Distinct routes of clonal progression in SF3B1-mutant myelodysplastic syndromes

Abstract

Myelodysplastic syndromes (MDS) are clonal stem cell disorders driven by heterogeneous genetic alterations leading to variable clinical course. MDS with splicing factor SF3B1 mutations is a distinct subtype with a favorable outcome. However, selected comutations induce poor prognosis and how these genetic lesions cooperate in human hematopoietic stem and progenitor cells (HSPCs) during disease progression is still unclear. Here, we integrated clinical and molecular profiling of patients with SF3B1 mutations with gene editing of primary and induced pluripotent stem cell-derived human HSPCs to show that high-risk comutations impart distinct effects on lineage programs of SF3B1-mutant HSPCs. Secondary RUNX1 or STAG2 mutations were clinically associated with advanced disease and reduced survival. However, RUNX1 and STAG2 mutations induced opposing regulation of myeloid transcriptional programs and differentiation in SF3B1-mutant HSPCs. Moreover, high-risk RUNX1 and STAG2, but not low-risk TET2, mutations expanded distinct SF3B1-mutant HSPC subpopulations. These findings provide evidence that progression from low- to high-risk MDS involves distinct molecular and cellular routes depending on comutation patterns.

Graphical abstract

Figure 1.

SF3B1 mutation constrains the spectrum of drivers of clonal progression. (A) Distribution of MDS/AML diagnostic categories in the SF3B1-mutant patient cohort based on WHO 2016 classification. Mutation analysis was restricted to the highlighted low-risk (BM blasts <5%) and high-risk groups (BM blasts ≥5%). (B) Frequency of SF3B1 comutated driver genes in the SF3B1-mutant patient cohort. n, number of mutated patients. Dashed box indicates genes mutated in ≥4% of patients. Only genes mutated in ≥1% of patients are shown. (C) Odds ratio distribution of recurrent SF3B1 comutated genes vs mutational frequency in the high-risk group. Red color indicates genes with the highest odds ratio of HR disease. (D) Hazard ratio for the risk of progression to AML or death, based on univariate Cox regression analysis. Red color indicates genes with the highest hazard ratio of leukemic transformation or death. (E) Frequency distribution of VAF of SF3B1 and RUNX1 (top), and SF3B1 and STAG2 (bottom) mutations in SF3B1-mutant MDS/AML. HR, high-risk group; LFS, leukemia free survival; LR, low-risk group; MDS-del5q, MDS with isolated del(5q); MDS-EB, MDS with excess blasts; MDS-RS, MDS with ring sideroblasts; MDS-SLD/MLD, MDS with single lineage/multilineage dysplasia; MDS-U, MDS, unclassifiable; WHO, World Health Organization.

Figure 2.

High-risk comutations induce divergent transcriptional changes. (A) Outline of the experimental approach for CRISPR/Cas9 editing of 5F-HPCs. (B) Proportion of frameshift mutations in AAVS1 (A), RUNX1 (R), and STAG2 (S) in SF3B1-mutant (S-A, S-R, and S-S) and WT (A, R, and S) patient-derived isogenic 5F-HPCs. Genotyping was performed by Sanger sequencing and ICE analysis. Data are presented as mean ± standard deviation (SD) from 3 independent experiments. (C) Normalized enrichment score of Hallmark gene sets (left) and human hematopoietic cell gene signatures from Laurenti et al and Hay et al (right) in gene set enrichment analysis (GSEA) analysis of S-R vs S-A and S-S vs S-A 5F-HPCs. FDR q < 0.05. (D) Relative expression of genes with divergent dysregulation in S-R and S-S 5F-HPCs. Values are shown as log₂FC relative to S-A 5F-HPCs; P < .05. (E) STRING protein–protein interaction network of genes with divergent dysregulation in S-R and S-S 5F-HPCs. Disconnected nodes were removed; line thickness proportional to interaction score of >0.40. (F) PU.1 protein level in CD34⁺ 5F-HPCs measured by intracellular flow cytometry. Representative flow plot (left). FC of MFI in S-R or S-S relative to S-A (right). Data are presented as mean ± SD from 4 independent experiments; 1-sample t test. (G) GSEA analysis of TRRUST PU.1 target genes in S-R vs S-A and S-S vs S-A 5F-HPCs. FDR q < 0.05. (H) Proportion of mis-spliced isoforms by category in SF3B1-mutant K562 cells edited for RUNX1 (S-R), STAG2 (S-S), or AAVS1 control (S-A). Mis-spliced events were categorized as tandem 3′ untranslated regions (tutr), cassette or skipped exons (se), retained introns (ri), mutually exclusive exons (mxe), alternative usage of normally constitutively spliced junctions (cj), alternative retention of normally constitutively spliced introns (ci), alternative 5′ss (a5ss), or alternative 3′ss (a3′ss). Events were restricted to ≥10% mis-splicing and Bayes factor of ≥5. (I) Spearman correlation matrix of the level of mis-splicing between S-A, S-R, and S-S. ∗∗∗∗P < .0001. (J) Proportion of a3′ss mis-spliced events shared between low-risk (S-A) and high-risk (S-R or S-S) genotypes (red), shared by high-risk (S-R and S-S) but not S-A (yellow), and unique to high-risk genotypes (blue). FC, fold change; FDR, false discovery rate; MFI, mean fluorescence intensity.

Figure 3.

High-risk comutations induce opposing myeloid lineage outcomes. (A) Differentiation efficiency of SF3B1-mutant 5F-HPCs CRISPR-edited for RUNX1 (S-R), STAG2 (S-S), or AAVS1 (S-A). Representative flow plot (left); S-R compared with S-A (center); S-S compared with S-A (right); same control group for both comparisons. Percent erythroid, myeloid, and Lin^– precursors of mCherry⁺DAPI^– lentivirus–transduced edited cells. Data are presented as mean ± SD from 5 independent experiments, each with S-A, S-R, and S-S groups; ratio paired t test. (B) Outline of the experimental approach used to introduce SF3B1 K700E knockin mutation with high-risk RUNX1 (S-R) or STAG2 (S-S) comutations, or AAVS1 control (S-A) into CB- or PB-derived CD34⁺ HSPCs. (C) Differentiation efficiency of CB CD34⁺ HSPCs gene edited for SF3B1 K700E and RUNX1 (S-R), STAG2 (S-S), or AAVS1 (S-A). Representative flow plot (left); S-R compared with S-A (center); S-S compared with S-A (right); the same control group was used for both comparisons. Percent erythroid, myeloid, and Lin^– precursors of total BFP⁺ cells is shown. Data are presented as mean ± SD from 3 independent experiments, each with S-A, S-R, and S-S groups; ratio paired t test. (D) Granulocytic maturation in BM samples of SF3B1-mutant patients. Representative flow plot with gating strategy (left) and quantification of immature CD13⁺CD16^– granulocytes (center) and mature CD16⁺ cells (right). Data are shown as mean with interval; 1-way analysis of variance (ANOVA). BFP, blue fluorescent protein; DAPI, 4′,6-diamidino-2-phenylindole; Lin^–, lineage-negative; ns, not significant.

Figure 4.

High-risk comutations expand distinct SF3B1-mutant HSPC compartments. (A) Representative flow plots and gating strategy used to measure the frequency of CD34⁺CD38^– and CD34⁺CD38^–CD133^+/− HSPCs during in vitro culture. (B-C) Frequency of SF3B1-mutant double-edited RUNX1 mutant (S-R) or STAG2 mutant (S-S) vs WT (S-A) CD34⁺CD38^–CD133⁺ (B) or CD34⁺CD38^–CD133^– HSPCs (C) for 14 days of in vitro culture. Data are presented as mean ± SD from 5 independent CB donor experiments (n = 3 with S-A, S-R, and S-S groups; n = 2 with S-A and S-R groups); 2-way ANOVA. (D) Representative flow plots and gating strategy to measure the frequency of HSCs, MPPs, and LMPPs during in vitro culture. (E) FC in the frequency of S-R or S-S phenotypic HSCs, MPPs, and LMPP relative to S-A at day 7 of in vitro culture. (F) Frequency of S-R or S-S vs S-A HSCs, MPPs, and LMPPs for 7 days of in vitro culture. Data are presented as mean ± SD from 2 independent CB donor experiments; 2-way ANOVA. (G) Frequency of SF3B1-mutant double-edited TET2 mutant (S-T) vs WT (S-A) CD34⁺CD38^–CD133⁺ (left) or CD34⁺CD38^–CD133^– (right) HSPCs for 14 days of in vitro culture. Data are presented as mean ± SD from 5 independent CB donor experiments; 2-way ANOVA. (H) Frequency of SF3B1-mutant triple-edited TET2 (S-T-A), RUNX1 (S-R-A), or TET2 + RUNX1 mutant (S-R-T) vs WT (S-A) CD34⁺CD133⁺ phenotypic HSCs during in vitro culture. Data are presented as mean ± SD from 2 independent CB donor experiments; 2-way ANOVA. ns, not significant.