FLI1 is associated with regulation of DNA methylation and megakaryocytic differentiation in FPDMM caused by a RUNX1 transactivation domain mutation

Familial platelet disorder with associated myeloid malignancies (FPDMM) is an autosomal dominant disease caused by heterozygous germline mutations in RUNX1. It is characterized by thrombocytopenia, platelet dysfunction, and a predisposition to hematological malignancies. Although FPDMM is a precursor for diseases involving abnormal DNA methylation, the DNA methylation status in FPDMM remains unknown, largely due to a lack of animal models and challenges in obtaining patient-derived samples. Here, using genome editing techniques, we established two lines of human induced pluripotent stem cells (iPSCs) with different FPDMM-mimicking heterozygous RUNX1 mutations. These iPSCs showed defective differentiation of hematopoietic progenitor cells (HPCs) and megakaryocytes (Mks), consistent with FPDMM. The FPDMM-mimicking HPCs showed DNA methylation patterns distinct from those of wild-type HPCs, with hypermethylated regions showing the enrichment of ETS transcription factor (TF) motifs. We found that the expression of FLI1, an ETS family member, was significantly downregulated in FPDMM-mimicking HPCs with a RUNX1 transactivation domain (TAD) mutation. We demonstrated that FLI1 promoted binding-site-directed DNA demethylation, and that overexpression of FLI1 restored their megakaryocytic differentiation efficiency and hypermethylation status. These findings suggest that FLI1 plays a crucial role in regulating DNA methylation and correcting defective megakaryocytic differentiation in FPDMM-mimicking HPCs with a RUNX1 TAD mutation.

untreated wild-type human iPSCs as a negative-control.RD refers to cells into which the RNP complex and donor DNA were delivered by electroporation.Left and right panels indicate the RUNX1 WT/R201Q and RUNX1 WT/Y287X mutation mimics, respectively.In the T7E1 assay, PCR was followed by denaturation, renaturing, and enzymatic treatment.If a mutation was present, a mismatched fragment was formed, which was cleaved by T7E1.
The wild-type control cells showed only 1205 bp fragments, which were consistent with the expected PCR product size (WT of left panel).However, cells that were electroporated with R201Q-specific genome editing machinery showed two shorter fragments of approximately 660 and 550 bp in addition to the 1205 bp fragments (RD of left panel).
Similarly, cells electroporated with Y287X-specific genome editing machinery also showed two shorter fragments (approximately 490 and 370 bp) in addition to the fragments matching the expected PCR product size (860 bp) (RD of right panel), whereas wild-type cells showed only 860 bp fragments (WT of right panel).

(b)
Overlap of genomic sites between predicted off-target candidate sites for crRNAs (blue for RUNX1 R201Q -crRNA and red for RUNX1  Wild-type (right) and FPDMM-mimicking RUNX1 WT/R201Q (abbreviated as R201Q; left) HPCs were dimension-reduced and visualized using a uniform manifold approximation and projection (UMAP) algorithm.As a result, these cells could be divided into three subclusters.Subclusters were colored red (cluster 0), green (cluster 1), or blue (cluster 2), respectively.
(d) Heatmap of the top ten marker genes for each subcluster.The color ranges from pink to yellow indicate the normalized and log-transformed expression levels from low to high, respectively.Based on the expression of known hematopoietic cell-type-specific genes, they were mapped into three cell types: erythroid lineage (cluster 0), early hematopoietic progenitors (clusters 1), and megakaryoid lineage (clusters 2).

(e)
Stacked bar plot of cell type proportions in wild-type and RUNX1 WT/R201Q HPCs.
X-axis indicates the sample names, and Y-axis indicates the percentage of each subcluster in each sample (percent clusters).These results show that cluster 0 comprises the majority of FPDMM-mimicking RUNX1 WT/R201Q HPCs compared to the wild-type cells (81.1% vs. 51.9%),with a decreasing percentage of clusters 1 (9.8% vs. 24.7%)and 2 (9.1% vs.
(g) A table of percentages of CD41 + CD42b + Mks per 20 000 of these HPCs.

Supplementary Figure 2 |
analysis.Shown left to right is the mutant iPSC line affected by the SNV, the name of gene affected by the SNV, the genomic location of the SNV, and the pattern of transversion mutation.(e) Bulk sequence traces of wild-type (left), RUNX1 WT/R201Q (middle), and RUNX1 WT/Y287X iPSCs (right) around the genomic regions where the putative SNVs validated by WES analysis are located.Stars (★) represent putative SNVs.Of the eight putative SNVs, three were specifically identified in RUNX1 WT/R201Q iPSCs and were located in the ADARB1, CNGB3, and NAALADL1 loci.Additionally, five were specifically identified in RUNX1 WT/Y287X iPSCs and were located in the ARHGEF2, GRM6, KAT2B, TAF5L, and ZFHX3 loci.(f) Confirmation of the expression of ADARB1, ARHGEF2, KAT2B, and TAF5L in wild-type HPCs (dark gray) compared with wild-type iPSCs (light gray) by qRT-PCR.The X-axis indicates the target genes, and the Y-axis indicates the fold-change.Data are presented as the mean ± SD of three biological replicates.Asterisks denote significant difference: *P < 0.05 and ns, not significant.Of the eight validated SNVs, four genes affected by SNVs (CNGB3, GRM6, NAALADL1, and ZFHX3) were not expressed in HPCs, whereas one gene affected by the SNV (TAF5L) was expressed at the same levels in wild-type iPSCs and HPCs, suggesting that those SNVs are irrelevant.The remaining three SNVs affected ADARB1, ARHGEF2, and KAT2B.The expression of these genes showed an upregulation tendency in HPCs compared with iPSCs.Evaluation of in vitro megakaryocytic differentiation processes of FPDMM-mimicking iPSCs.(a) Representative images of wild-type (upper left), RUNX1 WT/R201Q (R201Q, upper right), and RUNX1 WT/Y287X (Y287X, lower left) cells before passaging iPSCs (upper left of each group image) and on Day 1, 3, 5, 7, and 12 of in vitro hematopoietic differentiation using STEMdiff TM Hematopoietic Kit.(b) A table of average absolute numbers of CD34 + CD45 + HPCs in wild-type (WT), RUNX1 WT/R201Q (R201Q), and RUNX1 WT/Y287X (Y287X) iPSC samples per well of a 24well plate.(c) Confirmation of a potential bias in a cell state underlying wild-type and FPDMM-mimicking RUNX1 WT/R201Q HPCs by single-cell RNA sequencing analysis.
TFs in FPDMM-mimicking HPCs.(a) Enrichment distributions of ETS family TF-binding motifs.Solid lines represent probabilities at ± 5 kb from the commonly hypermethylated DMCs between RUNX1 WT/R201Q and RUNX1 WT/Y287X HPCs, and dashed lines represent probabilities at ± 5 kb from randomly selected CpGs.Red: GABPA-binding motif, light green: FEVbinding motif, orange: ELF1-binding motif, blue: ELF2-binding motif, and pink: ELK1-binding motif.(b) Distribution of ETS family TF (FEV)-binding motif enrichment.The solid line represents the probability at ± 5 kb from RUNX1 WT/Y287X HPC-specific hypermethylated DMCs, and the dashed line represents the probability at ± 5 kb from randomly selected CpGs.(c) Distributions of RUNX1-binding motif enrichment.Solid lines represent the probabilities at ± 5 kb for hypermethylated DMCs in RUNX1 WT/R201Q (upper left) and RUNX1 WT/Y287X (upper right) HPCs, and hypomethylated DMCs in RUNX1 WT/R201Q (lower left) and RUNX1 WT/Y287X (lower right) HPCs.Dashed lines represent the probabilities at ± 5 kb from randomly selected CpGs.(d) Confirmation of the expression of 16 ETS family TFs in RUNX1 WT/R201Q (blue) and RUNX1 WT/Y287X HPCs (red) compared with that in wild-type cells by qRT-PCR.The X-axis indicates the target genes, and the Y-axis indicates the fold-change.Data are presented as the mean ± SD of four biological replicates.Asterisks denote significant difference: *P < 0.05 and **P < 0.01.The expression of 12 of 28 ETS family TFs (EHF, ELF3, ELF5, ETV1, ETV2, ETV3L, ETV4, ETV7, FEV, SPIB, SPIC, and SPDEF) was undetermined.(e) Confirmation of the expression of ELF1 and FLI1 in cells hematopoietically differentiated from patient-derived iPSCs equivalent to RUNX1 WT/Y287X .The comparison was performed using two mutation-corrected FPDMM clones (isogenic controls clone E and clone F) by microarray analysis.The X-axis indicates the target genes, and the Y-axis indicates the fold-change.Data are presented as the mean ± SD of three biological replicates.The expression of FLI1 tended to be lower in patient-derived cells than in isogenic control clone E-derived cells (Y287X-equivalent vs. isogenic control clone E; P = 0.049, false discovery rate adjusted by the Benjamini-Hochberg method [FDR] = 0.18), as well as in patient-derived cells than in isogenic control clone F-derived cells (Y287Xequivalent vs. isogenic control clone F; P = 0.014, FDR = 0.19).Similarly, the expression of ELF1 tended to be lower in patient-derived cells than in isogenic control clone Ederived cells (Y287X-equivalent vs. isogenic control clone E; P = 0.052, FDR = 0.18), as well as in patient-derived cells than in isogenic control clone F-derived cells (Y287Xequivalent vs. isogenic control clone F; P = 0.077, FDR = 0.38).(f) (Left) The known HOCOMOCO v11 binding motif for FLI1 (top) and ETV6 (bottom).S: G/C, R: A/G, and M: A/C.(Right) Distribution of FLI1-and ETV6-binding motif-enrichment.The X-and Y-axes show the distance from DMC (bp) and probability of TF-binding motifs, respectively.Solid lines are probabilities at ± 5 kb for hypermethylated DMCs in RUNX1 WT/Y287X HPCs, and dashed lines are probabilities at ±