P-247: Increased WEE1 expression is predictive of short progression-free survival, independent of standard prognostic factors in multiple myeloma

Introduction: Current prognostic scores in multiple myeloma (MM) rely on disease burden and a limited set of genomic alterations. Gene expression panels have been reported to be prognostic but are not currently utilized in clinical settings. Recently, we applied a novel measure of network robustness, Ollivier-Ricci curvature, to examine patterns of gene-gene interactions in MM and identified novel pathways and genes associated with poor prognosis (Simhal, Maclachlan et al. BCJ 2023). To facilitate clinical application, we focused on the expression of the most prognostic gene highlighted — the tyrosine kinase WEE1, a G2M checkpoint regulator, which has a targeted therapy in trials in ovarian cancer.

Methods: We analyzed the MMRF CoMMpass RNA-Seq, WES, and WGS dataset (overlap; n=659) comparing with two independent datasets having microarray gene expression profiling (GEP) data from the Total Therapy 2 (N=341) and 3 (N=214) trials. WEE1 gene-neighborhood information was extracted from the STRING database for network analysis. Gene set enrichment analysis (GSEA) was conducted to compute hallmark pathway membership. Multivariate modeling was performed using Cox proportional hazards (CPH) models, random forests (RF) regression models to predict WEE1 expression, and random survival forests (RSF) to compute the concordance index (c-index).

Results: From the CoMMpass dataset, the mean age was 62.5 ± 10.7 years; 60% were male, ISS distribution was 35/35/30%, and 53% received an autologous stem cell transplant (ASCT). WEE1 expression defined a high-risk (HR) group by top tertile and a low-risk (LR) group by bottom tertile. PFS was significantly different (p < 1e-9) between the groups, which was validated in the GEP datasets. In multivariate CPH modeling, PFS prediction was independent of both standard biomarkers (hyperdiploidy, t(4;14), t(11;14), t(14;16), TP53 status) and emerging risk factors (the complex structural variant chromothripsis and APOBEC mutational activity). WEE1 expression also remained prognostic when stratifying by treatment type, including ASCT.

RSF modeling on WEE1 and its neighboring genes resulted in a c-index for PFS prediction of 0.63±0.04, comparable to either ISS or R-ISS with measurement of a single gene expression. The increase in WEE1 expression between the risk groups was not correlated with known interacting genes determined via RF. RF modeling showed a 3.2x rise in mean error when using interacting genes to predict WEE1 expression between the HR and LR groups. Differential GSEA showed dysregulation in the hallmark p53 pathway, suggesting elevated WEE1 expression is associated with TP53 aberration.

Conclusions: We showed that WEE1 is (1) prognostic independent of known biomarkers, (2) differentiates outcomes associated with known markers, (3) upregulated independently of its interacting neighbors, and (4) dysregulates P53 and proliferation pathways. Determining the causes of abnormal WEE1 expression may uncover therapeutic targets.