A comparison of a human DXZ4 3.0-kb tandem repeat monomer against the assembled mouse genome (mm9) with Blast-Like Alignment Tool (BLAT) produced no significant matches on the Ensembl genome browser 35 and a limited number of autosomal and X-linked matches on the UCSC Genome Browser 36 (data not shown). We therefore explored conserved gene order in human and mouse to identify a DXZ4 homolog 37. DXZ4 resides at Xq23 14 and is located between the t-Plastin gene (PLS3) and the Angiotensin II receptor, type-2 gene (AGTR2) (Figure 1a). Comparative analysis of human genes in the vicinity of DXZ4 in the mouse genome revealed several differences in the gene order (Figure 1a), including a break point between the mouse PLS3 and AGTR2 orthologs Pls3 and Agtr2 and between Pls3 and the mouse ortholog of HTR2C. In mouse, the nearest proximal gene to Pls3 is Rab39b (>200 kb proximal), whereas the nearest distal gene is Tbl1x, located 1.8 Mb distal to Pls3. In humans, the respective orthologs of these two genes are located >39 Mb distal to and >20 Mb proximal to PLS3, indicating that Pls3 alone defines the block of synteny for this region with the human X chromosome. In primates, DXZ4 is a homogenous tandem repeat 1730; we therefore performed pair-wise alignments of the genomic DNA sequence upstream of Agtr2 and downstream of Pls3 to look for evidence of tandem repeat DNA. Approximately 150 kb upstream of Agtr2, we identified an inverted repeat (Figure 1b) but no obvious tandem repeats. In contrast, pairwise alignments of the genomic DNA sequence distal to Pls3 identified an extensive tandem repeat spanning approximately 35 kb located 19 kb 3' to Pls3 (Figure 1c). In addition, an extensive minisatellite sequence spanning approximately 30 kb was located a further 24 kb downstream of the tandem repeat. The minisatellite was composed primarily of a novel gamma satellite sequence that is interrupted by several L1 and SYNREP repetitive elements, and the sequence itself displayed an inversion almost midway through the locus (Additional file 1a). The repeat showed significant sequence matches only to the mouse X chromosome, and no homologous repeat exists on the human or rat X chromosomes (data not shown). We therefore focused primarily on the tandem repeat.

We next checked to see how frequently a tandem repeat of comparable size (35 kb) occurred on the mouse X chromosome to see if detection of such a sequence downstream of Pls3 would likely occur by chance. Pair-wise alignments along the length of the mouse X chromosome indicated that large tandem repeats are not common (Additional file 2), supporting the possibility that this might be the mouse homolog of DXZ4.

Pair-wise alignment of the tandem repeat sequence revealed that, unlike DXZ4 in primates, where repeat units are very similar in size within a species 1730, the individual repeating units of the mouse tandem repeat varied from 3.8 to 5.7 kb (Figure 1d). Closer examination showed that the size variation was accounted for by the presence of an internal variable number tandem repeat (VNTR) of an approximately 900-bp sequence present as between one and three copies per monomer (Figure 1e). As in primate DXZ4 141730, less than 6% of the smallest monomer DNA sequence (3.8 kb) was repeat masked, and all of the masked regions corresponded to simple repeats. Examination of the largest monomer (5.7 kb) revealed that the first 147 bp of the internal VNTR was derived from an ERV class II long terminal repeat and that the other edge of the VNTR is defined by a simple repeat. The location of these repeat sequences may contribute to the observed copy-number variation. Three other defining features of human DXZ4 were examined for the novel mouse tandem repeat: CpG content, sequence variation between monomers, and size of the tandem array. Human DXZ4 DNA is 62.2% GC, contains 186 CpG dinucleotides per monomer 38, and shows less than 1% sequence divergence between adjacent monomers 17. In contrast, the mouse 3.8-kb monomer is 53.4% GC, contains 36 CpG dinucleotides, and shows greater than 5% sequence divergence from other monomers in the tandem array. In primates, DXZ4 is composed of as many as 100 repeat units spanning hundreds of kilobases on the X chromosome 1417. In the current build of the mouse genome, the tandem repeat is composed of approximately seven repeat units. Given the inherent difficulty with the computer-based assembly of tandem repeats 39, the actual array could be more extensive. We have previously used extended DNA fiber fluorescence in situ hybridization (FISH) to confirm tandem arrangement and copy-number variation of human DXZ4 17. We applied the same procedure to examine such variation in the mouse tandem repeat, revealing approximately six tandem repeats in two independent mouse cell lines (Figure 1f). This result suggested that the mouse tandem repeat is relatively small, and the presence of the tandem repeat and extensive flanking DNA sequences entirely within the inserts of at least ten independent mouse bacterial artificial chromosomes from three different libraries derived from two Mus musculus subspecies lends additional support (Additional file 3). The logical interpretation of these observations was that the mouse sequence downstream of Pls3 is a tandem repeat but that the overall copy number of repeat units is low, resulting in a smaller array than in primates. Despite these differences from primate DXZ4, the tandem repeat remains a good candidate for the mouse homolog and from this point forward is referred to as Dxz4.

Primate DXZ4 is expressed, and all regions of a monomer can be detected in complementary DNA (cDNA) 172230. Six regions of Dxz4 were assessed in cDNA from several different mouse total-RNA sources. The example shown in Figure 2a indicates that mouse Dxz4 was also expressed and that all parts of the Dxz4 monomer are transcribed into RNA. This result was confirmed by RNA FISH showing readily detectable Dxz4 primary transcript by means of direct-labeled Dxz4 probes (Figure 2b). In humans, DXZ4 is primarily transcribed from one strand (since designated the sense strand), but antisense transcript can be detected in females and is therefore interpreted as originating from the Xi 22. Our previous data showed that only sense transcript could be detected in macaque 30. To assess the relative frequencies of sense and antisense transcription of Dxz4, we primed male and female cDNA from total RNA using oligonucleotides that would prime sense or antisense cDNA synthesis. As in macaque, only sense transcript was readily detected (Figure 2c). In humans, DXZ4 transcript can be detected from the Xa and the Xi 1722, although expression in macaque is almost exclusively restricted to the Xa 30. RNA FISH was performed on female mouse cells with a direct-labeled Dxz4 probe and a probe to the X inactive specific transcript (Xist) 4041 to define the location of the Xi (Figure 2d). As in macaque, Dxz4 could only be readily detected from the Xa (Figure 2e). Collectively, our interpretation of these data is that expression of Dxz4 is restricted to the Xa allele and from one strand only.

Examination of the GenBank mouse mRNA annotation for the Dxz4 locus on the UCSC Genome Browser 36 revealed the presence of two alternatively spliced transcripts spanning Dxz4. Both transcripts originate at an exon almost 2.2 kb from the distal edge of the array (Figure 3a). The transcript then spliced to the same 163-nucleotide sequence within each of the monomers before splicing to an exon located 1.1 kb proximal to the array. One of the two spliced transcripts proceeded to be spliced to two additional exons approximately 16.0 kb downstream, whereas the other read through the splice site before terminating after a further 2.0 kb. To confirm the existence of the spliced forms of Dxz4, we performed reverse-transcription PCR (RT-PCR) between different combinations of the exons. The anticipated product was detected for each of the RT-PCR experiments (Figure 3b). Furthermore, the RT-PCR confirmed that the transcript contains multiple copies of the 163-nucleotide exon as can be seen from the laddered effect of progressively larger PCR products (see the PCR of exon 1 to 2 or 2 to 3 as examples). Furthermore, this exon was also alternatively spliced with some transcripts omitting one or more 163-nucleotide exons. This result could be observed as smaller laddered bands when RT-PCR was performed across the entire array (Figure 3b, exon 1 to 9/10).

Both the spliced and unspliced transcripts corresponded to the sense transcript, and therefore probably originated from a common promoter, unlike human DXZ4, which contains a region with promoter activity within each monomer 22. Examination of histone modification profiles from the Encyclopedia of DNA Elements (ENCODE) 42 revealed a distinct peak of histone H3 trimethylated at lysine 4 (H3K4me3) 43 in the vicinity of exon 1 (data not shown). H3K4me3 is a modification associated with transcriptional start sites 44. We therefore cloned the DNA sequence 5' of Dxz4 exon 1 immediately upstream of a promoterless luciferase reporter gene. Two constructs were generated. The first consisted of a 1.2-kb sequence that contained several repetitive elements that are located immediately upstream of exon 1 (Figure 3c). The second construct consisted of a 238-bp unique sequence 5' of exon 1. Robust promoter activity was detected for both constructs (Figure 3d); the highest activity consistently originated from the smaller unique sequence construct, confirming the location of the minimal Dxz4 promoter.

We next checked to see if the Dxz4 tandem repeat possessed intrinsic promoter activity like human DXZ4 22. Two overlapping fragments encompassing a complete Dxz4 monomer were PCR amplified (Additional file 4a), TA cloned and sequence verified. The DNA was then subcloned upstream of the promoterless luciferase reporter gene and assessed for promoter activity alongside the Dxz4 minimal promoter described above. Neither Dxz4 fragment showed obvious activity compared to the Dxz4 minimal promoter that consistently activated luciferase greater than 200-fold over the empty vector (Additional file 4b). Therefore, our interpretation of this result is that both the spliced and unspliced Dxz4 transcripts likely originate from transcription initiating from the minimal promoter. Consequently, it should be possible to detect by RT-PCR a transcript that spans exon 1 directly to the tandem repeat (Additional file 5a). Despite the relatively large size (approximately 2.5 kb) and proximity to the very 5' end of the message, this transcript can be detected in cDNA (Additional file 5b).

When the H3K4me3 profile of Dxz4 was examined, an additional major peak was noticed immediately distal to the downstream inverted tandem repeat (Ds-TR; data not shown), suggesting promoter activity within this region and the possibility that, like Dxz4, the Ds-TR is expressed. RT-PCR confirmed expression of Ds-TR in both male and female samples (Additional file 1b).

DXZ4 is unusual in that CpG dinucleotides are hypermethylated on the Xa but hypomethylated on the Xi 1422, a trait that is conserved in macaque 30. We selected several regions in and around Dxz4 at which to determine and compare the CpG methylation profiles of males and females (Figure 4a). These sites included the Dxz4 promoter, a region of relatively high CpG incidence within the Dxz4 internal VNTR, a CGI immediately downstream of the Dxz4 array (DD-CGI), and two regions in the vicinity of the H3K4me3 peak adjacent to the Ds-TR.

The Dxz4 promoter (Figure 4b, far right) showed a significantly higher percentage of CpG methylation in females than in males (P = 0.0052, two-sample t-test). This result is consistent with our expression analysis (Figure 2), suggesting that transcription of Dxz4 is subject to XCI 2021 and explaining why Dxz4 transcript was only detected from the Xa (Figure 2e).

Males and females did not differ significantly in methylation of the sequence closest to the Ds-TR (profile on far left in Figure 4b; P = 0.7580) or in the region immediately distal to it (P = 0.0577), but the two regions differed vastly; the proximal sequence was almost entirely methylated, and the distal sequence hypomethylated, on both X chromosomes. Both sites overlap a broad signal of H3K4me3 (data not shown), but examination of other ENCODE features 42 at these two regions revealed that the hypomethylated sequence overlapped a major peak of occupancy for Ctcf 45 and a DNaseI hypersensitive site 46, whereas the hypermethylated site did not (Additional file 6). Binding of Ctcf to target sites containing CpG is sensitive to methylation 4748. The hypomethylation in males and females suggests that Ctcf has the potential to bind this region on both the Xa and the Xi.

Males and females did differ significantly in CpG methylation at the Dxz4 array (P = 0.0027) similar to what we and others have reported for primate DXZ4 142230. However, many sites of CpG residues predicted on the basis of the reference genome sequence (mm9) are not conserved, as demonstrated by the numerous gaps in the bisulfite profiles. Methylated cytosine in CpG is prone to mutation by deamination, whereas mutation rates of unmethylated CpG are lower 49. As a consequence, hypomethylated CGIs are evolutionarily conserved 50. The apparent lack of conservation of CpG dinucleotides at Dxz4 is consistent with the overall hypermethylated profiles (Figure 4b). This situation differs from that of primate DXZ4, where CpG residues are highly conserved 2230, consistent with evolutionary maintenance of DXZ4 as an extensive CGI 50.

Furthermore, males and females did differ significantly in methylation at DD-CGI (P = <0.0001); more hypomethylated clones were obtained from the female samples (Figure 4b). Our interpretation of these data is that DD-CGI is hypomethylated on the Xi. DD-CGI spans 333 bp and contains 40 CpGs on the basis of the C57BL/6J reference genome sequence (mm9). None of the genomic feature annotations generated by ENCODE 42, including Ctcf, highlight DD-CGI, and therefore the significance of Xi hypomethylation remains unclear.

Next we sought to complement the DNA methylation analysis by examining histone methylation and Ctcf binding in and around Dxz4. Several sites were selected, including the Dxz4 promoter, the Dxz4 VNTR region, DD-CGI, Ds-TR, and the Pls3 promoter as a control for mouse genes subject to XCI 34 (Figure 5a).

Consistent with the expression analysis (Figure 2) and CpG methylation (Figure 4b), the Dxz4 promoter was characterized by the euchromatin mark H3K4me2 in male and female cells, whereas the facultative heterochromatin marker histone H3 trimethylated at lysine 27 (H3K27me3) was only a feature of the female samples (Figure 5b). The same profile is obtained for the Pls3 promoter, which is subject to XCI in mouse 34. Given that genes on the Xi are silenced by H3K27me3 5152, these data further support the conclusion that Dxz4 expression is subject to XCI.

In primates, H3K4me2 is a feature of DXZ4 on the Xi 2230, although this modification can be detected on the male X at low levels in some individuals and as a result of cellular transformation 53. In contrast, H3K4me2 was readily detected at Dxz4 in males and females (Figure 5b), another difference between mouse and primate DXZ4. Somewhat surprisingly, given the methylation profile at DD-CGI (Figure 4b), H3K4me2 could also be detected at this site in males and females. One possible explanation is that the DD-CGI is located within the transcriptional unit of one of the spliced Dxz4 transcripts (Figure 3a). Therefore, the detection of the euchromatin mark may reflect variable levels of H3K4me2 in the body of active genes 44.

A defining feature of primate DXZ4 is the association of CTCF with the Xi allele 2230. Ctcf was readily detected at Dxz4 in multiple independent female samples, but Ctcf was also detected, albeit at lower levels, in some but not all males (Figure 5b and data not shown). To investigate further the relationship between Ctcf and Dxz4 on the Xa and Xi, we examined DNA sequence reads from Ctcf chromatin immunoprecipitation (ChIP) combined with next generation sequencing (ChIP-Seq) performed on trophoblast stem cells (TSCs), which are derived from the extraembryonic material and undergo imprinted XCI with preferential inactivation of the paternal X chromosome 54. The TSCs were derived from a cross of a male C57BL/6J (BL6) with a female castaneous (cast) mouse. As a result, the BL6 X chromosome will be the Xi. ChIP-Seq reads were compared to BL6 and cast variant sequences for the Dxz4 interval assessed by ChIP-PCR and, where informative, were designated as originating from the Xa (cast) or Xi (BL6). Of 152 ChIP-Seq reads, almost half were assigned to the Xa and half to the Xi (Figure 5c), consistent with detection of Ctcf at the Xa in some males. One interpretation of these data is that Ctcf binds Dxz4 at the Xa and Xi equally, but not detecting Ctcf at Dxz4 in all males even when it is readily detected in the same samples at a known Ctcf binding site within the H19 imprinted control region 4748 suggests that binding of Ctcf to Dxz4 varies. This result could reflect subtle differences in CpG methylation (compare the two male bisulfite profiles in Figure 4b), strain or cell-type differences. Nevertheless, these observations are consistent with the differences we report above for Dxz4 chromatin organization at the Xa and Xi between mouse and primates. Notably, the association of Ctcf within the VNTR region means that although the array itself is relatively small, the potential Ctcf occupancy is higher than one per repeat monomer.

As mentioned above, the unique sequence (Ds-TR) located immediately distal to the large inverted satellite repeat (Figure 5a; Additional file 1) is characterized by DNaseI hypersensitivity and Ctcf binding (Additional file 6). Ctcf ChIP-PCR confirmed association with this sequence in males and females (Figure 5b), and as anticipated given the CpG hypomethylation (Figure 4b), the region was characterized by H3K4me2. To determine whether Ctcf at Ds-TR is associated with the Xa alone or with Xa and Xi, we used informative BL6 and cast SNPs to assign Ctcf ChIP-Seq reads to their X chromosome of origin. Unlike Dxz4, Ctcf at Ds-TR was biased toward the Xa but could also bind the Xi to a lower extent (Figure 5c).

Thus far we have shown that, as in primates 14172230, a large tandem repeat is present downstream of Pls3 on the mouse X chromosome despite extensive shuffling of the locations of genes from the same interval (Figure 1a). We sought to determine whether a tandem repeat was present downstream of PLS3 in a diverse set of mammals for which genome assemblies were sufficiently complete. Pairwise alignment of genomic sequence distal to PLS3 was performed for seven different mammals. Each revealed the presence of a tandem repeat within 28 to 110 kb of the 3' end of PLS3 (Figure 6).

Previously we have shown that a region encompassing the CTCF binding site is conserved in primates, but outside of this interval divergence of the sequence of DXZ4 and size of the individual tandem repeat unit increases substantially with distance down the primate tree 30. Focusing only on this region, we identified 74% nucleotide identity over 100 bp between human DXZ4 and the VNTR region within each mouse Dxz4 monomer. Similar levels of nucleotide identity over the same interval were identified within the tandem repeat DNAs shown in Figure 6. We used this interval to extract homologous DNA sequence entries from 25 different mammals before aligning all of the sequences. Most of the mammals examined formed clades corresponding to their respective orders and suborders, such as the primates, which all branch from a single node (Figure 7a). These data support evolution of DXZ4 from a common ancestor in a manner analogous to that of coding sequences. Close examination of the DNA sequence alignment revealed a subregion of the conserved DNA sequence in which several nucleotides were identical in all 25 mammals. A 34-bp sequence encompassing all invariable nucleotides was extracted from each sequence and used to generate a position weight matrix 55 that clearly revealed the nonrandom nature of this sequence (Figure 7b). Given that this sequence is entirely contained within the region assessed by PCR in primate CTCF ChIP 2230, mouse Ctcf ChIP (Figure 5b), and mouse Ctcf ChIP-Seq (Figure 5c), the position weight matrix sequence was compared with a previously defined Ctcf consensus sequence 47, and as can be seen in Figure 7b, the most conserved DXZ4 sequence in all mammals examined was a good match to this consensus.

The Ctcf match to the conserved sequence only accounts for bases 3 to 21, yet conservation of DNA sequence across the diverse group of mammals extends for an additional 13 bp. It is conceivable that this extended conservation reflects retention of an additional binding motif(s) for other DNA binding protein(s). To explore this possibility, the consensus sequence was compared to motifs in JASPAR 56. Two motifs showed good matches to this region. The first is a 9 out of 10 base match to the recently determined mouse consensus for the CCAAT/enhancer-binding protein alpha (Cebpa) 57, whereas the second is a match (9 out of 9) for the human consensus for ETS-domain protein 4 (ELK4) 58 (Additional file 7). Cebpa is an essential basic-leucine zipper DNA binding protein that performs essential roles in the development of myeloid cells 59 and in liver function 60. ELK4 is a ubiquitous serum response factor accessory protein 61 that is found at many locations in the genome 62. Whether either protein binds to Dxz4 has yet to be determined, but given the broad cross-species conservation of the DNA sequence and good matches with each DNA binding consensus sequence 5758, both are candidates worthy of further investigation.

Mouse male fibroblast cell line NIH/3T3 (CRL-1658) and female fibroblast cell line Balb/3T3 (CCL-163) were obtained from ATCC. Mouse female fibroblast cell line BC06 (hybrid C57BL/6J X castaneous) was obtained from Laura Carrel. Male and female CD-1 and C57BL/6J mouse embryonic fibroblasts were derived by standard techniques 69. All cells were maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum supplemented with 1× nonessential amino acids, 2 mM L-glutamine, 100 U/ml penicillin, and 0.1 mg/ml streptomycin. All medium components were obtained from Invitrogen (Life Technologies Corp, Grand Island, NY, USA); NIH/3T3 cells were cultured in media containing Hyclone bovine calf serum (Thermo Scientific, Rockford, IL, USA) in place of fetal bovine serum.

Genomic DNA was isolated from primary cells with the NucleoSpin Tissue kit (Machery-Nagel, Bethlehem, PA, USA). Genomic DNA was isolated from mouse tail snips by standard techniques 69. Unmethylated cytosines were converted to uracil with the EpiTect bisulfite modification kit (Qiagen, Valencia, CA, USA). Bisulfite-modified DNA was used as a template for PCR with OneTaq^® master mix (NEB, Ipswich, MA, USA) and the primers listed in Additional file 8. PCR products were cloned into pDrive TA vector (Qiagen), and positive clones sequenced (Eurofins MWG Operon, Huntsville, AL, USA) and analyzed with Sequencher 5.0 (Gene Codes Corp., Ann Arbor, MI, USA). Statistically significant differences in methylation between males and females were determined as follows. The percent methylation for individual clones (a single horizontal line in the profiles) was determined and the mean and standard deviation was calculated for the males and females. These were compared using the two-tailed t-test with differing variance as described previously for methylation profiles 70.

Mouse Dxz4 fragments were PCR amplified and cloned into the TA vector pCR2.1 (Life Technologies Corp.) before sequence verification. Direct-labeled FISH probes were generated from Dxz4-pCR2.1-isolated DNA with SpectrumOrange™ or SpectrumGreen™ and a nick translation kit (Abbott Molecular, Abbott Park, IL, USA). Probes were heat inactivated at 68°C for 10 minutes before ethanol precipitation and resuspension in Hybrisol VII (MP Biomedicals, Santa Ana, CA, USA). RNA FISH was performed on cells grown directly on microscope slides. Cells were rinsed with 1× phosphate-buffered saline (PBS) before being fixed and extracted for 10 minutes at room temperature in 3.7% formaldehyde, 0.1% Triton X-100 in 1× PBS. Slides were rinsed twice in 1× PBS before dehydration for 3 minutes in 70% and 100% ethanol before being air-dried. Probes were denatured in a thermal cycler at 72°C for 10 minutes before the temperature was reduced to 37°C, at which point the probe was applied directly to the slide, sealed under a cover glass, and hybridized overnight at 37°C. Cover slips were removed and the samples washed twice at room temperature for 2 minutes each in 50% formamide/2 SSC, once for 3 minutes at 37°C in 50% formamide/2× SSC, and once for 3 minutes at 37°C in 2× SSC before addition of ProLong^® Gold antifade reagent supplemented with DAPI (Life Technologies Corp.). Mouse extended DNA fibers were prepared and FISH performed essentially as previously described 17. Images were either collected with a Zeiss Axiovert 200 M fitted with an AxioCam MRm and managed with AxioVision 4.4 software (Carl Zeiss microimaging) or collected with a DeltaVision pDV. Delta Vision images were deconvolved with softWoRx 3.7.0 (Applied Precision, Issaquah, WA, USA) and compiled with Adobe Photoshop CS2 (Adobe Systems).

Total RNA was extracted from cells with the NucleoSpin RNA II kit (Machery-Nagel). For standard RT-PCR, first-strand cDNA was prepared from 2 μg of total RNA with random hexamers with and without M-MuLV reverse transcriptase (RT) according to the manufacturer's instructions (NEB). cDNAs prepared with and without RT were used as templates for PCR with either OneTaq^® master mix (NEB) or HotStar Taq (Qiagen) with the primers listed in Additional file 8. PCR was performed using an initial denaturation of 10 minutes at 94°C, followed by 35 cycles of: 94°C for 30 seconds, 58°C for 30 seconds and 72°C for 30 seconds for all products of up to 750 bp, 1 minute for all products up to 1,250 bp and 1 minute 30 seconds for products up to 2 kb. The cycling was followed by 10 minutes at 72°C before holding at 15°C. Strand-specific cDNA was prepared as above except that first-strand cDNA was primed with 1.5 pmol of a specified oligonucleotide (Additional file 8) in place of random hexamers and an additional control that included RT but no oligonucleotide that is used to determine the background levels of cDNA synthesized in the absence of a gene-specific primer. Strand-specific cDNA was assessed by quantitative RT-PCR using the primers given (Additional file 8) with a SYBR-Green qPCR Mastermix (SABiosciences, Qiagen) on a CFX96 (Biorad, Hercules, CA, USA). PCR was performed using an initial 10-minute denaturing step at 95°C followed by 40 cycles of: 15 seconds at 95°C, 30 seconds at 60°C and 30 seconds at 72°C. The cycle was followed by a melt-curve. PCR was performed in triplicate and the transcript level determined relative to background.

DNA fragments initiating in and extending upstream of Dxz4 exon 1 were generated by PCR with Platinum^®Taq (Life Technologies Corp.; 94°C for 2 minutes followed by 40 cycles of: 94°C for 30 seconds, 58°C for 30 seconds and 68°C for 1 minute 20 seconds for construct A or 68°C for 30 seconds for construct B) and cloned into pDrive (Qiagen). Inserts were verified by DNA sequencing before subcloning into the KpnI and XhoI sites of pGL4.10[luc2] (Promega, Madison, WI, USA). The Dxz4-promoter pGL4.10[luc2] firefly luciferase reporter constructs were co-transfected in triplicate on two separate occasions with the Renilla-luciferase expression vector pGL4.74[hRluc/TK] (Promega) into NIH/3T3 cells by means of Lipofectamine 2000 (Life Technologies Corp.). Cells were assayed for luciferase activity on a Glomax-20/20 Luminometer (Promega) 72 hours after transfection with the dual-luciferase reporter assay system, according to the manufacturer's recommendations (Promega).

Standard ChIP was performed on mouse cells essentially as described previously 22 except that formaldehyde cross-linking was with 0.75% formaldehyde rather than 1.0%. Chromatin was sheared with a Bioruptor (Diagenode, Denville, NJ, USA) set at 8 cycles of 30 seconds on and 30 seconds off on high setting. Rabbit polyclonal antibodies used were all obtained from Millipore (Billerica, MA, USA) and included anti-H3K4me2 (07-030), anti-H3K27me3 (07-449), and anti-CTCF (07-729). ChIP was assessed by quantitative PCR using the primers given (Additional file 8) with a SYBR-Green qPCR Mastermix (SABiosciences, Qiagen) on a CFX96 (Biorad). PCR was performed using an initial 10-minute denaturing step at 95°C followed by 40 cycles of: 15 seconds at 95°C, 30 seconds at 60°C and 30 seconds at 72°C. The cycle was followed by a melt-curve. Standard curves were prepared by making a 1:5 serial dilution of the input for each ChIP. ChIP and mock (rabbit serum) samples were assessed in triplicate and the percentage of quantitative PCR product normalized and determined from the standard curve using Bio-Rad CFX Manager 2.1 software (Biorad). Each ChIP experiment and all PCR assessments were replicated on at least three independent occasions. Anti-Ctcf ChIP on mouse TSCs derived from a C57BL/6J × CAST/EiJ cross was combined with next-generation sequencing (100-bp paired-end reads) as described in detail elsewhere (Calabrese JM and Magnuson T, in preparation). Briefly, ChIP was performed on 10 to 40 × 10⁶ feeder-free TSCs. Cells were crosslinked for 10 minutes at room temperature in 0.6% formaldehyde before quenching in 125mM glycine for 5 minutes. Cells were resuspended in 50 mM Tris-HCl pH 7.5, 140 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.1% Na-deoxycholate and 0.1% SDS. Cells were sonicated to generate fragments averaging 200 to 500 bp, cleared by centrifugation and resuspended at 20 × 10⁶ cells/ml in the buffer above supplemented with 1% Triton-X100. ChIP was performed with 10 μg of antibody. Post-ChIP, three washes with the buffer used for the ChIP were performed, followed by a wash in the same buffer but with 500 mM NaCl, once with 20 mM Tris pH 8.0, 1 mM EDTA, 250 mM LiCl, 0.5% Na-deoxycholate and once with TE buffer. Chromatin was eluted for 15 minutes at 65°C in 50 mM Tris pH 8.0, 10 mM EDTA and 1% SDS. A ChIP-Seq library was prepared according to Illumina instructions using 10 to 200 ng of ChIP DNA and sequenced on Illumina's Genome Analyzer IIx or HiSeq2000 instrument. Ctcf ChIP-Seq data have been deposited with Gene Expression Omnibus and assigned the provisional accession number GSE40667. The DNA sequence of the mouse Dxz4 array was used to extract ChIP-Seq hits with homology to Dxz4. An approximately 232-bp DNA fragment spanning the putative mouse Dxz4 Ctcf binding site was amplified from C57BL/6J and castaneous genomic DNA isolated from tail snips. PCR was performed using HotStar Taq (Qiagen) with an initial denaturation of 10 minutes at 94°C, followed by 35 cycles of: 94°C for 30 seconds, 58°C for 30 seconds and 72°C for 30 seconds. The PCR product was cloned into pDrive, and for each DNA source over 100 clones were isolated and sequenced. Sequence variants specific to C57BL/6J and castaneous were then used to manually align with 100% sequence identity over a minimum of 30 bp to the Ctcf ChIP-Seq Dxz4 sequences and designated either C57BL/6J or castaneous. All SNP variants have been deposited with dbSNP. Details can be found in Additional file 9.