Post-translational modification of Ngn2 differentially affects transcription of distinct targets to regulate the balance between progenitor maintenance and differentiation

Transcription factors undergo very dynamic binding cycles on promoters (Hager et al., 2009; Michel, 2009). ‘Slow’ promoters are thought to require more extensive chromatin modification for activation than ‘fast’ promoters and hence require greater promoter residence time for activation. Consistent with this, increased dwell time on promoters is associated with more active transcription (Stavreva et al., 2004; Hager et al., 2009; Voss et al., 2011). Could differential requirements for transcription factor occupancy be used to differentially control distinct targets of a master regulatory transcription factor in response to a changing cellular environment? We have explored whether such a mechanism might coordinate cell cycle length and differentiation during nervous system development.

The proneural basic helix-loop-helix (bHLH) transcription factor neurogenin 2 (Ngn2) is required for generation of glutamatergic neurons in many regions of the central nervous system (Bertrand et al., 2002) and has multiple downstream functional effectors (Seo et al., 2007). However, as well as controlling neuronal differentiation cell-autonomously, Ngn2 also controls progenitor maintenance, largely non-cell-autonomously, by direct transcriptional activation of the Notch ligand Delta (Ma et al., 1996). Classically, Ngn2-driven Delta expression in cells destined to become neurons activates the Notch pathway in adjacent cells. This results in expression of Hes proteins, which inhibit Ngn2 transcription and activity in these adjacent cells, maintaining their progenitor status (Chitnis et al., 1995), a process known as lateral inhibition. Recent evidence demonstrates an oscillatory relationship between Ngn2 and Notch signalling as a requirement for maintenance of progenitor status (Kageyama et al., 2008b; Shimojo et al., 2008), further defining the molecular basis of lateral inhibition. Escape from lateral inhibition-driven progenitor maintenance requires expression and activity of the bHLH transcription factor NeuroD, which is also a direct target of Ngn2 (Lee, 1997; Seo et al., 2007).

Thus, Ngn2 is required non-cell-autonomously for progenitor maintenance via Delta upregulation and, conversely, drives neuronal differentiation cell-autonomously via upregulation of NeuroD. The Delta promoter responds rapidly to Ngn2 expression and requires minimal chromatin modification, whereas upregulation of NeuroD transcription is considerably slower and requires chromatin modifiers such as p300 and Brg1 (Koyano-Nakagawa et al., 1999; Seo et al., 2005). This difference in promoter timing has been postulated to allow progenitor maintenance to be established before differentiation occurs (Koyano-Nakagawa et al., 1999), but how a difference in transcription factor binding requirements for promoter activation can act mechanistically as a developmental switch in response to the cellular environment has remained unclear.

Progenitor maintenance predominates at early stages of cortical development, whereas neuronal differentiation increases as the embryo develops, and this transition is coordinated by an average cell cycle lengthening (Lange and Calegari, 2010). We have previously shown that preventing Cdk-mediated phosphorylation of Ngn2, by mutating its serine-proline (SP) target sites to alanine-proline (9S-A Ngn2), significantly enhances the ability of Ngn2 to drive neuronal differentiation by enhancing its binding to downstream target promoters such as NeuroD, without apparently increasing non-cell-autonomous progenitor maintenance (Ali et al., 2011). However, un(der)phosphorylated Ngn2 also binds more tightly to the Delta promoter (Ali et al., 2011), which might be expected to simultaneously promote non-cell-autonomous progenitor maintenance. In practice, mutation of Cdk phosphorylation sites substantially enhances the ability of Ngn2 to drive neuronal differentiation in the cells in which it is expressed, both in vitro and in vivo (Ali et al., 2011). If un(der)phosphorylation of Ngn2 enhances occupancy of both the NeuroD and Delta promoters (Ali et al., 2011), why does it promote neuronal differentiation cell-autonomously preferentially over non-cell-autonomous Delta/Notch-driven progenitor maintenance?

We find that phosphorylation of Ngn2 differentially controls the activation of the NeuroD and Delta promoters by exploiting differences in the kinetics of activation in response to promoter occupancy by Ngn2. Hence, post-translational modification of Ngn2 in response to cellular conditions can directly control the balance between progenitor maintenance and differentiation by exerting different effects on the control of individual downstream targets.

Acquisition of Xenopus laevis embryos, preparation and injection of synthetic mRNA, staging of embryos and in situ hybridisation were as described previously (Vernon et al., 2003; Vosper et al., 2007; Vosper et al., 2009; Ali et al., 2011). Degradation assays were performed as described (Ali et al., 2011).

Cell culture and transfections were conducted as described previously (Ali et al., 2011).

cDNA was generated from mouse P19 cells or stage 15 Xenopus embryos and 50 ng used per qPCR reaction, which was performed in the LightCycler 480 PCR system (Roche) using SYBR Green mix (Roche), with β-actin and ornithine decarboxylase (ODC) as housekeeping genes for normalisation. Thermal cycling conditions were: 95°C for 5 minutes, then 45 cycles of 95°C for 10 seconds, 60°C for 10 seconds and 72°C for 10 seconds. Primer sequences (5′ to 3′): β-actin Fwd, CAGGGTGTGATGGTGGGAATG; β-actin rev, ATGGCTGGGGTGTTGAAGGTC; ODC Fwd, GAATCACCCGAATGCAAAGC; ODC rev, CCACTGCCAACATGGAAACTC.

Statistical analysis used a two-tailed Student’s t-test (P<0.05). Error bars indicate s.e.m.

Phosphorylation of Ngn2 has similar effects on its binding to the NeuroD and Delta promoters, as assayed by chromatin immunoprecipitation (ChIP), but dephosphorylated Ngn2 preferentially drives neuronal differentiation over progenitor maintenance (Ali et al., 2011). We hypothesised that the Delta and NeuroD promoters might respond differentially to changes in Ngn2 un(der)phosphorylation; a differential response of the Delta and NeuroD promoters to increased promoter occupancy, resulting from Ngn2 post-translational modification changes occurring on cell cycle lengthening, could result in an alteration in the balance between progenitor maintenance and differentiation within a population of neural progenitors (Vosper et al., 2009; Ali et al., 2011). To test this hypothesis, we compared the response of NeuroD and Delta (Neurod1 and Dll1 in mouse) expression to overexpression of mouse Ngn2 (mNgn2) and a mutant version thereof, 9S-A mNgn2, in which the nine SP sites that can be phosphorylated by Cdks have all been mutated to alanine-proline (Ali et al., 2011) (Fig. 1A; supplementary material Fig. S1). Delta expression was upregulated to a similar extent in response to Ngn2 or 9S-A Ngn2 in mouse P19 cells and Xenopus embryos. By contrast, un(der)phosphorylated 9S-A Ngn2 induced approximately twice as much NeuroD expression as wild-type Ngn2 (Fig. 1A,B). Thus, the Delta and NeuroD promoters respond differently to the phosphorylation status of Ngn2.

We have previously shown that mutation of an increasing number of Ngn2 phosphorylation sites progressively increases its affinity for E-box DNA in an environment with high Cdk kinase activity. This translates to an enhanced association of 9S-A Ngn2 with both the Delta and NeuroD promoters (Ali et al., 2011); ChIP assays demonstrate that 9S-A mNgn2 binding to the NeuroD1 and Delta promoters is increased ∼2- and 3-fold, respectively, over binding of wild-type Ngn2. Since the NeuroD promoter requires greater promoter remodelling for activation than the Delta promoter (Koyano-Nakagawa et al., 1999; Seo et al., 2005), we reasoned that the cumulative effect of individual phosphorylation events, each one having a small effect on DNA binding and, hence, promoter occupancy by Ngn2, would result in a quantitative shift in the ratio of NeuroD to Delta expression in response to progressively reduced Ngn2 phosphorylation. To test this, a panel of S-A mutant versions of mNgn2, in which mutation of SP sites was undertaken cumulatively from the N-terminus (i.e. mS-A1 has the most N-terminal SP site mutated; see supplementary material Fig. S1), was assayed for their ability to upregulate NeuroD and Delta expression.

In this assay (Fig. 1C), 9S-A mNgn2 drove NeuroD expression to twice the level of that achieved by the same amount of mNgn2. By contrast, expression of Delta driven by mNgn2 and 9S-A mNgn2 was similar, consistent with results presented in Fig. 1A,B. When assaying our cumulative mutant series, we observed that NeuroD expression showed an essentially linear increase for each additional SP site of Ngn2 that was mutated. However, Delta expression increased very little with progressive SP mutations, confirming that the Delta promoter is markedly less sensitive to Ngn2 phosphorylation status than the NeuroD promoter (Fig. 1C). We have previously shown that expression of the neuronal marker βIII-tubulin also gradually increases as additional SP sites are mutated (Ali et al., 2011).

Our results are consistent with a model whereby Delta transactivation requires little chromatin remodelling and can occur with both phosphorylated and un(der)phosphorylated Ngn2, whereas only enhanced promoter dwell time achieved by un(der)phosphorylated Ngn2 is sufficient to remodel the NeuroD promoter (Koyano-Nakagawa et al., 1999; Seo et al., 2005). To confirm this, we investigated the kinetics of activation of the Delta and NeuroD promoters in response to expression of Ngn2 in P19 cells, normalising to the maximum expression of each individual transcript after 10 hours. Both Delta and NeuroD expression showed a lag post-transfection, which was at least partly due to the time taken for Ngn2 proteins to accumulate from transfected plasmids. However, Delta expression was significantly elevated within 3-4 hours post-transfection and continued to rise (Fig. 2A), whereas the NeuroD promoter took 5-6 hours to respond to Ngn2 expression.

We next compared promoter responsiveness to mNgn2 and 9S-A mNgn2 over a longer timecourse. 9S-A mNgn2 induced Delta transcription at a somewhat higher rate than the wild-type protein, although this difference was small (see also Fig. 1); 9S-AmNgn2 induced on average 20-30% more Delta transcripts than mNgn2. After peaking at ∼18 hours, Delta expression in response to mNgn2 and 9S-A mNgn2 declined at similar rates. As previously seen, the onset of NeuroD transcription showed a longer lag than that of Delta after transfection of mNgn2 or 9S-A mNgn2. However, once NeuroD transcription had commenced, 9S-ANgn2 expression resulted in a more rapid accumulation of NeuroD transcripts than wild-type Ngn2, resulting in a 2-fold enhancement of expression by 18 hours. Moreover, whereas expression of NeuroD in response to Ngn2 peaked at ∼30 hours and then underwent slow decline, expression in response to 9S-A Ngn2 continued to rise until ∼35 hours before declining more rapidly (Fig. 2B).

Taken together, it is clear that 9S-A Ngn2 induces NeuroD expression more rapidly and for longer when compared with induction by the wild-type protein, whereas Delta expression profiles are similar after Ngn2 or 9S-A Ngn2 transfection. These data provide further confirmation that the Delta and NeuroD promoters have differential responses to distinct phosphorylated forms of Ngn2 protein. This is consistent with a model in which these differences are a result of the differing affinity of phosphorylated forms of Ngn2 for E-box-containing promoters (Ali et al., 2011), resulting in differential gene expression depending on the extent of chromatin remodelling required for promoter activation.

To test more directly whether manipulating the stability of Ngn2, and hence its ability to bind DNA, has different effects on the NeuroD and Delta promoters, we investigated whether artificially destabilised N-terminally ubiquitin-fused forms of Xenopus Ngn2 (UbxNgn2 and Ub9S-A xNgn2), which would target Ngn2 for rapid destruction via the ubiquitin fusion/degradation (UFD) pathway (Johnson et al., 1995), would differentially affect Delta and NeuroD transcription. First, we investigated the half-life of these proteins in in vitro degradation assays in Xenopus egg extracts (Vosper et al., 2007; Vosper et al., 2009).

As expected, in Xenopus egg extract, N-terminal ubiquitin fusion decreased the half-life of UbxNgn2 and Ub9S-A xNgn2, resulting in proteins that were approximately half as stable as the non-ubiquitylated forms (Fig. 3A). E protein addition has been shown to stabilise Ngn2 protein (Vosper et al., 2007) (Fig. 3A). Addition of E12 led to only a small increase in the UbxNgn2 and Ub9S-A xNgn2 half-lives, which were still considerably shorter than those of their non-ubiquitylated counterparts with or without E12 (Fig. 3A).

We then tested the ability of these Ngn2 derivatives to drive neuronal differentiation in developing Xenopus embryos. Whereas injected xNgn2 mRNA induced moderate ectopic neurogenesis, the same amount of UbxNgn2 mRNA was completely inactive in this assay, indicating that the Ngn2 protein half-life regulates its ability to drive neuronal differentiation. However, both 9S-A xNgn2 and Ub9S-A xNgn2 could induce ectopic neurogenesis (Fig. 3B,C), with 9S-A xNgn2 showing the greatest activity. This indicates that mutating phosphorylation sites, and hence promoting Ngn2-E protein complex stability and enhancing DNA binding, can at least partially compensate for constitutive Ngn2 ubiquitylation and reduced half-life when assaying the ability to induce neuronal differentiation.

Ubiquitylation of xNgn2 resulted in reduced Delta expression (Fig. 3D), which was not significantly rescued by mutation of xNgn2 phosphorylation sites, confirming that phosphorylation of xNgn2 does not play a major role in the regulation of Delta expression. By contrast, whereas UbxNgn2 was completely unable to upregulate NeuroD, Ub9S-A xNgn2 elevated NeuroD transcription ∼7-fold (Fig. 3D). Thus, mutation of phosphorylation sites in Ngn2 that enhance promoter binding can partially compensate for a shorter protein half-life when driving NeuroD expression but not when driving Delta transcription. This is consistent with the stability of promoter association playing a key role in regulating the level of NeuroD transcription but not that of Delta.

We find that un(der)phosphorylation of Ngn2 promotes transcription of NeuroD, driving cell-autonomous differentiation, while having little effect on the transcription of Delta that activates Notch signalling in adjacent cells, promoting progenitor maintenance. Mechanistically, as well as effecting Ngn2 transcriptional repression, Notch-induced Hes1 protein inhibits Ngn2 post-translationally by binding and sequestering E proteins, inhibiting proneural protein-mediated transcriptional activation directly (Kageyama et al., 2008a). Un(der)phosphorylated Ngn2 is stabilised by E protein binding to a greater extent than the wild-type protein (Ali et al., 2011) (Fig. 3A), and also binds to downstream target promoters more efficiently (Ali et al., 2011), indicating more stable E protein association. We speculated that the ability of Notch signalling to regulate Ngn2 protein function might depend upon the phosphorylation status of Ngn2, and a feed-forward loop might be set up when Ngn2 phosphorylation drops whereby the reduction of Cdk kinase activity results in both a Ngn2-driven relative enhancement of NeuroD expression over Delta expression and an increase in the resistance of Ngn2-dependent transcription to Notch-mediated inhibition.

To test this, we injected mRNA encoding the constitutively active Notch intracellular domain (NICD) (Chitnis et al., 1995) along with xNgn2, 9S-A xNgn2 or GFP (as a control), then investigated subsequent neuronal differentiation. As expected, NICD inhibited differentiation of endogenous primary neurons on the injected side (Fig. 4A,B). In addition, NICD inhibited the ability of xNgn2 mRNA to induce ectopic primary neurons, demonstrating post-transcriptional inhibition of xNgn2 function. However, 9S-A xNgn2 was essentially completely resistant to the inhibitory activity of Notch signalling (Fig. 4A,B). This might be a consequence of more stable Ngn2-E protein binding to DNA, resulting in resistance to the E protein sequestration by the Hes1 protein that is induced by Notch signalling.

Taken together, our results suggest a model (Fig. 4C) whereby phospho-Ngn2 would predominate in rapidly cycling cells, resulting in Ngn2-E protein dimers with weak promoter affinity. This is insufficient to activate the NeuroD promoter (Figs 1, 2), an activation that is required for cell-autonomous neuronal differentiation. However, phospho-Ngn2 is still able to activate the Delta promoter, resulting in non-cell-autonomous Notch-mediated progenitor maintenance.

By contrast, within individual cells, cell cycle lengthening favours un(der)phosphorylated Ngn2-E protein dimers that exhibit greater promoter occupancy (Ali et al., 2011), which would be permissive for both NeuroD and Delta transcription. This un(der)phosphorylated Ngn2 is resistant to Notch-mediated inhibition (Fig. 4), which further potentiates expression of downstream targets, including NeuroD (Figs 1, 2). In addition, the accumulating NeuroD protein thus produced is relatively insensitive to Delta-driven Notch-mediated inhibition (Chitnis and Kintner, 1996). Rising NeuroD expression against a backdrop of constant Delta-mediated Notch signalling (Fig. 1), coupled with a growing resistance of un(der)phosphorylated Ngn2 to Notch-mediated lateral inhibition (Fig. 4), favour differentiation over progenitor maintenance in response to cell cycle lengthening.

In reality, within a population of neural progenitors the cell cycle length may be variable while showing an average increase. This variation in cell cycle length and hence variation in Ngn2 protein activity might contribute to the ‘salt and pepper’ heterogeneity of neuronal differentiation seen during CNS development. Moreover, SP pairs can act as substrates for a variety of kinases, such as GSK3β [which have already been shown to target SP sites in Ngn2 (Ma et al., 2008)] and MAP kinases, so phosphoregulation of Ngn2 might act to integrate a number of cellular signalling events in addition to cell cycle length.

We thank Helen Wise, Christelle Fiore-Heriche, Romana Kucerova and Jon Vosper for important preliminary observations; Chris Hurley and Venkat Pisupati for technical assistance; and Luke Wylie, Gert Veenstra, Adele Murrell, Jeremy Gunewardena, Hanno Steen, Marc Kirschner and Bill Harris for helpful discussions.