Introduction

The coordinated expression of Hox genes clusters is critical for axial patterning to determine the positional identities during the development of many tissues, including the hindbrain (Frank et al., 2019; Parker et al., 2014; Parker et al., 2020; Pöpperl et al., 1993.). In mammals, there are 39 Hox genes organized in four clusters (HoxA-HoxD) and divided into 1-13 paralog groups (PG). During vertebrate development, the hindbrain forms eight metameric segmented units along the antero-posterior (A-P) axis known as rhombomeres (r), which later give rise to the cerebellum, pons and medulla later on (Ghosh et al., 2018; Parker et al., 2020.). In the embryonic hindbrain, Hox1-5 are expressed in a nested overlapping fashion with rhombomere-specific boundaries. This Hox code is critical for rhombomeric territories establishment and cell fate determination (Barsh et al., 2017; Parker et al., 2016; Parker et al., 2020.).

At later stages, the Hox genes also play vital roles in neuronal migration and neural circuit formation (Feng et al., 2021; Hockman et al., 2019; Kratochwil et al., 2017; X. Wang et al., 2013.). For example, the pontine nuclei (PN) in the pons acts as an information integration station between the cortex and cerebellum (Kratochwil et al., 2017; Maheshwari et al., 2020.). The PN neurons originate at the posterior rhombic lip of r6-r8 and migrate tangentially to the rostral ventral hindbrain (Di Meglio et al., 2008; Farago et al., 2006; Geisen et al., 2008; Kratochwil et al., 2017; Maheshwari et al., 2020.). The PN neurons express Hox 3-5 in a nested pattern determined by their rhombomeric origin, and maintained throughout migration and PN assembly (Di Meglio et al., 2013; Kratochwil et al., 2017; Lizen et al., 2017.). Hox factors are crucial for the specification, migration, cellular organization, and topographic input connectivity of PN neurons. Note that Hox gene expression is maintained up to postnatal and even adult stages in the hindbrain derivatives (Farago et al., 2006; Feng et al., 2021; Kratochwil et al., 2017.).

Hox pattern maintenance is generally governed by epigenetic regulators, especially the polycomb-group (PcG) and trithorax-group (trxG) complexes, which activate and repress Hox genes, respectively (Bahrampour et al., 2019; Chopra et al., 2009; Kang et al., 2022; Papp et al., 2006.). In mammals, the COMPASS (complex of proteins associated with Set1) family of TrxG complexes act as methyltransferases. They contain different SET/MLL proteins, and a common multi-subunit core module consisting of WDR5, RbBP5, Ash2L, and DPY-30 (WRAD) (Cenik et al., 2021; Jambhekar et al., 2019; Schuettengruber et al., 2017.). WDR5 plays a central scaffolding role in these complexes and has been shown to act as a key regulator of Hox maintenance (K. C. Wang et al., 2011; Wysocka et al., 2005.). It is involved in MLL-mediated HoxA9 activation in acute myeloid leukemia (AML) and acute lymphoblastic leukemia (ALL) and is a target for drug development (Chen et al., 2021; Yu et al., 2021.). The long noncoding RNA (lncRNA) HOTTIP also binds to WDR5 and recruits the MLL complex to stimulate HoxA9 and HoxA13 expression in prostate cancer and hepatocellular carcinoma (Fu et al., 2017; Malek et al., 2017; Quagliata et al., 2014; Wong et al., 2020.).

RNF220 is an E3 ubiquitin ligase widely involved in neural development through targeting a range of proteins (Kim et al., 2018; Kong et al., 2010; Pengcheng Ma & Mao, 2021; Pengcheng Ma et al., 2022; Y.-B. Wang et al., 2022.). Previous research has shown that Rnf220−/− mice exhibit neonatal lethality and Rnf220+/− mice develop severe motor impairments; however, the mechanisms responsible for these phenotypes remain incompletely understood (Pengcheng Ma, Li, et al., 2021; Pengcheng Ma et al., 2019.). In the present study, we observed marked up-regulation of Hox genes in the pons of both Rnf220+/−and Rnf220−/− mice during late embryonic development. In the PN of Rnf220+/− mice, both the Hox expression pattern and the projection pattern from the motor cortex to the PN were disrupted. We identified WDR5 as a direct target of RNF220 for K48-linked ubiquitination and thus degradation, a mechanism critical for Hox regulation by RNF220. These findings revealed a novel role for RNF220 in the regulation of Hox genes and pons development.

Results

Rnf220 insufficiency leads to dysregulation of Hox expression in late embryonic pons

Rnf20 is strongly expressed in the embryonic mouse mid/hind brain (Pengcheng Ma et al., 2019; Y.-B. Wang et al., 2022.). Based on microarray analysis, we explored changes in expression profiles in E18.5 Rnf220+/−and Rnf220−/− mouse brain. Interestingly, several Hox genes were up-regulated in the mouse brain of both genotypes (Tables 1 and 2). Considering that Hox expression in forebeain is to low (date not shown), we focus on hindbrain and found that this phenomenon was observed in late embryonic stages after E15.5 but not earlier (Figure 1-figure supplement 1). To determine the specific brain regions exhibiting aberrant Hox expression, we examined the expression levels of HoxA9 and HoxB9, which were among the most significantly up-regulated genes, across various brain and spinal regions (Figure 1-figure supplement 2A-D). Notably, both HoxA9 and HoxB9 were exclusively up-regulated in the brainstem (Figure 1-figure supplement 2C and D). The mouse brainstem comprises the midbrain, pons, and medulla (Figure 1-figure supplement 2E). Further validation confirmed that the dysregulation of these Hox genes was restricted to the pons in the Rnf220+/− mice (Figure 1-figure supplement 2F). RNA sequencing (RNA-seq) analysis of the pons in 2-month-old Rnf220+/−mice similarly revealed an overall increase in Hox expression (Figure 1A).

Up-regulation of Hox genes in pons of Rnf220+/− mice. (A) The heatmap of RNA-seq data showing Hox genes expression in pons of WT or Rnf220+/− mice (n=2 mice per group). (B-C) UMAP diagram showing 15 identified cell clusters annotated by snRNA-seq analysis of pons. Each dot represents a single cell, and cells are laid out to show similarities (n=3 mice per group). (D) Heatmap of snRNA-seq data showing Hox expression changes in each cell cluster. (E) Quantitative real time polymerase chain reaction, qRT-PCR) analysis showing mRNA levels of indicated Hox genes in P19 cells when endogenous Rnf220 was knocked down by siRNAs in the presence of RA. WT, wild-type. HE, heterozygote. RA, retinoic acid. **p < 0.01, ***p < 0.001.

Differently expressed genes identified using microarray between WT and Rnf220+/− mice.

Differently expressed genes identified using microarray between WT and Rnf220−/− mice.

To identify the specific cell population in which Hox genes were up-regulated, we conducted single-nucleus RNA-seq (snRNA-seq) analysis of the pons in 2-month-old wild-type (WT) and Rnf220+/− mice. In total, 125 956 cells were resolved in the WT mice and 122 266 cells were resolved in the Rnf220+/− mice, each containing an average of approximately 700 genes per cell. Subsequent uniform manifold approximation and projection (UMAP) identified 15 cell clusters, with no significant difference observed between the WT and Rnf220+/− groups (Figure 1B and C). These clusters were annotated based on their uniquely and highly expressed markers (Table 3). Among them, 10 clusters were identified as distinct neuronal groups, in addition to astrocytes, oligodendrocytes, oligodendrocyte precursors, ependymal cells, and a group not corresponding to any known cell type (Figure 1B). We analyzed changes in Hox gene expression within each cluster and found that up-regulation of Hox genes was most pronounced in clusters 2, 7, and 10 (Figure 1D), albeit with distinct profiles. For instance, Hox8–10 were the most up-regulated genes in cluster 2, while Hox3–5 and Hox1–3 were the most up-regulated genes in clusters 7 and 10, respectively (Figure 1D). These findings indicate that Hox genes were specifically de-regulated in the pons of Rnf220+/− mice during late developmental stages.

Differently expressed genes of pons identified using snRNA-seq between WT and Rnf220+/− mice.

P19 embryonal carcinoma cells can be induced to differentiate and express Hox genes upon retinoic acid (RA) administration and have been used to study Hox regulation (Figure 1-figure supplement 3A)(Heyden et al., 2003; Kondo et al., 1992; Pöpperl et al., 1993.). Here, we tested the effects of Rnf220 knockdown on RA-induced Hox expression in P19 cells. Without RA treatment, Rnf220 knockdown had no induction effect on Hox expression (Figure 1-figure supplement 3B). However, with RA treatment, Rnf220 knockdown increased RA-induced expression of HoxA1, B1, A9, and B9 (Figure 1E), suggesting a general role of Rnf220 in Hox regulation.

Rnf220 insufficiency disturbs Hox expression pattern and circuit formation in PN

Due to limitations in data resolution, the precise identification of each cluster was not fully resolved. We determined that cluster 10 likely represents cells derived from rhombomeres 2–5, given their expression of endogenous anterior Hox genes (Hox1–3), while cluster 7 likely represents cells derived from rhombomeres 6–8 cells, as evidenced by their expression of Hox3–5 (Figure 1D). After mapping known neuron-specific markers to the clusters, PN-specific markers, including Nfib, Pax6, and Barhl, were enriched in cluster 7 (Figure 2-figure supplement 1A and B). PN neurons serve as relay cells between the cerebral cortex and cerebellum, originating at the posterior rhombic lip and migrating tangentially to their final position in the ventral part of the pons (Di Meglio et al., 2013; Kratochwil et al., 2017; Maheshwari et al., 2020.). Hox genes are implicated in both PN patterning and circuit formation (Kratochwil et al., 2017; Maheshwari et al., 2020.). In this study, we explored the potential effects of Rnf220 insufficiency on PN patterning. The PN were first dissected to analyze the expression of Hox genes (Hox4 and Hox5) in both WT and Rnf220+/− mice (Figure 2-figure supplement 1C). The data indicated an up-regulation in Hox4–5 expression levels in the Rnf220+/− PN (Figure 2-figure supplement 1C). In the PN, endogenous Hox3–5 genes exhibit a nested expression pattern along the rostral-caudal axis, with Hox3 ubiquitously expressed throughout the PN, Hox4 localized to the middle and posterior regions, and Hox5 confined to the posterior segment (Kratochwil et al., 2017.). To ascertain whether this established pattern was disrupted in the Rnf220+/− PN, we dissected the PN of both WT and Rnf220+/− mice into rostral, middle, and caudal segments using a vibratome. Subsequent analysis of Hox35 expression revealed an up-regulation in Rnf220+/−PN, with uniform detection along the rostral-caudal axis (Figure 2A).

Hox gene expression was dysregulated and motor cortex projections were disorganized in PN of Rnf220+/− mice. (A) qRT-PCR analysis of relative expression levels of Hox3, Hox4, and Hox5 in rostral, middle, and caudal sections of PN in WT and RNF220+/− mice. Expression level of each gene in rostral section of WT PN was set to 1. (n=4 mice per group). (B) Diagram of experimental stereotactic injections. (C, D) Fluorescent signals showing projection from motor cortex to PNs in WT and Rnf220+/− mice (n=2 mice per group). WT, wild-type; HE, heterozygote; **p < 0.01; ***p< 0.001.

PN are predominantly innervated by the cerebral cortex and Hox expression is critical for the formation of proper projection patterns (Kratochwil et al., 2017; Maheshwari et al., 2020.). Considering the motor deficits of Rnf220+/− mice (Pengcheng Ma, Li, et al., 2021; Pengcheng Ma et al., 2019.), we next tested whether projection patterns were affected by anterograde tracing. A non-transneuronal tracing virus, rAAV-hSyn-EGFP-WPRE-hGH-poly A, was injected into the motor cortex of 2-month-old WT and Rnf220+/−mice to label axons sent from the motor cortex (Figure 2B). Green fluorescence intensity in the PN was measured after 3 weeks. In the PN, the projection range from the motor cortex was smaller in Rnf220+/− mice (Figure 2C and D). In addition, the projection pattern was shifted so that more fluorescence intensity was observed in the middle section in the Rnf220+/− PN compared to the wild type (Figure 2C and D).

Thus, Rnf220 insufficiency disturbed both the expression level and pattern of Hox genes in PN and affected the projection pattern from the motor cortex.

WDR5 is an RNF220 target for K48-linked ubiquitination and degradation

As Hox up-regulation was observed only in the late stage of embryonic development, we reasoned that Rnf220 may be involved in the maintenance of Hox expression. The Hox expression pattern is maintained epigenetically by the PcG and trxG complexes, which regulate the silencing and activation of Hox genes, respectively (Bahrampour et al., 2019; Chopra et al., 2009; Kang et al., 2022; Papp et al., 2006.). Indeed, we did observe local epigenetic modification changes in each Hox cluster in hindbrain of Rnf220+/− mice by ChIP-qPCR assays (Figure 3-figure supplement 1A and B). Therefore, we tested the expression of core components of PRC1/2 in the hindbrain (E18.5) and pons (2 months old) of Rnf220+/−mice and found no obvious change in protein levels (Figure 3-figure supplement 2A and B). However, when WDR5, a key component of the trxG complex, was tested, a clear increase was observed in the hindbrain of both Rnf220+/−and Rnf220−/− mouse embryos at E18.5 (Figure 3A), which was already evident at E16.5 (Figure 3B). We next tested the expression of WDR5 in the pons, cortex, and cerebellum of 2-month-old WT and Rnf220+/− mice. WDR5 was up-regulated in the pons but not in the cortex or cerebellum (Figure 3C, Figure 3-figure supplement 2C and D), consistent with the pons-specific Hox up-regulation. In addition, WDR5 also increased in the PN of Rnf220+/− mice (Figure 3D) and RA-treated P19 cells following Rnf220 knockdown (Figure 3E). Thus, the stabilization of WDR5 was correlated with Rnf220-regulated Hox expression.

RNF220 mediates WDR5 degradation. (A-D) Western blots analysis showing the protein level of WDR5 in the indicated brain tissues of mice with different genotypes at different ages. (E) Western blot analysis of protein levels of WDR5 in P19 cells with Rnf220 knockdown or not in the presence or absence of RA. IB, immunoblot. WT, wild-type. HE, heterozygote. KO, knockout. PN, pontine nuclei. NC, negative control. RA, retinoic acid.

We speculated that WDR5 might be a direct target of RNF220 for ubiquitination and degradation. Indeed, WDR5 efficiently pulled down endogenous RNF220 in mouse brain stem lysate (Figure 4A), and RNF220 and WDR5 pulled down each other when co-expressed in HEK293 cells (Figure 4B and C). Furthermore, co-expression of wild-type RNF220 but not the ligase-dead mutant (W539A) or the RING domain deletion (△Ring) form clearly reduced the expression level of WDR5 in HEK293 cells (Figure 4D), which was blocked by MG132, a proteasome inhibitor (Figure 4E). In ubiquitination assays, WDR5 ubiquitination was reduced in the lysate of the pons of Rnf220+/−mice (Figure 4F). Accordingly, the level of ubiquitinated WDR5 clearly increased when co-expressed with RNF220 but not the △Ring form in HEK293 cells (Figure 4G).

RNF220 interacts with and targets WDR5 for K48-linked polyubiquitination. (A) Endogenous co-immunoprecipitation analysis showing the interaction between RNF220 and WDR5 in hindbrains of WT mice. (B-C) Co-immunoprecipitation (co-IP) analysis of interactions between RNF220 and WDR5 in HEK293 cells. HEK293 cells were transfected with indicated plasmids and harvested after 48 h. Cell lysates were immunoprecipitated with anti-FLAG beads. Whole-cell lysate and immunoprecipitates were subjected to western blot analysis using indicated antibodies. (D) Western blots analysis shows the protein level of WDR5 when co-expressed with wild type or mutated RNF220 in HEK293 cells. (E) Western blots analysis shows the protein level of WDR5 when co-expressed with RNF220 in HEK293 cells in the presence of MG132 or not. MG132 treatment reversed the RNF220-induced decrease of WDR5 protein. (F) In vivo ubiquitination assays showing the ubiquitination status of WDR5 in hindbrains of WT or Rnf220+/− mice. (G) In vivo ubiquitination assays showing the ubiquitination status of WDR5 when co-expressed with WT or mutated RNF220 in HEK293 cells. (H) In vivo ubiquitination assays showing RNF220-induced polyubiquitination of WDR5 when the indicated ubiquitin mutations were used in HEK293 cells. (I) In vivo ubiquitination assays showing the ubiquitination status of the indicated WDR5 mutants when co-expressed with WT or ligase dead RNF220 in HEK293 cells. WT, wild-type. HE, heterozygote. KO, knockout. IB, immunoblot. UB, ubiquitin. WCL, whole cell lysate. △Ring, RNF220 Ring domain deletion. W539R, RNF220 ligase dead mutation. K48, ubiquitin with all lysines except the K48 mutated to arginine. K48R, ubiquitin with the K48 was substituted by an arginine. 3M, substitution of lysines at the positions of 109, 112, and 120 in WDR5 with arginines simultaneously.

We tested whether the polyubiquitination of WDR5 by RNF220 was K48 linked, which is the major form of ubiquitination modification leading to proteasomal degradation (Grice et al., 2016.). Indeed, the K48 ubiquitin mutant (in which all the lysines are substituted by arginines except lysine 48) was efficiently incorporated into the ubiquitinated WDR5 induced by RNF220, but not the K48R mutant (in which lysine 48 was substituted by an arginine), confirming that the polyubiquitination chains are K48-linked (Figure 4H). To determine the exact lysines ubiquitinated by RNF220, we tested the effects of RNF220 on the stability of different WDR5 truncates and found that the truncate remaining 1-127aa was enough to be degraded by RNF220 (Figure 4-figure supplement 1A). We then mutated the conserved lysines in this region to arginine individually, and found that mutation of K109, K112, or K120 reduced the ubiquitination level of WDR5 by RNF220 (Figure 4-figure supplement 1B). Simultaneous substitution of these three lysines with arginine (3M) greatly reduced WDR5 ubiquitination by RNF220 (Figure 4I). In conclusion, these findings indicate that RNF220 mediates K48-linked polyubiquitination of WDR5 primarily at residues K109, K112, and K120.

WDR5 is required for Hox regulation by RNF220

To investigate the requirement of WDR5 in RNF220-mediated Hox regulation, we first examined whether WDR5 knockdown could mitigate the impact of Rnf220 knockdown on Hox expression in RA-induced P19 cells. When P19 cells were transfected with siWdr5 together with siRnf220 and accompanied by RA treatment, the stimulation of Hox genes by RNF220 knockdown was significantly reduced (Figure 5A-D). Note that the expression of Hox genes had no clear change upon Wdr5 knockdown alone with or without RA treatment in P19 cells (Figure 5-figure supplement 1).

WDR5 recovered Rnf220 deficiency-induced up-regulation of Hox genes in P19 cell line. (A-B) qRT-PCR analysis showing the expression levels of Rnf220 (A) and Wdr5 (B) when transfected the indicated combinations of siRNAs against Rnf220 or Wdr5 in the presence or absence of RA. (C-D) qRT-PCR analysis showing the expression levels of HoxA1, HoxB1, HoxA9, HoxB9 when transfected siRnf220 or both siRnf220 and siWdr5 when treated with RA (D) or not (C). RA: retinoic acid. n.s., not significant. *p< 0.05, **p < 0.01, ***p < 0.001.

We also tested in mice. we first examined the potential for WDR5-IN-4, an established WDR5 inhibitor (Aho et al., 2019.), to rescue Hox up-regulation in Rnf220+/− embryos. Using in-utero microinjection, we injected WDR5-IN-4 or DMSO in mice at E15.5 and harvested the embryos at E18.5, with the hindbrains processed for qRT-PCR analysis (Figure 6A). Results showed that WDR5-IN-4 had no effect on RNF220 expression (Figure 6B). Interestingly, the up-regulation of HoxA9 and HoxB9 in Rnf220+/− mice was reversed in the WDR5-IN-4 injected embryos (Figure 6B).

Genetic and pharmacological ablation of WDR5 recovered Rnf220 deficiency-induced up-regulation of Hox genes. (A) Diagram of experimental strategy for in utero local injection of WDR5 inhibitors. (B) qRT-PCR analysis of expression levels of Rnf220, HoxA9, and HoxB9 in hindbrains of WT and Rnf220+/− mouse embryos treated with WDR5 inhibitors or not at E18.5 (n=3 mice per group). (C) qRT-PCR analysis of expression levels of Rnf220, Wdr5, HoxA9, and HoxB9 in hindbrains of adult mice with indicated genotypes (n=2 mice per group). WT, wild-type. HE, heterozygote. ***p < 0.001.

As loss of RNF220 stabilizes WDR5, we next tested whether WDR5 insufficiency would rescue the Hox expression pattern in Rnf220+/−;Wdr5+/−double heterozygous embryos. Rnf20fl/wt;Nestin-Cre mice were mated with Wdr5fl/wt mice to obtain embryos with different genotypes. The expression of Hox genes in the pons at P15 were analyzed by RT-PCR (Figure 6C, Figure 6-figure supplement 1). Results showed that the up-regulation of Hox in Rnf20fl/wt;Nestin-Cre mice were markedly recovered in the Rnf220fl/wt;Wdr5fl/wt;Nestin-Cre mice (Figure 6C, Figure 6-figure supplement 1).

Taken together, the above findings support the involvement of WDR5 in RNF220 insufficiency-induced Hox up-regulation.

Discussion

During early embryonic development, the Hox genes are collinearly expressed in the hindbrain and spinal cord along the A-P axis to guide regional neuronal identity. Later on, the segmental Hox gene expression pattern in the hindbrain is maintained till at least early postnatal stages. The Hox genes are also expressed in adult hindbrains with restricted anterior boundaries. In addition to their early roles in progenitor cell specification, cell survival, neuronal migration, axon guidance, and dendrite morphogenesis (Smith et al., 2024.), Hox genes also play key roles in the regulation of synapse formation, neuronal terminal identity and neural circuit assembly at late stages (Feng et al., 2020; Feng et al., 2021; Philippidou et al., 2013.). However, the mechanisms and functions of Hox expression during late hindbrain development remain poorly established. Our work revealed a dose-dependent role of RNF220/WDR5 in the maintenance of Hox expression in late hindbrain development which might have a functional role in the neural circuit organization of the pons.

In mammals, the COMPASS family of histone-lysine methyltransferase complexes is required for the maintenance of active chromatin states and thus Hox expression patterns (Bahrampour et al., 2019; Chopra et al., 2009; Kang et al., 2022; Papp et al., 2006.). As a key component of the COMPASS-related complexes, WDR5 is involved in the developmental regulation of Hox genes (Wysocka et al., 2005.). WDR5 is required for MLL-fusion protein (MLL-FP)-induced Hox activation in AML and ALL (Chen et al., 2021; Yu et al., 2021.). The lncRNA HOTTIP has been shown to interact with WDR5 and recruit WDR5-MLL complexes to HoxA to induce H3K4 trimethylation and gene activation (Fu et al., 2017; Malek et al., 2017; Quagliata et al., 2014; Wong et al., 2020.). Our findings suggest that increased WDR5 levels in the hindbrain were associated with aberrant Hox expression at the late developmental stages. This dysregulation could be ameliorated either by WDR5 inhibition or haploinsufficiency. The observed up-regulation of Hox genes occurred at the epigenetic level within the Hox cluster, as evidenced by qRT-PCR analysis and a generalized increase in active H3K4 methylation markers. Furthermore, the stabilization of WDR5 and stimulation of Hox expression upon RNF220 knockdown were also recapitulated in RA-treated P19 cells.

Interestingly, single-cell analysis revealed distinct profiles of Hox gene activation across different neuronal groups. Specifically, Hox13 was up-regulated in cluster 10, Hox35 was up-regulated in cluster 7, and Hox710 was up-regulated in cluster 2. Clusters 10 and 7 naturally expressed Hox1–3 and Hox3–5, respectively, suggesting that the observed Hox gene stimulation in these clusters could be attributed to elevated WDR5 levels induced by RNF220 insufficiency. Cluster 2 warrants particular attention, as it exhibited up-regulation of Hox7–10 genes, which were not expressed in the WT hindbrain. This cluster was characterized as glycinergic neurons expressing Slc6a5. The functional implications of this distinct cell group and the phenotypic consequences of Hox gene dysregulation require further investigation. The dysregulation of Hox genes could have extensive implications for the development of various nuclei within the pons, affecting aspects such as cell differentiation and the formation of neural circuits. For the PN, we showed that both their structural patterning and patterns of projections originating from the motor cortex were altered in Rnf220+/−mice. Therefore, we can fully suspect that the motor impairments of Rnf220+/−mice is caused by abnormal projection of the central nervous system, and of course, more convincing experimental verification are needed.

In summary, our data support WDR5 as an RNF220 target involved in the maintenance of Hox expression and, consequently, the development of the pons.

Materials and Methods

Mouse strains and genotyping

All procedures involving mice were conducted in accordance with the guidelines of the Animal Care and Use Committee of the Kunming Institute of Zoology, Chinese Academy of Sciences (IACUC-PA-2021-07-018). Mice were housed under standard conditions at a temperature range of 20–22 °C, humidity of 30%–45%, and 12-h light/dark cycle. All mice used were maintained on a C57BL/6 background. Vasa-Cre mice were used to mate with Rnf220 floxed mice (Rnf220fl/fl) to generate Rnf220 germ cell knockout (Rnf220−/−) or heterozygote (Rnf220+/−) mice. Nestin-Cre mice were used to mate with Rnf220fl/fland Wdr5 floxed mice (Wdr5fl/fl) to generate conditional neural specific knockout or heterozygote mice. Rnf220fl/wt;Wdr5fl/wt;Nestin-Cre mice were obtained by crossing Rnf220fl/wt;Nestin-Cre with Wdr5fl/wt mice.

Genotypes were determined by PCR using genome DNA from tail tips as templates. PCR primers were listed as follows: 5’-CTG TTG ATG AAG GTC CTG GTT-3’ and 5’-CAG GAA AAT CAA TAG ACA ACT T-3’ were used to detect Rnf220 floxP carrying. 5’-CTG TTG ATG AAG GTC CTG GTT-3’ and 5’-CTG ATT TCC AGC AAC CTA AA-3’ were used to detect Rnf220 knockout. 5’-GCC TGC ATT ACC GGT CGA TGC-3’ and 5’-CAG GGT GTT ATA AGC AAT CCC-3’ were used for Cre positive detection. 5’-GAA TAA CTA CTT TCC CTC AGA CC-3’ and 5’-CAG GCC AAG TAA CAG GAG GTA G-3’ were used to detect Wdr5 floxP carrying. 5’-GAA TAA CTA CTT TCC CTC AGA CC-3’ and 5’-AGA CCC TGA GTG AGG ATA CAT AA-3’ were used to detect Wdr5 knockout.

Cell culture

HEK293 and P19 cell lines were grown in Dulbecco’s Modified Eagle Medium (DMEM) (Gibco, C11995500BT) supplementary with 10% fetal bovine serum (FBS) (Biological Industries, 04-001-1A), 100 units/mL penicillin and 100 mg/ml streptomycin (Biological Industries, 03-031-1B). Cell cultures were maintained at 37°C in a humidified incubator with 5% CO2. 0.5μM RA (Sigma, R2625) was used to induce Hox expression in P19 cells.

To achieve gene overexpression or knockdown, HEK293 and P19 cells were transfected by Lipo2000 (Invitrogen, 11668500) according to the manufacturer’s instructions. The following small interfering RNAs (siRNAs) (RiboBio) were used for Rnf220 or Wdr5 knockdown in P19 cells: siG2010160325456075, siG2010160325457167, or siG2010160325458259; and siWdr5: siB09924171210, siG131113135429, or siG131113135419.

Total RNA isolation and qRT–PCR

Tissue and cells were homogenized with 1 mL TRIZOL (TIANGEN, DP424), after which 200 μL of chloroform was added to the lysates for phase separation. After centrifugation at 12 000 g for 15 min at 4 °C, the aqueous phase (500 μL) was transferred to a new tube and mixed with equal volumes of isopropanol for RNA precipitation. After 30 min, RNA pellets were harvested by centrifugation at 12 000 g for 15 min at 4 °C, twice washed with 75% ethanol, and dissolved in DNase/RNase-free water.

cDNA was synthesized with Strand cDNA Synthesis Kit (ThermoScientific, K1632) according to the manufacturer’s instructions. All reactions were performed at least triplicates with Light Cycler 480 SYBR Green Ⅰ Master (Roche, 04707516001). Primers used for real-time polymerase chain reaction list in Table 4.

ChIP-qPCR

Fresh mouse PN tissues were fully grinded, cross-linked with 1% formaldehyde for 25 min, and stopped with addition of glycine. After gentle centrifuging, samples were collected and lysed with ChIP lysis buffer (150 mM NaCl, 25 mM Tris, pH 7.5, 1% Triton X-100, 0.1% SDS, 0.5% deoxycholate, and complete protease inhibitor cocktail) for 30 min on ice. Genomic DNA was sheared into 200-500 bp fragments by sonication. After centrifugation, the supernatants containing chromatin fragments were collected and immunoprecipitated with anti-IgG (2 μg, ProteinTech, B900610), anti-H3K27me3 (2 μg, Abcam, ab192985), or anti-H3K4me3 antibodies (2 μg, Abcam, ab8580) at 4℃ overnight. Then the immunoprecipitates coupled with protein A/G agarose beads (Santa Cruz, sc-2003) were washed sequentially by a low salt washing buffer, high salt washing buffer, LiCl washing buffer, and TE buffer. The immunoprecipitated chromatin fragments were eluted by 500 μL elution buffer for reversal of cross-linking at 65 °C overnight. Input or immunoprecipitated genomic DNA was purified by the QIAquick PCR Purification Kit (Qiagen, 28104) and used as a template for quantification PCR. The primers we used were listed as follows:

In utero microinjection

Pregnant mice were administered isoflurane for deep anesthesia. Following this, a laparotomy was performed to carefully extract the embryos, which were then placed on a sterile surgical drape. WDR5-IN-4 (100 μg, MedChemExpress, HY-111753A), containing 0.05% Malachite Green reagent for tracing, was injected into the hindbrain of E15.5 embryos using a finely-tapered borosilicate needle. To minimize the risk of spontaneous abortion, injections were spaced for the selected embryos. Following the injections, the uterus was returned to the abdominal cavity and infused with 2 mL of 37 °C, 0.9% saline solution. The peritoneum and abdominal skin were then sutured. Finally, the mice were placed on a heating pad to facilitate recovery from anesthesia.

Ubiquitination assay, immunoprecipitation and western blot analysis

In vivo ubiquitination, immunoprecipitation and Western blot assays were carried out as previously described (Pengcheng Ma et al., 2023.). The following primary antibodies were used for botting: anti-RNF220 (Sigma-Aldrich, HPA027578), anti-WDR5 (D9E11) (Cell Signaling Technology, 13105), anti-Ring1B (D22F2) (Cell Signaling Technology, 5694S), anti-Sin3B (AK-12) (Santa Cruz, sc-768), anti-EZH2 (ProteinTech, 21800-1-AP), anti-SUZ12 (Bethyl, A302-407A), anti-RYBP (A-1) (Santa Cruz, sc-374235), anti-CBX6 (H-1) (Santa Cruz, sc-393040), anti-CBX7 (G-3) (Santa Cruz; sc-376274), anti-CBX8 (C-3) (Santa Cruz, sc-374332), anti-PHC1 (D-10) (Santa Cruz, sc-390880), anti-α-Tubulin (ProteinTech, 66031-1-Ig), anti-FLAG (Sigma-Aldrich; F-7425), anti-myc (ProteinTech; 16286-1-AP), and anti-Ub (P4D1) (Santa Cruz, sc-2007)

snRNA-seq library preparation, sequencing, and data analysis

snRNA-seq libraries were prepared using the Split-seq platform (Butler et al., 2018; Rosenberg et al., 2018.). Freshly harvested mouse pons tissues underwent nuclear extraction following previously described protocols (Butler et al., 2018; Rosenberg et al., 2018.). In brief, mouse brain tissues were transferred into a 2 mL Dounce homogenizer containing 1mL homogenizing buffer (250 mM sucrose, 25 mM KCl, 5 mM MgCl2, and 10 mM Tris-HCl [pH = 8.0], 1 mM DTT, RNase Inhibitor and 0.1% Triton X-100). The mixture was subjected to five strokes with a loose pestle, followed by 10 strokes with a tight pestle. The resulting homogenates were filtered through a 40-μm strainer into 5-mL Eppendorf tubes and subsequently centrifuged for 4 min at 600 g and 4 °C. The pellet was re-suspended and washed in 1 mL of PBS containing RNase inhibitors and 0.1% BSA. The nuclei were again filtered through a 40-μm strainer and quantified. These nuclei were partitioned into 48 wells, each of which contained a barcoded, well-specified reverse transcription primer, to enable in-nuclear reverse transcription. Subsequent second and third barcoding steps were carried out through ligation reactions. After the third round of barcoding, the nuclei were divided into 16 aliquots and lysed prior to cDNA purification. The resulting purified cDNA was subjected to template switching and qRT-PCR amplification, which was halted at the beginning of the plateau stage. Finally, the purified PCR products (600 pg) were used to generate Illumina-compatible sequencing libraries. Distinct, indexed PCR primer pairs were employed to label each library, serving as the fourth barcode.

The libraries underwent sequencing on the NextSeq system (Illumina) using 150-nucleotide kits and paired-end sequencing protocols. In this arrangement, Read 1 covered the transcript sequences and Read 2 contained the UMI and UBC barcode combinations. Initially, a six-nucleotide sequencing index, serving as the fourth barcode, was appended to the ends of Read 2. Subsequently, reads with more than one mismatching base in the third barcode were excluded from the dataset. Furthermore, reads containing more than one low-quality base (Phred score ≤ 10) within the UMI region were also discarded. The sequencing results were aligned to exons and introns in the reference genome (Genome assembly GRCm38, https://www.ncbi.nlm.nih.gov/datasets/genome/GCF_000001635.20/) and aggregated intron and exon counts at the gene level were calculated by kallisto and bustools software as described (https://bustools.github.io/BUS_notebooks_R).

Following export of the matrix, quality control measures were performed to remove low-quality cells and potential doublets, as described previously (Rosenberg et al., 2018.). After filtering, a total of 125,956 and 122,266 cells for wild type and Rnf220+/−pons respectively remained for subsequent analysis. Seurat v2 (Butler et al., 2018.) was used to identify HVGs within each group. Principal component analysis (PCA) and UMAP were performed to embed each cell from the reduced principal component space on a 2D map. Then clustering of cell populations and identification of differentially expressed genes were carried out as previously described (Ma et al., 2022.). We annotated the embryonic cell populations and lineages based on their DEGs (Table 3).

Mouse brain stereotactic injection and neuronal tracing

Mouse brain stereotactic injection and neuronal tracing were carried out as previously described with minor modifications (Xu et al., 2021.). Adult mice (2–3 months old) were used for anterograde monosynaptic neuronal tracing. The mice were first deeply anesthetized with isoflurane and placed into a stereotactic apparatus with the front teeth latched onto the anterior clamp. The mouth and nose of the mice were covered with a mask to provide isoflurane and keep them in an anesthetic state during the operation. The head was adjusted and maintained in horizontal by inserting ear bars into the ear canal. The hair of the head was shaved with an electric razor and cleaned using a cotton swab which was dipped in 75% alcohol. Cut the scalp along the midline with surgical scissors and make sure the bregma and the interaural line were exposed. The intersection between the bregma and interaural line was set to zero. The coordinates (−1.75, −4.9, −1.6) were applied for motor cortex localization. Each mouse received a single injection of 1 µL of rAAV-hSyn-EGFP-WPRE-hGH-poly A (Brainvta, PT-1990) viral fluid in the left hemisphere. After the injection, mice were bred for another 3 weeks for neuronal tracing before their brains were harvested for analysis.

Slice preparation

Mice subjected to stereotactic injection were euthanized, and whole brains were collected for the preparation of frozen sections. Following fixation in 4% paraformaldehyde (diluted in PBS) for 48 h, the brains were successively treated with 20% sucrose (diluted in PBS) for 48 h and 30% sucrose for 24 h. Brain tissues were then embedded in Optimal Cutting Temperature (OTC) compound (SAKURA, 4583) at −20 ℃ and sliced along the sagittal plane at a thickness of 40 μm.

Quantification and statistical analysis

The fluorescent and chemiluminescent intensities were quantified by Image J software (National Institute of Health). Statistical analyses were performed using GraphPad Prism software (GraphPad Software Inc., La Jolla, CA, USA). Comparisons were performed using the two-tailed Student’s t-test. p-values of less than 0.05 were considered statistically significant (*), 0.01 were considered statistically significant (**) and 0.001 were considered statistically significant (***).

Acknowledgements

We are grateful to all members of the Mao and Zhang laboratories for discussion of and comments on the manuscript. We would like to thank the Core Technology Facility of Kunming Institute of Zoology (KIZ), Chinese Academy of Sciences (CAS) for providing us with service.

Additional information

Funding

Author contributions

Huishan Wang, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing – original draft; Xingyan Liu, Data curation, Software, Formal analysis, Visualization, Methodology; Yamin Liu, Validation, Investigation, Visualization, Methodology; Chencheng Yang, Investigation, Visualization, Methodology; Yaxin Ye, Investigation, Methodology; Nengyin Sheng, Conceptualization; Shihua zhang, Data curation, Software, Formal analysis; Bingyu Mao, Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Visualization, Project administration, Writing – review and editing; Pengcheng Ma, Conceptualization, Resources, Formal analysis, Supervision, Funding acquisition, Visualization, Project administration, Writing – review and editing.

Data availability

All the snRNA-seq and RNA-seq raw data have been deposited in the GSA (https://ngdc.cncb.ac.cn/gsa/) with accession number is CRA013111.

Figure supplements

Hox genes exhibited de-repression in hindbrain of Rnf220−/−and Rnf220+/− mouse embryos at late developmental stages. Heatmap of Hox expression in hindbrain of Rnf220−/− mouse embryos at E16.5 and E15.5. Expression of each Hox gene was analyzed by qRT-PCR against GAPDH. Each Hox gene in wild-type controls was set to 1 (n=2 mice per group). Blank color represents the expression of that Hox is low and exceed detection range. KO, knockout.

Hox genes were up-regulated in brainstem of Rnf220+/− mice. (A) Diagram of central neural system (CNS) in mice. (B) qRT-PCR analysis of Rnf220 expression in CNS of Rnf220+/− and control mice. (C-D) qRT-PCR analysis of HoxA9 (C) and HoxB9 (D) expression in each CNS section of Rnf220+/− and control mouse. (E) Diagram of mouse brain stem. (F) qRT-PCR analysis of mRNA levels of HoxA9 and HoxB9 in indicated sections of WT and Rnf220+/− mouse brainstems (n=2 mice per group). WT, wild-type. HE, heterozygote. SC, Spinal cord. n.s., not significant. *p < 0.05, **p < 0.01, ***p < 0.001.

Expression levels of Hox genes were not affected by Rnf220 knockdown in P19 cell line without RA induction. (A) qRT-PCR analysis of Rnf220 and indicated Hox genes expression after RA treatment in P19 cells. (B) qRT-PCR analysis of expression levels of HoxA1, HoxB1, HoxA9, and HoxB9 in Rnf220 knockdown P19 cells without RA induction. RA: retinoic acid. n.s., not significant. *p < 0.05, **p < 0.01, ***p < 0.001.

Hox genes were up-regulated in PN of Rnf220+/− mice. (A) Heatmap of snRNA-seq data showing expression levels of PN markers (Pax6, Barhl1, Unc5b, and Nfib) among 15 identified cell clusters. (B, C) qRT-PCR analysis of expression levels of indicated PN markers (B) and Hox genes (C) in PN of Rnf220+/− and WT mice (n= 3 mice per group). WT, wild-type. HE, heterozygote. n.s., not significant. *p < 0.05, ***p < 0.001.

Repressive epigenetic modification was down-regulated while activated epigenetic modification was up-regulated in promoter regions of indicated Hox genes in hindbrains of Rnf220+/− mice. (A-B) ChIP-qRT-PCR analysis of repressive epigenetic modification (H3K27me3) (A) and activated epigenetic modification (H3K4me3) (B) levels in promoter regions of indicated Hox genes in hindbrains of Rnf220+/− and WT mice (n=2 mice per group). WT, wild-type. HE, heterozygote. n.s., not significant. *p< 0.05, **p < 0.01, ***p < 0.001.

Protein levels of the indicated core components of PRC1 and PRC2 complex in indicated mouse brain tissues of different genotypes. (A) Western blot analysis of protein levels of core components of PRC1 and PRC2 complexes in hindbrain of WT, Rnf220+/−, and Rnf220−/− mouse embryos at E18.5. (B) Western blot analysis of protein levels of core components of PRC1 and PRC2 complexes in pons of adult Rnf220+/− and WT mice. (C, D) Western blot analysis of protein levels of WDR5 in cerebellum and cortex of adult Rnf220 +/−and WT mice. WT, wild-type; HE, heterozygote; KO, knockout; IB, immunoblot.

RNF220 interacted with and targeted WDR5 for polyubiquitination at multiple lysine sites. (A) Western blot analysis of levels of three WDR5 truncated proteins in HEK293 cells when co-transfected with RNF220 or not. (B) In vivo ubiquitination analysis of ubiquitination status of indicated WDR5 KR mutants when co-expressed with RNF220 or not in HEK293 cells. IB, immunoblot. IP, immunoprecipitation. WCL, whole cell lysate. K31R, K52R, K109R, K112R, K120R, K123R, or K126R, substitution of lysine with arginine in WDR5 at indicated positions.

WDR5 knockdown had no effect on Hox genes in P19 cell line. qRT-PCR analysis of mRNA levels of WDR5, HoxA1, HoxB1, HoxA9, and HoxB9 when WDR5 was knocked down by siRNA transfection in P19 cells with or without RA treatment. RA: retinoic acid. n.s., not significant. *p< 0.05.

Genetic ablation of Wdr5 reversed Hox gene de-repression caused by Rnf220 deficiency. qRT-PCR analysis of mRNA levels of HoxB8, HoxC4, HoxC5, HoxC6, HoxD8, HoxC9, and HoxD9 in hindbrains of WT, Wdr5fl/wt;Nestin-Cre; Rnf220fl/wt;Nestin-Cre, and Rnf220fl/wt;Wdr5fl/wt;Nestin-Cre mice (n=2 mice per group). WT, wild-type. n.s., not significant. ***p < 0.001.