Epigenetic regulation of human SOX3 gene expression during early phases of neural differentiation of NT2/D1 cells

Sox3/SOX3 is one of the earliest neural markers in vertebrates. Together with the Sox1/SOX1 and Sox2/SOX2 genes it is implicated in the regulation of stem cell identity. In the present study, we performed the first analysis of epigenetic mechanisms (DNA methylation and histone marks) involved in the regulation of the human SOX3 gene expression during RA-induced neural differentiation of NT2/D1 cells. We show that the promoter of the human SOX3 gene is extremely hypomethylated both in undifferentiated NT2/D1 cells and during the early phases of RA-induced neural differentiation. By employing chromatin immunoprecipitation, we analyze several histone modifications across different regions of the SOX3 gene and their dynamics following initiation of differentiation. In the same timeframe we investigate profiles of selected histone marks on the promoters of human SOX1 and SOX2 genes. We demonstrate differences in histone signatures of SOX1, SOX2 and SOX3 genes. Considering the importance of SOXB1 genes in the process of neural differentiation, the present study contributes to a better understanding of epigenetic mechanisms implicated in the regulation of pluripotency maintenance and commitment towards the neural lineage.


Introduction
SOX3/Sox3 is an X-linked member of SOXB1 (SOX1-3) subfamily of transcriptional regulators [1][2][3]. Together with SOX1 and SOX2 it is expressed in neural progenitors where they counteract the activity of proneural proteins and maintain undifferentiated state of progenitor cells [4]. SOX2 gene, the closest relative of SOX3, is one of the core pluripotency factors involved in the regulation of stemness and differentiation [3,5,6]. SOX3 is recognized as one of the earliest neural markers in vertebrates; up to date the role of Sox3 in neural development has been the most studied aspect of the Sox3 action. It was shown that in murine telencephalon Sox3 is expressed in neural stem/progenitor cells (NP cells) during embryonic development and it is downregulated during neuronal differentiation [7]. In adult mice telencephalon, Sox3 expression is maintained only in progenitor cells of the adult neurogenic regions, subventricular and subgranular zones [7]. In contrast, during hypothalamic neurogenesis Sox3 expression is not restricted to neural progenitors, but to developing neurons and is maintained in a subset of differentiated hypothalamic cells through adulthood [7]. Consistent with its expression patterns, Sox3 plays important roles in the process of neural differentiation, as confirmed by genome-wide binding studies that verified its status as one of the earliest markers of vertebrate neurogenesis. It has been demonstrated that in mouse ES-derived NP cells Sox3 target genes have regulatory roles during development of the CNS [1]. While Sox3 mainly activates genes expressed in NP cells, it also binds to neuronal genes, preventing premature Sox11 binding and their consequent activation [1]. Recent studies have identified Sox3 target sites in murine NP cells in putative enhancers of neurodevelopmental genes, located primarily within the intergenic regions [8]. Furthermore, Sox3 acts as a pioneer factor whose binding to target enhancers establishes local epigenetic changes [1]. Due to functional redundancy between SoxB1 genes the expression of most NP genes is not affected in Sox3 null NP cells. Nevertheless, direct Sox3 targets have been identified with expression not rescued by other SoxB1 members [9].
Besides the prominent roles in the process of neural differentiation, there is evidence pointing at SOX3 as one of the players in the maintenance of human embryonal stem cells (hESCs) identity. Together with SOX2, SOX3 is implicated in the regulation of self-renewal and pluripotency of hESCs [10]. SOX3 is upregulated after the knockdown of SOX2 in hESC, keeping the cells in an undifferentiated state, while the self-renewal ability is reduced under these conditions [10]. Moreover, it was established that Sox1 and Sox3 can replace Sox2 during the process of iPSCs (induced pluripotent stem cells) generation from mouse embryonic fibroblasts (mEFs) [11]. Taken together, these data highlight the role of Sox3 in the selection and proper execution of developmental programs established through complex coordination between Sox3 and other SoxB1 genes and their partners.
Reports concerning the mechanisms of SOX3 regulation during neural differentiation are limited and mainly focused on the transcriptional control of human SOX3 expression [1,[12][13][14][15][16][17]. In recent years, it was revealed that regulation of developmental genes with dynamic expression patterns is not driven only by transcription factor networks, but also by the epigenome (reviewed in [18,19]). Epigenetic regulation of gene expression is achieved through genomic DNA methylation, post-translational modifications (PTMs) of histones, chromatin remodeling and non-coding RNAs [19]. The complex interplay between these mechanisms represents a mode in which genotype controls phenotype without changes in the DNA sequence. Special efforts are made in an attempt to delineate epigenetic processes underlining the formation of neurons, with an aim to improve stem cell based therapies in neurodegenerative diseases, and to control commitment of pluripotent cells [20]. Epigenetic profiles of pluripotency-associated genes, such as Oct4, Sox2 and Nanog have been investigated in several studies, and correlated with dynamic expression of these genes during development [21][22][23] while epigenetic control of SOX3 expression remained to great extent understudied.
In the present study we have analyzed epigenetic modifications of the promoters of SOXB1 genes during early phases of retinoic acid (RA) induced neural differentiation of human embryonal carcinoma NTera2/D1 (NT2/D1) cells with a special focus on SOX3 gene. We demonstrate that the human SOX3 gene promoter is extremely hypomethylated both in undifferentiated and RA-induced NT2/D1 cells and that it does not react to treatment with the demethylating agent 5-azacytidine (5-azaC). Furthermore, by employing chromatin immunoprecipitation (ChIP), we show that different regions of SOX3 gene are enriched in distinct histone PTMs switching throughout the course of RA-induced neural differentiation. The profiles of histone modifications on the SOX3 promoter differ from those on the SOX1 and SOX2 promoters, implying that SOXB1 genes are controlled by different epigenetic mechanisms.

5'-AAGCTGGGCCTGAAAGACTC -3' (forward)
All samples were measured in triplicate and the mean value was considered. The relative levels of SOX3, SOX2 and Nanog expression were determined using a comparative quantification algorithm where the resulting ΔΔCt value was incorporated to determine the fold difference in expression (2 −ΔΔCt ). Relative SOX3, SOX2 and Nanog mRNA levels were presented as a fold change in gene expression normalized to GAPDH and relative to the value in untreated NT2/D1 cells, which was set as 1. The expression level of Dbx1 was analyzed as 2 −ΔCt due to the low levels of Dbx1 mRNA in undifferentiated NT2/D1 cells.

RT-PCR analysis
Total RNA was isolated, subjected to DNaseI treatment and reversely transcribed as described in previous subsection. The synthetised cDNAs were used as templates for PCR amplifications with primers specific for SOX1 and GAPDH (as in previous subsection) as a loading control. Primers for SOX1 amplification were as follows: RT-PCRs were performed in 20 ul reactions using KAPA 2G Fast HotStart Ready Mix (Kapa Biosystems) according to manufacturer's protocol. Obtained products were separated electrophoretically on 2% agarose gel and visualized using ethidium bromide staining. Quantification of obtained bands was performed using ImageJ software (https://imagej.nih.gov/ij/).

Methylation-specific PCR
For MSP analyses, direct sodium-bisulfite conversion of uninduced NT2/D1 cells and cells induced with RA for 2, 4 and 7 days was performed using EZ DNA Methylation-Direct™ Kit (Zymo Research Corporation, CA, USA). For each conversion 10 5 cells were used and protocol provided by the manufacturer strictly followed. Upon conversion the reaction recovery rate was considered as 100% and hence concentration of DNA was not measured. Converted DNA samples were used as templates for PCR amplification using KAPA 2G Fast HotStart Ready Mix (Kapa Biosystems) and cycling conditions 95˚C for 5 minutes; 95˚C for 30 seconds, 60˚C for 20 seconds, 72˚C for 2 minutes, for 35 cycles. MSP primers were designed using MethPri- Obtained products were separated electrophoretically on 2% agarose gel and visualized using ethidium bromide staining.
Sodium-bisulfite conversion of isolated genomic DNA was performed using EZ DNA Methylation-Lightning™ Kit (Zymo Research Corporation, CA, USA). For each conversion 2 μg of DNA was used and protocol provided by the manufacturer strictly followed. Upon conversion the reaction recovery rate was considered as 100% and hence concentration of DNA was not measured.

PCR amplification of bisulfite converted SOX3 promoter
PCRs were performed in 60 μl reactions using KAPA 2G Fast HotStart Ready Mix (Kapa Biosystems) according to the manufacturer's protocol. Bisulfite converted DNA from uninduced NT2/D1 cells and cells induced for 2, 4 and 7 days with RA were used as a template to amplify 2nd CpG island within SOX3 promoter. PCR cycling conditions were 95˚C for 10 minutes; 95˚C for 20 seconds, 50-55˚C for 20 seconds, 60˚C for 2 minutes, for 40 cycles. PCR reactions were performed using non-modified forward primer and 5'-biotin-labeled reverse primer listed below. BSP primers were designed using MethPrimer tool [25]. Indicated primers positions were determined relative to TSS.

Pyrosequencing
Pyrosequencing of amplified biotinylated PCR products was performed by commercial service at Queen Mary University of London, Genome Centre, London on Pyromark MD system, Biotage (Qiagen Pyrosequencing for 2 nd CpG island was performed with primers: In the obtained pyrograms, the amount of C relative to the sum of the amounts of C and T at each CpG site is calculated as percentage of methylation level of indicated CpG site. Total methylation levels of 2 nd CpG island within SOX3 promoter were calculated as average values of every individual CpG sites methylation levels.

Statistics
Data were presented as mean ± S.D. and were analyzed using Student's T test. Ã P<0.05 was considered statistically significant.

Results
Analysis of human SOX3 gene expression during early phases of neural differentiation of NT2/D1 cells SOX3 is considered as one of the key regulators of neural development in vertebrates [27]. We analyzed its expression in the early stage of RA-induced neural differentiation of NT2/ D1 cells. Due to their similarity with hESC and the property to differentiate into morphologically and physiologically mature neurons after exposure to RA, NT2/D1 cells represent an appropriate in vitro model to study the process of human neural differentiation [24,28,29]. Although derived from human teratocarcinoma, these cells display properties of neural progenitor cells [30]. Following induction with RA, NT2/D1 cells loose expression of neuroepithelial markers and acquire expression of neuronal markers, yielding NT2N neurons that maintain a stable neuronal phenotype, form functional synapses, do not divide, and therefore have been used in various studies as an alternative graft sources in transplantation therapy for ischemia [30,31]. Moreover, in our previous studies and reports made by other groups, NT2/D1 cell line was used for the analyses of human SOX genes expression and regulation [12][13][14][15]17,23,[32][33][34][35][36][37]. We have previously shown that the treatment of NT2/D1 cells with RA triggered an early (48h) increase in the expression of SOX3 gene [35]. In the present study, we have expanded this analysis by following SOX3 expression in longer time points (4 and 7 days). By qRT-PCR and Western blot, we demonstrate significant upregulation of SOX3 after exposure of NT2/D1 cells to RA, at the mRNA ( Fig 1A) and protein levels ( Fig  1B). SOX3 shows a peak at 2 days of RA treatment, followed by decrease in protein and mRNA levels (Fig 1A and 1B).
To verify pluripotency of NT2/D1 cells and confirm the differentiation status after RA induction, we analyzed the expression of Nanog, one of the core pluripotency factors [38]. We also analyzed the expression of Dbx1, a homeodomain protein with a critical role in the establishment of V0 and V1 interneurons [39] and a direct target of Sox3 in neural precursors [9]. We observed a significant decrease in Nanog mRNA levels 4 days following RA treatment, confirming the exit of NT2/D1 cells from the pluripotency (Fig 1C). At day 4, the cells show upregulated expression levels of Dbx1 ( Fig 1C). Thus, the expression patterns of SOX3, Nanog and Dbx1 confirm the onset of neural fate commitment.
In order to further examine the temporal pattern of SOX3 protein expression during early phases of neural differentiation in our model system, we analyzed its co-expression with stemness marker OCT4 by immunocytochemistry. We have previously demonstrated that OCT4, together with Nanog, is downregulated during the first week of neural differentiation of NT2/D1 cells [34]. Herein, we confirmed that strong OCT4 immunoreactivity detected in undifferentiated NT2/D1 cells (Fig 2A2 and 2A4) declined in most of the cells following 2 (Fig 2B2 and 2B4) and 4 days (Fig 2C2 and 2C4), but almost completely disappeared after 7 days of RA induction (Fig 2D2 and 2D4). As shown in Fig 2, we detected a low level of SOX3 protein expression in NT2/D1 cells prior to the differentiation (Fig 2A1 and  2A4) that sharply increased after 2 days of RA treatment. Induction of SOX3 in most of the cells coincided with reduced level of OCT4 protein expression suggesting that these two transcription factors are oppositely regulated at initial periods of neural differentiation (Fig  2, arrowheads in B1, B2, B3, B4). However, as differentiation proceeded, only populations of cells which retained medium OCT4 protein expression level were also immunoreactive for SOX3 (Fig 2, arrows in C1, C2, C3, C4). Finally, at the end of the differentiation protocol, at day 7, expression level of SOX3 was barely detectable (Fig 2D1 and 2D4). Taken together, our results demonstrated that SOX3 protein is transiently expressed in specific developmental stage of NT2/D1 cells during initiating phases of in vitro neural differentiation. This prompted us to further investigate possible epigenetic mechanisms implicated in the regulation of SOX3 expression. In silico analysis of human SOX3 promoter Functionally, the promoter of human SOX3 gene is 713 bp in length [12] and we analyzed it with the MethPrimer tool [25]. The analysis was conducted using default criteria for CpG island prediction, and it has revealed the presence of 2 CpG islands within the human SOX3 promoter (Fig 3A). The island with 17 CpGs dinucleotides (from 209-423 bps) corresponds approximately to the minimal SOX3 promoter [12], while the second CpG island with 22 CpGs (434-657 bps) encompasses the SOX3 TSS ending downstream of the second ATG codon (Fig 3A) [12]. The methylation profile of SOX3 promoter during early phases of neural differentiation We sought to determine potential dynamic changes in the methylation levels of these CpG islands during the early phases of neural differentiation of NT2/D1 cells. First, we have employed methylation-specific PCR (MSP) to assess methylation status of the human SOX3 promoter in undifferentiated NT2/D1 cells and cells treated with RA for 2, 4 and 7. While the 1 st CpG island could not be subjected to MSP analysis due to the sequence-based difficulties in MSP primers design, 2 nd CpG island was successfully amplified. As shown in Fig 3B, products obtained with primer set corresponding to unmethylated DNA were highly abundant, in comparison with products obtained with primer set corresponding to methylated DNA. These results suggested that second CpG island within the SOX3 promoter is unmethylated in undifferentiated NT2/D1 cells and that this hypomethylation is sustained during RA-induction of these cells. In order to elucidate these findings in more detail, we employed bisulfite pyrosequencing.
We found 1 st CpG island to be difficult for the analysis due to GC-rich sequence which could give rise to homopolymers and secondary structures following bisulfite conversion of DNA. These structures prevent optimal amplification and subsequent pyrosequencing of the region encompassing 1 st CpG island within the SOX3 promoter and therefore we proceeded with the analysis of the second island. Genomic DNA isolated from undifferentiated NT2/D1 cells and cells treated with RA for 2, 4 and 7 days were subjected to the sodium-bisulfite conversion. Converted DNA was used as a template to amplify 2 nd CpG island within SOX3 promoter using non-modified forward primer and biotin-labeled reverse primer and PCR products were analyzed by pyrosequencing [40]. We demonstrated hypomethylation of the SOX3 promoter in undifferentiated NT2/D1 cells (Fig 3C), which is preserved during RAinduced neural differentiation (Fig 3C). Average methylation of SOX3 does not exceed 10% at all timepoints analyzed, as shown in Fig 3C. The data suggest that methylation of SOX3 does not correlate with the dynamic changes of SOX3 expression during the initial phases of neural differentiation.
In order to confirm these data, we treated NT2/D1 cells with 5-azaC, a demethylating agent acting during DNA replication and cell division [41,42] followed by the analysis of the SOX3 expression levels (Fig 3D). Endogenous caspase-3, previously shown to be upregulated following treatment of NT2/D1 cells with nucleoside drugs [43], was used as a positive control ( Fig  3D). As expected, treatment with 5-azaC did not induce any significant change in SOX3 protein levels (Fig 3D). The lack of upregulation of the SOX3 protein expression upon exposure to 5-azaC further supports the idea that methylation is not a mechanism governing RA-induced activation of SOX3 expression in NT2/D1 cells.
Histone modifications profiles on SOX3 gene during early phases of RAinduced neural differentiation of NT2/D1 cells Numerous reports highlighted the importance of histone covalent PTMs in the control of nucleosome dynamics during the process of differentiation induced by the environmental stimuli [44]. In order to check the status of histone modifications in the SOX3 gene in pluripotent cells, we examined ChIP-Seq reads in human ES cell line H1 (H1-hESC) using ENCODE datasets (Fig 4) (GRCh37/hg19; http://genome.ucsc.edu). ChIP-Seq signals for H3K4me2 and H3K4me3 in the promoter and the SOX3 coding region are high (Fig 4). Other markers are absent from the SOX3 promoter (H3K9ac and H3K27ac) and coding region (H3K36me3 and H4K20me1), and H3K79m2 is low within the coding region (Fig 4). Furthermore, H3K27me3 overlaps with the 3' region, but it is not present at the SOX3 promoter (Fig 4). These histone signatures indicate permissive promoter with low transcriptional activity in pluripotent cells, consistent with low expression levels of SOX3 in hESCs [10].
In order to investigate histone modifications on SOX3 during neural differentiation of NT2/D1, we employed ChIP using a range of specific antibodies, monitoring by qPCR three regions: a region~600bp upstream of the TSS (SOX3 upstream), the SOX3 core promoter and a region within SOX3 gene,~550bp downstream from TSS (SOX3 downstream) (Fig 5A). We detected enrichment of H3K4me3 on the core promoter region (Fig 5B). Upon RA-induction the level of H3K4me3 increased 1.6 times on the SOX3 core promoter within the first 2 days of RA induction, followed by a statistically significant drop at days 4 and 7 (Fig 5B). This profile mirrors the expression levels of the SOX3 gene (Fig 1A and 1B), providing evidence of a link between H3K4me3 and SOX3 transcription. Moreover, the region upstream of the SOX3 TSS is undergoing similar, but less prominent changes in H3K4me3 levels during the course of differentiation ( Fig 5B).
Next, we investigated profiles of H2B acetylation of lysine residues 5, 16 and 120 (Fig 5C-5E). It has been shown that acetylation of histones neutralizes lysine charges, thus affecting nucleosome stability and promoting DNA accessibility and transcriptional activation [45,46]. We demonstrate a statistically significant decrease in the levels of H2BK16ac, H2BK120ac and H2BK5ac on the core promoter of the SOX3 gene throughout the course of RA induction of Epigenetic regulation of human SOX3 gene NT2/D1 cells (Fig 5C-5E). Surprisingly, we did not observe an increase in H2B acetylation 2 days after RA introduction, the time point corresponding to the highest SOX3 promoter activity. We analyzed the region upstream of SOX3 TSS and observed changes in the H2B acetylation levels (Fig 5C-5E) correlating with the expression profile of SOX3 (Fig 1A and 1B).
In the SOX3 coding region, we detected a profile of H3K4me3 similar to the core promoter with more prominent increase on day 2 of RA-induction (Fig 5B). We speculate this increase is a reflection of equivalent changes in this mark on the promoter throughout the course of early differentiation. It has been demonstrated that activated transcription affects the levels of H3K4me3 in coding regions of genes and it is accompanied by the shift in the distribution of this mark on the promoter and coding region [47]. Moreover, the profiles of H2B acetylation in the coding region are similar to the ones in the core promoter, with lower abundance of all three marks at days 4 and 7 of RA-induction (Fig 5C-5E). Finally, in the coding region of SOX3 gene we analyzed the pattern of H3K79me2, a mark associated with elongating RNA Pol II [48,49], and observed an increase after the induction of neural differentiation (Fig 5F). These findings are consistent with the existing data indicating that this modification is enriched in exons and drives the elongation phase of the transcription [48]. Interestingly, the levels of H3K79me2 remain significantly higher, 4-fold approximately, even at day 7 of RAinduction, a time point with the lowest SOX3 expression.
Analysis of human SOX1 gene expression, methylation status and histone modifications profiles during the early phases of neural differentiation of NT2/D1 cells We analyzed the expression pattern of SOX1/SOX1 mRNA and protein levels in differentiating NT2/D1 cells. As shown in Fig 6, we detected low levels of SOX1 mRNA (Fig 6B) and SOX1 protein (Fig 6C) in uninduced NT2/D1 cells. At day 4 of RA treatment significant increase in SOX1/SOX1 mRNA and protein levels was observed and remained high in the day 7 of neural differentiation. This is in line with other studies which specified Sox1/SOX1 as an early responder to neural inducing signals and one of the early markers of neural induction [50][51][52]. Regarding the methylation status of SOX1 gene, data are limited to various types of cancers [53][54][55], while for neural differentiation experimental data are lacking. Thus, we proceeded with the MSP analysis of the methylation status of SOX1 gene promoter during the neural differentiation of NT2/D1 cells. We detected products obtained with primer set corresponding to unmethylated DNA, while the products obtained with primer set corresponding to methylated DNA were absent, as shown in Fig 6A. These results indicated that SOX1 promoter has a low methylation level in both undifferentiated NT2/D1 cells, as well as during the following days of RA-induced neural differentiation, similar to methylation profiles obtained for SOX3 promoter.
The analyses of histone PTMs revealed increase in H3K4me3 abundance on the promoter of SOX1 at day 2 of RA induction (Fig 7B). In following days of RA treatment we observed slight drop in the enrichment of H3K4me3, indicating that this histone mark is not in correlation with the detected transcriptional activation of SOX1. As for H2B modifications, prominent decline in H2BK16ac, H2BK120ac and H2BK5ac levels accompanied RA-induced neural differentiation of the cells (Fig 7B). These data suggest that selected H2B acetyl marks do not contribute to the enhanced SOX1 expression following RA induction.

Analysis of human SOX2 gene expression and histone modifications profiles during the early phases of neural differentiation of NT2/D1 cells
We analyzed the expression pattern of SOX2/SOX2 mRNA and protein levels in differentiating NT2/D1 cells. As shown in Fig 8, we detected a significant drop in SOX2/SOX2 mRNA and protein levels 2 days after exposure to RA, with no further decline in the following days ( Fig  8A and 8B). This is consistent with the exit from pluripotency and activation of the neural program [35]. It also coincides with the increase in SOX3 expression, indicating that SOX2 and SOX3 are differentially regulated. Numerous studies have demonstrated that the regulatory regions of SOX2 are nonmethylated during the course of neural differentiation of embryonal carcinoma cells [23,56] suggesting that the changes of SOX2 expression during neural differentiation are independent of DNA methylation.
As for histone PTMs, we detected high level of H3K4me3 on the human SOX2 promoter, consistent with the previous studies performed in mouse and human ESCs [57][58][59], as well as high levels of H2BK16ac, H2BK120ac and H2BK5ac. Following treatment of NT2/D1 cells with RA, we observed a slight decrease in H3K4me3, H2BK16ac and H2BK120ac (Fig 9). The most prominent change was in H2BK5ac, which declined on day 2 of differentiation (Fig 9). This coincides with the decrease of SOX2 expression (Fig 8), suggesting that deacetylation of H2BK5 is one of the marks associated to the response of the SOX2 promoter to the RA. The SOX2 promoter is showing an opposite profile of H2BK5ac compared to SOX3 at day 2, which implies that these two genes have different epigenetic regulation (Figs 5 and 9).

Bioinformatic analyses of human SOXB1 genes expression and histone modifications profiles in H1-ESCs and H1-derived neural progenitors
In order to assess expression and histone PTMs profiles of SOXB1 genes in another model system, we performed bioinformatics analysis of genome-wide data available from http://www. roadmapepigenomics.org/ for H1-ESCs and H1-derived neural progenitors. Data are presented in S1 Fig. Furthermore, expression data obtained in the present study are in line with RNA-Seq analyses of expression of SoxB1 mRNAs during RA-induced neuronal differentiation of mESCs [60].

Discussion
Despite the growing amount of data regarding the regulation of the human SOX3 gene expression and activity in pluripotent cells and their differentiated progeny, reports concerning the epigenetic mechanisms of its regulation during the process of differentiation are limited. Lindeman et al. revealed that regulatory sequences of the Sox3 zebrafish orthologue are methylated and low in positive histone PTMs in somatic cells, compared to midblastula transition (MBT) cells, characterized by high expression of Sox3, lack of DNA methylation and positive histone PTMs [61]. Epigenetic regulation of Sox3 has also been discussed as part of study by Azuara et al. [59] showing that markers of active and repressive chromatin are simultaneously present at silent tissue-specific genes, such as Sox3, in mESCs. Finally, hypermethylation of SOX3 was detected in chronic lymphocyte leukemia [62].
Embryonal carcinoma cell line NT2/D1 provided us with a valuable model system to study the epigenetic state of SOX3 gene in pluripotent cells as well as in cells responding to neural differentiation-inducing stimuli. Molecular events during RA induced differentiation of NT2/ D1 cells reflect the steps in the development of human brain [63]. We demonstrated that the initial phases of RA treatment are critical for the transcriptional activity of SOX3 [17,32,35]. Two days after RA introduction, we detected a significant increase in SOX3 mRNA and protein, thereafter, downregulation ensued. This profile resembles the one previously detected in the chicken and mouse neuroepithelium during early phases of neural induction, with transient upregulation of Sox3 gene [64][65][66].
Despite low SOX3 expression levels in undifferentiated cells, the SOX3 promoter is hypomethylated. This hypomethylation persists through neural differentiation induced by RA.  Similar results were obtained for the SOX1 promoter, which is also hypomethylated during the first 7 days of neural differentiation of NT2/D1 cells. While methylation of CpG-rich promoters is considered a hallmark of silent genes, nonmethylated promoters are not necessarily associated with active transcription [67,68]. Rather, hypomethylation allows a permissive platform on which other mechanisms mediating the recruitment of different transcription factors act [67]. Some of the germline-and pluripotency-associated genes with hypomethylated promoters, and low expression levels in stem cells, acquire methylation during lineage commitment [69]. Moreover, the regulatory regions of SOX2 which are unmethylated in undifferentiated NT2/D1 cells become methylated in neurons, with no detectable SOX2 expression [23]. At the same time, differentiation-induced repression of SOX2 is paralleled by shift from H3K4me3 to colocalization of both H3K4me3 and H3K27me3, as demonstrated in ES cells [58].
We demonstrate that H3K4me3 on the core promoter and upstream of TSS is associated with transcriptional induction of SOX3 gene. This finding is supported by the study by Lauberth et al. showing that H3K4me3 facilitates TFIID recruitment and enhances preinitiation compex (PIC) assembly, thus promoting transcription [70]. Many promoters in ES cells are positive for both H3K27me3 and H3K4me3 and there is evidence indicating that SOX3 and SOX1 are among the genes with bivalent promoters. In the study by Mikkelsen et al., Sox3 was one of the genes with bivalent promoter in mES cells, resolved into H3K4me3 + in lineage committed NP cells, while in mouse embryonic fibroblasts (MEFs) it is devoid of any histone marks [57]. In the same study, Sox1 was found to be bivalent in both mES and NP cells, and H3K4me3 -/H3K27me3 + in MEFs [57]. Furthermore, in the study by Pan et al. SOX1 was found to be marked by both H3K4 and H3K27 trimethylation during RA-and BMP4-induced differentiation of ES cells [58]. This state of Sox3 and Sox1 is typical for high CpG content promoters of genes with key roles in the development [57]. Further ChIP analyses are necessary in order to test this hypothesis.
Interestingly, activation of SOX3 at day 2 was not accompanied by an increase in H2B acetylation. There is limited literature data regarding the contribution of H2B acetylation in regulation of transcriptional regulation: a study by Kurdistani et al. demonstrated negative correlation of H2B acetyl marks with transcription in Saccharomyces cerevisiae [71]. Analyses of histone modification patterns in yeast suggested that acetylation of H2A and H2B has cumulative transcriptional effects together with H3 and H4 acetylations [72,73]. The lack of increase in H2B acetylation at the time point with the highest SOX3 expression implies that upregulation of the SOX3 gene upon RA treatment is associated to other histone marks. Nevertheless, substantial depletion of all three H2B acetyl marks on the promoter of SOX3 at day 7 coincides with the reduced expression of SOX3 gene. These findings suggest that H2BK5ac, H2BK16ac and H2BK120ac at the core promoter of SOX3 gene contribute to the regulation of SOX3 gene expression, primarily at the end of the first week of neural differentiation of NT2/ D1 cells, when they interact coordinately with other deposited histone marks and transcription apparatus or transcription factors. It would be interesting to evaluate chromatin marks on other potential regulatory regions of SOX3 gene. For example, recent analysis has revealed that distal enhancers of numerous genes are targeted by NF-Y, a transcriptional regulator of SOX3 gene and transcription factor with nucleosome-like properties that mediates establishment of permissive chromatin modifications [14,74,75]. Further analysis could reveal the existence and function of corresponding elements at distal regions of human SOX3 gene.
One of the interesting findings emerging from this study is that SOXB1 genes are differentially regulated in the initial phases of neural differentiation. Based on the presented data, early phase of neural differentiation of NT2/D1 cells could be divided into two stages. First stage falls within the first 2 days of RA induction and is characterized by the increase in SOX3 and reduction in SOX2 expression levels, whereas SOX1 remains unchanged. Second stage (days 4 and 7) is accompanied by the reduction in both SOX2 and SOX3 levels, while SOX1 is upregulated. In the first stage, H3K4me3 profile of the SOX2 promoter is not strongly correlated with the decline in the expression of SOX2, as opposed to SOX3 whose upregulation at day 2 is likely operated by H3K4me3. This is in concordance with the findings by Barrand et al. indicating that H3K4me3 is associated with SOX2 promoter in different cell types, regardless of SOX2 expression [56]. Moreover, H2BK5ac abundance at day 2 of RA induction is considerably lower for SOX2 than for SOX3 promoter, providing an additional proof of different states of SOX2 and SOX3 promoters in cells undergoing differentiation. As for SOX1 promoter, we detected raise in H3K4me3 at day 2, preceeding the actual increase in SOX1 expression levels at day 4 and 7 of RA induction.
In the second stage, at day 7 we detected significant decrease in all analyzed histone PTMs at promoters of SOX1, SOX2 and SOX3 genes. These implies that same epigenetic mechanisms could contribute to the observed reduction in SOX2 and SOX3 expression levels at later stages of neural differentiation. Conversely, detected decrease in H3K4me3 and H2B acetyl marks at SOX1 promoter suggests that these activating histone modifications do not have considerable role in transcriptional activation of SOX1. Therefore, other mechanisms such as a depletion of repressive mark H3K27me3 or changes in TFs activity and signaling pathways could be attributable to changes in SOX1 expression during the neural differentiation. It would be significant to analyze epigenetic mechanisms acting on SOXB1 genes in later stages of neural differentiation, as well as in cells undergoing the process of dedifferentiation, as this could have impact on our understanding of the exact roles of SOXB1 genes in pluripotentcy maintenance and on improvement of guided differentiation protocols.
Regarding the other mechanisms implicated in the regulation of SOXB1 genes, we have previously described the modulation of SOX2 and SOX3 gene expression during the early phases of neural induction of NT2/D1 cells by RA [35]. Also, we have performed extensive functional characterization of the regulatory regions within human SOX3 promoter involved in RA-responsiveness [12,14,15,17]. We have described several elements, including an atypical RAR/RXRα response element (RE), located at position -259 to -154 [15], and a DR-3-like RXR RE, positioned -68 to -54 relative to TSS [17]. We demonstrated the involvement of several transcription factors (NF-Y, PBX1 and MEIS1) in the regulation of expression of this gene during initial 48h of neural induction [14,15]. In addition, we analyzed whether some aspects of transcriptional regulation are preserved between human SOX2 and SOX3 genes during first 48h of RA induction [76]. Among many similarities (activation by Sp1, MAZ, PBX1, MEIS1 and liganded RXR) we observed that TGIF acted as a transcriptional repressor of the SOX3 gene, while no significant effect of this TF on the SOX2 expression has been observed. Also, in contrast to significant impact of NF-Y on RA-induced activation of SOX3, this TF had only mild effect on SOX2 expression in RA-induced NT2/D1 cells [76]. This observation is in concordance with the observation of Wiebe et al. that SOX2 promoter activity is down-regulated upon RA induction of mouse F9 EC cells [77]. Indeed, our results confirmed speculations of Wiebe et al. that epigenetic mechanisms could contribute to the silencing of Sox2 promoter upon differentiation of EC cells [77]. Apart from these nearby regions expression of both genes are regulated by far distance enhancers [27, [78][79][80] and multiple regulatory mechanisms (e.g signaling pathways) [81][82][83][84][85][86][87]. It is important to emphasize that SOX2 expression is also modulated through activity of miRNA and ncRNA. miRNA-134 and miRNA-145 repress SOX2 expression by targeting its coding region in mES cells and the 3 0 -UTR in hES cells, respectively [88,89]. Also, it has been postulated that ncRNA SOX2OT (SOX2 overlapping transcript) participate in SOX2 transcriptional regulation acting as an enhancer [90]. As for the SOX1, there are several studies demonstrating the roles of signaling pathways in SOX1 gene regulation [83,86,91,92], while data regarding TFs involvement in the regulation of this gene expression during the neural differentiation are lacking.
In summary, we provide a first map of the epigenetic landscape of SOX3 in pluripotent cells and during the early phases of neural differentiation. We found SOX3 gene to be non methylated from undifferentiated NT2/D1 to cells committed towards neural lineage. Furthermore, we presented data regarding epigenetic mechanisms acting on SOX2 and SOX1 genes during the initial phases of neural differentiation, pointing out to potential similarities and differences in the epigenetic control of SOXB1 genes. These findings could contribute to the elucidation of complex events during neural differentiation and ultimately provide better means for the development of therapies based on the use of epigenetic-modifying drugs.