Detection, quantification, and subtyping of adenoid cystic carcinoma (ACC) using methylation from liquid biopsies.

6118 Background: Despite the common indolent behavior of ACC, some patients will experience aggressive disease with short survival. These distinct clinical behaviors have been linked to two transcriptional profiles: ACC-I, defined by NOTCH1 activation, solid histology, and more aggressiveness, and ACC-II, characterized by predominant myoepithelial p63 expression and generally indolent behavior. DNA methylation has been widely used for tumor identification and subtyping; however, their relevance in distinguishing ACC-I and ACC-II has not yet been explored. Methods: Forty-one ACC tissue samples from the original subtype development cohort (41/54) and twenty-three cfDNA samples from a phase II trial of axitinib and avelumab in recurrent/metastatic ACC (NCT03990571) were profiled through target-enriched enzymatic methylation sequencing. ACC subtypes (ACC-I/II) in tissue samples were validated with unsupervised clustering of highly variable regions. To translate these signatures to liquid biopsy, we used the METER (METhylome AnalysER) pipeline to identify tumor-specific differentially methylated sites (DMS) and regions (DMRs) relative to 22 healthy donors, enabling ctDNA detection (METER-positive), tumor proportion score (TPS) quantification, and ACC subtype assignment in cfDNA. Methylation-derived TPS were validated with ichorCNA (copy number alterations) estimates, and cfDNA-assigned subtypes were benchmarked against matched tissue transcriptional signatures. Finally, we investigated if METER detection was associated with progression-free survival (PFS). Results: ACC tissue methylation identified two clusters that perfectly matched reference transcriptional subtypes (17/17 ACC-I and 24/24 ACC-II). ACC-I tumors were hypermethylated compared to ACC-II (11,205 hyper and 1,804 hypo DMR, p<0.05). Of 23 cfDNA samples, 15 (65.2%) were deemed as METER positive with a higher rate among ACC-I (8/9, 88.9%) compared to ACC-II (7/14, 50.0%), consistent with the more aggressive subtype of ACC-I. The median (range) estimated TPS among METER positive samples was (3.86%; range 0.7%-82%), with a numerical higher TPS among ACC-I (14.8% vs 2.0%, p=0.073). TPS had strong correlation with ctDNA estimation by ichorCNA (spearman 0.86, p<0.001). METER subtyping correctly identified ACC-subtypes in 15/15 (100%) METER positive and 4/8 (50%) METER negative samples. METER-positive patients had shorter PFS (HR = 4.32; 95%CI, 1.49–12.5, p=0.007), including among ACC-II patients (HR = 4.84; 95%CI, 1.20–19.5, p=0.026). Conclusions: Methylation-based subtyping of ACC is highly concordant with transcriptional profiles and enables detection and subtyping of ACC in liquid biopsies. Beyond subtyping, cfDNA detection may serve as an additional prognostic marker, identifying patients at high risk for progression even within traditionally indolent subgroups. Clinical trial information: NCT03990571 .