Germline variation in RASAL2 may predict survival in patients with RAS ‐activated colorectal cancer

Abstract Background Therapeutic agents that specifically target patients with RAS mutant colorectal cancer (CRC) are needed. We sought potential drug targets by relating genome‐wide association study and survival data in patients with advanced CRC profiled for mitogen‐activated protein kinase (MAPK) pathway mutations. Methods In total, 694 patients from the clinical trials COIN and COIN‐B had MAPK‐activated CRCs (assigned as KRAS, NRAS, or BRAF mutant). Genome‐wide single nucleotide polymorphism (SNP), gene, and gene‐set analyses were performed to identify determinants of survival. For rs12028023 in RAS protein activator‐like 2 (RASAL2), we studied its effect by MAPK pathway activation status (by comparing to 760 patients without MAPK‐activated CRCs), MAPK gene mutation status, surface area of the primary tumor (as a marker of proliferation), and expression on RASAL2. Results In MAGMA genome‐wide analyses, RASAL2 was the most significant gene associated with survival (p = 2.0 × 10−5). Patients carrying the minor (A) allele in the lead SNP, rs12028023 in intron 1 of RASAL2, had a median increase in survival of 167 days as compared with patients carrying the major allele. rs12028023 was predictive for survival by MAPK‐activation status (p Z‐test = 2.1 × 10−3). Furthermore, rs12028023 improved survival in patients with RAS mutant (hazard ratio [HR] = 0.62, 95% confidence intervals [CI] = 0.5–0.8, p = 3.4 × 10−5) but not BRAF mutant (p = 0.87) CRCs. The rs12028023 A‐allele was associated with reduced surface area of the primary tumor (Beta = −0.037, standard error [SE] = 0.017, p = 3.2 × 10−2) and reduced RASAL2 expression in cultured fibroblasts (p = 1.6 × 10−11). Conclusion Our data demonstrate a prognostic role for RASAL2 in patients with MAPK‐activated CRCs, with potential as a therapeutic target.


| INTRODUCTION
Monoclonal antibodies against the epidermal growth factor receptor, such as cetuximab, have shown benefit in KRAS and, KRAS and NRAS (RAS), wild-type advanced colorectal cancer (CRC) when either used as a monotherapy 1,2 or in combination with chemotherapy. [3][4][5] In contrast, targeted treatments for patients with RAS mutant disease are only just emerging. 6,7 Given that around half of all CRCs are RAS mutant, this represents a clear unmet clinical need. AMG 510 (Sotorasib), an inhibitor of KRAS G12C, traps mutant KRAS in its inactive GDP-bound state 8 and has shown effectiveness in a Phase 2 trial of nonsmall-cell lung cancer. 9 MRTX849 (Adagrasib) also binds KRAS G12C and inhibits intercellular signaling, 10 and has shown promising efficacy in patients with colorectal, nonsmall-cell lung, endometrial, pancreatic, and ovarian cancers. 11 However, both treatments are only effective in cancers harboring G12C, which occurs in just 1-3% of CRCs. Identifying drug targets for improved survival in patients with RAS mutant CRCs therefore remains challenging.
RAS protein activator-like 2 (RASAL2) encodes a RAS GTPaseactivating protein (GAP), which negatively regulates the RAS signaling pathway by converting RAS-GTP to RAS-GDP. 12 RASAL2 was identified as a tumor suppressor in prostate cancer 13 and its inactivation promotes progression and metastasis in colorectal, 14 lung, 15 ovarian 16 and luminal B breast 17 cancers. However, RASAL2 has also shown pro-oncogenic roles in triple-negative breast 18 and hepatocellular 19 cancers. Furthermore, RASAL2 is upregulated in metastatic CRCs with higher expression associated with lymph node involvement and distant metastasis. 12 Knockdown of RASAL2 in multiple CRC cell lines decreases cell proliferation, anchorage-dependent and -independent growth, cell invasion, and migration, 12 and may represent a potential candidate for targeted therapy.
Relating germline variation to outcome in patients with RAS mutant cancers offers the prospect of identifying novel therapeutic targets. To explore this possibility, we analyzed genome-wide association study (GWAS) and survival data on 1589 patients with advanced CRC from the clinical trials COIN 20 and COIN-B. 21 Patients' tumors were profiled for mutations in the mitogen-activated protein kinase (MAPK) and Akt pathways, to help stratify our survival analyses by MAPK pathway activation status.

| Patients and samples
In total, 2671 unrelated patients with metastatic or locally advanced CRC were recruited into the MRC clinical trials COIN (NCT00182715) 20 and COIN-B (NCT00640081) 21    of these mutations ( Figure 1 and Table 1). For comparison, we analyzed 760 patients without MAPK-activated tumors (i.e., those with KRAS, NRAS, and BRAF wild-type CRC) and a further subset whose CRCs carried PIK3CA mutations as a marker of Akt-activation (n = 87 patients with covariate data).

| Statistical analyses
We previously identified clinicopathological factors associated with survival in patients from COIN and COIN-B. 22 Due to the number of covariates added to the regression models, dimensionality reduction was performed using PCA to reduce the risk of overfitting. A threshold of 70% total variance explained was used to select the number of principal components to include, 29 the first five were selected (but only four were necessary when analyzing patients with NRAS mutations). We carried out the GWAS for OS under an additive model. All analyses performed by MAPK gene mutation status were multivariate.
Gene and gene-set analysis were performed on the summary statistics from the association analysis to identify genes containing significant numbers of highly associated SNPs and significantly enriched gene sets. The threshold for significance at gene level was p < 2.5 Â 10 À6 ,

| Bioinformatic analyses
Regional association plots were created using LocusZoom (http:// locuszoom.org). PCA, survival analyses, and manhattan/quantilequantile plots were performed using the psych (https://cran.r-project. org/web/packages/psych/index.html), gwasurvivr, 33 Table 1). Genomewide SNP, gene and gene-set analyses were performed to identify determinants of survival using the first five principal components as covariates, which explained 71.9% of the total variance for previously established prognostic factors. 22 No detectable genomic inflation was observed (lambda = 1.08). No SNPs passed the threshold for genome-wide significance (p < 5.0 Â 10 À8 ).
In MAGMA gene analysis, RASAL2 at 1q25.

| DISCUSSION
To help identify novel therapeutic targets in patients with MAPKactivated CRCs, we studied the relationship between germline variation and survival in patients with somatically profiled advanced CRC.
RASAL2 was the most significant gene associated with survival in patients with MAPK-activated CRCs. Although RASAL2 did not pass formal genome-wide significance in our screen, its direct interaction in two independent cohorts of patients with CRC, 12 although these were not molecularly stratified by MAPK-activation status. However, these data suggest that RASAL2 may represent a potential therapeutic target via modulation of its expression and warrant further investigation. Interestingly, we noted that the rs12028023 A-allele was associated with reduced surface area of the primary tumor in patients with MAPK-activated CRCs, potentially supporting a link between reduced RASAL2 expression and decreased proliferation. These data are consistent with in vitro models of RASAL2 knockdown. 12 Furthermore, given RASAL2's role in tumourigenesis in other cell types, 19 we speculate that it may represent a target for intervention in a broader range of cancers.

AUTHOR CONTRIBUTIONS
Jeremy P. Cheadle obtained funding for and directed this study.