Acquisition of NFKB1-Selective DNA Binding by Substitution of Four ...

MOLECULAR AND CELLULAR BIoLoGY, July 1993, p. 3850-38590270-7306/93/073850-10$02.00/0Copyright © 1993, American Society for Microbiology

Acquisition of NFKB1-Selective DNA Binding by Substitution ofFour Amino Acid Residues from NFKB1 into RelA

TIMOTHY A. COLEMAN,t CHARLES KUNSCH,t MAUREEN MAHER,STEVEN M. RUBEN,t AND CRAIG A. ROSENt*

Department of Gene Regulation, Roche Institute ofMolecular Biology,340 Kingsland Street, Nutley, New Jersey 07110

Received 5 February 1993/Returned for modification 11 March 1993/Accepted 1 April 1993

The subunits of NF-cB, NFKB1 (formerly p5O) and RelA (formerly p65), belong to a growing family oftranscription factors that share extensive similarity to the c-rel proto-oncogene product. The homology extendsover a highly conserved stretch of approximately 300 amino acids termed the Rel homology domain (RHD).This region has been shown to be involved in both multimerization (homo- and heterodimerization) and DNAbinding. It is now generally accepted that homodimers of either subunit are capable of binding DNA thatcontains a KB site originally identified in the immunoglobulin enhancer. Recent studies have demonstrated thatthe individual subunits of the NF-cB transcription factor complex can be distinguished by their ability to binddistinct DNA sequence motifs. By using NFKB1 and RelA subunit fusion proteins, different regions within theRHD were found to confer DNA-binding and multimerization functions. A fusion protein that contains 34N-terminal amino acids of NFKB1 and 264 amino acids of RelA displayed preferential binding to anNFKB1-selective DNA motif while dimerizing with the characteristics of RelA. Within the NFKB1 portion ofthis fusion protein, a single amino acid change of His to Arg altered the DNA-binding specificity to favorinteraction with the ReIA-selective DNA motif. Furthermore, substitution of four amino acids from NFKB1into RelA was able to alter the DNA-binding specificity of the RelA protein to favor interaction with theNFKB1-selective site. Taken together, these findings demonstrate the presence of a distinct subdomain withinthe RHD involved in conferring the DNA-binding specificity of the Rel family of proteins.

NF-icB, a DNA-binding complex consisting of NFKB1(formerly p5O) and RelA (formerly p65), was originallyidentified as being involved in the regulation of immunoglob-ulin K (IgK) light-chain expression (50) and is now known tofunction as a pleiotropic transcriptional activator (18, 33).The targets of NF-KB remain diverse and include genes thatencode cytokines, cytokine receptors, cell adhesion mole-cules, and viral proteins (1, 3). Recent cloning of the NFKB1(8, 17, 29) and RelA (39, 44) subunits of NF-KB has providedfurther insight into the mechanisms by which this transcrip-tion factor functions. DNA sequence analysis has revealed a

highly conserved region of approximately 300 N-terminalamino acids, termed the Rel homology domain (RHD),which is found in both the NFKB1 and the RelA subunits ofNF-KB, the v-rel and c-rel gene products, and the Dros-ophila maternal morphogen dorsal (5, 10, 24, 52). Similarconservation is seen in the amino-terminal domains of twonew Rel family members, NFKB2 (LytlO) (38) and RelB(I-Rel) (45, 47).

Regulation of NF-KB function occurs at different levels,including transcriptional and posttranslational control (1, 3).With respect to transcriptional regulation, the expression ofNFKB1 and RelB is induced following mitogenic stimula-tion, while the expression of RelA appears to be constitutivein all cells examined (45). Posttranslational control can beexerted at the level of cytoplasmic retention, as demon-strated by the ability of certain Rel family members toassociate with a group of proteins that contain repeat ele-ments similar to those found in cytoskeletal anchoring pro-

* Corresponding author.t Present address: Human Genome Sciences, Inc., 9620 Medical

Center Drive, Suite 300, Rockville, MD 20850.

teins (2, 20, 55). Several of these proteins, IKB-a, IKB-0, andIKB-y, have been shown to form specific associations withindividual Rel family members (6, 11, 23, 28, 39, 53). Inaddition, proteolytic processing of the Rel proteins repre-sents another level of regulation. Both NFKB1 and NFKB2(formerly p49) are processed from a larger precursor mole-cule (12, 48). Intramolecular masking of the nuclear local-ization signal of NFKB1 by ankyrin repeat elements presentin the preprocessed C terminus has also been demonstrated(7, 12, 21, 34, 43). The mechanism for subsequent release ofthis cytoplasmic complex to the nucleus remains unclear,though earlier studies with purified protein suggested a rolefor phosphorylation (16, 51).While the transcriptional regulatory elements of numerous

genes contain NF-KB DNA-binding motifs, their transcrip-tional regulation differs dramatically over a range of stimulithat induce NF-KB activity (1, 3). Because of the sequencevariation in the KB binding sites and the differential regula-tion of the target genes, it has been suggested that individualRel-related proteins might display a unique preference for asubset of DNA-binding motifs. Indeed, using a polymerasechain reaction (PCR)-based oligonucleotide selection assay,we have identified some DNA sequences preferred byNFKB1, RelA, and c-Rel homodimers (32). Several of theselected sequences exhibited absolute binding specificitytoward one member of the Rel protein family. Furthermore,NFKB1-RelA heterodimers (NF-KB) were unable to bindthese selective sequences (32). Taken together, these datasuggest that transactivation of different target genes byNF-KB-Rel proteins likely involves a complex regulatoryscheme that includes not only the formation of differenthetero- and homodimeric complexes but also the selectiveassociation of these complexes with different DNA-bindingsites (41).

3850

Vol. 13, No. 7

NFKB1-SELEClIVE DNA BINDING 3851

Classification of the NF-KB-Rel family of transcriptionfactors remains unclear. Transcriptional activation in vivohas been associated with the C termini of RelA, c-rel, anddorsal (4, 10, 24, 37, 46, 49). While the C terminus of p105,the NFKB1 precursor, has been shown to contain severalankyrin repeats thought to be involved in cytoplasmic reten-tion and protein-protein interaction (7, 12, 21, 34, 43), theRHD of this protein has been shown to confer transcriptionalactivation in vitro (14, 30). Initial mutagenesis studies of Relfamily members have demonstrated that the multimerizationand DNA-binding activities of these proteins were colocal-ized within the RHD. In NFKB1, deletion of amino acids 12to 200 within the RID resulted in expression of a proteinthat functioned as a transdominant negative in that it hadmaintained its ability to multimerize but could no longer bindDNA (35). Consistent with this idea, Kumar et al. (31) haveidentified a motif (RxxRxRxxC) in v-rel, conserved in all Relfamily members, that is required for DNA binding. Thepresence of a region involved in multimerization was sug-

gested by isolation of a naturally occurring splice variant ofRelA, RelAA (formerly p65A), which contains an internaldeletion of amino acids 222 to 231 (46). This protein isdeficient in its ability to dimerize with NFKB1. These datasuggest that the DNA-binding and multimerization proper-

ties of the Rel proteins are found in the amino- and carboxy-terminal regions of the RHD, respectively. Sequence analy-sis of the RHD, however, has failed to demonstrate featuressimilar to the well-characterized motifs involved in multim-erization or DNA binding (i.e., leucine zipper, zinc finger, or

helix-loop-helix motifs) (26, 36).To further define the regions within the RHD involved in

multimerization and DNA binding, we constructed a seriesof NFKB1-RelA fusion proteins. The DNA-binding specific-ity of the fusion proteins was determined by their ability tobind to either an NFKB1- or a RelA-selective KB motif (32).Because multimerization is prerequisite to DNA binding, wefurther assessed the multimerization specificity of thesefusion proteins in a coimmunoprecipitation assay. Our re-

sults demonstrate that the DNA-binding specificity andsequence-specific transcriptional activation of NFKB1 can

be conferred by fusion of 34 N-terminal amino acids ontoRelA and that a single amino acid change within the NFKB1sequence can alter the specificity back to that of RelA withinthe context of this fusion protein. Furthermore, substitutionof four amino acids from NFKB1 into RelA was sufficient tochange the DNA-binding specificity of this mutant protein to

that of NFKB1. These findings demhonstrate that the bindingspecificities of the Rel family of proteins, and their subse-quent ability to activate specific iKB motifs, are conferred bya minimal number of amino acid residues.

MATERIALS AND METHODS

Plasmid constructions. cDNAs encoding the RHD ofNFKB1 (amino acids 1 to 377) and RelA (amino acids 1 to

309) served as templates for construction of the fusion genes.

The fusion genes were generated by a two-step PCR method(22) followed by DNA sequence confirmation. In all cases,

the 5' positive-strand PCR primer contained a synthetic T7promoter and a unique HindIII site to facilitate in vitroexpression and/or enzyme digestion and cloning. Also, thesequence contained nucleotides that maintained a readingframe for cloning into a (His)6 bacterial expression system

(15). For example, in the T7 NFKB1 primer sequence,

5'-TGTAATACGACTCACTATAGGGCAAGCTTAGCCATG GCA GAA GAT GAT CC-3', underlined sequences

correspond to the T7 promoter sequence, the HindIII site isin italics, and 15 nucleotides corresponding to the first fiveamino acids of NFKB1 are shown starting at the ATG. The3' negative-strand PCR primer encodes a uniqueXbaI site aswell as a translational stop codon. After PCR amplification,the product was used directly as a template for in vitrotranscription. Alternatively, the amplicons were cloned intoBluescript SK (Stratagene) or the bacterial expression vec-tor pDS (15) for IPTG (isopropyl-13-D-thiogalactopyrano-side)-inducible overproduction of the chimeric proteins.Expression in mammalian cells was achieved by PCR am-plification of the fusion genes to include nucleotides encod-ing amino acids 309 to 550 of RelA and subsequent ligationinto a cytomegalovirus-directed expression vector as previ-ously described (46). The nomenclature for the fusion genesis detailed in the legend to Fig. 1. Mutagenesis of His-67(H-67) in NFKB1(1-377) was carried out by using comple-mentary oligonucleotides degenerate at three nucleotides[designed as follows: 5'(+)-GGCCCATCC(G/A/T)NNGGTGGACTAC-3' and 5'(-)-GTAGTCCACCNN(C/T/A)GGATGGGCC-3'] and a two-step PCR (22). Point mutationsin NFKB1(1-377), NFKB1(68)-RelA-1, and RelA(1-309)were generated in a similar manner.

In vitro transcription and translation and coimmunoprecip-itation. A 10-A1 volume of each 100-j±l PCR mixture gener-ated with the T7 5' positive-strand primers was used directlyas a transcription template, utilizing T7 RNA polymerase.The in vitro-transcribed RNA was used to program a rabbitreticulocyte translation lysate (Promega) that contained[35SJmethionine. A 2-,ul volume of each programmed lysatewas diluted with 5 ,ul of distilled water and an equal volumeof 2x sodium dodecyl sulfate (SDS) sample buffer. Proteinproducts were resolved on an SDS-10% polyacrylamide gelbefore being subjected to fluorography and subsequent au-toradiography.For coimmunoprecipitation, equivalent amounts of RNA

encoding either NFKB1 or RelA, tagged with an influenzavirus hemagglutinin (HA) epitope (25), were cotranslatedwith RNA encoding the fusion proteins as previously de-scribed (45, 46). A 4-,ul volume of the translation lysate wasincubated with 150 ,ul of immunoprecipitation buffer (20 mMHEPES [N-2-hydroxyethylpiperazine-N'-2-ethanesulfonicacid], pH 7.9; 250 mM NaCl; 4 mM EDTA; 0.1% NonidetP40), 2 pul of anti-HA immune serum (Babco), and 20 ,ul ofGamma-BIND G Sepharose (Pharmacia). After overnightincubation at 4°C, the complexes were washed four timeswith cold immunoprecipitation buffer. Agarose beads con-taining the immunoprecipitated complexes were suspendedin 20 pt1 of 2x SDS sample buffer and heated to 80°C for 10min, and then the protein products were analyzed on anSDS-10% polyacrylamide gel.

Gel mobility shift assays. In vitro-translated or bacteriallyexpressed proteins were used in binding reactions as de-scribed previously (45, 46), with the addition of dithiothreitol(1 mM), Nonidet P-40 (0.1%), and bovine serum albumin (1mg/ml). 32P-labeled probes were prepared as previouslydescribed (32). The KB probe contains the sequence 5'-GTAGGGGACTTTCCGAGCTCGAGATCCTATG-3',which corresponds to the KB motif present in the Ig light-chain and human immunodeficiency virus (HIV) enhancers(27). Probes 50-14 (5'-GTAGGGGGCCTCCCCGGCTCGAGATCCTATG-3') and 65-2 (5'-GTACCGGAAATTCCGGGCTCGAGATCCTATG-3') contain KB sequences that bindselectively to homodimers of NFKB1 and RelA, respec-tively (32).

Cell culture and transfection. Jurkat T cells were main-

VOL. 13, 1993

3852 COLEMAN ET AL.

RxxRxRxxCE

Rel Homology Domain A.I ~~~ ~~~~~~~~~~~~III NFKB1(1-377)

RelA (1-309)

NFKBI(68)/RelA

ReIA(44)/NFKB1

NFKB1(273)/RelA

ReIA(216)/NFKB1

FIG. 1. Schematic representation of NFKB1-RelA fusion genes.The RHD, with regions suggested to be involved in DNA binding(RxxRxRxxCE) and multimerization (dashed lines), is indicated.The fusion proteins were derived from NFKB1(1-377) and RelA(1-309). NFKB1(68)-RelA contains the first 68 amino acids of NFKB1fused to amino acids 45 to 309 of RelA. RelA(44)-NFKB1 containsthe first 44 amino acids of RelA fused to amino acids 69 to 377 ofNFKB1. NFKB1(273)-RelA contains the first 273 amino acids ofNFKB1 fused to amino acids 217 to 309 of RelA. ReIA(216)-NFKB1contains the first 216 amino acids of RelA fused to amino acids 274to 377 of NFKB1.

tained in RPMI 1640 medium containing 10% fetal calf serumand 50 ,ug of gentamicin per ml. For transient transfectionassays, 5 x 106 to 1 x 107 cells were transfected by aDEAE-dextran procedure (42). A 1-,ug sample of each chlor-amphenicol acetyltransferase (CAT) reporter construct,which contains four tandem copies of the Ig-HIV, 50-14, or65-2 KB motifs cloned upstream of the HIV long terminalrepeat and fused to the CAT coding region, was cotrans-fected with 2 jig of the mammalian cytomegalovirus-drivenexpression vectors. Cells were harvested 48 h after transfec-tion, and CAT assays were performed as previously de-scribed (19).

RESULTS

DNA binding and multimerization activities ofNFKBI-RelAfusion proteins. NFKB1 and RelA fusion genes were gener-ated as illustrated in Fig. 1. PCR products were used astemplates for in vitro transcription and translation, and theresulting proteins were analyzed by SDS-polyacrylamide gelelectrophoresis (SDS-PAGE) (Fig. 2A). To assess whetherthese fusion proteins were functional with respect to DNAbinding, the translation lysates were used in gel mobilityshift assays (Fig. 2B). Two of the fusion proteins were ableto bind DNA containing a consensus KB site derived fromthe Ig and HIV enhancers. The first protein, NFKB1(68)-RelA, contained amino acids 1 to 68 of NFKB1 fused toamino acids 46 to 309 of RelA. The second protein,RelA(216)-NFKBl, contained the first 216 amino acids ofRelA fused to amino acids 274 to 377 of NFKB1. By usingtwo previously identified oligonucleotides that demonstratedabsolute selectivity with respect to binding by NFKB1 andRelA (32), the DNA-binding preferences of the fusion pro-teins were examined. NFKB1(68)-RelA bound exclusivelyto the NFKB1-selective oligonucleotide (50-14), while nobinding to the RelA-selective site (65-2) was observed (Fig.2B, lanes 3). In contrast, RelA(216)-NFKB1 bound exclu-sively to the RelA-selective oligonucleotide but not to the

B.KB 65-2

'-I-

1 2 3 4 5 6 1 2 3 4 5 6 1 2 3 4 5 6

FIG. 2. DNA-binding specificity of in vitro-translated proteins.(A) In vitro-transcribed RNAs encoding the proteins depicted in Fig.1 were used to program a rabbit reticulocyte translation lysatecontaining [35S]methionine. Radiolabeled proteins were resolved bySDS-10% PAGE. M, molecular weight markers (in thousands). (B)Gel mobility shift analysis of in vitro-translated proteins. 32P-labeledoligonucleotides corresponding to the KB motif present in the Ig andHIV enhancers (KB), the NFKB1-selective site (50-14), and theRelA-selective site (65-2) were used as probes. Complexes wereresolved by 4% PAGE and visualized by autoradiography. Lanenumbers correspond to the positions of the proteins listed in panelA. The upper band seen with the KB probe corresponds to endoge-nous DNA-binding activity present in the rabbit reticulocyte lysate.

NFKB1-selective site (Fig. 2B, lanes 6). The other twofusion proteins, though expressed, were not active in DNAbinding with any of the KB sites tested.To analyze the ability of the fusion proteins to multimerize

with NFKB1 or RelA, coimmunoprecipitation assays wereperformed (Fig. 3). We have observed that the association ofNFKB1 with RelA in a heterodimeric complex is strongerthan the interaction of either homodimer (lOa). Influenzavirus HA epitope-tagged full-length proteins correspondingto NFKB1 (HA-NFKB1) and RelA (HA-RelA) werecotranslated with the C-terminally truncated fusion proteins.HA antibody was used to coimmunoprecipitate the proteincomplexes. The strong association of RelA(1-309) withHA-NFKB1 is clearly evident (Fig. 3A, lanes 3). As ex-pected, the homodimerization of NFKB1(1-377) with HA-NFKB1 was much weaker (Fig. 3A, lanes 2). NFKB1(68)-RelA dimerized very well with HA-NFKB1 (Fig. 3A, lanes4), suggesting that this protein maintains a RelA-like multi-merization domain. While not able to bind DNA, RelA(44)-NFKB1 interacts with HA-NFKB1 in a manner similar tothat of NFKB1(1-377) (Fig. 3A, lanes 2 and 5), suggestingthat this protein maintains an NFKB1-like multimerizationdomain. Interestingly, RelA(216)-NFKB1, which containsthe first 216 amino acids of RelA, also multimerized stronglywith HA-NFKB1 (Fig. 3A,' lanes 7).When HA-RelA was used to coimmunoprecipitate the

fusion proteins, different results were obtained. Consistentwith the results described above, the formation of a het-erodimer between NFKB1(1-377) and HA-RelA was stron-ger than the formation of a homodimer of RelA(1-309) withHA-RelA (Fig. 3B, lanes 2 and 3). RelA(44)-NFKB1, which

MOL. CELL. BIOL.

NFKB1-SELECTIVE DNA BINDING 3853

'bI.NA N,

%NA)$Ot

A.~~~~~~~~~~~~~~-69

*3~33 *HA-NFKB1 - -46

-30

B.HA-RelA - W _ 69

ii ~~~~46

-301 2 3 4 5 6 7 1 2 3 4 5 6 7

(-)4- ANTIBODY - *> (+)

FIG. 3. Multimerization analysis of the fusion proteins withHA-NFKB1 (A) and HA-RelA (B). Equal amounts of RNA encod-ing HA-tagged NFKB1 or HA-tagged RelA and the proteins de-picted in Fig. 1 were translated in a rabbit reticulocyte lysate asdescribed in the text. Cotranslation products (without antibody) areindicated on the left. Results of coimmunoprecipitation analysis(with antibody) are indicated on the right. , translation productfrom HA-NFKB1 or HA-ReLA alone; M, molecular weight markers(in thousands).

does not bind DNA and interacts weakly with HA-NFKB1,multimerized strongly with HA-RelA (Fig. 3B, lanes 5),suggesting the presence of an NFKB1-like multimerizationdomain. NFKB1(273)-RelA, which does not bind DNA, andRelA(216)-NFKB1 multimerized with HA-RelA to a similarextent (Fig. 3B, lanes 6 and 7). This interaction appears to bestronger than that of a homodimer but weaker than theassociation of the NFKB1-RelA heterodimer.DNA-binding properties of bacterially expressed fusion pro-

teins. To overproduce the fusion proteins, PCR-amplifiedmaterial corresponding to the two proteins functional inDNA binding [NFKB1(68)-RelA and RelA(216)-NFKB1;Fig. 2B] was subcloned into a bacterial expression vector.Two clones of each fusion protein were used to prepare cellextracts after induction by IPTG (15). Binding activitypresent in crude bacterial lysates was determined by gelmobility shift assays using the KB site and the NFKB1- andRelA-selective sites (Fig. 4). Bacterial extracts from bothRelA(216)-NFKB1 clones were able to bind the KB andRelA-selective sites but not the NFKB1-selective site (Fig.4, lanes 5 and 6), consistent with the results presented in Fig.2B. Bacterial lysates from both NFKB1(68)-RelA clonesbound the KB probe, albeit with slightly different mobilities(Fig. 4, lanes 3 and 4). Unexpectedly, these two proteinsdiffered in their selective binding to the NFKB1- and RelA-preferred DNA-binding sites. NFKB1(68)-RelA-1 bound tothe NFKB1- but not the RelA-selective site, consistent withthe results obtained with in vitro-translated proteins, whileNFKB1(68)-RelA-2 bound the RelA- but not the NFKB1-selective site. The difference between the binding specifici-ties of the two NFKB1(68)-RelA proteins can be attributedto a point mutation that alters a single amino acid residuewithin the NFKB1 portion of this fusion protein (see below).

Transcriptional activation of fusion proteins in vivo. Toexamine the functions of the fusion proteins in vivo, thegenes were cloned into a mammalian expression vector inwhich the cytomegalovirus promoter directs the expression

FIG. 4. DNA-binding specificity of bacterially expressed pro-teins. Transformants containing the fusion genes encoding theproteins active in DNA binding were identified. After induction withIPTG, bacterial cell extracts were prepared and used in gel mobilityshift assays. Binding specificity was determined by using 32P-labeledoligonucleotides containing the KB, NFKB1-selective (50-14), orRelA-selective (65-2) site.

of the fusion genes. The fusion genes were modified byaddition of amino acids 309 to 550 of RelA, a sequence whichcontains the transcriptional activation domain (46). Theseconstructs were cotransfected into Jurkat T cells with vari-ous KB-CAT reporter constructs in which CAT gene expres-sion is dependent upon transcriptional activation by NF-KB-Rel proteins. The reporter constructs contain four tandemcopies of the NFKB1- or RelA-selective KB binding sites anddemonstrate the same protein-binding specificity as theoligonucleotides used in the gel mobility shift assays (32). Allthree fusion proteins, NFKB1(68)-ReIA-1, NFKB1(68)-RelA-2, and RelA(216)-NFKB1, were able to transactivatethe p(Ig/HIVKB)4CAT construct (Fig. 5). NFKB1(68)-RelA-1 was able to activate the p(50-14KB)4CAT constructbut not the p(65-2KB)4CAT construct (Fig. 5, lane 4), con-sistent with the observed in vitro binding specificity. ThoughNFKB1(68)-RelA-2 was able to activate the p(Ig/HIVKB)4CAT construct, selective activation of the p(65-2KB)4CAT reporter was not observed in this cotransfectionassay (Fig. 5, lane 5). On the other hand, the RelA(216)-NFKB1 protein with the transactivation domain was apotent activator of both p(Ig/HIVKB)4CAT and p(65-2KB)4CAT gene expression (Fig. 5, lane 6).

In summary, the DNA-binding activities of the fusionproteins produced in vitro, in bacteria or in mammalian cells,were examined. RelA(216)-NFKB1 gave consistent results

VOL. 13, 1993

3854 COLEMAN ET AL.

IA <'- %N

N\OI%

p(ig/HIVKB)4 CAT

p(50-14 KB)4 CAT

p(65-2K1B)4 CAT

1 2 3 4 5 6

FIG. 5. Transcriptional activation by the NFKB1-RelA fusionproteins in Jurkat T cells. The fusion proteins active in DNA bindingwere further modified to contain the transcriptional activationdomain present in RelA. Jurkat T cells were transfected with theindicated CAT reporter plasmids containing four tandem copies ofthe KB [p(Ig/HIVKB)4CAT], NFKB1-selective [p(50-14KB)4CAT],or ReIA-selective [p(65-2KB)4CAT] motif (32). Cytomegalovirusexpression vectors encoding a previously described NFKB1-RelAfusion protein (46), RelA, or the fusion proteins identified hereinwere assayed for their ability to selectively activate CAT geneexpression. Lane 1, reporter alone.

independent of the source of the protein. When in vitrotranslated or bacterially expressed, this chimeric protein wasable to bind the KB and RelA-selective sites in gel mobilityshift assays. Consistent with these observations was theability of this protein to transactivate the RelA-selectiveCAT construct, p(65-2KB)4CAT, in the transient transfectionassay. Although in vitro-translated NFKB1(68)-RelA dem-onstrated that it was able to bind the KB and NFKB1-selective oligonucleotides, protein extracts from two inde-pendent bacterial clones showed altered DNA-bindingspecificity. NFKB1(68)-RelA-1 bound the KB and NFKB1-selective oligonucleotides, while NFKB1(68)-RelA-2 boundthe KB and RelA-selective oligonucleotides. Though thespecificity of clone NFKB1(68)-RelA-1 was maintained inmammalian cells, we were not able to demonstrate theselective activation of p(65-2KB)4CAT by NFKB1(68)-RelA-2. This could be due to a reduced DNA-bindingcapability of this protein, as suggested by the weak bindingof the bacterially expressed protein (Fig. 4, lane 4).A single amino acid change alters the DNA-binding speci-

ficity of the NFKB1-RelA fusion protein. To identify the causeof the altered binding specificity of the two bacteriallyexpressed NFKB1(68)-RelA proteins, DNA sequence anal-ysis of the N termini of both clones, through the fusion andinto the RelA coding region, was performed. An A--Gtransition in NFKB1(68)-RelA-2 resulted in a His-to-Argchange at position 67 (H67R) of the NFKB1 sequence (Fig.6A).The importance of residue 67 for DNA binding by wild-

A.NFKB1 __ R G F R F R Y V C E G P S H G

NFKB1 (68)/ReIA-1 - - H

CATCGT

NFKB1 (68)/Re]A-2 _ ..-._ R

RelA _ _ R G M R F R Y K C E G R S A G

B.

C.

FIG. 6. Identification of a single amino acid change involved inaltering DNA-binding specificity. (A) Amino acid sequence of theproposed DNA-binding region. The DNA sequence highlighting theA- G transition in NFKB1(68)-ReIA-2 which resulted in a His-to-Arg (H67R) change at position 67 of the NFKB1 sequence isindicated (arrow). (B) Translation products of mutants with muta-tions made at position 67 in NFKB1(1-377) and NFKB1(68)-RelA-1. NFKB1 H-67 degeneracy results in the expression of amixed pool of proteins with 18 different amino acid possibilities atposition 67. Changes of H67R, H67K, and H67A were made inNFKB1(1-377) (NFKB1 H67 R, K, and A) or NFKB1(68)-RelA-1(-1 H67 R, K, and A). An equivalent change in RelA(1-309) (RelAA43 R) was also made. (C) DNA-binding specificity determined byusing the KB, NFKB1-selective (50-14), and RelA-selective (65-2)oligonucleotide probes. The upper band seen with the KB probecorresponds to endogenous DNA-binding activity present in therabbit reticulocyte lysate.

*

S * 1

* .

0 0

MOL. CELL. BIOL.

NFKB1-SELECTIVE DNA BINDING 3855

type NFKB1 was determined by using degenerate oligonu-cleotides encoding 17 different amino acid substitutions atthis position in NFKB1. Codons for His (present at thisposition in the wild-type protein), Pro, and Gln were ex-cluded because of the degeneracy of the oligonucleotides atthis position. Transcription and translation of this templatewould result in the expression of a mixed pool of proteinscontaining different amino acids at position 67. Thoughprotein was made (Fig. 6B), DNA binding was completelyabolished by mutagenesis at this position (Fig. 6C, lane 3).Three specific changes in both NFKB1(1-377) andNFKB1(68)-RelA-1 at position 67 were then assayed. His-67to Arg (H67R) substitutes an Arg into either NFKB1(1-377)or NFKB1(68)-ReIA-1. This change in the fusion genegenerated a protein identical to NFKB1(68)-RelA-2.Changes from His-67 to Lys (H67K) and His-67 to Ala(H67A) were also introduced. The Lys (K) substitutionaddresses the importance of phosphate contact by the pro-tein (40), while Ala (A) is the amino acid present at theequivalent position in RelA (Fig. 6A). Similarly, because theH67R substitution is tolerated in the NFKB1(68)-RelA-2fusion protein and the protein maintains RelA selectivity, thecorresponding change of Ala-43 to Arg (A43R) in RelA(1-309) was generated. Each of these proteins was expressed invitro (Fig. 6B). All single amino acid changes in the contextof NFKB1(1-377) abolished DNA binding. As expected,substitution of H67R in NFKB1(68)-RelA-1 (-1 H67R)resulted in a chimeric protein that could bind the KB andRelA-selective sites but no longer bound the NFKB1-selec-tive site (Fig. 6C, lane 5). Substitution of H67K inNFKB1(68)-RelA-1 resulted in a protein with reduced DNA-binding affinity for both the KB and RelA-selective sites (Fig.6C, lane 7). The H67A change abolished DNA binding of thefusion protein, supporting a role for a basic amino acidresidue at this position. Mutagenesis of Ala-43 toArg (A43R)in RelA(1-309) resulted in a mutant protein that maintainedits DNA-binding specificity, though with a somewhat re-duced affinity compared with that of wild-type ReLA(1-309)(Fig. 6C, lanes 10 and 11). This supports the idea that an Argsubstitution can be tolerated for RelA-selective DNA bind-ing.Because multimerization is a prerequisite for DNA bind-

ing, we next examined whether the single amino acidchanges in NFKB1(1-377) affected multimerization. Each ofthe NFKB1 point mutants (H67R, H67K, and H67A) wasefficiently coimmunoprecipitated with HA-RelA (Fig. 7A).When the cotranslated proteins were incubated with a KBprobe, only wild-type NFKB1(1-377) and HA-RelA gener-ated an intermediate complex in gel mobility shift assays(Fig. 7B, lane 3). Though the NFKB1(1-377) point mutantswere able to form strong associations with HA-RelA, noneof these cotranslations resulted in formation of an interme-diate complex in the gel shift assay (Fig. 7, lane 3 versuslanes 5, 7, and 9). Taken together, these results demonstratethat mutations at His-67 in NFKB1 do not affect the abilityof these proteins to multimerize with ReLA. However, asthese mutations clearly abolish DNA binding, the datademonstrate that two functional DNA-binding domains, onecontributed by each subunit of the heterodimer, are requiredfor DNA binding.To further localize the region of NFKB1 involved in DNA

binding, we created two N-terminal deletions of 13 and 34amino acids corresponding to the second and third in-framemethionine residues, respectively. Amino-terminal deletionsof NFKB1(1-377), NFKB1(68)-RelA-1, and NFKB1(68)-RelA-2 resulted in the expression of proteins with reduced

A.

HA RelA +

A_

Ls

,~~~~~

111in1::j*4A-RSAR ____ _

69at_ -.-46

- 30

(-)*----ANTIBODY- -()

1 2 3 4 5 6 7 8 9

FIG. 7. Separation of multimerization and DNA-binding proper-ties of NFKB1. (A) RNAs encoding wild-type NFKB1(1-377) andthose with single amino acid changes (H67R, H67K, and H67A)were translated alone (-) or with HA-RelA (+) as indicated.Translation products are shown in the left panel, and results ofcoimmunoprecipitation analysis with anti-HA antibody are shown inthe right panel. (B) Gel mobility shift assay using a 2P-labeled KBoligonucleotide and the in vitro-translated proteins. Complexesrepresenting homodimers of HA-RelA and NFKB1(1-377) or het-erodimers of HA-RelA-NFKB1(1-377) are indicated by arrows.

molecular weights, as determined by SDS-PAGE analysis(Fig. 8A). As shown in Fig. 8B, amino-terminal deletions hadno effect on the DNA-binding specificity of either NFKB1(1-377) or the NFKB1(68)-RelA fusion proteins. Because thethird methionine is at position 35 in NFKB1(1-377) and theNFKB1(68)-RelA fusion proteins, the region of NFKB1involved in conferring DNA-binding specificity is limited toa 34-amino-acid stretch between residues 35 and 68.

Alteration of the DNA-binding specificity of RelA. To thispoint, all alterations in DNA-binding specificity were dem-onstrated in the context of NFKB1 and RelA fusion pro-teins. Several observations allowed us to predict that mini-mal amino acid changes in the context of RelA should beable to alter DNA-binding specificity. First, N-terminaldeletion of NFKB1(68)-RelA-1 to the third in-frame methi-onine resulted in expression of a fusion protein whichmaintained both DNA binding and selectivity (Fig. 8B).Second, a single amino acid change (H67R) could alter thespecificity of this fusion protein so that it bound the RelA-selective site. Finally, comparison of the amino acid se-quences of NFKB1 and RelA in this region demonstratedonly seven amino acid differences (Fig. 9A). On the basis ofthese observations, we generated a series of mutant RelAproteins with minimal amino acid substitutions, as shown inFig. 9A. The in vitro-translated proteins were used in gelmobility shift assays with the KB site and NFKB1- andRelA-selective sites. Mutation of V20L, E21Q, and I23L(RelA-I) had no effect on DNA binding or specificity (Fig.9B, lane 3). M32F and K37V substitutions (RelA-II) de-creased DNA binding but had no effect on selectivity in thatthis protein still maintained the ability to bind the KB and

VOL. 13, 1993

3856 COLEMAN ET AL.

=\ .j--b f

1 2 3 4 5 6 7 8 9

A.NFKB1RelA

RelA - IReIA - 11RelA - III

RelA - 1IIIRelA - J.IIIReIA - II/III

43 589,PYLQILEOPKQ)RGFHFRYVCEGPSHG 4

PYVE! IEQPKQRGMRFRYKCEGRSAG.~~~. _..., j1i III

-LQ IL.LFRFRYV

.- PSH.

LQ L --

LQ I L....FR FRY V.

PPSHFFRFRYV-V PSHH

B.

KB I

1 ".U. -

50-141 2 3 4 5 6 7 8 9

FIG. 8. Delineation of a minimal NFKB1 DNA-binding domain.(A) In vitro translation products of NFKB1(1-377), NFKB1(68)-RelA-1, and NFKB1(68)-RelA-2 from the first (Ml), second (M2),and third (M3) in-frame methionines. (B) Gel mobility shift analysisof in vitro-translated N-terminally truncated proteins using 32p-labeled oligonucleotides corresponding to the KB, NFKB1-selective(50-14), and ReIA-selective (65-2) sites.

RelA-selective sites (Fig. 9B, lane 4). R41P and A43Hchanges (RelA-III) led to expression of a protein with alteredDNA-binding specificity. Though this protein maintained theability to bind the KB and RelA-selective sites, it alsoacquired the ability to bind the NFKB1-selective site (Fig.9B, lane 5). These results suggested that complete DNA-binding selectivity might be achieved by the combination ofmore than one group of these changes. Proteins containingcombinatorial changes were therefore tested.The first of these proteins, RelA-I/II, contains five amino

acid changes. These changes have no effect on DNA-bindingspecificity but rather elicit a decrease in affinity equivalent tothat seen with RelA-II (Fig. 9B, lanes 4 and 6). RelA-I/III,which contains four amino acid changes, bound DNA in amanner similar to that of RelA-III (Fig. 9B, lanes 5 and 7).This is consistent with the idea that amino acids at positions20, 21, and 23 do not contribute to DNA binding. The proteincontaining the M32F-K37V and R41P-A43H changes (RelA-II/III) bound the KB and NFKB1-selective sites but not theRelA-selective site (Fig. 9B, lane 8). Therefore, a maximumof four amino acid changes in the context of intact RelA wereable to completely alter the DNA-binding specificity of thisprotein to favor interaction with this NFKB1-selectiveDNA-binding site.

65-2

1 2 3 4 5 6 7 8FIG. 9. Alteration of the DNA-binding specificity of ReIA. (A)

Amino acid sequences of NFKB1 and RelA in the region involved inDNA binding. Residues of NFKB1 substituted into the correspond-ing positions of RelA appear in boldface type. Clone designationsare listed to the left (i.e., RelA-I contains three amino acid changes,V21L, E22Q, and I24L). (B) Gel mobility shift analysis of invitro-translated RelA mutant proteins using 32P-labeled oligonucle-otides corresponding to the KB, NFKB1-selective (50-14), andRelA-selective (65-2) sites.

DISCUSSION

The genes for the NF-KB-Rel family of proteins encode agroup of well-studied transcription factors. Members of thefamily share a highly conserved stretch of approximately 300amino acids termed the RHD (18, 33). Transcriptional acti-vation is generally associated with the unique C termini ofthese proteins, whereas multimerization and DNA-bindingactivities have been colocalized within the RHD (1, 3).Unlike other well-characterized transcription factors, mem-bers of the NF-KB-Rel family do not contain motifs normallyassociated with these functions (i.e., leucine zipper, zincfinger, or helix-loop-helix motifs) (26, 36). The recent iden-tification of specific DNA sequence motifs that demonstrate

A.

B.

KB

50-14

- ..

MOL. CELL. BIOL.

NFKB1-SELECTIVE DNA BINDING 3857

preferential binding to three Rel family members provided areagent for the identification of residues involved in DNA-binding specificity. To identify these residues, NFKB1 andRelA fusion proteins were constructed and their ability tointeract with selective DNA-binding sites was examined.The first fusion protein, NFKB1(68)-RelA, which contains

68 amino-terminal residues of NFKB1 fused to residues 45 to309 of RelA, bound with the specificity of NFKB1, suggest-ing that residues within the amino-terminal domain ofNFKB1 confer DNA-binding specificity. This result is inaccord with earlier findings which showed that residueswithin the RxxRxRxxC domain of v-rel (amino acids 19 to28) were required for DNA binding (31). More recently,mutational studies have further defined a homologous region(amino acids 55 to 72) as critical for DNA binding by NFKB1(9, 54). A second chimera, ReIA(216)-NFKBl, in whichresidues 1 to 216 of RelA were fused to residues 274 to 377of NFKB1, bound the KB site and also demonstrated adistinct preference for the RelA-selective DNA motif. Thesefindings suggest that RelA-specific DNA binding requires anadditional protein sequence, specifically, amino acids lo-cated between positions 45 and 216 of the RelA protein.Indeed, Bressler et al. (9) have shown that mutations in twodistinct regions of NFKB1 abrogate DNA binding but havelittle effect on multimerization.The inability of RelA(44)-NFKB1 and NFKB1(273)-RelA

to bind DNA might be interpreted in several ways. Thefusion sites, though at the same relative positions as thoseused to generate proteins active in DNA binding, may induceconformational changes not suitable for DNA binding. Al-ternatively, DNA-binding activity may be conferred byresidues in multiple domains, and the alignment of theseresidues in the fusion proteins RelA(44)-NFKB1 andNFKB1(273)-RelA may have been significantly altered.Though neither of these proteins was functional in DNAbinding, both were able to demonstrate distinct multimeriza-tion capabilities. This implies that while DNA binding andmultimerization are colocalized, these functions likely residein separable domains within the RHD.Through the fortuitous generation of a mutation in the

NFKB1(68)-RelA fusion protein, a single amino acid changewas found to alter DNA-binding specificity. NFKB1(68)-RelA-2 contains an A--G transition which resulted in thesubstitution of His-67 by Arg. This protein no longer boundto the NFKB1-selective motif but, rather, acquired theability to bind the RelA-selective site. This result wasconfirmed and expanded by making designed changes inNFKB1(1-377) and NFKB1(68)-ReJA-1. Mutation of His-67to Arg, Lys, or Ala in NFKB1(1-377) produced proteinsinactive in DNA binding but competent in multimerizationwith RelA. Substitution of H67R in NFKB1(68)-RelA-1resulted in expression of a protein with the predicted DNA-binding specificity for a RelA-selective site, thereby confirm-ing the role of this amino acid in determining DNA-bindingspecificity. Though a Lys substitution (H67K) could betolerated in this fusion protein, substitution by Ala (H67A)was not. The importance of the equivalent residue in RelAwas also tested. Whereas the mutation of His-67 to Arg inNFKB1(1-377) resulted in complete loss of binding, thecorresponding change in RelA of Ala-44 to Arg had aminimal effect on DNA-protein interaction. Taken together,these observations suggested that it should be possible tomutate RelA in such a way as to alter its DNA-bindingspecificity. Comparison of a 26-amino-acid stretch ofNFKB1 (P-43--G-68) and RelA (P-19--)G-44) demonstratedthat these two proteins were 73% identical in this region,

containing only 7 amino acid differences. Complete alter-ation in DNA-binding specificity was achieved by the sub-stitution of four amino acids from NFKB1 (M32F-K37V andR41P-A43H) into RelA. This observation, together with theresults obtained with the NFKB1(68)-RelA chimera, sug-gests that although a single residue can alter DNA-bindingspecificity, DNA binding is most likely conferred by multiplecontacts of protein with DNA. Interestingly, and in supportof the idea that additional amino acids are required forRelA-selective DNA binding, a mutant NFKB1 proteincontaining the equivalent four amino acid substitutions fromRelA (F56M, V61K, P65R, and H67A) resulted in theexpression of a protein unable to bind DNA (data notshown).

It has been established that binding of Rel family proteinsto DNA requires either homo- or heterodimer formation.Earlier studies with a naturally occurring splice variant ofRelA that lacks amino acids 222 to 231 identified this regionas important in heterodimer formation (46). The exactboundaries of this suspected dimerization domain within theRel family members have not yet been established. The lackof DNA binding by the NFKB1 point mutants and even thealtered specificity of the chimeric fusion proteins couldreflect alterations in dimerization properties. To address thisissue, the NFKB1 His-67 mutants were examined. Theability of HA-tagged RelA to coimmunoprecipitate theNFKB1 His-67 mutants indicates that the multimerizationfunction of these proteins was not affected. Gel mobility shiftassays with these cotranslated proteins demonstrated thatcomplexes formed with these mutants were deficient in DNAbinding. Previous studies had suggested that two DNA-binding domains were required for efficient DNA binding.Logeat et al. (35) showed that deletion of residues 12 to 200from NFKB1 resulted in the expression of a trans-dominantnegative in that this protein maintained its ability to multi-merize but dimers formed with this protein were not active inDNA binding. Furthermore, it was demonstrated thatcotranslated or corenatured NF-KB (the NFKB1-RelA het-erodimer) was competent in binding a KB site but ineffectivewhen either the NFKB1- or the RelA-selective oligonucleo-tides were used (32). While heterodimer formation of theNFKB1(1-377) point mutants with HA-RelA could beclearly demonstrated, no intermediate complex representingNF-KB-like DNA binding was observed, supporting theargument for the involvement of both subunits in DNAbinding. Whether the multimerization properties of the fu-sion proteins are altered relative to those of the wild-typeprotein raises an issue more difficult to address. However,several observations suggest that the multimerization prop-erties of the fusion proteins do indeed reflect the residuesthat are present within the carboxy-terminal half of the RHDand are not affected by residues within the amino-terminalregion which confer DNA-binding specificity. This idea isbased on the observation that HA-NFKB1 forms a morestable association with RelA(1-309) than with NFKB1(1-377). Similarly, the association of HA-RelA withNFKB1(1-377) is stronger than its association with RelA(1-309). Therefore, the observations that HA-NFKB1 coim-munoprecipitates NFKB1(68)-RelA better than RelA(44)-NFKB1 and that HA-RelA coimmunoprecipitates RelA(44)-NFKB1 better than NFKB1(68)-ReLA support the premisethat the multimerization properties of these chimeras reflectthe contribution of residues within the carboxy termini of theRHD of these proteins.The fusion proteins generated at the downstream sites

gave ambiguous results in the co-immunoprecipitation as-

VOL. 13, 1993

3858 COLEMAN ET AL.

says. In these constructs, NFKB1(273)-RelA and RelA(216)-NFKB1, the fusions were made proximal to the region inRelA tentatively identified as critical for dimerization withNFKB1. These proteins demonstrated no clear dimerizationpreference for either HA-NFKB1 or HA-RelA. One possibleexplanation for the lack of specific dimerization would bethat the fusions were made in the middle of a bipartitemultimerization domain. The multimerization domains inthese fusion proteins would therefore share similarities witheach of the parental proteins, as is the case for these twochimeras.

In summary, regions involved in DNA binding and multi-merization of NFKB1 and RelA have been identified.NFKB1-selective DNA binding can be conferred upon RelAby inclusion of as few as 34 amino acids from NFKB1.RelA-selective DNA binding requires more than the 44N-terminal amino acids. The multimerization domain,though not clearly defined, seems to be associated withresidues nearer the carboxy terminus of the RHD. Moreimportantly, the results reported here suggest that minimalamino acid changes in the N terminus of the RHD can haveprofound effects on DNA-binding specificity. This idea issupported by recent studies of two different basic helix-loop-helix proteins. In the case of the yeast protein PHO4,binding specificity is altered by a single Glu-to-Asp substi-tution (13). More striking was the study by Davis andWeintraub (lla), which demonstrated that replacement ofthree amino acids in the non-tissue-specific basic helix-loop-helix protein E12 resulted in a protein capable of causingmyogenic transformation. In the case of the NFKB1(68)-RelA fusion protein discussed here, we found that a singleamino acid change of His to Arg altered DNA-bindingspecificity. Because the sequences of the NFKB1-selectiveand RelA-selective sites are known, the specific aminoacid-nucleotide contacts required for DNA-binding specific-ity by Rel-related proteins can now be determined. Further-more, because different DNA-binding motifs have beenshown to be present in promoters for cellular genes (1, 3),single amino acid changes that alter the DNA-binding spec-ificity of the Rel-related proteins could lead to transactiva-tion of different subsets of target genes responsive to the Relfamily members.

REFERENCES1. Baeuerle, P. 1991. The inducible transcription activator NF-KB:

regulation by distinct protein subunits. Biochem. Biophys. Res.Commun. 1072:63-80.

2. Baeuerle, P. A., and D. Baltimore. 1988. IKB: a specific inhibitorof the NF-KB transcription factor. Science 242:540-546.

3. Baeuerle, P. A., and D. Baltimore. 1991. The physiology of theNF-KB transcription factor. Hormonal control regulation ofgene transcription. Mol. Aspects Cell. Regul. 6:409-432.

4. Ballard, D., E. Dixon, N. Peffer, H. Bogerd, S. Doerre, B. Stein,and W. Greene. 1992. The 65-kDa subunit of human NF-KBfunctions as a potent transcriptional activator and a target forv-Rel-mediated repression. Proc. Natl. Acad. Sci. USA 89:1875-1879.

5. Ballard, D. W., W. H. Walker, S. Doerre, P. Sisra, J. A. Molitor,E. P. Dixon, N. J. Peffer, M. Hannink, and W. C. Greene. 1990.The v-rel oncogene encodes a KB enhancer binding protein thatinhibits NF-KB function. Cell 63:803-814.

6. Beg, A., S. M. Ruben, R. Scheinman, S. Haskill, C. Rosen, andA. Baldwin, Jr. 1992. IKB interacts with the nuclear localizationsequences of the subunits of NF-KB: a mechanism for cytoplas-mic retention. Genes Dev. 6:1899-1913.

7. Blank, V., P. Kourilsky, and A. Israel. 1991. Cytoplasmicretention, DNA binding and processing of the NF-KB p50precursor are controlled by a small region in its C-terminus.

EMBO J. 10:4159-4167.8. Bours, V., J. Vilalobos, P. R. Burd, K. Kelly, and U. Siebenlist.

1990. Cloning of a mitogen-inducible gene encoding a KBDNA-binding protein with homology to the rel oncogene and tocell-cycle motifs. Nature (London) 348:76-90.

9. Bressler, P., K. Brown, W. Timmer, V. Bours, U. Siebenlist, andA. S. Fauci. 1993. Mutational analysis of the p50 subunit ofNF-KB and inhibition of NF-KB activity by trans-dominant p50mutants. J. Virol. 67:288-293.

10. Bull, P., K. L. Morley, M. F. Hoekstra, T. Hunter, and I. M.Verma. 1990. The mouse c-rel protein has an N-terminal regu-latory domain and a C-terminal transcriptional transactivationdomain. Mol. Cell. Biol. 10:5473-5485.

10a.Coleman, T. A., S. M. Ruben, and C. A. Rosen. Unpublishedresults.

11. Davis, N., S. Ghosh, D. L. Simmons, P. Tempst, H.-C. Liou, D.Baltimore, and H. R. Bose, Jr. 1991. Rel-associated pp4O: aninhibitor of the Rel family of transcription factors. Science253:1268-1271.

11a.Davis, R., and H. Weintraub. 1992. Acquisition of myogenicspecificity by replacement of three amino acid residues fromMyo D into E12. Science 256:1027-1030..

12. Fan, C.-M., and T. Maniatis. 1991. Generation of p5O subunit ofNF-KB by processing of p105 through an ATP-dependent path-way. Nature (London) 354:395-398.

13. Fisher, F., and C. R. Goding. 1992. Single amino acid substitu-tions alter helix-loop-helix protein specificity for bases flankingthe core CANNTG motif. EMBO J. 11:4103-4109.

14. Fujita, T., G. Nolan, S. Ghosh, and D. Baltimore. 1992. Inde-pendent modes of transcriptional activation by the p50 and p65subunits of NF-KB. Genes Dev. 6:775-787.

15. Gentz, R., C. H. Chen, and C. A. Rosen. 1989. Bioassay fortrans-activation using purified human immunodeficiency virustat-encoded protein: trans-activation requires mRNA synthesis.Proc. Natl. Acad. Sci. USA 86:821-824.

16. Ghosh, S., and D. Baltimore. 1990. Activation in vitro of NF-KBby phosphorylation of its inhibitor IKB. Nature (London) 344:678-682.

17. Ghosh, S., A. M. Gifford, L. B. Riviere, P. Tempst, G. P. Nolan,and D. Baltimore. 1990. Cloning of the p5ODNA binding subunitof NF-KB: homology to rel and dorsal. Cell 62:1019-1029.

18. Gilmore, T. D. 1990. NF-KB, KBF1, dorsal, and related mat-ters. Cell 62:841-843.

19. Gorman, C. M., L. F. Moffat, and B. H. Howard. 1982.Recombinant genomes which express chloramphenicol acetyl-transferase in mammalian cells. Mol. Cell. Biol. 2:1044-1051.

20. Haskill, S., A. A. Beg, S. M. Tompkins, J. S. Morris, A. D.Yurochko, A. Sampson-Johannes, K. Mondal, P. Ralph, andA. S. Baldwin, Jr. 1991. Characterization of an immediate-earlygene induced in adherent monocytes that encodes IKB-likeactivity. Cell 65:1281-1289.

21. Henkel, T., U. Zabel, K. Zee, J. Muller, E. Fanning, and P.Baeuerle. 1992. Intramolecular masking of the nuclear locationsignal and dimerization domain in the precursor for the p50NF-KB subunit. Cell 68:1121-1133.

22. Higuchi, R., B. Krummel, and R. Saiki. 1988. A general methodof in vitro preparation and specific mutagenesis of DNA frag-ments: study of protein and DNA interactions. Nucleic AcidsRes. 16:7351-7367.

23. Inoue, J.-I., L. Kerr, A. Kakizuka, and I. Verma. 1992. IKB-y, a70 kd protein identical to the C-terminal half of pllO NF-KB: anew member of the IKB family. Cell 68:1109-1120.

24. Ip, Y. T., R. Kraut, M. Levine, and C. Rushlow. 1991. Thedorsal morphogen is a sequence-specific DNA-binding proteinthat interacts with a long-range repression element in Dros-ophila. Cell 64:439-446.

25. Janknecht, R., G. Martynoff, J. Lou, R. Hipskind, A. Nordheim,and H. Stunnenberg. 1991. Rapid and efficient purification ofnative histidine-tagged protein expressed by recombinant vac-cinia virus. Proc. Natl. Acad. Sci. USA 88:8972-8976.

26. Johnson, P., and S. McKnight. 1989. Eukaryotic transcriptionalregulatory proteins. Annu. Rev. Biochem. 58:799-839.

27. Kawakami, K.-T., C. Scheidereit, and R. G. Roeder. 1988.

MOL. CELL. BIOL.

NFKB1-SELECTIVE DNA BINDING 3859

Identification and purification of a human immunoglobulin-enhancer-binding protein (NF-KB) that activates transcriptionfrom a human immunodeficiency virus type I promoter in vitro.Proc. Natl. Acad. Sci. USA 85:4700-4704.

28. Kerr, L. D., J.-I. Inoue, N. Davis, E. Link, P. A. Baeuerle, H. R.Bose, Jr., and I. M. Verma. 1991. The rel-associated pp4Oprotein prevents DNA binding of Rel and NF-KB: relationshipwith IKB-,B and regulation by phosphorylation. Genes Dev.5:1464-1476.

29. Kieran, M., V. Blank, F. Logeat, J. Vandekerchove, F. Lott-speich, 0. LeBail, M. B. Urban, P. Kourilsky, P. A. Baeuerle,and A. Israel. 1990. The DNA binding subunit of NF-KB isidentical to factor KBF1 and homologous to the rel oncogeneproduct. Cell 62:1007-1018.

30. Kretzschmar, M., M. Meisterenst, C. Scheidereit, G. Li, and R.Roeder. 1992. Transcriptional regulation of the HIV-1 promoterby NF-KB in vitro. Genes Dev. 6:761-774.

31. Kumar, S., A. B. Rabson, and C. Gelinas. 1992. TheRxxRxRxxC motif conserved in all RelIKB proteins is essentialfor the DNA-binding activity and redox regulation of the v-Reloncoprotein. Mol. Cell. Biol. 12:3094-3106.

32. Kunsch, C., S. M. Ruben, and C. A. Rosen. 1992. Selection ofoptimal KB/Rel DNA-binding motifs: interaction of both sub-units of NF-KB with DNA is required for transcriptional activa-tion. Mol. Cell. Biol. 12:4412-4421.

33. Lenardo, M. J., and D. Baltimore. 1989. NF-KB: a pleiotropicmediator of inducible and tissue-specific gene control. Cell58:227-229.

34. Lion, H.-C., G. P. Nolan, S. Ghosh, T. Fujita, and D. Baltimore.1992. The NF-KB p50 precursor, p105, contains an internalIKB-like inhibitor that preferentially inhibits p50. EMBO J.11:3003-3009.

35. Logeat, F., N. Israel, R. Ten, V. Blank, 0. Le Bail, P. Kourilsky,and A. Israel. 1991. Inhibition of transcription factors belongingto the rel/NF-KB family by a trans-dominant negative mutant.EMBO J. 10:1827-1832.

36. Mitchell, P. J., and R. Tjian. 1989. Transcriptional regulation inmammalian cells by sequence-specific DNA binding proteins.Science 245:371-378.

37. Moore, P. A., S. M. Ruben, and C. A. Rosen. 1993. Conserva-tion of transcriptional activation functions of the NF-KB p50 andp65 subunits in mammalian cells and Saccharomyces cerevisiae.Mol. Cell. Biol. 13:1666-1674.

38. Neri, A., C. C. Chang, L. Lombardi, M. Salina, P. Corradini,A. T. Maiolo, and R. S. K Chaganti. 1991. B cell lymphoma-associated chromosomal translocation involves candidate onco-gene lyt-10, homologous to NF-KB p5O. Cell 67:1075-1087.

39. Nolan, G. P., S. Ghosh, H.-C. Liou, P. Tempst, and D. Balti-more. 1991. DNA binding and IKB inhibition of the cloned p65subunit of NF-KB, a rel-related polypeptide. Cell 64:961-969.

40. Pabo, C. 1992. Transcription factors: structural families andprinciples of DNA recogpition. Annu. Rev. Biochem. 61:1053-1095.

41. Perkins, N., R. Schmid, C. Duckett, K. Leung, N. Rice, and G.

Nabel. 1992. Distinct combinations of NF-KB subunits deter-mine the specificity of transcriptional activation. Proc. Natl.Acad. Sci. USA 89:1529-1533.

42. Queen, C., and D. Baltimore. 1983. Immunoglobulin gene tran-scription is activated by downstream sequence elements. Cell33:741-748.

43. Rice, N. R., M. L. MacKichan, and A. Israel. 1992. Theprecursor of NF-KB p5O has IKB-like functions. Cell 71:243-253.

44. Ruben, S. M., P. J. Dillon, R. Schreck, T. Henkel, C.-H. Chen,M. Maher, P. A. Baeuerle, and C. A. Rosen. 1991. Isolation of arel-related human cDNA that potentially encodes the 65-kDsubunit of NF-KB. Science 251:1490-1493.

45. Ruben, S. M., J. F. Klement, T. Coleman, M. Maher, C.-H.Chen, and C. A. Rosen. 1992. I-Rel: a novel rel-related proteinthat inhibits NF-KB transcriptional activity. Genes Dev. 6:745-760.

46. Ruben, S. M., R. Narayanan, J. F. Klement, C.-H. Chen, andC. A. Rosen. 1992. Functional characterization of the NF-KBp65 transcriptional activator and an alternatively spliced deriv-ative. Mol. Cell. Biol. 12:444-454.

47. Ryseck, R.-P., P. Bull, M. Takamiya, V. Bours, U. Siebenlist, P.Dobrzanski, and R. Bravo. 1992. RelB, a new Rel familytranscription activator that can interact with p5O-NF-KB. Mol.Cell. Biol. 12:674-684.

48. Schmid, R., N. Perkins, C. Duckett, P. Andrews, and G. Nabel.1991. Cloning of an NF-KB subunit which stimulates HIVtranscription in synergy with p65. Nature (London) 352:733-738.

49. Schmitz, M. L., and P. A. Baeuerle. 1991. The p65 subunit isresponsible for the strong transcription activating potential ofNF-KB. EMBO J. 10:3805-3817.

50. Sen, R., and D. Baltimore. 1986. Multiple nuclear factorsinteract with the immunoglobulin enhancer sequences. Cell46:705-716.

51. Shirakawa, F., and S. B. Mizel. 1989. In vitro activation andnuclear translocation of NF-KB catalyzed by cyclic AMP-dependent protein kinase and protein kinase C. Mol. Cell. Biol.9:2424-2430.

52. Steward, R. 1987. Dorsal, an embryonic polarity gene in Dros-ophila, is homologous to the vertebrate proto-oncogene, c-rel.Science 238:692-694.

53. Tewari, M., P. Dobrzansld, K. L. Mohn, D. E. Cressman, J.-C.Hsu, R. Bravo, and R. Taub. 1992. Rapid induction in regener-ating liver of RL/IF-1 (an IKB that inhibits NF-KB, RelB-p5O,and c-Rel-p5O) and PHF, a novel KB site-binding complex. Mol.Cell. Biol. 12:2898-2908.

54. Toledano, M. B., D. Ghosh, F. Trinh, and W. J. Leonard. 1993.N-terminal DNA-binding domains contribute to differentialDNA-binding specificities of NF-KB p5O and p65. Mol. Cell.Biol. 13:852-860.

55. Zabel, U., and P. Baeuerle. 1990. Purified human IKB can rapidlydissociate the complex of the NF-KB transcription factor withits cognate DNA. Cell 61:255-265.

VOL. 13, 1993