A novel approach to differentiate rat embryonic stem cells in vitro reveals a role for RNF12 in activation of X chromosome inactivation

X chromosome inactivation (XCI) is a mammalian specific, developmentally regulated process relying on several mechanisms including antisense transcription, non-coding RNA-mediated silencing, and recruitment of chromatin remodeling complexes. In vitro modeling of XCI, through differentiation of embryonic stem cells (ESCs), provides a powerful tool to study the dynamics of XCI, overcoming the need for embryos, and facilitating genetic modification of key regulatory players. However, to date, robust initiation of XCI in vitro has been mostly limited to mouse pluripotent stem cells. Here, we adapted existing protocols to establish a novel monolayer differentiation protocol for rat ESCs to study XCI. We show that differentiating rat ESCs properly downregulate pluripotency factor genes, and present female specific Xist RNA accumulation and silencing of X-linked genes. We also demonstrate that RNF12 seems to be an important player in regulation of initiation of XCI in rat, acting as an Xist activator. Our work provides the basis to investigate the mechanisms directing the XCI process in a model organism different from the mouse.

In mammals, X chromosome inactivation (XCI) ensures dosage compensation of sex chromosomal genes between females (XX) and males (XY) 1,2 . The process of XCI occurs early during female embryonic development and is mediated by a multitude of epigenetic mechanisms that result in the complete transcriptional inactivation of one entire X chromosome within the nucleus of every female somatic cell. In eutherians, initiation of XCI is mediated by the long non-coding RNA Xist 3-6 . During XCI, Xist RNA spreads in cis along the entire length of the X chromosome and triggers chromosome-wide silencing of X-linked genes by recruitment of a plethora of chromatin remodelers [7][8][9][10] . The study of XCI relies both on in vivo and in vitro models that allow genetic manipulation of the factors involved, and the vast majority of our current knowledge has been achieved by using the mouse as a model organism. In vivo studies have shown that XCI starts around the 4-8 cell stage of female mouse embryonic development and is initially imprinted (iXCI), resulting in exclusive inactivation of the paternal X chromosome (Xp) [11][12][13][14] . Later in development, at the blastocyst stage (~E4.5), the Xp becomes reactivated in the inner cell mass (ICM) of the embryo, whereas iXCI persists in the extra-embryonic lineages 12,13 . Reactivation of Xp in the ICM is then followed by random inactivation (rXCI) of either the paternal or maternal X chromosome in cells of the developing epiblast. In vitro, mouse embryonic stem cells (mESCs) have been extensively used to model rXCI. In fact, undifferentiated mESCs carry two active X chromosomes and faithfully mimic the pluripotent environment of the ICM, whereas their differentiation results in random inactivation of one of the two X chromosomes. Mouse ESC-based in vitro studies have led to the discovery of the long non-coding gene Tsix, which is transcribed antisense to Xist and represents the major repressor of Xist up-regulation at the onset of XCI [15][16][17][18] . XCI is tightly linked to loss of the pluripotent state 19,20 and several pluripotency factors including NANOG, SOX2, OCT4, REX1

Results
In vitro neuronal differentiation of rESCs. In vitro differentiation of mESCs towards different functional cell types including neurons, cardiomyocytes, hepatocytes and pancreatic cells can be efficiently achieved by several established protocols 73 . Usually, differentiation strategies are based on the formation of embryoid bodies (EB) followed by growth-factor-mediated induction of early progenitor cells to differentiate into their respective lineages. Despite of the growing list of differentiation protocols for mESCs, differentiation of rESCs is extremely difficult to achieve in vitro. To date, only two strategies have been described in which rESCs were triggered to differentiate into either cardiomyocytes or neuronal precursors and in these differentiation protocols MEK and GSK3β inhibitors, that are commonly used for ESC culture, are always present in low concentrations in the differentiation media 74,75 . XCI is closely linked to loss of pluripotency, and the presence of an inactive X chromosome provides a powerful readout for cell differentiation. Several rESCs derived from different rat inbred strains were differentiated, including three pure Lewis lines (LEW) (A4p20, A9p20, A10p20), and two lines of a mixed background of dark agouti (DA) and Sprague-Dawley (SD) (135-7, 141-6). All ESC lines displayed dome shaped morphology and expressed the pluripotency factors Rex1, Prdm14, and Esrrb (Fig. 1A,B). We initially set out to assess rat XCI after inducing rESCs differentiation according to protocols, which included MEK and GSK3β inhibitors. Although female cells appeared to be morphologically differentiated into neuronal precursors, we observed enrichment of the H3K27me3 histone modification (hallmark of gene silencing upon XCI) in only 20% of the cell population, (Supplementary Fig. 1A). Xist RNA FISH analysis, detecting Xist and Tsix, revealed absence of Xist accumulation, and confirmed the presence of small pinpoint transcription signals on both X chromosomes in most cells, likely representing Tsix transcription signals, as observed in mouse ( Supplementary Fig. 1B). These www.nature.com/scientificreports www.nature.com/scientificreports/ results indicated our rESCs to be compromised to initiate XCI. We therefore set out to optimize the protocol in such a way that differentiation would include proper initiation of XCI. The absence of Xist accumulation and XCI might be related to the presence of MEK and GSK3β inhibitors, stabilizing the pluripotent state, and potentially resulting in expression of factors that repress Xist 20,[22][23][24]57 . We therefore adapted the neuronal differentiation protocol initially described by Peng et al. and Vaskova et al. 72,75 as follows: (I) both 2i inhibitors were completely eliminated starting from day 1 of neuronal differentiation, (II) the concentration of ROCK (rho-associated protein kinase) inhibitor, shown to prevent dissociation-induced apoptosis in cultured human ES cells 46,76 , was increased, (III) cells were seeded on laminin, (IV), a greater number of rESCs was used for differentiation and finally (V) differentiation cultures were serum and FGF2 free. Using these modified conditions, we were able to maintain viable differentiating male and female rESCs in the absence of 2i inhibitors (Fig. 1A). Importantly, qPCR analysis of both pluripotency and differentiation marker expression levels at different time points upon differentiation confirmed  Female resCs undergo XCI upon in vitro neuronal differentiation. We then addressed the question of whether differentiating rESCs without the supplement of 2i inhibitors would facilitate XCI. To this end, four independent female rESC lines were differentiated and the Xist RNA expression level was assessed by qPCR analysis at different time points upon neuronal differentiation. Importantly, in order to assess the sex-specific regulation of Xist RNA, one male rESC line was also included in our analysis. As in mouse, we found that Xist upregulation occurs exclusively in female rat cells between day 2 and day 4 of differentiation ( Fig. 2A). In contrast, Tsix pinpoint signals decrease upon differentiation in all differentiating rESC lines that we tested ( Fig. 2A, and Supplementary Fig. 2). Next, we addressed the dynamics of Xist expression by performing Xist RNA FISH analysis at different time points upon neuronal differentiation. In undifferentiated rESCs, Xist RNA pinpoint signals were observed within the nuclei of both female and male cells (Fig. 2B). However, since the Xist RNA FISH probe can hybridize to both Xist and Tsix RNA, the pinpoint signal might represent Tsix expression instead of Xist. Around day 2 of neuronal differentiation, Xist RNA starts to accumulate exclusively on a single X chromosome within female nuclei, whereas Xist RNA accumulation was never observed in differentiating male rESCs (Fig. 2B, lower panel). Importantly, upon differentiation of A10p20 and A4p20 rESC female lines, more than 60% of the nuclei showed an Xist RNA-coated X chromosome at day 6 of differentiation ( Fig. 2C). Taken together, these observations show that neuronal differentiation of rESCs in absence of 2i inhibitors allows Xist RNA to be upregulated and spread in cis from a single X chromosome in female cells.
In mouse, the H3K27me3 histone modification associated with gene silencing represents one of the earliest histone modifications that accumulates on the Xi during XCI [77][78][79] . Therefore, we monitored enrichment of H3K27me3 by immunofluorescence analysis upon differentiation of both male and female rESCs. In undifferentiated rESCs, no H3K27me3 domains were observed in neither male nor female cells (Fig. 3A). However, starting from day 2 of differentiation and in line with female-specific upregulation of Xist RNA, H3K27me3 started to accumulate into specific nuclear domains within female cells. By day 6, more than 60% of the female nuclei showed one H3K27me3 domain, thus confirming that XCI is efficiently initiated upon female rESCs differentiation (Fig. 3B). Finally, to precisely assess the dynamics of X-linked gene silencing, we followed the Xist-mediated inactivation of the X-linked genes Pgk1 and Rnf12 by two-colour RNA-FISH analysis at different time points upon rESCs differentiation. While the single copy of Pgk1 and Rnf12 in male cells remains actively transcribed throughout differentiation, the transcriptional inactivation of one copy of both X-linked genes in female cells starts around day 2 of differentiation ( Fig. 3C and Supplementary Fig. 3). At day 6 of differentiation inactivation of X-linked genes is reached in up to 70% of the female nuclei ( Fig. 3C and Supplementary Fig. 3).
Overexpression of RNF12 leads to Xist activation. The X-linked E3 ubiquitin ligase RNF12 has been previously shown to activate Xist transcription at the onset of XCI 25,27 . Importantly, the pluripotency factor REX1 has been identified as a key target of RFN12, and dose-dependent degradation of REX1 by RNF12 has been proposed to act as a crucial mechanism directing initiation of XCI upon differentiation of female mESCs 26 . Since the RNF12-REX1 axis represents an important pathway for XCI initiation in mouse, we asked whether these factors play similar roles in rat XCI. To this end, we transiently overexpressed Rnf12 and Rex1 in rESCs, and determined the impact of overexpression on Xist RNA regulation. Based on findings in mouse, we expected REX1 overexpression to result in the inhibition of Xist transcription whereas overexpressing RNF12 would lead to Xist up-regulation 26,27,80 . Since the catalytic ring finger domain of RNF12 shows 100% of amino acid sequence identity between mouse and rat, and human RNF12 transgenes induce ectopic XCI in mouse ESCs 25 , we overexpressed the mouse RNF12 protein (mRNF12) in rESCs. Contrarily, as the zinc finger domain of REX1 is less conserved between the two species, overexpression of REX1 was achieved by transfecting rESCs with rat REX1 cDNA (rREX1). Xist RNA expression levels were determined by qPCR analysis, and the experiment was performed in three independent undifferentiated rESC lines as well as upon neuronal differentiation (Fig. 4A,B). Overexpression of mRNF12 consistently resulted in upregulation of Xist RNA in both male and female rESCs prior to and during rESCs differentiation, thus indicating RNF12 to act as important trans-acting activator of Xist in rat (Fig. 4A,B). Xist RNA FISH analysis performed at day 2 of neuronal differentiation upon mRNF12 overexpression further confirmed the impact of RNF12 on XCI initiation (Fig. 4C). However, the impact of rREX1 overexpression on Xist regulation, prior to and upon differentiation of rESCs, appeared less consistent (Fig. 4A-C). Although RNA-FISH and qPCR analysis indicated a decrease in Xist expression when rREX1 is overexpressed for one rESC line at day 2 of neuronal differentiation (line 135-7), Xist down-regulation was not significant for most comparisons. These results may suggest that Rex1 plays a less prominent role in XCI in rat, but could also be explained by our experimental setup, as previously observed in mouse ESCs 26 , where upregulation of Xist mediated by Rnf12 over-expression was also easier to detect than Rex1 over-expression mediated down-regulation of Xist. Although the role of Rex1 in XCI in rat needs further investigation, our findings indicate a role for Rnf12 in XCI in rat, providing a powerful new model system to elucidate the complex mechanisms directing initiation of XCI. the two groups (2 pinpoints versus cloud with or without pinpoint) within each differentiation time point. P-values are provided: p-value A10-d0 < 0.0001, p-value A10-d2 = 0.02, p-value A10-d3 = 0.03, p-value A10-d4 = 0.03 and p-value A4-d0 < 0.0001, p-value A4-d2 = 0.03, p-value A4-d4 = 0.05, p-value A4-d6 = 0.03. (2019) 9:6068 | https://doi.org/10.1038/s41598-019-42246-2 www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/

Discussion
Our knowledge concerning the regulation of XCI in developing rat embryos is limited and relies on conservation of the key regulators Xist and Tsix between mouse and rat and a few studies in which, similar to mouse, iXCI has been proposed to occur in early rat embryonic development 30,[37][38][39][40] . Studying the XCI process in rESCs offers the opportunity to explore species-specific epigenetic features and will help to reach a more comprehensive understanding of the XCI process in mammals. Although rESCs in vitro differentiation protocols have been previously established 72,74,75 , XCI studies were limited, only examining Xist expression and accumulation of chromatin modifications. Here, we confirm the accumulation of XCI-related epigenetic features in rESCs during differentiation according to the previously established protocols, and further show that this is accompanied by proper down-regulation of pluripotency markers 81 . We also showed that transcriptional inactivation of X-linked genes directly follows Xist RNA accumulation on one of the two X chromosomes. In fact, the exclusive enrichment of H3K27me3 loci in female nuclei starts around day 3 of neuronal differentiation, and Xist-mediated silencing of Pgk1 and Rnf12 occurs with similar dynamics.
Overexpression of mRNF12 protein in rESCs efficiently recapitulates RNF12 function in mouse as an activator of Xist expression. RNF12 is highly conserved among mammals 82 . The observed up-regulation of rat Xist upon mRNF12 overexpression is in line with our previous findings that overexpression of human RNF12 leads to ectopic Xist expression and XCI in mESCs 25 , indicating that the role of RNF12 in XCI is highly conserved in mammals. In contrast, rREX1 over-expression does not lead to consistent down-regulation of Xist levels as was expected based on our mouse studies. This result may indicate the presence of an alternative pathway by which RNF12 activates XCI in rat, a theory that is supported by the lower level of conservation in the REX1 zinc finger domain. On the contrary, REX1 represents the only target of RNF12 in mouse ESCs identified so far. In addition, measuring down-regulation of Xist expression in the present setting after transient transfection is technically more challenging than detecting up-regulation of Xist, and may require more sensitive approaches to experimentally define the exact role of Rex1 in XCI in rat. Additional alternative approaches, such as inducible over-expression of Rex1, and ChIP sequencing studies should shed more light on the exact role of Rex1 in XCI regulation.
In conclusion, we were able to set up a robust in vitro system to study the regulation of XCI in differentiating rESCs and our results suggest that the main steps of XCI in our rat in vitro system are highly similar to those of mouse XCI. The generation of hybrid F1 polymorphic rESCs together with the application of the CRISPR/Cas9 technology for genomic editing to the rat system will increase the use of rat as a model organism in basic epigenetic and biomedical research.

Methods
Cell culture and DNA transfection. rESCs were derived as previously described 56 and subsequently maintained in N2B27 medium supplemented with 3 μM CHIR99021 (Stemgent), 1 μM PD0325901 and 1000 U/ml mouse LIF on mouse feeders.
For monolayer differentiation culture plates were coated with 100 μg/ml laminin (Sigma-Aldrich) for at least 4 hours at 37 °C, followed by three PBS washes. Single rESCs were plated at a density of 10 5 /cm 2 for the female cell lines and 2 × 10 4 /cm 2 for the male cell lines in N2B27 supplemented with 10 μM of ROCK inhibitor (Sigma-Aldrich) for the first three days. Thereafter, the ROCK inhibitor was eliminated. Medium was refreshed daily.
For overexpression experiments, the mRex1, rRex1 and mRnf12 coding sequences were subcloned into pCAG-Flag, a CAG-driven expression vector containing a Flag-tag. RESCs were transfected using lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions, followed by 48 hours of puromycin selection (1,5 μg/ml). Overexpression in differentiating cells was performed as follows: rESCs were trypsinised and plated at a density of 1,3*10 5 /cm 2 in gelatinized 6-well plates in 2i media supplemented with 10 μM of ROCK inhibitor (Sigma-Aldrich) without feeders. The next day cells were transfected using lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Cells were left to recover in 2i media overnight and then a 48 hour-puromycin selection (0,25 μg/ml for the A10p20, A4p20 and A8p20 cell lines and 1 μg/ml for the 135-7 cell line) was initiated in N2B27 differentiation media.

probe preparation and Fluorescent in Situ Hybridization (FISH). For preparing probes detecting
Xist, Pgk1 and Rnf12 mRNAs, BACs harboring these genes were labelled as a whole, with digoxigenin and biotin (Roche) respectively, by nick translation following the manufacturer's instructions.
Immunocytochemistry. For immunofluorescence analysis of different time points of neuronal differentiation and of overexpression experiments, cells were grown on glass coverslips and then fixed with 3% PFA for 10 minutes at room temperature followed by three washes in PBS (3 × 5′). Thereafter, cells were permeabilised with 0.5% Triton, washed with PBS (3 × 5′) and blocked with 2% BSA, 5% donkey serum in PBS (blocking solution) for 30 minutes at room temperature. This was followed by anti-H3K27me3 rabbit (Diagenode, 1:500) incubation, diluted in blocking solution, at 4 °C overnight in a humid chamber. The next day, slides were washed in PBS (3 × 5′) and blocked with donkey anti-rabbit alexa fluor 488 (ThermoFischer Scientific, 1:250) secondary antibody, diluted in blocking solution for 1 hour at room temperature in a humid chamber. Slides were then washed in PBS (3 × 5′) and mounted with ProLong ® Gold Antifade Mountant with Dapi (ThermoFisher Scientific). Confocal imaging was performed on a Zeiss LSM700 microscope (Carl Zeiss, Jena). statistical analysis. Statistical significance between the different groups was assessed by student t-test.
Statistical significance was set at p < 0.05.