mRNA and miRNA expression profiles in an ectoderm-biased substate of human pluripotent stem cells

The potential applications of human pluripotent stem cells, embryonic stem (ES) cells, and induced pluripotent stem (iPS) cells in cell therapy and regenerative medicine have been widely studied. The precise definition of pluripotent stem cell status during culture using biomarkers is essential for basic research and regenerative medicine. Culture conditions, including extracellular matrices, influence the balance between self-renewal and differentiation. Accordingly, to explore biomarkers for defining and monitoring the pluripotent substates during culture, we established different substates in H9 human ES cells by changing the extracellular matrix from vitronectin to Matrigel. The substate was characterised by low and high expression of the pluripotency marker R-10G epitope and the mesenchymal marker vimentin, respectively. Immunohistochemistry, induction of the three germ layers, and exhaustive expression analysis showed that the substate was ectoderm-biased, tended to differentiate into nerves, but retained the potential to differentiate into the three germ layers. Further integrated analyses of mRNA and miRNA microarrays and qPCR analysis showed that nine genes (COL9A2, DGKI, GBX2, KIF26B, MARCH1, PLXNA4, SLC24A4, TLR4, and ZHX3) were upregulated in the ectoderm-biased cells as ectoderm-biased biomarker candidates in pluripotent stem cells. Our findings provide important insights into ectoderm-biased substates of human pluripotent stem cells in the fields of basic research and regenerative medicine.

biomarkers for defining and monitoring the pluripotent substates during culture, we established different substates in H9 human ES cells by changing the extracellular matrix from vitronectin to Matrigel. We then evaluated the expression patterns of various markers, examined the differentiation potential of the cells, and performed integrated analyses of mRNA and miRNA microarrays of the cells and exosomes to explore high-quality and reliable biomarkers. Our findings provided insights into the use of biomarker candidates during the earliest stages of ectodermal differentiation.

Results and Discussion
Changes in the extracellular matrix induced two different substates in H9 cells. Pluripotent stem cells require a balance between self-renewal and differentiation, and various factors (e.g., growth media, extracellular matrices, and environmental cues) are crucial for the maintenance of pluripotency 8,9 . To establish techniques for quality evaluation of human pluripotent stem cells using biomarkers, we first examined the conditions for pluripotency in H9 human ES cells by changing the extracellular matrix. H9 cells cultured in TeSR-E8 on Vitronectin XF were transferred to Matrigel. Typical colonies were observed under both conditions, and colonies with wide intercellular spaces appeared under Matrigel conditions, as observed by bright-field microscopy ( Fig. 1a). The epithelial-to-mesenchymal transition is observed during ES/iPS cell differentiation 21 . Accordingly, we performed immunocytochemistry with no permeabilisation using antibodies for the pluripotency marker R-10G epitope and the mesenchymal marker vimentin (Fig. 1a) 21,22 . Under vitronectin conditions, the R-10G epitope was highly expressed on the cell surfaces of almost all colonies. A few colonies showed vimentin expression. Under Matrigel conditions, colonies similar to those under vitronectin conditions showed strong expression of the R-10G epitope and almost no expression of vimentin. In contrast, colonies with wide intercellular spaces showed weak expression of R-10G epitope and strong expression of vimentin.
We then counted the colony numbers according to staining categories (Fig. 1b). The percentages of R-10G ++/vimentin− colonies under vitronectin and Matrigel conditions were 94.0% and 38.8%, respectively. R-10G + and vimentin + colonies, which may represent decreased pluripotency compared with R-10G ++ and vimentin− colonies, clearly increased to 60.4% under Matrigel conditions, showing a greater increase than that under vitronectin conditions. Thus, changes in the extracellular matrix from vitronectin to Matrigel induced two different substates. These results also indicated that there was an inverse correlation between R-10G epitope and vimentin expression. Vitronectin XF consists of recombinant human vitronectin protein, whereas Matrigel consists of extracellular matrix proteins, including laminin, collagen IV, heparin sulphate proteoglycans, entactin/nidogen, and growth factors. By changing the extracellular matrix, various factors in Matrigel may affect H9 cells. Importantly, Li et al. suggested that Matrigel induces ectoderm differentiation of embryoid bodies 23 . Matrigel is known to be suitable for pluripotent stem cell culture under feeder-free conditions 24 , whereas changing the extracellular matrix should be performed more carefully. Based on bright-field and immunohistochemical observations, we designated R-10G ++ and vimentin− typical colonies as "substate 1" and R-10G + and vimentin + colonies with wide intercellular spaces as "substate 2".

Substate 1 and substate 2 cells showed reversibility.
To investigate the reversibility between substate 1 and substate 2 cells, we examined whether substate 1 and substate 2 cells produced substate 2 and substate 1 cells, respectively. We performed live cell sorting of substate 1 and substate 2 cells by flow cytometry analysis using anti-vimentin antibodies. Substate 1 and substate 2 cells were separated from about 20% on the left and (a) Immunohistochemistry with no permeabilisation using antibodies for a pluripotency marker (R-10G epitope) and a mesenchymal marker (vimentin [VIM]). Nuclei were counterstained with DAPI. The expression levels of R-10G epitope and VIM were classified as "++" and "+" or "+" and "−". The scale bar represents 200 µm. (b) Colony distributions for R-10G epitope and VIM expression patterns. The colonies were manually counted (n = 100, in triplicate). Values are the means ± standard errors. *** indicates P < 0.001 by unpaired two-tailed t-tests. R-10G++/VIM− and R-10G+/VIM + colonies were classified as "substate 1" and "substate 2", respectively. right sides of vimentin signals, respectively (Fig. 2a). Substate 1 and substate 2 cells were subcultured after sorting and were then used for flow cytometry analysis. The subcultured substate 1 and substate 2 cells consisted of both vimentin + and vimentin− cells (Fig. 2b,c). The subcultured cells were also immunostained using anti-R-10G and anti-vimentin antibodies. Both substate 1 and substate 2 cells produced R-10G ++/vimentin− and R-10G +/vimentin + colonies (Fig. 2d). These results indicated that substate 2 cells did not deviate irreversibly from substate 1 cells and that substate 1 and substate 2 cells could show reversibility of their substates.

Substate 2 cells showed decreased expression of some cell surface markers related to pluripotency.
To characterise the two substates of H9 cells maintained on Matrigel with TeSR-E8 medium, we next performed immunocytochemistry with no permeabilisation using antibodies for cell surface markers of pluripotency and differentiation (Fig. 3a). The vimentin− and vimentin + colonies indicated substate 1 and substate 2 cells, as shown in Fig. 1, respectively (Fig. 3a). The cell surface markers SSEA-4 and TRA-1-60 characterise pluripotency, and the marker SSEA-1 characterises early differentiation in ES cells 25 . The epithelial marker E-cadherin has important functions in maintaining the pluripotency of iPS and ES cells 26 . Notably, the SSEA-4 signal was observed in both substate 1 and substate 2 colonies and could not be used to distinguish the substates (Fig. 3a). The TRA-1-60 signal was strong in substate 1 colonies and weak in substate 2 colonies, similar to the R-10G signal (Fig. 3a). E-cadherin was expressed only in substate 1 colonies but not in substate 2 colonies (Fig. 3a). The SSEA-1 signal for early differentiation was not detected in both substate 1 and substate 2 colonies (Fig. 3a), indicating that substate 2 cells were barely differentiated, as shown in Fig. 2. However, the expression patterns of the cell surface pluripotent markers SSEA-4, TRA-1-60, R-10G, and E-cadherin were different in substate 1 and substate 2 colonies. ZEB1 was specifically expressed in the nucleus of substate 2 cells. Next, we performed immunocytochemistry with permeabilisation using antibodies for transcription factors related to pluripotency and the epithelial-to-mesenchymal transition (Fig. 3b). Oct3/4 and Nanog are critical transcriptional regulators underlying pluripotency in ES cells 27,28 . OCT3/4 and NANOG were expressed in the nuclei of both substate 1 and substate 2 cells (Fig. 3b). The expression levels were slightly weaker in substate 2 cells than in substate 1 cells. The transcription factors ZEB, SNAI, and TWIST are known to contribute to the epithelial-to-mesenchymal  Substate 2 was an ectodermally biased pluripotent substate. To investigate differences in trilineage differentiation potential, substate 1 and substate 2 cells were concentrated by manually removing other cells and induced to differentiate into the three germ layers using a STEMdiff Trilineage Differentiation Kit. The induced cells were immunostained using antibodies for endoderm markers (SRY-box 17 [SOX17] and alpha-fetoprotein [AFP]), mesoderm markers (brachyury and α-smooth muscle actin [αSMA]), and ectoderm markers (orthodenticle homeobox 2 [OTX2] and neuron-specific class III beta-tubulin [TUJ1]) ( Fig. 4) 31,32 . Endoderm-or mesoderm-induced cells of both substate 1 and substate 2 expressed SOX17 or brachyury in the nucleus and AFP or αSMA in the cytoplasm, respectively (Fig. 4). There were only minor differences in the expression levels of these targets in substate 1 and substate 2 cells. OTX2 expression in the nucleus did not differ between ectoderm-induced substate 1 and substate 2 cells (Fig. 4). Interestingly, TUJ1 was detected in the cytoplasm in ectoderm-induced cells of both substate 1 and substate 2, although only substate 2 cells formed neuron-like structures (Fig. 4). These results indicated that both substate 1 and substate 2 cells were able to differentiate into the three germ layers, i.e., endoderm, mesoderm, and ectoderm. Because ZEB1 controls neuron differentiation through transcriptional activation 33 , substate 2 cells expressing ZEB1 may differentiate more easily into ectoderm  34 . These findings and our study indicated that Matrigel is an ectoderm-biased matrix.

Ectoderm-biased substate 2 cells exhibited neural differentiation signals.
We established ectoderm-biased substate 2 cells, which tended to differentiate into ectoderm. To explore new biomarkers for the ectoderm-biased pluripotent substate, a comparison of global gene expression profiles between substate 1 and substate 2 cells was performed by DNA microarray analysis using Z-score (Supplementary Table S1 and Supplementary Fig. S1). We found 1764 and 1932 mRNAs showing significant up-or downregulation in substate 2 cells compared with that in substate 1 cells, respectively (Figs 5a and S1). After conversion of HGNC symbols to NCBI Entrez Gene IDs, 902 upregulated and 1016 downregulated mRNAs were obtained (Fig. 5a). We then performed gene ontology (GO) enrichment analysis using the NCBI Entrez Gene IDs (false-discovery rate [FDR] ≤ 0.05, fold enrichment ≥ 1.5), and seven neuron-related GO terms were specifically detected in biological process in the upregulated mRNAs, as expected ( Fig. 5a and Supplementary Tables S2-S3).
Next, we analysed pluripotency, epithelial-to-mesenchymal transition, and trilineage markers, as described in Fig. 5b and Supplementary Table S4. E-cadherin and vimentin were expressed in substate 1 and substate 2 colonies, respectively, as shown in Fig. 3. As expected, CDH1 (E-cadherin) and VIM (vimentin) were expressed at significantly higher levels in substate 1 and substate 2 cells, respectively (Fig. 5b). There were no significant differences in expression between substate 1 and substate 2 cells in pluripotent genes (POU5F1, NANOG, SOX2, LIN28A, and LIN28B), endoderm markers (SOX17 and AFP), mesoderm markers (T and ACTA2), ectoderm markers (OTX2 and TUBB3), and epithelial-to-mesenchymal transition markers, except for ZEB1 and ZEB2 (SNAI1, SNAI2, TWIST1, TWIST2, and CDH2; Fig. 5b). Importantly, expression levels of ZEB1 and ZEB2 were significantly higher in substate 2 cells than in substate 1 cells (Fig. 5b). The high expression of ZEB1 mRNA in substate 2 cells may promote ZEB1 protein expression in the nucleus (Fig. 3b). These results indicated that ectoderm-biased substate 2 cells were undifferentiated, but exhibited neural differentiation signals. Unfortunately, approximately 1000 biomarker candidates for ectoderm-biased pluripotent substate were detected. miRNA analysis showed that substate 2 cells exhibited neural differentiation signals, similar to mRNA analysis. Not only mRNA but also miRNAs show promise as biomarkers for various diseases and cell quality evaluation 12,16,18 . Moreover, miRNAs derived from serum-and cell culture medium-exosomes could facilitate the noninvasive assessment of many diseases and cell qualities [16][17][18] . To explore new miRNA biomarkers for the earliest stages of ectodermal differentiation, we compared comprehensive expression profiles between substate 1 and substate 2 using miRNAs for cells and exosomes (Supplementary Table S1). miRNAs that showed 2-fold differences in expression between substate 1 and substate 2 cells were analysed. Significance analysis showed that 43 and 301 cell-derived miRNAs were upregulated and downregulated, respectively, in substate 2 cells compared with those in substate 1 cells (Fig. 5c and Supplementary Tables S5-S6). Moreover, we also identified 96 and 76 exosome-derived miRNAs that were upregulated and downregulated, respectively, in substate 2 cells compared with those in substate 1 cells (Fig. 5d and Supplementary Tables S7-S8). miRNA target prediction was carried out by performing a keyword search in TargetScanHuman 7.2 (http:// www.targetscan.org/vert_72/) using miRNA symbols to predict up-and downregulated mRNAs from the downand upregulated miRNAs, respectively. The numbers of predicted up-and downregulated mRNAs targeted by more than 50% of the miRNAs were 447 and 148 or 493 and 94 in cells or exosomes, respectively, from substate 2 (Fig. 5c,d and Supplementary Tables S9-S12). GO enrichment analysis was performed using NCBI Entrez Gene IDs (Supplementary Tables S13-S16). Overall, 24 and 12 neuron-related GO terms were detected in biological processes and cellular components, respectively, for the predicted upregulated mRNAs derived from cell miRNAs ( Fig. 5c and Supplementary Table S14). GO enrichment analysis of the predicted upregulated mRNAs derived from exosome miRNAs also showed 33 and nine neuron-related GO terms in biological process and cellular component, respectively ( Fig. 5d and Supplementary Table S16). Five transforming growth factor (TGF)-β pathway-related GO terms were detected in biological process of the predicted downregulated mRNAs derived from cell miRNAs ( Fig. 5c and Supplementary Table S13). Notably, ectoderm is differentiated from human and mouse ES cells in the absence of a TGF-β signal 35 . Accordingly, these results indicated that substate 2, which tended to differentiate into neurons, was characterised by specific expression patterns of mRNAs and miRNAs. However, in each analysis of cellular and exosomal miRNA microarrays, it was difficult to limit the number of biomarker candidates for the earliest stages of ectodermal differentiation.

Integrated analyses of mRNA and miRNA microarrays of substate 1 and substate 2 cells.
The ectoderm-biased substate 2 cells and substate 1 cells could differentiate into the three germ layers; therefore, these cells may have a few differential gene expression patterns. Integration of mRNA and miRNA approaches may provide new candidates for highly reliable biomarkers 19,20 . Accordingly, we attempted to conduct integrated analysis of mRNA and miRNA, as shown in Fig. 6a,b. We first discovered 20 upregulated and 48 downregulated miRNAs common to cells and exosomes in substate 2 using NCBI Entrez Gene IDs because miRNAs present in both cells and exosomes are reliable in microarray analysis (Fig. 6c,d and Supplementary Tables S17-S18). Twelve of the 20 upregulated miRNAs (miR-105-5p, miR-125b-5p, miR-130b-5p, miR-135b-5p, miR-181a-5p, miR-181b-5p, miR-218-5p, miR-324-5p, miR-338-3p, miR-421, miR-592, and miR-1306-5p)  www.nature.com/scientificreports www.nature.com/scientificreports/ 20 upregulated and 48 downregulated miRNAs were converted to predicted downregulated and upregulated mRNAs, respectively, using TargetScanHuman 7.2. We then searched for mRNA biomarker candidates based on a combination of mRNAs and predicted mRNAs targeted by miRNAs using NCBI Entrez Gene IDs. Fourteen upregulated mRNAs (BTBD9, CADM1, COL9A2, DGKI, GBX2, KCNC1, KIF26B, MARCH1, PLXNA4, SLC24A2,  Supplementary Table S4). The red and blue coloured fold change (FC) numbers indicate significantly higher and lower expression in substate 2 than in substate 1, respectively. (c and d) miRNA microarray analysis of the cells (c) and exosomes (d). The miRNAs that showed a 2-fold difference in expression between substate 1 and substate 2 cells were subjected to analysis of significance using paired t-test with multiple testing correction (q < 0.05; Supplementary Tables S5-8). The miRNA target prediction was performed by Targetscan 7.2. The predicted up-and downregulated mRNAs were derived from the down-and upregulated miRNAs, respectively (see Supplementary Tables S9-S12). The symbols were converted to NCBI Entrez Gene IDs using BioMart based on GRCh38.p12 (see Supplementary Tables S9-S12). GO enrichment analysis was performed using the IDs of genes targeted by more than 50% miRNAs (false-discovery rate [FDR] ≤ 0.05, fold enrichment ≥1.5; see Supplementary Tables S13-S16).  Supplementary Table S18. (e) Venn diagram of upregulated mRNAs and predicted upregulated mRNAs targeted by downregulated miRNAs. The mRNAs in Supplementary Fig. S1, Supplementary Table S10, and Supplementary Table S12 were used. The upregulated mRNAs common to upregulated cell mRNAs and predicted upregulated mRNAs from miRNAs of cells and exosomes are shown in Supplementary Table S19. (f) Venn diagram of downregulated mRNAs and predicted downregulated mRNAs targeted by upregulated miRNAs. The mRNAs in Supplementary Fig. S1 www.nature.com/scientificreports www.nature.com/scientificreports/ KIF26B, MARCH1, PLXNA4, SLC24A2, SLC24A4, TLR4, WNT2B, and ZHX3) and one downregulated mRNA (PPP1R12B) using RNAs of substate 1 and substate 2 cells (Fig. 7). Nine of 14 upregulated mRNAs (COL9A2, DGKI, GBX2, KIF26B, MARCH1, PLXNA4, SLC24A4, TLR4, and ZHX3) were significantly upregulated in substate 2 cells compared with substate 1 cells. PPP1R12B expression tended to be higher in substate 1 cells than in substate 2 cells (P < 0.1). Allison et al. recently reported that endoderm-biased ES cells expressed higher levels of GATA6 10 , whereas few reports have described biomarkers that can be used to evaluate pluripotency during the earliest stages of differentiation. The nine upregulated genes were the most likely biomarker candidates for the earliest stages of ectodermal differentiation in pluripotent stem cells.

Conclusions
Precise definition of the status of pluripotent stem cells during culture using biomarkers is essential for basic research and regenerative medicine. In this study, we first established the ectoderm-biased substate in H9 human ES cells. The ectoderm-biased substate was characterised by low and high expression of the pluripotency marker R-10G epitope and the mesenchymal marker vimentin, respectively. The ectoderm-biased cells could differentiate into all three germ layers. Finally, we identified nine upregulated mRNAs in the earliest stages of ectodermal differentiation by performing integrated analyses of mRNA and miRNA microarrays and qPCR analysis. Few reports have described biomarkers that can be used to evaluate pluripotency during the earliest stages of differentiation and in the germ layer-biased substates of human pluripotent stem cells. We provided new biomarker candidates that could be used in the earliest stages of ectodermal differentiation. Our findings may supply clues to develop high quality biomarkers to evaluate pluripotency. Immunocytochemistry. Immunocytochemical analysis was performed as previously described 61  Fisher Scientific), and mouse anti-TUJ1 antibodies (IgG; monoclonal; 1:500; cat. no. MMS-435P; BioLegend, San Diego, CA, USA). The cells were incubated with primary antibodies diluted in 1% bovine serum albumin and 5% serum in phosphate-buffered saline (PBS) at 4 °C overnight. Secondary staining was performed with appropriate secondary antibodies, i.e., donkey anti-mouse IgG antibodies conjugated to Alexa Fluor 488 (IgG; polyclonal; 1:300 dilution; cat. no. A21202; Thermo Fisher Scientific), goat anti-mouse IgM antibodies conjugated to Alexa Fluor 488 (IgG; polyclonal; 1:300 dilution; cat. no. A21042; Thermo Fisher Scientific), and donkey anti-rabbit IgG antibodies conjugated to Alexa Fluor 594 (IgG; polyclonal; 1:300 dilution; cat. no. A21207; Thermo Fisher Scientific), for 1 h at room temperature. The samples were then counterstained with 4′,6-diamidino-2-phenylindole (DAPI; 1:1000 dilution; cat. no. D523; Dojindo, Kumamoto, Japan). The images were collected with a BIOREVO BZ-9000 fluorescence microscope (Keyence, Osaka, Japan). Based on bright-field and immunohistochemical observations, R-10G++ and vimentin− typical colonies and R-10G + and vimentin + colonies with wide intercellular spaces were designated as "substate 1" and "substate 2", respectively. Flow cytometry. Flow cytometry was performed as previously described 62 . Briefly, H9 cells were digested with Accutase (cat. no. SF006; Millipore, Burlington, MA, USA) for generation of single-cell suspensions. The cells were incubated in MACS buffer (0.5% bovine serum albumin and 2 mM ethylenediaminetetraacetic acid in PBS) with mouse anti-vimentin antibodies (IgG; monoclonal; 1:33 dilution; cat. no. H00007431-M08; Abnova, Taipei, Taiwan) and then incubated with Alexa Fluor 488 donkey anti-mouse IgG antibodies (IgG; polyclonal; 1:300 dilution; cat. no. A21202; Thermo Fisher Scientific). Normal mouse IgG (1:33 dilution; cat. no. NI03; Merck) was used as an isotype control. Flow cytometry data were acquired using a SH800Z (Sony, Tokyo, Japan). The data were analysed using FlowJo v10 software (BD Biosciences).  3). mRNAs that showed a 1.5-fold difference in expression between substate 1 and substate 2 cells were subjected to analysis of significance using unpaired two-tailed t-tests (*P < 0.1; **P < 0.05; ***P < 0.01). Red and blue gene symbols indicate upregulated and downregulated mRNAs, respectively.