CircRNAome of Childhood Acute Lymphoblastic Leukemia: Deciphering Subtype-Specific Expression Profiles and Involvement in TCF3::PBX1 ALL

Abstract

Childhood B-cell acute lymphoblastic leukemia (B-ALL) is a heterogeneous disease comprising multiple molecular subgroups with subtype-specific expression profiles. Recently, a new type of ncRNA, termed circular RNA (circRNA), has emerged as a promising biomarker in cancer, but little is known about their role in childhood B-ALL. Here, through RNA-seq analysis in 105 childhood B-ALL patients comprising six genetic subtypes and seven B-cell controls from two independent cohorts we demonstrated that circRNAs properly stratified B-ALL subtypes. By differential expression analysis of each subtype vs. controls, 156 overexpressed and 134 underexpressed circRNAs were identified consistently in at least one subtype, most of them with subtype-specific expression. TCF3::PBX1 subtype was the one with the highest number of unique and overexpressed circRNAs, and the circRNA signature could effectively discriminate new patients with TCF3::PBX1 subtype from others. Our results indicated that NUDT21, an RNA-binding protein (RBP) involved in circRNA biogenesis, may contribute to this circRNA enrichment in TCF3::PBX1 ALL. Further functional characterization using the CRISPR-Cas13d system demonstrated that circBARD1, overexpressed in TCF3::PBX1 patients and regulated by NUDT21, might be involved in leukemogenesis through the activation of p38 via hsa-miR-153-5p. Our results suggest that circRNAs could play a role in the pathogenesis of childhood B-ALL.

Keywords: NUDT21; RNA-binding protein; TCF3::PBX1; acute lymphoblastic leukemia; childhood B-ALL; circular RNA.

Figure 1

Overview of the circRNAome in childhood B-ALL. (A) Number of circRNAs identified per chromosome before and after bioinformatic and biochemical filtering. (B) The mean size, average number of exons as well as the mean number of circRNA isoforms per gene of circRNAs identified. (C) Number of circRNAs identified in every ALL subtype and controls.

Figure 2

Characterization of circRNA profiles. (A) Unsupervised clustering analysis (n = 112) using t-SNE shows the stratification of childhood ALL subtypes using the 500 most variable genes and circRNAs. (B) Total number of circRNAs differentially expressed in discovery (Dis) and replication (Rep) cohorts and integration of the 2 cohorts (Int). (C) Venn diagram of overexpressed circRNAs for each subtype. (D) Venn diagram of underexpressed circRNAs for each subtype.

Figure 3

CircRNAome in TCF3::PBX1 ALL. (A) Hierarchical clustering using the circRNA signature (156 overexpressed and 134 underexpressed circRNAs) to cluster B-ALL subtypes and new patients with TCF3::PBX1 ALL. (B) FPKM values of NUDT21 in the discovery and validation cohorts were extracted from the transcriptome data and log transformed. Only significant p-values (<0.05) are indicated (ANOVA test, corrected for multiple testing with Bonferroni). (C) RT-qPCR analysis of 10 circRNAs and their cognate pre-mRNAs in 697 control cells and NUDT21-KD cells. Experiments were performed in triplicate. Comparisons assessed by t-test; * p-value < 0.05; ** p-value < 0.005; *** p-value < 0.0005.

Figure 4

Functional characterization of circBARD1 by loss-of-function studies. (A) Expression of circBARD1 and the linear BARD1 transcript in circBARD1 silenced cell line vs. control. Effects of the silencing of circBARD1 on (B) migration, (C) proliferation, and (D) oxidative stress induction. (E) Expression of circBARD1 in the cytoplasmic fraction of 697 cells. (F) Effects of the silencing of circBARD1 on protein phosphorylation. Experiments were performed in triplicate. Comparisons assessed by t-test; * p-value < 0.05; ** p-value < 0.005.