ERK2 protein expression with exogenous ERK2SR expressed at a similar level as non-doxycycline-treated, empty vector–expressing cells restored proliferation. Exogenous expression of ERK1SR also rescued the proliferation of these cells, suggesting that ERK1 overexpression is able to functionally compensate for ERK2 suppression in this assay.

Melanoma cells express more ERK2 than ERK1

Embryonic lethality of ERK2 (24, 25), but not ERK1 (26), knockout in mice is a function of expression differences of the two kinases, as ERK1 overexpression rescues ERK2 knockout (27). To determine whether expression level differences underlie the differential dependency on ERK1 and ERK2 in melanoma, we analyzed the relative expression of ERK1 and ERK2 from a proteomics study that analyzed relative protein expression across 400 cell lines from the CCLE by quantitative mass spectrometry (28). Melanoma cell lines clustered in the highest one-third of the ERK2 expression plot, suggesting relative overexpression of ERK2 in these cell lines compared with cell lines from other lineages (Fig. 3A; Supplementary Table S1). However, this trend was not seen for ERK1, suggesting that this protein is not expressed higher than in other cancer types. Similar results were observed in data from RNAseq performed on 633 CCLE cell lines (29): melanomas were significantly enriched in the top one-third of cell lines by MAPK1 expression but were slightly underrepresented in the top one-third of cell lines by MAPK3 expression (Supplementary Fig. S3A; Supplementary Table S1). In addition, the total expression of MAPK1 in transcripts per million (TPM) was higher in BRAFV600mut melanoma lines compared with MAPK3, indicating that ERK2 is the more highly expressed kinase of the pair (Supplementary Fig. S3B). When cell lines were divided into sensitive or insensitive to MAPK1 loss in at least one of the three dependency screens, the sensitive cell lines expressed significantly more ERK2 protein (Fig. 3B) or MAPK1 mRNA than insensitive cell lines (Supplementary Fig. S3C).

Figure 3.

Melanomas overexpress ERK2 compared with other cancer types. A, Quantitative mass spectrometry proteomics analysis of ERK2 and ERK1 protein expression in the CCLE (28) colored by cancer type (melanoma or nonmelanoma). The dashed lines separate the top one-third of cell lines by expression of the indicated protein. *, P < 0.001 for enrichment of melanoma cell lines in the top one-third of cell lines expressing ERK2 versus other cancer types by Fisher exact test. B, Comparison between ERK2 protein expression by quantitative mass spectrometry and sensitivity to MAPK1 (ERK2) loss in at least one of the three genetic dependency screens (DRIVE, ACHILLES, and AVANA). Black dots represent BRAFV600mut melanoma cell lines. **, P < 0.0001 by two-tailed Student t test. C and D, Western blot analysis (C) and quantification (D) of ERK1 and ERK2 protein expression in the indicated cell lines compared with recombinant protein standards of known amount. Data are representative of at least two independent replicate Western blots.

Figure 3.

Melanomas overexpress ERK2 compared with other cancer types. A, Quantitative mass spectrometry proteomics analysis of ERK2 and ERK1 protein expression in the CCLE (28) colored by cancer type (melanoma or nonmelanoma). The dashed lines separate the top one-third of cell lines by expression of the indicated protein. *, P < 0.001 for enrichment of melanoma cell lines in the top one-third of cell lines expressing ERK2 versus other cancer types by Fisher exact test. B, Comparison between ERK2 protein expression by quantitative mass spectrometry and sensitivity to MAPK1 (ERK2) loss in at least one of the three genetic dependency screens (DRIVE, ACHILLES, and AVANA). Black dots represent BRAFV600mut melanoma cell lines. **, P < 0.0001 by two-tailed Student t test. C and D, Western blot analysis (C) and quantification (D) of ERK1 and ERK2 protein expression in the indicated cell lines compared with recombinant protein standards of known amount. Data are representative of at least two independent replicate Western blots.

Close modal

To quantify the expression level of ERK1 and ERK2 in the six melanoma cell lines and transformed melanocytes we characterized for dependency, we performed a Western blot on a dilution series of lysates from these cell lines alongside recombinant ERK1 and ERK2 protein standards to quantify protein levels. We analyzed the six established melanoma lines side-by-side and analyzed the NHEM-TCB cells separately due to the varied expression of ERK1 and ERK2 in these lines (Supplementary Fig. S3D). Consistent with the mass spectrometry and RNAseq data, ERK2 is more highly expressed than ERK1 in all six melanoma cell lines (Fig. 3C and D). However, ERK1 and ERK2 levels are close to equal in NHEM-TCB cells. Interestingly, ERK2 is expressed at a very similar level in the four melanoma cell lines that are dependent on this protein and is lower in the two cell lines that are not dependent on ERK2. ERK1 expression is more variable among all six lines, but is highest in the two lines dependent on this protein and lowest in the SK-MEL-5 cell line that is not dependent on either ERK1 or ERK2. Likewise, total combined ERK1/2 protein is lowest in the two melanoma cell lines that are not dependent on either ERK1 or ERK2. Finally, total ERK1/2 expression is lowest in the NHEM-TCB cells, again suggesting that ERK upregulation may be an adaptation of advanced melanoma. Finally, though the phosphorylation levels of ERK1 and ERK2 varied from line-to-line, the ratio of phosphorylated ERK1-to-ERK2 reflected the ratio of total protein levels, and phosphorylation was lowest in the two least sensitive cell lines and the NHEM-TCB model (Supplementary Fig. S3D). Combined with the proteomic and transcriptional expression data, these results suggest that the dependency of BRAFV600mut melanoma lines on the ERK1 or ERK2 proteins is a function of their expression levels.

Inhibitor-resistant ERK1 and ERK2 mutants similarly rescue viability A375 cells treated with ERKi

As an alternative approach to test ERK1/2 redundancy, we engineered ATP-competitive inhibitor-resistant ERK mutants (ERK1IR and ERK2IR, respectively) through introduction of two point mutations (Y53A/Q122I in ERK1). Because constitutive ERK overexpression kills BRAFV600mut melanoma cells (12), we employed an inducible system orthogonal to the doxycycline/TET system. Using the ProteoTuner degron tag system, we encoded an N-terminal destabilization domain (DD) tag derived from FKBP12 onto ERK1IR and ERK2IR (30). The unstructured DD moiety is cotranslationally ubiquitinylated for proteosomal degradation and thus suppresses protein expression of the tagged ERK isoform. Addition of the cell membrane–permeable small molecule Shield1 stabilizes the DD tag and induces protein expression of the encoded ERK isoform. Using this system, we exogenously expressed DD-ERK1IR or DD-ERK2IR in A375 cells with or without Shield1 to induce ERKIR expression (Fig. 4A; Supplementary Fig. S4A) and assessed the viability of these cells after three days in the presence of an ERK inhibitor, ERKi-A1 (an analogue of the ERK inhibitor MK-8353). As expected, A375 cells are highly sensitive to ERK inhibition. To control for off-target effects of the Shield1 compound, we tested A375 cells infected with virus produced with empty vector and observed no change in viability upon the addition of Shield1 over a nine-point half-log dilution dose-response assay (Fig. 4B). Shield1-induced expression of either DD-ERK1IR or DD-ERK2IR led to a nearly complete rescue of viability in A375 cells (Fig. 4C; Supplementary Fig. S4B), suggesting that inhibition of endogenous ERK1 and ERK2 was compensated for by either ERK1IR or ERK2IR to a similar extent. Together with the rescue of viability upon ERK2 knockdown by either ERK1SR or ERK2SR, we conclude that ERK1 and ERK2 are functionally redundant in the viability and proliferation of A375 cells, and that the observed differences in dependency on these two kinases in melanoma cell lines is likely a function of higher ERK2 expression.

Figure 4.

Inhibitor-resistant ERK1 overexpression rescues viability upon ERK inhibition. A, Western blot analysis of A375 cells expressing empty vector (e.v.) or Shield1-inducible, DD-tagged (DD) inhibitor-resistant mutant (IR) DD-ERK1IR with no additional mutations (WT) or with a kinase-dead mutation (KD) or mutations disrupting the DRS or FRS. B–F, Dose response of cell viability in the presence of ERKi-A1 of A375 cells expressing e.v. (B) or DD-ERK1IR (C–F) with no additional mutations (WT; C) or with KD (D), DRS (E), or FRS mutations (F). Points and error bars represent the mean ± SEM of three or four replicate wells, and data are representative of at least three independent replicate experiments.

Figure 4.

Inhibitor-resistant ERK1 overexpression rescues viability upon ERK inhibition. A, Western blot analysis of A375 cells expressing empty vector (e.v.) or Shield1-inducible, DD-tagged (DD) inhibitor-resistant mutant (IR) DD-ERK1IR with no additional mutations (WT) or with a kinase-dead mutation (KD) or mutations disrupting the DRS or FRS. B–F, Dose response of cell viability in the presence of ERKi-A1 of A375 cells expressing e.v. (B) or DD-ERK1IR (C–F) with no additional mutations (WT; C) or with KD (D), DRS (E), or FRS mutations (F). Points and error bars represent the mean ± SEM of three or four replicate wells, and data are representative of at least three independent replicate experiments.

Close modal

To rule out off-target effects of ERKi-A1, we tested whether the kinase activity of ERKIR is necessary for rescue of A375 cell viability. Lys 71 of ERK1 is critical for kinase activity (8, 31), and mutating this residue to Arg in DD-ERK1IR abrogated its ability to rescue viability upon Shield1 treatment (Fig. 4D). Similar results were seen with the analogous mutation in ERK2 (Supplementary Fig. S4C). Thus, kinase activity is required for ERK1IR or ERK2IR to rescue A375 cell viability upon ERK inhibition.

The F-recruitment site is required for ERK-mediated viability of A375 cells

ERK1 and ERK2 possess two independent docking domains, the DRS and the FRS, that are distinct from the active site and recognize and interact with amino acid motifs on regulators and substrates (32). To determine whether one or both of these docking sites is required for ERK1IR to mediate viability in the presence of an ERK inhibitor, we mutated key residues in the FRS (L217A/L251A) or in the DRS (D338N) that reduce binding of substrates to these domains (32, 33). Induction of ERK1IR with the DRS mutation was able to partially rescue viability of A375 cells (Fig. 4E), whereas the FRS mutant failed to rescue viability in the presence of ERKi-A1 (Fig. 4F). Similar results were seen when the analogous mutants of ERK2IR were tested (Supplementary Fig. S4D and S4E). Thus, while the DRS of ERK1 and ERK2 was dispensable, the FRS was required to rescue A375 cell viability in the presence of an ERK inhibitor. These results suggest that the subset of ERK interactors that contain F-site motifs may be responsible for ERK-mediated viability of A375 cells and that allosteric inhibition of ERK targeting the FRS may provide a novel therapeutic modality.

ERK2 drives MAPK signaling in A375 and WM266–4 cells

To determine the mechanism by which WM266–4 cells are dependent on ERK2 but not ERK1, we analyzed the transcriptional changes in these cells upon treatment with MAPK inhibitors or induction of shERK1 or shERK2. WM266–4 cells were treated with inhibitors against BRAF, MEK, or ERK for 24 hours, or with doxycycline to induce shNT, shERK1, or shERK2 for seven days. We performed microarray gene transcriptional analysis of these cells, normalized the fold changes to DMSO treatment, and sorted the data by the largest transcriptional responses to ERK inhibitor treatment (Supplementary Fig. S5A). Calculated hierarchical clustering of the treatment groups revealed that the transcriptional response to ERK2 knockdown most closely resembles treatment with MAPK inhibitors, whereas knockdown of ERK1 closely resembles induction of non-targeting shRNA. Analysis of the most downregulated genes in this dataset revealed suppression of known MAPK transcriptional targets, including members of the ETV, ERG, and SPRY gene families, in the MAPK inhibitor-treated and shERK2 knockdown cells, whereas induction of shNT or shERK1 shRNA did not downregulate these genes (Supplementary Fig. S5B). Thus, ERK2 is the primary isoform driving the MAPK transcriptional program in WM266–4 cells. Notably, these results are in agreement with those observed by others studying the transcriptional effects of ERK1 and ERK2 in triple-negative breast cancer cell lines (34).

To further characterize the mechanism by which melanoma cells are dependent on ERK2 but not ERK1 in a second cell line, we analyzed the global transcriptional and signaling response changes due to ERK1 or ERK2 knockdown in A375 cells. We induced shNT, shERK1, or shERK2 for four days in A375 cells and performed RNA-seq expression analysis. In addition, ERK2 was knocked down for four days and DD-ERK1SR was induced with Shield1 for 24 hours. In contrast to the results seen in WM266–4 cells and reported by others in triple-negative breast cancer (34), analysis of these RNA-seq data revealed significant overlap between ERK1- and ERK2-regulated genes, but with some differences (Fig. 5A). Among the genes suppressed more by knockdown of ERK2 than ERK1, AREG, DUSP5, EGR1, FOSB, PDE2A, and SIRPB1 were validated by quantitative reverse-transcription PCR (qRT-PCR; Fig. 5B). In addition, expression of some genes increased upon knockdown of either ERK1 or ERK2 but increased to a greater extent with ERK2 knockdown (Supplementary Fig. S5C and S5D), and this effect was validated for COL21A1, NTRK2, and POSTN with qRT-PCR (Supplementary Fig. S5E). Among these genes, the ERK2 knockdown-induced expression changes of AREG, EGR1, FOSB, PDE2A, SIRPB1, and POSTN were rescued by overexpression of ERK1SR in both the original RNAseq and in qRT-PCR follow-up validation (Fig. 5B; Supplementary Fig. S5C–S5E). Thus, the expression of several genes including known MAPK target genes AREG (35), EGR1 (36), and FOSB (37) were selectively regulated by shERK2 and rescued by DD-ERK1SR, representing genes that may regulate the proliferation of A375 cells. Melanocytes differentiate from neural crest cells, and gene-set enrichment analysis performed on the RNA-seq data identified pathways involved in epithelial-to-mesenchymal transition and neurodevelopment (Supplementary Table S3), potentially representing a dedifferentiation program previously associated with MAPK inhibition in melanoma (38).

Figure 5.

Suppression of MAPK signaling is deeper and more sustained with ERK2 knockdown than with ERK1 knockdown. A, Heatmap of RNAseq analysis in A375 cells expressing the indicated doxycycline (Dox)-induced shRNA and/or Shield1-induced DD-ERK1SR and treated with doxycyline for 4 days and with or without Shield1 for 1 day normalized to no doxycycline controls. B, qRT-PCR of the indicated genes in cDNA libraries from A375 cells expressing the indicated doxycyline-inducible shRNA or Shield1-inducible DD-ERK1SR and treated with doxycycline for 4 days and with or without Shield1 for 1 day. Bars and error bars represent the mean ± SEM of three technical replicates. **, P < 0.0001 by two-way ANOVA with Tukey multiple comparisons test. Samples were obtained from biological replicates independently from those in A, and the data are representative of at least three independent replicate experiments. C, Western blot analysis of ERK targets in A375 cells after induction of the indicated shRNA with doxycycline for 4–10 days. Red arrows indicate media change days. Representative of at least three independent replicate experiments. Loading controls for each blot can be found in Supplementary Fig. S6.

Figure 5.

Suppression of MAPK signaling is deeper and more sustained with ERK2 knockdown than with ERK1 knockdown. A, Heatmap of RNAseq analysis in A375 cells expressing the indicated doxycycline (Dox)-induced shRNA and/or Shield1-induced DD-ERK1SR and treated with doxycyline for 4 days and with or without Shield1 for 1 day normalized to no doxycycline controls. B, qRT-PCR of the indicated genes in cDNA libraries from A375 cells expressing the indicated doxycyline-inducible shRNA or Shield1-inducible DD-ERK1SR and treated with doxycycline for 4 days and with or without Shield1 for 1 day. Bars and error bars represent the mean ± SEM of three technical replicates. **, P < 0.0001 by two-way ANOVA with Tukey multiple comparisons test. Samples were obtained from biological replicates independently from those in A, and the data are representative of at least three independent replicate experiments. C, Western blot analysis of ERK targets in A375 cells after induction of the indicated shRNA with doxycycline for 4–10 days. Red arrows indicate media change days. Representative of at least three independent replicate experiments. Loading controls for each blot can be found in Supplementary Fig. S6.

Close modal

Finally, we analyzed the signaling response to knockdown of ERK1 or ERK2 over a 10-day time-course in A375 cells by Western blotting for known MAPK targets (Fig. 2A). Knockdown of both ERK1 and ERK2 was stable over the entire 10-day period (Fig. 5C; Supplementary Fig. S6). Importantly, no compensatory increase in either protein level or phosphorylation of the remaining isoform was observed, indicating that A375 cells do not adapt to ERK2 suppression in the timeframe of this experiment. Although there is slight variability in the phospho-ERK2 levels between days in the ERK1 knockdown samples, we do not believe this is biologically relevant because it is not increased above the basal level and it is not sufficient to rescue proliferation. FRA1 and FOSB, AP-1 member proteins whose transcriptional activity and protein stability are induced by ERK phosphorylation (8), were suppressed by both ERK1 and ERK2 knockdown. However, the suppression was deeper and more stable with ERK2 knockdown than with ERK1 knockdown, indicating that ERK2 is the primary isoform responsible for the induction of these targets. Similarly, phosphorylation of p90RSK, a direct substrate of ERK1 and ERK2, decreased over 10 days of ERK2 knockdown, but remained steady in ERK1 knockdown samples. Finally, expression of the transcription factor EGR1 was suppressed by shERK2 but increased slightly with shERK1, an effect that was also observed in the RNA-seq dataset (Supplementary Fig. S5D). The deeper suppression of MAPK signaling in response to ERK2 knockdown compared with ERK1 knockdown likely contributes to the differential dependency of A375 and WM266–4 cells on these isoforms.

BRAFV600mut melanomas depend on continued signaling through the MAPK pathway (39). Analysis of three independent genome-wide screens revealed that dependency is not equal for all MAPK pathway members. While BRAFV600mut melanoma lines were heavily dependent on BRAF and ERK2, ERK1 was dispensable. These results were independently validated in six BRAFV600mut melanoma cell lines in which five of six lines depended more on ERK2 than ERK1. Furthermore, melanoma cell lines expressed more ERK2 than cell lines from other lineages at the mRNA and protein level, and expressed more ERK2 than ERK1 at the mRNA level, suggesting that ERK2 dependency is a function of its higher expression in BRAFV600mut melanoma. Transformation studies in primary human melanocytes revealed a similar dependency on ERK1 and ERK2, and expression levels were similar between the two kinases in this line and much lower than in the BRAFV600mut melanoma lines, suggesting that ERK2 overexpression may be an adaptive response to BRAF oncogene addiction in advanced melanoma rather than an innate trait of the melanocyte lineage.

Despite the close sequence homology between ERK1 and ERK2 (84% amino acid identity), distinct requirements for one kinase over the other have been reported (5–7, 40). In addition, the kinetics of nuclear translocation differ for these kinases (41). However, in some contexts, a requirement for one isoform of ERK can be compensated for through overexpression of the other isoform (5, 27, 42). We found that ERK1 expression can compensate for knockdown of ERK2 or pharmacologic inhibition of endogenous ERK1/2 in BRAFV600mut melanoma cells, supporting our hypothesis that expression level differences rather than distinct functions underlie the requirement for ERK2 but not ERK1 in this cancer.

The ERK1/2 MAPKs have over 200 identified substrates that promote cell growth, proliferation, and survival (43). ERK1 and ERK2 engage with many substrates and regulators through the D-recruitment site (DRS) and/or the FRS (32, 44–47). Here we report that the FRS but not the DRS of ERK1/2 is required for an inhibitor-resistant mutant to rescue viability upon pharmacologic ERK inhibition. Targeting the FRS to disrupt substrate binding may inhibit melanoma growth while sparing other ERK activities that require the DRS, potentially reducing dose-limiting on-target toxicity (10, 32, 44, 46, 48). Interrupting protein–protein interactions specifically at one or the other docking domain of ERK has been explored as a therapeutic strategy (9, 49, 50).

Analysis of transcriptional changes in two BRAFV600mut melanoma lines revealed that ERK2 knockdown suppressed canonical MAPK targets to a greater extent than ERK1 knockdown. In WM266–4 cells, global transcriptional changes upon ERK2 knockdown were similar to those seen upon treatment with inhibitors of RAF, MEK1/2, or ERK1/2, while ERK1 knockdown resembled the nontargeting shRNA control, findings that agree well with recent studies in triple negative breast cancer (34). In contrast, though consistent with other studies in A375 cells (51), significant overlap existed in the genes regulated by ERK1 or ERK2, but some differences were observed, including cases where gene expression changes were greater or in opposite directions upon ERK2 knockdown versus ERK1 knockdown.

Western blot analysis of A375 cells with knockdown of ERK1 or ERK2 over a 10-day time-course revealed that ERK2 knockdown led to a deeper and more robust suppression of MAPK signaling than ERK1 knockdown. Due to relief of feedback inhibition of ERK phosphorylation (52), suppression of one isoform of ERK may lead to increased activation of the remaining pool (53). Importantly, no compensation by the remaining isoform at the expression level or the phosphorylation level was observed when ERK1 or ERK2 were knocked down in our experiments. In addition, downstream MAPK signaling was suppressed upon ERK2 knockdown for the entire 10-day experiment with no compensation through ERK1 activity, suggesting that targeting ERK2 is sufficient to block MAPK signaling in A375 cells.

Current clinical ERK inhibitors display little selectivity for ERK2 over ERK1 and are limited by toxicity. The toxicity profiles of ERK inhibitors closely resemble those of other MAPK inhibitors such as BRAF and MEK inhibitors, suggesting that toxicity is due to on-target inhibition of MAPK signaling in normal tissue (4). Unlike in melanoma where we have described a relative overexpression of ERK2 compared to ERK1, analysis data in the Genotype-Tissue Expression (GTEx) project portal reveals that ERK1 is expressed close to evenly with or higher than ERK2 in normal human tissues (gtexportal.org), suggesting that ERK2-selective inhibition may be toxicity-sparing. Given the ability of ERK1 overexpression to rescue ERK2 suppression, we would expect upregulation of ERK1 to represent a potential resistance mechanism to ERK2-selective inhibition. However, whether tumors can evolve to upregulate ERK1 expression in response to ERK2 inhibition remains to be seen, and testing this hypothesis would require long-term acquired resistance experiments. The kinase domains of ERK1 and ERK2 are identical, thus preventing selective active site inhibition (54). However, ERK1 and ERK2 differ in allosteric sites, including in their N-termini where ERK1 contains an additional 17 amino acid residues, which was found to be responsible for differences in nuclear translocation kinetics (41). Nuclear translocation is key for full activation of the MAPK-induced transcriptional program, and targeting this function in BRAFmut melanoma cells has been explored (55). Given the widespread requirement for ERK2 in melanoma and the lack of compensation by ERK1 upregulation observed, an ERK2-selective inhibitor may provide better tolerated and more robust response in patients. Together, our findings describe a wide-spread dependency of BRAFV600mut melanoma on ERK2 but not ERK1 and support an isoform-selective strategy for the development of future clinical ERK inhibitors.

J.A. Engelman reports other compensation from Novartis during the conduct of the study; and Novartis employee and equity holder. D.D. Stuart reports personal fees from Novartis outside the submitted work. No disclosures were reported by the other authors.

M.S. Crowe: Conceptualization, resources, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. T. Zavorotinskaya: Conceptualization, data curation, validation, investigation, writing–review and editing. C.F. Voliva: Conceptualization, supervision, methodology, writing–review and editing. M.D. Shirley: Data curation, formal analysis, methodology, writing–review and editing. Y. Wang: Data curation, investigation, methodology, writing–review and editing. D.A. Ruddy: Data curation, supervision, methodology, writing–review and editing. D.P. Rakiec: Data curation, investigation, methodology, writing–review and editing. J.A. Engelman: Supervision, methodology, project administration, writing–review and editing. D.D. Stuart: Conceptualization, supervision, methodology, project administration, writing–review and editing. A.K. Freeman: Conceptualization, supervision, methodology, writing–original draft, project administration, writing–review and editing.

The authors would like to thank E. Deng, T. Feng, Y. Feng, J.M. Korn, G.P. Leung, Y.M. Mishina, and G.K. Yu for technical assistance and helpful discussion. All of the work described was done and supported by Novartis Institutes for BioMedical Research.

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

1.
Eblen
ST
. 
Extracellular-regulated kinases: signaling from Ras to ERK substrates to control biological outcomes
.
Adv Cancer Res
2018
;
138
:
99
142
.
2.
Davies
H
,
Bignell
GR
,
Cox
C
,
Stephens
P
,
Edkins
S
,
Clegg
S
, et al
Mutations of the BRAF gene in human cancer
.
Nature
2002
;
417
:
949
54
.
3.
Yaeger
R
,
Corcoran
RB
. 
Targeting alterations in the RAF-MEK pathway
.
Cancer Discov
2019
;
9
:
329
41
.
4.
Roskoski
R
 Jr
. 
Targeting ERK1/2 protein-serine/threonine kinases in human cancers
.
Pharmacol Res
2019
;
142
:
151
68
.
5.
Busca
R
,
Pouyssegur
J
,
Lenormand
P
. 
ERK1 and ERK2 map kinases: specific roles or functional redundancy?
Front Cell Dev Biol
2016
;
4
:
53
.
6.
Cohen-Armon
M
,
Visochek
L
,
Rozensal
D
,
Kalal
A
,
Geistrikh
I
,
Klein
R
, et al
DNA-independent PARP-1 activation by phosphorylated ERK2 increases Elk1 activity: a link to histone acetylation
.
Mol Cell
2007
;
25
:
297
308
.
7.
Kang
BS
,
Hwang
YJ
,
Dong
Z
. 
ERK1 Directly Interacts With JNK1 Leading to Regulation of JNK1/c-Jun Activity and Cell Transformation
.
J Cell Biochem
2017
;
118
:
2357
70
.
8.
Roskoski
R
 Jr
. 
ERK1/2 MAP kinases: structure, function, and regulation
.
Pharmacol Res
2012
;
66
:
105
43
.
9.
Miller
CJ
,
Muftuoglu
Y
,
Turk
BE
. 
A high throughput assay to identify substrate-selective inhibitors of the ERK protein kinases
.
Biochem Pharmacol
2017
;
142
:
39
45
.
10.
Sammons
RM
,
Ghose
R
,
Tsai
KY
,
Dalby
KN
. 
Targeting ERK beyond the boundaries of the kinase active site in melanoma
.
Mol Carcinog
2019
;
58
:
1551
70
.
11.
Barretina
J
,
Caponigro
G
,
Stransky
N
,
Venkatesan
K
,
Margolin
AA
,
Kim
S
, et al
The Cancer Cell Line Encyclopedia enables predictive modelling of anticancer drug sensitivity
.
Nature
2012
;
483
:
603
7
.
12.
Leung
GP
,
Feng
T
,
Sigoillot
FD
,
Geyer
FC
,
Shirley
MD
,
Ruddy
DA
, et al
Hyperactivation of MAPK signaling is deleterious to RAS/RAF-mutant melanoma
.
Mol Cancer Res
2019
;
17
:
199
211
.
13.
Smurnyy
Y
,
Cai
M
,
Wu
H
,
McWhinnie
E
,
Tallarico
JA
,
Yang
Y
, et al
DNA sequencing and CRISPR-Cas9 gene editing for target validation in mammalian cells
.
Nat Chem Biol
2014
;
10
:
623
5
.
14.
Dai
M
,
Wang
P
,
Boyd
AD
,
Kostov
G
,
Athey
B
,
Jones
EG
, et al
Evolving gene/transcript definitions significantly alter the interpretation of GeneChip data
.
Nucleic Acids Res
2005
;
33
:
e175
.
15.
Sato
M
. 
Phenotypic screening using large-scale genomic libraries to identify drug targets for the treatment of cancer
.
Oncol Lett
2020
;
19
:
3617
26
.
16.
McDonald
ER
 3rd
,
de Weck
A
,
Schlabach
MR
,
Billy
E
,
Mavrakis
KJ
,
Hoffman
GR
, et al
Project DRIVE: a compendium of cancer dependencies and synthetic lethal relationships uncovered by large-scale, deep RNAi screening
.
Cell
2017
;
170
:
577
92
.
17.
McFarland
JM
,
Ho
ZV
,
Kugener
G
,
Dempster
JM
,
Montgomery
PG
,
Bryan
JG
, et al
Improved estimation of cancer dependencies from large-scale RNAi screens using model-based normalization and data integration
.
Nat Commun
2018
;
9
:
4610
.
18.
Tsherniak
A
,
Vazquez
F
,
Montgomery
PG
,
Weir
BA
,
Kryukov
G
,
Cowley
GS
, et al
Defining a cancer dependency map
.
Cell
2017
;
170
:
564
76
.
19.
Dempster
JM
,
Pacini
C
,
Pantel
S
,
Behan
FM
,
Green
T
,
Krill-Burger
J
, et al
Agreement between two large pan-cancer CRISPR-Cas9 gene dependency data sets
.
Nat Commun
2019
;
10
:
5817
.
20.
Meyers
RM
,
Bryan
JG
,
McFarland
JM
,
Weir
BA
,
Sizemore
AE
,
Xu
H
, et al
Computational correction of copy number effect improves specificity of CRISPR-Cas9 essentiality screens in cancer cells
.
Nat Genet
2017
;
49
:
1779
84
.
21.
Schwanhausser
B
,
Busse
D
,
Li
N
,
Dittmar
G
,
Schuchhardt
J
,
Wolf
J
, et al
Global quantification of mammalian gene expression control
.
Nature
2011
;
473
:
337
42
.
22.
Reed
JA
,
Loganzo
F
 Jr
,
Shea
CR
,
Walker
GJ
,
Flores
JF
,
Glendening
JM
, et al
Loss of expression of the p16/cyclin-dependent kinase inhibitor 2 tumor suppressor gene in melanocytic lesions correlates with invasive stage of tumor progression
.
Cancer Res
1995
;
55
:
2713
8
.
23.
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
7
.
24.
Hatano
N
,
Mori
Y
,
Oh-hora
M
,
Kosugi
A
,
Fujikawa
T
,
Nakai
N
, et al
Essential role for ERK2 mitogen-activated protein kinase in placental development
.
Genes Cells
2003
;
8
:
847
56
.
25.
Saba-El-Leil
MK
,
Vella
FD
,
Vernay
B
,
Voisin
L
,
Chen
L
,
Labrecque
N
, et al
An essential function of the mitogen-activated protein kinase Erk2 in mouse trophoblast development
.
EMBO Rep
2003
;
4
:
964
8
.
26.
Selcher
JC
,
Nekrasova
T
,
Paylor
R
,
Landreth
GE
,
Sweatt
JD
. 
Mice lacking the ERK1 isoform of MAP kinase are unimpaired in emotional learning
.
Learn Mem
2001
;
8
:
11
9
.
27.
Fremin
C
,
Saba-El-Leil
MK
,
Levesque
K
,
Ang
SL
,
Meloche
S
. 
Functional redundancy of ERK1 and ERK2 MAP kinases during development
.
Cell Rep
2015
;
12
:
913
21
.
28.
Nusinow
DP
,
Szpyt
J
,
Ghandi
M
,
Rose
CM
,
McDonald
ER
 3rd
,
Kalocsay
M
, et al
Quantitative proteomics of the cancer cell line encyclopedia
.
Cell
2020
;
180
:
387
402
.
29.
Ghandi
M
,
Huang
FW
,
Jane-Valbuena
J
,
Kryukov
GV
,
Lo
CC
,
McDonald
ER
 3rd
, et al
Next-generation characterization of the Cancer Cell Line Encyclopedia
.
Nature
2019
;
569
:
503
8
.
30.
Banaszynski
LA
,
Chen
LC
,
Maynard-Smith
LA
,
Ooi
AG
,
Wandless
TJ
. 
A rapid, reversible, and tunable method to regulate protein function in living cells using synthetic small molecules
.
Cell
2006
;
126
:
995
1004
.
31.
Robinson
MJ
,
Harkins
PC
,
Zhang
J
,
Baer
R
,
Haycock
JW
,
Cobb
MH
, et al
Mutation of position 52 in ERK2 creates a nonproductive binding mode for adenosine 5′-triphosphate
.
Biochemistry
1996
;
35
:
5641
6
.
32.
Dimitri
CA
,
Dowdle
W
,
MacKeigan
JP
,
Blenis
J
,
Murphy
LO
. 
Spatially separate docking sites on ERK2 regulate distinct signaling events in vivo
.
Curr Biol
2005
;
15
:
1319
24
.
33.
Shin
S
,
Dimitri
CA
,
Yoon
SO
,
Dowdle
W
,
Blenis
J
. 
ERK2 but not ERK1 induces epithelial-to-mesenchymal transformation via DEF motif-dependent signaling events
.
Mol Cell
2010
;
38
:
114
27
.
34.
Gagliardi
M
,
Pitner
MK
,
Park
J
,
Xie
X
,
Saso
H
,
Larson
RA
, et al
Differential functions of ERK1 and ERK2 in lung metastasis processes in triple-negative breast cancer
.
Sci Rep
2020
;
10
:
8537
.
35.
Gusenbauer
S
,
Zanucco
E
,
Knyazev
P
,
Ullrich
A
. 
Erk2 but not Erk1 regulates crosstalk between Met and EGFR in squamous cell carcinoma cell lines
.
Mol Cancer
2015
;
14
:
54
.
36.
Kaufmann
K
,
Bach
K
,
Thiel
G
. 
The extracellular signal-regulated protein kinases Erk1/Erk2 stimulate expression and biological activity of the transcriptional regulator Egr-1
.
Biol Chem
2001
;
382
:
1077
81
.
37.
Inoue
D
,
Kido
S
,
Matsumoto
T
. 
Transcriptional induction of FosB/DeltaFosB gene by mechanical stress in osteoblasts
.
J Biol Chem
2004
;
279
:
49795
803
.
38.
Ahn
A
,
Chatterjee
A
,
Eccles
MR
. 
The slow cycling phenotype: a growing problem for treatment resistance in melanoma
.
Mol Cancer Ther
2017
;
16
:
1002
9
.
39.
Sharma
A
,
Trivedi
NR
,
Zimmerman
MA
,
Tuveson
DA
,
Smith
CD
,
Robertson
GP
. 
Mutant V599EB-Raf regulates growth and vascular development of malignant melanoma tumors
.
Cancer Res
2005
;
65
:
2412
21
.
40.
Ricard
N
,
Zhang
J
,
Zhuang
ZW
,
Simons
M
. 
Isoform-specific roles of ERK1 and ERK2 in arteriogenesis
.
Cells
2019
;
9
:
38
.
41.
Marchi
M
,
D'Antoni
A
,
Formentini
I
,
Parra
R
,
Brambilla
R
,
Ratto
GM
, et al
The N-terminal domain of ERK1 accounts for the functional differences with ERK2
.
PLoS One
2008
;
3
:
e3873
.
42.
Lefloch
R
,
Pouyssegur
J
,
Lenormand
P
. 
Total ERK1/2 activity regulates cell proliferation
.
Cell Cycle
2009
;
8
:
705
11
.
43.
Unal
EB
,
Uhlitz
F
,
Bluthgen
N
. 
A compendium of ERK targets
.
FEBS Lett
2017
;
591
:
2607
15
.
44.
Fantz
DA
,
Jacobs
D
,
Glossip
D
,
Kornfeld
K
. 
Docking sites on substrate proteins direct extracellular signal-regulated kinase to phosphorylate specific residues
.
J Biol Chem
2001
;
276
:
27256
65
.
45.
Jacobs
D
,
Glossip
D
,
Xing
H
,
Muslin
AJ
,
Kornfeld
K
. 
Multiple docking sites on substrate proteins form a modular system that mediates recognition by ERK MAP kinase
.
Genes Dev
1999
;
13
:
163
75
.
46.
Lee
S
,
Warthaka
M
,
Yan
C
,
Kaoud
TS
,
Ren
P
,
Dalby
KN
. 
Examining docking interactions on ERK2 with modular peptide substrates
.
Biochemistry
2011
;
50
:
9500
10
.
47.
Lee
T
,
Hoofnagle
AN
,
Kabuyama
Y
,
Stroud
J
,
Min
X
,
Goldsmith
EJ
, et al
Docking motif interactions in MAP kinases revealed by hydrogen exchange mass spectrometry
.
Mol Cell
2004
;
14
:
43
55
.
48.
Boston
SR
,
Deshmukh
R
,
Strome
S
,
Priyakumar
UD
,
MacKerell
AD
 Jr
,
Shapiro
P
. 
Characterization of ERK docking domain inhibitors that induce apoptosis by targeting Rsk-1 and caspase-9
.
BMC Cancer
2011
;
11
:
7
.
49.
Hancock
CN
,
Macias
A
,
Lee
EK
,
Yu
SY
,
Mackerell
AD
 Jr
,
Shapiro
P
. 
Identification of novel extracellular signal-regulated kinase docking domain inhibitors
.
J Med Chem
2005
;
48
:
4586
95
.
50.
Sammons
RM
,
Perry
NA
,
Li
Y
,
Cho
EJ
,
Piserchio
A
,
Zamora-Olivares
DP
, et al
A novel class of common docking domain inhibitors that prevent ERK2 activation and substrate phosphorylation
.
ACS Chem Biol
2019
;
14
:
1183
94
.
51.
Qin
J
,
Xin
H
,
Nickoloff
BJ
. 
Specifically targeting ERK1 or ERK2 kills melanoma cells
.
J Transl Med
2012
;
10
:
15
.
52.
Lake
D
,
Correa
SA
,
Muller
J
. 
Negative feedback regulation of the ERK1/2 MAPK pathway
.
Cell Mol Life Sci
2016
;
73
:
4397
413
.
53.
Frémin
C
,
Ezan
F
,
Boisselier
P
,
Bessard
A
,
Pages
G
,
Pouyssegur
J
, et al
ERK2 but not ERK1 plays a key role in hepatocyte replication: an RNAi-mediated ERK2 knockdown approach in wild-type and ERK1 null hepatocytes
.
Hepatology
2007
;
45
:
1035
45
.
54.
Shin
M
,
Franks
CE
,
Hsu
KL
. 
Isoform-selective activity-based profiling of ERK signaling
.
Chem Sci
2018
;
9
:
2419
31
.
55.
Plotnikov
A
,
Flores
K
,
Maik-Rachline
G
,
Zehorai
E
,
Kapri-Pardes
E
,
Berti
DA
, et al
The nuclear translocation of ERK1/2 as an anticancer target
.
Nat Commun
2015
;
6
:
6685
.

Supplementary data