operator search by mutant lac repressors

J. Mol. Biol. (1998) 278, 549±558

Operator Search by Mutant Lac Repressors

Andrew Barker*, Reimund Fickert, Stefan Oehler andBenno MuÈ ller-Hill

Institut fuÈ r Genetik derUniversitaÈt zu KoÈln, Weyertal121, D-50931 KoÈln, Germany

Present addresses: R. Fickert, SGPD-40237 DuÈ sseldorf, Germany; S. OGenetics, University of Cambridge,Cambridge CB2 3EH, UK.

Abbreviations used: PMSF, pheny¯uoride; BSA, bovine serum albumsulphoxide.

0022±2836/98/180549±10 $25.00/0/mb9

The Escherichia coli Lac and Gal repressors are two members of a largefamily of bacterial repressor proteins that share signi®cant sequence andstructural homology. Ef®cient repression by all family members requiresspeci®c binding to a site or sites close to the transcriptional start of thegenes regulated. Both LacR and GalR have to bind to at least two sitesfor ef®cient repression, yet they differ in one important respect: LacR is ahomotetramer whereas GalR is a homodimer. In an attempt to under-stand this difference, we studied the operator binding activity of a LacRvariant that has the DNA-binding speci®city of GalR (LacR-V17A18).A tetrameric version of this protein shows a 30-fold decrease in associ-ation rate to operator located on a long (l) DNA molecule, in comparisonto wild-type LacR, while a dimeric version of this protein shows an unal-tered association rate in comparison to dimeric LacR. This reduction inassociation rate correlates with a broadened DNA-binding speci®city forbase-pairs 4 and 5 of the operator: examination of an additional LacRvariant with an even broader DNA-binding speci®city indicates that atetrameric version also shows a 30-fold decrease in association rate incomparison to wild-type LacR, while a dimeric version again shows anunaltered association rate in comparison to dimeric LacR. This differencein association rate in vitro correlates with whether a tetrameric or dimericvariant of LacR of a given DNA-binding speci®city will repress lacZunder control of a single operator more ef®ciently in vivo. We thereforepropose that the formation of stable homotetramers becomes a distinctdisadvantage unless a high degree of DNA-binding speci®city is also pre-sent, and demonstrate that this in indeed the case for GalR-mediatedrepression of the gal operon. This functional constraint seems to havein¯uenced the evolution of the LacI-GalR family of repressors, most ofwhich have a relatively broad speci®city of DNA-binding and most ofwhich form only stable homodimers.

# 1998 Academic Press Limited

Keywords: lac operon; gal operon; DNA-protein interactions; operatorbinding speci®city; operator search mechanisms*Corresponding author

Introduction

Sequence-speci®c interactions of proteins andDNA are central to all aspects of the utilisation ofgenetic information in all organisms. The LactoseRepressor (LacR) of Escherichia coli has served as aparadigm for such interactions, even before the

, Herderstr. 18,ehler, Department ofDowning St.,

lmethylsulphonylin; DMSO, dimethyl

81729

chemical nature of the interacting partners wasclear (Jacob & Monod, 1961). LacR is one memberof a large family of related proteins, sharing signi®-cant sequence similarity with Gal Repressor (vonWilcken-Bergmann & MuÈ ller-Hill, 1982) and anumber of other repressor proteins and periplas-mic sugar-binding proteins found both in E. coliand other prokaryotes, known as the LacI-GalRfamily (MuÈ ller-Hill, 1983; Weickert & Adhya,1992). Structural studies have shed further light onthe similarities between the regulatory proteinswithin this family. LacR and the related PurineRepressor (PurR) both consist of a globular corethat contains the effector-binding site, and an N-terminal protrusion, containing the DNA-binding

# 1998 Academic Press Limited

Figure 1. Association rate (ka) constants in vitro for LacRvariants with lac operator fragments differing in length.The indicated lac operator sequences were cloned intoplasmid or l cloning vectors and association rates weredetermined with puri®ed LacR containing substitutionsat positions 17 and 18 as depicted. The operator bases atpositions 4 and 5 contacted by residues 17 and 18 ofLacR are in bold type. A dash in the centre of operatorsequences indicates the centre of symmetry. Wild-typeLacR has the sequence Y17Q18.

550 Operator Search by LacR Mutants

domain (Adler et al., 1972; Friedman et al., 1995;Lewis et al., 1996; Slijper et al., 1996, 1997;Schumacher et al., 1994, 1995). In both cases,sequence-speci®c binding occurs when a singlerepressor dimer contacts a quasi or fully symmetricsite on DNA, and this binding is suf®cient forrepression of transcription from the adjacent pro-moter by directly competing with RNA polymer-ase for binding (Schlax et al., 1995).

In many cases, this simplistic view is compli-cated by the need for multiple repressor dimers tocontact multiple operators to achieve full repres-sion. Extensive studies have indicated that coop-erative binding to at least two operators isessential for full repression by both LacR and GalR(e.g. see Bellomy et al., 1988; Haber & Adhya, 1988;Oehler et al., 1990, 1994; Choy & Adhya, 1992; Lawet al., 1993; Aki et al., 1996; MuÈ ller et al., 1996).Despite this need for cooperativity with both pro-teins, they differ in one important respect: LacR isa homotetramer (MuÈ ller-Hill et al., 1971), evenwhen diluted to at least 10ÿ13 M (Levandoski et al.,1996), while GalR is a homodimer (Nakanishi et al.,1973) with no evidence of higher-order oligomersin solution (Majumdar et al., 1987). This is due tothe absence, in GalR, of a sequence similar to theC-terminal heptad repeat, or mini leucine-zipper,responsible for tetramerisation of LacR dimers(Lehming et al., 1988; Alberti et al., 1991; Brenowitzet al., 1991; Chakerian & Mathews, 1991; Friedmanet al., 1995; Lewis et al., 1996).

We have sought to understand the reason forthis difference by studying the operator bindingactivity of a LacR variant that has the DNA-bind-ing speci®city of GalR, and show here that the tet-rameric version of this protein functions poorly incomparison to a dimeric version both in vitro andin vivo, while the opposite is true for wild-typeLacR. A similar defect is observed with other LacRvariants with broadened DNA-binding speci®city.

Results

Alteration in association kinetics of LacRrecognition helix mutants

We ®rst examined the kinetic parameters ofoperator binding of puri®ed tetrameric and dimericwild-type LacR-Y17Q18 (LacR has the sequenceY17Q18, and residues 17 and 18 are residues 1 and2 of the recognition helix, respectively) and two ofits variants using a ®lter-binding assay in vitro:LacR-V17A18, which has the same DNA-bindingspeci®city as GalR (Lehming et al., 1987, 1990); andan additional variant, LacR-H17Q18. The ideal oper-ator (Oid; Sadler et al., 1983; Simons et al., 1984)was used as a target DNA, cloned either into aplasmid DNA or l vector (Materials and Methods),in order to examine the dependence of length oftarget DNA on association rate ka (Fickert &MuÈ ller-Hill, 1992). In addition, the operator variant(O41) for which LacR-V17A18 has the highest af®nity

(Lehming et al., 1990) was also examined, for thatvariant only.

A marked dependence of ka on the length of theDNA in which the operator is presented wasmeasured for tetrameric, but not dimeric, versionsof LacR (Figure 1). As measured previously(Fickert & MuÈ ller-Hill, 1992), wild-type, tetramericLacR shows at least a fourfold increase in ka to Oid

when the operator is located on a long (49,000 bp)DNA, compared to a short molecule (2455 bp). Theassociation rates of dimeric LacR do not exhibitthis length dependence, being the same for bothlong and short target DNAs (Figure 1; and seeFickert & MuÈ ller-Hill, 1992). LacR-H17Q18 tetramershows a four-fold decrease in ka to operator locatedon a long DNA, compared to wild-type, while ka toa short DNA remains unaffected. LacR-V17A18 tet-ramer shows a 30-fold decrease in ka compared towild-type, both with Oid and its ``ideal'' operatorO41, again, only to operator located on a long DNA(Figure 1). No change in ka to a short DNA isobserved for either of these LacR variants, nor is achange in ka observed with dimeric forms of theseproteins to target DNAs of any length (Figure 1).

Additional experiments were performed toensure we were not observing a competitive effectdue to the excess (calculated in terms of base-pairs)of non-speci®c DNA when operator was presentedin a l molecule. Association rates were determinedfor both tetrameric and dimeric LacR (wild-type)in the absence or presence of an equimolar amountof l DNA not containing lac operator, or the pre-sence of an equal weight (i.e. an approximately200-fold molar excess) of an approximately 200 bpfragment containing either no operator-likesequences or seven copies of a half-operatorsequence: none of the competitor DNAs examinedhad a measurable effect on ka (Fickert, 1992). Fur-thermore, although the exact oligomeric state ofdimeric LacR at the concentration used in this

Operator Search by LacR Mutants 551

experiment is uncertain (Royer et al., 1990; but seealso Levandoski et al., 1996, for a con¯icting resultfor tetrameric LacR) we ®nd no evidence for a sig-ni®cant contribution of a monomer-dimer equili-brium to association of dimeric LacR of anyspeci®city to DNA of any length (not shown). Notealso that the values for ka presented here are con-sistently approximately 30% higher than thosereported previously (Fickert & MuÈ ller-Hill, 1992;Fickert, 1992); we have no explanation for thisdifference.

Specificity of operator binding in vivocorrelates with fast association rates in vitro

As the magnitude of the decrease in ka of tetra-meric LacR-V17A18 is unexpected, an explanationfor this decrease is required. An examination of thespeci®city of repression in vivo exhibited by wild-type LacR and all possible variants at positions 17and 18 (Lehming et al., 1990) offers an explanation.As indicated in Figure 2, a broadening of bindingspeci®city is observed for these variants, whichcorrelates with the decrease in ka to a long DNAmolecule. This explanation also offers a testableprediction: an additional variant, LacR-S17S18,which exhibits an even broader degree of speci-®city (Figure 2; Lehming et al., 1990), should also

Figure 2. Repression values for LacR variants in vivo witLacR variants, encoded on plasmid pBR322 derivatives (Lehpositions 17 and 18 as indicated. Operator (reporter) cons(Lehming et al., 1987). The values with ideal lac operator arestitutions at positions 4 and 5 (numbered from the centrebelow. The ideal operator is marked with an asterisk (*)(pred.) from the data presented in Table 2 of Lehming et al.for dimeric (di.) and tetrameric (tet.) variants, as indicated.accurately determined in this system, due to occasional loss

exhibit a strong decrease in ka to operator on along DNA if a tetramer, but not a dimer. Asshown in Figure 1, this is indeed the case. At leasta 30-fold decrease in ka for tetrameric LacR-S17S18 isobserved for both Oid and its ``ideal'' operator(O4151) as predicted by Lehming et al. (1990), incomparison to wild-type LacR. On the other hand,no change is apparent for dimeric LacR-S17S18 foreither operator, in comparison to dimeric LacR.Conversely, a variant that has a high degree ofbinding speci®city should also show an unalteredka to a long DNA as a tetramer, compared to wild-type. LacR-M17K18 is predicted to be such a variant(Lehming et al., 1990), but despite considerableeffort, we were unable to purify the tetrameric ver-sion of this protein in an active form (data notshown).

The data presented in Figure 2 were originallypresented in Lehming et al. (1990) as a theoreticalprediction of repression values for any combintionof mutations at positions 17 and 18, based on ananalysis of repression values obtained for all poss-ible single mutations at both positions: these valuesare shown as the predicted repression values inFigure 2. Such an analysis is possible because theeffects of single mutations at these two positionsare by and large additive, as has been experimen-tally veri®ed in a number of cases, including LacR-

h 90 � lacI� amounts of LacR (3600 monomers per cell).ming et al., 1987), contained amino acid substitutions attructs were located on plasmid pACYC184 derivativesgiven ®rst, and the values for all possible operator sub-of symmetry, which is indicated by a dash) are listed

where it appears for the second time. Values predicted(1990) are presented, as are experimental values obtainedNote that repression values greater than 200 cannot be

of the repressor-bearing plasmid (Lehming et al., 1987).

Figure 3. Half-life and dissociation rate (kd) constantsin vitro for dimeric LacR-lac operator complexes. Theindicated lac operator sequences were cloned into plas-mid vectors and dissociation rates were determinedwith puri®ed, dimeric LacR (Fickert & MuÈ ller-Hill, 1992)containing substitutions at positions 17 and 18 asdepicted. The operator bases at positions 4 and 5 con-tacted by residues 17 and 18 of LacR are in bold type.A dash in the centre of operator sequences indicates thecentre of symmetry.


H17Q18 and LacR-V17A18 (Lehming et al., 1990).However, as LacR-S17S18 had not previously beensubjected to a similar analysis, we chose to re-evaluate repression values in this system for allvariants used; these are the experimental data pre-sented in Figure 2. The data obtained for bothdimeric and tetrameric versions of wild-type LacR,LacR-H17Q18 and LacR-V17A18 are in agreementwith those obtained previously, with generally aless than twofold variation between dimeric andtetrameric versions (Figure 2; Lehming et al., 1990).The experimental values obtained for LacR-S17S18

differ from the predicted values in that the pre-dicted optimal operator is similar to others. As themeasured dissociation rate for LacR-S17S18±Oid and±O4151 complexes are the same, within experimen-tal error (see below), we chose to keep using O4151,the predicted optimal operator for LacR-S17S18

(Figure 2), in subsequent analyses.Previous experiments indicated that loop for-

mation by a LacR tetramer bound to two operatorson the same DNA, or on two seperate DNAs, ledto non-retention of the complexes on nitrocellulose®lters under the conditions used here (Fickert &MuÈ ller-Hill, 1992). Therefore, it is conceivable thattetramers of broad speci®city exhibit a higherdegree of loop formation due to contacts ofincreased stability to secondary sites than doeswild-type LacR, and that the decreased ef®ciencyof retention on nitrocellulose leads to the ``appar-ent'' reduction in ka observed in Figure 1. How-ever, an appraisal of the raw data gathered fordetermination of ka for all LacR variants precludesthis possibility. For all LacR variants, the amountof operator DNA retained on nitrocellulose oncethe association reaction has reached equilibrium isthe same, within experimental error, for bothdimeric and tetrameric versions (data not shown).

A broadened specificity does not affectdissociation rate

It is also conceivable that binding speci®citymight have an effect on the operator half-life, thatis the kinetic dissociation rate kd. Both LacR-H17Q18±Oid and LacR-V17A18±O41 complexes areknown to have signi®cantly longer half-lives, andtherefore signi®cantly reduced kd, in comparison towild-type LacR±Oid complexes (Lehming et al.,1987). Therefore, we re-determined half-life and kd

for these complexes, and also for the other repres-sor-operator complexes examined in Figure 1. Asshown previously (Lehming et al., 1987), LacR-H17Q18±Oid and LacR-V17A18±O41 are both extre-mely stable, while LacR-S17S18±O4151 has a rela-tively unaltered protein-DNA complex stability,compared to the wild-type (Figure 3). This pre-cludes any signi®cant contribution of bindingspeci®city to changes in kd. Furthermore, examin-ation of the half-life of LacR-V17A18±Oid indicatesthat the change in af®nity for Oid is due entirely todifferences in kd; ka is unaffected regardless of thestrength of the binding reaction (Figures 1 to 3).

On the other hand, LacR-S17S18±Oid shows thesame stability as LacR-S17S18±O4151, as notedabove.

The effect of operator binding specificity onrepression in vivo

The in vivo repression values presented inFigure 2 are from experiments performed with cel-lular concentrations of Lac repressor and lac oper-ator that are about 90 and 50-fold, respectively,higher than the wild-type (Lehming et al., 1987,1990). Under these conditions, events such as theability of repressor dimers to form tetramers, andthe contribution of the auxiliary operators to loopformation do not increase repression (Oehler et al.,1990). Therefore, in order to determine whether thereduced association rate of tetrameric LacR var-iants with broadened speci®city described above isrelevant in the in vivo situation, we chose to re-examine the ef®ciency with which they repress in asituation as close to wild-type as possible, but onein which auxiliary operators are absent (Oehleret al., 1990, 1994). Constructs in which lacZ isunder the control of a single, fully symmetric lacoperator were inserted into the chromosome ofhost cells and the repression values determined atclose to wild-type amounts of repressor (5 � lacI�).As shown in Figure 4, (wild-type) tetrameric LacR-Y17Q18 repressed most ef®ciently in this situation.A slight weakening of repression was observed fordimeric LacR-Y17Q18. The situation is dramaticallyaltered for the variants showing a broadening ofbinding speci®city. For both LacR-H17Q18 andLacR-V17A18, the dimeric form of LacR repressessigni®cantly better than the tetrameric form(Figure 4). In the case of LacR-S17S18, repression isweak for both dimeric and tetrameric LacR.Although there is some doubt that the operatorused in this experiment is the optimal one for thisvariant (Figure 2), consideration of the extremely

Figure 4. Repression values for dimeric and tetramericLacR variants in vivo with 5 � lacI� amounts of LacR(200 monomers per cell). LacR variants contained aminoacid substitutions at positions 17 and 18 as indicatedand were encoded by plasmid pACYC184 variants(Oehler et al., 1990). Reporter constructs contained syn-thetic operator sequences (the sequence is shown, withthe centre of symmetry marked by a dash and the basesat positions 4 and 5 contacted by residues 17 and 18 ofLacR in bold type) in place of O1, and lacked the auxili-ary operators O2 and O3 (Oehler et al., 1994). Constructswere located on the chromosome (Oehler et al., 1990,1994).

Figure 5. Repression by dimeric and tetrameric GalRin vivo. (a) Repression of the galE-lacZ fusion coded bychromosomally located lgE-Z in strain BMH8117. Notethat in this case, wild-type galR� is separated from the lintegration site by a map distance of approximately 44minutes. (b) Repression of the galE-lacZ fusion encodedby chromosomally located lgE-ZR99 (encoding wild-type, dimeric GalR) in strain CSH50�1. (c) Repressionof the galE-lacZ fusion encoded by chromosomallylocated lgEZ-R330I (encoding tetrameric GalR) in strainCSH50�1. Relevant genes in the reporter constructs areindicated by open or shaded boxes. The operators OE

and OI are indicated by ®lled boxes; gal promoters p1,p2 and the galR promoter, p, as well as the syntheticpromoter ps driving galR transcription in (b) and (c) areindicated by arrows. The ends of the relevant portionsof constructs, as well as the intervening DNA betweenthe reporter construct and galR� in (a) are indicated byvertical lines. The construction of all prophage used isdetailed in Materials and Methods.


broad binding speci®city of this variant and also ofthe in vitro stability of different repressor - operatorcomplexes (Figure 3) indicate that analysis of otheroperators in this system is unlikely to lead to adifferent outcome. Furthermore, analysis of LacR-M17K18 in this system, indicates that the tetramericversion of this variant represses approximatelytwofold better than the dimer (data not shown),although we could not compare this with ka.

GalR tetramers also function poorly in vivo incomparison to GalR dimers

The data presented above suggest that GalR,which has the same binding speci®city as LacR-V17A18 (positions 1 and 2 of the recognition helixare also valine and alanine, respectively; Lehminget al., 1987) should also function less ef®ciently as atetramer than as a dimer, despite ef®cient repres-sion requiring co-operative binding of two oper-ators by GalR dimers in the gal operon (Irani et al.,1983; Fritz et al., 1983). To determine whether thisis the case, we generated constructs in which thegal promoter region, including both OE and OI

operators and the ®rst 40 codons of galE, werefused to full-length lacZ, and then introduced thisreporter into the chromosome of host cells. We ®rstexamined repression of this construct by wild-type(dimeric) galR (Figure 5(a)), and compared therepression value to that obtained when the reporterconstruct also contained a copy of (dimeric) galRunder control of a synthetic promoter immediatelyupstream, but divergently transcribed from thegalE-lacZ fusion (Figure 5(b)). The repressionvalues obtained in both cases are equivalent, indi-cating that roughly wild-type amounts of GalR areproduced by our synthetic construct. We nextexamined a reporter construct in which a tetra-meric version of galR (Alberti et al., 1991) replacedthe dimeric version. The repression value in this

case is approximately 2.5-fold lower than thatobtained with dimeric GalR (Figure 5(c)). Thus, thepoor performance of tetramers with a broad DNA-binding speci®city is not limited to LacR variants.

Discussion

The correlation between binding specificityand ka

It has long been recognised that LacR-operatorassociation in vitro is much faster (about twoorders of magnitude) than can be accounted for bya simple diffusion mechanism (Riggs et al., 1970a).A generally accepted mechanism involves aninitial, random collision between LacR and oper-ator-containing DNA, resulting in a non-speci®-cally bound intermediary, followed by facilitateddiffusion, which can consist of either repressor slid-ing along a long DNA strand, or intersegmenttransfer (Berg et al., 1981; Winter et al., 1981) bestdescribed as ``in effect moving through the genomelike Tarzan swinging from vine to vine''(Lieberman & Nordeen, 1997). However, the exactcontribution of each to LacR (dimeric and tetra-meric) search patterns remains unclear. Publishedresults are con¯icting and contain the full range of

Figure 6. A representation of differing operator searchmechanism ef®ciencies exhibited by dimeric and tetra-meric LacR variants with tight or broad DNA-bindingspeci®city in vivo. (a) Dimeric LacR-Y17Q18 with Oid;(b) tetrameric LacR-Y17Q18 with Oid; (c) dimeric LacR-V17A18 with O41; (d) tetrameric LacR-V17A18 with O41.LacR-Y17Q18 monomers are indicated by open circles,while LacR-V17A18 monomers are indicated by shadedcircles. The complete E. coli chromosome is indicated bya line. Operators are indicated by ®lled boxes. Hypothe-tical low-af®nity sites are indicated by open boxes: thenumber and distribution of these sites is purely illustra-tive and the reader should note we have no direct evi-dence for their existence, nature, number or distribution.Furthermore, the low-af®nity site(s) recognised by LacR-Y17Q18 may or may not be the same as one (some) ofthe low-af®nity site(s) recognised by LacR-V17A18. Non-speci®c association of repressor with DNA is greaterthan 90% (von Hippel et al., 1974) in all cases, while theindicated lack of occupation of O41 by LacR-V17A18 tet-ramer symbolises that this occupation occurs four timesless frequently than with the corresponding dimer in vivo(Figure 4). Note also that the drawings are not to scale.


possibilities from: intersegment transfer for tetra-mer and at least a degree of intersegment transferfor dimer (Ruusala & Crothers, 1992); throughintersegment transfer for tetramer and sliding fordimer (Fickert & MuÈ ller-Hill, 1992); to sliding fordimer and ``accelerated sliding'' or combination ofsliding and intersegment transfer for tetramer(Hsieh & Brenowitz, 1997). While not addressingthis issue directly, the results presented here intro-duce two important, new concepts. First, by vary-ing the DNA-binding speci®city it is possible todrastically alter ka for operator located on a longDNA molecule, and at the same time to alter oper-ator binding as measured by repression of a genecontrolled by a single operator in vivo. Second,dimeric LacR variants are completely unaffectedby such changes.

To date, the contribution of non-speci®c sites ona DNA molecule have been considered purely interms of their number (e.g. see Berg et al., 1981)and not of their relative af®nities, nor their differ-ential effect on a protein dependent on its state ofoligomerisation. A simplistic interpretation of ourresults (one that enables it to be considered withinthe framework of existing models) would be toassert that mutants with broadened speci®city alsoexhibit an increased af®nity for non-speci®c DNA;we have found no evidence for such an increase.Instead, we suppose that tetrameric LacR locatesoperator predominantly by intersegment transferand dimeric LacR predominantly by sliding (butsee also above). Then, we can explain our resultsby proposing the existence of weak, pseudo sites,that tetrameric LacR is capable of binding to withone dimer, while searching the surrounding DNAfor a high-af®nity binding site with the seconddimer. Note that in a long DNA molecule such asl in solution or chromosomal DNA in vivo, thissurrounding DNA is not necessarily adjacent in thesequence. If there are too many of these pseudosites, as in the case of mutants with broadenedspeci®city, then the process of locating the operatoris drastically slowed, as represented in cartoonform in Figure 6. In the classical sliding model(Berg et al., 1981), the presence of such pseudo siteswould not be expected to signi®cantly hinder adimeric protein, which would recognise them as a`non' site and continue sliding. Thus, dimeric pro-teins should not be signi®cantly hindered by abroadening of speci®city (Figure 6), as is observedexperimentally. A full understanding of the contri-bution of such pseudo sites to the ka for tetramericLacR requires their complete characterisation,including determining binding constants; anattempt to do so is underway.

Analysis of ka and kd of LacR V17A18 with O41

and Oid also indicates that here, the strength ofDNA-binding is determined entirely by kd: despitewidely differing af®nities for these two operatorvariants, the measured ka is the same. A compar-able ®nding, that an increase in operator af®nity isdue entirely to changes in kd, while non-operatorDNA-binding remains unaffected, has been

reported for the X86 mutant of LacR (Jobe &Bourgeois, 1972) and it may well be a generalproperty of DNA-binding proteins.

Cooperativity and repression by LacR and GalR

In addition to alterations of kinetic properties,tetramer formation also obviously facilitates coop-erative binding of two adjacent operators (Haber &Adhya, 1988; Oehler et al., 1990). Cooperativity ofrepressor binding to two adjacent operators is cru-cial for ef®cient repression by both LacR and GalR,yet only LacR forms stable homotetramers (MuÈ ller-Hill et al., 1971; Nakanishi et al., 1973; Majumdaret al., 1987; Weickert & Adhya, 1992; Levandoskiet al., 1996). Nevertheless, despite intense efforts, adirect interaction between GalR dimers bound totwo operators has not been observed (Majumdar &


Adhya, 1984; Majumdar et al., 1987; Brenowitzet al., 1990; Choy & Adhya, 1992), although intro-duction of the LacR-tetramerisation domain to theC terminus of GalR leads to stable loop formationin vitro (Alberti et al., 1991). Additionally, replace-ment of both gal operators with ideal lac operatorsleads to ef®cient repression of the hybrid operonby tetrameric LacR (Haber & Adhya, 1988) andallows the direct observation of DNA loops formedby tetrameric, but not dimeric LacR (Mandal et al.,1990). A recent discovery demonstrates that theloop formation required for ef®cient repression byGalR is dependent on the presence of an additionalprotein, the histone-like protein HU (Aki et al.,1996). Why the necessity for additional factors? Weshow here that tetrameric GalR represses the galoperon poorly in comparison to dimeric GalR. Thisis probably because the decrease in ability to locatespeci®c DNA sequences for tetrameric proteinswith a broad speci®city of binding leads to the tet-rameric protein getting ``lost'', and therefore thetetramerisation required for ef®cient repression canoccur, in this case, only after the initial DNA-bind-ing event. In the case of GalR, this interaction(probably between residues in the C terminus ofeach dimer) is too weak to be detected andrequires the presence of at least one additional pro-tein to stabilise it (Aki et al., 1996).

Evolutionary consequences of specificity andtetramer formation

Are other DNA-binding proteins subjected tosimilar constraints? Most other members of theLacI-GalR family are dimers (Weickert & Adhya,1992); most also display a much broader DNA-binding speci®city than LacR (Lehming et al.,1990). An examination of the two known excep-tions is informative. FruR, a pleiotropic regulatorof genes involved in carbon metabolism in Escheri-chia coli (Chin et al., 1987; Ramseier et al., 1995) is ahomotetramer in solution (Cortay et al., 1994) andis predicted to have a broad DNA-binding speci-®city (Lehming et al., 1990). Binding is indeedobserved to a relatively degenerate palindromeboth in vitro and in vivo (Ramseier et al., 1995;NeÁgre et al., 1996). However, the maximal changein repression/activation on destruction of fruR isless than tenfold (Ramseier et al., 1995). Thus, thisparticular tetramer does not function ef®cientlyin vivo, a fact that is underpinned by sound physio-logical reasons (Ramseier et al., 1995), so in this

Table 1. Bacterial strains used in this study

Strain Genoty

BMH8117 Fÿ, �(lac-proAB), gyrA, thi, supE, lBMH8117F0 F0[lacIqÿ, Zÿ, Y�, proA�B�], �(lac-pDC41-2 Fÿ, �(lac-proAB), galE, rpsL, thi, reCSH50�1 Fÿ, ara, �(lac-proAB), �(galR-lysABHB2688 Fÿ, supo, recA, (l, imm434, cIt.s., b2,BHB2690 Fÿ, supo, recA, (l, imm434, cIt.s., b2,

particular case the poor functioning of a tetramermay even be of advantage.

RafR, the regulator of a plasmid-borne operonrequired for raf®nose uptake and utilisation inE. coli is also predicted to have a broad speci®cityof DNA-binding on the basis of its amino acidsequence (Aslanidis & Schmitt, 1990; Lehming et al.,1990). Analytical ultracentrifugation detected adimer-tetramer equilibrium at a protein concen-tration of approximately 50 mg/ml (Jaenicke et al.,1990), which leads us to suggest that RafR maybehave in a similar manner to that outlined forGalR above, with the important difference that thedimer-dimer interaction is suf®ciently strong not torequire auxiliary factors and therefore is detectablein vitro.

What of other families of DNA-binding proteins?The l CI repressor, which also shows relativelybroad operator speci®city, is a dimer at cellularconcentrations and the tetramerisation necessaryfor cooperative binding of two operators and ef®-cient repression occurs on the bound DNA(reviewed by Ptashne, 1992). Similarly, the coop-erative interactions necessary for the assembly oftranscription complexes at eukaryotic promotersoccur on the DNA (e.g. see Ptashne, 1992). How-ever, at least some eukaryotic transcription factorfamilies contain members that are capable of form-ing tetramers in vitro. As an example, we cite thehelix-loop-helix proteins (reviewed, for example,by FerreÂ-D'AmareÂ & Burley, 1994). In addition, arecent analysis indicates that the glucocorticoidreceptor also uses intersegment transfer as its pre-dominant target site search mechanism (Liebermanand Nordeen, 1997). It remains to be seen whethersuch proteins are subject to similar constraints, orwhether the kinetic pathway available to LacR is aspecial case among DNA-binding proteins.

Materials and Methods

Bacterial strains, growth media, plasmids andlll phages

Bacterial strains used are listed in Table 1. Growthmedia and conditions for reporter constructs were asdescribed by Lehming et al. (1987) and Oehler et al.(1990). For repressor puri®cation, cells were grown indYT medium (Miller, 1972).

Plasmids were constructed according to standard pro-cedures (Sambrook et al., 1989). LacR variants were gen-erated by cloning annealed oligonucleotides ofappropriate sequence into NaeI/XbaI-digested pWB1000or pWB100 DNAs, to produce tetrameric or dimeric ver-

pe Reference

ÿ Oehler et al. (1990)roAB), gyrA, thi, supE, lÿ Oehler et al. (1990)

cA, lÿ Lehming et al. (1987)), rpsL, thi, lÿ Alberti et al. (1991)redÿ, Eam, Sam) Hohn (1979)redÿ, Dam, Sam) Hohn (1979)


sions, respectively, of the desired LacR variant (Lehminget al., 1987, 1988). For experiments in which plasmidsexpressing 5 � i� amounts of LacR were desired, theselacI alleles were then subcloned as an EcoRI-BglII frag-ment into plasmid pSO1010-P1 (Oehler et al., 1990).Operator DNAs were generated by cloning self-annealed, fully symmetric oligonucleotides into theunique NheI site of pEE4, for 2455 bp operator fragments(Eismann et al., 1987). For 49,000 bp operator fragments,the same oligonucleotides were ®rst cloned into theunique SpeI site of plasmid pEsts00 (Oehler et al., 1994)and then the lacO-Z-ApR portion of this plasmid wascloned into lIP1 (Sieg et al., 1989) as described (Oehleret al., 1990, 1994). All plasmid and l DNAs used inin vitro experiments were puri®ed from BMH8117.pWB300 derivatives were as described by Lehming et al.(1990). lgE-Z was constructed as follows. The gal promo-ter region, containing both promoters (p1 and p2) andboth operators (OE and OI) as well as the ®rst 40 codonsof galE fused to full-length lacZ was isolated from plas-mid piWipGal (B. von Wilcken-Bergmann, unpublishedresults) and subcloned into lIP1 (Sieg et al., 1989). lgE-ZR99 and lgE-ZR330-I consist of the same reporter con-struct and contain in addition galR derived from pgR99(wild-type, dimeric) and pgR330-I (tetrameric), respect-ively (Alberti et al., 1991), located so that they are tran-scribed divergently from galE-lacZ. The identity of allDNAs used was con®rmed by DNA sequencing.

Purification of LacR variants

Puri®cation was a modi®cation of the method pre-sented by MuÈ ller-Hill et al. (1971). A minimum of 15 g(wet weight) of BMH8117 containing plasmids constitu-tively over-expressing LacR variants were harvestedfrom overnight cultures and frozen at ÿ20�C. The iden-tity of each preparation was con®rmed by performing aplasmid miniprep on a small portion of each culture anddetermining the nucleotide sequence. Frozen cells werethawed and resuspended to 0.5 g/ml in CSB (200 mMTris-HCl (pH 7.2), 200 mM KCl, 10 mM MgCl2, 5% (v/v)glycerol, 0.3 mM DTT, 1 mM NaN3, 1 mM PMSF) andlysed by grinding (slow stirring in the presence of one totwo heaped teaspoons of glass beads; Sigma G8893) inthe presence of 0.5 mg/ml lysozyme. After visible lysis,the mixture was treated with 0.4 mg of DNaseI forapproximately 30 minutes and cell debris was thenremoved by centrifugation. A crude LacR-containingfraction was then obtained as material precipitated fromthe supernatant by ammonium sulphate in the range of20 to 33% saturation and, after dialysis, applied to aphosphocellulose column in the presence of 75 mM KPG(75 mM KPO4 (pH 7.2), 0.1 mM EDTA, 5% (w/v) glu-cose, 0.3 mM DTT, 1 mM NaN3). LacR was eluted as a595% pure peak (as judged by SDS-PAGE) with a75 mM to 400 mM KPG linear gradient (400 mM KPG is400 mM KPO4 (pH 7.2), 0.1 mM EDTA, 5% (w/v) glu-cose, 0.3 mM DTT, 1 mM NaN3), concentrated by vac-uum dialysis in Collodion bags (Sartorius) againstrepressor storage buffer (200 mM KPO4 (pH 7.2), 2 mMEDTA, 400 mM NaCl, 1 mM DTT, 1 mM NaN3) withchanges, and then further (non-vacuum) dialysis againstthe same buffer containing 50% (v/v) glycerol. Proteinconcentration was determined from the A280 assumingan extinction coef®cient of 22,500 per monomer for tetra-meric LacR (Butler et al., 1977) or 33,180 for dimeric(lacIadi) LacR (determined using the Wisconsin sequenceanalysis package of the Genetics Computer Group:GCG) and ®nal active concentration was determined

from equilibrium ®lter-binding assays of each proteinwith its cognate operator (Riggs et al., 1970b). Activitywas in the range of 25 to 40% per dimer (not shown).

Filter-binding assay

Kinetic parameters for LacR variants were determinedusing a ®lter-binding assay (Riggs et al., 1970b). Bindingreactions were performed in 10 mM Tris-HCl (pH 7.2),0.1 mM EDTA, 10 mM KCl, 3 mM magnesium acetate,50 mg/ml BSA, 0.1 mM DTT, 5% (v/v) DMSO. TargetDNAs were linearised (plasmid DNAs by digestion withEcoRI, l DNAs by incubation at 70�C for 15 minutes tomelt the cohesive ends) and labelled by ®lling in over-hanging ends with all four nucleotides. For associationexperiments, protein concentration was normally5 � 10ÿ12 M calculated per dimer and DNA concen-tration was 3 � 10ÿ12 M. Dissociation experiments wereperformed by ®rst mixing 4 � 10ÿ10 M (per dimer) LacRwith 10ÿ10 M DNA and incubating at room temperaturefor 30 minutes. The reaction was then initiated by theaddition of a 150-fold excess of unlabelled, supercoiledoperator (plasmid) DNA. Dissociation rates were deter-mined only for dimeric versions of LacR, for reasonsdescribed previously (Fickert & MuÈ ller-Hill, 1992). Filter-ing conditions were exactly as described by Fickert &MuÈ ller-Hill (1992), and ka and kd were determined asdescribed by Riggs et al. (1970a). Results presented arethe average of a minimum of three independent exper-iments.

Reporter assays

The speci®city of repression was analysed using atwo-plasmid system (Lehming et al., 1987). Operatorderivatives were cloned into the lacZ-containingpACYC184 derivative pWB300 (Lehming et al., 1987)and transformed into DC41-2. After con®rming theiridentity by DNA sequencing, transformants were thenre-transformed with pWB1000 or pWB100 derivatives(Lehming et al., 1987, 1988) as used for LacR puri®cation.b-Galactosidase speci®c activity was then determined asdescribed by Miller (1972). Results presented are theaverages of at least four independent transformants gen-erated on at least two separate occasions The ef®cacy ofrepression by LacR was examined using chromosomallylocated reporter constructs. l DNAs used in the in vitroexperiments were re-packaged using extracts preparedfrom BHB2688 and BHB2690 (Hohn, 1979; Sambrooket al., 1989) and single lysogens of BMH8117F0 generatedand identi®ed as described by Oehler et al. (1990, 1994).Lysogens were then transformed with pSO1010-P1derivatives and b-galactosidase speci®c activity deter-mined from a minimum of three independent transfor-mants of at least three independently generated singlelysogens (Miller, 1972). For both plasmid and chromoso-mally located reporter constructs, the given repressionvalues are the quotient of speci®c b-galactosidase activityin the absence of active LacR and that in the presence ofthe indicated LacR variant: a repression value of 1 indi-cates no repression. GalR activity was assesed by gener-ating single lysogens of BMH8117 (galR�) with lgE-Z,and single lysogens of CSH50�1 with lgE-ZR99 or lgE-Z330I. b-Galactosidase speci®c activity was then deter-mined from each of three independently generated singlelysogens after growth in the presence or absence of2 mM D-fucose. The repression values given are the quo-tient of these values.


Acknowledgements

We are grateful to Daniela Tils for outstanding techni-cal support, and to Peter DroÈge, Jonathan Howard andKarin Schnetz for comments on previous versions ofthis manuscript. Supported by the Deutsche Forschungs-gemeinschaft through Normalverfahren grant no. Mu575/9.

References

Adler, K., Beyreuther, K., Fanning, E., Geisler, N.,Gronenborn, B., Klemm, A., MuÈ ller-Hill, B., Pfahl,M. & Schmitz, A. (1972). How lac repressor binds toDNA. Nature, 237, 322±327.

Aki, T., Choy, H. E. & Adhya, S. (1996). Histone-likeprotein HU as a speci®c transcriptional regulator:co-factor role in repression of gal transcription byGal repressor. Genes Cells, 1, 179±188.

Alberti, S., Oehler, S., von Wilcken-Bergmann, B.,KraÈmer, H. & MuÈ ller-Hill, B. (1991). Dimer-to-tetra-mer assembly of Lac repressor involves a leucineheptad repeat. New Biologist, 3, 57±62.

Aslanidis, C. & Schmitt, R. (1990). Regulatory elements ofthe Raf®nose operon: nucleotide sequences of oper-ator and repressor genes. J. Bacteriol. 172, 2178±2180.

Bellomy, G. R., Mossing, M. C. & Record, M. T., Jr(1988). Physical properties of DNA in vivo asprobed by the length dependence of the lac operatorlooping process. Biochemistry, 27, 3900±3906.

Berg, O. G., Winter, R. B. & von Hippel, P. H. (1981).Diffusion-driven mechanisms of protein transloca-tion on nucleic acids. 1. Models and theory. Bio-chemistry, 20, 6929±6948.

Brenowitz, M., Jamison, E., Majumdar, A. & Adhya, S.(1990). Interaction of the Escherichia coli Gal repres-sor protein with its DNA operators in vitro. Biochem-istry, 29, 3374±3383.

Brenowitz, M., Mandal, N., Pickar, A., Jamison, E. &Adhya, S. (1991). DNA-binding properties of a Lacrepressor mutant incapable of forming tetramers.J. Biol. Chem. 266, 1281±1288.

Butler, A. P., Rezvin, A. & von Hippel, P. H. (1977).Molecular parameters characterising the interactionof Escherichia coli lac repressor with non-operatorDNA and inducer. Biochemistry, 16, 4757±4768.

Chakerian, A. E. & Mathews, K. S. (1991).Characterisation of mutations in oligomerisationdomain of Lac repressor protein. J. Biol. Chem. 266,22206±22214.

Chin, A. M., Feucht, B. U. & Saier, M. H., Jr (1987).Evidence for the regulation of gluconeogenesis bythe fructose phosphotransferase system in Salmo-nella typhimurium. J. Bacteriol. 169, 897±899.

Choy, H. E. & Adhya, S. (1992). Control of gal transcrip-tion through DNA looping: inhibition of the initialtranscribing complex. Proc. Natl Acad. Sci. USA, 89,11264±11268.

Cortay, J.-C., NeÁgre, D., Scarabel, M., Ramseier, T. M.,Vartak, N. B., Reizer, J., Saier, M. H., Jr & Cozzone,A. J. (1994). In vitro asymmetric binding of thepleiotropic regulatory protein, FruR, to the ace oper-ator controlling glyoxylate shunt enzyme synthesis.J. Biol. Chem. 269, 14885±14891.

Eismann, E., von Wilcken-Bergmann, B. & MuÈ ller-Hill,B. (1987). Speci®c destruction of the second lac oper-ator decreases repression of the lac operon in Escher-ichia coli ®vefold. J. Mol. Biol. 195, 949±952.

FerreÂ-D'AmareÂ, A. R. & Burley, S. K. (1994). DNA recog-nition by helix-lop-helix proteins. Nucl. Acids Mol.Biol. 7, 285±298.

Fickert, R. (1992). Untersuchung zur beschleunigtenOperatorsuche durch Lac repressor in Escherichiacoli. PhD thesis, University of Cologne.

Fickert, R. & MuÈ ller-Hill, B. (1992). How Lac repressor®nds lac operator in vitro. J. Mol. Biol. 226, 59±68.

Friedman, A. M., Fischmann, T. O. & Steitz, T. A.(1995). Crystal structure of lac repressor core tetra-mer and its implications for DNA looping. Science,268, 1721±1727.

Fritz, H.-J., BicknaÈse, B., Gleumes, B., Heibach, C.,Rosahl, S. & Ehring, R. (1983). Characterisation oftwo mutations in the Escherichia coli galE gene inac-tivating the second galactose operator and compari-tive studies of repressor binding. EMBO J. 2, 2129±2135.

Haber, R. & Adhya, S. (1988). Interaction of spatiallyseparated protein-DNA complexes for control ofgene expression: operator conversions. Proc. NatlAcad. Sci. USA, 85, 9683±9687.

Hohn, B. (1979). In vitro packaging of l and cosmidDNA. Methods Enzymol. 68, 299±309.

Hsieh, M. & Brenowitz, M. (1997). Comparison of theDNA association kinetics of the Lac repressor tetra-mer, its dimeric mutant LacIadi and the nativedimeric Gal repressor. J. Biol. Chem. 272, 22092±22096.

Irani, M. H., Orosz, L. & Adhya, S. (1983). A controlelement within a structural gene: the gal operon ofEscherichia coli. Cell, 32, 783±788.

Jacob, F. & Monod, J. (1961). Genetic regulatory mechan-isms in the synthesis of proteins. J. Mol. Biol. 3,318±356.

Jaenicke, R., Muiznieks, I., Aslanidis, C. & Schmitt, R.(1990). Ultracentrifugal analysis of the quaternarystructure of the raf repressor from Escherichia coli.FEBS Letters, 260, 233±235.

Jobe, A. & Bourgeois, S. (1972). The lac repressor-oper-ator interaction VII. A repressor with unique bind-ing properties: the X86 repressor. J. Mol. Biol. 72,139±152.

Law, S. M., Bellomy, G. R., Schlax, P. J. & Record, M. T.,Jr (1993). In vivo thermodynamic analysis of repres-sion with and without looping in lac constructs.J. Mol. Biol. 230, 161±173.

Lehming, N., Sartorius, J., NiemoÈ ller, M., Genenger, G.,von Wilcken-Bergmann, B. & MuÈ ller-Hill, B. (1987).The interaction of the recognition helix of lac repres-sor with lac operator. EMBO J. 6, 3145±3153.

Lehming, N., Sartorius, J., Oehler, S., von Wilcken-Bergmann, B. & MuÈ ller-Hill, B. (1988). Recognitionhelices of lac and l repressor are oriented in oppo-site directions and recognise similar DNAsequences. Proc. Natl Acad. Sci. USA, 85, 7947±7951.

Lehming, N., Sartorius, J., Kisters-Woike, B., vonWilcken-Bergmann, B. & MuÈ ller-Hill, B. (1990).Mutant lac repressors with new speci®cities hint atrules for protein-DNA recognition. EMBO J. 9, 615±621.

Levandoski, M. M., Tsodikov, O. V., Frank, D. E.,Melcher, S. E., Saecker, R. M. & Record, M. T., Jr(1996). Cooperative and anticooperative effects inbinding of the ®rst and second plasmid Osym opera-tores to a LacI tetramer: evidence for contributionsof non-operator DNA binding by wrapping andlooping. J. Mol. Biol. 260, 697±717.


Lewis, M., Chang, G., Horton, N. C., Kercher, M. A.,Pace, H. C., Schumacher, M. A., Brennan, R. G. &Lu, P. (1996). Crystal structure of the lactose operonrepressor and its complexes with DNA and inducer.Science, 271, 1247±1254.

Lieberman, B. A. & Nordeen, S. K. (1997). DNA inter-segment transfer, how steroid receptors search for atarget site. J. Biol. Chem. 272, 1061±1068.

Majumdar, A. & Adhya, S. (1984). Demonstration of twooperator elements in gal: in vitro repressor bindingstudies. Proc. Natl Acad. Sci. USA, 81, 6100±6104.

Majumdar, A., Rudikoff, S. & Adhya, S. (1987).Puri®cation and properties of Gal repressor: pL-galRfusion in pKC31 plasmid vector. J. Biol. Chem. 262,2326±2331.

Mandal, N., Su, W., Haber, R., Adhya, S. & Echols, H.(1990). Genes Dev. 4, 410±418.

Miller, J. H. (1972). Experiments in Molecular Genetics,Cold Spring Harbor Laboratory Press, Cold SpringHarbor, NY.

MuÈ ller, J., Oehler, S. & MuÈ ller-Hill, B. (1996). Repressionof lac promoter as a function of distance, phase andquality of an auxiliary lac operator. J. Mol. Biol. 257,21±29.

MuÈ ller-Hill, B. (1983). Sequence homology between Lacand Gal repressors and three sugar-binding peri-plasmic proteins. Nature, 302, 163±164.

MuÈ ller-Hill, B., Beyreuther, K. & Gilbert, W. (1971). Lacrepressor from Escherichia coli. Methods Enzymol. 21,483±487.

Nakanishi, S., Adhya, S., Gottesman, M. E. & Pastan, I.(1973). In vitro repression of the transcription of galoperon by puri®ed gal repressor. Proc. Natl Acad.Sci. USA, 70, 334±338.

NeÁgre, D., Bonod-Bidaud, C., Geourjon, C., DeleÂage, G.,Cozzone, A. J. & Cortay, J.-C. (1996). De®nition of aconsensus DNA-binding site for the Escherichia colipleiotropic regulatory protein, FruR. Mol. Microbiol.21, 257±266.

Oehler, S., Eismann, E. R., KraÈmer, H. & MuÈ ller-Hill, B.(1990). The three operators of the lac operoncooperate in repression. EMBO J. 9, 973±979.

Oehler, S., Amouyal, M., Kolkhof, P., von Wilcken-Bergmann, B. & MuÈ ller-Hill, B. (1994). Quality andposition of the three lac operators of E. coli de®neef®ciency of repression. EMBO J. 13, 3348±3355.

Ptashne, M. (1992). A Genetic Switch, 2nd edit., BlackwellScienti®c Publications Inc., Cambridge, MA.

Ramseier, T., M., Bledig, S., Michotey, V., Feghali, R.,Saier & M., H., Jr (1995). The global regulatory pro-tein FruR modulates the direction of carbon ¯ow inEscherichia coli. Mol. Microbiol. 16, 1157±1169.

Riggs, A. D., Bourgeois, S. & Cohn, M. (1970a). The lacrepressor-operator interaction III. Kinetic studies.J. Mol. Biol. 53, 401±417.

Riggs, A. D., Suzuki, H. & Bourgeois, S. (1970b).lac repressor-operator interaction. I. Equilibriumstudies. J. Mol. Biol. 48, 67±83.

Royer, C. A., Chakerian, A. E. & Matthews, K. S. (1990).Macromolecular binding equilibria in the lac repres-

sor system: studies using high-pressure ¯uorescencespectroscopy. Biochemistry, 29, 4959±4966.

Ruusala, T. & Crothers, D. M. (1992). Sliding and inter-molecular transfer of the lac repressor: kinetic petur-bation of a reaction intermediate by a distant DNAsequence. Proc. Natl Acad. Sci. USA, 89, 4903±4907.

Sadler, J. R., Sasmor, H. & Betz, J. L. (1983). A perfectlysymmetric lac operator binds the lac repressor verytightly. Proc. Natl Acad. Sci. USA, 80, 6785±6789.

Sambrook, J., Fritsch, E. F. & Maniatis, T. (1989). Molecu-lar Cloning; A Laboratory Manual, 2nd edit., ColdSpring Harbor Laboratory Press. Cold Spring Har-bor, NY.

Schlax, P. J., Capp, M. W. & Record, M. T., Jr (1995).Inhibition of transcription initiation by lac repressor.J. Mol. Biol. 245, 331±350.

Schumacher, M. A., Choi, K. Y., Zalkin, H. & Brennan,R. G. (1994). Crystal structure of LacI member,PurR, bound to DNA: minor groove binding by ahelices. Science, 266, 763±770.

Schumacher, M. A., Choi, K. Y., Lu, F., Zalkin, H. &Brennan, R. G. (1995). Mechanism of corepressor-mediated speci®c DNA binding by the purinerepressor. Cell, 83, 147±155.

Sieg, K., Kun, J., Pohl, I., Scherf, A. & MuÈ ler-Hill, B. (1989).A versatile phage lambda expression vector systemfor cloning in Escherichia coli. Gene, 75, 261±270.

Simons, A., Tils, D., von Wilcken-Bergmann, B. &MuÈ ller-Hill, B. (1984). Possible ideal lac operator:Escherichia coli lac operator-like sequences fromeukaryotic genomes lack the central G-C pair. Proc.Natl Acad. Sci. USA, 81, 1624±1628.

Slijper, M., Bonvin, A. M. J. J., Boelens, R. & Kaptein, R.(1996). Re®ned structure of lac repressor headpiece(1-56) determined by relaxation matrix calculationsfrom 2D and 3D NOE data: change of tertiary struc-ture upon binding to the lac operator. J. Mol. Biol.259, 761±773.

Slijper, M., Boelens, R., Davis, A. L., Konings, R. N. H.,van der Marel, G. A., van Boom, J. H. & Kaptein,R. (1997). Backbone and side chain dynamics of lacrepressor headpiece (1-56) and its complex withDNA. Biochemistry, 36, 249±254.

Weickert, M. J. & Adhya, S. (1992). A family of bacterialregulators homologous to Gal and Lac repressors.J. Biol. Chem. 267, 15869±15874.

von Hippel, P. H., Revzin, A., Gross, C. A. & Wang,A. C. (1974). Non-speci®c binding of genome regu-lating proteins as a biological control mechanism. 1.The lac operon: equilibrium aspects. Proc. Natl Acad.Sci. USA, 71, 4808±4812.

von Wilcken-Bergmann, B. & MuÈ ller-Hill, B. (1982).Sequence of galR gene indicates a common evol-utionary origin of lac and gal repressor in Escherichiacoli. Proc. Natl Acad. Sci. USA, 79, 2427±2431.

Winter, R. B., Berg, O. G. & von Hippel, P. H. (1981).Diffusion-driven mechanisms of protein transloca-tion on nucleic acids. 3. The Escherichia coli lacrepressor-operator interaction: kinetic measurementsand conclusions. Biochemistry, 20, 6961±6977.

Edited by M. Yaniv

(Received 22 September 1997; received in revised form 9 February 1998; accepted 17 February 1998)

operator search by mutant lac repressors

Documents