PRC1 sustains the integrity of neural fate in the absence of PRC2 function

Polycomb repressive complexes (PRCs) 1 and 2 maintain stable cellular memories of early fate decisions by establishing heritable patterns of gene repression. PRCs repress transcription through histone modifications and chromatin compaction, but their roles in neuronal subtype diversification are poorly defined. We found that PRC1 is essential for the specification of segmentally restricted spinal motor neuron (MN) subtypes, while PRC2 activity is dispensable to maintain MN positional identities during terminal differentiation. Mutation of the core PRC1 component Ring1 in mice leads to increased chromatin accessibility and ectopic expression of a broad variety of fates determinants, including Hox transcription factors, while neuronal class-specific features are maintained. Loss of MN subtype identities in Ring1 mutants is due to the suppression of Hox-dependent specification programs by derepressed Hox13 paralogs (Hoxa13, Hoxb13, Hoxc13, Hoxd13). These results indicate that PRC1 can function in the absence of de novo PRC2-dependent histone methylation to maintain chromatin topology and postmitotic neuronal fate.


Introduction
Accurate control of gene expression is essential for the specification and maintenance of neural fates during development. Studies of cell-type-restricted transcription factors have illuminated the mechanisms by which spatial and temporal regulation of gene expression gives rise to identifiable neuronal subtypes (Doe, 2017;Fishell and Kepecs, 2020;Hobert and Kratsios, 2019;Sagner and Briscoe, 2019;Venkatasubramanian and Mann, 2019). A parallel and critical mechanism of gene regulation is through the post-translational modification of histones, which enables and restricts transcription by modifying chromatin structure Schuettengruber et al., 2017). The Polycomb group (PcG) is key family of histone-associated proteins that play evolutionarily conserved roles in restricting gene expression during development (Blackledge et al., 2015;Gentile and Kmita, 2020;Simon and Kingston, 2009;Soshnikova and Duboule, 2009). In embryonic stem cells, cell fate determinants are repressed through PRC activities, and PRC-associated histone marks are subsequently removed from loci as cells differentiate (Boyer et al., 2006;Farcas et al., 2012;Tavares et al., 2012). PRC repression is maintained through cell division and after differentiation and is thought to contribute to stable cellular memories of early patterning events (Ciabrelli et al., 2017;Coleman and Struhl, 2017). In vertebrates, much of our knowledge of how PRCs regulate gene expression has emerged from biochemical studies of PcG proteins or from the activity of these factors in proliferating cells. Despite an in depth understanding of the mechanisms of PRC action, how PcG proteins interact with gene regulatory networks in the CNS remains poorly understood.
The specification of neuronal fates in the vertebrate spinal cord provides a tractable system to elucidate the function of PcG proteins, as the pathways that determine identities are well-defined, and the molecular signatures of many subtypes are known (Butler and Bronner, 2015;Sagner and Briscoe, 2019). One neuronal class where fate specification has been closely examined is the spinal MN. A core set of transcription factors, including Mnx1, Isl1/2, and Lhx3, determines class-specific features of MNs, including neurotransmitter identity (Shirasaki and Pfaff, 2002). The subsequent diversification of MNs into hundreds of muscle-specific subtypes is achieved through a conserved network of Hox transcription factors differentially expressed along the rostrocaudal axis (Philippidou and Dasen, 2013). During neural tube patterning, opposing gradients of retinoic acid (RA) and fibroblast growth factors (FGFs) provide spinal progenitors with a positional identity (Bel-Vialar et al., 2002;Dasen et al., 2003;Liu et al., 2001). These morphogens act, in part, by temporally and spatially depleting Polycomb-associated histone marks from Hox clusters (Mazzoni et al., 2013). As progenitors exit the cell cycle, MNs continue to express Hox genes, where they regulate repertoires of subtype-specific genes (Catela et al., 2016;Dasen et al., 2005;Mendelsohn et al., 2017). Although the role of Hox proteins in the CNS are well-characterized (Parker and Krumlauf, 2020), and are known targets of PRC activities , the specific contributions of PRC1 and PRC2 to CNS maturation remain unclear, as few studies have directly compared their functions during embryonic development.
Polycomb repression is initiated by PRC2, which methylates histone H3 at lysine-27 (H3K27me3), permitting recruitment of PRC1 through subunits that recognize this mark, leading to chromatin compaction at genes targeted for repression (Margueron and Reinberg, 2011;Schuettengruber et al., 2017). In embryonic stem (ES) cells, Hox gene clusters are initially covered by H3K27me3 and loss of PRC2 function leads to reduced PRC1 binding and ectopic Hox expression (Boyer et al., 2006). During mouse development, H3K27me3 marks are progressively removed from Hox clusters, allowing for the temporal and spatial activation of more caudal Hox genes during axis extension (Soshnikova and Duboule, 2009). The progressive removal of PRC2-associated histone marks is also recapitulated in ES cell-derived MNs (ESC-MNs), where RA functions to deplete H3K27me3 from rostral Hox1-Hox5 paralogs, while FGF removes H3K27me3 from more caudal Hox genes through Cdx proteins (Mazzoni et al., 2013). Although loss of PRC2 affects the viability of ESC-MNs, a hypomorphic mutation in the PRC2 component Suz12 leads to ectopic Hox expression (Mazzoni et al., 2013). Thus, during early phases of embryonic development, PRC2 appears to have a critical role in establishing the early profiles of Hox expression in MNs along the rostrocaudal axis.
PRC1 and PRC2 can also exist in a variety of configurations, which may contribute to neuronal subtype-specific activities. The core subunit of PRC1, Ring1, binds to one of six Polycomb group Ring finger (Pcgf) proteins (Gao et al., 2012). PRC1 containing Pcgf4 interacts with Cbx proteins (canonical PRC1) which recognize H3K27me3 (Bernstein et al., 2006;Morey et al., 2012). Variant forms of PRC1 containing Rybp can inhibit incorporation of Cbx proteins into PRC1, and bind target loci independent of H3K27me3 (Tavares et al., 2012;Wang et al., 2010). We previously found that a PRC1 component, Pcgf4 (Bmi1), is required to establish rostral boundaries of Hox expression in differentiating MNs (Golden and Dasen, 2012). By contrast, deletion of a core PRC2 component, Ezh2, from MN progenitors had no apparent impact on fate specification, possibly due to compensation by Ezh1 (Shen et al., 2008), or through use of variant PRC1 containing Rybp. These observations raise the question of what are the specific roles of canonical PRC1, variant PRC1, and PRC2 during MN subtype diversification.
To determine the function of PRCs during neural differentiation, we removed core components of these complexes from MN progenitors. We found that depletion of Ring1 proteins, essential constituents of all PRC1 complexes, causes pronounced changes in transcription factor expression and a loss of Hox-dependent MN subtypes. By contrast, neither PRC2 nor variant PRC1 activities are required to maintain rostrocaudal positional identities at the time of MN differentiation. Deletion of Ring1 leads to increased chromatin accessibility and derepression of a broad variety of cell fate determinants, while class-specific features of MNs are preserved. The derepression of caudal Hox genes in Ring1 mutants leads to the suppression of MN subtype diversification programs. These findings indicate that PRC1 function is essential during terminal differentiation to specify the transcriptional identities of motor neurons.

PRC1 is essential for rostrocaudal patterning during neuronal differentiation
To determine the relative contributions of PRC1 and PRC2 to neuronal specification, we analyzed mice in which core subunit-encoding genes were selectively removed from MN progenitors. Ezh genes encode the methyltransferase activity of PRC2, while Eed is required to enhance this function. We first generated mice in which both Ezh genes are conditionally deleted, by breeding Ezh1 and Ezh2 floxed lines to Olig2 Cre mice, which targets Cre to MN progenitors (Ezh MNΔ mice) (Hidalgo et al., 2012;Su et al., 2003). We confirmed MN-restricted loss of PRC2 activity by examining the pattern of H3K27me3, which was selectively depleted from progenitors and post-mitotic MNs by E11.5 ( Figure 1A and B, Figure 1-figure supplement 1A,B). Expression of Mnx1 and Slc18a3 (Vacht), two general markers of MN identity, were grossly unchanged in Ezh MNΔ mice ( Figure 1C, D and G). We next analyzed expression of Hox proteins in Ezh MNΔ mice and found that the MN columnar subtype determinants Hoxc6, Hoxc9, and Hoxc10, were all expressed in their normal domains ( Figure 1E and F, Figure 1 Removal of Ezh genes via Olig2 Cre depletes PRC2 function in MN progenitors, raising the possibility that PRC2 regulates Hox expression at earlier stages. We therefore analyzed mice in which the core PRC2 component Eed was deleted using Sox1 Cre (Eed NEΔ mice), which targets Cre to neuroectoderm (Takashima et al., 2007;Yaghmaeian Salmani et al., 2018). In Eed NEΔ mice, H3K27me3 was depleted from spinal progenitors and postmitotic neurons by E11.5, with some H3K27me3 present in the floor plate ( Figure 1-figure supplement 1G). In Eed NEΔ mice, the number of Mnx1 + cells, and pattern of Hoxc6, Hoxc9, and Hoxc10 in MNs were unaffected, similar to Ezh MNΔ mice (Figure 1-figure supplement 1F, H-J). These observations indicate that depletion of PRC2 function at the time of differentiation does not affect MN class specification or rostrocaudal positional identities.
We next investigated the function of PRC1, which is thought to repress gene expression in a PRC2dependent manner. We generated mice in which Rnf2 (also known as Ring1B) is conditionally deleted from MN progenitors using Olig2 Cre in a global Ring1 (also known as Ring1A) mutant background (Ring1 MNΔ mice) (Calés et al., 2008;del Mar Lorente et al., 2000). In Ring1 MNΔ mice, Rnf2 (Ring1B) protein was selectively removed from progenitors and post-mitotic MNs ( Figure 1H,I). The number of Mnx1 + MNs and pattern of Slc18a3 expression were similar to controls in Ring1 MNΔ mice (Figure 1J,K and N). By contrast, Hox proteins normally expressed by forelimb-innervating brachial MNs were not detected, while thoracic and lumbar Hox proteins were ectopically expressed in more rostral MNs ( Figure 1L, M and O-T, Figure 1-figure supplement 2C-E). Brachial-level Hox proteins (Hoxc4, Hoxa5, Hoxc6 and Hoxc8), were selectively depleted from MNs, while thoracic and lumbar Hox determinants, (Hoxc9, Hoxc10, and Hoxd10) were derepressed in brachial segments ( Figure 1Q-T). Brachial MNs also co-expressed Hoxc9 and Hoxc10, which are normally restricted to thoracic and lumbar segments, respectively ( Figure 1-figure supplement 2B). At thoracic levels, Hoxc9 expression was attenuated and Hoxd10 was ectopically expressed (Figure 1-figure supplement 2D). These results show that loss of Ring1 causes a derepression of caudal Hox genes, leading to co-expression of caudal Hox proteins in brachial MNs, without affecting pan-MN molecular features ( Figure 1O).

Canonical PRC1 regulates Hox expression in MNs
Ring1 can interact with multiple Pcgf proteins, raising the question of which PRC1 configuration contributes to MN patterning. PRC1 containing Pcgf4 is required to establish Hox boundaries in MNs (Golden and Dasen, 2012), but can exist in two alternative configurations, depending on mutually exclusive incorporation of Cbx (canonical PRC1) or Rybp (variant PRC1) (Gao et al., 2012;Tavares et al., 2012). We examined the function of PRC1 isoforms first by manipulating Rybp Figure 1U). We used chick in ovo neural tube electroporation to express mouse Rybp in postmitotic MNs using the Mnx1 promoter (Mnx1-Rybp). Expression of Rybp under Mnx1 led to ectopic Hoxc9 and Hoxc10 expression at brachial levels ( Figure 1V). If Rybp acts by displacing Cbx, then elevating Cbx levels should restore normal Hox expression. To test this, we co-electroporated Mnx1-Rybp and Mnx1-Cbx2 at equivalent plasmid concentration, and no longer observed ectopic Hoxc9 in brachial segments ( Figure 1X). These observations are consistent with canonical PRC1 containing Pcgf4, Cbx, and Ring1 restricting Hox expression in MNs. Rybp is expressed by MNs at the time of their differentiation (Figure 1figure supplement 2F), raising the possibility that Rybp (or its paralog Yaf2) also plays regulatory roles during development. We therefore selectively deleted Rybp from MNs using Olig2 Cre mice in a Yaf2 -/background (Rybp/Yaf2 MNΔ mice). Combined deletion of Rybp and Yaf2 did not affect MN generation, Hoxc6, Hoxc9, or Hoxc10 expression ( Figure 1-figure supplement 2F). These findings indicate that canonical PRC1 maintains appropriate Hox expression, while variant PRC1 and PRC2 do not contribute to rostrocaudal patterning at the time of MN differentiation.

PRC1 is required in MNs for subtype diversification
Hox transcription factors play central roles in establishing neuronal subtype identities through regulating expression of subtype-specific genes. To determine the consequences of altered Hox expression in Ring1 MNΔ mice, we analyzed the molecular profiles and peripheral innervation pattern of MNs.
A key Hox target in MNs is the transcription factor Foxp1, which is essential for the differentiation of limb-innervating lateral motor column (LMC) and thoracic preganglionic column (PGC) neurons (Dasen et al., 2008;Rousso et al., 2008). In Ring1 MNΔ mice, Foxp1 expression is lost from brachial and thoracic MNs, and markedly reduced in lumbar MNs (Figure 2A We next assessed the impact of Ring1 deletion on two MN columnar subtypes that are specified independent of Hox function, the hypaxial and median motor columns (HMC and MMC) (Dasen et al., 2008;Jung et al., 2014). In Ring1 MNΔ mice, MMC neurons (Lhx3 + Mnx1 + ) were generated at normal numbers in thoracic segments, while the number of HMC neurons (Isl1/2 + Mnx1 + Lhx3 -) was increased in thoracic and brachial segments ( Figure 2E, F and H, Figure 2-figure supplement 1E,F). The increase in HMC neurons in Ring1 mutants is likely due to a reversion of presumptive Hox-dependent subtypes to an HMC fate, similar to mice in which Hox function is disrupted (Dasen et al., 2008;Hanley et al., 2016). We also observed a small population of medial neurons that coexpressed Isl1/2, Mnx1, and Lhx3 ( Figure 2E, F and H). These cells likely represent undifferentiated MN precursors, but did not appear to express progenitor markers such as Olig2 and Nkx6.1 (Figure 1-figure supple ment 2A). Mutation in Ring1 therefore depletes Hox-dependent subtypes, with the remaining MNs having a more ancestral axial or ambiguous subtype identity ( Figure 2P).
As we observed a dramatic loss of segmentally restricted MN subtypes in Ring1 MNΔ mice, we next assessed the impact on peripheral innervation pattern. To trace motor axon projections, we expression of Rybp and Cbx under Mnx1 fails to induce Hoxc9 in brachial MNs. Panels B-F show brachial sections from E11.5 embryos; I-M, P-T brachial sections from E12.5 embryos; V,X brachial sections of HH st25 chick.
The online version of this article includes the following source data and figure supplement(s) for figure 1: Source data 1. Counts of Mnx1-positive MNs in Ezh and Ring1 mutants.     Catela et al., 2016). In Ring1 MNΔ mice, only three nerve branches, radial, median, and ulnar were visible but appeared prematurely truncated and unbranched ( Figure 2K). In the trunk, innervation of sympathetic chain ganglia was lost, consistent with a loss of PGC fates ( Figure 2L and M). Projections to dorsal and ventral axial muscles by MMC and HMC subtypes were maintained in Ring1 MNΔ mice, and HMC axons were thicker and shorter than in controls ( Figure 2N and O, Figure 2-figure supplement 1G). Loss of Ring1 therefore causes severe defects in innervation pattern, while the trajectories of axial MNs are relatively spared.

Loss of Ring1 causes derepression of developmental fate determinants
Polycomb proteins regulate diverse aspects of differentiation by restricting gene expression during development. To investigate changes in gene expression after deletion of Ring1 genes in an unbiased manner, we performed RNAseq on MNs isolated from Ring1 MNΔ mice. We purified MNs from control and Ring1 MNΔ ::Mnx1-GFP embryos at E12.5 by flow cytometry (n = 4 Ring1 mutants, and n = 4 Crecontrols), and performed RNAseq. Because Ring1 mutants display segment-specific phenotypes, we collected MNs from brachial, thoracic, and lumbar levels and profiled each population independently. We identified a total of 1001 upregulated and 641 downregulated genes in Ring1 MNΔ mice (log 2 -FC > 2, FDR < 0.1), with 391 genes upregulated in all three segmental regions, compared to 46 genes that were commonly downregulated ( Figure 3A-C, Figure 3-figure supplement 1A,C, Supplementary file 1). Thus, many upregulated genes are shared between each segment, while downregulated genes tended to be segment-specific. Strikingly, 76% of the top 100 derepressed genes (by log 2 -FC) at brachial levels encode transcription factors, and include genes normally involved in the specification of spinal interneurons (e.g. Gata3, Shox2, Lhx2), brain regions (Foxg1, Six6, Lhx8), and nonneuronal lineages ( Figure 3B).
To validate these changes in gene expression, we performed mRNA in situ hybridization on a subset of upregulated or downregulated genes. We found that Gata3, Lhx8, and Lhx2 were markedly upregulated in MNs of Ring1 MNΔ mice ( Figure 4A-F), while other genes (Foxg1, Six6, Pitx2, Shox2, Nkx2.1) showed less prominent, but detectable, ectopic expression (Figure 4-figure supplement 1A-J). We also examined expression of previously uncharacterized genes that were downregulated in all three segmental regions. We found Rgs4, Uts2b, Vat1, and Gabra2 were expressed by MNs of control embryos, and were downregulated in Ring1 mutants ( Figure 4G-L, Figure 4-figure supplement 1K,L). These results indicate that despite the derepression of multiple fate determinants in Ring1 MNΔ mice, core features of MN identity are preserved, but that subtype diversification programs are selectively disrupted.        To further validate these findings, we analyzed Hoxc13 and Hoxb13 expression by in situ hybridization. In controls Hoxc13 is restricted to sacral segments, while Hoxb13 is not detected in MNs ( Figure 5C). In Ring1 MNΔ mice, both Hoxb13 and Hoxc13, were de-repressed in MNs throughout the rostrocaudal axis ( Figure 5C). In addition, in situ hybridization of Hoxc6 and Hoxc9 expression revealed reduced expression at brachial and thoracic levels, respectively (Figure 5-figure supplement 1D). By contrast, ectopic expression of Hoxb13 and Hoxc13 was not observed in Ezh MNΔ mice ( Figure 5-figure supplement 2A,B). Thus, ectopic expression of Hox13 paralogs in Ring1 mutants is associated with a loss of rostral Hox gene expression.

Ring1 is essential to maintain MN chromatin topology
Our findings indicate that in the absence of Ring1 genes, a broad variety of cell fate determinants are ectopically expressed in MNs, while only a subset of caudal Hox genes are derepressed. As PRC1 restricts gene expression through chromatin compaction, we investigated whether removal of Ring1 leads to changes in DNA accessibility at derepressed loci. We used Assay for Transposase-Accessible Chromatin with high-throughput sequencing (ATACseq) to identify genomic regions which have gained or lost accessibility in Ring1 MNΔ mice. We purified Mnx1-GFP MNs at E12.5 from control and Ring1 MNΔ embryos at brachial, thoracic, and lumbar levels, and performed ATACseq. In control samples, we observed a progressive opening of caudal Hox genes in more rostral segments. For example, the accessibility of Hoxc9 and Hoxa9 increases from brachial to thoracic segments, while Hoxc10, Hoxc11, and Hoxa10 are more accessible in lumbar segments ( Figure 6-figure supplement 1A, B).
To determine which genomic regions gained accessibility in Ring1 mutants, we compared ATACseq profiles between MNs of control and Ring1 MNΔ mice. We identified a total of 2305 loci that gained accessibility in Ring1 MNΔ embryos, 324 (14%) of which were common to all three segmental levels ( Figure 6A, Supplementary file 2). Common genes included caudal Hox genes (Hoxa13, Hoxc13) and other transcription factors (e.g. Foxg1, Lhx2, Pitx2) that were derepressed in our RNAseq analyses ( Figure 6C). We also identified 1264 loci that lost accessibility in Ring1 MNΔ mice, 39 (3%) of which were common to all three segments ( Figure 6A). Thus, similar to our RNAseq results, loss of Ring1 leads to increased accessibility in many genes that are shared among each segment, while genes that lose accessibility tend to be segment-specific.
We next examined the overlap between transcripts that were upregulated and loci that gained accessibility in Ring1 MNΔ mice. We found that 18% (148/813) of genes that were ectopically expressed at brachial segments also gained accessibility in Ring1 MNΔ mice ( Figure 6B). We found 19 genes, including Hoxa13 and Hoxc13, were derepressed and gained accessibly in all three segments of Slc18a3 (Vacht) is reduced (padj. = 0.026477) in Ring1 MNΔ mice. (E) DESeq2 counts of genes associated with specific MN subtype identities were reduced in Ring1 MNΔ mice. Counts for LMC and PMC markers are from Br segments, PGC from Th segments. Black bars shown in graphs indicate mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, unpaired t-test.
The online version of this article includes the following figure supplement(s) for figure 3:    ( Figure 6-figure supplement 1C). Since lumbar segments still retain features of Hox-dependent subtypes, we also compared the overlap between genes that were upregulated and gained accessibility in brachial and thoracic MNs. We identified 73 genes, 56 (77%) of which encode transcription factors, including each of the caudal Hox genes we found by RNAseq ( Figure 6B, Figure      . The gain of accessibility at transcription factor-encoding genes was prominent near transcription start sites ( Figure 6C), and regions that gained accessibility in Ring1 MNΔ neurons correspond to regions shown to be bound by Rnf2 (Ring1B) in ES cells (Figure 6-figure supplement 1D; Bonev et al., 2017).
In agreement with our RNAseq analysis, Hox13 paralogs showed pronounced increases in accessibility, with each of the four Hox13 paralogs gaining accessibility in each of the three segmental levels ( Figure 6D and E). By contrast, rostral Hox4-Hox8 paralogs, which are transcriptionally downregulated in Ring1 mutants, were not among that targets that lost accessibility ( Figure 6D). These results suggest that reduced expression of rostral Hox4-Hox8 paralogs in Ring1 mutants is not due to a loss in chromatin accessibility.

Hox13 paralogs repress rostral Hox genes by engaging accessible chromatin domains
As PRC1-mediated chromatin compaction is essential for Hox repression, it is surprising that Hox4-Hox8 paralogs are diminished in Ring1 mutants, without a noticeable reduction in chromatin accessibility at these loci. Because cross-repressive interactions between Hox genes themselves are an important regulatory mechanism determining Hox boundaries (Philippidou and Dasen, 2013), it is possible that ectopically expressed Hox13 paralogs directly repress multiple Hox genes in Ring1 MNΔ mice. To test this, we first analyzed MNs derived from ES cells (ESC-MNs) in which Hoxc13 expression can be induced upon doxycycline (Dox) treatment (Bulajić et al., 2020). ESC-MNs differentiated via RA and Sonic Hedgehog agonist are characterized by expression of anterior Hox paralogs. RNAseq analyses revealed that Hoxa4, Hoxa5, Hoxc4, and Hoxc5 expression were markedly reduced after Dox-induced Hoxc13 expression in MN progenitors, compared to non-induced ESC-MNs ( Figure 7A, Figure 7-figure supplement 1A-C). This repressive effect also appears to be direct, as ChIPseq analysis indicates that Hoxc13 can bind at multiple Hox genes ( Figure 7A).
Hoxc13 and Hoxa13 can target inaccessible chromatin (Bulajić et al., 2020;Desanlis et al., 2020), while repression by the Drosophila Hox protein Ubx is associated with chromatin compaction (Loker et al., 2021). To examine the possible effects of caudal Hox proteins on local chromatin structure, we compared accessibility at Hox loci in both control and Ring1 mutants, relative to the location of Hoxc13 binding sites. At Hoxc4, Hoxc5, Hoxa4, and Hoxa5, Hoxc13 bound sites correspond to regions that are accessible in both control and Ring1 mutant MNs ( Figure 7A, Figure 7-figure supplement 1A-C). Hox13 proteins therefore appear to repress rostral Hox genes, in part, through binding at pre-existing accessible regions.
To test whether Hox13 paralogs can repress multiple Hox paralogs in vivo, we used chick neural tube electroporation to express Hoxa13 and Hoxb13 within brachial, thoracic, and lumbar segments, and assessed Hox protein expression. We electroporated pCAGGs-Hoxa13-ires-nucGFP or pCAGGs-Hoxb13-ires-nucGFP and found that both cell-autonomously repressed expression of brachial, thoracic, and lumbar Hox proteins (Hoxc4, Hoxa5, Hoxc6, Hoxc8, Hoxc9, and Hoxc10) ( Figure 7B-D). By contrast, misexpression of Hoxa13 and Hoxb13 did not affect expression of the general MN markers Isl1/2 and Mnx1 ( Figure 7B-D). These results indicate that Hox13 paralogs can repress multiple Hox genes, and likely contribute to the MN fate specification defects of Ring1 MNΔ mice.

Discussion
The Polycomb group encompass a large and diverse family of proteins essential for maintaining epigenetic memory of early patterning events. Classically, PcG-mediated repression is thought to depend on recruitment of PRC1 through recognition of histone methylation marks deposited by PRC2 activity. Although alternative, H3K27me3-independent, mechanisms of PRC1 repression have been described (Tavares et al., 2012), the relative contribution of PRC1 and PRC2 to neuronal fate specification have not been resolved. In this study, we found that genetic depletion of PRC2 components     has no discernable impact on neural class or subtype diversification programs at the time of MN differentiation. By contrast, a core PRC1 subunit, Ring1, is required to restrict transcription factor expression in the CNS, maintain neuronal subtype-specific chromatin topology, and determine rostrocaudal boundaries of Hox expression. Our findings indicate PRC1 plays a key role in regulating expression of cell fate determinants during MN differentiation, safeguarding neurons from acquiring inappropriate gene regulatory programs.

PRC functions in Neural development
Polycomb proteins function in diverse aspects of CNS maturation, including the temporal transition from neurogenesis to gliogenesis, maintenance of adult neural stem cell fates, and restriction of gene expression in mature neurons (Di Meglio et al., 2013;Fasano et al., 2009;Hirabayashi et al., 2009;Molofsky et al., 2003;von Schimmelmann et al., 2016). The role of PRCs in neural subtype diversification has been challenging to study, in part, due the complexity and dynamic temporal regulation of its constituents. By selectively removing core PRC1 and PRC2 subunits from neural progenitors, we determined the relative contributions of these chromatin-associated complexes to neuronal diversification. We found that PRC1 represses a broad variety of cell fate determinants, with the majority of highly derepressed genes in Ring1 mutants encoding transcription factors. Ectopic expression of Ring1-regulated targets is associated with a disruption in chromatin topology, resulting in a gain in accessibility of derepressed genes.
Despite the derepression of multiple fate determinants in the absence of PRC1 function, markers of MN class identity (e.g. Mnx1, Isl1/2, and Lhx3) are largely unchanged, and MNs do not appear to acquire features of alternate neuronal class fates, beyond expression of fate-defining transcription factors. This observation suggests that once a class identity has been established, ectopic expression of fate determinants is insufficient to respecify basic neuronal features, such as neurotransmitter identity. In C. elegans the ability of transcription factors to reprogram neuronal class identity requires removal of histone deacetylase proteins (Tursun et al., 2011), suggesting that additional regulatory constraints, such as histone acetylation and/or DNA-methylation, restricts the ability of ectopic transcription factors to interfere with class-specific programs in Ring1 mutants.
While basic elements of MN fate are maintained in the absence of PRC1, their differentiation into molecularly distinct subtypes is severely compromised. The loss of subtype-specific programs in brachial and thoracic segments can be attributed to the attenuation of expression of rostral HoxA and HoxC cluster paralogs, as mice lacking these Hox genes show a similar loss in MN subtype features (Jung et al., 2010;Jung et al., 2014;Philippidou et al., 2012). Selective loss of Hox function in MNs also appears to account for the observation that the majority of downregulated genes in Ring1 mutants are segment-specific.
Our results indicate that in Ring1 mutants brachial MNs ectopically express Hox9-Hox13 paralogs, while expression of Hox genes normally expressed by brachial MNs is diminished. As a consequence, MNs fail to acquire a limb-innervating LMC molecular fate, and express markers indicative of MMC and HMC subtypes. These observations are consistent with a model in which the Hox code of MNs is scrambled in Ring1 mutants, reverting to the fate of ancestral axial subtypes. However, we cannot formally rule out the possibility that some MNs acquire a sacral Hox13 + fate, although we did not observe ectopic expression of markers of sacral PGC identity (e.g. nNos, Foxp1).
We previously found that acute depletion of the canonical PRC1 component Pcgf4 (Bmi1) in chick leads to ectopic Hoxc9 expression in brachial segments and a switch of LMC neurons to a thoracic PGC fate (Golden and Dasen, 2012). By contrast in Ring1 mutants, multiple caudal Hox genes (Hox9-Hox13 paralogs) are derepressed, and the remaining MNs have an axial subtype fate. While the exact mechanisms of these difference are unclear, they could reflect differences in the timing and spatial extent of the manipulations. In Ring1 mutants PRC1 function is depleted from MN progenitors throughout the rostrocaudal axis, whereas Bmi1 knockdown was performed in brachial segments just prior to differentiation.

PRC-dependent and independent mechanisms of rostrocaudal patterning in the CNS
Our results show that mutation in Ring1 genes leads to ectopic expression and increased accessibility of many transcription factor-encoding genes that are common to each segment, while genes that are downregulated and lose accessibility tend to be segment-specific. These observations indicate that PRC1 is required to restrict expression of multiple cell-fate determinants throughout the rostrocaudal axis. By contrast, the downregulation of segment-specific genes can be attributed to the loss of Hoxdependent MN columnar subtypes, due to altered Hox gene expression in PRC1 mutants.
We found that removal of PRC1 leads to the derepression and increased accessibility of caudal Hox genes. Loss of PRC1 does not result in a sustained derepression of all Hox genes, as Hox genes normally expressed by brachial-and thoracic-level MNs are attenuated in the absence of Ring1. These findings are reminiscent of the function of PRCs in Drosophila, where loss of PRC function leads to ectopic expression of the caudal Hox gene Abd-b and repression of Ubx expression (Struhl and Akam, 1985;Struhl and White, 1985), likely reflecting cross-repressive interactions between caudal Hox proteins and rostral Hox genes. Notably, reduced rostral Hox gene expression in Ring1 mutants is not a reflection of reduced chromatin accessibility at these targets, consistent with a model in which this form of repression is not associated with global changes in chromatin structure.
In Ring1 mutants, Hox13 paralogs are derepressed in MNs, and appear to contribute to the loss of MN subtypes. Derepression of Hox13 paralogs alone does not appear to account for all of the specification defects in Ring1 mutants, as Hox10 paralogs (Hoxa10, Hoxc10, and Hoxd10) were expressed normally in lumbar segments. Hoxc9 can also repress Hox4-Hox8 paralogs in MNs through direct binding (Jung et al., 2010), and likely contributes to the Ring1 mutant phenotype. Derepression of multiple caudal Hox paralogs therefore appears to have a cumulative effect on repressing brachial and thoracic Hox genes. In conjunction with studies in Drosophila, these finding indicate a conserved mechanism of rostrocaudal patterning, in which PRCs establish rostral Hox boundaries, while Hox cross-repressive interactions define caudal boundaries.

PRC1 functions in maintaining neuronal subtype fate
Studies of neuronal and non-neuronal development provide compelling evidence that PRC2 activity is required to restrict gene expression in early development (Ezhkova et al., 2009;Gentile et al., 2019;Snitow et al., 2015;Yaghmaeian Salmani et al., 2018). Our studies show that Hox boundaries are unaffected after genetic removal of PRC2 function, which leads to diminished H3K27me3 by the time of MN differentiation. Mutation of Ezh genes or Eed depletes H3K27me3 from progenitors and postmitotic neurons by E11.5, without appreciably affecting MN generation, Hox expression, or downstream Hox effectors. As the axial identity of progenitor cells appears to be shaped prior to neurogenesis (Metzis et al., 2018), our PRC2 manipulations likely remove H3K27me3 after Polycomb repression has been initiated. Consistent with this idea, our analyses of Ezh mutants as well as previous analyses of Eed NEΔ mice (Yaghmaeian Salmani et al., 2018), showed detectable levels of H3K27me3 in the MN domain up until E10.5, indicating that some H3K27me3 remains during the phase in which anterior Hox boundaries are established.
How does PRC1 restrict expression of fate determinants in the absence of PRC2 activity? One plausible scenario is that residual H3K27me3 is transmitted through cell division in PRC2 mutants and is sufficient to recruit canonical PRC1 and maintain target gene repression. As cells divide, newly synthesized histones are presumably devoid of H3K27me3 in PRC2 mutants, leading to replication-coupled dilution of H3K27me3. During neural differentiation, the rate of cell division decreases (Kicheva et al., 2014;Wilcock et al., 2007), potentially limiting H3K27me3 depletion, and enabling PRC1 to bind at target loci, even in the absence of de novo H3K27 methylation. Consistent with this idea, recent studies on the effects of PRC2 depletion in intestinal stem cells indicate that ~40% of residual H3K27me3 can maintain PRC repression (Jadhav et al., 2020), although the extent of derepression varies with gene target. Although we observed no MN phenotypes in Ezh or Eed mutants, later-born oligodendrocytes derived from the MN progenitor domain have been shown to depend on PRC2 function for their differentiation . This more pronounced effect on glial development likely reflects further dilution of H3K27me3 through cell division in PRC2 mutants.
Stabilization of PRC1 at repressed loci could also maintain target repression independently of de novo H3K27 methylation. Repression by canonical PRC1 has been shown to depend on the formation of phase-separated subnuclear structures (Polycomb bodies) assembled through polymerization of Polyhomeotic-like (Phc) and/or Cbx proteins (Isono et al., 2013;Plys et al., 2019;Tsuboi et al., 2018). Mutation of Phc2 in mice leads to ectopic expression of Hox genes, with Hoxb13 among the most robustly derepressed targets (Isono et al., 2013). One possibility is that PcG-mediated repression may not exclusively depend on anchoring of PRC1 through H3K27me3, but is also maintained through Phc-mediated chromatin compaction at specific loci. Alternatively, PRC2-independent repression by PRC1 could be facilitated through its H2A-ubiquitination activity, as has been suggested in other systems (Tsuboi et al., 2018).
We suggest that in the early phases of development PRC2 activity defines the sites of maintained H3K27 tri-methylation, initiates PRC1 recruitment, and restricts transcription factor expression in MNs. These activities are likely initiated prior to MN progenitor specification, as recent studies have demonstrated that the establishment of rostrocaudal positional identities can be specified before neural induction (Metzis et al., 2018). Patterning morphogens, such as RA and FGF, may act on stem cells to establish the pattern of H3K27me3 at Hox loci prior to neurogenesis. This early rostrocaudal patterning step may ultimately serve to coordinate Hox expression in the neural tube and surrounding mesodermally-derived tissues. The subsequent switch to reliance on PRC1 function could reflect a general mechanism of gene regulation in neuronal subtypes that become terminally differentiated during development.  (Hidalgo et al., 2012), Ezh2 flox/flox (Su et al., 2003), Olig2 Cre (Dessaud et al., 2007), Mnx1 GFP (Arber et al., 1999) mice have been previously described. Rybp flox/flox ::Yaf2 -/- (Hisada et al., 2012;Rose et al., 2016) mice were generated by microinjection of the mouse ESCs with these alleles into blastocysts followed by implantation into pseudopregnant female mice. Generation of Ezh MNΔ mice was performed by crossing Ezh1 flox/flox , Ezh2 flox/flox and Olig2 Cre mice. Ezh MNΔ mice are viable at birth but do not survive beyond P20. Generation of Ring1 MNΔ mice was performed by crossing Ring1 -

Slide immunohistochemistry
Embryos were fixed in 4% PFA for 1.5-2 hr at 4 °C, washed 5-6 times in cold PBS for 15-30 minutes each wash, and incubated overnight in 30% sucrose. Tissue was embedded in OCT, frozen in dry ice, and sectioned at 16 µm on a cryostat. For antibody staining of sections, slides of cryosections were placed in PBS for 5 min to remove OCT. Sections were then transferred to humidified trays and blocked for 20-30 min in 0.75 ml/slide of PBT (PBS, 0.1% Triton) containing 1% Bovine serum albumin (BSA). The blocking solution was replaced with primary staining solution containing antibodies diluted in PBT with 0.1% BSA. Primary antibody staining was performed overnight at 4 °C. Slides were then washed three times for 5 min each in PBT. Fluorophore-conjugated secondary antibodies were diluted 1:500-1:1000 in PBT and filtered through a 0.2 µm syringe filter. Secondary antibody solution was added to slides (0.75 ml/slide) and incubated at room temperature for 1 hr. Slides were washed three times in PBT, followed by a final wash in PBS. Coverslips were placed on slides using 110 µl of Vectashield (Vector Laboratories). Antibodies against Hox proteins and MN subtypes were generated as described (Dasen et al., 2008;Dasen et al., 2005). Additional antibodies were used as follows: goat anti-Rnf2 (Abcam, 1:2000), rabbit anti-Rnf2 (Abcam, 1:5000), rabbit anti-Rybp (Abcam1:2000), rabbit anti-Yaf2 (Abcam, 1:2000), rabbit anti-Cbx2 (Bethyl, 1:5000), rabbit anti-H3K27me3 (Cell Signaling, 1:2000).

In situ mRNA hybridization
Probe templates were generated by RT-PCR and incorporated a T7 promoter sequence in the antisense strand. Total RNA was first extracted from eviscerated E12.5 embryos using TRIzol (Invitrogen). Genes of interest were amplified with the One Taq One-Step RT-PCR kit (NEB) using 1 µg of RNA. After amplification by RT-PCR, a second PCR was performed to incorporate a T7 promoter sequence. Antisense riboprobes were generated using the Digoxigenin-dUTP (SP6/T7) labeling kit (Sigma-Aldrich). For in situ hybridization, sections were first dried for 10-15 min at room temperature, placed in 4% PFA, and fixed for 10 min at room temperature. Slides were then washed three times for 3 min each in PBS, and then placed in Proteinase K solution (1 µg/ml) for 5 min at room temperature. After an additional PFA fixation and washing step, slides were treated in triethanolamine for 10 min, to block positive charges in tissue. Slides were then washed three times in PBS and blocked for 2-3 hr in hybridization solution (50% formamide, 5 X SSC, 5 X Denhardt's solution, 0.2 mg/ml yeast RNA, 0.1 mg/ml salmon sperm DNA). Prehybridization solution was removed, and replaced with 100 µl of hybridization solution containing 100 ng of DIG-labeled antisense probe. Slides were then incubated overnight (12-16 hr) at 72 °C. After hybridization, slides were transferred to a container with 400 ml of 5 X SSC and incubated at 72 °C for 20 min. During this step, coverslips were removed using forceps. Slides were then washed in 400 ml of 0.2 X SSC for 1 hr at 72 °C. Slides were transferred to buffer B1 (0.1 M Tris pH 7.5, 150 mM NaCl) and incubated for 5 min at room temperature. Slides were then transferred to staining trays and blocked in 0.75 ml/ slide of B1 containing 10% heat inactivated goat serum. The blocking solution was removed and replaced with antibody solution containing 1% heat inactivated goat serum and a 1:5000 dilution of anti-DIG-AP antibody (Roche). Slides were then incubated overnight at 4 °C in a humidified chamber. The following day, slides were washed three times, 5 min each, with 0.75 ml/slide of buffer B1. Slides were then transferred to buffer B3 (0.1 M Tris pH 9.5, 100 mM NaCl, 50 mM MgCl 2 ) and incubated for 5 min. Slides were then developed in 0.75 ml/slide of B3 solution containing 3.5 µl/ ml BCIP and 3.5 µl/ml NBT for 12-48 hr. After color development, slides were washed in ddH 2 0 and coverslipped in Glycergel (Agilent). A more detailed in situ hybridization protocol is available on our lab website (http://www.dasenlab.com).

Wholemount immunohistochemistry
For wholemount immunohistochemistry embryos were fixed in PFA for 2 hr, then bleached for 24 hr at 4 °C in a 10% H 2 O 2 , 10% DMSO solution prepared in methanol. Embryos were washed three times for 10 min each in methanol, followed by five washes for 10 min in PBS. Primary antibodies were diluted in staining solution (5% BSA, 20% DMSO in PBS) and specimens were incubated in staining solution on a rotator overnight at room temperature. Samples were then washed three times for 5 min each in PBS, followed by four 1 hr washes in PBS. Specimens were then incubated in secondary antibodies diluted in staining solution overnight at room temperature. Samples were then washed three times for 5 min each in PBS, followed by four 1 hr washes in PBS, a single 10 min wash in 50% methanol, and three 20 min washes in 100% methanol. Samples were transferred to glass depression slides and tissue was cleared by incubating samples in BABB solution (1-part benzyl alcohol: 2-parts benzyl benzoate). Confocal images of embryos were obtained from Z-stacks using Zen software (Zeiss). Further details of wholemount staining protocols are available on our lab website: (http://www.dasenlab.com).

In ovo chick electroporation
In ovo electroporation were performed on Hamburger Hamilton (HH) stage 13-14 chick embryos and analyzed at HH stage 24-25. Fertilized chicken eggs (Charles River) were incubated in a humidified incubator at 39 °C for 40-48 hr until they reached HH13-14. The top of the egg shell was removed and a 1 µg/µl DNA (150-500 ng/µL expression plasmid and pBKS carrier DNA) containing ~0.02% Fast green was injected into the central canal of the neural tube using a sharpened glass capillary tube. Electrodes (Platinum/Iridium (80%/20%), 250 μm diameter, UEPMGBVNXNND, FHC Inc) were placed on both sides of the neural tube (4 mm separation) and DNA was electroporated using an ECM 830 electroporator (ECM 830, BTX; 25 V, 4 pulses, 50ms duration, 1 s interval). Eggs were sealed with parafilm and incubated for 48 hr prior to fixation. Results shown in figures are representative of at least three electroporated embryos from two or more experiments in which electroporation efficiency in MNs was above 60%.

RNA preparation and library preparation
RNA was extracted from FACS purified MNs dissected from E12.5 mouse embryo (10-20,000 cells/ segment), using the Arcturus Picopure RNA isolation kit. For on-column DNase treatment, Turbo DNase was used (ambion, AM2238). Each samples used separated bar code for libraries. RNA quality and quantity were measured with an Agilent Picochip using a Bioanalyzer, all samples had quality scores between 9-10 RIN. For library preparation 10 ng of total RNA was used to generate cDNA, which was amplified with SMARTer Stranded RNA-Seq kit. 100 ng of cDNA were used as input to prepare the libraries (Takara, #634837), and amplified by 10 PCR cycles. Samples were run in four 50-nucleotide paired-end read rapid run flow cell lanes with the Illumina HiSeq 4000 sequencer.

ATACseq
ATACseq was performed following previously described protocols (Buenrostro et al., 2015). DNA was extracted from purified dissected mouse embryonic MNs. Cells were aliquoted and washed twice in ice-cold 1× PBS. Cell pellets were resuspended in 10 mM Tris (pH 7.4), 10 mM NaCl, 3 mM MgCl2, 0.1% NP-40 (v/v), 0.1% tween20, 0.01% Digitonin and 1% BSA, centrifuged at 500 g for 5 min at 4 °C. Pellets were resuspended in 12.5 μl of 2× tagmentation DNA buffer, 1.25 μl Tn5 (Nextera DNA Sample Preparation Kit, FC-121-1030) and 11.25 μl of water, and incubated at 37 °C for 30 min. The sample was purified using the MinElute PCR Purification Kit (Qiagen, 28004). PCR enrichment of the library was performed with custom-designed primers and 2× NEB Master Mix. A qPCR reaction with 1× SYBR Green (Invitrogen), custom-designed primers and 2× NEB Master Mix (New England Labs, M0541) was performed to determine the optimal number of PCR cycles (one third of the maximum measured fluorescence) (Buenrostro et al., 2013). The libraries were purified using the AMPure XP beads (Beckman Coulter, A63880). High Sensitivity DNA ScreenTape (Agilent, 5067-5584) was used to verify the fragment length distribution of the library. Library quantification was performed using the KAPA Library Amplification kit on a Roche LightCycler 480. The libraries were sequenced on an Illumina NovaSeq (100 cycles, paired-end).

Statistics
Samples sizes were determined based on previous experience and the number of animals and definitions of N are indicated in the main text and figure legends. In figures where a single representative image is shown, results are representative of at least two independent experiments, unless otherwise noted. No power analysis was employed, but sample sizes are comparable to those typically used in the field. Data collection and analysis were not blind. Graphs of quantitative data are plotted as means with standard error of mean (SEM) as error bars, using Prism 8 (Graphpad) software. Unless noted otherwise, significance was determined using unpaired t-test in Prism eight software, or using adjusted p-values. Exact p-values are indicated, where appropriate, in the main text, figures, and figure legends.