Ezh2 Loss-of-Function Alters Zebrafish Cerebellum Development

1. Introduction

Polycomb Repressive Complex 2 (PRC2) is an essential chromatin-associated protein complex involved in the transcriptional silencing of gene expression programs during development and differentiation. EZH2 (or its paralog EZH1) is the PRC2 catalytic subunit trimethylating lysine 27 of histone H3 (H3K27me3). This epigenetic mark is responsible for heterochromatin formation which in turn represses the expression of numerous genes involved in various cellular and developmental processes, including nervous system development. In mouse, Ezh2 is highly expressed in both embryonic and adult neural stem cells (NSCs) and conditional Ezh2 inactivation leads to a significant reduction in both embryonic and adult NSC proliferation, outlining its crucial role in the maintenance of the proliferative capacity of NSCs [1]. In addition, loss of Ezh2 gene function in the embryonic cerebral cortex accelerates neurogenesis, with an early increase in the number of neurons and premature production of astrocytes, suggesting that Ezh2 regulates the timing of neuronal differentiation [2]. Moreover, in the embryonic cerebellum, Ezh2 controls the specification of GABAergic neurons by repressing genes that promote alternative differentiation, thereby ensuring the proper development of inhibitory circuits [3]. Besides its role in neural differentiation and development, a link between EZH2 dysregulation and brain tumorigenesis has been clearly established [4]. Diffuse midline gliomas (DMG) are one of the most aggressive pediatric brain cancers, characterized either by a lysine-to-methionine substitution at position 27 on certain histone H3 genes (H3F3A, HIST1H3B or HIST1H3C) or by the overexpression of EZHIP [5,6,7]. H3K27M mutation and EZHIP protein are both competitive inhibitors of PRC2 lysine methyltransferase activity causing a global reduction of H3K27me3 levels in DMGs. The reduction of H3K27me3 levels impairs the establishment of differentiation programs and promotes an undifferentiated cellular state fostering tumor formation in fine [8]. H3K27me3 levels are also affected in group 3 and 4 medulloblastoma due to alterations in the function of EZH2 or the corresponding histone demethylase KDM6A (UTX) [9]. The analyses of non-WNT/SHH medulloblastoma that includes group 3 and 4 revealed that about 47% of these tumors are H3K27me3-deficient [10]. Strikingly, H3K27me3 loss is associated with high rates of recurrence and poor overall survival compared to H3K27me3-proficient tumors. Together, these findings illustrate the fundamental role that EZH2 and H3K27me3 levels play in neural development and brain tumorigenesis.

PRC2-mediated gene control is evolutionary conserved and genomic analyses have revealed that Ezh2 recruitment at chromatin and H3K27me3 marks are conserved in zebrafish (Danio rerio) [11]. Numerous studies demonstrated that the zebrafish model provides new insight into the understanding of Polycomb repression in vertebrates (reviewed in [12]). On one hand, in contrast to mouse, where Ezh2-loss of function causes early lethality at the implantation stage [13], ezh2 zebrafish mutants survive up to 12 days post-fertilization (dpf) [14,15]. On the other hand, in zebrafish, the neural tube is formed at about 17 hours post-fertilization (hpf), primary neurogenesis starts at around 2 dpf and secondary neurogenesis at 3 dpf leading to a mature nervous system by 4 dpf [16]. Thus, zebrafish offers a unique opportunity to investigate Ezh2 and H3Kme3 role on brain development without the requirement of conditional mutagenesis strategies.

Here, we investigate the effects of zygotic ezh2 loss-of-function on brain development using the zebrafish ezh2(ul2) mutant line we have previously generated [15]. Our results indicate that ezh2 is required for the proper development of a limited number of cerebellar cells. Furthermore, locomotor activity assays highlight the role of PRC2 in zebrafish larval behavior. Taken together, our findings demonstrate that the ezh2(ul2) zebrafish line is a valuable model for studying the impact of reduced H3K27me3 levels and PRC2 loss-of-function on brain development and shed light on the cellular defects that could be involved in medulloblastoma genesis.

2. Results

2.1. Role of Ezh2 in Oligodendrocyte Development

Studies on embryonic stem cells (ESCs) showed that PRC2 is involved in maintenance of the balance between the self-renewal of neural progenitor cells and the onset of neurogenesis, promoting the transition from neurogenesis to gliogenesis, as well as regulating cell fate decisions during progenitor differentiation [17,18]. More precisely, in a murine neuronal stem cell (NSC) differentiation model, Ezh2 has been specifically implicated in oligodendrocyte lineage development, as demonstrated by its high expression in oligodendrocyte precursor cells (OPCs) compared to astrocytes and differentiating neurons. Overexpression of murine Ezh2 in differentiating NSCs is associated with an increase in oligodendrocytes and a reduction in astrocytes, whereas reduction of Ezh2 expression leads to opposite effects [19]. Ezh2 remains expressed during late stages of oligodendrocyte differentiation, suggesting a role for PRC2 not only in OPCs but also during oligodendrocyte maturation [20].

To further investigate the role of Ezh2 in oligodendrocyte development, we analyzed the consequences of its inactivation in zebrafish. A previous study in mice demonstrated that conditional inactivation of Ezh2 in oligodendrocyte progenitors does not impair OPC specification but results in delayed oligodendrocyte maturation [21]. Given this, we sought to determine whether ezh2 inactivation in zebrafish larvae similarly affects the development of oligodendrocyte lineage cells. In zebrafish, oligodendrocyte development begins around 10.5 hpf [22], with OPCs, characterized by expression of the transcription factor olig2, arising from the ventral progenitor domain of motor neurons (pMN). Olig2 is required for OPC specification and remains expressed throughout oligodendrocyte development, including in mature cells, even if Olig2 expression is reduced in later developmental stages [23]. Therefore, olig2 serves as a robust marker of the oligodendrocyte lineage [24]. We performed in situ hybridization for olig2 expression in wild-type and ezh2 mutant zebrafish larvae at 2, 3 and 5 dpf (Figure 1A).

In both wild-type and ezh2 mutant siblings, olig2-expressing cells are detected in the midbrain, hindbrain, and cerebellum at 2 dpf. At 3 dpf, olig2 expression increases in the cerebellum before declining during later stages of oligodendrocyte differentiation as expected. By 5 dpf, only a weak olig2 signal is retained in the cerebellum (Supplemental Figure S3), probably corresponding to eurydendroid cells, a teleost-specific neuronal population known to express olig2 [25,26,27]. The olig2 expression patterns are indistinguishable between wild-type and mutant larvae, suggesting that ezh2 loss-of-function does not affect olig2 expression, the presence of eurydendroid cells or the initial generation of oligodendrocyte lineage cells (Figure 1A-B). Since the loss of ezh2 function does not visibly disrupt early oligodendrocyte lineage specification in zebrafish, a potential role of ezh2 in later stages of oligodendrocyte maturation, rather than in progenitor specification, cannot be excluded.

To gain more insight into the role of ezh2 during oligodendrocyte development, we performed in situ hybridization experiments to label the expression of genes that are exclusively expressed in mature oligodendrocytes . The mag (myelin associated glycoprotein) and mpz (myelin protein zero) genes code for proteins that are components of the myelin sheath. At 5 dpf, these genes are expressed in myelinating oligodendrocytes located in the midbrain and hindbrain. Cells expressing mag and mpz are observed in the hindbrain, along the midline as well as on both sides of it. Additionally, a mag-associated signal is also detected in the cerebellum (Figure 1). The expression patterns observed in ezh2-deficient larvae are identical to those seen in wild-type larvae, indicating that ezh2 mutation does not notably affect oligodendrocyte development.

2.2. Loss of Ezh2 Function Selectively Impairs Cerebellar Progenitor Proliferation

In vitro studies on murine NSCs have shown that inhibition of Ezh2 expression reduces OPC proliferation and drastically decreases the number of oligodendrocytes [19]. In contrast, in zebrafish, zygotic loss of ezh2 function does not show an obvious reduction in the number of oligodendrocytes as assessed by whole-mount in situ hybridization, suggesting that ezh2 may not be necessary to oligodendrocyte differentiation in this model. However, the possibility that ezh2 plays a role in regulating neural cell proliferation during zebrafish development is not excluded. To investigate this matter, we analyzed the expression of pcna (proliferating cell nuclear antigen), a well-established marker of proliferating cells in the developing zebrafish brain [28] using in situ hybridization at 2, 3 and 5 dpf (Figure 2A).

At 2 dpf, several proliferative zones emerge, with pcna expression predominantly localized in the ventricular regions, the retina and the forebrain, particularly in the pallial and subpallial regions of both wild-type and ezh2 mutant larvae. At 3 dpf, pcna expression decreases in both ezh2^+/+ and ezh2^-/- siblings and becomes restricted to the optic tectum and the retina, with a faint signal persisting in the forebrain. No differences in expression can be detected between wild-type and ezh2 mutant larvae at this stage. At 5 dpf, pcna expression remains concentrated in the optic tectum and retina in wild-type and mutant larvae. A moderate signal is still detected in the forebrain in both groups. Notably, a robust pcna signal is observed in the cerebellum of wild-type larvae, whereas this signal is absent in ezh2 mutants (Figure 2A-B). To confirm this difference, we examined the expression of ccna2 (cyclin A2), another marker of cell proliferation, at 5 dpf (Figure 2A). In agreement with the pcna expression profile, ccna2 is expressed in the optic tectum, retina, and cerebellum of wild-type larvae. In contrast, ezh2 mutant larvae show no detectable ccna2 expression in the cerebellum. Together, our data indicate that ezh2 inactivation leads to a specific reduction in cell proliferation within the cerebellum, without affecting proliferation in other brain regions. This suggests a role for ezh2 in the development or maintenance of cerebellar progenitor cells during zebrafish neurodevelopment.

2.3. Loss of Ezh2 Function Selectively Affects Atoh1c-Expressing Cerebellar Progenitors

Like the mammalian cerebellum, the zebrafish cerebellum contains a diversity of neuronal subtypes that can be categorized according to the nature of their neurotransmitter: excitatory neurons are mainly glutamatergic, and inhibitory neurons, mainly GABAergic. The glutamatergic population includes granule cells, unipolar brush cells, and eurydendroid cells, whereas the GABAergic population comprises Purkinje cells and local interneurons. These neurons originate from two distinct progenitor domains: glutamatergic neurons come from progenitors expressing proneural atoh1 genes, while GABAergic neurons derive from ptf1a-expressing progenitors [27]. To investigate the effect of ezh2 loss-of-function on cerebellar neurogenesis, we analyzed the expression of key cerebellar progenitor markers in wild-type and ezh2 mutant zebrafish larvae using in situ hybridization (Figure 3A).

Zebrafish possesses 3 atoh1 (atonal bHLH transcription factor 1) paralogs - atoh1a, atoh1b and atoh1c - each expressed in distinct regions of the upper rhombic lip (URL) [29]. In contrast, ptf1a (pancreas-associated transcription factor 1a) is expressed in the ventricular zone, marking GABAergic progenitors [30]. At 2 dpf, both atoh1a and atoh1c are expressed in the URL in wild-type and ezh2 mutant larvae, consistent with previous reports [27]. Similarly, ptf1a expression was detected in both ezh2^+/+ and ezh2^-/- siblings at this stage, with no detectable differences. At 5 dpf, atoh1a expression becomes restricted to the valvula cerebelli (Va), with similar expression patterns observed in wild-type and ezh2 mutant larvae. However, atoh1c expression, normally observed at the midline of the corpus cerebelli (CCe), as well as in the lobus caudalis cerebelli (LCa) and the eminentia granularis (EG) (Figure 3B), is strongly reduced in ezh2 mutants compared to their wild-type counterparts. In contrast, ptf1a expression could no longer be detected at 5 dpf in either ezh2^+/+ or ezh2^-/- larvae (data not shown), likely reflecting the temporal restriction of expression of this gene to earlier stages of cerebellar development. Altogether, these results indicate that zygotic ehz2 loss-of-function does not affect early cerebellar progenitor populations, but specifically impairs atoh1c-expressing progenitor cells at 5 dpf. This suggests that ezh2 is required for the maintenance or continued identity of atoh1c-expressing glutamatergic progenitors in the developing zebrafish cerebellum.

2.4. Loss of Ezh2 Function Impairs the Differentiation of Cerebellar Granule and Purkinje Cells

Neural progenitors expressing atoh1 genes give rise to immature Neurod1⁺ granule cells. The neurod1 gene encodes a transcription factor required to initiate granule cell differentiation and remains expressed in immature granule cells [27]. The immature granule cells proliferate and migrate through the granule layer of the cerebellum to become mature granule cells. These mature granule cells are characterized by expression of slc17a7a (solute carrier family 17 member 7a; also known as vglut1, L-glutamate transmembrane transporter). To investigate the role of ezh2 in granule cell development and differentiation within the cerebellum, we analyzed the expression of neurod1 and slc17a7a in wild-type and ezh2-deficient zebrafish larvae at 3 and 5 dpf (Figure 4A).

At 3 dpf, when differentiation of GABAergic and glutamatergic neurons [26], neurod1 shows similar expression in the cerebellum of wild-type and ezh2-deficient zebrafish. However, at 5 dpf, its expression was markedly reduced in the cerebellum of ezh2 mutant larvae, being restricted to the retina and the eminentia granularis (EG) (Figure 4B). Similarly, slc17a7a, which marks mature granule cells, was initially detected in the lateral cerebellum (future EG) in both genotypes at 3 dpf. At 5 dpf, its expression expanded to all cerebellar lobes, the habenula, and the torus longitudinalis in wild-type larvae, while it remained restricted to the EG in ezh2 mutant larvae, with diminished expression in the habenula and torus longitudinalis (Figure 4A-B). These findings suggest that ezh2 is required for the proper late differentiation of granule cells, particularly in the central cerebellar lobe (CCe) and caudal lobe (LCa), as well as for the development of slc17a7a-expressing neurons in extracerebellar regions.

In addition, analysis of pvalb7 (parvalbulin 7), a marker of differentiated Purkinje cells derived from ptf1a-expressing progenitors [27], revealed a loss of medial cerebellar expression in ezh2^-/- mutants at both 3 and 5 dpf. At 5 dpf, the pvalb7 signal is confined to the EG in ezh2-deficient larvae, while it is present in both the CCe and EG in wild-type counterparts. These data indicate that ezh2 is also essential for the differentiation of Purkinje cells particularly in the CCe. Altogether, ezh2 loss-of-function disrupts the development of both glutamatergic (granular cells) and GABAergic (Purkinje cells) neurons in specific cerebellar subregions.

2.5. Loss of Ezh2 Function Does Not Affect The Development of Most Neurotransmitter-Specific Neuronal Populations

Expression profiles obtained using the glutamatergic-specific slc17a7a probe, revealed that cells located outside the cerebellum, specifically in the habenula and torus longitudinalis, are affected by ezh2 loss-of-function (Figure 4A). In order to obtain an overview of the effect of the ezh2 mutation on neuronal development, we performed in situ hybridization experiments to label the expression of markers for other main neurotransmitters in wild-type and ezh2^-/- zebrafish larvae at 5 dpf (Figure 5).

At In both mammals and zebrafish, GABAergic neurons express glutamate decarboxylase enzymes (Gad1 and Gad2), which catalyze the conversion of glutamate to GABA [26]. Zebrafish possess two paralogs of Gad1, gad1a and gad1b, which exhibit similar expression patterns [31]. At 5 dpf, gad1b is strongly expressed in several brain regions, including the retina, subpallium, optic tectum, and cerebellum. Expression patterns of gad1b are comparable between wild-type and ezh2^-/- larvae, suggesting that zygotic ezh2 loss-of-function does not disrupt the development of GABAergic neurons.

Catecholamines are organic compounds synthesized from tyrosine and act as neurotransmitters, with the most common being adrenaline, noradrenaline, and dopamine. In zebrafish, catecholaminergic neurons can be identified by the expression of slc18a2 (solute carrier family 18 member 2, an ATP-dependent transporter of monoamines) [32]. At 5 dpf, slc18a2 is expressed in a cell cluster located in the diencephalon, as well as in the raphe nuclei and hypothalamus. These expression domains are maintained in ezh2^-/- mutants, indicating that catecholaminergic neuron development is not impaired by the loss of zygotic ezh2 function.

Serotonergic neurons were examined using a probe against tph2 (tryptophan 5-monooxygenase), which encodes the enzyme tryptophan hydroxylase responsible for serotonin biosynthesis [33]. At 5 dpf, tph2 expression is detected in the raphe neurons in both wild-type and ezh2^-/- larvae, with no observable differences by whole-mount in situ hybridization, suggesting that serotonergic neuron development remains unaffected.

Dopaminergic neurons were visualized using th (tyrosine hydroxylase) expression as a marker. In zebrafish, th-positive clusters are found in the olfactory bulb, subpallium, preoptic area, pretectum, thalamus, hypothalamus, locus coeruleus, and medulla oblongata. These domains were preserved in ezh2^-/- larvae at 5 dpf, indicating that loss of ezh2 function does not induce major alterations in the dopaminergic system development.

Then, in situ hybridization experiments performed on ezh2^-/- larvae highlight a specific effect of ezh2 loss-of-function on the development of cerebellar neurons, particularly on subpopulations of granular cells and Purkinje cells, with little or no effect in other brain regions.

2.6. Loss of Ezh2 Function Alters Locomotor Activity

The cerebellum is involved in various motor functions, particularly in the regulation of locomotion through the activity of Purkinje cells [34]. Then, to investigate whether ezh2 loss-of-function could affect larval behavior, we performed locomotor assays using a Zebrabox chamber (ViewPoint Life Sciences) equipped with an infrared light-emitting floor and a top-mounted infrared camera allowing video recording of whole plates under both light and dark conditions. Locomotor activity assays conducted in 48-well plates following a protocol that consists of a 10 min initial acclimating period in the dark, followed by six alternating 10 min light and dark phases. As previously described [35], switching from dark to light dramatically decreases larval activity whereas return to darkness is associated with an increase in locomotor activity (Figure 6A).

Moreover, the assay reveals a hyperlocomotor phenotype and an increased total swimming distance (Figure 6B) for ezh2^-/- mutants at 5 dpf when compared to wild-type larvae. This indicates that ezh2 plays a role in the locomotor activity of zebrafish larvae at 5 dpf.

Finally, this hyperlocomotor phenotype is not observed in the heterozygous ezh2^+/- larvae, in which the expression of neurod1, slc17a7a and pvalb7 is also similar to wild-type (Supplementary Figure S27).

3. Discussion

Polycomb group proteins, and in particular the PRC2 complex member EZH2, have emerged as key epigenetic regulators of neural development. In this study, we investigated the role of ezh2 during zebrafish brain development, with a particular focus on oligodendrocyte specification, cerebellar neurogenesis, and neurotransmitter-specific neuronal populations.

Our results show that ezh2 is dispensable for early oligodendrocyte lineage specification in zebrafish. Expression of olig2, as well as mature oligodendrocyte markers such as mag and mpz, are preserved in ezh2^-/- larvae, suggesting that PRC2 activity is not required for oligodendrocyte precursor cell formation or terminal oligodendrocyte differentiation in this context. This contrasts with in vitro studies showing that Ezh2 promotes oligodendrocyte differentiation instead of the astrocyte fate [19], but aligns with in vivo data in mice showing normal oligodendrocyte precursor cell differentiation despite delayed maturation upon Ezh2 inactivation [21]. These findings underscore the importance of in vivo models for defining the context-dependent roles of PRC2 components.

Strikingly, the most profound effects of ezh2 loss-of-function in zebrafish were observed in the cerebellum. Our data demonstrate a selective reduction in proliferative markers (pcna and ccna2) within the cerebellum of ezh2^-/- larvae, while other proliferative brain zones remain unaffected, at least within the limit of sensitivity of whole-mount in situ hybridization. These results suggest that ezh2 is specifically required to maintain cerebellar progenitor proliferation during late stages of development. This is supported by the selective downregulation of atoh1c, a key marker of glutamatergic progenitors in the corpus cerebelli and caudal lobes, at 5 dpf in ezh2-deficient larvae, whereas earlier expression of atoh1a and ptf1a remains intact. In line with this, we observed an impaired differentiation of both major cerebellar neuronal subtypes, the glutamatergic granule cells and the GABAergic Purkinje cells. The expression of neurod1 and slc17a7a (vglut1), marking immature and mature granule cells respectively, was markedly reduced in the cerebellum of ezh2^-/- larvae, indicating a failure in granule cell maturation. Similarly, pvalb7, a marker of Purkinje cells, was diminished in the medial cerebellum of mutant larvae. These findings suggest that ezh2 is essential for the differentiation of multiple cerebellar neuron populations derived from distinct progenitor domains. The selective sensitivity of cerebellar neurons to ezh2 loss raises intriguing questions regarding the underlying mechanisms, whether due to epigenetic control of lineage-specific transcriptional programs or due to broader roles in cell cycle progression and survival.

Interestingly, neuronal cells outside the cerebellum were largely preserved in ezh2 mutants. The expression of gad1b, slc18a2, tph2, and th, markers for GABAergic, catecholaminergic, serotonergic, and dopaminergic neurons respectively, was indistinguishable between wild-type and ezh2-deficient larvae. This reinforces the notion that ezh2 exerts region- and lineage-specific effects, primarily targeting cerebellar neurogenesis rather than global neural differentiation. Within the cerebellum, despite the loss of a number of Purkinje cells, the gad1b signal appears similar in ezh2 mutants compared to wild-type larvae. In ezh2^-/- larvae, the reduction of gad1b expression could be masked by the signal originating from the GABAergic interneurons. Indeed, in situ hybridization for pax2a expression, a marker of interneurons [36], reveals that the number of interneurons is comparable between wild-type and ezh2^-/- larvae at 5 dpf (Supplementary Figure S25). Alternatively, one cannot rule out the possibility that ezh2 loss-of-function affects the maturation of certain Purkinje cells, giving rise to a Gad1b⁺, Pvalb7^- cell population.

The functional consequences of impaired cerebellar development were highlighted by behavioral assays, which revealed a hyperlocomotor phenotype in ezh2^-/- larvae at 5 dpf. The cerebellum plays a key role in fine-tuning motor coordination, and Purkinje cells, in particular, are crucial for modulating motor output [37,38]. Indeed, targeted ablation of Purkinje cells at 2 dpf induces a hyperlocomotor phenotype in zebrafish larvae [37]. Thus, the observed behavioral phenotype could reflect cerebellar dysfunction resulting from the loss of certain Purkinje cell populations, although we cannot exclude that this phenotype is due to other neuronal or metabolic defects not detected by our in situ hybridization approach.

Our data in zebrafish parallel findings in mice, where conditional inactivation of Ezh2 in the cerebellum causes transcriptional dysregulation leading to a reduction of Purkinje cells and impaired proliferation of granule precursor cells derived from the rhombic lip [3]. However, while cerebellar interneurons do not appear to be affected by ezh2 loss-of-function in zebrafish at 5 dpf (Supplementary Figure S25), this cell population is increased in Ezh2-deficient mouse cerebella [3]. Moreover, these mice exhibit marked cerebellar hypoplasia at postnatal day (P)8. Interestingly, deregulated expression of EZH2 has also been reported in congenital brainstem disconnection (CBSD), a rare developmental anomaly characterized by severe cerebellar hypoplasia and brainstem malformation [39]. Although we do not observe obvious cerebellar hypoplasia in ezh2-deficient zebrafish larvae at 5 dpf - likely because this developmental window precedes its manifestation - data from zebrafish, mice, and human disease collectively support the notion that EZH2 dysfunction impairs cerebellar development.

Beyond development, EZH2 and PRC2 dysfunction have also been implicated in pediatric brain tumorigenesis. Notably, diffuse midline gliomas (DMGs) harbor H3K27M mutations or overexpress the EZHIP, both of which competitively inhibit EZH2 methyltransferase activity and result in global H3K27me3 loss to impair neural differentiation and to promote a poorly differentiated, proliferative tumor state [7,8]. Oligodendrocyte progenitors are thought to be the cells of origin for the development of DMG [40,41]. However, our data do not show a role of ezh2 in oligodendrocyte differentiation suggesting that the model is not suitable for studying diffuse midline gliomagenesis. In contrast, our results in zebrafish show that ezh2 loss-of-function disrupts the progression of cerebellar progenitors, a population particularly relevant in the context of H3K27me3-deficient medulloblastoma. The parallel between impaired cerebellar development in ezh2^-/- mutants and H3K27me3 loss in human cerebellar tumors reinforces the importance of EZH2 in cerebellar lineage fidelity and highlights zebrafish as a tractable model to study the developmental origins of these pediatric brain cancers. Then, our study identifies ezh2 as a key epigenetic regulator required for cerebellar progenitor proliferation and neuron differentiation in zebrafish, with striking parallels to mechanisms disrupted in certain medulloblastomas.

4. Materials and Methods

4.1. Zebrafish Maintenance and Embryo Preparation

The zebrafish ezh2(ul2) line ([15]; ZDB-ALT-171009-3) harbors a 22 bp net insertion (insertion of 27 nucleotides together with a deletion of 5 bp) in the ezh2 exon 2, leading to a frameshift in the coding sequence and appearance of a premature stop codon. Hereafter ezh2^ul2/ul2 embryos will be referred as ezh2^-/-.

Zebrafish were maintained at 28°C in a 14/10h light/dark cycle. After spawning embryos or larvae are collected and staged according to Kimmel et al. [42]. The chorions were removed from embryos by the action of 1% pronase (Sigma) for 1 min. Zebrafish embryos or larvae were fixed overnight in 4% paraformaldehyde in PBS (phosphate-buffered saline, Invitrogen), dehydrated gradually to 100% methanol and kept at -20°C.

4.2. Whole-Mount in situ Hybridization

Antisense-RNA probes were generated using RT-PCR from total mRNA extracted from zebrafish larvae at 5 dpf using the RNeasy Mini Kit (Qiagen), following the manufacturer's instructions. After Reverse Transcription (Superscript III, Invitrogen), cDNAs were amplified by PCR using primers coupled to the T7 sequence for forward primers and to the SP6 sequence for reverse primers.

The primers used for probe generation were:

ISH_olig2_F:	TAATACGACTCACTATAGGGATGGACTCTGACACGAGC
ISH_olig2_R:	GATTTAGGTGACACTATAGGGGCTGAGGAAGGTTTGCCAT
ISH_mag_F:	TAATACGACTCACTATAGGGCCGTGAGGGTGTTCAGTGTGTGT
ISH_mag_R:	GATTTAGGTGACACTATAGCGTCTCCCGTGCCTTCCTCT
ISH_mpz_F:	TAATACGACTCACTATAGGGGTGGTGCTCTTGGGCATAGCCTCTC
ISH_mpz_R:	GATTTAGGTGACACTATAGGGAGCCCGTTATCACACCAGCC
ISH_pcna_F:	TAATACGACTCACTATAGGGGGCAACATCAAGCTCTCACA
ISH_pcna_R:	GATTTAGGTGACACTATAGAAATCCCACAGATGACAGGC
ISH_ccna2_F:	TAATACGACTCACTATAGGGGGAAGGATGTCAACACAAGGAAG
ISH_ccna2_R:	GATTTAGGTGACACTATAGGAGAGAACTGTCAGCACCAGATG
ISH_atoh1a_F:	TAATACGACTCACTATAGGGCCAACGTCGTGCAGAAA
ISH_atoh1a_R:	GATTTAGGTGACACTATAGAACCCATTACAAAGCCCAGATA
ISH_atoh1c_F:	TAATACGACTCACTATAGGGTTTCTCAGCGCACACGACCCT
ISH_atoh1c_R:	GATTTAGGTGACACTATAGTTTGGTCTCTTCGGTCATAGGCAAC
ISH_ptf1a_F:	TAATACGACTCACTATAGGGCACAGGCTTAGACTCTTTCTCC
ISH_ptf1a_R:	GATTTAGGTGACACTATAGCCCGTAGTCTGGGTCATTTG
ISH_neurod1_F:	TAATACGACTCACTATAGGGTCGAGACGCTCCGACTAGCCAA
ISH_neurod1_R:	GATTTAGGTGACACTATAGGCGTCGAGCCCGCGTAAAGA
ISH_vglut1_F:	TAATACGACTCACTATAGGGTGCCAGGGACTTGTGGAGGG
ISH_vglut1_R:	GATTTAGGTGACACTATAGCTGGCGTAGCGTGGTGCGA
ISH_pvalb7_F:	TAATACGACTCACTATAGGGTTATCCGTCTCTCACCTCCAGCCA
ISH_pvalb7_R:	GATTTAGGTGACACTATAGCGTGTTCGGTGGCTCTATCACAA
ISH_gad1b_F:	TAATACGACTCACTATAGGGTGAGCGGCATTGAGAGGGCA
ISH_gad1b_R:	GATTTAGGTGACACTATAGCGTAGGCGACCACTGAGCC
ISH_slc18a2_F:	TAATACGACTCACTATAGGGGCACTGGGAGGACTAGCAATGGG
ISH_slc18a2_R:	GATTTAGGTGACACTATAGGTTGGCGGGAGGATTTCGCAG
ISH_tph2_F:	TAATACGACTCACTATAGGGCGGACACCTGCCATGAACTGCTT
ISH_tph2_R:	GATTTAGGTGACACTATAGTGAGTAAGTCGATGCTCTGCGTGT
ISH_th_F:	TAATACGACTCACTATAGGGCCTGTCGGATGTTAGCACGCTGG
ISH_th_R:	GATTTAGGTGACACTATAGGGCCTCAACTGAAATCCTGTGCGT
ISH_pax2a_F:	TAATACGACTCACTATAGGGACACTGGAGCAGACGCAACCA
ISH_pax2a_R:	GATTTAGGTGACACTATAGAGGTCGCCGTCTCGCCTTGA.

The sequences corresponding to T7 and SP6 promoters for in vitro transcription are in italics.

The cDNAs were used to in vitro synthetize digoxigenin-labelled antisense RNA probes using the DIG RNA Labeling Kit (SP6) (Roche), following the manufacturer's protocol.

In situ hybridization was then performed as described by Thisse and Thisse [43]. Briefly, the fixed embryos were rehydrated and permeabilized with 10 µg/mL proteinase K for 10 min (2 dpf embryos) or 30 min (3-5 dpf larvae) at room temperature. Twenty-four to 50 embryos or larvae from ezh2^+/- in-crosses were hybridized with digoxigenin-labeled antisense RNA probes at 70°C. After extensive washing, the probes were detected with anti-digoxigenin-AP Fab fragment (Roche Diagnostics, 1093274, diluted at 1:10,000), followed by staining with BCIP/NBT (5-bromo-4-chloro-3-indolyl-phosphate/nitro blue tetrazolium) alkaline phosphate substrate.

The embryos and larvae were imaged using a Leica MZ10F stereomicroscope equipped with a Leica DFC295 digital camera. After imaging, the stained-embryos and -larvae are genotyped (see Supplementary Figures S1-S25).

4.3. Genotype Analyses

To genotype paraformaldehyde-fixed embryos and larvae, DNA was extracted using sodium hydroxide and Tris [15]. Single embryos or larvae were placed into microcentrifuge tubes containing 20 μL of 50 mM NaOH and heated for 20 min at 95 °C. The tubes were then cooled to 4 °C, and 2 μL of 1 M Tris-HCl, pH 7.4, was added to neutralize the basic solution. Genotype analysis was performed by PCR on 2.5 μL of samples using the primer set TAL_ezh2_En_5’ (AAATCGGAGAAGGGTCCTG) and TAL_ezh2_En_3’ (ACACACATGCAACTGGACTC) followed by a 2.5% agarose gel electrophoresis. The 22-bp insertion in the mutant allele allows to distinguish (+/+), (+/-) and (-/-) genotypes.

4.4. Locomotor Activity Assays

The locomotor activity assays were performed on 5 dpf larvae from ezh2^+/- in-crosses in 48-well plates that were handled minimally before placement in a Zebrabox chamber (ViewPoint Life Sciences, Lyon, France) equipped with an infrared light-emitting floor and a top-mounted infrared camera allowing video recording of the whole plate under both light and dark conditions. Larval behavior measurements, during a protocol consisting of a 10 min initial acclimating period in the dark, followed by six alternating 10 min light and dark phases, were achieved using the ZebraLab software with a detection threshold set at 35 and an xmin set at 3 (ViewPoint Life Sciences). After recording, the larvae were euthanized and their genotype determined as detailed before. Locomotor activity assays have been conducted 5 times from 3 independent ezh2^+/- in-crosses (see Supplementary Figure S26).

4.5. Statistical Analyses

Data analysis was carried out using the R/RStudio packages for statistical computing [44]. Data sets were tested for normal distribution by the Shapiro–Wilk’s test. Group differences were calculated by Kruskal-Wallis test followed by a Dunn’s post hoc test. ns, no statistical difference; *, p <0.01; **, p <0.05; ***, p <0.001.