Interaction of the Nuclear Matrix-associated Region (MAR)-Binding Proteins, SATB1 and CDP/Cux, with a MAR Element (L2a) in an Upstream Regulatory Region of the Mouse CD8a Gene*

Matrix-associated regions (MARs), AT-rich DNA segments that have an affinity for the nuclear matrix, have been shown to play a role in transcriptional regulation of eukaryotic genes. The present study demonstrates that a DNA element, called L2a, which has been implicated in the transcriptional regulation of the mouseCD8a gene encoding an important T cell coreceptor, is a MAR. Moreover, the identities of two nuclear proteins, L2a-P1 and L2a-P2, previously shown to bind to the L2a element, have been determined. The L2a-P1 protein found to be present in all CD8-positive T cell lines tested is SATB1, a known MAR-binding protein. The widely expressed L2a-P2 protein is CDP/Cux, a MAR-binding protein that has been associated with repression of gene transcription. Interaction of both proteins with the L2a element was studied using the missing nucleoside approach, DNase I footprinting, and electrophoretic mobility shift assays with wild type and mutant L2a elements. The data suggest that CDP/Cux bound to the L2a element is displaced by binding of SATB1 and the accompanying conformational change in the DNA lying between the primary binding sites of SATB1 and CDP/Cux. We suggest that displacement of CDP/Cux by SATB1 favors transcription of theCD8a gene, possibly by enhancing or altering its association with the nuclear matrix.

Among the DNA elements implicated in the regulation of gene transcription are matrix-associated regions (MARs), 1 which are defined as AT-rich DNA sequences that are preferentially retained by the nuclear matrix (1,2). The nuclear matrix is a proteinaceous subfraction that remains after extraction of nuclei with high salt concentrations, and it is thought to form a scaffold for chromosome attachment (3,4). MARs are typically 200 -300 bases in length, contain topoisomerase II cleavage sites, and occur on the average of one for every 30 kb of eukaryotic DNA (3,(5)(6)(7)(8). MARs have been increasingly observed near enhancer and promoter regions of genes (2, 8 -10), and the participation of MARs in regulating transcription of the immunoglobulin light chain (1) and heavy chain (11)(12)(13) genes has been described. Because MARs can often be shown to bind to nuclear matrices of a wide variety of cells, their ability to regulate transcription of specific genes might be questioned. However, cell type-specific MAR-binding proteins (12, 14 -18) could modulate the association of specific MAR sequences and their adjacent genes with the nuclear matrix and thereby repress or activate those genes.
The present study demonstrates that a DNA element (L2a; Ref. 19) implicated in the regulation of transcription of the mouse CD8a gene is a MAR. The CD8 molecule on the surface of T lymphocytes functions as an important coreceptor for class I major histocompatability/peptide complexes on stimulator and target cells (20 -22). In the mouse, CD8 is expressed on class I major histocompatability complex-specific T cells as a heterodimer of CD8␣ and CD8␤ chains that are encoded by the CD8a (or Lyt-2) and CD8b (or Lyt-3) genes, respectively (23,24), but as CD8␣␣ homodimers on some intraepithelial T lymphocytes (25). Little is known concerning the regulation of the mouse CD8a and CD8b genes, but some insights have come from studies of T-T cell hybrids produced by fusing cytotoxic T lymphocytes with the BW5147 thymic lymphoma. Such hybridomas are invariably CD8-negative (26) due to shut-off of CD8a gene transcription (27), but interestingly the CD8b gene, which is located 36 kb upstream of the CD8a gene (28) (see Fig. 1), continues to be transcribed (27,29). Working on the hypothesis that the shut-off of CD8a gene transcription in T-T cell hybridomas was due to cis-acting DNA elements upstream of the CD8a gene, stable clonal transfectants of BW5147 were produced using constructs of the CD8a gene containing varying amounts of 5Ј-flanking DNA and surface expression was monitored by flow microfluorometry. Initial results suggested that DNA sequences located between nucleotides Ϫ1,400 and Ϫ4,700 (the transcription start site is nucleotide ϩ1) (see Fig.  1) were sufficient to provide the negative effect (19). This same region is part of a DNase I-hypersensitive site specific to CD8positive T cells and contains two binding sites for the GATA-3 transcription factor (30).
Attention focused on a AccI/HpaI subfragment of this region (nucleotides Ϫ3800 to Ϫ4700) that lay upstream of the GATA-3 binding sites and that negatively affected CD8a gene expression (Fig. 1). A 270-nucleotide AccI/SstI subfragment of the AccI/HpaI fragment contained an element, designated L2a, which was protected in two places (called L and S) from DNase I digestion in vitro by nuclear extract of the CD8 ϩ T cell tumor line, VL3. A fragment (called 200(LϩS)) containing both pro-tected regions could be shown in electrophoretic mobility shift assays (EMSA) to bind nuclear proteins from thymus as well as from a variety of cell lines. Varying amounts of two prominent retarded bands (bands 1 and 2) were observed in EMSA using the 200(LϩS) fragment as probe and nuclear extracts from different cell lines and tissues (19). Band 1 appeared to be T cell-specific, and it reflects binding of a protein (designated L2a-P1) with an estimated molecular mass of 110 -140 kDa, which apart from its presence in two CD4 Ϫ CD8 Ϫ T cell lines correlated with CD8␣ expression. Band 2 was observed using nuclear extracts of all cell lines tested, and it was shown to reflect binding of a protein (designated L2a-P2) with an estimated molecular mass of Ͼ200 kDa. In preliminary competition experiments, the L2a-P1 and L2a-P2 proteins competed for binding to the 200(LϩS) fragment, and the L2a-P1 protein appeared to have the higher affinity (19).
A model was proposed in which proteins L2a-P1 and L2a-P2 play positive and negative roles, respectively, in regulating CD8a gene transcription by binding to the L2a element (19). The present results demonstrate that the L2a element is a MAR, that protein L2a-P1 is SATB1, a known MAR-binding protein (15), and that protein L2a-P2 is CDP/Cux, a second MAR-binding protein shown in several instances to have negative effects on gene expression (31)(32)(33). The results further demonstrate that the SATB1 and CDP/Cux proteins bind to overlapping sites on the L2a element. We suggest that binding of SATB1 to the L2a element displaces CDP/Cux and favors transcription of the CD8a gene, possibly by enhancing or altering its association with the nuclear matrix.
Probes and DNA Constructs-The 0.9-kb AccI/HpaI fragment located 3.8 -4.7 kb upstream of the mouse CD8a gene was treated with Klenow fragment and cloned into the SmaI site of pUC18 (p2211-1). The same fragment containing a deletion of 50 base pairs at its 5Ј end generated by Bal 31 digestion is present in plasmid p2211-1-5Ј-43. The various fragments and subfragments used in EMSAs were generated by digestion of these two plasmids with enzymes that cut within the polylinker site of the plasmid and/or within the fragment.
The 200(LϩS) probe used in gel retardation assays consists of a 200-nucleotide genomic AccI/BstXI fragment that contains the regulatory element described by Lee and co-workers (19). This fragment was subcloned into pUC18 and was excised using restriction sites in the polylinker. The 158-base pair fragment used for the DNase I footprinting of the bottom strand is included in the 200(LϩS) probe and was generated by digesting p2211-1-5Ј-43 with BamHI and BstXI followed by labeling with [␣-32 P]dNTPs at the BamHI end using Sequenase (Amersham Corp.). Similarly, the 150(LϩS) fragment used for the missing nucleoside and DNase I footprinting experiments was obtained by digesting p2211-1-5Ј-43 with HindIII and BstXI and subsequent end labeling at the HindIII end with Sequenase. The 100(LϩS) fragment was generated by subcloning a HincII/RsaI fragment from p2211-1-5Ј-43 into the SmaI site of pUC18, digesting the resulting plasmid with EcoRI and PstI, and end labeling at the EcoRI end with Sequenase.
Preparation of Nuclear Extracts-Nuclear extracts of cell lines were prepared essentially as described by Dignam et al. (36). All steps were performed at 4°C or on ice. Cells (2-5 ϫ 10 8 ) were washed once with ice-cold phosphate-buffered saline and resuspended in 5 ml of buffer A (10 mM HEPES, pH 7.9, 1.5 mM MgCl 2 , 10 mM KCl, 1 mM dithiothreitol (DTT), and 1 mM phenylmethylsulfonyl fluoride (PMSF)). DTT and PMSF were added to all buffers immediately before use. After incubation for 10 min, the cells were pelleted (1000 ϫ g for 10 min) and resuspended in 2 ml of buffer A. The cells were then lysed using 10 strokes in a homogenizer (Kontes homogenizer with a B pestle; Kontes, Vineland, NJ), and nuclei were pelleted (1000 ϫ g for 10 min), washed with 6 ml of buffer A, and pelleted again (30,000 ϫ g for 20 min). The nuclei were resuspended in 1-2 ml of buffer C (20 mM HEPES, pH 7.9, 25% glycerol, 0.42 M NaCl, 1.5 mM MgCl 2 , 0.2 mM EDTA, 1 mM DTT, 1 mM PMSF, 2.5 g/ml pepstatin A, 5 g/ml leupeptin, 12 g/ml aprotinin, and 12 g/ml soybean trypsin inhibitor), homogenized ( ϳ 30 strokes), and stirred on a magnetic stirrer for 30 min. Debris was pelleted (30,000 ϫ g for 20 min), and the supernatant was dialyzed against 75 volumes of buffer D (20 mM HEPES, pH 7.9, 20% glycerol, 0.1 M KCl, 0.2 mM EDTA, 1 mM DTT, and 0.7 mM PMSF) for 3 h. After dialysis, extracts were centrifuged (30,000 ϫ g, 20 min), and the supernatant was stored as aliquots in liquid nitrogen. The protein concentration in each extract was measured using the Bradford assay (37).
Antisera-Rabbit polyclonal anti-SATB1 antiserum was kindly provided by Dr. T. Kohwi-Shigematsu (Burnham Institute, University of California, Berkeley, CA) and also produced in our laboratory as described (15). Rabbit polyclonal anti-CDP/Cux was the kind gift of Dr. Ellis Neufeld (Harvard Medical School, Boston, MA).
For competition experiments, specific and nonspecific competitor DNAs (0.002-0.8 pmol) were added to the nuclear extracts before addition of radiolabeled probe. For distamycin A studies, distamycin A (Sigma) was added to nuclear extracts to the desired concentration before the addition of radiolabeled probe. For antibody inhibition ex-  (19) and represents a correction of the published sequence.
periments, 2 l of various dilutions of preimmune rabbit serum, anti-SATB1 antiserum or anti-CDP/Cux antiserum were added to the nuclear extracts before the addition of the radiolabeled probe.
Purification of an L2a-P1-like Protein from Calf Thymus-Calf thymus nuclear extract (600 mg) was prepared from fresh calf thymus using a modification of the protocols of Wang and Speck (40) and Dignam et al. (36) that employed additional protease inhibitors described by Thornell et al. (41). After ammonium sulfate precipitation (50% saturation), the extract was fractionated by four successive columns as described below. Protein was quantitated using the assay of Bradford (37), and individual fractions were assayed for L2a-P1-like activity by EMSA using a portion of the L2a element (the 100(L) fragment; Ref. 19) as the radiolabeled probe: 1) DEAE-cellulose chromatography (Sigma) (100-ml bed volume) run at pH 7.9 and 4°C and eluted with a linear gradient of from 0.1 M to 0.5 M KCl (200 ml each) containing 20 mM HEPES, 20% glycerol, and 2 mM EDTA; 2) heparin-Sepharose CL-6B chromatography (Pharmacia Biotech Inc., 17-046701) (40-ml bed volume) run under the same conditions with the same buffer and eluted with a linear gradient of 0.1 M KCl to 1.0 M KCl (100 ml each); 3) DNA-cellulose chromatography (Sigma, D8515) (20-ml bed volume) run as above and eluted with a gradient of from 0.1 M KCl to 1.0 M KCl (50 ml each); and 4) affinity chromatography on a column containing 200 mg of multimerized 200(LϩS) fragment containing the L2a element, which had been conjugated to Sepharose 4B. Trimers of the 200(LϩS) fragment were ligated to yield a mixture of fragments in the range of 20 -50 kb and conjugated to 2 g of CNBr-activated Sepharose essentially as described previously (42). The washed column (8-ml bed volume) was loaded with 20 ml of the above buffer containing 0.1 M KCl, 0.5 mg of protein, and 100 mg of poly(dI-dC)⅐poly(dI-dC). 20 ml of buffer containing 100 mg of poly(dI-dC)⅐poly(dI-dC) and 0.1 M KCl was then passed through the column at 30 ml/h, and 1 ml fractions were collected. 2-ml volumes of the same buffer containing increments of 0.1 M in KCl concentration (from 0.2 M to 0.9 M) were then passed through, followed by 6 ml of buffer containing 1.0 M KCl. Fractions 52-55 eluted from the L2a affinity column at 0.5-0.7 M KCl had significant probe binding activity and contained major components with molecular masses of approximately 100 kDa upon SDS-PAGE and silver staining. The peak fraction (fraction 53) was used for EMSA analysis and Western blotting (see "Results"), and fraction 54 was subjected to microsequence analysis as described below.
Microsequence Analysis of Fraction 54 -Following SDS-PAGE and detection of the stained bands with Coomassie Blue, fraction 54 was subjected to in-gel digestion, and peptides were prepared for HPLC separation essentially as described previously (43). The peptides were separated by reverse phase HPLC (Hewlett-Packard 1090) using a Vydac C18 column (2.1 ϫ 250 mm, 5 micron, 300A). Peptides were eluted with an increasing gradient of acetonitrile/2-propanol (3:1 mix) and absorbance was monitored at 214, 295, and 550 nm. Amino acids sequence analysis was performed on a Perkin-Elmer ABI model 477 sequencer with on-line HPLC analysis of phenylthiohydantoin derivatives.
For radiolabeling with [ 32 P]orthophosphate, cells (5 ϫ 10 6 /ml) were cultured for 4 h in phosphate-free medium (Sigma). [ 32 P]Orthophosphate (ICN Pharmaceuticals) was added to a concentration of 1 mCi/ml, and the cells were incubated for 2 h. After washing in phosphate-free medium, the cells were lysed in RIPA buffer containing the phosphatase inhibitors orthovanadate (0.1 mM) and ␤-glycerolphosphate (50 mM) and immunoprecipitated with preimmune or anti-SATB1 antiserum as described above.
Missing Nucleoside Interference Assay-The missing nucleoside experiment was performed essentially as described by Hayes and Tullius (44). Each DNA fragment was radiolabeled at one end of one strand only as indicated in the figure legends and subjected to hydroxyradical treatment by adding 10 l of 0.2 mM (NH 4 ) 2 FeSO 4 /0.4 mM EDTA/3% H 2 O 2 /10 mM L-ascorbate, to 70 l (300 ng) of labeled fragment. The reaction mixture was incubated at 37°C for 30 min. The reaction was quenched with 15 l of 0.1 M thiourea/32 l of 0.2 M EDTA. The reaction mixture was precipitated with 10 g of yeast tRNA, 20 l of 3 M sodium acetate, and 750 l of ethanol, redissolved in water, and precipited a second time with ethanol. The pellet was dissolved in 120 l of 50 mM Tris, pH 7.8. Multiple binding reactions were done as described above using 5 l of probe and 10 -15 g of nuclear extract and electrophoresed in 10 lanes of a 4% polyacrylamide gel at 180 V for the appropriate amount of time. The gel was exposed wet for 2-3 h, and the retarded bands and free probe were excised from nine lanes essentially as described previously (45). One lane was kept as reference. The probes were eluted from the gel by the crush and soak method using 0.2 M NaCl-TE as described by Dickinson et al. (15). The samples were electrophoresed on an 8% sequencing gel together with probes subjected to hydroxyradical treatment (control lane) or Maxam-Gilbert sequencing reactions (45) at 55 W for the appropriate amount of time. The gel was fixed in 10% methanol/acetic acid, dried, and exposed for 1-2 days to x-ray film at Ϫ70°C.
DNase I Footprinting-DNase I footprinting of the isolated bands was performed essentially as described by Landolfi et al. (46). Ten EMSA reactions were performed as described above using approximately 5 ng of labeled probe and 6 -7 g of nuclear extract per tube. Also added to each reaction mix was 2-3 l of 100 mM Mg 2ϩ /Ca 2ϩ for the subsequent DNase I digestion. The reactions were incubated for 20 min, 1-2 l of a 200 ng/ml DNase I solution was added to each tube, and the reaction mix was incubated for an additional 5 min. The DNase I digestion was subsequently quenched with 1 l of 100 mM EDTA. The reactions were electrophoresed on a 4% PAGE at 180 V for the appropriate amount of time and exposed wet for 2-3 h, and the retarded bands and free probe were excised from nine lanes with one lane being kept as a reference. The samples were eluted by the crush and soak method and electrophoresed on an 8% sequencing gel as described above for the missing nucleoside reaction.
Site-directed Mutagenesis-Site-directed mutagenesis was performed using the pALTER system (Promega, Madison, WI) essentially as described by the manufacturer. The SalI site in the polycloning site of pALTER-1 was changed to HpaI by site-directed mutagenesis. A 1.2-kb XbaI/HpaI fragment containing the L2a element was cloned into the XbaI and HpaI sites of pALTER-1. Site-directed mutagenesis was then performed, and AccI/BstXI fragments representing mutant 200(LϩS) regions were subcloned into pUC18 and subjected to automated nucleotide sequence analysis to verify their sequences. In some instances, further deletions were introduced into these subclones using StyI sites present in the genomic DNA and/or a BclI site introduced by site-directed mutagenesis. Fragments for use as probes in EMSA studies were excised using restriction sites in the polylinker.
RNA Isolation and Northern Hybridizaton-RNA isolation using guanidine isothiocyanate and Northern hybridizations were performed essentially as described previously (45).
Production of a SATB1-positive Transfectant of BW5147-An EcoRV/ XbaI fragment containing the full-length mouse SATB1 cDNA was excised from the pmAT plasmid (47), made blunt ended using Klenow fragment, and cloned into the SmaI site of the pECH expression vector to produce the pECH-mSATB1 construct. Both the pmAT and pECH vectors were the kind gifts of Dr. T. Kohwi-Shigematsu. The pECH vector contains the major PvuII fragment from pBluescript and the SV40 sequences and multicloning site from the pECE plasmid (48).
Electroporation was performed essentially as described previously (49). The hypoxanthine/aminopterin/thymidine medium-sensitive ␣ Ϫ ␤ Ϫ BW5147 cell line (35) (1 ϫ 10 6 cells/ml) was plated in DMEM containing 10% FCS and 0.1 mg/ml colcemid 12-16 h before electroporation. Cells were then pelleted and resuspended (5 ϫ 10 6 cells/0.5 ml) in 12 ϫ 75 mm polystyrene tubes. To each tube was then added linearized pECH-mSATB1 (40 g) and linearized pSV2neo (1 g; Ref. 50). After incubation for 10 min on ice, each mixture was transferred to a cuvette (Evergreen Scientific, Los Angeles, CA) containing electrodes 1 cm apart, and 1200 V were applied using a Zapper electroporation unit (University of Wisconsin Medical Electronic Lab, Madison, WI) with capacitance and fall time set at maximum and rise time set at minimum. After incubation on ice for 10 min, cells were transferred to a 100 ϫ 20 mm tissue culture dish containing 10 ml of DMEM/10% FCS and incubated for 48 h (37°C, 5% CO 2 in air). Cells were then plated in 24-well plates (2.5 ϫ 10 4 cells/well) in 1 ml of DMEM/10% FCS containing G418 (600 g/ml of active drug). Every 3-4 days an additional 0.25 ml of G418 containing medium was added. Clones expressing SATB1 mRNA were identified by Northern hybridization using a SATB1-specific probe.
Nuclear Matrix Isolation and MAR Binding Assay-Nuclei of the EL-4 T cell line were obtained by hypotonic lysis and purified by centrifugation through a cushion of 2 M sucrose, and nuclear matrices were isolated as described previously (1,8). Nuclear matrix binding was determined in an in vitro DNA-binding assay (8). Briefly, plasmids containing the wild type or mutant L2a elements were linearized with NdeI and end-labeled with [␥-32 P]ATP. 20 -50 ng of each plasmid were incubated with nuclear matrices (2 h, room temperature) in the presence of 37-40 g of unlabeled E. coli carrier DNA. Insoluble matrix proteins were pelleted and washed extensively to remove unbound DNA, treated with proteinase K, and extracted with phenol-chloroform, and the bound DNA was precipitated and electrophoresed on 0.7% agarose gels. Gels were dried on nylon membranes for autoradiography at Ϫ70°C. A 574-nucleotide BamHI fragment, BrMAR, that had previously been shown to be a strong MAR (8) was used as an internal control in every reaction, whereas the pUC19 vector DNA from which this fragment was excised served as an internal negative control. 25% of the total radioactivity added to each MAR binding assay (the input) was electrophoresed beside the bound fragments to show the relative intensities and sizes of the input fragments. Quantitation of the binding of the L2a constructs and the BrMAR was performed with a Molecular Dynamics PhosphorImager 425 and ImageQuant software (Sunnyvale, CA). The ratio of binding of wild type L2a element to BrMAR was taken as 1.0, and the relative binding of mutant L2a elements was calculated on this basis.

An L2a-binding Protein Isolated from Calf Thymus Shows
Homology to SATB1-EMSA analysis performed with radiolabeled L2a probe and nuclear extracts from mouse (19) or calf thymus (see below) yields a retarded band with the same mobility as band 1 shown by Lee et al. (19) to reflect binding of protein L2a-P1. A protein with properties similar to L2a-P1 was purified from calf thymus nuclear extract as described under "Experimental Procedures." The final purification step involved affinity purification on a portion of the 200(LϩS) fragment (called 100(L)), which binds L2a-P1 and gives rise to retarded band 1 in EMSA assays with VL3 nuclear extract (19). Three partially resolved bands of similar mobility (100 -110 kDa) were observed upon Coomassie Blue staining of the affinity-purified fractions, consistent with the estimated molecular mass of mouse L2a-P1 (19). EMSA analysis using the 100(L) probe and purified protein that had been subjected to SDS-PAGE and renatured from guanidine HCl gave rise to a specific retarded complex ( Fig. 2A). These results suggested that the isolated protein could be the calf homologue of L2a-P1.
Though only a very small amount of purified protein was obtained, the partial amino acid sequence of a peptide obtained from two incompletely resolved bands of fraction 54 showed significant homology in a data base search to SATB1, a protein whose cDNA was originally cloned from a testis cDNA library but that is reported to be expressed only in thymus (15, 47) (Fig. 2B). Although the SATB1 cDNA encodes a protein of molecular mass of 87 kDa, it is reported to migrate on SDS-PAGE as several closely spaced components of molecular mass of approximately 103 kDa, possibly due to phosphorylation or other post-translational modifications (47). This is comparable with the M r observed for the L2a-P1 protein (19). To test whether the isolated calf thymus protein was SATB1, Western blot analysis was performed using a polyclonal anti-human SATB1 antiserum. As shown in Fig. 2C, all three partially resolved components were detected by the anti-SATB1 antiserum but not by normal rabbit serum. Finally, to determine whether the retarded bands of mobility comparable with band 1 were due to calf SATB1, an EMSA assay was performed in the presence of polyclonal anti-SATB1 antiserum. As shown in Fig. 2D, anti-SATB1 antiserum abolished the appearance of band 1 and gave rise to a supershifted band. Taken together, these results demonstrate that calf SATB1 gives rise to re-tarded band 1 with the 200(LϩS) probe.
The L2a-binding Protein in VL3 Nuclear Extract Is Antigenically Related to SATB1 and Generates Band 1 with the 200(LϩS) Probe-To test whether the mouse L2a-P1 protein is antigenically related to SATB1, anti-SATB1 antiserum was used in an EMSA with mouse VL3 nuclear extract and the 200(LϩS) probe. As shown in Fig. 3A, anti-SATB1 antiserum but not normal rabbit serum supershifted band 1 but not band 2. These results suggest that mouse L2a-P1 is antigenically related to SATB1 and that the L2a-P2 protein that gives rise to band 2 is not (19).
Because the above experiment shows only that L2a-P1 is antigenically related to SATB1, it was necessary to test whether authentic mSATB1 gives rise to band 1 in an EMSA. The mouse BW5147 thymic lymphoma cell line, which gives no band 1 in EMSA assays (19) and does not produce mSATB1 mRNA (Fig. 3B), was therefore stably transfected with pECHmSATB1, an expression construct encoding mSATB1. Subclone A11 and a number of additional subclones expressed an appreciable amount of mSATB1 mRNA (though less than VL3) (Fig. 3B), and A11 nuclear extract gave rise to a complex of similar mobility to band 1 in an EMSA assay performed with the 200(LϩS) probe (Fig. 3A). Furthermore, anti-SATB1 antiserum but not normal rabbit serum abolished and supershifted the band 1 complex obtained with both A11 and VL3 nuclear extracts (Fig. 3A). These results strongly suggest that L2a-P1 is mSATB1. Importantly, band 2 was not supershifted by anti-SATB1 antiserum.
Binding of L2a-P1 to MAR Probes-The binding specificity of the L2a-P1 protein in VL3 nuclear extract was compared with that reported for human SATB1 (hSATB1) (15) in direct binding and competition EMSA assays. The hSATB1 was shown to bind a pentamer of the wild type binding site ( (25) 5 ), referred to as WT, derived from the 3Ј MAR of the immunoglobulin heavy chain enhancer, but not to a mutated version ((24) 4 ) (referred to as MUT) of the WT probe (15). In EMSA assays using VL3 nuclear extract and radiolabeled WT probe, all bands were efficiently competed by a 50-fold excess of unlabeled WT fragment (Fig. 4A) and incompletely competed by as much as a 500-fold excess of the 200(LϩS) fragment. The less efficient competition by the 200(LϩS) fragment may reflect the fact that it is a monomer, whereas the WT probe is a pentamer, but it is also possible that the affinity of L2a-P1 for the WT probe binding site is greater than that for its binding site in the 200(LϩS) fragment. No competition was obtained with the MUT fragment. In EMSA assays using radiolabeled 200(LϩS) fragment as probe (Fig. 4B), unlabeled 200(LϩS) and WT fragments but not MUT fragment competed for both bands with WT competing somewhat more efficiently than 200(LϩS) fragment.
The ability of L2a-P1 to bind to the WT probe and the ability of the WT but not the MUT probe to compete for L2a-P1 binding to WT and 200(LϩS) probes suggest that on the basis of these measures of binding specificity, L2a-P1 behaves like SATB1. This conclusion is strengthened by the finding that anti-SATB1 antiserum supershifts the band 1 complexes formed by VL3 nuclear extract and the WT MAR probe (Fig.  4C). Importantly, band 2 formed with either the 200(LϩS) or the WT MAR probe was not supershifted by anti-SATB1 antiserum (Figs. 3A and 4C).  (Fig. 5B). The size is consistent with that reported for mSATB1, and SATB1 has been reported to be phosphorylated (47). The failure to observe a corresponding band precipitated from A11 extract radiolabeled with either isotope may reflect the very low amount of SATB1 protein expressed. A low level of SATB1 protein expression in A11 cells is consistent with the weak intensity of retarded band 1 observed in EMSA assays as compared with that seen with VL3 nuclear extract (Fig. 3A).
L2a-P1 and SATB1 Have Similar DNA Sequence Require-ments for Interaction with the WT Probe-Dickinson et al. (15) used the missing nucleotide assay to determine the nucleotides required for interaction of bacterially produced SATB1 with a dimer of the WT MAR sequence. We have used both the missing nucleoside approach and DNase I protection assays to study the interaction of L2a-P1 and L2a-P2 with the L2a element.
In the missing nucleoside approach (44), double-stranded probe molecules are radiolabeled at one end of one strand only, and an average of one single-stranded nucleoside gap is introduced into each probe molecule. An EMSA is then performed with limiting amounts of probe, retarded bands are recovered, and the distribution of probe molecules in each band is displayed on a DNA sequencing gel. Failure of a probe of a certain length to appear in a retarded band suggests that a gap at that region interfered with binding of the protein that gives rise to that band. Generally, probes that fail to bind protein are overrepresented in the remaining unbound probe, which is also run on the sequencing gel (referred to as "free probe").
To compare the binding of L2a-P1 with that previously observed for SATB1, we performed the missing nucleoside assay using VL3 nuclear extract and the (25) 2 MAR probe used by Dickinson et al. (15). As shown in Fig. 6, the band 1 complex generated with the (25) 2 wild type MAR probe and VL-3 nuclear extract showed that L2a-P1 binding was inhibited by gapping at both repeats of the nucleoside sequence 5Ј-ATTATA-3Ј. Interestingly, gapping at the fourth nucleoside (the A shown above in italics) in either binding site did not appear to interfere with binding of L2a-P1. These results are very similar to those obtained for the bacterially produced mSATB1 and the (25) 2 WT probe (47), where gapping in any of the nucleosides AAGATTATAT interfered with mSATB1 binding. These similarities provide further evidence that the L2a-P1 protein is mSATB1. The imperfect match of footprints observed for L2a-P1 (Fig. 6) and bacterially produced mSATB1 (47) may reflect post-translational modification of L2a-P1 (e.g., phosphorylation) in the VL3 cells. Unfortunately we were unable to obtain a footprint with bacterially produced mSATB1 in our laboratory for direct comparison, and the A11 cell line (see above) produced too little mSATB1 protein for use in the missing nucleoside experiment.
Use of the Missing Nucleotide Experiment to Study the Binding of L2a-P1 (mSATB1) to the L2a Element-Previously, Lee and co-workers (19) had performed DNase I footprinting of the L2a element by mixing VL3 nuclear extract with the L2a probe and performing EMSA analysis without isolating the individual retarded bands 1 and 2. The two protected regions (L and S) and the intervening DNase I-hypersensitive region (referred to here as the INTER-LS region) identified in that study are shown in Fig. 1B. Also highlighted is a palindromic 12-mer present in the INTER-LS region.
In the present study using the missing nucleoside approach, the complex formed by L2a-P1 (i.e., band 1) was isolated before electrophoresis on the sequencing gel. As shown in Fig. 7A and summarized in Fig. 7E, nucleosides in the lower strand of the previously described L site (bracket I) were found to be important for L2a-P1 binding as seen by the absence of probe molecules in this region of the band 1 lane and enhancement of the corresponding fragments in the free probe lane. Loss of nucleosides in the L region of the upper strand (bracket VI in Fig. 7, B and E) also appeared to reduce binding. Interestingly, gapping at nucleosides in the middle of the previously described S site in the bottom strand also interfered with binding of L2a-P1 (bracket III in Fig. 7, A and E).
The bands corresponding to gapping at nucleosides in the INTER-LS region of both strands were enhanced in the band 1 lane as compared with the free probe and control lanes (bracket II of Fig. 7, A and E, bracket VII of Fig. 7, B and E). We interpret this to mean that gapping of probes in the INTER-LS region favors L2a-P1 binding, possibly by facilitating a conformational change in the DNA adjacent to the L2a-P1 interaction sites.
Use of DNase I Footprinting to Determine the Interaction of L2a-P1 (mSATB1) with the L2a Element-VL3 nuclear extract was used to footprint the L2a element, and in contrast to the previous study (19), the individual retarded bands were isolated before electrophoresis on the sequencing gel. The complex formed by L2a-P1 (retarded band 1) yielded a strong footprint in the L site of the bottom strand and a weak footprint in the top strand (brackets I and VI of Fig. 8, A and B, respectively; summarized in Fig. 8E), demonstrating that bound L2a-P1 contacted nucleotides in this region. Consistent with the results of the missing nucleoside experiment showing some binding in the S region, a short DNase I-protected region was apparent in the S region of the bottom strand (bracket III, Fig.  8, A and E).
The footprint of the complex formed by L2a-P1 demonstrated DNase I hypersensitivity in the INTER-LS region, which ex- tended several nucleotides into the S site (brackets II and VII of Fig. 8, A and B, respectively; summarized in Fig. 8E). This DNase I-hypersensitive region overlapped with the enhanced region seen in the missing nucleoside assay (brackets II and VI of Fig. 7E), and its presence is consistent with a local alteration of DNA conformation that is induced by L2a-P1 binding. In summary, the results of the DNase I footprinting studies of the complex of L2a-P1 with the L2a element demonstrate that most of the nucleosides required for L2a-P1 binding to the L site, as implicated by the missing nucleoside assay, are protected by L2a-P1 binding.
The Requirements for Binding of L2a-P2 to the L2a Element Are Distinct from Those of L2a-P1-Previous studies (19) suggested that a protein(s) of 200 kDa or greater gave rise to retarded band 2 in EMSA assays performed with VL3 nuclear extract and an L2a probe. Simultaneous with the studies of the band 1 complex described above, the interactions with the L2a element of protein(s) in retarded band 2 were studied using the missing nucleoside experiment and DNase I footprinting.
In the missing nucleoside experiment performed with VL3 nuclear extract, the requirement for binding of L2a-P2 appear different from L2a-P1. Nucleoside gapping in the lower strand of the S site significantly interferes with band 2 formation (brackets IV and V, Fig. 7A), whereas gapping in the L site, which interfered with L2a-P1 binding (bracket I, Fig. 7A), ap-pears to have no effect. Also, unlike the case with band 1, nucleoside gapping in the INTER-LS region does not appear to favor formation of band 2. As summarized in Fig. 7E, these results suggest that the S site of the L2a element is the primary interaction site of protein L2a-P2.
However, previous studies had suggested that L2a-P1 binds more strongly to the L2a probe than L2a-P2 (19). Under the conditions of limiting amounts of probe employed for the missing nucleoside experiment, the L2a-P1 and L2a-P2 proteins are competing for probe molecules, and in this situation an incomplete assessment of the binding specificity of the L2a-P2 protein may be obtained. We therefore performed the missing nucleoside experiment using nuclear extract derived from the BW5147 cell line, which does not express L2a-P1 and thus gives rise to only band 2 in the EMSA assay. As shown in Fig.  7C for the lower strand and summarized in Fig. 7E (brackets VIII and X), in the absence of L2a-P1 and band 1 formation, gapping in both the L and S regions interferes somewhat with the formation of band 2 complexes. Gapping in the L and S sites of the upper strand has little or no obvious effect on band 2 formation (Fig. 7D). However, gapping in either strand in the INTER-LS region appears to favor formation of band 2 complexes (bracket IX, Fig. 7C and bracket XI, Fig. 7D, summarized in Fig. 7E). Thus in the absence of competing L2a-P1, the L2a-P2 protein appears to show interaction with the L site as well as the S site.
Under the conditions of DNase I footprinting, probe DNA is in excess and proteins are not competing for binding to a limited number of probe molecules. Thus the footprints obtained should accurately reflect the nucleotides protected by the bound proteins. As shown by footprinting the lower strand of the L2a probe with VL3 nuclear extract (Fig. 8A), the proteins in bands 1 and 2 demonstrate contact with both the L site (brackets I and IV, respectively) and the S site (brackets III and V, respectively). Very slight protection of the L site of the upper strand is observed (bracket VIII, Fig. 8B). However, unlike the case for the L2a-P1 protein, the L2a-P2 protein in band 2 complexes results in no DNase I hypersensitivity of nucleotides in the INTER-LS region (Fig. 8, A and B, summarized in Fig. 8E).
The results of DNase I footprinting of the lower strand of band 2 obtained with BW5147 nuclear extract were similar to those obtained with VL3 nuclear extract. The L and S regions of the lower strand were protected (brackets IX and XI, respectively, of Fig. 8C), but the upper strand was not (Figs. 8D). Unlike the case with VL3, however, a short region of DNase I hypersensitivity was observed in both strands (brackets X and XII of Fig. 8, C and D, respectively, summarized in Fig. 8E).
Distamycin A Competes with L2a-P1 and L2a-P2 for DNA Binding-Because distamycin A, which binds to DNA in the minor groove, had been shown to abolish SATB1 binding to a WT MAR probe (15), we tested whether it would compete for complex formation in an EMSA using VL-3 nuclear extract and the 200(LϩS) probe. The addition of distamycin A (100 -200 M) abolished the formation of both bands 1 and 2 (Fig. 9). These results are consistent with both L2a-P1 and L2a-P2 binding to the minor groove of the DNA. Similar results were obtained when the wild type MAR fragment, (25) 5 , was used as probe (data not shown).
EMSA Analysis with Anti-CDP Antiserum Suggests That L2a-P2 Is the MAR-binding Protein, CDP/Cux-Based on its relative molecular mass of approximately 200 kDa, broad cell type distribution, and affinity for binding to AT-rich sequences as shown in the missing nucleoside and DNase I hypersensitivity studies (see above), we suspected that the L2a-P2 protein, which gives rise to retarded band 2, might be the MARbinding protein, CDP/Cux (51, 52). EMSA assays using VL3 nuclear extract and the 200(LϩS) probe were therefore performed in the presence of polyclonal anti-CDP antiserum. As shown in Fig. 10, anti-CDP antiserum but not normal rabbit serum abolished formation of retarded band 2 in EMSA assays performed with VL3, BW5147, and A11 nuclear extracts but had no effect on band 1, suggesting that the L2a-P2 protein may be CDP/Cux. Taken together with the finding that formation of band 2 was inhibited by distamycin A (Fig. 9), these results suggest that binding to the narrow groove contributes significantly to interaction of CDP/Cux with the 200(LϩS) probe.
EMSA Performed with Mutant L2a Elements Further Suggest That L2aP1 and L2a-P2 Bind Preferentially to the L Site and S Site, Respectively-To further analyze the sites in the L2a element required for formation of bands 1 and 2, mutations or deletions were introduced in regions implicated in L2a-P1 or L2a-P2 binding (Fig. 11A). In an EMSA performed using VL3 nuclear extract and wild type or mutant L2a elements as probes, mutations in the L site and in the middle of the S site had little effect on formation of band 2 but effectively abolished band 1 (Fig. 11B, lane M1). In results not shown, mutations in the L site alone nearly abolished band 1, although not as completely as when mutations in S were present as well. This suggests that L2a-P1 binds preferentially to the L site, whereas L2a-P2 binds binds to regions in the S site unaffected by the mutations. Also, as shown in Fig. 11B (lane M4), leaving the L site intact and deleting the S site and most of the INTER-LS region virtually abolished band 2 formation and appeared to enhance band 1 formation. These results support the above conclusion that L2a-P1 and L2a-P2 preferentially bind to the L and S sites, respectively.
The L2a Element Exhibits Properties of a MAR by Binding to the Nuclear Matrix-A direct nuclear matrix-binding assay was performed to test whether the L2a element behaves as a MAR (Fig. 12). A linearized pUC18 construct containing the wild type L2a element was radiolabeled, and its binding to nuclear matrices prepared from the mouse T cell line, EL4, was compared with that of a known MAR (8) and with vector alone as internal positive and negative controls, respectively (see "Experimental Procedures" and the legend to Fig. 12). The construct containing the wild type L2a element bound as strongly as the BrMAR positive control to the nuclear matrix, whereas the vector control showed no binding (Fig. 12A, lane B  under the WT bracket). This was a highly reproducible observation and strongly suggests that the wild type L2a element contains a MAR.
To determine whether the MAR involved sequences within the L2a element implicated in the above studies to interact with the L2a-P1 and L2a-P2 proteins, nuclear matrix binding of L2a elements containing mutations or deletions that affected binding of L2a-P1 and/or L2a-P2 (Fig. 11, A and B) was compared with binding of the wild type L2a element. As summarized in Fig. 12B, alterations in sites affecting either L2a-P1 or L2a-P2 binding reduced nuclear matrix binding to 15-28% of that observed for the wild type L2a element. Thus sequences required for binding of both L2a-P1 and L2a-P2 contribute to the nuclear matrix-binding properties of the L2a element. DISCUSSION Previous studies in which CD8a gene constructs containing varying amounts of 5Ј-flanking regions were stably transfected into BW5147 cells suggested the presence of an element between Ϫ3,800 and Ϫ4,700 nucleotides upstream of the gene, which was a target of negative regulation in that cell line (19). Negative regulatory activity appeared to reside in a 270-nucleotide subfragment of this region (nucleotides Ϫ4,430 to Ϫ4,700), which contained an element (L2a) that interacted with nuclear proteins. The L2a element lies within a DNase I-hypersensitive site specific to CD8-positive cells (30). Lee and co-workers (19) observed two retarded complexes (bands 1 and 2) in EMSA studies with the L2a element and nuclear extracts from all CD8-positive cell lines tested, mouse thymus, 2 and two of four double-negative T cell lines. In contrast, only the larger complex (band 2) was observed in EMSA assays using nuclear extract from a variety of CD8-negative cell lines. Evidence was presented that the L2a-P1 protein (molecular mass, approximately 100 kDa), which gives rise to retarded band 1, competes more strongly for binding to the L2a element than L2a-P2 (molecular mass, Ͼ200 kDa), which gives rise to band 2. A model was proposed in which interaction of the ubiquitously expressed L2a-P2 protein with L2a negatively affects CD8a gene expression, and its displacement by L2a-P1 has a positive effect on CD8a gene expression.
The present studies strongly suggest that the L2a element is a MAR and that the L2a-P1 and L2a-P2 proteins described by Lee et al. (19) are the MAR-binding proteins SATB1 and CDP/ Cux, respectively. That the L2a element is a MAR is suggested by its binding to T cell nuclear matrix preparations (Fig. 12A), and the observation that matrix binding is greatly reduced by mutants and deletions throughout the L2a element (Fig. 12, A  and B) suggest that the MAR properties reside in several regions of the L2a element. That L2a-P1 is SATB1 was demonstrated by results of EMSA supershift and immunoprecipitation studies using anti-SATB1 antiserum (Figs. 3A and 5, A and B), direct binding and competition EMSA studies using wild type and mutant MAR probes (Fig. 4, A and B), and inhibition of binding by distamycin A (Fig. 9). That L2a-P2 is CDP/Cux is suggested by EMSA supershift experiments using anti-CDP antiserum (Fig. 10), the general size of the protein that gives rise to the band 2 complex, and the similarity of the L2a-P2-binding sequences in the L2a element with the binding specificity reported for DNA-binding cut domains of the CDP/ Cux protein (51,52). The distamycin A sensitivity reported here for band 2 (Fig. 9) suggests that like SATB1, the CDP/Cux protein interacts significantly with the minor groove of DNA probes. Based on results of the present studies, the names SATB1 and CDP/Cux will be used below in place of L2a-P1 and L2a-P2, respectively.
The SATB1 protein was described as a special AT-rich DNAbinding protein present in high concentration in thymus (15,47). SATB1 was shown to bind in the shallow groove of DNA sequences in which one of the strands is ATC-rich. The site to which SATB1 binds in the 3Ј MAR of the IgH enhancer shows a propensity to become stably base-unpaired under conditions of torsional stress (53). Besides containing a proposed unique DNA-binding motif (47), SATB1 also contains a homeodomain 3 that may also participate in its interaction with DNA. Although no function for SATB1 has yet to be definitively shown, the studies of Lee and co-workers (19) together with the present results suggest that SATB1 may play a positive role leading to transcription of the mouse CD8a gene.
The CDP/Cux protein is named for the human (CDP; Ref. 54) and mouse (Cux; Ref. 55) homologues of the Drosophila homeodomain protein, Cut, which determines cell fate of several tissues of different embryonic origins (56,57). The CDP/Cux proteins contain a single homeodomain and three conserved DNA-binding repeats (called cut repeats) approximately 70 amino acids in length that exhibit subtle differences in DNA binding specificity (51,52,54). Although cut repeats 2 and 3 generally bind AT-rich sequences, they discriminate among similar nucleotide sequences. The CDP/Cux protein was suggested to negatively regulate the sea urchin histone H2B gene by displacing or competing for binding of a positively acting CCAAT box-binding factor (58). Negative regulation by CDP was demonstrated for the human cytochrome gene, gp91-phox (31), and CDP binding sites are present in the genes encoding human ␥-globin (59, 60) and rat neural cell adhesion molecule (55). Because not all CDP binding sites contain CCAAT boxes (31), the mechanism of repression by CDP/Cux is likely to have a more general basis.
We suggest that the L2a element, located in a region shown to be DNase I-hypersensitive and thus in an active and presumably exposed state in CD8-positive T lymphocytes (30), performs its function in regulating transcription of the mouse CD8a gene by behaving as a MAR. Based on studies of the interactions of SATB1 and CDP/Cux with the L2a element using the missing nucleoside experiment, DNase I footprinting, and EMSA studies using mutant L2a elements, we suggest that CDP/Cux represses CD8a gene transcription by binding to the L2a element and that its displacement by SATB1 favors transcription of the CD8a gene.
The L site defined previously by DNase I footprinting (19) appears to be the preferred binding site for SATB1 on the L2a element. This is suggested by missing nucleoside and DNase I protection experiments (summarized in Figs. 7E and 8E, respectively) and by EMSA assays using mutant L2a elements (Fig. 11B). Although SATB1 shows some evidence of interaction with the S site in the missing nucleoside experiment (region III in Fig. 7E) and in an EMSA assay performed with an L2a mutant probe with a wild type S region and altered L region (data not shown), the L site is its primary site of interaction with the L2a element. The CDP/Cux protein binds primarily to the S site as shown by the missing nucleoside (summarized in Fig. 7E) and DNase I protection experiments (summarized in Fig. 8E), but it can interact with the L site as shown in missing nucleotide experiments with BW5147 nuclear extract (Fig. 7E) and in DNase I protection experiments with BW5147 and VL3 nuclear extracts (Fig. 8E). When SATB1 is present, however (i.e., in VL3 nuclear extract), missing nucleoside experiments demonstrate that no binding of Cux/CDP to the L site is observed (Fig. 7E). Importantly, EMSA assays suggest that simultaneous binding of SATB1 and CDP/Cux does not occur because only two retarded bands are observed, one of which is shifted by anti-SATB1 antiserum and the other by anti-CDP/Cux antiserum. It cannot be ruled out, however, that both proteins may bind simultaneously to the L2a element under other conditions (e.g., in an intact chromosome).
The conformation of the region between the L and S sites appears to be affected by binding to the L site. The results of missing nucleoside experiments performed with VL3 nuclear extract suggest that gaps through the entire length of this region but especially in the palindromic 12-mer at the end of the INTER-LS region favor binding of SATB1 to the L site (summarized in Fig. 7E). In this experiment, such gaps appear to have little or no effect on Cux/CDP binding to the S site, but this could be misleading because under conditions of limiting probe, SATB1 may have bound to all such gapped molecules. Of interest in this regard is the finding that in studies with BW5147 nuclear extract that lacks SATB1, gaps in the intermediate region closest to the S site do appear to favor CDP/Cux binding (summarized in Fig. 7E). Nucleoside gaps in the IN- 3 T. Kohwi-Shigematsu, personal communication. TER-LS region may favor binding to the L or S site by facilitating a conformational change in the DNA that occurs upon SATB1 binding. That a major conformational change occurs upon SATB1 binding to the L site is indicated by the DNase I hypersensitivity of the entire INTER-LS region but especially the region closest to the S site in footprints of SATB1 (summarized in Fig. 8E). Similarly, DNase I hypersensitivity studies suggest that CDP/Cux binding can result in a significant but less dramatic conformational change in the INTER-LS region adjacent to the S site (summarized in Fig. 8E).
A model to explain the interaction of the SATB1 and CDP/ Cux proteins with the L2a element is shown in Fig. 13. In the absence of SATB1, we believe CDP/Cux binds primarily to the S site, but that it may also interact with the L site with one or more of its multiple cut domains (51,52). That CDP/Cux does not use the L site as its primary binding site is suggested by the significant reduction in band 2 in the EMSA performed with the M4 mutant L2a element (Fig. 11B). Some distortion of the INTER-LS region closest to the S site may result from binding of CDP/Cux. The interaction of CDP/Cux with the L site appears to be favored by gaps in the INTER-LS region as seen in the missing nucleoside experiment with BW5147 nuclear extract (Fig. 7D) and may lead to the distortion and resulting DNase I hypersensitivity observed with BW5147 nuclear extracts (Fig. 8, C and D). That such distortion is not a major factor in CDP/Cux association with its primary binding site in S is suggested by the missing nucleoside experiment with VL3 nuclear extract where CDP/Cux binds well to the S site of probe molecules that are not gapped in the INTER-LS region (Fig. 7A).
In the presence of SATB1, any CDP/Cux bound to or interacting with the L site is displaced by binding of SATB1. The present results suggest that the displacement of CDP/Cux is not due to simple competition for a common binding site, but results from the appreciable structural distortion induced by SATB1 throughout the INTER-LS region, particularly in the palindromic 12-mer adjacent to the S site (Fig. 8, A and B). Thus binding of SATB1 to the L site creates a structural distortion in the INTER-LS region that displaces CDP/Cux from the S site. We have called this mechanism a "displacement switch," and it may represent a general mechanism for switching some MAR elements from a repressed to an active state (Fig. 13). In this instance, the presence of SATB1 on the L2a element and the structural distortion resulting from SATB1 binding may increase the element's affinity for the nuclear matrix, thereby bringing the adjacent CD8a gene into an environment favorable to its transcription.
It is interesting that SATB1 binding is especially favored by nucleoside gaps in the palindromic 12-mer (Fig. 7, A, B, and E) and that DNase I hyperensitivity induced by either SATB1 or CDP/Cux binding is greatest in this region as well (Fig. 8). Because palindromes are frequently sites for protein interaction, it is possible that an as yet unidentified protein interacts with this site and may function in concert with SATB1 and/or CDP/Cux to modulate CD8a gene expression. For example, such a protein may bind to the L2a element, either alone or together with CDP/Cux, and be displaced by binding of SATB1 and the conformational change it induces in the palindromic binding site. Such a protein could function to further modulate association of the L2a element with the nuclear matrix, or it could affect CD8a gene transcription in some other way. Recently, EMSA analyses using a probe containing the palindromic 12-mer but lacking L and S sites suggest the presence in BW5147 nuclear extracts of a protein that specifically binds to this region. 4 The identity and binding properties of this protein are under investigation.
The model predicts that alterations in the L2a element that affect binding of SATB1 and/or CDP/Cux should have effects on transcription of the CD8a gene. We have recently shown that surface CD8␣ molecules are expressed by BW5147 cells stably transfected with a CD8a gene construct containing up to nucleotide Ϫ6,900 of the 5Ј-flanking region but lacking a 1.2-kb fragment containing the L2a element. 5 Thus the transfected CD8a gene is not subject to the negative regulation seen when this region is present (19). Interestingly, results of studies on mice in which the SATB1 gene has been "knocked out" by homologous recombination have shown that within the abnormal and greatly reduced population of T cells present in lymph nodes of SATB1 Ϫ / Ϫ mice, significant numbers of CD4-positive T cells are present but CD8-positive T cells are virtually absent. 6 These results suggest that SATB1 is required for production of mature CD8-positive T lymphocytes and are consistent with our suggestion that binding of SATB1 to the L2a element plays a role in regulating expression of the CD8a gene. Further information on the role of the L2a element and its associated proteins in CD8a expression should be forthcoming from stable transfection studies with constructs in which the wild type L2a MAR element has been replaced by mutant elements impaired in their interactions with SATB1 and/or CDP/Cux and from mouse knockout studies in which the L2a element has been deleted or modified. The model proposed for the displacement of CDP/Cux from the L2a element by SATB1 can be further tested by in vitro studies with additional mutant L2a probes and with modified SATB1 and CDP/Cux proteins and/or domains.