SRSF2 P95H induces monocytic priming of human hematopoietic stem and progenitor cells

Mutations in the splicing factor SRSF2 are commonly found in healthy individuals with clonal hematopoiesis (CH), where they confer a high annual clonal growth rate and an elevated risk for MDS/AML transformation [ 1 ]. Accordingly, SRSF2 mutations frequently occur in patients with myelodysplastic neoplasms (MDS, 10-15%), acute myeloid leukemia (AML, 10%), and chronic myelomonocytic leukemia (CMML, up to 40%) [ 2 ]. Recurrent mutations are exclusively heterozygous and primarily affect proline 95, with the P95H substitution being the most common. As part of the spliceosome, SRSF2 facilitates exon recognition by binding to exonic splicing enhancers (ESE). P95H mutations alter the ESE recognition motif of SRSF2 , leading to widespread missplicing of mRNAs [ 3 ]. While a few recurrent missplicing events could be directly linked to the development of myelodysplasia, the transformative impact of SRSF2 mutations on hematopoietic stem and progenitor cells (HSPCs) remains poorly understood [ 4 ]. Here, we specifically address this gap by capturing the immediate, cell-intrinsic consequences of SRSF2 mutation acquisition in primary human HSPCs using a bi-allelic endogenous knock-in (KI) approach. In contrast to prior studies relying on patient-derived samples or overexpression systems, this strategy enables temporal resolution of early molecular events independent of clonal evolution and co-mutation effects. We engineered primary human umbilical cord blood (UCB)-derived HSPCs to carry a heterozygous SRSF2 P95H/WT mutation using a CRISPR/Cas9 and rAAV6-mediated KI strategy (Fig. 1A, B ) and analyzed the earliest molecular changes. Because SRSF2 -mutant myeloid neoplasms arise predominantly in adult hematopoiesis, the use of UCB-derived HSPCs represents a limitation of this model. We therefore performed orthogonal validation experiments in adult mobilized peripheral blood (mPB)-derived CD34 + HSPCs where feasible, to test whether key molecular and functional findings were reproducible in an adult cellular context. Ribonucleoprotein (RNP) complexes composed of Cas9 protein and chemically modified sgRNA targeting SRSF2 exon 1 were electroporated into HSPCs, followed by rAAV6-mediated delivery of donor templates encoding either codon-optimized wildtype (WT) or mutant (P95H) SRSF2 cDNA (Fig. 1A ). This method ensured in-frame integration of WT and P95H sequences into the endogenous SRSF2 locus, preserving native regulatory elements (promoter and 5’ UTR). Including a 3’-fluorescent reporter cassette (GFP, BFP, or mCherry) in each rAAV6 vector enabled dual transduction (e.g., P95H-GFP + WT-BFP), with subsequent FACS purification of dual reporter positive cells, resulting in the production of exclusively heterozygous mutants ( SRSF2 P95H/WT ) (Fig. 1B ). The same bi-allelic KI strategy was used to generate matched WT control cells ( SRSF2 WT/WT ). Since homozygous or hemizygous SRSF2 mutations are lethal [ 5 ], our endogenous bi-allelic KI strategy was crucial for creating viable heterozygous SRSF2 P95H/WT cells. This approach also enabled the investigation of early SRSF2 mutation-driven alterations under controlled conditions, bypassing the confounding effects of prolonged in vitro selection or the genetic complexity of patient-derived samples. Our approach reliably produced WT and mutant HSPCs across independent donors (dual reporter-positive WT/WT: 8.0 ± 3.2%, P95H/WT: 7.6 ± 2.2%, n = 5), which were enriched to >95% purity via FACS before downstream analysis (Fig. 1C , Supplementary Fig. 1A ). Bi-allelic and locus-specific integration was confirmed by in-out PCR and Sanger sequencing (Fig. 1D , Supplementary Fig. 1B ). Importantly, total SRSF2 protein levels remained unchanged between engineered WT and mutant cells (Supplementary Fig. 1C ). Fig. 1: CRISPR-engineered SRSF2 P95H/WT primary human HSPCs share common splicing alterations with myeloid neoplasms. Full size image A Illustration depicting the CRISPR-mediated integration of heterozygous SRSF2 P95H/WT mutations at the endogenous gene locus on chromosome 17. SRSF2 WT and P95H sequences are coupled with different downstream fluorescent reporter cassettes driven by a spleen focus-forming virus (SFFV) promoter, which allows for the selection of a heterozygous genotype. B Schematic workflow for engineering SRSF2 mutant UCB-derived CD34 + HSPCs. Delivery of CRISPR reagents via nucleofection and P95H or WT SRSF2 cDNA via recombinant adeno-associated virus serotype 6 (rAAV6), followed by fluorescent reporter-based purification of cells with heterozygous genotype. C Efficiency of heterozygous SRSF2 P95H/WT and SRSF2 WT/WT HSPC generation, indicated by dual reporter positive cells. Bars show mean ± SEM of five independent biological replicates. D Confirmation of the heterozygous genotype by Sanger sequencing of genomic DNA. Red arrow indicates the genomic position of the point mutation, which alters the P95 codon. E Volcano plot visualizing differential splicing events in SRSF2 P95H/WT HSPCs five days after introduction of the mutation. The data was generated from isogenic HSPC pairs derived from five individual UCB donors. Significantly differentially spliced genes (p adj <0.05) are shown in black and highly differentially spliced genes (p adj <0.05, Δ PSI > 10%) were highlighted in red. F Gene set enrichment analysis ( Enrichr webtool, maayanlab.cloud/Enrichr) of significantly differentially spliced genes (p adj <0.05). Enriched terms from the TISSUES_Curated cell type (orange) and Reactome pathway (green) gene set databases are depicted. G Venn diagram showing differentially spliced genes in SRSF2 P95H/WT cells overlapping between AML-, MDS-, and CMML-derived patient samples and the engineered UCB-derived HSPCs of this study. The AML, MDS, and CMML gene sets were generated by combining data from four, four, and two publicly available datasets respectively. All datasets are listed in Supplementary Table 1 . The 14 genes and the respective splicing event type, which overlap between all four entities are listed in the table on the right. SRSF2 mutations are known to cause genome-wide missplicing in primary AML and MDS patient samples as well as in cell line models [ 4 ]. While SRSF2 mutations affect multiple types of splicing events, changes in exon skipping events are most common, likely due to altered ESE-motif recognition [ 3 ]. To capture the earliest effects of the SRSF2 P95H/WT mutation, we performed RNA sequencing on UCB-derived HSPCs five days after genetic engineering and analyzed alternative mRNA splicing events. Even at this early stage, we detected widespread differential splicing, including missplicing events in both previously reported (e.g., ATF2, TPM3, INTS3) and thus far unreported genes (e.g., EFCAB14, OSBPL3 ; Fig. 1E , Supplementary Fig. 1D , Supplementary Table 3 ). This demonstrates that SRSF2 P95H/WT induces rapid cell-intrinsic changes in splicing patterns in absence of confounding effects from clonal selection or additional mutations. While exon skipping, intron retention, and alternative splice-site usage occurred at similar frequencies across all significantly differentially spliced genes (DSGs), exon skipping was enriched among genes with high-magnitude changes (ΔPSI > 10%; Supplementary Fig. 1E ). The top hits from our differential splicing analysis, including two previously described missplicing events in EZH2 and INTS3 [ 6 ], were validated by RT-PCR in UCB-derived HSPCs and showed concordant directionality in independently engineered mPB-derived HSPCs (Supplementary Fig. 2A-B ). Interestingly, we did not observe a correlation between the degree of missplicing and mRNA abundance in these DSGs (Supplementary Fig. 2C ), indicating that, at least initially, SRSF2 P95H/WT mutations mainly alter isoform ratios rather than broadly triggering transcript degradation through nonsense-mediated decay. Enrichment analysis linked these DSGs to leukemic cell states and cellular processes, ranging from regulation of TP53 activity to RNA metabolism, underlining the wide range of affected genes and contributors to leukemic transformation (Fig. 1F ). Engineered SRSF2 P95H/WT HSPCs quickly developed a DSG landscape that substantially overlapped with ten publicly available human RNA-seq splicing datasets from SRSF2 -mutant AML, MDS, and CMML patient samples (Fig. 1G , Supplementary Table 1 ). This confirms that our engineered SRSF2 P95H/WT HSPCs effectively recapitulate a disease-relevant pathological splicing pattern soon after mutation acquisition. Specifically, 14 genes were recurrently misspliced across all clinical entities and our engineered HSPC model (Fig. 1G ). Consistent with this overlap, we verified the most prominent missplicing events also in an independent cohort of SRSF2 P95H/WT AML, MDS, and CMML patient samples (Supplementary Fig. 2D, E , Supplementary Table 2 ) using RT-PCR. Overall, these findings suggest that a large component of the disease-related SRSF2 P95H/WT splicing program is established within days of mutation acquisition in primary human HSPCs. Because splicing changes alone do not define the developmental trajectory of mutant cells, we next examined transcriptomic changes in SRSF2 P95H/WT HSPCs. These changes may arise downstream of specific missplicing events, through broader SRSF2-dependent effects on RNA biology, or through indirect consequences of early lineage priming. Differential expression analysis revealed upregulation of monocyte-related genes ( FCN1, NCF2, TLR2, CLEC12A ) and down-regulation of erythroid lineage-related genes ( HBB , HBA1/2 , HBG1/2 , ALAS2 ) in SRSF2 P95H/WT HSPCs, indicating early priming of mutant cells toward the myelo-monocytic lineage (Fig. 2A , Supplementary Table 4 ). Enrichment analysis further supported this shift by showing an upregulation of myeloid cell signatures (monocyte and dendritic cell), as well as TNF-α and inflammatory response signaling in SRSF2 P95H/WT HSPCs, along with a reciprocal down-regulation of erythroid precursor cell and heme metabolism signatures (Fig. 2B , Supplementary Fig. 2F ). Given that CRISPR/Cas9 and rAAV6-mediatedgenetic editing can induce cellular stress and inflammatory programs in HSPCs [ 7 ], these inflammatory signatures should be interpreted cautiously. However, both SRSF2 P95H/WT and SRSF2 WT/WT cells underwent the same bi-allelic editing, rAAV6 exposure, culture duration, and reporter-based purification, supporting that genotype-specific differences reflect the mutant P95H allele rather than the editing procedure alone. The observed transcriptomic changes could be secondary events caused by differential splicing of upstream regulators or might reflect broader non-canonical SRSF2 functions that could influence lineage output, including R-loop accumulation, replication stress, ATR pathway activation, or the reading of epitranscriptomic m 5 C marks on mRNA [ 4 , 8 , 9 ]. Consequently, in-depth investigations are warranted to determine whether and how the observed splicing alterations are mechanistically linked to the transcriptional signatures. Fig. 2: SRSF2 P95H/WT mutations transcriptionally prime HSPCs toward monocytic differentiation. Full size image A Volcano plot visualizing differentially expressed genes in SRSF2 P95H/WT HSPCs five days after introduction of the mutation. The data was generated from isogenic HSPC pairs derived from five individual UCB donors. Significantly differentially expressed genes (p adj <0.05) are shown in black, whereas erythroid and monocytic lineage related genes were highlighted in red and blue respectively. B Gene set enrichment analysis (Enrichr webtool, maayanlab.cloud/Enrichr) of significantly up- or down-regulated genes (p adj <0.05). Enriched terms from the HuBMAP and PanglaoDB cell type gene set databases are depicted. C Dot plot visualizing total (TPM) and differential (fold change) expression of the top 100 differentially expressed cell surface genes identified by SurfaceGenie ( https://www.cellsurfer.net/surfacegenie ). D Median fluorescence intensity (MFI) of CLEC12A cell surface expression on HSPCs measured via flow cytometry over a culture period of 16 days. N = 3 independent biological replicates (UCB donors). E CLEC12A expression (MFI) on immature and mature cell subsets analyzed on day 11 of culture. F CLEC12A expression (mRNA) from BM-derived CD34 + HSPCs of MDS patients . Analysis of public data (GEO: GSE58831) [ 12 ]. Comparison between patient samples without splicing factor mutations (SF WT ) and with SRSF2 mutations but no other splicing factor co-mutations ( SRSF2 MUT ). G CFU assay showing total number of colonies (left) and frequency of monocytic (CFU-M) and erythroid (BFU-E) colonies (right) from N = 5 independent biological replicates (UCB donors). To identify potential therapeutic targets or specific biomarkers for SRSF2 P95H/WT -driven disease, we filtered the differentially expressed genes for predicted cell-surface localization. This analysis identified CLEC12A (CLL-1/CD371) as one of the most consistently upregulated candidate surface markers (Fig. 2C ). CLEC12A is a myeloid marker with prognostic and therapeutic relevance in AML and is highly expressed in myeloid precursors, monocytes, granulocytes, and dendritic cells [ 10 ]. A recent study linked CLEC12A expression on HSPCs to myeloid lineage restriction [ 11 ], suggesting it as a traceable marker for cells with early myeloid priming. Flow cytometry confirmed a significant increase of CLEC12A on the surface of UCB- and mPB-derived engineered SRSF2 P95H/WT HSPCs and their myelo-monocytic progeny (Fig. 2D, E , Supplementary Fig. 3A, B ). Supporting our findings, CLEC12A was similarly upregulated in SRSF2 P95H/WT HSPCs from an independent public MDS-patient dataset (Fig. 2F ) [ 12 ]. This CLEC12A upregulation appears to be specific to SRSF2 P95H/WT , as it was not observed in SF3B1 and U2AF1 mutants (Supplementary Fig. 3C ). Recently, CLEC12A has been proposed as a promising immunophenotypic marker for MDS/AML blasts, raising its clinical relevance and resulting in the development of bispecific antibodies and CAR-T cells [ 13 ]. Given the increased CLEC12A expression observed in engineered SRSF2 P95H/WT HSPCs and in an independent MDS dataset, we hypothesize that SRSF2 -mutant myeloid neoplasms may warrant a targeted evaluation in future CLEC12A-directed diagnostic or therapeutic studies. This hypothesis requires further studies for functional validation and clinical correlation. To assess whether this SRSF2 P95H/WT -induced transcriptional priming translates into a functional lineage bias with altered differentiation capacity of HSPCs, we performed methylcellulose colony-forming unit (CFU) assays. Consistent with the molecular signatures, UCB- and mPB-derived SRSF2 P95H/WT cells showed a significant increase in monocytic colonies with a concomitant reduction in erythroid colony formation (Fig. 2G , Supplementary Fig. 3D ). These data support that early transcriptional changes (upregulation of myelo-monocytic and the suppression of erythroid programs) are associated with a functional bias toward monocytic differentiation, while the molecular mechanisms connecting SRSF2 P95H/WT -induced missplicing to altered lineage output remain unresolved. Importantly, these results not only replicate findings from lentivirally manipulated HSPCs and transgenic mouse models, which display increased monocyte production and signs of anemia [ 14 , 15 ], but also correspond with clinical observations in CMML, an entity where up to 40% of patients harbor SRSF2 mutations. In conclusion, we demonstrated that the heterozygous SRSF2 P95H/WT mutation rapidly induces disease-relevant alternative splicing and is accompanied by transcriptional and functional features of monocytic priming in primary human HSPCs. This priming can be tracked through CLEC12A cell-surface expression, highlighting it as a potential biomarker and a candidate for further therapeutic exploration. In addition, we provide a novel, controlled primary human HSPC-model to dissect early leukemogenic events and to facilitate the identification of mutation-specific vulnerabilities by linking cell-intrinsic molecular signatures to disease-relevant phenotypes.