Homologous Recombination Deficiency (HRD) in Non-Small Cell Lung Cancer: Genomic Analysis Using an RNA-Based HRD Algorithm

Authors Stephanie Thiede, Matthew Berginski, Akash Mitra, Timothy J. Taxter, Michelle M. Stein, Rotem Ben-Shachar, Halla Nimeiri, Charu Aggarwal, Jyoti D. Patel

Background:Recent evidence has suggested that some patients with non-small cell lung cancer (NSCLC) harbor a HRD signature that represents a distinct genomic subtype that could be targeted by PARP inhibitors (PARPi). However, there is little data on HRD prevalence in NSCLC or its genomic associations. Here, we evaluated the co-occurrence of driver mutations and established immune biomarkers with an RNA-based HRD signature in a large, real-world NSCLC cohort.

Methods:We analyzed data from 5119 NSCLC patients that underwent sequencing via the Tempus xT test (DNA-seq of 648 genes; RNA-seq with whole exome capture). HRD status was predicted by the Tempus HRD-RNA test, a pan-cancer logistic regression classifier that uses an RNA gene expression signature optimized to distinguish between BRCA-biallelic loss and homologous recombination repair (HRR)-WT samples (Leibowitz et al, 2022). Cohort samples were excluded from model training. All comparisons were tested via chi-squared or Fisher’s exact tests.

Results:An RNA-derived signature of HRD (HRD-RNA+) was observed in 3.53% (n=181/5119) of patients. HRD-RNA+ prevalence was higher in squamous cell carcinoma (84/1331, 6.3%) relative to adenocarcinoma (68/3015, 2.3%; p < 0.001). The prevalence of select alterations by HRD-RNA status are shown in Table. Alterations in BRCA1/2 and HRR genes (inclusive of BRCA1/2) were enriched in HRD-RNA+ vs. HRD-RNA- cases (8.8% vs. 2.5%, p < 0.001; 22% vs. 15%, p = 0.008 respectively). Notably, 141 (78%) HRD-RNA+ patients had no alterations in HRR genes. Of all NCCN targetable driver mutations assessed, KRAS G12C and ALK fusions were the only targetable drivers with significantly different prevalence in HRD-RNA+ vs. HRD-RNA- patients. Across the entire cohort, NCCN driver mutations were depleted in HRD RNA+ patients (18% in HRD-RNA+ vs. 30% in HRD-RNA-, p < 0.001). Immune biomarkers (TMB, PD-L1) did not vary by HRD-RNA status.

Conclusions:Compared to HRD-RNA- NSCLC, HRD-RNA+ NSCLC represents a unique, molecularly defined subset that has a decreased prevalence of NCCN-driver mutations and is not enriched for TMB-H or PD-L1 expression. Further, this signature increases the number of patients classified as HRD-RNA+ compared to HRR gene alterations alone. Functional characterization (e.g. RAD51 foci immunofluorescence assay) and clinical benefit of targeted HRD therapies such as PARPi should be explored in this HRD-RNA+ population.

HRD-RNA+, N = 181 HRD-RNA-, N = 4,938 p-value
BRCA1/2 loss or mutation, n (%) 16 (8.8%) 122 (2.5%) <0.001
HRR gene(s) loss or mutation, n (%) 40 (22%) 736 (15%) 0.008
KRAS G12C, n (%) 8 (4.4%) 529 (11%) 0.007
ALK Fusion, n (%) 0 (0%) 112 (2.3%) 0.034
BRAF V600E, n (%) 0 (0%) 60 (1.2%) 0.3
EGFR mutation, n (%) 22 (12%) 761 (15%) 0.2
ROS1/NTRK1/2/3/RET Fusion, n (%) 1 (0.6%) 45 (0.9%) >0.9
MET Exon 14 skipping, n (%) 1 (0.6%) 47 (1.0%) >0.9
NCCN Driver(s), n (%) 33 (18%) 1,617 (33%) <0.001