Loss of Polycomb Repressive Complex 2 Function Alters Digestive Organ Homeostasis and Neuronal Differentiation in Zebrafish

Polycomb repressive complex 2 (PRC2) mediates histone H3K27me3 methylation and the stable transcriptional repression of a number of gene expression programs involved in the control of cellular identity during development and differentiation. Here, we report on the generation and on the characterization of a zebrafish line harboring a null allele of eed, a gene coding for an essential component of the PRC2. Homozygous eed-deficient mutants present a normal body plan development but display strong defects at the level of the digestive organs, such as reduced size of the pancreas, hepatic steatosis, and a loss of the intestinal structures, to die finally at around 10–12 days post fertilization. In addition, we found that PRC2 loss of function impairs neuronal differentiation in very specific and discrete areas of the brain and increases larval activity in locomotor assays. Our work highlights that zebrafish is a suited model to study human pathologies associated with PRC2 loss of function and H3K27me3 decrease.


Genotype Analyses
For genotyping, embryos, larvae or caudal biopsies of larvae were incubated in 10 µL PCR extraction buffer (10 mM Tris-HCl pH 8.0, 2 mM EDTA, 0.2% Triton X-100, 100 µg/mL proteinase K) and placed at 50 • C for 4 h prior to proteinase K inactivation at 95 • C for 5 min. Genotype analysis was performed by PCR on 2.5 µL of samples using the primer set TAL_eed_5 (CAGCCTAAAGGTGAGGCCGC) and TAL_eed_3 a (CGATCTCGACC-CGGAACAC) followed by PCR product digestion with the BseNI restriction enzyme. An alternate genotyping strategy consists of the PCR amplification of genomic DNAs with the primer set TAL_eed5_new GTGTGCAGGAGGATCATGGTCAG) and TAL_eed3_new (GTCCGTCACTCACTCTGTTGCTG) followed by a 2% agarose gel electrophoresis without enzymatic digestion.
To genotype paraformaldehyde-fixed embryos and larvae, DNA was extracted using sodium hydroxide and Tris [46]. Single embryos or larvae were placed into microcentrifuge tubes containing 20 µL 50 mM NaOH and heated 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 on 2.5 µL of samples by PCR-BseNI digestion or PCR without digestion, as described above.

Kaplan-Meier Analysis
For the establishment of survival curves, the tip of the caudal fin of embryos from an eed +/− incross was transected within the pigment gap distal to the circulating blood for genotyping purposes [46]. The embryos were then pooled according to their genotypes, placed into separate 1-L tanks, and incubated at 28 • C. Larvae were fed from 5 dpf on, 3 times per day, with early-stage zebrafish nutrition (Gemma Micro ZF75, Planktovie, France). The larvae were regularly examined directly in the tanks using a Leica EZ4 stereomicroscope with no or minimal manipulation during a period of 18 days after the spawn. Larvae were declared dead when heartbeats could not be detected. Dead larvae were immediately removed from the tanks.

Oil Red-O Staining
Zebrafish larvae were fixed overnight in 4% paraformaldehyde and pigments were bleached 2 h in 1% KOH, 3% H 2 O 2 . Larvae were stained with oil red-O (Hamiya Biomedical Company, Tukwila, WA, USA) for 15 min. Stained larvae were washed with PBS, stored in Cells 2021, 10, 3142 4 of 23 70% glycerol, and imaged on a bright-field Leica MZ125 stereomicroscope equipped with a Leica DFC295 digital camera. After imaging, DNA was extracted using sodium hydroxide and Tris and genotyped.

Smurf Assay
Smurf assays were conducted according to Dambroise et al. [47] by placing zebrafish larvae in 50 mL Falcon tubes containing 2.5% blue #1 (erioglaucine disodium salt, Sigma) for 30 min at room temperature. They were then rinsed under aquarium water until no more blue coloration could be found in the eluate, anesthetized in 0.016% tricaine (MS-222; 3-amino benzoic acid ethyl ester, Sigma) and imaged on a bright-field stereomicroscope (Leica MZ125) equipped with a Leica DFC295 digital camera. After imaging, DNA was extracted for genotyping analyses.

Histological Analysis
For histological analyses, tail biopsies were used for genotyping before paraformaldehyde fixation of the larvae. Larvae were embedded in paraffin and cut into 5 µm-thick sections. These were mounted on sylanated glass slides, deparaffinated, rehydrated, and stained with hematoxylin and eosin for histological analysis as previously described [46].

Biometric Measurements
Standard length and eye diameter were measured according to Parichy et al. [48] on anesthetized larvae using ImageJ. After imaging and measurement, the larvae were genotyped.
Measurements of the surface of the pancreas labeled by in situ hybridization with a prss1 probe were performed by equal intensity pixel counting using ImageJ.
Quantification of macrovesicular in the liver was performed on at least 3 consecutive hematoxylin and eosin-stained slides from 4 wild-type and 4 eed −/− mutant larvae using ImageJ. Each image was anonymized and transformed into an 8-bit grayscale picture before counting the hepatic nuclei and the number of vesicles.
For all biometric measurements, the software GraphPad Prism was used to build the graphs and to perform the statistical analyses.

Confocal Microscopy
For confocal imaging, tail biopsies were transected for genotyping before paraformaldehyde fixation of the larvae. After fixation in 4% paraformaldehyde, zebrafish larvae were maintained on glass-bottom dishes in 1.5% agarose. Fluorescent images were captured with an LSM 880 microscope and an AiryScan detector (ZEISS, Oberkoden, Germany). A Z-stack slice was made every 5 µm.

Whole-Mount In Situ Hybridization
Antisense-RNA probes were generated by RT-PCR from total mRNA extracted from zebrafish larvae at 5 dpf using the RNeasy Mini Kit (Qiagen, Courtaboeuf, France), following the manufacturer's protocol. After reverse transcription (Superscript III, Invitrogen), cDNAs were amplified by PCR using the probe-specific primers, coupled to the T7 sequence for forward primers and the SP6 sequence for reverse primers. DIG-labeled antisense-RNA probes were synthetized using the DIG RNA Labeling Kit (SP6) (Roche-Sigma Aldrich Chimie S.a.r.l., Saint-Quentin-Follavier, France), following the manufacturer's instructions.
The primers used for probe generation were: Whole-mount in situ hybridizations were performed according to Thisse and Thisse [49]. Briefly, the fixed embryos were rehydrated and permeabilized with 10 µg/mL proteinase K for 30 sec (1-cell stage embryos), 10 min (24 hpf embryos), or 30 min (48 hpf embryos and 5 dpf larvae) at room temperature. Ten to 15 embryos from the TU strain of each time point, or about 50 larvae from eed +/− incrosses at 5 dpf, were hybridized with digoxigeninlabeled 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 were imaged using a Leica MZ10F stereomicroscope equipped with a Leica DFC295 digital camera.

Histone Extraction and Western Blot Analysis
Histone extraction and Western blot analyses were performed from 10 larvae as described previously [46].

RNA Extraction, RT-PCR and RT-qPCR
RNA extraction and RT-PCR experiments were conducted as previously described [50]. cDNA was synthesized using Superscript III (18080-044, Invitrogen, Carlsbad, CA, USA) following the manufacturer's instructions. A quantity of 1 µg of total RNAs were used to perform the reverse transcription experiments. Primers used were: Dr_eed_cDNA_5 : GGAAACGAGATGCCGAACAA Dr-eed_cDNA_3 a: CAGCCGGATCTCTCCCTGAG Dr_cDNA_ube2a_F: TGACTGTTGACCCACCTTACAG The quantitative qPCR reaction was performed in triplicate using a QuantStudio 3 Real-Time System (Thermo Fisher Scientific, Waltham, MA, USA) using SYBR Green Supermix (Bio-Rad, Marne-la-Coquette, France). Relative mRNA expression of each gene was normalized to ube2a levels.

Behavioral Studies
Behavioral studies were performed on 5 dpf larvae from eed +/− incrosses in 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 were achieved using the ZebraLab software with a detection threshold set at 35 and an xmin set at 3 (ViewPoint Life Sciences, Lyon, France). After recording, the larvae were euthanized, and their genotype was determined.
The locomotor activity assays were conducted in 48-well plates. The protocol consists of a 10 min initial acclimating period in the dark, followed by six alternating 10 min light and dark phases.
Thigmotaxis assays were conducted in 24-well plates. The timeline of the thigmotaxis assay protocol consists of a 6 min acclimatization step in the light followed by a 4 min dark challenge period. The arena was divided into two areas, inner and outer, of the equivalent surface. The diameter of the inner area corresponds to the total diameter of the well divided by √ 2. Thigmotaxis was presented as the percentage of the total distance moved or the total time spent by each larva into the outer zone during the challenge period.

eed Expression during Zebrafish Development
Human EED protein is encoded by 12 exons containing gene located on chromosome 11, upstream of the HIKESHI (heat shock protein nuclear import factor hikeshi) gene. The protein is composed of 441 amino acids and contains six WD40 domains adopting a β-propeller architecture forming multiple van der Waals interactions and hydrogen bonds with a domain located at the N-terminus of EZH2 [51]. EED is conserved across vertebrate species, both at the protein and genomic levels. In zebrafish, the eed gene is located on chromosome 1. It is organized in 12 exons positioned at the 5 of the hikeshi gene, thus defining an evolutionarily conserved synteny block. The gene also codes for a 441-amino-acids-long protein sharing 91.4% identity with its human ortholog and containing six WD40 domains as well (Figure 1a).
cies, both at the protein and genomic levels. In zebrafish, the eed gene is located on chromosome 1. It is organized in 12 exons positioned at the 5' of the hikeshi gene, thus defining an evolutionarily conserved synteny block. The gene also codes for a 441-amino-acidslong protein sharing 91.4% identity with its human ortholog and containing six WD40 domains as well (Figure 1a). In zebrafish, zygotic transcription is activated at about cell cycle 10-13, around 3.5 h post fertilization (hpf). Before this mid-blastula transition (MBT) stage, all developmental events depend on the maternally provided gene products [52,53]. Using in situ hybridization at 1-cell and 2-cell stages ( Figure 1b) and RT-PCR at 1 and 3 hpf (Figure 1c), we found in agreement with previous studies [54,55] that eed mRNA is maternally loaded into the zebrafish embryo. After MBT, during tissue specification, eed zygotic expression is detected by RT-PCR ( Figure 1c) and appears enriched in the anterior region of the embryo and in highly proliferative tissues such as the retina, the midbrain-hindbrain boundary, the pharyngeal arches, the pectoral fin buds or the intestine as shown by in situ hybridization at 24, 48 and 72 hpf (Figure 1b). Finally, in adult zebrafish, RT-PCR experiments showed that eed is ubiquitously expressed in all the tissues we tested (Figure 1d).

TALEN-Mediated Inactivation of eed in Zebrafish
To generate eed mutations in zebrafish, we designed TALENs targeting the exon 3 within the sequence coding for the first WD40 repeat (Figure 2a). The targeted site was chosen to cover a BseNI restriction site used to screen for TALEN-induced mutations and In zebrafish, zygotic transcription is activated at about cell cycle 10-13, around 3.5 h post fertilization (hpf). Before this mid-blastula transition (MBT) stage, all developmental events depend on the maternally provided gene products [52,53]. Using in situ hybridization at 1-cell and 2-cell stages ( Figure 1b) and RT-PCR at 1 and 3 hpf (Figure 1c), we found in agreement with previous studies [54,55] that eed mRNA is maternally loaded into the zebrafish embryo. After MBT, during tissue specification, eed zygotic expression is detected by RT-PCR ( Figure 1c) and appears enriched in the anterior region of the embryo and in highly proliferative tissues such as the retina, the midbrain-hindbrain boundary, the pharyngeal arches, the pectoral fin buds or the intestine as shown by in situ hybridization at 24, 48 and 72 hpf (Figure 1b). Finally, in adult zebrafish, RT-PCR experiments showed that eed is ubiquitously expressed in all the tissues we tested (Figure 1d).

TALEN-Mediated Inactivation of eed in Zebrafish
To generate eed mutations in zebrafish, we designed TALENs targeting the exon 3 within the sequence coding for the first WD40 repeat (Figure 2a). The targeted site was chosen to cover a BseNI restriction site used to screen for TALEN-induced mutations and for genotyping purposes (Figure 2b). Following TALEN mRNA injection in zebrafish embryos from the TU strain at the 1-cell stage, we isolated a line carrying an allele, eed ul4 , with a deletion of 14 base pairs (Figure 2c). This mutation produces a frameshift leading to the synthesis of a predicted protein of 136 amino acids lacking all the WD40 domains ( Figure 2d). Therefore, eed ul4 is very likely a null allele and henceforth referred to as eed − . for genotyping purposes (Figure 2b). Following TALEN mRNA injection in zebrafish embryos from the TU strain at the 1-cell stage, we isolated a line carrying an allele, eed ul4 , with a deletion of 14 base pairs (Figure 2c). This mutation produces a frameshift leading to the synthesis of a predicted protein of 136 amino acids lacking all the WD40 domains ( Figure  2d). Therefore, eed ul4 is very likely a null allele and henceforth referred to as eed − . The sequence targeted by the TALEN in exon 3 is indicated with left and right TALEN binding sites shown in red, while the BseNI restriction sequence is underlined; (b) identification of mutant embryos using restriction fragment length polymorphism. Genomic DNA was prepared from an uninjected (Control) and an eed TALEN injected (TAL-eed) embryo, amplified by PCR and subjected to BseNI digestion. The TAL-eed injected embryo contains undigested material (arrow at 407 bp), indicating that the BseNI diagnostic restriction site has been disrupted; (c) sequence of the mutant allele compared to its wild-type counterpart. Dashes show deleted nucleotides. The mutated eed ul4 allele possesses a deletion of 14 nucleotides; (d) schematic representation of wild-type (Eed wt ) and predicted mutant (Eed ul4 ) proteins. The gray line in the predicted mutant protein corresponds to residues read out of frame prior to encountering a premature STOP codon. The green triangles in the wild-type protein show the WD40 domains; (e) Kaplan-Meier survival curves over 18 days for eed +/+ (red curve) and eed -/-(green curve) siblings from a cross between heterozygous eed +/− fish. The number of fish considered is indicated; (f) measurement of the total length of eed +/+ (red) and eed −/− (green) siblings from a cross between heterozygous eed +/-fish at 11 dpf. Statistical significance was assessed using a t-test; (g) measurement of the eye diameter of eed +/+ (red) and eed -/-(green) siblings from a cross between heterozygous eed +/-fish at 11 dpf. Statistical significance was assessed using a Student t-test.
Adult heterozygous eed +/− fish are viable, fertile, and do not show any obvious phenotype. In contrast, and similarly to ezh2 −/− [37,38] and suz12a −/− ; suz12b −/− fish [39], homozygous eed −/− mutants die at around 10-12 dpf (Figure 2e). Analysis of the living eed -/-mutants at 11 dpf showed that they do not present gross morphological alterations while they appear significantly smaller than wild-type siblings (Figure 2f,g). Then, eed function is not required for early development and normal formation of the body plan but is necessary for zebrafish survival after 10-12 dpf. Genomic DNA was prepared from an uninjected (Control) and an eed TALEN injected (TAL-eed) embryo, amplified by PCR and subjected to BseNI digestion. The TAL-eed injected embryo contains undigested material (arrow at 407 bp), indicating that the BseNI diagnostic restriction site has been disrupted; (c) sequence of the mutant allele compared to its wild-type counterpart. Dashes show deleted nucleotides. The mutated eed ul4 allele possesses a deletion of 14 nucleotides; (d) schematic representation of wild-type (Eed wt ) and predicted mutant (Eed ul4 ) proteins. The gray line in the predicted mutant protein corresponds to residues read out of frame prior to encountering a premature STOP codon. The green triangles in the wild-type protein show the WD40 domains; (e) Kaplan-Meier survival curves over 18 days for eed +/+ (red curve) and eed −/− (green curve) siblings from a cross between heterozygous eed +/− fish. The number of fish considered is indicated; (f) measurement of the total length of eed +/+ (red) and eed −/− (green) siblings from a cross between heterozygous eed +/− fish at 11 dpf. Statistical significance was assessed using a t-test; (g) measurement of the eye diameter of eed +/+ (red) and eed −/− (green) siblings from a cross between heterozygous eed +/− fish at 11 dpf. Statistical significance was assessed using a Student t-test.
Adult heterozygous eed +/− fish are viable, fertile, and do not show any obvious phenotype. In contrast, and similarly to ezh2 −/− [37,38] and suz12a −/− ; suz12b −/− fish [39], homozygous eed −/− mutants die at around 10-12 dpf (Figure 2e). Analysis of the living eed −/− mutants at 11 dpf showed that they do not present gross morphological alterations while they appear significantly smaller than wild-type siblings (Figure 2f,g). Then, eed function is not required for early development and normal formation of the body plan but is necessary for zebrafish survival after 10-12 dpf.

Loss of eed Function Alters H3K27me2/3 Levels, but Not H3K27me1 nor H2AK119ub
To investigate the effect of eed loss of function on histone modifications, we performed a series of Western blot analyses using antibodies recognizing specific histone marks ( Figure 3). Histones were extracted from eed +/+ and eed −/− siblings at 9 dpf and subjected to Western blot analyses to determine global levels of several histone modifications.

Loss of Eed Function Alters H3K27me2/3 Levels, but Not H3K27me1 nor H2AK119ub
To investigate the effect of eed loss of function on histone modifications, we performed a series of Western blot analyses using antibodies recognizing specific histone marks ( Figure 3). Histones were extracted from eed +/+ and eed −/− siblings at 9 dpf and subjected to Western blot analyses to determine global levels of several histone modifications. Figure 3. Effects of eed loss of function on global levels of histone modifications: (a) after genotyping of the caudal extremity of larvae, 10 to 15 μg of bulk histones from a pool of 10 eed +/+ or eed −/− siblings at 9 dpf were analyzed by Western blot using specific antibodies recognizing H3K27me3, H3K27me2, H3K27me1, H2AK119ub, H3K27ac, H3K9me3 or H4K20me3. After stripping, the membranes were reprobed with an anti-histone H3 as a control; (b) comparison between eed +/+ (red) and eed −/− (green) siblings, of histone modification signals normalized by total histone H3 and expressed relative to the levels found in the wild-type fish. Quantified levels are the mean ± SD of at least three independent histone extractions, except for H3K9me3, where two histone extraction were performed, followed by Western blot analysis. Statistical significance was assessed using a Student t-test; ns, non-significant; *, p < 0.05; **, p < 0.01. (See Supplementary Materials).
Not surprisingly, a great reduction in bulk H3K27me3 and H3K27me2 levels is observed in eed −/− homozygous larvae at 9 dpf. However, faint signals for H3K27me2/3 methylations are still detected in homozygous mutant fish, presumably due to the activity of the maternal products initially deposited in the embryos. In contrast, we did not find significant differences in H3K27me1 levels between wild-type and mutant larvae. This puzzling observation might suggest that in the absence of PRC2 activity, other histone lysine methyltransferases could monomethylate H3K27 and compensate for PRC2 loss of activity. Because mammalian G9a (EHMT2) and Glp (EHMT1) histone lysine methyltransferases have been shown to target H3K27 in vivo [56,57], their zebrafish orthologs (Ehmt1a, Ehmt1b, and Ehmt2) are potential candidates that could maintain H3K27me1 levels in the Figure 3. Effects of eed loss of function on global levels of histone modifications: (a) after genotyping of the caudal extremity of larvae, 10 to 15 µg of bulk histones from a pool of 10 eed +/+ or eed −/− siblings at 9 dpf were analyzed by Western blot using specific antibodies recognizing H3K27me3, H3K27me2, H3K27me1, H2AK119ub, H3K27ac, H3K9me3 or H4K20me3. After stripping, the membranes were reprobed with an anti-histone H3 as a control; (b) comparison between eed +/+ (red) and eed −/− (green) siblings, of histone modification signals normalized by total histone H3 and expressed relative to the levels found in the wild-type fish. Quantified levels are the mean ± SD of at least three independent histone extractions, except for H3K9me3, where two histone extraction were performed, followed by Western blot analysis. Statistical significance was assessed using a Student t-test; ns, non-significant; *, p < 0.05; **, p < 0.01. (See Supplementary Materials).
Not surprisingly, a great reduction in bulk H3K27me3 and H3K27me2 levels is observed in eed −/− homozygous larvae at 9 dpf. However, faint signals for H3K27me2/3 methylations are still detected in homozygous mutant fish, presumably due to the activity of the maternal products initially deposited in the embryos. In contrast, we did not find significant differences in H3K27me1 levels between wild-type and mutant larvae. This puzzling observation might suggest that in the absence of PRC2 activity, other histone lysine methyltransferases could monomethylate H3K27 and compensate for PRC2 loss of activity. Because mammalian G9a (EHMT2) and Glp (EHMT1) histone lysine methyltransferases have been shown to target H3K27 in vivo [56,57], their zebrafish orthologs (Ehmt1a, Ehmt1b, and Ehmt2) are potential candidates that could maintain H3K27me1 levels in the absence of PRC2 function. In agreement with other studies using knockout embryonic stem cells (ESCs) and in pathological situations where the loss of H3K27me2/3 correlates with increased acetylation at H3K27 [58,59], we also found that eed −/− homozygous larvae display a slight increase in global H3K27ac levels at 9 dpf ( Figure 3).
Monoubiquitylation of H2AK119 contributes to polycomb repression and is achieved by the action of the catalytic subunits RING1/RNF2 of PRC1 [10,11]. PRC1 complexes consist of a variety of protein assemblies with diverse compositions and are classified as canonical (cPRC1) and non-canonical (ncPRC1) complexes [60,61]. This classification is mainly based on the presence of a chromobox (CBX) protein in cPRC1 complexes and YY1-binding protein (RYBP), or its homolog YAF2, in ncPRC1 complexes. Whereas cPRC1 is recruited to H3K27me3 via its CBX subunit, ncPRC1 complexes mediate H2A ubiquitylation at polycomb target sites independently of PRC2 action and of the presence of H3K27me3 epigenetic marks [60][61][62]. The examination of bulk H2AK119ub did not show differences between eed −/− and eed +/+ homozygous larvae at 9 dpf, indicating that in the absence of PRC2 function, ncPRC1 complexes are able to maintain normal levels of H2AK119ub ( Figure 3).
Finally, to check whether an increase in the heterochromatin marks H3K9me3 and H4K20me3 could compensate for the decrease in H3K27me2/3, we investigated their relative abundance in eed −/− homozygous mutant larvae at 9 dpf, compared to wild-type. If a decrease in H4K20me3 levels could be found in PRC2-deficient fish, we were not able to detect significant differences in bulk H3K9me3 ( Figure 3). Then, the normal formation of the body plan in eed −/− mutants cannot be explained by the compensation of H3K27me2/3 loss through the increase in other repressive marks.

Loss of eed Function Alters the Homeostasis of Digestive Organs in Zebrafish
Because it has been shown that ezh2-deficient zebrafish develop intestinal defects [38,63], we investigated the potential effect of eed loss of function on the development of digestive organs. First, we examined the expression of markers of the intestine (fabp2, fatty acidbinding protein 2, intestinal and ace2, angiotensin I converting enzyme 2), the liver (fabp10a, fatty acid-binding protein 10a, liver basic), and the exocrine pancreas (prss1, serine protease 1 also known as trypsin, try) in 5-dpf old larvae by in situ hybridization (Figure 4a). Pools of larvae from heterozygous eed +/− incrosses were subjected to in situ hybridization and individually pictured before being genotyped by RFLP.
absence of PRC2 function. In agreement with other studies using knockout embryonic stem cells (ESCs) and in pathological situations where the loss of H3K27me2/3 correlates with increased acetylation at H3K27 [58,59], we also found that eed −/− homozygous larvae display a slight increase in global H3K27ac levels at 9 dpf (Figure 3).
Monoubiquitylation of H2AK119 contributes to polycomb repression and is achieved by the action of the catalytic subunits RING1/RNF2 of PRC1 [10,11]. PRC1 complexes consist of a variety of protein assemblies with diverse compositions and are classified as canonical (cPRC1) and non-canonical (ncPRC1) complexes [60,61]. This classification is mainly based on the presence of a chromobox (CBX) protein in cPRC1 complexes and YY1-binding protein (RYBP), or its homolog YAF2, in ncPRC1 complexes. Whereas cPRC1 is recruited to H3K27me3 via its CBX subunit, ncPRC1 complexes mediate H2A ubiquitylation at polycomb target sites independently of PRC2 action and of the presence of H3K27me3 epigenetic marks [60][61][62]. The examination of bulk H2AK119ub did not show differences between eed −/− and eed +/+ homozygous larvae at 9 dpf, indicating that in the absence of PRC2 function, ncPRC1 complexes are able to maintain normal levels of H2AK119ub ( Figure 3).
Finally, to check whether an increase in the heterochromatin marks H3K9me3 and H4K20me3 could compensate for the decrease in H3K27me2/3, we investigated their relative abundance in eed −/− homozygous mutant larvae at 9 dpf, compared to wild-type. If a decrease in H4K20me3 levels could be found in PRC2-deficient fish, we were not able to detect significant differences in bulk H3K9me3 (Figure 3). Then, the normal formation of the body plan in eed −/− mutants cannot be explained by the compensation of H3K27me2/3 loss through the increase in other repressive marks.

Loss of Eed Function Alters the Homeostasis of Digestive Organs in Zebrafish
Because it has been shown that ezh2-deficient zebrafish develop intestinal defects [38,63], we investigated the potential effect of eed loss of function on the development of digestive organs. First, we examined the expression of markers of the intestine (fabp2, fatty acid-binding protein 2, intestinal and ace2, angiotensin I converting enzyme 2), the liver (fabp10a, fatty acid-binding protein 10a, liver basic), and the exocrine pancreas (prss1, serine protease 1 also known as trypsin, try) in 5-dpf old larvae by in situ hybridization (Figure 4a). Pools of larvae from heterozygous eed +/− incrosses were subjected to in situ hybridization and individually pictured before being genotyped by RFLP. The expression of fabp2, ace2, and fabp10a revealed that the intestine and the liver are normally developed in eed −/− larvae at 5 dpf. Furthermore, histological studies showed that the structure of the intestine is identical in eed +/+ and eed −/− zebrafish at this stage ( Figure 4b). In contrast, and similar to what was observed in ezh2 −/− mutants [38], the size of the pancreas appears smaller in eed-deficient fish at 5 dpf (Figure 4a,c).
To visualize the intestine structure in the whole larvae at a later developmental stage by confocal microscopy, we used the transgenic line Tg(actb2:GFP-Hsa.UTRN) e116 in which actin is labeled by the green fluorescent protein (GFP) [40,64,65]. Fluorescence imaging revealed an alteration of the intestine structure of the eed −/− transgenics at 9 dpf. Indeed, the number of cells forming the thickness of the intestinal wall and visible through GFP expression is decreased in eed-deficient zebrafish as of 9 dpf (Figure 5a). Then, while the intestine develops normally until 5 dpf, its integrity fails to be maintained at later developmental stages. Figure 4. Organization of the digestive organs at 5 dpf: (a) whole-mount RNA in situ hybridization to detect the expression of markers of the intestine (fabp2, ace2), the liver (fapb10a), and the exocrine pancreas (prss1) in eed +/+ and eed −/− siblings at 5 dpf. The red arrow shows the pancreas. Scale bar is 500 μm; (b) intestinal bulb sections from eed +/+ (left) and eed −/− (right) larvae at 5 dpf stained with hematoxylin and eosin. Scale bar is 50 μm; (c) measurement of the surface of the pancreas labeled by in situ hybridization using a prss1 probe at 5 dpf for eed +/+ (red) and eed −/− (green) larvae. Statistical significance was assessed by a Student t-test, and the corresponding p-value is indicated.
The expression of fabp2, ace2, and fabp10a revealed that the intestine and the liver are normally developed in eed −/− larvae at 5 dpf. Furthermore, histological studies showed that the structure of the intestine is identical in eed +/+ and eed −/− zebrafish at this stage ( Figure  4b). In contrast, and similar to what was observed in ezh2 −/− mutants [38], the size of the pancreas appears smaller in eed-deficient fish at 5 dpf (Figure 4a,c).
To visualize the intestine structure in the whole larvae at a later developmental stage by confocal microscopy, we used the transgenic line Tg(actb2:GFP-Hsa.UTRN) e116 in which actin is labeled by the green fluorescent protein (GFP) [40,64,65]. Fluorescence imaging revealed an alteration of the intestine structure of the eed −/− transgenics at 9 dpf. Indeed, the number of cells forming the thickness of the intestinal wall and visible through GFP expression is decreased in eed-deficient zebrafish as of 9 dpf (Figure 5a). Then, while the intestine develops normally until 5 dpf, its integrity fails to be maintained at later developmental stages. The analysis of histological sections of larvae at 11 dpf confirmed that the intestine wall of eed −/− mutants is strongly reduced and lacks folds at the level of the intestinal bulb (Figure 5b). By contrast, the structure of the proximal intestine does not show these strong defects, as previously observed in ezh2 −/− mutants [38]. Because the intestine looks normal at 5 dpf but altered at 9-11 dpf, we conclude that in the absence of eed and PRC2 functions, the development of the intestine occurs, but the integrity of its structure is not maintained. The analysis of histological sections of larvae at 11 dpf confirmed that the intestine wall of eed −/− mutants is strongly reduced and lacks folds at the level of the intestinal bulb (Figure 5b). By contrast, the structure of the proximal intestine does not show these strong defects, as previously observed in ezh2 −/− mutants [38]. Because the intestine looks normal at 5 dpf but altered at 9-11 dpf, we conclude that in the absence of eed and PRC2 functions, the development of the intestine occurs, but the integrity of its structure is not maintained. To investigate the permeability of the intestine in mutants, we performed a Smurf assay [47]. In this assay, the permeability of the intestine was assessed by evaluating the diffusion throughout the body of blue food dye present in the fish water. Figure 5c shows that in spite of a strong alteration of the structure of the intestine bulb, the intestinal permeability is not increased.
Neutral lipid dye oil red-O staining of larval zebrafish is commonly used to examine steatosis caused by consumption of a high-fat diet or induced by various drugs [66][67][68][69][70]. Oil red-O staining clearly shows a strong accumulation of neutral lipids in the liver of eed −/− larvae à 11 dpf when compared to wild-type larvae (Figure 6a). Such a strong oil red-O staining of the liver was also found in ezh2 −/− homozygous mutants [38], whereas exposure to Ezh1/2 inhibitor PF-06726304 acetate increases lipid accumulation in larval zebrafish [71]. Moreover, hematoxylin and eosin stain on histological sections of zebrafish larvae at 11 dpf showed signs of macrovesicular steatosis in eed −/− larvae (Figures 5b and 6b,c).
To investigate the permeability of the intestine in mutants, we performed a Smurf assay [47]. In this assay, the permeability of the intestine was assessed by evaluating the diffusion throughout the body of blue food dye present in the fish water. Figure 5c shows that in spite of a strong alteration of the structure of the intestine bulb, the intestinal permeability is not increased.
Neutral lipid dye oil red-O staining of larval zebrafish is commonly used to examine steatosis caused by consumption of a high-fat diet or induced by various drugs [66][67][68][69][70]. Oil red-O staining clearly shows a strong accumulation of neutral lipids in the liver of eed −/− larvae à 11 dpf when compared to wild-type larvae (Figure 6a). Such a strong oil red-O staining of the liver was also found in ezh2 −/− homozygous mutants [38], whereas exposure to Ezh1/2 inhibitor PF-06726304 acetate increases lipid accumulation in larval zebrafish [71]. Moreover, hematoxylin and eosin stain on histological sections of zebrafish larvae at 11 dpf showed signs of macrovesicular steatosis in eed −/− larvae (Figures 5b and  6b,c). (c) quantification of macrovesicular steatosis eed +/+ (red) and eed -/-(green) larvae at 11 dpf. The ratio of macrovesicular structures relative to total hepatic surface was calculated on 3 histological slices per larvae from 4 different larvae per genotype. Statistical significance was assessed using a Mann-Whitney t-test; (d) relative levels of fasn, srebf1, pparg, and nr1h3 RNA abundances in 9 dpf old eed +/+ (red) and eed -/-(green) larvae quantified by RT-qPCR. Three independent experiments were performed, and error bars represent standard deviation. Statistical analysis was performed using a one-way ANOVA test followed by a Dunn's multiple test comparison; ns, non-significant; *, p < 0.05; **, p < 0.01.
To investigate the impact of eed loss of function on the expression of lipogenic factors, we performed RT-qPCRs to determine RNA abundance for fasn (fatty acid synthetase), srebf1 (sterol regulatory element-binding transcription factor 1), pparg (peroxisome proliferator-activated receptor gamma), and nr1h3 (nuclear receptor subfamily 1, group H, member 3, also known as liver X receptor lxr) at 9 dpf. Then, in spite of showing an increase in lipid accumulation, RNA levels for lipogenic factors remain either unchanged (srebf1 and nr1h3) or reduced (fasn and pparg) in eed -/-larvae (Figure 6d). This suggests that (c) quantification of macrovesicular steatosis eed +/+ (red) and eed −/− (green) larvae at 11 dpf. The ratio of macrovesicular structures relative to total hepatic surface was calculated on 3 histological slices per larvae from 4 different larvae per genotype. Statistical significance was assessed using a Mann-Whitney t-test; (d) relative levels of fasn, srebf1, pparg, and nr1h3 RNA abundances in 9 dpf old eed +/+ (red) and eed −/− (green) larvae quantified by RT-qPCR. Three independent experiments were performed, and error bars represent standard deviation. Statistical analysis was performed using a one-way ANOVA test followed by a Dunn's multiple test comparison; ns, non-significant; *, p < 0.05; **, p < 0.01.
To investigate the impact of eed loss of function on the expression of lipogenic factors, we performed RT-qPCRs to determine RNA abundance for fasn (fatty acid synthetase), srebf1 (sterol regulatory element-binding transcription factor 1), pparg (peroxisome proliferator-activated receptor gamma), and nr1h3 (nuclear receptor subfamily 1, group H, member 3, also known as liver X receptor lxr) at 9 dpf. Then, in spite of showing an increase in lipid accumulation, RNA levels for lipogenic factors remain either unchanged (srebf1 and nr1h3) or reduced (fasn and pparg) in eed −/− larvae (Figure 6d). This suggests that eed loss of function-induced steatosis is likely the result of an increase in lipid storage in the liver rather than an increase in lipid synthesis.
Taken together, our study of eed-deficient zebrafish mutants demonstrates that PRC2 loss of function has various effects on the homeostasis of digestive organs, ranging from a delay in pancreatic development, a loss of maintenance of the intestine structures to hepatic steatosis.

Loss of eed Function Alters Neuronal Differentiation
To characterize in more detail our eed-deficient zebrafish mutants, we performed additional whole-mount RNA in situ hybridization experiments at 5 dpf with a focus on the brain. Mutant and wild-type siblings were treated in one batch for each probe and documented photographically. The genotype of each embryo was then determined by RFLP on genomic DNA extracted from the fixed and stained material. First, we showed that eed expression is strongly reduced in eed −/− larvae, presumably due to nonsensemediated decay (NMD) of the mutant transcript (Figure 7a). Second, we investigated the expression profile of phox2bb (paired like homeobox 2Bb), a transcription factor expressed by a subset of cells in the posterior regions of the brain. In eed −/− mutants, phox2bb appears additionally expressed ectopically in the retina (Figure 7a). This phox2bb misexpression in the retina was also observed in ezh2-deficient zebrafish larvae [38]. Then, eed loss of function recapitulates the defects found in ezh2 −/− mutants at the level of the digestive organs, but also in terms of phox2bb expression in the brain and the retina. Moreover, while eed is expressed in large territories of the brain, its loss of function impairs phox2bb specifically in the retina, indicating that PRC2 controls gene expression programs in a very subtle and discreet manner, depending on the cell type.
To gain insights into the possible role of PRC2 in brain development, we applied in situ hybridizations at 5 dpf on several genes known to be expressed in the midbrain-hindbrain boundary region that contains neural stem cells and dividing progenitors [72], and where eed transcripts are abundant (Figures 1a and 7a) [54]. The genes chosen for the study were the stemness markers bmi1a (bmi1 polycomb ring finger oncogene 1a) and sox2 (SRY-box transcription factor 2); the proliferation marker pcna (proliferating cell nuclear antigen); mycn (MYCN proto-oncogene, bHLH transcription factor), which is involved in several signaling pathways promoting cell growth, proliferation and metabolism of progenitor cells in different developing organs and tissues; two genes coding for cyclins, ccna2 (cyclin A2) and ccnd1 (cyclin D1); neurod1 (neuronal differentiation 1), a marker of neuronal precursors; and mag (myelin-associated glycoprotein), a marker of glial precursors (Figure 7b).
Analysis of expression profiles of the stemness markers bmi1a and sox2 revealed a strong increase in bmi1a labeling at the midbrain-hindbrain boundary of eed −/− mutants suggesting that the number of stem cells is increased, possibly due to differentiation defects in this region. In contrast, sox2 expression seems less affected by the loss of eed function, with eventually a slight increase in sox2 RNA abundance in the hindbrain. The RNA expression profile of pcna appeared globally similar in eed +/+ and eed −/− larvae indicating that PRC2 loss of function has little or no effect on cell proliferation in the developing brain at 5 dpf. However, while expression of the cyclin gene ccnd1 is unchanged in eed −/− larvae, transcription of ccna2 appears in part slightly deregulated. We observed that the expression of mycn is downregulated in the hindbrain and the retina of the eed mutants. Finally, using probes for neuronal precursors (neurod1) and glial precursors (mag), whole-mount RNA in situ hybridizations showed a reduction in the labeled areas indicating that the numbers of these precursors are diminished in the hindbrain or midbrain of eed −/− larvae, respectively. Altogether, our results suggest that PRC2 loss of function disturbs the differentiation and the development of a subset of neurogenic cells in the midbrain-hindbrain region and in the retina without dramatically affecting cell proliferation, as judged by pcna expression. Thus, remarkably, while eed loss of function drastically affects bmi1a expression, the effect of the mutation on the expression of other markers such as ccna2, mycn, neurod1, or mag remains relatively thin and affects discrete areas of the brain. Cells 2021, 10, x FOR PEER REVIEW 14 of 24 Figure 7. Whole-mount RNA in situ hybridization of the brain region of eed +/+ and eed -/-siblings at 5 dpf: (a) whole-mount in situ hybridization to detect eed (left) and phox2bb RNA expression. The red arrow shows the phox2bb ectopic expression in the retina of eed mutants. Scale bar is 500 μm; (b) whole-mount in situ hybridization to detect the expression profiles of stemness (bmi1a, sox2), proliferation (pcna, mycn), cell cycle (ccna2, ccnd1), and neuronal/glial precursor (neurod1, mag) markers. The red arrows emphasize expression profile differences between eed +/+ and eed -/-larvae. Scale bar is 500 μm.
To gain insights into the possible role of PRC2 in brain development, we applied in situ hybridizations at 5 dpf on several genes known to be expressed in the midbrain-hindbrain boundary region that contains neural stem cells and dividing progenitors [72], and where eed transcripts are abundant (Figures 1a and 7a) [54]. The genes chosen for the study were the stemness markers bmi1a (bmi1 polycomb ring finger oncogene 1a) and sox2 (SRYbox transcription factor 2); the proliferation marker pcna (proliferating cell nuclear antigen); mycn (MYCN proto-oncogene, bHLH transcription factor), which is involved in several signaling pathways promoting cell growth, proliferation and metabolism of progenitor cells in different developing organs and tissues; two genes coding for cyclins, ccna2 (cyclin A2) and ccnd1 (cyclin D1); neurod1 (neuronal differentiation 1), a marker of neuronal precursors; and mag (myelin-associated glycoprotein), a marker of glial precursors ( Figure  7b).
Analysis of expression profiles of the stemness markers bmi1a and sox2 revealed a strong increase in bmi1a labeling at the midbrain-hindbrain boundary of eed -/-mutants Figure 7. Whole-mount RNA in situ hybridization of the brain region of eed +/+ and eed −/− siblings at 5 dpf: (a) whole-mount in situ hybridization to detect eed (left) and phox2bb RNA expression. The red arrow shows the phox2bb ectopic expression in the retina of eed mutants. Scale bar is 500 µm; (b) whole-mount in situ hybridization to detect the expression profiles of stemness (bmi1a, sox2), proliferation (pcna, mycn), cell cycle (ccna2, ccnd1), and neuronal/glial precursor (neurod1, mag) markers. The red arrows emphasize expression profile differences between eed +/+ and eed −/− larvae. Scale bar is 500 µm.

Loss of eed Function Alters Locomotor Activity
Loss of eed function alters the differentiation of neurogenic cells at 5 dpf. Then, to investigate whether eed deficiency could also affect larval behavior, we performed locomotor and thigmotaxis assays. Zebrafish larvae from heterozygous eed +/− incrosses were individually disposed of into dish wells. The plates 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 were achieved using the ZebraLab software (Version 5.18.0.0, ViewPoint Life Sciences, Lyon, France). After recording, the larvae were euthanized and their genotype determined by RFLP.
The locomotor activity assays were 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 ( Figure 8a). As previously described [73], switching to light dramatically decreases larval activity, whereas return to darkness is associated with an increase in locomotor activity. Additional cycles of alternating light and dark produce alternating levels of low and high activity of the zebrafish larvae, respectively (Figure 8b). Behaviorally, eed −/− larvae are hyperactive compared with wild-type larvae at 5 dpf. This hyperactive phenotype of eed −/− mutant is particularly prominent and significant during the dark cycles (Figure 8b,c). Behavioral comparison between wild-type and eed -/-mutants at 5 dpf: (a) locomotor tracking for a 70 min recording of an eed -/-mutant (left) and a wild-type (right) larvae; (b) distance traveled throughout a 70 min session for wild-type (red, n = 11) and eed -/-mutant (green, n = 16). Data are presented as mean ± SD of the distance moved (in mm) in 2 min intervals. Black and white bars at the bottom indicate dark and light conditions, respectively. Statistical analysis was performed using two-way ANOVA followed by Bonferroni posttest comparisons; ns, non-significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001; (c) cumulative distance traveled for each wild-type (red) and mutant (green) larvae during the light (left) and dark (right) periods. Statistical analysis was performed using one-way ANOVA followed by Bonferroni posttest comparisons; ns, non-significant; ***, p < 0.001; (d) the experimental procedure for the thigmotaxis assay is composed of a 6 min acclimatization period with the light ON followed by a 4 min visual motor challenge period with the lights OFF. Experiments are performed in a 24-well plate format. Inner and outer zones are delineated by the green circle in a way that the two zones cover equivalent spatial areas. The distance traveled during the dark challenge period is shown for two larvae; (e) cumulative distance traveled in the outer zone of wild-type (red, n = 10) and eed -/-mutant (green, n = 16) larvae. Each point represents individual larvae. No statistical (ns) difference was found between wild-type and eed -/-mutants; (f) cumulative time spent in the outer zone of wild-type (red, n = 10) and eed -/-mutant (green, n = 16) larvae. Each point represents unique larvae. No statistical (ns) difference was found between wild-type and eed -/-mutants.
Thigmotaxis (or "wall hugging") is a common behavioral endpoint used in preclinical studies employing rodents but also applied to the zebrafish model [74,75]. Animals engaged in thigmotaxic behavior strongly avoid the center of an arena and reduce their exploration behavior to stay or move in close proximity to the boundaries of the environment. Thigmotaxis assays were conducted in 24-well plates using a protocol composed of a 6 min acclimatization period in the light followed by a 4 min dark challenge period. The arena was divided into two areas, inner and outer, of equivalent surface, and thigmotaxis was presented as the percentage of the total distance moved or the total time spent by each larva into the outer zone (Figure 8d). Our measurements showed that there is no significant difference in thigmotaxis behavior between eed −/− mutants and wild-type larvae Figure 8. Behavioral comparison between wild-type and eed −/− mutants at 5 dpf: (a) locomotor tracking for a 70 min recording of an eed −/− mutant (left) and a wild-type (right) larvae; (b) distance traveled throughout a 70 min session for wild-type (red, n = 11) and eed −/− mutant (green, n = 16). Data are presented as mean ± SD of the distance moved (in mm) in 2 min intervals. Black and white bars at the bottom indicate dark and light conditions, respectively. Statistical analysis was performed using two-way ANOVA followed by Bonferroni posttest comparisons; ns, non-significant; *, p < 0.05; **, p < 0.01; ***, p < 0.001; (c) cumulative distance traveled for each wild-type (red) and mutant (green) larvae during the light (left) and dark (right) periods. Statistical analysis was performed using one-way ANOVA followed by Bonferroni posttest comparisons; ns, non-significant; ***, p < 0.001; (d) the experimental procedure for the thigmotaxis assay is composed of a 6 min acclimatization period with the light ON followed by a 4 min visual motor challenge period with the lights OFF. Experiments are performed in a 24-well plate format. Inner and outer zones are delineated by the green circle in a way that the two zones cover equivalent spatial areas. The distance traveled during the dark challenge period is shown for two larvae; (e) cumulative distance traveled in the outer zone of wild-type (red, n = 10) and eed −/− mutant (green, n = 16) larvae. Each point represents individual larvae. No statistical (ns) difference was found between wild-type and eed −/− mutants; (f) cumulative time spent in the outer zone of wild-type (red, n = 10) and eed −/− mutant (green, n = 16) larvae. Each point represents unique larvae. No statistical (ns) difference was found between wild-type and eed −/− mutants.
Thigmotaxis (or "wall hugging") is a common behavioral endpoint used in preclinical studies employing rodents but also applied to the zebrafish model [74,75]. Animals engaged in thigmotaxic behavior strongly avoid the center of an arena and reduce their exploration behavior to stay or move in close proximity to the boundaries of the environ-ment. Thigmotaxis assays were conducted in 24-well plates using a protocol composed of a 6 min acclimatization period in the light followed by a 4 min dark challenge period. The arena was divided into two areas, inner and outer, of equivalent surface, and thigmotaxis was presented as the percentage of the total distance moved or the total time spent by each larva into the outer zone (Figure 8d). Our measurements showed that there is no significant difference in thigmotaxis behavior between eed −/− mutants and wild-type larvae at 5 dpf (Figure 8e,f).
Thus, in spite of having a hyperactive phenotype, the eed −/− mutants do not present an increase in their exploration behavior when compared to wild-type siblings.

Discussion
PRC2 is a histone-modifying protein complex conserved from fruit flies to humans and able to deposit H3K27me1/2/3 epigenetic marks responsible in turn to local chromatin compaction and gene repression. Genome-wide studies using human or mouse embryonic stem cells showed that PRC2 and H3K27me3 marks are enriched at promoters of numerous developmental genes, leading to the proposal that PRC2 could be involved in the maintenance of the pluripotency of stem cells by keeping developmental genes repressed [12][13][14][15][16][17][18][19]. Differentiation is then associated with the relocation of the PRC2 complexes turning stem cell genes expression off and switching on gene programs specific to given developmental lineages. PRC2 is composed of the core subunit EZH1/2, EED, SUZ12, and RBBP4/7 together with a wide range of substoichiometric subunits defining distinct alternate protein complexes [22]. However, PRC2 catalytic activity relies only on the presence of the two mutually exclusive histone methyltransferases EZH1 or EZH2, associated with EED and SUZ12. In zebrafish, as the result of a genome duplication that occurred in the teleost lineage, the Suz12 subunit is encoded by two ohnologues, suz12a and suz12b. Then, eed is the sole zebrafish gene coding for a PRC2 subunit present in all alternate complexes. Consequently, since EED is required for EZH1/2 methyltransferase activity, it is expected that inactivation of the zebrafish eed gene will result in the loss of activity of all PRC2 complexes.
Using the TALEN technology, we generated an eed allele harboring a 14 bp deletion within the third exon. This deletion conducts to a frameshift within the coding sequence leading to the production of a putative protein lacking all the WD40 domains, predicted to be non-functional, and therefore likely producing an eed null allele. Heterozygous eed +/− fish are viable, fertile, and do not present visible gross phenotypes. By opposition, eed − homozygosity leads to larval death at around 12 dpf. This time point is similar to what was observed for zebrafish lacking the function of other PRC2 components, such as for ezh2 mutants [37,38] or suz12a; suz12b double-mutants [39], but contrasts to Eed knockout mouse phenotype [25]. While mouse embryos homozygous for the Eed mutation fail to gastrulate normally, eed embryonic expression is not required for the establishment of the zebrafish body plan. This relatively late lethality makes the zebrafish a compelling model to study PRC2 function during the development of various tissues without the requirement of conditional knockout strategies. It also reinforces the idea that PRC2 function plays little or no role in the implementation of developmental decisions but is mainly required to maintain these developmental instructions.
Zebrafish eed −/− mutants display various alterations of the digestive organs. These defects include a smaller pancreas at 5 dpf, hepatic steatosis characterized by an increase in lipids and macrovesicles in the liver, and the loss of maintenance of the intestine structures particularly marked at the level of the intestinal bulb. Studies in mice also pointed out the role of Eed in intestinal development [76]. Indeed, the conditional knockout of Eed in postnatal intestinal crypts decreases the stem cell pools, reduces the proliferation potential, increases the secretory cell differentiation, and results in smaller mouse intestines. Zebrafish larvae hatch with a yolk sac, which is depleted at 6-7 dpf. Then, after that time, their growth relies on external food uptake. If left unfed, the larvae will usually die of starvation at around 12 dpf [77]. Thus, it is very likely that the defects of the digestive organs found in the eed −/− homozygous mutants account for their larval death at 12 dpf. Furthermore, the marked eed −/− phenotype at the level of the digestive organs closely resembles the phenotype of the ezh2 −/− homozygous mutants we have previously described [38]. These indistinguishable phenotypes exclude critical roles of Ezh1 and/or PRC2-independent Ezh2 functions in the development of the digestive organs, at least until 12 dpf.
Investigations on bulk histone modifications in eed-deficient larvae at 9 dpf revealed a dramatic decrease in H3K27me2/3 marks, as expected. However, eed −/− larvae retained some H3K27me2 and H3K27me3 residual methylation marks. We assume that the remaining H3K27me2/3 signals were deposited by the maternal PRC2 activity and retained in slowly dividing or quiescent cells. Maternal PRC2 contribution was previously demonstrated by showing that maternal-zygotic MZezh2 −/− mutants display a much stronger phenotype than zygotic ezh2 −/− mutants, with small eyes, accumulation of blood near the yolk extension, stringy heart, heart edema, and absence of pectoral fins at 2 dpf [37]. EZH1 and EZH2 are both able to deposit H3K27me1, but the role of PRC2 in the control of H3K27me1 monomethylation is still a matter of debate [20,58,78,79]. Studies using Eed and Suz12 knockout mouse ESC lines, as well as knockdown experiments in mouse ESCs, showed that the inhibition of PRC2 activity is responsible for a global loss of H3K27me2 and H3K27me3 levels, but at best for a partial reduction in H3K27me1 marks [24,80,81]. Our analysis in zebrafish revealed that H3K27me1 global levels are not significantly changed in eed −/− larvae indicating that in the absence of PRC2 function, H3K27 monomethylation is achieved by other histone methylation routes. It has been shown that the histone lysine methyltransferases G9a (EHMT2) and Glp (EHMT1), in addition to being potent H3K9 methyltransferases, are able to monomethylate H3K27 both in vitro and in mouse ESCs [56,57,82]. Thus, their zebrafish orthologs, Ehmt1a, Ehmt1b, and Ehmt2, appear as suitable candidates to account for monomethylation at H3K27 targets in the absence of PRC2 activity.
Polycomb repression is achieved by the combined action of both PRC2 and PRC1. While PRC2 methylates H3K27, PRC1 monoubiquitylates H2AK119. Multiple PRC1 subcomplexes have been shown to coexist in cells. These complexes are commonly referred to as canonical (cPRC1) or non-canonical (ncPRC1), based on their dependence on H3K27me3 recruitment to target loci [6,83]. We have shown that global H2AK119ub levels were unchanged in eed −/− larvae indicating H2K119 monoubiquitylation is maintained in the absence of PRC2 function, presumably thanks to ncPRC1 action. This situation is similar to what was observed in ESCs, where Ezh1/Ezh2 double knockout was reported to do not affect H2K119ub levels [58]. H2K119ub maintenance in eed −/− mutants also explains the milder phenotype observed in eed-deficient fish when compared to rnf2 −/− mutants. In zebrafish, Rnf2 is the only catalytic subunit common to all PRC1 subcomplexes [84]. Zebrafish deficient in rnf2 gene function lack H2K119ub marks, display several defects including jaw malformation, pericardial edema, diminished blood circulation, absence of pectoral fins, severe craniofacial phenotype, and die at around 4-5 dpf [85,86]. However, our results contrast with another study using maternal-zygotic MZezh2 zebrafish mutants [87]. In this report, it was shown that loss of Ezh2 function from maternal and zygotic origins resulted in dramatic impairment of PRC1 recruitment at the chromatin and complete loss of H3K119ub marks. This observation highlighted the role of the PRC2 in the functional recruitment of cPRC1 complexes and led us to propose that in the absence of zygotic Eed function, maternal PRC2 activity is sufficient to deposit enough H3K27me3 marks allowing cPRC1 recruitment at early developmental stages. Then, at later stages, H2AK119ub marks are propagated and transmitted through cell generations in a PRC2-independent fashion via the action of ncPRC1 complexes. This hypothesis would explain why rnf2 −/− and MZezh2 −/− mutants display more severe phenotypes than ezh2 and eed homozygous zygotic mutants do. Furthermore, it has been demonstrated that ncPRC1 complexes are able to propagate H2AK119ub during cell division via their recruitment to the chromatin through the binding of the RYBP/YAF2 subunits to H2AK119ub marks [88].
Sequencing studies on pediatric high-grade gliomas, including glioblastoma multiformes (GBMs) and diffuse intrinsic pontine gliomas (DIPGs), identify recurrent somatic lysine to methionine substitutions at position 27 (K27M) in histone H3. The H3 K27M mutation primarily occurs in H3F3A (>70%), one of the two genes coding for the histone variant H3.3, with a lower frequency in HIST1H3B and HIST1H3C encoding H3.1, and more rarely in HIST2H3C coding for H3.2 [89][90][91][92]. The H3.3 K27M mutated histone binds to EZH2 and inhibits the methyltransferase activity of the PRC2, leading to a dramatic decrease in global H3K27me3 marks and alterations in gene expression programs [93][94][95][96][97]. These changes in gene expression would sustain the undifferentiated phenotype of the tumor cells that contributes to the aggressiveness of the gliomas and to the dismal associated prognosis [94]. In addition to H3F3A K27M mutations, DIPGs also contain genetic alterations targeting canonical signaling pathways, most frequently TP53 loss of function (about 60%) and gain-of-function mutations or amplifications in platelet-derived growth factor receptor alpha (PDGFRA; about 40%) [89,98], as well as numerous other mutations that might contribute to significant inter-and intratumoral DIPG heterogeneity and found at lower-frequencies [99]. However, the individual contribution of each of the different mutations, including H3F3A K27M , in glioma genesis remains poorly understood. In this context, we took advantage of our eed-deficient zebrafish line to investigate the effects of PRC2 loss of activity on neuronal differentiation and proliferation. We found that in eed −/− larvae, expression of some stemness (bmi1a), as well as neuronal (neurod1) or glial (mag) precursor markers, is altered, indicating that neural differentiation is impaired in the absence of functional PRC2 activity. Interestingly, a number of human DIPGs have been shown positive for BMI1 expression, which then appears as a potential therapeutic target [100][101][102]. By contrast, our in situ hybridization experiments revealed that the pcna expression profile remained unchanged in the brain of eed −/− larvae suggesting that PRC2 loss of function and reduction in H3K27me3 marks do not affect the proliferation potential of neural cells. Thus, we propose that in DIPGs, the H3F3A K27M mutation is responsible for the maintenance of an undifferentiated state, whereas other mutations might contribute to proliferation.
Surprisingly, we also showed that while expressed ubiquitously in the brain of wildtype larvae, eed loss of function results in gene expression alterations in relatively thin and discrete areas of the brain. For instance, ectopic expression of phox2bb is specifically found in the retina, whereas neurod1 expression is lost in a subset part of the hindbrain. Then, the effects of PRC2 loss of function appear strictly cell-specific and context-dependent, differentially affecting cell fates. These subtle changes in gene expression would indubitably escape to global transcriptomic analyses. Thus, our observations also outline the importance of performing whole-mount in situ hybridizations, but not only global transcriptomic analyses, for the characterization of zebrafish lines harboring mutations in polycomb group genes. In this context, it is worth emphasizing that the loss of eed function gives different phenotypes depending on the considered organ. It affects neuronal differentiation in specific brain regions, the development of the pancreas, the maintenance of the intestine wall structures, or the metabolism of lipids in the liver. This variation in effects and phenotypes might reflect the fact that Eed controls different gene expression programs in the different organs. It also reinforces the idea that a better deciphering of PRC2 function would rely on transcriptomic analyses at the single-cell level.
Cohen-Gibson syndrome (COGIS, OMIM #617561) is an overgrowth disorder characterized by dysmorphic facial features, advanced bone age, skeletal abnormalities, associated with variable intellectual disability, and caused by de novo mutations in EED [30,103]. To date, COGIS (EED-related overgrowth) has been reported in eight individuals, with at least one patient showing hyperactivity [104]. Interestingly, our behavioral studies revealed that eed −/− larvae also present a hyperactive phenotype, raising the possibility that the eed-deficient zebrafish line could constitute a pertinent model to study the neurological defects leading to behavioral disabilities in COGIS.