The LARGE Principle of Cellular Reprogramming: Lost, Acquired and Retained Gene Expression in Foreskin and Amniotic Fluid-Derived Human iPS Cells

Human amniotic fluid cells (AFCs) are routinely obtained for prenatal diagnostics procedures. Recently, it has been illustrated that these cells may also serve as a valuable model system to study developmental processes and for application in regenerative therapies. Cellular reprogramming is a means of assigning greater value to primary AFCs by inducing self-renewal and pluripotency and, thus, bypassing senescence. Here, we report the generation and characterization of human amniotic fluid-derived induced pluripotent stem cells (AFiPSCs) and demonstrate their ability to differentiate into the trophoblast lineage after stimulation with BMP2/BMP4. We further carried out comparative transcriptome analyses of primary human AFCs, AFiPSCs, fibroblast-derived iPSCs (FiPSCs) and embryonic stem cells (ESCs). This revealed that the expression of key senescence-associated genes are down-regulated upon the induction of pluripotency in primary AFCs (AFiPSCs). By defining distinct and overlapping gene expression patterns and deriving the LARGE (Lost, Acquired and Retained Gene Expression) Principle of Cellular Reprogramming, we could further highlight that AFiPSCs, FiPSCs and ESCs share a core self-renewal gene regulatory network driven by OCT4, SOX2 and NANOG. Nevertheless, these cell types are marked by distinct gene expression signatures. For example, expression of the transcription factors, SIX6, EGR2, PKNOX2, HOXD4, HOXD10, DLX5 and RAXL1, known to regulate developmental processes, are retained in AFiPSCs and FiPSCs. Surprisingly, expression of the self-renewal-associated gene PRDM14 or the developmental processes-regulating genes WNT3A and GSC are restricted to ESCs. Implications of this, with respect to the stability of the undifferentiated state and long-term differentiation potential of iPSCs, warrant further studies.


Introduction
Human amniotic fluid cells (AFCs) represent a heterogeneous mixture of cells originating from different fetal tissues. They have been used for prenatal diagnosis of various congenital fetal abnormalities for more than fifty years [1]. Yet, especially within the last decade, molecular biology-based studies have revealed remarkable features of distinct subpopulations within bulk AFCs. For instance, in 1999, activity of the telomere-elongating enzyme telomerase was detected in young AFCs, with decreasing activity in aged AFCs [2]. Later on, the presence of cells exhibiting certain embryonic stem cell (ESC) features among bulk primary AFCs was reported [3,4]. Other groups have demonstrated the existence of mesenchymal stem cells (MSCs) within the amniotic fluid [5]. Based on these observations, several strategies have been developed to sort stem cell-like populations out of bulk AFCs and different subpopulations have been characterized in more detail [6][7][8][9][10]. Multipotent properties [6,7,11,12] and potential immune-privileged characteristics of particular AFCs [13,14] support the idea of utilizing amniotic fluid as a source of fetal stem cells, with feasible application in regenerative medicine, especially in fetal tissue engineering approaches [13]. However, there are drawbacks associated with the use of AFCs for such purposes. For instance, primary cultures of bulk AFCs, like primary cell lines in general, senesce after prolonged culture periods in vitro. Besides, the fact that AFCs do not form teratomas in vivo [6,15,16] implies that not even the stem cell-like cells within the amniotic fluid are bona fide pluripotent cells. Hence, their ability to form complex, mature differentiated cell types may be restricted. Indeed, the capacity of AFCs to form specialized cell types and to contribute to the formation of certain tissues or organs in vitro and in xenotransplantation experiments in vivo is a subject of debate [6,11,[17][18][19][20][21][22][23]. We believe that a means of assigning amniotic fluid cells greater value as an in vitro model system to investigate developmental processes, to conduct disease modeling, toxicological studies, drug research and exploitation in regenera-tive medicine could be achieved by cellular reprogramming of these cells to an undifferentiated ground state. As a result, AFCs acquire the ability to self-renew and become pluripotent.
During the course of this study, the generation of iPSCs from human AFCs were described [15,29,30]. However, these studies failed to characterize amniotic fluid-derived iPSCs beyond the standard assays required to confirm induced pluripotency. Yet, for potential application of iPSCs in basic and applied research, various fundamental aspects of iPSCs, in general, and of this new AF-derived iPSC type, in particular, remain to be understood. Our study aimed at a more detailed molecular characterization of AFiPSCs. To this end, we generated AFiPSCs and demonstrated their ability to differentiate into the extraembryonic trophoblast lineage. This study also highlights the potential of cellular reprogramming to avert replicative senescence observed in bulk primary AFCs. Furthermore, we have analyzed similarities and differences between AFiPSCs, ESC lines H1 and H9 and fibroblast-derived iPSCs (FiPSCs) on the basis of global gene expression. We discuss a fundamental principle of cellular reprogramming, which we have coined LARGE, the Lost, Acquired and Retained Gene Expression principle. This refers to specific genes, which are either switched off, activated or which remain expressed upon induction of pluripotency. In this context, we demonstrate the activation of a common self-renewal and pluripotency-associated gene regulatory network upon cellular reprogramming. Furthermore, we highlight putative implications of the loss of distinct donor cell signature genes and the activation and/or retention of genes implicated in development processes upon cellular reprogramming.

Ethics Statement
Auxiliary samples of human AFCs obtained during routine amniocentesis were kindly donated by the clinical laboratory of Prof. Dr. Wegner/PD Dr. Stumm (Zentrum für Prä nataldiagnostik, Kudamm-199, Berlin, Germany) after written informed consent. Utilization of these cells was approved by the ethics commission of the Charité Universitä tsmedizin Berlin.

Retroviral production and iPSC generation
OCT4, SOX2, KLF4 and c-MYC retroviruses were generated using pMX vectors as described previously [24]. Briefly, 7.5*10(6) Phoenix Ampho Cells were seeded onto gelatin-coated T75 cell culture flasks and grown in DMEM (Invitrogen) supplemented with 10% FBS (Biochrome, Berlin, Germany, www.biochrom.de) for 16 h. The cells in one flask were then transfected with 12 mg of one of the retroviral DNA vectors encoding either OCT4, SOX2, KLF4 or c-MYC using the FuGENE HD transfection reagent (Roche, Basel, Switzerland, www.roche.ch) according to the manufacturer's instructions. The retrovirus-containing medium was harvested 48 and 72 h post-transfection. For the generation of AFiPSCs, 180,000 AFCs were transduced with a cocktail of the respective retrovirus-containing media, supplemented with 4 mg/ ml Polybrene (Sigma-Aldrich, Munich, Germany, www.sigmaaldrich.com) at a rate of 1.25 or 2.5 MOI on days 1 and 2 after plating. Each time, directly after the addition of retroviruses, the plates were centrifuged at 8006g, at 37uC for 99 min before replacement of the infectious medium by fresh medium (DMEM/ 10% FBS). The next day, the infected cells were plated onto irradiated MEFs on Matrigel-coated dishes in DMEM/10% FBS. Another 24 h later, the medium was switched to ESC medium [31] for a total period of 10 d, with replacement on alternate days. Afterwards, the infected cells were grown in mouse embryonic fibroblast-conditioned medium (MEF-CM) [32], which was changed at an interval of 2 d until reprogrammed AFiPSC colonies were manually picked 24 d post-transduction and expanded under ESC conditions. Currently, we have AF-derived iPSC lines 4, 5, 6, 10, and 41 in culture passaged more than 25 times (P25).
The generation of FiPSCs used for the comparative transcriptome analysis has been described [31].

In vitro and in vivo differentiation of AFiPSCs
For in vitro differentiation, embryoid body (EB) formation of AFiPSC lines 4, 5 and 41 was induced in ESC medium without bFGF supplementation using the hanging-drop method [33]. After 2 to 3 d, EBs were placed onto low-attachment dishes. A week later, EBs were plated onto gelatin-coated dishes, allowed to differentiate for an additional 10 to 14 d and then stained. In vivo differentiation experiments were performed by EPO-Berlin GmbH (Germany, www.epo-berlin.de). Basically, approximately 2*10(6) cells of the AFiPSC lines 4 and 41 were collected by type IV collagenase-treatment or 0.05% Trypsin/EDTA-treatment, washed, pooled and injected s.c. into NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ mice, commonly known as NOD scid gamma (NSG). Teratomas were collected approximately 63 d after injection and processed according to standard procedures for paraffin embedding and hematoxylin and eosin staining. Histological analysis was performed by a pathologist.

Trophoblast differentiation of AFiPSCs
To induce differentiation into the trophoblast lineage, AFiPSC lines 5 and 41 were transferred onto Matrigel-coated cell culture dishes and grown in MEF-CM including 8 ng/ml bFGF (PeproTech, Rocky Hill, NJ, USA, www.peprotech.com) until they attained about 30 to 50% confluency. At this point, medium was changed to defined N2B27 medium (Invitrogen) lacking bFGF but including either 100 ng/ml BMP2 (PeproTech) or BMP4 (R&D Systems, Minneapolis, MN, USA, www.rndsystems.com) for a period of five days or a combination of 10 ng/ml BMP4 and 10 mM SB431542 (a TGFbRI inhibitor, Sigma-Aldrich) for 7 days. Undifferentiated controls were grown in N2B27 including 20 ng/ml bFGF only. After a period of 5 d or 7 d, including daily replacement of media, the cells were harvested for RNA isolation for qRT-PCR and global gene expression profiling analyses or fixed for immunofluorescence microscopy analysis.

DNA fingerprinting and karyotyping
The origin of AFiPSC cell lines 4, 5, 6, 10, and 41 was confirmed by fingerprinting analysis, as previously described [34]. The primer pairs D17S1290 and D21S2055 were used; sequences are provided in Table S1. For the detection of probable karyotypic abnormalities in AFiPSC lines 4, 5, 6, and 41, chromosomal analysis was performed after GTG-banding at the Human Genetic Center of Berlin. For each line, 25 metaphases were counted and 10 karyograms analyzed.

Illumina bead chip hybridization and data analyses
Total RNA was isolated using the RNeasy Mini Kit (Qiagen, Germantown, MD, USA, www.qiagen.com). In each case, 500 ng RNA were used as input for the bead chip hybridization (Illumina, San Diego, CA, USA, www.illumina.com). Processing of samples and the conversion of raw data was previously described [31]. For correlation coefficient analysis and the generation of Venn diagrams, detected gene expression was defined by a Detection P-Value ,0.01. To be considered as differentially expressed, genes had to be at least 1.5 fold up-or down-regulated in a group-wise comparison of all AFiPSC or FiPSC lines with either AFCs at passage 6 or 17 or fibroblasts (Fibs), respectively. Accordingly, the FDR-adjusted P-Value for differential gene expression had to be ,0.05. Human senescence-associated genes were derived using the AmiGO Browser version 1.7 of the Gene Ontology database (http://www.geneontology.org, 28 th of March 2010) [35]. Functional annotation and enrichment analyses were done using the DAVID platform version 6.7 (http://david.abcc.ncifcrf.gov/ home.jsp) [36,37]. Illumina ProbeIDs were used as input against the background of the Homo Sapiens species; analyses were executed based on DAVID default parameter settings (19 th of April 2010).

Quantitative real-time polymerase chain reaction and data analyses
Quantitative real-time PCR (qRT-PCR) was performed and analyzed using ABI PRISM SDS 2.1 software (Applied Biosystems, Foster City, CA, USA, www.applied biosystems.com) and Microsoft Excel as described elsewhere [31]. All primer sequences are provided in Table S1. For detection of pluripotency-associated genes in AFiPSCs, the undifferentiated ESC line H1 was used as a positive control. For the analysis of BMP2-or BMP4-induced trophoblast differentiation of AFiPSCs, placental RNA (Clontech, Mountain View, CA, USA, www.clontech.com) was used as a positive control. Data are presented as mean LOG2 ratios with respect to biological controls and standard deviation.

Immunofluorescence, alkaline phosphatase and cellular senescence staining
For the identification of ESC markers in undifferentiated AFiPSCs (lines 4, 5, 6, 10 and 41) and detection of lineage markers in AFiPSCs differentiated in vitro, cells were fixed, permeabilized and stained for immunofluorescent imaging as described by Prigione et al. [31]. The list of primary and secondary antibodies used is provided in Table S2. Nuclei were counter-stained with DAPI (100 ng/ml, Vector Laboratories, Burlingame, CA, USA, www.vectorlabs.com).
Alkaline phosphatase (AP) staining of all manually picked AFiPSC lines was performed following the manufacturer's instructions (Millipore, Billerica, MA, USA, www.millipore.com).
For staining of senescent cells, the Senescence beta-Galactosidase Staining Kit (Cell Signaling, Danvers, MA, USA, www. cellsignal.com) was used according to the manufacturer's protocol. All stainings were visualized and images were acquired using the confocal microscope LSM 510 Meta (Carl Zeiss, Oberkochen, Germany, www.zeiss.de). Processing of images was done with the help of AxioVision V4.6.3.0 (Zeiss) and Adobe Photoshop CS version 8.0 (Adobe, Munich, Germany, www.adobe.com) software.

Senescence is bypassed by the derivation of AFiPSCs from human AFCs
Under routine cell culture conditions, bulk primary human AFCs ( Figure 1A-I) senesce at approximately passage 17 as determined by decelerated proliferation. These cells also have an enlarged and flattened morphology and stain positive for the senescence-associated beta-galactosidase ( Figure 1A-II, -III). To bypass senescence and to enhance proliferation capacities of AFCs, we derived iPSCs from primary bulk AFCs by transduction with a retroviral cocktail consisting of OCT4, SOX2, KLF4 and c-MYC (OSKM) [24]. The resulting AFiPSC colonies appeared about seven days post-transduction ( Figure 1A-IV), which is approximately two weeks earlier than what we and others have observed for fibroblast-derived iPSCs [24,31]. Five clonal AFiPSC lines were expanded under ESC conditions and partly characterized. Of these, two lines underwent complete characterization. AFiPSCs were indistinguishable from ESCs (e.g. ESC line H1) in terms of morphology ( Figure 1A-V) and proliferation. These AFiPSCs also resembled ESCs with respect to alkaline phosphatase (AP) activity and expression of several markers of the undifferentiated state, including NANOG, OCT4, SOX2, SSEA4, TRA-1-60, TRA-1-81 as determined by immunocytochemistry ( Figure 1B). The AFiPSCs exhibited a normal karyotype several passages after their generation ( Figure 1C) and their genetic relatedness to primary AFCs cells was confirmed by DNA fingerprinting analysis ( Figure 1D).

Pluripotency and in vitro and in vivo differentiation of AFiPSCs
Microarray-based transcriptional analysis revealed up-regulation of self-renewal and pluripotency-associated genes [38][39][40] in AFiPSCs in contrast to primary AFCs (Figure 2A). qRT-PCR validations, performed for a selection of these pluripotencyassociated genes, confirmed the array-derived data ( Figure 2B). The ability of AFiPSCs to differentiate into derivatives representative of all three embryonic germ layers was assessed by embryoid body differentiation in vitro ( Figure 2C) and teratoma formation in vivo ( Figure 2D). Markers or histological structures representing endoderm-, mesoderm-and ectoderm-derived lineages were detected in both assays ( Figure 2C, D).

Trophoblast differentiation of AFiPSCs
To test if AFiPSCs, like ESCs, can undergo trophoblast differentiation [41][42][43][44], we stimulated two AFiPSC lines with 100 ng/ml BMP2 or BMP4 over a period of five days. As a result, a morphological change from densely packed ESC-like colonies ( Figure 3A-I, -II, -V, -VI) towards more loosely packed clusters of enlarged cells with typical cobblestone-like appearance ( Figure 3A-III,-IV,-VII,-VIII) was observed. This is a characteristic feature of trophoblast differentiation of ESCs [41,42]. Gene expression profiling and qRT-PCR analyses revealed down-regulation of the key pluripotency markers POU5F1 and NANOG and up-regulation of the trophoblast markers CDX2, KRT7, HAND1, FOXF1, GATA3, and ID2 ( Figure 3B). Both, BMP2 and BMP4, induced similar effects, however, BMP4 was more efficient. When we treated the AFiPSCs with a combination of 10 ng/ml BMP4 and 10 mM SB431542 over a period of seven days, the same morphological changes could be observed and human chorionic gonadotropin (hCG), a hormone secreted by trophoblastic cells of the placenta, was detected by immunofluorescence microscopy analysis ( Figure 3C).

Activation of a common ESC-like core transcriptional regulatory network in AFiPSCs and FiPSCs
In order to narrow down the self-renewal and pluripotency signature gene list obtained by comparing global gene expression patterns of AFCs, AFiPSCs and ESCs (1299 genes, Figure 4C), we compared the same ESC samples with FiPSCs and the respective parental fibroblast line HFF1 (Fibs) ( Figure 6A, the entire gene lists are presented in Table S5) [31]. Using the resulting equivalent selfrenewal and pluripotency gene signature, we could detect the overlap between the two self-renewal and pluripotency gene lists derived from the separate analyses (AFiPSCs/ESCs: 1299 genes in the self-renewal/pluripotency signature, FiPSCs/ESCs: 922 genes in the self-renewal/pluripotency signature). This revealed 525 genes expressed in common in all our pluripotent cell types (AFiPSCs, FiPSCs and ESCs), highlighting their role in maintaining self-renewal and pluripotency ( Figure 6B, the corresponding gene list is presented in Table S6). To gain further insight into the gene regulatory network (GRN) that induces and maintains pluripotency in AFiPSCs and FiPSCs and to define distinct functions of the 525 core self-renewal-associated genes in the undifferentiated embryonic stem cell state, we identified the overlap of these 525 genes with the list of genomic regions bound by either OCT4 alone or by OCT4, SOX2 and NANOG as identified in human ESCs by ChIP-on-chip analyses [46,47]. This, in turn, revealed a subset of genes expressed in all of our pluripotent cell lines, which are part of an ESC-specific transcriptional regulatory network, including, for example, POU5F1, SOX2, NANOG, DPPA4, LEFTY2 and CDH1 ( Figure 6C). To emphasize the established role of all the heatmap-listed genes in the regulation of the tightly controlled balance between the undifferentiated, self-renewing, pluripotent versus the differentiated ESC state, we combined the heatmap data in Figure 6C with gene expression data derived from siRNAmediated OCT4 knockdown in ESCs [39].

The LARGE Principle of Cellular Reprogramming
What can be gleaned from the global gene expression analyses presented here but also from other iPSC-based transcriptome analyses [48,49], is that induction of pluripotency is associated with the transcriptomes of the parental cells shifting towards a distinct ESC-like state, irrespective of the cell source. More precisely, for a distinct set of genes, which are expressed in the parental cell line, expression is lost (L), whereas the expression of another group of genes is acquired (A) in the process of iPSC generation. In turn, the expression of a third set of genes, detectable in the parental cells, is retained (R) in the corresponding iPSCs. We refer to this as the LARGE (Lost, Acquired, Retained Gene Expression) Principle of Cellular Reprogramming. Also referring to other studies [48][49][50], we propose that these particular LARGE patterns are the key to understanding similarities and differences between iPSCs and ESCs and their parental cell lines on the one hand as well the heterogeneity of different iPSC types on the other. As transcription factors normally influence gene expression of several downstream targets and, thus, are likely to play a fundamental role in this concept, we used gene expression patterns of transcription factors to illustrate the LARGE concept. For this purpose, we made use of the data from the Venn diagram analyses of AFCs/AFiPSCs/ESCs and Fibs/FiPSCs/ESCs ( Figures 4A and 6A, Tables S3 and S5). For each of the Lost (genes expressed in donor cells, but not in iPSCs), Acquired (genes expressed in iPSCs, but not in the donor cells) and Retained (genes expressed simultaneously in donor cells and iPSCs, excluding genes of the house keeping gene signature) Gene Expression sets, we arbitrarily picked out genes associated with the Gene Ontology term for transcription factor activity (GO0003700) [35]. Of these, the 12 transcription factors with the lowest (Lost), highest (Acquired) or least varying (Retained) expression change, when comparing AFiPSCs or FiPSCs with the corresponding parental cells, are depicted in the heatmaps in Figure 7. As a result, the group of lost transcription factor gene expressions included, for example, HOXB7, HOXA9, HOXA10, PAX8, DSCR1, MYC in AFiPSCs and EMX2, FOXF2, FOXF1, MYC, KLF4 in FiPSCs. The acquired gene expression set can be further divided into two groups on the basis of present or absent overlaps between the two analyses for AFiPSCs and FiPSCs: those, which are universally acquired self-renewal genes present in both, AFiPSCs and FiPSCs, or, more generally, in all pluripotent iPSCs (e.g. POU5F1, SOX2, NANOG), and those acquired gene expressions, which are rather iPSC type-dependent (e.g. SIX6, EGR2 (AFiPSCs) or PKNOX2, HOXD4, HOXD10 (FiPSCs); DLX5 (AFiPSCs & FiPSCs)). The retained gene expression sets included genes like PKNOX2 (AFiPSCs); HMBOX1, MGA (FiPSCs) or RAXL1 (AFiPSCs & FiPSCs).

Ground state pluripotency of AFiPSCs
We have shown that cellular reprogramming of primary AFCs results in a fully pluripotent iPSC type, which is in line with recent when compared to the undifferentiated cells (scale bars = 20 mm). (B) qRT-PCR and gene expression profiling (microarray) revealed down-regulation of pluripotency markers POU5F1 and NANOG, but up-regulation of trophoblast markers CDX2, KRT7, HAND1, FOXF1, GATA3 and ID2 upon BMP2-or BMP4-treatment of AFiPSCs. Data are presented as LOG2 ratios (BMP-treated versus untreated AFiPSCs) and standard deviation. (C) Immunofluorescence-based detection of the placental hormone human chorionic gonadotropin (hCG) in AFiPSCs after treatment with 10 ng/ml BMP4 and 10 mM SB431542 over a period of seven days (scale bars = 200 mm). doi:10.1371/journal.pone.0013703.g003  publications [15,29,30]. We have further demonstrated that AFiPSCs are, like ESCs, at an early developmental state, in which they are not only capable of forming derivatives of the three embryonic germ layers but also of the extraembryonic trophoblast lineage. This acquisition of key ESC characteristics during cellular reprogramming should be beneficial for the application of  Figure 4C) and ESC/FiPSC ( Figure 6A)-derived self-renewal signature gene lists. The overlap of 525 genes expressed in all analyzed pluripotent cells (AFiPSCs, FiPSCs, ESCs) represents the core self-renewal signature. (C) Of these 525 self-renewal-associated genes, those, bound by OCT4 or simultaneously by OCT4, SOX2 and NANOG as determined by ChIP-chip analyes [46,47], are depicted in the heatmap as LOG2 average expression signals. The heatmap is colored according to the color key on the bottom. Genes and samples were clustered by similar expression patterns using Eucledian distance measure. The table on the right identifies each gene to be bound by either OCT4 or by OCT4, SOX2 and NANOG (OSN) and shows expression changes upon siRNAmediated OCT4 knockdown in ESC line H1, including the differential expression P-value (P-Val) [39]. doi:10.1371/journal.pone.0013703.g006 AFiPSCs in basic and applied research, although it could be argued that the teratoma formation ability acquired by AFiPSCs hampers their use in cell replacement therapies. Yet, this is a feature of all kinds of iPSCs and ESCs, which still hold a lot of promise in this respect. Presumably, ways will be found to exploit the full differentiation potential of iPSCs while circumventing tumor formation risks, for instance, by developing accurate strategies to sort out differentiated cells of interest from potential tumorigenic stem cells.

Cellular reprogramming bypasses senescence of bulk primary AFCs
One of the great advantages of AFiPSCs over their bulk primary counterparts for any desirable application is their acquisition of the ability to propagate indefinitely. The data presented herein suggest, that this phenotypically rejuvenated appearance of AFiPSCs is based on a gene regulatory network, which averts or at least markedly delays the onset of senescence. This is based on the fact that primary AFCs and AFiPSCs and ESCs exhibit opposing expression patterns related to a large number of senescence-associated genes. In particular, we could detect high expression levels of various cell cycle and telomere elongationassociated genes, such as MAD2L2, PARP1, RPA3, DKC1, MSH6, CHEK1, PLK1, POU2F1, CDC2, LMNB1 and CDT1, as well as TERT itself in AFiPSCs in contrast to primary AFCs. The p53/ p21 pathway plays a pivotal role in inducing and maintaining senescence [51]. Accordingly, mRNA levels of several p53 target genes, which are known to be up-regulated in senescent cells [52][53][54], e.g. CDKN1A (p21), GDF15, and SERPINE1, were strikingly elevated in primary AFCs compared to AFiPSCs and ESCs. In contrast, low level gene expression of DNMT1 and DNMT3B were detected in bulk primary AFCs, whereas these genes are significantly up-regulated in AFiPSCs. Besides their function in establishing and maintaining CpG methylation patterns during embryonal development, they are also known to repress CDKN1A transcription in opposition to and potentially independent of p53 [54,55]. Hence, it could be anticipated that high expression levels the DNMTs may repress CDKN1A and, thus, senescence in AFiPSCs. Taken together, there is evidence that senescence is bypassed upon the activation of a self-renewal and pluripotency program in reprogrammed AFCs, which is in line with our previous findings [31]. However, further studies are needed to assess the actual ability of AFiPSCs to restore telomere restriction fragment length to an ESC level, a subject of controversial discussion in the iPSC field [45,56,57].  probable explanation could be that these gene expression patterns are due to an incomplete erasure of epigenetic imprints in iPSCs depending on the nature of chromatin modifications of the original cell type, or in other words, a kind of cell type-specific epigenetic memory [48,49]. However, viral integrations are probably the cause of the partially inconsistent gene expression patterns observed in the LARGE heatmap (Figure 7). In order to identify actual effects of viral integrations on the host cell genome and to avoid genomic alterations, the generation and comparative characterization of nonviral iPSCs, particularly from human AFCs, still remains.

The LARGE Principle of Cellular Reprogramming and ESCspecific gene expression signatures
In addition to the above-mentioned results of our LARGE analysis, we identified ESC-specific genes in the Venn diagrams, including, for example, PRDM14, WNT3A and GSC. PRDM14 has been implicated in maintaining the undifferentiated ESC state [73]. In contrast, WNT3A and GSC are primitive streak/ mesendoderm markers known to regulate developmental processes [43]. These genes distinguish our AFiPSCs and FiPSCs from ESCs, thus implying incomplete reprogramming and emphasizing general differences between ESCs and iPSCs despite the acquisition of the ESC phenotype in both iPSC types. Follow-up studies should be designed to identify functional consequences of this observation.

Conclusion
Both, primary AFCs, in particular stem cell-like subpopulations of primary AFCs, as well as AFiPSCs are considered to be valuable for the application in basic and applied research. Taken together, our results propose that, for these purposes, cellular reprogramming of AFCs is beneficial as it represses senescence and leads to a phenotype very similar, though not identical, to ESCs. These findings are even more significant, considering that due to the presence of fetal stem cells within bulk primary AFCs, amniotic fluid seems to be a very suitable source of cells for the realization of non-integrating reprogramming strategies. Yet, as a main result of this study, we identified gene expression signatures and LARGE patterns for different types of iPSCs, corresponding parental cells and ESCs. Two conclusions can be drawn from this. First, this kind of comparative transcriptome analysis should be extended integrating iPSC lines derived from several distinct cell sources and generated using various reprogramming techniques, as it would aid in enhancing our meagre understanding of mechanisms underlying cellular reprogramming. Secondly, the functional relevance of such distinct expression patterns, especially of AFCs, AFiPSCs and ESCs, will have to be investigated profoundly in order to estimate limitations and to exploit the full potential associated with putative future utilization of amniotic fluid-derived cells.    Table S4 Senescence-associated genes. List of 116 senescenceassociated genes derived from the Gene Ontology database [35], including those described by Vaziri et al. [45]. These genes served as input for the differential gene expression analysis between AFCs (P17) and the group of all AFiPSC lines.