Lsd1 interacts with cMyb to demethylate repressive histone marks and maintain inner ear progenitor identity

In the developing embryo, multipotent progenitors are able to maintain their identity and gene expression programmes despite ongoing changes in their signalling environment. We have recently uncovered a gene network that integrates signalling inputs and transcriptional activity during the transition from progenitor to committed inner ear cells (Anwar et al., 2017; Chen et al., 2017), providing a good basis to explore this problem in a well-defined system. The entire inner ear is derived from a simple epithelium, the otic placode, located in the ectoderm next to the hindbrain. Prior to placode formation, otic precursors are part of a sensory progenitor pool (the pre-placodal region, PPR), from which FGF signalling induces otic-epibranchial progenitors (OEPs) (Anwar et al., 2017; Ladher et al., 2000; Maroon et al., 2002; Martin and Groves, 2006; Sun et al., 2007; Urness et al., 2010; Wright and Mansour, 2003). Over time, OEPs segregate into committed ear and epibranchial cells. They do this gradually by responding to new signalling cues and by altering their transcriptional machinery (Chen et al., 2017; Freter et al., 2008; Jayasena et al., 2008; Park and Saint-Jeannet, 2008; Sun et al., 2007). Thus, during this transitional period, OEPs must maintain their identity before proceeding to lineage commitment. The molecular mechanisms that maintain the competence of OEPs to respond to signals and maintain their progenitor state have not been identified.

Epigenetic mechanisms such as histone modifications are known to maintain progenitor cell identity as well as prime them for differentiation (Martin and Zhang, 2005), and may therefore play a role in keeping cells in an OEP state. One such modification is histone methylation in gene bodies and regulatory regions (Layman and Zuo, 2014; Ooi and Wood, 2007) that are associated with either gene activation or repression. In placode-derived sense organs, virtually nothing is known about the epigenetic control of cell identity as cells transit from progenitor to lineage commitment. The histone demethylase Kdm4b (also known as JmjD2B), which specifically removes di- and trimethyl groups from lysine 9 of histone 3 (H3K9me2/3) (Labbé et al., 2014), is the only epigenetic enzyme identified in inner ear precursors to date and controls placode invagination at later stages (Uribe et al., 2015). Recent RNA-seq analysis (Chen et al., 2017) identified the demethylase Lsd1 as specifically enriched in OEPs, suggesting that it could be involved in ear specification.

Lsd1 plays a key role in maintaining the pluripotency of stem cells (Adamo et al., 2011), as well as acting as a molecular toggle switch that promotes or inhibits differentiation of progenitor cells (Laurent et al., 2015). Lsd1 (also known as Kdm1a/Aof2/BHC110) is a flavin-dependent demethylase that specifically removes active H3K4me1/2 or repressive H3K9me1/2 marks, and thus can function as a transcriptional repressor or activator (Forneris et al., 2006; Metzger et al., 2016, 2005). Consistent with this, it is found in both co-repressor and co-activator complexes such as REST-CoREST (Ballas et al., 2005; Lee et al., 2005; Shi et al., 2005) and androgen receptor (AR)-Chd1 complexes (Metzger et al., 2016, 2005). Here, we show that Lsd1 is indispensable for ear development. It is required to maintain the expression of key progenitor-specific genes (Sox8, Pax2, Etv4 and Zbtb16) that feed into a larger gene regulatory network that controls ear commitment. We identify cMyb as a novel Lsd1-interacting partner that recruits Lsd1 to promoter regions of actively transcribed genes to maintain the active histone mark H3K9ac by catalysing H3K9 demethylation.

The commitment of OEPs towards an otic fate is a gradual process that requires the integration and fine-tuning of gene regulatory networks (Chen et al., 2017). OEP specification by FGF signals begins around somite stage (ss) 1 in chick embryos (Groves and Bronner-Fraser, 2000). By ss5, OEPs express high levels of the FGF mediators Etv4 and Etv5 (Anwar et al., 2017; Betancur et al., 2011; Lunn et al., 2007), and develop along the otic lineage when cultured in isolation (Groves and Bronner-Fraser, 2000). However, it is not until ss10 that the otic lineage becomes molecularly distinct from the epibranchial cells (Chen et al., 2017), becoming fully committed to otic fate by ss11-12 (Groves and Bronner-Fraser, 2000). This progressive commitment is driven by a sequential activation of different signalling pathways (Freter et al., 2008; Jayasena et al., 2008; Park and Saint-Jeannet, 2008; Shida et al., 2015). The mechanisms whereby OEPs maintain their identity and remain poised for rapid response to new signals have not been identified.

We hypothesized that epigenetic mechanisms may be involved and, using our recent RNA-seq data (Chen et al., 2017), we identified Lsd1 as a candidate for maintaining OEP identity prior to lineage commitment. Lsd1 is expressed immediately after OEP specification but prior to molecular segregation of the otic-epibranchial lineages. We confirmed this by in situ hybridization. At ss5, Lsd1 is confined to the neural tube, with weak expression in the mesoderm, but is absent from OEPs (Fig. 1A,a). At ss8, it becomes expressed in OEPs (Fig. 1B,b), and by ss13, it is robustly expressed in the otic placode and weakly in more lateral ectoderm (Fig. 1C,c). To determine the onset of Lsd1 expression, we performed qPCR from dissected OEPs and placodes, and compared its expression with that of the transcription factor Pax2, an established, FGF-responsive OEP marker (Chen et al., 2017; Nechiporuk et al., 2007; Phillips et al., 2001; Yang et al., 2013). This analysis confirms the absence of Lsd1 from OEPs at ss5 and its upregulation at ss8 (Fig. 1D), preceding the segregation of otic and epibranchial lineages. Thus, Lsd1 expression is activated in OEPs after they have been specified, and its expression persists as otic cells become committed.

To test directly whether Lsd1 is required to maintain OEP identity, we performed knock-down experiments using two different morpholinos (MO) against Lsd1 (Lsd1-MO; both affect Lsd1 splicing; Fig. S1A). We find that Lsd1 knock-down with either MO leads to a dramatic reduction of otic vesicle size when compared with the control side (n=16, Fig. 2A-b′), indicating an important role for Lsd1 in ear development.

To understand global changes in the otic gene network after Lsd1 knockdown, we designed a NanoString probeset containing 220 transcripts based on our RNA-seq data (Chen et al., 2017), which includes 70 otic genes as well as factors specific for other placodes, neural, epidermal and neural crest cells (Table S1). Lsd1-MOs were electroporated into OEPs at ss5-6, and otic tissue was analysed after 4.5 and 9 h, corresponding to ss8-9 and ss11-12, respectively. Experiments were performed in triplicate and genes with a fold change of at least ±1.5 and P≤0.05 were considered to be deregulated (Table S2).

Several genes are rapidly upregulated after Lsd1 knockdown. These include the Notch targets Hes4 and Hey1 (a direct negatively-regulated Lsd1 target; Wang et al., 2007), the Notch ligand Jag1, and Cbx4, a component of the Lsd1-CtBP and the PRC2 co-repressor complexes (Shi et al., 2003; Tsai et al., 2010; Wang et al., 2007) (Fig. 3A).This is followed by increase of a number of lineage markers: the epidermal marker Gata2 (Sheng and Stern, 1999), the neural crest genes Snail1, Msx1 and Id4 (Kee and Bronner-Fraser, 2001; Nieto, 2002; Streit, 2002), and the lens-specific transcripts Pax6, Zeb2, L-Maf and C-Maf (Bailey et al., 2006; Bhattacharyya et al., 2004; Lecoin et al., 2004; Sheng et al., 2003) (Fig. 3B). These data suggest that Lsd1 normally represses these transcripts and may prevent the switch of OEPs to other fates. By contrast, Lsd1 knockdown leads to loss of either OEP-specific genes (e.g. Sox8, Pax2, Zbtb16, Etv4) or genes that are already expressed at PPR stages (e.g. Foxi3, Gbx2, Spry1/2, Six1, Eya2, Znf385c), with Sox8, Pax2 and Spry2 showing the greatest fold change (Fig. 3A). Most of these genes are dependent on FGF signalling (Chen et al., 2017; Yang et al., 2013), and continue to be downregulated 9 h after Lsd1 knock down, when expression of additional downstream genes such as Sox10 (a target of both Sox8 and Etv4) (Betancur et al., 2011) are also reduced. These data indicate that Lsd1 is crucially required to maintain the transcriptional profile of OEPs, while simultaneously preventing alternative fates.

To validate the NanoString data and as an alternative approach to perturb Lsd1 function, we cultured OEP explants in the presence of tranylcypromine (TCP) – a well-documented Lsd1 inhibitor (Schmidt and McCafferty, 2007) – and then performed qPCR for selected genes. This approach verifies the downregulation of OEP and posterior PPR genes at ss8-9 (Fig. 3C, Sox8, Pax2, Etv4, Zbtb16, Foxi3, Znf385c) but not the upregulation of other genes (Fig. S1B). There is no increase in activated caspase 3 in Lsd1-MO targeted OEPs (Fig. S4), indicating that reduction of Lsd1 does not promote apoptosis but that, instead, OEPs lose their identity. Analysing OEP markers by in situ hybridization 9-12 h post MO electroporation reveals a complete loss of Pax2 (n=8, Fig. 2Cc), Zbtb16 (n=6, Fig. 2Dd) and Etv4 (n=6, Fig. 2Ee) after Lsd1 knockdown, while upregulated genes cannot be verified (Fig. S1C,C′). Interestingly, although Lsd1-MO only target the ectodermal layer, Etv4 expression is also affected in the underlying mesoderm and endoderm, pointing to reciprocal signalling between these layers and the placode. The epibranchial expression of Pax2 is also abolished (Fig. 2C), and expression of other epibranchial markers such as Foxi2 is reduced (n=5, Fig. S2D-d″). Morphologically, the placode is thinner, having lost the characteristic epithelial thickening compared with the control side (Fig. 2C-E; inset in Fig. 2C; Fig. S3I-L). Only cells that escape being targeted by Lsd1-MO maintain Pax2 expression and coalesce to form a hypoplastic cup or vesicle (n=8, Fig. S2A-a; Fig. S3E-H), whereas control morpholinos (Ctrl-MO) do not affect normal placode involution into a cup (n=6, Fig. S2B,B′) or vesicle (n=4, Fig. S2C,C′). Pax2 expression, loss of epithelial thickening as well as the late morphological phenotype can be rescued by co-electroporating Lsd1-MO with full-length human Lsd1-RFP (Fig. S3). Together, these data strongly support a role for Lsd1 in maintenance of the OEP transcriptional state and suggest that in this context it may act as a co-activator of OEP genes that determine OEP fate.

If Lsd1 functions as a direct co-activator of OEP and PPR genes, it should associate with their promoter regions to modify histone tails. To investigate this, we performed Lsd1 ChIP-qPCR on otic placodes and found that the promoters of Sox8, Pax2, Etv4 and Zbtb16 were bound by Lsd1, whereas promoters of Spry2, Foxi3, Six1 and Znf385c did not show any significant Lsd1 recruitment (Fig. 4A, Fig. S5). Thus, OEP genes are direct Lsd1 targets, whereas PPR factors are likely to be regulated indirectly.

As a co-activator, Lsd1 acts by demethylating the repressive H3K9me2 found predominantly at silent promoters (Barski et al., 2007). H3K9 has a dual function at gene promoters, with opposing roles depending on its modification: acetylation is associated with gene activation, whereas methylation rapidly silences gene expression (Karmodiya et al., 2012). Thus, H3K9me2 demethylation by Lsd1 (Metzger et al., 2016, 2005) allows H3K9 acetylation, thereby maintaining an open chromatin state, while H3K9me2 accumulation in the absence of Lsd1 prevents H3K9ac and the chromatin remains closed (Fig. 4B). Indeed, when Lsd1 is knocked down using MOs or inhibited by TCP, H3K9me2 increases in otic placode cells (Fig. S6). Importantly, using ChIP-qPCR for H3K9me2 and H3K9ac, we find that Lsd1 inhibition in OEP explant cultures by TCP results in an increase in H3K9me2 (Fig. 4C) and a corresponding decrease in H3K9ac (Fig. 4D) at the promoters of Sox8, Pax2, Etv4 and Zbtb16. These data suggest that Lsd1 is normally recruited to the promoters of active OEP genes to prevent the accumulation of H3K9me2 and the loss of H3K9ac, thus allowing their continued expression.

Lsd1 itself lacks the ability to bind DNA or to recognize the H3 tail within nucleosomes (Shi et al., 2005). How is Lsd1 recruited specifically to the promoters of OEP genes? As a co-repressor, Lsd1 interacts with the co-factor of REST (CoREST) via its chromatin-interacting SANT2 domain (Ballas et al., 2005; Lee et al., 2005; Shi et al., 2005) or with SNAG-domain transcription factors (such as Snail1/2 and Gfi1) (Lin et al., 2010; Saleque et al., 2007), which recruit Lsd1 to their respective target genes. To investigate which factor is responsible for recruiting Lsd1 to the promoters of Sox8, Pax2, Etv4 and Zbtb16, we performed a motif enrichment analysis on these promoters using RSAT (Fig. 5A; Table S3). To narrow down potential Lsd1 partners in the ear, we removed factors belonging to the general transcriptional machinery and factors that are not expressed in OEPs. This strategy eliminates SNAG domain factors and hormone nuclear receptors, while returning Sox8, Pax2 as well as the SANT-domain containing coREST (via the presence of REST motif) and Myb as highly enriched motifs. Of those, Myb emerges as the top candidate to recruit Lsd1 to OEP gene promoters. Myb motifs are enriched in all four promoters (Fig. 5A), cMyb is expressed specifically in OEPs at ss5-6, just prior to Lsd1 (Betancur et al., 2011); it contains a SANT domain and, together with its transactivation domain, can function as a co-activator or a co-repressor (Boyer et al., 2004). Morpholino-mediated knockdown of cMyb in OEPs at ss3-4 leads to the loss of Lsd1-targets such as Pax2 (n=6, Fig. S7A-b) and Zbtb16 (n=5, Fig. S7B,B′), but not of Lsd1 expression (Fig. S7C-J).

To assess whether cMyb and Lsd1 physically interact, we used lysates from neuroblastoma cells, which express these proteins in abundance (Fig. 5B), to perform co-immunoprecipitation. Although cMyb antibodies do not work in immunoprecipitation, Lsd1 antibodies successfully precipitate Lsd1 protein (Fig. 5B″). We therefore used Lsd1 antibody for immunoprecipitation followed by western blotting with cMyb antibody; this reveals that both proteins interact (Fig. 5B′). We then asked whether cMyb is required for Lsd1 binding to the Sox8, Pax2, Etv4 and Zbtb16 promoters. Control or cMyb-MOs (Betancur et al., 2011) were electroporated into OEPs at ss3, targeted cells were harvested 10.5 h later at ss10 and processed for Lsd1 ChIP followed by qPCR for Sox8, Pax2, Etv4 and Zbtb16 promoter regions. Although Lsd1 binds all four promoters in controls, this is strongly reduced when cMyb is knocked down (Fig. 5C). Together, our data suggest that cMyb and Lsd1 proteins physically interact and that cMyb recruits Lsd1 to the promoters of the OEP genes Sox8, Pax2, Etv4 and Zbtb16. In turn, this keeps H3K9 demethylated and allows its acetylation, thus maintaining OEP gene expression.

Although the epigenetic mechanisms that maintain stem cell pluripotency and the exit from this state in vitro are fairly well characterized, their role in controlling cell fate decisions and maintaining cell identity in normal embryonic development is much less understood. Here, we take advantage of the recently established gene regulatory network for ear development to investigate the epigenetic mechanisms that maintain ear precursors in a progenitor state prior to lineage commitment. Ear commitment is a gradual process that begins with OEP specification from a pool of sensory progenitors. Exposed to a changing extrinsic environment as morphogenesis proceeds, OEPs must stabilize their identity and retain competence to respond to signals that ultimately commit them to their fate (Fig. 6A). Here, we establish a central role for the demethylase Lsd1 in this process, driven by the transcription factor cMyb (Fig. 6B,C).

The control of OEP specification and their subsequent commitment to either the otic or epibranchial lineages involves a hierarchical organization of gene regulatory networks and sub-networks. These network modules are interconnected and embedded within them are the signalling inputs that direct cell behaviour (Chen et al., 2017). Because of the hierarchical network organization and the crosstalk between regulatory modules, failure to maintain components of just one module will essentially shutdown the entire developmental programme. This is precisely what we observe in the absence of Lsd1 function. Lsd1 expression is activated in OEPs after specification, and its expression persists in otic cells as they commit to their fate. Therefore, by directly maintaining the expression of key players in the OEP module (Sox8, Pax2, Etv4 and Zbtb16) that precede its own expression, Lsd1 becomes part of a sub-network that acts like a central hub for the entire ear cascade. Interestingly, these Lsd1 targets are dependent on FGF signalling for their initial expression, and Lsd1 is known to regulate both Wnt and FGF signalling pathways (Chen et al., 2016; He et al., 2016; Huang et al., 2017, 2013; Lei et al., 2015; Takeuchi et al., 2015).

In the zebrafish lateral line, Lsd1 is required to activate Etv4, Axin2 and Tcf7l2 to promote supporting cell proliferation upon hair cell damage (He et al., 2016), whereas it downregulates Etv2 to promote commitment of hematopoietic progenitor cells to their fate (Takeuchi et al., 2015). Furthermore, in cancer cells, Lsd1 promotes the Wnt pathway by repressing negative regulators of Wnt/β-catenin signalling such as APC, Prickle, Sfrp5 and Dkk1 (Huang et al., 2017, 2013; Lei et al., 2015). In contrast, Lsd1 can also negatively regulate Wnt signalling by repressing Wnt pathway components such as Fzd1/2 to promote mesenchymal stem cell differentiation into brown adipocytes (Chen et al., 2016). In the ear, Wnt signalling is required for the transition of OEPs to committed otic cells (Freter et al., 2008; Jayasena et al., 2008; Park and Saint-Jeannet, 2008). Thus, it is possible that maintenance of the OEP module by Lsd1 is necessary for OEPs to retain competence to respond to Wnt signalling and that Lsd1 directly or indirectly promotes the expression of Wnt pathway components.

Virtually nothing is known about the upstream regulators of Lsd1, although in our NanoString data, the expression of Lsd1 itself is lost upon Lsd1-MO knockdown. As the expression of Sox8, Pax2, Etv4 and Zbtb16 in OEP cells precedes Lsd1 expression, it is likely that one or more of these factors regulate Lsd1 expression.

Lsd1 is a demethylase found in both activator and repressor complexes. Although the latter is well studied (Lin et al., 2010; Shi et al., 2005), its role as a co-activator is poorly defined. When Lsd1 is recruited by hormone nuclear receptors to their target promoters, its substrate specificity switches from H3K4me2 to H3K9me2 (Metzger et al., 2016, 2005). In turn, demethylated H3K9 provides a substrate for histone acetyltransferases, with H3K9ac at promoter regions being a hallmark for actively transcribed genes. In addition, the neuronal-specific isoform Lsd1+8a targets H3K9 for demethylation in the absence of nuclear receptors (Laurent et al., 2015). However, this isoform is found only in mammals and is absent in other vertebrates such as chick, fish and frog (Laurent et al., 2015; Zibetti et al., 2010). Our experiments reveal that Lsd1 can indeed switch substrate preference to act as a co-activator in the presence of cMyb.

In the otic placode, cMyb cooperates with Sox8 and Etv4 to control the expression of the late otic marker Sox10: cMyb knockdown results in Sox10 downregulation and the loss of placode thickening (Betancur et al., 2011). This phenotype is remarkably similar to that produced by Lsd1 knockdown. Here, we show that Lsd1 and cMyb physically interact and that Lsd1 recruitment to its target promoters depends on the presence of cMyb. This finding suggests that both are part of the same molecular machinery that controls otic lineage commitment. As Pax2 is expressed prior to Lsd1, we propose that Lsd1 can only function in OEP maintenance when both cMyb and Lsd1 are present. Pax2 expression is initiated by FGF and one study suggests that this pathway must be downregulated (Freter et al., 2008) for cells to commit to the otic lineage. This coincides with the onset of Lsd1 expression, and it is therefore possible that Lsd1 acts to maintain Pax2 when FGF signalling begins to decline.

At promoter regions of OEP genes, Lsd1 prevents accumulation of the repressive mark H3K9me1/2 through its demethylase activity, allowing H3K9ac accumulation. cMyb is likely to play a key role to facilitate this. Unlike the chromatin-interacting SANT domain, which facilitates histone deacetylation by HDACs (Boyer et al., 2004), the cMyb SANT domain binds DNA and cMyb recruits the histone acetyltransferase p300 via its transactivation domain (Mo et al., 2005). In contrast, high levels of H3K9 methylation are essential along the gene body of actively transcribed genes and this is facilitated by histone methyltransferases such as Setdb1 or Ehmt2 (also known as G9a) (Layman and Zuo, 2014; Ooi and Wood, 2007). During adipogenesis, Lsd1 opposes the activity of Setdb1 at target promoters to prevent premature differentiation (Musri et al., 2010). However, the precise mechanism of how this occurs and whether Lsd1 modifies Setdb1 itself or whether there is negative feedback between the complexes remains elusive. We therefore suggest that the Lsd1-cMyb-p300 co-activator complex at the promoter region of OEP genes is crucial to counteract HDAC and HMT activities, and as a result prevents deactivation of OEP transcripts.

In addition, Lsd1 may also be part of a repressor complex, as HDACs and REST/coREST are present in OEPs, albeit being ubiquitously expressed in the embryo (Chen et al., 2017). In this context, Lsd1 may be important to repress non-otic genes, which may in turn prevent ear development. We have previously shown that, in the absence of additional signals, sensory progenitors in the PPR activate the lens programme through expression of the early marker Pax6 and late lens genes such as L-maf and δ-crystallin. Lsd1 knockdown leads to the upregulation of lens genes, although at very low levels, and it is therefore tempting to speculate that Lsd1 is part of the machinery that prevents the lens default fate (Bailey et al., 2006). However, the upregulation of these genes is not maintained, possibly owing to inhibitory signals, and therefore the cells do not progress towards lens development. Together, our findings suggest that the Lsd1-cMyb-p300 co-activator complex lies at the heart of a mechanism that maintains OEP identity, and that Lsd1 is a molecular cell fate switch that ensures the execution of the ear programme.

Much of our understanding of the epigenetic regulation of cell fate decisions centres around maintenance of embryonic, neural or hematopoietic stem cells (Adamo et al., 2011; Foster et al., 2010; Sun et al., 2010; Takeuchi et al., 2015; Thambyrajah et al., 2016), cancer cells (Huang et al., 2017; Lei et al., 2015), and progenitor cell differentiation in the foetal or postnatal retina (Popova et al., 2016), anterior pituitary (Wang et al., 2007) or nervous system (Ballas et al., 2005). In all these cases, Lsd1 acts as a co-repressor to either maintain stem cell identity or to allow differentiation towards a particular fate, mostly targeting H3K4me1 at enhancer or H3K4me2 at promoter regions. However, there is limited evidence of the crosstalk between genetic and epigenetic regulation of cell fate in the developing embryo. To our knowledge, this is the first report that demonstrates a co-activator role for Lsd1 in cell fate maintenance. Our findings place Lsd1 within a gene regulatory network and show how a single epigenetic modifier maintains a network module that is in turn crucial to promote the ear programme. Many factors within this module are involved in other biological processes, cancer and/or mutated in human syndromes. Thus, from a therapeutic perspective, this study may provide insights into the mechanisms underlying the human syndromes caused by mutations in LSD1 (Chong et al., 2016; Tunovic et al., 2014), such as cleft palate, psychomotor retardation and distinct facial features (CPRF; OMIM # 616728), as well as overlapping features with Kabuki syndrome (OMIM # 300867).

Otic tissue collected in nuclear extraction buffer [10 mM Tris-HCl (pH 7.5), 3 mM CaCl₂, 250 mM sucrose, 0.5% NP-40, 0.25% Triton X-100, supplemented with 1 mM DTT and protease inhibitors (PIs)] was homogenized using a Dounce homogenizer on ice, crosslinked in 0.9% formaldehyde and quenched with 125 mM glycine. Nuclei were pelleted, washed in PBS (with DTT/PI) and lysed in SDS lysis buffer (50 mM Tris-HCl (pH 8), 1% SDS, 10 mM EDTA (pH 8), PIs] on ice for 1 h to release crosslinked chromatin. The chromatin was then sonicated using SorvicsVibra Cell and probe CV18 on ice in ChIP buffer without DTT [16.7 mM Tris-HCl (pH 8), 0.1% SDS, 1.2 mM EDTA (pH 8), 167 mM NaCl] to obtain DNA fragments of 0.2-1 kb. Fragmented chromatin was diluted in ChIP buffer with 1 mM DTT and PIs. Protein A magnetic beads (Invitrogen 10002D) were blocked in 0.5% BSA/PBS, incubated with 5 µg of antibody at 4°C overnight on a rotator and collected using a magnetic stand (Invitrogen 12321D). Immunoprecipitation was performed by resuspending antibody beads in fragmented chromatin at 4°C overnight on a rotator, then repeated washing in RIPA buffer [50 mM HEPES-KOH (pH 8), 500 mM LiCl, 1 mM EDTA (pH 8), 1% NP-40, 0.7% sodium deoxycholate, PIs] followed by a TE/50 mM NaCl wash. Elution was performed at 65°C for 15 min with elution buffer [50 mM Tris-HCl (pH 8), 10 mM EDTA (pH 8), 1% SDS] and reverse crosslinked at 65°C overnight together with input, then treated with RNAse A for 1 h at 37°C followed by 0.2 µg/ml proteinase K at 55°C for 1 h. DNA was recovered by phenol-chloroform extraction and ethanol precipitation. RT-qPCR was performed using SYBR green and AriaMx Real-Time PCR System (Agilent Technologies). Antibodies were from Abcam used at 5 µg per 25 µg of chromatin: Lsd1 (ab17721), H3K9ac (ab4441) and H3K9me2 (ab1220).

For immunoprecipitation assays, nuclear cell extracts from neuroblastoma cells were precleared in 0.5% BSA for 1 h on ice then with protein A magnetic beads for 30 min at 4°C to reduce non-specific binding and background. Precleared lysates were separated from the beads and incubated with 5 µg of antibody (anti-Lsd1, Abcam, ab17721) in co-immunoprecipitation (co-IP) buffer [20 mM Tris-HCl (pH 7.5), 100 mM NaCl, 0.2% TritonX-100, 0.2% NP-40, 2 mM β-mercaptoethanol, 0.2% sodium deoxycholate, 1 mM DTT, PIs] at 4°C overnight, then with beads for 4 h at 4°C followed by repeated washes in wash buffer [20 mM HEPES (pH 7.5), 1.25 mM MgCl₂, 0.1 mM EDTA (pH 8), 10% glycerol, 0.2% NP-40, 100 mM KCl, 1 mM DTT, PIs]. Beads were collected, resuspended in 2× SDS loading buffer, boiled for 5 min at 95°C and run on SDS-PAGE gel for western blotting.

Protein lysates from neuroblastoma cells were separated by SDS-PAGE and transferred onto PVDF membranes using a Bio-Rad Trans-blot Turbo transfer system. After blocking, the membranes were incubated with primary antibody (anti-cMyb, Abcam, ab45150 at 1:2500 or anti-Lsd1, Abcam, ab17721 at 1:1000) overnight at 4°C followed by PBST washes and incubation with HRP-conjugated secondary antibody (Jackson Labs, 1:10,000) for 2 h. Signals were visualized using Bio-Rad Clarity Western ECL Substrate and Bio-Rad ChemiDoc Touch imaging system.

Input Ct values were adjusted for the dilution factor and ΔCt calculated by normalizing Ct values to the adjusted input: ΔCt=Ct(input)-Ct(IP). The % input was calculated as % input=100×2e^−ΔCt. Fold enrichment was calculated using % input of IP/IgG: [ΔCt(IP)/ΔCt(IgG]. Error bars represent the standard error. Unless otherwise indicated, differences between experimental groups were compared using an unpaired two-tailed Student's t-test and P≤0.05 was considered statistically significant.

ImageJ was used to measure the integrated density (fluorescence intensity) in the region of interest for both electroporated and drug-treated samples. For electroporated cells, the mean fluorescence intensity of targeted cells/Hoechst was compared with non-targeted cells/Hoechst, whereas for drug-treated versus untreated samples, the entire placode/Hoechst was measured. To correct for different exposure levels between channels and individual samples, the integrated density from the mean of three background areas was also calculated. The formula used to estimate the corrected total cell fluorescence (CTCF) was as follows: integrated density–(area of selection×mean fluorescence of background reading). Statistical significance was estimated using an unpaired t-test.

Transcription factor-binding motifs were predicted using the JASPAR database (jaspar.genereg.net) and P-Match program from the Transfac database (www.gene-regulation.com/pub/programs.html). A profile matrix was generated to run a pattern matching scan using RSAT (rsat.sb-roscoff.fr/) on the promoters for enrichment against the background (chicken genome).

Dissected OEP tissues were cultured in collagen with trans-2-phenylcyclopropylamine (tranylcyclopromine, TCP, at 25 µm working concentration prepared in sterile water) in culture medium. Explants were incubated with drug (or no drug control) for 30 min at room temperature before culturing for either 4.5 or 9 h. For sectioning and immunofluorescence, the whole embryo or head was cultured in the presence or absence of drug for 12 h on polycarbonate membranes (Corning, 3422).

Fertilized hen's eggs (Winter Farm, Hertfordshire, UK) were incubated in a humidified incubator at 38°C to reach ss5-6. Eggs were windowed and embryos visualized using India ink (Pelikan; diluted 1:5 in Tyrode's saline). The positive (platinum) electrode was placed under the embryo at hindbrain (r4) level, DNA/morpholino was injected below the vitelline membrane, the negative (tungsten) electrode was placed on top and the current was applied using an Intracel electroporator (settings: 5 V, 5 pulses, 50 ms, 10 ms duration). Tyrode's saline supplemented with penicillin streptomycin was applied to the embryo, the egg was resealed with tape and incubated to the appropriate stage. Placodes were dissected as described previously (Anwar et al., 2017; Chen et al., 2017).

Embryos were fixed in 4% paraformaldehyde (PFA) overnight at 4°C and in situ hybridization performed using DIG-labelled riboprobes. For antibody staining, embryos were embedded in paraffin wax or gelatine and sectioned at 8 µm. Primary antibodies used were: anti-H3K9ac (Abcam, ab4441) 1:100; anti-H3K9me2 (Abcam, ab1220) 1:100; anti-FITC 488 (Thermo Scientific, 4-4-20) 1:500; anti-mCherry (Abcam, ab167453) 1:500; anti-cleaved caspase 3 (Cell Signaling, 9964B) 1:300; and anti-Lsd1 (Abcam, ab17721) 1:300. Secondary antibodies were: goat anti-rabbit or anti-mouse Alexa Fluor 568 (1:1000, Invitrogen) and Alexa Fluor 635 (1:500, Invitrogen).

Fluoresceinated morpholino antisense oligonucleotides were obtained from Gene Tools with standard Ctrl-MO (5′-CCTCTTACCTCAGTTACAATTTATA-3′) from the manufacturer and cMyb-MO (5′-ATGGCCGCGAGCTCCGCGTGCAGAT-3′) published by Betancur et al. (2011). Both Lsd1 morpholinos were splice morpholinos. Lsd1-MO 1 (5′-TTTGTGAAACTCACCAAGTCCTGTT-3′) deletes exon 8 (enzymatic domain). Lsd1-MO 2 (5′-ATCACTGGACAAAATCTGACCTTGT-3′) deletes exon 11 (substrate-binding domain). All morpholinos were used at a 1 mM final concentration. Lsd1-RFP expression constructs were generated by subcloning the human Lsd1 (a gift from Prof. Yang Shi, Harvard Medical School, Boston, MA, USA) open reading frame into pCAB.IRES.RFP vector.

Preparation of tissue for qPCR and NanoString analysis were conducted as previously described (Anwar et al., 2017; Chen et al., 2017).

We thank Rosalinda Guerra and Ewa Kolano for excellent technical support. We thank Prof. Claudio Stern, Dr Albert Basson, Dr Ravindra Prajapati and Dr Mark Hintze for comments on the manuscript; Dr Karen Liu and Daniel Doro Pereira for cell lines; and Prof. Yang Shi (Harvard Medical School). The authors declare no competing or financial interests.