Single-cell multi-omics defines the cell-type-specific impact of splicing aberrations in human hematopoietic clonal outgrowths

Abstract

RNA splicing factors are recurrently mutated in clonal blood disorders, but the impact of dysregulated splicing in hematopoiesis remains unclear. To overcome technical limitations, we integrated genotyping of transcriptomes (GoT) with long-read single-cell transcriptomics and proteogenomics for single-cell profiling of transcriptomes, surface proteins, somatic mutations, and RNA splicing (GoT-Splice). We applied GoT-Splice to hematopoietic progenitors from myelodysplastic syndrome (MDS) patients with mutations in the core splicing factor SF3B1. SF3B1^mut cells were enriched in the megakaryocytic-erythroid lineage, with expansion of SF3B1^mut erythroid progenitor cells. We uncovered distinct cryptic 3' splice site usage in different progenitor populations and stage-specific aberrant splicing during erythroid differentiation. Profiling SF3B1-mutated clonal hematopoiesis samples revealed that erythroid bias and cell-type-specific cryptic 3' splice site usage in SF3B1^mut cells precede overt MDS. Collectively, GoT-Splice defines the cell-type-specific impact of somatic mutations on RNA splicing, from early clonal outgrowths to overt neoplasia, directly in human samples.

Keywords: BAX; RNA-seq; SF3B1; clonal hematopoiesis; genotyping; long-read sequencing; multi-omics; myelodysplastic syndrome; single cell; splicing.

Figure 1.. Enrichment of SF3B1^mut cells in the megakaryocytic-erythroid lineage.

(A) GoT-Splice workflow combines GoT with CITE-seq and long-read full-length cDNA using ONT for the simultaneous single-cell profiling of protein and gene expression, somatic mutations, and alternative splicing. (B) Patient metadata and quality controlled GoT data for SF3B1-mutant MDS and CH samples. (C) Uniform manifold approximation and projection (UMAP) of CD34+ cells (n = 15,436 cells) from MDS patients with SF3B1 K700E mutations (n = 3 individuals), overlaid with cluster cell-type assignments. HSPC, hematopoietic stem progenitor cells; IMP, immature myeloid progenitors; MkP, megakaryocytic progenitors; MEP, megakaryocytic-erythroid progenitors; EP, erythroid progenitors; NP, neutrophil progenitors; E/B/M, eosinophil/basophil/mast progenitor cells; T/B cell progenitors; Mono, monocyte; DC, dendritic cells; Pre-B, precursors B cells; Mono DC, monocyte/dendritic cell progenitors. (D) Density plot of SF3B1^mut vs. SF3B1^wt cells, with genotypes (MDS01-03) for 12,494 cells (80.9 % of all cells). (E) Normalized frequency of SF3B1^mut cells in progenitor subsets with at least 300 genotyped cells. Bars show analysis of MDS01-03 with mean +/− s.e.m. of 100 downsampling iterations to 1 genotyping UMI per cell. Cell types with >300 cells were analyzed. P-value from likelihood ratio test of linear mixed model with or without mutation status. (F) Differential ADT marker expression between SF3B1^mut and SF3B1^wt cells. Red: higher expression in SF3B1^mut cells; blue: higher expression in SF3B1^wt cells. Dot size corresponds to the average ADT expression across cells in each cell-type. P-values determined through permutation testing. (G) Mutant cell fraction and ADT expression levels of CD36 and CD71 as a function of pseudotime along the megakaryocyte-erythroid differentiation trajectory for SF3B1^mut and SF3B1^wt cells in MDS01-03. Shading denotes 95% confidence interval. Histogram shows cell density of analyzed clusters, ordered by pseudotime. P-values were calculated by Wilcoxon rank sum test by comparing mutant cell fraction between pseudotime trajectory quartiles. (H) Differential gene expression between SF3B1^mut and SF3B1^wt EP cells in MDS samples. Genes with an absolute log₂(fold change) > 0.1 and P-value < 0.05 were defined as differentially expressed (DE). Cell cycle (red) and translation (blue) pathways (Reactome) are highlighted. (I) Expression (mean +/− s.e.m.) of TP53 pathway related genes (Reactome) between SF3B1^mut and SF3B1^wt cells in progenitor cells from MDS01-03 samples. Red: module score in SF3B1^mut cells; blue: module score in SF3B1^wt cells. P-values from likelihood ratio test of linear mixed model with or without mutation status. (J) Same as (I) for expression of cell cycle related genes (Reactome) between SF3B1^mut and SF3B1^wt cells in progenitor cells from MDS01-03 samples.

Figure 2.. Simultaneous profiling of gene expression, cell surface protein markers, somatic mutation status, and alternative splicing at single-cell resolution.

(A) Comparison of the percentage of ONT reads with either incorrect structure (double TSO, no adaptors, single R1 or single TSO) or correct structure (full-length reads) both before and after the inclusion of a biotin enrichment protocol step during preparation for sequencing. Bars show the aggregate analysis of n = 5 samples with mean +/− s.d. of the percentage for each category. (B) Scatter plot of the correlation between the number of UMIs/cell detected in long-read ONT vs. short-read Illumina data for cells sequenced across both platforms for sample MDS05. (C) Density plot of the correlation between the number of UMIs/gene detected in long-read ONT vs. short-read Illumina data for sample MDS05. (D) Number of splice junctions captured in the full-length long-read ONT data compared to short-read sequencing data (gene coverage >= 10 in both sequencing protocols, junction cluster coverage >= 600 and junction read support >= 1 read [see Methods]), demonstrating increased junction capture with GoT-Splice across cells. (E) Greater sequencing coverage uniformity of GoT-Splice compared to short-read sequencing over splice junctions, illustrated with the ERGIC3 gene. (F) Pie chart summarizing the distribution of different alternative splicing events detected after junction annotation. Inset: Differences in the usage of cryptic 3’ and 5’ splice site events between SF3B1^mut and SF3B1^wt cells measured with a dPSI (SF3B1^mut PSI - SF3B1^wt PSI). Associated with SF3B1^mut: +ve dPSI; associated with SF3B1^wt: -ve dPSI. (G) Comparison of dPSI values for shared cryptic 3’ splicing events identified in the MUT vs. WT cell comparison from GoT-Splice of SF3B1^mut MDS01-03 samples and in the SF3B1^mut vs. SF3B1^wt bulk comparison from bulk RNA-sequencing of CD34+ cells of MDS samples in Pellagatti et al.. Correlation coefficient ρ calculated using Spearman’s correlation and P-value derived from Student’s t-distribution.

Figure 3.. Progenitor cell-type specific mis-splicing in SF3B1^mut MDS.

(A) Differential splicing analysis between SF3B1^mut and SF3B1^wt cells across MDS samples. Junctions with absolute dPSI > 2 and BH-FDR adjusted P-value < 0.2 were defined as differentially spliced. Top: Bars showing percentage of genes differentially spliced in SF3B1^mut and SF3B1^wt cells in MDS and MDS validation cohorts. Inset: Expected peak in the number of identified cryptic 3’ splice sites at 15–20 base pairs upstream of the canonical 3’ splice site in SF3B1^mut cells. (B) Sashimi Plot of METTL17 intron junction with an SF3B1^mut associated cryptic 3’ splice site showing RNA-seq coverage in SF3B1^mut vs. SF3B1^wt cells within MDS samples. Inset: Expected increase in PSI value for the usage of this cryptic 3’ splice site in SF3B1^mut cells. (C) dPSI values between SF3B1^mut and SF3B1^wt cells for cryptic 3’ splicing events identified in main progenitor subsets across MDS samples. Columns- cryptic 3’ junctions differentially spliced in at least one cell-type, with P-value <= 0.05 and dPSI >= 2. Rows- cell-type. Genes highlighted for cell cycle (purple), heme metabolism (green), oxygen homeostasis (black), RNA processing (red) and erythroid differentiation (yellow) pathways. Left bar plots show the fraction of differentially spliced cryptic 3’ splice sites per cell. Top bar plots quantify the total number of cell types where an event is differentially spliced, with the cell-type specific events on the right.

Figure 4.. SF3B1^mut-associated mis-splicing changes along the continuum of erythropoiesis.

(A) Percent spliced-in (PSI) of junctions in SF3B1^mut cells along the hematopoietic differentiation trajectory (HSPCs, IMPs, MEPs, EPs). Rows (z-score normalized)- cryptic 3’ splice sites; columns- PSI for usage of a given cryptic 3’ splice site in each window (size of 3000 SF3B1^mut cells, sliding by 300 SF3B1^mut cells). Only junctions differentially spliced in at least one cell-type with a dPSI > 2 were analyzed. ADT expression of CD71 and cell type fractions are shown. Rows ordered according to PSI peak. Genes highlighted for cell cycle (purple), heme metabolism (green), oxygen homeostasis (black), RNA processing (red), erythroid differentiation (yellow) and apoptosis (blue) pathways. (B) Examples of mis-spliced genes at different erythroid maturation stages. Bars represent PSI in SF3B1^mut cells. Red lines represent junction ONT expression in SF3B1^mut cells. (C) Fold change (log₂) of gene expression between SF3B1^mut and SF3B1^wt EP cells in NMD-inducing vs. NMD-neutral genes. (D) BAX gene model and relevant isoforms. Characteristic domains are highlighted in main isoform BAX-ɑ. The cryptic 3’ splicing event on the terminal exon defines the BAX-ω isoform, characterized by frameshift disruption of the transmembrane domain (TM). (E) Western blot of TF-1 BAX/BAK knockout cells (DKO) with doxycycline-inducible expression of control GFP or FLAG-tagged BAX isoforms α, β, and ω after 24 hours. (F) Fold change in annexin V positive TF-1 DKO cells expressing different BAX isoforms under cytokine depleted conditions + doxycycline (1ug/mL) at 48 and 72 hours normalized to apoptotic cells - doxycycline treatment (black line). N=2 independent experiments performed in triplicate. (G) Representative annexin V/DAPI flow cytometry plots of different BAX isoforms after 72 hours under cytokine depleted conditions + doxycycline. Percent frequencies noted in relevant quadrants. Bars represent mean values. Error bars represent ±SD; ** P < 0.01 *** P < 0.001.

Figure 5.. SF3B1 mutations are enriched along the erythroid lineage in clonal hematopoiesis.

(A) UMAP of CD34+ (n = 9,007) cells from clonal hematopoiesis (CH) samples with SF3B1 K700E or SF3B1 K666N mutation (n = 2 individuals), overlaid with cluster cell-type assignments. See Figure 1A for cell-type descriptions. (B) Density plot of SF3B1^mut vs. SF3B1^wt cells. (C) UMAP of CD34+ cells from CH samples overlaid with pseudotemporal ordering. Inset: Pseudotime in SF3B1^mut vs. SF3B1^wt cells in the aggregate of CH01-02. P-value for comparison of means from Wilcoxon rank sum test. (D) Normalized ratio of mutated cells along pseudotime quartiles. Bars show aggregate analysis of samples CH01-CH02 with mean +/− s.e.m. of 100 downsampling iterations to 1 genotyping UMI per cell. Only cell types with >300 cells were analyzed. P-value from likelihood ratio test of linear mixed model with or without mutation status. Bottom: Fraction of cell types within each pseudotime quartile. (E) Differential gene expression between SF3B1^mut and SF3B1^wt HSPC cells in CH samples. Genes with an absolute log₂(fold change) > 0.1 and P-value < 0.05 were defined as differentially expressed (DE). DE genes in the translation pathway (red, Reactome) are highlighted (see Table S5). (F) Gene Set Enrichment Analysis of DE genes in SF3B1^mut HSPC cells across CH samples. Gene sets that overlap with SF3B1^mut EP cells in MDS highlighted (red). (G) Expression (mean +/− s.e.m.) of translation-related genes (Reactome) between SF3B1^mut and SF3B1^wt cells in progenitor cells from CH01-02 samples. P-values from likelihood ratio test of linear mixed model with or without mutation status.

Figure 6.. SF3B1^mut clonal hematopoiesis progenitor cells display cell-type specific cryptic 3’ splice site usage.

(A) Differential splicing analysis between SF3B1^mut and SF3B1^wt cells across CH samples. Junctions with an absolute delta percent spliced-in (dPSI) > 2 and BH-FDR adjusted P-value < 0.2 were defined as differentially spliced. (B) Sashimi Plot of ERGIC intron junction with an SF3B1^mut associated cryptic 3’ splice site showing RNA-seq coverage in SF3B1^mut vs. SF3B1^wt cells within CH samples, as well as compared to the CH samples when treated as bulk (pseudobulk of all cells regardless of genotype). PSI values showing the expected increase in usage of this cryptic 3’ splice site in SF3B1^mut cells alone when compared to both SF3B1^wt cells as well as all cells (pseudobulk of sample). (C) Venn Diagram of overlapping genes with cryptic junctions significantly differentially spliced in at least one erythroid lineage cell type (HSPCs, IMPs, MEPs, EPs) with a dPSI > 2 between MDS01-03 and CH samples. P-value for the overlap from Fisher’s Exact test. (D) Percent spliced-in (PSI) of junctions in SF3B1^mut cells along the hematopoietic differentiation trajectory of erythroid lineage cells. Rows (z-score normalized)- cryptic 3’ splice sites; columns- PSI for usage of a given cryptic 3’ splice site in each window (size of 600 SF3B1^mut cells, sliding by 60 SF3B1^mut cells). Only junctions differentially spliced in at least one cell type with a dPSI > 2 were analyzed. Pseudotime across each window shown. Rows are ordered according to the peak in PSI. Cryptic events also differentially spiced in MDS highlighted (red). (E) Bar plots of PSI values for usage of the BAX-ω isoform across each window of SF3B1^mut cells in the MDS, MDS validation and CH cohorts along the hematopoietic differentiation trajectory of erythroid lineage cells. Fraction of cell types in each window shown per cohort (MDS: SF3B1^mut cells (n = 6376) ordered by CD71 expression, MDS validation: SF3B1^mut cells (n = 987) ordered by pseudotime, CH: MUT cells (n = 1021) ordered by pseudotime).