Outlier expression of isoforms by targeted RNA sequencing as clinical markers of genomic variants in B lymphoblastic leukemia and other tumor types

Recognition of aberrant gene isoforms indicative of underlying DNA events can impact molecular classification and risk stratification of B lymphoblastic leukemia (B-ALL). Aberrant ERG isoforms have been proposed as markers of the favorable-risk DUX4-rearranged (DUX4r) subtype while deletion-mediated IKZF1 isoforms are associated with adverse prognosis in non-DUX4r B-ALL. The high-risk IKZF1plus signature depends on gene deletions including PAX5 while intragenic PAX5 amplifications (PAX5amp) are recurrent in the provisional B-ALL with PAX5-alteration subtype. In this study, we screened for outlier expression of isoforms within targeted RNA sequencing assays designed for fusions. Outlier analysis of known and novel IKZF1, ERG, and PAX5 isoforms was 97.0% (32/33), 90% (9/10), and 100% (6/6) sensitive and 97.8% (226/231), 100% (35/35), and 88.5% (23/26) specific for IKZF1 intragenic or 3 deletions, DUX4r, and PAX5 intragenic deletions respectively, where false positives were favored to represent low-level deletions below the limit of DNA-based detection. Outlier analysis also identified putative PAX5amp cases and revealed partial tandem duplication (PTD) spanning IKZF1 N159Y in the B-ALL with mutated N159Y subtype. To demonstrate utility in other tumor types, outlier analysis was 100% (9/9) sensitive and 100% (255/255) specific for KMT2A-PTD in hematologic samples and 100% (7/7) sensitive and 100% (79/79) specific for FGFR1 tyrosine kinase domain duplication in brain tumors. These findings support the use of aberrant isoform analysis in targeted RNA sequencing data as a robust strategy for the detection of clinically significant DNA events.