Transcriptional Differences between Rhesus Embryonic Stem Cells Generated from In Vitro and In Vivo Derived Embryos

Numerous studies have focused on the transcriptional signatures that underlie the maintenance of embryonic stem cell (ESC) pluripotency. However, it remains unclear whether ESC retain transcriptional aberrations seen in in vitro cultured embryos. Here we report the first global transcriptional profile comparison between ESC generated from either in vitro cultured or in vivo derived primate embryos by microarray analysis. Genes involved in pluripotency, oxygen regulation and the cell cycle were downregulated in rhesus ESC generated from in vitro cultured embryos (in vitro ESC). Significantly, several gene differences are similarly downregulated in preimplantation embryos cultured in vitro, which have been associated with long term developmental consequences and disease predisposition. This data indicates that prior to derivation, embryo quality may influence the molecular signature of ESC lines, and may differentially impact the physiology of cells prior to or following differentiation.


Introduction
Embryonic stem cells (ESC) derived from the inner cell mass (ICM) of preimplantation embryos have the potential to differentiate into any cell type of the three embryonic germ layers. ESC retain the ability to proliferate indefinitely, and maintain pluripotency through conserved regulatory networks; however require the provision of various extrinsic factors within the culture environment for continued growth and self-renewal capacity [1,2]. Loss of pluripotency results in changes in gene expression that include down-regulation of key pluripotency and repressive markers and the up-regulation of regulators of differentiation [3]. Recent studies have documented the transcriptional profiles of various embryonic stem cell lines [4][5][6][7], establishing a common stem cell regulatory program underlying pluripotency. However, ESC exhibit significant heterogeneity between and within lines, displaying differences in gene expression and differentiation capacity, as well as changes with increasing passage number and culture environment [8][9][10][11], largely attributed to adaptation with long term culture [12,13]. Significant differences have also been observed between human ESC lines attributed to differences in derivation techniques [14] and culture conditions [15][16][17]. Very little attention has been paid to other factors which may contribute to the overall normalcy of these cell lines, particularly the quality of the embryo from which a line is derived.
Preimplantation embryo development in vitro is associated with a number of perturbations in ultrastructure [18,19], gene expression [20][21][22][23][24][25] and post-transfer development [26][27][28][29][30], when compared with embryos derived in vivo. These differences likely underlie the significant variation between ESC lines. There is also considerable evidence that the environment to which the preimplantation embryo is exposed, particularly the in vitro culture environment, predisposes the resulting fetus to increased risk of adult onset diseases and imprinting disorders [28,[31][32][33][34][35][36]. Recently, Horii et al [37] reported retention of epigenetic differences in mouse ESC dependent on the in vivo or in vitro origin of the embryo from which they were derived. While ESC transcriptional profiles are known to differ from that of the ICM [38,39], these data raise the question as to whether ESC retain transcriptional memory of the embryos from which they were derived. Significantly, it is not clear whether current ESC models are similarly predisposed to developing disease characteristics post-transplantation, or whether they exhibit low levels of perturbation that are not easily distinguishable. To explore the hypothesis that differences exist between ESC derived from in vitro and in vivo embryos, gene expression profiles of rhesus macaque ESC generated from either in vitro cultured (Ormes series [40]) or in vivo derived (R series [41]) embryos were compared.

Results
Expression Profiling of rhesus ESC generated from in vitro or in vivo derived embryos The transcriptional profiles of undifferentiated ESC generated from either in vivo derived or in vitro produced rhesus embryos were compared using the Affymetrix GeneChip Rhesus Macaque Genome Array, enabling large scale gene expression profiling of 52,865 probe sets, representing over 20,000 genes. Initial data analysis using dChip software identified a total of 2537 transcripts as significantly different between in vitro and in vivo ESC, by a twofold or greater fold change (Table S2). Comparison between groups revealed 592 probe sets upregulated in rhesus ESC of in vitro origin. The reciprocal analysis identified 1945 probe sets upregulated in rhesus ESC of in vivo origin. Of the 2537, 1803 had known Entrez Gene IDs. As dChip is a model-based approach that only allows probe-level analysis, we undertook ChipInspector (Genomatix) analysis to assess differences at the level of each gene. ChipInspector identified a total of 3881 transcripts with differential expression of twofold or greater, of which 2706 were unique to the Genomatix analysis (Table S3), while 1175 transcripts overlapped with the dChip analysis. Of the 3881 transcripts, 560 genes were upregulated and 3321 were downregulated in in vitro ESC.
Further classification of the 3881 differentially expressed transcripts by biological function was undertaken using NetAffx (Affymetrix). Several significant (P,0.05) functional biological categories were represented including apoptosis, cell cycle, development and regulation of transcription ( Figure 1A). Of the 3321 downregulated genes and 560 upregulated genes, 797 and 129 were specific to in vitro ESC respectively ( Figure 1B). Hierarchical clustering demonstrated that gene expression profiles of in vivo ESC samples clustered together, separately from in vitro ESC samples ( Figure 1C), indicating that gene expression differences observed between in vivo and in vitro ESC were greater than differences within the experimental groups.
To identify functional relationships between transcripts, 3881 differentially expressed rhesus transcripts were uploaded into Bibliosphere (Genomatix) for literature based gene connection analysis. Bibliosphere identified 1388 transcripts significantly upor downregulated in rhesus ESC. Further analysis of the 1388 genes, identified 202 transcription factors (Table 1), and 40 significantly enriched pathways ( Table 2), involving a total of 544 genes.
Of the 202 transcription factors identified in Bibliosphere four known to be involved in the transcriptional control of pluripotency, POU5F1, Akt, SMAD2 and HIF1A, were further analyzed to establish literature based gene networks. The interactions of HIF1A and SMAD2 with other genes are presented in Figure 2. Regulatory mechanisms of the transcription factors HIF1A (Matrix family HIFF) and SMAD2 (Matrix family SMAD)'s were further studied as shown in Figure 2. The promoter regions of eleven genes were found to have HIFF binding sites. Likewise, the promoter regions of five genes contained SMAD binding sites.
Common framework, a pattern of transcription factor binding sites defined by a set of physical parameters such as order, distance, and strand orientation on the promoter region, is a promoter module that participates in transcription regulation in a certain context. The common frameworks were mined from the eleven genes' and five genes' promoter regions identified above. Frameworks CTCF-HIFF, ETSF-HIFF and SMAD-E2FF were identified in these two gene groups respectively and suggest that transcription factors CTCF and ETSF may work with HIFF, and that E2FF may work with SMAD, to regulate transcription ( Table  S4).

Expression of markers of pluripotency
Comparison of the 1388 significant differentially expressed genes with previous microarray data examining regulators of pluripotency [4][5][6]16,[42][43][44][45][46][47] identified 225 significantly different genes documented by at least one publication, with 68 of these genes documented by at least two or more publications ( Table 3). Among these genes FGF2 (basic FGF) and FGFR1 were significantly downregulated (2-fold) in in vitro ESC. Similarly, SOX2 expression was decreased more than 3-fold in in vitro ESC, while POU5F1 was reduced by 2-fold. Other genes, including those involved in transcriptional repression and TGFß signaling, were also identified. In particular TGFß1, FST, SMAD1, 4 and 5 and ID4 were downregulated in in vitro ES, while SMAD3 was upregulated (Table S3).

Differentially expressed genes correlate with differences observed in preimplantation embryos
Analysis was undertaken to determine whether ESC generated from in vitro cultured rhesus embryos displayed perturbations in gene expression reported in the literature as differentially expressed in in vitro and in vivo preimplantation embryos [19,23,26,28,31,[48][49][50][51][52], results of which are summarized in Table 4. These differences included significantly decreased expression of insulin-like growth factor receptor 1 and 2 (IGF-I, IGF-II), glucose transporters 3 and 5 (SLC2A3, SLCA2A5), activating transcription factor 1 (ATF1), cyclin D1, secreted phosphoprotein 1, and the antioxidant enzymes superoxide dismutase 1 (SOD1), peroxiredoxin 2 (PDX2) and glutathione peroxidase 4 (GPX4) was seen in in vitro ESC. Alterations in gene expression observed in mouse embryos as a result of the use of serum during embryo culture [52] were also detected, and included downregulation of platelet derived growth factor receptor (PDGFR), the metabolic genes pyruvate dehydrogenase isoenxyme 1, aldehyde dehydrogenase 2 (ADH2) and aldehyde dehydrogenase family 6 subfamily A1, and upregulation of solute carrier family 25 (mitochondrial carrier, citrate transporter) member 1.

Differential expression of oxygen-regulated and metabolic genes
Oxygen-regulated gene expression is known to be important for preimplantation embryo development [21]. The oxygen concentration in which the rhesus preimplantation embryo develops in vivo is reduced [53,54] [55], examining the effect of oxygen on preimplantation mouse embryos, resulted in the identification of an additional 23 genes that appear to be regulated by oxygen during early development [55] ( Table 6).
In addition to perturbed expression of metabolic genes previously reported in preimplantation embryos, including SLC2A1, SLC2A3, ALD2 and PDK1, regulatory genes controlling mitochondrial biogenesis were also identified as being downregulated in in vitro ESC, including mtSSB, POLG and TFAM, along with genes regulating mitochondrial dynamics (MFN1, KIF5C and OPA1; Table S3).

Confirmation of gene expression by RT-PCR
To confirm the fidelity of our results, we assessed the expression of 13 genes identified in the data analyses. Genes involved in metabolism and mitochondrial function (ATP5B, KIF5C, MFN1, PKM2, SLC2A3, UCP2), pluripotency (FGF2, POU5F1, SOX2, NANOG), transcriptional repression (PCGF2), aging (LMNA) and embryo development (FGF1R, IGF1R, IGFBP2) were examined in pooled ESC RNA from available cultures (Ormes 7 and R466) grown under the same conditions as the samples used for transcriptional profiling. Expression of these genes was confirmed   by RT-PCR, with all transcripts detected in both in vitro and in vivo ESC ( Figure S1).

Discussion
It is often overlooked that human ESC are generated from in vitro cultured, often surplus/'discard', embryos considered unsuitable for transfer in infertility clinics. While the classification of a good quality embryo is based largely on subjective criteria, it is well known that in vitro culture significantly perturbs embryo development, particularly in terms of gene expression, metabolism and subsequent development. With this in mind, we hypothesized that in vitro culture conditions would compromise gene expression in resulting ESC. To achieve this, we examined the transcriptional profiles of four different lines generated from in vivo derived embryos (R series) with that of four lines generated from in vitro derived embryos (Ormes series). Multiple passage numbers were analyzed to minimize passage related cell culture adaptation, with cells maintained under equivalent conditions known to support high quality ESC [56]. The data reported here represent selected passages between 8 and 37 for both in vitro and in vivo ESC. Transcriptional profiling of in vitro ESC and in vivo ESC identified a total of 3881 transcripts with twofold or greater differential expression, of which the majority were downregulated in in vitro ESC. Hierarchical clustering of ESC according to origin, irrespective of passage number, suggests that the differences in gene expression detected are stably maintained during long-term culture. It is important to consider that derivation of the R series (in vivo), and Ormes series (in vitro) carried out by different laboratories may contribute to some of the differences observed in the present study. However, as transcriptional profiles were compared over a range of early passage numbers, with all cell lines maintained under the same conditions by the same laboratory for each passage assessed, this contribution is likely to be minimal.
In vitro ESC and in vivo ESC differ in the expression of imprinted and cell cycle genes, a potential legacy of embryo culture Aberrant imprinting has been reported in a number of species following preimplantation embryo culture in vitro [57,58], including the rhesus macaque [59], with long-term consequences for fetal growth and adult health [29,33]. Bertolini et al [26] and Yaseen et al [60] have reported significantly decreased expression of IGF1R and IGF2R following in vitro culture of bovine embryos, conditions also associated with altered fetal and placental development and large offspring syndrome [27]. The expression of these genes was significantly lower in in vitro ESC when compared with in vivo ESC, suggesting that the altered expression of these genes in cultured embryos is preserved during ESC isolation. In support of this, a number of other genes involved in epigenetic regulation, including histones, histone deactylases and lysine-specific demethylase 3A were identified as differentially expressed between in vitro ESC and in vivo ESC (Table S3). Studies have also reported aberrations in imprinted genes in mouse [61], monkey [62,63] and human ESC [64][65][66][67], particularly that of IGF2 and IGF2R. Frost et al [68] reported genomic instability in human ESC, and suggested that derivation and ESC culture contributed to atypical methylation patterns, however it is possible that aberrant imprinting was inherent to the embryo from which the line was derived, in addition to any derivation and culture    The q-value is calculated as log2 fold change. doi:10.1371/journal.pone.0043239.t004 Table 5. Oxygen-regulated genes displaying differential expression between rhesus ESC generated from in vivo derived or in vitro cultured embryos compared with published data.  induced alterations. Significantly, epigenetic differences have been observed between mouse ESC generated from in vitro versus in vivo embryos [37], although these differences were lost by passage 5. Bioinformatic analysis of significantly different transcripts between in vitro and in vivo ESC also highlighted dysregulation of canonical pathways, particularly those regulating cyclins, cell cycle checkpoints and chromosomal stability ( Table 2), including genes involved in the G1 to S phase known to be important in ESC [69,70]. Mtango and Latham [71] have reported altered expression of cell cycle machinery in in vitro cultured rhesus embryos, suggesting that cell cycle control mechanisms may also be heritable from the embryo to resulting ESC. Misregulation of imprinted and cell cycle genes, previously documented following in vitro embryo culture, may therefore be preserved in resulting ESC, and may compromise the cells functionality during and/or following differentiation.  Table 6. Genes displaying differential expression between rhesus ESC generated from in vivo derived or in vitro cultured embryos and altered by oxygen in in vitro cultured preimplantation mouse embryos [55]. In vitro culture perturbs the expression of key pluripotency regulators Among the genes identified as significantly altered between ESC of different origin were known pluripotency markers, including POU5F1 (OCT4), basic FGF and SOX2. Basic FGF (FGF2) is an important component of primate ESC culture media required for propagation and colony maintenance. FGFs play several roles in vivo during early development [72] and are known to mediate IGF expression [73], representing a positive feedback loop. Sato et al [6] reported that FGF2 and FGFR1 were important genes enriched in the undifferentiated state, regulated by OCT4, SOX2 and NANOG. Activation of SMAD2/3 signaling is required for human ESC pluripotency [74] as both SMAD2/3 and FGF2 regulate NANOG gene expression. While NANOG is not significantly different between in vivo and in vitro ESC, in vitro ESC displayed significantly increased SMAD2 expression. Upregulation of SMAD2 may support ongoing culture in reduced levels of other pluripotency regulators. A reduction in the expression of OCT4 and SOX2, in addition to a reduction in FGF2 and FGF receptor expression, suggests that in vitro ESC may be more prone to spontaneous differentiation. Indeed, Byrne et al [55] reported significant variability in OCT4 expression across the same Ormes lines examined in the present study. Less than a two-fold difference in the level of OCT4 expression has been shown to have significant effects on ESC maintenance [75]. In support of this, Mtango et al [76] documented changes in pluripotency and differentiation marker expression during the early stages of rhesus macaque blastocyst outgrowth, and in Ormes 6 ESC, when compared with gene expression profiles of rhesus inner cell mass cells. Data therefore suggests that ESC derived from in vitro cultured embryos display alterations in pluripotency markers, however cells have potentially compensated by modulating other pathways to maintain self-renewal.
The effects of oxygen on in vitro cultured embryos are sustained in ESCs A significant difference between in vivo derived embryos and in vitro cultured embryos is the oxygen environment in which they develop. In vivo the oxygen concentration approximates 2-7% [52,53], with an oxygen concentration of 2% reported in rhesus macaque uteri, considerably lower than the atmospheric conditions commonly used for in vitro embryo culture, and lower than the 5% oxygen concentration used to generate the embryos from which the in vitro ESC were derived. The oxygen environment is known to alter blastocyst gene expression and embryo development [21,77]. Hypoxiainducible factors (HIFs) are oxygen-sensitive transcription factors that mediate cellular adaptation to reduced oxygen conditions. HIF1 protein levels increase exponentially at oxygen concentrations lower than 6% [78]. The response to hypoxia leads to the activation of signaling pathways involved in the regulation of mitochondrial function, glycolytic metabolism and cell survival. In the present study, HIF1 alpha was significantly reduced in in vitro ESC ( Table 1). Further analysis demonstrated enrichment (P = 0.0004) of HIF1 alpha regulated genes ( Table 5). Physiological oxygen concentrations also regulate human ESC pluripotency, proliferation, karyotypic stability and differentiation [15,[79][80][81][82], mediated by HIFs [83]. Consistent with our findings, significant differences in OCT4 levels [83,84] and SOX2 mRNA expression [83] have been reported in human ESC lines derived under 5% and 20% oxygen, or following transfer to reduced oxygen culture conditions. Significantly reduced expression of FGFR1 and FGFR2 [80] and SLC2A3, PKM2, ALDOC, and LGALS1 [17] have also been reported in human ESC in response to atmospheric oxygen conditions, and differences in SLC2A1, SLC2A3 and PGK1 have been reported between in vivo derived and in vitro produced rhesus macaque blastocysts [85]. These results suggest that underlying alterations in metabolism may exist. This is further supported by downregulation of regulatory genes controlling mitochondrial biogenesis and dynamics in in vitro ESC, including mtSSB, POLG and TFAM, as well as MFN1, KIF5C and OPA1 (Table S3). Differences in the expression of genes regulating mitochondrial biogenesis has also been reported between in vivo and in vitro rhesus blastocysts [86]. Significantly, Wale and Gardner [87] demonstrated that developmental perturbations observed following culture of preimplantation mouse embryo under atmospheric conditions were not restored by transferring cultures to a low oxygen environment, suggesting that adaptation of ESC will likewise not resolve underlying differences in ESC physiology. ESC properties may therefore be dependent on reduced oxygen conditions not only during derivation and subsequent expansion, but also during embryo culture prior to derivation.

Conclusions
Results of the present study document significant differences at the transcriptional level between embryonic stem cells derived from in vitro cultured embryos, and those derived from in vivo derived embryos. Data suggests that embryonic stem cells may retain a transcriptional memory representative of the environment of the preimplantation embryo from which the cells were derived. In vitro ESC exhibit transcriptional perturbations seen in in vitro cultured embryos, including alterations in markers of pluripotency and differences impacted by oxygen concentration. These differences may impact cell physiology, although it is unclear whether these differences will contribute to long-term functionality following ESC differentiation and transplantation. Further investigation into the differences between in vitro and in vivo ESCs, particularly in terms of imprinting, metabolism and functionality following differentiation, is warranted to ensure their therapeutic potential. Attention needs to be directed towards physiological measures of functionality, coupled with transcriptional, epigenetic and proteomic characterizations of pluripotency, to assess the impact the culture environment has throughout stem cell isolation, maintenance and differentiation. As methods become more refined and more efficient, and xeno-free isolation becomes routine, the examination of not only embryonic stem cells, but also induced pluripotent stem cells will be pivotal in establishing fundamental properties necessary to supply normal, safe and efficient cells for therapeutic translation.

Embryonic Stem Cell culture
Four rhesus (Macaca mulatta) ESC lines generated from in vitro cultured embryos cultured up to day 9 (Ormes 6, 7, 10 and 13, [40]; referred to as 'in vitro ESC') and four lines generated from in vivo derived embryos flushed from uteri 6 days post ovulation (Rseries 278, 366, 394 and 511, [41]; referred to as 'in vivo ESC') were cultured as previously described [56] and were generously provided by Dr Shoukhrat Mitalipov. Briefly, ESC were grown on mitotically inactivated mouse embryonic fibroblast feeder cells (MEF; cell line isolation was approved by the Oregon Health and Sciences University's Institutional Animal Care and Use Committee issued to S. Mitalipov) in Dulbecco's Modified Eagle Medium (DMEM/F12) (Invitrogen, Grand Island, NY) supplemented with 15% fetal bovine serum (FBS) (Hyclone, Logan, UT), 0.1 mM ß-mercaptoethanol, 1% nonessential amino acids (Invitrogen), 2 mM L-glutamine (Invitrogen), and 4 ng/ml FGF2 (Sigma), at 37uC under a 5% CO 2 -balance air atmosphere, and were passaged by manual scraping. To account for variability between derivation conditions, cultures were sampled from varying passage numbers (range  and cultures characterized to ensure that pluripotent ESC morphology, marker expression and karyotype were maintained. RNA extraction, microarray probe preparation and hybridisation ESC colonies were collected following manual removal of MEFs and careful dissection to ensure no feeder cell transfer prior to lysis. Total RNA was isolated from cultures for each respective ESC line using TRIZOL reagent (Invitrogen), followed by further purification with a RNeasy MinElute Cleanup Kit (Qiagen). The RNA samples were quantified using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE) and the quality of the RNA was assessed using Lab-on-a-Chip RNA Pico Chips and a 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). Samples with electropherograms showing a size distribution pattern predictive of acceptable microarray assay performance were considered to be of good quality. Twenty nanograms of total RNA from each line was amplified and labeled using a two-cycle cDNA synthesis and an in vitro transcription cRNA-RNA labeling system (GeneChip One-Cycle Target Labeling and Control Reagents; Affymetrix, Inc., Santa Clara, CA). Following successful cRNA amplification, 10 mg of labeled target cRNA was hybridized to Rhesus Macaque Genome Arrays (Affymetrix, Santa Clara, CA) using standard protocols, as described in the Affymetrix GeneChip Expression Analysis manual. Arrays were scanned using the GeneChip laser scanner (Affymetrix).

Bioinformatic analysis
All microarray data complies with MIAME guidelines, and all microarray information and individual cell intensity (CEL) files are available online at the Gene Expression Omnibus (GEO; GSE25198). Analysis of Affymetrix output files was performed with DNA-Chip Analyzer (dChip; Harvard School of Public Health, Boston, MA) and Genomatix (www.genomatix.de) software. In vivo ESC samples were used as the baseline for comparison. For dChip analysis, data normalization and model expression was undertaken using default dChip settings, with analysis of the False Discovery Rate (FDR) also performed. A gene was defined as significantly up-or down-regulated if the signal fold-change between the target samples was greater than 2, at a significance level of alpha = 0.05. For Genomatix data analysis, statistical significance of differential gene expression was assessed by computing a q-value (logarithm) for each gene. Genes were considered to be up-or down-regulated when the logarithm of the gene expression ratio was more than 1 or less than -1, that is, a 2fold or greater difference in expression, where alpha,0.05. Bibliosphere Pathway Edition (Genomatix), which combines literature analysis with genome annotation and promoter analysis, was used to create a directed regulatory network from transcripts identified by ChipInspector. To establish pathway and common framework information for significantly different transcripts, data was uploaded into GePS (www.genomatix.de). To further classify differentially expressed genes, Entrez gene IDs from the Genomatix analyses were used to search for over-represented biological processes against the rhesus and human genomes. Gene Ontology was performed using NetAffx (www.affymetrix.com).

RT-PCR validation
To validate the microarray results, RT-PCR was carried out on representative rhesus ESC samples (Ormes 7 in vitro and R475 in vivo) for 13 genes identified as significantly altered by the microarray analyses. RNA was extracted using an Absolutely RNA Nanoprep Kit (Stratagene, La Jolla, CA, USA), from which 1 mg was reverse transcribed into cDNA using SuperScript III reverse transcriptase (Invitrogen) and random primers (Applied Biosystems, Foster City, CA, USA) according to the manufacturer's instructions. Resulting cDNA was amplified with 1U Taq polymerase (Qiagen, Valencia, CA) in a final volume of 50 ml containing 16buffer, 1.5 mM MgCl2, 10 pmol of each sequence-specific primer and 10 mM of each dNTP. The mixture was amplified for 40 cycles in a BioRad DNA Engine thermal cycler (BioRad, Hercules, CA), where each cycle included denaturation at 94uC for 1 min, reannealing for 30 sec at 60uC, and primer extension at 72uC for 30 sec, followed by a final extension at 72uC for 7 min. PCR products were analyzed by electrophoresis through 2% agarose gels containing 0.5 mg/ml ethidium bromide and were photographed using a Kodak GL100 Imaging System equipped with Kodak Molecular Imaging software (Eastman Kodak Co., Rochester, NY). Primers were designed using Primer Express software (Applied Biosystems, Foster City, CA) and are listed in Table S1.