It is now well documented that human embryonic stem cells (hESCs) can differentiate into functional cardiomyocytes. These cells constitute a promising source of material for use in drug development, toxicity testing, and regenerative medicine. To assess their utility as replacement or complement to existing models, extensive phenotypic characterization of the cells is required. In the present study, we used microarrays and analyzed the global transcription of hESC-derived cardiomyocyte clusters (CMCs) and determined similarities as well as differences compared with reference samples from fetal and adult heart tissue. In addition, we performed a focused analysis of the expression of cardiac ion channels and genes involved in the Ca2+-handling machinery, which in previous studies have been shown to be immature in stem cell-derived cardiomyocytes. Our results show that hESC-derived CMCs, on a global level, have a highly similar gene expression profile compared with human heart tissue, and their transcriptional phenotype was more similar to fetal than to adult heart. Despite the high similarity to heart tissue, a number of significantly differentially expressed genes were identified, providing some clues toward understanding the molecular difference between in vivo sourced tissue and stem cell derivatives generated in vitro. Interestingly, some of the cardiac-related ion channels and Ca2+-handling genes showed differential expression between the CMCs and heart tissues. These genes may represent candidates for future genetic engineering to create hESC-derived CMCs that better mimic the phenotype of the cardiomyocytes present in the adult human heart.
- human embryonic stem cells
- gene expression
- Ca2+ handling
- ion channels
human embryonic stem cells (hESCs) have gained widespread utility during the last decade due to their unlimited proliferation capacity and their ability to differentiate into virtually any functional cell type in the body (1). Many studies have now documented differentiation of hESCs into cardiomyocytes displaying a phenotype resembling, in many aspects, that of their in vivo counterparts (reviewed in Ref. 42). Due to the shortage of donor organs, hESCs constitute a promising source of human cardiomyocytes for various applications such as drug discovery, toxicity testing, and regenerative medicine. Independent investigators have reported high similarity in both morphology and expression of specific cardiac biomarkers, as well as in electrophysiology, when comparing hESC-derived cardiomyocytes to human cardiac tissue (4, 8, 14, 18, 44). In addition, results from recent studies on cardiomyocytes derived from human induced pluripotent stem cells (hiPSCs) have also shown high similarities to the data obtained from hESC-derived cardiomyocytes (14, 45).
Despite similar phenotypic characteristics with adult cardiomyocytes, some limitations in the functionality of hESC-derived cardiomyocytes have also been reported, particularly regarding expression of specific ion channels (18, 30, 31). Furthermore, studies on hESC-derived cardiomyocytes have shown that Ca2+ handling, which is crucial for excitation-contraction coupling, is functional but appears immature in hESC-derived cardiomyocytes (12, 15, 16, 24, 25, 33), and a number of important Ca2+-handling proteins are differentially expressed in hESC-derived, fetal, and adult cardiomyocytes. The underlying mechanism for the defective excitation-contraction machinery in hESC-derived cardiomyocytes has been suggested to be a dysfunctional sarcoplasmic reticulum Ca2+ release capacity (12). However, much more research is needed to better understand the regulation of the maturation process in hESC-derived cardiomyocytes. One step toward this aim is a thorough characterization of the mRNA expression of ion channels and Ca2+-handling genes in hESC-derived cardiomyocytes. Here we use microarrays to globally screen for differences and similarities, at the mRNA level, between hESC-derived cardiomyocyte clusters (CMCs) and heart tissue.
In our previous study (40), we identified sets of genes that were significantly up- or downregulated in a concordant manner across all cardiac-related samples to find reliable markers of hESC-derived CMCs for characterization purposes. In the present study, we use a different approach, by directly comparing hESC-derived CMCs to samples of fetal and adult heart (FH and AH, respectively) tissue, with the aim of exploring differences and similarities between in vitro-generated CMCs and their in vivo counterparts. Our results show a strikingly high global transcriptional similarity between the hESC-derived CMCs and the cardiac tissue samples. However, sets of genes that deviated significantly across these samples were also identified and analyzed regarding their putative functions. Interestingly, prolonged culture of hESC-derived CMCs increased the expression of a number of genes that are associated with Ca2+ handling and cardiac ion channels, functions that are reported to be deficient in immature cardiomyocytes. This indicates that in vitro culturing, in some aspects, promotes the maturation of hESC-derived CMCs. Taken together, our results provide a comprehensive overview of the transcriptional activity in both in vitro derived CMCs and human cardiac tissue samples and suggest genes that may need to be induced/repressed to increase the functionality of the hESC-derived CMCs.
MATERIALS AND METHODS
Differentiation of hESCs and Microarray Experiments
Differentiation of hESCs (cell line SA002, Cellartis, http://www.cellartis.com) to spontaneously contracting CMCs was performed as previously described, and the cells were harvested at two time points (3 and 7 wk after onset of differentiation) (39, 40). Microarray experiments were performed using whole transcript Gene ST 1.0 arrays (Affymetrix, http://www.affymetrix.com) run in triplicate as described previously (40). The following five samples were included: undifferentiated hESCs (UD), hESC-derived CMCs cultured for 3 wk after onset of differentiation (CMC3w), hESC-derived CMCs cultured for 7 wk after onset of differentiation (CMC7w), and, as reference material, RNA samples from FH and AH (Yorkshire Bioscience, http://www.york-bio.com) (Fig. 1). The human reference RNAs were obtained from different donors. All donors were healthy, and both male and female sexes are represented. The FH samples were obtained from donors of age 33–35 wk, and the AH samples were obtained from donors of age 21–29 yr.
Analysis of similarities in the global gene expression.
To assess the similarity of the hESC-derived CMCs to the human heart tissues, the similarities in gene expression across the samples were investigated. Since there is a lack of statistical methods for the identification of similarly expressed genes in pairs of samples, we defined criteria for assessment of similarity by first calculating the number of genes with fold change (FC) < 1.5 between pair-wise comparisons of the cell and tissue samples. This threshold was selected since it has been shown that the sensitivity of microarrays allows for a detection of fold changes >1.5 with 90% power using three technical replicates (35). Hence, the FC threshold to indicate similarity was set to <1.5. In addition, to exclude genes with high variability between the biological replicates, only genes with coefficient of variation (CV) <30% between the replicates were included in this analysis. We investigated the robustness of the combined criteria of FC and CV by also applying more stringent thresholds (i.e., FC <1.3 and CV <20%).
Identification of differences in the global gene expression.
To investigate in more detail which genes differ significantly between the hESC-derived CMCs and the heart tissue samples, we used the Significance Analysis of Microarray data (SAM) (41), which estimates the false discovery rate (FDR). Significantly up- or downregulated genes in the various samples were identified, and a combined criterion of FDR <0.05 and FC >2 was applied to select differentially expressed transcripts between pairwise samples. Direct comparisons were performed between CMC3w-CMC7w, CMC3w-FH, CMC7w-FH, CMC3w-AH, CMC7w-AH, and FH-AH.
Gene Ontology enrichment analysis.
To investigate the functional properties of genes that showed upregulation in the heart tissue compared with hESC-derived CMCs, a Gene Ontology (GO) (3) enrichment analysis was performed using the DAVID bioinformatics resource (10). Significantly overrepresented GO annotations for biological process (BP), molecular function (MF), and cellular component (CC) were identified. All enrichment calculations were performed as described before (39). The genes that were upregulated in FH compared with CMC7w were used as input to the enrichment analysis, and all genes represented on the arrays were used as the reference list. Significantly overrepresented annotations (P < 0.01) from all three categories at level 5 in the GO annotation hierarchy were identified using the hypergeometric test (10).
Analysis of cardiac ion channel gene expression.
In total 85 genes encoding for ion channels and their subunits (Table 1) were selected from the literature (18, 23, 30, 31). Their mRNA levels in CMC3w, CMC7w, FH, and AH were determined. To identify differentially expressed ion channels in any of the four samples, SAM was applied and genes with a q-value < 0.05 were identified.
Analysis of cardiac Ca2+ handling-related gene expression.
Eight genes (CASQ2, RYR2, TRDN, ITPR3, PLN, ASPH, SLC8A1, and ATPA2) previously reported to be involved in Ca2+ handling in the heart were identified from the literature (11, 12, 15, 16, 24). Using STRING (43), we further extended this list to include 47 additional functionally linked genes (Table 2). Importantly, only genes with experimentally verified interactions were included. SAM was then applied to determine which of these genes were differentially expressed in CMC3w, CMC7w, FH, and AH using a q-value < 0.05.
The successful differentiation of hESCs to CMCs in the material analyzed here has previously been confirmed (40), and markers of pluripotency (POU5F1, NANOG, DNMT3B, and LIN28) were downregulated in the CMC samples to background levels similar to those observed in FH and AH. Moreover, cardiomyocyte markers (MYH7, MYH6, TNNT2, and PLN) showed a substantial upregulation in the CMC samples to levels comparable to the heart tissues (40). In the present study, we have further investigated the transcriptional differences and similarities between CMC3w, CMC7w, FH, and AH. A more detailed analysis was also performed involving selected ion channels and Ca2+-handling genes of functional relevance for cardiomyocytes.
Similarity in Gene Expression Between CMCs and Heart Tissue
To examine to what extent the transcriptional profile in the CMCs resembles what is observed in FH and AH, the percentage of genes that display a high similarity across the different samples was determined (Fig. 2). We used a stringent similarity threshold of FC <1.5, and transcripts with a high variation within the biological replicates (CV >30%) were excluded from the calculation. In line with the results from our previous study (40), the largest fraction of similarly expressed transcripts (93%) was observed between CMC3w and CMC7w, and a slightly smaller fraction (89%) was observed between the FH and AH samples. The CMC samples were more similar to the FH samples than to the AH samples. Between CMC7w and FH 82% of the transcripts showed similar expression, and between CMC3w and FH 80% of the transcripts were expressed at comparable levels. When the CMC samples were compared with AH, again CMC7w showed the highest transcriptional similarity (78%), and a slightly smaller fraction of similarly expressed transcripts (74%) was observed between CMC3w and AH. As a reference point, UD cells were compared with AH, and here only 67% of the genes were expressed at similar levels. Intuitively, these samples would be the most different ones in this study. The global expression profiles of hESC-derived CMCs were also compared with publicly available gene expression data from human skeletal muscle (36), and this analysis showed that the similarity was 63% between CMC3w and skeletal muscle and 65% between CMC7w and skeletal muscle. We also increased the stringency in the criteria for the selection of similarly expressed transcripts to FC <1.3 and CV <20% (Fig. 2). This resulted in a decrease in the fraction of similarly expressed transcripts across all samples, but the trend as described above was maintained.
Differentially Expressed Genes in CMCs and Heart Tissue
Despite the similarities between the samples analyzed here, important differences were also observed. By applying the SAM statistical algorithm with FDR <0.05 in combination with an FC threshold >2, we identified transcripts that showed differences between the cell and tissue samples (Fig. 3). These results were in line with the similarity analysis, and only 13 downregulated genes in CMC7w and no upregulated genes were observed in the comparison to the CMC3w. Notably, the 13 downregulated genes in the CMC7w were not cardiac markers, but instead the majority of these genes are typically associated with endoderm derivatives (Supplemental Table S1).1 This suggests that during the in vitro culture of CMCs a progressive decline in the content of early endoderm occur in the CMCs. Between the FH and AH samples only one significantly upregulated and one downregulated gene were identified, using our stringent criteria for differential expression.
On the other hand, when comparing the CMCs with the tissue samples we observed more differences. The total numbers of differentially expressed genes between all pairwise comparisons performed are shown in Fig. 3, and the corresponding gene lists from this analysis are available as supplementary material (Supplemental Table S1). Among these, there is a large variety of genes illustrating the fact that the tissue samples are composed by a mixture of cells, including cardiomyocytes, endothelial cells, vascular smooth muscle cells, and cardiac fibroblasts, that all contribute to the global gene expression profile. The largest number of differentially expressed genes was identified when comparing CMC3w and AH, with 1,842 upregulated and 1,907 downregulated transcripts in AH compared with CMC3w. Taken together, these results indicate, on a global scale, that the hESC-derived CMCs are more similar to FH than AH (Fig. 3).
Enriched GO Annotations
To classify systematically the biological functionality of the genes that displayed a lower expression in CMC7w than in FH, we performed a GO enrichment analysis including all the three categories BP, MF, and CC (Fig. 4). It is interesting to note that several of the significantly overrepresented annotations were related to heart development and function. In addition, processes associated to blood vessel formation were also overrepresented. This is, however, not unexpected since the heart tissue is highly vascularized. Thus, the global gene expression profile of the tissue will be affected by the presence of vascular wall cells, which are not major constituents of the hESC-derived CMCs.
Gene Expression of Cardiac-related Ion Channels in CMCs and Cardiac Tissue
The gene expression of a selection of ion channels, known to be associated with cardiac function, were determined in CMC3w, CMC7w, FH, and AH. Using the SAM algorithm, genes for 85 ion channels and their subunits were explored. Eighty-two of these genes showed a q-value <0.05, and these are listed in Table 3. However, despite a significant q-value, most of these genes only showed small quantitative changes in expression values. Fourteen genes had an FC >3, and their expression profiles are shown in Fig. 5. In particular, it is interesting to note that KCNJ2, KCNIP2, KCNAB1, KCNJ8, and KCND3 appear to be expressed at close to base line levels (UD cells) in the hESC-derived CMCs, while in the heart tissue samples these genes are expressed at severalfold higher levels.
Expression of Genes Involved in Ca2+ Handling in CMCs and Cardiac Tissue
The expression levels of Ca2+-handling genes was, in a focused effort, explored in a similar way as for the ion channels above. Initially, we identified eight genes from the literature based on their association with Ca2+ handling in the heart. These genes were subsequently used as input to STRING to identify additional genes that directly, or indirectly, may be involved in Ca2+-handling processes. The protein interaction network (PIN) generated by STRING illustrates 47 additional genes (Fig. 6). SAM was used to detect Ca2+-handling genes that differed significantly in expression between the CMCs and tissue samples. Our approach identified 46 genes with a q-value <0.05 (Table 4). However, only two of the initial eight Ca2+-handling genes, RYR2 (Fig. 5) and CASQ2 (Fig. 7), showed an FC >3 between any two samples (UD excluded). Moreover, only seven of the 46 genes with interactions to the eight Ca2+-handling genes and with a q-value < 0.05 identified by SAM, showed an FC >3 (Figs. 6 and 7) between any pairs of the cardiac samples.
The capacity of hESCs to differentiate to cells of the cardiac lineage opens possibilities for development of unique cell-based assays. However, to fulfill the promise of this technology, it is important that the in vitro generated cardiomyocytes display a phenotype that highly resembles their in vivo counterparts. During recent years, several investigators have demonstrated useful functionality of human pluripotent stem cell (hPSC)-derived cardiomyocytes (7, 18, 26, 28, 29, 37). A number of studies, including work from our group, have used microarrays to investigate the global gene expression of in vitro derived cardiomyocytes (5, 8, 20, 38–40). However, only a few studies have included FH or AH tissue as reference samples (8, 14, 40, 44).For the first time, here we determined similarities in the global gene expression profiles between hESC-derived CMCs, FH, and AH. When exploring similarities in gene expression data including a series of samples, the obvious approach is to cluster the data into groups of genes with similar expression profiles (19). However, assessing similarity between pairs of single samples, where no profiles are available, is more demanding and has not, to our knowledge, been performed in any previous studies on hPSC-derived cardiomyocytes. We defined a similarity measure based on CV across the biological replicates and FC between mean values in each group and calculated the fraction of genes that displayed high reproducibility and small transcriptional differences between pairs of samples. When hESC-derived samples were compared with tissue samples, CMC7w and FH showed the highest similarity. This result is in line with a previous study, which suggests that in vitro differentiated cardiomyocytes have a higher similarity to FH than to AH (44).
Although we observed high transcriptional similarities between the four sample groups we also identified genes that showed significantly differential expression between pairs of samples (Fig. 3 and Supplemental Table S1). In concordance with results from our previous study (40) and with the results from the similarity analysis, we identified only a few differentially expressed genes between the pairs of samples that showed highest similarity (i.e., between CMC3w and CM7w as well as between FH and AH). The largest number of differentially expressed genes was observed between the CMC3w and AH samples, which were the cardiac samples that also demonstrated least similarity. To investigate how our data compare to results from other global transcriptional studies on hPSC-derived cardiomyocytes, conducted both in our group (39) as well as by other investigators (8, 14, 18, 44), we explored the overlap of enriched genes across studies. In a previous study, we identified 530 genes that were significantly more highly expressed in CMCs than in hESCs (39). In total 335 (63%) of these genes were also identified as upregulated in either CMC3w or CMC7w (or in both) in the present study. It should be noted that different types of Affymetrix GeneChips were used in these two studies, which may have implications on the results. In a study by Beqqali and coworkers (5), 15 genes were identified as cardiac enriched, and 14 (93%) of these genes are also significantly upregulated in CMCs compared with UD in our present study. Moreover, 83 cardiac-enriched genes reported by Gupta and coworkers (14) were investigated in our data; 72 (87%) of these genes were significantly upregulated in our CMC samples. A large overlap was also observed with data from Xu et al. (44), even though they applied a different experimental set-up and used another microarray platform. In their study, they identified 622 transcripts, representing 578 unique genes that were at least twofold upregulated in both hESC-derived cardiomyocytes and in FH and AH compared with hESCs. In total, 347 (60%) of these overlapped with our list of significantly upregulated genes in hESC-derived CMCs compared with UD. There are several possible explanations for the discrepancies observed among these studies ranging from biological differences (e.g., differences in cell lines, differentiation protocols, culture conditions etc.) to technical/analytical differences (e.g., inclusion/exclusion criteria, statistical methods, array platforms, etc.).
The differences between the CMCs and heart tissue with regard to global gene expression profiles are complex to analyze. It is obvious that the cellular compositions of the CMCs and the tissue samples are to some extent different, and this will complicate the interpretation of the data. Nevertheless, in an attempt to analyze systematically the difference between CMC7w and FH we performed GO enrichment analysis, which revealed several BPs related to heart and blood vessel formation. In addition, in the MF category the Ca2+-binding annotation was highly enriched, suggesting impaired Ca2+ handling in the hESC-derived CMCs.
Recent reports on functional studies of cardiomyocytes derived from hPSCs indicate that these cells have the potential to provide a cell-based alternative to animal experimentation (18, 28). However, the electrophysiological properties of hPSC-derived cardiomyocytes have been reported to be similar to the embryonic/fetal phenotype. This prompted us to analyze, in detail, the gene expression of cardiac-related ion channels known to be expressed in cardiac tissue and explore their expression in hESC-derived CMCs and compare that to FH and AH. Interestingly, our results demonstrate a strikingly high transcriptional similarity between hESC-derived CMCs and fetal and adult cardiac tissue samples for the majority of the 85 investigated ion channels. Nevertheless, 14 of these 85 genes displayed a deviating expression (FC >3) in hESC-derived CMCs compared with the cardiac tissue samples. As shown in Fig. 5, KCNJ2, KCNIP2, KCNAB1, KCNJ8, KCND3, CACNA2D1, and RYR2 demonstrate a significantly lower expression in hESC-derived CMCs compared with the tissue samples. On the other hand, the expression of KCNJ3, KCNJ5, CACNA1G, KCNH7, CACNA1D, and HCN1 is higher in CMCs. These results suggest that overexpression or inhibition of specific ion channels in the in vitro-derived CMCs may help to better mimic the transcriptional patterns observed in FH and AH tissue.
Our results regarding the ion channel expression are in line with data reported by Rust and coworkers (31). In their study, high expression of the nucleotide-gated K+ channels HCN1, HCN4, and the L-type Ca2+ channel CACNA1C was demonstrated. These ion channels are also highly expressed in immature cardiomyocytes in the murine heart (9). Moreover, Rust et al. (31) also report low expression of the voltage-gated Na+ channel SCN5A and voltage-gated inwardly rectifying K+ channels KCNJ2, KCNQ1, and KCND3, which are highly expressed in mature cardiomyocytes (17). Notably, our results show that KCNQ1 is higher expressed in CMC7w than in both FH and AH tissue. Moreover, the expression of KCNQ1 increased during in vitro culture and shows 2.6 times higher expression in CMC7w than in CMC3w (Table 3), indicating that prolonged culture of hESC-derived CMCs to some extent promotes the maturation process. A similar pattern is observed for SCN5A for which the expression clearly increased during in vitro culturing, and after 7 wk in culture the level is comparable with the levels of FH and AH (Fig. 5).
In addition to the early-stage electrophysiological properties of hESC-derived cardiomyocytes, hPSCs have also been reported to display an immature phenotype with regard to Ca2+ handling and limitations of the sarcoplasmic reticulum during heart contraction (12, 15, 16, 24, 25). As indicated above, our GO enrichment analysis also suggests impaired Ca2+ handling in the CMCs. To decipher the molecular status of in vitro derived CMCs we investigated in detail the expression pattern of a selection of Ca2+-handling key genes (ASPH, ATP2A2, CASQ2, ITPR3, PLN, RYR2, SLC8A1, and TRDN) in these cells. All eight genes showed significantly differential expression (q < 0.05) when analyzed with the SAM algorithm. However, despite a statistically significant dissimilarity, only two of them, CASQ2 (Fig. 7) and RYR2 (Fig. 5), displayed differences with FC > 3 when we compared CMCs to heart tissue. To further investigate the molecular machinery associated with cardiomyocyte contraction, we extended the panel of genes that are linked to Ca2+ handling, either directly or indirectly. By using STRING and the eight input genes we created a PIN and identified 47 gene products, with experimentally verified protein interactions with gene products from the input genes. Notably, only a few of these 47 interacting gene products showed significant transcriptional differences between the CMCs and one or both of the heart tissue samples, when we used a q-value < 0.05 and FC > 3. The seven genes that deviated most strongly were S100A1, PPP1R3A, CAV3, HRC, AKAP6, TRPC6, and SLN. All of these have previously been reported to be involved in Ca2+-handling processes and thus required for cardiac muscle excitation-contraction coupling (2, 6, 13, 21, 22, 27, 32, 34).
In conclusion, we have in the present study investigated the global transcription in hESC-derived CMCs and compared that to the gene expression in FH and AH. For the first time, the global transcriptional similarities between in vitro derived CMCs and FH and AH tissue have been investigated, and our results demonstrate that hESC-derived CMCs express most of the cardiac-specific markers at similar levels as in heart tissue. The CMCs show a slightly higher similarity to FH than to AH, although the differences between the two tissue samples were small. In particular, we also investigated the transcription of genes associated with cardiac-related ion channels and Ca2+ handling, which are groups of genes connected with functions that previously have been reported to be immature in hPSC-derived cardiomyocytes. Importantly, our results show that only a small group of genes involved in these functions have a deviating expression pattern compared with the heart tissues, and the majority of the known cardiac-specific genes exhibit a strikingly high similarity in hESC-derived CMCs and in heart tissues at the transcriptional level. Taken together, this work provides an important starting point for further functional studies aimed at improving the maturation processes of hPSC-derived CMCs.
This work was supported by Cellartis (Göteborg, Sweden), the EU-COLIPA funded project “SCR&Tox” (grant agreement no. 266753), and the University of Skövde, Sweden, under a grant from the Knowledge Foundation (2010/0069).
C. Améen and P. Sartipy are employed by Cellartis (http://www.cellartis.com). No competing financial interests exist for J. Synnergren and A. Jansson.
Author contributions: J.S., A.J., and P.S. conception and design of research; J.S. and C.A. performed experiments; J.S., C.A., and P.S. analyzed data; J.S. and P.S. interpreted results of experiments; J.S. prepared figures; J.S. drafted manuscript; J.S., C.A., A.J., and P.S. edited and revised manuscript; J.S., C.A., A.J., and P.S. approved final version of manuscript.
↵1 The online version of this article contains supplemental material.
- Copyright © 2012 the American Physiological Society