Cancers 2020, 12, 723 3 of 27 Whereas spliceosome mutations were infrequent in ALL [32], BCP-ALL cells did display global aberrant splicing when compared with non-malignant controls [33,34]. Interestingly, we previously showed that aberrant splicing of folylpolyglutamate synthetase (FPGS), a key determinant of methotrexate (MTX) efficacy, is associated with MTX resistance in childhood ALL [35,36]. Notably, high levels of one particular aberration, i.e., FPGS intron 8 partial retention, were also associated with increased resistance to GCs. This suggests that leukemic cells of GC-resistant patients carry a more profound splicing dysregulation which could be exploited for therapeutic purposes.Building on these data, the aim of the current study is to characterize the global alternative splicing profiles associated with ex vivo GC resistance in childhood ALL. Specifically, we investigated differential splicing profiles in 38 primary childhood ALL samples by using RNA sequencing [37]. Finally, we tested whether resistant cell lines and primary specimens can be sensitized to GC treatment by using SF3B modulators.
Results
Differential Splicing Landscape Associated with GC Resistance in Pediatric ALLIn order to evaluate whether GC resistance is associated with specific splicing patterns in childhood ALL, we used RNA sequencing to profile transcriptomes of specimens obtained from 36 newly diagnosed and two relapsed pediatric ALL patients ( Figure 1A). This study cohort was well characterized with respect to clinical features and ex vivo drug resistance, including Dex and Pred (Supplemental Data S1). The patient specimens were classified either as GC-sensitive (N = 15) or GC-resistant (N = 23) based on ex vivo Dex and Pred LC 50 values according to the previously established cut-offs of 0.01 µg/mL and 0.1 µg/mL, respectively [9] ( Figure 1B). Furthermore, we determined the immunophenotype and genetic profile of the samples, including mutations in GR and recurrent genetic alterations associated with ALL [21,38] (Figure 1C), which allowed us to account for possible confounders in the analysis.The global differential splicing profiles of GC-sensitive and GC-resistant samples were determined using the rMATS algorithm (Figure 2A). This software identifies sequencing reads which support a certain splice event (e.g., the inclusion or skipping of a certain exon in a gene of interest) and calculates the inclusion levels (or Percentage Spliced-In, Ψ). Ψ is computed as the proportion of reads supporting the inclusion of the exon in question divided by the sum of reads supporting the inclusion and skipping of this exon. Subsequently, it compares the average Ψ values of GC-sensitive specimens with that of GC-resistant samples by computing the Inclusion Level Difference (∆Ψ) and the corresponding p-value and false discovery rate (FDR) for each splice event [37]. We found, in total, 994 significant differential splicing events (FDR < 0.05, Figure 2B and Supplemental Data S2) affecting 762 genes. Hierarchical clustering ( Figure 2C) and ...