Abstract
X chromosome inactivation (XCI) is an epigenetic process that almost completely inactivates one of two X chromosomes in somatic cells of mammalian females. A few genes are known to escape XCI and the mechanism for this escape remains unclear. Here, using mouse trophoblast stem (TS) cells, we address whether particular chromosomal interactions facilitate escape from imprinted XCI. We demonstrate that promoters of genes escaping XCI do not congregate to any particular region of the genome in TS cells. Further, the escape status of a gene was uncorrelated with the types of genomic features and gene activity located in contacted regions. Our results suggest that genes escaping imprinted XCI do so by using the same regulatory sequences as their expressed alleles on the active X chromosome. We suggest a model where regulatory control of escape from imprinted XCI is mediated by genomic elements located in close linear proximity to escaping genes.
- X chromosome inactivation
- chromosome conformation
- topologically associated domain
- trophoblast stem cell
THE three-dimensional shape of chromosomes has a direct impact upon gene regulation, as chromatin looping mediates the interaction of enhancers with transcriptional start sites (TSSs) (Lieberman-Aiden et al. 2009; Li and Reinberg 2011; Krivega and Dean 2012). Analysis of genome-wide interactions suggests that chromosomes self-organize into topologically associated domains (TADs) that are ∼0.8–1 Mb in linear length (Dixon et al. 2012). Loci within a TAD are more likely to interact with each other as opposed to forming interactions with loci residing in other TADs.
Chromosomal interactions are thought to play a pivotal role in the epigenetic process of X chromosome inactivation (XCI) (Splinter et al. 2011). During XCI, mammalian females transcriptionally inactivate one X chromosome (Xi) per somatic cell to balance X-linked gene dosage with males (Chow and Heard 2009). Whereas genes on the active X chromosome (Xa) are thought to form stable interactions with other loci on the Xa (cis) and other chromosomes (trans), the interactions formed by Xi-linked loci are relatively less established, suggesting that Xi chromatin folds in a random manner (Splinter et al. 2011).
In the mouse, two forms of XCI are observed: imprinted XCI and random XCI. Imprinted XCI occurs within extra-embryonic tissues and is characterized by the exclusive inactivation of the paternally derived X chromosome (Xp) (Takagi and Sasaki 1975). Random XCI occurs within somatic tissues of the developing embryo and adult (Lyon 1961). While imprinted and random XCI may initiate via distinct mechanisms (Kalantry et al. 2009), the genetic programs required for the maintenance of both forms appear similar (Marahrens et al. 1997; Kalantry et al. 2006; Jonkers et al. 2009; Shin et al. 2010).
Interestingly, a few genes are known to escape both imprinted and random XCI and are expressed from both X chromosomes (Berletch et al. 2011; Calabrese et al. 2012). Profiles of XCI escape vary among different cell types; the number of escape genes, termed escapers, ranges from 3% to 25% of all X-linked genes (Berletch et al. 2011). The molecular mechanism underlying escape has proven elusive. It is suggested that escapers physically associate with each other to facilitate their expression from the Xi (Splinter et al. 2011).
To date, no study has correlated escape from XCI with any genomic regulatory element. It is possible that a single genomic element could act as a locus control region (LCR) for escape. Alternatively, escape-specific sequences may be regionally scattered along the X chromosome, licensing escape on a region-by-region basis. Finally, a sequence specific to escape may not exist and mechanisms of escape may vary from gene to gene.
To distinguish among these possibilities, we used allele-specific circular chromosome conformation capture 4C-sequencing (4C-Seq) to identify genomic interactions occurring at escapers within female F1 hybrid trophoblast stem (TS) cells. TS cells undergo imprinted XCI. Therefore, F1 hybrid TS cells serve as an ideal system for allele-specific analysis of mechanisms underlying escape from XCI, as no mutations are necessary to bias XCI.
Our results suggest that escape from imprinted XCI happens on a gene-by-gene basis. We demonstrate that escapers do not converge upon a single LCR, and we did not identify sequences consistent with regionally dispersed escape regulatory elements. Rather, regardless of escape status, genomic regions in close linear proximity tend to share regions of contact. Furthermore, we show that unlike genes subject to XCI, escapers are located in close linear proximity to putative active enhancer elements that are also found on the active X chromosome. We suggest that genes escaping imprinted XCI utilize regulatory elements in close linear proximity and that mechanisms of escape may vary from gene to gene.
Materials and Methods
TS cell derivation and culture
F1 hybrid TS cells were derived and cultured as described previously (Himeno et al. 2008; Calabrese et al. 2012).
Allele-specific 4C-Seq
Allele-specific 4C was based upon previously published work (Splinter et al. 2011), with several changes (see Figure 1A and Supporting Information, File S1). 4C anchor primer sequences were designed to capture informative single nucleotide polymorphisms (SNPs) during sequencing. To generate 3C library templates, chromatin from 3 × 107 TS cells was isolated, digested, ligated, and purified using lysis conditions adapted to TS cells. Each 4C template was generated with 12.5 μg of 3C library and was performed as previously described (Splinter et al. 2011). One microgram of 4C template was amplified per 4C primer pair using optimized PCR conditions. The 4C PCRs were then size selected between 150 bp and 650 bp, purified, and further amplified in a linear range with outer sequencing adapter primers. Amplified 4C libraries were purified with AmpureXP beads (BioRad) and submitted for paired-end 100-bp sequencing on Illumina HiSeq 2000 sequencers (Illumina). Biological replicates for each anchor were performed and sequenced separately.
The active and inactive X chromosomes form distinct conformations. Cis interactions generated by Xa and Xi anchors are mapped to their chromosomal positions and are depicted by black vertical lines. Horizontal blue, brown, and gray bars indicate the chromosome to which interactions mapped: the Xi (B6 genome), the Xa (Cast genome), and no allelic call, respectively. The number of interactions detected for each allelic call is noted to the right. Asterisks indicate anchor point location. Anchor points are listed along the y-axis according to their position along the X chromosome with Xa and Xi anchors colored in blue or brown, respectively. Anchor points from genes subject to XCI or escaping XCI are highlighted in light red and light green, respectively. Genomic position along the X chromosome is listed on the x-axis in megabases. Also see Figure S1 and Figure S2.
Filtering and statistical analysis of allele-specific 4C-Seq reads
See File S1 for in-depth details of all filtering and statistical analysis. Briefly, raw sequencing reads were filtered through custom Perl and R scripts (available upon request). Anchor fragment portions of reads were filtered by known SNPs to identify the anchor point of origin for any given read pair (either Xa or Xi). The unknown portion of reads was then mapped to the B6 and Cast genomes using Bowtie (version 0.12.7) (Langmead et al. 2009). The Cast genome was generated by substituting identified SNPs (Yalcin et al. 2011) into the mm9 B6 consensus genome. Reads were paired and files were generated for statistical analysis. PCR amplification bias was not detected in any sequencing dataset, allowing the use of the full range of sequencing reads.
Statistical enrichment of sequencing reads (Williams et al. 2014) was performed on each biological replicate. Briefly, a sliding window analysis was performed using window size corresponding to three 3C restriction fragment lengths. Analysis was performed per chromosome and raw reads were shuffled randomly across chromosomes 1000 times to generate significance thresholds. Read-containing windows passing empirically determined thresholds were called interactions. Per anchor, interactions from each biological replicate were then compared to generate a final list of genomic coordinates. Interactions were assigned to the B6, Cast, B6/Cast (equal contribution of both alleles), or NoCall (no allelic data), based upon the presence of informative SNPs within reads contributing to interactions.
RNA-Seq, DNAse-Seq, ChIP-Seq, genomic repeat analysis
All allele-specific RNA-Seq, DNAse-Seq, and ChIP-Seq datasets in C/B TS cells were obtained (GEO accession GSE39406) and analyzed for the whole genome, based upon previously described methods (Calabrese et al. 2012). A table of genomic repeats and their locations in build mm9 were obtained from the University of California Santa Cruz genome browser (Meyer et al. 2013). Genomic feature indices were generated by dividing the total number of identified features over the total number of bases covered by all interactions in a dataset. Paired, two-tailed t-tests or paired, two-sample t-tests were used to determine P-values when comparing between homologous alleles or among anchors on the same chromosome, respectively. Custom Perl scripts (available upon request) were used to identify the overlap of genomic features. Enrichment of any feature was measured by calculating the ratio of random occurrences over number of permutations (P ≤ 0.05, FDR ≤ 0.05, see File S1 for further details).
Binning of 4C interaction data for comparison and correlation analysis
To properly compare the genomic coordinates of interaction profiles among anchors, each anchor interaction profile was transformed into a binned profile. For Pearson correlation analysis of cis interactions, allelic data for each anchor was divided into bins of 500 bp. For common shared genomic regions, bins were also set at 500 bp, and allelic data were analyzed separately from nonallelic data. Per anchor, reads contributing to called interactions were sorted into appropriate bins. Binned data were then binarized to 1 or 0, depending upon the presence or absence of data within a bin.
Because interaction frequency is expected to decay over linear distance, an adjustment value for each 4C pair was determined using
where a and b are the coordinates (in base pairs) of the TSS of the genes being compared and chrX is the total length of the X chromosome (based on mm9). Positive and negative Pearson R-values for each 4C anchor profile comparison were then divided or multiplied, respectively, by the corresponding adjustment value.
Identification of putative active enhancer elements near genes
Allelic peaks for H3K27Ac ChiP-Seq, H3K4me1 ChiP-Seq, and DNAse in C/B TS cells were compared to generate a conservative list of putative active enhancer elements. A putative enhancer element was called only at regions of overlap when all three features contained allelic data. If overlap occurred, but any of the features lacked an allelic assignment, the putative active enhancer was identified, but given a “no allele” assignment. When locating putative enhancer elements in close proximity to genes, we only used genes where allelic contribution to overall levels could be determined, based on RNA-Seq in C/B TS cells (Calabrese et al. 2012). Genes were then divided into two categories: genes subject to XCI and escaper genes. Fifty kilobases were then added to the annotated TSS and transcriptional termination of each gene. The number of putative enhancer elements located within each gene body (±50 kb) was normalized per gene by dividing the total number of enhancers found by the total kilobases searched. Mouse embryonic stem (ES) cell TAD boundaries (Dixon et al. 2012) were obtained (GEO accession GSE35156) and compared to the coordinates of identified enhancer elements.
Fluorescence in situ hybridization confirmation of interactions
RNA/DNA fluorescence in situ hybridization (FISH) was performed identically to Calabrese et al. (2012). See File S1 for additional information.
Results
Generation of allele-specific interactions
We performed allele-specific 4C-Seq (Splinter et al. 2011) at the TSS of escapers to test if the physical association of escape promoters with regulatory elements governs escape from XCI. A female F1 hybrid TS cell line between the strains Mus musculus castaneous (Cast/EiJ or Cast) and Mus musculus domesticus (C57BL6/N or B6) were used for our analysis (Calabrese et al. 2012). TS cells undergo imprinted XCI (Mak et al. 2002), therefore our Cast/B6 (C/B) TS cells harbor a paternally inherited B6 Xi and a maternally inherited Cast Xa (Calabrese et al. 2012).
We modified a previously published 4C-Seq protocols and statistical analysis for our study (Splinter et al. 2011) (Figure S1, A and B and Materials and Methods). For all gene promoters (termed anchors), one primer within each 4C primer pair hybridizes upstream of a known SNP (Yalcin et al. 2011), allowing for the identification of the anchor allele of origin for every sequencing read (Figure S1A, red lines). Appropriate 4C-Seq anchor points were identified using allele-specific RNA-Seq datasets in C/B TS cells (Calabrese et al. 2012). Seven anchor points were chosen: five escapers (Nkap, Taf1, Ogt, Ftx, and Kdm5c) and two X-inactivated genes (Rlim and Huwe1).
Statistically significant genomic interactions were categorized by their location and the presence of informative SNPs (see Materials and Methods). Cis and trans interactions that could confidently be assigned to an allele were termed allelic, whereas interactions lacking sufficient informative SNPs were termed non-allelic. Depending upon the anchor, allelic interactions account for ∼17–37% of all data (Table S1).
FISH of randomly selected interactions in C/B TS cells was used to confirm our 4C-Seq pipeline (Figure S1, B and C). On both the Xi and Xa, measured distances between FISH signals located within interactions were shorter than distances measured between FISH probes located outside of interactions (Figure S1C), suggesting that we reliably detected allele-specific interactions within the genome.
The Xi in TS cells generates stable interactions
A broad analysis of our 4C-Seq data was performed and we tested if the contact profiles generated by Xa anchors differed from those generated by Xi profiles. We found no difference in the overall number (P = 0.722) and proportion of cis to trans (P = 0.215) interactions generated by TS cell Xi and Xa anchors (Figure 1, Figure S2, and Table S1). Additionally, anchor origin and transcriptional activity had no effect on the linear distance bridged by cis interactions (Figure 1). In agreement with previous chromosome conformation studies (Lieberman-Aiden et al. 2009; Dixon et al. 2012), Xa and Xi anchors preferred to interact with their X chromosome of origin (Figure S2 and Table S1), with most interactions occurring within a few megabases of the anchor point (Figure 1). Correlation analysis of cis interactions demonstrated that Xa and Xi contact profiles were largely different (Figure 1 and Table S2). Additionally, Xi loci were located closer together than Xa loci (Figure S1C). Taken together, our data suggest that the Xi in TS cells behaves in a similar manner to other chromosomes, though it likely adopts a different overall structure as compared to the Xa.
Contact with transcriptionally active genes does not correlate with escape from XCI
In general, transcriptionally active genes tend to interact with each other, while silent genes interact with other silent genes (Lieberman-Aiden et al. 2009). Therefore, we sought to test if genes escaping XCI tended to interact with other active genes.
We classified the transcriptional activity within interacting regions using RNA-seq data from C/B TS cells (Calabrese et al. 2012). Normalized indices (genes/kilobase) were used to compare anchor points since the number and median width of interactions varied among anchors (Table S1). Genes found within interactions were grouped into three classes corresponding to their expression status: expressed (transcriptionally active), repressed (transcriptionally silenced due to non-XCI mechanisms), and inactivated (transcriptionally inactivated due to XCI).
The transcriptional profiles found within the interactions generated by X-inactivated and escaper genes did not differ (Table S3 and Figure 2). X-inactivated loci and escaper loci were equivalent in their ability to contact active genes (P = 0.737, Figure 2A), silent genes (P = 0.788, Figure 2B), and genes escaping XCI (P = 0.913, Figure 2C). Thus, in TS cells, the association of a TSS with other actively transcribed genes is not a likely mechanism for escape from imprinted XCI.
Interactions with active genes do not correlate with transcriptional status. Gene indices (genes/kilobase) per Xi anchor point for specific gene classes are plotted. (A) Plot of expressed gene indices. (B) Plot of silent gene indices (repressed and X-inactivated genes). (C) Plot of escape gene indices. Anchor points are listed on the x-axis in their order along the X chromosome. Anchor points from genes subject to XCI or escaping XCI are highlighted in light red and light green, respectively.
Lack of evidence for an escape LCR or a common escape motif
It is possible that escape from XCI is facilitated by an LCR upon which escapers converge. Alternatively, specialized escape-specific enhancers, or other genomic features, could be dispersed across the length of the X chromosome and utilized on a regional basis. Finally, an escape-specific enhancer sequence may not exist, and mechanisms of escape from XCI may vary from gene to gene.
To test the hypothesis that escapers converge upon an LCR, we first preformed a correlation analysis of Xi cis-interaction profiles (Figure 3A). Xi anchors clustered according to linear position on the X chromosome, not by expression status. Next, we directly compared the genomic locations of cis and trans interactions of Xi anchors and searched for regions within the B6 and Cast genomes where Xi anchor profiles overlapped (Figure 3B and Figure S3). Escapers only converged with each other when they were located within 1 Mb of each other (Figure 3B). Consistent with our correlation analysis of interaction profiles (Figure 3A), escapers and X-inactivated genes in close linear proximity shared common regions of contact (Figure 3B).
Escapers do not converge upon a common genomic location. (A) Scaled Pearson correlation of cis interactions generated by Xi anchor points. Anchor points are listed along the x- and y-axes according to their position along the X chromosome. Pearson correlations are scaled for linear distances between the anchor points. (B) Bar graph of shared genomic regions of Xi anchor interactions. The combination of anchor points tested is listed on the x-axis with escaper anchors in green and inactivated anchors in red. Brown and blue bars indicate shared genomic regions in the B6 or Cast genomes, respectively. Also see Figure S3.
We next tested if escapers converged on a dispersed sequence or class of sequences. In an attempt to identify a common sequence motif among escape genes, we performed de novo sequence analysis using multiple EM for motif elicitation (MEME) (Bailey et al. 2009) on ±5 kb of all 30 genes known to escape XCI in C/B TS cells (Calabrese et al. 2012). Identified nonrepetitive sequences were then passed to Cis-eLement OVERrepresentation (CLOVER) (Frith et al. 2004) to identify if these motifs were enriched in escaper interaction profiles vs. X-inactivated interaction profiles. We detected no enrichment of any motifs examined (data not shown). Further, CLOVER analysis of the JASPAR database of transcriptional regulators (Bryne et al. 2008) in escaper interaction profiles and X-inactivated interaction profiles did not associate a particular biological pathway with escape from XCI (data not shown).
In comparison to the autosomes, the X chromosome is enriched for LINE elements (Meyer et al. 2013) and these features may play a role in the initiation of XCI (Chow et al. 2010). We hypothesized that if LINEs facilitate XCI, then genes escaping XCI may form fewer contacts with repeat elements. Upon testing this possibility, we found no difference in the presence of all repetitive elements (P = 0.347) or LINE elements only (P = 0.932, Table S3) found within interactions generated by X-inactivated and escaper anchors.
Taken together, our 4C-Seq data suggest that genes escaping imprinted XCI likely do not converge upon a genomic region, or particular sequence, to facilitate escape. Rather, genes within close linear proximity have similar contact profiles (Figure 3 and Figure S3), suggesting a model whereby escape from imprinted XCI may be governed by regulatory elements found within close linear proximity of escape genes.
Evidence for active enhancers in close proximity to escapers
We next tested the possibility that escaper interaction profiles were generally enriched for genomic features associated with enhancer elements (H3K4me1, H3K27Ac, and DNAse) (Ernst et al. 2011; Shen et al. 2012) or formation of chromatin loops (CCCT-binding factor) (Dixon et al. 2012). Our analysis of datasets in C/B TS cells (Calabrese et al. 2012) did not detect any significant difference of these features between escapers and X-inactivated genes (Figure S4). These results suggest that a general strategy of simultaneous interaction with individual epigenetic features does not facilitate escape from XCI.
We next searched gene bodies, plus an additional 10 kb, 15 kb, and 50 kb upstream and downstream (Figure 4A and data not shown), for the presence of putative active enhancer elements to determine if they were correlated with escape from XCI. Putative active enhancers in TS cells were identified as genomic regions enriched for the combination of H3K4me1, H3K27Ac, and DNAse. We then used our RNA-Seq data (Calabrese et al. 2012) to generate a full set of genes subject to XCI (n = 276) and escapers (n = 30).
Putative active enhancers are found in close proximity to escaper genes. (A) Boxplots of the number of putative active enhancer elements per kilobase found within gene bodies ±50 kb. Schematic above the plot indicates generic genomic regions searched. XCI = Xa allele is expressed and the Xi allele is subject to XCI. Escape = Xa allele is expressed and the Xi allele escapes XCI. Blue (Xa) and brown (Xi) bars indicate the chromosome to which the putative active enhancer elements mapped. (B) Map of putative active enhancers located near Taf1 and Ogt. Putative enhancers mapped to the Xa, Xi, or no allelic call are indicated by blue, brown, or black boxes, respectively. Gray boxes indicate Taf1 and Ogt gene bodies. Orange denotes the mouse ES cell TAD for this genomic region. Note, one Xi enhancer falls outside of this TAD. Also see Figure S4.
Our analysis demonstrates that putative active enhancers mapped to the Xi were found at higher frequencies surrounding escapers vs. X-inactivated genes (Figure 4A, brown bars). This difference was not observed on the Xa where all 296 genes are expressed (Figure 4A, blue bars). For any given escape gene, we noted that the Xa allele was associated with more putative enhancers as compared to the Xi allele (Figure 4). Interestingly, if a putative active enhancer was in close proximity to an escaper, the homologous region on the Xa was also identified as a putative active enhancer (Figure 4B and Table S4). This latter observation was not observed in genomic regions surrounding Xist, a gene exclusively expressed on the Xi (Table S4).
It is possible that identified putative enhancers license escape of all genes located within close proximity. If true, we would not expect to find inactivated genes in close proximity to putative enhancers. To that end, we assessed the transcriptional activity of genes found within ±50 kb of putative enhancers identified in close proximity to escapers. We found that, while escaping genes were found at the highest frequencies, inactivated and silenced genes were also found (Table S5). This suggests that proximity to a putative enhancer does not license escape and an additional layer of transcriptional control is likely required.
We tested to see if escapers formed contacts with putative enhancer elements. With the exception of Nkap, all escapers formed at least one cis contact with a putative enhancer element (Table S5). We note that while several putative enhancer elements surround Nkap, there is insufficient SNP data to make a proper call as to their location on the Xa or Xi (data not shown). Therefore, while it is likely that Nkap also contacts a putative enhancer on the Xi, we cannot properly demonstrate it with the available data.
In general, TADs are thought to be conserved across cell types (Dixon et al. 2012). Because TADs for TS cells have not been defined by any study, we used mouse ES TAD boundaries (Dixon et al. 2012) as a proxy to test if the identified putative active enhancers resided in the same TAD as the escape genes we tested. Our analysis revealed that 85% of such enhancers resided in the same TAD as the escape gene associated with them.
Taken together, escaper genomic interactions are not enriched for individual factors associated with active enhancers. However, putative active enhancers mapped to identical genomic coordinates on the Xa and Xi are located in close proximity to escapers and are likely within the same TAD as the escape genes, suggesting that escape from imprinted XCI is facilitated by promoter proximal regulatory elements.
Discussion
We have used allele-specific 4C-Seq to understand mechanisms of escape from XCI. Our data are consistent with a model where escape from imprinted XCI is facilitated by regulatory elements proximal to escaping genes.
The imprinted Xi interacts with the genome
A previous study using neural progenitor cells (NPCs), which undergo random XCI, found that the NPC Xi did not form a predictable structure and was less likely to interact with other chromosomes (Splinter et al. 2011). In that same work, the two escaping genes tested were found to interact with other escaping genes more frequently than X-inactivated genes (Splinter et al. 2011). Our 4C-Seq data in TS cells, which undergo imprinted XCI, suggest that these findings are not universally true, highlighting a potential difference between imprinted and random XCI.
Another possibility is technical differences between the two studies. Primarily, Splinter et al. (2011) used a median of 2.3 × 106 raw reads per anchor with an “N” of 1 to draw their conclusions. The small number of reads forced the authors to “binarize” their data, reducing the mapped dataset to a series of 1s and 0s, depending on the presence of mapped reads. This reduction of the data, while protecting against PCR amplification artifacts, forces the use of large window sizes (100 3C fragments) during analysis and decreases resolution.
Our study utilized biological replicates per anchor, and depending on the replicate, had a median of 6.54 × 106 and 5.87 × 106 mapped/processed reads per anchor. The barcodes included in our primer design allow for the elimination of PCR artifacts during analysis. Further, our substantial increase in mapped reads allowed for the use of the full dynamic range of each replicate, a significantly smaller window size (three 3C fragments) during analysis, and increased resolution. Finally, our ability to compare biological replicates for each anchor allowed for the exclusion of interactions generated due to random collisions in the nucleus. Regardless of the source of the differences between the two studies, our data indicate that escapers may or may not interact with other escapers. Furthermore for TS cells, genes escaping imprinted XCI do so by using local regulatory sequences that are the same as their expressed alleles on the active X chromosome (see below).
Significant cis and trans interactions between anchor points and other genomic coordinates were found for all genes examined, regardless of the location of the anchor (Xa vs. Xi) or the transcriptional status of the gene. While the Xa and Xi likely adopt different structures, Xi alleles preferred to interact with other genomic loci in close linear proximity, an identical behavior noted for loci on the autosomes (Dixon et al. 2012). Together, our data suggest that Xi-linked loci in TS cells physically interact with the genome in a manner similar to the Xa and potentially other chromosomal regions.
Escape from XCI is likely mediated within topological domains by active enhancers located proximal to escapers
The observance of TADs within the genome suggests that the majority of regulatory elements required for the expression of any given gene are likely located in close proximity to the gene (Dixon et al. 2012). Our 4C-Seq data are consistent with this model. Escaper genes do not always contact other escaper genes; an observation that would be predicted if escape LCRs existed. In the same vein, overlap of interactions was independent of transcriptional activity and only occurred when genes were within a few megabases of each other. This latter result is consistent with a recent study showing that Huwe1 and Kdm5c, an X-inactivated and escaper gene pair separated by ∼500 kb, adopt similar positions relative to the Xist cloud in nuclear space (Calabrese et al. 2012).
Consistent with the TAD model, we find that putative active enhancer elements contact escape genes. In addition, they are found within close proximity to, and likely within the same TADS as escaper genes. In contrast, X-inactivated genes are rarely found in close proximity to putative active enhancers. Interestingly, with the exception of Xist, the genomic coordinates of escaper-associated putative active enhancer elements on the Xi are identical to a subset of the putative enhancers on the Xa. The larger number of enhancers found on the Xa may explain why, among escapers, the Xa-linked allele is transcribed at a higher level than the corresponding Xi-linked allele (Calabrese et al. 2012). That inactivated genes are found in close proximity to putative enhancer elements suggests that these sequences regulate individual escaper genes and do not license wholesale escape from XCI of any gene located within a close linear distance.
Taken together, our observations support a model where regulatory control of escape from imprinted XCI is possibly governed within TADs. While it is possible that our identified putative enhancer elements are not causing escape, our evidence strongly supports their role in at least maintaining escape from XCI. Recently, it was shown that in TS cells, XCI appears to be maintained independently of a chromosome-scale nuclear compartment dedicated to transcriptional silencing (Calabrese et al. 2012). Our model for escape is consistent with this conclusion, as it places the regulatory elements necessary for escape in close proximity to escaping genes. Our model also may explain cell-type-specific escape profiles. If transcription from the Xi does not require any additional regulatory elements other than those used on the Xa, then any gene is capable of escape, so long as the appropriate mechanisms are in place to license usage of the necessary regulatory elements.
Acknowledgments
The authors thank members of the Magnuson lab, particularly Andrew Fedoriw and Jesse Raab, for their many helpful comments and suggestions and for their critical reading of this manuscript. We also thank E. deWitt for providing us with code used as a basis for our statistical analysis. This work was funded by National Institutes of Health grants (R01GM101974 to T.M. and F32-CA144389 to J.W.M.).
Footnotes
Communicating editor: J. Schimenti
- Received February 9, 2014.
- Accepted March 13, 2014.
- Copyright © 2014 by the Genetics Society of America
