Fragile X syndrome (FXS) is the most common heritable cause of intellectual disability and the most common known cause of autism. Most cases of FXS result from the expansion of a CGG·CCG repeat in the 5′ UTR of the FMR1 gene that leads to gene silencing. It has previously been shown that silenced alleles are associated with histone H3 dimethylated at lysine 9 (H3K9Me2) and H3 trimethylated at lysine 27 (H3K27Me3), modified histones typical of developmentally repressed genes. We show here that these alleles are also associated with elevated levels of histone H3 trimethylated at lysine 9 (H3K9Me3) and histone H4 trimethylated at lysine 20 (H4K20Me3). All four of these modified histones are present on exon 1 of silenced alleles at levels comparable to that seen on pericentric heterochromatin. The two groups of histone modifications show a different distribution on fragile X alleles: H3K9Me2 and H3K27Me3 have a broad distribution, whereas H3K9Me3 and H4K20Me3 have a more focal distribution with the highest level of these marks being present in the vicinity of the repeat. This suggests that the trigger for gene silencing may be local to the repeat itself and perhaps involves a mechanism similar to that involved in the formation of pericentric heterochromatin.
Fragile X syndrome (FXS; MIM no.: 300624) is an X-linked developmental disorder whose symptoms include intellectual disability, autistic-like symptoms, attention deficits, hyperactivity, insomnia, anxiety, depression, aggressive behavior, seizures and hypersensitivity to sensory stimuli (1). FXS results from mutations in the Fragile X Mental Retardation 1 (FMR1) gene. This gene encodes a translational regulator, FMRP, that is ubiquitously expressed at high levels early in embryonic development (2). The most common FXS mutation is an expansion of a CGG·CCG repeat tract in the 5′-untranslated region of the Fragile X mental retardation 1 (FMR1) gene from a premutation (PM) allele with 55–200 repeats to an allele having >200 repeats (3–5). These alleles, referred to as full mutation (FM) alleles, are transcriptionally silenced with a consequent deficiency of FMRP (6–8).
Although indirect evidence suggesting that FX alleles are transcriptionally silent and associated with heterochromatin preceded the identification of the FMR1 gene in 1991, and the promoter and exon 1 of FX alleles have since been shown to have extensive CpG methylation and to be associated with certain marks of silent chromatin, the silencing mechanism remains the subject of much debate (3,9–25). Transcripts with long CGG tracts are known to form substrates for the RNA interference (RNAi) pathway (23). In addition, below the silencing threshold, an increase in the repeat number is associated with an increase in the amount of both the sense and antisense transcripts produced from the FMR1 gene (21,26). Increases in repeat number also affect splicing and lead to the use of transcription start sites that are located further 5′ on the gene, in some cases as much as 10 kb away (21,26,27). Thus, during early embryogenesis when FX alleles are still active (20), unique transcripts may be produced or the existing transcripts may be made in sufficient quantity to trigger silencing. This silencing could occur via an RNAi-dependent or RNAi-independent process (28). In addition, a variety of CGG·CCG-repeat-binding proteins are found in human cells (29,30). By analogy with sequence-specific DNA-binding proteins such as MEF2 (31) and Sp1 (32), such proteins may recruit enzymes responsible for establishing and reinforcing gene silencing. Possible models for FX gene silencing thus include RNA-dependent and RNA-independent mechanisms in which the histone modifications are either nucleated within the repeat or initiate elsewhere in the gene as a downstream consequence of repeat expansion (21,23,33,34).
Distinct histone ‘signatures' have been identified for various silenced regions of the genome that may reflect differences in the way that transcriptional repression is accomplished at these loci (35–39). To date, only two repressive histone modifications, H3K9Me2 and H3K27Me3, have been studied with respect to the silenced FMR1 gene in individuals with FXS (16,19,40,41). H3K9Me2 and H3K27Me3 are modifications that are typically seen in unique regions of the genome that are developmentally repressed (38,42,43). Histone modifications, such as H3K9Me3 and H4K20Me3, that characterize repetitive DNA sequences like Satellite 2 (Sat2), the long array of 1 kb+ imperfect repeats found in pericentric heterochromatin (36,42–44), have yet to be examined. To better understand FX gene silencing, we have extended our previous studies on the FMR1 gene to examine the FMR1 gene of normal and silenced alleles for the presence and distribution of such modified histones. We show that FX alleles are enriched for both H39Me3 and H4K20Me3. The distribution of these modifications suggest that FX gene silencing may arise from events intrinsic and local to the repeat and result in a repressive histone profile reminiscent to that seen on human pericentric heterochromatin.
A localized region of elevated H3K9Me3 and H4K20Me3 occurs on FX alleles coincident with the repeat
We examined three regions of both active and silenced FMR1 alleles for association with H3K9Me3 and H4K20Me3. Since FMR1 is located on the X chromosome, we used cells from males to avoid complications arising from the presence of two X chromosomes in females. Two of the regions we tested were located within the body of the FMR1 gene. Because of the technical difficulties associated with the PCR of long CGG·CCG-repeat tracts and the vast difference in the efficiency of amplification of short (active) and long (silenced) alleles, we used primer pairs from the region immediately downstream of the repeats in exon 1 to examine chromatin changes in the repeat region. We also used a primer pair in intron 1, ∼1 kb downstream of exon 1 to examine chromatin changes downstream of the repeat (Fig. 1A). A region 5′ of the FMR1 promoter and ∼1 kb upstream of the major transcription start site of the FMR1 gene, referred to as the upstream region, was also examined. Exon 1 of GAPDH, a robustly expressed housekeeping gene was used as a negative control. Sat2 was used as a positive control.
Normal cells had levels of H3K9Me3 and H4K20Me3 that were not significantly above GAPDH in any of the three regions tested. In contrast, exon 1 in FXS cells showed levels of these histones that were comparable to Sat2 and significantly higher than GAPDH (Fig. 1B and C). The level of both modified histones then dropped off significantly in both the 5′ and 3′ directions. This would be consistent with the idea that the events that nucleate the deposition of these repressive histone marks initiate within or very close to the repeat.
A zone of developmentally regulated heterochromatin is located 5′ of both active and silenced FMR1 alleles
To assess whether the repressive histone marks previously shown to be associated with FX alleles, H3K9Me2 and H3K27Me3, have a similar distribution, we analyzed the association of these histones with the same three regions of the FMR1 locus in cells from unaffected individuals and those with FXS. As with H3K9Me3 and H4K20Me3, these two modified histones were present at elevated levels on FX alleles. However, in contrast to what was seen with the first pair of modified histones, H3K9Me2 and H3K27Me3 showed a similar elevated distribution on all three regions tested in FX cells (Fig. 2A and B) with the levels of these marks being comparable to that seen on Sat2.
H3K9Me2 and H3K27Me3 were also present at similar levels on the upstream region of normal alleles (Fig. 2A and B). However, significantly lower levels of these histones were found on exon 1 or intron 1 of such alleles (Fig. 2). While this work was in progress, DNA methylation of the upstream region of normal alleles was reported (33). Previous work had suggested that the FMR1 gene was embedded in a much larger region of altered chromatin (45). Our work, together with the other recently published study, demonstrates that this region of altered chromatin actually abuts the 5′ end of the FMR1 gene. Furthermore, our data suggest that this upstream heterochromatin zone (UHZ) has elevated levels of H3K27Me3 and H3K9Me2, but not H3K9Me3 or H4K20Me3.
Thus, our data suggest that the complement of histone modifications present on the UHZ of normal and FX alleles is different from that seen within the body of FX alleles. The UHZ histone profile is reminiscent of normal, developmentally repressed single copy genes, whereas the profile of all four repressive marks on exon 1 of FXS alleles was indistinguishable from that seen on Sat2. Thus, it may be that the mechanism of FX gene silencing is different from the mechanism responsible for establishing the UHZ.
Since the histone modification ‘signature' of the UHZ is typical of developmentally repressed genes, we examined the deposition of H3K27Me3 on this region in normal human embryonic stem cells (hESCs) before and after differentiation. We found that in undifferentiated ESCs, H3K27Me3 levels were low (Fig. 2C). However, on differentiation of the hES cells into embryoid bodies (EBs), an increase in the levels of H3K27Me3 occurred. This increase was confined to the UHZ as it is in lymphoblastoid cells from unaffected individuals (Fig. 2C). A similar situation was seen in mouse (data not shown) Thus, the deposition of H3K27Me3 on the UHZ is an early step in normal human and murine embryonic development.
The FMR1 locus produces a complex mixture of long and short RNAs
A number of different RNA species have been implicated in gene silencing (46). One species consists of long non-coding transcripts that act in cis or trans to target silencing complexes to the target gene independent of the RNAi machinery. The second species are small RNAs produced from longer sense–antisense pairs by the RNAi machinery. We have used a combination of northern blotting and RNase protection to examine the RNAs produced by the FMR1 locus. Our northern blotting data on total RNA suggests that a variety of long antisense transcripts are made including very high levels of a ∼5 kb transcript that hybridizes strongly with the UHZ probe (Fig. 3A). This is presumably not the previously described FMR4 gene which initiates in a similar location, since the FMR4 transcript is only 2.5 kb (47). It is also not one of the splice isoforms generated from the antisense transcript, ASFMR1, since it hybridizes only weakly with sequences from the promoter region and exon 1. Unlike the previously described antisense transcripts (21,47), these RNAs are seen at similar levels in carriers of normal alleles having <54 repeats, as well as carriers of PM and FM alleles. Since the RNAs are the same size in all of these individuals, these transcripts likely do not contain the repeats. Failure to detect the previously described ASFMR1 and FMR4 antisense transcripts by northern blotting would be consistent with previous reports and presumably reflects the much lower levels of these transcripts (21). Since the RNAs we describe are found at equivalent levels in both normal and FX cells, it seems unlikely that they are involved in FX gene silencing, although it may be that they contribute to the formation of the zone of heterochromatin in which the FMR1 gene is embedded in both normal and FXS cells.
In addition to these long RNAs, small RNAs, ∼20 nt in length, that are derived from the FMR1 locus were also detected using probes from the UHZ, promoter and exon 1. Similar levels of these RNAs were seen in normal cells as well as in patient cells with or without gene reactivation with 5-azadeoxycytidine (5-aza), a DNA methyltransferase inhibitor (Fig. 3B and data not shown). These RNAs correspond to the antisense strand of the FMR1 locus and are slightly smaller than the small RNAs produced from the DMPK locus in both normal cells and cells from patients with myotonic dystrophy type I (DM1; MIM no.:160900), a CTG·CAG-repeat expansion disorder (Fig. 3B) (48). Moreover, in contrast to what is seen in DM1, where the small antisense RNAs are limited to the region that includes the repeat and that becomes heterochromatinized, these small RNAs extend far beyond the repeat and are not associated with heterochromatin in normal cells, despite being present at levels similar to the small DMPK RNAs.
This leaves the possibility that small RNAs produced from the repeats themselves, either because of their higher molar amount or because of some as yet unknown biological property, are particularly efficient at inducing gene silencing when the repeat number exceeds 200. This possibility is appealing since the distribution of H3K9Me3 and H4K20Me3 suggests that nucleation of the heterochromatin occurs close to or within the repeat. We have been able to detect small RNAs corresponding to the CGG repeats in both normal cells and patient cells where the FMR1 gene has been reactivated (data not shown). However, because CGG repeats are found elsewhere in the human genome, it is unclear whether these small repeat containing RNAs are derived from the FMR1 region or not. Whatever their origins, we have found no evidence of any RNA species unique to patient cells even when silenced alleles were reactivated with 5-aza. Nor do we see altered levels of any of these RNAs. Thus, although it is tempting to think that transcripts from the FMR1 locus play a role in FX gene silencing, we have been unable to find any clear evidence that they do. Our data thus raise the possibility that RNA-independent mechanisms may be more likely to be involved in FX gene silencing.
We have extended the work from a variety of laboratories, including our own, to more extensively characterize the FMR1 gene and the region of the X chromosome in which it is embedded. We showed that the FMR1 gene lies immediately downstream of a UHZ that is established during early embryonic development in both normal and FX alleles. The UHZ of active alleles has a histone profile similar to what is seen on developmentally repressed single copy regions of the genome being enriched primarily for H3K9Me2 and H3K27Me3. In contrast, the body of the silenced FMR1 gene contains, in addition to the modified histones seen on the UHZ, elevated levels of H3K9Me3 and H4K20Me3, two histone modifications not previously known to be associated with FX alleles (Figs 1 and 2). The fact that the histone profiles differ as they do suggests that gene silencing in FXS is unlikely to simply reflect the spread of silencing marks from the UHZ into the body of the FMR1 gene as has recently been suggested (33).
All four modified histones are present on exon 1 of FX alleles at levels comparable to that found on Sat2. However, whereas H3K9Me2 and H3K27Me3 are broadly distributed across the whole 5′ end of silenced alleles, H3K9Me3 and H4K20Me3 levels were highest on exon 1 which contains the CGG-repeat tract as represented diagrammatically in Fig. 4. The difference in the spread of the H3K9Me2/H3K27Me3 histones and H3K9Me3/H4K20Me3 histones seen on FX alleles may be related to the tendency of H3K27Me3-enriched heterochromatin to spread widely in the absence of effective chromatin boundaries, whereas ‘focal heterochromatin' with a much more limited ability to spread is often seen on genes enriched for H4K20Me3 and H3K9Me3 (49–51).
On normal alleles, spreading of the repressive histone marks from the UHZ into the body of the active FMR1 gene may be prevented by factors such as CTCF and USF1/USF2. These proteins bind to the promoter of active FMR1 alleles (21,52) and have the potential to enforce a chromatin boundary (53,54). In FXS, the putative chromatin boundary at the 5′ end of the FMR1 gene appears to have been lost, and the UHZ and the repeat-induced heterochromatin have merged, with limited spread of the H3K9Me3/H4K20Me3 marks from the body of the FMR1 gene into the UHZ (Fig. 4). However, since the histone profile on exon 1 of FMR1 is similar to Sat2 with respect to all four repressive heterochromatin marks, it may be that the FX gene silencing mechanism is similar to the one responsible for silencing of Sat2. Thus, it is unnecessary to invoke a corresponding spread of histone modifications in the other direction to explain the histone profile in the body of the silenced FMR1 gene. However, it is possible that there are histone modifications that remain to be identified that originate in the UHZ and spread into the body of the FMR1 gene when the chromatin boundary is lost, thereby contributing to FX gene silencing.
The fact that the body of the silenced gene shares a heterochromatin profile with Sat2, along with the fact that the highest levels of H3K9Me3 and H4K20Me3 are seen on exon 1, supports the idea that gene silencing in FXS is related to events directly triggered by the repeats perhaps in a manner analogous to the way that Sat2 is silenced. This narrows the range of likely gene silencing mechanisms. The fact that no unique transcripts are seen in FX cells even when the FMR1 gene is reactivated and that the levels of small RNAs corresponding to the repeat are similar in normal and FX cells raises the possibility that FX gene silencing occurs via an RNA-independent mechanism. Indeed, while Sat2 was once thought to be silenced by an RNAi-based mechanism like the centromeric repeats of fission yeast (55), it is now known that the chromatin modifications on these repeats are unaffected in Dicer-null ES cells (56). This is consistent with an RNAi-independent mechanism (57). A similar lack of the effect of Dicer depletion has been reported for a transgene tandem array that normally undergoes silencing in mammalian cells (58). Interestingly, in the yeast mating-type region, heterochromatinization of the cenH DNA element, which has 96% homology to the yeast centromeric repeats, is regulated by both RNA-dependent and independent mechanisms. The RNA-independent mechanism is initiated by sequence-specific binding by members of the ATF/CREB family of transcription factors, Atf1 and Pcr1 (59,60). In the context of FX alleles, silencing may be initiated by one of the many CGG·CCG-repeat-binding proteins that have been described (29,30). It may be that these proteins recruit enzymes such as histone deacetylases and methylases that initiate gene silencing. Why 200 repeats would be required before silencing is seen is not clear. It may be that a critical level of binding of these proteins is only reached when the repeat number reaches 200 or that a factor that keeps the gene active also binds the repeat and is only titrated out when the repeat number reaches this threshold. The latter idea is appealing since it may also explain the hyperexpression of PM alleles. It is also possible that silencing is related instead to the length-dependent induction of the DNA damage response by FX repeats (61–64), since DNA damage leads to the recruitment of proteins involved in gene silencing such as SIRT1, EZH2, DNMT1 and DNMT3B (65,66).
Whatever the trigger for gene silencing, in addition to the shared histone profiles, both Sat2 and FX gene silencing involve the protein deacetylase SIRT1 (67,68). SIRT1 is thought to act on pericentric heterochromatin via recruitment of the SUV39H1 histone methyltransferase rather than G9a, the methylase more commonly active on developmentally repressed regions of the genome (67,68). Since these methylases represent different potential druggable targets, understanding how far the similarities between Sat2 and FX silencing extend may be useful in the development of strategies to reactivate FX alleles.
MATERIALS AND METHODS
Cell lines and reagents
Lymphoblastoid cell lines from normal (GM06865, GM06895) and fragile X patients (GM03200B, GM04025E, GM07924, GM09145) were obtained from Coriell Cell Repository (Camden, NJ, USA). Cells were grown in RPMI 1640 supplemented with 10% FBS and 1× antibiotic–antimycotic liquid consisting of penicillin, streptomycin and fungisone (Invitrogen, Carlsbad, CA, USA). To reactivate FX alleles, cells were treated with azadC as previously described (54). Chromatin immunoprecipitation assay kits and normal mouse and rabbit IgG were purchased from Upstate (Temecula, CA, USA). Antibodies against histone H3K9Me2 (ab1220), H3K9Me3 (ab8898), H3K27Me3 (ab6002) and H3K4Me2 (ab7766) were from Abcam (Cambridge, MA, USA). Antibodies against H4K20Me3 (39180) were from Active Motif (Carlsbad, CA, USA). Mouse embryonic stem cell chromatin immunoprecipated with antibodies to H3K27Me3 was a kind gift of Joseph Landry (NCI). H1 hESCs and differentiated EBs produced from these cells were a gift of Barbara Mallon (NIH Stem Cell Unit). The ESCs were propagated as described in http://stemcells.nih.gov/research/nihresearch/scunit/culture.asp#mef. The EBs were differentiated using DMEM supplemented with 10% FBS as described in http://stemcells.nih.gov/research/nihresearch/scunit/ebd.htm, except that the EBs were harvested after 4 days.
Chromatin immunoprecipitation assays
The chromatin immunoprecipitation (ChIP) assay kit from Upstate was used according to the manufacturer's instructions. Real-time PCRs on the immunoprecipitated DNAs were carried out using the Power SYBR™ Green PCR master mix (Applied Biosystems) and the primer pairs shown in Table 1. The location of the FMR1 amplicons is shown in Figure 1A. Regions of GAPDH, Sat2 and MYOD were amplified using the primers recommended by Abcam, Inc. for use as negative and positive controls for ChIP. Their sequences are listed in Table 1. For quantitation, the comparative threshold (Ct) method was used. The immunoprecipitated material was normalized to 5% of input and was expressed relative to the appropriate positive control. Student's t-test was used to calculate statistical significance.
|Primer name||Primer sequence|
|FMR1 exon1-R||5′-GTACCTTGTAGAAAGCGCCATTGG AG-3′|
|Primer name||Primer sequence|
|FMR1 exon1-R||5′-GTACCTTGTAGAAAGCGCCATTGG AG-3′|
Detection of antisense RNA by northern blot analysis
Total RNA was isolated using Trizol (Invitrogen) as per the manufacturer's instructions. Five micrograms of total RNA was mixed with equal volume of NorthernMax®-Gly sample loading buffer (Ambion, Austin, TX, USA) and incubated at 50°C for 30 min. The samples were then run on a 1% agarose gel in 1X NorthernMax®-Gly Gel Prep/Running buffer (Ambion). The gel was stained with ethidium bromide to check RNA integrity. Blotting to Hybond N+ nylon membrane (Amersham) was carried out according to standard procedures. The blot was then rinsed with 5× SSC and crosslinked using the UVStratalinker (Stratagene, Santa Clara, CA, USA). For northern blot analysis, three templates were generated to make sense riboprobes for the detection of the antisense RNAs. The template for the 5′ flank region was prepared by linearizing a plasmid containing bases −510 to −950 of the FMR1 gene. The template for the promoter region was prepared by linearizing a plasmid containing bases −19 to −121 of the FMR1 gene. A PCR fragment containing T7 sequence linked to bases +137 to +412 of the FMR1 gene was used as a template to generate sense probe in the exon 1/intron 1 region. Templates were transcribed using T7 polymerase plus (Ambion) and α-32P-CTP (8000 Ci/mmol, MP Biomed, Solon, OH, USA) and the RNA purified according to standard procedures. After prehybridization in Ultrahyb buffer (Ambion), hybridization with the riboprobe was carried out in Ultrahyb buffer at 68°C overnight. The blot was then washed twice with 2× SSC/0.1% SDS for 5 min at 68°C followed by two 15 min washes with 0.1× SSC/0.1% SDS. The blot was then exposed to a Kodak Biomax XAR X-ray film.
RNase protection assays
RNase protection assays were performed using the mirVana™ miRNA detection kit (Ambion). Briefly, the riboprobes were transcribed as described above and purified by running on a 5% denaturing polyacrylamide gel. The gel was exposed to X-ray film and the band containing the full-length probe was cut and RNA was eluted in elution buffer provided in the kit overnight at 37°C. The template for the DM1 probe was prepared as described elsewhere (48). Five micrograms of total RNA was hybridized with 5 × 105c.p.m. of the riboprobe at 50°C overnight. The hybridized probe/RNA was then digested with RNase A/T1 and precipitated. The protected fragments were detected by running the samples on a 15% (19:1, acrylamide:bis) polyacrylamide denaturing gel and exposing it to an X-ray film. Labeled pBR322 digested with MspI was used for sizing of the protected fragments.
This work was supported by a grant to K.U. from the Intramural Program of NIDDK, National Institutes of Health, USA (grant number DK057602).
The authors wish to thank Thomas Jenuwein (Max Planck Institute, Freiburg, Germany) for the kind gift of antibodies to H3K9Me3 and H3K27Me3 that were used for pilot experiments.
Conflict of Interest statement. None declared.