Comparison of gene expression between human and mouse iPSC-derived cardiomyocytes for stem cell therapies of cardiovascular defects via bioinformatic analysis

Abstract

Background

Preclinical studies have demonstrated the potential use of induced pluripotent stem cells (iPSCs) to treat cardiovascular disease (CVD). In vivo preclinical studies conducted on animal models (murine, porcine, guinea pig, etc.) have employed either syngeneic or human-derived iPSCs. However, no study has been carried out to investigate and report the key genetic differences between the human and animal-derived iPSCs. Our study analysed the gene expression profile and molecular pathway patterns underlying the differentiation of both human and mouse iPSCs to iPSC-cardiomyocytes (iPSC-CMs), and the differences between them via bioinformatic analysis.

Method

Data sets were downloaded from the Gene Expression Omnibus (GEO) database and included both human and mouse models, and the data for undifferentiated iPSCs and iPSC-CMs were isolated from each. Differentially expressed genes (DEGs) were screened and then analysed. The website g:Profiler was used to obtain the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment. Protein-protein interaction (PPI) networks of the DEGs were constructed using the Search Tool for the Retrieval of Interacting Genes (STRING) database and Cytoscape software. The subclusters were then extracted from the PPI network for further analysis.

Results

iPSC-derived cardiomyocytes expressed many genes related to vascular, endothelial, and smooth muscle repair in the human iPSC-CMs, and prevention of calcification in the mouse iPSC-CMs with clear differences in gene expression, which will affect how iPSCs act in research. Especially in the human iPSC-CMs, and also prevention of calcification processes in the mouse data. The identified differences in gene expression of iPSCs derived from the two species suggests that in vivo studies using mouse iPSC-CMs may not reflect those in humans.

Conclusion

The study provides new insights into the key genes related to the iPSCs, including genes related to angiogenesis, calcification, and striated muscle, endothelium, and bone formation. Moreover, the clear differences between both mouse and human-derived iPSCs have been identified, which could be used as new evidence and guidance for developing novel targeted therapy strategies to improve the therapeutic effects of iPSC treatment in cardiovascular defects.

Background

Cardiovascular diseases (CVDs) are the number one cause of mortality around the world with 31% of all global deaths in 2016 [1]. Several heart and blood vessel diseases are encompassed by this. Aortic stenosis is the most prevalent heart valve disease and the third most prevalent CVD in developed countries. It is characterised by progressive narrowing of the aortic valve due to fibrosis and calcification [2], producing stiffness and disruption of normal flow of blood through the valve. This leads to insufficient outflow from the ventricle, leading to left ventricular hypertrophy to compensate which eventually becomes maladaptive [3]. A brief review of valvular heart disease, including the current treatments and unmet clinical outcomes, is summarised in Table S1 in Supplementary Information. Therefore, there is a need to find alternative therapeutic options, with one possibility being stem cell therapy or cell-based tissue engineered heart valve (TEHV).

Different parts of the cardiovascular system have different anatomical arrangements, including different cells which make up their structure as well as important markers which are used to identify them (Table S2 and Fig. S1 in Supplementary Information). These are important to understand what cells are needed to form fully functional structures to treat CVDs. In order to produce the necessary cells, cell-based TEHV can help to generate biomimetic and viable tissue either before (in vitro) or after (in situ) implantation. Induced pluripotent stem cells (iPSCs) are a cell type that has been used in cell therapy and modelling of TEHVs and other CVD diseases (Table S3 in Supplementary Information). The iPSCs have several advantages over other cell types, especially in disease modelling. When iPSCs differentiate the genetic makeup of the donor cells remains, meaning there is the potential to reproduce essential aspects of genetic diseases in vitro using donor cells from patients. However, iPSCs for this purpose still carry limitations hindering their clinical translation. Firstly, the potential for cell survival and proliferation and thus any long-term benefits are poor. Secondly, iPSCs display a more fetal CM phenotype [4] which not only inaccurately replicates the electrophysiological responses but also produces a risk of arrhythmias on transplantation. Thirdly, enrichment and maintenance processes often produce a heterogenous population of iPSC-CMs [5] which exhibit phenotypic and genomic differences. Moreover, studies have shown that iPSC-CMs are associated with epigenetic pattern retention, producing differences in differentiation and physiological responses [6]. The potential of undifferentiated iPSCs also poses a tumorigenic risk, proving that these processes cannot account for the heterogeneity present [7].

The aim of this study was to conduct bioinformatic analysis for identifying key genetic factors and signalling pathways involved in the differentiation of iPSCs to cardiomyocytes (iPSC-CMs). We anticipate that identification of key genes and pathways will help lead to the understanding and optimisation of iPSC-based therapy for cardiovascular defects. Moreover, cells derived from different species have been widely used as replacements for studying human diseases. It has been shown that the key genes and the signaling pathways for a certain disease could vary between species, which is a common factor that confines much success to laboratories. Comparison is undoubtedly needed to minimise the inaccuracy when transforming achievements from the bench to clinical trials. We noticed the importance of this but found highly limited studies comparing at the gene level. Thus, we wish to compare the results from bioinformatic analyses between the human and mouse iPSC-CMs, and we hypothesise that the comparison will provide an understanding of iPSC-CM gene expression and guidance on potential different markers and therapeutic effects in in vivo mouse models and the human condition.

Database and method

Database searching

Gene Expression Omnibus (GEO) Database [8] is an international public archive with functional genomic data, and using this, two gene expression data sets which conducted transcriptomic profiling, GSE17579 [9, 10], and GSE18514 [11] were found and downloaded from the site. The data for undifferentiated iPSCs and iPSC-CMs were isolated from each data set. GSE17579 used a human model and was based on the GPL6947 (Illumina HumanHT-12 V3.0 expression beadchip), while GSE18514 used a murine model and was based on the GPL6997 platform (Illumina MouseWG-6 v2.0 expression beadchip). Ribonucleic acid (RNA) samples were prepared from three independent biological replicates for both undifferentiated iPSCs and iPSC-CMs for each dataset.

The human iPS cell line was derived by the data set GSE17579 from foreskin fibroblasts, and the human ES cell line HES-2 were differentiated to CMs using the END2 co-culture system. They were experimented on day 18 of their development. The data set GSE18514 developed the mouse iPSC cell line using a transgenic murine iPSC cell line “which expresses puromycin resistance protein N-acetyltransferase and EGFP under the control of the cardiomyocyte-specific α-myosin heavy chain promoter (alphaMHC-Puro-IRES-GFP, aPiG)” [11]. Murine aPIG-iPS and aPIG-ES cells differentiated into spontaneously beating CMs and were analysed on day 16 of their development.

Identification of DEGs

Network analyst [12] is a website used for gene expression analysis which was used to identify differentially expressed genes (DEGs) in the samples collected. The data was firstly normalised in Network Analyst using a log2 transformation, and an adjusted p-value of 0.01 and log2 fold change of 2.0 were used as thresholds to identify DEGs. Using the matrices downloaded from GEO, significant gene matrices of the DEGs were produced in Microsoft Excel. The DEGs were then converted into their equivalent official gene symbols, and their log2 fold changes calculated from their count values so heatmaps could be produced using a program called TBtools [13].

GO and KEGG pathway enrichment analysis of DEGs

g:Profiler [14] was used to perform pathway enrichment analysis of the DEGs using their official gene symbols. Gene ontology (GO) is a data source which focuses on the functions of genes and the GO terms are divided into molecular function (MF), biological processes (BP) and cellular component (CC). Moreover, Kyoto Encyclopedia of Genes and Genomes (KEGG) terms focus on the biological pathways of the DEGs. Together they can provide overall functional profiling of the DEGs.

The input parameter of either homo sapiens or mus musculus could be chosen depending on the data set analysed. g:Profiler has a few threshold algorithms to enrich terms, the one used was the g:SCS method to correct for the multiple testing problem. This occurs in pathway enrichment analysis of genes due to the high numbers of alternative hypotheses being tested at once (whether the GO and KEGG terms are significant), causing a higher chance of false positive results. Therefore, this algorithm was used as it provides a more accurate threshold between significance and non-significance [14]. A p-value threshold of < 0.05 was used to screen for the statistically significant results.

PPI network construction and functional subcluster analysis

Protein-protein interaction (PPI) networks were produced using a website called Search Tool for the Retrieval of Interacting Genes (STRING) [15] and a program called Cytoscape (Version 3.8.2) [16], a software for visualising molecular interactions and biological pathways. The PPI networks were created using STRING and imported into Cytoscape, where they were annotated and edited as appropriate. The add-on MCODE was also downloaded through Cytoscape, and this was used to produce subclusters of the densely connected parts of the networks. The genes in these subclusters also underwent pathway enrichment analysis to find their GO and KEGG terms.

Analysing the GO and KEGG terms and up- or down-regulation of genes

Using g:Profiler, the genes related to the GO and KEGG terms could be retrieved. To confirm if they were up- or down-regulated, this was done manually by locating these genes from the significant gene’s matrices, produced from the GEO matrices. Their log2 fold changes were then calculated to find if they were up- or down- regulated in the iPSC-CMs compared to the undifferentiated iPSCs. Using this data, they were then ranked to show which genes had the highest and lowest log2 fold changes, and which genes were up- or down-regulated.

Results

Identification of DEGs

The normalisation box plots after log2 normalisation for both the human and mouse data sets show that the data has been correctly normalised with an appropriate range of values for data analysis, as shown in Fig. 1.

Fig. 1

Normalisation box plots- Box plots of normalised data sets of iPSCs and iPSC-CMs produced after the data was normalised in Network Analyst using a log2 transformation. Black lines in the boxes represent the median values. a the human data set, GSE17579. b mouse data set, GSE18514

The heatmaps produced by TBtools are shown in Fig. 2. By undergoing differential expression analysis, in GSE17579 (human model) 471 DEGs were identified (Fig. 2a), and in GSE18514 (mouse model) 972 DEGs were identified (Fig. 2b). They show the DEGs which are significant at the threshold of a log2 fold change of 2.0. These verify that the undifferentiated iPSCs and the iPSC-CMs have different gene expression profile compared to one another for the DEGs. The volcano plots (Fig. 3) also show the non-significant genes and significant up- and down-regulated genes.

Fig. 2

Heatmaps- Heatmaps which show hierarchical clustering analysis results of the DEGs from both the mouse and human data matrices and show those genes which are significant at a threshold of a log2 fold change of 2.0. Each column represents a sample, and each row represents a DEG. Blue shows a lower value for gene expression while red shows a higher value for gene expression. a the human data set, GSE17579. b mouse data set, GSE18514

Fig. 3

Volcano plots: Volcano plots of all genes in the data sets, with each dot representing an individual gene. Blue dots show the genes which are downregulated, red dots show the genes which are upregulated, and the grey dots are genes which are non-significant. The X-axis is the log2-base fold change, and the Y-axis is the -log10-base adjusted P-value. a the human data set, GSE17579. b mouse data set, GSE18514

GO and KEGG pathway enrichment analysis of DEGs

Table 1 shows the top 10 ranked GO and KEGG terms for the DEGs, for both the human and mouse models, produced in the pathway enrichment analysis from g:Profiler. Looking at the human data set (Table 1a), enrichment analysis showed that: for MF, DEGs were most enriched in relation to extracellular matrix (ECM) and muscle structure, and binding of proteins, growth factors, signalling receptors, platelet-derived growth factor, actin, and glycosaminoglycans; for BP, DEGs were most enriched in relation to the development of the circulatory system, heart, and other tissues and organs, and developmental and system processes; for CC, DEGs were most enriched in relation to structural components of the ECM and muscle; for KEGG, DEGs were most enriched in relation to ECM-receptor interactions, focal adhesion, signalling pathways of TGF-beta, PI3K-Akt, regulation of pluripotency of stem cells, and cardiac muscle contraction.

Table 1 Top 10 ranked GO and KEGG terms for DEGs for a) human or b) mouse data sets

In comparison, pathway enrichment analysis in the mouse data set in Table 1b showed that: for MF, DEGs were most enriched in relation to exo-alpha- and alpha-sialidase activity and protein tyrosine/serine/threonine phosphatase activity; for BPs, DEGs were most enriched in relation to ganglioside catabolic, oligosaccharide catabolic, and nervous system processes; for CC, DEGs were enriched in relation to intermediate filaments and the intermediate filament cytoskeleton; for KEGG, DEGS were enriched in relation to the estrogen signaling pathway.

Table 2 shows a table of the shared DEGs between the two data sets, allowing them to be compared with one another. There are genes which are involved: in cardiomyocyte formation, including TMOD1, which encodes a member of the tropomodulin family and has roles in regulating the organisation of actin filaments; in angiogenesis, including VEGFC and VASH2 (as well as VEGFC being involved in endothelial formation); bone formation and remodelling, including FRZB and SPP1 respectively; oxidation of low-density lipoproteins, including OLR1; stem cell proliferation and renewal, including NANOG.

Table 2 The shared DEGs between the humans and mouse data sets

PPI network construction and functional subcluster analysis

Figure 4 shows the PPI networks of iPSC-CMs produced from Cytoscape, and Table 3 shows the top 10 hub genes present in each of the PPI networks. The PPI network produced from the human data set (Fig. 4a) is composed of 398 nodes and 2271 edges, and the PPI network from the mouse data set (Fig. 4b) is composed of 663 nodes and 1780 edges.

Fig. 4

The PPI networks of all the DEGs. They were edited in Cytoscape with continuous mapping so the colour and size of the node reflected the connection degree. This meant the redder, larger nodes represent a higher connection degree, while the bluer, smaller nodes represent a lower connection degree. For those DEGs, the maximum threshold for coexpression was set as 0.999 and the minimum threshold as 0.0, with values around 0.5019 coloured whiter. The top 10 nodes with the highest connection degrees are known as the hub genes in the PPI networks. a the human data set, GSE17579. b mouse data set, GSE18514

Table 3 Top 10 most significantly up-regulated genes (Hub Genes) from the PPI Networks of a) human or b) mouse data sets

The top 3 subclusters for both the human and mouse networks, which included hub genes, were retrieved, and can be seen in Fig. 5. For each subcluster, their top 5 genes with the highest coexpressions are shown in Table 4. Pathway enrichment analysis was also undertaken on the genes in these subclusters to find their GO and KEGG terms, and these can be seen in Table 5.

Fig. 5

Subclusters produced from the PPI network using MCODE. These show the top three most significant modules in which hub genes from the original PPI networks are present. The colour and size of the node reflects the connection degree, with the redder, larger nodes having a higher connection degree, while the bluer, smaller nodes having a lower connection degree. a Subclusters Ha (Human a), Hb (Human b), and Hc (Human c) from the human data set, GSE17579 (b) Subclusters Ma (Mouse a), Mb (Mouse b), and Mc (Mouse c) from the mouse data set, GSE18514. These show which subclusters and genes are related to which processes

Table 4 Top 5 Ranked genes in each subcluster from a) human or b) mouse data sets

Table 5 Pathway enrichment analysis to show the significantly enriched GO and KEGG terms for each subcluster from a) human or b) mouse data sets

The top 3 human subclusters contain: Human a (Ha)- 40 nodes and 285 edges, human b (Hb)- 26 nodes and 57 edges, human c (Hc)- 10 nodes and 19 edges (Fig. 6a). Pathway enrichment analysis in these subclusters showed that enriched genes were related to: Ha- ECM structure, muscle contraction, and myopathies; Hb- structural, signalling receptor and growth factor binding, developmental processes, platelets, diabetes pathway, and focal adhesion; Hc- protein ubiquitination, sarcomeres, myofibrils, and contractile fibres. The top 3 mouse subclusters contain: Mouse a (Ma)- 13 nodes and 78 edges; mouse b (Mb)- 14 nodes, and 44 edges, mouse c (Mc)- 55 nodes and 178 edges (Fig. 5b). Pathway enrichment analysis showed that the enrichments were related to: Ma- GPCR and transmembrane receptor activity, calcium and chemical ion homeostasis, and neuron/axon/ligand-receptor interactions; Mb- RNA/nucleic acids, RNA and mRNA processing and spliceosomes and RNA splicing; Mc- DNA repair, cell cycle and chromosomes.

Fig. 6

Bubble plot showing up- and down-regulation of genes- This shows the top 10 human and mouse hub genes, as well as the shared DEGs between the data sets, and their count values for both the undifferentiated iPSCs and the iPSC-CMs. A blue glow behind the dots represents down-regulation of the DEG, while an orange glow represents up-regulation of the DEG. The four shared DEGs which express opposite trends between the human and mouse data sets are labelled (TMOD1, FRZB, SH3RF2, NANOG). Note: For the top 10 mouse hub genes, SREK1 is not plotted because it did not have an equivalent Official Gene Symbol in the conversion table, so its count values were unable to be retrieved

Analysing the GO and KEGG terms and up- or down-regulation of genes

Tables were produced to show the top 5 up- and down-regulated genes for each GO and KEGG term in the networks, as well as each of the subclusters analysed, as shown in Table 6, Additional file 2 Table S4, and Additional file 3 Table S5 in the Supplementary Information. Table 6 summarises the full data from the Additional file 2 Table S4 and Additional file 3 Table S5 by showing most commonly up- and down-regulated genes from the GO and KEGG terms for the human and mouse data sets. All the software and websites used in this paper can be found in the Additional file 4 Table S6. Figure 6 also shows a summary of the top 10 human and mouse hub genes (Table 3), as well as the shared DEGs (Table 2), and their count values to show if they are up- or down-regulated.

Table 6 Top 5 most commonly up- and down-regulated genes for data sets

Discussion

Several key pathways and genes related to iPSC-CM development and various physiological processes, such as vascular regeneration and prevention of calcification, were identified. These provide guidance and hypotheses for the development of iPSC therapies, such as for aortic stenosis, where cells could not only regenerate endothelium and muscle cells, but also coded for proteins which helped to prevent calcification. Angiogenesis is one of the key challenges in tissue engineering [17], and there are many novel methods which are being used to try and increase the vascularisation in angiogenesis, including using stem cells. Studies by Geiger et al. showed that transfection of mesenchymal stem cells with VEGF-plasmid DNA increased vascularisation and resorption of bone substitute scaffolds, as well as leading to a more homogenous vascularisation, demonstrating the potential of angiogenic factors [18].

From the human data, there are several pathways relating to angiogenesis. Firstly, the PI3K-Akt pathway is involved in the development of blood vessels during both normal and tumour development, where it mediates various angiogenic factors, especially VEGF [19]. Looking at Table 6, we can also see that COL3A1 is a commonly up-regulated gene, and it is involved in the production of type III collagen, a major component of larger blood vessels [20]. Moreover, the results show upregulation of platelet-derived growth factor (PDGF), which is involved in angiogenesis, especially PDGF receptor beta [21, 22]. Moreover, VASH2, a potent angiogenic growth factor, is downregulated in the human model for the GO term actin binding. This may be attributed to its potential inhibition of angiogenesis along with its family member VASH1 reported in [23]. In unsurprised contrast, most of the upregulation of angiogenesis from VASH2 has been shown in tumours rather than normal blood vessel formation [24,25,26,27]. VEGFC is also upregulated in Table 2, which encodes for vascular endothelial growth factor C, a protein which promotes angiogenesis [28].

In terms of striated muscle, the human data shows upregulation of significant GO and KEGG terms in relation to the development of the circulatory system, heart, and general anatomy, and the formation of myofibrils, sarcomeres, and contractile fibers. Moreover, key genes are upregulated such as MYOM1, MYH6 and MYH7, as shown in Table 6, with MYH7 being a key marker for ventricular cardiomyocytes (Table S2). We can identify the cells as cardiomyocytes due to the marker TNNT2 which is upregulated in the human iPSCs-CMs, and in contrast, has a low expression in smooth muscle cells [29]. Upregulation of genes related to the formation of smooth muscle or endothelial development are also important because these are key parts of vascular structure. By comparing key genes involved in smooth muscle development with the human data, one of the most upregulated genes, LUM, is also expressed in aortic smooth muscle cells [30]. Here it has a key role in regulating collagen matrix assembly, as well as aiding in cell migration and proliferation. Other related genes found in this study include ACTA2, which encodes α-smooth muscle actin, involved in smooth muscle contractility and blood pressure homeostasis [31, 32]. This is a known marker of smooth muscle cells (Table S2), indicating the possibility of the presence of smooth muscle cells. MYH11 which codes for myosin in smooth muscle cells [33], MYOM1 which codes for the protein titin, and the aforementioned protein COL3A1 [29], are all shown to be upregulated in the data, and are all also present in smooth muscle cells.

There are fewer genes in the human data related to endothelium formation, apart from the only notable one being VEGFC, which is also involved in endothelial cell growth [28]. In comparison, looking at the top 10 hub genes in the mouse data, TNNC2 (which encodes troponin, a key protein in striated muscle contraction), is the only hub gene to show specific expression to cardiomyocytes. The other hub genes show several different pathways expressed in the GO and KEGG terms, such as RNA processing and splicing, as well as chromosome, ribosome, and cytoskeletal formation, showing less gene expression in terms of angiogenesis, smooth muscle, and endothelial formation. However, several terms can still be linked to cardiomyocytes. Firstly, the estrogen pathway is upregulated, which is known to promote cardiac regeneration and have cardioprotective effects [34]. Moreover, the terms Cellular calcium ion homeostasis and Cellular chemical homeostasis show that there are processes regulating calcium and transmitting calcium ion currents throughout the mouse iPSC-CMs. These results reflect that the mouse iPSC-CMs have more pathways related to the prevention of calcification compared to human iPSC-CMs. From the human data, the KEGG term TGF-beta signaling pathway is upregulated, which may relate to calcification and aortic stenosis. The TGF-beta signaling pathway is involved in elongating glycosaminoglycan (GAG) chains, which causes lipoproteins to accumulate. Lipids will then take part in the calcification process by being modified by enzymes into various lipid-derived compounds [2].

Comparing the shared DEGs in Table 2, most of them are expressed with similar trends, except for TMOD1, FRZB and NANOG. TMOD1 is gene coding for the protein tropomodulin 1, involved in capping the end of actin filaments in sarcomeres in the mouse. The CVDs associated with TMOD1 include hypertrophic and dilated cardiomyopathy [35, 36]. Opposite trends of TMOD1 in both iPSCs and iPSC-CMs are detected, which is downregulated in the mouse iPSC-CM but upregulated in human’s, and vice versa for undifferentiated iPSCs. Moreover, expression of the KEGG terms Hypertrophic cardiomyopathy and Dilated cardiomyopathy in subcluster Ha shows that upregulation of TMOD1 might be associated with a higher rate of CVDs in the human iPSC-CMs. This may give a warning that potential conflicting results could have been obtained using TMOD1 as a marker in mice and humans.

CVDs such as aortic stenosis are also associated with the “osteoblastic transition of cardiac valve interstitial cells” [3], so it is also important to look at genes related to bone formation. FRZB is a gene whose overexpression is linked to osteogenesis. Looking at Table 2, it is upregulated in the human data set but downregulated in the mouse data, showing that perhaps the human cells are more prone to osteogenesis in the aortic stenosis process [37]. However, for both data sets, SPP1 is upregulated, which is related to bone remodelling [38], with the human data set having a higher up-regulation. Overall, both genes show the human iPSC-CM’s networks are associated with increased bone remodeling compared to the mouse iPSC-CM’s.

While the human data downregulates NANOG for stem cell proliferation and renewal, the mouse data has upregulation, which could be a reflection on their relative times in culture, which will be discussed in the limitations section.

In comparison, the genes upregulated in angiogenesis and endothelial formation show consistent trends in both mouse and human. Both iPSC-CMs stimulate these processes, as shown by upregulation in both data sets for VEGFC and downregulation in both for VASH2, showing similarities between the data sets in Table 3, in agreement with previous studies in literature [23,24,25,26,27,28].

For both data sets there is upregulation of OLR1. OLR1 is a risk factor for coronary artery disease (CAD) because it encodes lectin-like oxidised low-density lipoprotein receptor-1 (LOX-1), which increases the absorption and degradation of oxidised low-density lipoproteins (ox-LDLs) in cells [39, 40]. Therefore, both the human and mouse data sets can be linked to also being involved in calcification and aortic stenosis since many studies show a link between higher levels of ox-LDL and fibro-calcific modelling in aortic valves [41, 42].

There were a couple of limitations with this study, the first being that each data set downloaded had limited biological replicates, with only 3 each for both the undifferentiated iPSCs and differentiated iPSC-CMs, so there is limited in vitro data. Moreover, there was no in vivo data or data from disease states, which means that we would need further data to verify conclusions on these topics.

Moreover, since the experiments were conducted by two separate groups, there were differences in the methodology between the groups. For example, the human iPSC-CMs were produced via co-culture with the murine visceral endoderm-like cell line END2 [9, 10], while the mouse iPSC-CMs were produced via introducing a transgene into the murine iPSCs. Furthermore, slight differences such as the fact that the genetic expression in mouse iPSC-CMs was tested on day 16, while the human iPSC-CMs were tested on day 18, could be the reason why the shared DEG NANOG was more highly expressed in the mouse iPSC-CMs, since it could reflect that they had a more immature phenotype compared to the human iPSC-CMs.

Another issue in our study includes the varying efficacy of differentiation of the iPSCs. Other methods could be investigated so that a higher proportion of the iPSCs can be successfully differentiated into the desired cells. Several biophysical factors and the concentration, timing and the mixture of factors/molecules will affect the efficiency of the differentiation pathways. Our data provides a starting point in helping to identify the genetic makeup of the iPSC-CMs, and by looking at how this can affect their behavior and how they differentiate into the desired cells, this could be utilised in this further research. This variation in differentiation is shown by the fact that the murine iPSC-CMs display fewer myosin related genes, and genes related to things such as sarcomere assembly or muscle function. Typically, iPSC-CMs are cultured for 35-40 days post differentiation to allow for maturation into CMs before any studies are done, so the biological and physiological differences shown in the data could have been due to the mouse and human iPSC-CMs being at different points of maturation than normal. Other possible causes of this could have been due to the differences in the species themselves, any possible epigenetic pattern retention, or the different methods of their differentiation which could have led to the cells going down different developmental routes.

In our future work something that would be pertinent firstly would be to confirm the results of the bioinformatics analysis using qPCR or western blotting to look at the gene expression profiles of the iPSC and iPSC-CMs. Moreover, following on from this we can then start to investigate specific systems discussed above, such as confirming the mechanisms and genes involved in angiogenesis, calcification, sarcomere assembly and muscle function etc. and be able to see the complex dynamic systems we have only partly touched on so far. For example, the mouse iPSC-CMs have more pathways in the data related to calcification prevention than human iPSC-CMs, but due to the complex regulatory systems involved in this, things such as the ion channels involved would need to be clarified using experiments such as qPCR. In addition, comparing the iPSC-CMs of mouse and humans to primary cardiomyocytes from the respective organisms is also valable to see how genetically similar they are.

Further research about the cell delivery to diseased tissues in vivo is also important. As shown in Table S3, there are studies investigating the way these cells can actually be delivered/applied to diseased tissues, for instance using heart patches [43]. In future research, we can use iPSCs to produce an iPSC-loaded heart-valve scaffold, and then investigate how this can help to treat CVDs. We can investigate the effect of the scaffold on the delivery and survival of the cells, and the efficacy of treatment in vivo. This allows us to further investigate how iPSC-CMs can be more efficiently utilised in cardiovascular diseases, using our current study as a foundation for how iPSC-CMs may differ in their genetic makeup, and thus their behavior in research.

Conclusion

This study analysed the gene expression profiles between human and mouse iPSC-derived cardiomyocytes. DEGs were identified, pathway enrichment analysis was performed to find GO and KEGG terms, and PPI networks and subclusters were constructed and analysed. The results show that iPSC-derived cardiomyocytes could be used as a potential therapeutic for CVD, with the expression of key cardiomyocyte genes. They also show potential in vascular, endothelial, and smooth muscle repair, especially from the human iPSC-CMs, and prevention of calcification in the mouse iPSC-CMs, due to the genes expressed. It also indicates how in vivo studies using mouse iPSC-CMs may not reflect those in humans due to the clear differences in gene expression between the species, which affects how iPSCs will act in research. This is the first study which has shown comprehensive information on the genetic differences between human and mouse iPSC-CMs and highlighting key areas where they may be used for design and validation of iPSC therapeutics.