Landscape of homozygous copy number (CN) losses with established and emerging clinical actionability across 430,467 pan-tumor tissue samples

Background: Loss-of-function (LOF) alterations (alts) in BRCA1/2 and PTEN are associated with approved therapies in multiple tumor types, RB1 and STK11 are established prognostic biomarkers, and LOF of a growing number of additional genes (e.g. MTAP, NF1 ) have emerging actionability and are under investigation in clinical trials. LOF alts include inactivating mutations, rearrangements, and homozygous CN losses, which can be analytically challenging to detect. With the growing clinical importance of CN losses, we assessed their prevalence in a cohort of pan-tumor samples submitted for tissue comprehensive genomic profiling (CGP). Methods: 430,467 pan-tumor tissue samples underwent NGS-based CGP on FoundationOne ® CDx. A genome-wide CN model for each sample was generated, segmented and combined with variant allele frequencies of heterozygous SNPs to estimate tumor purity (TP), ploidy and CN for each segment; the presence of homozygous CN losses was assessed for 310 genes. Results: CN losses were detected in 31% of pan-tumor samples and were most prevalent in glioma (53%), pancreas (50%), and gallbladder (50%). The most prevalent pan-tumor homozygous losses were in CDKN2A/ B (15/17%), MTAP (11%), PTEN (4.2%), SMAD4 (2.2%), and RB1 (1.5%). Disease-specific CN loss enrichments were observed, including PTEN in prostate (22%), SMAD4 in GI tumors (12% pancreas, 8.2% gallbladder, 8% small intestine, 4.9% CRC), RB1 in SCLC (12%), STK11 in cervix (5.0%) and NSCLC (2.9%), NF1 in ovarian (4.2%), and KEAP1 in NSCLC (0.6%). HRD associated losses were also detected in BRCA2 (2.8%, 0.5%, and 0.3%), BRCA1 (0.06%, 0.3%, and 0.6%), and PALB2 (0.02%, 0.03%, and 0.02%) in prostate, breast, and ovarian cancers, respectively. CN losses comprised >98% of LOF alts in CDKN2B and MTAP , and >20% of LOF alts in many genes including CDKN2A , PTEN, RB1 , and STK11 . 27% of samples harbored a CN loss associated with a clinical trial enrolling based on LOF alts (across 63 genes), and 22% harbored a CN loss (across 16 genes) associated with trials specifically requiring a CN loss or deletion vs. any LOF alt. The median TP of a sample with a CN loss detected was 50% (IQR 33-68%). The pan-tumor prevalence was 16% in samples with TP < 20% (n=65,881), increased to 34% in samples with 20-30% TP, and remained relatively stable at higher TPs: 36% at 30-40% TP and 33% at >40% TP. Conclusions: CN losses were detected in >30% of pan-tumor tissue samples and included losses associated with approved or investigational therapies such as BRCA1/2, PTEN, MTAP , and NF1 , and losses with prognostic implications such as RB1 and STK11 . Prevalence trended with increasing TP and reached the prevalence of the overall cohort in samples with TP ≥ 20%. CN losses are a diverse class of analytically challenging but increasingly clinically relevant biomarkers, and reliable detection of these events will be important for optimal therapy selection.