Determining how genes are epigenetically regulated to ensure their correct spatial and temporal expression during development is key to our understanding of cell lineage commitment. Here we examined epigenetic changes at an important proneural regulator gene Mash1 (Ascl1), as embryonic stem (ES) cells commit to the neural lineage. In ES cells where the Mash1 gene is transcriptionally repressed, the locus replicated late in S phase and was preferentially positioned at the nuclear periphery with other late-replicating genes (Neurod, Sprr2a). This peripheral location was coupled with low levels of histone H3K9 acetylation at the Mash1 promoter and enhanced H3K27 methylation but surprisingly location was not affected by removal of the Ezh2/Eed HMTase complex or several other chromatin-silencing candidates (G9a, SuV39h-1, Dnmt-1, Dnmt-3a and Dnmt-3b). Upon neural induction however, Mash1 transcription was upregulated (>100-fold), switched its time of replication from late to early in S phase and relocated towards the interior of the nucleus. This spatial repositioning was selective for neural commitment because Mash1 was peripheral in ES-derived mesoderm and other non-neural cell types. A bidirectional analysis of replication timing across a 2 Mb region flanking the Mash1 locus showed that chromatin changes were focused at Mash1. These results suggest that Mash1 is regulated by changes in chromatin structure and location and implicate the nuclear periphery as an important environment for maintaining the undifferentiated state of ES cells.
Mash1 (Ascl1) is a key transcription factor essential during embryogenesis for the production of neural precursor cells and for promoting the commitment of these multipotent progenitors to different neuronal fates including noradrenergic, serotonergic and olfactory neurons, whilst inhibiting their astrocyte potential (Bertrand et al., 2002; Guillemot et al., 1993; Hirsch et al., 1998; Pattyn et al., 2004). Identified through its homology and shared function to the Achaete-Scute family in Drosophila, Mash1 is a member of a small group of basic helix-loop-helix proteins known as `proneural' factors which bind to and activate target promoters through interaction with E-proteins and the E-box hexanucleotide motif (Johnson et al., 1990). Mash1 expression during development is tightly regulated (Bertrand et al., 2002; Kageyama et al., 2005) and factors responsible for the transcriptional upregulation (retinoic acid and neurotrophins) and downregulation (Hes1) have been identified (Ishibashi et al., 1995; Ito et al., 2003; Itoh et al., 1997; Shoba et al., 2002) as well as cis-acting regulatory elements within the locus (Meredith and Johnson, 2000).
Gene transcription is regulated by an integrated hierarchy of sequence elements, transcription factors and local modifications to chromatin structure (van Driel et al., 2003). In addition, the spatial organisation of chromatin within the nucleus can also influence gene expression. For example, positioning of individual gene loci relative to specific nuclear landmarks - such as heterochromatin, the nuclear periphery, chromosome territories and nuclear bodies - has been associated with different transcriptional states in a variety of cell types (Baxter et al., 2002; Gasser, 2001; Kosak and Groudine, 2004; Williams, 2003).
Replication timing during S phase is known to broadly correlate with transcriptional activity and chromatin structure: accessible chromatin including expressed genes replicates early whereas constitutive heterochromatin and some facultative heterochromatin replicates later (Azuara et al., 2003; Gilbert, 2002; Schubeler et al., 2002). We have recently shown that Mash1 replicates late during S phase in undifferentiated ES cells, but much earlier when these cells were induced to generate neural progenitors after treatment with retinoic acid (Perry et al., 2004). These results indicate that global changes in chromatin structure of the Mash1 locus occur in response to neural induction. To define the full nature of such changes and how they are important for regulating Mash1, we have systematically compared the epigenetic status of Mash1 in undifferentiated ES cells and ES-derived neural progenitors. Here we show that in ES cells the late-replicating, transcriptionally inactive Mash1 locus is positioned at the nuclear periphery but that upon neural induction the locus relocates towards the nuclear interior. Repositioning of Mash1 is neural specific and is coupled with transcriptional upregulation as well as local changes in the abundance of modified histones at the Mash1 promoter. Given the extent of repositioning, neighbouring genes were also found to relocate towards the interior. However, evidence is provided that chromatin reorganisation initiates at the Mash1 gene.
Mash1 is transcriptionally upregulated upon neural differentiation
Efficient neural induction of OS25 ES cells was achieved using a standard protocol (Billon et al., 2002) and verified by immunofluorescence. Undifferentiated (day 0) ES cell cultures contained, on average, 90% Oct4-positive cells, whereas neural-induced cultures (day 8) contained 85% nestin-positive cells with less than 5% of cells Oct4 positive. Using this system, we have previously shown that upon neural induction Mash1 switched from late to early replication (Perry et al., 2004). We assessed to what extent Mash1 might be upregulated upon neural induction of ES cells. Mash1 transcripts were not detected in undifferentiated ES cells, even after 30 cycles of RT-PCR, but were abundant in day 8 neural cells (Fig. 1A). Control ES- and neural-specific markers showed the expected pattern of expression. To determine whether very low levels of Mash1 transcription or alternative RNA products could account for a previous report that Mash1 can be transcribed in undifferentiated ES cells (Ying et al., 2003) we also performed semi-quantitative RT-PCR and northern blot analysis. The latter revealed a single band for Mash1 in retinoic-acid-differentiated OS25 cells which was undetectable in undifferentiated ES cells (data not shown), whereas the RT-PCR revealed that a low level of Mash1 can be detected in ES cells after 35 cycles: at least 125-fold lower in abundance than that detected in day 8 samples (Fig. 1B). These data confirm that Mash1 is dramatically upregulated during neural differentiation of ES cells in vitro.
The Mash1 locus relocates in the nucleus following neural differentiation
A number of genes are recruited to centromeric heterochromatin when silent (Brown et al., 1999; Brown et al., 1997; Dernburg et al., 1996; Francastel et al., 1999; Grogan et al., 2001), whereas positioning of silent genes towards the nuclear periphery has been shown in yeast (Andrulis et al., 1998; Gasser, 2001) and mammalian nuclei (Kosak and Groudine, 2004; Kosak et al., 2002; Li et al., 2001). We investigated the spatial relationship of Mash1 relative to both centromeric heterochromatin domains and the nuclear periphery. By 3D FISH using a probe for γ-satellite (pericentric DNA) and the Mash1 locus (BAC RP24-130P7), we found that Mash1 is rarely associated with pericentric heterochromatin in either ES or neural cells (Fig. 2A). Thus, late replication and transcriptional repression of Mash1 in ES cells is not due to association with pericentric heterochromatin. We next investigated the spatial relationship of Mash1 to the nuclear periphery using 2D analysis. The positions of Mash1 FISH signals were defined as ratios of the nuclear radius. Values greater or equal to 0.8 were scored as peripheral (where the nuclear centre=0 and the periphery=1) (Kosak et al., 2002). We found that in ES cells Mash1 was preferentially associated with the periphery. However, after neural differentiation the locus was significantly more often located away from the periphery (P<0.001 by the chi-square test) (Fig. 2B,C). This suggests that neural induction results in a coordinated switch in replication timing, transcriptional upregulation and relocation of the Mash1 locus from the nuclear periphery to the nuclear interior. The repositioning of Mash1 was also confirmed by additional 3D experiments. In 3D-preserved nuclei we observed 84% of Mash1 signals adjacent to the nuclear periphery in ES cells whereas only 28% of signals were peripheral in ES-derived neural cells (Fig. 2D).
Movement of Mash1 away from the nuclear periphery is specific to the neural fate
The Mash1 gene is known to be transcriptionally upregulated by retinoic acid (Shoba et al., 2002). Since our chosen method of neural induction involves treatment with retinoic acid we were interested in whether the movement of Mash1 away from the nuclear periphery could be recapitulated using alternative neural differentiation methods. We differentiated MR-7 ES cells into neural cells by growth on a PA6 stromal cell layer (Kawasaki et al., 2000). By day 8 of differentiation these cultures contained 80% nestin-positive cells, and RT-PCR analysis confirmed an upregulation of Mash1 (Fig. 3A). FISH analysis revealed that Mash1 was located at the nuclear periphery in MR-7 ES cells and following differentiation was again relocated towards the nuclear interior. The shift was slightly less dramatic than in OS25 cells but still significant (P<0.001) (Fig. 3B,C). Movement of the Mash1 locus away from the periphery and its transcriptional upregulation is therefore a consistent feature of neural induction of ES cells, rather than simply being a result of retinoic acid treatment. Furthermore, we extended this analysis to normal ex vivo neural progenitor cells. Neural cells, isolated from the forebrain of day 11.5-12.5 Sox1-lacZ mouse embryos, were labelled for lacZ expression and positive cells were isolated by FACS. We found that Mash1 was located towards the nuclear interior in ex vivo neural cells (Fig. 3B,C) and furthermore was early replicating (data not shown). Again positioning of Mash1 in ex vivo neural cells was significantly different to that in OS25 ES cells (P<0.001).
To address whether the spatial relocation of Mash1 was specific for neural commitment or results from generalised changes associated with ES cell differentiation we examined Mash1 in a variety of non-neural cells. OS25 ES cells were differentiated by growth on collagen IV-coated plates without LIF. Mesodermal cells (FLK+/Ecad-) were then purified from the differentiated population by FACS as described previously (Fraser et al., 2003). In these cells Mash1 was found to remain at the nuclear periphery. Mash1 was also located at the nuclear periphery in ex vivo keratinocyte progenitors and ex vivo T cells (Fig. 3B,C). P values were not significant for these three cell types compared to OS25 ES cells (P=1, 0.599 and 0.299 for OS25 ES vs mesoderm, keratinocytes and T cells, respectively) confirming that repositioning of Mash1 to the nuclear interior is selective for neural-committed cells.
Epigenetic changes are focussed at the Mash1 locus
As relocation of Mash1 appears to be specific for neural cells, we asked whether neural-specific epigenetic changes are nucleated at the Mash1 locus or are a passive reflection of genomic events at a neighbouring gene. For example, the entire chromosome territory may move position during neural differentiation, or the regulation of upstream or downstream genes may influence positioning. In silico analysis of a 2 Mb region surrounding the Mash1 locus on chromosome 10 did not reveal any neural-specific genes close to Mash1 (Fig. 4A). RT-PCR analysis of the genes around Mash1 show an interesting pattern of expression during ES to neural differentiation (Fig. 4B). Strikingly, the Pah gene which is just 30 kb telomeric to Mash1 is silent in both ES and neural cells and so too is the Igf1 gene, a further 300 kb telomeric. Nup37 on the other hand is expressed robustly in ES cells and is also expressed in neural progenitors. Centromeric to Mash1 however all four genes analysed were transcriptionally upregulated in response to neural induction, in common with Mash1. Given that none of these genes are known to be neural specific, it seems unlikely that they are the targets for dominant chromatin cues involved in repositioning Mash1, although of course we cannot completely rule this out.
To examine how far around Mash1 the replication timing switch and nuclear repositioning extends, we performed a `genome walk' either side of Mash1, from Timp3 to Nup37. As shown in Fig. 4C, the difference in replication timing between ES and neural progenitors was most marked in the Mash1-LOC380647 (Gm1554) region but extended to 1.2 Mb centromeric (Timp3) and 0.65 Mb telomeric (Nup37) where replication was at mid S phase in both cell types. To look at nuclear repositioning of genes around Mash1, given the large number of loci to study and the potential need to be able to detect subtle movements, we used a high-throughput, semi-automated system previously reported (Roix et al., 2003). The map positions of the probes used for this analysis are shown in Fig. 4A (indicated in green). We looked at loci at distances from Mash1 up to the point of no replication timing change in both directions. All six probes covering the 2 Mb region around the Mash1 locus showed significant movement away from the nuclear periphery towards a more internal location (Fig. 4D). These data indicate that despite movement being specific to neural cells (and therefore probably coordinated at Mash1), the region as a whole relocates during neural commitment and confirms that transcriptional status of certain genes is independent of position relative to the nuclear periphery (Pah, Igf1, Nup37). Interestingly, a similar analysis of two unlinked genes: Neurod (involved in neurogenesis, located on chromosome 2C3) and Sprr2a (involved in epidermal differentiation, located on chromosome 3F1), which like Mash1 are inactive and late replicating in undifferentiated ES cells, showed that these loci also occupied a peripheral position. In contrast, the expressed and early replicating genes Sox2 and Utf1 were more centrally located in undifferentiated ES cell nuclei (Fig. 4D). These data raise the possibility that positioning of developmentally regulated genes towards the nuclear periphery might be important for maintaining their silent state and thus the pluripotency of ES cells. In day 8 differentiated neural cells both Neurod and Sprr2a are inactive and late replicating and, consistent with this, both remain closely associated with the nuclear periphery. Sox2 and Utf1 on the other hand, which are both early replicating in neural cells, do not associate with the periphery (Fig. 4D).
To define the chromatin changes at the Mash1 promoter that accompany neural induction, we performed chromatin immunoprecipitation (ChIP) analysis. The specificity of these changes was assessed by comparing Mash1 to its two nearest neighbouring genes, Pah and 1700113H08rik. In undifferentiated OS25 ES cells we observed abundant Me3H3K27 at the Mash1 promoter. Upon neural differentiation the level of Me3H3K27 declined and we observed a reciprocal increase in levels of H3K9 acetylation (Fig. 5A). Although both of these modifications, AcH3K9 and Me3H3K27 showed some change at the neighbouring Pah and 1700113H08rik genes, neither showed levels comparable to Mash1 giving evidence that these chromatin modifications associated with neural induction are orchestrated at, and specific to, Mash1.
Low levels of both Me3H3K9 and Me3H4K20 at all three genes, compared with pericentromeric heterochromatin (positive control; data not shown) indicates that neither are likely to be involved in the silencing and peripheral localisation of Mash1.
Peripheral positioning of Mash1 does not require HMTases or DNMTases
The abundance of H3K27 methylation at the Mash1 promoter in ES cells, suggests that this modification could be important for facultative heterochromatin formation and tethering Mash1 to the nuclear periphery. To test this we examined the positioning of Mash1 in ES cells lacking the Eed/Ezh2 HMTase complex, which is required for H3K27 methylation (Montgomery et al., 2005; Silva et al., 2003). In these cells, Mash1 was preferentially positioned at the nuclear periphery, similar to the OS25 and MR7 ES cells (Fig. 5B,C). To investigate whether other chromatin-silencing factors might be responsible for tethering the Mash1 locus to the nuclear periphery we analysed a number of other ES cell lines targeted for the following candidates: the HMTases, Suv39h-1 and G9a and the DNA methyltransferases, Dnmt1 and Dnmt3a and 3b. None of the ES cell lines analysed showed Mash1 signals in a preferentially non-peripheral location in the nucleus (Fig. 5B,C), i.e. when compared with OS25 ES cells, none were significantly different (P=0.426, 0.637, 0.326, 0.619 and 0.619 for ES vs Eed-/-, G9a-/-, SuV39h1-/-, Dnmt1-/- and Dnmt3ab-/-, respectively). The results therefore suggest that these chromatin silencing factors are not uniquely required for tethering Mash1 to the nuclear periphery.
Mash1 relocates in response to neural induction
Dissecting the epigenetic mechanisms regulating key developmental genes is important for understanding both how ES cell pluripotency is maintained, and how cell fate is assigned in the developing embryo. Using ES cells as a model system, we characterised the epigenetic changes that occur at the Mash1 locus in response to neural commitment. From a previous survey of approximately 50 transcription factor genes, Mash1 was shown to be one of only three that replicates late in ES cells (Perry et al., 2004). Upon neural induction of ES cells Mash1 was shown to switch from late replication to early and we show here that in parallel Mash1 is transcriptionally upregulated, acquires H3K9 acetylation across its promoter, declines in H3K27 methylation and is relocated from the nuclear periphery towards the interior.
We demonstrate that Mash1 repositioning is selective for neural differentiation because Mash1 was centrally located in ES-derived neural cells and primary neural cells but not in T cells, keratinocytes or ES-derived mesodermal cells. This neural specificity argues that general features of differentiation such as the reorganisation of large blocks of heterochromatin are unlikely in themselves to be the cause of Mash1 relocation. The finding that changes in histone modifications were prominent at the Mash1 promoter compared with its neighbouring genes and that maximal changes in replication timing were centred at the Mash1-LOC380647 region, argue that these epigenetic changes are nucleated at Mash1. A recent report that loci which change replication timing during ES differentiation are often AT rich and LINE dense (Hiratani et al., 2004) fits well with the occurrence of an approximately 300 kb cluster of full-length LINE elements, spanning the region from Mash1 to LOC380647.
What might constrain Mash1 at the periphery in ES cells? In the Drosophila embryo the genome is non-randomly associated with the nuclear periphery (tethered) at regular intervals (Marshall, 2002; Marshall et al., 1996). Consistent with this, live-cell analysis of chromatin dynamics in human nuclei has revealed that loci nearer to the nuclear periphery have a smaller radius of confinement (indicating restraint by tethering) than loci positioned towards the interior (Chubb et al., 2002). Our data show that Mash1 is associated with the nuclear periphery in undifferentiated ES cells and that, in these cells, H3K27 methylation is readily detected at the Mash1 promoter. Although this result suggests that facultative heterochromatin could play a role in Mash1 tethering, analysis of a panel of ES cells deficient for specific chromatin silencing factors - including Eed, a component of the H3K27-specific HMTase - did not reveal any major changes in Mash1 location. This could either mean that none of these chromatin modifications are required, or that the cells are able to compensate for loss of an individual factor by alternative mechanisms (redundancy). Alternatively, the nuclear periphery may represent a `default' location for Mash1 in ES cells, where relocation is a response to locus activation. In this regard, it has previously been shown that the transcriptional activator VP16 can drive the spatial relocation of a lac-operator repeat array from a preferentially peripheral nuclear location towards the interior (Tumbar et al., 1999). Also, in lymphocytes and neuroblastoma cells the cystic fibrosis gene (CFTR) relocates away from a poorly acetylated chromatin environment of the nuclear periphery, to an internal position in response to treatment with the histone deacetylase (HDAC) inhibitor trichostatin A (Zink et al., 2004). Although the observation that neural differentiation results in increased H3K9 acetylation across the Mash1 promoter, could fit this interpretation, HDAC inhibitors dramatically affect the global organisation of constitutive heterochromatin in differentiating cells as well as other cell types (Taddei et al., 2001) which complicates their use in our present model system.
Peripheral positioning and transcriptional activity
Recently, CFTR and two neighbouring genes that are differentially expressed were shown to localise to discrete areas relative to the nuclear periphery, in a transcription-dependent manner (Zink et al., 2004). In our study, repositioning of Mash1 resulted in the relocation and replication time switching of a number of neighbouring genes. Two of these, Pah and Igf1, remained transcriptionally silent in both ES and ES-derived neural cells. This shows that certain epigenetic features of accessible chromatin such as a favourable nuclear position or time of DNA replication, are insufficient in themselves to drive gene expression: a conclusion that has also been highlighted by studies of the α-globin locus in non-erythroid cells (Brown et al., 2001; Smith and Higgs, 1999). Further from Mash1 - 1.2 Mb centromeric (Timp3) and 0.65 Mb telomeric (Nup37) - replication was in middle S phase in both cell types. Nup37 was transcribed in both cell types, however, Timp3 was silent in ES cells and upregulated in neural cells, indicating that change in replication timing is not a general feature of changing transcriptional status.
The fact that Pah and Igf1 remain silent, whereas other Mash1 neighbouring genes such as Timp3, Syn3, LOC380647 and 1700113H08rik are coordinately upregulated in response to neural induction, could indicate the presence of a so-called boundary element or insulator (Labrador and Corces, 2002) between Mash1 and Pah. Although no potential CTCF binding sites have yet been identified, a BLAST analysis of the 30 kb separating the these genes revealed a region of approximately 1.2 kb which was highly conserved between human and mouse (87% conserved sequence, located 6 kb upstream of the Mash1 promoter) that contains a functional enhancer element (Meredith and Johnson, 2000; Verma-Kurvari et al., 1998).
The nuclear periphery is generally considered a region of transcriptional repression, being enriched for late-replicating, gene-poor, hypoacetylated chromatin (O'Keefe et al., 1992; Croft et al., 1999; Ferreira et al., 1997; Sadoni et al., 1999; Tanabe et al., 2002). Positioning of silent genes at the nuclear periphery has been shown in yeast (Andrulis et al., 1998; Gasser, 2001) and mammalian nuclei (Kosak et al., 2002; Li et al., 2001), whereas actively transcribed genes and early replicating DNA tend to locate towards the nuclear interior (O'Keefe et al., 1992; Carter et al., 1993; Kosak and Groudine, 2004). In yeast, telomeres cluster at the nuclear periphery and recruit SIR proteins, known to be involved in silencing (Cockell et al., 1995; Cockell and Gasser, 1999). By artificially tethering a reporter gene to the yeast nuclear periphery, silencing of the reporter was induced and shown to be SIR dependent (Andrulis et al., 1998), providing a clear mechanism in which loci are both repressed and physically tethered to the nuclear periphery. Recent studies indicate a more complex role of the nuclear periphery in gene regulation. A genome-wide analysis in yeast has shown that nuclear pore proteins associate with a subset of highly transcribed genes (Casolari et al., 2004). The nuclear pore complex has also been implicated in boundary activity (Labrador and Corces, 2002), a finding supported by the observation that the gypsy insulator element localises to the periphery in Drosophila nuclei (Gerasimova et al., 2000). In mammalian cells, however, a role for the nuclear periphery in actively regulating gene repression remains unproven. We have recently shown that the interferon-γ gene was constitutively located at the nuclear periphery in T cells, apparently irrespective of its expression (Hewitt et al., 2004). Here we show that certain genes (Pah, Igf1) change their location in differentiating ES cells despite remaining transcriptionally inactive. Probably these closely linked genes are simple `passengers' of chromatin changes that are centred at the Mash1 locus and that drive spatial repositioning of the entire region. Whatever the explanation, the demonstration that two other developmentally important but unlinked genes - Neurod and Sprr2a (selected as representatives of inactive and late-replicating genes in undifferentiated ES cells) - in addition to Mash1, both localise at the nuclear periphery in undifferentiated ES cells is intriguing. These data suggest that the nuclear periphery could function as a repressive compartment within undifferentiated ES cells, important for maintaining their uncommitted status.
Materials and Methods
OS25 ES cells, in which the β-geo gene is inserted at the Sox2 locus and the hygromycin-thymidine kinase gene is inserted at the Oct4 locus, were cultured as previously described (Billon et al., 2002; Perry et al., 2004). MR-7 ES cells were maintained on irradiated primary embryonic fibroblasts plated on gelatin-coated dishes in G-MEM supplemented with 1% FCS (Globepharm), 10% KSR, 2 mM glutamine, 0.1 mM nonessential amino acids, 1 mM pyruvate, 0.1 mM 2-mercaptoethanol (2-ME), penicillin-streptomycin (GIBCO-BRL) and 1000 U/ml LIF (Chemicon). Sox1-positive ex vivo neural cells (a kind gift from C. Andonaidou and R. Lovell-Badge, NIMR, London, UK) were isolated from day 11.5-12.5 Sox1-lacZ mouse embryos by dissociating forebrains into single cell suspension and labelling lacZ-positive cells using the DetectaGene Green CMFDG lacZ Gene Expression kit (Molecular Probes), followed by FACS. T lymphocytes were isolated from lymph nodes of 6- to 8-week-old C57BL/6 mice and were cultured for 3-4 days in IMDM supplemented with 10% FCS, 2.5×10-5 M 2-mercaptoethanol and antibiotics on culture plates coated with 2 μg/ml anti-TCRb (H57-597; BD Pharmingen) and 10 μg/ml ati-CD28 (BD Pharmingen) with Il2 (1 ng/ml). Primary keratinocytes were isolated from 3-day-old mice as described previously (Hennings, 1994) and cultured in complete `low-calcium FAD' (0.1 mM Ca2+) (one part Ham's F12 and three parts Dulbecco's modified Eagle's medium) supplemented with FBS and growth factors (Hobbs et al., 2004). Cells were plated on collagen coated plates and were grown as attached, undifferentiated precursor keratinocytes.
ES cell differentiation
OS25 ES cells were differentiated into neural ectoderm as previously described (Billon et al., 2002; Perry et al., 2004). Briefly, cells were trypsinised and plated onto 10 cm bacterial dishes without LIF to allow embryoid bodies to form. 1 μM retinoic acid was added on day 4, removed on day 6 and replaced with DMEM-F12 (containing N2) and neurobasal media (containing B27) in a 1:1 ratio. From day 6-8 selection for expression of Sox2 and against expression of Oct4 was achieved using G-418 (100 μg/ml) and gancyclovir (2.5 μM) respectively. On day 8, cells were harvested for RNA extraction, immunofluorescence as previously described (Perry et al., 2004) and for FISH fixation.
MR-7 ES cells were differentiated into neural cells by plating on irradiated PA6 stromal cells in G-MEM medium supplemented with 10% KSR, 2 mM glutamine, 1 mM pyruvate, 0.1 mM nonessential amino acids, and 0.1 mM 2-ME (Kawasaki et al., 2000). Again, day 8 cells were harvested for FISH fixation, immunofluorescence and RNA extraction.
Differentiation of OS25 cells into mesoderm was carried out as previously described (Fraser et al., 2003) with minor modifications. Cells were trypsinised and plated at low density (0.5×103 cells/cm2) on collagen IV-coated plates in differentiation medium without LIF. After 5 days cells were harvested using cell dissociation buffer (Sigma) and labelled with PE-conjugated anti-FLK and FITC-conjugated anti-E-cadherin (Pharmingen). FLK+/E-cadherin- mesoderm cells were collected by FACS for FISH fixation.
RNA extraction was performed using RNAzol B (Biogenesis). Cells were resuspended in 0.2 ml RNAzol per 106 cells, mixed thoroughly, and then subjected to chloroform extraction and isopropanol precipitation before resuspension in RNAse-free water. One μg total RNA was then reverse transcribed using Superscript™ First Strand Synthesis kit (Invitrogen) and cDNAs of interest were PCR amplified in 20 μl reaction containing 2 U Taq polymerase (New England Biolabs) and 0.5 μM of each primer. The following cycling conditions were used: 94°C for 2 minutes, then 30 cycles of 94°C for 30 seconds, 60°C for 30 seconds and 72°C for 1 minute, finishing with 72°C for 5 minutes. For semi-quantitative RT-PCR, cDNA from OS25 cells (undifferentiated and day 8 differentiated), and mouse E15 embryonic heads was serially diluted and 2 μl of 1:1, 1:5, 1:25 and 1:125 dilutions were amplified by 35 cycles of PCR. Primer sequences are available on request.
Fluorescence in situ hybridisation (FISH)
BAC probes for FISH were checked for the presence of the appropriate gene by PCR and then labelled with digoxygenin using the Nick Translation kit (Invitrogen). Specific hybridisation of BACs was confirmed by metaphase FISH. For 2D Interphase, FISH cells were harvested by trypsinisation, and hypotonically treated in 75 mM KCl for 5 minutes at room temperature before fixing in ice-cold methanol: acetic acid (3:1). Fixed cells were dropped onto slides, denatured, hybridised and washed as previously described (Williams et al., 2002). For 3D FISH cells were harvested, washed in PBS and attached to poly-L-lysine-coated coverslips. Fixation, denaturation, hybridisation and washing were all carried out as previously described (Brown et al., 1997).
Microscopy and measurements
FISH on 3D-preserved nuclei was viewed under a Leica laser-scanning confocal microscope, using a 100× oil-immersion objective. Optical slices across the z-axis of nuclei were captured every 0.25 μm to create z-stacks for analysis. Mash1 was scored as `associated' with γ-satellite when the two signals were touching or overlapping (see Williams et al., 2002). FISH on 2D-fixed nuclei was viewed under a Leica epifluorescence microscope, using a 100× oil-immersion objective. Images were captured with a CCD camera and analysed using IP lab software. The position of loci relative to the nuclear periphery was determined by measuring the distance from the nuclear centroid to the FISH signal as a ratio of the nuclear radius (radius=distance of nuclear centroid to nuclear periphery through the FISH signal). FISH signals at ≥80% of nuclear radius were considered peripheral (Kosak et al., 2002). For each cell type at least two experiments were performed and approximately 100 FISH signals were scored. The chi-square test for significance was performed on the pooled totals for each cell type. High-throughput analysis of nuclear positioning was performed as previously described (Roix et al., 2003), and for these the Wilcoxon-Mann-Whitney test for significance was performed.
Replication timing analysis
OS25 cells were BrdU labelled, fixed in 70% ethanol and sorted into cell-cycle fractions as previously described (Azuara et al., 2003; Perry et al., 2004). BrdU-labelled nascent DNA was isolated by immunoprecipitation and the abundance of nascent DNA in each fraction for the genes of interest was determined by real-time PCR (primer sequences available on request).
Chromatin immunoprecipitation (ChIP) analysis
OS25 cells were harvested on day 0 (ES) and day 8 (neural) of differentiation and processed for ChIP analysis as previously described (Baxter et al., 2004). Briefly, 140 μg chromatin was immunoprecipitated with 2 μl anti-histone H3 (input control, Abcam 1791), 3 μl anti-acetyl-histone H3K9 (Upstate, 07-352), 5 μl anti-acetyl-histone H4 (Serotec, AHP418), 5 μl anti-trimethyl-histone H3K9 and anti-trimethyl-histone H4K20 (both from Thomas Jenuwein) and 2 μl rabbit anti-mouse-IgG antiserum (negative control, Dako). DNA was then eluted from immune complexes and quantification of recovered DNA was performed using real-time PCR (primer sequences available on request).
We thank Francois Guillemot for Mash1 BAC, RP24-130P7, Cynthia Andoniadou and Robin Lovell-Badge for the Sox1/LacZ-positive primary neural cells and Simon Broad for the mouse embryonic keratinocyte precursors. We thank Bernard Ramsahoye and En Li, Yoichi Shinkai, Thomas Jenuwein and Rosalind John for the Dnmt, G9a, SuV39h-1 and Eed mutant ES cell lines, respectively. We thank Mikhail Spivakov for help with the replication timing analysis. This work was supported by the Medical Research Council, UK.
- Accepted October 4, 2005.
- © The Company of Biologists Limited 2006