Design and properties of a Myc derivative that e�ciently homodimerizes

Laura Soucek1, Manuela Helmer-Citterich2, Alessandra Sacco1, Richard Jucker1, Gianni Cesareni2

and Sergio Nasi1

1UniversitaÁ `La Sapienza', Centro Acidi Nucleici CNR, 00185 Rome; 2UniversitaÁ `Tor Vergata', Dip. Biologia, 00173 Rome, Italy

bHLH and bHLHZip are highly conserved structuraldomains mediating DNA binding and speci®c protein-protein interactions. They are present in a family oftranscription factors, acting as dimers, and their selectivedimerization is utilized to switch on and o� cellproliferation, di�erentiation or apoptosis. Myc is abHLHZip protein involved in growth control and cancer,which operates in a network with the structurally relatedproteins Max, Mad and Mnt. It does not formhomodimers, working as a heterodimer with Max;Max, instead, forms homodimers and heterodimers withMad and Mnt. Myc/Max dimers activate gene tran-scription, while Mad/Max and Mnt/Max complexes areMyc/Max antagonists and act as repressors. Modifyingthe molecular recognition of dimers may provide a toolfor interfering with Myc function and, in general, fordirecting the molecular switches operated viabHLH(Zip) proteins. By molecular modelling andmutagenesis, we analysed the contribution of singleamino acids to the molecular recognition of Myc,creating bHLHZip domains with altered dimerizationspeci®city. We report that Myc recognition speci®city isencoded in a short region within the leucine zipper;mutation of four amino acids generates a protein,Omomyc, that homodimerizes e�ciently and can stillheterodimerize with wild type Myc and Max. Omomycsequestered Myc in complexes with low DNA bindinge�ciency, preventing binding to Max and inhibiting Myctranscriptional activator function. Consistently with theseresults, Omomyc produced a proliferation arrest inNIH3T3 cells. These data demonstrate the feasibilityof interfering with fundamental biological processes, suchas proliferation, by modifying the dimerization selectivityof a bHLHZip protein; this may facilitate the design ofpeptides of potential pharmacological interest.

Keywords: Myc; bHLHZip domain; molecular model-ling; protein dimerization; cell proliferation

Introduction

The Max network is a cellular regulatory circuit,involved in cell fate determination (Amati and Land,1994; Henriksson and Luscher, 1996). The circuit isoperated by dimers of the bHLHZip (basic helix ±loop ± helix/zipper) proteins Myc, Max, Mad (Mad1,Mxi or Mad2, Mad3, Mad4), Mnt (Rox); Max holds acentral position, participating in the formation of avariety of protein complexes which bind DNA and

regulate gene transcription. Myc/Max dimers induceproliferation or apoptosis, while Max/Mad and Max/Mnt complexes cause cell growth arrest and differentia-tion (Amati et al., 1992; Ayer and Eisenman, 1993; Ayeret al., 1993; Evan et al., 1992; Hurlin et al., 1995, 1997;Meroni et al., 1997; Zervos et al., 1993). All dimersrecognize in vitro the same consensus binding site, theCACGTG E-box, but have distinct transcriptionalactivities: Myc/Max dimers activate, while Max/Mador Max/Mnt complexes repress transcription from thissite in arti®cial promoters (Amati and Land, 1994; Ayeret al., 1993; Blackwell et al., 1990; Gu et al., 1993;Kretzner et al., 1992; Laherty et al., 1997; Murre et al.,1989; Prendergast et al., 1991; Sommer et al., 1997). Theequilibria among the various dimers are mainlycontrolled through extra-cellular signal induced mod-i®cations in Myc or Mad expression levels, the Maxconcentration remaining constant. Mad genes expres-sion is induced during cell di�erentiation, while c-myctranscription is regulated by tyrosine kinase signallingand is rapidly turned on during the G1 phase of the cellcycle, upon-mitogenic stimulation; DNA binding ofMyc/Max complexes can also be regulated by post-translational modi®cation, e.g. Max phosphorylation(Ayer and Eisenman, 1993; Berberich and Cole, 1992;Henriksson and Luscher, 1996; Zu et al., 1997).Heterodimers with Max are the functionally active

form of the Myc protein, which homodimerizes andbinds DNA poorly (Amati et al., 1993; Blackwoodand Eisenman, 1991; Blackwood et al., 1992; Little-wood and Evan, 1994). Myc/Max complexes activatetranscription of several genes containing theCACGTG site, such as the genes for the polyamminebiosynthesis enzyme ornithine decarboxylase, prothy-mosine a, the elongation factor eIF4E, the embryonicmarker ECA39, the helicase MrDb, the growth-relatedfactor Rcl and two cell cycle regulators: the CDC25Aphosphatase and the E2F transcription factor (Bello-Fernandez and Cleveland, 1992; Benvenisty et al.,1992; Eilers et al., 1991; Galaktionov et al., 1996;Grandori et al., 1996; Jones et al., 1996; Leone et al.,1997; Lewis et al., 1997; Sears et al., 1997). Myc canalso repress transcription of cell cycle regulators, likecyclin D1 and Myc itself, and of growth arrest genes,like gas1 (Lee et al., 1997; Penn et al., 1990; Peukertet al., 1997; Philipp et al., 1994). It is unclear whichgenes represent the physiologically relevant Myctargets.The Myc family proteins c-Myc, L-Myc and N-Myc

have a role in the genesis of a wide range of neoplasias,including lymphomas, neuroblastomas and small celllung carcinomas (Cory and Adams, 1988; Lombardi etal., 1987; Weiss et al., 1997). Myc/Max dimers controlan important step in cell cycle progression in the G1/Sand, possibly, the G2/M transition (Seth et al., 1993).CyclinE/Cdk2 complexes represent a major target for

Correspondence: S NasiReceived 14 April 1998; Revised 5 June 1998; accepted 5 June 1998

Oncogene (1998) 17, 2463 ± 2472ã 1998 Stockton Press All rights reserved 0950 ± 9232/98 $12.00

http://www.stockton-press.co.uk/onc

G1/S transition regulation by Myc. The activity ofMyc/Max dimers neutralizes the p27 cell cycleinhibitor, activating cyclinE/CDK2 complexes, andovercomes growth inhibition by the p16/Rb pathway;this suggests that bypassing the tumour suppressorfunction of this pathway may be an essential feature ofMyc oncogenic potential (Alevizopoulos et al., 1997;Steiner et al., 1995; Vlach et al., 1996). Oncogenicactivation of myc genes occurs mainly through aderegulated expression that leads to a shift of theequilibria among dimers towards Myc/Max complexes(Koskinen et al., 1995). Interfering with Myc dimeriza-tion is therefore biologically relevant and it should bepossible to direct the operation of the Max network bycontrolling the equilibria among di�erent dimers.In this study, we investigated molecular recognition

of Myc protein bHLHZip domain. We identi®ed bymolecular modelling the critical amino acid residues forMyc/Max protein-protein interaction and constructed amutant Myc with altered dimerization speci®city,Omomyc, able to homodimerize and to e�cientlyform complexes with wild type Myc. Omomycexpression in mammalian cells provided an alternativebHLHZip domain to the Max network and was able tointerfere with Myc transactivation and growth control.This may constitute an initial step to obtain peptides ofpharmacological interest which control Myc and Maxdimerization.

Results

Four amino acids in the leucine zipper are critical forMyc dimerization speci®city

Myc and Max proteins from di�erent species share asigni®cant degree of sequence similarity in theirbHLHZip domains (Littlewood and Evan, 1994). Toidentify the amino acids that are most important fordimerization, we utilized the crystallographic data ofthe DNA bound Max homodimer (FerreÁ -D'AmareÁ etal., 1993) for building, by homology, model structuresof a hypothetical Myc homodimer and a Myc/Maxheterodimer (Figure 1a). The Max bHLHZip domainconsists of two long a-helices, separated by a loop; theN-terminal a-helix is composed of the basic region andof helix 1 (H1), the second a-helix includes helix 2 (H2)and the leucine zipper. H1 and H2 of the twomonomers form a parallel four-helix bundle, a proteinfold shared by bHLHZip and bHLH proteins; the twobasic regions project from the bundle and bind DNAin the major groove (Ellenberger et al., 1994; FerreÁ -D'AmareÁ et al., 1993). The two leucine repeat portionsform a parallel coiled coil that closely resembles theleucine zipper of the yeast transcription factor GCN4(FerreÁ -D'AmareÁ et al., 1993). We focused our attentionon the H2Zip region as our previous work showed thatthe N-terminal a-helix (bH1) is not involved in thespeci®city of dimerization, which is dictated by H2Ziponly (Marchetti et al., 1995). The classical leucinezipper structure is characterized by the heptad repeat(abcdefg)n, where leucines usually occupy d andhydrophobic amino acids a position, and is mainlystabilized by hydrophobic interactions between aminoacids in a and d position of the two helices (Figure 1c);interhelical salt bridges between an e residue of one

strand and the preceding g residue of the other strandmay give a further contribution to the stability of thestructure. However, a prevalence of hydrophobicresidues in a position is not observed in Myc andMax zippers (Figure 1c). The dimeric model structureswere analysed by visual inspection and by theESCHER docking procedure (Ausiello et al., 1997),evaluating electrostatic interactions and surface com-plementarity, in order to identify the amino acidsinvolved in recognition speci®city.By comparing Myc dimer model structure and Max

dimer crystallographic structure, three regions (Figure1) were identi®ed as an obstacle to Myc homodimer-ization, as they displayed major steric and electrostaticclashes. They correspond to four charged amino acids,two glutamates (E57, E64) and two arginines (R70,R71), located in the leucine zipper, the region ofgreater diversity between Myc and Max; all fourcorrespond to amino acids involved in dimerizationaccording to the Max dimer crystallographic data.Three of them (E57, E64 and R71) are located in aposition, involved in interactions stabilizing the leucinezipper dimer; the corresponding positions in Max areoccupied by asparagine, isoleucine and asparagine(N,I,N), respectively. In the-Max homodimer, more-over, glutamine and asparagine residues at positions 70and 71 of the two monomers form a remarkably stabletetrad (QN/QN) of major importance for structurestabilization (FerreÁ -D'AmareÁ et al., 1993). Conse-quently, the poor Myc homodimerization is explainedby the disruption of this tetrad, since positions 70 and71 in Myc are occupied by aminoacids (two arginines)with the same polarity, and by the presence of threecharged residues (the two glutamate 57,64 and arginine71) at three consecutive a positions, that destabilize theZipper region. Four di�erent Myc model structures,with mutations in these four aminoacids (mut1-mut4),were built and the stability of the corresponding homo-and heterodimeric (with Max and wild type Myc)structures evaluated (Figure 1).In Myc-mut1 (E57N, E64I), the glutamates 57 and

64 were substituted with asparagine and isoleucine, asin the Max protein (Figure 1c). In mut2 (R70Q, R71N)the two arginines 70 and 71 were substituted with thecorresponding Max amino acids (glutamine andasparagine), reconstituting the Max tetrad. Theanalysis of mutant model structures indicated thatMyc-mut1 was unlikely to be able to dimerize becauseof the presence of the highly unfavourable RR/RRtetrad, while mut2 displayed a much nicer shape andelectrostatic complementarity, even if the two Eresidues in positions 56 and 57 still represent apotentially destabilizing tetrad. Myc-mut3 (E57N,E64I, R70Q, R71N) has all four amino acids (57, 64,70 and 71) substituted with the Max ones. This mutantwas predicted to homodimerize only weakly, due to anunfavourable shape complementarity of amino acidsaround position 57. Therefore a fourth mutantstructure, named mut4 or Omomyc (E57T, E64I,R70Q, R71N), was modeled. In this mutant, MycE57was substituted with a T, while the other three aminoacids were substituted with the amino acids present inMax. The T substitution provided a better shapecomplementarity between the two monomers (Figure1d). As mut2 (where only the Max tetrad isreconstituted) and mut4 (where all four amino acids

OmomycL Soucek et al

2464

were mutated) model structures appeared favourablefor homodimerization, the corresponding mutationswere generated by PCR mutagenesis of the mycbHLHZip domain and introduced in the pC135

expression vector. The mutated Myc domains arefused in frame to the DNA binding domain of lphage cI repressor, producing a chimerical repressor(Marchetti et al., 1995).

Figure 1 Molecular modeling of Myc and Max dimers. (a) Three dimensional models of DNA bound human Myc homodimer(left) and Myc/Max heterodimer (right). Myc is coloured by atom type (green: C; red: O; blue: N), while Max is in yellow. Themodels were obtained by superposing the backbone structure of the two bHLHZip domains onto the three-dimensional crystalstructure of a DNA bound Max homodimer (FerreÁ -D'AmareÁ et al., 1993). Amino acids E57, E64, R70, R71 destabilize the putativeMyc homodimer. (b) Stability of human Max, Myc (Mycmod) and mutant Myc (mut1-4) bHLHZip dimers. Predictions wereperformed with the ESCHER procedure, evaluating surface complementarity, hydrogen bonding and electrostatic interactions.Phenotype refers to the experiments reported in Figures 2 and 3. (c) Sequence comparison of Myc, Max and Myc mutants 1 ± 4leucine zippers. The four aminoacids that were mutated are in bold character. (d) Details of the 3D structure of Myc, Max andOmomyc bHLHZip domains around the four residues 57, 64, 70 and 71. These four positions were mutated in the Myc domain inorder to obtain Omomyc: three of the four residues were substituted with corresponding aminoacids in Max, while glutamate (E) 57was substituted with threonine (T), that appeared to ®t better than asparagine the structural model

OmomycL Soucek et al

2465

Omomyc homodimerizes as e�ciently as Max

Dimerization of Myc and Max bHLHZip domains wastested by the chimeric l repressor assay (Castagnoli etal., 1994; Hu et al., 1990; Marchetti et al., 1995) inbacteria transformed by the plasmids producing cI-bHLHZip domain fusion proteins. Dimerization offusion proteins allows the formation of functionalrepressor molecules, since dimerization is essential forstable DNA binding and repression activity of the cIprotein. The repressor assays reported in Figure 2speci®cally re¯ect the dimerization ability of the

bHLHZip domains fused to the l repressor, becausedi�erent domains were produced at the same level inbacteria (Figure 2d).Repressor function was ®rst qualitatively assessed,

by determining sensitivity to l phage infection ofbacteria producing di�erent chimeric proteins: dimer-ization of bHLHZip domains confers resistance tophage infection, by allowing formation of repressing cIdimers (Figure 2a). As the Myc bHLHZip domain isunable to dimerize e�ciently, bacteria transformed bythe cI-Myc fusion expression plasmid (pC135Myc)were sensitive to phage infection; by contrast, cI-Max, which e�ciently forms dimers, conferred on hostbacteria complete resistance to l infection (Figure 2b).The cI-Mycmut2 chimera did not dimerize, as it wasunable to confer immunity to phage infection (data notshown). The cI-Mycmut4 (cI-Omomyc) chimera,instead, dimerized e�ciently because it behaved likeMax, rendering bacteria immune to phage infection.Chimeric repressor dimerization was then quanti-

tated, by measuring repression of b-galactosidaseactivity in GC421 bacteria transformed by plasmidsdirecting the synthesis of cI-bHLHZip fusion proteins;in the GC421 bacterial strain, the b-galactosidase geneis under control of lPROR (right promoter ± operator ofphage l) and is expressed constitutively. Dimerizationof cI-bHLHZip fusion proteins causes repression of thereporter gene, thus inhibiting b-galactosidase activity.The cI-Mycmut2 chimera, carrying only two of thefour amino acid substitutions present in Omomyc, wasunable to a�ect enzymatic activity (Figure 2c). This®nding con®rmed that the two mutations at positions70 and 71 were not su�cient for homodimerization,which was obtained only when all four destabilizingamino acids were changed. As a matter of fact,Omomyc lowered b-galactosidase expression as effi-ciently as Max (Figure 2c). The ®nding that Mycmut2,which full®lled the modelling criteria, was unable todimerize, can be accounted for by the uncertainty ofhomology modelling procedures. It can be consideredthat, for a 50% identity for a known structure builtfrom another known structure, the mean error in themodelled coordinates can be as large as 1.5 AÊ ngstrom,with considerably larger local errors (Chothia andLesk, 1986; Sippl, 1993). Since the human MycbHLHZip domain shares only a 36% identity withMax aminoacid sequence, our modelled structures,overall, showed a good ®t with experimental data.Therefore, only Omomyc was further analysed.

Omomyc forms dimeric complexes with Myc and Max

Heterodimer formation by Omomyc was assayedindirectly, through the ability of Myc and Max tocompete with repressing cI-Omomyc dimers as aconsequence of the formation of heterodimericcomplexes with Omomyc. To this purpose, we haveutilized an expression vector derived from pC135 todirect the synthesis of bHLHZip domains fused to a cIdomain defective in operator recognition, cI* (Longoet al., 1995; Marchetti et al., 1995). We have thentested the ability of these constructions to relieve l Prtranscription repression by a compatible plasmid,derived from pACYC184 and containing an active cI-Omomyc gene (Figure 3a). As pACYC184 is a lowcopy number plasmid, it drives a lower level of

Figure 2 Omomyc forms homodimers as e�ciently as Max. (a)Rationale of the chimerical repressor system. The DNA binding,N-terminal domain of the l phage cI repressor requires the C-terminal dimerization domain to form dimers that strongly bindthe phage operator Or. Substituting the cI repressor C-terminaldomain with a heterologous domain will generate a functionalrepressor only if the heterologous domain dimerizes e�ciently. (b)Phage immunity test. Phages lDHK54 were spotted, atconcentrations varying from 105 ± 101 phages per spot, on lawnsof bacteria transformed with pC135Myc, pC135Max andpC135Omomyc. Lysis plaques are evident in bacteria expressingthe cI-Myc chimerical protein, while no plaques are visible onbacteria harboring cI-Max or cI-Omomyc. (c) b-galactosidaserepression assay. pC135 -Myc, -Max, -Mycmut2 and -Omomycexpression plasmids were introduced into GC421 bacterial strain,which constitutively produces b-galactosidase under lPrOrcontrol. Fusion protein dimerization was quanti®ed as repressionof b-galactosidase activity. cI-Omomyc repressed it as well as cI-Max. Mean values and standard deviations from at least fourindependent experiments are presented. (d) Expression of Myc,Max and Omomyc proteins in bacteria transformed with pC135vectors. Equal amounts of total cellular extracts transformedbacteria were analysed by Western blots with antisera speci®c forMyc (left) and Max (right)

OmomycL Soucek et al

2466

expression as compared to the high copy numberplasmid pC135. Therefore, the b-galactosidase activityrepression reported in Figure 3, which is relative tobacteria transformed by pACYC Max or Omomyc, isabout sevenfold lower than reported in Figure 2, wherethe pC135 vector was employed. In GC421 cells co-transformed by pACYC184Omomyc and pC135*Mycor pC135*Max, heterodimer formation was easilydetected by an increase in b-galactosidase activity ascompared to cells transformed by pACYC184Omomyconly. In fact, the formation of cI-Omomyc/cI*-Mycand cI-Omomyc/cI*-Max heterodimers, which cannotbind the phage operator and do not repress b-galactosidase, interfered with the formation of repres-sing Omomyc homodimers.These results demonstrate that Omomyc can form

heterodimers with Myc and with Max; its binding toMax, however, is not as strong as the binding betweenMyc and Max (Figure 3b, last two columns).

Omomyc binds the CACGTG E box with low a�nityand interferes with Myc/Max dimer formation

To determine whether Omomyc binds the consensusbinding site for Myc/Max and Max/Max complexes,

gel shift assays were performed using puri®edOmomyc, Myc and Max GST-fusion proteins and theradiolabelled DP oligonucleotide, containing the E-boxsequence CACGTG. A shift was observed withOmomyc-GST and its speci®city was established bycompetition with di�erent E-boxes (Figure 4a). DNAbinding of Omomyc dimers was ine�cient as it wasobserved only with relatively high protein amounts(greater than 250 ng), about tenfold higher than thoserequired to observe a shift with Max or with otherbHLH protein dimers. We then compared DNAbinding of Omomyc dimers, Max dimers and Omomyc/Myc or /Max heterodimers. DNA bound GST-Maxrun as a doublet in our hands (lanes 1 and 2). Omomycformed heterocomplexes with Max which bound DNA(Figure 4b, lanes 7 ± 10) with a high e�ciency, similarto Myc/Max dimers (lanes 3 ± 6) and considerably

Figure 3 Omomyc forms dimeric complexes with Max and Myc.(a) Rationale of the cI/cI* heterodimerization system. cI* is a lrepressor DNA binding domain with three mutations in the cIDNA binding domain. As a consequence, it is unable to recognisel PrOr and acts as dominant negative in heterodimers with cI. (b)b-galactosidase assay. Myc or Max bHLHZip domain, fused tocI* (pC135* vector), was introduced in bacteria expressing cI-Omomyc or cI-Max (from the low copy number vector pACYC).Heterodimerization was assayed by loss of b-galactosidaserepression, due to cI homodimers: cI*-bHLHZip/cI-bHLHZipcomplexes are unable to bind the l operator DNA but competewith the formation of repressing cI-Max or cI-Omomychomodimers

Figure 4 DNA binding, (a) Omomyc binds speci®cally, with alow a�nity, the Myc/Max recognition site CACGTG. Binding ofa GST-Omomyc fusion protein (500 ng) to 32P-labelled DPoligonucleotide, containing the CACGTG E box, was analysedby gel shift in the absence (lane 1) or presence of an excess of coldoligonucleotides containing wild type (DP, lane 2) and variant Eboxes (VZ, lane 3, CATTTG; ZT, lane 4, CAGGTT). Omomycbinding was observed only at protein concentration higher than250 ng. (b) Omomyc competes with binding of Max/Max andMyc/Max complexes to their preferred binding site (CACGTG Ebox). Mobility shift assays with the DP probe were performedwith di�erent amounts of Myc, Max and Omomyc GST-fusionproteins. The locations of gel shifts corresponding to di�erentdimers are indicated by arrows. Lanes 1 ± 10: a constant amountof Max (20 ng), in the presence of increasing amounts (20 ±200 ng) of Myc or Omomyc (lanes 3 ± 6 and 7 ± 10, respectively),was incubated with the DP oligonucleotide. Max antiserum C124was used in the supershift experiments (lanes 1, 6 and 10). In ourhands, this antiserum caused both supershift (lanes 6 and 10) andDNA binding inhibition (lanes 1, 6 and 10). Although DNAbinding inhibition only was observed in lane 1 in this particularcase, the identity of the Max/Max complex has been con®rmed inseveral other independent experiments. Lanes 11 ± 13: DNAbinding of Omomyc (400 ng, lane 11) and Omomyc/Myccomplexes (200 ng each, lane 12; 400 ng each, lane 13) wasmeasured. Lanes 14 ± 16: increasing amounts (100, 200, 400 ng) ofOmomyc were incubated in the presence of constant amounts ofMyc (100 ng) and Max (20 ng) proteins. Shifts corresponding toheterodimers have a di�erent mobility with respect to homo-dimers. Myc/Max (lanes 3 ± 5) and Omomyc/Max dimers (lanes7 ± 9) bind the E box more e�ciently than Omomyc/Myc dimers(lanes 12 and 13)

OmomycL Soucek et al

2467

higher than Omomyc dimers (lane 11). DNA bindinga�nity of Omomyc/Myc heterodimers was instead verylow, as we were unable to observe a speci®c shift evenat relatively high protein concentration (lanes 12 ± 13).We ®nally tested the capacity of Omomyc to interferewith Myc/Max dimerization and DNA binding. In thepresence of increasing amounts of Omomyc protein,DNA binding of Myc/Max complexes was decreased,in parallel with the appearance of DNA boundOmomyc/Max complexes (lanes 14 ± 16).These results indicated that the Omomyc protein

interferes with Myc/Max dimer formation andcompetes, as homodimer and as heterodimer withMax, for DNA binding on the same sites targeted bythe various Max network dimers. Moreover, theysuggest that Omomyc can bind Myc and sequester itinto complexes inactive in DNA binding and,consequently, in transcriptional activation.

Omomyc interferes with Myc induced transactivation

The ability of Omomyc to interfere with dimerizationand DNA binding of Myc/Max complexes suggestedthat it might a�ect transcriptional regulation by Myc/Max dimers in mammalian cells.To establish the potential of Omomyc to a�ect

transcriptional regulation by Myc/Max complexes,BOSC23 cells were transfected with expressionvectors and the plasmid M4minCAT, containing theCAT reporter gene linked to a minimal thymidinekinase promoter with four repeats of the CACGTG E

box. CAT activity of the reporter was increased aboutfourfold by a cotransfected Myc expression plasmid,compared with control vector (Figure 5). Omomyccaused instead a 50% repression of the basal level ofCAT activity and a signi®cant inhibition of transcrip-tional activation by Myc. This e�ect was dependenton the E box as no repression or activation wasobserved with the minCAT reporter, devoid of the Eboxes (data not shown). As Myc transactivationdepends on the formation of Myc/Max complexes,this result might be explained by Omomyc inhibitionof both Myc and the endogenous Max protein,present in relatively large amounts. To distinguishbetween the two possible e�ects of Omomyc, weperformed transfections with a vector expressing atruncated Myc protein (DMyc), containing thebHLHZip domain only. DMyc is unable to transacti-vate, because it lacks the Myc activation domain, anddimerizes with Max, but not with Myc. This proteinneither repressed CAT activity or inhibited Mycmediated transactivation (Figure 5), suggesting that,although Omomyc can bind Myc and Max equallywell (Figure 4), transcriptional repression is mainlydue to Myc inhibition. This was further con®rmed bythe ®nding that repression of CAT activity byOmomyc was reversed by an excess of Myc protein(Figure 5, last column).

Omomyc inhibits cell proliferation in mammalian cells

Since the Myc protein is required for cell cycleprogression, the ability of Omomyc to a�ect cellproliferation was tested, by a colony formation assayin NIH3T3 cells. These cells were transfected withOmomyc and Myc expression vectors together with aneomycin resistance plasmid; number and size ofneomycin resistant colonies were scored 2 weeks later(Figure 6). Myc expression plasmid did not a�ectcolony formation, whereas Omomyc caused a *tenfoldreduction in colony number; the Omomyc colonieswere on the average one third the diameter of thoseobtained with Myc expression or empty vectors. Thise�ect is Myc dependent, at least in part, since DMyc,which does not interact with Myc, a�ected colonyformation in an intermediate way between Myc andOmomyc and since Myc overexpression rescued theproliferation arrest caused by Omomyc. These datastrongly suggest that Omomyc a�ects cell proliferationby interfering with transcriptional regulation of Myc/Max target genes.

Discussion

The basic cellular processes of proliferation, differ-entiation and death, are controlled by complexregulatory pathways triggered by extracellular signalsand converging to regulation of gene expression. In thiscontext, bHLH(Zip) transcriptional regulators (MyoD,Achaete-Scute, Mash, NeuroD, E2A, Myc, Max etc)are often utilized as a crucial switch, for instance inmuscle cells and in cells of the nervous and immunesystems. Our aim was to understand the rulesunderlying protein-protein recognition in this class ofproteins in order to develop tools for interfering withsuch transcription regulatory networks.

Figure 5 Repression of c-Myc mediated trascription byOmomyc. BOSC23 cells were cotransfected with the M4minCatreporter plasmid, containing four CACGTG binding sites, andwith the indicated expression plasmids. CAT activity wasdetermined from cell lysates 72 h later and normalized to b-galactosidase activity derived from a pRSV b-gal co-transfectedplasmid. CAT values are relative to that for the M4minCATreporter and an empty expression vector. Omomyc e�ectivelyinhibited CAT activity and prevented transactivation by Myc, in adose-dependent manner. The Myc bHLHZip domain (DMyc) didnot have the same e�ect. Repression by Omomyc was reversedwhen Myc was overexpressed. Mean values and standarddeviations from at least three independent experiments arepresented

OmomycL Soucek et al

2468

The signal transduction pathways that have animpact on transcription factors networks are intricateas well as branched. Therefore, it proves to be di�cultcontrolling these networks from the outside, e.g. byacting on the extracellular signal, their receptors or ona component of the transduction machinery, such asthe Ras protein. Therefore we have decided to follow adi�erent strategy, that consists in the design, bymolecular modelling, of a new protein domain whichacts from within, by altering the balance amongdi�erent factors of one of these networks, the Maxnetwork. Naturally occurring dominant negativemutations in bHLHZip proteins, like those in the Migene (Moore, 1995), as well as previously designeddominant mutations (Krylov et al., 1997) a�ect thefunction of the basic region, thus abolishing DNAbinding. Our target was instead protein dimerization.Myc/Max dimers are key players of the Max network.Previous works (Amati et al., 1993; Marchetti et al.,1995; Muhle-Goll et al., 1995) showed the importanceof the zipper region for Myc dimerization speci®city;the Myc HLHZip exchange mutant MycRX, a Myccontaining the HLHZip domain of Max, was shown tohave a dominant negative activity on wild type Mycfunction (Amati et al., 1993). The present workrepresents a further advancement for two reasons: (i)We were able to identify four amino acids (E57, E64,R70, R71) that, when mutated, modify Myc molecularrecognition; (ii) These critical residues are all includedin a region of only 14 amino acids within the leucine

zipper and two of them (E57 and R70) are conservedamong all members of the Myc protein family (c-Myc,v-Myc, N-Myc, S-Myc, L-Myc, d-Myc; Table 1). Thisimplies that this small region may represent a lesscomplex structural target for Myc activity inhibition.Myc E57 and E64 are thought to be involved ininterhelical electrostatic interactions, together withMax H60, responsible for heterodimerization specifi-city (Lavigne et al., 1995). Our data are only partlyconsistent with this model, as substitution of the twonegatively charged glutamic residues with a polar and anonpolar aminoacid (T and I) a�ected Myc hetero-dimerization with Max only weakly (Figure 3b), whileit turned out to be essential for Myc homodimeriza-tion. By substituting the four amino acids we obtaineda new bHLHZip domain, Omomyc, that homodimer-ized as strongly as Max and e�ciently formedheterodimers with wild type Myc, as well as with Max.Since the proteins in the Max network act on a

speci®c DNA target site (the CACGTG E box), wetested the ability of Omomyc to compete with othernetwork dimers for binding to this site; speci®c bindingof this protein to the E box was demonstrated. TwoMyc basic regions (present in Omomyc/Omomyc andOmomyc/Myc dimers) were ine�cient in binding to theE box (Figure 4), pointing to a speci®c role of the Maxbasic region in target site recognition (see also Amati etal., 1993). DNA binding of the Omomyc/Max dimerwas instead observed at lower protein concentration,but these dimers, which lack any transactivatingregion, are transcriptionally inactive. On the basis ofthese observations we predicted that Omomyc expres-sion in mammalian cells should result in a dominantphenotype as a consequence of interference with Mycphysiological functions: transcriptional regulation andcell growth control. Transient transfections demon-strated that Omomyc a�ects transcription of apromoter containing the E box CACGTG and inhibitsMyc transactivating capacity in a dose dependentmanner. These e�ects are mainly due to complexformation with Myc, rather than to sequestration ofMax. This is shown by the inability of the DMycprotein, unable to bind Myc while still interacting withMax, to have a comparable e�ect and by the ®ndingthat Myc overexpression overcomes the transcriptionalinhibition exherted by Omomyc. Finally, and coher-ently with the results of the transactivation assays,Omomyc was able to arrest cell proliferation in acolony formation assay in NIH3T3 ®broblasts. Thedistinctive dimerization property of Omomyc is a toolto better dissect Myc mechanism of action in cell cycle

Figure 6 Inhibition of cell growth by Omomyc ±ColonyFormation E�ciency. NIH3T3 cells were transfected with theempty vector, Omomyc and Myc expression vectors, together witha neomycin resistance plasmid. For each individual experiment,cells were split 18 ± 24 h post-transfection and selected for 2weeks. (a) shown are representative plates stained after 2 weeksselection. Control corresponds to cells transfected with emptypCS vector. (b) indicated are average number and diameter ofneomycin resistance colonies from independent experiments. TheOmomyc vector was transfected alone or together with a Mycexpression vector in a threefold excess

Table 1 Comparison of Myc family proteins leucine zippers

Leucine zipper domains of Myc family proteins are aligned toemphasize identical and homologous residues (respectively high-lighted by black and grey boxes)

OmomycL Soucek et al

2469

control. Furthermore our results support the view thatsequestration of endogenous Myc is an e�ectivestrategy to control cell proliferation.This encourages us to utilize Omomyc or a

Omomyc-like peptide for interfering with the mostrelevant biological aspect of Myc protein function: celltransformation. To this end, retroviral or adenoviralvectors can be exploited for peptide delivery inside thecell. Alternatively, shorter peptides with Omomyc-likeproperties could be designed and delivered to thecytoplasm of Myc dependent tumor cells by fusion toan Antennapedia homeodomain. Finally these shorterpeptides could be used as leads for the design ofsmaller peptide-mimetics with better pharmacologicalproperties.The results that we have obtained with Omomyc

prove the validity of the strategy we adopted and openthe way to the design of other mutants, displaying anincreased selectivity in their interaction with Myc.

Materials and methods

Molecular modelling

Model structures of human Max, Myc and mutant MycbHLHZip domain homo- and heterodimers were built byhomology using the Max dimer three-dimensional crystalstructure (FerreÁ -D'AmareÁ et al., 1993) as a template.Modelling was carried out on a Silicon Graphics O2workstation R5000SC, using the Insight II program.Models were subjected to 100 cycles of energy re®nement(Discover program, steepest descent algorithm, BiosymTechnologies). Stability of dimeric structures was evaluatedby the ESCHER automatic procedure (Ausiello et al.,1997) and by visual inspection.

Plasmids, mutagenesis and protein expression

pC135Myc and pC135Max plasmids, containing thebHLHZip domains of human myc (amino acid residues348 ± 439) and max (aminoacids 11 ± 99) genes were alreadydescribed (Marchetti et al., 1995). pC135Myc was used as atemplate for PCR mutagenesis, by the StratageneQuikChange site directed mutagenesis kit, primed byspeci®c oligonucleotides containing the desired mutations.The following oligonucleotides were used for R70Q andR71N substitutions (Myc-mut2): 5'-GTT GCG GAA ACAAAA CGA ACA GTT GA-3' and 5'-TCA ACT GTT CGTTTT GTT TCC GCA AC-3'. pC135Myc-mut2 was thenused as a template for a second mutagenesis round, primedby the following oligonucleotides introducing E57N andE64I substitutions: 5'-CAA GCA GAG ACG CAA AAGCTC ATT TCT GAA ATC GAC TTG TTG-3' and 5'-CAA CAA GTC GAT TTC AGA AAT GAG CTT TTGCGT CTC-3'. The plasmid containing the Myc mutantwith the four substitutions was called pC135-Omomyc(Mycmut4). The bHLHZip domains in all these constructsare fused in frame with lambda phage cI repressor N-terminal domain (aminoacids 1 ± 132) at the 5' end and theb-galactosidase a-peptide gene at the 3' end; theirexpression is driven by the lac promoter. All constructswere sequenced by the Sanger method to con®rm themutations identity and the proper reading frame. pC135* issimilar to pC135, but has three mutations that replacethree residues in the l repressor DNA binding domain withthe corresponding amino acids of the 434 phage repressor(Longo et al., 1995; Marchetti et al., 1995). pACYC-bHLHZip plasmids were constructed by replacing the 1690bp HindII-HindIII fragment of pACYC184 (Chang andCohen, 1978) with PvuII-PvuII fragments (containing the

Plac-cI-bHLHZ-a-peptide fusions) excised from pC135-bHLHZip plasmids. The chimerical plasmids were intro-duced by electroporation into Sure cells (from Stratagene),which were selected in ampicillin and/or chloramphenicolsupplemented medium. The GST (glutathione-S-transfer-ase)-Omomyc fusion was constructed in the pGEX-2TKvector (Pharmacia). The Omomyc DNA sequence wasampli®ed from pC135Omomyc by polymerase chainreaction (PCR) primed by the following oligonucleotides:5'GGA GAG CCC GGG GTC GAC CGA GGA GAA3',containing a SmaI site, and 5'ATG TCA ATC ATA TGTACC CCG GTT3', located downstream of the codingsequence. The omomyc gene was excised as a SmaI-EcoRIfragment and cloned in the pGEX-2TK vector, in framewith the C terminus of GST.

For the detection of cI-bHLHZip proteins, bacterial cellstransformed with di�erent expression vectors were lysed insample bu�er; the extracts were used in immunoblotting,probed with Max (C-124) and Myc (237/6) speci®c antisera.

Chimeric repressor dimerization assays

Phage Immunity Test l phage sensitivity of bacterial cellstransformed by pC135bHLHZip plasmids was tested byspot tests, at concentrations varying from 105 to 101 phages(lDHK54) per spot on lawns of transformed bacteria.

b-galactosidase assay For a quantitative assay of chimericrepressor dimerization, the pC135bHLHZip plasmids wereintroduced into the GC421 strain, which constitutivelyexpresses b-galactosidase under the control of lPrOr (rightpromoter-operator). Chimerical repressors dimerizationwas measured by % inhibition of b-galactosidase activityin trasformed bacteria with respect to b-galactosidaseactivity in non trasformed ones, after bacterial lysis asdescribed (Marchetti et al., 1995; Miller et al., 1972).

Heterodimerization Assay Heterodimerization ability wasestimated by the increase of b-galactosidase activity inGC421 cells co-trasformed with pACYCbHLHZip andpC135*bHLHZip, with respect to the inhibition obtainedin cells transformed with pACYCbHLHZip only.

Electrophoretic mobility shift assays (EMSA)

GST-bHLHZip fusion proteins were a�nity-puri®ed onglutathione-agarose beads and then eluted with reducedglutathione; amounts of puri®ed GST fusion proteins wereused in DNA binding assays. A 30 min incubation at 378Cwas performed to allow dimers exchange; speci®c (DP) andnon-speci®c (ZT, VZ) competitor oligonucleotides wereadded and a further 15 min incubation at room tempera-ture was performed to allow the DNA-protein complex toform. One ml of anti-Max antiserum (C124, from SantaCruz) was added to the reaction mixture for the supershiftexperiments. 104 c.p.m. of 32P end-labelled double-strandedDP oligonucleotide was used per reaction; speci®c and nonspeci®c competitors were added at a 100-fold excess. TheDP oligonucleotide (5'-CGGGGCGAGACCACGTGAC-CCCGC-3', Marchetti et al., 1995) includes the CACGTGE box. In ZT and VZ oligonucleotides the E box sequenceCACGTG is replaced respectively by the sequencesCAGGTT and CATTTG.

CAT assays

For the construction of expression vectors, Omomyc andDMyc (Myc bHLHZip domain) sequences were ampli®edby PCR with Pfu polymerase, primed by oligonucleotidesproviding an ATG and a ribosome binding site, andinserted downstream of the CMV promoter in the pCS+plasmid (from D Turner); the Myc expression vector was

OmomycL Soucek et al

2470

pBabe-neo-c/vMyc (Crouch et al., 1990). BOSC23 cellswere transfected with calcium phosphate (36105 cells in a35 mm dish and 5 mg DNA per experiment). One mg of thechloramphenicol acetyltransferase (CAT) reporter plasmidM4minCAT (from R Eisenman) was co-transfected with0.3 mg of pRSVb-gal in order to assess transfectione�ciencies; total amount of transfected DNA wasbalanced in all experiments by empty pCS plasmid. CATactivity was assayed by the di�usion method of Neuman etal. (1987); CAT values were corrected for transfectione�ciency by normalization to b-galactosidase values. Allassays were performed at least three times; average valuesand standard deviations are reported.

Colony Forming E�ciency (CFE) assays

NIH3T3 cells were transfected with calcium phosphate(26106 cells in a 10 cm dish and 15 mg DNA perexperiment). The expression vectors were co-transfected

with the neomycin resistance plasmid pSV2neo (3 mg). Foreach individual experiment, cells were split into at leasttriplicate dishes (1/4, 1/8, 1/16 dilutions) 18 ± 24 h post-transfection. After 48 h, 800 mg/ml G418 was added. Twoweeks later, colonies were ®xed in methanol and stainedwith Giemsa.

AcknowledgementsWe gratefully acknowledge R Eisenman, G Meroni and FTatoÁ for generously providing reagents and advice. Wethank S Burley for providing the atomic coordinates of theMax dimer. We thank Nicola Rizzo for expert technicalassistance and R Gargamelli for photographs. LS wassupported by a fellowship from A Buzzati Traversofoundation. This work was supported by grants fromAIRC (Italian Association for Cancer Research), CNR,MURST and Telethon to GC and SN.

References

Alevizopoulos K, Vlach J, Hennecke S and Amati B. (1997).EMBO J., 16, 5322 ± 5333.

Amati B, Dalton S, Brooks MW, Littlewood TD, Evan GIand Land H. (1992). Nature, 359, 423 ± 426.

Amati B, Brooks MW, Levy N, Littlewood TD, Evan GI andLand H. (1993). Cell, 72, 233 ± 245.

Amati B and Land H. (1994). Curr. Opin. Genet. Dev., 4,102 ± 108.

Ausiello G, Cesareni G and Helmer-Citterich M. (1997).Proteins, 28, 556 ± 567.

Ayer DE and Eisenman RN. (1993). Genes Dev., 7, 2110 ±2119.

Ayer DE, Kretzner L and Eisenman RN. (1993). Cell, 72,211 ± 222.

Bello-Fernandez C and Cleveland JL. (1992). Curr. Top.Microbiol. Immunol., 182, 445 ± 452.

Benvenisty N, Leder A, Kuo A and Leder P. (1992). GenesDev., 6, 2513 ± 2523.

Berberich SJ and Cole MD. (1992). Genes Dev., 6, 166 ± 176.Blackwell TK, Kretzner L, Alt FW, Blakwood EM,Eisenman RE and Weintraub H. (1990). Science, 250,1149 ± 1152.

Blackwood EM and Eisenman RN. (1991). Science, 251,1211 ± 1217.

Blackwood EM, Luscher B and Eisenman RN. (1992). GenesDev., 6, 71 ± 80.

Castagnoli L, Vetriani C and Cesareni G. (1994). J. Mol.Biol., 237, 378 ± 387.

Chang ACY and Cohen SN. (1978). J. Bacteriol., 134, 1141 ±1156.

Chothia C and Lesk AM. (1986). EMBO J., 5, 823 ± 826.Cory S. and Adams J. (1988). Annu. Rev. Immunol., 6, 25 ±48.

Crouch DH, Lang C and Gillespie DA. (1990). Oncogene, 5,683 ± 689.

Eilers M, Schirm S and Bishop JM. (1991). EMBO J., 10,133 ± 141.

Ellenberger T, Fass D, Arnaud M, and Harrison SC. (1994).Genes Dev., 8, 970 ± 980.

Evan GI, Wyllie AH, Gilbert CS, Littlewood TD, Land H,Brooks M, Waters CM, Penn LZ and Hancock DC.(1992). Cell, 69, 119 ± 128.

FerreÁ -D'AmareÁ AR, Prendergast GC, Zi� EB and BurleySK. (1993). Nature, 363, 38 ± 44.

Galaktionov K, Chen X and Beach D. (1996). Nature, 382,511 ± 517.

Grandori C, Mac J, Siebelt F, Ayer DE and Eisenman RN.(1996). EMBO J., 15, 4344 ± 4357.

Gu W, Cechova K, Tassi V and Dalla-Favera R. (1993).Proc. Natl. Acad. Sci. USA, 90, 2935 ± 2939.

Henriksson M and Luscher B. (1996). Adv. Cancer Res., 68,109 ± 182.

Hu JC, O'Shea EK, Kim PS and Sauer RT. (1990). Science,250, 1400 ± 1403.

Hurlin PJ, Que va C, Koskinen PJ, Steingrimsson E, AyerDE, Copeland NG, Jenkins NA and Eisenman RN.(1995). EMBO J., 14, 5646 ± 5659.

Hurlin PJ, Que va C and Eisenman RN. (1997). Genes Dev.,11, 44 ± 58.

Jones RM, Branda J, Johnston KA, Polymenis M, Gadd M,Rustgi A, Callanan L and Schmidt EV. (1996). Mol. Cell.Biol., 16, 4754 ± 4764.

Koskinen PJ, Ayer DE and Eisenman RN. (1995). CellGrowth Di�er., 6, 623 ± 629.

Kretzner L, Blackwood EM and Eisenman RN. (1992).Nature, 359, 426 ± 429.

Krylov D, Kasai K, Echlin DR, Taparowsky EJ, ArnheiterH and Vinson C. (1997). Proc. Natl. Acad. Sci. USA, 94,12274 ± 12279.

Laherty CD, Yang WM, Sun JM, Davie JR, Seto E andEisenman RN. (1997). Cell, 89, 349 ± 356.

Lavigne P, Kondejewski LH, Houston Jr ME, SonnichsenFD, Lix B, Sykes BD, Hodges RS and Kay CM. (1995). J.Mol. Biol., 254, 505 ± 520.

Lee TC, Li L, Philipson L and Zi� EB. (1997). Proc. Natl.Acad. Sci. USA, 94, 12886 ± 12891.

Leone G, DeGregori J, Sears R, Jakoi L and Nevins JR.(1997). Nature, 387, 422 ± 426.

Lewis BC, Shim H, Li Q, Sun Wu C, Lee LA, Maity A andDang CV. (1997). Mol. Cell. Biol., 17, 4967 ± 4978.

Littlewood TD and Evan G. (1994). Protein Pro®le, 1, 639 ±709.

Lombardi L, Newcomb EW and Dalla-Favera R. (1987).Cell, 49, 161 ± 170.

Longo F, Marchetti MA, Castagnoli L, Battaglia PA andGigliani F. (1995). Biochem. Biophys. Res. Commun., 206,326 ± 334.

Marchetti A, Abril-Marti M, Illi B, Cesareni G and Nasi S.(1995). J. Mol. Biol., 248, 541 ± 550.

Meroni G, Reymond A, Alcalay M, Borsani G, Tanigami A,Tonlorenzi R, Lo Nigro C, Messali S, Zollo M, LedbetterD, Brent R, Ballabio A and Carrozzo R. (1997). EMBO J.,16, 2892 ± 2906.

OmomycL Soucek et al

2471

Miller JH. (1972). Experiments in Molecular Genetics, ColdSpring Harbor Laboratory Press: Cold Spring Harbor,New York.

Moore KJ. (1995). Trends Genet., 11, 442 ± 448.Muhle-Goll C, Nilges M and Pastore A. (1995). Biochem.,34, 13554 ± 13564.

Murre C, McCaw PS, Vaessin H, Caudy M, Jan LY, Jan YN,Cabrera CV, Buskin JN, Hauschka SD, Lassar AB,Weintraub H and Baltimore D. (1989). Cell, 58, 537 ± 544.

Neuman JR, Morency CA and Russian KO. (1987). Bio.Techniques, 5, 444 ± 447.

Penn LJZ, Brooks MW, Laufer EM and Land H. (1990).EMBO J., 9, 113 ± 121.

Peukert K, Staller P, Schneider A, Carmichael G, Hanel Fand Eilers M. (1997). EMBO J., 16, 5672 ± 5686.

Philipp A, Schneider A, Vasri I, Finke K, Xiong Y, Beach D,Alitalo K and Eilers M. (1994).Mol. Cell. Biol., 14, 4032 ±4043.

Prendergast GC, Laue D and Zi� EB. (1991). Cell, 65, 395 ±407.

Sears R, Ohtani K and Nevins JR. (1997). Mol. Cell. Biol.,17, 5227 ± 5235.

Seth A, Gupta S and Davis RJ. (1993). Mol. Cell. Biol., 13,4125 ± 4136.

Sippl MJ. (1993). Proteins, 17, 355 ± 362.Sommer A, Hilfenhaus S, Menkel A, Kremmer E, Seiser C,Loidl P and Luscher B. (1997). Curr. Biol., 7, 357 ± 365.

Steiner P, Philipp A, Lukas J, Godden-Kent D, Pagano M,Mittnacht S, Bartek J and Eilers M. (1995). EMBO J., 14,4814 ± 4826.

Vlach J, Hennecke S, Alevizoupolos K, Conti D and AmatiB. (1996). EMBO J., 23, 6595 ± 6604.

Weiss WA, Aldape K, Mohapatra G, Feuerstein BG andBishop JM. (1997). EMBO J., 16, 2985 ± 2995.

Zervos AS, Gyuris J and Brent R. (1993). Cell, 72, 223 ± 232.Zu X, Rudchenko S, Kwon-kin W and Calame K. (1997).

Genes Dev., 11, 654 ± 662.

OmomycL Soucek et al

2472