Splice-site mutations in POU2AF1 are associated with B-cell lymphomagenesis and therapeutic response

Abstract

BOB.1, encoded by POU2AF1, is one of many factors regulating physiological B-cell maturation in the germinal center. Recently, several studies have described recurrent mutations in a three-nucleotide region in the POU2AF1 splice site in the two most common B-cell non-Hodgkin lymphomas: diffuse large B-cell lymphoma and, more frequently, follicular lymphoma. In this study, we introduced a C→G mutation at the + 1 position of the POU2AF1 splice site in two B-cell lymphoma cell lines (WSU-NHL and SUDHL4) using CRISPR/Cas9 gene editing. Our results demonstrate how point mutations in the POU2AF1 splice site decreased BOB.1 expression levels. The mutation did not produce significant changes in cell proliferation, migration, or invasiveness, but did affect cell morphology, aggregation, and cell survival in a cell-line-dependent manner. Lastly, we found that the POU2AF1 mutation c.16 + 1G > C increased BCR activation, especially in SUDHL4 cells, downregulated oxidative phosphorylation (OxPhos) metabolism, and modified therapy sensitivities in both cell lines. Mutated B-cells were more sensitive to the BTK inhibitor ibrutinib. In conclusion, mutations in the POU2AF1 splice site impact B-cell lymphomagenesis at multiple levels and represent a potential therapeutic target for patients with tumors harboring this mutation.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-43710-6.

Introduction

Lymphoid malignancies comprise one of the most heterogeneous sets of diseases classified as a single type of malignancy. Diffuse large B-cell lymphoma (DLBCL) and follicular lymphoma (FL) are the two most common B-cell non-Hodgkin lymphomas (NHLs). DLBCL is an aggressive and heterogeneous disease with a variable clinical outcome. It is classified into two main subtypes based on the cell-of-origin (COO): germinal center (GC) and activated B-cell (ABC) DLBCL, defined by their gene expression profiles^1,2. This classification holds prognostic significance, with ABC-DLBCL associated with a worse clinical prognosis. More recent studies have identified new genetic subgroups of DLBCL, showing strong consistency across different studies: BN2/C1, A53/C2, EZB/C3, ST2/C4, MCD/C5, and N1^3–6. These subtypes, defined by specific mutations, are associated with different prognoses, with MCD and N1 linked to poorer outcomes, and ST2 showing a better prognosis.

On the other hand, FL is a heterogeneous though generally indolent disease with a slow progression and long clinical course; however, around 20% of FL patients could suffer early progression or transform into a high-grade lymphoma, more frequently into DLBCL. The standard treatments for both lymphomas are rituximab-based regimens⁷. However, ∼20% of FL and ∼30% to 40% of DLBCL patients do not respond to treatment, depending on the disease stage, international prognostic index, or molecular characteristics^8,9. Hence, there is a relevant unmet clinical need for targeted therapies for these patients.

During the physiological B-cell maturation process in the GC, DNA damage and cell proliferation checkpoints must be attenuated to ensure correct immunoglobulin affinity maturation, rendering these cells at higher risk of malignant transformation¹⁰. In recent years, high-throughput single-cell transcriptomic studies revealed a desynchronization of human GC transcriptional programs in B-cell lymphomas^11,12. Indeed, most B-cell lymphomas originate from GC-experienced B-cells¹³.

Lymphomagenesis is driven by genetic alterations that evade normal physiological restrictions on growth, differentiation, proliferation, cell death, and invasiveness¹⁴. It is also shaped by the interactions with the tumor microenvironment and immune escape mechanisms¹⁵. GCs are micro-anatomical structures in which antigen-activated B-cells proliferate and mature upon activation of the B-cell receptor (BCR)¹⁶. Throughout the GC reaction, a well-orchestrated series of changes in the levels of transcription factors, surface receptors, and the activation or repression of downstream pathways leads to B-cell maturation¹⁷. One of the most significant factors is B-cell lymphoma 6 (BCL6), a transcriptional repressor that acts at multiple levels and is regulated by Interferon Regulatory Factor 4 (IRF4). BCL6 enables the establishment of the GC B-cell program, facilitating B-cell migration into the GC and repressing B-cell differentiation.

Additionally, it can act as a repressor of the tumor suppressor protein p53 (TP53), conferring survival, protection, and maintenance of lymphoma cells¹⁸.

C-X-C Chemokine Receptor 4 (CXCR4) is a surface receptor whose expression is determined by the B-cell maturation stage¹⁷. It mediates the spatial segregation of GC B-cells into the dark and light zones, is implicated in disseminating DLBCL cells, and confers resistance to BCR signaling inhibitors in B-cell malignancies^19–21.

Enhancer of Zeste homolog 2 (EZH2) is an essential repressor of the B-cell transcriptional program, regulating antibody production and secretion, and maintaining a transient, immature, proliferative state^22,23.

B-cell-binding Octamer protein 1 (BOB.1, encoded by POU domain class 2-associating factor 1, POU2AF1), is a lymphocyte-specific coactivator that acts upon binding to octamer-binding transcription factors OCT1 (POU2F1) and OCT2 (POU2F2). The ternary complex enhances the transcription of essential genes involved in GC formation and dynamics, chromatin modification, B-cell differentiation, and surface receptor expression, as well as the BCR signaling pathway, activation of the NF-κB pathway, apoptosis, and B-cell survival^4,24,25. BOB.1 plays a role in the maturation of B and T-cells in the GC, and alterations in its expression levels have been described in various lymphoid malignancies^26–30.

Several studies have described recurrent mutations in a specific three-nucleotide region of the splice site of POU2AF1 in FL and DLBCL^3,4,31–36. Here, we evaluate the physiological relevance of one of these splicing mutations in B-cell lymphomagenesis, as well as in the activation state and drug sensitivities of malignant B-cell-derived cell lines.

Methods

Cell culture

B-cell lymphoma cell lines SUDHL4 and WSU-NHL were provided by Dr. Giovanna Roncador, Head of the Monoclonal Antibodies Unit at the Spanish National Cancer Research Center (CNIO, Madrid, Spain). They were authenticated by Short Tandem Repeats (SRTs) analysis in the Genomics Unit at CNIO. The cell lines were grown in Roswell Park Memorial Institute 1640 (RPMI) Medium (Lonza Biologics, Porriño, Spain), supplemented with 10% heat-inactivated fetal bovine serum (FBS) (Cultek S. L., San Fernando de Henares, Spain) and penicillin-streptomycin (P/S). YK6 cells were generated by Professor Chan-Sik Park (University of Ulsan College of Medicine, Seoul, Korea), and donated by Dr. Patricia Pérez-Galán (Institut d’Investigacions Biomèdiques August Pi i Sunyer, IDIBAPS, Spain), a member of the HK-FDC User Group. YK6 cells were grown on RPMI medium supplemented with 20% FBS. Mycoplasma testing was performed regularly using the Mycoplasma Gel Detection Kit (Biotools B&M Labs, S.A., Madrid, Spain).

CRISPR/Cas9 gene editing

CRISPR/Cas9 genetic editing was carried out using the Neon Transfection System (Invitrogen - Thermo Fisher Scientific, Spain) to deliver two different constructs introducing a C→G substitution at the + 1 position on the splicing site of POU2AF1 (Figure S1A-B). Construct referred to as POU^mut−3’ carried an eGFP after the 5’UTR region and before the POU2AF1 coding region. Construct POU^mut−5’ carried an eGFP after the mutation sequence in the intronic region of the gene (Supplementary Figure S1A-B). To avoid potential biases due to eGFP-induced off-target changes, eGFP was inserted into noncoding regions in both constructs.

Cells were resuspended at 2·10⁷ cells/mL density in R buffer. Electroporation was performed with 2 µg of construct per condition in 10 µL Neon tips at 1200 V, 20 ms, 2 pulses for SUDHL4 cells or at 1500 V, 20 ms, single pulse for WSU-NHL. Control cells were electroporated on the same conditions without the constructs. Cells were transferred to 24-well plates with recovery medium (20% hiFBS RPMI).

GFP+ cells were sorted in a FACSAria II cell sorter (BD Biosciences, NJ. USA) at a pressure of 20 Pounds per Square Inch (psi) and a rate of 1500 events per second. The purity of the selected populations was over 98%. Insertion of the mutation was detected by Sanger sequencing.

Spheroid generation

SUDHL4 and WSU-NHL spheroids were generated using maleimide (MAL)-functionalized polyethylene glycol (PEG)-based immune tissues with integrin αvβ3 binding RGD and integrin α4β1 binding REDV, in the presence or absence of stromal follicular dendritic cells (YK6 cells), as previously described³⁷. Four-arm PEG-MAL (20 kDa) was obtained from Laysan Bio (Arab, AL) with > 90% purity. Integrin αvβ3-binding RGD peptide (GRGDSPC, > 90% purity), integrin α4β1-binding REDV peptide (GREDVGC, > 90% purity), and matrix metalloproteinases (MMP)-9 degradable VPM peptide (GCRDVPMSMRGGDRCG, > 90% purity) were obtained from AAPPTec (Louisville, KY). PEG-MAL macromers were functionalized with thiolated adhesive peptides RGD or REDV with a 4:1 MAL-to-peptide molar ratio for 30 min at 37 °C. Cells were resuspended in DPBS pH 7.4, 1% HEPES containing VPM peptide or a 50:50 mix of VPM with non-degradable dithiothreitol (DTT) at a 4:1.5 MAL-to crosslinker molar ratio. Then, 5 µL of PEG-MAL macromer solution was placed in the middle of a well of a non-treated 96-well plate, and 5 µL of cell-containing crosslinker solution was injected into the droplet and mixed by pipetting 5 times. Hydrogel droplets were prepared and cured for 15 min at 37 °C for complete crosslinking. Fresh RPMI 1640 medium supplemented with 10% FBS and 1% P/S was added to each well. Half of the media was replenished every 2–3 days.

RNA purification of cell lines and qPCR

Total RNA was extracted from 5·10⁶ cells using the Maxwell^® 16 Cell LEV Total RNA Purification Kit (Promega Biotech Ibérica SL, Alcobendas, Spain). cDNA synthesis was done using AffinityScript One-Step RT-PCR Kit (Agilent Technologies, Madrid, Spain) following standard cycling parameters (Supplementary Figure S1A-1 C). It was amplified by real-time PCR in a 15 µl volume reaction in a LightCycler^®480 (Asinteg S.R.L) with RT² SYBR^® Green qPCR Mastermix reagent (Qiagen). All analyzed gene cycle threshold (Ct) values ranged between 16 and 29. Data were represented using the comparative C_t method. For a valid ΔΔC_t value, the amplification efficiency of the target and the reference gene was approximately equal (the absolute value of the slope of ΔC_t versus log relative concentration should be between − 0.1 and 0.1). The C_t was determined for each target gene in duplicate. ΔC_t was calculated by the difference between the C_t of each target gene and the C_t of an artificial BestKeeper reference gene based on the C_t values of two independent reference genes, SDHA and PPIA, calculated using the BestKeeper© software (http://gene‐quantification.com/bestkeeper.html).

Splice variant prediction

To analyze the potential effects of splice variants in silico, we use SpliceAI Lookup (https://spliceailookup.broadinstitute.org), developed by the Broad Institute. This tool integrates the SpliceAI algorithm³⁸, which provides insights into how genetic variants may impact splicing, and Pangolin³⁹, a deep-learning-based method for predicting splice-site strengths.

The SpliceAI Lookup provides several scores to assess the impact of splice variants: (i) Δ (Delta) Score: The Δ score ranges from 0 to 1, where a higher absolute delta score indicates a stronger predicted impact on splicing due to the variant. A large Δ score suggests that the variant may significantly alter splicing, either by creating a new splice site or disrupting an existing one: (ii) Pangolin Score: This score predicts potential changes in splicing, similar to SpliceAI but based on a different deep learning model. The Pangolin score helps further assess the predicted effect of the variant on splice site strength and its possible functional consequences. (iii) Combined Annotation-Dependent Depletion (CADD) Score: The CADD score estimates the potential deleteriousness of genetic variants. It ranges from + 8 (strongly pathogenic) to -8 (strongly benign)⁴⁰. A higher CADD score suggests a greater likelihood that the variant may be harmful to gene function, and thus more likely to be involved in disease.

RNA sequencing and bioinformatics analysis

Libraries for RNA-seq were generated with QuantSeq 3’ mRNA Seq Library Prep Kit-FWD (Lexogen, Vienna, Austria), according to the manufacturer’s guidelines. The cDNA libraries were quantified using Qubit and an Agilent 2100 BioAnalyzer. The libraries were sequenced using 75 bp single-end reads on an Illumina NextSeq 550 (Genomics Unit, CNIO, Madrid, Spain). Three biological replicates were performed independently for each condition.

Quality control of the raw reads of FASTQ files was conducted with Fastqc v0.12.1. Adapter sequences were removed with cutadapt v4.4. Trimmed reads were further selected with trimmomatic v0.39 for sizes bigger than 20nt and mapq > 30. Only reads that satisfied these criteria were used in ulterior analyses. Reads were aligned with HISAT2 v2.2.1 to the human genome (GRch38.p13 primary assembly).

For the 3′ mRNA-Seq data analysis, ~ 7–10 million (M) of total reads were generated from each library. FASTQ files were uploaded into the Lexogen Data Analysis Portal, and primary QC and transcript counts extraction were performed. Reads were aligned with HISAT2 v2.2.1 to the human genome (GRch38.p13 primary assembly). Mapped reads were annotated to protein-coding genes from Ensemble (GRch38 v109) and quantified using HTseq v2.0.3. Gene counts were analyzed using principal components with custom R scripts (R Project for Statistical Computing, v.4.3.1, RRID: SCR_001905). Differential Gene Expression analysis (DEG) was performed using DESeq2 v1.43.5 (RRID: SCR_015687). All samples were used to estimate size factors and dispersions. Shrinkage of effect size was performed using the ashr method⁸. Adjusted P value (Q) < 0.10 and absolute log2-transformed fold change > 0.5 were used to identify differentially expressed genes (DEGs). DESeq2 output was used for drawing volcano plots using ggplot2 v.3.5.0 (RRID: SCR_014601)^42,43.

Differential Gene Expression analysis was performed using DESeq2 v1.43.5 (RRID: SCR_015687), and Gene-set enrichment analyses (GSEAs) were conducted with GSEA software (version 4.3.3)^41,42 using RNA normalized counts as input data. GSEA was run using T-test analyses based on gene-set permutation. We used a custom-curated Lymphoma-Enriched gene set database (Supplementary Table S1), which contains selected pathways related to lymphoid cells and lymphoma processes from different sources, mainly Biocarta⁴³. Other publicly available Molecular Signature Databases (MsigDB), like Hallmark genes, curated gene sets, ontologies, and oncogenic and immunologic signatures, were also used.

Single score GSEA (ssGSEA) was done using ssGSEA (v10.1.x) in the GSEA module in GenePattern with the custom-curated Lymphoma-Enriched gene set database. Plotting Hierarchical clustering and ssGSEA projection (importance score per pathway) were performed using the Euclidean distance and the Ward.D2 clustering method with the pheatmap package v.1.0.12⁴⁴, and removing the pathways presenting an overlap in gene content with pathways specific to other cell types differing from our cell lines. The GSEA dot-plot graphs were generated using the enrich-plot (v.1.22.0) with ggplot2. Pathways not related to B cells were removed from the analysis.

Western blot

5·10⁶ cells were washed and lysed using RIPA Buffer with Halt Protease and Phosphatase inhibitor cocktail (EDTA-free). After sonication and centrifugation, the protein-containing supernatants were stored at -20 °C until use. Protein samples (50µL) were run into 15% SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and wet-transferred for 90 min to an Amersham™ Protran™ 0.2 μm NC nitrocellulose blotting membrane (Sigma-Aldrich). Membranes were blocked in Intercept ^® (TBS) Blocking Buffer (LI-COR Biosciences) 1 h at 4 °C and incubated at 4 °C with the corresponding primary antibody (overnight for Anti-BOB.1 antibody [SP92] (ab99891) at 1:100 dilution, or 2 h for α-Tubulin at 1:5000 dilution (Sigma-Aldrich). After washing, the membranes were incubated 1 h at 4 °C with IRDye 680RD Goat anti-Rabbit IgG Secondary Antibody or IRDye 800CW Goat anti-Mouse IgG Secondary Antibody (1:5000, LI-COR), washed and processed in a LI-COR Odyssey. Protein bands were quantified using the Image Studio software (LI-COR, Lincoln, NE, USA).

Proliferation assays

CMRA clearance assay Cells were incubated with CMRA (ThermoFisher) 10 µM 45 min 37 °C. The number of CMRA+ cells over time was determined by flow cytometry.

Competition assays To determine if the introduced mutation affected cell proliferation in the presence of non-mutated cells, B-cell lines bearing (POU^mut) or not the mutation were grown at different control vs. POU^mut proportions during 28 days for SUDHL4 cells or 48 days for WSUNHL cells.

Long-term cell proliferation B-cell or complex spheroids, including B-cells and FDCs were functionalized with RGD or REDV ligands. YK6 cells were treated with mitomycin C for 45 min before encapsulation with lymphoma cells into the spheroids to avoid cell division. SUDHL4 and WSU-NHL were encapsulated at 12,000 cells per gel, while FDCs were encapsulated at 6,000 cells per spheroid. The hydrogels were crosslinked with a 50% VPM with 50% DTT crosslinker mix to analyze long-term proliferation, as previously described⁴⁵. B-cell proliferation was quantified using Cell-Titer 96 AQueous One Solution Cell Proliferation Assay (Promega), according to manufacturer’s instructions, directly on spheroids at 4, 6, 10, and 14 DIV.

Fas assays

Fas-ligand-induced B-cell death in SUDHL4 and WSU-NHL cells, bearing or not the mutation, was determined in cell suspension or spheroids (with integrin αvβ3 binding RGD and integrin α4β1 binding REDV). Fas stimulation was achieved using a Fas ligand mimic (anti-mouse anti-FAS antibody, BD Pharmingen, Jo2) at 5 or 15 µM, and an antigen mimic (IgM antibody at 10 µM) was used to crosslink BCR like an antigen and avoid Fas-L induced apoptosis in B-cells.

Aggregation assays

Cell clustering parameters were determined by growing 2·10⁵ cells/well in 6-well plates and performing five acquisitions per field at 24, 48, and 72 h. The number of clusters per area, cluster size, cell circularity, skewness, roundness, density, area, and cell number were determined using the Fiji Software⁴⁶.

Transwell cell migration and invasion assay

10⁵ cells were added to the filter membrane in a 24-well transwell insert and incubated for 10 min at 37 °C and 5% CO₂ to allow the cells to settle down. After cell settlement, 600 µl of RPMI, CXCL13 (20 ng/mL; Peprotech, ThermoFisher Scientific, Spain) or CCL20 (20 ng/mL; Peprotech, ThermoFisher Scientific, Spain) were added into the bottom of the lower chamber in a 24-well plate⁴⁷. Cell migration was determined as the number of cells in the lower chamber at 24, 48 and 72 h. For invasion assays, 40 µl of Matrigel^® Matrix (Corning, Fisher Scientific, Spain) was added per insert and solidified at 37 °C for 30 min to form a thin gel layer before the addition of the cells. Cell invasion was determined as the number of cells in the lower chamber at 7 and 14 days.

Flow cytometry

For spheroids, hydrogel degradation was achieved by two hours of incubation with Collagenase IV (50 U/mL) in HBSS at 37 °C. Cells from cell suspension or disaggregated spheroids were centrifuged and resuspended in DPBS. The cells were stained with fluorochrome-conjugated antibodies for intracellular markers after fixation and permeabilization with the specific Fix and Perm kit (Immunostep, Spain). BCR activation was determined by staining the cells with IRF4-PEVio770 (Miltenyi, Spain), phospho-SYK-PE, phospho-BTK-Alexa Fluor 647, and phospho-ERK1/2-PE-Cy7 (BD Biosciences). GC differentiation was evaluated by staining with EZH2-FITC, BCL6-PE, CXCR4-APC and CD20-PerCPVio700 (Miltenyi, Spain). Before staining, all monoclonal antibodies were titrated, and cells were pre-incubated using the FcR Blocking Reagent (Miltenyi Biotec). Cell viability was determined by incubation with 0.5 µL of LIVE/DEAD TM Fixable Violet Dead Cell Stain Kit (Thermo Fisher Scientific).

Data were acquired in a digital flow cytometer MACSQuant 10 (Miltenyi Biotec) and analyzed using MACSQuantify (Miltenyi Biotec) and Infinicyt 2.0 (BD Biosciences) software.

Drug sensitivity tests

R-CHOP mix was prepared following the proportions used in clinical practice: 375 mg Rituximab (Truxima, Celltrion HealthCare), 750 mg Cyclophosphamide (Genoxal, Baxter), 50 mg Doxorubicin (Accord Healthcare), 1.4 mg Vincristine (Pfizer), and 100 mg Methylprednisolone (Urbason) in RPMI medium. Lenalidomide (Sigma-Aldrich, Spain) and ibrutinib (Deltaclon S.L., Spain) were reconstituted in DMSO. The cell lines were grown on RPMI supplemented with human serum for R-CHOP and Rituximab tests. Human serum was obtained from healthy blood donors in the Blood Bank Unit at the Hospital Universitario Puerta de Hierro-Majadahonda.

Drug responses were determined in 4-day spheroids or cell suspension, grown on flat-bottom white 96-well plates (CellStar, Greiner, Spain). For R-CHOP and Rituximab tests, the cell lines were grown on RPMI supplemented with human serum to allow complement-mediated ADCC/complement recognition. After incubation with the treatment (24 h for rituximab and R-CHOP; 72 h for Ibrutinib; 96 h for Lenalidomide), cell proliferation in the hydrogels was determined with Cell-Titer 96 AQueous One Solution Cell Proliferation Assay (Promega).

Statistical analysis

Data are summarized as the mean ± standard deviation (SD). Unpaired Student’s t-tests were used to compare the two groups’ overall differences in variable means. Group differences were then examined further using 2-way ANOVA or a mixed-effects model (REML) when values were missing, followed by Tukey’s post hoc multiple-comparison test.

In competition assays, differences in the slopes and intercepts were determined using linear regression, followed by the F test.

Principal component analysis (PCA)⁴⁸ was conducted on the raw variables to identify principal components capturing the major sources of variation across samples, using the FactoMineR package in R (version 4.3.3) to assess whether morphological and aggregation parameters distinguished the POU2AF1 splicing-mutation group from controls.

GI50 was calculated using the Best-fit values obtained on the “Sigmoidal, 4PL, X is log(concentration) equation”. Differences in the LogGI50 between selected groups were determined by Nonlinear Regression comparison using the Extra sum-of-squares F test. All tests were performed using GraphPad Prism v8 (GraphPad Software Inc., San Diego, CA, USA), and P-values < 0.05 were considered statistically significant.

Results

Recurrent mutations on the splice site of POU2AF1 affect B-cell malignancies

We first reviewed the literature on POU2AF1 mutations in the FL and DLBCL cohorts. Five FL and three DLBCL cohorts included POU2AF1 mutational data^{3,4,31–35,49,50}. The incidence of POU2AF1 mutations was higher in the FL than in the DLBCL samples (6.4% and 2.6%, respectively, Fisher exact test, Odds ratio = 2.51, P < 0.0001), mainly in the splicing region of POU2AF1 (82.61% vs. 53.13%, related to total mutations on POU2AF1, Odds ratio = 4.19, P = 0.0063, see Supplementary Figure S2). We also identified POU2AF1 splice site mutations in 2 of 29 DLBCL cell lines (OCI-LY-19 and SUDHL6)³⁵(Supplementary Figure S2). These data indicate that POU2AF1 contains a mutational hotspot at the splice-site, with these alterations occurring significantly more often in FL than in DLBCL. Moreover, while coding mutations were also detected in POU2AF1, splice-site mutations are more frequent, again more commonly in FL.

SUDHL4 and WSU-NHL cells differ in their POU2AF1 and BOB.1 basal levels

BOB.1 is essential for B-cell differentiation and germinal center function²⁹. For this reason, we selected two B-cell lymphoma-derived cell lines, SUDHL4 and WSU-NHL, which were previously described as COO-GC^51,52. The SUDHL4 cell line harbors mutations in BCL2, EZH2, and TNFRSF14 (cBioPortal), and is classified as EZB genetic subtype. The WSU-NHL cell line exhibits several EZB-characteristic mutations in the BCL2, CREBBP, and KMT2D genes, as well as a mutation in the PEST domain of NOTCH1 (COSMIC), and is, therefore, classified as N1. The presence of these mutations has been validated in our cell line clones (data not shown).

We first determined the gene expression profiles (GEPs) of the two cell lines using RNA-seq. The basal level of POU2AF1 transcript and one of its partners in the OCT-BOB.1 complex, POU2F2 (OCT2), was lower in WSU-NHL than in SUDHL4 cells (P < 0.05 and P < 0.001, respectively), whereas the level of POU2F1 (OCT1) expression was similar in both cell lines (Supplementary Figure S3A).

A C→G substitution at the + 1 position of the splice site of POU2AF1 decreases mRNA and BOB.1 protein expression

We performed an in silico analysis using the SpliceAI tool^38,39 to evaluate the effect of the mutations described in FL samples at the POU2AF1 splice site. For the mutation at POU2AF1 donor splice site of intron 1, c.16 + 1G > C, the tool predicted a donor loss with a SpliceAI delta score value of 0.99, a Pangolin delta score of 0.65, and a CADD score of + 2 points, indicating that this mutation is moderately pathogenic. The promoter AI sores, with a value of -0.63, indicate probable under-expression. This splice-site mutation likely led to skipping exon 1³⁵. Therefore, we introduced it in SUDHL4 and WSU-NHL cell lines using CRISPR/Cas9 following the strategy described in the Methods section. This method enables specific genome editing while preserving key regulatory regions, such as the promoter, intact. Several studies have shown that maintaining the endogenous promoter and avoiding functional mutations in regulatory or coding regions does not significantly affect expression levels^53–56. Furthermore, introducing a fluorescent marker, such as a fusion protein, does not alter its expression⁵⁴. Therefore, the effects observed can be attributed solely to the mutation.

The mutation led to a reduction in POU2AF1 mRNA expression in SUDHL4 cells for the POU^mut−3’ (P < 0.001) but not for POU^mut−5’ (P = 0.096). In WSU-NHL cells, POU2AF1 mRNA levels were significantly reduced in both mutated clones (P = 0.028 for POU^mut−3’; P = 0.013 for POU^mut−5’) as measured by RNA-seq (Fig. 1A). In contrast, the expression of the transcriptional complex partners OCT1/2 (encoded by POU2F1 and POU2F2) was not affected (Fig. 1B and C).

Fig. 1.

Effect of the POU2AF1 c.16 + 1G > C splice-site mutation on BOB.1 expression in SUDHL4 and WSU-NHL cell lines. (A–C) Gene expression levels of POU2AF1 (BOB.1) and its transcriptional partners POU2F1 (OCT1) and POU2F2 (OCT2), quantified by RNA-seq. Data represent mean ± SD of normalized gene expression from three independent replicates. Statistical significance was determined using an unpaired t-test. (D–F) Relative mRNA expression of POU2AF1 splice variants 1, 2, and 3 in control and mutant SUDHL4 and WSU-NHL cells, as determined by RT-qPCR. Data represent mean ± SD from three independent replicates. Statistical significance was assessed using an unpaired t-test. (G) Representative western blot showing BOB.1 protein expression in control and POU2AF1-mutated (POU^mut-3′ or POU^mut-5′) SUDHL4 and WSU-NHL cell lines. (H) Relative quantification of BOB.1 protein variants. Uq, ubiquitinated; sv1, splice variant 1; sv2, splice variant 2. Data represent mean ± SD from three replicates for control cells and four replicates for mutated cells. Statistical significance was determined by two-way ANOVA followed by Tukey’s post hoc test. P < 0.05; *P < 0.01; **P < 0.001. a.u.: arbitrary units; F.I.: fold increase.

To assess the impact of the splice-site mutation on POU2AF1 splicing, splice variant-specific qPCR was performed (see Supplementary Information and Supplementary Figure S1). This analysis revealed a significant reduction in the expression of splice variant sv.1 in both cell lines (P < 0.001), and of sv.2 in SUDHL4 cells (P < 0.05), whereas sv.3 expression was unchanged (Fig. 1D − 1 F). Overall, in SUDHL4 cells, RNA-seq revealed a significant decrease in total POU2AF1 mRNA only for the POU^mut−3′ clone; however, qPCR showed significant reductions in sv.1 and sv.2 for both mutated clones, with no effect on sv.3. In WSU-NHL cells, total POU2AF1 mRNA was significantly reduced for POU^mut−3′ and POU^mut−5′ by RNA-seq, and qPCR demonstrated a significant decrease only in sv.1, with no significant changes in sv.2 or sv.3.

The decrease in POU2AF1 mRNA expression was also associated with a drop in total BOB.1 protein expression (2-way ANOVA P < 0.001 for SUDHL4 and P < 0.001 for WSU-NHL) (Fig. 1G and H, Supplementary Table S2 and Supplementary Figure S4). Indeed, there was a significant decrease in all the BOB.1 isoforms: 40 KDa (P = 0.009 for POU^mut−3’; P = 0.001 for POU^mut−5’ for SUDHL4, and P = 0.021 for POU^mut−3’; P = 0.018 for POU^mut−5’ for WSU-NHL), 36 KDa (P < 0.001 in all cases), 28 KDa (P = 0.005 for POU^mut−3’; P = 0.002 for POU^mut−5’ for SUDHL4, and P = 0.001 for POU^mut−3’; P < 0.001 for POU^mut−5’ for WSU-NHL), and 14 KDa (P = 0.001 for POU^mut−3’; P = 0.001 for POU^mut−5’ for SUDHL4, and P = 0.001 for POU^mut−3’; P < 0.001 for POU^mut−5’ for WSU-NHL).

These results confirm that the c.16 + 1G > C splice-site mutation decreases the overall level of POU2AF1 mRNA, affects specific splicing events, and reduces the expression of all the BOB.1 protein isoforms.

POU2AF1 splice site mutation affects survival and cell clustering but not proliferation, migration, or invasiveness of the cells

Tumor cell survival, proliferation, clustering, migration, and invasion are interconnected hallmarks of cancer progression. To understand the relevance of the c.16 + 1G > C POU2AF1 splice site mutation in the proliferative capacity of the cells, we estimated the division rates of mutated and non-mutated cells. There were no significant differences in the doubling time between the POU2AF1-mutated and non-mutated clones (Fig. 2A). In a competition assay, in which cells were grown at varying ratios of mutated to control cells, mutated SUDHL4 cells did not exhibit a significant decrease in survival compared to control SUDHL4 cells (Fig. 2B). However, mutated WSU-NHL cells showed significantly lower survival compared to control WSU-NHL cells for the three different proportions of WT/mutated clones (P < 0.001 for both POU^mut−3’ and POU^mut−5’ in the three ratios) (Fig. 2C).

Fig. 2.

Effect of POU2AF1 c.16 + 1G > C mutation on cell proliferation of SUDHL4 and WSU-NHL cell lines. (A) Doubling time of the control and mutated SUDHL4 and WSU-NHL cells. Data represent mean ± SD of doubling times obtained in 4 independent replicates. Statistical relevance was determined using an unpaired t test. (B, C) Division rate of (B) SUDHL4 and (C) WSU-NHL cells grown at varying ratios of mutated versus non-mutated clones. Data represent mean ± SD from three independent experiments. Differences in slopes and intercepts were assessed by linear regression followed by an F test. (D–G) Long-term proliferation of SUDHL4 and WSU-NHL cells grown as (D, F) simple spheroids or (E, G) complex spheroids in the presence of follicular dendritic YK6 cells. Data represent mean ± SD of the percentage of viable cells from three independent replicates for controls and four for mutated cells. Statistical significance was determined by two-way ANOVA followed by Tukey’s post hoc test. P < 0.05; *P < 0.01; **P < 0.001.

We next performed an LTP assay in hydrogel 3D structures over 14 days as previously described⁴⁵. It showed that the POU^mut−3’ mutation increased LTP survival in SUDHL4 REDV simple spheroids (P = 0.001 for POU^mut−3’) and in RGD-YK6 complex spheroids (P = 0.005 and P < 0.001 for POU^mut−5’ and POU^mut−3’, respectively) (Fig. 2D-E). In WSU-NHL cells, the mutation reduced cell LTP regardless of the hydrogel conditions (RGD P < 0.001 and REDV P = 0.032) and the presence of YK6 cells (RGD n.s. and REDV P = 0.017) (Fig. 2F-G). Comparable levels of LTP were observed in cells mutated with either of the two constructs (POU^mut−3’ or POU^mut−5’).

These results show a potential discrepancy between the competition assay and spheroid experiments in SUDHL4. The lack of a proliferative difference in standard culture, in contrast to the increased growth of the mutant cells in spheroid assays (Fig. 2D–E), indicates that the detectable impact of the splice mutation in SUDHL4 may be dependent on the growth context. Hence, we cannot exclude additional factors that may be necessary for the mutation’s effect to be noticeable.

CXCL13 and CCL20 are B-cell chemoattractant chemokines that enhance the migration and proliferation of cancer cells, promoting lymphoma pathogenesis and progression⁴⁵. Transwell migration assays showed that the mutation did not change the basal migration capacity of SUDHL4 (P = 0.057 for POU^mut−3’; n.s. for POU^mut−5’) or WSU-NHL (n.s.) cells. SUDHL4 cells showed enhanced migration upon exposure to CXCL13 (P < 0.001) and CCL20 (P = 0.048) (Fig. 3A), while WSU-NHL cells showed enhanced migration when exposed to CXCL13 (P < 0.001) but not to CCL20. POU2AF1 mutation c.16 + 1G > C did not affect stimulus-induced cell migration in any cell line (except for SUDHL4 POU^{mut −5’} exposed to CCL20, P = 0.034).

Fig. 3.

Effect of the mutation in POU2AF1 in migration, invasiveness and aggregation capacity of SUDHL4 and WSU-NHL cell lines. (A) Cell migration in the absence of chemokine stimulation or in response to CCL20 and CXCL13. Data represent mean ± SD of fold change in migration relative to basal migration of non-mutated control cells, obtained from 4–6 independent experiments. (B) Cell invasiveness assessed in the absence of chemokine stimulation or in the presence of CCL20 and CXCL13. Data represent mean ± SD of the number of cells in the lower chamber from 4–6 independent experiments. Statistical significance was determined using an unpaired t-test. (C, D) Principal component analysis (PCA) of morphological and aggregation parameters in (C) SUDHL4 and (D) WSU-NHL cells with or without the POU2AF1 splice-site mutation. PCA was based on 47 quantified parameters measured in five fields per condition, and the data were analyzed across independent triplicates in two independent experiments. P < 0.05; **P < 0.001. D1: 24 h; D2: 48 h; D3: 72 h.

Next, we determined the cell invasiveness using the same stimuli, CCL20 and CXCL13. Results showed no differences in the invasiveness capacity of the cell lines, regardless of exposure to chemoattractant or the presence of the mutation (Fig. 3B).

The number of cells involved in cell dissemination and whether they disseminate as single cells or multicellular clusters determine whether the tumor spreads efficiently in different niches. We studied the effects of the mutation on morphological and aggregation parameters (cluster size, number, and morphology, among others). A principal components analysis (PCA), showed that POU2AF1-mutated clones differed from the POU2AF1 wild-type ones (Fig. 3C-D and Supplementary Table S3). The main shared variables contributing to this dimension were cell density, cell area, cluster solidity, and circularity.

Mutated SUDHL4 and WSU-NHL cells have different gene expression profiles

We next explored the changes in gene expression profiles induced by the POU2AF1 splice-site mutation.

The basal expression of genes involved in B-cell development and maturation processes, were different in both cell lines, including MTOR (P = 0.01), MS4A1 (CD20) (P = 0.002), CXCR4 (P = 0.038), IRF4 (P = 0.046), BCL2 (P = 0.003) and BCL6 (P < 0.001) (Supplementary Figure S3B).

We explored the differential expression (DE) between the control SHDHL4 and WSU-NHL cell lines. ssGSEA analysis revealed basal differences in GC- and BCR-related gene sets (including GC-CCG, PIM family and BLIMP targets (P-adj > 0.05) (Figures S3C), and also BCR-receptor, CD40 signaling during GC development, enriched in SUDHL4 (Figure S3D). On the contrary, plasma cell differentiation gene set (P-adj < 0.001, Supplementary Figure S3C) and others, such as the CCR5 and TNFR2 or FAS pathways, are enriched in WSU-NHL (Supplementary Figure S3D). These show different phenotypic backgrounds of both cell lines, showing that WSU-NHL, although described as GCB, has a more activated phenotype.

We then analyzed DGE between the control and the paired mutated cell lines. We identified 632 POU^mut−3’ and POU^mut−5’ common DE transcripts for SUDHL4 and 1104 for WSU-NHL cell lines. We also performed GSEA using a curated B-cell lymphoma (Supplementary Table S1) and the h.all.v2023.2.Hs of the Human collection MsigDB gene sets (https://www.gsea-msigdb.org/gsea/msigdb/index.jsp).

Analysis of the gene expression profile of mutated SUDHL4 cells (POU^mut−3’ and POU^mut−5’) showed downregulation of genes involved in the regulation of metabolic processes (RNH1, EIF6), MAPK activation (MCRIP2, LAMTOR1, PSMC4), DNA transcription and replication (GRWD1, PINX1) and cell migration (SH3BP1), regardless of the mutation construct used (Fig. 4A and B).

Fig. 4.

Alterations in the SUDHL4 and WSU-NHL gene expression profile induced by POU2AF1 c.16 + 1G > C splicing site mutation. (A–D) Volcano plots showing differentially expressed genes in (A, B) SUDHL4 and (C, D) WSU-NHL cells expressing POU^mut-3’ or POU^mut-5’ constructs compared with control cells. Upregulated genes are shown in red and downregulated genes in blue. Genes highlighted in bold are significantly altered in both POU^mut-3’ and POU^mut-5’ conditions. (E, F) Gene Set Enrichment Analysis (GSEA) dot plots comparing mutant versus control cells: (E) SUDHL4 and (F) WSU-NHL. Analyses were performed for POU^mut-3’ (left) and POU^mut-5’ (right) constructs using curated lymphoma-associated gene sets. Dot size reflects statistical significance, and color intensity represents the normalized enrichment score (NES). Statistical significance was estimated by t-test based on gene-set permutation analysis of triplicate samples.

Conversely, mutated WSU-NHL cells (POU^mut−3’ and POU^mut−5’) showed decreased expression of genes involved in inflammatory responses (KCNMB3, NMRAL1, MTOR, TLR10), mitochondrial metabolism (MT-ND4, MT-CO2, MT-ATP6), gene transcription (ZNF3A), and autophagy (ATG10) (Fig. 4C and D).

Functional analysis with GSEA showed depletion of the POU2AF1 pathway in mutated WSU-NHL, but not in mutated SUDHL4 (Supplementary Table S5 and S6 and Fig. 4E and F). The downregulated gene sets in the clones bearing the POU2AF1 mutation c.16 + 1G > C differed between the two cell lines, likely due to their distinct genetic and phenotypic backgrounds.

Mutated SUDHL4 cells showed decreased expression of pathways related to B-cell differentiation and activation (Toll and Rho pathways and XBP1 targets), cell survival and proliferation (PI3K-AKT-MTOR, ROS, apoptosis, DNA repair responses, hypoxia), transcription factors (such as ETS), immune interactions (Hallmarks IL2, IL6, IFNγ, TGFβ), inflammatory response, T cell cytokine signaling, apical junctions, focal adhesion, related to poor prognosis (coagulation cascades, angiogenesis, myogenesis, p38, Myc, epithelial-mesenchymal transition), cellular processes (proteasome, endoplasmic reticulum stress) and metabolism (glycolysis, lipid metabolism, and OxPhos). On the contrary, they showed enrichment of HM-INFα and K-RAS signaling pathways (Fig. 4E, Supplementary Figure S5A, and Table S5).

WSU-NHL mutated clones exhibited enrichment in pathways related to B-cell infiltration and immune response (e.g., IL-17, IL-22 22 (Fig. 4F and Supplementary Table S6), and TNFA and JAK/STAT signaling pathways, among others, as well as TP53 and apoptosis-related pathways associated with poor prognosis⁴⁶(Figure S5B). Mutated cells showed decreased expression in metabolism-related pathways (OxPhos, glycolysis, adipogenesis, fatty acid metabolism), of gene sets/pathways related to cellular processes (autophagy, spliceosome, peroxisome, proteasome) cell survival and proliferation (cytosolic DNA sensing, DNA repair, G2M checkpoint, CK1 pathway, apoptosis), among others (Fig. 4F, Supplementary Figure S5B and Table S6).

POU2AF1 donor splice site mutation affects GC differentiation

B-cell terminal differentiation depends on BCR signaling⁴⁷. To test how POU2AF1 mutation affects this process, we studied BCR activation using a Fas ligand mimic, to crosslink Fas, and an antigen mimic (IgM antibody) to crosslink the BCR⁴⁸. The results showed no differences in cell survival, regardless of the concentration of the Fas ligand mimic, IgM stimulation, the mutation or growth conditions (cell suspension or spheroids) in any of the cell lines (Supplementary Figure S6A-B).

To determine the effect on GC differentiation status (GCDS) of the cell lines, we measured the expression of the GC markers CXCR4, EZH2 and BCL6, the BCR activation markers IRF4/MUM1, and phospho-SYK (pSYK), pBTK, and pERK1/2, under basal conditions and upon IgM stimulation. Two-way ANOVA tests revealed differences in GC marker expression in mutated SUDHL4 cells vs. control (P = 0.0029 for POU^mut−3’). These differences were not apparent when cells were grown in RGD or REDV spheroids (Fig. 5A). In WSU-NHL cells, however, the mutation did not affect GC marker expression (Fig. 5B).

Fig. 5.

Effect of the mutation in POU2AF1 in germinal center (GC) differentiation status and BCR activation in SUDHL4 and WSU-NHL cell lines. (A, B) Relative expression (fold increase, FI) of germinal center markers (CXCR4, EZH2, and BCL6) and (C, D) BCR activation markers (IRF4, phosphorylated SYK, BTK, and ERK1/2) in control (C) and POU2AF1-mutated (POU^mut−3’ or POU^mut−5’) cells. Analyses were performed in SUDHL4 (A, C) and WSU-NHL (B, D) cells grown in suspension or within hydrogel matrices functionalized with RGD or REDV motifs, under basal conditions or following BCR stimulation with IgM. The Y-axis represents relative expression normalized to the basal expression of each cell line grown in suspension. Statistical significance was determined by two-way ANOVA followed by Tukey’s post hoc test from four or five independent experiments. Mutated versus control cells: P < 0.05; *P < 0.01; **P < 0.001. Basal versus IgM-stimulated conditions: # P < 0.05; ## P < 0.01; ### P < 0.001.

IgM stimulation induced differences in the GC markers of control SUDHL4 cells, both in cell suspension (P = 0.0067), and when grown on spheroids (P = 0.013 for RGD and P = 0.025 for REDV spheroids). It only affected GC markers expression of mutated cells when grown on spheroids (P = 0.004 for RGD and P = 0.048 for REDV spheroids in POU^mut−3’; and P = 0.05 for RGD and P = 0.0211 for REDV spheroids in POU^mut−5’). IgM stimulation did not induce significant changes in GC markers in WSU-NHL cells.

Concerning BCR activation, the mutation increased the phosphorylation of proteins downstream the BCR pathway in SUDHL4 cells growing in cell suspension (P = 0.0004 in POU^mut−3’ and P = 0.008 in POU^mut−5’), or in RGD spheroids (P = 0.037 in POU^mut−3’ and P = 0.019 in POU^mut−5’). WSU-NHL cells showed increased BCR activation only when grown on RGD (P = 0.0037 in POU^mut−5’), or REDV (P = 0.0019 in POU^mut−3’ and P = 0.0079 in POU^mut−5’) spheroids.

Upon IgM stimulation, we found a significant increase in BCR activation markers in control SUDHL4 cells (P = 0.018) and in all experimental groups when grown on spheroids (P < 0.001 in all cases). WSU-NHL cells showed increased BCR activation when grown on RGD (P < 0.001 in POU^mut−5’), or REDV (P = 0.0019 in POU^mut−3’ and P = 0.03 in POU^mut−5’). We found no significant changes after IgM stimulation. These results suggest that a POU2AF1 splice site mutation induces alterations in the GCDS, particularly in SUDHL4 cells, resulting in a GC-like phenotype with increased BCR activation. The effects are more limited in WSU-NHL cells, maybe due to the basal differences in the expression of GC-related pathways and higher BCR activation in this cell line compared to SUDHL4 cells.

POU2AF1 splice site mutation affects drug response in the cell lines

We investigated the impact of the POU2AF1 mutation c.16 + 1G > C on the response of cells to treatment with rituximab (R), R-CHOP, which is the standard treatment for DLBCL and FL, ibrutinib, or lenalidomide, two treatments used in both refractory or relapsed DLBCL and FL patients, with different mechanisms of action. We tested them on cell suspension and 3D spheroid models.

POU2AF1 splice site mutation did not seem to induce differences in the GI50 of rituximab in SUDHL4 cells when grown in cell suspension. GI50 was lower in SUDHL4 POU^mut−3’ cells when grown in simple (P < 0.001) or complex spheroids (P < 0.001 for POU^mut−3’ and POU^mut−5’ (Figs. 6A - D). WSU-NHL did not respond to rituximab because it lacked CD20 expression (Figure S3B).

Fig. 6.

Effect of directed mutations in POU2AF1 on SUDHL4 and WSU-NHL cell line drug sensitivities. Drug sensitivity was assessed by determining the concentration required to inhibit cell growth by 50% (GI50) at 24 h (rituximab and R-CHOP), 72 h ( ibrutinib), or 96 h (lenalidomide) after treatment. Dose–response curves are shown for non-mutated (solid lines) and POU2AF1-mutated cells (dashed lines for POU^mut−3’ and dotted lines for POU^mut−5’) in SUDHL4 (blue) and WSU-NHL (purple) cell lines: (A–D) rituximab, (E–H) R-CHOP, (I–L) lenalidomide, and (M–P) ibrutinib. Data were fitted using a four-parameter logistic (4PL) sigmoidal model with log-transformed drug concentrations, based on at least three independent experiments. Differences in LogGI50 values between groups were evaluated using nonlinear regression with an extra sum-of-squares F test (P ≤ 0.05). P < 0.05; * P < 0.01; ** P < 0.001. Panels (D, H, L, P) show representative heatmaps of GI50 values across experimental conditions. Higher drug concentrations are indicated in dark blue, and lower concentrations in white, as shown in the accompanying scale bar.

When chemotherapeutic agents were added to rituximab (R-CHOP), we found mutation-induced differences in the GI50 values of SUDHL4 cells grown in cell suspension (P < 0.001 for POU^mut−3’ and POU^mut−5’), simple spheroids (P < 0.001 for POU^mut−3’ and POU^mut−5’), and complex spheroids (P < 0.001 for POU^mut−3’ and POU^mut−5’). For WSU-NHL cells, POU2AF1 mutation c.16 + 1G > C only seemed statistically significant in the context of complex spheroids for one of the constructs (P < 0.001 for POU^mut−3’) (Figs. 6E-H).

None of the cell lines responded to the treatment with lenalidomide when grown in suspension. POU2AF1 mutation c.16 + 1G > C reduced the sensitivity to lenalidomide of SUDHL4 cells when grown in simple (P = 0.043 for POU^mut−3’; P < 0.001 for POU^mut−5’) or complex spheroids (P = 0.001 for POU^mut−3’; P < 0.001 for POU^mut−5’). In WSU-NHL cells, the mutation increased lenalidomide GI50 in monolayer (P < 0.001 for POU^mut−3’ and POU^mut−5’), and this effect was maintained when cells were grown on spheroids, although without reaching significance (Fig. 6I-L).

Finally, the POU2AF1 splice site mutation increased ibrutinib GI50 values of SUDHL4 cells grown in cell suspension (P < 0.001 for POU^mut−3’ and POU^mut−5’), but decreased it when the cells were grown in simple (P < 0.001 for POU^mut−3’ and POU^mut−5’) or complex spheroids (P < 0.001 for POU^mut−3’and POU^mut−5’). WSU-NHL also showed higher GI50 values in cell suspension for POU^mut−3’ (P < 0.001), and lower ones in simple (P = 0.002 for POU^mut−5’) and complex spheroids (P < 0.001 for POU^mut−3’ and POU^mut−5’) (Fig. 6M-P).

These results demonstrate a stronger response to the BTK inhibitor ibrutinib in both cell lines bearing the mutation, which may be of relevance for targeted therapy.

Discussion

The complex formed by BOB.1 (POU2AF1) and its partner OCT2 (POU2F2) is essential for GC formation and B-cell fate determination. Their expression is restricted to the lymphocyte lineage, making these molecules especially relevant for targeted therapies.

Recurrent splice site mutations in POU2AF1 suggest a mutation hotspot in this region. To investigate the effect of recurrent + 1 splice site mutation (c.16 + 1G > C), we introduced it into two B-cell-derived cell lines using CRISPR/Cas9. The mutation was evaluated using constructs that utilize the endogenous promoter, with fusions at both the N- and C-terminal domains of the reporter protein, and produced consistent changes across both constructs in the two cell lines, supporting the reliability of the data.

A reduction in BOB.1 had been previously linked to alterations in B cell differentiation and/or germinal center formation³⁵. Despite the basal differences in POU2AF1 mRNA and BOB.1 expression between the cell lines, the splice-site mutation reduced BOB.1 protein levels in both. Moreover, we observed decreases in total mRNA expression and cell line-dependent specific splicing events. Overall, our findings indicate that the c.16 + 1G > C splice-site mutation impacts POU2AF1 expression at multiple regulatory levels. Differences between RNA-seq, splice-variant–specific qPCR, and protein analyses likely reflect post-transcriptional and translational regulation, as well as potential differences in RNA and protein stability. While the relationship between individual splice variants and specific BOB.1 protein isoforms remains unresolved, the consistent reduction of sv.1 across both cell lines may contribute to the observed decrease in total BOB.1 protein levels.

We found negative enrichment of the POU2AF1 pathway in WSU-NHL-mutated cell lines. The reasons why this pathway does not appear to be significantly affected by the mutation in SUDHL4 cells remain unclear. Although the pathway was negatively enriched in the mutated SUDHL4 POU^mut−5′ clone (NES = − 1.187), this change did not reach statistical significance (nominal P-value = 0.14). Genetic and phenotypic differences between the two cell lines may partially account for the distinct effects observed on gene expression. In addition, differential dependence on IRF4, recently reported by He et al.⁵⁷(IRF4 was not detected in SUDHL4 cells) (Figure S3B), may provide a possible explanation for these findings.

A remarkable GSEA finding was that the mutation seems to downregulate metabolism-related pathways in SUDHL4 and WSU-NHL cell lines, including OxPhos, glycolysis, fatty acid metabolism, and xenobiotic metabolism. Metabolic reprogramming is a hallmark of cancer. Some metabolic intermediates are crucial in immune cell differentiation, proliferation, and function^50,51. The previously described OxPhos-DLBCL and BCR-DLBCLs subtypes displayed different metabolic signatures⁵². OxPhos has been linked to GC B-cell proliferation and positive selection⁵⁸. Both mutated cell lines showed lower OxPhos and glycolysis, as well as increased BCR pathway activity, suggesting a role for BOB.1/POU2AF1 in regulating B-cell metabolism. These results align with the BCR activation we found when characterizing the GCDS after introducing the mutation. Other downregulated pathways in the mutated clones included BLIMP1 and XBP1 targets, which are involved in plasma cell differentiation, as well as the TOLL pathway.

We assessed the GDCS by measuring GC markers expression (CXCR4, EZH2, and BCL6) and BCR activation markers (IRF4, p-SYK, p-BTK, and p-ERK1/2) in wild-type and mutated clones under basal and IgM-stimulation conditions. The mutation altered GC marker expression, especially in SUDHL4. No IgM-induced BCR activation changes were observed in WSU-NHL, but SUDHL4 cells showed increased BCR activation in spheroid cultures. BTK phosphorylation was the primary driver of this increased activation (Fig. 5 and Supplementary Figure S6C-D), regardless of the integrin-binding peptide used in the matrix (RGD or REDV). These results suggest that cell interactions play a crucial role in BCR activation and that the mutation induces an activated state in SUDHL4 and, more weakly in WSU-NHL. Overall, mutated cells display a GC-like phenotype with BCR activation.

POU2AF1 c.16 + 1G > C splice-site mutation did not affect migration or invasiveness but seemed to alter clustering parameters in both mutated clones, as shown by the PCA. Notably, the primary contributing variables were cell density, area, cluster solidity, and circularity. This observation, along with the transcriptomic alterations in pathways involved in cell interactions (e.g., apical surface, apical junctions, see Supplementary Figure S5), suggests morphological changes and disrupted cell interactions due to the mutation.

Another key finding is the impact of the POU2AF1 mutation on drug sensitivities. Recent studies have shown that decreased BOB.1 supports a dependency shift toward aberrant chromatin remodeling, rendering tumors particularly reliant on the mSWI/SNF complex, which is functionally linked to IRF4 through its role in enabling IRF4-dependent transcriptional programs⁵⁷. We found differences in treatment responses related to growth conditions and spheroid complexity. These differences may be partly explained by intercellular interactions, and stronger BCR activation in 3D cultures (as shown by our GCDS analysis). Notably, the mutation has different effects depending on the drug. Remarkably, we found differences in drug sensitivities of mutated cell lines, which could be explained by their different genetic backgrounds. Rituximab-based therapies remain the standard of care for B-cell lymphoma. Still, given the refractoriness of some patients, new approaches, including lenalidomide and BTK inhibitors, are currently being tested in clinical trials (https://www.clinicaltrials.gov/). Lenalidomide shows preferential activity in ABC/non-GCB disease. Still, it is broadly used at relapse across DLBCL subtypes, including GCB cases, providing a biologically relevant context to assess whether POU2AF1 alterations affect sensitivity to immunomodulatory stress. Although ibrutinib is most effective in ABC-DLBCL, it has also been administered in clinical trials and real-world practice irrespective of cell-of-origin. These agents represent clinically relevant therapeutic options and distinct mechanistic classes, enabling us to assess whether the POU2AF1 splice-site mutation influences drug responses in FL and DLBCL.

Mutated WSU-NHL cells were more sensitive to chemotherapy than the wild-type clone. The mutation increased R-CHOP sensitivity in SUDHL4 cells in complex spheroids, highlighting the relevance of cellular interactions in drug responses. In this context, lenalidomide, which exerts its immunomodulatory effects by altering cytokine production and targeting cellular interactions, has shown antineoplastic and antiproliferative effects in B-cell lines, especially in ABC-DLBCL⁵³. We found that lenalidomide treatment affected only the SUDHL4 and WSU-NHL cell lines when grown in 3D and showed reduced efficacy in mutated lines. This may be partially due to the decreased expression of genes involved in immune checkpoints, cell interactions, and metabolism as revealed by GEP analysis. The reduced sensitivity to lenalidomide observed in the mutated clones suggests that patients harboring this mutation may exhibit resistance to lenalidomide therapy.

BTK plays a central role in B-cell lymphoma pathogenesis and progression by chronic activation of BCR signaling⁵⁴. Ibrutinib, an irreversible BTK inhibitor, has shown promising results in clinical trials involving patients with relapsed/refractory DLBCL^59,60 or relapsed FL⁶¹. We found that mutated SUDHL4 cells are more sensitive to BTK inhibition when grown as spheroids, with or without dendritic cells. Conversely, mutated WSU-NHL cells grown in complex spheroids were also more sensitive to the treatment. The most prominent mutation-related effect in spheroids was enhanced ibrutinib response. Recent studies have shown that OxPhos blockade increases sensitivity to ibrutinib. Our findings of OxPhos downregulation and increased BCR activation, especially via BTK, may explain this enhanced drug response (Supplementary Figure S6C-D).

In conclusion, our findings suggest that POU2AF1 c.16 + 1G > C splice mutation reduces BOB.1 expression and affects its splicing events. These alterations increase BCR activation and downregulation of their cellular metabolism. These findings prompt changes in therapeutic sensitivity, rendering cells more responsive to ibrutinib and less susceptible to lenalidomide, thereby highlighting a potential therapeutic opportunity for patients harboring these mutations.

Our results enhance our understanding of the role of POU2AF1 in B-cell lymphoma and support the development of targeted therapies to overcome conventional treatment resistance.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

N.Y.C. designed the study, performed experimental procedures and analyzed data, performed the statistical analysis, and wrote the paper; L.P., B.H. and S.G. performed experimental procedures; A.G.G. was involved in the flow cytometry assays; R.M.V., I.F.M. and M.P.A. performed the bioinformatics analysis and data processing; R.T.R and S.R.P were involved in the design of genetic edition procedures; M.S.B. designed the study, discussed and analyzed results and wrote the paper. All the authors reviewed and approved the manuscript.

Funding

This work was supported by: Spanish Ministry of Economy and Competence (MINECO) and Instituto de Salud Carlos III (ISCIII), ISCIII-MINECO AES-FEDER (PI17/00272, PI20/00591, PI23/01587); Dirección General de Universidades e Investigación de la Consejería de Educación e Investigación de la Comunidad de Madrid (CM) (B2017/BMD-3778); and Fundación de Investigación Biomédica H. U. Puerta de Hierro-Majadahonda (FIB HUPHM), Madrid. L.P. received an iPFIS predoctoral fellowship (IFI18/0004) ISCIII-MINECO AES-FEDER, Plan Estatal I + D+I 2014–2020). I.F.M. was supported by B2017/BMD-3778 and the FIB HUPHM. N.Y.C. is supported by the Fundación Científica de la Asociación Española Contra el Cáncer (POSTD18029SANC). B.H. and M.P.A. are supported by the Plan de Empleo Juvenil de la CM (PEJ-2020-TL/BMD-19530, and PEJ-2023-AI/SAL-GL-28806 respectively).

Data availability

The datasets generated and/or analysed during the current study are available in the Sequence Read Archive (SRA) repository, BioProject code PRJNA1289409, [https://dataview.ncbi.nlm.nih.gov/object/PRJNA1289409?reviewer=37slsedlut6pfr1rbgp8bkd2os](https:/dataview.ncbi.nlm.nih.gov/object/PRJNA1289409?reviewer=37slsedlut6pfr1rbgp8bkd2os) .

Declarations

Competing interest

The authors declare no competing interests.