Murtaza, Muhammed

Abstract We demonstrate development and technical validation of a novel approach for analysis of plasma circulating-tumor DNA (ctDNA) for residual disease detection and monitoring that offers improvements over existing technologies. ctDNA in blood can be leveraged to monitor treatment response, detect cancer recurrence, and inform treatment decisions for cancer patients. However, accurate detection and quantification of ctDNA is technically challenging due to low concentration of plasma DNA, especially in the adjuvant setting. Most current assays rely on interrogating multiple patient-specific point mutations to achieve exquisite limits of detection but struggle with polymerase errors, sequencing artifacts and limited conversion of input DNA. To overcome these challenges, we developed an approach called Structural Variant Enrichment and Normalization (SVEN) that targets patient-specific genomic rearrangements to achieve sensitive detection and quantification of plasma ctDNA. This method leverages intrinsic features of tumor-derived structural variants (SVs) to improve enrichment and confidence of detection at low tumor fraction through multiplexed PCR and amplicon-based next generation sequencing. To evaluate the assay, we performed whole genome sequencing of DNA from two breast cancer cell lines, as well as matched FFPE tumor and buffy coat samples from 21 early-stage breast cancer patients. We identified consensus SVs, and designed multiplexed SV-specific primers for each subject. We then conducted PCR amplification of cell-free DNA followed by sequencing, amplicon identification, and quantification. Analytical validation was conducted using 284 replicates of cell line dilutions from 1% to 0.003%, and 5 to 40 ng of input. We developed multiplexed assays targeting 25 and 32 SVs across two cell lines, and 3 - 43 SVs across breast cancer patients (median 23 per patient). 91% and 42% of PCR targets were successfully validated for cell lines and FFPE samples respectively. Cell line dilutions demonstrated a sensitivity of 89% at 0.003% tumor fraction with 10 ng input or more. Tumor fractions measured using SVEN agreed with known values (R2 = 0.989). Application of validated panels to patient plasma samples and tumor/germline dilutions is ongoing. SVEN is a novel approach for tumor-guided ctDNA detection and quantification with potential relevance for residual disease detection. On-going studies are investigating the performance of this approach for ctDNA detection in patients with early and locally advanced breast cancer treated with neoadjuvant therapy. Citation Format: Bradon R. McDonald, Kirsten L. Dennison, Amanda L. Schussman, Stephanie M. McGregor, Barbara A. Pockaj, Muhammed Murtaza. Personalized minimal residual disease detection using tumor-derived structural variants in cell-free DNA [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2416.

Genome-wide fragmentation patterns in cell-free DNA (cfDNA) in plasma are strongly influenced by cellular origin due to variation in chromatin accessibility across cell types. Such differences between healthy and cancer cells provide the opportunity for development of novel cancer diagnostics. Here, we investigated whether analysis of cfDNA fragment end positions and their surrounding DNA sequences reveals the presence of tumor-derived DNA in blood. We performed genome-wide analysis of cfDNA from 521 samples and analyzed sequencing data from an additional 2147 samples, including healthy individuals and patients with 11 different cancer types. We developed a metric based on genome-wide differences in fragment positioning, weighted by fragment length and GC content [information-weighted fraction of aberrant fragments (iwFAF)]. We observed that iwFAF strongly correlated with tumor fraction, was higher for DNA fragments carrying somatic mutations, and was higher within genomic regions affected by copy number amplifications. We also calculated sample-level means of nucleotide frequencies observed at genomic positions spanning fragment ends. Using a combination of iwFAF and nine nucleotide frequencies from three positions surrounding fragment ends, we developed a machine learning model to differentiate healthy individuals from patients with cancer. We observed an area under the receiver operative characteristic curve (AUC) of 0.91 for detection of cancer at any stage and an AUC of 0.87 for detection of stage I cancer. Our findings remained robust with as few as 1 million fragments analyzed per sample, demonstrating that analysis of fragment ends can become a cost-effective and accessible approach for cancer detection and monitoring.

Abstract Purpose Accurately predicting the clinical breast cancer subtypes could be extremely helpful for radiologists, pathologists, surgeons, and clinicians and inform future treatment prediction algorithms. Therefore, we evaluate and compare the accuracy of radiomic features extracted from contrast enhanced mammography (CEM) and magnetic resonance imaging (MRI) scans to make predictions to subtypes of breast cancer. Methods This HIPAA-compliant prospective single institution study was approved by the local institutional review board with written informed consent. Women with breast tumors 2 cm or larger underwent dynamic contrast-enhanced MRI and/or CEM for surgical staging. Semi-manual regions of interest were drawn by radiologist using Cancer Imaging Phenomics Toolkit (CaPTk). Radiomic features were obtained using PyRadiomics and MR- and CEM- based classification models were built on a low-dimensional representation of the features obtained via kernel principal component analysis. We subscribed to an ensemble tree-based learning approach called extremely randomized trees (ERT) to predict cancer subtypes captured via immunohistochemistry markers. Results For MRI analysis, 124 women with newly diagnosed breast cancer were included in the analysis comprising 49 HR+HER2-, 37 HR+HER2+, 11 HR-HER2+, and 27 HR-HER2- cases. For CEM analysis, models were built using data from 170 female patients including 74 HR+HER2-, 41 HR+HER2+, 14 HR-HER2+, and 43 HR-HER2-. CEM based model resulted in accuracies of 55%, 72%, 88%, and 71% respectively for HR+HER2-, HR+HER2+, HR-HER2+, and HR-HER2- whereas MRI based model alone led to accuracies of 54%, 62%, 89%, and 76% respectively for HR+HER2-, HR+HER2+, HR-HER2+, and HR-HER2-. Conclusions Radiomic features extracted from CEM and MR were strong predictors of breast cancer subtypes with CEM-based radiomic features performing slightly better, though not statistically significantly better (p = 0.82), than its MRI counterpart.

e16185 Background: Comprehensive sequencing efforts have identified fibroblast growth factor receptor (FGFR) translocations in numerous cancers. FGFR2 fusions were recently identified as an actionable therapeutic target in ̃17% of cholangiocarcinoma patients, predominantly found in patients with intrahepatic cholangiocarcinoma. Genomic partners for FGFR fusions are variable. In addition, obtaining high yield tissue biopsies for CCA remains challenging and tumor tissue available for molecular analyses is scarce. When combined, these two factors make detection of FGFR fusions in CCA challenging. An accurate molecular assay to detect FGFR fusions in CCA could improve utilization of FGFR-targeted therapies. Methods: The objective of our study was to develop an approach for targeted RNA analysis to improve sensitivity for FGFR fusion detection from limited tumor material. We adapted the sensitive targeted digital sequencing (TARDIS) approach recently developed in our lab for analysis of RNA. This approach utilizes a combination of open-ended targeted amplification followed by ligation, enabling detection of tyrosine kinase fusions without prior knowledge of the precise sequence of the fusion breakpoint or identity of the fusion partner. For evaluation of analytical performance, we analyzed RNA from a CCA organoid model with a known FGFR2 fusion, and Seraseq Fusion RNA Mix v3 Reference Material with two known FGFR3 fusions. Results: Without prior knowledge the partner or coordinates of the fusion breakpoint, we detected an FGFR2-KIF5C fusion in RNA from a CCA organoid model. The fusion breakpoint coordinates predicted by TARDIS-RNA were validated by comparison with RNA-Seq. To assess sensitivity and quantitative performance of the assay, we analyzed a serial dilution of the reference RNA sample from Seraseq with concentration of RNA molecules representing candidate fusions verified by digital PCR. Using TARDIS-RNA to analyze replicates with 5 ng RNA input, we were able to detect candidate fusions represented by as few as 2.7 fusion molecules on average. Concentrations of fusion molecules measured using the two methods were highly correlated (R = 0.96). Conclusions: These results demonstrate sensitivity and quantitative performance of a targeted RNA sequencing assay to detect and quantify FGFR fusions in tumor specimen from intrahepatic cholangiocarcinoma. On-going work is focused on further evaluating assay performance and characterizing FGFR fusions using additional tissue specimen.

Abstract Background: Accurate circulating biomarkers for detecting residual disease can help guide therapy decisions, particularly in early-stage cancer patients. However, currently available methods lack the sensitivity required to confidently assess the presence of residual disease in patients with low tumor burden. To address this need, we have developed TARDIS (Targeted Digital Sequencing), a personalized, multiplexed amplicon sequencing method capable of tracking as many as 100 or more mutations simultaneously. Methods: We obtained tumor biopsies and longitudinal plasma samples from patients with early-stage breast cancer, glioblastoma, and pancreatic cancer. Each tumor biopsy was analyzed whole-exome sequencing. Founder mutations were selected, accounting for copy number alterations (analyzed using sequenza) and a consensus allele fraction approach that combined pyclone and custom in-house methods. Patient-specific TARDIS primers were designed to detect these mutations in plasma cfDNA. Error suppression in TARDIS was achieved using a combination of unique molecular identifiers and fragment sizes to group sequencing reads into read families. Results: In 33 patients with early-stage breast cancer treated with neoadjuvant therapy, we targeted between 3 and 116 (mean 30) mutations per patient and analyzed between 1 and 4 longitudinal plasma samples using TARDIS. Prior to treatment, we detected ctDNA in 100% patients with Stage I-III breast cancer (n=32, 95% CI= 89%-100%). We detected tumor-specific mutations in 100% of baseline breast cancer plasma samples. After completion of neoadjuvant therapy and before surgery, ctDNA levels were significantly lower in patients with pathologic complete response (pathCR, no evidence of disease at surgery) compared to patients with residual disease (median tumor fractions 0.003% and 0.017%, respectively, p=0.0058, AUC=0.83). Conclusions: TARDIS enables highly sensitive detection of ctDNA in patients with nonmetastatic cancers. Analysis of longitudinal plasma samples using TARDIS holds promise for personalizing the extent of treatment in patients with curable disease. Multiple clinical validation studies across cancer types are ongoing to define quantitative thresholds for changes in ctDNA levels that could improve clinical decision making. Citation Format: Bradon R. McDonald, Tania Contente-Cuomo, Stephen-John Sammut, Michelle D. Stephens, Ahuva Odenheimer-Bergman, Brenda Ernst, Nieves Perdigones, Suet-Feung Chin, Maria Farooq, Rosa Mejia, Patricia A. Cronin, Karen S. Anderson, Heidi E. Kosiorek, Donald W. Northfelt, Ann E. McCullough, Bhavika K. Patel, Jeffrey N. Weitzel, Thomas P. Slavin, Carlos Caldas, Barbara A. Pockaj, Muhammed Murtaza. Personalized monitoring of treatment response using Targeted Digital Sequencing of circulating tumor DNA [abstract]. In: Proceedings of the AACR Special Conference on Advances in Liquid Biopsies; Jan 13-16, 2020; Miami, FL. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(11_Suppl):Abstract nr A51.

A robust personalized ctDNA test, TARDIS, achieves high accuracy for residual disease after completion of neoadjuvant therapy.

Abstract Background: Circulating tumor DNA (ctDNA) analysis holds potential for minimal residual disease (MRD) detection in early stage breast cancer. However, sensitivity for MRD is limited due to low ctDNA levels in early stage patients and limited blood volumes. Loss of input DNA during library preparation, limited multiplexing or low sensitivity of current molecular methods further limit accuracy. To address this gap, we have developed TARgeted DIgital Sequencing (TARDIS), a novel method for simultaneous analysis of multiple patient-specific mutations in plasma DNA. Methods: Using tumor exome sequencing, we identify and prioritize somatic founder mutations, design nested primers and evaluate them for multiplex performance. Using 5-10 ng input plasma DNA, we perform 1) targeted linear pre-amplification to improve downstream molecular conversion, 2) single-stranded adapter ligation to incorporate unique molecular identifiers (UMIs) and 3) targeted PCR to prepare sequencing-ready libraries. The resulting sequencing reads have fixed target-specific ends and variable ligation ends. We utilize fragment size and UMIs to group sequencing reads into read families. To ensure specificity, we require targeted mutations are supported by 2 or more read families. Results: To assess analytical performance, we targeted 8 mutations in cell-free DNA reference samples with 0.25%-2% mutation allele fractions (AFs). Precision across 7-16 replicates at each AF level agreed with expectations of Poisson distribution, demonstrating effective analysis of ˜70% of input DNA. At 2%, 1%, 0.5% and 0.25% AFs, variant-level sensitivity was 96.4%, 96.4%, 91.1% and 65.8%, approaching the theoretical limit given input DNA. At 0.25% AF, 3-7 mutations were detected per sample, achieving 100% sample-level sensitivity. In 16 wild-type replicates, no targeted mutations were called (100% specificity). Averaging multiple mutations improved precision in sample-level AF estimates. Mean AFs from 8 mutations for the 2% sample were 2.34%-2.80% (5.8% CV). In 6 patients with breast cancer treated with neoadjuvant therapy (NAT), we analyzed 8-18 patient-specific mutations (mean 11.8). Before treatment, ctDNA was detected in 5/6 patients at mean AFs of 0.02%-1.19% (mean 0.40%), supported by 2-10 mutations (mean 5.6). Of these 5 patients, 4 had residual disease after NAT and ctDNA was detected pre-operatively or during NAT in 3/4 patients. 1 patient achieved pathological Complete Response and ctDNA was undetectable after NAT. Conclusions: Preliminary results suggest TARDIS enables accurate MRD detection after neoadjuvant therapy in patients with early stage breast cancer. On-going work is expanding this analysis to include additional patients and investigate the clinical validity of peri-operative ctDNA monitoring. Summary of clinical resultsPatientPre-NAT Stage (TNM)SubtypeNo. of Mutations TargetedBaseline ctDNA (AF%, No. of Mutations)ctDNA after or during NAT (AF%, No. of Mutations)Residual Tumor (TNM)1T3 N1ER+ PR+ HER2-8+ (0.02%, 2)-T2 N12T3 N0TNBC12+ (0.29%, 6)+ (0.01%, 1)T1a N03T2 N1TNBC18+ (1.19%, 10)+ (0.01%, 1)T1mi N04T3 N1TNBC10+ (0.02%, 3)+ (0.05%, 3)T3 N15T2 N0TNBC9+ (0.46%, 7)-pathCR6T1c N1TNBC14--pathCR Citation Format: McDonald BR, Contente-Cuomo T, Sammut S-J, Ernst B, Odenheimer-Bergman A, Perdigones N, Chin S-F, Farooq M, Cronin PA, Anderson KS, Kosiorek H, Northfelt D, McCullough A, Patel B, Caldas C, Pockaj B, Murtaza M. Multiplexed targeted digital sequencing of circulating tumor DNA to detect minimal residual disease in early and locally advanced breast cancer [abstract]. In: Proceedings of the 2018 San Antonio Breast Cancer Symposium; 2018 Dec 4-8; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2019;79(4 Suppl):Abstract nr P4-01-21.

TPS532 Background: Recent studies have shown that FGFR2 fusions may represent recurrent (prevalence ~10-15% in larger cohorts) driver events amenable to therapeutic targeting in patients with advanced, intrahepatic cholangiocarcinoma. Methods: This single arm pilot study (NCT02265341) is being conducted at 3 centers (Mayo Clinic Arizona, Mayo Clinic Florida and Mayo Clinic Rochester; enrollment initiated December 2014). Next-generation sequencing based assays and a FISH-break apart assay are used to identify tumors with FGFR2 fusions. Patients must additionally have histologically/cytologically confirmed cholangiocarcinoma, ECOG performance score 0-2, measurable disease, preserved hematologic/renal and hepatic function and absence of cardiac events/stroke in the preceding 6 months. Ponatinib is administered at 45 mg once daily. Patients will be treated until disease progression, intolerable toxicity or patient decision to not pursue further therapy. The primary objective of the study is to estimate the clinical benefit rate (defined as composite of partial response rate, complete response rate and stable disease rate ≥ 4 months; as defined by RECIST criteria v1.1). Secondary endpoints include estimation of progression-free survival, overall survival, CA19-9 response rate (defined as > 50% reduction from baseline) and evaluation of the overall adverse event profile of ponatinib in this patient population. Translational objectives include serial circulating DNA (ct-DNA) monitoring, target knockdown assessments in tumor tissue (p-FGFR2, p-FRS2, p-Akt and p-ERK) and evaluation of mechanisms of resistance through comparison of baseline and progression biopsy samples using multi-platform sequencing (exome, RNA-Seq and methyl-Seq). Additionally, quality of life assessments will be evaluated using EORTC-QLQ-C30 and BIL21 tools. A total of 30 patients will be enrolled into the study (12 of 30 patients have been enrolled). The trial is designed using an underlying assumption of clinical benefit rate (CBR) of 5% with significance level of 0.15 and 80% power to detect a 20% CBR. Clinical trial information: NCT02265341.

293 Background: Cholangiocarcinoma (CCA) is a deadly cancer with ~20% five year survival. Obtaining tissue to confirm diagnosis of CCA is often challenging. This also limits the use of tumor genotyping that could guide potential molecularly targeted treatment (such as FGFR inhibitors for a subset of cases). We investigated whether circulating tumor DNA (ctDNA) analysis in plasma is feasible for noninvasive tumor genotyping in patients with advanced CCA. Methods: We collected and processed plasma samples for ctDNA analysis from 17 patients with CCA at presentation and from 11 healthy volunteers. Following cell-free DNA extraction and quality assessment using digital PCR, we performed low-pass, shallow whole genome sequencing (sWGS). We analyzed sequencing data to identify copy number aberrations and quantify tumor fraction in plasma using recently published computational approaches. Results: Median total cfDNA concentration in patients with CCA was 7,003 genome equivalents (GEs)/mL plasma (range: 1,512- 107,707 GEs/mL) and significantly higher than healthy volunteers (two tailed t-test p < 0.04, median: 1,658 GEs/mL, range: 661-4,938 GEs/mL). We successfully sequenced all tumor libraries, generating a median of 173 million paired-end reads. ctDNA was detectable in 10/17 samples using sWGS. In these 10 samples, median tumor fraction in plasma DNA was 7.1% (range: 3.9%-44.6%). We found recurrent copy number aberrations expected in CCA including 1q gain in 80%, 7p gain in 50%, 1p loss in 70% and 3p loss in 50% of cases. Conclusions: ctDNA analysis for noninvasive tumor genotyping is feasible in patients with advanced cholangiocarcinoma. On-going work is focused on expanding these results using targeted and whole-exome sequencing.