Splicing Factor Mutations & the Leukemia Proteome

A probabilistic, isoform-resolved DIA-MS proteogenomic analysis

← Research

Background

Recurrent mutations in the splicing factor genes SF3B1, U2AF1, and SRSF2 are among the most common drivers of myelodysplastic syndromes (MDS) and secondary acute myeloid leukemia (sAML). All three disrupt pre-mRNA splicing and are thought to drive disease through dysregulated gene expression. However, while their transcriptomic consequences are well characterized, their downstream impact on the proteome—the functional layer where splicing-driven changes are ultimately executed—remains comparatively underexplored. This project profiles the proteomes of CRISPR/Cas9-engineered K-562 leukemia cells carrying the SF3B1-K700E, U2AF1-S34F, or SRSF2-P95H mutations, alongside an isogenic wild-type control, using label-free DIA mass spectrometry.

Approach

Two key features distinguish this analysis from a standard differential-expression study:

First, quantification utilizes {limpa}, a probabilistic, imputation-free framework (Li, Cobbold & Smyth). By modeling a Detection Probability Curve for each sample, it derives protein abundance estimates directly from the pattern of observed and missing data, circumventing the imputation assumptions that can severely bias results.

Second, identification relies on a customized proteogenomic database search. By integrating RNA-seq-informed splice-junction and variant sequences with PEAKS’ de novo, Sequence Variant, and Novel Peptide search modules, we capture splicing-associated proteoforms directly at the peptide level.

Building on this, I developed a three-tier differential-usage framework to maximize biological resolution:

  • Differential Protein Group Abundance (DPGA): Asks whether a gene’s total translational output shifted.
  • Differential Isoform Group Usage (DIGU): Asks whether the relative usage of its known isoforms changed.
  • Differential Peptide Usage (DPU): Asks whether individual peptides deviate significantly from their parent protein’s overall abundance trend, exposing localized structural dynamics—such as cryptic splicing events, altered proteolytic processing, or changes in post-translational modifications (PTMs)—that protein-level rollups obscure.

Key findings

  • Across 6,113 quantified protein groups, all three mutants display widespread, mutation-specific remodeling relative to wild-type, resolved simultaneously across a full set of pairwise mutant-vs-mutant contrasts.
  • Strikingly, 39 protein groups—including HNRNPA2B1, HNRNPK, HNRNPM, the splicing kinase SRPK1, and two distinct proteoforms of PTBP1—are supported exclusively by novel splice-junction peptides that would be completely invisible to a canonical-only database search.
  • Isoform-resolving peptides enable us to distinguish differentially regulated isoforms from their unchanged counterparts within the same gene. For instance, HNRNPH3 is consistently down-regulated across all three mutants, while NDUFV3 exhibits a genuine isoform-usage shift accompanying its abundance change specifically in SRSF2-P95H cells.
  • Pathway analysis highlights mutation-specific remodeling concentrated within the spliceosome and RNA-processing machinery, alongside strong signatures in innate-immune signaling and hematologic-malignancy pathways.
Status

Presented as an oral talk at ASMS 2025 (Baltimore, Cancer Research session); manuscript in preparation with the Manley lab. Collaborators: Pedro Bak-Gordon, David S. Johnson, Chloe J. Jones, Claire E. Sattler, Lewis M. Brown, James L. Manley (Columbia University).