Proc. Natl. Acad. Sci. USAVol. 92, pp. 10242-10246, October 1995Biochemistry

Interactions of a Rel protein with its inhibitor(ankyrin repeats/protein-protein interaction/Rel domain)

NORBERT LEHMING, SEAN MCGUIRE, JOSHUA M. BRICKMAN, AND MARK PTASHNEDepartment of Molecular and Cellular Biology, Harvard University, 7 Divinity Avenue, Cambridge, MA 02138

Contributed by Mark Ptashne, June 28, 1995

ABSTRACT Cactus, a Drosophila homologue ofIcB, bindsto and inhibits Dorsal, a homologue of the p5O and p65components of NF-#cB. We describe experiments in yeast withvarious Dorsal and Cactus derivatives showing that Cactusblocks the DNA binding and nuclear localization functions ofDorsal. In contrast, Dorsal's transcriptional activating regionis functional in the Dorsal-Cactus complex. We identify twoDorsal mutants, Dorsal C233R and Dorsal S234P, that escapeCactus inhibition in vivo, and we show that these mutants failto interact with Cactus in vitro. From this and data of others,we identify the likely surface of Dorsal that binds Cactus. Wealso describe a modified PCR mutagenesis procedure, easierto use than conventional methods, that produces a library ofhigh complexity.

The Drosophila morphogen Dorsal, a transcriptional regulator,activates the genes twist (twi) and snail (sna) with the help ofcoactivators (1) and represses the genes zen and dpp with thehelp of corepressors (2-4). Cactus sequesters Dorsal in thecytoplasm (5) and in a graded fashion releases Dorsal intoventrally positioned nuclei as determined by a signal emanat-ing from the ventral part of the embryo (6). Deletion analyseshave shown that the C-terminal half of the Rel domain ofDorsal is necessary for interaction with Cactus and that aCactus fragment containing all six of its ankyrin repeats issufficient to bind to Dorsal in vitro (7). IKB prevents nuclearlocalization ofNF-KB by masking its nuclear localization signal(NLS; refs. 8 and 9), and p65 with a mutated NLS no longerinteracts with lKB (8). In vitro, IKB inhibits the DNA bindingactivity of NF-KB, and Cactus has a similar effect on Dorsal(10, 11).

Dorsal functions as a transcriptional activator in yeast, andthe experiments described here and previously (2) show thatCactus can inhibit this activation. We here describe Dorsalmutants that escape Cactus inhibition and a Cactus mutantthat inhibits these Dorsal mutants. The positions of theseCactus-resistant mutations, taken with the recently solvedx-ray structure of a Rel domain (12), define a likely pocket onthe Rel domain that interacts with Cactus. We present exper-iments with various fusion proteins in yeast and show thatwhereas Cactus has no effect on Dorsal's activating function,it inhibits both the nuclear localization and DNA bindingfunctions. We also describe a modified PCR mutagenesisprocedure that facilitates production of large numbers ofcloned mutants.

MATERIALS AND METHODSIn Vivo Methods. f3-Galactosidase assays were performed as

described (13, 14). The yeast expression libraries derived from0- to 4-hr-old and 4- to 8-hr-old Drosophila embryos have beendescribed (2). The yeast expression plasmids for Dorsal,Cactus, Dif, p50, and p65 are derivatives of pADNS, a leu2-

marked 2-,tm plasmid containing the ADHI promoter andterminator (15). The yeast integrating plasmids for Dorsal andCactus are derivatives of pGlA2p., a trpl-marked integratingplasmid containing the GPD promoter and terminator (2). TheGAL4-(1-147) fusions to Dorsal and Cactus were made usingpYl, a trpl-marked ArsCen plasmid containing the ADHIpromoter and terminator (16). The LexA-(1-202) fusion toDorsal was cloned into pADNS.

In Vitro Methods. Insoluble material from lysates of theEscherichia coli strain BL21/(DE3)pLysS overexpressing Dor-sal derivatives in pETlia (Novagen) were solubilized in 20mMHepes, pH 7.5/50 mM NaCl/10 mM MgCl2, 6 M urea/i mMdithiothreitol/l mM phenylmethylsulfonyl fluoride and elutedfrom an S Sepharose fast flow (Pharmacia) column with alinear gradient of sodium chloride (from 50 mM to 1 M) in thesame buffer and renatured by stepwise dialysis against 50 mMTris, pH 8.0/600 mM NaCl/10 mM MgCl2/20% (vol/vol)glycerol/0.1% Nonidet P-40/10mM dithiothreitol/i mM phe-nylmethylsulfonyl fluoride containing decreasing concentra-tions of urea. Insoluble bacterial lysate material containinghistidine-tagged Cactus-(217-500) in pET16b (Novagen) wasresuspended in 20 mM Tris, pH 8.0/100 mM KCl/20%glycerol/6M urea/i mM phenylmethylsulfonyl fluoride/1 mMimidazole. Cactus-(217-500) was eluted from a Nj2+_nitrilotriacetic acid (Qiagen) column with the same buffercontaining 250 mM imidazole and renatured as describedabove. Electrophoretic mobility shift assays were done asdescribed in ref. 17. Cactus-(215-500) was cloned into pGEX-5X-1 (Pharmacia). Wild-type and mutant forms of Dorsal andDif were transcribed and translated in vitro (TNT; Promega).The different translation products were added to 2 ,ug ofglutathione S-transferase (GST)-Cactus-(215-500) coupled toglutathione Sepharose beads (Pharmacia) in 20 mM Tris, pH7.4/150 mM NaCl/0.2% Nonidet P-40/2 mM dithiothreitol/0.25% bovine serum albumin in 0.5 ml of binding reactionmixture. Binding reactions were incubated at 25°C for 2 hr withgentle rocking and washed twice with 1 ml of the same bufferand then twice without bovine serum albumin.

RESULTSFig. 1A describes the method we used to isolate two Dorsalmutants that are not inhibited by Cactus, as well as a Cactusmutant that inhibits those Dorsal mutants. The scheme ex-ploits our findings that in yeast Dorsal activates transcriptionof a reporter containing a part of the Drosophila zen promoterand that Cactus inhibits that activation. An example of Cactusinhibition of activation by wild-type Dorsal in yeast is shown inFig. 2B; the figure also shows that two mutants of Dorsal(C233R and S234P) activate transcription but are insensitive toinhibition by Cactus. The two Dorsal mutants were isolatedindependently multiple times (20 and 5 times, respectively). Itis thus unlikely that other Dorsal mutants that are defective for

Abbreviations: NLS, nuclear localization signal; GST, glutathioneS-transferase; X-Gal, 5-bromo-4-chloro-3-indolyl 3-D-galactoside;5-FOA, 5-fluoroorotic acid.

10242

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement" inaccordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 7,

202

0

Proc. Natl. Acad. Sci. USA 92 (1995) 10243

Yeast cell Yeast cell

blue white5-FOAs 5 FOAR

B

FIG. 1. Isolation of mutant Dorsal and Cactus proteins. (A) Yeast strain used for detecting the mutants. The yeast strain contains the ura3 andGALl-lacZ fusion genes each under the control of a 600-bp fragment from the Drosophila zerknullt (zen) promoter. This promoter fragment, calledthe ventral repression element (VRE; ref. 18), contains four strong Dorsal binding sites (17). Activation of transcription of the templates containingthe ventral repression element enables the strain to grow on uracil-depleted medium and to form blue colonies on plates containing5-bromo-4-chloro-3-indolyl f3-D-galactoside (X-Gal) indicator. Inhibition of that activation results in white colonies resistant to the drug5-fluoroorotic acid (5-FOA; ref. 19). 5-FOAS, 5-FOA susceptible; 5-FOAR, 5-FOA resistant. (B) Mutagenesis of Dorsal and Cactus. Themutagenesis scheme utilizes the yeast gap repair system and mutant DNA molecules generated by PCR and avoids passage of the molecules throughE. coli as in the usual procedure. A 1-kb fragment containing the Dorsal Rel domain was subjected to 30 rounds of PCR using Taq polymerase(which results in one mutation per molecule; ref. 20) and then cotransformed directly into yeast with a gapped Dorsal expression plasmid. The yeaststrain also expressed Cactus and bore the depicted reporter genes, so only Dorsal mutants insensitive to Cactus grew on medium depleted for uraciland formed blue colonies on X-Gal indicator plates. One million colonies were screened, and the 25 plasmids rescued revealed two differentmutations. The change at codon 233 from TGC to CGC was found independently 20 times with several different libraries, and the change at codon234 from TCG to CCG was found independently 5 times with several different libraries. One plasmid carried a double mutation, C233R/I243V,reflecting an additional change at codon 243 from ATC to GTC. To obtain Cactus suppressors, genes expressing Dorsal C233R and Dorsal S234Pwere separately integrated into the chromosome of the yeast strain containing the two zen reporters. The two resulting strains are phenotypicallyura+ and thus sensitive to 5-FOA and form blue colonies on X-Gal indicator plates, even if transformed with a plasmid encoding Cactus. A 1.5-kbCactus fragment containing all six ankyrin repeats was subjected to 30 rounds of PCR and then cotransformed with a gapped Cactus expressionplasmid. This library was screened for Cactus mutants that would inhibit Dorsal C233R or Dorsal S234P and thus allow the yeast strains to growon medium containing 5-FOA and to form white colonies on X-Gal indicator plates. One Cactus mutant was isolated from each strain, and sequenceanalysis showed that both of the corresponding genes contained the change AAG to ATG at codon 477, thereby substituting methionine for lysinein the protein.

binding to Cactus but retain the abilities to enter the nucleus,bind DNA, and activate transcription exist. The facts that themutant Dorsal proteins activate as efficiently and bind to DNAin vitro as efficiently as does wild type (Figs. 2B and 3B)indicate that the Cactus-resistant mutants are otherwise wildtype; they are not, for example, deficient in the dimerizationand DNA binding functions. These results are consistent withthe fact that, as indicated by the structure of a Rel domain, themutations lie on a surface of the protein well separated fromthe dimerization interface and the DNA binding surface (seebelow). We also isolated a Cactus mutant, Cactus K477M, thatinhibits the two Dorsal mutants. As shown in Fig. 2B, theCactus mutant inhibits wild-type Dorsal somewhat more effi-

ciently than it inhibits the Dorsal mutants. All of these mutantswere generated by PCR mutagenesis, but unlike in previouslydescribed procedures, the mutagenized pool was transformeddirectly into yeast cells without prior growth in bacteria (seeDiscussion). The amino acid sequence in and around the sitesof the Cactus-resistant mutations of Dorsal is shown in Fig. 24,along with the corresponding sequences from several closerelatives.

Fig. 3A shows that, in vitro, Cactus-(217-500), a fragmentcontaining all six ankyrin repeats, prevented binding of Dorsal-(1-340), a fragment containing the Rel domain, to DNA, buthad no effect on the corresponding Dorsal fragment carryingthe C233R mutation. Challenge of the Dorsal-DNA complex

Biochemistry: Lehming et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 7,

202

0

10244 Biochemistry: Lehming et alt

ADNA

BindingDorsal _

rel-Domain

NLS Activation

t 47 340

233 234

255 L4I1TRLCSCAATAN CGDETIMLCC 279

251 L IM RMDRTAGCVI O,13EEIYLLCD K 275

194 LKIICRVNRNSGSCLEEIFLWCEK 218

185 jRICRVKNCGSVR--7DEIFLjLCD 209

285 &ICRINKESGPCT-EELYL C K309

678aa

Z GALt f/acZaI I I*

Zen promoter fragment

p-Gal. activityActivator -Cactus + Cactus + Cactus(K477M)--1 <1 <1

Dorsal(WT) 1000 50 20Dorsal(C233R) 1000 1000 50Dorsal(S234P) 1000 1000 50

FIG. 2. Two Dorsal mutants that escape Cactus inhibition. (A) Comparison of partial amino acid sequences of Dorsal, Dif, the NF-KB subunitsp5O and p65, c-Rel, and RelB. h, human. (B) Response of wild-type (WT) and mutant Dorsal to Cactus. Cells bore the indicated reporter andexpressed mutant and wild-type Dorsal and in some cases also Cactus. 13-Galactosidase (3-Gal.) activities were measured from cell extracts.

with competitor oligonucleotide revealed no difference be-tween the DNA binding affinities of wild-type and mutant

A BDorsal(1 -340)

WT C233RCactus(217-500) Cactus(21 7-500)

Dorsal proteins (Fig. 3B). Fig. 3C shows that GST-Cactus-(215-500) bound to Dorsal-(1-340) but not to Dorsal-(1-

Dorsal(l -340)

WT C233R

GGGAATTCCC GGGAATTCCC

Dorsal(1-340)/ --__-DNA __ A-

Free DNA -_

'U

Input

I

C.)

U) U)

o o

ncs

ln~-

CY) flr)cm N

3: o-) oC)c

Free DNA

GST GST-Cactus(215-500)

r

0

-n~ -a

U)

o o

0 0 0

I1

cn CY -crCM cm OjL)3H gC)gc

a

CZ)

-n -n

0-o

Er aCCco co

coC') flC

Fici. 3. Interactions of Dorsal nlnd Dorsail mutants with DNA aindwith Cictus in vitro. (A) Dorsal-( 1-340) C233R but not its wild-type(WT) palrent binds DNA in tlhe presence of Caictus-(217-o00). Elec-trophoretic mobility shift issays were performed uwith a purifiedDorsal fratgment (residucs 1-340) in its wild-tvpe and mutant forms,usiiie a libeled oliconuclcotide containinti a Doisal consellsus site.Purified Cactus-(017-500) piotecin was addeid (in 2-fold steps). (B)Dorsal-( 1 -34(0) C233R binds DNA with the saimc iiffiniitv as wild type.A competition assay was performed is in A. escept that insteaid ofCactuis an unlabeled oligonucleotide containing the Dorsal consensussite wts aidded (in 33-fold stcps). (C) ThenmutatLtions C233R and S2341Pabolish hindinoL of Dorsatl to GST-Cactus-(2l$-50t)}) in solution. Dor-sal-(l-340). Dorsal-(1-340) C233R. Dorsail. Dorsatl C233R. DorsalS234P. aind Dif wire trainslatted in litrno and laibeled with 35S. (Rig/il)Proteins thit bound to GST-Cactus-(215-500) oni beads. (Left) Inlputof ini 0itro-translated and 3'S-labeled protein. (C(enter) No proteini walsboLndblv GST alone.

B

Dorsal:Dif:h-p50:h-p65:hc-Rel:h-RelB:

C

Dorsal(1 -340)/DNA

n

to

WpU

Proc. Natl. Acad. Sci. USA 92 (1995)

Dow

nloa

ded

by g

uest

on

Nov

embe

r 7,

202

0

Proc. Natl. Acad. Sci. USA 92 (1995) 10245

-100

GAL4 sites

|IGALl AacZ -100

I LJ.LexA sites

-200 IGALl /lacZI sites..GAL4 sites

P-Gal. activity+ GAL4(1-147) + GAL4(1-147)- Fold

Cactus(215-500) Activationc1 5 _

Dorsal(WT) <1 500 100Dorsal(C233R) <1 15 3Dorsal(S234P) <1 10 2Dif <1 500 100p50 <1 10 2p65 <1 500 100

C -100 IGAL1 /lacZ.1 s.I t .-.-.I

Rel sites

P-Gal. activity Fold

! -Cactus + Cactus Inhibition

Dorsal(WT)Dorsal(C233R)Dorsal(S234P)Difp50p65

<1 <1 -

100 5 20100 100 1

100 100 1100 10 1050 50 1

200 20 10

FIG. 4. Dorsal-Cactus interaction in yeast. (A) Cactus inhibits two functions of Dorsal. ,B-Galactosidase (,3-Gal.) activities of cell extracts bearingthe depicted reporter genes and expressing the indicated Dorsal fusion proteins were measured in the absence and in the presence of Cactus protein.WT, wild type. (B) Dorsal, Dif, and p65 convert GAL4-Cactus-(215-500) into an activator. f3-Galactosidase activities of cells bearing the depictedreporter gene and expressing the indicated Rel proteins were measured in the presence of GAL4-(1-147) and in the presence of GAL4-Cactus-(215-500). (C) Dorsal, Dif, and p65 are inhibited by Cactus. ,B-Galactosidase activities of cells bearing the depicted reporter and the indicated Relproteins were measured in the presence and absence of Cactus. The activating domain of p65 has been shown to contain the same sequences as

assayed in human cells and in yeast (24).

340)C233R. The figure further shows that GST-'500) did not bind to the two full-length Dorsalbound to wild-type Dorsal and to Dif, a Drosophilof Dorsal that is expressed at larval stages (21).The differential response of two Dorsal fus

Dorsal and GAL4-Dorsal) to Cactus reveals that tof Dorsal are inhibited by Cactus in vivo. The GAIfusions, residues 1-147, binds DNA and bears an Pan activating region (22, 23). The LexA moiety, re

also binds DNA but lacks both an NLS and;

NLSCactus

A"_

NLS

Monomer I

FIG. 5. Model of Cactus binding to Dorsal. Ca(indicated would cover the site of the mutations and the D'is the structure of the Rel domain as found in the p50-](12); the DNA is indicated by a black circle. The piCactus-resistant mutations as found in the Rel domairmarked by asterisks, and the positions of the NLS are

Cactus-(215- region. As shown in Fig. 4A, in the absence of Cactus, bothmutants but fusions activate transcription on promoters bearing bindinglahomologue sites either for Dorsal or for the heterologous DNA binding

region (i.e., that of LexA or GAL4). Cactus inhibited activa-sions (LexA- tion by GAL4-Dorsal of a template bearing Dorsal bindingtwo functions sites but did not affect activation of a template bearing GAL4L4 part of the binding sites. In contrast, Cactus prevented LexA-Dorsal from4L8 but lacks activating transcription from templates bearing either Dorsalsidues 1-202, or LexA sites. We imagine that, in the case of the LexA fusion,an activating the protein is sequestered in the cytoplasm and hence cannot

activate either reporter. In contrast, because of the additionalNLS (8), the GAL4-Dorsal-Cactus complex moves to thenucleus where it binds to the GAL4 site (but not to the Dorsalsite) and activates transcription. Thus Cactus must inhibit boththe nuclear localization and DNA binding functions of Dorsalbut evidently leaves unaffected its activating region.From the experiment of Fig. 4A described above, we con-

cluded that in the GAL4-Dorsal-Cactus complex, the Dorsal-

Monomer II activating region is free to activate transcription. The experi-ment of Fig. 4B reinforces this picture. In this case Cactus isfused to GAL4 and bound to a GAL4 site on DNA; additionof Dorsal activates transcription, evidently because Dorsal'sactivating region is tethered to DNA by virtue of Dorsal'sinteraction with DNA-bound Cactus. We repeated this exper-iment substituting Dif, p65, p50, Dorsal C233R, and DorsalS234P in place of Dorsal and found that the former two, butnot the latter three, activated transcription, a result suggestingthat Cactus interacts with p65 and Dif but not with p50 or thetwo Dorsal mutants. This conclusion is further supported bythe results of the experiments of Fig. 4C. This figure shows that

ctus bound as Dorsal, Dif, p65, p50, and the two Dorsal mutants all activated

DNA cocrystal transcription in yeast from a template containing two copies of

ositions of the a Rel consensus site and that Cactus inhibited activation byi of Dorsal are Dorsal, p65, and Dif but had no effect on p50 or the two Dorsalindicated. mutants.

A -100

Rel sites

l-Gal. activity Fold p-Gal. activity Fold f-Gal. activity _ Fold-Cactus + Cactus Inhibition -Cactus + Cactus Inhibition -Cactus + Cactus Inhibition

<c1 <1 <1 _ c -

Dorsal(WT) 100 5 20 <1 <1 - <1 ctGAL4(1-147)-Dorsal 60 4 15 60 60 1 <1 <1 -LexA-Dorsal 60 2 30 <1 <1 - 120 8 15

B

Biochemistry: Lehming et al.

L1

Dow

nloa

ded

by g

uest

on

Nov

embe

r 7,

202

0

10246 Biochemistry: Lehming et al.

DISCUSSION

Fig. 5 shows the structure of the Rel domain as found in ap50-DNA cocrystal (12) and the likely site of Cactus inter-action. Note that although the NLS is separated by more than100 amino acids from the sites of the Cactus-resistant muta-tions, the protein folds so that the two regions are in nearjuxtaposition. As mentioned above, the integrity of the NLS isrequired for Rel-inhibitor binding, and the model shows howCactus would cover both the NLS and the sequence defined byour mutations. The activating region on relatives of p50 (e.g.,that on p65) extends from the carboxyl end of the Rel domain,and it is therefore not surprising that Cactus does not occludethis function. It is possible, however, that Cactus directly blocksDNA binding; this result is not predicted by the model of Fig.5. One possibility is that, despite our inability to find Cactus-resistant mutants elsewhere in Dorsal, Cactus extends suffi-ciently to cover the DNA binding surface. If so, a fragmentcontaining only the six ankyrin repeat must extend similarly,because we and others (7) have found that such a fragmentblocks DNA binding of Dorsal. A more likely explanation forCactus' effect on Dorsal's DNA binding, we believe, is that theinhibition works indirectly in the regard, causing a conforma-tional change in Dorsal that is not consistent with DNAbinding.From the inhibition and activation experiments of Fig. 4, we

deduce that Cactus binds best to Dorsal, less well to Dif andp65, and not at all to p50. Fig. 2A shows that Dif and p65 eachhave one residue in common with Dorsal at positions 233 and234, the site of our Cactus-insensitive Dorsal mutations. p50differs in both positions, consistent with our finding that thesepositions are important for specifying the Rel domain-ankyrinrepeat interaction.Our method of mutagenizing genes in yeast exploits two

previously published findings. First, Taq polymerase lacks aproofreading function, and PCR performed with Taq results inabout one mutation per kilobase (20). Second, yeast efficientlyrecombines two pieces ofDNA by homologous recombination(25). We have found that a fragment produced by PCRcotransformed with a gapped plasmid will recombine in yeastwith high efficiency to form an intact plasmid. In experimentsin which a 1-kb fragment was used for the PCR reaction, everyplasmid on average carried a single mutation. A conventionallibrary is formed by ligation of different components andpassage through E. coli. Our method avoids the use of E. coliand thus has three advantages. First, the complexity of thelibrary is not limited by ligation frequency. (The transforma-tion efficiency of this method is 105 yeast colonies per micro-gram of DNA). Second, the DNA sequences typically elimi-nated by E. coli may survive. Third, the number of moleculesused to start the PCR reaction (108) far exceeds the number oftransformants screened (106), and so all mutants recovered arehighly likely to be independent events. The fact that we did notfind Cactus-insensitive mutations in Dorsal's NLS (which, asnoted above, is required for Cactus binding) was expected,because the screening procedure requires that Dorsal move tothe nucleus and be transcriptionally active.The system described in Fig. 1 is a powerful tool for isolating

regulatory proteins and studying their functions. We used thismethod previously (2) to isolate cDNAs encoding the activatorDorsal and two Dorsal inhibitors, Cactus and DSP1, from anexpression library derived from early Drosophila embryos (2).

The method can be modified by fusing a promoter other thanzen to our reporters and screening an expression libraryderived from the organism of choice in order to search forother positive and negative regulators.

We thank D. Baltimore, S. C. Harrison, T. Maniatis, B. Muller-Hill,and D. Thanos for comments on the manuscript; C. W. Muller andS. C. Harrison for Fig. 5, showing the p50-DNA cocrystal structure;V. Palombella and T. Maniatis for the yeast vectors expressing p50 andp65; Y. T. Ip and M. Levine for the cDNA encoding Dif; P. Chang forthe GST-Cactus-(215-500) fusion; J. Pearlberg for the GAL1-lacZreporter containing unique restriction sites at -100; S. Passmore forintroducing us to the yeast gap repair system; R. Hellmiss for graphics;and L. Barberis-Maino and S. Bassiri for technical assistance. Thiswork was supported by a grant from the National Institutes of Healthto M.P. and a fellowship from the American Chemical Society to N.L.

1. Ip, Y. T., Park, R. E., Kosman, D., Yazdanbakhsh, K. & Levine,M. (1992) Genes Dev. 6, 1518-1530.

2. Lehming, N., Thanos, D., Brickman, J. M., Ma, J., Maniatis, T. &Ptashne, M. (1994) Nature (London) 371, 175-179.

3. Jiang, J., Cai, H., Zhou, Q. & Levine, M. (1993) EMBO J. 12,3201-3209.

4. Kirov, N., Zhelnin, L., Shah, J. & Rushlow, C. (1993) EMBO J.12, 3193-3200.

5. Roth, S., Stein, D. & Niisslein-Volhard, C. (1989) Cell 59,1189-1202.

6. Steward, R. & Govind, S. (1993) Curr. Opin. Genet. Dev. 3,556-561.

7. Kidd, S. (1992) Cell 71, 623-635.8. Beg, A. A., Ruben, S. M., Scheinnman, R. I., Haskill, S., Rosen,

C. A. & Baldwin, A. S. (1992) Genes Dev. 6, 1899-1913.9. Henkel, T., Zabel, U., van Zee, K., Muller, J. M., Fanning, E. &

Baeuerle, P. A. (1992) Cell 68, 1121-1133.10. Haskill, S., Beg, A. A., Tompkins, S. M., Morris, J. S., Yurochko,

A. D., Sampson-Johannes, A., Mondal, K, Ralph, P. & Baldwin,A. S. (1991) Cell 65, 1281-1289.

11. Geisler, R., Bergmann, A., Hiromi, Y. & Nusslein-Volhard, C.(1992) Cell 71, 613-621.

12. Muller, C. W., Rey, F. A., Sodeoka, M., Verdine, G. L. & Har-rison, S. C. (1995) Nature (London) 373, 311-317.

13. Miller, J. H. (1972) Experiments in Molecular Genetics (ColdSpring Harbor Lab. Press, Plainview, NY).

14. Saha, S., Brickman, J. M., Lehming, N. & Ptashne, M. (1993)Nature (London) 363, 648-652.

15. Colicelli, J., Birchmeier, C., Michaeli, T., O'Neill, K, Riggs, M.& Wigler, M. (1989) Proc. Natl. Acad. Sci. USA 86, 3599-3603.

16. Sadowski, I., Bell, B., Broad, P. & Hollis, M. (1992) Gene 118,137-141.

17. Ip, Y. T., Kraut, R., Levine, M. & Rushlow, C. A. (1991) Cell 64,439-446.

18. Doyle, H., Kraut, R. & Levine, M. (1989) Genes Dev. 3, 1518-1533.

19. Boeke, J. D., LaCroute, F. & Fink, G. R. (1984) Mol. Gen. Genet.197, 345-346.

20. Zhou, Y., Zhang, X. & Ebright, R. (1991) Nucleic Acids Res. 19,6052.

21. Ip, Y. T., Reach, M., Engstrom, Y., Kadalayil, L., Cai, H.,Gonzalez-Crespo, S., Tatei, K. & Levine, M. (1993) Cell 75,753-763.

22. Silver, P. A., Keegan, L. P. & Ptashne, M. (1984) Proc. Natl.Acad. Sci. USA 81, 5951-5955.

23. Silver, P. A., Brent, R. & Ptashne, M. (1986) Mol. Cell. Biol. 6,4763-4766.

24. Moore, P. A., Ruben, S. M. & Rosen, C. A. (1993)MoL Cell. Biol.13, 1666-1674.

25. Rothstein, R. (1991) Methods Enzymol. 194, 281-301.

Proc. Natl. Acad. Sci. USA 92 (1995)

Dow

nloa

ded

by g

uest

on

Nov

embe

r 7,

202

0