Erythroid Differentiation Enhances RNA Mis-Splicing in SF3B1-Mutant Myelodysplastic Syndromes with Ring Sideroblasts

Abstract

Myelodysplastic syndromes with ring sideroblasts (MDS-RS) commonly develop from hematopoietic stem cells (HSC) bearing mutations in the splicing factor SF3B1 (SF3B1mt). Direct studies into MDS-RS pathobiology have been limited by a lack of model systems that fully recapitulate erythroid biology and RS development and the inability to isolate viable human RS. Here, we combined successful direct RS isolation from patient samples, high-throughput multiomics analysis of cells encompassing the SF3B1mt stem-erythroid continuum, and functional assays to investigate the impact of SF3B1mt on erythropoiesis and RS accumulation. The isolated RS differentiated, egressed into the blood, escaped traditional nonsense-mediated decay (NMD) mechanisms, and leveraged stress-survival pathways that hinder wild-type hematopoiesis through pathogenic GDF15 overexpression. Importantly, RS constituted a contaminant of magnetically enriched CD34+ cells, skewing bulk transcriptomic data. Mis-splicing in SF3B1mt cells was intensified by erythroid differentiation through accelerated RNA splicing and decreased NMD activity, and SF3B1mt led to truncations in several MDS-implicated genes. Finally, RNA mis-splicing induced an uncoupling of RNA and protein expression, leading to critical abnormalities in proapoptotic p53 pathway genes. Overall, this characterization of erythropoiesis in SF3B1mt RS provides a resource for studying MDS-RS and uncovers insights into the unexpectedly active biology of the "dead-end" RS.

Significance: Ring sideroblast isolation combined with state-of-the-art multiomics identifies survival mechanisms underlying SF3B1-mutant erythropoiesis and establishes an active role for erythroid differentiation and ring sideroblasts themselves in SF3B1-mutant myelodysplastic syndrome pathogenesis.

Graphical abstract

Figure 1.

Erythroid differentiation and enucleation remain active in SF3B1^mt MDS-RS erythroblasts. A, Flow cytometry strategy for staging of erythroblast populations from BM MNC of patients with MDS-RS. Gating steps identify live and terminally differentiating erythroid cells (Lin⁻7AAD⁻GPA⁺), from which erythroblasts (EB) are staged according to Band 3 and integrin α4 expression (26). Integrin α4–negative cells are excluded from quantification to avoid skewing by anucleate cells. B, Mean (±SEM) cell population frequencies within flow cytometry parent populations (singlets > terminally differentiating erythroid > erythroblast subsets), quantified in NBM donors (n^NBM = 4) and patients with MDS-RS (n^MDS-RS = 6). Erythroid cells are quantified within the singlet population. Erythroblast subsets are quantified within the GPA⁺ population. C, Mean (±SEM) RS frequencies per sorted erythroblast subset and compared with frequencies in matched diagnostic BM smears (n^MDS-RS = 6). D, Gating and quantification [mean (±SEM)] of DNA content (Draq-5) and intracellular Ki-67 abundance in GPA⁺ magnetically sorted cells (n^NBM = 3; n^MDS-RS = 6). E, Mean (±SEM) RS frequencies per sorted Ki67-expressing subset and compared with frequencies in matched diagnostic BM smears (n^MDS-RS = 6). F, Mean (±SEM) CD71 (transferrin receptor, TFRC) median fluorescence indices per erythroblast subset (n^NBM = 4; n^MDS-RS = 6). G, Mean (±SEM) enucleation frequencies after 28-day 3D culture of MDS-RS BM MNCs (n^MDS-RS = 4) and separated by iron granule visibility upon morphological analysis. Statistical comparison was performed by paired t test analysis. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, nonstatistically significant.

Figure 2.

Reagent-free MACS enables direct characterization of viable SF3B1^mt RS. A, Method for RS and siderocyte purification from BM aspiration material. A representative flow cytometric diagram plots RNA/DNA content (Thiazole Orange) against DNA content (Hoechst 33342) in Lin⁻GPA⁺ singlets after MACS of HD cells. Representative micrographs are shown at the right (blue, iron granules; brown, hemoglobin; pink, DNA). Scale bars, 10 μm. B, Mean (±SEM) RS frequencies before and after MACS alone in three MNC and four HD samples, and further purification with FACS (M+FACS) of the same HD samples. HD RS quantification before enrichment steps identifies only 0.1% to 0.001% as potential RS due to high RBC proportions. C, Isolated RS numbers in MACS-enriched cells from 5 × 10⁶ MNCs (n = 3) or M+FACS-enriched HD cells (n = 26, 19 unique biological replicates + 7 repeat visits). D, Correlation of log₁₀-converted isolated RS numbers and RS frequencies in matched BM aspirates (n = 17). E, Mean (± SEM) SF3B1 mutation (SF3B1^mt) VAF in unfractionated MNCs (baseline) and MACS-enriched or M+FACS-enriched HD cells, as determined by ddPCR (n = 3 per enrichment method, 5 patients in total). The dashed line indicates complete heterozygosity (VAF = 50%). F, Mean (±SEM) CD71 staining indices (MFI of the cell population – MFI of the CD71-negative RBC population divided by 2 × SD of the RBC population; n^NBM = 15, n^MDS-RS = 8, n^RS = 14). G, Immunofluorescence of Ki-67 detection in NBM erythroblast and an MNC-derived RS isolate, co-labeled for DNA (DAPI; cyan), Ki-67 (yellow), and mitochondria (MitoTracker; magenta). Individual grayscale channels and a composite image of all three markers are shown. A Ki-67^neg RS is shown with an outlined arrow; a Ki-67^hi RS with a filled arrow. Scale bars, 20 μm. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, nonstatistically significant.

Figure 3.

Peripherally circulating RS are common and clinically relevant in MDS-RS. A, Isolation steps from the PB HD fraction of patients with MDS-RS through reagent-free magnetic separation and representative flow cytometry diagram, where RS are identifiable and validated as present through morphological analysis. B, Correlation of RS abundances isolated from matched BM and PB samples (leftmost subpanel, n = 16). C–E, Correlation of log₁₀-converted isolated RS numbers obtained from anemic (Hb < 12.0 g/dL) patients with MDS-RS with BM RS percentages from clinical BM smears (n^PB = 15, n^BM = 17; C), serum erythropoietin levels (untreated patients only, n^PB = 8, n^BM = 8; D), and hemoglobin levels (n^PB = 18, n^BM = 19; E). F, Flow cytometry example of BM and PB RS with increased DNA content comparing two visits of the same patient to the clinic, before and after ESA treatment. A cell population of increased DNA content is highlighted with dark red arrows. G, Mean (SD) frequency of RS with elevated DNA content, separated by EPO treatment status and cell fraction of origin. Arrows indicate binucleate RS identified during morphological analysis of EPO-treated and RS-enriched samples. Scale bars, 10 μm. H, Morphologic visualization of binucleated RS in ESA-treated, RS-enriched samples. Scale bars, 10 μm. ***, P < 0.001.

Figure 4.

SF3B1 mutations have limited impact on the gene expression of true MDS-RS HSPC. A, Principal component analysis (PCA) plots of a full-length bulk RNA-seq experiment encompassing sorted cell populations from NBM donors (n^CD34 = 7, n^GPA = 4, n^Ret = 4) and SF3B1^mt MDS-RS patients (n^CD34 = 6, n^GPA = 5, n^RS = 4, n^Sid = 4). Sample distribution along PC1 is visualized against PC2/PC3. B, Global overview of two integrated 10x Genomics scRNA-seq experiments encompassing sorted cell populations from NBM donors and SF3B1^mt patients with MDS-RS (n^NBM = 3, n^MDS-RS = 5, n^RS = 1). A UMAP-based bidimensional projection is displayed and separated per sample group, where each cell is visualized as one point. The dotted circle indicates the RS-enriched cell subset, which is absent in NBM. Cell types were annotated according to gene expression signatures per cluster set. Sample and cell type composition in the total dataset are shown below the UMAP plots. C, Volcano plot (left) displaying differentially expressed genes (DEG) in CD34⁺ MACS-enriched BM cells comparing SF3B1^mt MDS-RS versus NBM. Cut-offs for significance were log₂ FC > 0.5, adjusted P value <0.01. Genes were overexpressed (OE; red), underexpressed (UE; blue), or not significantly different (NS; gray). Gene set enrichment analysis (GSEA) of overexpressed genes (middle) was performed with the Enrichr Human Gene Atlas. The right UMAP heatmap displays expression of bulk OE genes in scRNA-seq. The dotted rectangle highlights the HSPC transcriptomic cluster. D, AUCell erythroid score [based on erythroid markers from An et al. (30)] mapped in the UMAP overlays and separated by mutational background. The erythroid score is similarly displayed in violin plots (gray, NBM; orange, SF3B1^mt) and grouped per cell population (excluding cell subsets unrelated to erythroid development, e.g., macrophages). E, Representative CD34 and GPA FACS plots from CD34⁺ MACS-separated BM MNCs isolated from an patient with SF3B1^mt MDS-RS and from a patient with non-SF3B1^mt non-RS MDS. Lin⁻CD34⁻GPA⁺ cells are gated in blue and connected to representative micrographs. Scale bars, 10 μm. F, Mean (±SEM) percentage of Lin⁻CD34⁻GPA⁺ cells in CD34⁺-enriched cells (n = 3 per group). G, Mean (±SEM) cell frequencies based on morphologic analysis of Lin⁻CD34⁻GPA⁺ in MACS-purified CD34⁺ MDS-RS samples. H, UMAP projection of CD34 RNA-positive HSPCs in the scRNA-seq dataset. I, Mean (±SEM) frequencies of transcriptomically identifiable HSPC subsets as set out in H and compared between NBM and MDS-RS samples. J, Gene set enrichment analysis results for Gene Ontology biological process (GO BP) enrichment of differentially expressed genes identified in the HSPC cluster between MDS-RS and NBM cells (n^NBM = 432, mean 144 cells/donor; n^MDS-RS = 510, mean 102 cells/donor). ***, P < 0.001; ns, nonstatistically significant.

Figure 5.

SF3B1^mt RS activate a prosurvival transcriptomic program against oxidative and RNA splicing stress. A, Volcano plot displaying differentially expressed genes in bulk data comparing M+FACS-purified SF3B1^mt RS against MACS-purified NBM GPA⁺ erythroblasts with an absolute log₂ FC cut-off of 2 and an adjusted P value cut-off of 10⁻⁴. B, UMAP overlays of RS overexpressed transcript percentages per cell (top row) and AUCell scores of RS identity (bottom row), separated by sample type. Transcript percentages are mapped with an initial baseline cut-off of 4% of total transcripts. AUCell scores were based on the RS overexpressed gene set. C, Gene Ontology biological process (GO BP) enrichment analysis of differentially expressed genes identified in the RS-enriched cluster comparing HD fraction-derived RS versus MNC-derived RS (nonspecifically present among MDS-RS MNCs and with presumably decreased iron load). D, Heatmap of all differentially expressed genes between RS from SF3B1^mt MDS-RS patients versus NBM, subclustered by cell subset. The top bar above the heatmap identifies each sample type, and cells are further clustered according to cell type, identified by the bottom bar. Dashed lines highlight cell type separation. E, Metascape gene ontology term network generated from all differentially expressed genes identified through comparison of RS and NBM samples at each transcriptomically identified differentiation stage cluster (HSPC, ProEB + Early EB, Differentiating EB, and LateEB + RS-enriched), correcting for differentiation stage skewing. Gene ontology subterms (small circles) are organized and clustered by major functional terms (numbered black circles). Clusters are annotated in the table (right), including adjusted P values from Metascape analysis. F,ABCB7 RNA expression in bulk RNA-seq displayed in log normalized counts from all assayed cell populations (left) and overlaid in the 10x UMAP projection (right). G, Sashimi plots for canonical (normal font) and mis-spliced (bold) read counts of the ABCB7 alternative 3′ splice site associated with targeting by NMD. H,GDF15 expression based on RNA sequencing of purified populations (quantified in log normalized counts; left) and single cells (UMAP overlay; right). I, Mean (±SEM) GDF15 concentration in culture supernatants obtained from 28-day erythroid culture of BM MNCs (n^NBM = 3, n^MDS-RS = 3), as determined by ELISA. One empty scaffold was kept in the same media and culture conditions to evaluate GDF15 levels in base media. J, Mean (±SD) erythroid and myeloid colony formation from MACS-enriched CD34 cells (n^NBM = 3, n^MDS-RS = 5), normalized to untreated numbers. Minimum total colonies counted were 254 among NBM donor conditions and 124 among MDS-RS donor conditions. Cells were treated with either recombinant GDF15 peptide at a concentration of 100 ng/mL (gray squares) or with an equal volume of water (vehicle, black circles). *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, nonstatistically significant.

Figure 6.

Distinct RNA dynamics in erythroid differentiation intensify SF3B1^mt mis-splicing. A, Proportional Venn diagram of genes undergoing statistically significant AS [FDR < 0.001, absolute difference in percentage spliced in levels (Abs. ΔPSI) > 0.2] in the HSPC (green; MDS-RS CD34⁺ vs. NBM CD34⁺), nucleated erythroid [red; MDS-RS RS vs. NBM EB, Erythroid (N)] and anucleate erythroid [blue; MDS-RS Siderocytes vs. NBM Ret^PB, Erythroid (A)] populations. B, Gene Ontology biological process (GO BP) enrichment analysis results comprising genes mis-spliced in both the HSPC and erythroid (N) populations are shown in the top bar chart. The Human Gene Atlas enrichment results with the lowest adjusted P value are shown in the bottom table. C, Frequency of AS events split by rMATS category in each sample group comparison. SE, skipped exon; RI, retained intron; MXE, mutual exon exclusion; A5SS, alternative 5′ splice site; A3SS, alternative 3′ splice site. Statistical comparisons of A3SS+RI frequencies were performed with Fisher exact test. D, Box plots of percent spliced-in (PSI) values of literature-validated SF3B1^mt-induced A3SS events in MDS-RS samples, separated by sample type (CD34, GPA, RS). Known targeting by NMD is indicated with a red circle; in-frame events without a PTC are indicated by a blue circle. E, Box plots of PSI values in newly identified ASEs affecting known driver genes implicated in the pathogenesis of MDS and CSA. Known targeting by NMD is indicated with a red circle; PTC detection with unverified NMD is indicated with an orange circle; in-frame events without a PTC are indicated by a blue circle. F, Distribution of base pair distances from cryptic A3SS sites to canonical splice sites (horizontal axis) in HSPC and Erythroid (N). Further detail is provided in −400 bp to 0 bp for increased contrast. Lines at −30 bp and −10 bp demarcate the interval associated with SF3B1 mis-splicing (38). Additional lines at −140 bp and −330 bp demarcate additional intervals of interest. G, Sequence logos of canonical and A3SS sequences encompassing the 3′ splice site (starting at −35 bp upstream of the AG motif) and statistical comparison through a two-sample logo. H, Frequency of A3SS events per rMATS cell type comparison where the splice site shift remains in-frame (blue) or induces a frameshift event (orange). I, Frequency of exon insertion events per rMATS cell-type comparison where the splice site shift incorporates a new PTC (pink) or remains in-frame with no PTC induction (green). J and K, RNA velocity analysis of transcriptomically identified HSPC and erythroblast subsets in 10x scRNA-seq, visualizing the percentage of spliced transcripts along pseudotime in the total cell populations (violin plots) or separated by sample group (scatter plots). The analysis in J includes all transcripts, whereas K excludes ribosomal and globin transcripts. L, UMAP overlay of FACS-purified HSPC subsets and GPA+ EB from 1 SF3B1^mt MDS-RS patient after Smart-seq3xpress (SS3x), visualizing true versus predicted cell-type identity. M, RNA velocity analysis of spliced RNA read percentages in the FACS-sorted SS3× experiment, analyzed independently of the 10x dataset. This graph excludes ribosomal and globin transcripts. N, Mean (±SEM) differences in PSI after 3 hours cycloheximide treatment (70 μg/mL) versus DMSO (1:1,000, vehicle) in MDS-RS CD34 and GPA cells. SF3B1^mt-associated NMD-targeted ASEs with sufficient coverage are shown at far left, SF3B1^mt-associated in-frame ASEs at middle-left, and endogenous NMD-targeted transcripts at middle-right. The far-right plot visualizes all ASEs. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, nonstatistically significant.

Figure 7.

Proteomic analysis of SF3B1^mt RS defines RNA/protein uncoupling downstream of RNA mis-splicing with severe dysregulation of proapoptotic genes. A, Design of a combined transcriptomic and proteomic analysis of SF3B1^mt RS. Erythroblast (EB) samples from five NBM donors (separated into three biologically distinct batches) and paired EB + RS samples from three patients with MDS-RS were subjected to semiquantitative proteomics. RNA-seq DE genes are compared against differentially expressed proteins to obtain four major signatures of differential expression, which are highlighted in each quadrant and correspond to the gene ontology enrichment results provided in Supplementary Table S4. B, Scatter plot of SF3B1^mt mis-spliced genes with uncoupled RNA and protein expression in RS as compared with NBM EB. RNA and protein expression cut-offs were applied to limit analysis only to genes with RNA Log₂FC > 0.5, Protein Log₂FC < 0.5, and a difference between RNA and protein Log₂FC values > 1 (RNA Log₂FC – Protein Log₂FC). AS detection cutoffs were followed as indicated in Fig. 6A (FDR < 0.001; Abs. ΔPSI > 0.2). Pink, known SF3B1^mt mis-spliced genes (Fig. 6D). Blue, MDS/CSA-implicated driver genes (Fig. 6E). TP53 pathway/proapoptotic genes were detected through enrichment analysis and are highlighted in orange. C, Mean (±SEM) protein expression levels of ABCB7 and MAP3K7, normalized to mean NBM expression. D, 10x single-cell RNA-seq cross-validation of increased RNA expression for TP53 pathway genes MDM2, BAX, and FAS, with gene expression values overlaid in the HSPC/erythroid UMAP projection and separated by sample type. Below each gene, sashimi plots display major mis-spliced transcript regions. Black, canonical splice junction counts (SJ); red, cryptic SJ counts. A full legend for the sashimi plots is provided below the graph. The asterisks indicate sites corresponding to transcripts canonically targeted by NMD. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ns, nonstatistically significant.