Electrostatic activation ofEscherichia coli methionine repressor

Kathryn Phillipst and Simon EV Phillips*Department of Biochemistry and Molecular Biology, University of Leeds, Leeds, LS2 9JT, UK

Background: The three-dimensional structure of theEschericbia coli methionine repressor (met repres-sor) is relatively unperturbed by the binding of itscorepressor, Sadenosylmethionine (SAM), and of oper-ator DNA. The positively charged corepressor binds tosites on the repressor remote from the DNA-binding site,and despite the lack of induced structural change is ableto raise the affinity for operator DNA by a factor of upto 1000. Neutral corepressor analogues also bind to therepressor, but do not increase operator affinity.Theseobservations suggest that the corepressor effect may beelectrostatic.Results: Using the program DELPHI, we have calcu-lated electrostatic potentials for the repressor and itscomplexes, and have obtained results consistent with

an electrostatic model for repressor activation. The pos-itive potential originating from the corepressor is prop-agated through the repressor-operator complex, andis significant at DNA phosphate groups buried in theprotein-DNA interface. The rank order of calculatedelectrostatic interaction energies for complexes withSAM, and two closely-related analogues, is in agreementwith experimental measurements of the correspondingrepressor-operator affinities.Conclusion: Long-range (> 10A) electrostatic interac-tions between bound corepressor and operator phos-phate groups in the repressor-operator complex maybe sufficient to explain repressor activation. Met repres-sor could, therefore, be an electrostatically triggered ge-netic switch.

Structure 15 April 1994, 2:309-316Key words: electrostatic potential, protein-DNA interactions, repressor, S-adenosylmethionine

IntroductionExpression of genes encoding the enzymes for methio-nine biosynthesis in Escherichia coli is regulated by adelicate balance of activation and repression (reviewedin [1 ]). Much of the repressive control is exercised bythe product of the metJgene, the methionine repressor(met repressor), a dimer of identical 104 amino acidsubunits. The free repressor (aporepressor) is acti-vated by the non-cooperative binding of two moleculesof Sadenosylmethionine (SAM) to form the holore-pressor. The free repressor does not bind methion-ine, and SAM is almost certainly the true corepressorin vivo, although there is no direct evidence for this.SAM carries an activated methyl group, and is the donorin many biological processes involving the transfer ofone carbon units. As it cannot be transported acrossthe bacterial cell wall, the only source of SAM in thecell is the activated methyl cycle, which generates SAMfrom methionine. Measurements of repressor-operatorinteraction in vitro, using filter binding assays withaporepressor and synthetic operator fragments, showaffinities in the micromolar range. Operator affinity is,however, increased up to 1000-fold in the presence ofSAM [2].

Met repressor regulates at least nine genes in themethionine biosynthetic pathway by recognizing, andbinding to, at least eight different operator sites inthe E coli genome. A number of these sites havebeen sequenced, and consist of between two and fivetandem repeats of homologous eight base-pair (bp)

motifs known as 'met boxes' [3]. Alignment of theknown operator sequences yields the met box con-sensus sequence, AGACGTCT. The shortest naturallyoccurring operator, metC has two met boxes (16 bp),which show high overall homology to the consensussequence (81 % identity). Binding assays in vitro showthat two tandem met boxes are the minimum require-ment for a viable operator site ([2]; Y-Y He, T McNally,I Manfield, I Parsons, SEV Phillips and PG Stockley,unpublished data), and that operator binding is highlycooperative with respect to protein concentration.

Three-dimensional structure of met repressorThe three-dimensional structure of met repressor hasbeen determined by X-ray crystallography [4]. Twoidentical subunits, each consisting of one 3-strand andthree oc-helices (A, B and C), come together to forman intertwined symmetrical dimer (Fig. 1). Three c-he-lices, A (residues 30-45), B (52-66) and C (86-94)form the lower part of the molecule. The top surfaceis formed by a two-stranded antiparallel -sheet (3-rib-bon, residues 22-28), which binds to the major grooveof the DNA in the repressor-operator complex. Thesubunit interface is formed by the two symmetry-re-lated 3-strands, which loop over one another to formthe P3-ribbon, and the B-helices, which pack togetherto form the central hydrophobic core of the dimer.Two molecules of SAM bind to the repressor in siteslying alongside the B-helices, each with its adeninering inserted into a pocket otherwise occupied by theside-chain of Phe65 in the aporepressor. The positively

'Corresponding author. tPresent address: Melvin Calvin Laboratory, University of California at Berkeley, Berkeley, CA 94720, USA.

( Current Biology Ltd ISSN 0969-2126 309

310 Structure 1994, Vol 2 No 4

Fig. 1. Overall fold of the holorepres-sor backbone, with the molecular dyadvertical. One subunit is shown in redwith regions of secondary structure la-belled, and the other in pink. The SAMmolecules are shown as ball-and-stickmodels (green) with the sulphur atomsyellow, and bind to two symmetry-related sites at the bottom. Drawn withprogram MOLSCRIPT [16].

charged tertiary sulphur atom of each SAM lies at thecarboxyl terminus of a B-helix, 3.4 A from the carbonyloxygen of Ala64, placing it at the negative end of thelocal helix dipole [5].

Crystal structures have now been solved for fivedifferent crystal forms, two aporepressor ([4]; SDStrathdee and SEVP, unpublished data), two holore-pressor ([4]; KP and SEVP, unpublished data) and onerepressor-operator complex [6]. Least squares super-position of all these structures shows that the coreof the repressor is relatively unperturbed by bindingDNA or SAM, or by packing in different crystal lattices.However, two surface loops, residues 12-20 of eachsubunit (Fig. 1), are flexible, and their conformation isaffected by the local crystal environment, but not by thepresence or absence of SAM. This is in contrast to thestructure of the trp repressor, where corepressor bind-ing can induce large conformation changes, reorientingits two DNA 'reading heads', thereby directly affectingoperator affinity [7].

Structure of the repressor-operator complexIn the crystal structure of the repressor-operator com-plex [6], two repressors are bound to a 19 bp oligonu-cleotide fragment corresponding to two tandem con-sensus met boxes flanked by two extra bases at the5' end and one at the 3' end (Fig. 2). Each repres-sor is bound at the centre of a met box, with its [3-ribbon inserted into the major groove (Fig. 3), andfour side-chains, two from each strand of the ribbon,forming direct hydrogen bonds to base pairs. Theseinteractions make the major contribution to sequencespecificity. The conformation of the operator is closeto that of ideal B-form DNA, except for 25° bends inthe helical axis at the centre of each met box, andoverwinding of the T8pA9 step in the centre of the

operator. The overwinding is linked to a shift of thephosphate 3' to A9 by 2 A from its expected position,allowing it to make hydrogen bonds to main chainamide hydrogens at the amino terminus of a repressorB-helix. Since this distortion is favoured or facilitated bythe presence of the TpA step, and replacement of thisstep by other sequences leads to reduction in operatoraffinity ([8]; Y-Y He, T McNally, I Manfield, I Parsons,SEV Phillips and PG Stockley, unpublished data), thishas been interpreted as a sequence-specific interaction[6]. The bends in the operator result in narrowingof the major groove around the 3-ribbon, and allowthe DNA to wrap more closely around the repressorsurfaces. The cooperativity of binding arises from anextensive hydrophobic contact between A-helices ofadjacent bound repressors.

The origin of the corepressor effectThe crystal structure of the repressor-operator com-plex yields little information about the mechanism ofactivation by the corepressor. In allosteric proteins,binding of small effector molecules results either instructural change, or displacement of equilibrium be-tween different structures, which affects activity, butthe structure of met repressor shows little change inresponse to SAM binding. Activation could also beachieved if the corepressor interacted directly with theDNA by binding in the protein-DNA interface, as in trprepressor, but this is clearly not the case for met. In fact,the closest approach of non-hydrogen atoms betweenSAM and DNA in the repressor-operator complex is11.6 A for N6 of the SAM adenine ring and a non-es-terified phosphate oxygen atom of the DNA. In theabsence of significant conformation change inducedby SAM binding, it is necessary to consider alternativemechanisms that could operate at long range.

-1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

T AGACGTCT AGACGTA TCTGCAGA TCTGCA

115 16 17

'CTTA6GA TT

Fig. 2. Sequence and numbering schemefor the operator oligonucleotide in thecrystal structure of the complex. Thetwo strands are related by a cen-tral crystallographic two-fold axis. Theboxed regions correspond to consensusmet boxes.

- -

_

-

Activation of methionine repressor Phillips and Phillips 311

Fig. 3. Stereo view of repressor-operatorcomplex. Two repressors (in red) arebound to the DNA at upper left andlower right respectively, and related bya crystallographic two-fold axis pass-ing through the centre of the complexperpendicular to the plane of the pa-per. The SAM molecules (green) lie onthe outer surface of the complex, dis-tant from the DNA. The DNA is shownin a surface representation, colouredsuch that all atoms lying within the+lkTe -1 potential contour due toSAM partial charges are blue, while allatoms with a potential < +kTe - 1 arewhite.

Filter binding and gel mobility assays ([2]; Y-Y He,T McNally, I Manfield, I Parsons, SEV Phillips and PGStockley, unpublished data) have shown that while SAMbinding increases affinity of the repressor for the op-erator in vitro by a factor of up to 1000, bindingof S-adenosylhomocysteine (SAH) leaves it unchangedfrom that of the aporepressor. SAH is the derivativeof SAM produced when it is used in a methylationreaction. For instance, SAM is the methyl donor forprokaryotic DNA-methyltransferases, and is then regen-erated from SAH by the activated methyl cycle. It istherefore necessary for the repressor to distinguish be-tween SAM and SAH. The chemical structures of thesetwo molecules are shown in Fig. 4, the only differencesbeing the charge on the sulphur atom and the Smethylgroup. Since the methyl group is exposed to the solventin the holorepressor structure, and makes no contacts

to the protein, the implication is that repressor acti-vation is electrostatic, and depends on the charge onthe sulphur. In the repressor-operator complex, thissulphur lies at the carboxyl terminus of the B-helix,while the amino terminus of the same helix interactswith a phosphate group of the operator.

We considered three possible models for repressor ac-tivation. In the first model, corepressor binding wouldstabilize the repressor conformation found in the op-erator complex, thus favouring DNA binding. Thebound conformation differs significantly from that offree aporepressor and holorepressor only in the loopsformed by residues 12-20, which swing away from theprotein surface to wrap around DNA phosphates. Theloops are, however, solvent exposed and quite dis-tant from the corepressor site. In the second model,

Fig. 4. Chemical structures o -aoeno-sylmethionine (SAM), adenosylornithine(AO) and S-adenosylhomocysteine(SAH). The partial charges used in thecalculation of the corepressor and core-pressor analogue interactions with theoperator DNA are shown in full forSAM. For AO and SAH, atom names

re -hnwn nlv wharp thpv differ frnm

SAM. The charge sets were derived fromthe AMBER charges for methionine and

dArnncino

SAM

0.33N-

-0.706

0. 1371

-0.70

0.339 0.339E a

-0. 793

-0.599 0.813- 10.48 N -0.760

~0.333 0 ~CH .571330 o333 -0.457 1 C

0.695N 0.009 0.042 -0.413 0

7C82 0 -0.717

CT '~.227 cm ~Cu 0.522

6 0.037 0 2ICE CE 0. 082

0. 70.5140 0-0.512

3 E0.313 0.313

SAE

0.97

Guru vo lo

312 Structure 1994, Vol 2 No 4

a long range electrostatic interaction between the pos-itively charged corepressor and the negatively chargedphosphate backbone of the operator would stabilizethe complex. In the third model, a long-range electro-static effect would increase the cooperativity of bind-ing of adjacent repressors in the protein-DNA com-plex. To test these hypotheses, the finite differencePoisson-Boltzmann equation [9] was used to calcu-late the electrostatic potential due to SAM and its ana-logues in the environment of the holorepressor andrepressor-operator complex.

Results and discussionExperimental observations of binding affinityThe affinity of met repressor for synthetic operators,consisting of two consensus met boxes embedded ina longer DNA fragment, determined by filter binding as-says, has been reported ([2]; Y-Y He, T McNally, I Man-field, I Parsons, SEV Phillips and PG Stockley, unpub-lished data). The repressor-operator complex formedunder the conditions of the experiment was shownto consist of two repressors bound to the met boxes,in an arrangement corresponding to that observed inthe crystal structure of the complex [6]. The experi-ments were carried out at pH 7.6 in a buffer containing0.1 M KCI, and yielded an apparent dissociation con-stant for the binding reaction, Kd, of 4( ±3) x 10-9Min the presence of saturating levels of SAM. In the ab-sence of SAM, saturation binding to the DNA was neverachieved, and Kd is approximately 10-6M.

Equilibrium dialysis showed that SAH is a competitiveinhibitor of SAM binding to aporepressor. It binds witha similar affinity, presumably in the same orientation.The results of filter binding experiments performed inthe presence of saturating levels of SAH, however, areindistinguishable from those using aporepressor in theabsence of SAM (Kd - 10-6M), with saturation bind-ing never achieved (Y-Y He, T McNally, I Manfield, I Par-sons, SEV Phillips and PG Stockley, unpublished data).Since SAH is electrically neutral, this implies that the in-crease in operator affinity in the presence of SAM mightbe due to the additional positive charge on the sulphur.The methylase inhibitor sinefungin (adenosylomithine,AO), shown in Fig. 4, is also an analogue of SAM, andcarries a net positive charge. When it was used as core-pressor in a gel shift assay, operator affinity was higherthan for SAH, and saturation binding was achieved, al-beit at higher protein concentrations than necessaryin the presence of SAM (PG Stockley and T McNally,personal communication). These results are consistentwith the suggestion that corepressor charge is impor-tant for activation, but it should be noted that the sul-phur atom is remote from the DNA in the complex,its nearest approach to any DNA atom being 19.4Afor a non-esterified phosphate oxygen. Very recently,crystals have been obtained for complexes of repressorwith AO and SAH (JP Porter and SEVP, unpublished

data) suitable for structure determination to confirmthe mode of binding.The differences in free energy of formation of com-plexes of two repressors on two met box opera-tors, in the presence of different corepressor ana-logues, AAG, can be directly estimated from theratio of their dissociation constants. For SAM andSAH, for instance, AAGholo(SAH) - holo(SAM) is approx-imately 4 kcal mol- 1 or 7 kT, with a similar value forAAGapo -holo(SAM)' Proposed models for the core-pressor activation mechanism should predict energydifferences of this order.

Electrostatic potential of repressor in the presence andabsence of SAMThe calculated electrostatic potential of the fully sol-vated aporepressor (holorepressor structure with SAMremoved - see Materials and methods) is markedlydipolar. A region of positive potential extends out intothe solvent from the -ribbon face (top in Fig. 5),while the opposite face of the molecule is surroundedby negative potential. The net charge on the repres-sor is - 6e at neutral pH, somewhat unexpected for aDNA-binding protein, but the distribution of chargedside chains is such that positively charged groups arefound mostly on the 3-ribbon (DNA-binding) face,and negatively charged ones on the opposite C-helixface. The positive potential arises mainly from Lys23and Lys25 on the -ribbon, and Arg40 on the up-per face of the A-helix. These three side chains con-tribute to the protein-DNA interface in the complex,Lys23 making two sequence-specific contacts to oper-ator bases, and Lys25 and Arg40 interacting with thephosphate backbone. This constellation of positivelycharged side chains forms two ridges over the -rib-bon such that, when DNA is bound, the phosphatebackbone is buried within corresponding ridges of thepositive potential. The conformations of these long andflexible side chains vary in the different crystal forms,but they are mostly fully extended. Such variations af-fect the exact shape of the positive potential.Two other repressor proteins, Arc and Mnt from bacte-riophage P22, show sequence homology to met repres-sor and bind DNA via 3-strands [10]. The NMR struc-ture of Arc repressor [11] shows it has a similar DNA-binding motif, consisting of the 3-ribbon and helices Aand B. Calculation of the electrostatic potential of Arcrepressor (data not shown) showed it to be similarlydipolar, with positive potential around the P-ribbon,and negative potential around the carboxy-terminal endof the B-helix and the carboxyl terminus, despite themolecule carrying a net positive charge of -8e. Thesimilarity of the met and Arc repressor potentials isremarkable in view of the large difference in their netcharges.Inclusion of the SAM molecules in the calculation of themet repressor potential reduces the net overall chargeto - 4e, but has little effect on the positive potential onthe 3-ribbon face (Fig. 5). Solvent shielding prevents

Activation of methionine repressor Phillips and Phillips 313

Fig. 5. Contours of electrostatic potential of the holorepressor dis-played as a slice through the molecule in the same orientation asFig. 1. The + lkT e - 1 contour is blue, and the - kT e - contourred. The repressor is green. A region of positive potential extendsinto the solvent from the a-ribbon DNA-binding face at the topof the molecule. The potential map for aporepressor (not shown)is superficially very similar.

the positive potential extending further into solution.The two extra net positive charges of bound SAM do,however, attenuate the negative potential between theC-helices, such that the -1 kTe- 1 potential contourfollows their surfaces more closely than for aporepres-sor.

Although the structure of the repressor core is verysimilar in the different crystal forms, the flexible loops(residues 12-20) immediately preceding the p3-ribbonshow some differences. In the absence of DNA theylie on the protein surface, but change conformationdramatically when DNA is bound (as shown in Fig.3), looping away from the protein surface to wraparound DNA phosphate groups [6]. If this loop confor-mation was favoured by corepressor binding it wouldincrease operator affinity. The loops are not in contactwith bound corepressor, however, so the effect wouldhave to be electrostatic. This proposal was tested bycalculating the electrostatic interaction energy of par-tial charges on the atoms of the loops in the holore-pressor structure with the potential originating fromthe two bound SAM molecules. The calculation wasrepeated with a repressor model extracted from therepressor-operator complex, where the loops have theother conformation. Both interaction energies are ofthe order of 0.5 kT, indicating that the SAM potentialdoes not selectively favour one loop conformation.

Electrostatic interactions of corepressor in therepressor-operator complexTo investigate the direct electrostatic interaction be-tween bound corepressor molecules and the DNA, thepotential generated by the partial atomic charges of thefour SAM molecules in the repressor-operator com-

Fig. 6. Map of the potential generated by SAM molecules whenburied in the low dielectric cavity of the repressor-operator com-plex found in the crystal structure in the same orientation asFig. 3. The +1kTe - 1 (blue) potential crosses the interface be-tween the two repressor dimers, and extends through the com-plex to lie over an extended region of the phosphate backbone.The -kTe -1 contour is shown in red, DNA in white.

plex was calculated (Fig. 6). It is positive everywhereexcept for localized negative regions around the elec-tronegative SAM atoms. The + 1 kT e- 1 contour ex-tends throughout the complex, covering part of thephosphate backbone of the operator. The oxygens ofeight phosphate groups lie within this contour, and theinteraction energy of all partial atomic charges on theDNA with this potential is -6.2kT (Table 1), domi-nated by the contribution of negatively charged phos-phate groups. This free energy difference correspondsto a DNA binding affinity constant for holorepressorapproximately 500 times greater than that for apore-pressor, which is in the range observed experimentally.

The phosphate group lying between bases A9 and G10is displaced by 2 A from the position it would be ex-pected to take in ideal B-form DNA. This has beeninterpreted as a sequence-specific effect, and if this de-viation from ideality shifted the phosphate group to

Table 1. Calculated electrostatic interaction energies between partialcharges of corepressor analogues and operator DNA.

Interaction energy (kT)0.1 M Salt Zero salta Coulombicb

SAM-MetJ-DNA -6.2 -11.4 -10.3SAM-MetJ-DNAc -4.5 - -

AO-MetJ-DNA - 4.8 - -SAH-MetJ-DNA - 3.4 - 4.4 --1.3

alonic strength set to 0. blnterior and exterior dielectrics set to 78.5.Idealized 10.6-fold B-DNA.

314 Structure 1994, Vol 2 No 4

a more positive region of the SAM-induced potential,SAM binding would enhance sequence-specificity. Fig.7 shows the position of this phosphate relative to thepotential contours. A model of ideal B-DNA was super-imposed on the operator, and the electrostatic inter-action of the corresponding idealized phosphate posi-tion compared with that observed in the complex. Thevector between the two positions approximately fol-lows an equipotential, so that the corepressor potentialdoes not appear to discriminate between the differentphosphate positions and cannot, therefore, contributeto sequence specificity.

Fig. 7. Comparison of the +1kTe -1 potential contours due toSAM (pink), SAH (blue) and AO (red) at phosphate P10 (3' to A9) ofthe operator (white). The SAM contour extends furthest over theoperator backbone, resulting in the most favourable interaction.

Since the positive potential induced by the four SAMmolecules is continuous between the two repressors, itcould affect the cooperativity. The potential was there-fore calculated for the two SAM molecules in one of therepressors, and its interactions with the partial chargesof the other repressor were calculated. The resultingtotal interaction energy is +0.4kT, implying a slightreduction in cooperativity. Taken together, the aboveresults suggest that the increase in affinity of repressorfor operator could arise solely from long-range elec-trostatic interactions between SAM and the DNA, prin-cipally the phosphate groups.

The large buried protein-DNA interface excludes bulksolvent, although a few ordered water molecules arevisible in electron density maps, and additionally ex-cludes cations that must be present in solvated DNAto neutralize the negative charge of the phosphates.Measurements of the salt dependence of complex for-mation show the apparent affinity constant to be ap-proximately two orders of magnitude lower in 400 mMthan 50mM KC1, and that five or six cations are dis-placed from the DNA as a consequence of binding thetwo repressors (Y-Y He, T McNally, I Manfield, I Par-sons, SEV Phillips and PG Stockley, unpublished data).

The interface is formed by the insertion of the -rib-bon into the major groove, deformation of the 12-20loops to wrap around and interact with the phosphatebackbone, and bending of the DNA to close the ma-jor groove around the -ribbon. The effect of DNAbending on the electrostatic interactions was assessedusing the model complex with an ideal (straight) B-form DNA With ideal DNA, the SAM-DNA electrostaticinteraction energy is reduced by 1.7kT to -4.5kT(Table 1). The overall DNA conformation, therefore,not only serves to optimise the large number of short-range interactions at the interface, but also enhancesthe SAM-DNA electrostatic interaction.The effects of SAM analogues were modelled by mod-ifying the charge set used for SAM (Fig. 4). For AOthe net positive charge was displaced from the sul-phur (carbon atom in AO) along the S-Me bond tothe methyl group (NH3+ in AO). For SAH, the sul-phur charge and methyl group were simply removed.Interestingly, the electrostatic potential produced bythe SAH partial charges is again positive throughoutthe complex, even though SAH itself is neutral, andthe interaction energy of this potential with the oper-ator is -3.4kT (Table 1). This was unexpected, butcloser inspection of the arrangement of partial chargesshows that the positively charged amide group of SAH(and SAM) faces into the protein surface, and this isresponsible for the positive potential. The negativelycharged carboxylate group, however, is oriented outinto the solvent, and the negative potential it generatesis attenuated by solvent shielding. When SAM is bound,its sulphur atom also faces in towards the protein, andis partially protected from the solvent by the presenceof the Smethyl group. This reduces solvent screeningof its charge, and enhances its long-range interactionwith the DNA For bound AO the positive charge lieson the nitrogen (Fig. 4), and is shifted away from theDNA along the C-N bond, relative to the position ofthe charge in SAM. The potential generated by AOagain has a similar shape, with net interaction energyof - 4.8 kT with the DNA (Table 1). This is lower thanfor SAM due to the increased distance of the chargefrom the DNA, and increased solvent screening due toits more exposed position.The SAM potential, therefore, extends furthest into theDNA, with that from AO less far, and SAH the least (Fig.7). This reproduces the interaction hierarchy foundfrom experimental measurements of affinity. An ap-parent discrepancy with experiment is that SAH has asmall favourable electrostatic interaction with the DNA,- 3.4 kT, while it shows no corepressor effect in bind-ing assays. The difference between calculated electro-static interaction energies for DNA-SAM and DNA-SAHis consequently reduced to 2.8 kT. The potential gradi-ent at the phosphate backbone is, however, very steep(Fig. 7) and small changes in atomic positions couldhave large effects on the interaction energy. All the cal-culations were carried out using the observed structure

Activation of methionine repressor Phillips and Phillips 315

of a repressor-operator complex containing SAM, as-suming that there would be no change in structure ifSAM were replaced by its analogues. It is possible thatthe reduced electrostatic attraction for the phosphatesin the SAH and AO complexes could result in the DNAbackbone being less tightly bound, and the phosphategroups lying a little further from the protein surface.Quite small shifts of the charged phosphates could re-sult in significant reductions in interaction energy.

When the calculation is repeated at zero ionic strengthfor the SAM and SAH complexes, the interaction en-ergy with SAM is 4.8kT greater (Table 1), consistentwith the observed salt dependence, while that for SAHis less affected. The steep gradient of potential acrossthe phosphate backbone is maintained. This impliesthat exclusion of salt from the protein-DNA interfaceis important. A simple calculation using Coulomb's lawwith constant high dielectric, shows there is a smalleffect due to removal of the dielectric discontinuity,amounting to 1.1 kT for SAM (Table 1). The steep po-tential gradient across the phosphate backbone is lost,and replaced by a shallow one depending only on thereciprocals of the distances between the charges. Thedielectric discontinuity stemming from the shape of thecomplex surface therefore makes a small, but signifi-cant, contribution to the overall electrostatic corepres-sor effect.

While the quantitative results of these electrostatic cal-culations must be treated with some caution, they cor-rectly reproduce the rank order of activation by thethree different corepressor molecules, and give reason-able values for interaction energies. The overall formof the potential surfaces is produced by the overallshape of the repressor-operator complex, and seemsdesigned to cover the phosphate backbone. The elec-trostatic model now requires further experiments totest its validity.

Biological implicationsThe function of the methionine (met) repressorof Eschericia coli is to regulate transcription ofgenes in the methionine biosynthetic pathwayand the activated methyl cycle, in response tothe intracellular level of its corepressor S-adeno-sylmethionine (SAM). Met repressor binds in ar-rays to tandem repeats of eight base-pair metbox sequences in met operators, and a minimumof two bound repressors are necessary for ef-fective repression. The repressor must discrimi-nate between SAM and the very similar moleculeS adenosylhomocysteine (SAH), the product ofmethylation reactions where SAM is the cofactor.SAM and SAH bind to the repressor with approxi-

mately equal affinity, but only SAM enhances DNAbinding affinity.In the structure of a repressor-operator complex,the SAM-binding site lies on the outside surfaceof the repressor molecules, and more than 10Afrom the DNA. Furthermore, SAM binding inducesno significant structural change in the repressorto explain the enhancement of operator affinity.SAM, however, carries a formal charge of + le,whereas SAH is neutral, and our calculations ofelectrostatic potential in the repressor-operatorcomplex now suggest that the corepressor ef-fect of SAM is the result of long-range electro-static interactions between its positive charge andthe negatively charged DNA phosphates. This isonly possible where there is a tightly packedprotein-DNA interface that excludes solvent andions. Other regulatory processes, such as phos-phorylation, also involve charge changes, and themet repressor results show that these could, inprinciple, be effected at long range as an alterna-tive to induced structural changes or direct inter-atomic contacts.

Materials and methodsElectrostatic calculationsAll of the calculations described were conducted with the finitedifference Poisson-Boltzrnann method of Nicholls and Honig[12] using the program DELPHI, and the results displayed withINSIGHT II (Biosym Technologies, San Diego). The dielectricconstants for solute and solvent were set at 2 and 78.5 respec-tively. The ionic strength of the solvent continuum was set to0.1 M, corresponding to the conditions of the filter binding ex-periments ([2]; Y-Y He, T McNally, I Manfield, I Parsons, SEVPhillips and PG Stockley, unpublished data). The coordinates ofthe holorepressor crystal structure solved in space group P3221[4] (Brookhaven Protein Data Bank entry lcmc) were used tocalculate the electrostatic potential of the holorepressor. Thecoordinates of the aporepressor crystal structure [4] were notused to calculate the aporepressor potential, since small differ-ences in the dielectric shape between different experimentallydetermined structures might have resulted in differences in thepotentials for aporepressors and holorepressors, not due to theeffect of the corepressor alone. The aporepressor potential was,therefore, calculated using the holorepressor model with theSAM molecules omitted.

Repressor potentialThe net charge of the holorepressor is dependent on the proto-nation state of the six histidine side chains (in the dimer). Theexpected pK of each histidine was assessed by inspection of itshydrogen bonding interactions with other groups in the struc-ture. It was not possible to assign a charge by inspection for thetwo solvent-exposed histidine side chains in the repressor, andthey were therefore assumed to be neutral at pH 7.6. The dielec-tric boundary between the molecule and the bulk solvent was as-signed using a solvent accessible surface calculated with a rollingsphere probe of radius 1.4 A [ 13]. Points inside this surface wereassigned a low dielectric constant (2) and those outside a solvent

316 Structure 1994, Vol 2 No 4

dielectric (78.5). A focusing protocol, suggested by Klapper etal. [14], was used to increase the accuracy of the calculation,with the final boundary 10A beyond the largest dimensions ofthe repressor molecule. Electrostatic potentials were calculatedusing a charge set including partial charges on all polar groups[15]. The partial atomic charges for the corepressor, SAM andits analogues SAH and AO (Fig. 4) were based on those reportedfor methionine and adenine [15].

The electrostatic potential of the structurally related Arc repres-sor from bacteriophage P22 [11] was calculated using a similarprotocol, using coordinates kindly given to us by Prof. R Kaptein.

Repressor-operator complex potentialThe coordinates of the repressor-operator complex [6](Brookhaven Protein Data Bank entry lcma) were used to as-sign the dielectric boundary, and the potential due to the fourcharged SAM molecules was calculated. A focusing protocol wasused again, with the final boundary set to 1 A beyond the largestdimension of the complex. The electrostatic interaction energyof the SAM molecules with all the partial charges on the DNAwas then calculated. The SAM charge set (Fig. 4) was modifiedto model the two corepressor analogues, SAH and AO. The in-teraction energies between the corepressor analogues and theDNA were calculated, with the boundary conditions and focusingprotocol used for the SAM, and are shown in Table 1. For theSAM and SAH complexes the calculation was repeated for zeroionic strength, and also for the simple Coulombic model usingzero ionic strength and setting the dielectric constant to 78.5throughout the complex.

Acknowledgments: We thank Julia Goodfellow for advice on electro-static potential calculations, Peter Stockley and Teresa McNally formaking available unpublished results from operator binding exper-iments, and other colleagues for stimulating discussions. We thankRobert Kaptein for sending us the coordinates of Arc repressor. Wethank the referees for suggesting the calculation using Coulomb's lawalone. KP was supported by a grant from the University of Leeds, andby grants from Departments of Health and Environmental Research,Energy and NIH (grant #CA45593) (USA) [to Sung-Hou Kim]. SEVPis an SERC Senior Fellow and International Research Scholar of theHoward Hughes Medical Institute.

References1. Old, I.G., Phillips, S.E.V., Stockldey, P.G. & Saint-Girons, I.

(1991). Regulation of methionine biosynthesis in the enter-obacteriaceae. Frog Biophys Mot BioL 56, 145-184.

2. Phillips, S.E.V., et at, & Stockley, P.G. (1989). Cooperative tan-dem binding of met repressor of Escherichia coli. Nature 341,711-715.

3. Belfaiza, J., et al, & Saint-Girons, I. (1986). Evolution in biosyn-thetic pathways: two enzymes catalyzing consecutive steps inmethionine biosynthesis originate from a common ancestorand possess a similar regulatory region. Proc. NatL Acad Sc.USA 83, 867-871.

4. Rafferty, J.B., Somers, W.S., Saint-Girons, I. & Phillips, S.E.V.(1989). Three dimensional crystal structure of Escherichia colimet repressor with and without corepressor. Nature 341,705-710.

5. Hol, W.GJ., van Duijnen, P.T. & Berendsen, HJ.C. (1978). Thea-helix dipole and the properties of proteins. Nature 273,443-446.

6. Somers, W.S. & Phillips, S.E.V. (1992). Crystal structure ofthe met repressor-operator complex at 2.8A resolution revealsDNA recognition by -strands. Nature 359, 387-393.

7. Zhang, R.-G., Joachimiak, A, Lawson, C.L, Schevitz, R.W.,Orwinowski, Z. & Sigler, P.B. (1987). The crystal structure oftrp aporepressor at 1.8A shows how binding tryptophan en-hances DNA affinity. Nature 327, 591-597.

8. Phillips, S.E.V., et at, & Stockley, P.G. (1993). E. coli metrepressor: DNA recognition by P-strands. In Nucleic Acids andMolecular Biology Vol. 7 pp 28-46, (Eckstein, F. & Lilley, D.MJ.eds.) Springer-Verlag, Berlin.

9. Warwicker, J. & Watson, H.C. (1982). Calculation of the elec-trical potential in the active site cleft due to a-helix dipoles.J. MoL Biol 157, 671-679.

10. Knight, K.L, Bowie, J.U., Vershon, A.K., Kelly, RD. & Sauer, RT.(1989). The Arc and Mnt repressors: a new class of sequence-specific DNA-binding proteins. J BioL Cbemn 264, 3639-3642.

11. Breg, J.N., Van Opheusden, J.HJ., Burgering, MJ.M., Boelens,R & Kaptein, RI (1990). Structure of Arc repressor in solution:evidence for a family of P-sheet DNA-binding proteins. Nature346, 586-589.

12. Nicholls, A & Honig, B.H. (1991). A rapid finite differ-ence algorithm utilizing successive over-relaxation to solve thePoisson-Bolmann equation. J. Comput Chem. 12, 435-445.

13. Connolly, M. (1985). Computation of molecular volume. J. AmChem. Soc 107, 1118-1124.

14. Klapper, I., Hagstrom, R., Fine, R., Sharp, KA & Honig, B.H.(1986). Focusing of electric fields in the active site of Cu-Znsuperoxide dismutase: effects of ionic strength and amino-acidmodification. Proteins 1, 47-59.

15. Weiner, S.J., et al, & Weiner, P. (1984). A new force field formolecular mechanical simulation of nucleic acids and proteins.J. Am. Cbem. Soc 106, 765-784.

16. Kraulis, PJ. (1991). MOLSCRIPT: a program to produce bothdetailed and schematic plots of protein structures. J. AppL Crystallogr. 24, 946-950.

Received: 11 Feb 1994; revisions requested: 1 Mar 1994;revisions received: 4 Mar 1994. Accepted: 7 Mar 1994.