Pradeep, Pankaj

Abstract While missense somatic mutations in the serine/arginine rich splicing factor 2 (SRSF2) occur early in the clonal development of myeloid neoplasms, they are uniquely enriched in chronic myelomonocytic leukemia (CMML), an MDS/MPN overlap neoplasm, where their estimated prevalence is ∼50%. Short-read RNA-sequencing experiments have demonstrated that missense alterations of the proline 95 residue hotspot (SRSF2P95) globally dysregulate exon inclusion, though the full transcript structures induced by such alternative splicing events remain uncharacterized. Given the striking enrichment of SRSF2 mutations in CMML, we hypothesized that there exist unique splicing alterations contributing to the biology of this disease, and sought to profile the landscape of alternative transcription induced by SRSF2 mutations in CMML using long-read RNA-Seq. We conducted bulk long-read RNA-Seq on bone marrow mononuclear cells from 5 CMML patients with SRSF2P95H (n=3) and SRSF2P95L (n=2) mutations, with VAFs ranging 20-48%, along with 5 CMML splicing factor wildtype patients. Libraries were prepared with the Pacbio Kinnex full-length RNA kit and sequenced on a Revio SPRQ sequencer. Reads were processed and aligned using the IsoSeq pipeline, and isoform collapsing, quality control, reference comparison, and quantification were performed with IsoQuant and SQANTI3. Differential transcript and gene expression was assessed with DESeq2, while differential transcript usage was assessed with the DRIM-Seq and stageR packages. After long-read collapsing and quality control, we identified 255,505 unique transcript isoforms, of which 14,845 (∼5.8%) were novel to the reference transcriptome annotation (GENCODE v39). We identified 23 transcripts differentially expressed in SRSF2-mutant samples, including 9 novel transcripts, the majority of which (n=7) were upregulated in SRSF2-mutant cases. A higher number of transcripts (n=80) and genes (n=65) demonstrated significant changes in proportional transcript usage. Among transcripts with increased expression and usage in SRSF2-mutant cases, we identified a novel transcript of the gene YBX1, which encodes an RNA- and DNA-binding transcription factor required for cell survival and self-renewal in myeloid neoplasms. This novel transcript, which we termed YBX1ΔEx4, has an in-frame loss of exon 4, and is predicted to encode a full-length protein missing a 30 amino acid region of the cold-shock (DNA-binding) domain. Using short read RNA-Seq data from an additional 29 CMML patients, as well as an independent dataset (GSE251806) comprising 43 blast-phase CMML patients, we confirmed increased exclusion of YBX1 exon 4 in SRSF2-mutant cases, with an even greater loss of exon 4 in blast-phase SRSF2-mutant cases. We hypothesize an important role of this isoform in CMML disease progression and are currently performing functional studies to characterize the effect of YBX1ΔEx4. Citation Format: Nickolas Steinauer, Terra Lasho, Christy Finke, Pankaj Pradeep, Jenna Fernandez, Alejandro Ferrer, Moritz Binder, Abhishek Mangaonkar, Mrinal M. Patnaik, . Long-read RNA-Seq reveals a novel isoform landscape of SRSF2-mutant chronic myelomonocytic leukemia [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 1497.

Abstract Introduction: VEXAS (vacuoles, E1 enzyme, X-linked, autoinflammatory, somatic) is an adult-onset autoinflammatory disorder defined by somatic UBA1 mutations in hematopoietic stem cells (HSCs). Given the higher median age at onset and prevalent inflammation, approximately 60% of patients have concurrent clonal hematopoiesis (CH) involving epigenetic regulators, DNMT3A and TET2. We have previously demonstrated (Gutierrez-Rodrigues, Blood 2023) that DNMT3A and TET2 mutations in VEXAS are independently associated with increased mortality, without an increased risk of myelodysplastic syndrome (MDS) or acute myeloid leukemia (AML). To define the role of CH in VEXAS, we investigated the impact of CH on mortality and related epigenetic dysregulation in primary VEXAS patient samples and validated these findings using an adenine base-editing (ABE) strategy to model cooccurring UBA1 and DNMT3A mutations in CD34+ HSCs. Methods: After IRB approval, 114 VEXAS patients from the Mayo Clinic and the NIH were retrospectively assessed, including 104 screened for CH by error-corrected DNA sequencing (ECS). DNA methylation changes were identified using the Illumina Infinium MethylationEPIC BeadChip and genome-wide using Biomodal 6-base sequencing on peripheral blood samples (PB). Cellular indexing of transcriptomes and epitopes (CITE)-seq was performed using the 10x Chromium 3' GEX platform; 21,114 pooled cells were subjected to UMAP dimensionality reduction, followed by Azimuth based cell identification. HSCs from different donors were base edited for UBA1 M41T mutation with and without cooccurring DNMT3A splice-site mutations. Levels of cytokines in cell culture supernatants were measured using Olink technology, transcriptomic changes were assessed using bulk RNA-seq. Statistical analyses were carried out using R. Results: Patients' median age was 68 (62-74) years; 95% were white and 99% were males. 39% had a pathogenic CH variant, with DNMT3A (18%; median VAF: 42%; range 6-46%) and TET2 (11.5%; median VAF: 11%; range: 2-37%) being most common (2% had both). While the presence of any CH mutation did not impact survival (7.7 vs 11.5yrs, p =0.69), DNMT3Amt CH had an age-independent negative impact on mortality (9.2 vs 5.6yrs, p=0.02), without an increased risk of MDS/AML (p=0.31). Changes in DNA methylation were profiled using the methylation array in 19 VEXAS patients without CH (Median Age: 68 years) and in 9 with DNMT3Amt CH (Median Age: 74 years), respectively. VEXAS samples with DNMT3Amt CH displayed a hypomethylation phenotype compared to VEXAS samples with no CH, with 4,096 unique hypomethylated CpGs, occurring mainly at gene enhancers (Enh) and transcriptional states (Tx, TxWk). This loss of DNA methylation was enriched in inflammatory pathways with extensive hypomethylation observed in the promoters and gene bodies of IL-1 family genes. CITE-seq performed on 7 VEXAS samples (4 DNMT3Amt CH vs 3 no CH) demonstrated an upregulation of gene expression predominantly in T-cell subsets. This increased expression was enriched in inflammatory genes, largely centered around IL-1 and IL-18 family signaling. Pathway analysis revealed an enrichment in respiratory electron transport and oxidative phosphorylation (oxphos), with the VEXAS samples with DNMT3Amt CH demonstrating significantly increased oxphos gene signatures. Base editing of CD34+ cells was done using an adenine base-editing (ABE) strategy to model dual UBA1 and DNMT3A splice-site mutations. We observed that the HSCs base edited for the UBA1 M41T recapitulated the cellular hallmarks of VEXAS demonstrating reduced ubiquitination, unfolded protein response (UPR) activation, and increased inflammation. The addition of cooccurring DNMT3A mutations resulted in decreased DNMT3A expression, further increasing inflammatory markers, confirming altered inflammatory and metabolic pathways. Conclusion Our study validates the age-independent, negative prognostic impact of DNMT3Amt CH on survival in VEXAS syndrome. DNMT3A mutations mediate epigenetic and transcriptional priming associated with upregulation of inflammatory and mitochondrial function pathways. These findings have been validated in a base edited cell model with UBA1 and DNMT3A co-mutations.

Abstract Introduction: ASXL1-mutant (mt) Chronic Myelomonocytic Leukemia (CMML) is an aggressive myeloid neoplasm characterized by increased proliferation, resistance to epigenetic therapies, and poor survival. Epigenetic remodeling and up-regulation of leukemogenic driver genes are hallmarks of ASXL1mt CMML. Truncating mutations in ASXL1 activate enhancers that drive the expression of key leukemogenic driver genes such as HOXA9 and its co-factor MEIS1. Here, we investigate the chromatin accessibility of these leukemogenic enhancers by ASXL1 genotype at single-cell resolution (GoT-ChA-seq). Methods: After IRB approval, we performed bulk RNA-seq and ChIP-seq (H3K4me1, H3K4me3, H3K27ac, H3K27me3, H2AK119ub, H2BK20ac) on bone marrow (BM) mononuclear cells from 40 patients with CMML (19 ASXL1mt, 21 -wildtype/wt). After quality control, differential analyses of gene expression and histone modification occupancy between ASXL1mt and wt patients were conducted. A consensus peak calling framework was used to define genomic regions with ASXL1mt-specific histone modifications. Canonical enhancers were defined as genomic regions with a co-occupancy of H3K4me1 and H3K27ac. We then added regions marked by H2BK20ac to capture regulatory regions associated with p300 (Narita et al. Nat Genet 2023). We used an in silico prediction strategy to associate regulatory regions marked by strong histone modification occupancy with highly up-regulated leukemogenic driver genes (HOXA6-10, MEIS1). The ReMap database (Hammal et al. Nucleic Acids Res 2022) was queried for transcription factor binding in the genomic regions of interest. We then performed GoT-ChA-seq (Izzo et al. Nature 2024) to investigate chromatin accessibility for ASXL1mt (n=3) and wildtype BM single cells from CMML patients. In vitro colony forming assays were conducted on primary CMML BM cells (11- ASXL1mt and 9 wt CMML), assessing the efficacy of EP31670, a novel, oral, dual inhibitor of p300/BRD4. Results: This study included 40 patients with CMML, 19 ASXL1mt and 21 wt. The variant allele frequencies for ASXL1 were compatible with heterozygosity (median 0.41) and TET2 mutations were balanced between the two groups. We identified 149 candidate regulatory regions correlating with the expression of up-regulated target genes of interest (HOXA6-10, MEIS1) by bulk ChIP- and RNA-seq. Publicly available ChIP-seq data from myeloid cell lines demonstrated enrichment of p300 (in THP-1 cells, enrichment effect size +2.58, p=0.002) and BRD4 (in K-562 cells, enrichment effect size +1.58, p=0.045) in these candidate regulatory regions. Next, we interrogated the chromatin accessibility of ASXL1mt (n=13620) and ASXL1wt (n=24269) single cells from 3 CMML patients using GoT-ChA-seq. Differential accessibility analysis revealed 144 genomic regions, all of which were preferentially accessible in ASXL1mt cells compared to ASXL1wt cells (p<0.05 for all regions). Four of these preferentially accessible genomic regions (p<0.001 for all regions) overlapped with the previously identified candidate regulatory regions for MEIS1. The H3K27ac and H2BK20ac occupancy levels in these regions was strongly associated with the expression of MEIS1 (median Pearson r=0.73, p=1.16x10-7). These candidate regulatory regions were annotated in ENCODE as distal and proximal enhancers. Targeting this biology with the dual p300/BRD4 inhibitor EP31670 in progenitor colony forming assays led to a preferential therapeutic effect in ASXL1mt BM samples (median IC50 = 25nM) compared to in ASXL1wt samples (median IC50 = 74nM, p=0.03). Treatment with EP31670 was also associated with increased differentiation (increased blast forming units-erythroid; BFU-E) and organization of colonies in ASXL1mt vs wt CMML. Conclusions: ASXL1mt CMML is characterized by the up-regulation of several leukemogenic driver genes including HOXA9 and its co-factor MEIS1, secondary to the activation of genotype-specific enhancers. The preferential accessibility of these enhancers can be defined between ASXL1mt and ASXL1wt cells within patients. The enhancers of interest interact with p300 and BRD4 in myeloid cell lines and targeting them with the novel dual p300/BRD4 inhibitor, EP31670, induces therapeutic responses at low nanomolar drug concentrations in vitro, preferentially in ASXL1-mutant cells. These data solidify the biological rationale for targeting p300/BRD4 in ASXL1mt chronic myeloid neoplasms for therapeutic benefit (NCT05488548).

Abstract Introduction: Protein truncating somatic ASXL1 mutations in CMML are associated with high-risk features and poor outcomes. We have previously demonstrated that ASXL1-mutant (mt) CMML is characterized by permissive chromatin states and overexpression of leukemia stemness genes (Binder et al. Nat Commun; 2022). DLK1 (delta-like homolog 1) is a pseudo-NOTCH like ligand that is maternally imprinted and paternally expressed and is physiologically not expressed beyond embryogenesis. Aberrant DLK1 expression has been associated with poor outcomes in solid tumors and we have shown differential overexpression in ASXL1mt versus wild-type (wt) CMML. Both the promoter and the last exon of DLK1 contain CpG islands that play a role in regulating DLK1. The CpG island within the fifth exon of DLK1, known as the DLK1-differentially methylated region (DMR) [chr14:100,726,892-100,738,224] is completely unmethylated on the maternal allele and partially methylated on the paternal allele. Previous work (Developmental Cell; 2021) has also identified a DLK1 imprinting control region (ICR) known as the IG-DMR [chr14:100,811,001–100,811,037]. DNA methylation state of the IG-DMR has been shown to impact the expression of MEG3 mRNA (lncRNA) which functions as a cofactor to direct PRC2 to the DLK1 promoter, suppressing DLK1 expression. Since the regulation of DLK1 in ASXL1mt CMML is not well understood, we conducted this multiomic analyses to define epigenetic mechanisms of DLK1 overexpression in ASXL1mt CMML. Methods: After IRB approval, we performed bulk RNA-seq, ChIP-seq (H3K4me1/me3, H3K27ac, H3K27me3) and Illumina MethylationEPIC array on bone marrow (BM) mononuclear cells of 40 CMML patients (19 ASXL1mt, 21 wt). scRNA-seq was conducted on 5 ASXL1mt and 3 wt cases. All libraries were sequenced on an Illumina NovaSeq instrument. We assessed these samples for differential gene expression (RNA-seq) and changes in histone modifications (ChIP-seq) between ASXL1mt and wt CMML, specifically at the DLK1-DMR and IG-DMRregulatory loci. We conducted ddPCR (digital droplet PCR) to determine allelic status of DLK1 in ASXL1mtand wt cases. Global and allele/sequence specific DNA methylation changes were determined using Biomodal's duet EvoC methodology (6 base sequencing, including 5mC and 5hmC) in 5 ASXL1mt and 5 wt CMML cases. Results: We validated the overexpression of DLK1 in 19 ASXL1mt versus 21 wt (log2 fold change +4.85, FDR 4.67e-13) CMML BM samples. scRNA-seq on 5 ASXL1mt versus 3 wt samples demonstrated that DLK1 is only expressed in mutant cases, with the expression being limited to HSPCs. ChIP-seq analysis at the DLK1 regulatory loci validated (a) reduced occupancy of H3K27me3, correlating with ASXL1mt associated disruption of the PRC2-complex, and (b) increased co-occupancy of H3K4me1 and H3K27ac in mt cases, validating a permissive chromatin state in ASXL1mt cases at the DLK1 regulatory regions. Through correlation analysis of RNA-seq and ChIP-seq, we defined the promoter for DLK1, characterized by high Pearson correlation coefficients (R > 0.5) and enrichment of H3K27ac and H3K4me3. Differential methylation analysis identified global hypomethylation around the DLK1-DMR and promoter regions. We observed a mean reduction of ~40% in methylation status at the DLK1 promoter in ASXL1mt versus wt cases, suggesting a loss of epigenetic silencing at this locus. This was corroborated by significant enrichment of 5hmC at the fifth exon of DLK1 indicating active DNA demethylation. Conversely, we observed a gain in methylation at the IG-DMR, suggesting loss of imprinting control through MEG3 silencing and DLK1 overexpression. To further determine the impact of imprinting deregulation on overexpression of DLK1 in ASXL1mt CMML, allele-specific methylation was profiled using the Biomodal whole genome methylation data. This demonstrated that the methylation patterns in ASXL1mt CMML cases were significantly deregulated at the DLK1-DMR, with >30% loss of methylation compared to GRCh38 reference alleles. Furthermore, ddPCR was able to confirm bi-allelic expression of DLK1 in ASXL1mt CMML, suggesting that the observed methylation dysregulation leads to loss of imprinting of the maternal allele. Conclusion: In summary, we demonstrate that ASXL1 mutations in CMML disrupt epigenetic pathways, promoting DNA demethylation, resulting in permissive chromatin states and an aberrant and oncogenic re-expression of maternal DLK1, due to loss of imprinting control.

Abstract The importance of bone marrow (BM) microenvironment and its immune interactions within and their role in chronic myelomonocytic leukemia (CMML) disease progression have not been well studied. Our previous work in CMML identified type 1 helper T cell (Th1) and CD4 terminal effector cell suppression in peripheral blood from untreated patients, in comparison to age-matched normal controls (Mangaonkar A et al. Blood Adv 2020). Furthermore, clonal dendritic cell (DC) aggregates in CMML are present in approximately 20-30% patients at diagnosis, display significant cell heterogeneity (plasmacytoid DCs + myeloid DCs and other myeloid/immune cells), express indoleamine 2,3-dioxygenase-1 (IDO), associate with T regulatory cell (Treg) expansion, and transformation to AML [blast transformation or BT, Lucas N et al. Leukemia 2019, Mangaonkar A et al. Blood Adv 2020, and Mangaonkar A et al. Blood 2023 (abstract)]. The purpose of this study was to define the myeloid-immune cell interactions within the BM microenvironment critical for CMML progression to AML. To phenotypically identify cell types and their interactions within the CMML BM, we used single-cell type spatial omics through two distinct modalities on archival formalin-fixed paraffin-embedded (FFPE) BM biopsy specimens. Ten serial CMML samples at chronic phase [CP, n=5, median BM blast% 5 (2-15)] and at BT [n=5, median BM blast% 60 (25-74)] were used for both techniques. Spatial transcriptomics was conducted through GeoMx Digital Spatial Profiler (DSP) platform that uses UV-photocleavable, spatial-barcoded oligo probes to profile the whole transcriptome in defined regions of interest (ROI). CD123 (marker for DC-myeloid aggregates) immunofluorescence antibody was used to identify and segment ROI. 16-color flow cytometry was done to validate cell-specific expression. Spatial cellular phenotyping of immune cells on the same samples (serial cuts) was done through multiplex imaging by co-detection by indexing (CODEX) using 27 markers: CD31, Ki67, CD11b, CD20, CD44, Ecad, CD123, CD107a, CD8, CD11c, CD45RO, FoxP3, CD14, PDL1, CD56, CD68, CD163, PD-1, CD38, HLA-DR, CD4, IDO, CD45, PanCK, Na/K/ATPase, CD3e, and DAPI. CD8T+ cell anergy was demonstrated through CFSE cell proliferation assay. Benjamini-Hochberg correction was used to adjust for multiple hypotheses. After filtering poor quality ROIs [ROIs with total read count < 1000 were removed, ROIs with trim, stitch, alignment < 80% or saturation < 50% were removed], 3rd-quartile normalization [BiocGenerics (v. 0.48.1)] and batch correction [limma (v. 3.58.1)], we compared transcriptomics signatures between CD123+ and CD123- ROIs in all samples. Results showed that LIPA, SLC4A1, HBA2, TIMD4 and HMOX1 werethe top five genes upregulated in CD123+ areas. Among all gene transcripts, TIMD4 expression increased the most (Log2FC=0.6, P-value=8.79E-10, FC=1.52, FDR=1.68E-07) when CD123+ ROIs were compared between CMML-CP and CMML-BT serial samples, suggesting a biological relevance in disease progression. Previous work in metastatic peritoneal cancer has established that TIM-4 expressing macrophages suppress Th1/Th2 ratios and impair CD8+ T cell function (Chow A et al. Cancer Cell 2021). To confirm expression at the protein level, we used a multicolor flow cytometry panel and confirmed that TIM4 is expressed on macrophages (Lineage-negative, HLA-DR+, CD68+ cells) in independent CMML BM aspirate samples (n=3). Next, we conducted spatial phenotyping using CODEX on serial cuts from the same serial samples used for spatial transcriptomics. Anatomical co-localization of macrophages with leukemic cells, monocytes and T cells was visually inspected and confirmed. Cell-interaction analysis (after Benjamini-Hochberg correction) indicated that macrophages-monocyte, macrophage-leukemic cells, and macrophage-memory CD4+T cells were significantly (P<0.05) increased in CMML-BT versus CMML-CP. CD8T+ cell anergy was confirmed through a CFSE cell proliferation assay in untreated CMML PBMC samples. In conclusion, our data suggests that TIM4-expressing macrophages and their interactions with monocytes/monocyte precursors and immune cells within a DC aggregate create a protective leukemic stem cell niche and T cell tolerance, promoting blast transformation in CMML.

Abstract Introduction: Somatic mutations in splicing factors (SF) - SF3B1, SRSF2, U2AF1, and ZRSR2 – represent one of the most common drivers of premalignant clonal hematopoesis. While SF mutations are potent drivers of clonal expansion (Fabre et al., Nature 2022), multiple SF mutations are rarely observed in the same individual (negative epistasis). Prior studies have identified rare cases of concurrent SF mutations (<4%) and described specific hotspot combinations that escape negative epistasis (Taylor et al., Blood 2020). We conducted this study to further characterize the frequency and clonal relationship between concurrent SF mutations using a larger patient cohort. Methods: After IRB approval, we assessed 3,369 myeloid next generation sequencing (NGS) results that identified somatic variants in SF3B1, SRSF2, U2AF1, or ZRSR2, in 2,500 adult patients from 08/2015 to 07/2025. Patient demographics, ICD-9/10 codes, and complete blood counts with differentials were obtained, retaining only those datapoints assigned 30 days before or after NGS testing. To allow comparison of clone size between autosomal and the X-linked SF ZRSR2, VAFs were multiplied by two to estimate the cancer cell fraction (CCF), except for ZRSR2 mutations in males. For patients harboring concurrent SF mutations, we categorized them as co-dominant if the difference in respective CCFs was < 10% of the greater CCF. Specific SF variants were categorized into common hotspots (n=13), except for ZRSR2 variants, which do not occur in defined hotspots. Fisher's exact testing was used to assess the enrichment of SF hotspot variants in patients with concurrent SF mutations, as well as combinations of SF variants in patients with co-dominant SF mutations. Results: We identified 2,331 patients with at least 1 pathogenic SF mutation, of which 6.7% (n=156) had > 1 SF mutation (multi-SF MT). The total cohort comprised cases of AML (n=549, 24%), MDS (n=977, 42%), MDS/MPN (n=233, 10%), MPN (n=228,10%), and uncategorized ICD codes (“Other”, n=326), of which the largest category was clonal cytopenia(s). Notably, the prevalence of multi-SF cases was higher than previously described and our cohort included 5 patients with 3 SF mutations and 1 with 4 SF mutations. Of the multi-SF MT patients, we identified 42 (27%) with co-dominant CCFs among their respective SF mutations. Thirty-three multi-SF MT patients had multiple NGS sampling timepoints (median 2, range 2-4) spanning a median of 0.97 years (range 0.06-5.69 years); we found that 17 of these 33 demonstrated correlative SF CCFs over multiple timepoints, suggesting their SF mutations occurred in the same clone (“co-clonal”). Of these 17 patients, 82% had co-dominant CCFs at the single-timepoint level, whereas none of the 16 non-co-clonal demonstrated co-dominance. Thus, we used SF co-dominance as a surrogate for cases likely to be co-clonal. ZRSR2 (p=8.8x10-29), U2AF1Q157 (p=9.4x10-9), SF3B1K666 (p=1.4x10-4) and infrequent SF3B1 (“SF3B1-Other”, p=.0029) mutations were overrepresented in the multi-SF MT cohort compared to patients with a single SF mutation. Within the multi-SF MT cohort, 3 SF mutation combinations showed overrepresentation in codominant cases – SRSF2P95R/H/L+ ZRSR2 (p=0.014), SRSF2P95R/H/L+ SF3B1K666 (p=0.044), and U2AF1S34 + U2AF1Q157 (p=0.024), although not significant after multiple hypothesis correction. Dual ZRSR2 mutations were the most common combination seen in the multi-SF MT cohort but were underrepresented in codominant cases (p=0.025). Although there was no disease subtype uniquely associated with total or co-dominant multi-SF MT cases, occurrence of multiple SF mutations was more common in males (8% vs. 3% in females), an effect driven by ZRSR2 mutations. Hemoglobin and WBC were lower in multi-SF MT cases, with the lower hemoglobin strongly driven by the codominant multi-SF MT cases (median 8.3 g/dL vs 9.4 g/dL, p=0.00046). Conclusions: These findings suggest that SF mutation co-occurrence in myeloid neoplasms is more prevalent than previously estimated (~7%). Furthermore, we identified 6 individuals harboring > 2 SF mutations, which has not been previously described. We validate specific SF variant pairs that appear to exist in the same clone and are working to confirm the clonal architecture and characterize global splicing aberration in selected cases with single-cell studies to fully uncover the mechanism by which negative epistasis is overcome.