Comparison of primary versus metastatic tumor tissue sources when designing panels for whole genome–based tumor-informed ctDNA assays in clear cell renal cell carcinoma.

3039 Background: Due to low shedding and low tumor-mutation burden, metastatic clear cell renal cell carcinoma (ccRCC) is poorly modeled by current cell-free circulating tumor DNA (ctDNA) assays. Second generation tumor-informed ctDNA approaches may demonstrate improved sensitivity in ccRCC, where somatic variants identified by whole-genome sequencing (WGS) of tumor tissue are tracked in probe panels to monitor molecular residual disease (MRD) in patient plasma. However, intertumoral genomic heterogeneity of input tissue may hinder optimal panel design. Here we evaluated the concordance of ctDNA results using panels designed from primary nephrectomy or metastatic tumors in ccRCC. Methods: In five patients with oligometastatic ccRCC (≤5 lesions of metastatic disease; median of 1, range 1-4), 13 primary tumor, 24 metastatic lesion subsites and matched buffy coat samples were collected. Baseline plasma was collected prior to radiotherapy. Samples were tested with Myriad Genetics High-Definition MRD assay. Briefly, tumor and normal DNA were whole-genome sequenced (median coverage 30x). A sub-panel with up to 1000 somatic variants was designed for each lesion subsite; variants could be common across sub-panels. Sites were enriched using hybridization capture from patient plasma-derived cfDNA and sequenced to high depth. A statistical model of variant allele counts was used to assess the presence or absence of ctDNA (MRD status) and infer quantitative tumor fraction. Results: Median time from nephrectomy or most recent metastasis sample collection to baseline plasma collection was 146 months (71-169) and 32 months (7-48), respectively. Median panel probe count per patient was 4,311 (3,400-5,000) and per tumor sample was 1,545 (102-2,158). Confident calls were emitted for tumor fractions ranging from 0.0004% to 0.027%. Baseline MRD status was fully concordant across sub-panels derived from primary tumor lesion subsites and metastases in three of five patients (3/5; 60%). Of the patients with discordant calls, a single metastasis-derived sub-panel (comprising 20%-25% of metastatic sub-panels) called MRD discordantly in each. Of the three patients MRD positive at baseline, only one was alive at last follow up. All patients MRD negative at baseline are still alive. Conclusions: A tumor-informed, WGS MRD assay for ccRCC showed comparable performance between probe capture panels designed from primary tumor and metastatic lesion subsites despite differences in anatomical location and time from plasma collection. This result suggests that intertumoral genomic heterogeneity may not be consequential in MRD assays that leverage large panels to detect ctDNA variants, obviating the need for repeat biopsies of metastatic sites. Work is underway to validate this MRD assay in a larger cohort of patients with ccRCC before and after treatment.