Notch inhibition allows oncogene-independent generation of iPS cells.

The reprogramming of somatic cells to pluripotency using defined transcription factors holds great promise for biomedicine. However, human reprogramming remains inefficient and relies either on the use of the potentially dangerous oncogenes KLF4 and CMYC or the genetic inhibition of the tumor suppressor gene p53. We hypothesized that inhibition of signal transduction pathways that promote differentiation of the target somatic cells during development might relieve the requirement for non-core pluripotency factors during induced pluripotent stem cell (iPSC) reprogramming. Here, we show that inhibition of Notch greatly improves the efficiency of iPSC generation from mouse and human keratinocytes by suppressing p21 in a p53-independent manner and thereby enriching for undifferentiated cells capable of long-term self-renewal. Pharmacological inhibition of Notch enabled routine production of human iPSCs without KLF4 and CMYC while leaving p53 activity intact. Thus, restricting the development of somatic cells by altering intercellular communication enables the production of safer human iPSCs.

Use of the potent oncogenes KLF4 and CMYC in the generation of induced pluripotent stem cells (iPSCs) limits their translational utility [1,2]. Currently, elimination of these genes during human iPSC reprogramming requires suppression of p53 activity [2-16], which in turn results in the accumulation of genetic mutations in the resulting iPSCs [8]. Therefore, there remains a real need for reprogramming approaches that enable iPSC generation without the use of KLF4 and CMYC while leaving p53 activity intact.

In part to address this need, several groups have undertaken chemical screens to identify small molecules that can improve reprogramming [17-21]. Thus far, the majority of active compounds are thought to improve reprogramming by inhibiting chromatin-modifying enzymes or by reinforcing the transcriptional network associated with the pluripotent state [17-22]. Consistent with their proposed mechanisms of action, these chemicals generally function in cellular intermediates that arise late in reprogramming, catalyzing their final conversion into iPSCs [19,22]. It is currently unclear whether known chemicals are sufficient for generating iPSCs from adult human cells, which are consistently more difficult to reprogram than mouse embryonic fibroblasts [23].

Given the likely need for additional reprogramming chemicals and the knowledge that most known compounds act late in this process, we reasoned it would be valuable to identify small molecules that improve reprogramming by acting early, perhaps within the somatic cells themselves. We reasoned that one approach towards this goal would be to identify chemicals that could modulate signal transduction cascades in somatic cell populations to enrich for those cells with an enhanced capacity for reprogramming. We reasoned that if such compounds could be identified, they might expand the translational utility of chemical reprogramming.

It has been recognized that the extent of a target cell's differentiation is an important determinant of the efficiency by which it can be reprogrammed [24-26]. We therefore hypothesized that chemically driving somatic cells into a more potent “stem cell” state might improve their reprogramming. To test this hypothesis, we chose to ask whether known chemical inhibitors of the Notch signaling pathway could aid in reprogramming.

The Notch signaling pathway is highly conserved and regulates the proliferation and differentiation of many distinct progenitor cell and stem cell types [27]. Notch ligands are generally transmembrane proteins that require contact between two cells in order to mediate signal transduction [28]. In skin, Notch promotes differentiation by directly activating p21 expression, which in turn blocks proliferation and induces the differentiation of keratinocyte stem cell populations [29,30]. We therefore hypothesized that inhibition of Notch in keratinocytes might enhance iPSC generation by inhibiting differentiation and enriching more easily reprogrammed progenitor cells. We also felt that keratinocytes were an attractive model for testing our hypothesis because if Notch inhibition did have an effect, it could be immediately translated to the production of patient-specific iPSCs [31,32].

Here, we show that Notch inhibition significantly improves the efficiency of iPSC generation from mouse and human keratinocytes by suppressing p21 and thereby enriching undifferentiated cells with increased reprogramming potential. In addition, pharmacological inhibition of Notch enabled the efficient production of human iPSCs without KLF4 and CMYC while leaving p53 activity intact, resulting in the production of safer human iPSCs.

Results

DAPT treatment promotes keratinocyte reprogramming

Notch signaling is activated by the γ-secretase complex, which cleaves the membrane-tethered Notch receptor upon ligand binding and generates a free intracellular domain that can translocate to the nucleus and modulate transcription [27]. It has previously been shown that the γ-secretase inhibitor DAPT (Fig. 1a) can block Notch signaling in mouse keratinocytes [33]. As expected, 10 μM DAPT treatment of both neonatal mouse and human keratinocytes transduced with the iPSC reprogramming factors increased abundance of the full-length Notch receptor, reduced levels of cleaved Notch intracellular domain (NICD) (Supplementary Results, Supplementary Fig. 1a), and decreased expression of the Notch-target genes Hey1, Hes1, Hes5, and Col6a1 (Supplementary Fig. 1b).

Figure 1

DAPT treatment promotes mouse and human keratinocyte reprogramming

a, Chemical structure of DAPT. b, The efficiency of iPSC generation from mouse and human keratinocytes transduced with Oct4, Sox2, Klf4, and cMyc with DMSO or DAPT treatment (DAPT used at 10 μM in mouse experiment). c, The efficiency of iPSC generation from mouse and human keratinocytes transduced with all combinations of 2 reprogramming factors with DMSO or 2.5 μM DAPT treatment from days 1-18 post-transduction. d, A P0 mouse and human iPSC colony generated using OCT4, SOX2, and DAPT, scale bars = 100 μm. g, Teratoma generated by iPSCs derived from human neonatal kerationcytes using OCT4, SOX2, and DAPT, scale bar = 50 μm. j, NANOG+/TRA-1-81+ iPSC line generated from human adult keratinocytes using OCT4, SOX2 + DAPT, scale bars = 100 μm. For all experiments, error bars represent the standard deviation between two or three biological replicates and statistical significance was determined using a two-tailed homoscedastic Student's t-test.

To determine whether inhibition of Notch could increase the efficiency of reprogramming, we transduced Oct4::GFP mouse or human keratinocytes with Oct4, Sox2, Klf4, and cMyc and cultured the resulting cells for 25 days either in the presence or absence of DAPT. We found that the addition of 10 μM DAPT led to a significant, 4-fold increase in the number of resulting Oct4::GFP+ mouse and NANOG+/TRA-1-81+ human iPSC colonies (Fig. 1b).

We wondered whether this increase in reprogramming activity might allow the generation of iPSCs from keratinocytes without Klf4 and cMyc. Indeed, although transduction of Oct4 and Sox2 alone were not sufficient to induce keratinocyte reprogramming, Oct4 and Sox2 combined with DAPT treatment routinely yielded mouse and human iPSC colonies (Figs. 1c,d and Supplementary Fig. 2a). This effect was specific to Oct4 and Sox2-transduced cells because other 2-factor combinations did not yield iPSCs in the presence of DAPT (Fig. 1c).

To determine whether these putative iPS cell lines were pluripotent, we subjected them to a “scorecard” assay for pluripotency that we recently developed [34]. We found that these cell lines were indeed composed of pluripotent cells and that they performed comparably to human embryonic stem cells (ESCs) in their expression of pluripotency-associated genes and differentiation propensities (Supplementary Figs 2b, c). To further confirm their differentiation capacity, we also injected the OCT4, SOX2 + DAPT human cells into immunocompromised mice and found that they readily formed teratomas containing differentiated cells (Fig. 1e). Moreover, when injected into blastocysts, the Oct4, Sox2 + DAPT mouse cells contributed to the development of chimeric mice (Supplementary Fig. 2d), including the germ-line (Supplementary Fig. 2e).

Many applications of iPS cells would require the DAPT-dependent generation of KLF4 and CMYC-free iPSCs from adult keratinocytes. Therefore, we determined if DAPT treatment increased the reprogramming potential of adult human keratinocytes. As with mouse and human neonatal keratinocytes, we found that DAPT treatment of KLF4, SOX2, OCT4, and CMYC-transduced adult human keratinocytes significantly improved their rate of reprogramming (Supplementary Fig. 2f) and also enabled the generation of iPSCs with just OCT4 and SOX2 (Fig. 1f, Supplementary Fig. 2g). The scorecard assay again verified that these 2-factor iPSCs were pluripotent (Supplementary Figs 2b, c). Together, these results demonstrate that DAPT reliably enables the generation of bona fide mouse and human iPSCs from keratinocytes without KLF4 and CMYC.

Notch inhibition promotes reprogramming

Our results thus far suggest that antagonizing Notch signaling in keratinocytes may promote their conversion into iPSCs. To begin verifying that NOTCH was indeed the functional target of DAPT during reprogramming, we tested a structurally distinct γ-secretase inhibitor, DBZ [35](Fig. 2a), for activity in iPSC generation. When we treated human keratinocytes with DBZ, we observed significant reductions in the levels of the intracellular domain of the NOTCH receptor (Supplementary Fig.1a) and the NOTCH-dependent genes HES1 and HES5 (Supplementary Fig. 3a), indicating that DBZ administration inhibited NOTCH signaling. Consistent with the notion that NOTCH inhibition increases the rate of reprogramming, DBZ significantly stimulated the formation of human iPSC colonies (Fig. 2b).

Figure 2

γ-secretase inhibition promotes reprogramming by blocking Notch signaling

a, Chemical structure of DBZ. b, The efficiency of NANOG+/TRA-1-81+ iPSC generation from human neonatal keratinocytes transduced with OCT4, KLF4, SOX2, and CMYC and treated with different concentrations of DBZ from days 1-18 post-transduction. c, The efficiency of NANOG+/ TRA-1-81+ iPSC generation from human neonatal keratinocytes transduced with OCT, SOX2, KLF4, and CMYC and GFP or NOTCH ICD and treated with DMSO or 10 μM DAPT from days 1-18 post-transduction. Cells were transduced with NOTCH ICD or GFP lentivirus 1 day after transduction with the reprogramming factors. d, qPCR analysis of expression levels of NOTCH-dependent gene HES1 in human neonatal keratinocytes transduced with dominant-negative MASTERMIND-LIKE-1 (dnMAML) or RFP. e, The efficiency of NANOG+/TRA-1-81+ iPSC generation from human neonatal keratinocytes transduced with OCT, SOX2, KLF4, and CMYC and RFP or dnMAML and treated with DMSO or 10 μM DAPT from days 1-18 post-transduction. For all experiments, error bars represent the standard deviation between two or three biological replicates and statistical significance was determined using a two-tailed homoscedastic Student's t-test.

Both DBZ and DAPT could have effects on the processing of unidentified γ-secretase substrates that are distinct from NOTCH, which might also impact reprogramming efficiency. If the beneficial effects of DAPT on reprogramming were being mediated through the specific inhibition of NOTCH signaling rather than through some other target of γ-secretase, then we reasoned that constitutive activation of NOTCH signaling should eliminate the beneficial effect of DAPT. Consistent with this notion, we found that overexpression of the NOTCH intracellular domain (Supplementary Fig. 3b) stimulated the expression of NOTCH-target genes (Supplementary Fig. 3c) and completely blocked the positive effects of DAPT on reprogramming (Fig. 2c). Conversely, we reasoned that antagonizing the transcriptional activity of NOTCH should increase the rate of keratinocyte reprogramming. Indeed, when we suppressed NOTCH activity by overexpressing a dominant-negative form of MAML1 (Fig. 2d), a transcriptional co-activator for NOTCH [36,37], we observed an increase in iPSC generation from keratinocytes transduced with all 4 reprogramming factors (Fig. 2e). Therefore, we conclude that the inhibition of NOTCH signaling promotes the reprogramming of both human and mouse keratinocytes.

In order to understand how Notch inhibition promotes iPSC generation, we first determined when in the reprogramming process it was required. We treated mouse keratinocytes with DAPT either before or both before and after transduction with reprogramming factors. While treatment both before and after transduction yielded a 4-fold increase in iPSC generation, we found that pre-treatment alone resulted in a significant 2.5-fold enhancement in reprogramming efficiency (Supplementary Fig. 4a). To more precisely pinpoint the effective post-transduction treatment window, we transduced human keratinocytes with KLF4, OCT4, SOX2, and CMYC and administered DAPT or DBZ from days 1-6, 6-11, 11-16, or 1-16 after viral infection (Figs 3a-c). Chemical inhibition of NOTCH signaling was most effective during early time points, significantly increasing iPSC generation when used from days 1-6 and 6-11 (Figs 3b, c). In contrast, a later treatment from days 11-16 had little effect on reprogramming (Figs 3b, c). Together, these results indicate that Notch inhibition can act on the starting keratinocytes and at early time points just after the initiation of transcription factor overexpression to enhance reprogramming.

Figure 3

Notch inhibition promotes keratinocyte reprogramming by suppressing p21

a, Schematic of the DAPT treatment time course on human neonatal keratinocytes. b, Efficiency of NANOG+/TRA-1-81+ iPSC generation from human neonatal keratinocytes transduced with OCT4, SOX2, KLF4, and CMYC and treated with intervals of 10 μM DAPT or c, 2 μM DBZ. d, Western blot for p21 in human neonatal keratinocytes transduced with OCT4 and SOX2 and treated with DMSO or 10 μM DAPT. Full blot shown in Supplementary Figure 7c. e, Western blot for INVOLUCRIN in human neonatal keratinocytes treated with DMSO, 10 μM DAPT, or 1.2 mM calcium chloride for 6 days. Calcium was used as a positive control to induce keratinocyte differentiation. Full blot shown in Supplementary Figure 7d. f, Efficiency of NANOG+/TRA-1-81+ iPSC generation from human neonatal keratinocytes transduced with OCT4, KLF4, SOX2, and CMYC and a scrambled shRNA or a p21 shRNA at day 0 of reprogramming. DAPT was added at 10 μM. g, Efficiency of NANOG+/TRA-1-81+ iPSC generation from human neonatal keratinocytes transduced with OCT4 and SOX2 and a scrambled shRNA control or a p21 shRNA at day 0 of reprogramming. DAPT was added at 2.5 μM. h, Efficiency of NANOG+/TRA-1-81+ iPSC generation from human neonatal keratinocytes transduced with OCT, SOX2, KLF4, and CMYC and GFP or p21 and treated with DMSO or 10 μM DAPT from days 1-18 post-transduction. For all experiments, error bars represent the standard deviation between two-three biological replicates and statistical significance was determined using a two-tailed homoscedastic Student's t-test.

Notch inhibition acts by suppressing p21 expression

One way that Notch inhibition could promote iPSC formation is by activating the expression of the reprogramming transcription factors from their endogenous loci. However, when we treated human keratinocytes with DAPT and analyzed their gene expression, we found that levels of KLF4, OCT4, and CMYC actually decreased and SOX2 did not significantly change (Supplementary Fig. 4b).

In the mammalian epidermis, Notch signaling functions as a switch that directly activates p21 transcription, which in turn forces keratinocytes to exit the cell cycle and begin differentiating [38]. To determine if chemical inhibition of Notch signaling in keratinocytes might be enhancing their reprogramming potential by suppressing p21, we measured p21 levels in human keratinocytes in the presence and absence of DAPT. Consistent with previous reports [38], we found that Notch inhibition decreased the levels of p21 mRNA and protein in these cells (Supplementary Fig. 4c, Fig. 3d). In addition, DAPT treatment slightly decreased the level of FLAG-tagged p21 protein expressed by an exogenous retrovirus, indicating that Notch may also regulate p21 post-transcriptionally (Supplementary Figs. 4d, e). Consistent with these observations, Notch inhibition suppressed expression of INVOLUCRIN, which is expressed in more differentiated keratinocytes (Fig. 3e).

To verify that Notch inhibition promotes iPSC reprogramming by suppressing p21, we performed 2-factor (Oct4 and Sox2) and 4-factor reprogramming in keratinocytes with p21 siRNA/shRNA in the presence or absence of DAPT. Mouse keratinocytes transduced with Klf4, Sox2, Oct4, and cMyc showed a similar increase in iPSC generation when treated with either 2.5 μm DAPT or p21 siRNA (Supplementary Fig. 4f). The efficiency of reprogramming with these two methods was not significantly different (Supplementary Fig. 4f), and treating with DAPT in the presence of the p21 siRNA did not produce a significant increase in iPSC formation (Supplementary Fig. 4f). Similarly, suppression of p21 by shRNA (Supplementary Figs 4g, h) enabled the generation of iPSCs from human keratinocytes transduced with 2 or 4 factors at rates equivalent to DAPT treatment (Figs 3f, g). Again, supplementing p21 knockdown with DAPT treatment did not result in a significant increase in iPSC formation (Figs 3f, g). These results indicate that p21 suppression and DAPT have similar effects on iPSC generation from keratinocytes and that DAPT does not provide an additional advantage over p21 suppression alone.

If Notch inhibition and p21 suppression indeed blocks keratinocyte differentiation, the p21-treated keratinocytes would be predicted to display an increase in their long-term proliferative capacity [39]. The ability to form large colonies on collagen demonstrates the ability of keratinocytes to self-renew extensively and is a functional property unique to undifferentiated cells of this lineage [39]. In contrast, differentiated keratinocytes senesce after only a few rounds of division and do not form colonies [39]. DAPT treatment of human keratinocytes for 6 days significantly increased the number of cells capable of forming large colonies when cultured for an additional 14 days in the absence of the chemical (Supplementary Fig. 4i). The resulting 4-fold increase in colony formation rate was similar in magnitude to the elevation in iPSC generation with DAPT treatment (Fig. 1b). To determine if this increased self-renewal capacity was indeed promoting reprogramming, we transduced keratinocytes with p21 to limit their replication and attempted to reprogram them either with or without DAPT. The forced p21 expression severely impaired the self-renewal potential of the keratinocytes (Supplementary Fig. 4j) and inhibited iPSC formation after transduction with the 4 reprogramming factors and treatment with DAPT (Fig. 3h).

Because Notch inhibition does not promote fibroblast replication [40], if this is the mechanism by which DAPT improves reprogramming, we would not expect chemical treatment to affect mouse embryonic fibroblast [41] reprogramming [41]. Indeed, DAPT treatment of MEFs transduced with all 4 reprogramming factors did not affect the rate of iPSC generation (Supplementary Fig. 4k). Together, these results demonstrate that Notch inhibition promotes iPSC generation from keratinocytes by repressing their differentiation and enhancing their long-term replicative potential through p21 suppression.

Efficient reprogramming with Notch and DOT1L inhibition

Knowing that Notch inhibition enhances iPSC generation through this unique mechanism, we next wanted to compare its activity to previously described reprogramming molecules that act through other mechanisms [17,42-44] and identify any that DAPT might synergize with. When we transduced human neonatal keratinocytes with KLF4, SOX2, OCT4, and CMYC and treated them with various combinations of compounds shown to enhance reprogramming in other reports, including an activator of 3’-phosphoinositide-dependent kinase-1 (PDK1)[17], inhibitors of TGF-β, MEK, and GSK3β signaling [17], histone deacetylase inhibitors [17,42], histone methyltransferase inhibitors [17,44], and a DNA methyltransferase inhibitor [43], we found that DAPT treatment was the most potent at enhancing reprogramming (Fig. 4a). This remained true when we attempted reprogramming with only OCT4 and SOX2 (Fig. 4b).

Figure 4

Highly efficient reprogramming with NOTCH and DOT1L inhibition

a, Comparison of NANOG+/TRA-1-81+ iPSC generation from OCT4, SOX2, KLF4, and CMYC-transduced human neonatal keratinocytes using 10 μM DAPT versus other published reprogramming chemicals. “A83” = A8301 (.5 μM), “PD” = PD0325901 (.5 μM), “All from ref (13)” = A8301 (.5 μM), PD0325901 (.5 μM), PS48 (5 μM), sodium butyrate (.25 mM), Parnate (2 μM), CHIR99021 (3 μM), “AZA” = 5-aza-cytidine (.5 μM), “VPA” = valproic acid (.5 mM), “iDOT1L” = EPZ004777 (3 μM). b, Comparison of NANOG+/TRA-1-81+ iPSC generation from OCT4- and SOX2-transduced human neonatal keratinocytes using 2.5 μM DAPT versus other published reprogramming chemicals. c, iPSC line generated from human neonatal keratinocytes using OCT4, SOX2, DAPT, and iDOT1L. scale bars = 100 μm. For all experiments, error bars represent the standard deviation between two-three biological replicates and statistical significance was determined using a two-tailed homoscedastic Student's t-test.

However, an inhibitor of the histone methyltransferase DOT1L (iDOT1L) synergized OCT4, SOX2 and DAPT to elevate the rate of iPSC generation by 10-fold over the rate with OCT4, SOX2 and DAPT alone, making it even more efficient than 4-factor reprogramming either with or without DAPT (Fig. 4b). The OCT4 + SOX2 + DAPT + iDOT1L colonies could be readily expanded and maintained NANOG and TRA-1-81 expression (Fig. 4c). These data indicate that Notch inhibition is a potent enhancer of reprogramming in keratinocytes that can synergize with chromatin-modifying compounds to induce pluripotency at a high efficiency with only OCT4 and SOX2.

Notch inhibition does not compromise p53 activity

Previous studies of p53 and p21 in reprogramming have suggested that ectopic overexpression of reprogramming transcription factors can activate p53, which then induces either apoptosis or the expression of p21, thus inhibiting reprogramming [3,6]. Because suppression of the p53 pathway greatly facilitates iPSC generation, this approach has become an important part of reprogramming methods that reduce or eliminate integrating exogenous transcription factors [3,4]. However, because p53 inhibition allows the accumulation of genetic mutations during reprogramming [8], alternative approaches for increasing reprogramming efficiencies would be desirable. We therefore next asked whether Notch inhibition promotes reprogramming through a p53-dependent or independent pathway by analyzing the effects of DAPT and DBZ treatment on p53 and its target genes. First, we confirmed the finding that transduction with the iPSC reprogramming factors stimulated p53 activity (Fig. 5a). Chemical inhibition of Notch signaling in both human and mouse keratinocytes did not reduce the expression of p53 at the protein or mRNA level either before or after transduction with the reprogramming factors (Figs 5b,c and Supplementary Figs 5a, b). Moreover, transcriptional analysis of DAPT-treated human and mouse keratinocytes revealed that the mRNA levels of the p53 target genes Dr5, Puma, and Fas were not decreased (Fig. 5c and Supplementary Figs 5a, b), supporting the notion that p53 activity was not suppressed by Notch inhibition.

Figure 5

NOTCH inhibition suppresses p21 without reducing p53 activity

a, qPCR analysis of p53-dependent genes in human neonatal keratinocytes 3 days after transduction of GFP or OCT4 and SOX2. b, Western blot of p53 levels in human neonatal keratinocytes with DMSO or 10 μM DAPT treatment for 3 days. Full blot shown in Supplementary Figure 7g. c, qPCR analysis of p53-dependent genes after 10 μM DAPT or 2 μM DBZ treatment for 3 days in OCT4, SOX2-transduced human keratinocytes. d, The efficiency of NANOG+/TRA-1-81+ iPSC generation in OCT4, SOX2, KLF4, and CMYC-transduced human neonatal keratinocytes transduced with p53DD or GFP with or without exposure to UV irradiation. e, γH2AX immunostaining in human neonatal keratinocytes 10 days after transduction with OCT4, SOX2, KLF4, and CMYC and treatment with DAPT, p53DD, or p53 shRNA. Scale bars = 50 μm. f, The percentage of TUNEL-positive cells in human neonatal keratinocyte reprogramming cultures with active or inactive p53 (p53DD expression) 10 days after transduction with OCT4, SOX2, KLF4, and CMYC. g, The number of insertions or deletions (indels) per iPSC line derived under normal, DAPT, or p53DD conditions, as determined by array CGH. For all experiments, error bars represent the standard deviation between two biological replicates and statistical significance was determined using a two-tailed homoscedastic Student's t-test. * denotes significance p-value < .05.

To further confirm that DAPT treatment did not suppress p53 activity, we performed reprogramming experiments with and without DAPT after UV irradiation. UV exposure causes DNA damage, which in turn reduces reprogramming efficiencies by inducing p53-dependent apoptosis [8]. However, p53-deficient cells are resistant to the negative effects of UV irradiation on reprogramming [8]. Therefore, if p53 activity was maintained in DAPT-treated cultures, then we would expect a sharp decrease in reprogramming efficiency after UV irradiation. As a control for p53-deficiency, we performed 4-factor reprogramming with or without UV irradiation using keratinocytes in which we overexpressed a dominant-negative form of p53 (p53DD) [3] that suppressed p53 activity as evidenced by a decrease in the expression levels of p53-dependent target genes (Supplementary Fig. 5c). As expected, UV exposure did not impact the rate of iPSC generation when p53DD was expressed, functionally demonstrating that p53 activity was indeed impaired (Fig. 5d). In contrast, in the absence of p53DD overexpression, UV exposure sharply reduced the number of iPSCs generated in DMSO-treated cultures (Fig. 5d). Similarly, UV irradiation severely diminished the number of iPSC colonies in DAPT-treated cultures again suggesting that Notch inhibition does not suppress p53 activity during reprogramming (Fig. 5d).

Although the difference in reprogramming efficiency in p53-deficient versus DAPT-treated keratinocytes was clearly evident when UV irradiation was used to induce DNA damage, we next determined whether DNA damage was measurably influenced by DAPT treatment under normal reprogramming conditions. To test this, we quantified phosphorylated histone H2AX (γH2AX) expression in 4-factor-transduced human keratinocytes treated with DAPT, p53DD, or p53 shRNA. Histone H2AX becomes phosphorylated in response to double strand DNA breaks, making it a reliable marker of DNA damage [8]. Pan-nuclear γH2AX expression results from replication-induced damage and could indicate insults sustained during reprogramming [8]. We found that 10 days after transduction, pan-nuclear γH2AX staining was significantly elevated in cultures treated with p53DD or p53 shRNA, which is consistent with a previous study in which elevated rates of DNA damage were observed in p53-deficient cells during reprogramming and in the resulting iPSCs [8] (Fig. 5e and Supplementary Figs 5d-f). The DAPT-treated cells, however, maintained low cell numbers with pan-nuclear γH2AX expression that were similar to the control cultures (Fig. 5e and Supplementary Fig. 5f). These results suggest that, in contrast to p53-deficiency, DAPT treatment did not promote the survival and reprogramming of cells with DNA damage.

To confirm that Notch inhibition does not prevent the apoptosis of compromised cells during reprogramming, we measured the fraction of TUNEL-positive nuclei in DAPT-treated cultures. Despite high rates of DNA damage in the p53-deficient reprogramming cultures, the percentage of TUNEL-positive nuclei was greatly reduced compared to a wild-type control, indicating that inactivation of p53 permitted the survival of cells with compromised genomes (Fig. 5f). In contrast, the percentage of TUNEL-positive cells was not significantly reduced by DAPT treatment (Fig. 5f).

In order to determine if DAPT enabled the efficient generation of iPSCs that displayed improved genomic integrity relative to their counterparts made through p53 suppression, we measured the copy number variation in iPSC lines made with DAPT or p53DD. Consistent with the γH2AX and TUNEL staining results, we found that iPSC lines derived in the presence of p53DD possessed an average of 4 indels/line, while iPSCs derived with a control GFP vector or 10 μm DAPT contained only 1 or .5 indels/line, respectively (Fig. 5g and Supplementary Fig. 6). Together, these experiments show that DNA damage is present during normal reprogramming conditions and that inhibition of p53 allows cells with damaged genomic material to persist. In contrast, Notch inhibition enhances reprogramming without compromising genomic integrity or promoting the survival of iPSCs that have undergone DNA damage.

Discussion

In summary, our findings suggest that signaling through the Notch pathway is a significant impediment to the early stages of the reprogramming of both mouse and human keratinocytes into iPSCs (Figure 6). Importantly, the mechanism by which Notch signaling likely inhibits reprogramming of mouse and human cells is by activating p21 independently of p53. Consistent with this hypothesis, treatment of reprogramming cultures with the γ-secretase inhibitors DAPT and DBZ reduced the levels of intracellular Notch and increased colony forming potential, leading to an increase in the rate of iPSC formation. Suppression of p21 expression by siRNA/shRNA was sufficient to replace Notch inhibition in reprogramming, and exogenous p21 blocked the beneficial effects of DAPT. Importantly, the resulting improvement in reprogramming activity did not come at the expense of a reduction in p53 activity or increased genomic instability (Figure 6).

Figure 6

Model of iPSC generation from human keratinocytes

Notch inhibition allows the production of safer oncogene-free iPSCs by suppressing p21 in a p53-independent manner.

Our findings have immediate and practical ramifications for the improved production of patient-specific human iPSCs. When taken together, our studies show that through pharmacological inhibition of NOTCH, it is routinely possible to produce human iPSCs with only OCT4 and SOX2, rendering CMYC and KLF4 dispensable and thereby reducing the oncogenic potential of the resulting cells. Furthermore, our findings enabled CMYC and KLF4-free iPSC production without inhibition of p53 or its target genes involved in apoptosis, allowing pro-apoptotic pathways that ensure genomic integrity to be engaged [8,10,11]. Thus, in this approach, the production of oncogene-free iPSC lines does not come at the expense of an increase in mutational load [8,10,11,45].

Studies using nuclear transplantation and defined transcription factors have shown that nuclei become less amenable to reprogramming as they advance developmentally [24-26]. Our study demonstrates that intercellular communication in somatic cultures can cause them to differentiate and lose their reprogramming potential, but that with small molecule treatment, it is possible to force them to remain in an undifferentiated, highly reprogrammable state. This approach synergized potently with chemical inactivation of the histone H3 methyltransferase DOT1L, allowing two-factor reprogramming at higher efficiency than with four transcription factors. This indicates that while histone methyltransferase inhibition had almost no effect on the reprogramming of differentiated keratinocytes, it had a profound ability to enhance the reprogramming of undifferentiated keratinocytes. Thus, somatic cells at different developmental stages respond differentially to chromatin-modifying signals during reprogramming. The combined chemical inhibition of NOTCH and DOT1L provides a new approach for boosting the reprogramming potential of keratinocytes and is an attractive starting point for the identification of a small molecule reprogramming cocktail for human cells.

Online Methods

iPSC reprogramming experiments

The IACUC committee of Harvard University approved the use of mice for all experiments included in this paper. Oct4:GFP neonatal mouse keratinocytes were isolated from P1-P2 pups using an overnight digestion in either .25% trypsin/EDTA or TrypLE (Life Technologies) at 4 degrees Celsius. They were cultured in SFM medium (Life Technologies) on collagen IV-coated plates. Neonatal human epidermal keratinocytes (Lonza) were cultured in Epilife medium (Invitrogen) on collagen-coated plates. Keratinocytes were reprogrammed using retroviruses containing either mouse or human OCT4, SOX2, KLF4, and CMYC produced in the pMXs backbone. Chemical treatment was initiated 1-2 days after viral transduction and re-administered every other day until the end of the experiment unless otherwise specified. DAPT (EMD Millipore) was used at 10 μM for reprogramming experiments using OCT4, SOX2, KLF4, and CMYC and 2.5 μM for OCT4, SOX2 reprogramming experiments unless otherwise noted. DBZ was used at 2 μM. Irradiated mouse embryonic fibroblast feeders were added 6 days after transduction and the media was changed to mouse or human embryonic stem cell medium at that time. Colonies were scored as iPSC colonies if they were Oct4::GFP+ in mouse experiments or NANOG+/TRA-1-81+ in human experiments.

Gene expression analysis of iPSCs

Nanostring (Nanostring Technologies) and scorecard analysis was performed as described [34]. iPSCs were cultured in mTesr1 medium (Stem Cell Technologies) prior to RNA isolation. To measure their differentiation propensities, iPSCs were dissociated into embryoid bodies and cultured in human embryonic stem cell medium without bFGF for 16 days. Cells were then lysed and total RNA was extracted using Trizol (Life Technologies) and purified using the RNeasy kit (QIAGEN). 300 ng to 500 ng of RNA was profiled on the Nano-String nCounter system (Nanostring Technologies) according to manufacturer's instructions. A custom nCounter codeset covering 500 genes that monitor cell state, pluripotency, and differentiation was used [34]. Data analysis was performed with the R statistics package as in [34]. Briefly, the lineage scorecard performs a parametric gene set enrichment analysis on t scores obtained from a pairwise comparison between the cell line of interest and the reference of ES cell-derived EBs.

Differentiation of iPSCs

For teratoma formation, 1-2 million human iPSCs were injected into the kidney capsule of nude mice and harvested 2 months later. Teratomas were sectioned and stained with hematoxylin and eosin for visualization. For the mouse iPSC chimera assay, 10 Oct4::GFP+ iPSCs were injected per ICR blastocyst, and 20 blastocysts were transplanted into each pseudopregnant female. Embryos were either allowed to develop to term or harvested at day E12.5 and dissected for genital ridge analysis using a stereomicroscope.

Gene expression analysis of reprogramming cultures

Illumina MouseRef-8 microarrays (Illumina) were used for genome-wide mRNA expression analysis of reprogramming mouse keratinocyte cultures treated with DMSO or 10 μM DAPT. For QPCR analysis, RNA was isolated using Trizol, cDNA synthesis was performed using the iScript cDNA synthesis kit (Bio-rad), and the SYBR Green qPCR Supermix (Bio-rad) was used for PCR product detection.

Western blots and immunofluorescence

Antibodies detecting mouse Notch (Santa Cruz Biotechnology, sc-6015), human NOTCH (Abcam, ab27526, and Santa Cruz Biotechnology, sc-23307), cleaved human NOTCH (Cell Signaling Technology, 2421), p53 (Santa Cruz Biotechnology, sc-56182), Involucrin (Abcam, Ab53112), and p21 (Cell Signaling Technology 05-345) were used for western blots. Blots were quantified using ImageJ software. Antibodies specific for NANOG (Abcam, AF1997) and TRA-1-81 (Chemicon, MAB4381) were used to identify human iPSCs. A γH2AX (Abcam, ab11175) antibody was used to detect γH2AX foci. Cells in which γH2AX staining covered greater than half the nucleus were scored as positive for γH2AX foci.

UV irradiation assay

UV irradiation was performed at a dosage of 30 J. TUNEL staining was performed using a TUNEL kit (Pharmacia Biosciences).

shRNA/siRNA knockdown experiments

shRNAs and siRNAs were purchased from Sigma and added to reprogramming cultures within 1 day after addition of the reprogramming retroviruses. shRNAs (TRCN0000003753, p53 and TRCN0000287021, p21) were expressed in the pLKO.1 lentiviral backbone. siRNAs were used at 80nM and were transfected into reprogramming cultures using RNAiMAX (Life Technologies).

Array CGH analysis of iPSC lines

Cell Line Genetics performed array CGH analysis of iPSC lines at passage 5 using 4×180K+SNP analysis.

Statistical analysis

For all experiments, error bars represent the standard deviation between two-three biological replicates and statistical significance was determined using a two-tailed homoscedastic Student's t-test.