REX1 is the critical target of RNF12 in imprinted X chromosome inactivation in mice

In mice, imprinted X chromosome inactivation (iXCI) of the paternal X in the pre-implantation embryo and extraembryonic tissues is followed by X reactivation in the inner cell mass (ICM) of the blastocyst to facilitate initiation of random XCI (rXCI) in all embryonic tissues. RNF12 is an E3 ubiquitin ligase that plays a key role in XCI. RNF12 targets pluripotency protein REX1 for degradation to initiate rXCI in embryonic stem cells (ESCs) and loss of the maternal copy of Rnf12 leads to embryonic lethality due to iXCI failure. Here, we show that loss of Rex1 rescues the rXCI phenotype observed in Rnf12−/− ESCs, and that REX1 is the prime target of RNF12 in ESCs. Genetic ablation of Rex1 in Rnf12−/− mice rescues the Rnf12−/− iXCI phenotype, and results in viable and fertile Rnf12−/−:Rex1−/− female mice displaying normal iXCI and rXCI. Our results show that REX1 is the critical target of RNF12 in XCI.

The data provided basically suggest that Rnf12-mediated degradation of Rex1 plays a crucial role in imprinted as well as random XCI. The study brings about another piece of in vivo evidence that Rnf12 targets Rex1 for degradation to initiate not only random but also imprinted XCI. Although this study contains some interesting and important findings, I provide some suggestions below to further strengthen the paper.
Major points 1. They examined the effect of loss of Rex1 in Xi reactivation in E11.5 PGCs as well as in ICM of E3.5 and E4.5 embryos by monitoring the timing of H3K27me3 disappearance. However, I wonder if the idea that the loss of H3K27me3 in ICM cells and PGCs can be referred to as Xi reactivation has been previously shown to be appropriate or widely accepted. If it has been the case, then the authors should cite the relevant references. If not, they should avoid to definitely mention Rex 1 is dispensable for Xi reactivation, or otherwise, should perform RNA-FISH to show expression of Xlinked genes at two sites in the nucleus of ICM cells or PGCs. In addition, although comparison of H3K27me3 localization in ICM cells between E3.5 and E4.5 allows them to suggest no difference in the timing of Xi reaction, they would not be able to evaluate the effect of functional loss of Rex1 on Xi reactivation in PGCs by looking at just one stage (E11.5). Figure 3d only tells that the accumulation of H3K27me3 on Xi has been lost by E11.5 in both wildtype and Rex1-/-PGCs. It is totally unclear if there is a difference between wildtype and Rex1-/-PGCs in the timing when H3K27me3 starts or completes to be lost. The authors should compare localization of H3K27me3, preferably in combination with RNA-FISH for X-linked genes, in PGCs at different stages. I would also like to point out that the quality of immunostaining for H3K27me3 in PGCs should be improved.
2. For allele-specific expression analysis, how did the authors prepare the extraembryonic yolk sac？ Did they separate the visceral endoderm from the yolk sac mesoderm? If they did not isolate the extraembryonic component of the yolk sac, they could not yet conclude that imprinted XCI was not affected. 3. Based on the observation that Rnf12-/Y males were underrepresented in the male pups born to Rnf12+/-females crossed with wildtype males, they concluded Rnf12-/Y display partial embryonic lethality. It is not clear, however, how the sex and genotype of the offspring were identified. If the number of respective genotypes was based on the number of pups at weaning, the loss of Rnf12-/Y males could be attributable to not only embryonic but also postnatal lethality. The authors should clarify this.
4. In Discussion, the authors describe a scenario that Rnf12-/+ cells in the preimplantation embryos carrying the maternal deletion of Rnf12 initiate imprinted inactivation of their paternal wildtype X by upregulating Xist on it. This would lead to functional loss of Rnf12 as the paternal Rnf12 is effectively downregulated, resulting in accumulation of Rex1 protein, which they propose represses Xist initially upregulated on the paternal X. The authors expect that this ends up with derepression of the initia lly inactivated paternal X and lack of imprinted XCI after all. A series of previous work from the Gribnau lab suggest that dose of Rnf12 is critical for upregulation of Xist and one copy of Rnf12 is insufficient for stochastically upregulating Xist in random XCI in differentiating ESCs. If this is also the case in the cells of preimplantation embryos, wouldn't Rnf12-/+ cells upregulate paternal Xist anyway? Furthermore, I am also wondering if Rex1 protein accumulates in those cells lacking maternal Rnf12 in the preimplantation Rnf12-/+ embryos as were seen in the trophectoderm of Rnf12-/-and Rnf12-/Y blastocysts. 5. The authors previously suggested that Rex1 was a potential activator of Tsix and Rnf12-mediated degradation of Rex1 abolished Tsix transcription, enabling Xist on the same chromosome to be expressed in cis to initiate XCI in ESCs. I am wondering if maternal transmission of Rnf12 deletion causes ectopic activation of paternal Tsix in Rna12-/+ as well as Rnf12-/-preimplantation embryos. If this is the case, one can ascribe the failure of imprinted XCI upon maternal transmission of Rnf12 deletion to ectopic expression of Tsix on the paternal X. The mice generated in this study provides a good opportunity to address this issue.
3. In Figure 3a, 4b, and 5b, how did the authors confirm the genotype of the embryos? Did they extract DNA from each embryo after the observation by fluorescence microscopy for PCR? 4. Images of RNA-FISH and immunofluorescence are too small in general.
Reviewer #2 (Remarks to the Author): The manuscript from Gribnau and colleagues addresses the mechanism by which RNF12, an E3ubiquntin-ligase, functions in imprinted X chromosome inactivation. While RNF12 has been shown to be required for rXCI in mESCs, it is dispensable for rXCI in vivo. However, maternally-transmitted Rnf12-deletion causes iXCI defects and female embryonic lethality, indicating a specific role of Rnf12 in iXCI. This manuscript is a step forward in showing that REX1, a protein target of RNF12, is directly downstream of Rnf12 in the regulatory pathway for iXCI. By showing that iXCI defects in the Rnf12deletion mutant embryo is associated with REX1 stabilization and that double-KO mice of Rnf12-/-; Rex1-/-are rescued, being viable and fertile irrespective of gender, the authors conclude that REX1 is downstream of RNF12 responsible for iXCI. In addition, the authors showed similar rescuing effects of REX1 removal in Rnf12-/-mESCs, which exhibit the normal rXCI. Their model summary therefore includes mechanisms of Rnf12-Rex1 regulation for both iXCI and rXCI. Overall, this paper contains a large amount of work and data, and provides new insight into the regulatory pathways controlling Xist and XCI, which represent an important area of research that is still poorly understood. The knockout mutant mice and the analyses from this work would be a valuable resource for researchers in the field. This paper can make an important contribution to our understanding of regulatory factors in controlling the mouse XCI, if the authors clarify some specific points.
Major points:

2.
As the authors explained in the introduction, "iXCI takes place during preimplantation development in the embryo and in the extra-embryonic tissues, where the paternal X chromosome is always inactivated." However, Fig. 3f shows bi-allelic expression of X-linked genes in the yolk sac, for both WT and Rex-/-samples. Similar observations are presented in Fig. 5d. Although the skewing pattern in the yolk sac is obviously distinct from the expression pattern in the embryo, for a normal situation, shouldn't iXCI in yolk sac reflect an exclusively paternal-X silencing? A clarification would be helpful.
3. Fig. 2d shows that Xist level is increased in Rex1+/-mESCs. Does it suggest that the "dosesensitive" repressive role of Rex1 on Xist could be effective as a heterozygous Rex+/-possibly leading to an ectopic expression of Xist? An Xist RNA FISH on these cells would be informative. Fig. 3a: the image resolution is low. H3K27me3 staining in the extraembryonic cell looks confusing in the E3.5 Rex1-/-sample. For E4.5 Rex1-/-embryo, there seems to be two foci of H3K27me3 in a cell (e.g. cells surrounding the ICM in the zoom-in panel for Rex1-/-show H3K27me3 pattern different from the WT example). It would be supportive to have Xist RNA FISH on these embryos, which should help clarify the XCI status.

5.
Figs. 4b & 5b: What are Xist expression patterns in Rnf12-/-embryos and in Rnf12-/-;Rex1-/embryo? Xist RNA FISH in these embryos should provide supportive and more direct evidence for the mutants' effect on Xist and iXCI. 6. Fig. 6: The "trans" repressive effect of RNF12 on REX1 for iXCI is confusing. With a trans effect, a paternal copy of RNF12 should be sufficient for REX1 degradation and iXCI in the Rnf12-/+ embryo. In the figure legend, Line 564: "This is followed by increased RNF12 expression in the developing epiblast, resulting in increased turnover of REX1, concomitant with its transcriptional downregulation, facilitating rXCI. In Rnf12-/+ and Rnf12-/-embryos, REX1 levels are stabilised in the extra-embryonic tissues, preventing Xist upregulation and iXCI." The authors should incorporate these in the illustration -the model figure is too simplified and not very helpful for understanding.

7.
In the discussion, the authors should address the possible involvement of other factors, which are identified in the same region of Xist. Specifically for imprinted XCI, transgenic studies have been reported for factors in the X-inactivation center, which is directly responsible for the mouse iXCI.

8.
It would be helpful if the authors could clarify the usage and analysis of different KO ESC lines, i.e. generated by CRISPR or HR-targeting, in regard to the consistence or possible difference in phenotypes 1.

2.
Line 105: (Fig. 2d,e) does not support "Rnf12 is required for rXCI in vivo" as the figured was referred.

3.
Supplementary Figure 4: This figure title does not match the content.
Reviewer #3 (Remarks to the Author): In this work, Gontan et. al. extend their previous findings (Ref 11) on the regulation of Xist and X inactivation by RNF12 and REX1 from mES cells to a mouse model. The present study shows that the removal of Rex1 is sufficient to rescue embryonic lethality of Rnf12 knockout, suggesting that REX1 is required for the maintenance of iXCI during development.
I recommend publication after the following points are addressed in the manuscript.

1.
In figure 1, the figure labels of 1d and 1e are swapped, leading to inconsistency between the description within the main text, figure legend and the figures; The legend for Fig 1. is also wrong.

2.
Line 102-3 Listing of figures does not match the text in Fig 1. (e.g. western blot is 1e, not 1d, also for qPCR and FISH)

3.
For SILAC results, it is common practice in the proteomics field to present all of the data points on the scatter-plot Fig 1b, rather than just a single quadrant and schematized figure. Doing so will provide readers a clearer understanding of the quality of the SILAC pulldown experiment and readers can visualize if the target protein is differentially expressed. This is especially so when no statistic test was performed as was the case in this manuscript.

Dear reviewers,
We would like to thank you for your constructive comments and questions. Besides the requested experiments and clarifications we have also included novel data related to the Rnf12 and Rex1 knockout ES cell lines described and studied in this manuscript. In our present manuscript we have studied Rnf12 -neo/-puro ESCs previously generated by homologous recombination (Barakat 2011) with alleles that leave the 5'region of Rnf12 intact, resulting in a 333aa peptide which lacks the catalytic Ring finger and interaction domain. In addition, for our study, we have now generated by CRISPR/Cas9 targeting new Rnf12 CR-/CR-ESCs that lack the complete open reading frame of Rnf12 (Fig.  1b,c and Supplementary Fig. 2). A recent paper published by the Bach laboratory (Wang Cell Reports 2017) suggests that the presence of the 333aa peptide in our Rnf12 neo-/puro-ESCs results in nuclear accumulation of REX1, possibly explaining discrepancies with respect to the XCI phenotype observed between both laboratories. To clarify this issue, we performed Western blotting (WB), IF and XCI studies comparing wild type (WT) ESCs with Rnf12 neo-/puroand Rnf12 CR-/CR-ESCs in the same genetic background. Our WB analysis confirms the presence of the truncated 333aa RNF12 peptide in Rnf12 neo-/puro-ESCs and the complete absence of RNF12 protein in Rnf12 CR-/CR-ESCs and that REX1 is accumulated at equally high levels in Rnf12 -neo/-puro and Rnf12 CR-/CR-ESCs as compared to the WT ESCs (Supplementary Figure 2a). In contrast to what is reported by Wang et al., immunocytochemistry analysis indicates nuclear accumulation of REX1 both in Rnf12 neo-/puroand Rnf12 CR-/CRknockout ESC lines ( Fig. 1d and Supplementary Figure 3b). Finally, examination of XCI dynamics indicates no difference between both Rnf12 knockout ESC lines, showing loss of XCI upon monolayer as well as embryoid body differentiation assays (Supplementary Fig. 3c-g). These findings indicate that the Rnf12 neo-/puroalleles behave as the full Rnf12 knockout alleles, which we think is important to show and discuss in the revised manuscript. We have included these results in Fig1 and Supplementary Fig2 and 3, and describe our results in the accompanying text and discussion.
For this revision, we have generated a new homozygous null Rex1 -/-ESC line (129/Sv:Cast/EiJ) from Rex1 −/− blastocysts. We have performed WB and XCI studies comparing WT ESCs with Rex1 +/and Rex1 -/-ESCs. (Fig. 2c,e and f, and Supplementary Fig.5). We observe an increase in the percentage of cells with two Xist clouds during differentiation in both Rex1 +/and Rex1 -/-ESCs. This result highlights the importance of Rex1 as a repressor of XCI.
Below we address these comments in a point by point rebuttal.

Reviewer 1
Major points 1. They examined the effect of loss of Rex1 in Xi reactivation in E11.5 PGCs as well as in ICM of E3.5 and E4.5 embryos by monitoring the timing of H3K27me3 disappearance. However, I wonder if the idea that the loss of H3K27me3 in ICM cells and PGCs can be referred to as Xi reactivation has been previously shown to be appropriate or widely accepted. If it has been the case, then the authors should cite the relevant references. If not, they should avoid to definitely mention Rex 1 is dispensable for Xi reactivation, or otherwise, should perform RNA-FISH to show expression of Xlinked genes at two sites in the nucleus of ICM cells or PGCs. In addition, although comparison of H3K27me3 localization in ICM cells between E3.5 and E4.5 allows them to suggest no difference in the timing of Xi reaction, they would not be able to evaluate the effect of functional loss of Rex1 on Xi reactivation in PGCs by looking at just one stage (E11.5). Figure 3d only tells that the accumulation of H3K27me3 on Xi has been lost by E11.5 in both wildtype and Rex1-/-PGCs. It is totally unclear if there is a difference between wildtype and Rex1-/-PGCs in the timing when H3K27me3 starts or completes to be lost. The authors should compare localization of H3K27me3, preferably in combination with RNA-FISH for X-linked genes, in PGCs at different stages. I would also like to point out that the quality of immunostaining for H3K27me3 in PGCs should be improved.  2007), thoroughly studied the relationship between loss of H3K27me3, Xist and Xreactivation in PGCs. Their work indicates that loss of H3K27me3 happens concomitantly with loss of Xist, preceding reactivation which happens shortly afterwards. On a gene by gene basis X reactivation is heterogeneous with some genes being reactivated shortly after loss of H3K27me3/Xist, others later during PGC development (Sugimoto and Abe Plos Genetics 2007), indicating that a global marker such as H3K27me3 will be more useful than focussing on single genes.
Nevertheless, we have attempted to follow X reactivation of a single transgene by crossing in Xlinked GFP and tdTomato alleles located in the Hprt locus (Wu Neuron 2014). Reactivation of this reporter was not detected in migrating PGCs both in a WT and Rex1 -/background indicating that reactivation of this reporter happens later after they reach the female gonads. Unfortunately, expression of the transgene in germ cells was down-regulated in the ovary and therefore reactivation (presence of yellow cells-GFP + /Tomato + ) could not be assessed in this setting (see rebuttal Fig. 1). and TRA98 antibodies, respectively. No reactivation of Hprt-reporters(presence of yellow cells-GFP+/Tomato+) is detected in migrating PGCs at E9.5 (top panels) and 11.5 (not shown). In E14.5 ovaries the Hprt-reporter is silenced, and therefore reactivation cannot be studied.
With respect to the timing of reactivation we agree with this reviewer that we should have looked at different time points to conclude on X reactivation dynamics. We therefore also examined PGCs in E9.5 WT, Rex1 -/and Rnf12 -/-Rex1 -/embryos determining the percentage of cells containing H3K27me3 coated X chromosomes. This study indicates that X chromosome reactivation is not delayed in the PGCs, and even shows a trend towards enhanced reactivation in the Rex1 -/-. We have included these new panels in Fig. 3c,d, Supplementary Fig. 6d,e and supplementary Fig. 9b Fig.6 b,c). Our results show that GFP detection in the epiblast, signaling the reactivation of the paternal Xi, is tightly linked to the loss of H3K27me3 and no difference was observed between Rex1 -/and WT female blastocysts.
2. For allele-specific expression analysis, how did the authors prepare the extraembryonic yolk sac？ Did they separate the visceral endoderm from the yolk sac mesoderm? If they did not isolate the extraembryonic component of the yolk sac, they could not yet conclude that imprinted XCI was not affected.
We agree with the reviewer. In the revised manuscript, we have repeated our experiments and performed the trypsin-pancreatin method (Nagy 2003) to enzymatically separate the mesoderm (VYSM, embryonic layer with random XCI) and endoderm (VYSE, extra-embryonic layer with imprinted XCI) layers of the visceral yolk sac (VYS). RT-PCR using the placental marker Afp and vascular endothelial marker Flk-1 confirmed that the separation had been successful (data not shown). We again find that Rex1 -/and Rnf12 -/+ :Rex1 -/have identical imprinted XCI compared to WT. 3. Based on the observation that Rnf12-/Y males were underrepresented in the male pups born to Rnf12+/-females crossed with wildtype males, they concluded Rnf12-/Y display partial embryonic lethality. It is not clear, however, how the sex and genotype of the offspring were identified. If the number of respective genotypes was based on the number of pups at weaning, the loss of Rnf12-/Y males could be attributable to not only embryonic but also postnatal lethality. The authors should clarify this.
Sexing and genotyping of pups was performed at day 5 after birth. We therefore agree with this reviewer that loss of Rnf12 -/y pups could have happened perinatally or early postnatally. We have adapted the manuscript accordingly.
4. In Discussion, the authors describe a scenario that Rnf12-/+ cells in the preimplantation embryos carrying the maternal deletion of Rnf12 initiate imprinted inactivation of their paternal wildtype X by upregulating Xist on it. This would lead to functional loss of Rnf12 as the paternal Rnf12 is effectively downregulated, resulting in accumulation of Rex1 protein, which they propose represses Xist initially upregulated on the paternal X. The authors expect that this ends up with derepression of the initially inactivated paternal X and lack of imprinted XCI after all. A series of previous work from the Gribnau lab suggest that dose of Rnf12 is critical for upregulation of Xist and one copy of Rnf12 is insufficient for stochastically upregulating Xist in random XCI in differentiating ESCs. If this is also the case in the cells of preimplantation embryos, wouldn't Rnf12-/+ cells upregulate paternal Xist anyway?
Our previous work has indicated that the robustness of XCI is affected in Rnf12 +/cells. Nevertheless, in vitro and in vivo many cells are still capable of initiating XCI, indicating that more XCI activators are involved in XCI initiation (Jonkers Cell 2009, Barakat PLoS Genetics 2010, Shin Nature 2010). We also showed that at least one intact copy of Rnf12 is required for XCI to be initiated in vitro (Barakat PLoS Genetics 2010, Barakat Mol Cell 2014) and that at least one copy of Rnf12 is required for sufficient break down of REX1. Our present study confirms this model for both in iXCI in the embryo and rXCI in ESCs. In iXCI, loss of XCI in Rnf12 -/mice can be rescued by removal of Rex1 resulting in live Rnf12 -/-:Rex1 -/animals, indicating that accumulation of REX1 is the main factor in iXCI mediated lethality. In Rnf12 -/+ blastocysts, iXCI is mostly abrogated as can be observed by the absence of H3K27me3 domains in trophoblasts cells. However, in some cells, Xist manages to escape REX1 repression and inactivates the X chromosome. This is not sufficient to sustain proper embryonic development, and Rnf12 -/+ eventually die. With respect to rXCI, we also find a rescue of the XCI phenotype observed in differentiating Rnf12 -/-ESCs upon removal of Rex1. However, for this revised manuscript we also established Rex1 -/-ES cell lines and determined and found that the percentage of cells with two clouds was significantly increased in differentiating Rex1 -/-ESCs. This finding, included as panels e and f in Figure 2 of the revised manuscript, and described in the accompanying text and discussion, supports a role for the RNF12:REX1 axis in negative feedback, preventing XCI of the second X chromosome. We also modified the model presented in Figure 6 to provide a better overview of our findings.
Furthermore, I am also wondering if Rex1 protein accumulates in those cells lacking maternal Rnf12 in the preimplantation Rnf12-/+ embryos as were seen in the trophectoderm of Rnf12-/-and Rnf12-/Y blastocysts.
As shown in Supplementary Figure 8b, Rnf12 -/+ preimplantation embryos do not show increased levels of REX1 as Rnf12 -/and Rnf12 -/y embryos. This might be due to the fact that the paternal X chromosome cannot be properly inactivated otherwise the cells would become Rnf12 -/which cannot upregulate Xist. Indeed, most of the cells in the trophoblast of Rnf12 -/+ blastocysts do not show H3K27me3 domains, which suggest these embryos have problems with inactivating their paternal Xi. Essentially, most of the trophoblast cells still keep the paternal X active and express RNF12 which will keep the levels of REX1 too low to enable the detection of REX1 upregulation by immunostaining. In the few cells that start Xi on the paternal X, as soon as RNF12 gets inactivated REX1 levels will be transiently upregulated leading to repression of Xist and ineffective imprinted XCI.
This negative feedback loop that prevents imprinted XCI to take place in the Rnf12 -/+ embryos is facilitated by the proximity along the X chromosome of Rnf12 and Xist genes, Rnf12 is one of the first genes to get compliantly inactivated on the Xp, as early as at the 8 cells stage only maternal Rnf12 expression is detected (Borensztein Nature Communications 2017). In addition, according to the results of our SILAC experiments, RNF12 and REX1 have a high turn-over rate, which will make the regulation of imprinted XCI via the RNF12-REX1 axis a very dynamic and fast process.
5. The authors previously suggested that Rex1 was a potential activator of Tsix and Rnf12-mediated degradation of Rex1 abolished Tsix transcription, enabling Xist on the same chromosome to be expressed in cis to initiate XCI in ESCs. I am wondering if maternal transmission of Rnf12 deletion causes ectopic activation of paternal Tsix in Rna12-/+ as well as Rnf12-/-preimplantation embryos. If this is the case, one can ascribe the failure of imprinted XCI upon maternal transmission of Rnf12 deletion to ectopic expression of Tsix on the paternal X. The mice generated in this study provides a good opportunity to address this issue.
We have addressed this question in the new Supplementary Fig. 8c. of our revised manuscript, and have modified the text accordingly. E3.5 hybrid embryos (129:cas) were generated by crossing 129 Rnf12 +/females mice with cas WT male mice. Allele specific Tsix expression was studied by RT-PCR making use of a Tsix RFLP (MnlI)(Lee Cell 2000) that produced a 129-specific band (198 bp) and a cas band (151 bp) (see materials and methods section for more details). During the revision period, we encountered breeding problems with our Cast/EiJ Rnf12 KO mice, which is why we could only confirm that there is no ectopic expression of the paternal Tsix (cas allele) in Rnf12 -(129)/+(cas) female embryos. Therefore, the failure of iXCI is due to direct repression of Xist and not to the upregulation of Tsix from the paternal allele. During rXCI in the differentiation of Rex1 -/+ and Rex1 -/-ESC lines Xist expression is up-regulated and Tsix expression is down-regulated in a REX1 dose-dependent mode ( Supplementary Fig. 5).
3. In Figure 3a, 4b, and 5b, how did the authors confirm the genotype of the embryos? Did they extract DNA from each embryo after the observation by fluorescence microscopy for PCR?
Yes, when necessary, the genotyping of the blastocysts was performed by PCR after immunostaining/FISH and confocal microscopy. Briefly, embryos were individually recovered, washed in PBS and lysed in 10 μl of lysis buffer (AM1722, Cells-to-cDNA™ II Kit, Thermo Fisher Scientific) for 15min at 75°C. 4. Images of RNA-FISH and immunofluorescence are too small in general.
As requested by this reviewer we have enlarged our Figure panels when possible.
2. As the authors explained in the introduction, "iXCI takes place during preimplantation development in the embryo and in the extra-embryonic tissues, where the paternal X chromosome is always inactivated." However, Fig. 3f shows bi-allelic expression of X-linked genes in the yolk sac, for both WT and Rex-/-samples. Similar observations are presented in Fig. 5d. Although the skewing pattern in the yolk sac is obviously distinct from the expression pattern in the embryo, for a normal situation, shouldn't iXCI in yolk sac reflect an exclusively paternal-X silencing? A clarification would be helpful.
We agree with the reviewer. Our previous visceral yolk sac (VYS) isolation protocol resulted in contamination of the extraembryonic tissue with embryonic material. In the revised manuscript, we have repeated our experiments and performed the trypsin-pancreatin method (Nagy 2003) to enzymatically separate the mesoderm (VYSM, embryonic layer with random XCI) and endoderm (VYSE, extra-embryonic layer with imprinted XCI) layers of the VYS. RT-PCR using the placental marker Afp and vascular endothelial marker Flk-1 confirmed that the separation had been successful (data not shown). We again find that Rex1 -/and Rnf12 -/+ :Rex1 -/mice have identical imprinted XCI compared to WT. Fig. 3e,f; 5c,d and Supplementary Fig. 7c and 9d have been modified accordingly. In the newly isolated VYSE there is a total skewing pattern where Xist expression is exclusively paternal and G6pdx and Mecp2 are maternally expressed.
3. Fig. 2d shows that Xist level is increased in Rex1+/-mESCs. Does it suggest that the "dosesensitive" repressive role of Rex1 on Xist could be effective as a heterozygous Rex+/-possibly leading to an ectopic expression of Xist? An Xist RNA FISH on these cells would be informative.
As suggested by this reviewer, we investigated Xist expression in WT, Rex +/and the newly generated Rex1 -/-ESCs by qPCR and Xist RNA FISH. Our qPCR results show a dose-dependent increase in Xist expression during differentiation of Rex1 +/and Rex1 -/-ESCs cells as compared to WT ( Supplementary  Fig. 5). In line with this result, we observed by Xist RNA FISH that, Rex1 +/and Rex1 -/-ESCs cells display significantly more Xist clouds at day 3 and 6 of differentiation. More importantly, we also find ectopic Xist expression with a significant increase in the number of cells displaying two clouds (Fig. 2e,f). This indicates that Rex1 acts as an important break on the system involved in the feedback mechanism preventing XCI of too many X chromosomes. This finding is described in the results section and the discussion of the revised manuscript.
4. Fig. 3a: the image resolution is low. H3K27me3 staining in the extraembryonic cell looks confusing in the E3.5 Rex1-/-sample. For E4.5 Rex1-/-embryo, there seems to be two foci of H3K27me3 in a cell (e.g. cells surrounding the ICM in the zoom-in panel for Rex1-/-show H3K27me3 pattern different from the WT example). It would be supportive to have Xist RNA FISH on these embryos, which should help clarify the XCI status.
We apologize for the limited quality of our Figures, which was related to the size limits of this journal. We have improved the quality of the pictures, and included panels with a larger magnification. The blastocyst confocal images in these figures are representative maximum intensity projections of a Zstack. The counting of the H3K27me3 domains has been performed by ImageJ on the separate optical sections out of the confocal stack . After careful analysis of the separate optical sections we can conclude that the different clouds belong to separate cell nuclei. To show iXCI in a more clear way, we have also performed Xist RNA-FISH on WT, Rex1 -/and Rnf12 -/-Rex1 -/blastocyst outgrowths as performed by others (Shin 2014 and Wang 2017), which show no differences with WT blastocysts outgrowths in terms of iXCI in the trophoblast cells ( Supplementary Fig. 7a,b) .

Figs. 4b & 5b
: What are Xist expression patterns in Rnf12-/-embryos and in Rnf12-/-;Rex1-/embryo? Xist RNA FISH in these embryos should provide supportive and more direct evidence for the mutants' effect on Xist and iXCI.
We have performed Xist RNA-FISH on Rnf12 -/-Rex1 -/blastocyst outgrowths and show that there are no differences with WT embryos (Supplementary Fig. 7d,e). The Xist expression pattern in Rnf12 -/embryos has been already studied by others (Bach Nature 2010).
6. Fig. 6: The "trans" repressive effect of RNF12 on REX1 for iXCI is confusing. With a trans effect, a paternal copy of RNF12 should be sufficient for REX1 degradation and iXCI in the Rnf12-/+ embryo. In the figure legend, Line 564: "This is followed by increased RNF12 expression in the developing epiblast, resulting in increased turnover of REX1, concomitant with its transcriptional downregulation, facilitating rXCI. In Rnf12-/+ and Rnf12-/-embryos, REX1 levels are stabilised in the extra-embryonic tissues, preventing Xist upregulation and iXCI." The authors should incorporate these in the illustration -the model figure is too simplified and not very helpful for understanding.
We apologize for the confusion and have adapted Figure 6 accordingly. 7. In the discussion, the authors should address the possible involvement of other factors, which are identified in the same region of Xist. Specifically for imprinted XCI, transgenic studies have been reported for factors in the X-inactivation center, which is directly responsible for the mouse iXCI.
We agree with this reviewer that, besides Rnf12 and Rex1, other factors are involved in the regulation of iXCI. Among these factors are the repressive imprint of the maternal Xist copy (as transgene and other studies have indicated), but may also include Yy1 mediated activation of the paternal copy of Xist, analogous to the role of Yy1 in regulation of rXCI. We have included this possible mechanism in our revised discussion section. Minor points: 1. Higher-resolution images should be used in the figures panel: e.g. Fig. 3a&c, Supplementary Fig.  3a, Supplementary Fig. 5a.
We did.
2. Line 105: (Fig. 2d,e) does not support "Rnf12 is required for rXCI in vivo" as the figured was referred.