OCT4 alone was insufficient to reprogram PBMCs into iMSCs directly
Previously, we reported that lentivirally expressed OCT4 could directly reprogram human cord blood CD34+ hematopoietic progenitor cells into iMSCs with very high efficiency11. Therefore, we first tried to convert human PBMCs into iMSCs by overexpressing OCT4 alone using a clinically relevant vector system. Isolated human PBMCs were cultured in a Stemline-based erythroid medium for six days to expand erythroid progenitors. Using the nucleofection method, 2106 expanded PBMCs were transfected with our modified oriP/EBNA1-based episomal vector, which expressed OCT4 under a strong SFFV promoter (Fig.1a), as we previously described11. Cells were then cultured in MSC medium11 supplemented with small molecules that promote reprogramming (3M CHIR99021, 10M forskolin, 10M ALK inhibitor (SB431542), and 5M tranylcypromine hydrochloride)15. However, there was no MSC-like colony formation 2 weeks later, indicating that OCT4 alone was insufficient to convert human PBMCs into iMSCs directly (Fig.1b).
a Schematic diagram of the episomal vector plasmids. SFFV is the spleen focus-forming virus U3 promoter; WPRE, posttranscriptional regulatory element; SV40PolyA, polyadenylation signal from SV40 virus; OriP, EBV (EpsteinBarr virus) origin of replication; EBNA1, EpsteinBarr nuclear antigen 1. b Colony formation at day 14 after nucleofection with 2106 PBMCs and maintenance in MSC culture conditions. c Reprogramming efficiency with different combinations of reprogramming factors. Error bars indicate standard deviation. n=3 biologically independent samples for each group. d Fluorescence-activated cell sorting (FACS) analysis of iMSCs 8 days after reprogramming with different factor combinations. SOX2 induced iPSCs generation (TRA-1-60+ cells). However, SOX9 did not induce detectable TRA-1-60+ cells. e Colony formation at day 14 after nucleofection with 1106 PBMCs (control) or CD34+-depleted PBMCs followed by maintenance in MSC culture conditions.
Our previous studies showed that BCL-XL is a critical reprogramming factor in blood cell reprogramming9,16, which increased the reprogramming efficiency by 10-fold when converting PBMCs into iPSCs using Yamanaka factors16. Here, we observed that transfection of PBMCs with OCT4, BCL-XL, and MYC (OBM) led to the formation of MSC-like colonies 2 weeks later (Fig.1b), albeit at low efficiency. The combination of any two of the OBM factors failed to generate iMSC colonies (Fig.1b). To improve the reprogramming efficiency further, we examined OBM with different combinations of other factors for generating iPSCs, including KLF4 and SOX2. KLF4 moderately improved iMSC generation, whereas SOX2 increased reprogramming efficiency by ~5-fold (Fig.1c and Supplementary Data1). However, the presence of SOX2 in the reprogramming cocktail resulted in ~12% of reprogrammed cells expressing iPSC markers, e.g., TRA-1-60 (Fig.1d) and NANOG (Supplementary Fig.1), even in MSC expansion culture conditions. Since iPSCs may induce teratomas, the SOX2-containing approach is not clinically prudent.
We decided to replace SOX2 with SOX9 because SOX9 plays an important role in skeletal development and chondrogenesis17,18. Surprisingly, SOX9 showed greater potency than SOX2 in iMSC reprogramming (Fig.1c). As expected, SOX9 virtually abolished the generation of TRA-1-60-expressing cells (Fig.1d). To ensure the absence of undetectable levels of iPSCs after reprogramming with SOX9, we cultured iMSCs in iPSC medium for 1 week. Phenotyping analysis of cultured cells showed no expression of iPSC markers. These data suggested that SOX9 restricted cell fate to iMSCs, whereas SOX2 would overshoot the reprogramming of a proportion of PBMCs beyond the stage of iMSCs. Moreover, after reprogramming with SOX9, PBMCs transformed morphologically to spindle-like cells resembling MSCs within 46 days, whereas SOX2-reprogrammed cells did not display spindle-like morphology (Supplementary Fig.2a).
Although PBMCs are composed of many different cell types, based on our previous studies3,16,19, we hypothesized that the CD34+ cell subset in peripheral blood was the most amenable to reprogramming to iMSCs. After six days of culture in hematopoietic stem cell expansion medium, the percentage of CD34+ cells in PBMCs increased from <1% to ~45%. When we depleted CD34+ cells from PBMCs before inducing reprogramming, no MSC-like colonies were observed (Fig.1e). These results suggested that the five reprogramming factors converted the CD34+ hematopoietic stem cells and progenitors but not the matureblood cells into iMSCs.
Having observed that the combination of OCT4, BCL-XL, MYC, KLF4, and SOX9 (named as 5F) induced the highest levels of PBMC conversion without overshooting the iMSC reprogramming process, we used the five factors (5F) for reprogramming in subsequent experiments. In all, 57 days after nucleofection of PBMCs with 5F, dozens of MSC-like colonies were observed. At approximately 2 weeks, reprogrammed cells resembled MSCs with typical spindle-like morphology (Fig.2a). The expression of MSC markers such as CD90 and CD73 increased from ~5% of reprogrammed cells by ~1 week to ~15% and 40% of the cells, respectively, by week 2 and >75% by week 3 (Fig.2b and Supplementary Data2). Four weeks after reprogramming with 5F, almost all cells expressed typical MSC markers: CD29 (99.7%), CD73 (95.3%), CD90 (96%), and CD166 (80%) (Fig.2c, d). The expression of hematopoietic markers such as CD45 and CD34 was negligible (Fig.2e). In addition, OCT4+ cells were not detectable (Supplementary Fig.3). Next, we evaluated the immunomodulatory potential of the iMSCs. We found that our 5F iMSCs were able to significantly suppress T-cell proliferation (CD4+ and CD8+ T-cell subsets) after 3 or 6 days (Fig.2f, Supplementary Fig.4a, and Supplementary Data3) co-culture with PBMCs. To further determine if the reprogramming to iMSCs or their expansion in culture may cause any chromosomal abnormalities, we performed digital karyotyping using SNP arrays. We did not identify any chromosomal abnormalities after either 1 week or 4 weeks of in vitro culture (Supplementary Figs.57). These data demonstrated that human PBMCs can be efficiently reprogrammed into iMSCs using our nonintegrating episomal vector system.
a Representative images of human PBMCs and iMSCs 14 days after reprogramming with five factors (5F). Scale bar represents 100m. b Changes in the percentage of cells expressing the MSC markers CD73 and CD90 as measured by flow cytometry of 5F-transfected PBMCs over time.c,d Flow cytometry plots of typical MSC marker expression (CD29, CD73, CD90, CD166)at 4 weeks after reprogramming. n=3 biologically independent samples for time point. eBlood cell markers (CD45 and CD34) were assessed 4 weeks after transfection of reprogramming factors. f iMSCs significantly inhibited T-cell proliferation after 3 days of co-culture with PBMCs. **P=0.0007. Error bars indicate standard deviation. n=3 biologically independent samples for each group.
To assess the essentiality of the five factors, we performed reprogramming by omitting a single factor in separate experiments. PBMCs from various donors were used. Surprisingly, we found that skipping OCT4, a critical factor for blood cell reprogramming, still allowed the generation of a considerable number of MSC-like colonies (Fig.3a and Supplementary Data4). In addition, PBMCs could be converted to iMSCs without KLF4, although at a ~35% decreased efficiency (Fig.3a). Omitting SOX9 not only significantly reduced the number of colonies formed but the reprogrammed cells were round in shape instead of spindle-like MSCs suggesting that SOX9 played a pivotal role in determining the MSC fate (Supplementary Fig.2b). By comparison, hardly any colonies were formed in the absence of BCL-XL or MYC. Taken together, SOX9, BCL-XL, and MYC were indispensable for reprogramming PBMCs into iMSCs.
a Reprogramming efficiency with the five-factor combination and removing one of the five factors. One-way ANOVA and Dunnetts multiple comparisons test, *P<0.05 vs. 5F group, ***P<0.001 vs. 5F group. ns: not significant. Error bars indicate standard deviation. n=5 for each group from biological independent donors. b Flow cytometry analysis of the MSC marker CD73 4 weeks after transfection with 5F, 4FnoO (no OCT4), and 4FnoK (no KLF4). c Flow cytometry analysis of the MSC markers CD73 and CD90 at 2, 3, and 4 weeks after transfection with 5F, 4FnoO (no OCT4), or 4FnoK (no KLF4). df RTqPCR analysis of osteogenesis-, adipogenesis-, and chondrogenesis-related genes in iMSCs reprogrammed with 5F, 4FnoO, and 4FnoK 2 weeks after multilineage differentiation. Tukeys multiple comparisons test, *P<0.05, 4FnoO vs. 5F and 4FnoK group. #P<0.05, 4FnoO vs. 4FnoK group. n=4 biologically independent samples for each group. Error bars indicate standard deviation (SD). g Multilineage differentiation of iMSCs reprogrammed with 5F, 4FnoO, or 4FnoK. Cells were cultured in osteogenic, adipogenic, or chondrogenic induction medium for 24 weeks and stained with Alizarin Red (osteogenesis), Oil Red O (adipogenesis), or Alcian blue (chondrogenesis), respectively. Scale bars represent 200m.
The iMSCs generated with the three different combinations of reprogramming factors, 5F, 4FnoO (5F minus OCT4), and 4FnoK (5F minus KLF4), were morphologically similar: they were all spindle-shaped, resembling MSCs (Supplementary Fig.2b). We evaluated the proliferation of the iMSCs generated from different conditions and compared it with primary human bone marrow MSCs (BMMSCs) (Supplementary Fig.2c). Primary human BMMSCs showed slowed proliferation after ~1 month in culture. The iMSCs reprogrammed from PBMCs displayed an enhanced in vitro proliferative capacity compared with BMMSCs. While the 5F iMSCs and 4FnoK iMSCs have similar proliferation ability, the 4FnoO iMSCs showed slower proliferation compared with the other two types of iMSCs (5F iMSCs and 4FnoK iMSCs). More than 100-fold more 5F iMSCs were generated than the human primary BMMSCs after ~1 month culture. In addition, >90% of the reprogrammed cells expressed the MSC marker CD73 (Fig.3b) 4 weeks after vector transfection. To monitor the reprogramming process in more detail, we evaluated the expression of the MSC markers CD73 and CD90 at 2-, 3-, and 4-week post-transfection (Fig.3c). We found that more than 60% of cells reprogrammed from either 5F or 4FnoK conditions became CD90+ by week 2, whereas only ~6% of cells from 4FnoO were CD90+, suggesting that OCT4 promoted the formation of CD90+ cells.
A characteristic feature of MSCs is the potential for trilineage differentiation into osteoblasts, adipocytes, and chondrocytes20. To assess the functionality of iMSCs reprogrammed with 5F, 4FnoO, or 4FnoK, we cultured iMSCs in three lineage-specific induction media, followed by RTqPCR analysis on the marker genes of osteogenesis, adipogenesis, and chondrogenesis.
The expression levels of runt-related transcription factor 2 (RUNX2), an early marker of osteogenic commitment, as well as the later osteogenic markers SP7 and alkaline phosphatase (ALP), were significantly decreased in the 4FnoO-reprogrammed iMSCs compared with 5F- or 4FnoK-reprogrammed iMSCs (P=0.01 and 0.03, respectively; Tukeys multiple comparisons test, Fig.3d and Supplementary Data5). To confirm the osteogenic commitment, we assessed calcium deposits by Alizarin Red S staining. Mineralization was observed in iMSCs reprogrammed with either 5F or 4FnoK but not in 4FnoO-reprogrammed iMSCs (Fig.3g).
Regarding chondrogenic differentiation, there was no significant difference in the expression of chondrogenic marker genes such as ACAN among the three groups (Fig.3e and Supplementary Data5). Alcian blue staining, which stains for aggrecans associated with MSC chondrogenic potential, also showed no significant difference among the three groups (Fig.3g). However, SOX9 expression was significantly reduced in 4FnoO iMSCs (4FnoO vs. 4FnoK, P=0.005). These data suggested that omitting OCT4 also impaired the chondrogenic differentiation potential of iMSCs. Taken together, these five factors were necessary for the generation of iMSCs with unbiased differentiation potential. Conversely, reprogramming without OCT4 led to the formation of dysfunctional iMSCs.
After the induction of adipogenic differentiation, lipoprotein lipase (LPL) and fatty acid-binding protein 4 (FADP4) were expressed at substantially lower levels in 4FnoO iMSCs than in either 5F or 4FnoK iMSCs (Fig.3f and Supplementary Data5). We used Oil Red O staining to visualize lipid droplets in functional adipocytes. Consistent with the adipogenic gene expression data, iMSCs reprogrammed without OCT4 failed to differentiate into functional adipocytes (Fig.3g). Of interest, omitting KLF4 led to the expression of higher levels of adipocyte markers and the formation of larger oil droplets, suggesting that KLF4 played a role in restricting adipogenic-biased MSCs.
To evaluate the immunomodulatory potentials of iMSCs reprogrammed with 5F, 4FnoO, or 4FnoK, we compared a list of major immunoregulatory cytokines, chemokines, and soluble factors secreted by MSCs21,22 using the normalized gene counts from the RNA-seq data (Supplementary Fig.4b and Supplementary Data6). We found that compared with 5F iMSCs, in addition to impaired trilineage differentiation potential, the 4FnoO iMSCs showed significantly reduced gene expression on many immunoregulatory cytokines/chemokines, such as IL-10, HGF, VCAM1, CCL2, CXCL14 (Supplementary Fig.4b). Both 5F and 4FnoK iMSCs showed comparable levels of immunoregulatory cytokines/chemokines gene expression compared to the primary human bone marrow-derived MSCs23.
To investigate the mechanisms underlying the distinct features of iMSCs reprogrammed with different factors (i.e., 5F, 4FnoO, and 4FnoK), we conducted transcriptome analysis 4 weeks after reprogramming factor transfection. We chose 4 weeks because >90% of the reprogrammed cells expressed MSC markers at this time point, and the nonintegrating episomal viral vectors were cleared from the reprogrammed cells7. First, we investigated the differentially expressed genes (DEGs) between the 5F, 4FnoO, or 4FnoK iMSCs. DEG analysis identified 827 significantly down- and 538 significantly upregulated genes in 4FnoO iMSCs compared to 5F iMSCs (FDR<0.05 and fold change (FC)>2, Fig.4a and Supplementary Data7). Of note, 5F and 4FnoK iMSCs showed similar transcriptomes with only 24 DEGs, consistent with their seemingly identical differentiation potentials (Supplementary Fig.8). Hierarchical clustering analysis identified a set of genes highly enriched in 5F and 4FnoK iMSCs, some of which were reported as MSC lineage signature genes, such as SRPX, S1PR3, ROBO2, NCAM1, COL5A1, and COL4A1 etc24,25,26 (Fig.4b and Supplementary Data7). Furthermore, the 4FnoO iMSCs displayed a significant decrease in the expression of mesoderm-regulating genes, including SOX4, SALL4, and TWIST1 (Supplementary Data7). We speculated that these downregulated genes might be associated with the impaired functionality of 4FnoO iMSCs. We then performed Gene Ontology (GO) enrichment analyses to explore the pathways associated with genes expressed at low levels in 4FnoO iMSCs. We found that 1365 DEGs were enriched in the biological processes of axonogenesis, extracellular structure organization, ossification, and cartilage development (Fig.4c). The top identified Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways were the PI3K-Akt signaling and calcium signaling pathways (Fig.4d). These data helped explain the functional defects in osteogenesis of 4FnoO iMSCs and further understanding of the role of OCT4 in reprogramming PBMCs into iMSCs.
a Volcano plot showing differentially expressed genes identified in 4FnoO iMSCs compared with 5F iMSCs. Each dot represents a gene. The red dots are genes significantly upregulated (right) or downregulated (left) in 4FnoO iMSCs (Cutoff: P<10e6, fold change>2). b Heatmap showing the top 30 differentially expressed genes between 5F iMSCs and 4FnoO iMSCs (ranked by p-value). c, d Dot plots showing the top Gene Ontology (GO) biological process (BP) terms (c) and KEGG pathways (d) enriched from DEGs in 4FnoO iMSCs compared to 5F iMSCs. e PCA of RNA-seq from iMSCs 4 weeks after reprogramming with 5F, 4FnoO, or 4FnoK, primary human bone marrow-derived MSCs (BMMSC) and primary adipose-derived MSCs (AdMSC). For each condition, iMSCs were reprogrammed from PBMCs derived from three biologically independent donors. f Pearson correlation analysis of iMSCs and primary MSCs. g Comparison of twenty-four genes previously determined to be specific to the MSC lineage between primary MSCs and iMSCs.
To compare the iMSCs reprogrammed from PBMCs with primary human MSCs, we downloaded RNA-seq data generated from primary human bone marrow-derived MSCs (BMMSC)23 and primary human adipose-derived MSCs (AdMSC). First, we analyzed the transcriptional similarity of the iMSCs in our study to the primary human MSCs using principal component analysis (PCA) (Fig.4e). The reduction of the multi-dimensional dataset into two principal component (PC) dimensions enables the unbiased comparison and visualization of the transcriptomes between samples. As expected, the results showed that 4FnoO iMSCs were distinct from the other two iMSC groups (Fig.4e), consistent with the impaired differentiation potential of 4FnoO iMSCs when compared with 5F and 4FnoK iMSCs. The transcriptomes of human BMMSC and AdMSC were very similar to each other. Furthermore, the variation captured in PC1 demonstrated closer similarity of 5F and 4FnoK iMSCs with the primary MSCs compared to 4FnoO iMSCs, which tended to cluster further away from BMMSC and AdMSC (Fig.4e). Pearson correlation analysis confirmed that the 4FnoK and 5F iMSCs retained strong transcriptome correlation with the primary MSCs, while the 4FnoO iMSCs had less correlation with the primary MSCs (Fig.4f). A panel of 24 MSC lineage genes25,26 were compared between the primary MSCs and our iMSCs (Fig.4g). The 4FnoO iMSCs showed distinct expression patterns of these MSC signature genes that contrasted strongly with other groups. Noteworthy is that COL4A1, COL5A1, LOX, NNMT, which are known to be upregulated in MSCs versus fibroblasts24,27, were downregulated in 4FnoO iMSCs.
Genome-wide chromatin accessibility can provide mechanistic insights at the molecular level into cell fate decisions, especially during the reprogramming process. Thus, we performed ATAC-seq28 analysis on iMSCs 4 weeks after reprogramming PBMCs with 5F, 4FnoO, or 4FnoK. Open chromatin regions were identified as peaks in the ATAC-seq dataset. Furthermore, after peak calling, the relative genomic distribution of ATAC peaks showed reduced peaks within promoter regions in iMSCs generated without OCT4 (Fig.5a). In contrast, these cells had more open chromatin at intron regions. These results suggested that OCT4 may preferentially bind promoter regions to promote chromatin accessibility during reprogramming.
a Genomic location of ATAC-seq peaks from 5F, 4FnoO, and 4FnoK iMSCs. b PCA using normalized ATAC-seq counts from 5F, 4FnoO, and 4FnoK iMSCs, and two datasets from bone marrow-derived CD34+ cells (SRR2920489 and SRR2920490). For each condition, the chromatin accessibility was profiled from iMSCs that were reprogrammed from two biologically independent donors. c Heatmap showing ATAC-seq signals with the top 200 most different peaks (ranked by padj). Red represents chromatin regions with more mapped reads, suggesting possible chromatin openness. Gray represents chromatin regions with fewer mapped reads, suggesting closed chromatin. d Selected genomic views of the ATAC-seq data using IGV (2.8) for the indicated groups. For each gene, all genome views are on the same vertical scale. e The bar plot showing RNA-seq gene expression values for the respective genes shown above in the genome view. RNA-seq gene expression levels are shown as log2() normalized read counts. n=3 biologically independent samples for each group. *P0.05; error bars indicate standard deviation.
Similar to what was observed in the RNA-seq transcriptomic data, PCA of normalized ATAC-seq read counts showed that chromatin accessibility of three groups of iMSCs (5F, 4FnoO, and 4FnoK) were well-separated from each other, in which the accessible chromatin regions were mainly different in 4FnoO cells (PC1=52% variance, Supplementary Fig.9). However, in contrast to the similar transcriptomes between 5F and 4FnoK iMSCs (Supplementary Fig.8 and Fig.4a), ATAC-seq analysis showed that therewas aclear separation between 5F and 4FnoK iMSCs (PC2=19%, Supplementary Fig.9). These data suggested that both OCT4 and KLF4 facilitate chromatin remodeling during reprogramming. To compare the changes in chromatin accessibility during reprogramming, we downloaded the ATAC-seq data of primary CD34+ cells from bone marrow (SRR2920489, SRR2920490)29, which are similar to our reprogramming-initiating cells in this study. The datasets were processed using the same analysis pipeline. PCA revealed that CD34+ hematopoietic progenitor cells clustered separately from the three groups of reprogrammed iMSCs (Fig.5b), whereas 5F iMSCs and 4FnoK iMSCs were clustered closely with each other.
We also noticed that some chromatin regions remained closed in both CD34+ and 4FnoO iMSCs, whereas the same regions were in an open configuration in the 5F and 4FnoK iMSCs (Fig.5c). These data suggested that OCT4, but not KLF4, played a critical role in opening chromatin during the reprogramming process. More specifically, OCT4 opened the chromatin of the stemness-associated gene SALL4, Wnt signaling-related genes such as SFRP4, microtubule-binding and glutamate receptor binding-related genes JAKMIP2 and SYNDIG1, and MSC lineage signature gene NNMT (Fig.5d). These genes with reduced ATAC-seq peaks in 4FnoO iMSCs also showed significantly reduced mRNA expression, indicating a consistency between transcriptome and chromatin accessibility data (Fig.5e and Supplementary Data8).
DNA methylation is the most common epigenetic modification of the genome to control gene expression in mammalian cells30 and the differentiation or self-renewal of MSCs13. To determine the effects of reprogramming factors on methylation levels and patterns in iMSCs, we assessed genome-wide CpG methylation profiles in 5F, 4FnoO, and 4FnoK iMSCs at week four using RRBS. First, we profiled CpG methylation patterns on five different genomic features (all sites, promoters, exons, introns, and transcription start sites (TSSs) (Fig.6a, b and Supplementary Data9). We found that iMSCs reprogrammed without OCT4 showed a globally hypermethylated CpGs compared to iMSCs reprogrammed with OCT4 (Fig.6a, b). Specifically, when reprogramming in the absence of OCT4, we identified 10,760 differentially methylated cytosines (DMCs) (20%, q=0.1, Supplementary Data10), of which 9004 DMCs were hypermethylated and 1756 DMCs were hypomethylated (4FnoO vs. 5F). Among these sites, 7.7% were within promoter regions, and 7.9% werewithin exon regions (Fig.6c). In contrast, there was no significant difference in CpG methylation within all five genomics features in the iMSCs when reprogrammed in the absence of KLF4 (Fig.6a, b). Of the 3849 CpG sites significantly different (20%, q=0.1) between the 5F and 4FnoK groups, 3698 CpG sites were hypermethylated, and 151 sites were hypomethylated. When measuring the average methylation against the distance to the TSS, there was a global hypermethylation pattern in the iMSCs reprogrammed without OCT4 (Fig.6d, p<0.0001), suggesting that OCT4 was critical for global demethylation during reprogramming of PBMCs to iMSCs.
a The bar graph showing the methylation levels of all sites, promoters, exons, and intron regions from 5F, 4FnoO, and 4FnoK iMSCs. n=2 biologically independent samples for each group. b The methylation levels of the TSS region. n=2 biologically independent samples for each group. c The percentage of differentially methylated CpGs (DMCs) between 5F and 4FnoO iMSCs annotated within the promoter, exon, intron, and intergenic regions shown in the pie chart. d The average methylation levels surrounding the TSSs (5000 to +5000bp) in 5F, 4FnoO, and 4FnoK iMSCs. e Hierarchical clustering and heatmap analysis of 13,974 DMCs. f The bar plot showing the log2() normalized read counts from RNA-seq. n=3 biologically independent samples for each group. *P<0.05; error bars indicate standard deviation.
We performed hierarchical clustering on six RRBS datasets and generated a heatmap using the beta value of all common CpG sites. As expected, two datasets from 4FnoO clustered together, enriched a set of hypermethylated DMCs that were not observed in the 5F and 4FnoK datasets (Fig.6e). Since the cells reprogrammed from 5F and 4FnoK were very similar in their transcriptomes, chromatin openness, and methylation levels, we focused on our comparisons in the iMSCs programmed using 5F vs. 4FnoO. We annotated 10,760 DMCs and identified 665 differentially methylated genes (DMGs) between 5F and 4FnoO iMSCs (Supplementary Data10) which were subject to GO enrichment analysis (Supplementary Fig.10). Similar to the GO enrichment analysis based on RNA-seq data, DMGs were enriched in axonal guidance signaling and mesenchyme development. Of note, POU5F1, SALL4, NCAM1, HDAC4, and MSC lineage signature gene COL5A1 were significantly hypermethylated in iMSCs reprogrammed using 4FnoO compared with the iMSCs programmed using 5F (Supplementary Data10), suggesting that these genes might be associated with the impaired functionality in the 4FnoO iMSCs.
Demethylation may occur passively. DNMT1 is the most abundant DNA methyltransferase in mammalian cells and is considered the key methyltransferase responsible for DNA methylation maintenance, and its inhibition will result in passive demethylation. We found that the expression levels of DNMT1 in iMSCs reprogrammed with or without OCT4 were similar (Fig.6f and Supplementary Data8), suggesting minimal role of DNMT1 in OCT4-mediated demethylation. We then suspected that active DNA demethylation might have contributed to the global hypomethylation. Active DNA demethylation is mainly regulated by ten-eleven translocation (TET) enzymes31. We observed that the expression of TET1, but not TET2, was significantly reduced when reprogramming without OCT4 (Fig.6f), suggesting that TET1 might have contributed to OCT4-induced global demethylation. Meanwhile, the expression level of DNMT3B was significantly increased when reprogramming without KLF4, suggesting a role of KLF4 in regulating DNA methylation homeostasis via de novo DNA methyltransferase DNMT3B (Fig.6f).
To assess the influence of methylation on gene expression, we performed integration analysis of DMGs and DEGs datasets. We found the co-occurrence of 67 genes between 5F and 4FnoO iMSCs (Fig.7a and Supplementary Table1). Hypergeometric test was applied to show that the overlap is significant. Our analysis suggested that the observed difference in functionality between 5F and 4FnoO iMSCs might be a consequence of the difference in the methylation status of these 67 genes. Among these genes, ZFHX4, SLC8A2, NCAM1, TFPI2, and SALL4 were the most differentially expressed (Fig.7b). When PBMCs were reprogrammed without OCT4, not only were these genes significantly hypermethylated on either promoters or exons compared to PBMCs reprogrammed with OCT4 (Supplementary Data10), but some chromatin regions of these genes also remained inaccessible/closed (Fig.7c). Consistent with the hypermethylation of the four genes, their transcription levels were close to zero (Fig.7d and Supplementary Data8).
a Venn diagram illustrating the overlap between the differentially expressed genes (DEGs) and differentially methylated genes (DMGs) between 5F iMSCs and 4FnoO iMSCs. A total of 1365 DEGs and 665 DMGs were identified; 67 of these were both differentially expressed and differentially methylated. b Volcano plot showing 67 overlapping genes between the DEG and DMG. pCutoff=10e6, log2 FC>1). c Selected genomic views of the ATAC-seq data using IGV (2.8) for the indicated groups. For each gene, all genome views are on the same vertical scale. d The bar plot showing the RNA-seq gene expression values for the respective genes, which are shown above in the genome view. RNA-seq gene expression levels are shown as log2() normalized read counts. *, P<0.05; error bars indicate standard deviation. n=3 biologically independent samples for each group. e Heatmap showing the normalized gene read count after log2() transformation from RNA-seq.
ZFHX4, a transcription-related zinc finger protein involved in the mesodermal commitment pathway, is upregulated in both embryonic stem cell-derived and bone marrow (BM)-derived MSCs32,33. These reports, together with our findings, indicate that ZFHX4 may serve as an MSC marker. In addition, neural cell adhesion molecule (NCAM), also called CD56, is expressed on human MSCs and was proposed as a marker for human MSC isolation34,35. Also, CD56+ cells showed increased colony formation ability, suggesting CD56 expression enriches MSCs with self-renewal potency36. On the other hand, BM-MSCs from NCAM-deficient mice exhibited defective migratory ability and significantly impaired adipogenic and osteogenic differentiation potential37.
Many genes have been proposed as MSC surface marker genes, but no consensus has been reached yet. To screen possible trilineage differentiation function associated MSC markers, we compared ten well-established MSC surface markers between primary MSCs and our iMSCs (Fig.7e and Supplementary Fig.11). We found that other than NCAM1, four additional MSC surface markers (CD90, PDGFRB, CD82, and FZD5) were highly expressed in both primary MSCs and 5F/4FnoK iMSCs but downregulated in 4FnoO iMSCs (Fig.7e). Taken together, integrated analysis of multiomics data lead to the identification of putative functional MSC markers, and our dataset enables the mining for additional MSC surface markers that co-associate with functional potential.
The rest is here:
Reprogramming of human peripheral blood mononuclear cells into ... - Nature.com
