Human Molecular Genetics Advance Access originally published online on February 24, 2005
Human Molecular Genetics 2005 14(7):953-965; doi:10.1093/hmg/ddi089
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Published by Oxford University Press 2005.
Familial cases of point mutations in the XIST promoter reveal a correlation between CTCF binding and pre-emptive choices of X chromosome inactivation
,*1Molecular Pathology Section, Laboratory of Immunopathology, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Rockville, MD 20852, USA, 2Department of Development and Genetics, Uppsala University, Norbyvägen 18A, S-752 36 Uppsala, Sweden and 3Department of Psychiatry and Behavioral Science, State University of New York at Stony Brook, Stony Brook, New York 11794-81012, USA
* To whom correspondence should be addressed at: Molecular Pathology Section, LIP NIAID NIH, Twinbrook I, Room 1417, MSC-8152, 5640 Fisher Lane, Rockville, MD 20852, USA. Tel: +1 301 435 1690; Fax: +1 301 402 0077; Email: vlobanenkov{at}niaid.nih.gov
Received December 12, 2004; Revised February 4, 2005; Accepted February 14, 2005
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSSUPPLEMENTARY MATERIALREFERENCES |
|---|
The choice mechanisms that determine the future inactive X chromosome in somatic cells of female mammals involve the regulated expression of the XIST gene. A familial C(–43)G mutation in the XIST promoter results in skewing of X chromosome inactivation (XCI) towards the inactive X chromosome of heterozygous females, whereas a C(–43)A mutation found primarily in the active X chromosome results in the opposite skewing pattern. Both mutations point to the existence of a factor that might be responsible for the skewed patterns. Here we identify this factor as CTCF, a conserved protein with a 11 Zn-finger (ZF) domain that can mediate multiple sequence-specificity and interactions between DNA-bound CTCF molecules. We show that mouse and human Xist/XIST promoters contain one homologous CTCF-binding sequence with the matching dG-contacts, which in the human XIST include the –43 position within the DNase I footprint of CTCF. While the C(–43)A mutation abrogates CTCF binding, the C(–43)G mutation results in a dramatic increase in CTCF-binding efficiency by altering ZF-usage mode required for recognition of the altered dG-contacts of the mutant site. Thus, the skewing effect of the two –43C mutations correlates with their effects on CTCF binding. Finally, CTCF interacts with the XIST/Xist promoter only in female human and mouse cells. The interpretation that this reflected a preferential interaction with the promoter of the active Xist allele was confirmed in mouse fetal placenta. These observations are in keeping with the possibility that the choice of X chromosome inactivation reflects stabilization of a higher order chromatin conformation impinging on the CTCF–XIST promoter complex.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSSUPPLEMENTARY MATERIALREFERENCES |
|---|
To mediate dosage compensation, one of the two X chromosomes of female mammals is inactivated (1
,2). This process, which is called X inactivation, is normally random in the soma of eutherian mammals whereas it is parent-of-origin-dependent in the extra-embryonic tissues (3,4). Genetic dissection has revealed that X inactivation involves a complicated set of events that culminates in the activation of XIST on the future inactive X chromosome (5). The XIST transcript subsequently coats the chromatin fiber to irreversibly induce the inactive state along the bulk of the chosen X chromosome in somatic cells (6,7).
Although the mechanisms underlying the activation of XIST remain poorly understood, they are likely to involve the antisense non-coding transcript Tsix (8), because deletion of the Tsix gene or regulatory elements controlling Tsix expression leads to skewed X inactivation (8,9). In a more recent study, Ogawa and Lee (10) functionally characterized the ‘X-inactivation intergenic transcription elements’ (Xite) region. Xite is candidate enhancer element possibly regulated by tandem CTCF-binding sites in a chromatin-insulator within the Tsix/DXPas34 region (11). The mouse Tsix/DXPas34 region contains more than 40 CTCF motifs, whereas the corresponding region of the human X chromosome has more than 10 similar sites (11). X Chromosome inactivation (XCI) is not affected only by the regions of Xist. Further characterization of the 5' end of the Xist locus revealed specific zones of non-coding transcription where a series of targeted deletions and mutations also results in a pattern of XCI skewing similar to that resulting from deletions in the Xite or Tsix/DXPas34 regions (12). Thus, there are perhaps several additional candidate regions on both sides of Xist, which together may define a regulated balance between sense and antisense transcription across Xist prior to the onset of random X inactivation.
However, a similarly skewed X inactivation could be achieved more directly by modulating the activity of the XIST promoter itself. Although familial XIST mutations are rarely found in humans, it has been nevertheless possible in a few cases to link an extreme skewing for inactivation of one X chromosome to germline mutations of a particular region in the XIST promoter. For example, all females belonging to two unrelated families demonstrated preferential inactivation of X chromosomes carrying identical cytosine to guanine germline substitutions at the same –43 position upstream of the XIST transcriptional start site C(–43)G mutation (13). While suggesting that this mutation may hit a binding site for a transcriptional factor(s) important for proper regulation of the XIST promoter, the value of these findings remains unclear until such a factor(s) have been identified.
Intriguingly, another substitution of adenosine at the same cytosine at –43 position of the XIST promoter [C(–43)A mutation], on the ring X chromosome of a 3.5-year-old girl with short stature, facial dimorphism and developmental delay, was associated with escape from X inactivation (14). Because this mutation occurs at the same dC-base as the C(–43)G mutation, but is associated preferentially with the active X chromosome, we hypothesized that the opposite choices of XCI converged on the regulation of affinity for a trans-acting factor specifically recognizing the XIST promoter. Using an experimental protein-binding screen for the region of XIST promoter (from +31 to –1080 position) and chromatin immunoprecipitation (ChIP) analysis, this factor is identified here as CTCF.
CTCF is a highly conserved chromatin protein already implicated in a remarkably diverse number of functions including regulation of imprinted genes in soma and of XCI in mice (11,15,16). Different functions of CTCF appear to be mediated by binding to highly dissimilar 
50 bp-long target sequences by engaging varying combinations of its 11 Zn-fingers (ZFs) in establishing specific DNA base contacts. Because CTCF-contacting DNA base motifs have no single consensus that would allow one to reliably predict novel CTCF target sites (15), we initially utilized recombinant CTCF to map and characterize its interactions with the wild-type and mutant human XIST promoters in vitro. In addition, we studied binding of CTCF to the wild-type human and mouse XIST/Xist promoters in vivo. We demonstrate here that CTCF binds with a similar, relatively modest affinity to the same region of sequence homology in mouse, rabbit, horse and human XIST wild-type promoters (17), although the C(–43)G and C(–43)A mutations display increased and absence of CTCF binding, respectively.
On the basis of the fact that CTCF efficiently interacts with itself to mediate site-specific heterodimerization of CTCF–DNA complexes (18) and the fact that CTCF is in vivo associated with the active Xist promoter of only the inactive X chromosome, we discuss the possibility that the inherited point mutations at position –43 of the XIST promoter may pre-empt XCI choice by facilitating the formation of a higher order chromatin conformation permissive for XIST transcription.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSSUPPLEMENTARY MATERIALREFERENCES |
|---|
The mouse/human Xist/XIST promoters are specifically recognized by the multivalent 11 ZF DNA binding domain of CTCF
Attempts to predict CTCF-binding sequences, solely based on homology with the known targets, often results in finding of such sequences among numerous genomic regions previously tested experimentally negative for CTCF binding (19
). Therefore, we carried out a direct systematic search for CTCF-binding sites by using recombinant CTCF protein an electrophoretic mobility shift assays (EMSA) with two sets of eight consecutive overlapping DNA probes. One set of EMSA probes spanned the XIST promoter of human origin (Fig. 1B), whereas a similar set of probes covered the promoter from Mus musculus domesticus (Fig. 1C).
|
CTCF sites can be missed by conventional EMSA with DNA probes made as 20–60 bp long double-stranded oligonucleotides. These probes are too short to accommodate binding of the 11 ZF domain that generates
50–60 bp long DNAse I footprints on each strand with each individual ZF (plus the linker) requiring up to 5–6 bases of DNA for the efficient binding (20). Therefore, each set of 32P-end-labeled DNA probes used in Figure 1 was designed to include a 120–250 bp long overlapping DNA fragments. In addition, the overlapping of DNA fragments were designed such that an additional flanking DNA around 50 bp of a putative CTCF-binding sequence would become included in one of two consecutive fragments.
In both screens (Fig. 1B and C), sequence-specific binding of CTCF was detected only to DNA fragment no. 8, which in both human and mouse species contains the XIST/Xist transcriptional start site. The DNA sequences of the binding fragments showed 78% overall homology in several species, including human, mouse, rabbit and horse (17). A multi species sequence alignment of mammalian XIST promoter presented by Hendrich et al. (17) revealed within fragment 8 several near-identical nucleotide motifs, or phylogenetic footprints, that are the likely binding sites for conserved transcriptional factors (21). Because CTCF alone cannot account for all of these potential binding sites, it is possible that several other conserved factors in addition to CTCF may interact with the XIST promoter region around and immediately upstream of the transcriptional start site.
CTCF is preferentially associated in vivo with the active XIST promoter on the inactive X chromosome in human somatic cells, and with the homologous region of mouse Xist allele of the paternally inherited inactive X chromosome in female mouse placentas
To ascertain the in vivo relevance of our in vitro observations, we examined the interaction between CTCF and the XIST promoter of the inactive X in human somatic cells of female and male origin. Figure 2A shows the results of the ChIP assays with primary cultures of human mammary epithelial cells (HMEC) and normal human dermal fibroblast (NHDF) derived from healthy adult men and women. As positive control for in vivo CTCF binding, we used human c-MYC oncogene 5'-insulator site (N-site), known to be constitutively occupied by CTCF in normal cells (15,22). This experiment revealed that only female human cells, regardless of their origin, had CTCF bound to Xist promoter in vivo. In contrast, the control (N-site) has been pulled down equally well by CTCF antibody from both male and female cells. This result could be extended to freshly isolated peripheral blood leukocytes (PBLs) from healthy male and female donors for ChIP analyses. Figure 2B illustrates a 6–7-fold enrichment of DNA ChIPed with CTCF antibody from female PBLs, whereas there was no enrichment in ChIP analyses of male PBLs.
|
To examine more closely the possibility of a specific in vivo association between CTCF and the Xist allele of only inactive X chromosome in mammalian female cells, we turned our attention to the female mouse placenta, a tissue that exhibits selective inactivation of the paternal X chromosome. Dispersed, formaldehyde-crosslinked placental cells were subjected to ChIP analysis using affinity-purified rabbit polyclonal antibodies against the C-terminal portion of CTCF (23
). Figure 2C shows results of the ChIP assays comparing a positive, internal control, the maternal H19 ICR allele that interacts constitutively with CTCF in vivo (23), with the Xist promoter region positive for CTCF binding in the in vitro EMSA (Fig. 1C). These results show that the CTCF antibody pulled down both the H19 ICR and the Xist promoter fragments from female placenta, whereas only the H19 ICR could be recovered in CTCF ChIP material derived from the male placenta. These results agree well with the ChIP data from the human cells, reinforcing the possibility that CTCF was associated with the Xist promoter sequence of only the inactive X chromosome. To directly examine this contention, CTCF ChIP DNA samples derived from female placenta of interspecific M. musculus and M. spretus crosses were sequenced directly. This approach exploited an A/C polymorphism at position –88 upstream of the transcriptional start site of the Xist gene in M. musculus (C57BL/6J) compared with M. spretus allowing us to distinguish a maternally inherited allele from a paternally inherited allele in mouse placentas from female F1 mice. Figure 2D shows that the CTCF antibody pulled down only the paternally inherited allele (M. spretus), confirming our supposition that CTCF is associated only with the Xist promoter of the inactive X chromosome in female mouse placentas and, by extrapolation, also to female humans.
Germline mutations of the CTCF target sequence in the XIST promoter in families with skewed X inactivation
The cytosine at position –43 of the XIST promoter may play an important role in the X inactivation process: X chromosomes harboring a cytosine to guanine C(–43)G mutation at this position were preferentially inactivated in females of two unrelated families (13). Conversely, a cytosine to adenosine mutation at the same position, C(–43)A on the ring X chromosome of a young girl, was associated with escape from X inactivation (14). Because these mutations are located in the middle of fragment 8 of the XIST promoter, which is conserved among mammals and contains a target for CTCF, it seemed possible that these mutations might affect CTCF binding.
To examine this possibility, we used EMSA to test a 140 bp fragment of the human XIST promoter from –108 to +31 containing a wild-type sequence or the same fragment with either the C(–43)G or the C(–43)A mutation (Fig. 3A). While the C(–43)G mutation dramatically enhanced the affinity of CTCF binding when compared with the wild-type probe, CTCF binding was abolished by the C(–43)A mutation (Fig. 3B). This result observed with in vitro translated CTCF protein was also seen with nuclear extracts from various cells (Fig. 3B and data not shown). The specificity of CTCF binding in nuclear extracts was confirmed by a super-shift EMSA with CTCF antibodies (Fig. 3B). Promoter fragments with either the wild-type sequence or the C(–43)G mutation were efficiently bound by CTCF and the resulting CTCF–DNA complexes were super-shifted with antibodies against CTCF (Fig. 3B). In sharp contrast, no CTCF–DNA interactions could be detected with a fragment containing the C(–43)A mutation or the ‘dG-contacts mutation’ shown in Figure 3. The latter mutation was designed to eliminate CTCF binding by mutating a string of six major CTCF-contacting guanines (Fig. 3A). Those guanines were identified as bases involved in recognition of XIST by the 11 ZF domain of CTCF by DMS methylation interference assay (Fig. 5). Thus, the C(–43)A mutation eliminates CTCF binding as efficiently as the designed dG-contacts mutation that destroys the CTCF-recognition sequence.
|
-anti-CTCF antibodies (NE+
|
Figure 4 presents data from competitive EMSA carried out to compare binding affinities of CTCF to the wild-type and to the C(–43)G-mutated XIST promoter regions. A
100-fold excess of unlabeled cold DNA fragment bearing the C(–43)G mutation to the labeled wild-type DNA probe completely abolished formation of the normal CTCF DNA complex (Fig. 4A, lane 3). Under similar conditions, the fragment with the C(–43)A mutation had no effect on CTCF binding to the wild-type XIST site (Fig. 4A, lane 8). Similar experiments performed with a labeled EMSA probe containing the C(–43)G mutation showed that an excess of the wild-type competitor DNA fragment resulted in only a marginally decreased formation of the complex produced by the probe with the mutation (Fig. 4A, lane 13). No competitive effect on this complex was observed with the cold C(–43)A DNA (Fig. 4A, lane 16). Moreover, in EMSA experiments with increasing amounts of cold DNA fragment containing a high-affinity CTCF-binding site, DMD4 from the mouse H19 ICR (23), the labeled mutant C(–43)G XIST probe was much more efficient in competing for CTCF than the labeled wild-type XIST probe. On the basis of pair-wise comparisons of the DNA-bound CTCF bands in EMSA lanes 4–7 of Figure 4A with lanes 12–15, it is seen that an estimated 100-fold more cold H19 DMD4 competitor DNA was required to inhibit CTCF from binding to the C(–43)G mutant probe than to the wild-type XIST probe. These data indicate that the C(–43)G mutation results in a significantly enhanced binding of CTCF to the XIST minimal promoter.
|
These data were additionally supported by the results shown in Figure 4B. In that set of experiments, a radioactive high-affinity probe of the chicken FII insulator (24
) was titrated with cold wild-type and C(–43)G mutant fragments. The wild-type fragment was not very efficient in competing with the FII fragment, whereas the C(–43)G mutant fragment completely abolished formation of hot probe–CTCF complexes, even at lowest concentration used. Titration with a cold FII fragment is shown as a control in Figure 2C.
Differences in the base-contacting points in CTCF–DNA complexes formed with the wild-type and mutated XIST promoters
To identify potential changes in contact guanine nucleotides recognized by CTCF in human wild-type and mutant XIST fragments, we carried out DMS-methylation interference assays for both strands of each DNA fragment. Guanine residues essential for the CTCF–DNA interaction are significantly reduced or excluded from the bound (B) DNA probe lane when compared with the unbound or free (F) probe lane (Fig. 5A). CTCF-contacting guanines for both wild-type and mutant probes are located within the DNA sequence from positions –44 to –36 upstream from the transcription start site (+1), revealing that the C(–43)G mutation hits the core of the CTCF-binding site. In comparing major CTCF-contacting guanines in the wild-type sequence and in the C(–43)G-mutated sequence, we noted that the C to G substitution creates a new contact nucleotide for CTCF binding in the top strand, and also results in elimination of one of the many dG-contacts in the bottom strand (Fig. 5A). Thus, this mutation alters the pattern of CTCF-contacting DNA bases, thereby yielding a new CTCF-binding sequence. Taken together with the observation that the XIST promoter fragment with the C(–43)A mutation did not bind CTCF, this result documents that the two types of germline mutations at position –43 of the XIST promoter are associated with drastic but opposite effects on CTCF–DNA interactions at this site.
To define which region of the EMSA positive fragment is covered by CTCF, we performed DNase I footprinting assays. Figure 5B shows that with the wild-type probe, CTCF protects a region of 30 bp from nuclease attack on the top strand and 40 bp on the bottom strand. Both top and bottom strands of the footprints showed several sites of accessibility to nuclease attack within the normal DNA–CTCF complex, which are not detected in the DNA–CTCF complex formed with DNA bearing the C(–43)G mutation. Nuclease hypersensitive sites, marked ‘HS’ in Figure 5B, are induced by CTCF binding at two positions inside and at the flanks of the footprints on both the wild-type and the mutant probes. CTCF protects more sequence on the top strand of DNA containing the C(–43)G mutation than is found with the wild-type XIST promoter region.
The observation that different amounts of CTCF protein were required for protection of wild-type and C(–43)G mutant promoters is in further keeping with increased CTCF binding to the C(–43)G mutant XIST promoter. Although one footprint unit of recombinant CTCF protein (see Materials and Methods for definition of the ‘footprint unit’) resulted in the complete protection of the C(–43)G mutant site from nuclease attack, the same amount of CTCF protein did not protect the wild-type XIST promoter region from the DNase I attack (data not shown). However, increase in CTCF to DNA ratio, as shown in Figure 5B, generated a visible DNase I footprint on the bottom and top strands of the wild-type XIST promoter region, but the pattern of protection was not as strong as with the C(–43)G mutant XIST promoter region, thereby suggesting a difference in CTCF–DNA complex formation and in affinity of the two sites. These results are summarized in Figure 5B–D.
Figure 5E represents the alignment of mouse and human Xist promoters in the region of the CTCF footprint. The alignment was obtained by the GCG-package Best Fit program for maximal nucleotide identity in the two sequences, with a 10 bp incremental extension in the 5' direction starting from the first 10 bp window set for the +1 to +11 positions. The program generated this alignment for the two CTCF-binding DNA fragments from human and mouse promoters used in our EMSA analyses (Fig. 1). The alignment is displayed only for the top strands in the 5' to 3' direction corresponding to the 5' flanking regions of XIST/Xist promoters immediately upstream of the transcriptional initiator (INR) start site conserved in human, mouse, rabbit and horse Xist promoters (17).
In contrast to the alignment of longer Xist-promoter sequences by Hendrich et al. (17), the alignment optimized for 84 bp demonstrates that all dG-contacts essential for CTCF binding on both strands in the human XIST promoter are conserved in the mouse Xist promoter, including the guanine at position –43 marked in red as a capital C in the top strand and underlined. Moreover, the nucleotide affected by the germline mutations belongs to the core of critical CTCF ZF-contacting bases of the wild-type human site, whereas the best fit alignment indicated that the same critical core motif of CTCF-contacting nucleotides is conserved in the wild-type mouse (M. musculus) Xist promoter. Note that both the number and arrangement of the aligned ZF-contacting bases depicted in capital letters in Figure 5E follow very well the stereo-chemical rules for DNA recognition by CTCF-like ZFs (25) provided that (i) approximately 4–5 nucleotides can indeed be engaged by one single finger of CTCF into binding to both mouse and human Xist/XIST sites, as was previously demonstrated for another CTCF site by analyzing the effects of CTCF ZF-deletions on the total length of DNase I CTCF-footprint in the APP-promoter (20), and (ii) a certain flexibility for having one or two non-contacting nucleotides between the groups of the dG-contacts made by the two neighboring ZFs is allowed by the presence and by the fitting positioning of non-canonical long inter-finger linker-regions in CTCF, as previously suggested by Kim and Pabo (26).
Reasoning that a different pattern of DNA protection by CTCF seen with wild-type and C(–43)G mutant promoter sequences might result from different CTCF conformations upon binding to the wild-type and to the mutant XIST DNA fragments, we next examined which particular sets of individual ZF from the 11 ZF DNA-binding domain are engaged in the binding of CTCF to these two target sequences.
Formation of CTCF–DNA complexes on the XIST promoter containing the wild-type sequence or the ‘skewing mutation’ in the –43 target site is mediated by dissimilar contributions of different ZF subsets
We have previously demonstrated that CTCF uses different sets of ZFs to bind different DNA sequences (15). Here, we performed EMSA of the human and mouse wild-type and human C(–43)G mutant XIST promoters with five N-terminally ZF truncated and six C-terminally ZF truncated in vitro translated products of the 11 ZF domain of CTCF, one-by-one from each side of this DNA-binding domain (27) (Fig. 6). The results are summarized in three panels (Fig. 6 A–C) with each of the 11 boxes representing one individual ZF for C- and N-terminal consecutive ZF deletions separately. A black box depicts that a particular ZF is dispensable for the binding to the particular site, a gray box indicates incomplete loss of binding and hence only modest contribution for binding, whereas white boxes indicate ZFs which are absolutely essential for the CTCF binding to each site.
|
The wild-type promoters of both humans and mice showed practically identical modes of ZF utilization for CTCF binding (Fig. 6). For instance, C-terminal finger 8 is required for binding to both human and mouse XIST, because deletion of this ZF completely abolished CTCF binding of constructs 1–7, 1–6 and 1–5 to the human and mouse wild-type promoters. In contrast, C-terminal fingers 8 and 7 are not important for binding to the promoter with the C(–43)G mutation, whereas deletion of ZF 6 only partially reduced CTCF binding to this fragment (Fig. 6). On the other hand, N-terminal fingers 1 and 2 are dispensable for CTCF binding to the promoter with the C(–43)G mutation, but they significantly reduced CTCF binding to the human and mouse wild-type promoters. We, therefore, conclude that although the pattern of ZF utilization has been stable since the divergence of a common ancestor between mouse and human, a single mutation at the XIST promoter associated with skewed X inactivation dramatically changes the ZF utilization. We propose that this change in pattern directly or indirectly leads to conformational changes of CTCF that may be of functional importance.
Increased CTCF affinity to the C(–43)G mutant XIST promoter is paralleled by gain of enhancer-blocking activity
We have earlier demonstrated that the binding of CTCF to a single target generally correlates with the enhancer-blocking activity of a target in an affinity-dependent manner (28). By extrapolation, changes in CTCF-binding efficiency induced by the –43 XIST site mutations identified in vitro by DNAse I footprinting and EMSA might be accompanied with similar changes in insulator properties. To this end, fragments (–308 to +31) with human and mouse wild-type, with C(–43)G and C(–43)A germline mutations and with the designer dG-contact mutations were inserted (in the same orientations) between the ß-globin LCR enhancer and the neomycin reporter gene (Fig. 7). Following transfection and neomycin selection, we scored for numbers of colonies. In this assay, the insulator strength is inversely proportional to the numbers of surviving cells that register as neomycin-resistant colonies (24). Figure 7 shows that mouse and human wild-type CTCF-binding region, or the C(–43)A and dG-contacts mutant XIST promoter fragments, do not noticeably reduce the number of colonies when compared with the transfection control used in the enhancer-blocking assay. In a sharp contrast, the same transfection-selection assay with the same plasmid except introducing of the C(–43)G mutation results in a significantly reduced colony numbers. This result strengthens the link between CTCF-binding efficiency and functional properties at the XIST promoter.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSSUPPLEMENTARY MATERIALREFERENCES |
|---|
The convergence of C
A and CG mutations at the same (–43) position upstream of the XIST+1 transcription start site and their association to opposite patterns of XCI (13,14) potentially provides an important opening with respect to our understanding of the X inactivation mechanism. Indeed, the identification of a factor that binds to this region, which is affected by these mutations, will provide an important handle to uncover this phenomenon. On the basis of ChIP, EMSA, DNase I footprinting and methylation interference assays, we argue here that this factor is identical to CTCF. Not only do these mutations map to an 40 bp long CTCF-binding sequence within a string of guanines at –44 to –36 positions involved in establishing the DNA–protein contacts necessary for the specific interaction with CTCF, but CTCF also interacts with the Xist promoter in only female human and mouse cells. This could be explained by the demonstration here that CTCF associates to the preferentially active Xist allele in female mouse placenta. Taken together with a cluster of CTCF-binding sites in the chromatin insulator region earlier mapped at the 3' end of Xist (11), these findings suggest a dual function for CTCF in the X inactivation process.
The changes in the affinity of CTCF for the wild-type and mutant XIST promoters might be explained by our observations of how distinctively each of the two mutations affected critical CTCF-contacting DNA bases. The C(–43)A mutation removed one base from the recognition site, whereas the C(–43)G mutation altered two CTCF-contacting bases by creating a new dG-contact in the top strand and removing another dG-contact from the bottom strand. The competitive EMSA titration experiments showed that the newly created XIST site with the C(–43)G mutation has a higher affinity for CTCF than two previously characterized insulator sites with extremely strong CTCF binding. One explanation to account for this effect is based on the observation that enhanced affinity of artificial poly-ZF proteins could be achieved by adjusting and selecting slightly longer linkers between six CTCF-like C2H2-class fingers (26). Because CTCF contains several similarly longer and non-canonical linkers, these may allow CTCF to bind to different DNA sequences in distinctly different ways (15) providing necessary flexibility to make different number of contacts, and to form structurally distinct complexes with varying stability. Moreover, the combination of ZF used for binding the wild-type and C(–43)G mutant XIST promoters differed rather significantly. Therefore, it is most likely that the interaction with the C(–43)G mutant site involves different ZF subsets for reaching a near-ideal ‘fitting’ of the new dG-contacts to the stereo-chemical rules (25).
Although neither the wild-type nor the C(–43)A mutant XIST promoter displayed any significant chromatin insulator function, the C(–43)G mutant XIST promoter was very potent in preventing enhancer–promoter communications. Because the chromatin insulator function could be assigned to the mutant XIST promoter fragment that is associated solely with the inactive X chromosome and hence an active XIST, it is possible that the insulator protects from repressive, upstream cis elements in the affected patients (17). Alternatively, and not mutually exclusive, the CTCF target site in the XIST promoter might facilitate the recruitment of the pre-initiation RNA polymerase complex to enhance XIST transcription. This possibility is supported by the demonstration that CTCF is a component of the RNA polymerase II holoenzyme complex and that a range of intergenic CTCF-binding sites throughout the mouse genome display extensive binding to pol II, as determined by double ChIP on chip assays (Chernukhin et al., unpublished data). However, as indicated by any direct effect of the mutations on XIST promoter-directed transcription using transient transfection assays of plasmid DNA (data not shown), this possibility might apply only in the context of higher order chromatin conformations.
The skewed X inactivation profiles in patients with the point mutations at the pivotal –43 position of the XIST promoter suggest an interference with the random choice of XCI in affected patients. By extrapolation, the choice mechanism in normal somatic cells might be governed by the regulated affinity between CTCF and the wild-type XIST promoter. The Tsix transcript itself may be an important component of this process, because the Tsix transcriptional process which extends beyond the Xist promoter (8) could interfere with CTCF–Xist promoter interactions on both X chromosomes in females prior to the onset of the inactivation process.
Another parameter in this process might be the cooperativity of CTCF target sites in this region, as CTCF target sites within the H19 imprinting control region have been shown to interact in a cooperative manner via heterodimerization (18). If so, the choice process could involve the transformation of an initially weak CTCF target site into a strong site by cooperative interactions between CTCF in Xist and neighboring CTCF target sites at the choice/imprinting center in the Tsix gene (11). This possibility is in keeping with the demonstration here that CTCF is stably associated with the Xist promoter of only the inactive X chromosome and that CTCF has an essential role in mediating higher order chromatin conformations. The CTCF-dependent close physical proximity between the H19 ICR and a differentially methylated region (DMR)1 might be essential for the silencing of the maternal Igf2 gene as well as the coordination of regional epigenetic marks (Tiwari et al., submitted for publication). The existence of CTCF target sites within the DMR1 (Tiwari et al., submitted for publication) might enable formation of the H19 ICR–DMR1 complex via CTCF heterodimers. This proposal is in keeping with our demonstration that heterodimeric CTCF–DNA complexes organize interactions within the H19 ICR (18). By extrapolation, CTCF might be involved in the choice mechanism by generating one kind of chromatin conformation on the active X chromosome from Tsix and another on the inactive X chromosome from Xist.
Another possibility for the choice mechanism is involvement of testis-specific CTCF-paralog BORIS (29). It shares the same 11 ZF-DNA recognition domain with CTCF, but has absolutely no homologies to CTCF in ZF-flanking N- and C-terminal domains. Male germ cells deficient in both CTCF and me-dC content and undergoing resetting of epigenetic marks at H19 ICR CTCF sites (18,23,29) are positive for BORIS. Moreover, BORIS is active in undifferentiated but not in differentiated ES cells (data not shown). Taken together, these observations raise an interesting possibility that BORIS may participate in the choice mechanism for XCI, being expressed at the ‘right place at right time’.
In conclusion, our results indicate a close association between the choice of the future inactive X chromosome and binding properties of a CTCF target site in the XIST promoter. Regardless of the underlying mechanisms, it is remarkable that a point mutation at a single but pivotal CTCF target site on one of the X chromosomes can be linked to the activation or inactivation of the opposite chromosome in humans. We postulate here that this process requires the CTCF-dependent organization of higher order chromatin conformation that impinges on the stability of the CTCF–XIST promoter complex during early stages of X inactivation.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSSUPPLEMENTARY MATERIALREFERENCES |
|---|
Tissue and cell materials
Fetal mouse placenta derived from M.m. musculusxM. spretus interspecific crosses was dissected under microscope and tissue cells dispersed prior subsequent analyses. The male and female fetal placentas (E14.5) were discriminated by PCR-based genotyping. Live primary cultures (at passages 2–4) of HMEC and NHDF derived from healthy adult men and women, purchased from Clonetics and Cambrtex, were used in ChIP assay to show an association between CTCF and the human XIST promoter. In addition, we used freshly isolated PBLs from normal adult XX versus XY blood. Blood (100 ml) was drawn by authorized personnel at the Red Cross Donor Facility (Rockville, MD, USA) from healthy male and female volunteers of 30–40 years of age, and PBL fraction purified using LSM (Mediatech Cellgro, VA, USA) according to the manufacturer's instructions. Approximately 100 million cells from isolated male and female PBL suspensions were taken for each preparation of fixed chromatin for each ChIP. Minor modifications for ChIP analyses with human cells included addition to all buffers immediately before use of protease inhibitors as recommended by a manufacturer (Sigma; Cat. no. P-8340).
Chromatin immunopurification analyses
DNA and protein components of mouse or human chromatin were cross-linked in situ by incubation of dispersed cells in 1% formaldehyde for 10 min at 37°C. The DNA–protein complexes were immunopurified using anti-CTCF antibody (Upstate) or a mixture of nine anti-CTCF MABs (Supplementary Material) and protein A4 Fast Flow Sepharose beads (Pharmacia-Upjohn) as described earlier (23).
The immunopurified DNA from mouse tissues was PCR amplified with primers spanning the Xist promoter: sense primer 5'-CACGCGTCATGTCACTGAGCTTACGTACCTC-3'; antisense primer 5'-GAACCGCACATCCACGGGAAACGAG-3'. PCR conditions were 95°C for 5 min, 29x(94°C for 45 s, 62°C for 1 min, 72°C for 45 s). As an internal control to allow comparison between the formaldehyde cross-linked samples, the CTCF target site no. 3 within the H19 ICR was amplified using primers: sense primer 5'-CTCAGTGGTCGATATATGGTTT-3'; antisense primer 5'-TGAGTCAAGTTCTCTTGGTTC-3'. PCR conditions were 95°C for 3 min, 28x(94°C for 40 s, 54°C for 40 s, 72°C for 40 s). The PCR products were analyzed on a 1.5% agarose gel and visualized by SYBR® Green. To distinguish the maternally inherited allele from the paternally inherited one in mouse placentas from female progeny of two different strains, the CTCF chromatin immunopurification (ChIP) material from the interspecific cross was subjected to sequencing exploiting an A/C polymorphism at position –88.
The immunopurified DNA of human origin was PCR amplified with primers spanning the XIST promoter: sense primer 5'-caaatcacaaagatgtccggctttcaatcttctaggc-3' and antisense primer 5'-aagcttccagccccgagagagtaagaaatatggctg-3'. As positive controls for the human XIST site in CTCF ChIP analyses, we used well known human c-MYC oncogene 5'-insulator site (N-site) that was amplified by using primers: sense primer 5'-cctgaaagaataacaaggaggtggctggaaacttg-3'and antisense primer 5'-gcaaattactcctgcctccaggcctttg-3'.
Quantitation of ChIP DNA was performed by real-time PCR analysis using the ABI Prism 7900 Sequence Detection System according to Applied Biosystem's SYBR® Green PCR Master Mix Protocol. Real-time PCR was carried out in triplicate of ChIP, control and input DNA samples at the following thermal cycling parameters 95°C for 10 min and 40 cycles of 95°C for 15 s and 60°C for 1 min. Data were collected at 60°C and analyzed by applying comparative CT method as described in ABI User Bulletin (Biosystems, updated 04/2001 #933) and in Litt et al. (30). Fold enrichment for a particular target sequence was determined by calculation of ratio of the amount of the target sequence in immunoprecipitation (IP) to the amount of the target sequence in input (In) DNA. Briefly, we used the equation
 |
Sequencing of mouse Xist promoter
To find a sequence polymorphism(s) at or near the CTCF-binding site, we analyzed the DNA sequence of mouse Xist minimal promoter from five strains of mice [129X1/SvJ mouse_(Xce-a); C57BL/6J mouse_(Xce-a); BALB/cJ mouse_(Xce-b); CAST/EiJ mouse_(Xce-c); and SPRET/Ei mouse_(Xce-d)] with the previously characterized strain-specific haplotypes of the ‘X-controlling element’ (Xce) from isolated DNA purchased from the Jackson Laboratory. Mouse DNA from all mouse strains was amplified from position –32 to +690 relative to the Xist transcriptional start site by sense primer: 5'-GACAGTTCTTTTAAGTTAGCAGTGTCTCTGGGG-3' and antisense primer 5'-GAACCGCACATCCACGGGAAACGAGC-3'. The amplification protocol was as follows: denaturation at 95°C for 5 min, followed by 35 cycles at 95°C for 1 min, 66°C for 1 min, 72°C for 1 min and final extension 72°C for 10 min. The primers used for PCR amplifications were designed based on the high-quality wild-type XIC/Xic sequence annotations (31). The nucleotide sequences were submitted to GenBank with accession nos AY618353 for 129X1/SvJ mouse_(Xce-a); AY618354 for C57BL/6J mouse_(Xce-a); AY618355 for BALB/cJ mouse_(Xce-b); AY618356 for CAST/EiJ mouse_(Xce-c); AY618357 for SPRET/Ei mouse_(Xce-d).
Plasmid constructs
The PCR fragment of the human, CTCF-positive XIST minimal promoter was cloned into the pCRII-TOPO vector (Invitrogene) resulting in the wtXIST-pCRII-TOPO plasmid. The PCR fragment was amplified from position –309 to +31 relative to the XIST transcriptional start site by sense primer: 5'-GTTATGGAGGATTTTAGCATTAATTATTG-3' and antisense primer 5'-CCAGCCCCGAGAGAGTAAGAATATG-3'. The SacI sites at the 5' end of primers were introduced to facilitate the subcloning of the fragment into either the enhancer-blocking plasmid or the luciferase-reporter pGL-2 plasmid. The amplification protocol was as follows: denaturation at 95°C, 5 min, followed by 35 cycles at 95°C for 30 s, 64°C for 30 s, 72°C for 1 min, final extension 72°C for 10 min.
Both the naturally occurring C(–43)G and C(–43)A mutations and the artificial GCCCCCC/AAAGCTT mutation (named dG-contacts mutation) were generated in the wtXIST-pCRII-TOPO plasmid by the QuikChange site-directed mutagenesis kit (Stratagene). DNA sequencing on Li-Cor Sequencer confirmed the accuracy of the cloning as well as the mutagenesis of the insert. These plasmids [wild-type, the C(–43)G and C(–43)A mutations and the dG-contacts mutation] from the XIST-pCRII-TOPO-based vectors were used for generating of insulator constructs, and as the templates for DNase I footprinting and methylation interference analyses.
The insulator assay series of plasmids were generated on the basis of pJC 5-4 plasmid as has been described earlier (32). The HS4 globin insulator was inserted into SacI site of the parent plasmid between an enhancer and the mouse 
-globin promoter–Neo reporter gene. To substitute this SacI HS4 insulator with the wild-type, C(–43)G mutant, C(–43)A mutant and ‘dG’ mutant XIST promoter fragments, the inserts were cut out by SacI enzyme from the corresponding pCRII-TOPO-based constructs described earlier, and ligated into the SacI site of the pJC 5-4 plasmid. To normalize the colony number, we used pJC 5-4 vector with deleted HS4 globin insulator.
Nuclear extracts and in vitro transcription–translation
Full-length human CTCF and the 11 ZF CTCF-binding domain were in vitro translated from pET-7.1 and pET-11 ZF, respectively (33), using the TnT reticulocyte lysate-coupled in vitro transcription–translation system (Promega). Twelve plasmids with sequentially truncated ZFs of the CTCF-binding domain (from amino acid position 236–622 of the human CTCF protein) were in vitro synthesized in the presence of 35S-Met (27). Nuclear extracts were prepared from K562 and other cell lines by using the 2xNUN extraction solution as described elsewhere (34). Partially purified recombinant CTCF was obtained from Pichia pastoris extract by single step SP cation exchange chromatography as described previously (20). Fractions of 0.5 ml were collected and CTCF-binding activity was monitored by EMSA with CTCF target from the APP promoter (20). The fraction containing peak binding activity was used for DNase I footprinting to determine the minimal amount (1 µl) of recombinant CTCF protein sufficient to fully protect CTCF-binding site in the APP promoter from the nuclease attack. In the experiments with the XIST promoter, 1 µl of this fraction was taken as ‘one CTCF footprint unit’.
Electrophoretic mobility shift assays
Eight consecutive fragments of the human XIST minimal promoter from position –1080 to +31 relative to +1 start site and eight consecutive fragments of the mouse Xist minimal promoter from position –1020 to +34, one-by-one covering partially overlapping promoter DNA sequences of interest, were PCR amplified. The fragments were simultaneously end-labeled on either strand during amplification by using pairs of 20–28 bp long primers that were end-labeled at their 5' ends using 32P--ATP and T4 polynucleotide kinase. The DNA–protein binding incubation was carried out in a buffer containing standard phosphate-buffered saline with 5 mM MgCl2, 0.1 mM ZnSO4, 1 mM dithiothreitol, 0.1% Nonidet P-40 and 10% glycerol in the presence of poly (dI–dC) plus poly (dG)·poly (dC). The nuclear extract-based EMSA was performed in the presence of unlabeled double-stranded oligonucleotides containing Sp1 like sites and ‘G-string’-binding factors as described (35). To reach the same shift mobility for in vitro translated CTCF and nuclear extract's CTCF, we added 10 µl of a TnT reticulocyte lysate to the nuclear extract-based binding reaction. For ‘super-shift’ assays, an antibody against the C-terminal domain of CTCF (Upstate Biotechnology) was used. The EMSA reaction mixtures of a 20 µl final volume were incubated for 30 min at room temperature followed by electrophoresis on 5% non-denaturing polyacrylamide gels.
The competition assay was performed with wild-type, C(–43)G and C(–43)A mutant XIST fragments PCR amplified by sense primer: 5'-CAAATCACAAAGATGTCCGGCTTTC-3' and antisense primer 5'-CCAGCCCCGAGAGAGTAAGAATATG-3' from wt XIST-, C/G- and C/A-mut.Xist-pCRII-TOPO plasmids, respectively. The specificity of the protein–DNA interactions were assessed by adding 100-fold molar excess of unlabeled wild-type or C/G mutant probes in reactions with full-length CTCF and end-labeling C/G mutant or wild-type PCR fragments, respectively. The C(–43)A mutant was used as cold competitor in reaction with full-length CTCF and end-labeling wild-type and C/G mutant PCR fragments in 100-fold molar excess. Finally, the well-characterized CTCF-binding site, site no. 3 from the human H19 imprinting control region (23), was used with a 500-, 100-, 10- and 1-fold molar excess of cold probe.
DNase I footprinting and methylation interference assay
We performed DNase I footprinting and methylation interference analyses with in vitro translated, full-length CTCF and the 141 bp XIST wild-type and C(–43)G mutant promoter fragments amplified by sense primer: 5'-CAAATCACAAAGATGTCCGGCTTTC-3' and antisense primer 5'-CCAGCCCCGAGAGAGTAAGAATATG-3' from wtXIST- and C/G-mut.XIST-pCRII-TOPO plasmids, respectively. Both fragments were labeled at their 5' ends on either the top or the bottom strand. Following gel purification, the fragments were either incubated with CTCF and then partially digested with DNase I or partially methylated at guanine residues by dimethyl sulfate and then incubated with CTCF, as has been described (27).
Insulator assay
Insulator assays were performed in human K562 erythroleukemia cells as has been described (24). All plasmid DNAs, prepared by using Qiagen Maxiprep Kit, were linearized with SalI, phenol–chloroform extracted and ethanol-precipitated. In total, 100 ng of linear plasmid DNA was mixed with 107 of K562 erythroleukemia cells and electroporated at a pulse field strength of 200 V and 960 µF capacitance with no resistor. The transfected cells were grown in 20 ml DMEM medium containing 10% fetal bovine serum without G418 at 37°C for 24 h. The following day, the cells were seeded on 150 mm tissue culture plates in soft agar prepared by standard procedure with DMEM+FBS containing 750 µg/ml of G418. Finally, the numbers of G418-resistant cell colonies formed during 3 weeks of neomycin selection were counted. In this assay, which was repeated at least three times, the relative numbers of colonies reflect the potency of an insert to block enhancer-mediated activation of the Neo-driving promoter.
|
SUPPLEMENTARY MATERIAL |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSSUPPLEMENTARY MATERIALREFERENCES |
|---|
Supplementary Material is available at HMG Online.
|
ACKNOWLEDGEMENTS |
|---|
We wholeheartedly thank Dr Herbert Morse III for critical reading of the manuscript, Dr Anita Mattsson for the mouse placenta work, Drs Chen-Feng Qi and Shao Xiang for performing the immunoblot analysis of nine CTCF monoclonal antibodies, and Drs Yoo-Wook Kwon and Hanlim Moon and other members of the Lobanenkov Lab for encouragement and support during the course of this work. We would also like to acknowledge an anonymous reviewer suggested our search for sequence polymorphisms near CTCF site by sequencing Xist promoters from five particular mouse strains. This work was supported by intramural research funding to V.V.L. and grants from the National Institutes of Health (Grant NS30994) to W.Q., the Swedish Science Research Council (VR) to R.O., the Juvenile Diabetes Research Foundation International (JDRF) to R.O., the Swedish Cancer Research Foundation (CF) to R.O., the Swedish Pediatric Cancer Foundation (BCF to R.O. and the Wallenberg and Lundberg Foundations to R.O.
|
FOOTNOTES |
|---|
These authors contributed equally. 
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSSUPPLEMENTARY MATERIALREFERENCES |
|---|
- Lyon, M.F. (1961) Gene action in the X-chromosome of the mouse (Mus musculus L.). Naturwissenschaften, 190, 372–373.
- Lyon, M. (1999) Imprinting and X chromosome inactivation. In Ohlsson, R. (ed.), Genomic Imprinting: An Interdisciplinary Approach. Springer-Verlag, Berlin/Heidelberg/New York, Vol. 25, pp. 73–90.
- Takagi, N. and Sasaki, M. (1975) Preferential inactivation of the paternally derived X chromosome in the extraembryonic membranes of the mouse. Nature, 256, 640–642.[CrossRef][Medline]
- West, J.D., Frels, W.I., Chapman, V.M. and Papaioannou, V.E. (1977) Preferential expression of the maternally derived X chromosome in the mouse yolk sac. Cell, 12, 873–882.[CrossRef][ISI][Medline]
- Penny, G.D., Kay, G.F., Sheardown, S.A., Rastan, S. and Brockdorff, N. (1996) Requirement for Xist in X chromosome inactivation. Nature, 379, 131–137.[CrossRef][Medline]
- Brown, C.J., Hendrich, B.D., Rupert, J.L., Lafreniere, R.G., Xing, Y., Lawrence, J. and Willard, H.F. (1992) The human XIST gene: analysis of a 17 kb inactive X-specific RNA that contains conserved repeats and is highly localized within the nucleus. Cell, 71, 527–542.[CrossRef][ISI][Medline