Insulation of the Ubiquitous Rxrb Promoter from the Cartilage-specific Adjacent Gene, Col11a2*

The retinoid X receptor β gene (Rxrb) is located just upstream of the α2(XI) collagen chain gene (Col11a2) in a head-to-tail manner. However, the domain structures of these genes are unknown. Col11a2 is specifically expressed in cartilage. In the present study, we found Rxrb expression in various tissues with low expression in the cartilage. Col11a2 1st intron enhancer directed cartilage specific expression when linked to the heterologous promoter in transgenic mice. These results suggest the presence of enhancer-blocking elements that insulate Rxrb promoter from the Col11a2 enhancer. So far, most of insulators examined in vertebrates contain a binding site for CTCF. We found two possible CTCF-binding sites: one (11P) in the intergenic region between Rxrb and Col11a2 by electrophoretic mobility shift assays, and the other in the 4th intron of RXRB by data base search. To examine the function of these elements, we prepared bacterial artificial chromosome (BAC) transgene constructs containing a 142-kb genomic DNA insert with RXRB and COL11A2 sequences in the middle. Mutation of 11P significantly decreased the RXRB promoter activity in muscular cells and significantly increased expression levels of RXRB in chondrosarcoma cells. In transgenic mouse assays, the wild-type BAC transgene partly recapitulated endogenous Rxrb expression patterns. A 507-bp deletion mutation including 11P enhanced the cartilage-specific activity of the RXRB promoter in BAC transgenic mice. Chromatin immunoprecipitation analysis showed that CTCF was associated with RX4, but not with 11P. Our results showed that the intergenic sequence including 11P insulates Rxrb promoter from Col11a2 enhancer, possibly associating with unknown factors that recognize a motif similar to CTCF.

that survived were sterile due to oligoasthenoteratozoospermia (11). RXR␤ and RXR␥ are expressed in the striatum and control the function of the dopaminergic mesolimbic pathway (12). RXR forms heterodimers with liver X receptor and farnesoid X-activated receptor and regulates cholesterol balance (13). Thus, the fidelity of the spatiotemporal expression of RXR␤ is crucial for the normal development of a wide variety of tissues.
The organization of eukaryotic genomes necessarily results in the proximity of domains with distinct functions. The identity of domains is maintained by classical transcriptional regulatory elements, such as enhancers, silencers, and upstream activating sequences. In some cases, specific DNA sequences, referred to as insulators (14), and their associated binding proteins have a role in establishing or maintaining discrete interdomain boundaries. Insulators have been divided into two classes: enhancer-blocking insulators, which prevent distal enhancers from activating a promoter when placed between an enhancer and promoter, and barrier insulators, which block heterochromatinization and the consequent silencing of a gene (15). In vertebrates, the enhancer blocking activity of insulators is associated with a binding site for the CCCTC-binding factor (CTCF), which recognizes long and diverse nucleotide sequences. CTCF is a ubiquitously expressed nuclear protein with 11 zinc finger DNA-binding domains. Thus far, CTCF remains as the only major protein implicated in the establishment of insulators in vertebrates (16). There has been great interest in identifying other binding sites for CTCF. Several recent high-throughput ChIP-chip analyses and comparative genomic studies have identified tens of thousands of potential CTCF-binding sites in the human and mouse genomes (17)(18)(19). The sensitivity of the ChIP-chip analysis for CTCF is 88%; thus, there are some false-negative results (18). Recently, it was discovered that cohesin binds to many of the CTCF-binding sites, co-localizing with CTCF to insulate promoters from distinct enhancers (20,21).
Despite the proximity of Rxrb and Col11a2, the dissimilarities in their expression patterns and functions suggest the existence of an intergenic boundary. To investigate this possibility, we first clarified differences in the expression patterns of the two genes and examined whether the cartilage-specific enhancer of Col11a2 could affect transcriptional activities of heterologous promoters. We then searched for a CTCF-binding site between the two genes. We examined whether the intergenic sequence and CTCF-binding site affected the activities of the Rxrb promoter by using bacterial artificial chromosome (BAC) transgene constructs that cover the entire Rxrb and Col11a2 genes.

EXPERIMENTAL PROCEDURES
Northern Hybridization-Total RNA was extracted from various mouse tissues at 16.5 days postcoitus and from cell lines by using RNeasy Mini Kits (Qiagen, Santa Clara, CA). Ten micrograms of total RNA was fractionated by electrophoresis through formaldehyde-agarose gels and then transferred onto Nytran membranes (Amersham Biosciences). Complementary DNAs (cDNAs) were labeled with [ 32 P]dCTP by using the Rediprime II Random Prime Labeling System (Amersham Bio-sciences). The membranes were hybridized with 32 P-labeled mouse Col11a2 and Rxrb cDNAs.
Real-time RT-PCR-Total RNAs were digested with DNase to eliminate any contaminating genomic DNA before real-time quantitative RT-PCR. One microgram of total RNA was reverse transcribed into first-strand cDNA by using QuantiTect Reverse Transcription (Qiagen). The PCR amplification proceeded in a 20-l reaction mixture containing 2 l of cDNA, 10 l of SYBR PremixExTaq (Takara, Japan), and 4 pmol of primers specific for rat Rxrb and Col11a2. The quantified individual RNA expression levels of rat Col11a2 and rat Rxrb were normalized to the respective rat Gapdh expression levels. The primers used are listed in supplemental Table S1.
Generation of Transgenic Mice Bearing the Col11a2 Enhancer Linked to Heterologous Promoters-pAD-LacZ was the expression vector pADbeta (Clontech number 6176-1), which contains the adenovirus II major promoter, SV40 RNA splice site, the ␤-galactosidase reporter gene and the SV40 polyadenylation signal. The fragment of 2.3 kb of the first intron sequence of Col11a2 as an enhancer was cloned into the SalI/ PstI polylinker site located downstream of the SV-40 polyadenylation signal of ADbeta to create pAD-LacZ-Int. Nf-LacZ-Int was created by inserting the 2.3-kb fragment of the first intron sequence of Col11a2 as an enhancer into the SalI/PstI polylinker site located downstream of the SV-40 polyadenylation signal of pNf-LacZ (27), which contains the neurofilament promoter, SV40 RNA splice site, the ␤-galactosidase reporter gene, and the SV40 polyadenylation signal.
The plasmids AD-LacZ, AD-LacZ-Int, and Nf-LacZ-Int were digested with EcoRI and PstI to release the inserts from their vector sequences. Transgenic mice were produced by microinjecting each of the inserts into the pronuclei of fertilized eggs from F1 hybrid mice (C57BL/6x DBA) as described previously (8). Generation 0 (G0) embryos were sacrificed at 13.5 days postcoitus and processed for expression analysis of the reporter gene. Transgenic embryos were identified by PCR analysis of genomic DNA extracted from the placenta or tail as described previously (8). X-Gal staining of mouse bodies and sections was performed as previously described (28).
Electrophoretic Mobility Shift Assays-The full-length coding region of mouse CTCF cDNA was PCR amplified with primers NT364 and NT365 (supplemental Table S2) and cloned into pCR-BluntII-TOPO using the Zero Blunt TOPO PCR cloning kit (Invitrogen catalog number K2800-20). CTCF cDNA was transferred to pTNT vector and used for in vitro synthesis of CTCF protein with the TNT T7 Quick Coupled Transcription/Translation System (Promega, Madison, WI, catalog number L1170). Nuclear extracts from RCS cells were prepared using CelLytic NuCLEAR Extraction kits (Sigma) according to the manufacturer's instructions. Probes were prepared as described under supplementary data.
DNA fragments were denatured, end-labeled at the 5Ј-end with [␥-32 P]ATP and T4 polynucleotide kinase, and annealed to prepare radiolabeled probes. The radiolabeled probes were gel purified and combined with equal amounts of in vitro-synthesized CTCF protein or equal amounts of nuclear extracts. For binding reactions, we used a buffer containing phosphate-buffered saline with 5 mM MgCl 2 , 0.1 mM ZnSO4, 1 mM dithiothreitol, 0.1% Nonidet P-40, and 10% glycerol in the presence of 10 ng/l poly(dI-dC) and unlabeled 10 nM double-strand synthesized oligonucleotides containing an Sp1-binding site, as SP1like factors can bind to GC-rich segments and obscure the binding of CTCF (29). The reaction mixtures of a 10-l final volume were incubated for 20 min at room temperature followed by electrophoresis on 6% non-denaturing polyacrylamide gels. We performed electrophoretic mobility shift assay (EMSA) with nuclear extracts in the presence of 0.2 ng/l salmon sperm DNA. For super-shift assays, the reaction mixture was combined with 2 l of anti-CTCF antibody (Upstate number 07-729) or normal rabbit IgG and incubated for 2 h at 4°C before the addition of radiolabeled probes. The primers and synthesized oligonucleotides are listed in supplemental Table  S2.
ChiP Assay-We used the ChIP Assay Kit (Upstate number 17-295), according to the manufacturer's protocol as described under supplementary data.
We performed quantitative real-time PCR in triplicate with DNA prepared with anti-acetylated histone H3 antibody and input DNA. We used the LightCycler FastStart DNA Master plus HybProbe kit (Roche number 03-515-575001) in a LightCycler (Roche). Amplification reactions were performed in a volume of 20 l with 20 pM (each) primers and FRET probes (0.4 l of each primer per reaction), 5 l of template DNA, and 1ϫ Master mixture (4 l/reaction).
We performed quantitative real-time PCR in triplicate with DNA prepared with anti-CTCF antibody, DNA prepared with anti-Rad21 antibody, DNA prepared with anti-BORIS antibody, and input DNA by using primer pairs designed for 11P and RX4. We used LightCycler and SYBR Green reagent (Takara), according to the manufacturer's instructions.
Data were analyzed with LightCycler software by the Fit Point method to minimize noise and obtain the best possible correlation coefficient between standards. The -fold difference for a particular target sequence was determined by calculating the ratio of the amount of the target sequence in the immuno-precipitation to the amount of the target sequence in the input DNA. Primers and FRET probes were selected from human and rat retinoid X receptor and ␣2(XI) collagen chain gene sequences using Roche Primer design software. Primers and FRET probes were obtained from Sigma and Nihon Gene Research Laboratories Inc. (Sendai, Japan), respectively. The primers and FRET probes are listed in supplemental Table S3.
BAC Transgene Construction-A human BAC clone, CTD-2054I15 (WT), which contains RXRB and COL11A2 in the middle, was purchased from Invitrogen. Modifications of a BAC clone were performed as described under supplementary data. Constructs were checked with digestion with various restriction enzymes and PCR analysis. The primers used in BAC modifications are listed in supplemental Table S4.
Establishment of Transgenic RCS Cells-The BAC transgene constructs (SV40-Pur/WT, SV40-Pur/11Psub, and SV40-Pur/ 11Pdel) were transfected into 1 ϫ 10 6 RCS cells in 10-cm dishes using Lipofectamine 2000 (Invitrogen) (supplemental Fig. S2). After 24 h, cells were selected by puromycin (3 g/ml). After 14 days of selection, colonies were picked-up and replated in 24-well dishes. We established 12, 14, and 18 independent RCS cell lines for SV40-Pur/WT, SV40-Pur/11Psub, and SV40-Pur/ 11Pdel transgenes, respectively. Total RNAs were extracted from each transgenic cell and subjected to real-time RT-PCR to analyze the relative expression levels of rat Col11a2, rat Rxrb, human COL11A2, and human RXRB mRNAs. We normalized the quantified individual RNA expression levels of rat Col11a2 and rat Rxrb to the respective rat Gapdh expression levels, and those of human COL11A2 and human RXRB were normalized to the respective puromycin expression levels. The primers used are listed in supplemental Table S1.
Production and Genotyping of Transgenic Mouse Lines-BAC transgene constructs (WT-LacZ, 11Psub-LacZ, 11Pdel-LacZ, and RX4sub-LacZ) were purified with the Qiagen large construct kit (Qiagen, number 12462), following the manufacturer's instructions. The BAC DNA constructs were linearized by digestion with PacI. The linearized transgenes (2-5 ng/l) in injection buffer (10 mM Tris-HCl, pH 7.5, and 0.1 mM EDTA, pH 8.0) were microinjected into fertilized ova (supplemental Fig. S2). Founder embryos at E13.5 were sacrificed and genotyped with three sets of PCR primers (JM281/JM282, JM283/ JM284, and NT418/419) located at both ends and the middle of the linearized BAC transgene DNA. The primers are listed in supplemental Table S4.

Different Expression Patterns between the Rxrb and the
Col11a2 Genes-To delineate differences in the transcriptional activities of Rxrb and Col11a2 in various tissues, we analyzed the expression of mRNAs of these genes in samples from identical tissues or cells. Northern hybridization analysis on tissue

Insulation of Rxrb from Cartilage-specific Col11a2
OCTOBER 10, 2008 • VOLUME 283 • NUMBER 41 samples showed that Col11a2 mRNAs were abundant in limb buds but not in the intestine, brain, liver, or lung of mouse embryos at 16.5 days postcoitus (Fig. 1A). The Rxrb mRNAs were clearly detected in all of these tissues. Northern hybridization analysis on cultured cells showed that Col11a2 mRNA was abundant in RCS but not in L6 (rat skeletal muscle), FR (rat fibroblast), PC12 (rat adrenal pheochromocytoma), and H4IIE (rat hepatoma) cells. The Rxrb mRNAs were clearly detected in all these cells except for the RCS cells. Real-time RT-PCR analysis on cultured cells revealed that relative mRNA expression levels of Col11a2 to Gapdh were much higher in RCS as compared with the levels in L6, FR, PC12, and H4IIE cells (Fig. 1B).
The relative mRNA expression levels of Rxrb were low in RCS cells as compared with the levels in other cells. In situ hybridization performed on semi-serial sections showed that Col11a2 mRNA was highly and specifically expressed in cartilage (Fig. 1C). Rxrb mRNA was detected in various tissues, with abundant expression in muscle. Signals for Rxrb mRNA were present in cartilage; these signals were weak compared with the signals in other tissues. In summary, we confirmed the previous reports that the Col11a2 gene was transcribed specifically in cartilage. We found that Rxrb was expressed in a variety of tissues; interestingly, Rxrb was weakly expressed in cartilage.
Lysine residues on the N-terminal tails of histones H3 and H4 are more highly acetylated in the neighborhood of transcriptionally active promoters and enhancers than those in regions of transcriptional inactivity (30). We examined the acetylation levels and patterns of histone H3 from the Rxrb to Col11a2 genes using RCS and L6 cells ( Fig. 2A). A ChIP assay with anti-acetylated histone H3 antibody showed that the promoter region of Rxrb and the promoter and first intron regions of Col11a2 were hyperacetylated as compared with other regions in RCS cells. Acetylation levels of histone H3 were low except for the promoter region of Rxrb in L6 cells; these results are consistent with the expression levels of Rxrb and Col11a2 in L6 cells.
Effects of the First Intron Sequence of Col11a2 on Transcriptional Activities of Heterologous Promoters-Some enhancers are specific for certain promoters (31,32). We examined whether the 1st intron sequence of Col11a2 serves as an enhancer for heterologous promoters by generating transgenic mice bearing LacZ reporter gene constructs. The adenovirus late II major promoter (AD) directed LacZ expression without tissue specificity. We obtained 16 founder mice bearing an AD-LacZ transgene. Eight of the mice showed X-gal staining, and the patterns of X-gal staining varied among the founder mice (Fig. 2B), suggesting the influence of integration sites of the transgene in the genome. Addition of the 1st intron sequence of Col11a2 to the AD promoter resulted in the production of transgenic mice with cartilage-specific expression of LacZ (Fig. 2C). We showed no X-gal staining. Neurofilament gene promoter (Nf) directs neural tissue-specific expression of the LacZ reporter gene in transgenic mice (33). The addition of the 1st intron sequence of Col11a2 to the neurofilament promoter resulted in the production of transgenic mice with cartilage-specific expression of LacZ (Fig. 2D). We obtained 7 founder mice bearing the Nf-LacZ-Int transgene. Three of these mice showed cartilage-specific LacZ expression, 1 showed LacZ expression in various tissues including cartilage and brain, and three showed no X-gal staining. Although many of mice bearing transgenes with the Col11a2 1st intron sequence directed expression primarily in cartilage, some mice showed expression in non-cartilaginous tissues. Because expression of the transgene is affected by the site where the transgene is integrated in the genome, it is possible to speculate that the sequences around the integrated sites might have enhancer activities in non-cartilaginous tissues in the mice showing LacZ expression in non-cartilaginous tissues. Overall, the results suggest that the 1st intron of Col11a2 could serve on heterologous promoters as an enhancer in cartilage, because AD-LacZ and Nf-LacZ transgenic mice do not show the cartilage-specific expression pattern of LacZ. These results raise the possibility of the existence of a mechanism that prevents the enhancer of the 1st intron sequence of Col11a2 from affecting the Rxrb promoter in the genome.

CTCF-binding Sites in the Intergenic Region between Rxrb and
Col11a2-We performed EMSAs to identify CTCF binding sites in the intergenic region between the rat Col11a2 and Rxrb genes. We first prepared 12 radiolabeled, overlapping probes that cover the stretch from the termination codon of Rxrb to the translational start site (ATG codon) of Col11a2 (Fig. 3A). We incubated the probes with in vitro translated CTCF protein. We detected the shifted bands reproducibly only in samples containing probe 8 and DNA fragments with the CTCF-binding  (n ϭ 3). B-D, transgenic mice bearing the LacZ reporter gene constructs driven by the Col11a2 enhancer linked to a heterologous promoter. B, four of eight AD-LacZ transgenic founder embryos with LacZ expression are shown. C, we obtained six AD-LacZ-Int transgenic founder embryos with LacZ expression. Four embryos showed cartilage-specific LacZ expression (4, 8, 19, and 22), and 2 embryos showed LacZ expression in other tissues (9 and 27). D, we obtained 4 Nf-LacZ-Int transgenic founder embryos with LacZ expression. Three embryos showed cartilage-specific LacZ expression (2, 17, and 19), and 1 embryo showed LacZ expression in various tissues including cartilage (12). Scale bars, 2 mm.

Insulation of Rxrb from Cartilage-specific Col11a2
site DMD4 (Fig. 3B). We concluded that among the 12 probes, probe 8, which corresponds to Ϫ748 to Ϫ422 of the Col11a2 translational start site, interacted with CTCF, although we could not deny the possibility that other probes interact with CTCF under different conditions. Next, we prepared four overlapping probes (A-D) covering this stretch (Fig. 3A). Among them, only probe B corresponding to Ϫ679 to Ϫ576 interacted with CTCF (Fig. 3C). The probe B-CTCF complex was supershifted by the addition of anti-CTCF antibodies but not by the addition of normal rabbit IgG.
To determine the CTCF target sequence within probe B, we prepared a series of radiolabeled probes (M1 to M8) with substitution mutations (Fig. 3D). Probe M5 did not interact with CTCF. Probe M4 interacted with CTCF to a lesser extent than the other mutated probes (Fig. 3E). Competition experiments with unlabeled mutated probes showed that an excess of unlabeled probes M4 and M5 failed to block CTCF binding to probe B, whereas the other probes abolished this binding interaction (Fig. 3F). These results suggest that the sequences (Ϫ641 to Ϫ630 and Ϫ629 to Ϫ616) used to create the M4 and M5 substitution mutations are critical for CTCF binding in the rat sequence. We designated this element as "11P." Nuclear extract from RCS cells interacted with probe B, but not with the M5Bgl probe (Fig. 3G). M5Bgl was identical to M5 except that the mutated sequence in M5Bgl contained the BglII cleavage site. Formation of the complex of probe B with the nuclear extract was disrupted by the addition of anti-CTCF antibodies but not the addition of normal rabbit IgG, suggesting that the complexes contained CTCF.
We then examined whether CTCF binds to the human sequence. Alignment of the human and rat sequences revealed that the rat CTCF-binding site, 11P, corresponds to Ϫ682 to Ϫ656 of the human COL11A2 translational start site. Twenty of 26 nucleotides were identical in the human and rat sequences. Fourteen nucleotides at the 3Ј-end showed 86% similarity between the human and rat sequences. We created human probes h-B and h-M5Bgl that correspond to rat probes B and M5Bgl, respectively (Fig. 3D). Nuclear extract prepared from RCS cells interacted with h-B, but not with the h-M5Bgl probe (Fig. 3G). The formation of the complex of h-B with the nuclear extract was disrupted by addition of anti-CTCF antibodies but not by addition of normal rabbit IgG, suggesting that the complexes contained CTCF. These results suggest that human and rat 11P sequences interact with CTCF.
CTCF Binding Sites within the Rxrb Gene-By accessing a CTCF-binding site data base (17), we found CTCF binding sites upstream of RXRB, in the 4th intron of Rxrb, and downstream of COL11A2 in the human sequence (Fig. 3H). The 11P site we found as a CTCF-binding site by EMSA was not classified as a CTCF-binding site in the data base. Because insulators interfere with enhancer-promoter interactions when placed between the two, we considered the CTCF-binding site in the 4th intron of RXRB as another candidate for the insulator between the RXRB promoter and the COL11A2 enhancer. In the RXRB 4th intron sequence, we found that a 20-mer element (nucleotides ϩ2810 to ϩ2829 from the translational start site of RXRB) had a similar sequence to the consensus binding sequence for CTCF. We designated this element "RX4" (Fig. 3H). We performed EMSAs to determine whether the RX4 sequence interacts with CTCF. We prepared a 276-mer probe (probe 13) with a 20-mer RX4 sequence in the middle and a mutated probe (probe 13M) that contained a 14-mer substitution mutation at the RX4 sequence. We incubated the probes with in vitro translated CTCF protein. Probe 13 showed a retarded band, but probe 13M did not; this result suggests that CTCF specifically interacts with the RX4 sequence (Fig. 3I).
Mutation of 11P Decreased Rxrb Promoter Activities in Colony Assays with L6 Cells-To investigate whether the cartilagespecific enhancer activities of Col11a2 are blocked from affecting the Rxrb promoter and whether possible CTCF-binding sites or the intergenic sequence are involved in this blockage, we performed colony assays. We tested the 11P site and the RX4 site by introducing 14-bp substitution mutations at these sites. In addition, we tested the longer fragment (507 bp) containing the 11P site within the intergenic region.
We prepared human BAC clone CTD-2054I15 (Invitrogen) that contained a 142-kb genomic DNA insert with RXRB and COL11A2 sequences in the middle. By using BAC recombineering techniques, we introduced a 14-bp substitution mutation at 11P, a 507-bp deletion mutation including the 11P site, and a 14-bp substitution mutation at the RX4 site to prepare 11Psub, 11Pdel, and RX4sub constructs, respectively (Fig.  4A). The sequences of the 14-bp substitution mutations in 11Psub and RX4sub were identical to those of probes h-M5Bgl and 13M, respectively, which we used in EMSA and thus contained BglII cleavage sites. The 507-bp deleted sequence in 11Pdel corresponded to Ϫ1148 to Ϫ642 of the translational start site of human COL11A2. Previous reports showed tissuespecific regulatory elements (27) and a Sox9-binding element (34) in the intergenic sequence between mouse Rxrb and Col11a2. The human sequences corresponding to these elements reside at Ϫ1148 to Ϫ1125 and Ϫ1123 to Ϫ1107 from the translational start site, respectively. Thus, these elements in addition to 11P were deleted in 11Pdel. Next, we inserted a puromycin-resistance gene sequence into the 1st exon of RXRB in the original BAC clone (WT), 11Psub, 11Pdel, and RX4sub

Insulation of Rxrb from Cartilage-specific Col11a2
OCTOBER 10, 2008 • VOLUME 283 • NUMBER 41 constructs to generate WT-Pur, 11Psub-Pur, 11Pdel-Pur, and RX4sub-Pur transgenes, respectively (Fig. 4B). The BAC construction and the amounts of BAC DNAs used for transfection were checked by electrophoresis after digestion with BglII (supplemental Fig. S1). Lipofectamine 2000 was used to introduce 2 g of each transgene into L6 cells, and the cells were cultured in the presence of 5 g/ml of puromycin for 7 days. The mean numbers of colonies formed in cultures transfected with 11Psub-Pur or 11Pdel-Pur were significantly decreased compared with the colonies in cultures with WT-Pur (Fig. 4C). The mean number of colonies formed in cultures transfected with RX4-Pur was not significantly different from the number in WT-Pur cultures. These results suggest that 11P is an enhancer for the RXRB promoter in L6 cells. Alternatively, 11P might block the silencer activity of the Col11a2 gene from affecting the Rxrb promoter in L6 muscular cells.
Mutation of 11P Increased RXRB Expression in Transgenic RCS Cells-We also performed colony assays on RCS cells after introducing these transgenes; however, few colonies formed, probably due to the low transfection efficiency of RCS cells and the weak transcriptional activities of Rxrb in RCS cells. Therefore, we employed another approach for RCS cells. We inserted the SV40 early promoter-puromycin resistance gene cassette at the end of the insert of the WT, 11Psub, and 11Pdel human BAC clones to generate SV40-Pur/WT, SV40-Pur/11Psub, and SV40-Pur/11Pdel (Fig. 4D). We introduced these constructs into RCS cells that were cultured in the presence of puromycin until colonies formed. Isolated colonies were picked up and transferred to separate dishes to establish stably transformed RCS cells bearing the BAC transgene constructs. We extracted total RNAs from each transformant and performed real-time RT-PCR. PCR with human primers showed that there were no significant differences in the mean relative expression levels of human COL11A2 mRNAs to puromycin mRNAs between RCS cells bearing SV40-Pur/11Psub or SV40-Pur/11Pdel and RCS cells bearing SV40-Pur/WT. The mean relative expression levels of human RXRB mRNAs to puromycin mRNAs were significantly higher in RCS cells bearing SV40-Pur/11Psub or SV40-Pur/11Pdel than in RCS cells bearing SV40-Pur/WT (Fig. 4E). There was no amplification with human primers and cDNAs from the original RCS cells as templates, suggesting the specific amplification of human transgene products (Fig. 4F). PCR with rat primers demonstrated that there were not significant differences in the mean relative expression levels of rat endogenous Col11a2 or rat endogenous Rxrb mRNAs to rat Gapdh mRNAs between RCS cells bearing SV40-Pur/11Psub or SV40-Pur/ 11Pdel and RCSs bearing SV40-Pur/WT (data not shown). These results suggest that 11P is a silencer for the RXRB promoter in RCS cells. Alternatively, 11P might block the enhancer activity of the Col11a2 gene from affecting the Rxrb promoter in RCS cells.
Effect of CTCF-binding Sites and the Intergenic Sequence on RXRB Promoter Activity in Transgenic Mice-To examine the effects of the intergenic sequence between RXRB and COL11A2 and CTCF-binding sites on the RXRB promoter activities in vivo, we generated transgenic mice bearing BAC DNA containing RXRB and COL11A2 genes. We inserted a LacZ sequence into the 1st exon of RXRB in each of the WT, 11Psub, 11Pdel, and RX4sub constructs to generate WT-LacZ, 11Psub-LacZ, 11Pdel-LacZ, and RX4sub-LacZ transgenes, respectively (Fig.  5A). LacZ was used to monitor RXRB promoter activity. The A, schematic representation of mutations introduced into the BAC clone, CTD-2054I15. We introduced a 14-bp substitution mutation at 11P, a 507-bp deletion mutation including the 11P site, and a 14-bp substitution mutation at the RX4 site to prepare 11Psub, 11Pdel, and RX4sub constructs, respectively. Mutated sequences in 11Pub and RX4sub contained BglII cleavage sites. B and C, colony assays in L6 cells. B, to monitor transcriptional activities of the RXRB gene in colony assays, we inserted the puromycin-resistance gene sequence into the 1st exon of RXRB in each construct to generate WT-Pur, 11Psub-Pur, 11Pdel-Pur, and RX4sub-Pur transgenes, respectively. C, 2 g of each transgene was introduced into L6 cells. Cells were then cultured in the presence of puromycin and stained with crystal violet, and the numbers of colonies were counted. Error bars indicate the mean Ϯ S.D. (n ϭ 3). *, p Ͻ 0.01 between WT-Pur and constructs bearing a mutation, as determined by the Student's t test. D-F, establishment of transgenic RCS cells. D, we inserted a SV40 early promoter-puromycin resistance gene cassette into each construct to generate SV40-Pur/WT, SV40-Pur/11Psub, and SV40-Pur/11Pdel, respectively. E, we introduced transgene constructs into RCS cells and cultured them in the presence of puromycin, picked up colonies, and established stably transformed RCS cells. RNAs were extracted from each transformant and subjected to real-time RT-PCR. Bars indicate mean values. *, p Ͻ 0.05 and **, p Ͻ 0.01 as determined by the Student's t test. F, RT-PCR amplification with human Col11a2 or RXRB primers and cDNAs extracted from transgenic RCS cells bearing SV40-Pur/WT or from original RCS cells as templates. Specific products were amplified with human primers only from transgenic cells, suggesting no amplification of endogenous rat sequences with the human primers.
four BAC transgenes were linearized by digestion with PacI restriction enzyme. Thirteen days after microinjection of transgenes, embryos were recovered, and the forelimbs on one side were dissected, frozen-sectioned, and subjected to X-gal staining. The remaining bodies were also stained with X-gal. Based on the patterns of X-gal staining, the founder mice were divided into two groups: 1) mice with X-gal staining in a variety of tissues, and 2) mice with X-gal staining mainly in cartilage. We obtained 13 founder transgenic embryos bearing the WT-LacZ transgene with positive X-gal staining. Seven embryos (54%) showed X-gal staining in a variety of tissues, whereas 6 embryos (46%) had X-gal staining mainly in the cartilage (Fig. 5B). X-Gal staining patterns of WT-lacZ transgenic founder embryos are shown in supplemental Fig. S3. To examine the possibility that the transgene constructs were degraded and only part of the BAC DNAs had integrated into the genome, we performed genomic PCR with three sets of primers that are located at the 5Ј-end, 3Ј-end, and the middle of the transgenes (Fig. 5A). The 3 sets of primers produced PCR products of the correct size from genomic DNAs extracted from all transgenic founder mice; but, no products were amplified from the genomic DNA of the non-transgenic embryos (Fig. 5C). On the assumption that BAC DNAs were integrated into the genome without degradation, the staining patterns of WT-LacZ transgenic mice suggest that the 150-kb BAC construct was not sufficient to recapitulate endogenous RXRB expression patterns. This broad domain structure of Rxrb contrasts the compactness of the Col11a2 domain, as the 2.3-kb sequence of Col11a2 was able to recapitulate expression patterns of Col11a2, even when linked to heterologous promoters.
The percentage of mice with cartilage-specific LacZ expression increased to 62 and 75% by introducing a 14-bp mutation at the 11P or RX4 sites, respectively (Fig. 5B). We cannot judge whether these increases were substantial or not. Deletion of a 507-bp stretch within the intergenic region increased the proportion to 90% (9 of 10 mice) (Fig. 5B). The X-gal staining patterns of 11Pdel-LacZ transgenic founder embryos are shown in supplemental Fig. S4. These results suggest that the 507-bp deletion in the intergenic region might convert the expression pattern of Rxrb to that of Col11a2 in the BAC transgenic mice to some extent. The integrations of correctly modified BAC constructs were examined by digestion of the genomic PCR products with BglII (Fig. 5D).
CTCF Associated with RX4, but Not with 11P in Vivo-To examine whether CTCF binds to 11P or RX4 sites in vivo, we prepared nuclei from L6 cells and performed ChIP assays with anti-CTCF antibodies. The amount of CTCF-attached DNA fragments were quantified by real-time PCR. The primers designed for RX4 amplified a product from samples immunoprecipitated with two commercially available anti-CTCF antibodies (Fig. 6). These two antibodies were raised against different domains of the CTCF protein. On the other hand, primers designed for 11P amplified only a small amount of product from  A) and genomic DNA extracted from each mouse (as indicated at top). Non-Tg, normal mice that did not bear a transgene; DW, distilled water. D, BglII digestion of the PCR products amplified with mouse genomic DNA as templates. 188-bp products amplified with DI001/JM412 primers from RX4sub-LacZ transgenic mice were cleaved with BglII to 107-and 81-bp fragments, whereas the PCR products from WT-LacZ or 11Psub-LacZ or 11Pdel-LacZ were not cleaved. 534-bp products amplified with NT418/419 primers from the 11Psub-LacZ transgenic mouse were cleaved with BglII to 232-and 302-bp fragments, whereas the PCR products from WT-LacZ or RX4sub-LacZ were not cleaved. There were no amplified products from 11Pdel-LacZ due to the absence of annealing sites for NT 418. FIGURE 6. CTCF associated with RX4, but not with 11P in vivo. After treatment of RCS cells with formaldehyde, lysates were subjected to chromatin immunoprecipitation with or without antibodies as indicated. Co-immunoprecipitating sequences were detected using real-time RT-PCR. The -fold difference of a sequence at 11P precipitated by the antibodies indicated was set as 1. Error bars indicate mean Ϯ S.D. (n ϭ 3).

Insulation of Rxrb from Cartilage-specific Col11a2
OCTOBER 10, 2008 • VOLUME 283 • NUMBER 41 samples immunoprecipitated with the two anti-CTCF antibodies. A substantial amount of products were amplified by primers designed for RX4 but not by primers designed for 11P from samples immunoprecipitated with anti-Rad21 antibodies. Rad21 is one of the subunits of cohesin. These results suggest that RX4 associates with CTCF and cohesin and that 11P does not associate with either CTCF or cohesin in nuclei. It has been reported that Brother of the Regulator of Imprinted Sites (BORIS) possesses an 11-zinc finger region that is highly homologous to that of CTCF, suggesting similar DNA-binding potential (35). To examine whether BORIS interacts with 11P sites in nuclei, we performed ChIP assays with anti-BORIS antibodies. Primers designed for 11P amplified a small amount of product from samples immunoprecipitated with anti-BORIS antibodies that was less than the amplified amount without anti-BORIS antibodies, suggesting that BORIS did not interact with 11P sites in vivo.

DISCUSSION
RXRB and COL11A2 are located in the major histocompatibility complex region of chromosome 6p in humans. This genomic organization is preserved in mice, as Rxrb and Col11a2 are located in the major histocompatibility complex region of chromosome 17. It has not been investigated how the Rxrb and Col11a2 domains are separated.
CTCF is the only example in vertebrates where an identified protein and binding site have unambiguously been implicated in enhancer blocking activities. But it was reported that elements between T cell receptor genes and the Dad1 gene can block enhancers but do not harbor CTCF sites, suggesting the presence of CTCF-independent insulators (36). In the present study, we demonstrated that CTCF binds to the 11P sequence by EMSA. Colony assays showed that 11P was functional. However, ChIP assays showed that CTCF, Rad21, or BORIS did not bind to the intergenic region of Col11a2. It remains to be determined if 11P associates with other unknown transcriptional factors that recognize a motif similar to CTCF and regulates the transcription of Rxrb as demonstrated by colony assays. This indicates the presence of novel enhancer blocking elements required for tissue-restricted expression potentially by a previously unrecognized mechanism.
We employed BAC transgenic mice to examine the functions of the intergenic sequence between Rxrb and Col11a2 and the CTCF binding site within the Rxrb gene. The BAC clone we used contained entire COL11A2 and RXRB genes. But the WT-LacZ BAC transgene failed to recapitulate the endogenous expression patterns of RXRB. The WT-LacZ transgene directed substantial LacZ expression in the cartilage of all transgenic mice; in fact, in nearly half of the mice, the LacZ expression was most prominent in the cartilage, although the expression of endogenous Rxrb in cartilage was weak compared with its expression in other tissues. These results suggest that the Col11a2 domain affected the transcriptional activity of Rxrb in the BAC transgene, rather than that the sequences around the integrated sites directed the expression of LacZ to cartilage. These results suggest that insulation of the Rxrb promoter from the Col11a2 enhancer may be regulated by an unknown remote regulatory sequence that the BAC transgene did not cover, at least in part. The percentage of transgenic mice expressing LacZ predominantly in cartilage increased to 90 from 46% by deleting a 507-bp section of the intergenic sequence. In these mice, the LacZ expression decreased in non-cartilaginous tissues. These results suggest that a 507-bp region of the intergenic sequence may be involved in blocking the Col11a2 enhancer and silencer from affecting the Rxrb gene.
Our results suggest that a 507-bp region of the intergenic sequence is involved in the insulation of the Rxrb promoter from the Col11a2 enhancer/silencer. This 507-bp stretch spans from Ϫ530 to Ϫ90 of the transcriptional start site of Col11a2 in mice, being within the promoter of Col11a2. Indeed, elements within the 507-bp stretch contain a basal promoter (8) and cisacting elements that increase the transcription in cartilage (27,34) or neural tissues (27). To assay enhancer-blocking activities, typical colony assays use experimental constructs containing promoter, enhancer, and blocking elements. An element with enhancer-blocking activity interferes with enhancer-promoter communication when inserted between the enhancer and the promoter. The same element has little or no effect on transcriptional activation when present in a position flanking the promoter-enhancer pair (37). Because the 507-bp stretch contains cis-acting elements, it may affect the promoter when present in the flanking region of a promoter-enhancer pair. Thus, we did not employ typical colony assays for testing the 507-bp stretch. Instead, we performed colony assays using BAC transgenes. Deletion of the 507-bp stretch from the intergenic sequence and a substitution mutation at the 11P site decreased the promoter activities of RXRB in L6 cells, in which Col11a2 expression is suppressed, and increased the promoter activities of RXRB in RCS cells, in which Col11a2 expression is activated. These results may be consistent with the characteristics of enhancer-blocking elements; however, we cannot classify these intergenic sequences as a true insulator because we did not perform typical colony assays. The mingling of a promoter with a possible insulator reminds us of the idea that there might be considerable overlaps in the functions of promoters, enhancers, silencers, and insulators in vivo (14). Insulators may share the same bag of tricks as other regulatory elements, such as enhancers, silencers, and upstream activating sequence, which they combine in various ingenious ways to acquire specific properties.
The cartilage-derived retinoic acid-sensitive protein (Cdrap/ Mia) gene is primarily expressed in cartilage and located closely flanked with housekeeping genes Snrpa and Rab4b, which are expressed ubiquitously (38). Removal of enhancer/suppressor sequences of the Cdrap gene resulted in broad tissue expression (39). These findings and our data will contribute to understanding the mechanism that regulates cartilage-specific expression from multigenic loci.
In summary, we identified a 507-bp intergenic sequence that insulates the Rxrb promoter from the Col11a2 enhancer/silencer in cell culture and transgenic mouse experiments. 11P was found to be active in cell culture experiments. Our results suggest that the intergenic sequence including 11P insulates the Rxrb promoter from the Col11a2 enhancer, probably associating with unknown factors that recognize a motif similar to CTCF.