Structural and functional analysis of NF-kappa B. Determinants of DNA binding specificity and protein interaction.

The NF-kappa B transcription factors display a high degree of sequence conservation in a domain initially described in the rel oncogene. Two family members, NF-kappa B1 and NF-kappa B2, have distinct DNA binding properties and functionally distinct effects on different enhancers. NF-kappa B1, for example, binds to the kappa B site from the human immunodeficiency virus (HIV) with approximately 15-fold higher affinity than NF-kappa B2. In this study, we have defined regions within the Rel domain which determine DNA binding specificity and interaction with other proteins. We find that the COOH-terminal putative Rel dimerization domain of NF-kappa B1 is required for preferential binding to the HIV kappa B site. In contrast, preferential stimulation of the HIV enhancer by NF-kappa B2 with RelA(p65) is determined by both the NH2- and COOH-terminal Rel domains of NF-kappa B2. These two regions of NF-kappa B2 also mediate preferential synergy with Bcl3. These data suggest that a specific subdomain of the Rel conserved region has evolved to control the fine specificity of DNA binding, and two distinct subregions within the Rel domain determine the specificity of interaction with other transcription factors. These specific Rel-conserved domains therefore determine the specificity of NF-kappa B interactions and contribute to selective gene activation.

terminal R e 1 domains of Mi"KB2. These two regions of NF-&2 also mediate preferential synergy with Bc13. These data suggest that a specific subdomain of the Re1 conserved region has evolved to control the fine specificity of DNA binding, and two distinct subregions within the Re1 domain determine the specificity of interaction with other transcription factors. These specific Rel-conserved domains therefore determine the specificity of N F -& interactions and contribute to selective gene activation.
!kanswiptional regulation by the NF-ttB/Rel family of proteins is regulated at multiple levels. The diversity of this family has been well established and includes at least five independent gene products in mammalian cells 11-11). Although they are related in their Rel-conserved region, N F -K B~~ gene products display Werences in DNA binding specificity and transcriptional activation (12-14). We have prex6ously described an NF-KB family member which is an -100-kDa protein and can give rise to an -60-kDa form which is capable of dimerizing to RelA(p65) and other Re1 family members (9,12,13). This gene product, W -K B~, binds less avidlyto the canonical KB site from the immunoglobulin (Ig) or HIV-1' enhancers than NF-KB~ (13). N F -K B~ nonetheless p r e f e r e n t i~y stimu-* This work was supported in part by Grant R 0 1 AI29179 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked ' '~~e r t~e~e~~~ in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. $ Present address: University of Ulm, Dept. of Internal Medicine I, 89081 Ulm,Germany. 5 Recipients of fellowships and grants from the Deutscbe Forschungs-g~m e~s c h~ (74013-1 (to R. M. S.) and 516/2-1 (to S. L.)). lates the HW enhancer in combination with RelA(p65) compared with N F -K B~ when transfmted into Jurkat cells (9). It is unlikely that this preferential activation is due to increased DNA binding affinity, since in the case of NF-&2/RelA, the DNA binding affinity of the heterodimer does not differ significantly from NF-KB~/R~LA (13,151. In addition, it has been found that NF-KBZ preferentially s t i m~a~ transcription in combination with Bc13 (16,17).
Despite their functional differences, N F -K B~ a n d N F -K B~ a r e nearly 55% identical at the amino acid level in their Rel-related regions. Within this region are two subdomains which show even higher sequence conservation ( Fig. 1). From previous studies, these subdomains, here termed A and B (Fig. 11,were shown to contribute to DNA binding and dimerization, respectively (4, 18-23). Mutations within the NH,-terminal region abrogate DNA binding but not dimerization, whereas those in the COOH-terminal region also affect dimerization (4,(12)(13)(14).
Although the physical structure of these molecules is not yet known, we sought to define the relative contributions of Re1 subdomains to further understand the structure and function of these proteins.
Because they are highly related, a variety of swap mutations between the two NF-KB subunits can be generated between conserved amino acid residues, and the contribution of these domains to the DNA binding specificity and interaction with RelA(p65) or Bc13 can be determined. In this report, we show that the COOHterminal (putative dimerization) region of N F -K B~ mediates the preferential binding to canonical KB, whereas both NH,and COOH-terminal regions of N F -K B~ a r e required to confer specificity of interactions with other proteins, presumably through protein-protein interactions of both the dimerization and DNA binding domains. These findings suggest that the putative dimerization domain of this DNA-binding protein determines its specificity of DNA binding, and i n t e~c t i o~s of the NE,and COOHterminal Re1 domains of N F -K B~ with adjacent proteins contribute to the specificity of transcriptional activation.

MATERIALS AND METHODS
Plasmids-The RSV eukaryotic expression plasmids containing the human NF-KB~ (RSV p105 truncated at XbaI, truncated at RsaI a h r RSV p100 truncated at XhoI), ReMRSV p65) and Bcl3(RSV Bc13) introduction of a stop codon at amino acid 434), NF-KB~ (RSV p49 and cDNAs have been previously described (9, 12, 13). A stop codon was introduced at amino acid 440 of NF-KB~ to generate a protein containing only the NH,-terminal part of N F -~B 2 ( p l~) using site-directed mutagenesis. To create swap mutations, silent restriction sites were introduced in the RSVp49 and RSVp50 plasmids by site-directed NF-RBZ has more amino acid sequence upstream of the Rel-conserved domains, and in some cases, numbering is not consecutive, although all fusions were made precisely at conserved residues. B.

c.
D. In addition, the RSVp49 and RSVp5O conserved SfaNI restriction site and the SphI site in RSVpBO were used to clone chimeric cDNAs. Fragments generated by polymerase chain reaction and cloning junctions were confirmed by sequencing analysis (9, 12, 13).

CZaI in
Proteins-293 human embryonal kidney cells or Jurkat T leukemia cells were transfected with eukaryotic expression vectors encoding indicated NF-KBI, NF-KB~, and mutants. Nuclear extracts were prepared as described earlier (24). For EMS&, proteins were analyzed and quantitated by SDS-polyac~lamide gel electrophoresis and s~d a~i z e d by Western blot analysis using a rabbit antisemm against a synthetic peptide c o~s p o n d i n g to the NHpterminal 18 amino acid sequence of N F -K B~ (25). A second antiserum was generated against a COOH-terminal peptide, corresponding to the COOH-terminal 376-392 amino acids of NF-~B2(p49). Using both antisera, all chimeric proteins were standardized to NF-~B2(p49). Expression in Jurkat cells of these vectors has been shown previously to be comparable (9, 12).
In vitro translations were performed by adding plasmid DNAs encoding the cDNAs for NF-KB1, NF-KB~, and RelA to a coupled transcriptiodtranslation rabbit reticulocyte lysate system (Promega), using T7 RNA polymerase and incubation temperatures recommended by the manufacturer. The translation products were radioactively labeled by including 20 pCi of [35S]methionine (Amersham Corp.) per reaction. Labeled proteins were analyzed on 10% SDS-polyacrylamide gels.
~~SAs-Proteins expressed in 293 cells were incubated with radiolabeled DNA probes in a 20-pl reaction mixture containing 10 m M Tris (pH 7.8, 50 rnM NaCl, 1 m M dithiothreitol, 3% glycerol, 50 p~ MgCl,, and 1 pg polyfdI.dCj poly fdI.dC). Nucleoprotein complexes were resolved by electrophoresis on 4% nondenaturing poIyacrylamide gels in 1 x Tris glycine. Dried gels were exposed to Kodak XAR-5 film at -70 "C with intensifying screens. The DNA probes used in this study were a double-stranded oligonucleotide probe encoding a single HIV KB site Oligonucleotides were radiolabeled using T, kinase and [Y-~~PIATP, -1 ng was used in an EMSA reaction. Immunoprecipitations-Immunoprecipitations were performed on in vitro synthesized proteins in a buffer containing 10 m M Tris (pH 7.5), 150 m M NaC1, 0.1% Nonidet P-40, and protease inhibitors. An antibody was used directed against RelA(p65) (Santa Cruz). The antigen-antibody complexes were precipitated with protein G bound to Sepharose beads, Laemmli buffer was added, complexes were dissociated by boiling, and separated on a 10% SDS-polyac~lamide gel. Dried gels were exposed to Kodak XAR-5 film.
~ansfec~ions and CAT Assu~s-~ans~ections using DEAF.-dextran with Jurkat T leukemia cell and CaPO, transfection of 293 cells were performed and standardized for transfection efficiency as described previously (24,26,27). CAT assays were also performed as described previously (24).

RESULTS
The Rel-conserved regions of N F -K B~ and N F -K B~ were examined. A series of swap mutations at conserved amino acids spanning the Rel-conserved region (Fig. 1) were expressed using a eukaryotic expression plasmid with the RSV enhancer/ promoter. Swap mutants which showed comparable leveis of DNA binding activity following transfection of 293 cells (Fig.   2 A ) and Western blotting (data not shown) were analyzed. Nuclear protein extracts were prepared, and their DNA binding activity determined by EMSA. Similar to the processed wild type NF-KB1 gene product, swap mutants which contained the COOH-terminal Rel-conserved region of N F -K B~ bound to the canonical KB site of the HWDg enhancers. These proteins also bound well to the &-related site in the class I MHC gene ( Fig. 2A, lunes 2-7 versus 10-15). In contrast, the N F -K B~ gene product (p52 or p49) bound poorly to the HlV/Ig KB site (Fig. 2.4, lune 8) type prottlin (Fig. 213. lnncs 2 to . 5 vprsus 9-12 I. These findings different swap mutant protcbins upon binding to the >IF!<' demonstrate that the COOII-terminal rrk~on of NF-KRI is KR site wcrc' diff+rrnt among thwr mutants. l'ossillly. these differences could be due to variation in conformation among these mutant proteins. Regardless of these differences, however, the complexes formed with the H N KB sites suggest that the COOH-te~inal putative dimerization domains of N F -K B~ or NF-KBS determine their fine specificity of DNA binding, presumably by indirectly altering the conformation of the region which contacts DNA. Another difference between N F -K B~ and N F -K B~ is their ability to functionally transactivate reporter genes in combination with RelA(p65). The efficiency may vary among cell types and is dependent upon the relative concentration of RelA(p65) (9, 12); however, in the presence of a constant and suboptimal amount of RelA(p651, N F -K B~ stimulates gene expression of the H N enhancer more effectively than NF-KB~ (Fig. 3 A ; see also Refs. 9 and 12). This ability of N F -K B~ to more efficiently synergize with RelA was similar whether a processed form of pl00 or the alternatively spliced transcript, p49, was used ( Fig. 3 A , NF-KBZ ip100; 1446) versus NF-KB2  (p4.9; 1444). The difference in cooperativity with RelA(p65) was not due solely to sequences in the COOH-terminal Re1 domain of NF-KBB (Fig. 3B, G J ) . In fact, all swap mutants showed reduced ability to act in concert with RelA(p65) (Fig.  3B). These results indicated that the NH,-or COOH-terminal Re1 region of NF-KBS alone did not specify its ability to preferentially activate transcription in concert with RelA(p65) and suggested that more than one part of the Rel-conserved domain of NF-KBB may be required.
To define the regions of the NF-KB~ and NF-KB~ responsible for these effects, swap mutants of both the COOH-and N H 2terminal Re1 regions were used (Fig. 4A). Several of these proteins demonstrated a higher level of ~a n~~i v a t i o n in combination with RelA(p65) (Fig. 4A, NF-KB 2-1-2, 1-V). Interestingly, two of these chimeric proteins transactivated the HIV enhancer better than wild type NF-KB~ in combination with RelA(p65) (Fig. 4. 4, NF-KB 2-1-2,III and N ) . Possibly, this difference is due to changes in the binding of these double swap chimeric proteins to the HIV KB site (Fig. 4 B , lanes 4 and 5 ) . Alternatively, the conformation of these swap mutants which improves DNAbinding might facilitate other protein-protein interactions that promote transcriptional initiation. NF-KB and swap mutants showed comparable levels of protein expression following transfection both in Jurkat cells (Fig.  4C, see also Ref. 12) and in 293 cells by Western blot analysis (data not shown), c o~~n g that the functional differences were not the resuit of differentia1 protein expression.
A similar effect was seen with Bc13, which had been shown previously to function more effectively with NF -KBB (16,17). In Jurkat cells, co-transfection of N F -K B~ and Bc13 expression vectors showed greater stimulation of CAT activity than NF-K B~ and Bc13, and the NH,-or COOH-terminal Re1 region of NF-KBS was not sufficient to restore function with Bc13 (Fig. 5). However, double swap mutants 111 and V were able to stimulate HIV enhancer activity to levels comparable with wild type NF-K B~ with Bc13 (Fig. 5). This synergistic effect was seen in a dose-dependent fashion, peaking when 1 pg of RSV Bc13 was used (data not shown). Of interest, similar subregions of the N F -K B~ Re1 domains conferred preferential activation for both RelA(p65) and Bc13.
TO determine whether NF-KB~ dimerizes more efficiently with RelA compared with NF-KB~, immunoprecipitations were performed with in. vitro synthesized proteins. This analysis showed comparable abilities of both proteins to interact with RelAin solution (Fig. 6). These results suggest that dimerization properties alone are unlikely to account for the differential effects of NF-KB subunits. It is thus likely that NH,-and COOHterminal domains contribute to the specificity of interactions with RelA(p65) and Bc13 by affecting protein conformation.

DISCUSSION
The role of NF-KB in transcriptional regulation is complex. A number of different KB sites have been found in the regulatory regions of multiple viral and cellular genes. It has been shown that specific interactions of different NF-KB/Rel subunits can allow selective activation of &-regulated genes (12). In addition, it has recently been realized that a variety of proteinprotein interactions occur between NF-KB and other transcription factors. In this report, we have determined the contributions of different regions of the Re1 domains of N F -K B~ and N F -K B~ to their DNA binding specificity and interaction with other transcription factors. In a separate study, we have determined that an interaction of the RelA(p65) transcription factor with Spl, also mediated through the Rel-conserved domain, is These data show that specific subdomains of the Rel-conserved region of NF-KB determine the specificity of interactions with DNA and with other transcription factors. Remarkably, the putative dimerization domain of N F -K B~ influences the specificity of DNA binding. This protein represents an example wherein hinding specificity can he mapped to a regpon thought not to be directly involved in contacting DNA. Despite the fact that the COOH-terminal domain confers DNA hinding specificity, both subunits, N F -K R~ a n d N F -K R~, s e e m to interact with RelA with comparable strength in solution (Fig. 6). These data suggest that Re1 domains which show differences in DNA hinding of the homodimer do not differ significantly in their ability to form heterodimers with another family member, RelAfp65). Differences in homodimer DNA binding specificity are there-fore likely determined by t h r conformation of the suhunits of the homodimer. In contrast to the NF-KR homodimrr hinding, we find that additional regions promote optimal functional interactions with RelA(p65) or Rcl3, suggrsting that hrtcrologous protein interactions can occur through the NH,-trrrninnl rrpion of the Rel-conserved domain. I t is likely that the. NH,-tc.rmin:~I subdomain (A) is required for interactinn with other transrription factors a s well (29).
Whether DNA binding is requirrd for this intcaraction in c.icw is unknown, although it is not required for thrsr intcrartions a t high protein concentration in r*ifro. Physiolo~~cic:~lly. within a cell, it is likely that DNA is required to stahilizc. thrsr intcvactions and promote the appropriate conformation of t h r RelA(p65) protein. These findings ancl swernl rccrnt ohsrwntions have begun to suggest that thr IINA hinding domains of
transcription factors are not limited simply to an interaction with nucleic acid.
Our data suggest that these regions of the protein, in addition to their associations with nucleic acids, allow them to intrract with other transcription factors. It is prohahle that the spacing of adjacent transcriptional regulatory sites helps to determine whether such interactions will facilitate productive assemhly of the transcription complex. In the case of the HIV enhancer, we have shown previously that spacing of the KB and Spl sites is essential for transcriptional activation and is not compensated by mutations which restore the dyad symmetry ofthe DNA(26). In this sense, the DNA serves as a template to facilitate specific protein-protein interactions which will subsequently allow cooperative hinding of TRP-associatrd factors. It is likely that different combinations of NF-KR gene products and cis-acting regulatory srquences therefore can be used to activate specific genes during cellular activation and in development. The ability to interfere with such protein interactions selectively may also provide for mechanisms to selectively regulate gene expression in rlivo and lead to further understanding of their roles in differentiation and devrlopment.