A comprehensive genome-wide CRISPR–Cas9 study in Nature Genetics has uncovered a synthetic lethal interaction between two paralogous genes, CDS1 and CDS2. The findings point to a potential therapeutic vulnerability in metastatic uveal melanoma—a disease with historically poor outcomes and limited treatment options.

Despite breakthroughs in immune checkpoint inhibitors for cutaneous melanoma, metastatic uveal melanoma has seen only modest treatment advancements. Tebentafusp, the first approved systemic therapy, is limited by HLA-A*02:01 allele specificity, and other targeted therapies have shown restricted antitumor efficacy. For the current study, researchers sought to systematically identify novel tumor-intrinsic vulnerabilities to guide therapeutic development.

Led by Pui Ying Chan of the Wellcome Sanger Institute in Hinxton, UK, the investigators profiled 10 human uveal melanoma cell lines using both genome-wide single-gene and combinatorial paired-gene CRISPR–Cas9 libraries. These models underwent extensive genomic, transcriptomic, and proteomic characterization to align closely with The Cancer Genome Atlas (TCGA) uveal melanoma cohort.

The CRISPR library targeted 514 gene pairs—210 putative synthetic lethal, 262 paralog pairs, and 42 uveal-specific gene sets. Each gene pair was interrogated with 32 unique sgRNA pairings, which resulted in more than 25,000 combinatorial constructs. Guide-level effects on cell fitness were rigorously assessed across triplicates and 28-day culture periods.

The researchers identified 105 synthetic lethal gene pairs that were significantly depleted in at least one cell line. Among these, the CDS1/CDS2 pair emerged as a top candidate:

• CDS2 encodes CDP-diacylglycerol synthase 2, an enzyme that is involved in phosphoinositide synthesis, which is crucial for cell signaling pathways such as MAPK and AKT.
• In uveal melanoma cells with low CDS1 expression, knockout of CDS2 led to marked reductions in cell viability.

Single-cell RNA sequencing of 26 uveal melanoma tumors confirmed that malignant cells express CDS2 highly while CDS1 is expressed at minimal levels. In TCGA data, CDS1 had a median expression of log₂(RSEM+1) = 6.1 compared to CDS2 at 10.8.

Functional loss of CDS2 led to:

• Accumulation of phosphatidic acid (PA), the precursor for phosphoinositides
• Depletion of PI, PIP, and PIP2 species (confirmed via LC–MS)
• Formation of supersized lipid droplets
• Increased apoptosis (Annexin V+/DAPI+ cells) rather than cytostasis
• Disruption of GPCR signaling pathways

Importantly, CDS1 re-expression via lentiviral transduction rescued colony formation in CDS2-deficient cells, which validated the synthetic lethality mechanism.

Analysis of 937 cell lines from the Broad Institute’s DepMap showed that CDS2 is significantly more essential in cancers with low CDS1 expression (P=8.59×10⁻⁹), including cutaneous melanoma, glioblastoma, hepatocellular carcinoma, and sarcoma.

Hypermethylation of the CDS1 promoter correlated with gene silencing, which suggested an epigenetic mechanism that may be underpinning the low expression.

Overall, the findings suggest:

• CDS2 represents a promising drug target for tumors with low CDS1 expression.
• CDS1 expression may serve as a predictive biomarker for CDS2 inhibitor sensitivity.
• Lack of bypass mechanisms or resistance in suppressor screens indicates sustained therapeutic potential.

Further, protein structure models from AlphaFold/AlphaFill suggest druggable cavities in CDS2, which further supports its tractability as a small-molecule target.

“An important hurdle,” the authors wrote, “will be to evaluate the effect of CDS2 loss/depletion in normal cells where CDS1 expression is low, including hepatic and heart tissues.”

A full list of author disclosures can be found in the published research. RP