Whole-genome CRISPR-Cas9 knockout screens identify SHOC2 as a genetic dependency in NRAS-mutant melanoma

Mutations in the oncogene NRAS that induce constitutive RAS-GTPase activity lead to unchecked cell proliferation and migration through downstream activation of the mitogen-activated protein kinase (MAPK) and phosphoinositide 3-kinase (PI3K) signalling pathways [1]. These mutations occur in approximately 20% of melanomas and very rarely coexist with BRAF V600 mutations. NRAS-mutant melanoma is associated with poor survival [2] and represents an unmet clinical need, with no effective therapies available following immunotherapy failure.

Identification of contextual essential genes that exert stronger fitness effects on NRAS-mutant melanoma cells presents an opportunity for the discovery of targeted therapies. In this study, we employed CRISPR-Cas9-mediated whole-genome dropout screens to identify genetic dependencies in NRAS-mutant melanoma. Typically, melanoma cell lines are cultured under ambient (∼20%) O₂ conditions, despite O₂ concentrations of < 8% at the epidermal-dermal junction where melanocytes reside, resulting in adaptations in gene and protein expression [3]. Therefore, for our screens, we used a panel of early-passage New Zealand Melanoma (NZM) cell lines that were established and cultured under physiological (5%) O₂ conditions [4].

Six NRAS-mutant and seven NRAS-wildtype (five BRAF-mutant, two BRAF/NRAS/NF1-wildtype) NZM cell lines (Supplementary Table S1) were transduced in multiple replicates with the Brunello single guide RNA (sgRNA) library at a multiplicity of infection of approximately 0.3 and screened at 5% O₂ for up to 35 days (Supplementary Methods and Materials). All NZM lines were cultured for transduction at fewer than 10 passages from derivation. The representation of the sgRNA libraries was assessed to evaluate transducibility, with any cell lines exhibiting poor sgRNA representation (< 80% of sgRNAs detected with ≥1 count) excluded from further analyses (Supplementary Table S2). Moderate to high representation was observed in nine of the 13 NZM cell lines, whereas four cell lines were excluded due to < 80% of sgRNAs being detected (Figure 1A). Reduced sgRNA representation was accompanied by dropout of non-targeting control (NTC) sgRNAs (Figure 1A), greater read count inequality (Supplementary Figure S1, Supplementary Table S2) and reduced correlation with the Brunello library plasmids and between individual cell line replicates (Supplementary Figure S2), suggesting stochastic evolution rather than knockout-induced fitness effects.

We used BAGEL2 to estimate gene essentiality relative to reference sets of common essential and nonessential genes. Typically, tissue-agnostic gene sets are used for this purpose [5], but we established a combined essential gene set incorporating both tissue-agnostic and melanoma-specific essential genes (Supplementary Table S3) that demonstrated high precision-recall (Supplementary Figure S3A), with a distinctive separation between the distributions of common essential and nonessential genes in NZM37 cells (Figure 1B, Supplementary Figure S3B). Supplementing tissue-agnostic genes with tissue-specific genes can broaden the pool of common essential genes and enhance the sensitivity of BAGEL2 analysis, making it a valuable consideration when designing CRISPR dropout screens. BAGEL2 analysis using the combined essential gene set revealed differences in the distribution of common essential and nonessential genes across the NZM cell lines (Supplementary Figure S4). Cell lines with high sgRNA representation exhibited a distinct separation between the distributions of common essential and nonessential genes (Figure 1C), whereas the four excluded lines and NZM74 showed > 60% overlap in gene set distributions (Supplementary Table S4).

Comparative analysis between the five NRAS-mutant and four NRAS-wildtype cell lines revealed 59 candidate genes that were significantly more essential in NRAS-mutant melanoma cell lines than in NRAS-wildtype cell lines (P < 0.01; Supplementary Table S5). As expected, given its oncogenic role, NRAS was the top hit, validating our approach for identifying genetic dependencies in NRAS-mutant melanoma. The next most prominent hit was SHOC2, a positive regulator of MAPK signalling (Figure 1D).

To verify our findings, we conducted BAGEL2 analysis of 54 melanoma cell lines screened with the whole-genome Avana sgRNA library for 21 days at 20% O₂, using data extracted from DepMap 24Q2 (Supplementary Table S1). Since DepMap applies stringent criteria for screen quality, the Avana melanoma screens exhibited high sgRNA representation (Supplementary Table S6), with less than 30% overlap between the reference gene set distributions in all but two cases (Supplementary Figure S5, Supplementary Table S7). Consistent with the NZM screens, the top two ranked genes that were significantly more essential in NRAS-mutant than in NRAS-wildtype cell lines were NRAS and SHOC2 (Figure 1E-F). Moreover, these were the only genes that ranked in the top 30 across both sets of screens (Supplementary Table S5). Reactome pathway analysis revealed an over-representation of genes related to DNA repair, RNA metabolism, translation, and ribosomal RNA processing, but not MAPK signalling (Supplementary Figure S6).

To confirm its genetic dependency in NRAS-mutant melanoma cells, SHOC2 was individually knocked out in three NRAS-mutant and three NRAS-wildtype NZM cell lines. Reduced cell proliferation, relative to either nonessential gene knockout cells or NTC cells, was observed following SHOC2 depletion with three separate sgRNA in all replicate transductions of the NRAS-mutant cell lines (Figure 1G, Supplementary Figure S7). In contrast, SHOC2 knockout did not impair the proliferation of NRAS-wildtype cells, which grew similarly to nonessential gene knockout and/or NTC cells (Figure 1H). By day 28, SHOC2-knockout NRAS-mutant cell lines exhibited significantly reduced growth compared to NTC cells for at least two of the three SHOC2 sgRNAs in each cell line, while no significant differences were observed in NRAS-wildtype cell lines (Supplementary Figure S8).

Additionally, SHOC2 knockout was associated with reduced ERK phosphorylation at day 28 in two NRAS-mutant NZM17 replicates but not in NRAS-wildtype NZM37 cells (Figure 1I). As pooled knockouts, the third NZM17 replicate population had regained some SHOC2 expression by day 28. However, at earlier timepoints, it showed no detectable SHOC2 expression and exhibited reductions in ERK phosphorylation similar to those observed in the 28-day NZM17 R1 and R2 cultures, as well as in SHOC2-wildtype cells treated with the RAF dimer inhibitor belvarafenib (Supplementary Figure S9). To confirm that the effect of SHOC2 knockout was not limited to NZM cells under physiological oxygen conditions, we transduced SK-MEL-2 NRAS-mutant melanoma cells with three SHOC2 sgRNAs and observed similar reductions in cell proliferation and ERK phosphorylation, comparable to those seen in NRAS-mutant but not NRAS-wildtype NZM cell lines. Additionally, SHOC2 knockout increased sensitivity to the RAF dimer inhibitors belvarafenib and naporafenib (Supplementary Figure S10).

Although SHOC2 knockout reduced proliferation over 28 days in NRAS-mutant but not NRAS-wildtype cultures, we cannot confirm whether these effects are maintained long-term or in vivo. While SHOC2 depletion has been reported to prevent tumour growth in KRAS-mutant pancreatic and lung tumour xenografts [6, 7], such studies were not attempted here. In pooled knockout cultures, positive selection of cells expressing SHOC2—due to their growth advantage—could occur either in vitro or in vivo [8], potentially underestimating the true impact of SHOC2 knockout on NRAS-mutant tumour growth. This issue could be mitigated by using clonal knockouts; however, these clones would likely have very slow growth, if any at all.

SHOC2 plays a critical role in the regulation of MAPK signalling by forming a holophosphatase complex with MRAS and PP1C, which activates RAF proteins through dephosphorylation at an inhibitory phosphorylation site [9]. The induced essentiality of SHOC2 specifically in NRAS-mutant cells likely reflects their dependence on constitutive MAPK signalling for cell survival (oncogene addiction), in contrast to NRAS-wildtype cells [1]. Additionally, mutant NRAS may substitute for MRAS in the SHOC2-RAS-PP1C complex and, due to its constitutive activity, promote activation of RAF to a greater extent than the SHOC2-MRAS-PP1C complex does in NRAS-wildtype cells [10]. SHOC2 has previously been identified as a genetic dependency in other RAS-related cancers [6, 7], and now our study suggests it could also be a promising therapeutic target for NRAS-mutant melanoma.

Some of the NZM screens were adversely affected by poor sgRNA representation, which limited the ability to distinguish known essential and nonessential genes. The lack of high concordance among replicate libraries within the same cell line likely reflects the heterogeneity present in early-passage NZM cell lines. While early-passage status preserves tumour heterogeneity and enhances clinical relevance, cell lines with greater heterogeneity may be more susceptible to clonal evolution over the screening period, as different cell populations grow at varying rates, leading to random clonal expansion and dropout —similar to what we have observed in in vivo screens [8]. This issue may have been further compounded by physiological O₂ levels, which not only slowed proliferation and necessitated longer duration screens, but also reduced transduction efficiency. Nonetheless, this raises an important dilemma in the design of functional genomics screens regarding prioritisation of clinical relevance versus high sgRNA representation and reproducibility. By analysing two independent datasets, our study aimed to balance both priorities.

In summary, whole-genome CRISPR-Cas9 knockout screens in early-passage NZM cell lines grown under physiological O₂ identified SHOC2 as a genetic vulnerability in NRAS-mutant but not NRAS-wildtype melanoma cells. This finding was subsequently validated through analysis of whole-genome DepMap melanoma screens, and through individual SHOC2 knockout, which resulted in reduced proliferation and ERK phosphorylation in NRAS-mutant melanoma cell lines. Our results highlight the potential of SHOC2 as a therapeutic target for the unmet clinical need of NRAS-mutant melanoma.

Conceptualisation: Tet Woo Lee, Dean C Singleton, Francis W Hunter, and Stephen MF Jamieson. Data curation: Andrea Y Gu, Tet Woo Lee, Aziza Khan, Xuenan Zhang, and Stephen MF Jamieson. Funding acquisition: Francis W Hunter and Stephen MF Jamieson. Investigation and methodology: all authors. Resources: Andrea Y Gu, Tet Woo Lee, Aziza Khan, and Francis W Hunter. Visualisation: Andrea Y Gu and Stephen MF Jamieson. Supervision: Tet Woo Lee, Dean C Singleton, and Stephen MF Jamieson. Writing–original draft: Andrea Y Gu and Stephen MF Jamieson. Writing–review and editing: all authors.

The authors declare that they have no competing interests.

This research was funded by the Cancer Society of New Zealand (18.14), Cancer Research Trust New Zealand (2005-PGS), University of Auckland Genomics into Medicine Strategic Research Initiatives Fund and Maurice Wilkins Centre for Molecular Biodiscovery Flexible Research Programme. Dean C Singleton and Stephen MF Jamieson were supported by Senior Fellowships from the Cancer Society Auckland/Northland.

Not applicable