PCGF5 is required for neural differentiation of embryonic stem cells

Polycomb repressive complex 1 (PRC1) is an important regulator of gene expression and development. PRC1 contains the E3 ligases RING1A/B, which monoubiquitinate lysine 119 at histone H2A (H2AK119ub1), and has been sub-classified into six major complexes based on the presence of a PCGF subunit. Here, we report that PCGF5, one of six PCGF paralogs, is an important requirement in the differentiation of mouse embryonic stem cells (mESCs) towards a neural cell fate. Although PCGF5 is not required for mESC self-renewal, its loss blocks mESC neural differentiation by activating the SMAD2/TGF-β signaling pathway. PCGF5 loss-of-function impairs the reduction of H2AK119ub1 and H3K27me3 around neural specific genes and keeps them repressed. Our results suggest that PCGF5 might function as both a repressor for SMAD2/TGF-β signaling pathway and a facilitator for neural differentiation. Together, our findings reveal a critical context-specific function for PCGF5 in directing PRC1 to control cell fate.

1. The authors have showed that Pcgf5 promotes monoubiquitination of H2AK119 by Ring1B, in addition, they have showed that Pcgf5 loss of function results in reduction of ubH2A enrichment on the promoter of Nodal gene. However, the authors should perform ChIP-seq experiments in the not only mESCs but also PCGF5-/-mESCs, even in the neural differentiation process at the early stage to see Pcgf5 loss of function on the global effects and distribution of histone H2AK119ub1 level in both mESCs and differentiated cells. 2. The discussion part is too short. Nodal/TGF-β signaling pathway has been previously reported and has a negative effect in regulating neural differentiation, while this manuscript mainly focus on the upstream of pcgf5 loss-of-fuction interfere with ESC neural differentiation by activating SMAD2 phosphorylation and the Nodal/TGF-beta signaling pathway. As the members of Pcgfs, except Pcgf5, the other members of pcgfs, such as pcgf1, pcgf2, pcgf6 have been previously , but these papers were neither cited nor discussed. 3. The manuscript has a very short background as well, more like a Nature Letter, it should be enlarged here. 4. RNA-seq has been performed in this manuscript. However, there is no GEO number contained. GEO number should be added in a revised version of manuscript for RNA-seq experiments. 5. In supplementary figure 1a, capital letters for different epigenetic genes should be used as small letters. 6. It is hard to assess how many technical/experimental/biological replicates experiments have been done for some experiments. The authors could add this information into the figure legends. 7. The authors should explain how they generated Pcgf5-depleted 46C mESCs and they should provide the detailed information in the paper.
Reviewer #2 (Remarks to the Author): In this manuscript, Yao and co-authors dissect the role of the non-canonical Polycomb Repressive Complex 1 member Pcgf5 in the regulation of neural differentiation of embryonic stem (ES) cells. They used the TALEN gene editing system to generate Pcgf5 knockout ES cells. The loss of Pcgf5 did not lead to any change in ES cell maintenance or the expression of any of the key pluripotency factors. However, intriguingly the Pcgf5 null cells displayed a failure to silence these pluripotency factors and to activate neural specific genes during induced neural differentiation. They go on to link this block in neural differentiation to an increase in TGFB pathway activity. They show that inhibition of the TGFB pathway in Pcgf5 null cells partially rescues the neural differentiation phenotype; however the rescue is not very strong. Finally they find that there is a reduction in Polycomb activity at the Nodal promoter in Pcgf5 null cells and this correlates with the observed increase in Nodal transcription during neural differentiation.
Major points: 1. The partial rescue of Sox1, Nestin and Pax6 gene expression upon TGFB pathway inhibition in Figure 2C is mild at best. This would suggest that the role of hyper active TGFB pathway in Pcgf5 null cells is not actually causing the neural differentiation defect. 2. In Figure 2F and 2G, inhibition of the TGFB pathway by LY2109761 in wild type cells seems to lead to an increase in expression of neural markers (Nestin, Pax6 and Sox1). This suggests that the observed increase in expression of these same genes upon treatment with LY2109761 is solely a result of TGFB pathway inhibition and is independent of the loss of Pcgf5. 3. The authors should perform Western blots of H2AK119ub1 and H3K27me3 in Pcgf5 null ES and NSC. If Pcgf5 is essential for the presence of these histone marks in either cell type then you may expect a change in global levels of these marks. This would reinforce the ChIP evidence for a reduction in these marks at the Nodal promoter in figure 4. 4. In addition to Western blotting, the authors should also carry out ChIP-RX (with a drosophila chromatin spike-in for correct quantification) of H2AK119ub1 and H3K27me3 in ES and NSC. This would give a more comprehensive overview of the role of Pcgf5 in mediating these histone modifications. The analysis presented in the manuscript only focusses on one gene (Nodal), which is not a thorough approach to determine the effect of Pcgf5 loss on histone modifications at the other thousands of Polycomb target genes. 5. The ChIP evidence presented in figure 4 suggests that Pcgf5 has a role in repressing Polycomb targets (albeit only at one gene). A previous report (Gao et al, Nature, 2014) provided evidence that Pcgf5 functions as an activator of transcription following artificial recruitment to the luciferase promoter. For this reason the authors should perform Pcgf5 ChIP seq to determine its direct target genes. This will allow a proper assessment of the direct role of Pcgf5 in transcription and mediation of H2AK119ub1 and H3K27me3. The effects exhibited in figure 4 at the Nodal promoter may be an indirect loss of repression of Nodal caused by transcriptional changes at a different gene which could be the real Pcgf5 target. Therefore, coupling the existing RNA-seq analysis from figure 1 to a Pcgf5 ChIP-seq would go a long way to uncovering the genome-wide role of Pcgf5. 6. The model presented in figure 4J should be amended to reduce the focus on competition between Pcgf5 and the TGFB pathway, since this link does not seem to be all that strong. Instead the focus should be shifted to the role of Pcgf5 in regulating H2AK119ub1 and PRC2 recruitment/H32K7me3 deposition at Polycomb target genes and also illustrate the block in differentiation exhibited in the absence on Pcgf5. The authors have some very compelling preliminary evidence in Figure 1 particularly and I have made suggestions above on how to extend this.
Minor points 1. Figure 1B and 1C do not contribute to our understanding of Pcgf5 function since they only illustrate how the authors made the useful null cell line to study its function. Therefore these panels could be moved to a supplementary figure. 2. Figure 2F could be better presented as a bar chart showing the levels of Sox1 alone. The yaxis in this FACs analysis does not inform the reader. 3. The immunoprecipitation experiment presented in Figure 3A is not informative. Performing a Flag IP of Pcgf5 and then probing for H2A when the control IP does not have overexpressed H2A is not a properly controlled experiment  figure 3J should be quantified by quantifying Pcgf5 and H2AK119ub1 in cells in several fields and then quantifying the correlation between high Pcgf5 and high H2AK119ub1. 6. In Figures 3C-F, the authors nicely map the interaction regions between Ring1B and Pcgf5. To boost the ubiquitination assay presented in Figure 3I, they could include a mutant Pcgf5 incapable of interaction with Ring1B. One would expect that the Ring finger mutant of Pcgf5 would have greatly reduced ubiquitination activity. 7. In the figure 4 title, the authors use the phrase "Loss of Pcgf5 reduces the recruitment of histone H2AK119ub1 onto the promoter of nodal…". Histone modifications are not recruited, they are catalysed in situ, and therefore this title should be rephrased to reflect this. 8. Some nice data is left in the supplentary figure and could be moved to the main figure. The RT-PCRs in FS1A show that Pcgf5 is highly up-regulated during neural differentiation. It is the only non-canonical Pcgf that seems highly expressed in NSCs. The quantification of Pcgf1-6 expression in this supplementary figure could be moved to the main figure 1 to illustrate the potential importance of Pcgf5 in NSC prior to the authors characterising their Pcgf5 null cells.

Response to Reviewers' comments:
Reviewers' comments:  We also added the following sentences into the discussion.  Fig. 9c). These data suggest that PCGF5 may function as an activator of NPC-specific genes during mESC neural differentiation.  figure 1a (previous as supplementary figure 1a). 6. It is hard to assess how many technical/experimental/biological replicates experiments have been done for some experiments. The authors could add this information into the figure legends.

Answer: We thank the reviewer#2's comments on this manuscript. We have performed more experiments and address the questions as below.
Major points: 1. The partial rescue of Sox1, Nestin and Pax6 gene expression upon TGFB pathway inhibition in Figure 2C is mild at best. This would suggest that the role of hyper active TGFB pathway in Pcgf5 null cells is not actually causing the neural differentiation defect. SMAD2 phosphorylation (Fig.   2d) Fig. 2 to make the manuscript more clear. Figure 2F and 2G, inhibition of the TGFB pathway by LY2109761 in wild type cells seems to lead to an increase in expression of neural markers (Nestin, Pax6 and Sox1).

In
This suggests that the observed increase in expression of these same genes upon treatment with LY2109761 is solely a result of TGFB pathway inhibition and is independent of the loss of Pcgf5. Fig. 1a, b) Fig. 1c from the cited paper 7 below). Therefore, we Fig. 1d, e).

For Nestin and Pax6, when we treated the cells with another TGF-β signaling pathway inhibitor SB431542, we could not see the increase of RNAs for both Nestin and Pax6 in wild type cells during neural differentiation (see below
Together, we speculate that selectivity of LY2109761 or other unknown mechanisms might result in the increase of gene expression for these two markers. We removed previous Fig.2f, g.

The authors should perform Western blots of H2AK119ub1 and H3K27me3 in Pcgf5
null ES and NSC. If Pcgf5 is essential for the presence of these histone marks in either cell type then you may expect a change in global levels of these marks. This would reinforce the ChIP evidence for a reduction in these marks at the Nodal promoter in figure 4. Fig.7a, b). These are consistent with previous reports that the abundance of H3K27me3 mark decreased during neural differentiation 8

. We would like to emphasize that it is a common observation that the loss of epigenetic regulators does not necessarily lead to an obvious change in the global level of an epigenetic mark 8 , instead there are often relatively small context-specific changes, which we characterized with our ChIP-seq data.
4. In addition to Western blotting, the authors should also carry out ChIP-RX (with a drosophila chromatin spike-in for correct quantification) of H2AK119ub1 and H3K27me3 in ES and NSC. This would give a more comprehensive overview of the role of Pcgf5 in mediating these histone modifications. The analysis presented in the manuscript only focusses on one gene (Nodal), which is not a thorough approach to determine the effect of Pcgf5 loss on histone modifications at the other thousands of Polycomb target genes. Fig. 2a,b), therefore, we did not show these data in the manuscript; However, PCGF5 ChIP-seq would go a long way to uncovering the genome-wide role of Pcgf5.

Answer: We have followed the reviewer's suggestions and performed PCGF5 ChIP-seq in NPCs at day 6 after neural differentiation. At the beginning, we performed ChIP-seq
by using commercial anti-PCGF5 antibodies; unfortunately, these antibodies did not work for ChIP. Therefore, we generated Flag-tagged PCGF5 knockin stably expressing in Sox1-GFP mESCs (Fig. 4a-c and Supplementary Fig. 6a) (Fig.4d). About 25.04% of PCGF5 sites are near promoter regions.

35.87% of the PCGF5-binding sites are located in the intergenic regions, a significant number of PCGF5-binding sites fall within genes, with 33.55% in the introns and 5.54
in the exons (Fig.4e). We further focused on the effects of PCGF5 on the genes in NPCs. Among 80 genes in the TGF-β signaling pathway, 35 of them were directly targeted by PCGF5.

ChIP-seq by using anti-H2AK119ub1 and anti-H3K27me3 antibodies in both wild type and Pcgf5-depleted NPCs. We investigated whether PCGF5 co-bound to specific genes with histone repressive marks, H2AK119ub1 and H3K27me3, we found that only small fraction of PCGF5-binding sites overlapped with repressive marks H2AK119ub1-and
H3K27me3-binding sites (Supplementary Fig. 9a) Fig. 9b). Interestingly, we further found that about 39.5% (1024) of PCGF5-binding sites targeted to the active genes and only 6.6% (170) of PCGF5-binding sites bound to silent genes (Supplementary Fig. 9c). PCGF5 loss-of-function resulted in 521 downregulated PCGF5 target genes that were associated with axon development, neuron project development (Fig. 4f,g). And the upregulated genes targeted by PCGF5 were involved in mRNA processing, ribonucleoprotein complex biogenesis (Fig. 4g).

Included within these 99 upregulated PCGF5 target genes were Nodal, Lefty1 and
Lefty2, three TGF-β signaling pathway genes. Hence, we performed ChIP-qPCR experiments to examine the recruitments of PCGF5, H2AK119ub1 and H3K27me3 onto the promoters of these genes. Our data indicated that PCGF5 was indeed specifically recruited to the promoters of Nodal, Lefty1 and Lefty2 at day 6 during neural differentiation of mESCs (Fig. 4h). On the other hand, PCGF5 loss-of-function prevented the reduction of H2AK119ub1 and H3K27me3 around neural specific genes and kept them repressed (Fig. 5e-h), further suggesting that PCGF5 might be required for activation of NPC-related genes. We added these data into the newly updated manuscript. 6. The model presented in figure 4J should be amended to reduce the focus on competition between Pcgf5 and the TGFB pathway, since this link does not seem to be all that strong. Instead the focus should be shifted to the role of Pcgf5 in regulating H2AK119ub1 and PRC2 recruitment/H3K27me3 deposition at Polycomb target genes and also illustrate the block in differentiation exhibited in the absence on Pcgf5. The authors have some very compelling preliminary evidence in Figure 1 particularly and I have made suggestions above on how to extend this.

H2AK119ub1 and H3K27me3 at key TGF-β signaling genes, and that this is required for the repression of these genes, but also facilitates the reduction of H2AK119ub1 and
H3K27me3 around the promoters of neural-specific genes during neural differentiation, suggesting PCGF5 might play dual functions in regulating mESC neural differentiation, acting as a repressor for TGF-β signaling pathway and functioning as a facilitator for neural-related genes.

Minor points
1. Figure 1B and 1C do not contribute to our understanding of Pcgf5 function since they only illustrate how the authors made the useful null cell line to study its function.
Therefore these panels could be moved to a supplementary figure. Figure 1B and 1C to Supplementary Fig. 1f, g. 2. Figure 2F could be better presented as a bar chart showing the levels of Sox1 alone.

Answer: We have followed the reviewer's suggestion and moved
The y-axis in this FACs analysis does not inform the reader. Fig. 2h (previous as Figure   2f) with a bar chart showing Sox1 levels in different treatments in Fig. 2i. 3. The immunoprecipitation experiment presented in Figure 3A is not informative.

Answer: We followed the reviewer's suggestion and added
Performing a Flag IP of Pcgf5 and then probing for H2A when the control IP does not have overexpressed H2A is not a properly controlled experiment. H2A (Fig. 3a).  figure 3J should be quantified by quantifying Pcgf5 and H2AK119ub1 in cells in several fields and then quantifying the correlation between high Pcgf5 and high H2AK119ub1.

Answer: We have followed the reviewer's suggestion and quantified PCGF5 and
H2AK119ub1 levels in cells in multiple fields. We also analyzed the correlation between high Pcgf5 and high H2AK119ub1. We added these data into supplementary Fig. 5c, d.

In Figures 3C-F, the authors nicely map the interaction regions between Ring1B and
Pcgf5. To boost the ubiquitination assay presented in Figure 3I, they could include a mutant Pcgf5 incapable of interaction with Ring1B. One would expect that the Ring finger mutant of Pcgf5 would have greatly reduced ubiquitination activity.
Answer: We sincerely accepted reviewer's comments. We further performed in vitro ubiquitination assay and found that histone H2A ubiquitination activity have greatly reduced with a mutant Pcgf5 without Ring-finger. We added this data into The RT-PCRs in FS1A show that Pcgf5 is highly up-regulated during neural differentiation. It is the only non-canonical Pcgf that seems highly expressed in NSCs.