Ten-eleven-translocation 2 (TET2) belongs to the TET protein family that catalyzes the conversion of 5-methylcytosine into 5-hydroxymethylcytosine and plays a central role in normal and malignant adult hematopoiesis. Yet the role of TET2 in human hematopoietic development remains largely unknown. Here, we show that TET2 expression is low in human embryonic stem cell (ESC) lines and increases during hematopoietic differentiation. shRNA-mediated TET2 knockdown had no effect on the pluripotency of various ESCs. However, it skewed their differentiation into neuroectoderm at the expense of endoderm and mesoderm both in vitro and in vivo. These effects were rescued by reintroducing the targeted TET2 protein. Moreover, TET2-driven differentiation was dependent on NANOG transcriptional factor. Indeed, TET2 bound to NANOG promoter and in TET2-deficient cells the methylation of the NANOG promoter correlated with a decreased in NANOG expression. The altered differentiation resulting from TET2 knockdown in ESCs led to a decrease in both the number and the cloning capacities of hematopoietic progenitors. These defects were due to an increased apoptosis and an altered gene expression profile, including abnormal expression of neuronal genes. Intriguingly, when TET2 was knockdown in hematopoietic cells, it increased hematopoietic development. In conclusion, our work suggests that TET2 is involved in different stages of human embryonic development, including induction of the mesoderm and hematopoietic differentiation. Stem Cells2014;32:2084–2097
DNA methylation is implicated in the control of cellular differentiation, embryogenesis, and oncogenesis [1, 2]. DNA methyltransferases (DNMT) add a methyl group to a cytosine, generating 5-methylcytosine (5mC) that can be removed either in a passive way via inhibition of methylation in the newly synthesized DNA or by active enzymatic process. However, the existence of an active demethylase activity in mammals has been elusive .
Proteins of the ten-eleven-translocation (TET) family (TET1, 2 and 3) are 2-oxoglutarate- and Fe(II)-dependent oxygenases that convert 5mC into 5-hydroxymethylcytosine (5-hmC) [4, 5], or into 5-formylcytosine (5-fC) and 5-carboxylcytosine (5-caC) [6, 7], possibly initiating DNA demethylation. TET1 may control DNA methylation by binding to CpG-rich regions to prevent unwanted DNMT activity . Alternatively, TET1 cooperates with base excision repair activities to actively remove 5mC, uncovering a novel epigenetic regulation mechanism [9, 10].
High frequency of 5-hmC and abundant TET proteins have been detected in various cell types including embryonic stem cells (ESCs) [11, 12], neurons  or oocytes, and zygotes [14, 15]. 5-hmC are especially enriched in promoter and exons of highly transcribed genes [16-19]. In mouse ESCs, Tet1 plays a dual role in transcription regulation by either promoting demethylation at CG-rich promoters or repressing transcription through the binding to Polycomb group target genes [18, 19]. In mouse ESCs, Tet1/2 are the most abundant Tet proteins and tet1−/− or/and tet2−/− ESCs are pluripotent and develop the three germ layers [8, 12, 16, 19]. Moreover tet1−/− mice and a fraction of tet1−/−tet2−/− mice can survive [20-22]. However tet1−/− and tet1−/−tet2−/− ESC-derived teratomas are skewed toward the trophoblast differentiation [20, 21]. This suggests that Tet1 and Tet2 are involved in mouse development, but that other epigenetic regulators such as Tet3 can compensate their loss in vivo.
The precise role of TET2, particularly in human, is incompletely understood. We and others have identified alterations in the TET2 gene in various hematological malignancies [22-27]. A decrease in 5-hmC levels was found in TET2-mutated patients displaying myeloid disorders associated with hypomethylated or hypermethylated genomic DNA [4, 28, 29]. TET2 knockdown affects both murine and human adult hematopoiesis by allowing the self-renewing potential of hematopoietic stem cells and favoring monocytic differentiation at the expense of granulopoiesis [4, 29]. With age, tet2+/− and tet2−/− mice develop a myeloid malignancy resembling human myelomonocytic leukemia [22, 30, 31]. In addition, it has been recently shown that acquired TET2 mutations can be found in elderly people with a clonal hematopoiesis in the absence of hematological malignancies . Altogether, these data suggest that TET2 plays a major role in normal hematopoiesis by limiting the self-renewal capacities of hematopoietic stem cells and by regulating differentiation.
Even if tet2−/− mice develop normally, the role of TET2 in human development and hematopoietic emergence remains unknown. In this study, we investigated the role of TET2 in human ESCs (hESCs), including their pluripotency and hematopoietic differentiation potential. We show that TET2 silencing in hESCs decreased the development of the mesoderm germ layer through a downregulation of NANOG but without altering their pluripotency. Consequently, the development of hematopoiesis was greatly impaired. In contrast, TET2 silencing targeted to hematopoietic cells led to an increased hematopoietic differentiation. Altogether, these results suggest that TET2 regulates human development.
Materials and Methods
Liquid culture media, including Iscove's Modified Dulbecco Medium, were from Invitrogen (Cergy Pontoise, France, https://www.lifetechnologies.com/fr/en/home/brands/invitrogen.html). Human recombinant cytokines erythropoietin (EPO), thrombopoietin (TPO), and stem cell factor (SCF) were generous gifts from Amgen (Thousand Oaks, CA), Kirin (Tokyo, Japan), and Biovitrum AB (Stockholm, Sweden), respectively. IL-3 was from Miltenyi Biotec (Paris, France, https://www.miltenyibiotec.com/en/). Vascular endothelial growth factor (VEGF) and basic fibroblast growth factor (bFGF) were from Peprotech (Neuilly-sur-Seine, France, http://www.peprotech.com/en-GB)
H9 hESCs (NIH code WA09) were obtained from WiCell Research Institute (Madison, WI) and HUES 1, 2, 7 from DA Melton (Harvard Stem Cell Institute, Boston, MA). Cells were cultured as described previously . Cells were maintained in an undifferentiated state on mouse embryonic fibroblasts (MEF) in Dulbecco modified Eagle medium (DMEM)/F-12 supplemented with 20% knockout serum replacement, 0.1 mM of minimum essential media nonessential amino acids solution, 2 mM l-glutamine, 50 U/ml penicillin, 50 µg/ml streptomycin (Invitrogen), 2 ng/ml of 0.1 mM beta-mercaptoethanol (Sigma, St. Louis, MO, http://www.sigmaaldrich.com/france.html), and 10 ng/ml human bFGF. Cells were fed daily and replated weekly to maintain the undifferentiated growth. For this study, we used H9 hESCs from passage 30 to 66, HUES1 cells from passage 35 to 38 and HUES2 cells from passage 25 to 28.
Culture and Quantification of Hematopoietic Progenitors
Hematopoietic differentiation was performed on OP9 stromal cells in the presence of VEGF (20 ng/ml) . On day 7, we added EPO (1 U/ml), TPO (20 ng/ml), SCF (25 ng/ml), and IL-3 (10 ng/ml) and on days 10–14, cells were enzymatically dissociated. The recovered cells were cultured or sorted for expression of green fluorescent protein (GFP), CD43, and CD34. Sorted cell populations were plated in triplicate at a density of 1.5 × 103 cells per milliliter in methylcellulose medium (H4100 methocult, Stem Cell Technologies, Grenoble, France, http://www.stemcell.com/en/Products/Area-of-Interest/Stem-cell-biology.aspx?tab=1&gclid=CIXGtJaM070CFUoJwwodU5MAzQ) supplemented with 25 ng/ml SCF, 100 U/ml IL-3, 20 ng/ml TPO, and 2 U/ml EPO. Hematopoietic colonies were counted after 10–14 days of culture. Alternatively, clonal differentiation of ESCs was also performed on OP9 cells at single-cell level. ESCs were sorted on GFP+ and TRA1–81+ at one cell/well in 96-well plates using a FACSDiva cell sorter (Becton Dickinson [BD], Le Pont de Claix, France, http://www.bd.com/fr/). The percentage of GFP+CD34+CD43+ cells in each clone was measured by flow cytometry at days 11–13.
Antibodies and Flow Cytometry Analysis
For cell surface markers identification, cells were incubated with different fluorochrome-conjugated antibodies and analyzed on a FACSCanto II (BD) or sorted on a FACSAria III cell sorter (BD) or on an Influx flow cytometer (BD). Directly conjugated monoclonal antibodies were used for sorting and characterization of either pluripotent cells (SSEA-4, eBioscience, San Diego, CA, http://www.ebioscience.com/ and TRA1–81, [BD]) or hematopoietic cells (anti-CD34, Beckman, Villepinte, France, https://www.beckmancoulter.com/wsrportal/wsr/research-and-discovery/products-and-services/centrifugation/index.htm; anti-CD43, glycophorin A (GPA), Invitrogen and anti-CD41, CD15, and CD14, Pharmingen, San Diego, CA, https://www.bdbiosciences.com/). For apoptosis analyses, cells were stained with Annexin V (BD) and DAPI (Sigma, Lyon, France). Cell cycle was analyzed by BrdU (BD) and DAPI incorporation. hESCs colonies were stained by an alkaline phosphatase kit using StemTAG Alkaline Phosphatase Staining kit (Cell Biolabs, Euromedex, France, http://www.euromedex.com/).
Constructs and Production of Viral Particles
Short hairpin (sh)RNA strategy to regulate TET2 expression based on lentivirus was used as previously described . Sequences for human shTET2 and shTET2.b were (5′GGGTAAGCCAAGAAAGAAA3′) and (5′AAACAAAGAGCAAGA GATT3′), respectively. Lentivirus particles containing pRRLsin-PGK-eGFP-WPRE vector (Genethon, Evry, France) were produced as previously described  and H9, HUES1, and HUES2 cells were infected with lentivirus containing human shTET2 and shTET2.b. GFP+ cells were sorted on a FACSDiva cell sorter (BD). Alternatively H9 cells were infected with a lentivirus containing commercial human shTET2puro or a SCRpuro. Cells were selected by puromycin (1 µg/ml) (Thermoscientific, Saint Rémy-les Chevreuses, France, http://www.thermoscientific.com/en/home.html).
Flag-TET2wt expression vector was generated by site-directed mutagenesis mutated to be unrecognized by shTET2 with 5′GGA GAA AGA CGT AAC TTC GGG GTA TCG CAA GAA AGA AAT CCA GGT GAA AGC3′ primers and was cloned in the lentiviral vector pRRLsinEF1a-PGK-cherry-WPRE vector. Stable shTET2 ESCs were infected twice with Flag-TET2wt-expressing virus and the GFP+/cherry+ cells were sorted on an Influx cell sorter (BD). For NANOG expression, we transduced hESCs with pSIN-EF2-Nanog-Puro (Addgene plasmid #16578, Cambridge, MA, https://www.addgene.org/) and selected them with puromycin.
hESCs were scrapped and SSEA-4+TRA1–81+ cells (1 × 106) were sorted and resuspended in 140 µl ES medium. Undiluted matrigel (60 µl) was added prior to subcutaneous injection into 15 Rag2−/− γC−/− mice for each condition. After 8–12 weeks, tumors were isolated and fixed in formalin (10%). Sections were stained for germ layers analysis. For quantification of the three germ layers, entire slides were scanned and each layer was delineated with Adobe Photoshop software and each area was quantified. Spontaneous differentiation was generated by embryoid body (EB) formation . To generate EB, hESCs were treated with 1 mg/ml collagenase IV (Sigma) for 7 minutes and washed before gently harvesting by scraping. The clusters were plated in low adherence six-well plate in medium consisting of DMEM KO (Invitrogen) supplemented with 15% FBS (Hyclone Laboratories Stem Cell Technologies, Meylan, France). Every 2 days, half of the media was replaced.
Methylation Study by Bisulfite DNA Modification and PCR
DNA (200 ng) was treated by sodium bisulfite using Methyl Detector (Active Motif) according to the manufacturer's protocol and stored at −80°C until use. Converted promoters were amplified by PCR using converted primers and direct sequencing reaction was performed according to manufacturer's instructions (Applied Biosystems, http://www.appliedbiosystems.com/absite/us/en/home.html). Primer sequences have been listed in Supporting Information Table S1.
Day 14 EBs were dissociated by incubation in collagenase IV (Invitrogen) for 30 minutes at 37°C before a 2 minutes trypsin/EDTA treatment at 37°C. Cells were cross-linked for 10 minutes in 1% formaldehyde and terminated by the addition of 125 mM glycine. Cells were washed, collected, and resuspended in 100 µl of SDS lysis buffer (Millipore Billerica, MA, http://www.millipore.com/) containing 1× of protease inhibitors cocktail (Roche, https://www.roche-applied-science.com/shop/home) for 15 minutes. DNA were fragmented by sonication using Covaris (S220, LGC Genomics GmbH, Berlin, Germany), shearing was confirmed by agarose gel electrophoresis to have an average size of 0.5 kb. ChIP reaction was performed using ChIP-IT Express kit according to the manufactor's protocol (Active Motif). Briefly, 25 µl of chromatin was incubated with a rabbit IgG (CliniSciences, Santa-Cruz, Nanterre, France sc-2027, http://www.clinisciences.com/index.php?langue=en) or an anti-TET2 antibody (Santa Cruz, sc-136926) or an anti-H3K4me3 antibody (Active Motif). ChIP samples and input (2 µl) were analyzed by qPCR, with specific primer (hNanogprom-S1/AS1) (Supporting Information Table S1) using the Power SYBR Green PCR Master Mix (Invitrogen). Fold enrichments were calculated by comparing ChIP samples to genomic DNA control and IgG.
mRNA Expression Analysis
Total RNA was isolated using RNeasy Mini Kit (Qiagen, Courtaboeuf, France, http://www.qiagen.com/) and cDNA was synthesized by SuperScript II Reverse Transcriptase (Invitrogen). PCRs were carried out in the ABI Prism GeneAmp 7500 Sequence Detection System (Applied Biosystem, Invitrogen), using the Power SYBR Green PCR Master Mix (Invitrogen) and specific primers (Supporting Information Table S1). The expression levels of all genes were expressed relative to PPIA or HPRT. For microarray analysis, RNA was hybridized on Agilent 4X44K arrays following manufacturer's procedures. Analysis was performed using Bioconductor and Rosetta Resolver (Microsoft Corp., NY). Genes selected for heatmap were described by Vodyanik et al. . Gene ontology term enrichment (DAVID) was used to classify regulated genes.
Western Blot Analysis
Western blot analysis was done using an anti-TET1 or anti-TET2 rabbit polyclonal (produced in house)  or an anti-TET2 mouse serum against the C-terminal part of the protein (produced in house)  or anti-TET3 (CliniSciences, Santa-Cruz, Nanterre, France). Normalization was done using HSC70 antibody (Enzo Life Sciences, Villeurbanne, France, http://www.enzolifesciences.com/contact-us/). Fold change was calculated after quantification using IMAGEJ software.
Results are means ± SEM of at least three independent experiments. Statistical analysis of the cell lines was performed using the Student's t test.
TET2 Transcript Is Upregulated During Hematopoietic Differentiation of hESCs
We first checked the gene expression of TET1, TET2, and TET3 in the undifferentiated human H9−, HUES1−, and HUES2− TRA1–81+ cells by qRT-PCR. All cell lines expressed high levels of TET1 compared with TET2 (eightfold lower) and TET3 (fivefold lower). Spontaneous differentiation by EB formation was associated with a reduction in 5-hmC, as shown by dot blot, without obvious change in the levels of 5mC (Fig. 1F, 1G), confirming the association between high 5-hmC level and pluripotent state [11, 12]. Overall, these results are in agreement with a recent report on human-induced pluripotent stem cells (IPSC) . As expected, expression of the pluripotent markers POU5F1 (OCT4) and NANOG rapidly decreased during differentiation (Fig. 1D). TET2 and TET3 expression levels progressively increased while TET1 expression remained stable (Fig. 1A–1C). These observations were also confirmed at the protein level (Fig. 1E).
H9 hESCs were next induced toward hematopoietic differentiation. We measured the mRNA levels of the TET genes in cultured cells at different steps of differentiation: at day 4 (before emergence of hematopoiesis), at days 10 and 14 in the hematopoietic progenitor-enriched CD34+CD43+ cell population , and between days 14 and 18 in cell supernatant (maturing hematopoietic cells). Hematopoietic differentiation was associated with a gradual decrease in TET1 expression, a small increase in TET3 transcription, and a time-dependent accumulation of TET2 mRNA (Fig. 1H). H9-derived hematopoietic cells showed a 20-fold decrease in TET1 and a 30-fold increase in TET2 expression compared to undifferentiated cells. The small increase in TET3 expression did not reach statistical significance (Fig. 1I). These results underscore the association between hematopoiesis and TET2.
TET2 Knockdown Does Not Affect the Pluripotent Markers of hESCs
To study the role of TET2 during human differentiation, we transduced H9 hESC lines with a lentivirus expressing either shRNA targeting the third exon present in all described TET2 isoforms (shTET2 or shTET2.b) or a scramble sequence (SCR), and GFP as a selection marker. All cell lines had a normal karyotype (Supporting Information Fig. S1). Alternatively, we used a lentivirus vector conferring puromycin resistance to hESC H9 cells transduced with shTET2 (shTET2puro) or SCR (SCRpuro). Among the three shRNA used in the study, shTET2 was the more potent by inducing a 60% reduction in TET2 transcripts and was used hereafter (Fig. 2A). TET2 knockdown did not affect the levels of TET1 and TET3 transcripts (Fig. 2A) and did not induce changes in colony morphology (not shown). SCR and shTET2 cells cultured on MEF layers in the presence of FGF2 showed equal expression of TRA1–81 and SSEA-4 pluripotency markers (Fig. 2B) and phosphatase alkaline staining remained unchanged (Fig. 2C). 5-hmC content did not differ significantly between SCR and shTET2 cells (0.9 ± 0.2-fold) (Fig. 2D), even after several passages (Supporting Information Fig. S2A, S2B). No difference in the cell cycle and apoptosis was observed (Fig. 2E, 2F). qRT-PCR analysis showed similar expression of the pluripotent markers NANOG, POU5F1 (OCT4), and SOX2 (Fig. 2G). Microarray analysis showed very similar gene expression profiles in SCR and shTET2 cell lines, except for seven genes (Supporting Information Fig. S2C). Five of these seven genes were confirmed in qRT-PCR in five independent samples (Supporting Information Fig. S2D). Altogether, TET2 knockdown did not alter the pluripotent markers of hESCs.
TET2 Knockdown Skews Spontaneous Differentiation of hESCs Toward Neuroectoderm
We further explored the spontaneous differentiation of hESCs into the three embryonic germ layers by analyzing gene expression during EB formation until day 14. In shTET2 H9 cells, we observed a marked decrease expression of genes involved in mesoderm and endoderm specification, whereas expression of genes involved in neuroectoderm differentiation increased (Fig. 3A, 3B and Supporting Information Fig. S3A). Accordingly, DAVID gene functional classification showed a marked increase in gene sets defining neuroectoderm differentiation whereas those defining the mesoderm (“heart development”) and endoderm (“lipid homeostasis” with hepatic gene expression) differentiation were downregulated (Supporting Information Table S2). The TET2 knockdown-induced decreased mesoderm managed by BRACHUYRY (T) gene expression was also observed in HUES1 cells early during differentiation (at day 2) (Supporting Information Fig. S3B). These results were reproduced with another shRNA targeting TET2 (shTET2puro) (Supporting Information Fig. S3C). Moreover, when we expressed an ectopic TET2wt (unrecognizable by shTET2) in TET2 knockdown ESCs, we restored a normal spontaneous differentiation as shown by gene expression analysis (Fig. 3A–3C). These results showed that TET2 depletion inhibits mesoderm development and consequently skewed differentiation toward neuroectoderm. Importantly, we performed promoter methylation analyses of several gene markers of neuroectoderm (PAX6, TFAP2B, and ZIC1) and mesoderm/endoderm (APOA1 and EOMES) during EB formation. As shown in Figure 3C and Supporting Information Figure S3D, APOA1 and EOMES were hypermethylated in shTET2 cells whereas PAX6, TFAP2B, and ZIC1 were hypomethylated correlating with gene expression analyses. To confirm the skewed differentiation, we performed xenograft transplantations of ESCs into immunodeficient NOG (NOD/Shi-scid/IL-2Rγnull) mice. Both ESC lines were able to generate teratomas with all three embryonic germ layers. However, TET2 knockdown teratomas showed decreased mesoderm/endoderm and elevated neuroectoderm development compared to control teratomas (Fig. 3D). Altogether, these results show that TET2 knockdown negatively regulates meso-endoderm differentiation at the benefit of neuroectoderm through changes in promoter methylation of key regulatory genes.
NANOG Is Implicated in TET2-Mediated Spontaneous Differentiation of hESCs
NANOG has been shown to direct the differentiation of mesoendoderm [40, 41], and TET2 associates with NANOG promoter  or interacts with NANOG protein  during reprogramming of somatic cells to IPSC. Thus, we asked if TET2 could affect NANOG expression in our system. We found that shTET2 induced a decreased expression of NANOG during EB differentiation compared to SCR (Figs. 3A, 4A). Concomitantly, NANOG promoter was methylated in shTET2 cells (Fig. 4B). ChIP experiments with an anti-TET2 antibody revealed an enrichment of TET2 (fivefold relative to IgG) at the NANOG promoter, which was abolished in shTET2 cells (Fig. 4C). Moreover, the active transcription mark H3K4me3 antibody was greatly reduced by shTET2 at the NANOG promoter (Fig. 4C).
We asked whether an ectopic NANOG expression could rescue the decrease in meso-endoderm induced by TET2 deficiency. We developed H9-shTET2 hESC lines transduced with a lentiviral vector encoding NANOG and performed gene expression during EB formation (Fig. 4D). Interestingly, NANOG overexpression rescued the shTET2 phenotype by increasing the expression of genes involved in meso-endoderm and by decreasing those involved in neuroectoderm. These results suggest that TET2 controls NANOG gene expression and show that NANOG overexpression could overcome the effect of TET2 knockdown on neuroectodermal versus meso-endodermal differentiation.
TET2 Is Required for Hematopoietic Differentiation of hESCs
We studied the hematopoiesis of TET2-depleted H9 hESCs. The shRNA induced a constant decrease (60%) in TET2 mRNA all along hematopoietic differentiation (Fig. 5A). Immunoblot analysis confirmed the increased expression of TET2 during hematopoietic differentiation, and showed a 70% decrease in protein expression by the shRNA (Fig. 5B). Hematopoietic cells, as defined by the CD43 antigen , showed a significant reduction in 5-hmC content upon TET2 knockdown using dot blot (0.6 ± 0.2-fold decrease) (Fig. 5C) and immunofluorescence (Supporting Information Fig. S5A). TET2 depletion induced a marked (∼70%) reduction in the percentage of cells expressing hematopoietic differentiation markers (Supporting Information Fig. S4) and in the total number of megakaryocytes/monocytes (White) and erythroid (EryP) colonies generated in semisolid medium (Supporting Information Fig. S5B) [33, 44]. Similar results were obtained with another shRNA (shTET2puro) (Supporting Information Fig. S5C).
To precisely quantify hematopoietic cell generation, we explored the progeny of individual hESCs by plating individual TRA1–81+ H9 hES per well in OP9 stroma-coated plates, in the presence of VEGF and hematopoietic cytokines. Although SCR and shTET2 cells showed equal cloning efficiency on OP9 layer, the number of CD34+CD43+ hematopoietic cells generated by TET2 knockdown hESC was markedly reduced compared to SCR-infected cells (median 0.5% vs. 4.1%, p < .05) (Fig. 5D, 5E). Similar results were obtained with another ESC line (HUES1) (Fig. 5F).
Moreover, TET2 depletion inhibited blast colony (colony forming unit (CFU)-blast) formation by preventing the emergence of hemangioblasts, which are common precursors of hematopoietic and endothelial cells (Fig. 5G). Next, we investigated the clonogenic potential of the CD34+CD43+ cells after 10 and 14 days of OP9 coculture . At day 10, SCR CD34+CD43+ cells generated small clusters of White (megakaryocytes and monocytes) or red cells (EryP), characteristic of the first wave of hematopoiesis. TET2 depletion generated a lower number of white clusters and virtually no EryP. At day 14, shTET2 CD34+CD43+ cells generated fewer colonies from the second wave of hematopoiesis, including monocytes/macrophages (M), granulocytes (G), and both (GM) than SCR cells (Fig. 6A). These results were confirmed in two other ESC lines, HUES1 and HUES7 (Fig. 6B, 6C). The results were finally reproduced in the H9 hESC lines infected with shTET2.b even though the decrease in TET2 protein levels in these cells (30%) was less efficient than with shTET2 (Supporting Information Fig. S5D). The cell cycle and apoptosis analysis of hES-derived hematopoietic cells showed a 2.5-fold increase in Annexin V+ DAPI− cells and a decrease in cycling cells in shTET2 compared to SCR-infected cells (Fig. 6D, 6E).
hESC-derived hematopoiesis was analyzed by global gene expression profiling of CD34+CD43+ cells. Upon TET2 knockdown, there were 95 and 38 genes at day 10, and 118 and 224 genes at day 14 significantly upregulated or downregulated, respectively, The expression pattern of six genes (ANK1, CD36, S100A10, GDF3, FAM124B, and PF4) was confirmed by qRT-PCR (Supporting Information Fig. S5D). Using gene ontology term enrichment (DAVID), we observed that downregulated genes in CD34+CD43+ cells were related to hematopoietic functions and listed as “defense response,” “myeloid leukocyte activation,” and “inflammatory response” clearly reflecting the hematopoietic defect. Of note, we did not observe any change in master genes of hematopoiesis such as SCL/TAL1, RUNX1, GATA1/2, or LMO2. Up-regulated genes were involved in “cholesterol homeostasis” and associated with liver and endoderm differentiation, and in “lung development” and “neuron differentiation,” suggesting that TET2 depletion profoundly affected hESC fate (Supporting Information Table S3).
Since TET2 knockdown and/or knockout in human and mouse hematopoietic cells promote myelomonocytic differentiation and induce an increased pool of hematopoietic stem cells [4, 22, 29-31], we wanted to check the effect of TET2 knockdown directly in hematopoietic progenitors derived from hESC. Therefore, we transduced hematopoietic CD43+ cells at day 13 with shTET2 or SCR, sorted with GFP+ at day 15 and seeded in semisolid medium. We did not find a decrease in hematopoietic colonies but rather an augmentation of the number of progenitors (Fig. 6F). These data indicate that TET2 knockdown in primitive hematopoiesis results in the amplification of hematopoietic progenitors.
Previous studies have investigated the effect of Tet2 in mice including its role in pluripotent capacities and development using mESCs as well as mouse models [11, 12, 20, 22, 31]. The role of TET2 in the adult hematopoiesis has also been studied in both mouse and human cells [4, 22, 29, 31]. In this work, we investigated the role of TET2 during the spontaneous differentiation of hESCs into the different germ layers and during the development of hematopoiesis.
Similarly to previous studies with murine and hESCs [11, 12, 38, 45], we found that three different undifferentiated hESCs expressed high levels of TET1 compared to TET2 and TET3. In murine ESCs, TET2 was shown to be expressed at higher levels than TET3, while the TET2 transcript was the least abundant in hESCs [11, 12]. Spontaneous differentiation of the three ESC lines by EB formation was associated with a gradual increase in TET2 and TET3 transcripts but with sustained TET1 transcript levels. These data were also confirmed at protein levels. This observation differs from the reported decrease in Tet1 expression during murine ESCs differentiation  and suggests that TET1 regulates the appropriate differentiation of ESCs rather than their pluripotency [8, 12, 16, 18]. The divergences between murine and hESCs for the regulation of TET1 expression might be related to interspecies differences, their respective cellular origins (inner cell mass in mouse vs. epiblast in human) , and/or the culture conditions (hESC are dependent on FGF whereas mESC are cultured in the presence of leukemia inhibitory factor (LIF)) . We measured high levels of 5-hmC in undifferentiated ESCs and a reduction of 5-hmC during spontaneous differentiation, confirming other studies that reported the association between high levels of 5-hmC and the pluripotent state [11, 12]. This high level of 5-hmC in hESCs and IPSC is mostly mediated by TET1 and not by TET2 .
To assess the effect of TET2 knockdown on ES proliferation and pluripotency, we used a shRNA knockdown strategy. Similarly to mESCs, TET2 knockdown did not alter the pluripotent markers of hESCs [11, 12]. However, TET2 knockdown downregulated the meso-endoderm differentiation at the benefit of the neuroectoderm through changes in promoter methylation of key regulatory genes. These effects were observed both by spontaneous in vitro EB formation and by in vivo teratoma formation and were confirmed by two different TET2 shRNA (shTET2 and shTET2puro) in two different cell lines in vitro. Moreover, the shRNA-induced phenotype was rescued by reintroducing a wild-type TET2 form (TET2wt). Accordingly, TET2 deficiency in mESCs also induced a biased differentiation toward neuroectoderm . TET1 is thought to activate neurogenesis by increasing PAX6 or NEUROD1, thus TET1 and TET2 might play antagonist roles during neurogenesis . However, constitutive tet2 knockout in mice does not induce embryonic lethality [22, 30, 31]. In contrast, most tet1 and tet2 double knockout mice died perinatally and displayed a variety of malformations, such as exencephaly, hemorrhage in the head, or profound growth retardation . Moreover, tet1−/− tet2−/− mice that developed normally displayed a compensatory increase in TET3 expression. TET1 and TET2 gene expression differs in mouse and hESCs [16, 48], thus we cannot rule out interspecies differences. Moreover, it is possible that some epigenetic regulators may compensate for tet2 deficiency in vivo in mice. In addition, the consequences of TET2 depletion in cultured ESCs may be exaggerated compared to in vivo embryonic development, as in vitro differentiation relies on specification molecules such as retinoic acid, bone morphogenetic proteins (BMP), VEGF, and activin. These pathways may be fine-tuned by TET-regulated genes [8, 12, 16, 18].
Our data also show that NANOG overexpression could overcome TET2 knockdown-mediated increased neuroectoderm and decreased meso-endoderm differentiation. NANOG has been shown to play an important role in directing the differentiation of meso-endoderm [40, 41]. Some recent works have underscored that TET1/2 proteins could control NANOG expression by inducing interaction and regulating methylation of its promoter [11, 42]. Indeed, in shTET2 cells, we observed a strong methylation of the NANOG promoter as well as a decrease in H3K4me3, associated with the decreased transcription of NANOG. Moreover, interaction between TET2 and the promoter of NANOG decreased in shTET2 cells. Finally, our results are in line with the fact that tet1/2-deficient MEF failed to be reprogrammed into IPSC and displayed hypermethylation of the NANOG promoter [38, 42]. Altogether, our results confirm the role of TET2/NANOG axis in the differentiation properties of hESCs.
Our results underline the association between hematopoiesis and TET2. First, we showed that TET2 specifically increased in hematopoietic cells compared to TET1 and TET3 transcripts. Accordingly TET2 protein was expressed at lower level in ESCs than in hematopoietic cells. Second, using shTET2 ESCs, we present evidence that TET2 deficiency results in deregulation of lineage specification. Especially, the decrease emergence of mesoderm led to a subsequent defect in hematopoietic lineage as early as the hemangioblast step, leading to profound impairment of hematopoiesis. Particularly, TET2 knockdown induced a decreased emergence of hematopoietic progenitors (CD34+CD43+) in two cells lines and with two different TET2shRNA. Consequently, CD34+CD43+ cells from shTET2 ESCs were less able to give rise to hematopoietic cells (Ery-P, megakaryocytes, granulocytes, and monocytes) compared to control and were largely apoptotic. These results were also reproduced with three different TET2shRNA and in three different cell lines. This could be the consequences of an uncontrolled mesoderm differentiation since we detected an aberrant program of differentiation with high expression of lung, hepatic and neuronal genes in CD34+CD43+ progenitors cells. Surprisingly, we did not observe any change in master genes of hematopoiesis such as SCL/TAL1, RUNX1, GATA1/2, or LMO2 at the stage of differentiation analyzed (days 10–14). This result could be explained either by an earlier tight regulation of these genes during differentiation or by an indirect role of TET2 in their activities or by a direct control of TET2 on their specific target genes. In addition, we cannot exclude that the TET2 deficiency induces an upstream molecular defect altering the hematopoiesis by mainly decreasing its emergence. Whichever the case is knockdown of TET2 as early as the hESC state profoundly affects the differentiation and let the stigma of inappropriate genic expression that definitively impairs the hematopoietic progenitors. Altogether, these results suggest that TET2 affects the early stage of development and subsequent hematopoiesis. These results were rather unexpected as TET2 knockout in mouse hematopoietic cells promote myelomonocytic differentiation and induce an increased pool of hematopoietic stem cells that display higher self-renewal capacities [22, 30, 31]. An explanation might be interspecies differences between mouse and human with different TET gene expression [16, 49]. Alternatively, TET2 depletion could affect the transient and primitive hematopoiesis, shown during the differentiation of hESCs in vitro but could differentially affect the definitive hematopoiesis in mouse and in human, not explored in this study.
These data also contrast with the TET2 knockdown in human cells and the acquired loss-of-function mutations found in patients with hematological malignancies that drive progenitors amplification and monocytic differentiation at the expense of granulocytic differentiation [4, 24, 29]. However, the TET2 knockdown directly performed in hematopoietic progenitors (CD34+CD43+ cells) derived from hESCs led to an augmentation in the number of progenitors compared to control. Therefore, this result showed that it is possible to reproduce the role of TET2 in human adult hematopoiesis in this cellular model but at later stage of hES differentiation.
The fact that one patient carrying a heterozygous germ line mutations of TET2 in the C-terminal suffered from a myeloproliferative neoplasms  could be contradictory with our findings. However, the precise role of this mutation in the human development is not known. Thus, it will be important to establish IPSC from TET2 mutated-patients to further investigate the role of different TET2 mutations on the human development and hematopoietic differentiation.
In conclusion, our work suggests that during human embryonic development, TET2 is involved at different stages including induction of the mesoderm and hematopoietic differentiation. First, TET2 may favor meso-endodermic differentiation of hESCs, thereby promoting hemangioblast and hematopoietic differentiation. Second, TET2 could regulate the hematopoietic progenitors later during differentiation. This novel role for TET2 during early development will likely be important for further understanding the epigenetic mechanisms that might be deregulated in human hematopoietic malignancies.
We thank Olivia Bawa and Paule Opolon for the histopathological analysis of teratoma. We are grateful to Françoise Wendling for improving the manuscript. We thank Justine Guegan from the “Plateforme de génomique fonctionnelle” of Institut Gustave Roussy for all the microarrays and Philippe Rameau and Yann Lecluse from “Plateforme de cytométrie.” We also thank Elodie Pronier for technical development of 5-hmC measurement. This work was supported by grants from Agence Nationale de la Recherche (ANR-blanc Megon 2009 and thrombocytosis 2011; ANR-Blanc 2010 Epigenome) and from the Association pour la Recherche contre le Cancer (ARC) (ontogenèse et mégacaryopoïèse normale et pathologique, and ARC libre 2012) and the Ligue Nationale Contre le Cancer (Equipe labellisée). Labex GR-Ex (I.P., V.W.) is funded by the program “Investissements d'avenir.” B.C.R.M.M. is a research fellow from the Fondation de la recherche médicale. T.L. was supported by a fellowship from ANR-Megon. G.L. was supported by a postdoctoral fellowship from the Ile de France Cancéropôle and INCA. C.M. was funded by a postdoctoral fellowship from ANR Blanc and then supported by a labex GR-Ex fellowship. N.D. and W.V. are recipients of a research fellowship from H Bordeaux-INSERM, AP-HP-INSERM and IGR INSERM (contrats d'interface). I.P. was supported by grants from ARC (projet libre 2012). The work in the laboratory of KH was supported by the ERC and the Danish Cancer Society. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
W.V. and I.P.: conceived and designed the experiments, analyzed the data, and wrote the manuscript; B.D.C.R.M.-M., T.L., and G.L.: conceived and designed the experiments, performed the experiments, and analyzed the data. They equally contributed; F.F., C.A., J.C., and K.H.: contributed reagents/materials/analysis tools; C.M., J.P.L.C., and N.A.: performed the experiments; N.D.: performed the experiments and contributed experimental/intellectual input; O.B. and E.S.: wrote the manuscript and contributed experimental/intellectual input; H.R., F.D., and T.M.: contributed experimental/intellectual input. W.V. and I.P. contributed equally to this article.
Disclosure of Potential Conflicts of Interest
The authors indicate no potential conflicts of interest.