Abstract A013: Multi-omic profiling of plasma circulating tumor DNA (ctDNA) detects and differentiates Wilms tumors and malignant rhabdoid tumors of the kidney

Purpose: Pediatric renal tumors (RT) represent a range of histologies with varying prognoses and therapy recommendations, complicating initial treatment decisions without a histologic diagnosis when surgery or biopsy are contraindicated. In these cases, non-invasive diagnostic assays could guide disease-specific therapy. We previously showed that circulating tumor DNA (ctDNA) is detectable in patients with Wilms tumor (WT) and can identify somatic features when tumor tissue is unavailable. However, conventional ctDNA profiling methods, such as copy number alteration (CNA) detection, are unlikely to distinguish RT histologies. For example, malignant rhabdoid tumors of the kidney (MRTK) are driven by biallelic INI1 loss which is difficult to detect in ctDNA and typically lack other somatic events. Recent advances now allow whole-genome methylation sequencing (WGMS) on samples with low DNA input, such as cell-free DNA. Here, we examined if methylation and fragment features could improve ctDNA detection and differentiate diagnoses in patients with WT and MRTK as proof of concept for liquid biopsy diagnostics. Methods: Pretreatment plasma and tumor tissue from patients with WT (14 matched pairs) and MRTK (7 pairs, 8 unmatched plasma), and plasma from 20 healthy donors were profiled by enzymatic WGMS at 30X target coverage. ichorCNA was used to detect ctDNA based on CNAs. Differentially methylated regions (DMRs) were identified using R methylKit. Fragment length and end motif frequencies were analyzed in WT and healthy plasma. Machine learning prediction models were built using cross-validation to classify healthy, WT and MRTK plasma. Results: ctDNA was detected based on aneuploidy in 8/14 (53%) patients with WT and only 3/15 (20%) with MRTK. Matched WT plasma and tumors had 85.7% agreement in detecting 1p loss, and 71.4% agreement in detecting 1q gain or 16q loss. For matched MRTK samples, 5/7 tumor and 1/7 plasma had detectable INI1 loss. Compared to healthy donor plasma, WT tissue had 2146 DMRs and MRTK tissue had 65 DMRs. Separate unsupervised analyses using these DMRs clustered WT and MRTK plasma with their respective tumor types compared to healthy controls, including 3 WT and 7 MRTK plasma where ctDNA was not detected based on aneuploidy. Generalized linear models (GLM) were trained to differentiate WT and MRTK tissue based on these DMRs and classified 12/14 WT and 11/15 MRTK plasma as their assigned diagnoses. WT cell-free DNA also had higher frequency of short fragments and distinct end motif profiles than healthy plasma (p<0.01). GLMs that were trained to differentiate WT from healthy plasma using DMRs, fragment lengths and end motifs predicted tumor signals in 14/14 WT plasma. Conclusions: Methylation profiling improves ctDNA detection in WT and MRTK and can differentiate these diagnoses in most cases. For WT, using fragment data improved sensitivity of ctDNA detection in cases where detecting aneuploidy was insufficient. Our findings support further development of WGMS for ctDNA detection and minimally invasive diagnosis of pediatric RTs. Citation Format: Nensi M Ruzgar, Cora Ricker, Kelly Klega, Catherine Clinton, Aleksi Halme, Natalie Greenwood, Laura Madanat-Harjuoja, Elizabeth A Mullen, Brian D Crompton. Multi-omic profiling of plasma circulating tumor DNA (ctDNA) detects and differentiates Wilms tumors and malignant rhabdoid tumors of the kidney [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Discovery and Innovation in Pediatric Cancer— From Biology to Breakthrough Therapies; 2025 Sep 25-28; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2025;85(18_Suppl_2): nr A013.