An Alternatively Spliced Gain-of-Function NT5C2 Isoform Contributes to Chemoresistance in Acute Lymphoblastic Leukemia

Abstract

Relapsed or refractory B-cell acute lymphoblastic leukemia (B-ALL) is a major cause of pediatric cancer-related deaths. Relapse-specific mutations do not account for all chemotherapy failures in B-ALL patients, suggesting additional mechanisms of resistance. By mining RNA sequencing datasets of paired diagnostic/relapse pediatric B-ALL samples, we discovered pervasive alternative splicing (AS) patterns linked to relapse and affecting drivers of resistance to glucocorticoids, antifolates, and thiopurines. Most splicing variations represented cassette exon skipping, "poison" exon inclusion, and intron retention, phenocopying well-documented loss-of-function mutations. In contrast, relapse-associated AS of NT5C2 mRNA yielded an isoform with the functionally uncharacterized in-frame exon 6a. Incorporation of the 8-amino acid sequence SQVAVQKR into this enzyme created a putative phosphorylation site and resulted in elevated nucleosidase activity, which is a known consequence of gain-of-function mutations in NT5C2 and a common determinant of 6-mercaptopurine resistance. Consistent with this finding, NT5C2ex6a and the R238W hotspot variant conferred comparable levels of resistance to 6-mercaptopurine in B-ALL cells both in vitro and in vivo. Furthermore, both NT5C2ex6a and the R238W variant induced collateral sensitivity to the inosine monophosphate dehydrogenase inhibitor mizoribine. These results ascribe to splicing perturbations an important role in chemotherapy resistance in relapsed B-ALL and suggest that inosine monophosphate dehydrogenase inhibitors, including the commonly used immunosuppressive agent mycophenolate mofetil, could be a valuable therapeutic option for treating thiopurine-resistant leukemias. Significance: Alternative splicing is a potent mechanism of acquired drug resistance in relapsed/refractory acute lymphoblastic leukemias that has diagnostic and therapeutic implications for patients who lack mutations in known chemoresistance genes.

Graphical abstract

Figure 1.

Splicing signature of relapsed leukemias. A, Oncoprint showing distribution of acquired mutations in known r/r genes in 48 paired TARGET samples. The blue rectangle on the right marks samples with no identifiable relapse-specific mutations. B, Oncoprint showing distribution of mutations in select SF genes in 48 TARGET patients under investigation. C, Gene set enrichment analysis of transcripts comprising one of the unsupervised clusters. Green rectangles, Gene Ontology categories involving RNA splicing. D, Heatmap showing changes in splicing in a cluster of 17 relapse vs. diagnostic B-ALL samples (light blue box). The dendrograms on top represent the results of unsupervised clustering without excluding samples with known relapse-specific mutations. The adjacent sample PARLAF with several profoundly mis-spliced transcripts is shown for comparison. Red and blue, increases/decreases in inclusion of exonic segments, respectively.

Figure 2.

LOF AS events affecting CREBBP and FPGS transcripts. A, Box plot showing percentages of reads connecting CREBBP exon 24 to exon 27 in diagnosis/relapse pairs from Fig. 1D. PSI, percent spliced-in. B, Sashimi plots visualizing the same CREBBP events in a representative PAPZST sample. C, RT-PCR analysis of CREBBP transcripts in 697 B-ALL cells transfected with exons 25- and 26-specific single-guide RNAs packaged into Cas9 particles. D, Immunoblotting analysis of the same samples using antibodies recognizing indicated proteins and post-translational modifications of histone H3. β-Actin was used as a loading control. E, Box plot showing percentages of reads connecting the 5′ splice site in FPGS exon 8 to the downstream exon 9 in diagnosis/relapse pairs from Fig. 1D. F, Sashimi plots visualizing the same FPGS event in a representative PARLAF sample. G, RT-PCR analysis of FPGS transcripts in Nalm6 cells transfected with control or FPGS exon 8 5′ splice site-specific MO. H, Immunoblotting analysis of the same samples using an anti-FPGS antibody. I, IC₅₀ plot representing survival of these cells after exposure to increasing concentrations of MTX. FL, full-length isoform; PE, poison exon-containing isoform.

Figure 3.

The NT5C2ex6a isoform as a driver of resistance to thiopurines. A, Box plot showing percentage of read connecting NT5C2 exon 6 to exon 6a in six diagnosis/relapse pairs depicted in Fig. 1C. PSI, percent spliced-in. B, Sashimi plots visualizing the same NT5C2 events in a representative PANYGB sample. C,In vitro nucleotidase assays assessing the enzymatic activity of the canonical (WT) and E6a NT5C2 isoforms in the presence of increasing concentrations of ATP. Data are shown as the mean ± SD. Asterisks indicate statistical significance per Student t test, with P values calculated using two-way ANOVA. ^∗∗, P < 0.0058; ^∗∗∗∗, P < 0.0001. D, Expression levels of transduced NT5C2 isoforms in REH cells. WT, E6a, and R238W denote the canonical isoform, the NT5ex6a splice variant, and the R238W hotspot mutant, respectively. Empty, vector-only cells. E, The IC₅₀ plot representing survival of these cells exposed to increasing concentrations of 6-MP. F, The IC₅₀ plot representing survival of these cells exposed to increasing concentrations of mizoribine. G, Bioluminescent detection of the same cells additionally expressing the firefly luciferase gene. Cells were xenografted into NSG mice and imaged on days 10 and 14.