Valproic acid silencing of ascl1b/Ascl1 results in the failure of serotonergic differentiation in a zebrafish model of fetal valproate syndrome

Fetal valproate syndrome (FVS) is caused by in utero exposure to the drug sodium valproate. Valproate is used worldwide for the treatment of epilepsy, as a mood stabiliser and for its pain-relieving properties. In addition to birth defects, FVS is associated with an increased risk of autism spectrum disorder (ASD), which is characterised by abnormal behaviours. Valproate perturbs multiple biochemical pathways and alters gene expression through its inhibition of histone deacetylases. Which, if any, of these mechanisms is relevant to the genesis of its behavioural side effects is unclear. Neuroanatomical changes associated with FVS have been reported and, among these, altered serotonergic neuronal differentiation is a consistent finding. Altered serotonin homeostasis is also associated with autism. Here we have used a chemical-genetics approach to investigate the underlying molecular defect in a zebrafish FVS model. Valproate causes the selective failure of zebrafish central serotonin expression. It does so by downregulating the proneural gene ascl1b, an ortholog of mammalian Ascl1, which is a known determinant of serotonergic identity in the mammalian brainstem. ascl1b is sufficient to rescue serotonin expression in valproate-treated embryos. Chemical and genetic blockade of the histone deacetylase Hdac1 downregulates ascl1b, consistent with the Hdac1-mediated silencing of ascl1b expression by valproate. Moreover, tonic Notch signalling is crucial for ascl1b repression by valproate. Concomitant blockade of Notch signalling restores ascl1b expression and serotonin expression in both valproate-exposed and hdac1 mutant embryos. Together, these data provide a molecular explanation for serotonergic defects in FVS and highlight an epigenetic mechanism for genome-environment interaction in disease.


INTRODUCTION
Valproate (VPA) is a fatty acid derivative widely prescribed for its anticonvulsant, mood-stabilising and pain-relieving properties, but it has teratogenic and neuropsychiatric side effects upon in utero exposure, collectively termed fetal valproate syndrome (FVS). The underlying molecular cause of FVS is unknown, but candidate mechanisms are the dysregulation of transcription factors important for brain development, disruption of signal transduction pathways, inositol depletion and direct inhibition of epigenetic regulators such as the histone deacetylases (HDACs) (Chen et al., 1997;Detich et al., 2003;Einat et al., 2003;Marchion et al., 2005;Milutinovic et al., 2007;Phiel et al., 2001;Williams et al., 2002).
Significantly, 5HT is also implicated in autism pathogenesis. In ASD, 30% of subjects have elevated 5HT blood levels (Mulder et al., 2004;Schain and Freedman, 1961), central 5HT homeostasis is altered (Chugani et al., 1999;Chugani et al., 1997) and an association with stereotyped behaviour has been reported (Kolevzon et al., 2010;Sacco et al., 2010). Selective serotonin reuptake inhibitors (SSRIs) improve some manifestations of autism, including stereotypical behaviours (Hollander et al., 2003;McDougle et al., 2000), whereas depletion of the 5HT precursor tryptophan exacerbates these symptoms (Bauman et al., 2006). Genetic or pharmacological perturbation of the 5HT system is associated with autism-like behaviours in humans and in rodents (Bauman et al., 2006;Cook et al., 1997;Kane et al., 2012;Klauck et al., 1997;Nabi et al., 2004;Nakatani et al., 2009;Sutcliffe et al., 2005;Veenstra-VanderWeele et al., 2012). In particular, an allelic polymorphism of the serotonin transporter gene (SERT), which is a determinant of Valproic acid silencing of ascl1b/Ascl1 results in the failure of serotonergic differentiation in a zebrafish model of fetal valproate syndrome 5HT activity, is associated with ASD (Devlin et al., 2005). Furthermore, a mouse model of one of the human gain-of-function SERT genetic variants displays ASD-like behaviours and hyperserotonaemia (Veenstra-VanderWeele et al., 2012). Therefore, increases and decreases in central 5HT activity seem to produce common behavioural phenotypes, which is consistent with the view that autism can result from positive and negative changes in neurotransmitter signalling (Zoghbi and Bear, 2012).
We investigated the molecular pathophysiology underlying serotonergic deficits in a zebrafish FVS model because this could provide mechanistic insight into the genesis of core autism behaviours. Importantly, hindbrain development and neuronal subtype diversity and serotonergic differentiation show strong conservation (Lillesaar, 2011). Furthermore, zebrafish exposed to VPA display morphological defects similar to those described in FVS, suggesting the validity of our system for modelling features of human FVS (Gurvich et al., 2005;Herrmann, 1993). We show that VPA specifically blocks hindbrain 5HT expression in zebrafish. Acting via Hdac1, VPA silences the zebrafish ortholog of mammalian Ascl1, ascl1b, by unmasking tonic Notch repression. Moreover, Ascl1b is sufficient to rescue 5HT expression in VPAtreated embryos.

VPA impairs central 5HT neuronal differentiation
To assess the effect of VPA on brainstem development, we exposed zebrafish gastrulae at 50% epiboly to 0.625 mM VPA until 27 hours post-fertilisation (hpf), at which time the drug was removed and the embryos were allowed to develop until 48 hpf. Treatment from gastrulation was based on the heightened risk of teratogenicity in human infants exposed to VPA during the first trimester of pregnancy (Ornoy, 2009). Immunostaining for a range of hindbrain neuronal subtypes, specifically motor neurons, 5HT neurons, GABA-ergic neurons and Mauthner neurons, revealed a failure of 5HT neuronal differentiation marked by absence of 5HT expression (Fig. 1A). Additionally, Mauthner neurons, which express a neurofilament-associated antigen that is detected by the 3A10 monoclonal antibody, were also absent (Hatta, 1992;Schier et al., 1996) (Fig. 1A). Isl1-positive motor neurons and GABA-ergic neurons were present. The spatial distribution of motor neurons in VPA-treated embryos appeared subtly altered, suggesting migratory defects, but we did not pursue these changes further (Fig. 1A). Instead we focused on the striking serotonergic phenotype.
We speculated that exposure to VPA from gastrulation could affect common steps in the differentiation of multiple neuronal lineages. Therefore, we treated embryos with 0.625 mM VPA from 24-48 hpf and used appropriate markers to detect brainstem neuronal subtypes (Fig. 1B,C). Somatic motor, GABA-ergic and Mauthner neurons were still present (Fig. 1C), but there was a specific deficit of 5HT neuronal differentiation, marked by the absence of 5HT and tph2 expression (Fig. 1B). Moreover, there was an absence or severe reduction of pet1 expression, which indicates that VPA acts at a step proximal to or at the early stages of 5HT neuronal differentiation (Fig. 1B). Lower doses than this did not consistently lead to the loss of 5HT expression (see supplementary material Fig. S1).
Having established that brainstem 5HT expression was specifically lacking in embryos treated with VPA from 24 hpf, we addressed whether there was a delay in the differentiation of 5HT neurons. VPA was removed at 48 hours and embryos were analysed after a further 24 hours of incubation (Fig. 1D). There was limited recovery of 5HT expression at 72 hpf [mean number of 5HT neurons in controls=26.7±1.7 (s.d.), n=3; mean number of 5HT neurons in VPA-treated condition=1.8±1.3, n=4], which suggests that VPA does not merely retard the differentiation of 5HT neurons. We conclude that VPA specifically blocks the differentiation of 5HT neurons in embryos exposed to the drug between 24 and 48 hpf. This period coincides with the onset of differentiation of 5HT neurons between 25 and 30 hpf (Lillesaar et al., 2007). Because the post-mitotic differentiation of 5HT neurons is well underway from 48 hpf onwards (Lillesaar et al., 2007;McLean and Fetcho, 2004), we asked whether VPA could abolish 5HT expression after 48 hpf. VPA exposure between 48 hpf and 72 hpf had no effect on 5HT expression (Fig. 1D)

Clinical issue
The drug valproate is used worldwide as an anticonvulsant agent, as a mood stabiliser and for its pain-relieving properties. Valproate is teratogenic (interferes with early development) and fetal exposure causes fetal valproate syndrome (FVS), which is characterised by a spectrum of morphological, cognitive and behavioural deficits. Recent population-based epidemiological studies have highlighted the significantly increased risk of autism spectrum disorders (ASDs) in children exposed to valproate in utero. The in vivo mechanism of valproate action that is pertinent to its neuropsychiatric side effects is not clear. Multiple in vitro mechanisms have been described, including inhibition of histone deacetylases. Studies on animal models of FVS have identified biochemical and cellular perturbations of the central serotonergic system. Altered serotonin homeostasis is also a feature of idiopathic autism; therefore, uncovering the pathogenesis of serotonin deficits in FVS could reveal the molecular underpinnings of core behavioural abnormalities in autism.

Results
Zebrafish are highly suited to pharmacological and genetics approaches that can be combined to provide novel insights into disease mechanisms. In this article, the authors describe a zebrafish model of FVS that displays a failure of serotonergic differentiation in the brainstem in response to valproate treatment. They show that a critical proneural gene, ascl1b, is silenced by valproate through a mechanism that depends on inhibition of the histone deacetylase Hdac1. Their experiments further show that valproate unmasks tonic repression of the ascl1b gene by the Notch pathway. If the Notch pathway is blocked, valproate is no longer able to silence ascl1b. Importantly, restoration of Ascl1b in the presence of valproate rescues the expression of serotonin in the zebrafish brainstem.

Implications and future directions
This study shows that valproate represses expression of ascl1b, leading to defects in the serotonergic system in zebrafish. The failure in serotonergic differentiation in this new model is reminiscent of defects reported previously for mice exposed to valproate, suggesting that the mechanism unveiled herein is likely to provide a common molecular explanation for serotonergic abnormalities in this disorder. Indeed, the conservation of the differentiation pathways of serotonergic neurons in zebrafish and in humans suggests that these findings will be relevant to understanding the complex pathophysiology of FVS. More broadly, the authors highlight an epigenetic mechanism at work in an iatrogenic form of ASD that could also be relevant to idiopathic, common forms of autism.
n=4). These data suggest that VPA acts on serotonergic progenitors, rather than on post-mitotic neurons.
To test whether specific neuronal subtypes are vulnerable to VPA application from 24 hpf because they only differentiate at or after 24 hpf, we extended the range of neuronal subtypes assayed. Cerebellar Purkinje neurons begin to differentiate from 3 dpf and are marked by expression of Parv7 (Bae et al., 2009). Wild-type embryos were incubated in VPA for a prolonged period, from 24 hpf to 4.5 dpf (see supplementary material Fig. S2). Immunostaining against Parv7 showed the persistence of this cell type in VPA-treated embryos (supplementary material Fig. S2). The differential sensitivity of 5HT and Purkinje neurons suggests that vulnerability to the effects of VPA is not linked directly to the timing of neuronal differentiation.
Inhibition of Hdac1 by VPA accounts for the failure of 5HT neuronal differentiation Next, we sought to identify the molecular pathway targeted by VPA in blocking 5HT neuronal differentiation. Previous studies have shown both in vivo and in vitro that VPA is an inhibitor of HDACs at therapeutic concentrations (Göttlicher et al., 2001;Gurvich et al., 2005;Kook et al., 2003;Phiel et al., 2001;Tremolizzo et al., 2002;Yildirim et al., 2003). To test the involvement of Hdac1, we used a zebrafish hdac1 mutant line, hdac1 s436 , that was previously isolated in a forward genetic screen (Noël et al., 2008). Immunostaining of hdac1 s436 mutants revealed a lack of 5HT expression in the brainstem at 48 hpf ( Fig. 2A). However, by 72 hpf there was a partial recovery of 5HT neuronal differentiation ( Fig. 2B) that was more complete than in VPA-treated embryos at the same stage ( Fig. 1D, middle panel) [mean number of 5HT neurons in control siblings (sibs)=31±2.5, n=4; mean number of 5HT neurons in mutants=25.3±5.4, n=4]. Apart from a severe reduction in the differentiation of somatic motor neurons, marked by Isl1 expression, which has been reported previously (Cunliffe, 2004), other neuronal subtypes, GABA-ergic and Mauthner neurons, appeared intact ( Fig. 2A).
To further test the involvement of HDACs in 5HT neuron production, wild-type embryos were treated with the HDAC inhibitor trichostatin A (TSA), a potent inhibitor of class I and class II HDACs (Yoshida et al., 1990). Pharmacological blockade of HDACs with TSA from 24-48 hpf recapitulated the effect of VPA treatment, and resulted in loss of 5HT expression at 48 hpf, with preservation of somatic motor, GABA-ergic and Mauthner cell differentiation ( Fig. 2A). This suggests that the hdac1 s436 neuronal phenotype differs from the effect of VPA most likely because of differences in the timing of Hdac1 inactivation. Moreover, the closely similar phenotypes that result from hdac1 mutation and VPA treatment are consistent with the idea that VPA blocks 5HT neuronal differentiation by inhibiting Hdac1. VPA downregulates expression of the proneural gene ascl1b in serotonergic progenitors through a mechanism consistent with Hdac1 inhibition Next we examined the expression of transcription factors that specify serotonergic progenitors. We reasoned that, because 5HT neurons are absent, any disruption of serotonergic fate determinants might occur before the first 5HT neurons are born. We therefore exposed 24 hpf embryos to VPA and analysed gene expression by in situ hybridisation 8 hours later. There was a striking reduction of ascl1b expression by 32 hpf that was not apparent at 28 hpf, just 4 hours earlier (Fig. 2C). Colocalisation of ascl1b and nkx2.2 in the hindbrain by fluorescent double in situ hybridisation confirmed the expression of ascl1b in serotonergic progenitors (Fig. 2D). Expression of the related gene ascl1a and other fate determinants, nkx2.2 and foxa2, and the signalling molecule shh were unchanged, which implies that the loss of ascl1b expression is not due to depletion of progenitors (Fig. 2E).
Further evidence implicating Hdac1 as the mediator of the effects of VPA on 5HT neurons is that, in hdac1 s436 mutants, expression of ascl1b is also severely reduced, a finding that has been reported previously in a different hdac1-null mutant line (Fig. 2C) (Cunliffe, 2004).
In contrast to the downregulation of ascl1b, another bHLH proneural gene, ptf1a, which is required for the generation of cerebellar Purkinje neurons (Hori et al., 2008), was not decreased by VPA. Ptf1a expression in the cerebellar primordium is first detected at 48 hpf (Kani et al., 2010). We monitored ptf1a expression by in situ hybridisation for ptf1a transcripts in wild-type zebrafish embryos and by an eGFP reporter in a transgenic ptf1aegfp line at 52 hpf in embryos continuously exposed to VPA from 24 hpf (see supplementary material Fig. S2) (Pisharath et al., 2007). The persistent ptf1a expression suggests that proneural genes are differentially sensitive to VPA and this is consistent with the continued differentiation of Purkinje neurons after VPA exposure (see supplementary material Fig. S2).
Notch signalling represses ascl1b in the presence of VPA VPA has previously been reported to upregulate Notch signalling in a variety of systems (Greenblatt et al., 2007;Stockhausen et al., 2005). In zebrafish embryos this effect seems likely to be mediated by Hdac1 (Cunliffe, 2004). One possibility, therefore, was that Hdac1 was inhibiting the Notch pathway and thus decreasing the Notch-mediated inhibition of ascl1b via Her genes, which are the major effectors of Notch signalling in zebrafish (Cunliffe, 2004;Louvi and Artavanis-Tsakonas, 2006). A second possible mechanism is through direct upregulation of ascl1b by Hdac1 (Cunliffe, 2004;Harrison et al., 2011;Wang et al., 2009). Finally, Hdac1 repression of specific members of the Her family, in particular her6, which in turn represses ascl1b, is a third possibility (Cunliffe, 2004;Harrison et al., 2011). her4 is a direct target of Notch signalling that is widely used as a readout of Notch activity Yeo et al., 2007). The downregulation of ascl1b in embryos following an 8-hour treatment with VPA from 24 hpf (Fig. 2C) was not accompanied by obvious elevation of her4 expression (Fig. 3A,B). Quantitative reverse-transcriptase PCR confirmed that there was a modest decrease in her4 expression relative to untreated control embryos (Fig. 3A). her6 is also a target of the Notch pathway and to date is the only Her gene reported to undergo negative regulation by Hdac1 (Cunliffe, 2004). In situ hybridisation for her6 and a panel of additional Her family members in wild-type embryos similarly treated with VPA also showed no difference in expression compared with controls (data not shown).
To investigate the involvement of the Notch pathway in ascl1b repression by VPA, we inhibited Notch signalling over the same period and determined whether VPA could still repress ascl1b. Embryos at 22 hpf were treated with a small-molecule inhibitor of Notch signalling, the γ-secretase inhibitor DAPT, which is known to have potent activity in vivo in zebrafish (Geling et al., 2002;Imbimbo, 2008). DAPT treatment alone from 22-32 hpf effectively downregulated her4 expression, whereas ascl1b expression was maintained (Fig. 3A,C). Strikingly, following the addition of VPA at 24 hpf, her4 remained repressed and ascl1b expression persisted ( Fig. 3A,C). Therefore, the repression of ascl1b by short-term exposure to VPA is prevented by concomitant blockade of Notch signalling, which implies that a basal level of Notch signalling represses ascl1b when Hdac1 activity is blocked.
To confirm that the repression of Ascl1b by VPA is not due to increased Notch signalling secondary to blockade of Hdac1 function, we analysed her4 expression in hdac1 s436 mutants at 32 hpf (Fig. 3B). There was a marked reduction of her4 expression in mutant hindbrains, indicative of reduced Notch signalling (Fig. 3B). Therefore, VPA inhibition of Hdac1 does not repress ascl1b through the upregulation of Notch signalling, or via Hdac1-mediated repression of her4 or her6. Instead, Hdac1 acts independently of Notch activity to regulate ascl1b. Moreover, a parallel, tonic level of Notch signalling was sufficient to repress ascl1b transcription when Hdac1 was inhibited by VPA.

Hdac1 and Notch have opposing effects on ascl1b
To confirm the parallel roles for Hdac1 and Notch signalling, we tested whether DAPT could restore ascl1b expression in hdac1 s436 mutants. hdac1 s436 mutants and sibs were treated with DAPT from 22 hpf (Fig. 4A). At 32 hpf, sibs treated with DAPT showed strong expression of ascl1b, which was virtually absent by 48 hpf presumably because of depletion of the progenitor pool caused by the blockade of Notch signalling. Remarkably, DAPT restored ascl1b expression in hdac1 s436 mutants, and expression was maintained at least up to 48 hpf (compare Fig. 2C and Fig. 4A). The other progenitor markers, nkx2.2, foxa2 and shh, were not altered by DAPT in either sibs or hdac1 s436 embryos at 48 hpf (Fig. 4B). These data confirm that Hdac1 and Notch exert positive and negative regulatory effects, respectively, on ascl1b. Given the finding that DAPT could prevent the repression of ascl1b by VPA, the expression of ascl1b was then analysed in a Notch-signalling mutant, mindbomb (mib). These mutants lack an E3 ubiquitin ligase, which is crucial for Notch signalling (Itoh et al., 2003;Schier et al., 1996). Consistent with the data from chemical inhibition of Notch signalling, the expression of ascl1b was restored in 32 hpf mib mutants treated with VPA from 50% epiboly to 27 hpf (Fig. 4C).
Failure of 5HT neuronal differentiation in VPA-treated embryos is due to repression of ascl1b Finally, we asked whether replacement of ascl1b was sufficient to rescue 5HT neuronal differentiation in VPA-treated embryos. To test this, a plasmid encoding a Myc-tagged full-length Ascl1b (ascl1bmyc) under the control of UAS was injected into a stable transgenic zebrafish line, ubi:ERT2-GAL4 (Gerety et al., 2013), which allows the regulation of ascl1b-myc by 4-hydroxytamoxifen (4-OHT). In control embryos treated with VPA alone, Ascl1b-Myc could not be detected by immunostaining for Myc antigen (Fig. 5A) and 5HTexpressing neurons were absent. Addition of 4-OHT resulted in strong and widespread expression of Ascl1b-Myc and the rescue of 5HT neuronal differentiation in 16% of embryos (Fig. 5A, P=0.0058).
If repression of ascl1b is the reason for the failure of 5HT neuronal differentiation upon VPA exposure, then restoration of ascl1b expression in hdac1 s436 and mib mutants should be sufficient to rescue 5HT expression. Indeed, immunostaining of DAPT-treated hdac1 s436 embryos at 48 hpf revealed the rescue of 5HT expression in the raphe (Fig. 5B). We also exposed mib mutants to VPA from gastrulation (50% epiboly) to 27 hpf, and then incubated the embryos in normal medium until 48 hpf, at which time they were fixed and immunostained for 5HT. We had already established that an identical VPA regime abolishes 5HT expression on a wild-type background (Fig. 1A). Untreated mib mutants lacked 5HT neurons but, remarkably, 5HT neurons were generated in mib mutant embryos exposed to VPA (Fig. 5C). These findings confirmed that the loss of 5HT expression in VPA-treated embryos was not due to neuronal death, but instead was due to a block in their differentiation.

DISCUSSION
Using a zebrafish model of FVS, we have identified a molecular mechanism to explain the loss of 5HT neurons, which have been implicated in the neuropsychiatric manifestations of FVS. Two different VPA treatment regimes caused the failure of serotonergic differentiation, and the late-dosing regime (24-48 hpf) seemed to have greater selectivity for the serotonergic system than the earlydosing schedule (50% epiboly to 27 hpf). There is wide variation in the teratogenic dose of VPA in human and animal studies (Christensen et al., 2013;Ornoy, 2009), which makes direct comparisons of dosing regimes between different model organisms and with humans exposed to VPA in utero difficult to interpret. Nevertheless, the altered serotonergic differentiation in this zebrafish model is reminiscent of similar deficits that are found in rodent models, which correlate with autism-like behavioural abnormalities (Dufour-Rainfray et al., 2010;Dufour-Rainfray et al., 2011).
Recovery of ascl1b expression in VPA-treated hdac1 s436 or mib mutant embryos is associated with rescue of 5HT neuronal differentiation. However, direct replacement of Ascl1b is sufficient to rescue 5HT expression in only 16% of embryos. Suboptimal timing or level of the ectopically expressed Ascl1b expression might explain the lower efficiency. Alternatively, non-physiological, persistent high-level expression of Ascl1b in post-mitotic neurons might account for the low rate of rescue (Cai et al., 2000). Nevertheless, taken together these findings indicate that the failure Lines with an orthogonal bar represent inhibition. The inhibitory arrow with a question mark above it, from Hdac1 to Her, takes account of the possibility that another Her gene (or another transcription factor of unknown identity) that represses ascl1b is in turn repressed by Hdac1.
of 5HT neuronal differentiation in our model of FVS is most probably due to the loss of ascl1b expression. We cannot, however, exclude the possibility that VPA disrupts the expression of other, unknown, genes that are crucial for 5HT neuronal differentiation. We anticipate that the finding of downregulated expression of ascl1b in our zebrafish model will prompt closer analysis of rodent FVS models for changes in the expression of the mammalian ortholog Ascl1 in serotonergic progenitors. Ascl1 is known to be required for neuronal-subtype specification, which is separable from its betterknown proneural function (Goridis and Brunet, 1999). In mammals, a requirement for Ascl1 in serotonergic differentiation has previously been demonstrated, including for the expression of 5HT itself, and this subtype-specification function cannot be substituted by other proneural genes (Jacob et al., 2009;Parras et al., 2002;Pattyn et al., 2004). Additionally, Ascl1 has been shown to be essential for the acquisition of noradrenergic traits in all mammalian central and peripheral neurons (Blaugrund et al., 1996;Guillemot et al., 1993;Hirsch et al., 1998). At least in part, Ascl1 regulates neurotransmitter phenotype in noradrenergic neurons in collaboration with Phox2 genes (Pattyn et al., 2000). However, Phox2 genes are not expressed in the mammalian serotonergic lineage. Instead, Ascl1 regulation of the zinc-finger transcription-factor-encoding gene Insm1 in postmitotic serotonergic precursors is a crucial component of the mammalian serotonergic transcriptional programme (Jacob et al., 2009).
Using the HDAC inhibitor TSA and an hdac1 mutant line, we have provided evidence that the effect of VPA on serotonergic differentiation is mediated through Hdac1 inhibition. These findings are consistent with previous reports that VPA inhibits HDACs in vivo (Gurvich et al., 2005). Moreover, Hdac1 is one of the key molecular factors that regulate the transcription of ascl1b and these data are consistent with the previously identified positive regulatory role of Hdac1 (Cunliffe, 2004;Harrison et al., 2011). The regulation of ascl1b might be direct given that Hdac1 binds to the ascl1b promoter (Harrison et al., 2011). It has been proposed that Hdac1 promotes activation of ascl1b directly by participating in cycles of deacetylation and histone-acetyltransferase-dependent acetylation of transcription-unit-associated histones, and/or by maintaining the ascl1b promoter in a transcriptionally poised state (Wang et al., 2009). Nevertheless, we do not rule out the possibility that Hdac1 functions indirectly by inhibiting the expression of a transcriptional repressor of ascl1b (Fig. 4D).
A second transcriptional regulator of ascl1b that is unmasked by VPA is repressive regulation by the Notch pathway. It was shown previously that VPA upregulates Notch signalling and this is associated with increased expression of Notch effector genes (Greenblatt et al., 2007;Stockhausen et al., 2005). Therefore, we expected that repression of ascl1b by VPA might be mediated by enhanced Notch signalling. However, short-duration VPA treatment is sufficient to downregulate ascl1b without concomitantly enhancing Notch signalling. Consistent with this, the expression of a panel of Her genes, including her4, a target of Notch signalling, and her6, are not appreciably altered under these conditions. Nevertheless, our findings using chemical and genetic blockade of Notch function demonstrate that parallel Notch signalling participates in the VPA-mediated downregulation of ascl1b (Fig. 4D). Importantly, only a basal level of Notch signalling is required for VPA downregulation of ascl1b. Together, our findings lead to a revised view of ascl1b regulation based on a derepression model, because removal of the basal inhibitory Notch input when Hdac1 function is blocked by VPA is sufficient for ascl1b transcription.
The regulatory relationship between Hdac1 and the Notch pathway in the control of ascl1b expression reported in a previous study (Cunliffe, 2004) provided an opportunity to evaluate further whether VPA blockade of serotonergic differentiation occurs via Hdac1 inhibition. In hdac1 mutants, we found that her4 expression is reduced (Fig. 3B). Our result differs from the latter study that proposed increased Notch signalling in hdac1 mutants based on enhanced her6 expression (Cunliffe, 2004). A possible explanation for the difference is that her6 expression in hdac1 mutants is developmental-stage-dependent. The loss of ascl1b expression in hdac1 s436 mutants and in VPA-treated embryos (Fig. 2C) occurs in the absence of enhanced Notch pathway activity during the period of treatment. This is consistent with Hdac1 inhibition as the salient in vivo mechanism of VPA. Moreover, these data are also consistent with direct regulation of ascl1b by Hdac1.
Because VPA has been reported to have a variety of modes of action, identifying which of these mechanisms is relevant in vivo for the behavioural manifestations of FVS is challenging. This is an important question, however, because identification of the molecular mechanism has potential therapeutic significance. The involvement of Hdac1 in the serotonergic deficit in FVS strongly suggests that an epigenetic mode of action could mediate at least some of the behavioural manifestations of FVS. This is not an isolated finding, because epigenetic contributions to other ASDs have been noted. Methyl-CpG-binding protein 2 (MECP2), which is implicated as the cause of the ASD-associated disease Rett syndrome contains a transcriptional repression domain, which physically interacts with the transcriptional co-repressor Sin3A (Amir et al., 1999). In turn, Sin3A recruits HDAC1 and HDAC2, which are believed to mediate transcriptional repression by MECP2 (Jones et al., 1998;Nan et al., 1998). More broadly, epigenetic modifications seem to regulate autism susceptibility; for example, autism is associated with duplications of 15q11-13, which is an imprinted region of the genome where DNA methylation status is associated with Prader-Willi syndrome and Angelman syndrome (Dykens et al., 2011). In conclusion, epigenetic regulation of gene expression seems to be an emerging point of convergence for environmental agents and hereditary factors in ASD pathogenesis.

Transgenic constructs and transient transgenic zebrafish
DNA encoding Ascl1b tagged with six copies of Myc at the C-terminus was made by gene synthesis (Genwiz). ascl1b-myc was subcloned into a 5× UAS plasmid containing a miniTOL2 backbone to facilitate genomic integration in zebrafish (Balciunas et al., 2006). Transient transgenic embryos were generated by co-injecting 10-20 pg of UAS:ascl1b-myc plasmid DNA with 25 pg of tol2 transposase mRNA into one-cell-stage embryos obtained by intercrossing UBI:ERT2-GAL4 transgenic zebrafish.

Reverse transcriptase PCR (RT-PCR) and primer sequences
Between 30 and 40 embryos at 32 hpf were used for each condition and experiments were performed in triplicate. At the end of the period of drug treatment embryos were placed in Trizol, homogenised, chloroform was added, the sample was centrifuged at 4°C, glycogen and isopropanol were added, the sample was then centrifuged again at 4°C, 75% ethanol was added, centrifugation at 4°C was repeated and the pellet was re-suspended in water. RNA purification was performed using the RNeasy Micro Kit (Qiagen) according to the manufacturer's instructions. cDNA synthesis was performed using the Superscript First-Strand Synthesis System for RT-PCR (Invitrogen), according to the manufacturer's instructions. Quantitative RT-PCR was performed using the 7900HT Fast Real-Time PCR System (Applied Biosystems).

Immunohistochemistry and in situ hybridisation
For immunofluorescence and in situ hybridisation, embryos were fixed in 4% paraformaldehyde (PFA) either for 2-3 hours at room temperature or overnight at 4°C. For immunofluorescence, embryos were rinsed three times in 0.1% Tween in phosphate buffered saline (PBST), incubated for 20 minutes in 0.01 mg/ml proteinase K (Roche) in PBST/0.2% BSA/2% heatinactivated goat serum solution (PBT), post-fixed in PFA, washed three times in PBT, blocked in PBT containing 1% DMSO, 0.5% Triton-X and 1% BSA (PBDT), and then incubated for 24 hours at 4°C in primary antibody in PBST. The following day, embryos were washed four times in PBT, blocked again in PBDT and then incubated in secondary antibody overnight in PBDT. Further washes in PBT were performed the next day, and embryos were then transferred to 70% glycerol and mounted under coverslips for viewing. Immunofluorescence staining was visualised by confocal microscopy (Leica TCS SP2). For single and double in situ hybridisation, embryos were processed as described (Lauter et al., 2011). Fluorescent signals were developed with Fast Blue and Fast Red. Differences between embryos were analysed using Fisher's exact test.

Drugs
Embryos were dechorionated manually prior to drug treatment. Sodium valproate (Sigma) was dissolved in water as a 2 M stock solution and was used at a final concentration of 0.625 mM. This dose is comparable to doses found to inhibit HDACs in vivo in other studies (Gurvich et al., 2005). Lower doses did not consistently abolish 5HT expression in the hindbrain (see supplementary material Fig. S1). DAPT (Calbiochem) was dissolved in

RESEARCH ARTICLE
Disease Models & Mechanisms (2014) doi:10.1242/dmm.013219 DMSO as a 46 mM stock solution and used at 200 μM final concentration, which is similar to doses previously reported to block Notch signalling in zebrafish (Geling et al., 2002). A 5 mM TSA solution in DMSO (Sigma) was diluted to a final concentration of 2 μM in fish water prior to use, which is comparable to doses used in vivo to inhibit HDACs (Gurvich et al., 2005). 4-OHT (Sigma) was dissolved in ethanol at 12.5 mg/ml. Dilutions of 4-OHT were made to 1.5 μM in standard fish water just prior to use. In all drug conditions, controls were treated with an equivalent amount of vehicle diluted in fish water. Stock solutions of all these drugs were stored at −20°C.