Previous studies of the hdac1 mutant embryonic CNS revealed extensive defects of neuronal specification and patterning, along with an absence of differentiated oligodendrocytes 9 10 11 12 . When the levels of Hdac1 protein, histone acetylation and histone methylation were compared in hdac1 mutant and wild-type sibling embryos between 27 hpf and 96 hpf, hdac1 mutants exhibited consistently reduced levels of Hdac1 protein, together with a large, persistent increase in global histone acetylation and a stable reduction in H3K9 methylation, in comparison to wild-type siblings (Figure 1). By contrast, levels of H3K4 methylation were mostly unaffected by loss of hdac1 function. The major differences in the abundance of acetylated histones and methylated H3K9, that were detected between hdac1 mutant and wild-type sibling embryos, demonstrated that Hdac1 plays a key role in the epigenetic regulation of chromatin structure, and implied that there would be major impacts of Hdac1 function on transcriptome composition. Therefore, to develop a systems-level understanding of Hdac1 function during embryonic development, we used a global gene expression profiling platform to identify the Hdac1-regulated transcriptome in developing zebrafish embryos. By defining temporal changes in the composition of the Hdac1-regulated transcriptome from the beginning of neurogenesis onwards, we sought to identify genes whose normal transcript levels exhibited a continuous dependence on Hdac1 function throughout neurogenesis, and which could thus be important direct effectors of Hdac1 function in the developing CNS.

As a starting point, 27 hpf hdac1 mutant and wild-type sibling embryos were sorted into morphologically distinct groups. RNA was extracted from sorted embryos, labelled and hybridised to Agilent arrays comprising 43427 oligonucleotide probes, which corresponded to 18636 unique Unigene clusters and approximately 2000 additional Ensembl-annotated genes 20 . Using a p-value threshold of <10^-5 to define statistically significant expression differences between hdac1 mutant and wild-type sibling samples, we observed that 4345 probes exhibited increased expression in the 27 hpf hdac1 mutant samples, whereas 4142 probes exhibited decreased expression in these samples (Figure 2A, B). 629 of these Hdac1-regulated probes exhibited a 2-fold or greater increase or decrease in transcript abundance as a result of the hdac1 mutation.

Comparison of hdac1 mutant and wild-type transcriptomes at 27 hpf identified large numbers of Hdac1-regulated genes, many of which were likely to be indirect, downstream effectors of the primary target genes for Hdac1. To identify candidate Hdac1 direct target genes, we sought to identify genes whose expression was regulated by Hdac1 during early developmental stages that preceded the emergence of a morphologically recognisable hdac1 mutant phenotype. Since morpholinos can be used to create defined batches of embryos that are all deficient in a specific gene function, we compared the phenotypes of embryos microinjected with a translation-blocking morpholino, Hdac1ATG1 9 or a splice-blocking morpholino Hdac1SPL1, with the phenotype of hdac1 mutant embryos. Microinjection of either Hdac1ATG1 or Hdac1SPL1 morpholinos into embryos caused morphological abnormalities that closely resembled the hdac1 mutant phenotype (data not shown), and which were first detectable at ~24 hpf. Moreover, both the Hdac1ATG1 and Hdac1SPL1 morpholinos significantly reduced Hdac1 protein levels in the embryo, although the Hdac1ATG1 morpholino induced a larger reduction in Hdac1 protein levels than did the Hdac1SPL1 morpholino (Figure 2C). By contrast, neither an Hdac1 Mismatch Control morpholino (Hdac1 Control), nor a Standard Control morpholino, exhibited any appreciable effect on embryo morphology, or on levels of Hdac1 protein (Figure 2C).

In order to begin to elucidate the effects of Hdac1ATG1 and Hdac1SPL1morpholinos on embryonic transcription, the transcriptomes of these morphant embryos were analysed at 27 hpf. Both the Hdac1ATG1 and Hdac1SPL1 morpholinos elicited major effects on the embryonic transcriptome (Additional file 1). A set of 6117 probes was identified, from within the group of probes whose expression is affected in the hdac1mutant, that also exhibited statistically significant (p < 10^-5) Hdac1-regulated gene expression in 27 hpf morphant samples. This set of 6117 probes was further analysed with the Explore Gene Ontology (eGOn) software, to identify Gene Ontology (GO) categories that were significantly enriched in the Hdac1-regulated probe set (Additional file 2). The Biological Process category was found to be considerably over-represented in the annotations of Hdac1-regulated genes, within which the sub-categories of GO terms related to Transcriptional Regulation (highlighted in purple, Additional file 2) or Developmental Processes occurred particularly frequently. Other prominent GO terms that are enriched in the Hdac1-regulated probe set include nervous system development, neuronal differentiation and photoreceptor differentiation (Additional file 2). Thus, both hdac1 mutant and morphant transcriptomes exhibit similar alterations to the expression of genes that encode transcription factors and developmental regulators, with an emphasis on development of the nervous system. These results implied that Hdac1 performs a central role in regulating expression of these classes of genes and further validated the morpholinos as tools for perturbing Hdac1 function during nervous system development.

In order to determine which of the two Hdac1 morpholinos most closely phenocopied the hdac1 mutation, hierarchical cluster analysis 21 was used to order the transcriptome datasets obtained from 27 hpf hdac1 mutant, Hdac1ATG1 and Hdac1SPL1 morphant embryos, according to the similarities in the patterns of gene expression changes observed in each of the three different manipulations (Additional file 1). This analysis showed that overall, the set of Hdac1-regulated genes that was identified using the Hdac1ATG1 morpholino, more closely resembled the group of genes identified by analysis of 27 hpf hdac1 mutant embryos than did those identified using the Hdac1SPL1 morpholino. The Hdac1ATG1 morpholino was therefore chosen to investigate Hdac1 function during early stages of CNS development before a morphologically distinct phenotype can be recognised. As a first step in this analysis, we determined whether this morpholino could reduce Hdac1 protein levels from early stages of CNS development onwards. Hdac1ATG1 morphant embryos and individually genotyped wild-type and hdac1 mutant embryos, aged 12 hpf, 18 hpf and 27 hpf (corresponding to early, mid- and late neurogenesis stages, respectively) were immunostained with an anti-Hdac1 antibody and analysed by confocal imaging (Figure 2D). Whilst Hdac1 protein was abundant in all cell nuclei of wild-type embryos at each of the stages analysed, both hdac1 mutant and Hdac1ATG1 morphant embryos exhibited a substantially lower level of Hdac1 protein, which was more pronounced at later developmental stages (Figure 2D). The results of the transcriptomic analysis, together with the experiments to compare the distribution of Hdac1 protein in hdac1 mutant and morphant embryos, revealed a close correspondence between the 27 hpf hdac1 mutant and morphant phenotypes and validated the Hdac1ATG1 morpholino as a precision tool with which to investigate Hdac1 function in the embryo between 12 hpf and 27 hpf.

To identify genes whose expression was regulated by Hdac1 during early stages of neurogenesis, pools of embryos were prepared that were injected with either the Hdac1ATG1 morpholino or the Standard Control morpholino, and samples were collected at 12 hpf and 18 hpf for RNA extraction and analysis. At 12 hpf, neurogenesis has just begun, by 18 hpf it is well underway, and by 27 hpf this process is widespread throughout the nervous system. Transcriptomes were analysed and the Hdac1-regulated gene datasets were compared to those obtained from analysis of 27 hpf Hdac1ATG1 morphants (Figure 3). The results of this cluster analysis revealed that considerably more differential gene expression was detectable between the transcriptomes of Hdac1ATG1 and Standard Control morphant embryos at 12 hpf and 18 hpf, than was detectable at 27 hpf, even though there are no appreciable morphological abnormalities in Hdac1ATG1 morphant embryos before 22 hpf.

The cluster analysis identified several discrete gene classes with temporally distinct, differential expression profiles in Hdac1ATG1 and control morphant embryos, including a small subset of genes that exhibited consistent, statistically significant transcriptional sensitivity to loss of Hdac1 at all three stages analysed (arrows in Figure 3). Thus, in Hdac1ATG1 morphant embryos, 199 probes were up-regulated at 12 hpf, 18 hpf and 27 hpf, whereas 335 probes were down-regulated at all three time points, in comparison to their expression levels in Standard Control morphants. In order to identify the genes exhibiting the greatest degree of sustained, differential gene expression at all three time points, probes exhibiting a >2-fold change in expression level at all three time points were selected. 84 of these probes exhibited a >2-fold change in expression in Hdac1ATG1 morphants at all three time points analysed (Additional file 3). However, in view of the fact that morpholinos can sometimes elicit off-target effects that are not recapitulated by a corresponding mutant, we applied an additional exclusion criterion to eliminate probes that exhibited less than a 1.2-fold change in expression (of the same polarity) in 27 hpf hdac1 mutants, as compared to wild-type embryos (Additional file 3). Applying this criterion removed 17 of the 84 probes that exhibited differential expression in Hdac1ATG1 and Standard Control morphants, thus reducing the number of probes in this group to 67. Of these Hdac1-regulated probes, 56 exhibited decreased transcript abundance in hdac1 mutant and Hdac1ATG1 morphant embryos at each of the time points analysed, whereas 11 of the 67 probes exhibited consistent increases in transcript abundance (Figure 4). The 56 down-regulated probes correspond to 43 distinct down-regulated genes, of which 28 are specifically expressed in the CNS or have a known CNS-oriented function. Of these 28 genes, 18 encode transcription factors with known or likely roles in programming of CNS development: dbx2, bhlhe22, neurod4, ascl1a, ascl1b, crx, gsx1, foxn4, lbx1b, pou23, rx2, fezf1, neurod, mab21l1, hoxb6b, lhx9, hoxa3a, and nr4a2. In addition, the dlb gene, encoding the nervous system-specific Notch ligand, DeltaB, is another robustly Hdac1-dependent gene. Other genes implicated in CNS development that were down-regulated in Hdac1-deficient embryos include genes encoding the Retinoic Acid-metabolising enzyme Cyp26b1, GFAP, Fads2 and Atp1a1a.2. The two genes exhibiting the greatest fold reduction in transcript abundance in Hdac1ATG1 morphants across the three time points analysed are nsun5, which encodes a protein of unknown, potentially nuclearly-localised function, with a putative methyltransferase-encoding NOL1/NOP2/Sun domain, and a novel gene with no significant sequence similarity to known genes (Unigene ID Dr.136125). Of the 11 up-regulated probes, 4 are specifically expressed in the CNS or have a known CNS-oriented function, such as trim9 and fjx1, but none encode known transcription factors. Taken together, these results demonstrate that Hdac1 promotes the co-ordinated expression of a core set of neurogenic transcriptional regulators, which we define here as the Hdac1 neurogenic programme, from early stages of neurogenesis onwards. These results are fully consistent with the Gene Ontology analysis of the hdac1 mutant (Additional file 2), which identified genes involved in transcriptional control and developmental mechanisms, including nervous system development, as the principal targets of Hdac1-mediated transcriptional regulation.

To document the spatial extent of Hdac1 function in the neural plate during neurogenesis, the expression patterns of a subset of components of the Hdac1 neurogenic programme, neurod4, ascl1b, neurod, lhx9 and dlb, were characterised by whole-mount in situ hybridisation of 6 somite-stage (12 hpf) and 27 hpf Hdac1ATG1 morphant and control morphant embryos (Figure 5). We observed that Hdac1ATG1 morphants exhibited considerably reduced expression of these genes throughout the developing CNS at both 12 hpf and 27 hpf. Coinjection of a p53-targeted morpholino, which eliminates off-target effects caused by morpholino injection, had no observable effects on any of the gene expression patterns analysed (Figure 5). By contrast, the expression pattern of sox2, which is expressed widely in the developing nervous system, was unaffected by loss of hdac1 function at 12 hpf (Figure 5). Consistent with the observations in Hdac1ATG1 morphant embryos, the expression pattern of ascl1b was also appreciably reduced in individually genotyped hdac1 mutant embryos at 12 hpf and 18 hpf, in comparison to its expression level in wild-type sibling embryos (Figure 6A,B). When the expression patterns of neurod4, ascl1b, neurod, lhx9 and dlb were then compared in 27 hpf hdac1 mutant embryos, expression of all five genes was dramatically reduced throughout the CNS, as was observed in 27 hpf Hdac1ATG1 morphant embryos (Figure 6C). Taken together, these results suggest that, from early stages of neurogenesis onwards, Hdac1 promotes the co-ordinated transcription of developmental regulatory genes, at the core of which lies a set of transcription factors that play key roles in regulating neuronal specification and patterning throughout the CNS.

Previous studies from our laboratory demonstrated that Hdac1 antagonises Notch signalling and attenuates expression of the Notch target gene her6 during CNS development 9 . Moreover, the time-resolved microarray analysis confirmed that statistically significant increases in her6 expression could be detected in hdac1 mutants and Hdac1ATG1 morphants at 27 hpf, but by contrast, statistically significant changes in her6 expression were not detected in 12 hpf and 18 hpf hdac1 morphant embryos, which excluded it from the group of robustly Hdac1-regulated genes defined by the microarray analysis (Additional file 3). However, when her6 expression was analysed by in situ hybridisation, increased expression was observed in the caudal hindbrain and developing optic vesicles of 12 hpf Hdac1ATG1 morphants, and the increased her6 expression persisted in the morphant hindbrain at 18 hpf (Figure 7A-F). Thus, although these changes were too small to be reliably detected by the whole-embryo microarray analysis, our in situ data show that in discrete regions of the CNS, Hdac1 represses her6 transcription from early stages of neurogenesis onwards, which is consistent with previous observations in older embryos 9 .

To determine whether HDAC activity is required for correct expression of Hdac1-regulated genes specifically during early neurogenesis, wild-type embryos were incubated with the HDAC inhibitor Trichostatin A (TSA) between 10 hpf and 14 hpf, then fixed and analysed for gene expression by in situ hybridisation. These experiments revealed that expression of neurod4, ascl1b, neurod, dlb and lhx9 in the embryonic CNS was almost completely extinguished after incubation in TSA for 4 hours (Figure 8). Moreover, TSA treatment of embryos from 10-14 hpf caused increased expression of her6 in the caudal hindbrain and optic vesicles, as was observed in 12 hpf Hdac1ATG1 morphants (Figure 7G-I). Thus, we conclude that HDACs promote transcription of neurod4, ascl1b, neurod, dlb and lhx9 throughout the CNS, and attenuate expression of her6 in the hindbrain and optic vesicles, during early stages of neurogenesis. In view of the fact that all of these genes are specifically regulated by Hdac1, it seems likely that the effects of TSA during early neurogenesis are mediated via direct inhibition of Hdac1.

Genome-wide studies of HDAC binding in mammalian T cells have revealed a strong, discrete peak of HDAC1 binding within the chromatin encompassing the transcription start site of many transcriptionally active genes as well as a smaller number of inactive genes, suggesting a direct role for this chromatin modifying enzyme in local control of chromatin structure at both transcribed and silent target gene promoters 7 . The specific enrichment of HDAC1 at the promoters of active genes enables chromatin structure to be reset for subsequent re-acetylation of histones during transcription elongation 7 . Our transcriptomic analysis revealed key roles for Hdac1 in promoting expression of proneural genes and other neural fate-determining factors in the developing CNS, as well as in limiting expression of other genes such as her6, within the CNS. Furthermore, we found that pharmacological inhibition of HDAC function caused a rapid shut-down of the neurogenic programme throughout the CNS of 12 hpf embryos, as well as increasing her6 expression in restricted regions of the CNS at this stage. We therefore wondered whether Hdac1 might directly bind to the promoter elements of neural fate-determining genes during neurogenesis, in order to directly regulate their transcription. To investigate this possibility, 12 hpf and 27 hpf wild-type embryos were subjected to chromatin immunoprecipitation analysis with the anti-Hdac1 antibody, which was validated to specifically and efficiently immunoprecipitate Hdac1 from embryonic chromatin (Additional file 4). Immunoprecipitated DNA samples were then analysed for the presence of specific sequences within the proximal promoter region, the first exon and the first intron of ascl1b, neurod4, deltab and her6 (Figure 9). These four genes were chosen for this analysis because their expression patterns in the developing CNS were consistently Hdac1-regulated between 12 hpf and 27 hpf. Surprisingly, robust and specific binding of Hdac1 was detected throughout the region between -785bp and -85bp upstream of the ascl1b transcription start site in both 12 hpf and 27 hpf wild-type embryos (Figure 9A,B). In 12 hpf embryos, Hdac1 binding was also detected in chromatin fragments spanning +69 bp to + 175 bp and +686 bp to +790 bp downstream of the ascl1b transcription start site. In striking contrast, no specific binding of Hdac1 to promoter-proximal sequences of neurod4, dlb or her6 was detectable. Comparative analysis of Hdac1 binding to the ascl1b promoter region in 27 hpf hdac1 mutant and wild-type siblings further revealed that all of the statistically significant Hdac1 binding to the ascl1b promoter, in the region between -785bp and -85bp upstream of the ascl1b transcription start site, was lost in hdac1 mutant embryos (Figure 9C). Our results demonstrate that Hdac1 specifically binds to a region of the ascl1b promoter between -785 and -85 bp upstream of the transcription start site and promotes ascl1b expression likely in response to neural signals. These results also indicate that no binding of Hdac1 to DNA sequences close to the corresponding transcription start sites of dlb, neurod4 or her6, could be detected.

In zebrafish and mouse embryos, Hdac1 promotes the transformation of neural progenitors into differentiated neurons and glia. The experiments reported here further elucidate the mechanism of Hdac1 function in zebrafish CNS development. We initially sought to identify all of the genes whose expression levels were robustly Hdac1-regulated in 27 hpf embryos. We found that approximately 20% of all the probes on the microarrays exhibited statistically significant differences in transcript abundance in 27 hpf hdac1 mutant and wild-type sibling embryos, although only 7% of these Hdac1-regulated probes exhibited more than a two-fold change in transcript abundance. However, as this comparative analysis was performed on morphologically distinct 27 hpf hdac1 mutants and wild-type siblings, it seemed likely that many of the genes identified in this way as being Hdac1-regulated are indirect downstream effectors of a smaller programme of gene expression that is directly regulated by Hdac1 from earlier stages of neurogenesis. Therefore, to identify transcriptional changes that were likely to be more direct consequences of loss of Hdac1 function, we used the Hdac1ATG1 morpholino to generate defined batches of Hdac1-deficient embryos, whose transcriptomes were then analysed well before any morphological differences appeared in these embryos. Gene Ontology analysis of Hdac1-regulated genes, in both hdac1 mutant and Hdac1ATG1 morphant embryos, revealed a consistent and robust enrichment of genes with known functions in both transcriptional control and CNS development. We then discovered within this group a set of 54 genes that exhibited consistently increased or decreased changes in transcript abundance in Hdac1-deficient embryos at 12 hpf, 18 hpf and 27 hpf. More than half of these genes are implicated in CNS development and down-regulated in Hdac1-deficient embryos. Within this set of co-regulated genes, we identified a core subset of 18 genes encoding sequence-specific DNA binding transcription factors, whose down-regulation could readily account for many of the neural specification defects that characterise the Hdac1-deficient CNS. Interestingly, the gene that exhibited the greatest-fold decrease in gene expression in hdac1-deficient embryos encodes a novel member of the methyltransferase-encoding NOL1/NOP2/Sun domain family of proteins, Nsun5, but nothing currently is known about the function of this protein.

Many of the 18 Hdac1-regulated CNS-specific transcription factor genes that were identified in our experiments play important roles in the specification of neuronal subtype identities in the spinal cord, brain and retina. Thus, dbx2, lhx9, lbx1b, gsx1, bhlhe22, ascl1a, ascl1b or their vertebrate orthologues have been implicated in neuronal specification within the spinal cord 22 23 24 25 26 . In the brain, neurod, neurod4 and bhlhe22 are are required for the production of cortical projection neurones 27 28 , whilst upper motoneurones of the corticospinal tract depend on fezf1 29 and specification of branchiomotor neurones of the mouse hindbrain requires the combined activities of ascl1 and neurod4 orthologues 30 . In the vertebrate retina, formation of amacrine cells is positively regulated by orthologues of ascl1a/b, neurod, neurod4, foxn4, bhlhe22 and nr4a2 31 32 33 34 . Moreover, mammalian photoreceptor specification also requires neurod, crx and rx functions 35 36 37 . Taken together with these observations, our genome-wide expression analysis indicates that Hdac1 is deployed widely throughout the developing CNS to specify a wide variety of neuronal subtypes, by promoting the transcription of this core group of developmental regulatory genes. This conclusion was confirmed by our in situ hybridisation analysis of the expression patterns of proneural/bHLH genes ascl1b, neurod4, and neurod, along with lhx9 and the Notch ligand gene dlb. Transcription of these genes was considerably reduced throughout the Hdac1-deficient CNS, both at 12 hpf and at 27 hpf, and transient incubation of embryos in the HDAC inhibitor TSA between 10 hpf and 14 hpf almost completely extinguished their expression. By contrast, widespread expression of the neural transcription factor gene sox2 was not appreciably altered at 12 hpf, demonstrating that at this early stage, the reduced expression of ascl1b, neurod4, neurod, dlb and lhx9 in Hdac1-deficient embryos is not due to reduction in the pool of neural precursors, but rather the result of a failure to specify particular neuronal identities. Intriguingly, forced co-expression of the murine proneural gene Ascl1 (the orthologue of zebrafish ascl1a and ascl1b) with a small set of neuronal transcription factors in cultured murine fibroblasts, was recently shown to directly re-programme these cells to differentiate into distinct neuronal subtypes 38 . Remarkably, this group of collaborating murine transcription factors, defined by Vierbuchen et al., includes close relatives of brn1 (Brn2), neurod (Neurod1), lhx9 (Lhx2) and nr4a2 (Nr2f1), all of which we have found to exhibit stringent Hdac1-dependent co-expression throughout zebrafish neurogenesis. Interestingly, the murine Ascl1 and Brn1 proteins co-regulate a neurogenic programme by co-operative binding to a conserved DNA sequence motif in the cis-regulatory regions of Delta genes 39 . Thus, loss of brn1 and ascl1b expression in hdac1 mutant CNS could be responsible for the reduced expression of dlb that was also observed in hdac1 mutant zebrafish embryos. Overall, our analysis has uncovered a group of co-regulated neurogenic transcription factors that may function as a regulatory network driving neuronal patterning and differentiation in the CNS. It will now be of interest to investigate whether expressing particular combinations of these transcription factors in hdac1 mutant embryos can rescue defects of neuronal specification that are caused by loss of Hdac1 function.

Somewhat surprisingly, relatively few genes were identified in the time-resolved transcriptome analysis whose expression was robustly repressed by Hdac1 between 12 hpf and 27 hpf. Overall, 11 genes were identified, of mostly unknown functions, which exhibited a consistent, two-fold or greater increase in expression in Hdac1-deficient embryos across this period of development. Of these genes, 4 have been implicated previously in aspects of CNS development, raising the possibility that their increased transcription could contribute to the defects of neural development in Hdac1-deficient embryos. TRIM9, for example, encodes a brain-specific E3 ubiquitin ligase 40 and fjx1 encodes a Notch-regulated inhibitor of neuronal dendrite branching that is expressed in both the central and peripheral nervous systems 41 . Our previous work demonstrated that Hdac1 repressed the Notch target her6 at 26hpf and 33hpf 9 , but her6 was not initially identified as an Hdac1-regulated gene in the time-resolved transcriptome analysis because its transcript abundance was not statistically significantly changed in 12 hpf and 18 hpf morphant embryos. Nevertheless, whole mount in situ hybridisation to Hdac1ATG1 morphant embryos revealed that her6 expression was appreciably increased in the hindbrain and optic vesicles at 12 hpf, which was confirmed in TSA-treated embryos, and that the increased her6 expression in the hindbrain persisted in 18 hpf hdac1 morphants. Thus, we conclude that Hdac1 attenuates her6 expression in restricted regions of the CNS from early stages of neurogenesis onwards.

A recent microarray-based study of HDAC function in the differentiation of mouse retinal explants revealed that transcription of genes involved in promoting photoreceptor specification, such as Crx, Neurod4 and Neurod, as well as Otx2 and Nrl was rapidly suppressed within 3 hours of administering TSA to retinal explants 42 . Moreover, attenuation of Crx and Nrl expression by TSA was independent of a requirement for protein synthesis, implying a direct role for HDAC enzymes in promoting the transcription of these genes. We observed a similarly rapid and near complete extinction of ascl1b, neurod4, dlb, neurod and lhx9 expression, when 10 hpf embryos were incubated in TSA. Interestingly, many HDACs, including HDAC1, are abundant at the promoters of active genes in mammalian T lymphocytes, and closely associated with both HATs and phosphorylated RNA Polymerase II in the transcribed regions of these genes 7 . In yeast, HDACs biochemically oppose HAT functions by removing acetylation marks from chromatin at active genes, which facilitates their HAT-mediated re-acetylation in subsequent cycles of re-transcription. Accordingly, in yeast HDAC mutants, genes that are embedded within hyperacetylated chromatin are transcriptionally impaired 4 5 6 , and moreover, similar observations have been made in mammalian T-cells 7 . Thus, it seems possible that some of the hyperacetylated histones we observed in hdac1 mutant zebrafish embryos are associated with genes whose transcription has been impaired by loss of Hdac1 function. The finding that Hdac1 is stably and specifically associated with sequences within the promoter region and first exon of ascl1b in 12 hpf and 27 hpf zebrafish embryos, transcription of which is Hdac1-dependent throughout this period, identifies ascl1b as a novel in vivo direct target of Hdac1. To our knowledge, this is the first such direct in vivo target to be defined for Hdac1 in vertebrate embryos. Taken together, our results suggest that Hdac1 could promote transcription of this proneural gene directly, by removing acetyl modifications from transcription unit-associated histones to enable their transcription-coupled re-acetylation, and/or by maintaining the ascl1b promoter in a deacetylated, H3K4 methylated, transcriptionally poised state. Indeed, the fact that global H3K4 methylation levels were unchanged by loss of hdac1 function (Figure 1) suggests that Hdac1 may bind to promoters within H3K4 methylated chromatin of zebrafish embryos, thus poising them for activation 7 . In view of the fact that our experiments were performed on whole embryos, it is possible that Hdac1 also binds to the ascl1b promoter in non-neuronal cells, although the significance of such possible interactions is unclear, as no ectopic expression of ascl1b was observed in Hdac1-deficient embryos. Future studies will aim to identify the DNA sequences within the ascl1b promoter, and the cognate DNA binding proteins, which recruit Hdac1 to this target gene, and determine whether these interactions are specific for the neuronal lineage in the developing CNS. We also intend to investigate how the distribution of epigenetic modifications across the genome is regulated by Hdac1.

Our in vivo chromatin immunoprecipitation studies detected no interactions between Hdac1 and the DNA sequences immediately upstream and downstream of the transcription start sites of neurod4, neurod, deltab and her6. Whilst it is possible that these genes are regulated by Hdac1 indirectly, it is also conceivable that Hdac1 binds directly to other cis-regulatory elements that are located further away from the transcription start sites of these genes. Another possibility is that the level of Hdac1 binding to these promoters in vivo may be below the threshold for detection in the chromatin immunoprecipitation assay. Despite the lack of Hdac1 binding to her6 promoter sequences, this gene remains a candidate direct target for direct repression by Hdac1, as Hdac1 repressed her6 expression in restricted regions of the CNS from 12 hpf onwards. Indeed, it remains possible that Hdac1-mediated repression of her6 facilitates the transcription of proneural genes such as ascl1b independently of the physical interaction between Hdac1 protein and the ascl1b promoter that we describe here. More extensive studies of Hdac1 binding to the chromatin in which these and other genes are embedded, and identification of proteins that recruit Hdac1 to its direct target genes, will allow the Hdac1-regulated genes identified by our transcriptome analysis to be evaluated further as candidate direct targets for Hdac1 binding. This information will help to define better the functional interrelationships between components of the Hdac1-regulated genetic network that promotes neurogenesis.

hdac1 ^hi1618 mutant zebrafish were maintained at University of Sheffield. Individual embryos were genotyped using primers for hdac1 (Hdac1F: 5'-GGC AGG CGC AGG CTG TAA TT-3'; Hdac1 Intron1R: 5'-GGC TAA ACC CGG CTA ACA AT-3'); the 3' Long Terminal Repeat of the MLV vector (MLV 3LTR F: 5'- AAA GAC CCC ACC TGT AGG TTT G-3') 43 . Animal care and use was in accordance with the UK Animals (Scientific Procedures) Act 1986.

Morpholino sequences were as follows: Hdac1ATG1: 5'-ttg ttc ctt gag aac tca gcg cca t-3'; Hdac1SPL1: 5'-ata ttc tta ccg tca taa taa tag c-3'; Standard Control (human β-globin): 5'-cct ctt acc tca gtt aca att tat a-3'; Hdac1 Mismatch control: 5'-ttg ctc gtt gag aac tct gca caa t-3'. 1-2nl of 0.3 mM morpholino solution in milli-Q water was microinjected into embryos at the 1-2-cell stage.

Trichostatin A (TSA) was dissolved in DMSO to 3 μM and added to E3 medium to a final concentration of 1 μM. 10 hpf wild-type embryos were incubated in E3 medium containing 1 mM TSA for four hours until 14 hpf, then samples were fixed and analysed by in situ hybridisation

Whole-mount in situ hybridisation was performed using standard procedures.

For immunostaining with anti-Hdac1 antibody (abcam ab41407), embryos were fixed overnight, dehydrated in methanol and stored at -20°C. After rehydrating and permeabilizing with acetone at -20°C for 7 minutes, embryos were blocked in PBS containing 0.5% Triton-X, 1% DMSO, 1% BSA and 2% sheep serum (PBDTss) for 2 hours at 4°C and incubated with Hdac1 antibody (ab41407, 1:100, abcam) overnight. The next day, embryos were rinsed in PBDTss and incubated with alexa488-conjugated rabbit IgG (1:500, Invitrogen) before mounting for confocal microscopy.

200 μg of zebrafish embryo protein extract, corresponding to 5 whole 24hpf embryos, was separated by SDS-PAGE, transferred to Nitrocellulose (Amersham), and incubated with the following antibodies: anti-Hdac1 (ab41407, abcam), Pan-acetyl lysine (gift from C. Crane-Robinson, University of Portsmouth 44 , H3acetylK9 (ab4441, abcam), H3dimethlyK9 (ab1220, abcam), H3trimethlyK4 (ab8580, abcam), H3 (ab1791, abcam). Signals were visualized using Horse Radish Peroxidase-conjugated secondary antibodies and the ECL system (GE Healthcare).

40-60 embryos were treated with RNAlater (Ambion) and RNA was extracted using TRIzol. Traces of DNA were removed with RNase-free DNase I and RNA was purified on RNeasy columns (QIAGEN). Cy-dye (Perkin Elmer)-labelled amino Allyl-Modified aRNA probes (Ambion Amino Allyl MessageAmp II) were synthesised and hybridized to a custom Agilent 4 × 44 K array. Hybridizations were scanned using the standard Two-Color Microarray-Based Gene Expression Analysis Protocol (publication number G4140-90050; Agilent Technologies). The custom 44 K design contained a 22 K Agilent probe set together with a 16 K Sigma-Compugen probe set and bespoke oligo list 20 . In total the array contained probes corresponding to 18636 unique Unigene clusters and approximately 2000 additional Ensembl-annotated genes.

For each batch of chromatin, three hundred wild-type (AB), hdac1 mutant, wild-type sibling, or morphant embryos were enzymatically dechorionated at 12 hpf or 27 hpf. Embryos were fixed immediately in 1.5 mM ethylene glycolbis[succinimidyl succinate] (EGS; Sigma-Aldrich) for 45 minutes, followed by 1.5% formaldehyde for 20 minutes 48 . Glycine (0.125 M) was added to quench the formaldehyde and embryos were washed in ice cold PBS. Embryos were deyolked 49 and homogenised on ice in extraction buffer with protease inhibitors. Chromatin was sheared by sonication to give DNA fragments of approximately 300-700 bp in size. Sonicated samples were centrifuged at 12,000 × g at 4°C for 30 minutes, and insoluble material was discarded. The supernatant was incubated at 4°C for 3 hours with 6 μg of either anti-Hdac1 antibody (Abcam ab41407) or control IgG antibody (Abcam ab46540). 80 μl of protein G magnetic beads (Pierce) were added to each sample and the samples rotated at 4°C overnight. Beads were washed six times with RIPA buffer at 4°C and bound complexes were eluted from the beads in Elution Buffer at 65°C for 20 minutes with vortexing. Cross-links were reversed overnight at 65°C and DNA fragments were purified by treatment with RNase A, followed by proteinase K digestion, phenol:chloroform:isoamyl alcohol extraction and ethanol precipitation. Quantitative real-time PCR (QPCR) of immuno-precipitated DNA was used to detect DNA sequences around the transcription start sites (+1) of ascl1b: between -785bp and +790 bp; dlb: between -964bp and +1018 bp; neurod4: between -803bp and +856 bp; her6: between -838bp and +825 bp. QPCR analyses were performed in triplicate and DNA abundance was normalised against Input values.