HNRNPU mutations redirect cell cycle control to E2F in MYC-driven lymphomas

Key Points

• •HNRNPU loss-of-function mutations are a common feature of B-cell lymphomas with MYC rearrangements.
• •Reduced hnRNPU expression causes lower MYC abundance and shifts dependence to E2Fs, leading to unique targetable vulnerabilities.

Visual Abstract

Abstract

Heterogeneous nuclear ribonucleoprotein U (hnRNPU) is a ubiquitously expressed, pleiotropic DNA and RNA binding protein involved in RNA metabolism. Heterozygous HNRNPU nonsense mutations have been observed in a variety of B-cell lymphomas, but the functional consequence of these mutations remains largely unknown. Through a large meta-analysis of genome- and exome-wide sequencing data, we find that these mutations are more common among tumors with MYC rearrangements, including high-grade B-cell lymphoma with MYC and BCL2 and/or BCL6 rearrangements (12.1%), and Burkitt lymphoma (5.2%). Using isogenic cell line models, we demonstrate that HNRNPU is a haploinsufficient tumor suppressor, with inactivation of a single allele promoting cell cycle entry through widespread alteration of the gene expression and splicing landscape. We show that reduced hnRNPU expression consistently lowers MYC levels while simultaneously enhancing E2F-driven signaling. This creates a cellular state in which MYC-induced stress may be buffered while cell cycle progression is maintained. Finally, we show that owing to this increased dependence on E2Fs for proliferative signaling, HNRNPU-mutated lymphomas are more sensitive to E2F inhibitors. These results highlight hnRNPU-mediated regulation of MYC and its downstream effects as possible new avenues for therapeutic intervention in MYC-driven lymphomas.

Introduction

B-cell lymphomas are a diverse group of malignancies with varied genetic and clinical features. Numerous comprehensive genomic studies have sought to delineate the genetic alterations that underlie each of these entities; however, functional characterization of these features has not kept pace with expanding lists of genes recurrently mutated in B-cell lymphomas.1, 2, 3, 4 Among the B-cell non-Hodgkin lymphomas (NHLs), MYC represents 1 of the most frequently dysregulated genes, typically induced through chromosomal rearrangements.^5,6 MYC rearrangements are a genetic hallmark of Burkitt lymphoma (BL) but also appear in other B-cell lymphomas. The most common form of these rearrangements is a juxtaposition of potent enhancers belonging to the immunoglobulin heavy or light chain locus with the MYC locus.^7,8 MYC rearrangements are also observed in ∼10% of tumors with diffuse large B-cell lymphoma morphology (DLBCL) and, in conjunction with B-cell lymphoma 2 (BCL2) rearrangements, predict a more aggressive clinical course.⁹ In recognition of their distinct biology, tumors with rearrangements of MYC and BCL2 and/or BCL6 have been assigned to a separate entity: high grade B-cell lymphoma with MYC and BCL2 and/or BCL6 rearrangements (“double hit” or “triple hit” [HGBCL-DH/TH]).¹⁰ Patients with tumors harboring these genetic features more frequently experience immunochemotherapy failure, with most events occurring within the first year of diagnosis.^9,11 Hence, these patients stand to benefit from a clearer understanding of the unique molecular underpinnings of MYC-translocated B-cell lymphomas.

Previous work led to the identification of mutations affecting heterogenous nuclear ribonucleoproteins (hnRNPs) in several different B-cell neoplasms including silent and noncoding mutations affecting HNRNPH1 in mantle cell lymphoma,¹² and coding and truncating mutations affecting HNRNPU in BL.¹³ hnRNPs are among the most abundantly expressed RNA-binding proteins in cells and play major roles in all aspects of RNA metabolism. They are increasingly recognized as essential for mammalian B cell development by regulating the precise expression of specific genes. Given the tight regulation of the hnRNP-RNA network, it is unsurprising that an increasing number of genes encoding hnRNPs have been causally linked to the onset or progression of many cancers, including B-cell neoplasms.¹⁴

hnRNP U (hnRNPU) is the largest protein in the hnRNP family and is 1 of the most abundant and ubiquitously expressed hnRNP proteins.¹⁵ It is a pleiotropic DNA and RNA binding protein that plays various important roles in RNA metabolism and nuclear organization.^16,17 The domain structure of hnRNPU reflects these diverse functions, with an N-terminus DNA binding domain; a central actin binding domain; and a C-terminus, intrinsically disordered, RNA binding domain.¹⁸

We previously demonstrated that HNRNPU mutations are a common genetic feature of BL¹³; however, the functional consequence of these mutations remains unclear. Here, we investigate the pattern and functional role of hnRNPU mutations across the MYC-driven B-cell lymphomas. Using isogenic cell line models, we show that HNRNPU is haploinsufficient, with mutations promoting cell cycle entry by altering the gene expression and splicing landscape of affected B cells. This reduced hnRNPU expression lowers MYC expression, shifting the dependence from MYC to E2Fs for cell cycle progression. In turn, HNRNPU-mutated lymphomas are more sensitive to E2F inhibitors. This highlights a potential novel therapeutic avenue for MYC-driven lymphomas, particularly those harboring HNRNPU mutations.

Methods

Sequencing data analyses

Somatic variant and structural rearrangement profiles were generated in an ongoing meta-analysis of mature B-cell neoplasms that includes whole-genome and -exome sequencing from NHLs. Simple somatic mutations were identified using the SLMS-3 pipeline,¹³ and MYC rearrangements were detected with GRIDSS¹⁹ and Manta.²⁰ Public short hairpin RNA knockdown and enhanced crosslinking immunoprecipitation (eCLIP)-sequencing data sets for HNRNPU (HEPG2, K562) were obtained from Encyclopedia of DNA Elements (ENCODE),²¹ and CLIP peaks were resolved to nucleotide resolution with PureClip.²² RNA-sequencing (RNA-seq) libraries were prepared using ribodepletion, sequenced on an Illumina NovaSeq, and processed with STAR, featureCounts, and DESeq2²³; pathway enrichment was conducted with FGSEA and EnrichR, and differential splicing was evaluated with rMATS.²⁴ Full details for clinical samples and analytical procedures are provided in the supplemental Methods.

Cell lines and culture conditions

HEK293 and B-cell lymphoma cell lines (Raji, Ramos, JVM2, Rec-1, SUDHL-6, HT, WSU-NHL, OCI-LY3/10, Seraphina) were cultured under standard conditions in RPMI 1640 or Dulbecco modified Eagle medium supplemented with 10% fetal bovine serum. Protein lysates were prepared using radioimmunoprecipitation assay buffer and analyzed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and immunoblotting with standard antibodies (hnRNPU, MYC, histone H3, E2F family). Image quantification was performed using ImageJ. Full culture conditions and reagent lists are described in the supplemental Methods.

Cell cycle synchronization and functional perturbations

Cells were synchronized by serum starvation, then released into complete media before RNA and protein analyses. Transient transfections were performed using either Amaxa Nucleofector (lymphoma lines) or Lipofectamine 3000 (HEK293). CRISPR-CRISPR–associated protein 9 editing of HNRNPU in Raji cells used Alt-R RNP complexes with single-cell cloning and Sanger validation. Small interfering RNA (siRNA) knockdown and plasmid overexpression procedures are detailed in the supplemental Methods.

RNA analyses

Total RNA was isolated using standard column-based extraction with DNase treatment. Quantitative polymerase chain reaction (PCR) was performed using SYBR Green chemistry, normalizing to β-actin and glyceraldehyde-3-phosphate dehydrogenase genes. RNA immunoprecipitation was performed using hnRNPU antibody or immunoglobulin G control followed by droplet digital (dd) PCR or western blot. mRNA stability assays were performed after actinomycin D treatment. Primer sequences, ddPCR probe designs, and full protocols are provided in the supplemental Methods.

Minigene and expression constructs

HNRNPU–enhanced green fluorescent protein (GFP) and MYC minigene plasmids (full-length and Δintron-1 variants) were generated by PCR amplification, restriction digestion, and ligation into GFP- or reporter-based backbones. All constructs were sequence-verified by Sanger sequencing. Complete cloning strategies are available in the supplemental Methods.

Flow cytometry and viability assays

GFP-positive cells were isolated using FACSAria Fusion. Cell viability was measured using CellTiter-Blue fluorescence assays under baseline conditions, after siRNA or CRISPR perturbations, and after inhibition with HLM00647 (E2F inhibitor). The 50% inhibitory concentration (IC₅₀) values were derived from 4-parameter logistic fits. Additional assay conditions and gating strategies are provided in the supplemental Methods.

This study was approved by the BC Cancer Research ethics committee and was conducted in accordance with the Declaration of Helsinki.

Results

Mutations affecting hnRNPU are common in MYC-driven lymphomas

Proper mRNA maturation is essential to many biological pathways, with dysregulation of this process putatively linked to oncogenesis.^25,26 To investigate the role of splicing factor mutations in B-cell lymphomas, we reanalyzed sequencing data that had been analyzed and harmonized as part of an ongoing project. We detected somatic mutations in numerous genes encoding splicing factors²⁷ across HGBCL-DH/TH, DLBCL not otherwise specified, BL, follicular lymphoma, and mantle cell lymphoma (Figure 1A), and identified several hnRNPs (HNRNPU, HNRNPH1, HNRNPQ, and HNRNPM) among the 10 most commonly mutated splicing factors. HNRNPU was the most recurrently mutated splicing factor across all the lymphomas screened, with HGBL-DH/TH (12.1%) and BL (5.2%) pathologies showing the greatest prevalence (Figure 1B). Considering the significant enrichment of mutations among pathologies with high rates of MYC rearrangements (Figure 1C), we hypothesized that HNRNPU mutations were positively selected for in the context of MYC rearrangements. Indeed, stratifying on MYC rearrangements status revealed a significant enrichment of HNRNPU mutations in MYC rearrangement–positive cases (P < .0001; χ² test; Figure 1D), pointing to the functional importance for hnRNPU particularly in MYC-driven lymphomagenesis. These results lead us to survey the relative hnRNPU and MYC protein expression levels in various lymphoma cell lines, with high hnRNPU protein levels representing a unifying feature among the cell lines with higher MYC expression (Figure 1E).

Figure 1.

HNRNPU expression and mutations are a feature of MYC-driven B-cell lymphomas. (A) An overview of mutations detected in splicing factor genes in tissue biopsies from patients with B-cell lymphoma. Cases with no mutations in any splicing factor gene are excluded. A total of 5261 tumors were analyzed (3086 DLBCL, 474 BL, 189 MCL, 428 HGBCL-DH/TH). We find that splicing factors are recurrently mutated in B-cell NHLs with HNRNPU being the most frequently mutated splicing factor. (B) HNRNPU mutations were the most common in HGBCL-DH/TH (12.07% cases), followed by BL (5.2%), DLBCL (2.01%), and FL (1.6%). (C) Presence of a MYC rearrangement in the cases from panel B, determined using a combination of fluorescence in situ hybridization and sequencing. (D) HNRNPU mutations are significantly more common among cases with a MYC rearrangements (P < .0001; χ² test). (E) Lymphoma cell lines with high hnRNPU expression tend to have high expression of MYC protein. Western blot is representative of 3 biological replicates. ABC, activated B cell; BL, Burkitt lymphoma; COO, cell of origin; DHITsig, double-hit signature; DLBCL, diffuse large B-cell lymphoma; EBV, Epstein-Barr virus; FL, follicular lymphoma; GCB, germinal center B-cell; HGBL-DH/TH, high-grade B-cell lymphoma with double hit/triple hit; MCL, mantle cell lymphoma; MUT, mutant; NEG, negative; POS, positive; WT, wild-type.

Mutations affecting HNRNPU were predominantly nonsense mutations and were clustered at the region encoding amino acids 100 to 300 (Figure 2A). These mutations were heavily weighted toward C>T mutations, which could be obtained through the enzymatic activity of activation-induced cytidine deaminase (AID). Comparing the spectrum of mutation contexts between HNRNPU and 2 targets of aberrant AID activity in DLBCL (MYC and BCL2), we found that the prevalence of C>T mutations were at comparably high levels in all 3 genes (Figure 2B). Considering the sequence context preferred by AID (WRCY),²⁸ we found that a large proportion of HNRNPU mutations occur in the WRCY motif, with levels similar to canonical AID targets (Figure 2B). Together, these data support the notion that HNRNPU mutations result from AID-mediated aberrant somatic hypermutation.

Figure 2.

Nonsense mutations in HNRNPU cluster in coding regions encoding the N-terminal portion of the protein and are enriched within AID-associated WRCY motifs. (A) HNRNPU frameshift and nonsense mutations are clustered at the N terminus of the protein in BL and HGBL-DH/TH. Plot generated with ProteinPaint. (B) All mutations from the meta-cohort in HNRNPU, MYC, and BCL2 were annotated according to their trinucleotide context and annotated according to their presence in the AID binding site, WRCY. In the WRCY motif, the primary AID target is the central cytidine (the third position), but the terminal Y can also be a cytidine and subject to deamination, illustrating that the motif defines the broader sequence context whereas the site refers to the specific base modified. AAA, ATPases associated with diverse cellular activities; RBM, RNA-binding motif; SAP, SAF-A/B, Acinus and PIAS; SPRY, splA kinase and ryanodine receptor.

HNRNPU mutations reduce MYC dependance while sustaining cell cycle programs

To investigate the role of HNRNPU mutations in MYC-driven lymphomas, we used CRISPR to introduce mutations in the BL cell line, Raji. These cells were deemed to best reflect patients harboring HNRNPU mutations, because they contain an IGH::MYC rearrangement, have only 2 copies of the HNRNPU genetic locus, and normally express high levels of both proteins (Figure 1E). We selected 2 CRISPR-edited clones (CR-1 and CR-2) that each contained a frameshift deletion in 1 allele and an in-frame deletion in the other (supplemental Figure 1A). The in-frame deletions remove 2 to 4 amino acids from an intrinsically disordered region. These were chosen because they were deemed unlikely to affect the function of the hnRNPU protein. The frameshift deletions occur at residues 150 (CR-1) and 155 (CR-2), with downstream stop codons predicted to generate 203– and 199–amino acid truncated hnRNPU variants, respectively. We found that although there was variable expression of HNRNPU at the mRNA level (supplemental Figure 1B), protein expression was consistently and significantly lower in the edited cells (supplemental Figure 1C-D). In addition, the presence of a truncated isoform was not observed (supplemental Figure 1D), indicating that the transcripts bearing frameshift mutations are removed by nonsense-mediated decay. Consistent with these findings, we did not recover any viable clones harboring biallelic frameshift mutations in HNRNPU despite extensive CRISPR screening, indicating that complete loss of HNRNPU is not tolerated. This interpretation is supported by genome-wide CRISPR essentiality screens showing that HNRNPU is required for cell survival across multiple human cancer cell lines, including BL models.²⁹ Together, these data support the interpretation that tumor-associated mutations likely represent partial loss-of-function events rather than complete gene inactivation.

We performed RNA-seq on the CRISPR-modified Raji cell lines and identified 1296 significantly differentially expressed genes, between HNRNPU mutant and wild-type samples (P value <.05, log₂ fold change >0.58; Figure 3A). We performed pathway-based enrichment analysis using the hallmark gene sets from the Molecular Signatures Database,³⁰ which represent well-defined biological states or processes derived from coherent gene expression patterns. This analysis revealed that the most significantly upregulated pathways were associated with cell cycle progression, including the G2/M checkpoint, mitotic spindle, and DNA repair (Figure 3B). Strong enrichment (Figure 3B-C) and higher expression (Figure 3D) of MYC- and E2F-targeted genes was also observed, which is notable considering MYC and E2F family members are thought to cooperatively drive cell cycle dysregulation in BL and HGBL-DH/TH.³¹ Considering our observed correlation between elevated hnRNPU and MYC protein levels in various lymphoma cell lines (Figure 1E), we assessed MYC levels in the CRISPR-edited cells. Strikingly, RNA-seq (Figure 3E), quantitative PCR (supplemental Figure 1E), and western blotting (Figure 3F) data all showed a marked reduction in MYC mRNA/protein, suggesting that impairment of HNRNPU repressed the expression of this oncogene. The E2F protein family contains activating (E2F1 and E2F2) and repressive (E2F6) members, and our data indicate that E2F1 and E2F2 are significantly upregulated (Figure 3E-F; supplemental Figure 1E) and E2F6 is significantly downregulated (Figure 3E-F; supplemental Figure 1E). Although the E2F expression profiles conceptually aligned with the enrichment in E2F target genes (Figure 3B-C), the downregulation of MYC (Figure 3E-F; supplemental Figure 1E) and upregulation of MYC target genes (Figure 3B-D) were somewhat unexpected. Pathway analysis identified 19 significantly enriched MYC target genes, of which 18 were upregulated, whereas 1 (MYC itself) was downregulated (Figure 3D). Notably, nearly half of these genes (9/19 [47%]) are also annotated as E2F targets, including genes involved in DNA replication and cell cycle progression, such as MCM2, CDC20, PCNA, and MAD2L1 (Figure 3D). This substantial overlap suggests that the observed enrichment of MYC targets may reflect activation of a broader cell cycle transcriptional program driven by E2F rather than direct MYC activity alone. Consequently, increased E2F activity may drive expression of genes that are also categorized as MYC targets, resulting in apparent enrichment of MYC-regulated genes despite reduced MYC transcript levels. This may explain the similar viability of HNRNPU mutant and wild-type cell lines (Figure 3G), despite reduction in MYC expression.

Figure 3.

HNRNPU mutations alter the gene expression landscape of Raji cells. (A) Heat map showing hierarchical clustering of the expression of 1269 differentially expressed genes in wild-type (WT) (HNRNPU^+/+) and mutant (HNRNPU^+/−) Raji lines. (B) Gene set enrichment analysis was performed using FGSEA with the Hallmark pathways data set. A positive normalized enrichment score (NES) value indicates enrichment in the HNRNPU-mutant phenotype, a negative NES indicates enrichment in the WT phenotype. Significance threshold: adjusted P value >.05. (C) Enrichment plots for top enriched pathways, showing the profile of the running enrichment score and positions of gene set members on the rank-ordered list. (D) Heat map showing expression of top significantly differentially expressed MYC and E2F targets. (E) MYC and E2F family members that are found to be significantly differentially expressed with Deseq2, P value <.05, log₂ fold change of >0.58. (F) Western blot showing reduction in hnRNPU protein in mutant lines, downregulation of MYC protein and induction of E2F1 protein. Histone H3 used as a loading control. (G) Cell viability approximated with resazurin reduction. HNRNPU loss minimally alters cell viability. ns, not significant.

To extend these observations, we investigated whether there was any evidence of E2F dysregulation in tumors from patients with mutant HNRNPU, using RNA-seq data from HGBCL-DH/TH tumors carrying HNRNPU mutations. HGBCL-DH/TH cases frequently involve MYC rearrangements to nonimmunoglobulin partners, likely owing to the high levels of somatic hypermutation that increase opportunities for these events.³² To minimize confounding from differing rearrangement partners, we restricted our analysis to samples with immunoglobulin::MYC rearrangements and excluded those with MYC rearrangements involving nonimmunoglobulin loci. Our data set compared a total of 10 HNRNPU-mutant samples with 28 samples that possessed no mutations in any other splicing factor. This analysis showed a notable enrichment in E2F targeted and G2/M checkpoint genes (supplemental Figure 2A) and upregulation (E2F1 and E2F2) or downregulation (MYC) of the critical cellular factors (supplemental Figure 2B). To determine whether transcriptional changes associated with HNRNPU mutation were accompanied by differences in tumor proliferative index, we examined Ki67 expression in lymphoma tumor samples. No significant differences in Ki67 staining were observed between HNRNPU-mutant and wild-type tumors (supplemental Figure 2C). These findings indicate that overall proliferative activity is broadly comparable between groups despite reduced MYC expression, potentially due to compensatory upregulation of E2F target genes. Given the limited sample size, it was encouraging to see this similar trend in patient samples and HNRNPU-mutant Raji cell lines. HNRNPU knockdown data were also reanalyzed in chronic myelogenous leukemia cell line K562, showing an enrichment in MYC-targeted, E2F-targeted, and G2/M checkpoint genes despite decrease in MYC expression (supplemental Figure 2D).

Next, we sought to determine whether these changes in MYC and E2F expression were consistent across additional lymphoma cell lines. To test this, we performed HNRNPU knockdown in the HGBCL-DH cell line, WSU-NHL. Consistent with our observations in mutant Raji cells, HNRNPU depletion resulted in decreased MYC protein expression alongside increased expression of E2F1 and E2F2 (supplemental Figure 2E). Knockdown of hnRNPU in HEK293 cells also showed reduced MYC expression, but changes in E2F targets differed from those observed in lymphoma cell lines, indicating that E2F regulation by hnRNPU may be cell type–specific whereas MYC regulation appears conserved across cell lines (supplemental Figure 2E).

HNRNPU mutations induce widespread splicing alterations with enrichment in cell cycle programs

Given the prominent role of hnRNPU in RNA splicing, we compared the abundance of isoform-distinguishing events between the HNRNPU-mutant and wild-type cell lines. This revealed 891 alternative splice variants (false discovery rate of 0.01, delta percent spliced in of >0.1), with 77% representing skipped exon events (Figure 4A). Using the Reactome2022 data set and EnrichR,³³ we determined which pathways were significantly affected by these events. Similar to the differential expression results, this revealed pathways related to cell cycle regulation (Figure 4B). Notably, a skipped exon event was observed in the protein phosphatase CDC25A, which is crucial for promoting G1/S transition.³⁴ Specifically, HNRNPU-mutant cell lines exhibited increased skipping of exon 6 in CDC25A transcripts, accompanied by increased CDC25A mRNA levels (Figure 4C). Previous studies suggest that CDC25A isoforms lacking exon 6 may exhibit altered stability due to loss of regulatory phosphorylation and ubiquitination sites.³⁵

Figure 4.

HNRNPU mutations alter the splicing landscape of Raji cells. (A) A total of 891 alternative splice variants were identified from the RNA-seq data when comparing HNRNPU WT cell lines to HNRNPU mutants. False discovery rate of 0.1; delta percent spliced in (dPSI), >0.1. Among these, the most overrepresented class is that of skipped exon events. (B) Pathway enrichment analysis performed with EnrichR using the Reactome 2022 database. Genes involved in cell cycle progression are differentially spliced. (C-E) Differential splice events in CDC25A (C), MDM2 (D), and CARD8 (E). dPSI calculated based on inclusion reads/exclusion reads. AURKA, aurora kinase A; NLP, ninein-like protein; NuMA, nuclear mitotic apparatus protein; PLK1, polo-like kinase 1.

This analysis also identified altered splicing in MDM2 involving a mutually exclusive exon (MXE). Inclusion of MXE exon 1 produces an out-of-frame isoform that is annotated as being degraded by nonsense-mediated decay (MDM-002, ENST00000393417), whereas inclusion of MXE exon 2 results in a functional transcript (MDM-013, ENST00000393416). In the HNRNPU-mutant cell lines, MXE exon 1 was less likely to be included, and was associated with higher levels of MDM2 mRNA (Figure 4D). MDM2 is an oncogene encoding an E3 ubiquitin ligase, which acts as a negative regulator of p53 and Rb to allow for apoptosis evasion and growth stimulation.^36,37

Regulators of apoptosis were also shown to be affected, with a member of the caspase-associated recruitment domains family (CARD-8) found to be differentially spliced in these CRISPR-modified cells. CARD8 triggers apoptosis in response to cell death signals by physically interacting with caspase 1.³⁸ In HNRNPU-mutant cell lines, we observe a skipping event in exon 4, predicted to result in a truncated protein isoform (Figure 4E).

Although consistent changes at the protein level were not robustly detected, likely due to the inherent instability of these proteins, the observed splicing alterations suggest modulation of pathways involved in cell-cycle regulation and apoptotic signaling. Taken together, these findings indicate that HNRNPU mutations are associated with widespread splicing changes affecting genes linked to cell cycle control, which may contribute to the phenotypic differences observed in mutant cells.

HNRNPU mutations regulate MYC stability and may modulate MYC-induced cellular stress

Our findings have consistently indicated that mutation or decreased expression of HNRNPU leads to reduced MYC levels. Considering the dependence of BL and HGBCL-DH/TH on MYC, we questioned how HNRNPU mutations could provide a selective advantage to tumor cells. Under normal circumstances, high MYC levels are known to trigger programmed cell death39, 40, 41 and the unfolded protein response.⁴² Accordingly, a reduction of MYC facilitated by HNRNPU mutations could help titrate MYC to tolerable levels. Directly testing this concept in established lymphoma models is challenging, because MYC-driven lymphomas are likely to have acquired mechanisms that allow them to tolerate MYC-associated stress early during lymphomagenesis. We instead ectopically expressed GFP-tagged hnRNPU that resulted in increased MYC mRNA and protein levels (Figure 5A-B). These cells exhibited a significant reduction in viability (Figure 5C) and reduced splicing of X-box–binding protein 1 (Figure 5D), a commonly used marker of endoplasmic reticulum stress and unfolded protein response activation.⁴³ These observations show that elevating MYC through hnRNPU can introduce additional cellular stress. Based on this, we speculate that loss-of-function mutations affecting HNRNPU may promote fitness by buffering against deleterious effects of high MYC.

Figure 5.

hnRNPU-mediated modulation of MYC expression. (A-B) Exogenous hnRNPU expression in Raji cells results in increased MYC mRNA (A) and protein (B) expression. (C) hnRNPU overexpression significantly decreases cell viability in overexpressed cells relative to those transfected with empty vector. (D) In the overexpressed cells we observed an increase in the splicing of XBP-1, indicating an induction of the unfolded protein response and proteotoxic stress. (E) Assessment of hnRNPU-bound RNA transcripts in Raji cells was performed using RNA immunoprecipitation followed by ddPCR. We observed a significant enrichment of MYC transcripts containing intron 1 but not of MYC transcripts lacking the intron. Primers specific for intron 1 (unspliced MYC) or the exon 1-2 junction or exon 2-3 junction (spliced MYC) were used to identify the fold enrichment of these transcripts. TBP and PGK1 were used as negative controls, because they showed no binding to hnRNPU in Encode e-CLIP data. (F) MYC minigene transcript transfected into HEK-293 cells. After transfection, hnRNPU-IP was performed. Bound RNA was assessed through a minigene-specific ddPCR primer probe. Enhanced enrichment of full-length MYC minigene relative to MYC minigene lacking intron-1. Significance was calculated by Student t test (∗P ≤ .05). Results from 3 or 4 biological replicates. (G) By stratifying patients by the location of the MYC rearrangements relative to the locus, we observed that HNRNPU nonsense mutations were restricted to samples with either upstream or downstream breaks (n = 183) and absent from MYC-rearranged samples with intronic breaks (n = 85). DDX3X, DEAD-box helicase 3 X-linked; IgG, immunoglobulin G; PCBP1, poly(rC)-binding protein 1; PGK1, phosphoglycerate kinase 1; TBP, TATA-box binding protein; XBP-1, X-box–binding protein 1.

Next, we sought to understand the mechanism underlying hnRNPU-associated changes in MYC expression. Initial insights from eCLIP experiments conducted in K562 and HepG2 cell lines,²¹ showed direct association between hnRNPU and the MYC transcript, specifically within poly-G tracts in intron 1 (supplemental Figure 3A). Notably, this region is predicted to form G-quadruplexes, secondary structures that are preferred binding sites for hnRNPU.⁴⁴ To confirm this interaction in Raji cells, we performed RNA immunoprecipitation of hnRNPU (Figure 5E, inset), followed by assessment of coimmunoprecipitated RNA transcripts using ddPCR (Figure 5E). Consistent with the eCLIP data, hnRNPU interacted with the unspliced version of MYC (intron 1) but not the mature transcript (exon 1-2 and 2-3 junctions) or negative controls (TBP and PGK1; Figure 5E). To directly resolve this interaction, we introduced HEK-293 cells with 2 minigene constructs corresponding to the full-length transcript (MYC 5′ untranslated region and all introns and exons up to the stop codon), and a version lacking intron 1. After co-RNA immunoprecipitation of hnRNPU (Figure 5F, inset), we observed greater enrichment of the full-length minigene transcripts, relative to the version lacing intron 1. Control and siRNA-depleted hnRNPU (HEK293) or HNRNPU-mutant (Raji) cells were treated with the transcriptional inhibitor actinomycin D, and RNA was collected at 30-minute time points. The results showed a ∼30% reduction in the MYC mRNA half-life in both HEK293 (control, 104 minutes; si-hnRNPU, 70 minutes) and Raji (wild type, 73 minutes; CRISPR, 54 minutes) cells (supplemental Figure 3B). Collectively, these data show that hnRNPU directly influences MYC protein abundance by interacting with the intron 1 region, reducing mRNA stability. This implies that hnRNPU-mediated regulation would be disrupted by rearrangements that removes this region from the MYC gene. Consistent with this, nonsense mutations in HNRNPU were never observed in tumors with intronic MYC rearrangements, and instead, only appear in tumors with upstream or downstream MYC rearrangements (Figure 5G).

HNRNPU expression correlates with sensitivity to E2F inhibition

The culturing and selection of CRISPR-modified cell lines provide time for cells to compensate for the disruption of targeted genes. This may explain why we did not observe a reduced viability in mutant Raji cell lines (Figure 3G), whereas transient overexpression of hnRNPU led to a significant increase in cell death (Figure 5C). Because our data demonstrate that MYC levels are decreased in HNRNPU-mutant cell lines (Figure 3F), we hypothesized that these cells were more dependent on E2F activity to promote cell cycle entry. To test this, we used a pan-E2F inhibitor, HLM006474, which has shown efficacy in blocking cell growth in a melanoma culture model.⁴⁵ This resulted in reduced cell viability in the 2 HNRNPU-mutant cell lines, with the IC₅₀ values for CR-1 (18.6μM) and CR-2 (13.8μM) significantly lower than that of the parental wild-type cells (31.6μM; Figure 6A). We expanded this to Seraphina, a BL cell line with a 50–base pair deletion in 1 allele of HNRNPU in exon 1. This deletion leads to a frameshift of 24 amino acids followed by a stop codon (supplemental Figure 4). We show that relative to Raji, Seraphina has lower expression of hnRNPU and MYC, but increased expression of E2F1 and E2F2, as predicted (Figure 6B). Treatment with HLM006474 showed that Seraphina (IC₅₀: 19.5μM,) was significantly more sensitive to the E2F inhibition compared with wild-type Raji cells (Figure 6C). Overall, these findings demonstrate that cells with mutations in HNRNPU exhibit a greater reliance on E2F activity to propel the cell cycle, causing an increased vulnerability to E2F inhibition. Considering the limited efficacy of MYC inhibitors⁴⁶ and the shared pathway between MYC and E2F for cell cycle initiation and proliferation, E2F could represent a promising therapeutic target for these lymphomas.

Figure 6.

HNRNPU mutations render cells more sensitive to E2F inhibitors. (A) Comparison of IC₅₀ for Raji WT and HNRNPU CRISPR-mutant (CR-1/2) cell lines with the E2F inhibitor HLM006474. HNRNPU-mutant cell lines were significantly more sensitive to the inhibitor. (B) Western blot showing protein expression of MYC and E2F family members in Raji and Seraphina cell lines. These data reflect that cell lines with HNRNPU mutations have corresponding low MYC expression but increased expression of E2Fs. (C) Comparison of IC₅₀ for Raji with Seraphina cell line for HLM006474 (E2F inhibitor). Cell lines with low hnRNPU expression and high MYC expression are significantly more sensitive to the E2F inhibitor. Representative dose response curve for HLM006474 is shown and IC₅₀ calculated from 3 independent drug dose-curve assays. All data represent the mean of 3 independent experiments ± standard deviation. Significance was calculated by Student t test (∗P ≤ .05, ∗∗P ≤ .01).

Discussion

Many RNA binding proteins, particularly hnRNPs, have the capacity to regulate both tumor suppressive and oncogenic pathways, and overexpression and knockdown can result in cell proliferation and apoptotic defects.⁴⁷ Given the tissue-specific functional roles of hnRNPs, different repertoires of RNA targets predicts that their mutation may have pleiotropic and highly context-specific effects. Although hnRNPU has been reported to act as an oncogene,^48,49 we observed a pattern of mutations in B-cell lymphomas more consistent with tumor suppressor activity. We found that loss of a single allele of HNRNPU reduced the amount of hnRNPU protein in the affected B cells along with a reduction of MYC protein. This supports our model that HNRNPU is haploinsufficient in these cell lines. The HNRNPU mutations result in significant alterations to the transcriptomic and splicing landscapes of the mutant cells. Overall, most of these changes indicated a propensity toward cell cycle promotion by upregulating genes that facilitate the G1/S and G2/M transitions, particularly those belonging to the E2F family.

E2Fs contribute to oncogenesis in a variety of cancer types, and high expression is associated with a poor prognosis for patients with cancer.^50,51 In BL, E2F1 is highly expressed relative to nontransformed B cells, and reduction in its expression decreases tumor formation and proliferation.⁵² Ectopic expression of E2F1 is sufficient to drive quiescent cells to S phase.⁵³ In plasmablastic lymphoma, the upregulation of E2F1, E2F3, and E2F8, correlates significantly with an increased expression of MKI67 (Ki67), indicating that E2F exerts a crucial role in abnormal proliferation.⁵⁴ Although dysregulation of the E2F pathway is typically a result of alterations in the RB1 pathway, this was not immediately apparent in our model. Although this may suggest an RB1-independent form of E2F dysregulation in these lymphomas, our alternative splicing analysis points to relieving of RB inhibition as a potential alternative. Specifically, mutant cells showed significant differences in the splicing of CDC25A and MDM2, whose gene products facilitate inactivation of the Rb repressor to promote G1/S transition.^34,55 The combined effects of elevated E2F expression and enhanced stability of MDM2 and CDC25A may contribute to increased expression of cell cycle–associated genes in the HNRNPU-mutant setting.

In addition, we observed a consistent positive correlation between MYC and HNRNPU expression across all tested cell lines. Reduced MYC signaling would typically be expected to impair cell cycle progression in MYC-dependent lymphoma cells^56,57; however, despite reduced MYC expression, we did not observe marked differences in overall viability in HNRNPU-mutant cells (Figure 3G) or proliferative indices in HNRNPU-mutant primary tumors (supplemental Figure 2C). This suggests that increased E2F activity may help sustain cell cycle gene expression programs in the HNRNPU-mutant context. In MYC-addicted cancers, there is a requirement for E2F activity to induce proliferation.^31,58 Because E2Fs are known transcriptional targets of MYC,⁵⁹ we speculate that hnRNPU modifies the gene expression profile in a manner that uncouples MYC from E2F expression. This would allow the cell cycle to proceed through sustained E2F signaling, without activating the proapoptotic effects of MYC.

We hypothesize that HNRNPU mutations might represent an early event in lymphomagenesis that provide a selective advantage by lowering MYC expression to tolerable amounts to help buffer MYC-induced proteotoxic stress. This is comparable with the effect imparted by DDX3X, which is among the most commonly mutated genes in BL.⁴² Notably, we did not observe co-occurrence of HNRNPU and DDX3X nonsense mutations in the same tumor, which could imply they provide redundant roles in MYC-driven lymphomagenesis. Buffering MYC expression allows cells to evade MYC-induced cell stress but maintain the ability to enter the cell cycle by means of upregulated E2F activity.

Current therapies for MYC-driven lymphomas are based on highly aggressive, toxic, combination immunochemotherapy regimens. B-cell lymphomas with MYC rearrangements such as HGBCL-DH/TH are more often refractory to conventional therapy options or relapse early. There is, thus, a need for new therapeutic strategies that exploit molecular vulnerabilities specific to MYC-driven lymphomas. Identifying and targeting molecular vulnerabilities that aid in MYC-driven lymphomagenesis could help bypass the “undruggable”⁶⁰ MYC and create new avenues for intervention in these lymphoma types. The interplay between hnRNPU and MYC represents a novel survival mechanism in these cancers. The consistently increased dependence on E2F activities in HNRNPU-mutant models was associated with increased sensitivity to E2F inhibitors.

Conflict-of-interest disclosure: The authors declare no competing financial interests.