A Nuclear Matrix Attachment Region Upstream of the T Cell Receptor β Gene Enhancer Binds Cux/CDP and SATB1 and Modulates Enhancer-dependent Reporter Gene Expression but Not Endogenous Gene Expression*

We have previously identified a DNase I-hypersensitive site in the T cell receptor β locus, designated HS1, that is located 400 base pairs upstream of the transcriptional enhancer Eβ and is induced during CD4−CD8− to CD4+CD8+thymocyte differentiation. Using electrophoretic mobility shift assays, we show that HS1 induction correlates with increased binding of two nuclear factors, Cux/CDP and SATB1, to a 170-base pair DNA sequence within HS1. Furthermore, we demonstrate that HS1 is a nuclear matrix attachment region, referred to as MARβ. These findings demonstrate that an analogous organization of cis-regulatory elements in which a nuclear matrix attachment region is in close proximity to an enhancer is conserved in the immunoglobulin and T cell receptor loci. In addition, we show that MARβ represses Eβ-dependent reporter gene expression in transient transfection assays. However, the targeted deletion of MARβ from the endogenous locus does not change T cell receptor β gene transcription in developing T cells. These contrasting results suggest a potential pitfall of functional studies of nuclear matrix attachment regions outside of their natural chromosomal context.

We have previously identified a DNase I-hypersensitive site in the T cell receptor ␤ locus, designated HS1, that is located 400 base pairs upstream of the transcriptional enhancer E ␤ and is induced during CD4 ؊ CD8 ؊ to CD4 ؉ CD8 ؉ thymocyte differentiation. Using electrophoretic mobility shift assays, we show that HS1 induction correlates with increased binding of two nuclear factors, Cux/CDP and SATB1, to a 170-base pair DNA sequence within HS1. Furthermore, we demonstrate that HS1 is a nuclear matrix attachment region, referred to as MAR ␤ . These findings demonstrate that an analogous organization of cis-regulatory elements in which a nuclear matrix attachment region is in close proximity to an enhancer is conserved in the immunoglobulin and T cell receptor loci. In addition, we show that MAR ␤ represses E ␤ -dependent reporter gene expression in transient transfection assays. However, the targeted deletion of MAR ␤ from the endogenous locus does not change T cell receptor ␤ gene transcription in developing T cells. These contrasting results suggest a potential pitfall of functional studies of nuclear matrix attachment regions outside of their natural chromosomal context.
The rearrangement and expression of the T cell receptor ␤ (TCR␤) 1 gene is essential to early T lymphocyte development (1). Prior to TCR␤ gene rearrangement, germline transcripts are initiated upstream of the D ␤ 1 gene segment in CD4 Ϫ CD8 Ϫ (double negative, DN) thymocytes (2,3). Recombination of TCR␤ gene occurs exclusively during the DN stage of thymocyte development by an ordered process wherein D ␤ -J ␤ rearrangement occurs prior to joining of a V ␤ gene segment. Allelic exclusion operating at the level of V ␤ to D ␤ J ␤ rearrangement ensures that only one of the two possible TCR␤ alleles are expressed by an individual T cell (4,5). After V ␤ D ␤ J ␤ rearrangement, a mature transcript is initiated from the V ␤ promoter in a T cell-specific manner (6). To achieve the lineage-, stage-, and allele-specific TCR␤ gene rearrangement and transcription, many cis-acting elements and their associated transacting factors are likely to be involved (4,5).
To date, the TCR␤ gene enhancer (E ␤ ) is the only cis-regulatory element demonstrated to be required for both the lineage-and stage-specific transcription and rearrangement of the TCR␤ gene (7)(8)(9)(10)(11)(12). Although the V ␤ promoter is required for lineage-specific TCR␤ transcription, its role in regulating V ␤ gene rearrangement remains unclear (13)(14)(15)(16)(17). In addition to V ␤ promoters and E ␤ , there are likely other cis-regulatory elements involved in the control of various aspects of TCR␤ gene rearrangement and transcription. In particular, nuclear matrix attachment regions (MAR) are a class of cis-regulatory elements found in many genetic loci that are distinct from transcriptional promoters and enhancers, and yet are often closely associated with these regulatory elements (18 -20). MARs are typically AT-rich DNA sequences that bind to the nuclear matrix, often contain topoisomerase II cleavage sites, and exhibit a propensity for base unpairing when subjected to superhelical strain (21,22). They have been proposed to be involved in transcription, DNA recombination, replication, and repair (23). In the immunoglobulin heavy chain (IgH) locus, MARs flank the intronic enhancer E and are in close proximity to V H promoters (22)(23)(24)(25)(26). Reporter gene assays in cell lines and transgenic mice have suggested that these MARs exert both positive and negative effects on IgH gene transcription and promote long range chromatin accessibility (27)(28)(29)(30)(31). A highly conserved MAR is also found 200 base pairs (bp) upstream of the intronic immunoglobulin (Ig) enhancer in mouse, human, and rabbit (22,32). Together, the Ig MAR and enhancer promote demethylation, transcription, recombination, and somatic hypermutation of the locus although no specific function has been attributed to the MAR alone (33)(34)(35)(36). The presence of MARs at other antigen receptor loci has not been reported.
To characterize novel cis-regulatory elements involved in controlling TCR␤ gene rearrangement and/or transcription, we previously screened a 100-kb region of the TCR␤ locus and identified along with E ␤ 10 additional DNase I-hypersensitive sites (HS) (37). HS1 was previously shown to be just 400 bp upstream of E ␤ and to be strongly induced during DN to DP thymocyte differentiation. In this report, we localize nuclear factor binding sites within HS1, characterize two nuclear factors that bind to HS1, demonstrate that HS1 is a nuclear matrix attachment region, and reveal the potential pitfalls of functional analyses of MARs outside of their natural chromosomal context.
Luciferase reporter constructs were prepared using the pGL2 promoter vector (Promega, Madison, WI). The existing SV40 promoter in pGL2 was deleted by BglII-HindIII digestion, and a 424-bp EcoRI-NcoI fragment containing the V ␤ 13 promoter was inserted to generate construct 1 (Fig. 8A). Construct 1 was then used to generate constructs 2 and 3. An 830-bp BglII-NcoI fragment containing E ␤ was cloned into the BamHI site located downstream of the poly(A) site of the luciferase gene, generating construct 2. A 1000-bp BsgI-NcoI fragment containing HS1 and E ␤ in their natural configuration was cloned into the same position of construct 1 to generate construct 3. Therefore, construct 3 differed from construct 2 only by having the 170-bp HS1 (Fig. 8A).
Nuclear Extracts-Nuclear extracts were prepared from DN and DP thymocytes of RAG-deficient mice and RAG-deficient mice complemented with an activated lck transgene, respectively (37). Nuclei from thymocytes were prepared following the method of Forrester et al. (38). Nuclei were resuspended in 0.2 ml of buffer A containing 10 mM Tris-Cl, pH 7.5, 10 mM HEPES, 10 mM MgCl 2 , 1 mM DTT, 50 mM NaCl, and 20% glycerol. An equal volume of buffer A supplemented with 420 mM NaCl was then added slowly to the nuclei suspension and mixed immediately. The resulting mixture was incubated on ice for 10 min and then centrifuged at 14,000 ϫ g for 5 min at 4°C. Supernatants were then dialyzed against a low salt buffer containing 20 mM Tris-Cl, pH 7.5, 10 mM MgCl 2 , 1 mM DTT, 50 mM NaCl, and 10% glycerol, and were dispersed into small aliquots and stored at Ϫ80°C. Protein concentrations of nuclear extracts were determined using a protein assay kit from Bio-Rad.
Electrophoretic Mobility Shift Assays (EMSA)-All DNA probes for EMSA were end-labeled with [␣-32 P]dCTP using Klenow fragment of DNA polymerase I. Reaction mixtures were passed through a Nuc-trap column (Stratagene, La Jolla, CA) and precipitated. The labeled DNA fragments were further purified on a 5% polyacrylamide gel. For EMSA, 0.2-20 g of nuclear extract or 1.25-5 ng of purified Cux/CDP protein were incubated with 2 ng of labeled DNA probes (ϳ20,000 cpm) in 24 l of reaction buffer containing 10 mM HEPES-KOH, 10 mM Tris glutamate, pH 8.0, 100 mM NaCl, 10 mM magnesium glutamate, 50 mM potassium glutamate, 10% glycerol, 2 mM DTT, and 100 g/ml of poly(dI-dC) (Sigma) at room temperature for 10 min. For EMSA in the presence of antiserum, probes were first incubated with the nuclear extracts as above and then 5 l of diluted antiserum or control preimmune serum were added and incubated for another 10 min. Anti-Cux/ CDP antiserum was kindly provided by Dr. Ellis Neufeld of Children's Hospital (Boston, MA), anti-SATB1 antiserum by Dr. Terumi Kohwi-Shigematsu of Lawrence Berkeley Laboratory (Berkeley, CA), and purified Cux/CDP protein by Dr. Richard Scheuermann of the University of Texas Southwestern Medical Center (Houston, TX). For EMSA in the presence of competing oligonucleotides, probes were mixed with competing oligonucleotides first and then nuclear extract was added into the mixture and incubated as above. DIST (5Ј-GCTTTTCAGTTGACC-GGTGATTATTAGCCAATTTCTGATAAAAAGAAAAGGAAACCGATT-GC-3Ј) and ␥-globin (5Ј-TGCCTTGACCAATAGCCTTGACAAGGCAAA-CTTGACCAATAGTCTTAGAGTATCCAGTG-3Ј) oligonucleotides were from Dr. Ellis Neufeld, and CD8␣ MAR was isolated as a 200-bp PstI-EcoRI fragment from a plasmid kindly provided by Dr. Paul Gottlieb of the University of Texas (Austin, TX). Fifteen l of the incubation mixture were electrophoresed in a 4.5% polyacrylamide gel (acrylamide: bisacrylamide, 40:1) in 0.5ϫ TBE (Tris borate-EDTA) buffer at 15 V/cm for 4 h in the cold room. Gels were dried and autoradiographed.
MAR-binding Assay-Soluble nuclear matrix was isolated from DP thymocytes following the standard protocol (22). The matrix binding assay was performed following the method described by Zong and Scheuermann (39) with minor changes. Briefly, 10 l of nuclear matrix (corresponding to 1 ϫ 10 7 nuclei) was incubated with 2 ng of each labeled probe, and 20 or 40 g of sonicated Escherichia coli DNA at the room temperature for 1 h in a total volume of 100 l containing 10 mM Tris-Cl, pH 7.4, 50 mM NaCl, 2 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin, and 0.25 mg/ml bovine serum albumin. The mixture was centrifuged for 5 min and the pellet was washed in 1 ml washing buffer containing 10 mM Tris-Cl, pH 7.4, 50 mM NaCl, 2 mM EDTA, and 0.25 mg/ml bovine serum albumin. The nucleoprotein complex was treated with proteinase K in the presence of 10 mM Tris-Cl, pH 7.4, 2 mM EDTA, and 5 g/ml salmon sperm DNA at 37°C overnight. The mixture was extracted with phenol-chloroform, and DNA was recovered by alcohol precipitation. Bound DNA was resolved on a 4.5% acrylamide gel and visualized by autoradiography.
Screening Expression cDNA Library-The expression cDNA library was kindly provided by Dr. Linda Clayton of the Dana-Farber Cancer Institute (Boston, MA). The library was constructed in ZAP Express vector (Stratagene, La Jolla, CA) using cDNA prepared from mRNA of DP thymocytes. To identify cDNA clones that encode proteins capable of binding probe IV (40), 5 ϫ 10 4 phages were plated per 150-mm plate (a total of 13 plates were screened). Plates were incubated for 3 h at 42°C until tiny plaques were visible. Plates were then moved to a 37°C incubator, overlaid with nitrocellulose filters that have been soaked in 10 mM isopropyl-1-thio-␤-D-galactopyranoside, and incubated for 6 h. Filters were blocked with BLOTTO containing 5% Carnation nonfat milk powder, 50 mM Tris-Cl, pH 7.5, 50 mM MaCl, 1 mM EDTA, and 1 mM DTT and then hybridized with 32 P-labeled probe IV in the binding buffer containing 50 mM Tris-Cl, pH 7.5, 50 mM NaCl, 1 mM EDTA, 1 mM DTT, and 5 g/ml sonicated calf thymus DNA. Filters were washed in the binding buffer and exposed to x-ray film. Putative positive plaques were isolated and screened for two more rounds as above until single positive plaques were obtained.
Transient Transfection and Luciferase Assay-Thymoma line P4890 and P4833 were derived from DN thymocytes of mice deficient in both p53 and RAG1. Thymoma line 4b was derived from a DP thymocyte of a mouse deficient in both p53 and TCR␣ (41). Thymoma EL4 was derived from a CD4 ϩ thymocyte. M12, a murine B cell lymphoma, was also used in the assays. Cells were maintained in Dulbecco's modified Eagle's medium containing 10 mM HEPES, 10% heat-inactivated fetal bovine serum, 50 M ␤-mercaptoethanol, 100 units/ml penicillin, 100 g/ml streptomycin, and 2 mM L-glutamine. For transient transfection, 2 ϫ 10 7 cells were mixed with 10 g of construct DNA and 10 g of control CMV-␤-galactosidase plasmid in 0.5 ml of RPMI. The mixture was electroporated at 300 V and 960 microfarads at room temperature. Cells were immediately resuspended in 10 ml of Dulbecco's modified Eagle's-HEPES medium and cultured for 48 h. Cell lysates were prepared, and luciferase activity was measured using a luciferase assay kit from Promega (Madison, WI). ␤-Galactosidase activity was measured by O-nitrophenyl-␤-D-galactopyranoside assay. Luciferase activity of individual transfection was normalized to ␤-galactosidase activity of the same sample.
Northern Blotting Assay-Total RNA was isolated from thymocytes of wild type and homozygous mutant mice that were deleted of HS1 from the TCR␤ locus (37). RNA was fractionated on a formaldehyde gel and transferred onto Zeta-probe filters. Filters were hybridized individually with a C ␤ 2 probe, a V ␤ 8 probe, or a D ␤ 1-J ␤ 1 intronic probe and exposed to x-ray films. The probes used were as follows: C ␤ 2 probe, a 430-bp cDNA fragment; V ␤ 8 probe, 190-bp EcoRI-StuI fragment containing the V ␤ 8 gene segment; and D ␤ 1-J ␤ 1 intronic probe, a 663-bp PCR product.

RESULTS
Localization of Two Nuclear Factor Binding Sites within HS1-HS1 was previously mapped to a 780-bp BstXI-BglII fragment located 400 bp upstream of E ␤ (Fig. 1A) (37,42). To identify protein factors that bind to HS1, we performed EMSAs with nuclear extracts from DP thymocytes because HS1 was most prominent in these cells (37). Two subfragments of HS1 were used as probes in EMSA, a 480-bp BstXI-AccI fragment (probe I) and a 300-bp AccI-BglII fragment (probe II, Fig. 1A). While probe I failed to yield any complex, probe II readily generated two shifted complexes in EMSA (Fig. 1B, lanes 1-4). The same complexes were also observed when the entire 780-bp BstXI-BglII fragment was used as a probe, although at a substantially reduced level probably due to the large size of the probe (data not shown).
We next sought to narrow down the location of the factor binding sites within probe II. Previously, sequence analyses showed the presence of consensus motifs of AP-1, Oct-1, and c-Myb binding sites in the 5Ј region of the 300-bp probe II in both mouse and human (37,42). Therefore, two subfragments of probe II, a 128-bp AccI-BsgI fragment (probe III) and a 170-bp BsgI-BglII fragment (probe IV), were used in EMSA to test if factor binding occurred at any of these consensus sites.
Despite the presence of the consensus nuclear factor binding sites within probe III, it did not generate any complex in EMSA (Fig. 1B, lanes 5 and 6). Likewise, a slightly larger AccI-BsmAI fragment did not produce any complex ( Fig. 1A and data not shown). In contrast, probe IV detected the two shifted complexes as seen when probe II was used (Fig. 1B, lanes 7 and 8).
Thus, although the 128-bp fragment (probe III) contained the various sequence motifs (42), the complexes detected under our reaction conditions resulted from nuclear factor binding to cis elements in the 170-bp BsgI-BglII region of HS1 (probe IV). The observed nuclear protein-DNA interactions were specific. First, the two complexes were stable in the presence of an excess amount of nonspecific competitor poly(dI-dC) (100 g/ ml). Second, a greater amount of complexes was formed with increasing amounts of DP thymocyte nuclear extract (Fig. 1B). Finally, the formation of both complexes was specifically abolished by non-radiolabeled probe IV but not probe III in a competition experiment (Fig. 2).
To further delineate the location of the nuclear factor binding sites within probe IV, smaller probes were generated by deleting sequences from the 5Ј and 3Ј flanks of probe IV and used in EMSA (Fig. 1). Probe V, which was 30 bp shorter than probe IV at the 3Ј end formed the lower complex at a level comparable to that of probe IV, but did not efficiently form the upper complex, as only a miniscule amount was detected (Fig. 1B, lanes 9 and  10). This observation suggested that the upper and lower complexes were formed by factor binding to two different regions of probe IV. Factor binding at the far 3Ј region of the probe produces the upper complex and factor binding somewhere else in the probe produces the lower complex. Supporting this notion, probe VI, which has 21 bp removed from the 5Ј end of probe IV, formed only approximately a third as much of the lower complex as probe IV (Fig. 1B, lanes 11 and 12). An additional 3-fold reduction in the formation of the lower complex resulted with probe VII, which had additional 33 bp removed from the 5Ј end of probe VI (Fig. 1, A and B, lanes 13 and  14). In contrast, in both of these cases, the upper complex was not significantly reduced, confirming its probable binding to the far 3Ј region of probe IV. The binding site for the formation of the lower complex was probably broad because deletion from 5Ј end of probe IV substantially reduced the amount of the complex and probe IX containing only the 5Ј third of probe IV did not form a complex (Fig. 1, A and B, lanes 17 and 18). The 3Ј region of probe IV appears to stabilize the lower complex FIG. 1. Localization of cis-regulatory elements of HS1 to a 170-bp sequence. A, schematic diagrams of HS1 at the TCR␤ locus and of the probes derived from the 780-bp BstXI-BglII fragment encompassing HS1. The numbers above the probe fragments indicate the sizes in base pairs of the probes. Formation or lack of formation of shifted complexes of the probes in EMSA are indicated by ϩ or Ϫ, respectively, and the number of ϩ indicate the relative amounts of complexes formed as shown in Fig. 1B. Some restriction enzymes used in the probe isolations are indicated (see "Experimental Procedures" for details). B, formation of protein-DNA complexes by different probes and nuclear extracts from DP thymocytes. Two ng of each 32 P-labeled probe were incubated with either 10 or 20 g of nuclear extract in the presence of 100 g/ml poly(dI-dC) and electrophoresed on a 4.5% polyacrylamide gel. The upper (up) and lower (low) complexes and unbound probes (Free) are indicated, whereas the uppermost bands represent trapped probes in the wells. because no complexes were formed using probe VIII that had 30 bp deleted from the 3Ј end of probe VI (Fig. 1B, lanes 15 and  16). In summary, these data suggest that there are two major factor binding sites in the 170-bp BsgI-BglII fragment (probe IV) of HS1.
Correlation of Factor Binding with the Induction of HS1 during DN to DP Thymocyte Differentiation-We have previously shown that HS1 is induced during DN to DP thymocyte differentiation. In DN thymocytes of RAG2-deficient mice, HS1 is barely detectable, whereas HS1 is a major DNase I-hypersensitive site in DP thymocytes of RAG2-deficient mice that have been complemented with a functional TCR␤ or activated lck transgene, or have been treated with anti-CD3⑀ antibody (37). We reasoned that if the formation of the complexes in EMSA correlated with HS1 formation, we should detect less complex formation when nuclear extracts from DN thymocytes are used in EMSA. To test this, we performed EMSA using probe IV and nuclear extracts prepared from DN thymocytes of RAG2-deficient mice and from DP thymocytes of RAG2-deficient mice that have been complemented with an activated lck transgene. As shown in Fig. 3, both the upper and lower complexes were formed using nuclear extracts from DN thymocytes; however, the amounts of complexes yielded were only a quarter of those formed by the equivalent amounts of nuclear extracts from DP thymocytes. Thus, the formation of both the upper and lower complexes was induced in DP thymocytes, correlating with the induction of HS1 during DN to DP thymocyte differentiation. These findings support the possibility that the nuclear factors detected by probe IV in EMSA are directly involved in the formation of HS1.
Identification of Proteins That Bind to Cis Elements in HS1-To determine the identity of protein factors in the upper and lower complexes, we screened an expression cDNA library constructed in phage using mRNA from DP thymocytes with the 170-bp probe IV. Five positive plaques were obtained from screening 6.5 ϫ 10 5 phages. DNA was prepared from all five phage clones and analyzed by restriction enzymes. Of the five clones, two contained cDNA inserts that gave distinct restriction patterns (data not shown), indicating that they were unique. The other three clones contained related cDNA inserts, as indicated by the same sizes and restriction patterns (data not shown). Sequence analysis revealed that one of the unique cDNA inserts encoded the deoxycytidine kinase. Considering the function of the deoxycytidine kinase, its binding to probe IV was probably nonspecific. The second unique phage clone contained a 4-kb cDNA insert identical to the 3Ј region of the mouse Cux gene (CDP in human) (43)(44)(45)(46). The other three phage clones contained a cDNA whose sequences did not match any known genes, although identical sequence fragments were found in EST data base. Thus, this cDNA encodes a novel protein; whether it binds to HS1 specifically is currently under investigation.
Specific Binding of Cux/CDP to HS1-Cux/CDP is a homeodomain protein originally identified in Drosophila (called Cut) and later in mouse and human (43)(44)(45)(46). In addition to its homeodomain, Cux/CDP also contains three cut repeats that can independently bind DNA of relatively degenerate consensus sequences (47)(48)(49). To determine whether Cux/CDP binds to HS1 specifically, we first performed a competition assay using DIST oligonucleotide derived from the promoter of cytochrome b heavy chain gene (gp91 phox ), which was previously characterized to efficiently and specifically bind Cux/CDP in EMSA (50). As shown in Fig. 4A (lanes 1-6), increasing amounts of the unlabeled DIST oligonucleotide progressively abolished the formation of both the upper and lower complexes detected by probe IV. Under the same conditions, an oligonucleotide derived from the ␥-globin promoter that was previously shown to only weakly bind Cux/CDP (51) had minimal effect on the formation of the complexes (Fig. 4A, lanes 7-10). These data indicate that the 170-bp HS1 contains binding sites for the same protein complexes as gp91 phox fragment and the binding affinity is relatively high.
To directly determine if both complexes detected by probe IV contained Cux/CDP, we examined if these complexes were immunoreactive with anti-Cux/CDP antiserum (46). As shown in Fig. 4A (lanes 12-14), the upper complex was first supershifted and then abolished completely in the presence of increasing amounts of specific antiserum. In contrast, the formation of the lower complex was enhanced initially with increasing amounts of antiserum, and only at the highest antiserum concentration was this complex supershifted. The observed immunoreactivity was specific since the same amounts of control preimmune serum did not supershift or abolish the complexes (Fig. 4A,  lanes 15-17). These data were most consistent with the interpretation that the upper complex contains Cux/CDP and a different but probably related protein contributes to the formation of the lower complex. Supporting this conclusion, when probe IV was incubated with purified Cux/CDP protein, only the upper complex was generated (Fig. 4B, lanes 1-6). In contrast, no complex was detected when the purified Cux/CDP was incubated with probe II*, which has deleted 30 bp from the 3Ј end of probe II and preferentially forms the lower complex when incubated with DP thymocyte nuclear extract (Fig. 4B, lanes 7-9, and data not shown). Finally, thymocyte nuclear extract was prepared from a mutant mouse that expresses a truncated Cux lacking the strongest cut binding repeat. 2 When used in EMSA with probe IV, the mobility of the upper complex was altered and the amount of the complex was significantly reduced, whereas the formation of the lower complex was enhanced (Fig. 4B, lanes 10 -12). Together, these data suggest that Cux/CDP specifically binds to HS1 to from the upper complex.

FIG. 3. Formation of protein-DNA complexes is increased during DN to DP thymocyte differentiation.
Two ng of 32 P-labeled probe IV were incubated with 5, 10, 20, or 40 g of nuclear extract from DN or DP thymocytes. Lane 1, probe alone without extract; lanes 2-5, probe plus DN thymocyte extract; lanes 6 -9, probe plus DP thymocyte extract. The numbers specify the g of nuclear extract used. Complexes (up/low) and unbound probes (Free) are indicated.

Specific Binding of SATB1 to HS1-Previous studies have
shown that Cux/CDP binds to elements located within MARs. For example, Cux/CDP binds to a MAR in the CD8␣ promoter, in the long terminal repeat of mouse mammary tumor virus, and the MARs flanking the enhancer E where SATB1 binds to nearby sites (39,(53)(54)(55). 3 SATB1 is a major MAR-binding protein and is most abundantly expressed in thymocytes (53). Although within HS1 there were no long stretches of AT-rich sequences as found in many MARs, a stretch of 150-bp sequence was present with high ATC on one strand that may permit SATB1 binding (Fig. 10) (57). Moreover, the region resided entirely within the 170-bp probe IV, and the putative core element was contained within the 115-bp probe VII, the smallest probe shown to form the lower complex (Fig. 1).
To determine whether SATB1 binds to HS1, EMSA was carried out with probe IV and DP thymocyte nuclear extracts in the presence of anti-SATB1 antiserum. With increasing amounts of anti-SATB1 antiserum, the formation of the lower complex was progressively abolished, while the formation of the upper complex was increased initially and reduced only when a large amount of antiserum was used (Fig. 5, lanes 3-8).
In contrast, a comparable amount of preimmune serum did not abolish the lower complex formation, indicating that the result was due to specific anti-SATB1 immunoreactivity. We also conducted competition EMSA with unlabeled DNA fragments containing MAR from the CD8␣ promoter region, which has been previously shown to bind SATB1 as well as Cux/CDP (55). With the increasing amounts of unlabeled CD8␣ MAR, both the upper and the lower complexes were competed away while the 3 Wang, Z., Goldstein, A., Neufeld, E. J., Scheuermann, R. H., and Tucker, P. W. (1998) Mol. Cell. Biol., in press. Cux/CDP protein; lanes 10 -12, probe IV plus 5, 10, or 20 g of nuclear extract from mutant thymocytes.

FIG. 4. Upper complex formation results from Cux/CDP binding to the 3 end of probe IV.
A, 2 ng of 32 P-labeled probe IV were incubated with 20 g of DP thymocyte nuclear extract in the presence of unlabeled DIST or ␥-globin oligonucleotides, anti-Cux/CDP antiserum (␣Cux/CDP), or control preimmune serum. DIST oligonucleotide was derived from the cytochrome b heavy chain gene (gp91 phox ) promoter and is known to bind to Cux/CDP strongly. ␥-Globin oligonucleotide was derived from ␥-globin gene promoter and is known to bind to Cux/CDP weakly. Lane 1, probe alone without extract; lanes 2 and 11, probe plus extract; lanes 3-6, probe plus extract in presence of 25, 50, 100, and 200 ng of DIST oligonucleotide; lanes 7-10, probe plus extract in the presence of 25, 50, 100, and 200 ng of ␥-globin oligonucleotides; lanes 12-14, probe plus extract and 5 l of anti-Cux/CDP antiserum that was diluted 125-, 25-, and 5-fold, respectively; lanes 15-17, probe plus extract in presence of 5 l of preimmune serum that was diluted 125-, 25-, and 5-fold, respectively. B, 2 ng of 32 P-labeled probe IV or probe II* were incubated with different amounts of nuclear extract or purified Cux/CDP protein. Probe II* was the same as probe II, except 30 bp was deleted from the 3Ј end as in probe V (Fig. 1A). Nuclear extracts were prepared from DP thymocytes of RAG-deficient mice complemented with an activated lck transgene or thymocytes (85% DP) of mutant mice that express a truncated Cux lacking cut repeat 1. A variant Cux/CDP without CR1 was expressed in the mutant mice (see Footnote 2). Lanes 1-3, probe IV plus 5, 10, or 20 g of DP thymocyte nuclear extract; lanes 4 -6, probe IV plus 1.25, 2.5, or 5 ng of purified Cux/CDP protein; lanes 7-9, probe II* plus 1.25, 2.5, or 5 ng of purified FIG. 5. SATB1 is present in the lower complex. Two ng of 32 Plabeled probe IV were incubated with 20 g of nuclear extract from DP thymocytes in the presence of either anti-SATB1 antiserum (␣SATB1) or preimmune serum. Lane 1, probe alone; lane 2, probe plus nuclear extract; lanes 2-8, probe plus extract in presence of 5 l of anti-SATB1 antiserum that was diluted 320-, 160-, 80-, 40-, 20-, and 10-fold, respectively; lanes 9 and 10, probe plus extract in the presence of 5 l of preimmune serum that was diluted 20-and 10-fold, respectively. nonspecific pUC18 DNA did not have any effect (Fig. 6). Together, these data indicate that SATB1 binds specifically to HS1 and contributes to the formation of the lower complex.
HS1 Is a MAR-Given the above findings, experiments were conducted to determine whether the 170-bp probe IV exhibits specific binding to nuclear matrix, a characteristic inherent to MARs (20,22). Nuclear matrices were prepared and used in a MAR binding assay with radiolabeled probe IV of HS1 (39,55). As a positive control, a 200-bp PstI-EcoRI fragment containing the CD8␣ MAR was used and the linearized pUC18 DNA was included in the assays to measure nonspecific binding. Fig. 7 shows that in the absence of E. coli competitor DNA, significant levels of pUC18, probe IV, and the CD8␣ MAR were associated with the nuclear matrix (lanes 2 and 6). In the presence of 0.2 or 0.4 g/ml E. coli competitor DNA, pUC18 binding was abolished, consistent with its nonspecific binding to nuclear matrix. Under the same condition, the binding of probe IV and the CD8␣ MAR persisted (Fig. 7, lanes 3, 4, 7, and 8), indicating their specific binding to the nuclear matrix. Together with the finding that CD8␣ MAR competed away the complex formation by probe IV and nuclear extract (Fig. 6), these data show that the 170-bp HS1 is a MAR.
HS1 Represses E ␤ -dependent Transcription in Reporter Gene Assays-Cux/CDP is known as a transcriptional repressor. Many studies have shown that the binding of Cux/CDP and/or SATB1 to a MAR can alter the transcriptional activity of a nearby transcriptional regulatory element (50, 54, 55, 58 -62). 3 Considering that HS1 is just 400 bp upstream of the E ␤ enhancer, we tested whether HS1 modulates transcriptional activity of the E ␤ by transient transfection assays using reporter gene constructs. Construct 1 was a basal vector in which the luciferase gene was placed behind a 424-bp promoter derived from the V ␤ 13 gene segment. Construct 2 had an additional 830-bp BglII-NcoI fragment containing the E ␤ inserted downstream of the luciferase gene. Construct 3 had a 1.0-kb BsgI-NcoI fragment containing HS1 and E ␤ in their natural configuration inserted downstream of the luciferase gene. Transient transfection assays of construct 1 gave only low levels of luciferase activity, indicating a strict requirement for E ␤ in transcriptional activation. Construct 2 consistently resulted in much higher levels of luciferase activity, ranging from 80-to 160-fold increases over the construct 1 in four thymoma lines tested (Fig. 8B). In contrast, construct 3 yielded luciferase activities that were approximately 2.5-fold lower than those of construct 2 in all four thymoma lines (Fig. 8B). Thus, in this assay system, HS1 and the associated trans factors appear to be able to interact with E ␤ enhancer and repress its transcriptional activity.
Mutational Analysis of HS1 in TCR␤ Transcription at the Endogenous Locus-Simultaneous to our present studies, we have generated mutant mice in which the 780-bp BstXI-BglII fragment containing HS1 was deleted from the TCR␤ locus (37). To examine the effect of the deletion on TCR␤ gene transcription, we isolated RNA from thymocytes of wild type and homozygous mutant mice and assayed for the levels of transcripts initiated from different regions of the TCR␤ locus by Northern blotting. As shown in Fig. 9, no apparent difference was detected in the level of mature TCR␤ transcripts by a C ␤ 2 probe, a V ␤ 8 probe, or a V ␤ 14 probe (data not shown). Similarly, no difference was detected in the level of germline transcripts by a D ␤ 1-J ␤ 1 intron probe. Analyses of methylation status in the promoter region of the V ␤ 14 gene segment revealed similar levels of demethylation between wild type and mutant mice (data not shown), correlating to the similar levels of V ␤ 14 transcription. These data suggest that HS1 is not required for transcription and demethylation at the endogenous TCR␤ locus under physiological conditions.  7. The 170-base pair HS1 DNA fragment (probe IV) specifically binds to the nuclear matrix. Nuclear matrix binding assays were performed with 32 P-labeled probe IV plus pUC18 or CD8␣ MAR plus pUC18 in the presence or the absence of unlabeled E. coli DNA. Two ng of probe IV or CD8␣ MAR and 20 ng of pUC18 were used in a 100-l reaction. Twenty or 40 g of unlabeled E. coli DNA was added per reaction as nonspecific competitor. DNA bound to nuclear matrix was extracted with phenol-chloroform, precipitated, and run on a 4.5% polyacrylamide gel. Lanes 1-4, control binding assays with CD8␣ MAR as specific probe and pUC18 as nonspecific probe; lanes 5-8, MAR binding assays with probe IV as specific probe and pUC18 as nonspecific probe.

DISCUSSION
Cux/CDP and SATB1 Bind to HS1-Using thymocyte nuclear extracts and EMSA, we have localized two nuclear factor binding sites, one for Cux/CDP and the other for SATB1, to within a 170-bp region of HS1. Of the two nucleoprotein complexes detected, the upper complex is most likely due to Cux/ CDP binding to the 3Ј end of the 170-bp HS1. First, deletion of 30 bp from the 3Ј end of the probe IV almost completely abolished the formation of the upper complex without significantly reducing the lower complex formation (Fig. 1). Second, the formation of the upper complex was preferentially abolished by the presence of anti-Cux/CDP antiserum in EMSA (Fig. 4). Finally, only the upper complex was detected in EMSA using the purified Cux/CDP protein and probe IV (Fig. 4). Cux/CDP is a homeodomain protein containing a homeodomain and three cut repeats, all of which can independently bind DNA (43)(44)(45)(46). Based on binding of individual cut repeat polypeptides to oligonucleotides in vitro, cut repeats 2 and 3 share similar consensus binding sequences while cut repeat 1 displays a more restricted pattern of DNA sequence recognition (Fig. 10) (47,48). Sequence comparison revealed a stretch of sequence at the 3Ј end of the 170-bp HS1 that shares a high homology (83%) to the cut repeats 2 and 3 consensus sequence (Fig. 10B). In addition, an overlapping stretch of sequence shares a modest homology (50%) to cut repeat 1 consensus binding site. Thus, Cux/CDP may interact with cis DNA elements in HS1 through its multiple DNA-binding domains. Supporting this notion, an altered upper complex was observed when probe IV was incubated with thymocyte nuclear extract from mutant mice that express a truncated Cux lacking cut repeat 1 (Fig. 4B). 2 SATB1 contributes to the formation of the lower complex, as revealed by its preferential abolishment by anti-SATB1 antiserum. The core SATB1 binding site apparently does not overlap substantially with the Cux/CDP binding site at the 3Ј end of the 170-bp probe IV because deletion of 30 bp from the 3Ј end abolished the upper complex but not the lower complex (Fig.  1B). SATB1 binding in HS1 is probably broad because deletion of 54 bp from the 5Ј end did not abolish but substantially reduced the formation of the lower complex (Fig. 1B). SATB1 is known to lack strict sequence specificity and prefers a stretch of ATC-rich sequence confined to one strand of DNA (21,57,63). Sequence analysis revealed that the leading strand (5Ј to 3Ј) of probe IV is highly ATC-rich (156 of 174 nucleotides) (Fig.  10A). In particular, there is a stretch of 37 ATC nucleotides that probably forms the core unwinding element for SATB1 binding. The characteristic feature of ATC-rich on one strand in this region is highly conserved between mouse and human (Fig. 10A). Interestingly, as the very 5Ј end of the 170-bp HS1 is not required for the lower complex formation (Fig. 1), there is more sequence divergence in this region between mouse and human. In the rest of the sequence, mouse and human share 89% identity.
As discussed above, binding of HS1 by Cux/CDP and SATB1 appears to be relatively independent; however, their binding sites may overlap. Supporting this notion, inhibition of SATB1 binding to HS1 by anti-SATB1 antiserum actually caused an increase in the amount of the Cux/CDP-containing upper complex in EMSA (Fig. 5). Conversely, inhibition of Cux/CDP binding to HS1 by anti-Cux/CDP antiserum promoted the formation of the SATB1-containing lower complex (Fig. 4A). Moreover, when truncated Cux/CDP that lacks the cut repeat 1 was used in EMSA the formation of Cux/CDP-containing upper complex was reduced, whereas the formation of the SATB1-containing lower complex was markedly enhanced (Fig. 4B). Thus, Cux/ CDP and SATB1 appear to interfere with each other in binding to HS1. In addition to its originally defined MAR-binding domain, SATB1 was recently found to contain a separate homeodomain and two cut-like repeats that share homology to those in Cux/CDP (57,63). These structural homologies between the FIG. 8. HS1 represses E ␤ -dependent reporter gene expression in transient transfection assays. A, schematic diagrams of constructs 1-3 (not drawn to scale). Construct #1, luciferase gene driven by a 424-bp V ␤ 13 promoter; construct #2, the same as construct 1 but with a 830-bp BglII-NcoI fragment containing E ␤ inserted downstream of the luciferase gene; construct #3, the same as construct 1 but with a 1-kb BsgI-NcoI fragment containing both HS1 and E ␤ inserted downstream of the luciferase gene (see "Experimental Procedures"). B, relative levels of luciferase activities by different constructs in different thymoma lines. Thymoma P4890 and P4833 were derived from DN thymocytes, thymoma 4b from a DP thymocyte, and thymoma EL4 from a CD4 ϩ thymocyte (41). Cells (2 ϫ 10 7 ) were electroporated with 10 g of construct DNA plus 10 g of control CMV-␤-galactosidase plasmid. Forty-eight h after transfection, luciferase and ␤-galactosidase activities in cell lysates were assayed. Luciferase activity of individual transfection was normalized to ␤-galactosidase activity of the same sample. Luciferase activities of constructs 2 and 3 are expressed as -fold of increase over construct 1. When the reporter constructs were assayed in a B cell lymphoma line M12, luciferase activity remained at the background levels, consistent with the lineage specificity of the V ␤ promoter and E ␤ enhancer (data not shown).
FIG. 9. Northern blotting analyses of transcripts initiated from different regions of the TCR␤ locus in thymocytes of HS1 wild type and homozygous mutant mice. Mice homozygous for the deletion of a 780-bp region containing HS1 at the TCR␤ locus were generated by targeted mutagenesis (37). Total RNA was isolated from thymocytes of wild type and homozygous mutant mice, and the levels of TCR␤ transcripts were assayed by Northern blotting. For each sample, 20 g of total RNA was fractionated on a 1.2% formaldehyde gel, transferred to Zeta-probe filters, and hybridized with a C ␤ 2 probe, a V ␤ 8 probe, or a D ␤ 1-J ␤ 1 intronic probe. Ethidium bromide staining of 28 S RNA was shown for loading controls. two proteins may provide an explanation for competitive binding of overlapping sites in HS1 and the minor cross-reactivities of antisera.
HS1 Is a MAR-Cux/CDP and SATB1 are well-characterized MAR-binding proteins. Their binding to the 170-bp HS1 suggests that HS1 is a MAR. Supporting this conclusion, the binding of Cux/CDP and SATB1 to HS1 was competed away by a MAR isolated from an upstream regulatory element of the CD8␣ gene (Fig. 6). Furthermore, the 170-bp HS1 itself bound specifically to the nuclear matrix (Fig. 7). Thus, HS1 is a MAR and will be referred to as MAR ␤ . Compared with the MARs from the IgH and Ig loci, MAR ␤ shares with them interesting similarities as well as differences. In the Ig locus, a highly conserved MAR is located 200 bp upstream of the intronic iE enhancer (22,32). In the IgH locus, the intronic E enhancer is flanked on both sides by intimately associated MARs (24). In B cells, they are associated with MAR-BP1 and Bright (39,52), whereas in non-B cells they are bound by SATB1 and NF-NR which is likely Cux/CDP (28,53). 3 Similarly, MAR ␤ is just 400 bp upstream of E ␤ and is also associated with Cux/CDP and SATB1. However, MAR ␤ does not possess the large stretches of ATs present in the Ig MARs (18,32). Although AT-rich is a common feature of MARs (21), recent studies suggested that a stretch of ATC-rich sequence confined to one strand may be a more characteristic feature of MARs and MAR ␤ contains this characteristic sequence (21,53,57). These findings suggest that an analogous organization of cis-regulatory elements in which a MAR is in close proximity to an enhancer is conserved in the IgH, Ig, and TCR␤ loci and probably in other antigen receptor gene loci. The conserved organization of MAR and enhancer and the association of the same nuclear factors with the MARs may provide a structural basis for the similar functions of these regulatory elements in the control of rearrangement and/or transcription of the respective antigen receptor genes.
Potential Function of MAR ␤ -MAR ␤ appeared to be a likely candidate that may function as an additional cis-regulatory element, besides V ␤ promoters and E ␤ , to exert control on TCR␤ gene rearrangement and transcription. It is a major DNase I-hypersensitive site induced during DN to DP thymocyte differentiation, and its induction coincides with the cessation of TCR␤ gene rearrangement in DP thymocytes (allelic exclusion) (37). We demonstrated that MAR ␤ repressed E ␤ -dependent reporter gene expression in transient transfection assays (Fig. 8), suggesting that nuclear factors that bind to MAR ␤ modulate in cis the enhancing influence of E ␤ on transcription in T cell lines. However, targeted deletion of a 780-bp region containing MAR ␤ did not affect TCR␤ transcription at the endogenous locus in developing T cells. Moreover, in the absence of MAR ␤ , demethylation in the promoter region of the V ␤ 14 gene segment and recombination of the TCR␤ gene also occurred to the same extent as on the normal allele (Ref. 37 and data not shown). Thus, our findings demonstrate that MAR ␤ is not absolutely required for these processes at the TCR␤ locus under physiological conditions.
The contrasting results obtained from transcriptional analyses of MAR ␤ in reporter gene assays and at the endogenous locus could have different underlying mechanisms. One possibility is that MAR ␤ does modulate the contribution of E ␤ to transcriptional activity in the endogenous locus; however, its function is compensated for by additional MARs within the locus in the mutant mice. It is also possible that the 780-bp region deleted in the mutant mice may contain an as yet unidentified cis-regulatory element that has an opposite function as MAR ␤ . Deletion of both simultaneously might result in functional neutralization and no apparent phenotype. Alternatively, MAR ␤ may have no role in the regulation of TCR␤ gene transcription. Its effect on E ␤ -dependent transcription in transient reporter gene assays may be an artifact resulting from taking MAR ␤ out of its natural chromosomal context. In this regard, it is worthwhile to consider the putative functions of MARs at the IgH and Ig loci. By reporter gene assays in cell lines or transgenic mice, MARs flanking the enhancer E were shown to modulate IgH gene chromatin accessibility (31) and transcription in both positive and negative manners (27)(28)(29)(30). Similarly, MAR associated with the Ig intronic enhancer is thought to be involved in Ig gene transcription, demethylation, recombination, and somatic hypermutation (18,(33)(34)(35)(36). However, similar to targeted deletion of MAR ␤ , targeted deletion of MAR from Ig locus did not have any obvious effect on Ig gene recombination and transcription. 4 Although MARs are widely found in diverse eukaryotic organisms and have been implicated in the regulation of transcription, DNA recombination, replication, and repair, their true function in most of these processes is still unclear (56). The uncertainty arises because putative functions of MARs are usually studied using reporter gene assays. Targeted mutagenesis of MAR ␤ represents a first attempt to elucidate a MAR's function in its endogenous locus. The different results we have obtained from assaying MAR ␤ in reporter gene assays and at 4 Y. Xu, personal communication.
FIG. 10. Sequence comparison of HS1 between mouse and human and identification of putative Cux/CDP and SATB1 binding sites. A, the 170-bp DNA region in HS1 (probe IV) from mouse and human are aligned. Both mouse and human sequences are ATC-rich on the top (leading) strand. The stretch of 37 ATC (overlined) could potentially serve as the core unpairing region for SATB1 binding. Cux/CDP binding sites may reside in the second overlined region (dotted line). B, putative Cux/CDP binding sequences from MAR ␤ are compared with cut repeat 1 (CR1) and cut repeat 2/cut repeat 3 (CR2/CR3) consensus binding sequences. the endogenous locus clearly point out the potential pitfalls of analyzing the function of MARs outside of their natural chromosomal context.