|Home | About | Journals | Submit | Contact Us | Français|
Transcription-factor-directed reprogramming from somatic cells to induced pluripotent stem cells (iPSCs) is by nature an epigenetic process of cell fate change. Previous studies have demonstrated that this inefficient process can be facilitated by the inclusion of additional factors. To gain insight into the reprogramming mechanism, we aimed to identify epigenetic enzymes capable of promoting iPSC generation. Here we show that Kdm2b, a histone H3 Lys 36 dimethyl (H3K36me2)-specific demethylase, has the capacity to promote iPSC generation. This capacity depends on its demethylase and DNA-binding activities, but is largely independent of its role in antagonizing senescence. Kdm2b functions at the beginning of the reprogramming process and enhances activation of early responsive genes in reprogramming. Kdm2b contributes to gene activation by binding to and demethylating the gene promoters. Our studies not only identify an important epigenetic factor for iPSC generation, but also reveal the molecular mechanism underlying how Kdm2b contributes to reprogramming.
Direct reprogramming from somatic cells to a pluripotent state can be achieved by the introduction of defined transcription factors, such as Oct4, Sox2, Klf4 and c-Myc (ref. 1). The resultant iPSCs are molecularly and functionally similar to embryonic stem cells (ESCs) derived from the inner cell mass of a blastocyst2,3. However, the process of iPSC generation is highly inefficient with a low frequency and a long latency before the establishment of pluripotency2,3. A number of factors have been shown to affect reprogramming efficiency through cell-cycle-dependent or -independent mechanisms2,4. For example, inhibition of the p53–p21 pathway and the Ink4a/Arf locus increases the reprogramming efficiency and accelerates the reprogramming dynamics by affecting cell proliferation5–10; whereas, ectopic expression of Nanog enhances reprogramming presumably through an epigenetic mechanism without changing the proliferation status6.
A recent study indicates that reprogramming from fibroblasts to iPSCs involves a series of transcriptional changes11. At the beginning, epithelial genes that alter the morphology of fibroblasts to an ESC-like state are first activated, followed by the activation of Nanog and other pluripotent factors. After these waves of activation, mesenchymal genes are repressed, followed by the activation of ‘mature’ pluripotent genes11. Consistent with these sequential molecular events, factors facilitating mesenchymal-to-epithelial transition, such as BMPs, Tgf-β inhibitors and microRNA-200s, have been shown to promote iPSC generation11–14, indicating that activation of an epithelial transcription program early in reprogramming is crucial for the establishment of pluripotency. However, how such a program is activated is unknown at present.
Given that cell fate reprogramming is essentially a resetting of epigenetic states15, it is not surprising that chemical inhibitors of epigenetic enzymes16–19, and certain chromatin-remodelling factors, are capable of promoting iPSC generation20. Through studying epigenetic factors specifically enriched in ESCs, we found that Kdm2b (also known as Jhdm1b and Fbxl10), an H3K36me2-specific demethylase21,22, is able to facilitate iPSC generation. This property is independent of its effect on cell proliferation, but relies on its demethylase and DNA-binding activities that contribute to the activation of the reprogramming transcription program in the early stage.
To identify epigenetic factors that facilitate iPSC generation, we focused on the epigenetic factors enriched in ESCs. We found that isoform 1 (IF1) of Kdm2b (Supplementary Fig. S1) is highly expressed in ESCs (Fig. 1a) and progressively upregulated during reprogramming by Oct4, Sox2 and Klf4 (OSK; Supplementary Fig. S2a). To explore its potential role in iPSC generation, we constructed a doxycycline-inducible lentiviral plasmid expressing carboxy-terminal Flag-tagged Kdm2b-IF1 (hereafter referred to as Kdm2b). On doxycyclin induction, Kdm2b is expressed more than 50 times above the ESC level (Supplementary Fig. S2b). When introduced into Oct4–IRES–GFP/Rosa26-M2rtTA mouse embryonic fibroblasts (MEFs) together with the retroviral factors OSK, it can increase Oct4–GFP+ colony numbers by 4–6-fold (Fig. 1b). Furthermore, when c-Myc is included in the factor cocktail (OSKM), Kdm2b also increases Oct4–GFP+ colony numbers (Fig. 1c). Together, these data show that Kdm2b is capable of enhancing OSK- and OSKM-mediated iPSC generation.
By following the reprogramming kinetics, we found that although Kdm2b promotes iPSC generation, it does not significantly shorten the latency time, as the first Oct4–GFP+ colony appears around day 10 (OSK) or day 8 (OSKM) of reprogramming regardless of the presence of Kdm2b (Fig. 1d). This observation differs from the documented cell-cycle-based enhancement of iPSC generation, including depletion of p53 and p21 (refs 6,8,9) and inhibition of Ink4a/Arf (refs 5,7), which significantly shorten the latency time until the first appearance of iPSCs, but it is similar to the cell-cycle-independent enhancement by Nanog overexpression6.
To further confirm the role of Kdm2b in reprogramming, we depleted Kdm2b by small hairpin RNA (shRNA; ref. 21; Fig. 1e) and induced reprogramming with OSK using the Kdm2b-depleted cells. Kdm2b depletion greatly reduces the reprogramming efficiency (Fig. 1f), indicating that Kdm2b is required for optimal induction of iPSCs. Collectively, these studies demonstrate that Kdm2b facilitates iPSC generation.
Next, we set out to characterize the iPSCs generated with OSK in the presence of exogenous Kdm2b. After 16–18 days of induction, Oct4–GFP+ colonies were picked up and propagated in the absence of doxycyclin. The resulting cell lines exhibit typical ESC morphology with Oct4–GFP expression (Fig. 2a), show alkaline phosphatase activity and express ESC markers SSEA-1 and pluripotent transcription factors Nanog and Sox2 (Fig. 2b). The endogenous loci of Oct4, Sox2 and Nanog in the iPSCs are activated to a level similar to that in ESCs (Fig. 2c); whereas the transgenes are mostly silenced (Fig. 2d). When implanted into immunodeficient mice, these iPSCs generate teratomas with tissues belonging to three germ layers (Fig. 2e). Importantly, they are competent for chimaera generation when injected into blastocysts (Fig. 2f). These data support that the iPSCs generated in the presence of exogenous Kdm2b are pluripotent.
To understand how Kdm2b facilitates iPSC generation, we attempted to determine the domains of Kdm2b important for this property. To this end, we carried out reprogramming with point mutations in the catalytic JmjC domain21,22 and the DNA-binding CXXC-type zinc finger (ZF) domain23,24. We first confirmed that both the mutants and wild-type (WT) Kdm2b are expressed at similar levels on induction (Fig. 3a). Consistent with the enzymatic activity of Kdm2b, overexpression of the WT Kdm2b and the ZF mutant, but not the catalytic mutant, leads to a specific decrease of the H3K36me2 level (Fig. 3a). Importantly, mutations in either JmjC or ZF abrogate the capacity of Kdm2b in promoting iPSC generation (Fig. 3b), indicating that both the demethylase and DNA-binding activities of Kdm2b are essential for reprogramming enhancement.
Kdm2b is able to promote cell proliferation by repressing senescence21,25,26. To determine whether the capacity of Kdm2b in enhancing reprogramming is mediated by promoting cell proliferation, we examined whether ectopic expression of Kdm2b affects cell proliferation in the context of OSK-induced reprogramming. We found that although the ZF domain of Kdm2b is critical for promoting iPSC generation (Fig. 3b), it is not required for promoting cell proliferation (Fig. 3c), indicating that the ability of Kdm2b to enhance proliferation is not sufficient for promoting iPSC generation. To separate its proliferation effect from the reprogramming effect, we normalized the reprogramming efficiency by dividing the Oct4–GFP+ colony numbers with the total cell numbers in the reprogramming populations. We found that exogenous WT Kdm2b increases the reprogramming efficiency by more than fourfold after normalization (Fig. 3d). These analyses not only confirm that the demethylase activity and DNA-binding capacity of Kdm2b are necessary for enhancing iPSC generation, but also indicate that cell proliferation stimulated by Kdm2b is not a major contributing factor for its role in promoting reprogramming.
To further ascertain that Kdm2b-mediated suppression of senescence is not a major contributor to its role in reprogramming enhancement, we examined the transcript levels of Ink4a, Arf and Ink4b during reprogramming involving exogenous Kdm2b. We found that Ink4a, but not Arf or Ink4b, is significantly downregulated by Kdm2b in the first 12 days of OSK reprogramming (Fig. 4a). However, the protein level of Ink4a is less affected by Kdm2b (Fig. 4b), probably owing to the stability of this protein. To examine whether repression of this locus contributes to the effect of Kdm2b on reprogramming, we introduced shRNA that depletes Ink4a and Arf (Fig. 4c,d) to the reprogramming cells transduced with OSK and OSK plus Kdm2b. We found that Kdm2b was still able to increase the Oct4–GFP+ colony number following the depletion of Ink4a/Arf (Fig. 4e), indicating that the effect of Kdm2b on reprogramming is largely independent of its role in downregulating Ink4a/Arf. Consistent with previous findings that senescence suppression promotes reprogramming5,7,10, we also observed a great enhancement of reprogramming efficiency when Ink4a/Arf are knocked down regardless of whether Kdm2b is introduced (Fig. 4e). Collectively, our data indicate that the ability of Kdm2b to suppress cellular senescence is not the main reason for its role in enhancing reprogramming.
Given that the role of Kdm2b in suppressing senescence and/or promoting cell proliferation is not a main contributing factor, and that introduction of Kdm2b does not enhance expression of transduced reprogramming factors (Supplementary Fig. S3), we searched for alternative mechanisms. As reprogramming can be divided into stages with distinct molecular features4,11,27,28 and factors that promote iPSC generation can function at different stages during reprogramming14,29,30, we sought to determine the time window in which Kdm2b enhances iPSC generation. To this end, we induced Kdm2b expression for different durations and found that enforced expression of Kdm2b in the first 8 days constantly enhanced the reprogramming efficiency (Fig. 5a, upper panels). As a complement, we also induced Kdm2b expression beginning at different days after the initiation of reprogramming and found that the increase in reprogramming efficiency is largely proportional to the length of the doxycyclin treatment (Fig. 5a, lower panels). Together, these studies indicate that Kdm2b promotes iPSC generation by functioning early during the reprogramming process.
To understand the effect of Kdm2b during reprogramming at the molecular level, we performed microarray analyses on cells collected at days 4, 8 and 12 during the OSK reprogramming in the presence or absence of Kdm2b. These analyses revealed that 418 probes are upregulated at least twofold in at least one of the three time points, and 143 probes show at least twofold downregulation (Supplementary Table S1). Hierarchical analysis revealed that the Kdm2b-upregulated genes can be grouped into three distinct clusters (I, II and III) on the basis of their activation timings (Fig. 5b), whereas Kdm2b-downregulated genes are distributed more ambiguously in terms of affected timings.
Gene ontology analysis revealed that Kdm2b-affected genes are exceptionally enriched for adhesion processes (P value = 10−10 ~ 10−12). Other terms with a significant P value (<10−3) include those related to cell morphology, development and epithelium-related processes (Fig. 5c and Supplementary Table S2). By comparing Kdm2b-affected genes (Fig. 5b) with a list of ‘signature’ reprogramming genes that are dynamically regulated in reprogramming11, we found that the former include early-activated epithelial genes (Cdh1, Cldn3, −4 and −7, Epcam, Esrp1 and Ocln) and pluripotent genes (Nanog, Dppa5a and Tdgf1; Fig. 5d). Indeed, all of the previously identified early-activated genes11 have enhanced expression in the presence of Kdm2b, whereas the expression of late-activated genes or repressed genes is not significantly altered (Supplementary Fig. S4). Combined with the fact that Kdm2b functions from the beginning of reprogramming (Fig. 5a), Kdm2b seems to facilitate activation of early responsive genes during reprogramming.
To further dissect the expression changes, we grouped the Kdm2b-affected genes on the basis of the timings of their expression changes (Supplementary Table S1). We found that most of the day 4 or day 8 affected genes are also affected at later time points, whereas a substantial portion of day 8 and/or day 12 affected genes are not affected at an earlier time point (Fig. 5e). Such a distribution pattern prompts us to reason that the introduction of Kdm2b probably amplifies a reprogramming transcriptional cascade. Moreover, gene ontology analyses revealed that genes upregulated from day 4 through 12 are enriched in adhesion molecules, whereas those exclusively upregulated at one or two time points seem to undergo a functional transition from cell adhesion to development processes (Fig. 5f and Supplementary Table S3). This observation indicates that a developmental-related transcription program is triggered sometime after an initial adhesion-related program, and these two programs may constitute a transcription cascade. Given that the developmental program seems to start at day 8 (Fig. 5f) and Kdm2b exerts its effect from the beginning of reprogramming (Fig. 5a), it is likely that Kdm2b directly contributes to the earlier adhesion-related program, whereas its effect on the later developmental program is the result of an amplified cascade.
Quantitative PCR with reverse transcription (RT–qPCR) analysis confirmed the upregulation of genes encoding epithelial markers, such as Cdh1 (also known as E-cadherin), Crb3 and Epcam, as well as desmosomal components Dsg2 and Dsp (Fig. 6a–c). Enhanced expression of these genes starts at day 4 (Fig. 6b), and is affected more obviously at days 8 and 12 (Fig. 6a,c). We also note that genes activated earliest at day 4 include those encoding transcription factors implicated in cell adhesion and development, including Irf6, which is important for cleft palate development31,32, and Insm1, which is important for pancreatic and neuronal development33–35. Interestingly, these genes are upregulated by OSK and introduction of Kdm2b further augments their activations (Fig. 6a,b). The potential function of these early responsive genes in reprogramming remains to be analysed. Furthermore, the expression levels of pluripotent genes, such as Nanog and Tdgf1, are also enhanced by Kdm2b at days 8 and 12, but not at day 4 (Fig. 6a,b), indicating that their activation might be an indirect effect of Kdm2b.
We also examined the expression of mesenchymal-specific transcription factors, which are shown to be downregulated following the activation of Nanog and other pluripotent genes11. We observed a mild downregulation of Snai1, Snai2, Zeb1 and Zeb2 at days 8 and/or day 12 (Supplementary Fig. S5a), although their downregulation is less than twofold and consequently not picked up by the microarray analyses (Fig. 5d). Finally, we did not observe activation of endogenous Oct4 and Sox2 during the first 12 days of reprogramming (Supplementary Fig. S5b), consistent with the fact that even in the case of Kdm2b-assisted reprogramming, only a very small fraction of the starting cells gained pluripotency in the initial 12 days of reprogramming (Fig. 1b). Collectively, the data presented above indicate that Kdm2b facilitates activation of genes related to epithelial adhesion, which may in turn activate downstream genes including pluripotent genes. Thus, it seems sensible that Kdm2b promotes iPSC generation by facilitating the initiation of a putative transcription cascade.
As mutations in the JmjC domain and the ZF domain abrogate the capacity of Kdm2b in promoting iPSC generation, we investigated whether these mutations affect the capacity of Kdm2b to augment the gene activation in reprogramming. RT–qPCR and western blot analysis demonstrate that both mutations abrogate the ability of Kdm2b to activate Cdh1 and Epcam (Fig. 6d,e), as well as the subsequent activation of Nanog (Fig. 6d), consistent with the notion that Kdm2b enhances iPSC generation by facilitating early gene activation.
To understand how Kdm2b facilitates early gene activation in reprogramming, we investigated whether Kdm2b activates these genes alone or in concert with the key reprogramming factors. To this end, we introduced individual factors and different factor combinations into MEFs and examined their effects on the expression of early responsive epithelial genes, such as Cdh1, Crb3 and Epcam, as well as the later activated Nanog. We found that, overall when any of the OSK factors is omitted, activation of these genes is either greatly compromised or completely abolished despite the presence of Kdm2b (Fig. 6f). This observation is consistent with the fact that Kdm2b cannot replace any of the OSK in iPSC generation (data not shown). However, we note that Kdm2b alone does exhibit an about fivefold activation on the Cdh1 gene when compared with non-transduced MEF cells. Nevertheless, this activation seems to be minor (~5%) when compared with the activation by OSK (Fig. 6f). Consistent with a previous report13, transduction of Klf4 alone can partially activate Cdh1 when compared with OSK transduction. Its activation can be further boosted when Kdm2b is combined with Klf4 (Fig. 6f). For Crb3, Epcam and Nanog, Kdm2b barely activates them in the absence of any of the OSK (Fig. 6f). On the basis of these results, we conclude that Kdm2b acts in concert with the key reprogramming factors OSK to upregulate early responsive genes.
To determine whether Kdm2b directly contributes to the activation of the early responsive genes, we investigated whether Kdm2b binds to these genes. Taking advantage of the Flag epitope on the Kdm2b constructs, we performed chromatin immunoprecipitation (ChIP) analysis using cells at day 4 of reprogramming. This analysis demonstrates that Flag–Kdm2b is enriched in the promoter of the early-activated genes, including Cdh1, Epcam, Dsg2, Dsp and Irf6; however, no enrichment is detected at the Nanog promoter, consistent with the notion that Kdm2b contributes to Nanog activation indirectly (Fig. 7a). ChIP analysis also revealed the presence of exogenous Kdm2b at the promoter and part of the gene body of Cdh1 (Fig. 7b,c). Parallel ChIP experiments demonstrate that the level of H3K36me2 at the promoter of Cdh1, Dsp and Irf6 is decreased on the introduction of Kdm2b (Fig. 7d), consistent with the fact that Kdm2b preferentially removes H3K36me2 (ref. 21). These results support the notion that Kdm2b contributes to the activation of early responsive genes by binding to and demethylating H3K36me2 on their promoters.
Epithelial genes, which are activated early11,13 and whose activation is enhanced by Kdm2b during reprogramming, are subjected to repression by Tgf-β signalling36. To demonstrate that increased expression of these genes contributes to Kdm2b-enhanced reprogramming, we performed reprogramming using OSK plus Kdm2b in the presence of Tgf-β. We found that, Kdm2b-mediated upregulation of epithelial genes, such as Cdh1 and Epcam, is abrogated on Tgf-β treatment (Fig. 8a). Importantly, Kdm2b-enhanced iPSC generation is also abrogated by Tgf-β (Fig. 8b), consistent with the notion that Kdm2b-enhanced expression of the epithelial genes contributes to its effect on reprogramming. To directly address the role of these epithelial genes in mediating Kdm2b-enhanced reprogramming, we focused on Cdh1, one of the epithelial genes important for reprogramming13,37,38 and directly regulated by Kdm2b. We found that depletion of Cdh1 by shRNAs (Fig. 8c,d) greatly compromised the capacity of Kdm2b to enhance OSK reprogramming (Fig. 8e), indicating that Cdh1 is one of the downstream targets that mediate the Kdm2b reprogramming effect.
Despite the fact that cellular senescence is a ‘roadblock’ for reprogramming5,7 and that Kdm2b is capable of suppressing senescence21,25, we demonstrate here that Kdm2b promotes iPSC generation largely independently of its role in senescence and/or proliferation. First, although Kdm2b increases cell proliferation during reprogramming (Fig. 3c), when the proliferation effect is subtracted, Kdm2b still exhibits a more than fourfold increase in reprogramming efficiency (Fig. 3d). Second, although an intact ZF domain is required for Kdm2b to promote iPSC generation, mutation on this domain does not alter its ability to stimulate cell proliferation (Fig. 3b–d), indicating that the roles of Kdm2b in promoting proliferation and iPSC generation are independent and separable. Third, when Ink4a/Arf, the key senescence regulators and documented targets of Kdm2b (ref. 25), are depleted, Kdm2b is still capable of promoting iPSC generation (Fig. 4e), indicating that antagonizing senescence is not a major contributor for Kdm2b-mediated reprogramming enhancement. Last, Kdm2b does not significantly shorten the latency time from induction until the first appearance of iPSCs (Fig. 1d), which is consistent with the reprogramming kinetics of cell-cycle-independent enhancement of reprogramming6.
While our manuscript was under review, a report was published demonstrating that Kdm2b greatly enhances iPSC generation in the presence of vitamin C and indicating that the effect of Kdm2b may be mediated through promoting cell cycle progression and activation of microRNA cluster 302/367 (ref. 39). However, as most of the experiments in this report were performed in the presence of both vitamin C and Kdm2b, the specific contribution of Kdm2b to iPSC generation is not addressed. Indeed, the same group previously found that vitamin C is able to alleviate the blockage of cell cycle progression imposed by p53–p21 (ref. 40). As vitamin C has a more pronounced effect on iPSC generation than Kdm2b (ref. 39), it is possible that the effect of vitamin C on cell cycle progression might mask the effect of Kdm2b and consequently the contribution of Kdm2b to iPSC generation might be overshadowed.
By analysing the gene expression changes within or immediately after the functioning time window of Kdm2b, we found that exogenous Kdm2b enhances the expression of a set of early-activated genes during reprogramming (Fig. 5d and Supplementary Fig. S4). The enhanced expression of epithelial genes (Cdh1, Epcam and so on) and other uncharacterized genes (Irf6 and Insm1) takes place in the functioning time window of Kdm2b (Figs 5a and 6a,b). Following this first wave of activation, the expression of Nanog and other pluripotency factors is upregulated (Fig. 6a,b) concomitant with enrichment of developmental genes in the upregulated gene group (Fig. 5f). Meanwhile, some mesenchymal genes, whose downregulation follows Nanog activation11, begin to exhibit magnified downregulation in the presence of Kdm2b (Supplementary Fig. S5a). Soon after Nanog activation (day 8; Fig. 6a), the first Oct4–GFP+ colonies are observed at day 10 (Fig. 1a). The observation that Kdm2b amplifies these sequential transcription events prompted us to propose a transcription cascade model to explain how Kdm2b might contribute to reprogramming (Fig. 8e). We propose that Kdm2b facilitates initial gene activation that occurs on the epithelial genes, causing an amplified transcription cascade, which in turn enhances the activation of pluripotent genes such as Nanog, eventually resulting in an increase in reprogramming efficiency (Fig. 8e).
This model is consistent with previous observations that Tgf-β inhibitors induce Nanog expression and enhance reprogramming12,14, as Tgf-β signalling inhibits epithelial gene expression13,36 and suppression of Tgf-β potentially enhances the activation of epithelial genes. It also agrees with the notion that Nanog is not required for initiating reprogramming but plays a key role in driving the ‘pre-iPSCs’ to pluripotency30. Uncovering the potential links between transcription events is crucial for attesting the transcription cascade model.
The results of our study indicate that Kdm2b contributes to iPSC generation by facilitating activation of early responsive genes. We found that the ability of Kdm2b in promoting gene activation depends on OSK (Fig. 6f), supporting an axillary role for Kdm2b in enhancing gene expression. Furthermore, we demonstrate that Kdm2b binds to the promoters of early-activated genes and maintains a low level of H3K36me2 at these loci (Fig. 7a,c), indicating that Kdm2b directly participates in the regulation of these genes. A recent study indicates that depletion of H3K36me2 is a feature of CpG island promoters23. It is possible that Kdm2b-mediated removal of H3K36me2 at promoters facilitates OSK-directed gene activation by creating a chromatin environment favourable for cofactor recruitment. Such a scenario explains why Kdm2b can promote gene activation (Fig. 6) and facilitate reprogramming by OSK (Fig. 1) but fails to substitute for any of OSK (data not shown). Future studies should reveal the molecular mechanisms underlying how demethylation of H3K36me2 contributes to gene activation.
Mouse Kdm2b (isoform 1) was amplified from complementary DNA, fused with a C-terminal Flag tag and cloned into a doxycyclin-inducible lentiviral vector pTYF-TRE. The JmjC (H211A, D213A) and ZF (C573A C576A C579A) mutants were constructed by mutagenesis PCR and confirmed by sequencing. Target sequences of control shRNA and shRNAs against Kdm2b (ref. 21), Ink4a/Arf and Cdh1 are described in Supplementary Table S4. These shRNAs were expressed in lentiviral plasmid pTY-U6-Pgk-Puro. Lentivirus was prepared by co-transfection of pTY/pTYF plasmids with pHP, pHEF1α-VSVG and pCEP4-Tat into 293T cells, and collected at 24, 36 and 48 h after transfection. Viral supernatant was filtered through a 0:45-µm membrane and concentrated by a spin column before being applied to MEFs. Retroviral plasmids pMXs-Oct4, Sox2, Klf4 and c-Myc were obtained from Addgene, and retrovirus was prepared as previously described18.
MEFs for iPSC generation were prepared from E13.5 embryos of Oct4–IRES–GFP/Rosa26-M2rtTA double knock-in mice. To derive iPSCs, MEFs at the first 2 passages were seeded onto 6-well plates at a density of 1 × 105 cells per well, 16 h before viral infection. Two doses of retrovirus and/or one dose of lentivirus was applied within 48 h in the presence of Polybrene (10 µg ml−1). At 24 h after the second retroviral transduction, the virus supernatant was withdrawn and the day was designated as day 0 post-transduction. Subsequently, iPSCs were induced for 12–18 days in mouse ESC medium (DMEM with 15% FBS, non-essential amino acid, GlutaMax, sodium pyruvate, β-mercaptoethanol, penicillin/streptomycin and 1,000 U ml−1 leukaemia inhibitory factors) in the presence of doxycyclin (1 µg ml−1). If indicated, Tgf-β treatment was carried out by applying Tgf-β1 (R&D Systems) at 2 ng ml−1. Oct4–GFP+ colonies were counted on selected days from day 6 to day 18. Reprogramming efficiency was presented as the number of Oct4–GFP+ colonies derived from 1 × 105 MEFs. Relative reprogramming efficiency over a control induction is also used in some cases. At day 18, Oct4–GFP+ colonies were manually picked, trypsinized and seeded onto mitomycin-C-treated feeder MEF cells. The derived iPSC lines were propagated in mouse ESC medium in the absence of doxycyclin for at least 8 passages before being characterized.
For immunofluorescent staining, antibodies against SSEA-1 (Chemicon mAB4301, clone MC-480), Nanog (Bentyl, IHC-00205) and Sox2 (Millipore, AB5603) were applied at a concentration of 1:500, 1:250 and 1:1,000, respectively. Alkaline phosphatase staining was carried out with an alkaline phosphatase detection kit (Millipore). Teratoma analysis was performed as previously reported18. For generation of chimaeric mice, 12-week-old Albino B6 (C57Bl/6J-Tyr < c-2J >) female mice were stimulated to superovulation by injection with pregnant mare serum gonadotropin (2.5 IU) followed by administration of human chorionic gonadotropin (5 IU) 47 h later. The female mice were subsequently mated with Albino B6 stud males and blastocysts were collected on gestation day 3.5. On the day of microinjection, iPSCs (line 8 and 17) were rinsed twice with PBS, dissociated with 0.05% trypsin, washed once with and then resuspended in Knockout DMEM supplemented with 15% FBS. Each blastocyst was injected with 10–15 iPSCs using a piezo impact micromanipulator. Injected embryos were then implanted into the uterus of pseudopregnant Swiss Webster recipient females.
Quantitative and semi-quantitative RT–PCR was carried out using primers in Supplementary Table S5. Western blotting was performed using antibodies against Arf (Santa Cruz sc-32748, clone 5-C3-1, 1:200), Cdh1 (Cell Signaling 3195, clone 24E10, 1:1,000), Epcam (Abcam ab32392, clone E144, 1:500), Flag (Sigma F1804, clone M2, 1:5,000), histone H3 (Abcam ab1791, 1:5,000), H3K4me2 (Active Motif 39141, 1:1,000), H3K36me1 (Abcam ab9048, 1:1,000), H3K36me2 (ref. 22; 1:1,000), H3K36me3 (Abcam ab9050, 1:1,000), Ink4a (Santa Cruz sc-1207, 1:200) and α-tubulin (Sigma T6199, clone DM1A, 1:2,000).
RNA samples were extracted from cells transduced with OSK or OSK plus Kdm2b at post-transduction day 4, 8 and 12. The reverse transcription and hybridization procedure was carried out as previously described18. The microarray data were analysed with GeneSpring software and are available in the ArrayExpress database (http://www.ebi.ac.uk/arrayexpress/) with accession number E-MEXP-3433.
Cells transduced with OSK plus Flag-tagged Kdm2b were collected for ChIP at post-transduction day 4. ChIP was performed using Imprint ChIP Kit (Sigma) according to the manufacturer’s instruction. Chromatin was prepared by sonication at 4 °C on a Bioruptor 300 (Diagenode) with a high magnitude for 10 cycles with 30 s on and 30 s off. For each precipitation reaction, chromatin from 2 × 105 cells was applied to a Stripwell plate pre-bound with antibodies against Flag (Sigma F1804, clone M2), H3K36me2 (ref. 22), H3K4me2 (Active Motif 39141) or mouse IgG. If necessary, immunoprecipitated and purified DNA fragments were subjected to amplification using a Whole Genome Amplification Kit (Sigma). Immunoprecipitated or amplified DNA was analysed by qPCR using primers listed in Supplementary Table S6.
We thank UNC Animal Models Core for chimaera generation and UNC Functional Genomic Core for microarray analysis. We thank S. Yamaguchi and K-H. Hong for helpful discussions. This work is supported by U01DK089565 from the NIH. Y.Z. is an Investigator of the Howard Hughes Medical Institute.
Note: Supplementary Information is available on the Nature Cell Biology website
AUTHOR CONTRIBUTIONSG.L. and Y.Z. designed all of the experiments and wrote the manuscript. G.L. performed most of the experiments. J.H. constructed the Kdm2b plasmids and lentiviral doxycyclin inducible system.
COMPETING FINANCIAL INTERESTS
The authors declare no competing financial interests.