Distinct and convergent effects of SF3B1 mutations in human breast cancer

Authors

Bergman RE

Keeton CI

Sinha IR

Ling JP

Wang Y

Ha L

Mehta EP

Reed SC

Fox EK

O’Grady TM

Yu Y

Fragkogiannis S

Chao C

Davidson BA

Miranda AX

Croessmann S

Zhao S

Liu Q

Shyr Y

Dalton WB

Hurley PJ

Park BH

Doi

10.1073/pnas.2505374122

PMID: 22158541 · DOI: 10.1073/pnas.2505374122 · Journal: Proceedings of the National Academy of Sciences (2025)

TL;DR

This study assembled the largest clinical cohort of SF3B1-mutant breast cancer patients (N=420 from the Tempus database) and developed isogenic knock-in cell line models (MCF7, T47D, MCF10A) of the most common SF3B1 hotspot mutations (K700E, R625C, R625H). SF3B1 driver mutations occur in approximately 2.5% of HR+ HER2- breast cancer and are associated with significantly improved real-world overall survival (HR=0.67, P=0.046), attributable to enrichment in Luminal A disease and mutual exclusivity with TP53 mutations. Paradoxically, SF3B1 hotspot mutations are deleterious to cell growth in vitro and in vivo, with cells reverting to wildtype over time through copy number deletion of the mutant allele, providing a rationale for the low frequency and low VAF of these mutations in clinical samples.

Cohort & data

  • Clinical cohort: 420 patients with SF3B1-mutant breast cancer identified by Tempus xT targeted sequencing; 218 classified as Pathogenic/Likely Pathogenic (P/LP), 202 as VUS.
  • Cancer type: Breast cancer (BRCA), specifically HR+ HER2- subtype (N=215 total, 125 P/LP).
  • Survival analysis: 82 SF3B1 P/LP mutant vs. 3,474 WT patients with HR+ HER2- breast cancer; multivariate CoxPH adjusted for race, sex, age at diagnosis.
  • Cell models: Isogenic knock-in panel in MCF7, T47D, and MCF10A cell lines using CRISPR-Cas9 and AAV-mediated gene targeting.
  • Sequencing: Whole exome sequencing (WES) via Twist Exome on Illumina NovaSeq XP at 100X coverage; RNA-seq via Illumina NovaSeq X Plus targeting 50M reads/sample. Data deposited: SRA PRJNA1299110, GEO GSE304350.
  • Methods: Whole exome sequencing, RNA-seq, droplet digital PCR (ddPCR), orthotopic xenograft models.

Key findings

  • SF3B1 P/LP mutations occur at 2.48% frequency in HR+ HER2- breast cancer, with K700E accounting for 84.86% of all P/LP mutations.
  • Patients with SF3B1 P/LP mutations in HR+ HER2- breast cancer had significantly improved rwOS compared to WT (median OS 34 mo vs. 27 mo; HR=0.67, 95% CI 0.45-0.99, P=0.046).
  • SF3B1 mutations were significantly enriched in Luminal A disease; when stratified by PAM50 subtype, survival differences were no longer significant.
  • TP53 alterations are significantly less common in SF3B1-mutant breast cancer (mutual exclusivity confirmed across multiple cBioPortal datasets including ER+ breast, pan-cancer, melanoma, and AML).
  • Heterozygous SF3B1 hotspot mutations (K700E, R625C, R625H) significantly decreased cell growth in MCF7, T47D, and MCF10A cell lines.
  • R625 mutations were not tolerated in nontumorigenic MCF10A cells, while K700E and K666N were stably maintained, providing a rationale for lineage-specific hotspot enrichment.
  • SF3B1 mutant cells reverted to wildtype over serial passage (MCF7 most rapidly), with mutant VAF decreasing significantly by passage 5-10.
  • Reversion occurred through copy number deletion of the mutant SF3B1 allele; reverted WT clones retained the growth deficit and showed significantly increased CNVs genome-wide.
  • WT SF3B1 overexpression rescued the growth deficit in both mutant and reverted WT cells, supporting a loss-of-WT-function mechanism.
  • Convergent hallmark pathway enrichment across all hotspot mutants: negative enrichment of E2F targets, G2M checkpoint, MYC targets, oxidative phosphorylation; positive enrichment of inflammatory signaling, apoptosis, p53 pathway.
  • Top alternatively spliced transcripts shared across hotspots: DLST, GCC2, TMEM14C, ENOSF1.
  • Alternative 3’ splice site usage correlated more strongly with mutant VAF than with specific hotspot identity, supporting dose-dependent splicing effects.

Genes & alterations

  • SF3B1 — Hotspot mutations K700E (84.86% of P/LP), R625C/G/H/L (5.05%), K666E/N/Q (2.29%), K741E (1.38%). All mutations heterozygous; VAF never exceeds 0.5. Median VAF 0.2 overall; significantly higher in Stage IV (0.21) vs. Stages I-III (0.15). Deleterious to cell growth; reverts to WT via copy number deletion.
  • TP53 — Significantly mutually exclusive with SF3B1 across breast cancer and pan-cancer datasets. Comutation does not result in synthetic lethality (tested by CRISPR-Cas9 knockout of p53 in MCF10A K700E). Mutual exclusivity partially drives the improved prognosis of SF3B1-mutant tumors.
  • PIK3CA — Comutation with SF3B1 does not lead to significantly worse survival in advanced-stage HR+ HER2- breast cancer when TP53 status is controlled for. Endogenous PIK3CA E545K VAF remained stable in xenografts while SF3B1 mutant VAF declined.
  • MKI67 — Expression significantly lower in SF3B1 P/LP mutant HR+ HER2- breast cancer, consistent with Luminal A enrichment.
  • MYC — MYC targets negatively enriched in SF3B1 mutants; positively enriched in reverted WT clones. Convergent pathway across all hotspots.

Clinical implications

  • Prognosis: SF3B1 P/LP mutations are associated with improved real-world overall survival in advanced-stage HR+ HER2- breast cancer, but this is attributable to enrichment in Luminal A disease and mutual exclusivity with TP53 rather than a direct protective effect of SF3B1 mutation.
  • Biomarker context: The low frequency (~2.5%), low VAF, and tendency toward reversion suggest SF3B1 mutations may be transient events in breast tumorigenesis that require permissive secondary events (e.g., copy number changes) to be maintained.
  • Therapeutic relevance: The paper does not identify actionable therapeutic targets, but notes that the convergent pathway alterations (inflammatory signaling, p53, MYC) and alternatively spliced transcripts (DLST, GCC2, TMEM14C, ENOSF1) may guide future therapeutic strategies. Synthetic lethality approaches (e.g., PARP inhibitors for SF3B1-mutant BRCA-like phenotypes, referenced from other studies) remain an area of interest.

Limitations & open questions

  • The Tempus clinical cohort is enriched for Stage IV disease; Stage I-III disease is underrepresented, limiting comparisons between early and advanced-stage disease.
  • Non-K700E hotspot mutations had very low sample sizes (n=9 for non-K700E in survival analysis), preventing definitive conclusions about hotspot-specific prognosis.
  • MCF10A cells are triple-negative; no ER+ nontumorigenic cell models exist, limiting the translational relevance of findings in untransformed cells.
  • The mechanism underlying lineage-specific enrichment of individual hotspots (e.g., R625 in melanoma vs. K700E in breast) remains unresolved.
  • Clinical data are proprietary (Tempus); restricted access limits independent replication.
  • Immunocompromised xenograft models cannot assess the role of immune selection, which may be important given SF3B1-driven neoantigen production.
  • Whether the observed reversion and copy number changes occur in patient tumors (vs. cell culture artifacts) requires single-cell resolution clinical studies.

Citations from this paper used in the wiki

  • “SF3B1 driver mutations in approximately 2.5% of HR+ HER2- breast cancer, with strong enrichment of K700E substitutions” (Abstract)
  • “rwOS…was significantly improved [HR = 0.67, 95%CI (0.45, 0.99), P = 0.046] in patients with SF3B1 hotspot mutations (median OS = 34 mo) compared to WT (median OS = 27 mo)” (Results, p.2)
  • “The K700E hotspot mutation accounted for 84.86% of all P/LP SF3B1 mutations” (Results, p.2)
  • “TP53 alterations are significantly less common in SF3B1 mutants” (Results, p.3)
  • “heterozygous incorporation of SF3B1 hotspot mutations led to a significant decrease in cell growth” (Results, p.4)
  • “R625C and R625H mutations are only transiently tolerated” in MCF10A cells (Results, p.4)
  • “the mutation was lost through copy number deletion” (Discussion, p.9)
  • “WT overexpression (OE) rescues the observed growth deficit” (Results, p.2)

This page was processed by paper-compiler on 2026-05-06.