Pan-Cancer Analysis of Copy-Number Features Identifies Recurrent Signatures and a Homologous Recombination Deficiency Biomarker to Predict Poly (ADP-Ribose) Polymerase Inhibitor Response

PURPOSE

Copy-number (CN) features reveal the molecular state of cancers and may have predictive and prognostic value in the treatment of cancer. We sought to apply published CN analysis methods to a large pan-cancer data set and characterize ubiquitous CN signatures across tumor types, including potential utility for treatment selection.

METHODS

We analyzed the landscape of CN features in 260,333 pan-cancer samples. We examined the association of 10 signatures with genomic alterations and clinical characteristics and trained a machine learning classifier using CN and insertion and deletion features to detect homologous recombination deficiency signature (HRDsig) positivity. Clinical outcomes were assessed using a real-world clinicogenomic database (CGDB) of comprehensive genomic profiling linked to deidentified, electronic health record-derived clinical data.

RESULTS

CN signatures were prevalent across cancer types and associated with diverse processes including focal tandem duplications, seismic amplifications, genome-wide loss of heterozygosity (gLOH), and HRD. Our novel HRDsig outperformed gLOH in predicting BRCAness and effectively distinguished biallelic BRCA and homologous recombination-repair wild-type (HRRwt) samples pan-tumor, demonstrating high sensitivity to detect biallelic BRCA in ovarian (93%) and other HRD-associated cancers (80%-87%). Pan-tumor prevalence of HRDsig was 6.4%. HRRwt cases represented a significant fraction of the HRDsig-positive cohort, likely reflecting a population with nongenomic mechanisms of HRD. In ovarian and prostate CGDBs, HRDsig identified more patients than gLOH and had predictive value for poly (ADP-ribose) polymerase inhibitor (PARPi) benefit.

CONCLUSION

Tumor CN profiles are informative, revealing diverse processes active in cancer. We describe the landscape of 10 CN signatures in a large pan-cancer cohort, including two associated with HRD. We trained a machine learning-based HRDsig that robustly identified BRCAness and associated with biallelic BRCA pan-tumor, and was predictive of PARPi benefit in real-world ovarian and prostate data sets.

Tumor Fraction Correlates With Detection of Actionable Variants Across > 23,000 Circulating Tumor DNA Samples

PURPOSE

Profiling of circulating tumor DNA (ctDNA) is increasingly adopted in the management of solid tumors, concurrent with increased availability of more comprehensive ctDNA panels. However, variable ctDNA shed can result in variable assay sensitivity. We studied the relationship between ctDNA tumor fraction (TF) and detection of actionable alterations across cancer types.

METHODS

A total of 23,482 liquid biopsies (LBx) submitted between September 2020 and October 2021 were sequenced using a hybrid capture panel that reports genomic alterations (GAs) and genomic biomarkers across 324 cancer-related genes. The primary end points were the prevalence of targetable GAs by cancer type and detection in relationship to ctDNA TF. Sensitivity of detection in LBx was assessed in 1,289 patients with available tissue results.

RESULTS

94% (n = 22,130) of LBx had detectable ctDNA, with a median TF of 2.2%. LBx profiling detected GAs in National Comprehensive Cancer Network category 1 genes in 37% of lung, 30% of prostate, 36% of breast, and 51% of colon cancer cases. Potential germline GAs flagged on clinical reports were detected in genes including <i>BRCA1/2</i>, <i>PALB2</i>, <i>CHEK2</i>, and <i>ATM.</i> Polyclonal mutations in genes associated with resistance such as <i>AR</i>, <i>ESR1</i>, <i>RB1</i>, and <i>NF1</i> were detected. The sensitivity of LBx to detect driver alterations identified in tissue biopsy from the same patient ranged from 58% to 86% but was consistently at or near 100% in cases with TF ≥ 10%.

CONCLUSION

Elevated ctDNA shed is associated with both high sensitivity and negative predictive value for detection of actionable GAs. The presence of elevated TF suggests adequate tumor profiling and may reduce the value of subsequent reflex to confirmatory tissue testing in patients with negative LBx results.

Genomic Analysis of Circulating Tumor DNA in 3,334 Patients with Advanced Prostate Cancer Identifies Targetable BRCA Alterations and AR Resistance Mechanisms

PURPOSE

Comprehensive genomic profiling (CGP) is of increasing value for patients with metastatic castration-resistant prostate cancer (mCRPC). mCRPC tends to metastasize to bone, making tissue biopsies challenging to obtain. We hypothesized CGP of cell-free circulating tumor DNA (ctDNA) could offer a minimally invasive alternative to detect targetable genomic alterations (GA) that inform clinical care.

EXPERIMENTAL DESIGN

Using plasma from 3,334 patients with mCRPC (including 1,674 screening samples from TRITON2/3), we evaluated the landscape of GAs detected in ctDNA and assessed concordance with tissue-based CGP.

RESULTS

A total of 3,129 patients (94%) had detectable ctDNA with a median ctDNA fraction of 7.5%; BRCA1/2 was mutated in 295 (8.8%). In concordance analysis, 72 of 837 patients had BRCA1/2 mutations detected in tissue, 67 (93%) of which were also identified using ctDNA, including 100% of predicted germline variants. ctDNA harbored some BRCA1/2 alterations not identified by tissue testing, and ctDNA was enriched in therapy resistance alterations, as well as possible clonal hematopoiesis mutations (e.g., in ATM and CHEK2). Potential androgen receptor resistance alterations were detected in 940 of 2,213 patients (42%), including amplifications, polyclonal and compound mutations, rearrangements, and novel deletions in exon 8.

CONCLUSIONS

Genomic analysis of ctDNA from patients with mCRPC recapitulates the genomic landscape detected in tissue biopsies, with a high level of agreement in detection of BRCA1/2 mutations, but more acquired resistance alterations detected in ctDNA. CGP of ctDNA is a compelling clinical complement to tissue CGP, with reflex to tissue CGP if negative for actionable variants.See related commentary by Hawkey and Armstrong, p. 2961.