Reprogramming to pluripotency does not require transition through a primitive streak-like state

Pluripotency can be induced in vitro from adult somatic mammalian cells by enforced expression of defined transcription factors regulating and initiating the pluripotency network. Despite the substantial advances over the last decade to improve the efficiency of direct reprogramming, exact mechanisms underlying the conversion into the pluripotent stem cell state are still vaguely understood. Several studies suggested that induced pluripotency follows reversed embryonic development. For somatic cells of mesodermal and endodermal origin that would require the transition through a Primitive streak-like state, which would necessarily require an Eomesodermin (Eomes) expressing intermediate. We analyzed reprogramming in human and mouse cells of mesodermal as well as ectodermal origin by thorough marker gene analyses in combination with genetic reporters, conditional loss of function and stable fate-labeling for the broad primitive streak marker Eomes. We unambiguously demonstrate that induced pluripotency is not dependent on a transient primitive streak-like stage and thus does not represent reversal of mesendodermal development in vivo.

of Nodal activities induce the expression of mesendodermal marker genes such as Eomes, Mixl1, Tdgf1 (Cripto), Lhx1 and Foxh1 (reviewed in 6,7 ). In particular, Eomes is critically required for the specification of early mesendoderm (DE and anterior mesoderm) 1,4,8 , and all cells of the early PS transiently express Eomes. Eomes-deficiency in the epiblast also dramatically perturbs PS formation due to defective EMT leading to early embryonic arrest 4,8 . Similarly, the sequential formation of different cellular subtypes of the streak can be mimicked by in vitro cell differentiation and monitored by the expression of marker genes. The in vitro differentiation of pluripotent stem cells towards the three germ layers can be guided by similar signaling stimuli and mRNA expression profiles usually reflect the in vivo situation 9,10 .
During reprogramming to induced pluripotency through forced expression of the core pluripotency factors SOX2, OCT4, KLF4, and C-MYC, somatic cells lose their differentiated state 11,12 . Several reports suggested that reprogramming follows distinct stages resembling a reversal of embryonic development 13,14 . Fibroblasts, as the most common starting cell type for reprogramming, represent cells of mesoderm origin. Thus, the reversal of their cellular ontogeny during reprogramming would most likely involve the passage through a PS-like stage 14 . Accordingly, it was proposed that a sequential cascade of EMT-MET facilitates the reprogramming process 13,15 .
To initiate the transcription factor networks and signaling pathways that are characteristic for pluripotent cells, extensive alterations in the epigenetic landscape take place such as broad changes of chromatin modifications, chromatin architecture, and gross changes in the cellular transcriptome 16 . Although several studies have explored mechanisms and stages during the reprogramming process 17 , the question concerning an analogy of the reprogramming process as reversal of physiological embryonic development, including gastrulation is controversial 14 . Moreover, it is debatable why also cells derived from ectodermal lineages, such as astrocytes or keratinocytes would show a PS-like global gene expression pattern during reprogramming 14 , given that ectodermal cells developmentally never ingress through the PS. Thus, it is questionable, whether these events indeed reflect reverted embryonic development or might rather represent changes in transcriptional programs induced by the forced expression of reprogramming factors. To address these developmental aspects of reprogramming, we used different reprogramming approaches including somatic cells from different germ layers and organisms, namely murine and human cells as well as different reporter alleles and fate analysis tools. Thereby, we provide evidence that somatic cell reprogramming neither follows a reversed mesendoderm development nor that occurring mesendodermal gene signatures reach physiological and functionally relevant levels during differentiation.

Results
Gene expression patterns during reprogramming of human somatic cells of mesoderm and ectoderm origin. To investigate if cells during human reprogramming follow stages of reversed embryonic development, we transduced keratinocytes and fibroblasts which have ectodermal and mesodermal origin, respectively, with a polycistronic OKSM (OCT3/4, KLF4, SOX2, c-MYC) construct to monitor and directly compare gene expression signatures during reprogramming (Fig. 1A, Supplemental Fig. 1A). Consistent with previous reports 14 , a transcriptional signature resembling a PS-like and mesendodermal program was observed during reprogramming of both cell types representing different germ layer origin (Fig. 1B, Supplemental Fig. 1B,C). Expression patterns of key markers of PS formation and subsequent early mesendoderm differentiation (EOMES, T, CER, LHX1, FGF4, FGF8, MIXL1) were similarly regulated in both keratinocytes and fibroblasts. However, reprogramming of keratinocytes appeared delayed compared to fibroblasts as shown by the expression profile of the pluripotency marker NANOG (Fig. 1B). Of note, particularly NANOG expression, previously shown to reinforce mesendoderm differentiation during pluripotency exit 18 , coincides with the mesendodermal signature (Fig. 1B, Supplemental Fig. 1B,C). Both mesodermal fibroblasts and ectodermal keratinocytes displayed an increase in mesendoderm and primitive streak markers starting at day 6-8 (fibroblasts) or 9-12 (keratinocytes) with a peak between day 12-14 (fibroblasts; except for CER1) or 15-18 (keratinocytes), followed by the downregulation of these genes (except FGF8) until the induced pluripotent stem cell (iPSC) state (Fig. 1B, Supplemental  Fig. 1B,C). Next, we aimed to determine the expression range of this mesendodermal gene signature by comparing mRNA levels of cells during reprogramming with cells undergoing directed mesendoderm differentiation in vitro. (Fig. 1C). This direct comparison showed that mRNA levels of mesendoderm markers (EOMES, LHX1, CER1) were several magnitudes higher in differentiating cells compared to the expression during reprogramming (Fig. 1D). Given that EOMES expression is critical for PS and subsequent mesendoderm formation 1,4,8 , we evaluated EOMES protein levels during reprogramming. However, no EOMES protein was detected, neither by immunocytochemistry nor by Western blot (in fibroblasts), during reprogramming of fibroblasts and keratinocytes ( Fig. 1E-G). As a control, expression of the pluripotency marker NANOG was analyzed in parallel showing increasing protein levels during reprogramming ( Fig. 1E-G). As control, mesendodermal differentiation of hiP-SCs displayed the expression of EOMES protein on day 3 as analyzed by immunofluorescent staining (Fig. 1H). In summary, these results indicate that PS and mesendoderm markers are significantly upregulated during the course of reprogramming. At the same time NANOG reaches its expression peak, suggesting that the establishment of pluripotency networks triggers a PS-like expression phenotype. However, mRNA levels are detected at much lower levels compared to those found during mesendodermal differentiation (Fig. 1D).

Murine fibroblasts undergoing reprogramming do not express Eomes protein.
To corroborate the finding that PS markers are frequently expressed only at low mRNA level during reprogramming, independent of the parental germ layer origin, we investigated the expression of Eomes as one of the key TFs for PS and mesendoderm development 1,4,8 . Since we didn't detect EOMES protein during human somatic cell reprogramming ( Fig. 1E-G), we sought to apply mouse embryonic fibroblasts (MEFs) harboring a Eomes GFP/+ reporter allele 19 to track eventually arising Eomes-expressing, GFP-positive cells using very sensitive FACS techniques ( Fig. 2A). Eomes GFP/+ MEFs were transduced with a polycistronic OKS (OCT3/4, KLF4, SOX2) construct harboring a Td-tomato expression cassette to visualize cells that undergo reprogramming 20 ( Fig. 2A,B). During 21 days of reprogramming, FACS analyses and immunocytochemistry for both the GFP and Tomato signal were conducted at intervals of 2 or 3 days. FACS-analysis did not reveal any GFP-positive cells, while the red tomato-signal from the reprogramming cassette expectedly got silenced when reaching the iPSC state on day 21 20 (Fig. 2B,C). In line with human data, we observed Eomes and other mesendodermal marker up-regulation on mRNA level but in a far lower range than observed in spontaneous, differentiating mouse iPSC cultures (Supplemental Fig. 2; Fig. 2D,E). To control for the efficiency of the Eomes GFP/+ reporter, we differentiated resulting Eomes GFP/+ iPSCs using high doses (50 ng/ml) of Activin A to drive mesendoderm formation and could detect high levels of GFP-expression. Thus, we could successfully validate the functionality of the reporter allele during differentiation ( Fig. 2F-H). Despite the inability to detect Eomes GFP/+ reporter expression during reprogramming, transient Eomes expression cannot be entirely excluded, as cell samples were harvested at time-intervals of 2-3 days.

Eomes expression remains undetectable during cell lineage tracing. To rigorously test if Eomes
is significantly expressed at any stage during reprogramming, we used MEFs carrying a 4-hydroxytamoxifen (4-OHT)-inducible CreERT in the Eomes locus (Eomes CreERT ) and a Cre-inducible fluorescent reporter cassette (Rosa26 Tom/GFP ) to permanently lineage-label Eomes-expressing cells 21 (Fig. 3A,B). Ssea1 was used in FACS analysis to mark pluripotent cells and asses the efficiency of reprogramming. Among the detected Ssea1-positive cells, no GFP-positive cells were detected after tamoxifen treatment on days 3-15 during reprogramming (Fig. 3C). The absence of GFP-positive cells in successfully reprogrammed cells was further confirmed by immunofluorescence in picked and expanded iPSC cultures (Fig. 3D, upper image), indicating the lack of Eomes expression during and at the end of the reprogramming process. To validate the linage labeling tool used in these experiments, we induced mesendoderm differentiation of resulting iPSCs in the presence of tamoxifen and Activin A, which resulted in the appearance of GFP-positive, Eomes expressing cells within the differentiating Tomato-positive embryonic bodies (Fig. 3D, lower image).

Eomes is dispensable for reprogramming of somatic cells of mesoderm origin.
To exclude that very low amounts of Eomes protein are being expressed at levels undetectable via FACS or immunofluorescent staining, we used MEFs carrying the Eomes GFP reporter allele and a floxed Eomes allele 8 (Eomes GFP/fl ) in combination with a ubiquitously expressed 4-hydroxytamoxifen (4-OHT)-inducible CreERT (Rosa26 CreERT ) to inducibly delete Eomes function during reprogramming (Fig. 4A,B). Given the critical role of Eomes for PS formation in the early embryo, we reasoned that the reprogramming of cells lacking Eomes expression would be impaired if the transition of cells through the PS-like intermediate state would be a crucial step during the reprogramming process. Cells were transduced with the OKS construct 20 and treated with tamoxifen at different time points to induce the genetic deletion of Eomes (orange letters; Fig. 4B,C). Efficient reprogramming was assessed by alkaline phosphatase staining (Fig. 4D), FACS staining for Ssea1-positive cells (Fig. 4E) and Oct3/4 expression (Fig. 4F). The loss of Eomes did not result in any significant change in number, morphology, or marker expression of arising iPSC colonies, irrespective of the time-point of induced deletion (Fig. 4D-F), albeit the slight Eomes expression peak (Supplemental Fig. 2) could be ablated upon tamoxifen treatment (Fig. 4G). This indicates that Eomes is functionally entirely dispensable for the reprogramming to pluripotency, despite its prominent role during PS formation and gastrulation initiation.

Discussion
Cellular events during reprogramming were extensively studied over the past years. However, the different stages during reprogramming need to be further defined and exact molecular mechanisms remain to be resolved. Several studies of changes in gene expression during reprogramming have suggested that cells undergo a reversal of embryonic development including MET-EMT events and transiently acquire a PS-like gene expression signature 13,14,15 . Gene expression patterns of PS formation, as well as EMT-MET events, occur independently of an epithelial or mesenchymal origin of the starting cell population during reprogramming 13,14,15 . Thus, it remains questionable whether reprogramming indeed follows stages of "reverse embryonic development", or if observed gene signatures solely represent the spurious activation of developmental programs, or if genetic programs are indeed necessary to establish the pluripotency network. The latter view was recently supported by studies indicating that EMT-related transcription factors cooperate with core factors of the pluripotency circuitry to induce pluripotency 22,23 . Here, we apply different genetic tools including fate-analysis and reporter alleles at the Eomes gene locus, as one of the central transcription factors with important functions in the gastrulating mouse embryos for PS formation, EMT and specification of the mesendoderm lineages. None of the applied genetic tools and analyses indicated significant expression of Eomes on the route to iPSC reprogramming. Additionally, we tested if Eomes was functionally required during reprogramming by genetically deleting Eomes in starting cells. Indeed, the genetic deletion of Eomes had no effect on reprogramming efficiency, suggesting that the induction of a transient PS-like state is no crucial step during reprogramming.
We propose that observed mesendodermal/PS-like expression profiles in cells that are undergoing reprogramming reflect a non-physiological transcriptional response to the reprogramming factors. Thus, it is likely that high level expression of reprogramming factors by the lentiviral transduction and subsequent re-activation of the pluripotency network triggers the transcription of mesendodermal genes. In addition to maintaining the pluripotent state, transcription factor of the pluripotency network such as Nanog, Oct3/4, Klf4, and Tbx3 share functions  during the early phase of exit from pluripotency. Thus, they contribute to the initiation of transcriptional programs to guide mesendoderm cell fate determination, e.g. by regulating Eomes expression 5,24-26 . In that way, core pluripotency factors govern the first steps of differentiation and cell fate determination 5,26,27 . This hypothesis is underlined by the fact that mesendoderm transcriptional signatures can be likewise found in cells of ectodermal origin during reprogramming, although these cells never go through a mesendodermal/PS-like state during in vivo embryogenesis. Finally, the extent of mesendodermal gene transcription levels did not reach the range of physiological lineage differentiation as shown for murine and human somatic cell reprogramming. In summary, our data confirm the previously described mesendodermal fingerprint arising during reprogramming in cells of mesoderm and ectoderm origin 14 . While previous reports interpreted these findings as a reversed process of embryonic development [13][14][15] , our data instead favors a non-physiological transcriptional response resulting from the forced induction of the pluripotency network, which does not reflect the magnitude of gene expression as seen during mesendodermal lineage commitment in vivo 5,24-26 .

Material and Methods
Cell cultures. Rat embryonic fibroblasts (REFs) from embryonic day 14 Sprague Dawley rats were generated according to the protocol previously described in 28 and were cultured in Dulbecco's Modified Eagle's Medium (DMEM, Sigma Aldrich) containing 10% fetal bovine serum (FBS, Sigma Aldrich/Biochrom), 1% GlutaMAX, 1% nonessential amino acids (NEAA), and 1% antibiotic-antimycotic (all from Life Technologies). REFs were treated with 7.5 μg/mL mitomycin C (Biomol) for 2.5 hours for mitotic inactivation. All animal experiments were performed in compliance with the guidelines for the welfare of experimental animals issued by the Federal Government of Germany, the National Institutes of Health and the Max Planck Society. The experiments in this Cells for reprogramming. Mouse embryonic fibroblasts (MEFs) were cultured according to standard methods at 5% CO2 and 37 °C as described previously in 29,30 . Briefly, DMEM was supplemented with 15% FBS, 1% P/S, 1% GlutaMAX, 1% NEAA, 1 mM Sodium Pyruvate (Sigma Aldrich), 1% β-Mercaptoethanol (Merck Millipore) and 0.05 mg/ml Vitamin C. The cultivation of keratinocytes from plucked human hair was performed according to 28,[31][32][33] . In brief, keratinocytes were cultured on 20 μg/mL collagen IV (Sigma Aldrich) coated dishes in EpiLife medium with HKGS supplement (Gibco ® Life Technologies) until they reached about 70% confluency. Human foreskin fibroblasts (HFFs) (System Biosciences) were cultivated in DMEM supplemented with 10% FBS and 1% GlutaMAX, 1% NEAA, and 1% antibiotic-antimycotic.
Mouse. MEFs were seeded on a gelatine coated plate (4 × 104 cells/12-well) one day prior to infection. Next day, 5 µl concentrated polycistronic OKS (OCT3/4, KLF4, SOX2) lentivirus harboring a Td-tomato 20 together with 8 µg/ml polybrene (Sigma Aldrich) in 1 ml ES Feeder Medium was added to each 12-well. After 8 h of incubation at 37 °C, medium was removed, cells were washed with PBS and ES-Feeder medium was added and refreshed daily. At day 6, medium was changed to ES Feeder KOSR, where FCS was exchanged by KOSR. On day 20, cells were either stained for alkaline phosphatase (AP) expression according to standard protocols or cells were analyzed by flow cytometry.
Tamoxifen treatment. 4-Hydroxytamoxifen (Sigma Aldrich) was added to the cell culture medium to a final concentration of 1 µg/ml for the respective time frames. After tamoxifen treatment cells were washed with PBS once and received fresh medium.
Gene expression Analysis. Total RNA was isolated from cell lysates using RNeasy Mini Kit according to the manufacturer's instructions (Qiagen). First, cDNA synthesis was performed using 80 ng RNA with RT Buffer (Promega), dNTPs (GE Healthcare), Hexanucleotide Mix (Roche) and MMLV RT (Promega). For the preamplification step PreAmp Master Mix, SuperScript III First-Strand Synthesis SuperMix (both Thermo Fisher), TE buffer (Ambion) was used according to the manual. To quantify the amount of the genes of interest QuantiTect Primer Assays (Qiagen) were used on the BioMark HD System with 96.96 Dynamic Arrays (both Fluidigm). Relative gene expression was calculated as a ratio of target gene concentration to the housekeeping gene concentration. Details have been described in 38,40 . FACS analysis. While Ssea1 surface staining and Tomato/GFP auto-fluorescence staining was performed on living cells, cells were fixed in 4% paraformaldehyde, 10% sucrose for 20 minutes on ice for intranuclear Oct3/4 staining. Stainings were performed according to standard methods. Briefly, adherent cells were washed with PBS and dissociated into single cell suspension by incubation with 0.25% trypsin/EDTA (Millipore). For staining of cells were blocked in PBS supplemented with 10% FBS, incubated with primary antibody α-Ssea1 (1:1600) (Cell Signaling MC480) for 1.5 h on ice in the dark, and incubated with secondary antibody α-mouse AlexaFluor 647 nm (1:600) (Invitrogen A21238) for 30 min on ice in the dark. Washing steps were performed with PBS with 2% FBS, and 1% P/S (FACS Buffer). For staining of Oct3/4 paraformaldehyde-fixation was followed by permeabilization of cells for 30 min in 0.5% Saponin (Sigma Aldrich) in FACS Buffer, blocking in 5% normal goat serum (Sigma Aldrich), and 0.5% Saponin in PBS for 20 min on ice. Antibodies were diluted in the blocking dilution, the primary antibody α-Oct3/4 (Santa Cruz sc-5279) 1:100, and the secondary antibody α-mouse AlexaFluor 488 nm (Invitrogen A11029) 1:200. Washing steps were performed with FACS Buffer supplemented with 0.5% Saponin. Cells were analyzed with a FACSAria II or III flow cytometer (BD). All events were gated with forward scatter and side scatter profiles.

Statistical analysis.
All experiments were independently repeated at least 3 times and Error bars in the graphs show calculated Standard Error if not otherwise stated. Statistical significance was calculated using Students t-test. p-values have been calculated where appropriate and now illustrated by asterisks according to the following definitions: *p < 0.05; **p < 0.01; ***p < 0.001. GraphPad Prism 5 was used for statistical and graphical data evaluations.