Alternative RNA Splicing Defects in Pediatric Cancers: New Insights in Tumorigenesis and Potential Therapeutic Vulnerabilities

Abstract

Background:

Compared to adult cancers, pediatric cancers are uniquely characterized by a genomically stable landscape and lower tumor mutational burden. However, alternative splicing, a global cellular process that produces different mRNA/protein isoforms from a single mRNA transcript, has been increasingly implicated in the development of pediatric cancers.

Design:

We review the current literature on the role of alternative splicing in adult cancer, cancer predisposition syndromes, and pediatric cancers. We also describe multiple splice variants identified in adult cancers and confirmed through comprehensive genomic profiling in our institutional cohort of rare, refractory and relapsed pediatric and adolescent young adult cancer patients. Finally, we summarize the contributions of alternative splicing events to neoantigens and chemoresistance and prospects for splicing-based therapies.

Results:

Published dysregulated splicing events can be categorized as exon inclusion, exon exclusion, splicing factor upregulation, or splice site alterations. We observe these phenomena in cancer predisposition syndromes (Lynch syndrome, Li-Fraumeni syndrome, CHEK2) and pediatric leukemia (B-ALL), sarcomas (Ewing sarcoma, rhabdomyosarcoma, osteosarcoma), retinoblastoma, Wilms tumor, and neuroblastoma. Within our institutional cohort, we demonstrate splice variants in key regulatory genes (CHEK2, TP53, PIK3R1, MDM2, KDM6A, NF1) that resulted in exon exclusion or splice site alterations, which were predicted to impact functional protein expression and promote tumorigenesis. Differentially spliced isoforms and splicing proteins also impact neoantigen creation and treatment resistance, such as imatinib or glucocorticoid regimens. Additionally, splice-altering strategies with the potential to change the therapeutic landscape of pediatric cancers include antisense oligonucleotides, adeno-associated virus gene transfers, and small molecule inhibitors.

Conclusions:

Alternative splicing plays a critical role in the formation and growth of pediatric cancers, and our institutional cohort confirms and highlights the broad spectrum of affected genes in a variety of cancers. Further studies that elucidate the mechanisms of disease-inducing splicing events will contribute toward the development of novel therapeutics.

Introduction

By 2040, approximately 27.5 million new cancer cases, with nearly 200,000 in children, and 16.3 million cancer deaths are expected annually worldwide¹. Cancer is a leading cause of death due to disease in children^2–4, and there remains a need for improved understanding of childhood cancer biology and the development of more effective therapeutics.

Genomic alteration is an initiating event in tumorigenesis. In comparison to adult cancers, pediatric cancers are typically associated with epigenetic dysregulation rather than numerous genetic alterations^5–9. Some pediatric cancers are characterized by copy number variants or alterations in master regulatory factors, such as transcription factors or epigenetic regulators, that affect expression of a broad range of downstream genes^6,10–13. Between 6 and 18% of pediatric cases arise in the context of a genetic predisposition syndrome¹⁴. Recent reports also suggest that alterations in pre-mRNA splicing contribute to tumorigenesis^15,16.

Splicing has been repeatedly described as a critical driver in adult cancers. The Cancer Genome Atlas (TCGA) studies on soft tissue sarcomas and bladder urothelial cancer have identified well over 35,000 alternative splicing events in thousands of analyzed genes^17,18. The extent to which splicing plays a role in pediatric cancers, however, is largely unknown.

Herein, we review what is currently known about alternative mRNA splicing in pediatric cancers. We categorize various splicing events in tumorigenesis and identify commonalities between cancer types. We also highlight recent splicing variations in a series of patients at our own institution, emphasizing how the study of alternative splicing mechanisms is essential to better understand pediatric oncogenesis. Furthermore, we discuss opportunities for the development of novel therapeutics targeting splicing-driven cancer.

Normal Alternative Splicing

Splicing is an essential cellular mechanism in the nucleus that results in distinct protein isoforms that can be structurally and functionally different from the parent protein. In constitutive splicing, introns are removed, and exons are retained sequentially. During alternative splicing, sequences are selectively excluded or included. Alternative splicing is categorized into distinct events—exon skipping, mutually exclusive exons, alternative 3’ and 5’ splice site selection, and intron retention¹⁹. The process of splicing is sequential, requiring the recruitment and assembly of various ribonucleoprotein complexes and splicing factors (Figure 1).

Figure 1: Normal splicing through spliceosome assembly.

Pre-mRNA splicing is a sequential, two-step transesterification process that requires the recruitment of various ribonucleoprotein complexes (U1, U2, U2AF, U4, U5, U6) and RNA binding proteins (RBPs), such as hnRNPs or SRSFs. Splice sites on the 5’ (donor – GU) and 3’ (acceptor – AG) ends of the introns dictate the exact locations where the spliceosome, the macromolecular machine that mediates this process, cuts the pre-mRNA. Following the binding of the U1 snRNP at the 5’ intronic splice site and the U2 snRNP at the branch site (denoted by adenosine near the 3’ end of the intron), the U4-U6 snRNPs are recruited to form the mature spliceosome. Next, the removal of the 5’ intronic end and subsequent ligation of the 5’-GU site to the branch site creates a lariat. Finally, cleavage of the 3’ intronic splice site removes the intron completely, and the exons are ligated together to form a mature RNA transcript.

Splicing in Adult Cancers: A Broad Overview

Splicing defects are a feature of many adult cancers^15,16,20. To provide context for the pediatric landscape, we briefly examine a selection of dysregulatory splicing events in the four most prevalent adult cancers: prostate, breast, lung, and colorectal²¹ (Table 1).

Table 1:

Key Aberrant Splicing Events in Prevalent Adult Cancers

Adult Cancer Splicing Events
Exon Inclusion
Description	Outcome	References
Inclusion of exon 4A or 4B of KRAS gene in colorectal cancer	Pro-apoptotic (exon 4A) or anti-apoptotic effect (exon 4B) with overall effects on cell migration and tumorigenesis	22,23,25,26,28
Exon Exclusion
Partial or complete skipping of BRCA exon 11 in breast cancer	Dysfunction of tumor suppressor and drug resistance to PARP inhibitors and cisplatin	30
Skipping of part of KLF6 exon 2 in breast cancer	Localization to cytoplasm instead of nucleus, leading to cancer cell proliferation and metastasis	31,32
Skipping of exons 4–8, containing the ligand-binding domain, in AR in prostate cancer	Produce AR-V7 alternative transcript → Constitutive activation of androgen receptor in absence of androgen and cell cycle dysregulation Produce AR-V9 alternative transcript (different cryptic terminal exons than AR-V7) → Acquired drug resistance against abiraterone	33–36
Skipping exon 11 of INSR gene in breast cancer, prostate cancer, non-small cell lung cancer, colorectal cancer	Increased cancer cell proliferation Drug resistance (colorectal cancer) Increased binding of IGF2, which promotes cancer cell proliferation and drug resistance (breast cancer specifically)	43–46
Splicing Factor Upregulation
SRSF1 overexpression in lung cancer (induced by MYC overexpression)	More aggressive cancer phenotype, increased oncogenic potential, drug resistance	47,48
SRSF1, SRSF3, hnRNPA1/A2/I overexpression in breast cancer	Promote oncogenesis, drug resistance Differential expression of PKM2 isoform→ tumorigenic, Warburg effect	29,49–51
Splice Site Alterations
5’ or 3’ splice site mutations in MET exon 14	Increased stability and oncogenic potential of METΔ14 isoform	52,53

AR: androgen receptor, INSR: insulin receptor

Exon Inclusion Events

KRAS has two isoforms (KRAS4A, KRAS4B) characterized by the inclusion of alternative fourth exons. Perhaps influenced by epigenetic modifications, the ratio of KRAS4A:4B expression is altered in colorectal cancer and affects apoptosis, cell migration, and overall tumorigenesis^22–28.

Exon Exclusion Events

Aberrant exon exclusion events are numerous in adult cancers, especially in breast cancer²⁹. BRCA1 exon 11 can be partially or completely skipped, resulting in dysfunction of the tumor suppressor and drug resistance to common therapies³⁰. In KLF6, another tumor suppressor, skipping of the nuclear localization signal creates multiple isoforms that aberrently localize to the cytoplasm, inducing proliferation and metastasis^31,32. In prostate cancer, the androgen receptor (AR)’s ligand-binding domain is encoded in exons 4–8, and skipping of this region creates isoforms that constitutively activate the AR and dysregulate the cell cycle (AR-V7) or promote abiraterone resistance (AR-V9)^33–36.

Notably, the insulin receptor (INSR) is alternatively spliced in multiple cancers, including breast, prostate, non-small cell lung, and colorectal cancer^37–41. INSR has two isoforms—INSR-A (excludes exon 11) and INSR-B (includes exon 11)—that are differentially expressed in a cell- and development-specific manner⁴². In adults, INSR-B is preferentially expressed in highly metabolic tissues, whereas INSR-A is predominant in the embryo and in sarcomas because of its high-affinity binding of insulin-like growth factor II (IGF-II), which promotes increased cancer cell growth, migration, and viability^43,44. INSR exon 11 exclusion facilitates cell proliferation and contributes to drug resistance^44–46.

Splicing Factor Upregulation

Overexpression of splicing factors can increase splicing events. In lung cancer, upregulated MYC leads to overexpression of serine-rich splicing factor 1 (SRSF1), which causes a more aggressive cancer phenotype^47,48. In breast cancer, SRSFs and heterogenous nuclear ribonucleoprotein (hnRNP) proteins, known to promote oncogenesis and drug resistance, are commonly upregulated^29,49,50. hnRNPA1/A2/I all contribute to the differential expression of pyruvate kinase (PKM), which has two mutually exclusive alternative exons⁵¹. PKM2 expression is favorable as a tumorigenic isoform and is necessary for the Warburg effect⁵¹.

Splice Site Alterations

Splice site alterations can introduce a premature termination codon (PTC) that truncates the mRNA transcript (functionally similar to a nonsense or other classical truncating mutation) or promote alternative sequence inclusion/exclusion. In lung cancer, where splice site alterations are common, the proto-oncogene MET can harbor mutations in both 5’ or 3’ splice sites surrounding exon 14 that result in exon exclusion. The alternative MET isoforms lack a key phosphorylation site to bind CBL, a ubiquitin ligase that degrades MET^52,53. The METΔ14 splice variant is more stable and potentiates oncogenesis, making it a therapeutic target for tyrosine kinase inhibitors⁵².

Cancer Predisposition Syndromes

Pediatric cancer patients may have a cancer predisposition syndrome, which originates from germline pathogenic variants associated with splicing defects^54,55. Patients with Lynch syndrome have a predisposition to colon and endometrial cancers, some of which can occur during childhood. Many cases are associated with germline alterations of the mismatch repair genes MSH2 and MLH1 that result in low or absent expression of these proteins⁵⁶. An intronic variant in MSH2 (c.212–553_c.212–479) creates a novel splice donor site that introduces a PTC⁵⁷. Additionally, splice site alterations in MSH2 (c.211+1G>C) and MLH1 (c.1731+5G>A) can induce nonsense-mediated decay (NMD) of the mRNA transcripts⁵⁸. Other studies have shown that several MLH1 variants in the exon 10 coding region lead to aberrant splicing⁵⁹. Additionally, the key genes associated with Lynch syndrome (MLH1, MSH2, MSH6 and PMS2) have been reported to be highly enriched for splicing alterations⁶⁰.

Li-Fraumeni syndrome is characterized by alterations in the tumor suppressor TP53, increasing the risk for childhood/young adulthood sarcomas, breast cancer, leukemia, etc.^61–64. Familial studies have identified both intronic and exonic splice site alterations that generate PTCs and aberrant TP53 transcripts^65,66. Loss of the initiation codon, synonymous variants, or missense variants in TP53 that localize in exonic splicing enhancers can contribute to the creation of novel splice sites and the expression of different TP53 isoforms^67,68.

Additionally, alterations in the Checkpoint Kinase 2 (CHEK2) gene on chromosome 22, encoding the CHK2 protein, are associated with an increased risk of inherited breast cancers and therapy resistance^69,70. In an analysis of 13 meningiomas, frameshift alterations resulting in alternative splicing and lack of the CHK2 kinase domain are hypothesized to contribute to the loss of DNA repair mechanisms and increased chromosomal instability⁷¹. Within our institutional cohort of rare, refractory, and relapsed pediatric and adolescent young adult (AYA) cancer patients on whom comprehensive genomic profiling was performed (see Supplementary Materials), we identified a novel CHEK2 splice variant in a seven-year-old diagnosed with a subependymal giant cell astrocytoma, World Health Organization (WHO) grade I (Figure 2). Paired somatic disease/germline comparator enhanced exome sequencing revealed a germline splice site variant in CHEK2 (NM_007194.3:c.444+1G>A), also referred to as IVS3+1G>A or IVS2+1G>A⁷². This variant has been classified as a pathogenic alteration by other clinical laboratories (ClinVar Variation ID:128075) and is associated with cancer predisposition, most frequently breast cancer⁶⁹. This splice donor variant in intron 3 is a founder variant in the Polish population (general minor allele frequency (MAF): 0.0001, enriched in European population MAF: 0.0005)⁷³. Functional studies have described the use of an alternative 5’ splice site, four base pairs (4bp) within intron 3, predicted to encode a PTC that truncates the transcript and result in reduced CHK2 protein levels⁷².

Figure 2: CHEK2 splice variant in a seven-year-old diagnosed with a subependymal giant cell astrocytoma.

A) Schematic of CHEK2 alternative splicing (adapted from Dong et al.⁷² and Dufault et al.²⁰⁴). The G>A single nucleotide variant in intron 3 results in the use of an alternative splice site, which leads to the insertion of four extra nucleotides between exons 3 and 4 and an out-of-frame transcript. The frameshift creates a premature termination codon (PTC) in exon 4 and reduced CHK2 protein. B) Enhanced exome sequencing was performed on tissues collected from consenting patients enrolled on an institutional translational research protocol (see Supplementary Materials for more information). Aberrant RNA splicing due to germline/somatic variation is shown for each patient (red) compared to a reference cohort of tissue- and disease-matched controls (blue) in a Sashimi plot (methodology described in²⁰⁵). The corresponding genomic sequence is shown (top: reference sequence; bottom: alternative sequence) with the variant highlighted in orange. Coordinates are provided for GRCh38. The patient had a germline splice site variant in CHEK2 (NM_007194.3:c.444+1G>A) confirmed to alter RNA splicing through intron inclusion.

Aberrant Splicing in Pediatric Cancers

While the relationship between alternative splicing and cancer formation is well established in adult cancer literature, the discovery and mechanistic study of disease-causing splice variants in pediatric cancer is a growing field. Most splicing studies have been conducted in the context of leukemias and sarcomas (Table 2).

Table 2:

Summary of Aberrant Splicing Events in Pediatric Cancers

Pediatric Cancer Splicing Events
Exon Inclusion
Description	Outcome	References
Inclusion of exon 9i in TP53 transcript in pediatric B-ALL	Increased expression of p53β isoform that is regulated by SRSF3	74,75
Inclusion of alternative exon 4A in NT5C2 transcript in pediatric B-ALL	Dysfunction in metabolism of nucleoside analog drugs → drug resistance	74,76–78
Exon Exclusion
Skipping exon 11 of hnRNPA1 gene in pediatric B-ALL	NMD activated → changes in cell metabolism, DICER1 deregulation	74
Skipping of exons 4–11 of MDM2 in rhabdomyosarcoma	MDM2-ALT1 expression → loss of p53 regulation	79–82
Skipping of exons 4–9 of MDM2 in patient with metastatic aggressive alveolar rhabdomyosarcoma (institutional cohort)	MDM2-A expression → diminished direct p53 interaction	* 201,202 ** 203
Skipping exon 11 of INSR gene in rhabdomyosarcoma and osteosarcoma	Increased INSR-A: INSR-B ratio → increased cell growth, migration and viability	44,86,88
Skipping exon 6A of DHX9 gene	Increased DHX9 and enhanced EWSR1-FLI1 activity	91,95
Skipping exon 8 of EWSR1 gene in certain Ewing sarcoma translocations	Produce functional EWSR1-FLI1 protein	92,93
Skipping of exons 6 or 9 in MDM4 in retinoblastoma	MDM4-S (exon 6) and MDM4-A (exon 9) expression → oncogenesis	101
Skipping of various exons of RB1 (some induced by mutations) in retinoblastoma	Aberrant function of RB1 protein, inheritance of retinoblastoma	103–109
Splicing Factor Upregulation
SRSF3 upregulation in osteosarcoma	FoxM1, PLK1, CDC25B upregulation and ILF3 alternative splicing → increased proliferation, loss of cell cycle regulation	110–112
hnRNPa1 and PTBP1 upregulation caused by N-myc upregulation in neuroblastoma	Increased neuroblastoma cell proliferation, differential splicing of PKM gene (isoform 2, inclusion of exon 10)	114
Splice Sites and Splice Site Alterations
Use of alternative splice sites after exons 5 or 9 of WT1 in Wilms tumor	Production of four WT1 isoforms (−17AA/−KTS, +17AA/−KTS, −17AA/+KTS and +17AA/+KTS) → undifferentiated tumor cells, upregulation of VEGF and other proliferation genes, altered DNA/RNA binding, RNA splicing/processing, functional abnormalities	119,121–128
Altered splice donor site of intron 6 of TP53 in patient with osteosarcoma (institutional cohort)	Inclusion of 5 bp of intron 6 and altered splicing	* 129,130
Altered splice donor site of intron 11 of PI3KR1 in patient with osteosarcoma (institutional cohort)	PI3KR1 exon 11 skipping → constitutive activation of PI3K signaling pathway	* 132 ** 131
Indel event spanning intron 10 splice acceptor site and adjacent exonic sequence of KDM6A in patient with WHO grade IV medulloblastoma	12 bp deletion (including 10 bp of intron 10 and 2 bp of exon 11) and 2 bp insertion → loss of splice acceptor site and reintroduction of new splice acceptor site → predicted shift of open reading frame, termination codon	Present paper
Altered splice donor site of intron 33 of NF1 in patient with WHO grade I ganglioglioma	Exon 33 skipping event	140–142

ALL: acute lymphoid leukemia, INSR: insulin receptor, NMD: nonsense-mediated decay

^*reference in adult literature that describes the splice variant

^**reference in pediatric literature (cancer or other disease type) that describes the splice variant or a functionally similar variant

Exon Inclusion Events

Dysregulated splicing resulting in exon inclusion has been identified in pediatric B-cell acute lymphoblastic leukemia (B-ALL). B-ALL datasets reveal increased inclusion of TP53 exon 9i, resulting in a TP53 transcript that promotes the p53β isoform regulated by SRSF3^74,75. Another example involves the cytosolic 5’-nucleosidase II gene NT5C2, which encodes a protein that dephosphorylates purine monophosphates (IMP, GMP). Cancers with NT5C2 transcripts harboring alternative exon 4A transcripts produce an enzyme necessary for inactivating nucleoside-analog drugs used in ALL treatment, leading to drug resistance and cancer relapse^74,76–78.

Exon Exclusion Events

Most splicing events in pediatric cancers appear to be exon exclusion events in splicing factor or disease-driving genes. In pediatric B-ALL, exon 11 skipping in the splicing factor hnRNPA1 gene activates NMD of the hnRNPA1 transcript⁷⁴. The decay and decreased expression of hnRNPA1 then affects cell metabolism/DICER1 regulation⁷⁴.

In sarcomas, exon exclusion within critical genes promotes oncogene function and increased cell growth. Rhabdomyosarcoma is the most common soft tissue tumor in children. Skipping of exons 4–11 in the mouse double minute 2 homolog (MDM2) proto-oncogene, an E3 ubiquitin ligase that targets p53 toward degradation, produces the alternative transcript MDM2-ALT1⁷⁹. This isoform is constitutively expressed in alveolar and embryonal rhabdomyosarcoma tumors and is unable to fulfill the full-length transcript’s role of negatively regulating p53⁸⁰. In vitro and in vivo studies have shown that MDM2-ALT1 expression leads to accelerated tumorigenesis and increased incidence of rhabdomyosarcoma^80,81. These splicing responses are mediated in part through the splicing factor SRSF1, which binds an exonic splicing silencer element in MDM2 exon 11 and acts as a negative regulator to promote MDM2-ALT1 expression⁸². In our institutional cohort, we identified a one-year-old patient diagnosed with metastatic aggressive alveolar rhabdomyosarcoma harboring a PAX3-FOXO1 fusion and a hotspot missense variant in the TP53 gene (NM_000546.5:c.742C>T (p.Arg248Trp)). Additionally, in the same patient, we identified alternative splicing of MDM2 (NM_002392.5), specifically deletion of exons 4–9 (known as MDM2-A)⁸³ (Figure 3A). MDM2 and TP53 interact though an autoregulatory negative feedback loop to inhibit cell growth⁸⁴. The MDM2-A transcript lacks its TP53 binding domain, resulting in diminished direct p53 protein interaction⁸⁵. Multiple dysregulation events within the MDM2-TP53 pathway are hypothesized to contribute to this tumor’s aggressive nature.

Figure 3: Sashimi plots of splicing events in pediatric cancers.

A) Schematic of MDM2 alternative splicing seen in a one-year old rhabdomyosarcoma patient harboring a hotspot missense TP53 variant (NM_000546.5:c.742C>T). MDM2 exons 4–9 are skipped to generate the MDM2-A variant. B) A 16-year old osteosarcoma patient had a somatic splice site variant in TP53 (NM_000546.5:c.672+1G>T) that was confirmed to alter RNA splicing through intron inclusion in the primary tumor. This patient also had a metastatic lung nodule with an additional somatic splice site variant in PIK3R1 (NM_181523.3: c.1425+1G>C) confirmed to result in exon 11 skipping. (C) A 27-year old medulloblastoma patient had a somatic splice site variant in KDM6A (NM_001291415.2:c.876–10_877delinsAG), which generated a novel splice site 2bp within exon 11. D) A 16-year old ganglioglioma patient presented with a somatic variant in NF1 (NM_000267.3:c.4514+1G>T), which was confirmed to result in exon 33 skipping.

In rhabdomyosarcoma and osteosarcomas, multiple studies have noted the increased expression of INSR^44,86,87. An analysis of 19 osteosarcoma patient samples showed that INSR-A transcripts, which are formed through exon 11 exclusion and demonstrate higher-affinity binding to IGF-II, were consistently expressed at higher levels than INSR-B transcripts in all samples⁸⁶. Similarly, a recent study of 30 embryonal and alveolar rhabdomyosarcoma patient samples found that all tumors expressed significant levels of INSR-A compared to normal muscle control samples⁸⁸.

Interestingly, exon exclusion events in Ewing sarcoma affect both the hallmark oncoprotein and genes enhancing its function. The EWSR1-FLI1 translocation (t(11;22)(q24;q12)) encodes a namesake aberrant transcription factor that has a widespread impact on DNA regulation, transcription, and post-transcriptional processes, including splicing^89,90. The EWSR1-FLI1 chimeric interacts with the spliceosome by inducing alternative splicing of oncogenic regulatory genes or associating with splicing factors, a phenomenon observed in Ewing sarcoma cell lines and patient tumors before chemotherapy⁹¹. The most common EWSR-FLI1 translocation is composed of EWSR1 exons 1–7 and FLI1 exons 6–9. One fusion variant includes EWSR1 exon 8, which must be spliced out by hnRNPH1 to generate an in-frame EWSR1-FLI1 transcript and promote Ewing sarcoma pathogenesis^92–94. In another example, expression of DHX9, which participates in RNA metabolism and maintenance of genomic stability, enhances EWSR1-FLI1 activity and correlates with worse patient prognosis^91,95. DHX9 exon 6A contains a PTC, and splicing factors hnRNPM and SRSF3, among others, have been shown to splice this exon out, ensuring high expression of the DHX9 transcript⁹¹.

Retinoblastoma, an ocular cancer, presents with all categories of aberrant splicing^96–99. Out of 1453 differentially spliced genes in retinoblastoma tumor samples, exon exclusion and mutually exclusive exons were the most predominant events¹⁰⁰. Furthermore, these differential splicing events enriched the activity of E2F family transcription factors, the visual sense gene ABCA4, and the splicing factor DAZAP1¹⁰⁰. One other well known event in retinoblastoma occurs in MDM4, where exclusion of exons 6 or 9 produces the oncogenic isoforms MDM4-S or MDM4-A respectively¹⁰¹. Additionally, the Dab1 and RB1 genes undergo both exon inclusion and exclusion events, and various exon skipping events, some resulting from alterations, in RB1 have also been reported through genetic analyses of families with retinoblastoma^102–109.

Splicing Factor Upregulation

In osteosarcoma, SRSF3 upregulation contributes to the subsequent upregulation of cell cycle regulators FoxM1, PLK1, and CDC25B, all of which increase cell proliferation and transformation¹¹⁰. Furthermore, SRSF3 promotes alternative splicing of interleukin enhancer-binding factor 3 (ILF3) isoforms 1 and 2 and loss of cell cycle regulation^110–112.

In neuroblastoma, the most common extracranial solid tumor in children, the splicing factors hnRNPA1 and PTBP1 are upregulated due to N-Myc overexpression^113,114. Both splicing factors increase neuroblastoma cell proliferation and the differential splicing of PKM2 through the inclusion of exon 10¹¹⁴. Like INSR, the PKM2 splice variant is primarily expressed during fetal development and in several cancers due to its favorable effect on cell metabolism¹¹⁵.

Splice Sites and Splice Site Alterations

Wilms Tumor, a pediatric nephroblastoma characterized by the WT1 gene, has four major isoforms generated through the use of alternate splice sites after exons 5 or 9, though other variants have been observed^116–120. The resulting functional isoforms and their ratios are involved in undifferentiated tumor cells, upregulation of proliferation genes, DNA/RNA binding, and post-transcriptional regulation, and functional abnormalities (seen in Frasier syndrome)^{119,121–128}.

Within our institutional pediatric and AYA cancer cohort, we found evidence of the contribution of splice site alterations. In a 16-year-old female diagnosed with osteosarcoma, we identified a variant in TP53 affecting the splice donor site of intron 6 (NM_000546.5:c.672+1G>T) (Figure 3B). This variant is predicted to result in the inclusion of 5bp of intron 6 as previously described (MutSpliceDB, https://brb.nci.nih.gov/splicing/) and has been documented in multiple cancer types in the Catalogue of Somatic Mutations in Cancer (COSMIC) and the International Agency for Research on Cancer (IARC) TP53 databases^129,130. The metastatic lung nodule harbored the same TP53 variant as the primary tumor, as well as a newly acquired PIK3R1 variant affecting the splice donor site of intron 11 (NM_181523.3:c.1425+1G>C) (Figure 3B). This PIK3R1 variant results in exon 11 skipping and is predicted to constitutively activate the PI3K signaling pathway^131,132. Activating alterations within the PI3K/AKT/mTOR signaling pathway are a newly described mechanism of cell growth and proliferation in osteosarcoma¹³³.

Variants within genes involved in epigenetic regulation are frequent within medulloblastoma¹³⁴. KDM6A, a histone demethylase, undergoes recurrent frameshift, splice-site, and predicted truncating variants that are well documented in cancer, including medulloblastoma^135–138. In our cohort, a 27-year-old WHO grade IV medulloblastoma patient was found to harbor a novel KDM6A complex insertion-and-deletion (indel) event spanning the intron 10 splice acceptor site and adjacent exonic sequence (NM_001291415.2:c.876–10_877delinsAG) (Figure 3C). This complex alteration includes a 12bp deletion and a 2bp insertion. The deletion encompasses 10bp of intron 10, including the splice acceptor site (AG), and 2bp of exon 11 (CCCTTCTCAG|GT). A 2bp insertion (AG) then reintroduces a splice acceptor site. Using ExPASy Translate, we found this change is predicted to result in a shift of the reading frame, leading to introduction of a PTC (p.(Cys293LeufsTer19)) and truncated transcript¹³⁹.

We also identified a splice variant in a 16-year-old patient diagnosed with a ganglioglioma, WHO grade I. The tumor was found to harbor an NF1 splice site variant affecting the donor site of intron 33 (NM_000267.3:c.4514+1G>T) (Figure 3D). Disruption of this splice site has been previously reported in clinical databases and literature as a germline finding in patients with neurofibromatosis type 1 (NF1) (ClinVar Variation ID 576239) and as a somatic finding in NF-1 associated tumors^140–143. Similar to predictions in the literature, this splice site variant results in an exon 33 skipping event^140,142.

Alternative Splicing and Neoantigens

The multitude of transcripts generated from alternative splicing can create novel epitopes and antigens that diversify a cancer’s transcriptional profile¹⁴⁴. Widespread analysis of the TCGA has demonstrated the creation of novel exon-exon junction sites (coined “neojunctions”) that are independent of tumor mutational burden¹⁴⁵. These sites, and others generated through noncoding regions, intron retention, indels, or splicing errors, can generate polypeptides that bind the MHC complex and become cancer-specific neoantigens, which contribute to a cancer’s immunogenicity and response to immunotherapy^145–150. For example, non-small cell lung cancer patient samples had a higher expression of neoantigens derived from aberrantly spliced gene isoforms¹⁵¹.

Within pediatric cancers, the contribution of neoantigens arising from alternative splicing is an emerging field; however, studies in hematologic malignancies and medulloblastoma have demonstrated integral roles of neoantigens in cancer progression. An analysis of RNA-seq data from 400 pediatric B-ALL samples showed an increased prevalence of the CD22 exon 5–6 skipping event, which was further validated in 18 primary B-ALL samples¹⁵². Furthermore, this CD22 isoform, and not the full-length isoform, was targetable with a novel, highly specific monoclonal antibody in BALB/c mice¹⁵². Neoantigens in medulloblastoma, a cancer characterized by low mutational load, were created primarily through aberrant splice junctions that joined two non-exonic sequences^153,154. Pharmacologic modulation of splicing also affects neoepitope production. The in vitro and in vivo use of splicing modulators like indisulam on various cancer types has resulted in splicing-derived neoantigens that favorably alter tumor immunogenicity¹⁵⁵. Such an approach suggests that alternative splicing in pediatric cancers could be clinically targeted to enhance current immune therapies.

Alternative Splicing and Chemoresistance

One of the first clinically used small molecule inhibitors, imatinib effectively abrogates signaling of the BCR-ABL chimeric fusion protein in chronic myelogenous leukemia (CML)^156,157. BCR-ABL is a constitutively activated tyrosine kinase resulting from a t(9;22) gene fusion between the ABL1 and BCR genes on chromosomes 9 and 22 respectively. By binding close to the ATP binding site of BCR-ABL, imatinib sterically competes with ATP and hinders the downstream signaling cascade that would normally promote oncogenesis. However, imatinib is less effective when BCR-ABL acquires alterations in the kinase domain or is alternatively spliced to include intronic nucleotides between exons 8 and 9^158–160. These changes to BCR-ABL structurally alter the kinase domain or induce conformational shifts in the BCR-ABL protein, effectively decreasing the response to imatinib^158–162. A second potential mechanism of imatinib resistance in CML and other tyrosine kinase-driven cancers is the expression of a BCL2-like 11 (BIM) alternatively spliced isoform, which leads to loss of the apoptosis-inducing BH3 domain on BCL2^161,163,164. Notably, this splicing event can be circumvented through antisense oligonucleotide (ASO) therapy, and tyrosine kinase inhibitor response can be successfully restored¹⁶⁴.

Another common example of chemoresistance resulting from alternative splicing can be found in BRCA1 isoforms¹⁶¹. Mice implanted with BRCA1 exon 11 mutant tumors expressed high levels of BRCA1-Δ11q, which correlated with treatment resistance to PARP1 inhibitors and cisplatin³⁰. Interestingly, BRCA1-Δ11q levels can be decreased via the spliceosome inhibitor PB-1 or ASOs to resensitize cells to PARP1 inhibitors^30,165.

Likewise, there are examples of alternative splicing leading to therapy resistance in pediatric cancers. Chimeric antigen receptor T-cell (CAR-T) immunotherapy involves the infusion of autologous T cells that targeted against CD19, commonly expressed by neoplastic and non-neoplastic B cells¹⁶⁶. A subset of pediatric B-ALL patients who relapse following CAR-T therapy have downregulated CD-19 expression on their cancer cells, which sometimes results from CD19 pre-mRNA alternative splicing^167,168. When CD19 exon 2 is skipped, the final protein is truncated at the N-terminus and loses its CAR recognition site, rendering the CAR-T therapy ineffective¹⁶⁸. Another example of splicing-mediated chemoresistance is seen in glucocorticoid treatment resistance in pediatric ALL. An RNA sequencing study of 38 ALL samples identified several splicing pathways that were altered in glucocorticoid-resistant ALL cells, including variants in hnRNP and SRSF protein families that were specific to ALL subtypes¹⁶⁹. However, further investigation in a lineage-specific manner would be necessary to elucidate the exact splicing changes and their roles in steroid-resistant ALL cells¹⁶⁹.

Therapies for Splicing-Induced Diseases

Antisense Oligonucleotides and Alternative Vehicles

Therapeutic development for pediatric diseases in recent years has been pioneered by the success of ASO therapies that modulate mRNA splicing, via exon inclusion or exclusion, in spinal muscular atrophy (SMA) and Duchenne muscular dystrophy (DMD). For SMA, nusinersen is an 18-mer ASO that blocks hnRNPA1 intronic binding sites to promote SMN2 exon 7 inclusion and full-length SMN2 mRNA levels^170–174 (Figure 4A–C). In DMD, the morpholino treatment eteplirsen induces exon 51 skipping in dystrophin pre-mRNA and restores the open reading frame to produce functional dystrophin¹⁷⁵. New therapies have since been developed to restore dystrophin through exon 45 and exon 53 skipping¹⁷⁶. Both treatments have improved motor function and clinical outcomes in patients, suggesting that RNA-based therapeutics can be viable approaches for splicing-induced diseases like cancer^16,177,178.

Figure 4: Mechanisms of ASO therapy to modulate alternative splicing in SMA and cancer.

Humans have two copies of the SMN gene, which are nearly identical except for a C vs. T nucleotide difference in exon 7¹⁷¹. SMA is caused by low levels of SMN protein due to A) loss of function of SMN1 and B) exon 7 skipping in SMN2. C) ASOs targeting the intronic splicing silencer (ISS) located in intron 7–8 can prevent hnRNPA1 binding, which promotes exon 7 inclusion and functional SMN2 transcripts/SMN protein production^172–174. D) Full-length MDM2 expression is facilitated through SRSF2 binding at exon 11⁸². E) In the setting of wild-type p53 in cancer, ASOs can block the SRSF2 binding site and promote skipping of MDM2 exons 4–11, resulting in MDM2-ALT1 and loss of MDM2-induced p53 degradation⁸².

In in vitro rhabdomyosarcoma models, ASOs can block splicing factor binding sites to switch the splicing of cancer-driving genes, such as INSR. The splicing factors SRSF3 and CUGBP-1 mediate INSR alternative splicing in rhabdomyosarcoma, and blocking these binding sites with splice switching oligonucleotides in rhabdomyosarcoma cell lines reduces the levels of INSR-A, resulting in decreased angiogenesis and proliferation¹⁷⁹. In the case of MDM2, assessing p53 status of the tumor is essential to determining therapeutic ASO design. For example, in the setting of wild-type p53, ASOs can target the SRSF2 binding site on MDM2 exon 11 to induce skipping of exons 4–11⁸² (Figure 4D–E). This exon skipping results in the production of the MDM2-ALT1 isoform, which can no longer degrade wild-type p53 in cancer cells.

By binding to the mRNA and affecting mRNA/protein levels, ASOs are an attractive alternative used to target large proteins. In diffuse large B-cell lymphoma, STAT3 is one such protein against which ASOs have shown therapeutic efficacy. In a phase 1B trial, AZD9150, a 16-mer ASO that downregulates STAT3, showed 13% clinical benefit¹⁸⁰. Additionally, several pilot clinical trials have been launched to explore ASO use in solid tumors, leukemias, and lymphomas, though many of these are not in pediatric cancers¹⁸¹.

The barriers of ASO therapy in vivo lie in targeted delivery to tumors, lack of efficacy due to degradation, and toxicity^182–184. Systemic intravenous administration, for example, deposits the ASO compound in both target and non-target tissues, such as the liver and kidney. The ASO’s decreased bioavailability might not simply be overcome with increased dosing due to accumulation, immune, or aptameric binding toxicities. Some off-target effects, instability, and dosing issues are avoided through different chemical formulations, which may include various backbones, ribose modification, or polymer/peptide conjugations. Nanoparticles and other technologies are also a growing field to mitigate obstacles in ASO delivery^185–187. However, each chemical class of ASOs or delivery systems has its own associated toxicities that must be considered during therapeutic design.

Additionally, vehicles such as non-pathogenic adeno-associated viruses (AAVs) offer a solution for issues with drug delivery. AAVs are single-stranded DNA viruses that contain rep and cap genes between inverted terminal repeats (ITRs) on the 5’ and 3’ ends of the viral genome. Recombinant vectors can be engineered by replacing the rep and cap genes with a promoter followed by transgenes and a poly-A tail¹⁸⁸. Multiple studies have explored the insertion of ASO sequences instead of a transgene in the recombinant AAV vector^189,190. In DMD, sustained exon 51 skipping in the dystrophin gene was achieved by encoding antisense sequences following a U7 promoter in an AAV vector, and this approach was extremely effective in inducing long-term exon 51 skipping and therapeutic dystrophin levels in the mdx52 mouse model^189,191. Therefore, similar principles might be applicable to pediatric cancers with splicing defects, where the antisense sequence necessary to therapeutically alter splicing is expressed via a U7 promoter in an AAV vector.

Small Molecule Inhibitors

The dynamic assembly of the spliceosome can be difficult to target, but small molecule inhibitors (SMI) can inhibit the individual proteins within the complex, making them effective choices for diseases driven by splicing factors^15,192,193. Multiple SMIs, including spliceostatin A or pladienolides, directly target SF3B1, a commonly altered component of the spliceosome identified in uveal melanoma, AML, and myelodysplastic syndromes^194–198. Upon treatment with SF3B1 inhibitors, in vitro and in vivo assays show arrested cell growth and increased tumor cell death¹⁹⁹. Though early attempts at clinical trials were unsuccessful due to toxicity, SMIs are still being explored as promising agents against myeloid tumors with splicing factor mutations and defects (www.clinicaltrials.gov, NCT02841540)¹⁹³. To mitigate the unfavorable side effects often seen with SMIs, examining and repurposing currently approved FDA drugs with therapeutic effects on splicing could be an alternative approach²⁰⁰.

Conclusions

A growing body of work in the splicing field has shown that events such as exon inclusion/exclusion, splicing factor upregulation, and splice site alterations are prevalent in pediatric cancers and contribute to oncogenesis, neoantigens, and chemoresistance. Through a better understanding of how splicing and other post-transcriptional mechanisms act in tumorigenesis, we can develop more effective treatments outside of the standard regimens for pediatric patients.

Highlights.

• Alternative splicing has been increasingly implicated in pediatric cancer, neoantigens, and chemoresistance.

• Exon inclusion/exclusion, splicing factor upregulation, or splice site alterations encompass most of these splicing events.

• Splice variants previously observed in several cancers have also been observed in our institutional cohort of pediatric patients.

• Current therapies against splicing events use ASO, AAV, and small molecule inhibitors.

• Better understanding of splicing biology will contribute toward the development of novel therapeutics for pediatric cancer.