Cooperative Binding between Factors RFX and X2bp to the X and X 2 Boxes of MHC Class I1 Promoters*

histocompatibility complex (MHC) class I1 genes is controlled primarily by the promoter, which contains several conserved cis-acting ele- ments, including the X, X2, and Y boxes. We show here that RFX, the X box-binding protein that is deficient in certain MHC class I1 regulatory mutants, binds coopera- tively with an X2 box-binding protein (X2bp) to form an RFX-X2bp*DNA complex in which the interaction of the two factors with their target sites is strongly stabilized. A functional role of this RFXeX2bp complex is consistent with mutational analysis of the X and X2 boxes of the DRA and DRBl class I1 promoters. Together with previ- ous results demonstrating cooperative binding between RFX and the Y box-binding protein NF-Y, our results indicate that RFX plays a central role in promoting cooperative binding interactions required for stable occu- pation of the MHC class I1 promoter. This may explain why the highly specific defect in binding of RFX ob- served in certain MHC class I1 regulatory mutants is associated in vivo with a bare promoter in which all of the cis-acting elements, including the X, X2, and Y boxes, are unoccupied.

The abbreviations used are: MHC, major histocompatibility complex; BLS, bare lymphocyte syndrome; CRE, CAMP-responsive element; CREB, CRE-binding protein; TRE, TPA responsive element; ATF, activation transcription factor; EMSA, electrophoretic mobility shift assays. iting a n immune response. Precise regulation of MHC class I1 gene expression is crucial for the control of the immune response, as demonstrated by the fact that inappropriate or unusually high expression is associated with autoimmune disease (5) and a lack of expression leads to severe immunodeficiency (6). MHC class I1 genes are expressed constitutively in only a limited number of cells, including B lymphocytes, macrophages, dendritic cells, thymic epithelium, and activated T lymphocytes (4). Their expression can be modulated or induced by a variety of stimuli (41, of which the most potent is interferon-y (7-9). Finally, in mature plasma cells they are maintained repressed (10,11).
In transient transfection experiments the 150-base pair promoter region of MHC class I1 genes is sufficient to reproduce normal cell-specific and inducible transcription (for reviews, see . All MHC class I1 promoters contain conserved cis-acting sequence motifs referred to as the W (also called Z, H, or S), X, X2, and Y boxes (4,12-14). The Y box is the target of NF-Y (or YEBP), a heterodimeric CCAAT-binding transcription factor, which has been shown to be essential for class I1 promoter activity in in vitro transcription systems (15)(16)(17). Proteins binding to the X box include a novel family of DNA-binding proteins called  and a distinct nuclear protein called RFX (21,22). While the role of the cloned X box-binding proteins (RFX1-RFX4) in MHC class I1 gene transcription is not clear, there is strong evidence that RFX is crucial for MHC class I1 promoter function. Indeed, RFX binding activity is deficient in several B cell lines derived from patients suffering from MHC class 11-deficient immunodeficiency (also called the bare lymphocyte syndrome or BLS), a disease known to be due to a regulatory defect leading to a complete lack of MHC class I1 gene expression (6, [21][22][23]. The X 2 box is a relatively degenerate sequence element, which, depending on the MHC class I1 promoter examined, is related to either a CRE or a TRE (4,(12)(13)(14), and can hence be recognized in vitro by members of the Jun/Fos and CREB/ATF families (24-28).
The MHC class I1 promoter appears to function as a single unit in which spacing between the individual cis-acting elements is crucial (29,30) and in which the W, X, X2, and Y boxes all contribute to both B cell-specific and interferon-y-induced expression (4, [12][13][14]. This suggests that proteins binding to these sequences interact cooperatively to activate the MHC class I1 promoter. We have confirmed this recently by showing that B cell-specific and interferon-y-induced expression requires cooperative binding between RFX and NF-Y (31). Several lines of evidence suggested that RFX was also likely to bind cooperatively with proteins binding to the X 2 box. First, the X and X 2 boxes are always immediately adjacent to each other. Second, the X 2 box is quite degenerate and in several MHC class I1 promoters, such as in the DRBl and DRB3 genes, consists of essentially only one TRE or CRE half-site, a se-quence to which JuniFos and CREBIATF proteins would not be expected to bind efficiently on their own and might require stabilization by a protein such as RFX binding to a n adjacent site. Finally, the strongest indication comes from the analysis of BLS cells exhibiting a regulatory defect in MHC class I1 gene expression (6, 23). Several of these mutant cell lines are characterized by a defect in RFX binding activity in vitro (21, 22). Although in vitro this binding defect is restricted to RFX and does not concern proteins binding to the other promoter elements (21, 22), the entire promoter, including the X2 box, is unoccupied in vivo (32, 33). This discrepancy suggests that occupation of the X2 box in vivo requires stabilizing interactions between RFX and X2-binding proteins.
Here we demonstrate that RFX and an as yet unidentified X2-binding protein (X2bp) bind together to MHC class I1 promoters to form a complex in which the interaction of both proteins with their respective target sites is stabilized. The RFX.X2bp complex forms efficiently even on the DRBl promoter in which the X and X2 boxes taken individually are only low affinity target sites for RFX and X2bp, respectively. Together with the bare promoter phenotype observed in RFXnegative BLS cells (32, 33) and previous data on cooperative binding between RFX and NF-Y (31), these results suggest that RFX plays a central role in promoting cooperative binding interactions required for MHC class I1 promoter occupation and activity.

MATERIALS AND METHODS
Plasmids and Oligonucleotides-The DRA x/x2 (-144 to -701, DRA X2 (gtcagtcATGCGTCATCTagtcag), DRAW (-144 to -lOl), and DRAY (-89 to -49) oligonucleotides were prepared as described (21). Mutated DRA X/X2 oligonucleotides were prepared and provided by J. M. Boss (Atlanta). The Py oligonucleotide contains the EF-C site from the enhancer of polyoma (19,20) inserted into the EcoRV site of the polylinker of a Bluescript plasmid (Stratagene). This Py oligonucleotide is a high affinity binding site for RFX but contains no sites for X2bp. The TRE and CRE oligonucleotides were prepared and provided by M. Yaniv (Pasteur Institute, Paris, France) (34). The wild type DRB1-CAT reporter construct containing nucleotides -176 to +19 of the DRBl promoter has been described (35). The DRBlXm and DRBlX2m mutations were introduced into this plasmid by the same method described previously (35). Wild type and mutated DRBl fragments used in EMSA (-160 to -73 of the DRBl promoter) were derived from their respective plasmids by polymerase chain reaction.
Electrophoretic Mobility Shift Assays (EMSA) a n d Methylation Interference Assays-Methylation interference assays were done as described (21, 22, 36). EMSA was performed as described (31) with the following modifications. All binding reactions were done at 15-20 "C. Unless indicated otherwise ( Fig. 2 A ) , binding mixtures contained 8 pg of nuclear extract. All binding mixtures contained 100 ng of a methylated pBR322 oligonucleotide (19,20). This oligonucleotide eliminates complexes due to binding of the cloned RFX1-RFX4 proteins (19, 20) but does not affect binding of the nuclear RFX complex. The amounts of various competitor oligonucleotides added in each experiment are indicated in the figure legends. For analysis of the dissociation rates, binding mixtures optimized for RFX alone or for RFX.X2bp (see Fig. 2) were first incubated for 30 min at 15 "C, supplemented with 100 ng of the DRA X / X 2 oligonucleotide, and then pursued a t 15 "C for the indicated times prior to gel electrophoresis.
Nuclear Extracts and Affinity Purification of RFX-Nuclear extracts (37) were prepared from B cell lines (Robert and SJO) or from fresh B lymphocytes (Dietto) obtained by leukopheresis of blood from a B cell lymphocytic leukemia patient. The protocol for purification of RFX will be described in detail elsewhere.2 Briefly, a crude B lymphocyte nuclear extract was first fractionated by hydroxylapatite chromatography to eliminate X box-binding proteins other than RFX, such as the RFX1-RFX4 proteins (20). The 0.2 M NaCl eluate containing RFX was then subjected to affinity chromatography using three different consecutive techniques: 1) binding with a biotinylated X box oligonucleotide and purification using streptavidin-coupled magnetic beads, 2) binding with Durand, B., Kobr, M., Reith, W., and Mach, B. (1994) Mol. Cell. Biol., in press. a biotinylated X box oligonucleotide and fractionation on a streptavidinagarose column, and 3) fractionation on a covalently coupled X box oligonucleotide-agarose column. The final fraction was estimated to be enriched approximately 1500-fold and contains no detectable X2 boxbinding proteins or X box-binding proteins other than RFX. Cell Culture, Tkansfections, and CAT Assays-The B lymphoma cell line Raji and the Epstein-Barr virus-transformed B cell lines Robert and SJO were grown in RPMI 1640 medium, 2 m~ L-glutamine, 100 IU penicillin, 100 pg/ml streptomycin, 10% fetal calf serum, at 37 "C in 5% CO,. Raji cells (1 x lo7) were cotransfected with 30 pg of wild type or mutated DRB1-CAT plasmids and 3 pg of pSVPAP (38) by electroporation with a 250-V, 960-millifarad pulse (Gene Pulser, Life Technologies, Inc.). Cell extracts were prepared 42 h after transfection, assayed and corrected for alkaline phosphatase activity (381, and then assayed for chloramphenicol acetyltransferase activity (39). Results were quantified by excision and counting of the acetylated and nonacetylated [14Clchloramphenicol from the chromatograms.

RFX Forms n o Distinct Complexes with the XIX2 Box
Region-In EMSA experiments performed with a DRA X / X 2 oligonucleotide and nuclear extracts from MHC class II-positive B cells, two major complexes are observed ( Fig. 2 A ) . The lower complex is due to binding of RFX (21, 22). This lower complex is favored at low protein concentrations and when binding is performed a t 0 "C (21,22). The upper complex, on the other hand, is favored at high protein concentrations and when binding reactions are carried out at 15-20 "C ( Fig. 2 A ) . Like RFX (lower complex), the upper complex does not form in nuclear extracts from RFX-negative regulatory mutants such as Robert and SJO ( Fig. 2 A ) , suggesting that it also contains RFX. To determine whether this is indeed the case, an SJO extract was complemented with an affinity-purified fraction of RFX. This RFX fraction has been purified to near homogeneity,' and in EMSA generates only the band characteristic of RFX (lower complex), even when essentially all of the probe is bound (Fig.  2 A ) . Both complexes are restored in the complemented SJO extract, thus confirming that the upper complex also contains RFX ( Fig. 2 A ) . Further evidence for this is provided by the finding that both complexes are eliminated by competitor oligonucleotides ( X K 2 and Py) specific for RFX (Fig. 2B).

RFX Binds Together with a Protein
Binding to the X 2 Box-The lower RFX complex corresponds to binding of RFX alone, because it comigrates with affinity-purified RFX ( Fig. 2 . 4 ) and exhibits methylation interference contact points that are identical to those previously characterized for RFX ( Fig. 1) (21, 22,  36). The upper complex also contains the major contact points of RFX, but in addition exhibits two additional contact points ( Fig. 1) that are characteristic of X2 box-binding proteins (361, suggesting that it contains RFX together with a second protein (X2bp) bound to the X2 box. The target site of X2bp was confirmed by competition experiments (Fig. 2B) and binding experiments using DRA X K 2 oligonucleotides containing point mutations in the X 2 box (Fig. 3A). The upper complex was not formed in the presence of an excess of unlabeled X2 oligonucleotide, while addition of oligonucleotides flanking the X K 2 region (Y and W) had no effect (Fig. 2B). Moreover, mutations falling within the X2 box, but not outside the X 2 box, specifically inhibited formation of the upper band (Fig. 3A).
X2bp Is a TREICRE-binding Protein-To determine whether X2bp could belong to the Jun/Fos or CREB/ATF families, of which several members are known to be able to bind t o the X2 box (24-281, we performed competition experiments with various TRE and CRE oligonucleotides. Formation of the RFX.X2bp complex is specifically inhibited by several of these oligonucleotides (Fig. 3B), suggesting that X2bp is indeed a member of the above mentioned transcription factor families. The two CRE oligonucleotides show the most efficient competition (oligonucleotides 4 and 5 in Fig. 3B), while a perfect TRE site does not compete (oligonucleotide 3 in Fig. 3B).

Cooperative Binding between RFX and
Binding of RFX and X2bp Is Stabilized in the RFX.X2bp Complex-No complexes attributable to X2bp bound on its own are detected, while even in the presence of an excess of free probe and low protein concentrations, the RFX.X2bp complex is always the most abundant (Fig. 2). Moreover, elimination of the RFX-X2bp complex can only be obtained by high concentrations of competitor oligonucleotides (250-500-fold molar excess) (Figs. 2 and 3B). Clearly, binding of X2bp to the DRAX2 box is strongly dependent on and stabilized by RFX.
The RFX.X2bp complex is also more abundant than RFX bound on its own (Fig. 2), suggesting that binding of RFX is also enhanced by X2bp. To confirm this, we determined dissociation rates for the RFX.DNA and RFX.X2bp.DNAcomplexes (Fig. 4). The comparison of the half-lives of these complexes shows that stability of the RFX-DNA interaction is enhanced at least 100fold by the presence of X2bp (Fig. 4).
The methylation interference experiments suggest that stabilization of RFX by X2bp results in a change in the interaction of RFX with the X box (Fig. 1). Two weak inhibitions observed for the lower complex (RFX) at the 5' end of the X box are consistently missing for the upper complex (RFX.X2bp). This change in the interaction of RFX with the X box is likely to be due to the stabilizing effect of X2bp, which might be expected to render the RFX.X2bp complex less sensitive to methylated G residues that already exhibit only a weak inhibition on binding of RFX on its own.
Enhanced Stability of the RFX.X2BP Complex Allows Its Formation on the Low Affinity X and X2 Boxes of the DRBl Promoter-The X and X 2 boxes of the DRB 1 promoter are both only very low affinity target sites for RFX and X2bp, respectively. RFX binds very poorly to the DRBl X box (351, such that little or no specific complex is detected using a DRBl oligonucleotide containing only a functional X box (Fig. 5A). Similarly, no =-binding complexes are observed with a DRBl oligonucleotide containing only a functional X 2 box (Fig. 5A). Nevertheless, the RFX.X2bp complex forms readily on the wild type DRBl promoter fragment containing intact X and X 2 boxes (Fig. 5A). This RFX.X2BP complex is in all respects identical to the one formed on the DRA promoter. First, it is eliminated by an excess of the DRA X 2 box competitor oligonucleotide (Fig.  5A). Second, it has very similar methylation interference contact points within the X and X 2 boxes (Fig. 1). Third, it does not form in extracts from RFX-deficient regulatory mutants (35). Finally, it has the same relative affinities for the different TRE and CRE oligonucleotides tested in Fig. 3B (data not shown). The interaction between RFX and X2bp thus stabilizes their binding sufficiently to allow them to bind together efficiently to the DRBl promoter.
Functional Relevance of the RFX.X2bp Complex-It has been shown previously that many of the DRA X box mutations that inhibit binding of RFX also compromise DRApromoter function (40, 41). Mutations of the DRA X 2 box also lead to a reduction in promoter activity (41). Interestingly, DRA X 2 box point mu- tations that inhibit formation of the RFX.X2bp complex (Fig.  3 A ) correlate well with those that affect promoter function (41). Taken together, these results suggest that the RFX.X2bp complex is functionally relevant for the DRA promoter.
The respective roles of the X and X 2 boxes of the DRBl gene have not been studied previously, although the entire x/x2 region is clearly crucial for DRBl promoter activity (35). The DRBl X box is a poor target site for RFX (Fig. 5A). Moreover, the DRBl X 2 box is only a poorly conserved CRE (Fig. 1) which does not even have one perfect half-site (GTCA), and by itself does not appear to be bound by =-binding proteins (Fig. 5A). Yet even these poor X and X 2 sites should be functionally important if the RFX-X2bp complex formed on the DRBl X / X 2 region is relevant for promoter function. We therefore constructed two mutated DRBl promoter-CAT reporter gene con- structs. One, DRBlXm, contains clustered point mutations that completely disrupt the X box but leave the X 2 box intact. The second, DRBlX2m, contains X 2 box mutations that eliminate all homology to TRE or CRE sequences. To eliminate all potential TGA half-sites, two additional point mutations were introduced into the 3' end of the X box (Fig. 11, but these mutations are known from a previous study on the DRA gene to have no effect on binding of RFX (40) and do not eliminate weak binding of RFX to the DRBl X box (data not shown). Disruption

Cooperative Binding between RFX and X2bp
of the X box (DRBlXm) results in an activity that is only 4% of wild type (Fig. 5B 1. Mutation of the X2 box (DRBlX2m) leads to an activity that is only 40% of wild type (Fig. 5B). Thus, both the X and X2 boxes contribute to activity of the DRBl promoter, although taken individually they constitute only low affinity sites for RFX and X2bp, respectively.
The effect of the X2 box mutation is, as expected, less strong than that of the X box mutation. Residual activity of the DRBlX2m promoter can be explained by the fact that disruption of the X 2 box does not affect formation of second functionally relevant composite complex (RFX.NF-Y) on the DRA (31) and DRBl (data not shown) promoters. Mutation of the X box (DRBlXm), on the other hand, completely eliminates promoter activity because it disrupts both the RFX.X2bp (Fig. 5) and RFX.NF-Y (31) complexes.

DISCUSSION
It is becoming increasingly clear that protein-protein interactions play a key role in determining the transcriptional activity of eukaryotic genes. Despite detailed analysis of the individual promoter elements and the proteins that bind to them (4, 12-14), few studies have addressed the question of proteinprotein interactions that allow MHC class I1 promoter-binding proteins to cooperate in activating transcription. In other systems, cooperation between transcription factors has been shown to occur indirectly at the level of transcription activation by synergistic interactions with the basal transcription machinery (3,(42)(43)(44) or directly at the level of DNA binding (4547). Activity of MHC class I1 promoters appears to depend to a large extent on cooperative binding interactions in which the X box-binding protein RFX is a central player. In a previous study, we have shown that RFX binds cooperatively with the Y box-binding transcription factor NF-Y to form a strongly stabilized RFX.NF-Y.DNA complex that is crucial for promoter activity (31). We show here that RFX also binds cooperatively with X2bp, an X2 box-binding protein, to form a n RFX.X2bp.DNA complex in which the interaction of both proteins with their target sites is strongly stabilized. Formation of the RFX.NF-Y and RFX.X2bp complexes is not mutually exclusive because the trimeric RFX.X2bp.NF-Y complex can form efficiently on promoter fragments containing the X, X2, and Y boxes (data not shown).
The stabilizing RFX.X2bp interaction is evident on both the DRA promoter and on the promoter of the DRBl gene. In addition, a complex now known to represent RFX.X2bp has previously also been shown to bind to the DRB3 promoter (35). The RFX.X2bp complex thus forms on at least three different coregulated MHC class I1 promoters. Moreover, the efficiency of RFX.X2bp complex formation, DRA> DRBl> DRB3, correlates with the relative strengths of these promoters (35).
An interaction between proteins binding to the X and X2 boxes has been postulated on the basis of mutational analysis of the X and X2 boxes of the DRApromoter (41). The cooperative binding between RFX and X2bp described here provides direct biochemical evidence for this model. Further evidence for a functional role of the RFX.X2bp complex comes from analysis of the DRBl promoter. Although taken individually the X and X2 boxes of the DRBl promoter are only very low affinity sites for RFX and X2bp, they are nevertheless crucial for promoter activity. These results are fully consistent with the observation that a cooperative interaction between RFX and X2bp is required to recruit these proteins to the DRBl X K 2 region. This situation is reminiscent of the composite response element of the proliferin gene, at which the glucocorticoid receptor interacts with members of the A F ' 1 family (48,49). I t is becoming more and more evident that such composite cis-acting elements, which integrate diverse regulatory signals by permitting interactions between different DNA-binding proteins, play an important role in the transcriptional control of eukaryotic genes (2, 50).
The results presented here and in a recent report by Reith et al. (31) demonstrate that the analysis of cis-acting elements by the introduction of point mutations must be interpreted and evaluated with caution. Indeed, single point mutations may disrupt binding of individual proteins on their own without affecting significantly the recruitment of these same proteins to the mutated elements by cooperative binding interactions. Three observations emphasize this point. 1) A mutation of the X box that completely eliminates binding of RFX on its own has no effect on formation of the RFX.NF-Y complex, and hence does not adversely affect promoter activity (mutant M2 in Reith et al., 1994 (31)). 2) With respect to the DRA promoter, the DRBl X-X2 region contains several nucleotide differences that abolish binding of both RFX and X2bp on their own, yet the RFX.X2bp complex forms efficiently (Fig. 5) such that DRBl promoter activity is only 2-fold lower than that of the DRA promoter (35).
3) The C residue immediately preceding the GTCA sequence of the consensus CREB/ATF binding site has been shown to be essential both for the function of this site and binding of CREB/ATF proteins (51, 52). However, mutation of the corresponding C residue in the X2 box (position -94) only partially inhibits formation of the RFX.X2bp complex (Fig. 3 A ) and promoter activity (41). Stabilization by RFX thus appears to allow X2bp to bind and function even at very unusual CRE sites. Clearly, meaningful conclusions concerning the role of the individual MHC class I1 promoter-binding proteins can only be extrapolated from the effects of promoter mutations if cooperative binding interactions are taken into account. Several different proteins have been shown to be able to bind in vitro t o the X2 boxes of certain MHC class I1 genes. These include members of the Jun/Fos and CREBIATF families such as c-Jun, c-Fos, hXBP1, and mXBP (24-28). We have as yet not been able to identify which, if any, of the above mentioned proteins are present in the X2bp complex detected here. Antibodies specific for CREB, c-Jun, c-Fos, and hXBPl have proved unable t o affect formation of the RFX.X2bp complex (data not shown). I t is, however, not clear whether this indicates that X2bp does not contain these proteins or whether the epitopes recognized by these antibodies are masked in the RFX.X2bp complex. Nevertheless, the data presented here should help identification of the proteins functioning at the MHC class I1 X 2 box. The functionally relevant X2-binding protein(s1 must exhibit the characteristic ability to bind cooperatively with RFX at the coregulated D M , DRB1, and DRB3 promoters. This criterion is clearly more stringent and functionally relevant than those based solely on specificity for the X2 box of a given MHC class I1 promoter. The latter approach can be misleading because the X 2 box is extremely variable in sequence, such that in the absence of stabilization by RFX the X2 box of different MHC class I1 genes may be bound preferentially by different proteins (24-28, 36). The study of MHC class I1 gene regulation has been greatly facilitated by the availability of regulatory mutants (BLS cells) deficient in expression of MHC class I1 genes. These mutants have been classified into at least three complementation groups (53)(54)(55). In vitro binding studies have shown that two of these complementation groups are characterized by a deficiency in RFX binding activity (21,22,(56)(57)(58). This binding defect is specific for RFX and does not affect X2-binding proteins or 22,56,57). Surprisingly, footprint experiments have demonstrated that in these mutants, the entire promoter, including the X 2 and Y boxes as well as the X box, is unoccupied in vivo (32,33). The finding that RFX stabilizes binding of both X2bp and NF-Y to the X2 and Y boxes, respectively, may provide an explanation for this apparent discrepancy between the in vitro and in vivo binding studies. In vitro, formation of the 22. Herrero Sanchez, C., Reith, W., Silacci, P., and Mach, B. (1992) Mol. Cell. Biol. higher Order RFX'X2bp, RFX'NF-Y, and RFX'X2bp'NF-Y ' Om-