The Role of Protonation and Metal ChelationPreferences in Defining the Properties ofMercury-binding Coiled Coils

Gregg R. Dieckmann1,2, Donald K. McRorie3, James D. Lear2

Kim A. Sharp2, William F. DeGrado2* and Vincent L. Pecoraro1*

1Department of ChemistryUniversity of Michigan930 North UniversityAnn Arbor, MI 48109-1055USA2The Johnson ResearchFoundation, Department ofBiochemistry and BiophysicsUniversity of Pennsylvania37th & Hamilton WalkPhiladelphia, PA 19104-6059, USA3Department of Research andApplications, BeckmanInstruments, Inc., 1050 PageMill Road, Palo Alto, CA 94304USA

To de®ne the delicate interplay between metal chelation, protein foldingand function in metalloproteins, a family of de novo-designed peptideswas synthesized that self-assemble in aqueous solution to form two andthree-stranded a-helical coiled coils. Each peptide contains a single Cysresidue at an a or d position of the heptad repeat. Peptide associationthus produces a Cys-rich coordination environment that has been used tobind Hg(II) ions. These peptides display a pH-dependent association,with trimers observed above the pKa of Glu side-chains and dimersbelow this value. Finite-difference Poisson-Boltzmann calculationssuggest that the dimeric state decreases the unfavorable electrostaticinteractions between positively charged Lys side-chains (relative to thetrimer). The Cys-containing peptides bind Hg(II) in a position-dependentfashion. Cys at a positions form three-coordinate Hg complexes at highpH where the trimeric aggregation state predominates, and two-coordi-nate complexes at lower pH. A d position Cys, however, is only able togenerate the two-coordinate complex, illustrating the difference in coordi-nation geometry between the two positions in the coiled coil. The bindingof Hg(II) was also shown to substantially increase the stability of the heli-cal aggregates.

# 1998 Academic Press

Keywords: de novo design; molecular modeling; coiled coils; Hg binding;three-coordinate Hg*Corresponding authors

Introduction

De novo protein design has emerged as an attrac-tive approach for addressing questions in proteinstructure/function. Through the design of modelpeptides and proteins it has been possible to assessthe contributions of hydrogen bonding, hydropho-bicity, and electrostatics to the stabilization of pro-tein three-dimensional structure (Beasley & Hecht,1997; Bryson et al., 1995). De novo design has beenused to probe the structural basis of the function of

metalloproteins (Berg, 1990; Lu & Valentine, 1997;Regan, 1995). Zn(II)-binding sites have been incor-porated within designed proteins to determinehow metal-ligand interactions affect protein stab-ility and dynamics (Handel & DeGrado, 1990;Handel et al., 1993; Regan & Clarke, 1990; Walkup& Imperiali, 1997). Also, heme (Arnold et al., 1997;Benson et al., 1995; Choma et al., 1994; D'Auriaet al., 1997; Nastri et al., 1997; Robertson et al.,1994), iron and iron-sulfur clusters (Coldren et al.,1997; Gibney et al., 1995, 1996; Pinto et al., 1997;Scott & Biggins, 1997), and other redox-active cen-ters (Winkler & Gray, 1992) have been introducedinto designed or natural proteins to determine howthe protein matrix affects the rates and thermodyn-amics of electron transfer reactions.

We have recently begun to use model proteinsto explore the interplay between metal chelationand protein folding in de®ning the structures ofmetalloproteins with unusual metal ion chelation

Present address: G. R. Dieckmann, Johnson ResearchFoundation, Department of Biochemistry andBiophysics, University of Pennsylvania, 37th &Hamilton Walk, Philadelphia, PA 19104-6059, USA.

Abbreviations used: CD, circular dichroism; FDPB,®nite-difference Poisson-Boltzmann; GuHCl, guanidine-HCl; [y]222, mean residue ellipticity; Mz, Z-averagemolecular mass.

Article No. mb981891 J. Mol. Biol. (1998) 280, 897±912

0022±2836/98/300897±16 $30.00/0 # 1998 Academic Press

sites (Dieckmann et al., 1997). Metal ion bindingsites often serve as nucleation sites to direct thefolding of small proteins, such as Zn(II) ®ngers(Berg, 1995). In such cases, the protein derivesmuch of its stability from the formation of metal-ligand interactions, with ligation geometries thatare highly symmetrical and well precedented inthe structures of small molecule model complexes.By contrast, in blue copper proteins such as plasto-cyanin and azurin, it is the fold of the proteinthat imposes a functionally important chelationenvironment on the metal ion (Malmstrom, 1994).These chelation geometries are often dif®cult toreproduce in small molecule complexes, and Wil-liams has referred to them as ``entatic'' or strainedstates (Vallee & Williams, 1968). How a proteinmay induce a seemingly strained geometry, whilealso maintaining excellent af®nity and speci®cityfor a metal ion, remains a major area of interest.

The chelation geometries of most metalloproteinslie along the continuum between these twoextremes. The metalloregulatory protein MerR hasan unusual three-Cys Hg(II)-binding site, whichhas been proposed to adopt a trigonal geometrybased on spectroscopic (Utschig et al., 1995; Wattonet al., 1990; Wright et al., 1990b) and mutagenesisstudies (Ross et al., 1989). This transcription factoris a sensor for the toxic Hg(II) ion and relates theconcentration of Hg(II) to the expression level of anumber of Hg-detoxi®cation proteins in prokar-yotes (Ralston et al., 1989). Hg(II) has a strong pro-pensity to form linear, two-coordinate complexeswith thiolate ligands; however, studies with smallmolecular thiolates in organic solvents have shownthat higher-order complexes can also be formed(Utschig et al., 1993; Wright et al., 1990a). Thermo-dynamic studies show that the ®rst two thiolatesbind to Hg(II) with high af®nity, while a thirdligand binds with signi®cantly reduced af®nity(Cheesman et al., 1988). Thus, while only 1:2 Hg-thiolate complexes are formed in dilute solution,three-coordinate complexes can be formed inmore concentrated solutions containing an excessof thiolate. The structures of these 1:3 Hg-thiolatecomplexes show roughly trigonal symmetry, andtheir spectroscopic properties coincide with thoseof MerR (Utschig et al., 1993; Wright et al.,1990a).

In previous work we designed peptides that selfassemble in dilute aqueous solution into a helicalcoiled coil with three Cys residues positionedappropriately to generate a three-coordinate Hg-thiolate complex (Dieckmann et al., 1997). In thesecomplexes, the free energy of folding of the peptideis used to drive the formation of the desired three-coordinate site in preference to the more common,linear two-coordinate geometry. As a structuralframework, we used Tri, a peptide patterned afterthe coiled coil peptide Coil-Ser (Betz et al., 1995;Lovejoy et al., 1993; O'Neil & DeGrado, 1990), andthe polyhepta-peptides described by Hodges andco-workers (Hodges et al., 1981, 1990; Lau et al.,1984; Figure 1(a)). Tri consists of four repetitions of

the heptad Leua-Lysb-Alac-Leud-Glue-Gluf-Lysg, inwhich the Leu residues at positions a and d areintended to form a hydrophobic core (Figure 1(b)).Further, the Glu and Lys residues at positions eand g, respectively, are positioned to form favor-able interhelical charged interactions if they inter-act in the desired parallel coiled coil, but should bedestabilizing if Tri forms an antiparallel (up-up-down) coiled coil. Molecular modeling was used tode®ne positions within Tri that might form a three-coordinate site (Dieckmann, 1995). In a parallelthree-stranded coiled coil, the a and d positionsproject toward the center of the structure wherethey lie in roughly planar layers perpendicular tothe helical axis. These layers are good candidatesfor the formation of the desired three-coordinatemetal-binding site. In addition, because peptidesrelated to Tri can adopt antiparallel as well as par-

Figure 1. (a) Amino acid sequences of the peptides usedin this study. The letters a through g at the top denotethe heptad positions. (b) Helical wheel diagram of paral-lel (left) and antiparallel (right) three-stranded coiledcoils, showing the locations of the heptad positions inthe structures. Residues in the top hydrophobic layerare circled.

898 Hg-Binding by De Novo Coiled Coils

allel structures (Lovejoy et al., 1993; Ogihara et al.,1997), we initially focused on the peptide L16C,which contains a Cys residue near the middle ofthe Tri peptide sequence (Dieckmann et al., 1997).In models, this peptide can form three-coordinateHg(II)-binding sites in both parallel and anti-parallel orientations. In the parallel structure(Figure 2(a)), a 3-fold symmetrical site is formed bythe convergence of three a residues within a singlelayer; a less symmetrical three-coordinate site canalso be formed between two layers if the peptideadopts an antiparallel structure (Figure 2(b)).

The designed peptides Tri and L16C were bothfound to exist in monomer/dimer/trimer equili-bria favoring either dimers or trimers, dependingon the solution conditions (Dieckmann et al., 1997).Dimers are preferentially formed at low pH, whilethe trimeric state predominates at pH 8.5. Further,at pH 8.5 L16C bound Hg(II) in a manner that wasdependent on the Hg:L16C stoichiometry; a three-coordinate Hg(II) complex was formed with a 1:3Hg:L16C ratio. At a higher Hg:peptide ratio (1:2),however, only two-coordinate complexes wereobserved, although the Hg(II)-linked dimerstended to further associate into higher order aggre-gates.

These studies opened a number of intriguingquestions which we address in this work. First, wewished to determine the thermodynamic basis forthe preferential formation of dimers at low pH.A second goal was to determine whether the three-coordinate complex is formed within a parallel orantiparallel coiled coil of a-helices. To determinethe coiled coil topology (parallel or antiparallelorientation), we prepared L9C (Figure 1(a)). As inL16C, the Cys residue of L9C occupies an a site,but its position is displaced towards the N termi-nus away from the center of the helix. In a parallel3-helical structure, the three Cys residues of (L9C)3

should converge about the 3-fold symmetry axis toform a three-coordinate site as in (L16C)3. How-ever, an antiparallel arrangement would place oneCys residue approximately 15 AÊ away from theother two Cys residues based on our models(Figure 2(c)).

Finally, we wished to determine whether athree-coordinate geometry for binding Hg(II)requires precise alignment of the Cys residues inthe trimer, or whether it is suf®cient to simply pos-ition them close together in space. To address thisissue, Cys was placed in the d position, formingL12C and L19C (Figure 1(a)). Studies with thisseries of peptides probe whether the formation of athree-coordinate site has strict geometric require-ments that are ful®lled only when the Cys residueoccupies the a position of a coiled coil.

While these studies were in progress, the crys-tal structure of CoilVaLd was determined to 2.1 AÊ

resolution (Ogihara et al., 1997). This parallelthree-stranded coiled coil is closely related toboth Coil-Ser and Tri, but contains Val instead ofLeu at each a position (Figure 1(a)). Unlike Coil-Ser, CoilVaLd trimerizes in a highly cooperative

process, with dimers being minimally populatedat neutral pH (Boice et al., 1996). A Cys residuewas placed at an a position of this peptide (Coil-

Figure 2. MOLSCRIPT (Kraulis, 1991) ®gures showingeight residue slices through (a) parallel and (b) antipar-allel three-stranded coiled coil models of L16C, showingthe ability of each to form a three-coordinate Hg(II)complex. The antiparallel helix in (b) is on the right.Hg(II) is represented as a sphere. Cys residues areshown as ball-and-stick representations. The third helixin the back is shown as a thin ribbon for clarity. (c) Rib-bon diagram showing parallel (left) and antiparallel(right) three-stranded coiled coil models of L9C. TheCys residues are shown as CPK surfaces.

Hg-Binding by De Novo Coiled Coils 899

VaLd-V16C) to con®rm that a three-coordinatesite could indeed be introduced within a parallelthree-stranded coiled coil of known three-dimen-sional structure.

Results

CD spectroscopy

The CD spectra of the peptides used in this studyare typical of highly helical peptides, with minimaat 208 nm (mean [y]208 � ÿ29,200(�2100) deg.cm2 dmolÿ1 resÿ1) and 222 nm (mean [y]222 �ÿ29,400(�1600) deg. cm2 dmolÿ1 resÿ1), and amaximum at 192 nm ([y]192 � 69,800(�2900) deg.cm2 dmolÿ1 resÿ1) observed for each peptide atpH 2.5 and 8.5. The Cys substitutions at either a ord heptad positions had little effect on the CD spec-tra and, therefore, the secondary structure of thepeptides. Little change was also seen in the CDspectra when in the presence of Hg(II) (data notshown).

Sedimentation equilibrium ultracentrifugation

Previously, sedimentation equilibrium ultracen-trifugation was used to demonstrate that Tri andL16C existed in a monomer/dimer/trimer equili-brium, which favored dimers at low pH (2.5) andtrimers at high pH (8.5) (Dieckmann et al., 1997).L12C was similarly examined as a representativepeptide with the Cys substitution at a d rather thanan a position (Table 1). As reported for Tri andL16C, L12C also forms trimers at pH 8.5 anddimers at pH 2.5, both in the presence and absenceof added salt. At intermediate pH (5.5), Tri wasfully trimeric, while L12C and L16C showed inter-mediate degrees of association, with the trimericstate being favored in the presence of salt. Thus, achange in the protonation state of Glu side-chains(vide infra) is accompanied by a switch in aggrega-tion state from dimer at low pH to trimer at higherpH. This pH-induced change in aggregation statehas not been previously reported for other coiledcoil systems.

The pH-dependence of GuHCl-inducedunfolding transitions

The stabilities of the dimeric and trimeric formsof the peptides were determined from theirGuHCl-induced unfolding transitions. Figure 3illustrates the variation in the mean residue ellipti-city, [y]222 (a measure of the helical content), of thepeptides as a function of the denaturant concen-tration at pH 2.5. In agreement with previous work(Zhou et al., 1993), the replacement of Leu forreduced Cys resulted in destabilization of thecoiled coil, with the effect being more pronouncedwhen the substitution occurred at a d position; thed position (L12C and L19C) and a position substi-tutions (L9C and L16C) had denaturation mid-points of 4.8 and 5.0 M GuHCl, respectively, atpH 2.5 (Figure 3). Thus, the Cys substitutionsresulted in changes in stability that are well prece-dented and predictable based on previous studies.We therefore focused further pH-dependent ther-modynamic studies on Tri and L12C, which rep-resent the most and least stable of the series ofpeptides, respectively.

Figure 4(a) illustrates GuHCl titrations for Tri atvarious pH values. As has been described for var-ious designed and natural coiled coil peptides(Hodges et al., 1994; Kohn et al., 1995; Lowey, 1965;Noelken & Holtzer, 1964; O'Shea et al., 1992; Zhouet al., 1992, 1993, 1994), Tri shows increased resist-ance to GuHCl-induced denaturation as the pH islowered through the pKa of the Glu residues. Thethermodynamic stability of L12C shows a similarpH dependence (Figure 4(b)), although the pH sen-sitivity continues above pH 6.5 for L12C, whereasTri shows no change above this pH. Thisadditional change in stability with pH for the Cys-

Table 1. Sedimentation equilibrium results for Tri, L12Cand L16C

pH [NaF] (mM) Tri L12C L16C

2.5 0 2.3 2.0 NAa

50 2.3 2.1 2.05.5 0 3.3 2.1 1.8

50 3.0 2.5 3.08.5 0 3.5 2.9 3.0

50 3.5 2.7 3.1

Association states determined by dividing the apparent Z-aver-age molecular mass by the theoretical monomeric molecularmass of each peptide. The 95% con®dence levels were within10% of the stated values and are given in the Supplemen-tary Material.

a Data not available.

Figure 3. GuHCl denaturation titrations at pH 2.5. Theratio of [y]222 at 25�C versus 0�C ([y]T/[y]o) is plottedagainst GuHCl concentration for Tri (open circles), L9C(®lled squares), L12C (®lled triangles), L16C (opensquares), L19C (open triangles) and CoilVaLd-V16C(®lled diamonds). Curves are monomer/dimer ®ts tothe data (see Table 2 for details).

900 Hg-Binding by De Novo Coiled Coils

containing peptides is attributed to titration of theCys side-chain. Figure 5 shows the pH dependenceof [y]222 for Tri and L12C measured at 4.7 MGuHCl for Tri and 3.1 M GuHCl for L12C (theseGuHCl concentrations were selected to provide amaximal change in [y]222 over the pH range). Thecurves show transitions with midpoints in thevicinity of the Glu carboxylate, which is the onlyside-chain with a pKa near this value (Stryer, 1988).Thus, the transition involves a change in the proto-nation state of one or more Glu side-chains (proto-nation of the Cys side-chain is ruled out by the factthat the same transition is seen for Tri, which lacksCys).

The stabilities of Tri and L12C were quantitat-ively evaluated at pH 2.5 and pH 8.5 using mono-mer/dimer and monomer/trimer formalisms,

respectively (Boice et al., 1996) and summarized inTable 2 and Figure 6. At intermediate pH, the dataare less easily interpretable due to the populationof both dimers and trimers. For instance, the con-centration dependence of [y]222 for Tri at a constantconcentration of denaturant (5.5 M GuHCl) and atintermediate pH (5.5) shows a cooperativity of 2.5(Supplementary Material), similar to that observedpreviously for Coil-Ser (Betz et al., 1995). A similarconcentration dependence is observed for the Cys-substituted peptides (data not shown).

The GuHCl denaturation curves at pH 2.5 arewell described by cooperative monomer/dimerequilibria, and the pH 8.5 data are similarlydescribed using monomer/trimer equilibria. Theextrapolated dimer dissociation constants (KD) atpH 2.5 for Tri and L12C were 1.3 � 10ÿ15 and1.1 � 10ÿ12 M, respectively, corresponding todimerization free energies per monomer of 10.2(Tri) and 8.2 (L12C) kcal/mol (1 M standard state).The analysis of the trimerization reactions at pH 8.5gave overall trimer dissociation constants (KT) forTri and L12C of 5.9 � 10ÿ18 and 8.9 � 10ÿ14 M2,respectively (calculations in which the cooperativ-ity of oligomerization was less than 3.0 failed tosigni®cantly change this result). These KT valuescorrespond to free energies of trimerization permonomer of 7.8 and 5.9 kcal/mol for Tri andL12C, respectively. Thus, on a per monomer basis,protonation of the Glu residues leads to an averagestability difference of 2.3 kcal/mol. It should benoted that comparisons of the energetics of bimole-cular and termolecular processes at other concen-trations require consideration of the differences inthe dimensions of the equilibrium constants fromwhich the free energies are calculated (Brady &Sharp, 1997; Janin, 1996). A simple calculationshows that the stability difference between the

Figure 4. GuHCl denaturation titrations of (a) Tri and(b) L12C at various pH values are plotted as mean resi-due ellipticity at 222 nm versus GuHCl concentration.For (a) Tri: pH 2.5 (open circles), pH 3.6 (®lled circles),pH 4.6 (open squares), pH 5.0 (®lled squares), pH 5.6(open diamonds), pH 6.6 (®lled diamonds) and pH 8.5(open triangles). For (b) L12C: pH 2.5 (open circles),pH 4.0 (®lled circles), pH 4.6 (open squares), pH 5.0(®lled squares), pH 6.0 (open diamonds), pH 7.5 (®lleddiamonds) and pH 8.5 (open triangles).

Figure 5. Mean residue ellipticty at 222 nm versus pHfor Tri (open circles) and L12C (®lled circles). The plotwas made by taking [y]222 at 4.7 M GuHCl (for Tri, seeFigure 4(a)) or 3.1 M GuHCl (for L12C, see Figure 4(b))and plotting those values against pH.

Hg-Binding by De Novo Coiled Coils 901

dimer and trimer is signi®cant at all experimentallyaccessible peptide concentrations. For instance, if apeptide showed a KD of 1.3 � 10ÿ15 M and an over-all KT of 5.9 � 10ÿ18 M2 (corresponding to the KD

and KT for Tri), then the dimer would be the pre-dominant species at all peptide concentrationsbetween 10ÿ15 M and 109 M.

Calculation of the electrostatic contribution toaggregate formation

In order to understand the possible role thatelectrostatics play in determining the preference byTri and its Cys-substituted derivatives for thedimeric aggregation state at low pH, ®nite-differ-ence Poisson-Boltzmann (FDPB) calculations wereperformed on dimeric and trimeric models of Triusing DelPhi (Gilson et al., 1988; Nicholls et al.,1991; Sharp & Honig, 1990). Because it is probablethat the positively charged Lys side-chains play akey role in the electrostatic interactions betweenpeptide monomers, atomic charges were placedonly on the atoms in the Lys side-chains. Theresults of the DelPhi calculations are given inTable 3. For both the dimeric and trimeric aggrega-tion states, the electrostatic free energy for protona-tion of Lys side-chains is positive (unfavorable), asis expected in a system where only positive formalcharges are interacting. However, the magnitudeof this penalty is 0.4 kcal/mol larger for each Lysin the trimeric versus dimeric state. There are twocontributions to this unfavorable energy: thedesolvation of the Lys ammonium groups and theLys/Lys electrostatic repulsions. Both are greaterin the trimer versus the dimer.

Hg(II)-binding studies of peptides with Cyssubstitutions at position a

In previous work, we demonstrated that UV/visible spectroscopy is a convenient and reliablemethod to determine the Hg(II) chelation proper-

ties of L16C (Dieckmann et al., 1997). Titrations ofthe other a-substituted peptides (L9C in Figure 7(a)and CoilVaLd-V16C in Figure 8(a)) were quite simi-lar. Their UV spectra indicate that they form simi-lar two- or three-coordinate complexes at pH 8.5depending on the Hg-peptide ratio (Figure 7(a)and (b)). As an example, a titration experiment inwhich CoilVaLd-V16C (an a substitution) is addedto an aqueous solution of HgCl2 is shown inFigure 8(a). Under conditions where the peptideis fully trimeric in the apo-state (pH 8.5), thetitration is biphasic. A peak of low intensity(�e240 � 2700 Mÿ1 cmÿ1) is observed after theaddition of 0.2 to 2.0 equivalents of peptide. Thisspectral feature is typical of two-coordinateHg-thiolate complexes, as has been con®rmedby EXAFS and 199Hg NMR for L16C (Dieckmannet al., 1997). Addition of further aliquots of peptideleads to a new set of spectral features at 247,265 and 295 nm (�e247 � 16,800 Mÿ1 cmÿ1;�e265 � 10,600 Mÿ1 cmÿ1; �e295 � 5000 Mÿ1 cmÿ1),which are indicative of three-coordinate Hg-thio-late complexes. Again, this assignment has beencon®rmed for L16C by EXAFS and 199Hg NMR(Dieckmann et al., 1997). The intensity of thesepeaks increases linearly with respect to peptideconcentration between 2.0 and 3.0 equivalents ofadded peptide, after which no further spectralchanges are observed.

To further demonstrate that the associative equi-libria are critical for de®ning the three-coordinateHg(II) chelation site, the binding of Hg(II) by L9C,L16C and CoilVaLd-V16C was also examined atlow pH (2.5 and 5.5), where the peptides are pre-dominantly dimeric (Table 1). Under these con-ditions, titrations of the peptides into 10 mM HgCl2(as shown for L16C in Figure 8(b)) gave rise tospecies with spectral characteristics generallyassociated with two-coordinate complexes(Figure 7(a) and (b)). This assignment has beencon®rmed for L16C by EXAFS and 199Hg NMRspectroscopy (unpublished results). At these low

Table 2. Results of ®ts to GuHCl denaturation curves of Tri and Cys-containing peptides

Folded baseline Unfolded baselinePeptide pH na [y]nmer

0 b bnmer [y]mon0 b bmon �G0 c mG

Tri 2.5 2 ÿ29100 450 ÿ2300 330 20.3 (10.2) ÿ2.38.5 3 ÿ29300 620 ÿ5700 990 23.5 (7.8) ÿ2.7

L12C 2.5 2 ÿ29300 790 ÿ6100 950d 16.3 (8.2) ÿ2.08.5 3 ÿ29200e 700e ÿ7200 1300 17.8 (5.9) ÿ2.5

L9C 2.5 2 ÿ29000 ÿ10 ÿ6600 950d 15.9 (8.0) ÿ1.8L16C 2.5 2 ÿ29100 610 ÿ6400 950d 16.5 (8.3) ÿ1.9L19C 2.5 2 ÿ29300 810 ÿ6500 950d 16.1 (8.1) ÿ1.9V16C e 2.5 2 ÿ29100 1000 ÿ3800 460 13.1 (6.6) ÿ1.7

Calculations assumed a 1 M standard state.a Aggregation state of folded species.b Mean residue ellipticity at 222 nm. Units are deg cm2 dmolÿ1 resÿ1.c Units are kcal/mol. Error is �0.5 kcal/mol, based on the sensitivity of the value when the folded and unfolded baselines were

®xed at reasonable extremes. The values in parentheses are per monomer.d These values were ®xed during the ®tting procedure because the pre or post-transition baselines were not adequately de®ned.

The values used were chosen based on the slopes of analogous peptides, and a sensitivity analysis showed that their values couldbe varied over a reasonable range without signi®cantly affecting the value of �G0.

e CoilVaLd-V16C.

902 Hg-Binding by De Novo Coiled Coils

pH values, no evidence for the formation of athree-coordinate species was observed for the a-substituted peptides, even at the highest pepti-de:Hg(II) ratio examined (5:1).

Hg(II)-binding studies of peptides with Cyssubstitutions at position d

We also examined the Hg(II)-binding propertiesof the peptides L12C and L19C, in which a Cys

residue substitutes for Leu at a d position(Figure 7(c) and (d)). Although these peptides aretrimeric at pH 8.5, UV titrations indicate that theyonly form two-coordinate complexes, with no evi-dence for the formation of a three-coordinatespecies (as shown for L12C in Figure 8(c)). Similarresults were observed at lower pH (Figure 7(c) and(d)). The inability of d-substituted Cys-peptides toform three-coordinate Hg(II) complexes, even atpH 8.5 where the trimeric aggregate is favored,indicates that the formation of a three-coordinateHg(II) complex requires not only the formation ofa trimeric structure, but also the proper orientationof the Cys side-chains for productive interactionswith Hg(II); this proper orientation is onlyachieved when the Cys is at an a position(Figure 7).

GuHCl denaturations of Cys-peptides in thepresence of Hg

Having shown that Hg(II) binds to the Cys-con-taining peptides, we next investigated the stabilitychanges of the peptides in the presence of Hg(II).The addition of Hg(II) increases the thermodyn-amic stability of the folded forms of L16C (Figure 9)and L12C (data not shown) as assessed from theirGuHCl-induced unfolding transitions. For a 1:2Hg:peptide ratio at pH 2.5, a single unfolding tran-sition was observed near approximately 6 MGuHCl, suggesting that this peptide formed ahomogeneous complex consisting of one Hg(II) perdimer. At lower Hg:peptide ratios, two unfoldingtransitions were observed, presumably re¯ectingthe presence of both free and complexed dimers.For instance, at Hg(II):peptide ratios of 1:3 and 1:4(pH 2.5), the unfolding curves are biphasic(Figure 9(a)). Similar behavior is observed atpH 8.5, where the peptide is able to form three-coordinate complexes (Figure 9(b)), except thecurves become biphasic at different Hg(II):peptideratios, re¯ecting the ability of this peptide to formdifferent Hg(II) complexes at the two pH values.Thus, at high pH the denaturation curves for the1:2 and 1:3 Hg:peptide samples are very similar,re¯ecting the fact that this peptide can form bothtwo-coordinate and three-coordinate complexes.Only upon addition of a fourth equivalent of pep-tide do we observe a large change in the denatura-tion curve, consistent with the presence ofuncomplexed trimers.

A second difference in the denaturation curvesat low versus high pH is that the binding of Hg(II)

Figure 6. Fits of GuHCl denaturation curves for (a) Triand (b) L12C. The pH 2.5 data (open circles) were ®t tomonomer/dimer equilibria, and the pH 8.5 data (®lledcircles) were ®t to monomer/trimer equilibria. The ®tsare shown as continuous curves through the data.Folded and unfolded baseline ®ts for the pH 2.5 and 8.5data are shown as broken and dotted lines, respectively.Fit results are reported in Table 2.

Table 3. Electrostatic free energies in kcal/mol calculated by DelPhi

Aggregation Lys-H� a Lys b (Lys-H�)ÿ(Lys) per Lys c

Dimer 1.78 0.02 1.76 0.11Trimer 13.27 1.34 11.93 0.50

a Lys side-chains protonated.b Lys side-chains deprotonated.c Eight Lys residues per monomer.

Hg-Binding by De Novo Coiled Coils 903

shows a larger effect on the stabilities of thepeptides at pH 8.5. The difference in the GuHCldenaturation midpoints (presence versus absenceof 0.5 equivalents Hg(II) per peptide) wasapproximately 2.0 M at pH 2.5; the correspond-ing difference at pH 8.5 was approximately4.5 M. This difference may be associated withdifferences in the chelation properties of theunfolded states. At high pH, where thiolates areeasily formed, two-coordinate complexes areformed, even in dilute aqueous solution(Cheesman et al., 1988). Thus, the unfolded pep-tides are expected to form Hg(II)-linked dimersat high pH. Crosslinking of the unfolded pep-tides should decrease the entropic requirementsfor folding, leading to a large increase in thepeptides' resistance to GuHCl-induced denatura-tion in the presence of the Hg(II) (Tamura &Privalov, 1997). By contrast, at low pH, wherethe formation of Cys thiolates is less favorable,the binding of Hg(II) should be weakened, andthe Hg(II) may be able to dissociate in theunfolded state. In this case, addition of Hg(II)would stabilize folding by a distinct mechanism

involving preferential binding to the folded ver-sus the unfolded state.

Discussion

The primary goals of this work were to deter-mine the source of the pH-dependent shift in theassociation state of the Tri family of peptides, andto explore the geometric requirements for for-mation of three-coordinate Hg(II)-thiolate com-plexes. The observed shift in association state to adimer at low pH could be a consequence of desta-bilizing the trimer or stabilizing the dimer; ourresults show that the dimer is very strongly stabil-ized at low pH. The increased stability of coiledcoil proteins at lower pH has also been observed inseveral other systems, including tropomyosin(Lowey, 1965; Noelken & Holtzer, 1964), the Fos-Jun leucine zipper (O'Shea et al., 1992), and severalsynthetic models of coiled coils (Zhou et al., 1992).As discussed previously, this phenomenon rep-resents the balance of a large number of opposingforces. At low pH, the Glu residues are protonated,which increases their helix propensity thereby sta-

Figure 7. UV spectra of Hg-peptide complexes of (a) L9C, (b) L16C, (c) L12C and (d) L19C. Broken curve, 1:2 Hg:pep-tide at pH 2.5; continuous curve, 1:3 Hg:peptide at pH 2.5; Bold broken curve, 1:2 Hg:peptide at pH 8.5; bold con-tinuous curve, 1:3 Hg:peptide at pH 8.5. The data for L16C are taken from Dieckmann et al. (1997) and are includedfor comparison to L9C and CoilVaLd-V16C (Figure 8(a)).

904 Hg-Binding by De Novo Coiled Coils

bilizing the folded state. Also, protonationdecreases the polarity of the Glu side-chains,which may decrease their free energy of desolva-tion when placed next to hydrophobic Leu resi-dues. However, protonation should also decreasethe favorable electrostatic interactions betweenionized Glu and Lys side-chains. These electrostaticinteractions are expected to be partially screenedby exogenous electrolytes, particularly in the dena-turations which employ high concentrations of thesalt GuHCl. Thus, although a large number ofopposing forces in¯uence the free energy balance,folding is favored at low pH.

These explanations provide a rationale for thegeneral stabilization of coiled coils at low pH, butwhy are two-helix coiled coils preferred over three-helical aggregates as the pH is lowered? One factoris that the dimeric state has a larger surface-to-volume ratio than the trimer, leading to lower elec-trostatic repulsions between (and less desolvationof) the Lys side-chains in the low pH form. Tofurther test these ideas, we constructed trimericand dimeric computer models of Tri with fully pro-tonated Glu residues and calculated the total elec-trostatic energy associated with the Lys side-chainsinteracting with each other using DelPhi. It wasfound that this electrostatic free energy is signi®-cantly more unfavorable for the trimer versus thedimer. Thus, it appears that electrostatics may playan important role in de®ning the preference for adimeric state at low pH. Other factors mightinclude the preferential packing of protonated Gluside-chains against the apolar residues at positionsa and d, although an in depth analysis wouldrequire high resolution structures of the low andhigh pH forms of the peptide in both dimeric andtrimeric conformations.

A second goal of this research was to de®ne thetopologies of the Hg(II)-binding peptides related toTri. They appear to adopt a parallel coiled coil con-formation as evidenced by the Hg-binding proper-ties of the a-substituted and d-substituted familesof peptides. Within a given family, the chelationproperties of each peptide are invariant withrespect to the position of the Cys-substitutionalong the chain, ruling out an antiparallel model(Figure 2(b)).

Our results demonstrate that the default coordi-nation state for this class of peptides is two-coordi-nate, and that the coordination of three thiolatesrequires relatively stringent geometric and chemi-cal conditions. At pH 2.5 and 5.5, all the Cys-derivatives form complexes with spectral charac-teristics similar to two-coordinate Hg(II) species.Also, at pH 8.5, all of the peptides studied were tri-meric, yet only the a-substituted Cys-derivatives ofTri and CoilVaLd were capable of producing three-coordinate Hg(II) species. Further, the a-substitutedpeptides formed three-coordinate complexes onlywhen they were in greater than twofold molarexcess over Hg(II), forming two-coordinate com-plexes at lower peptide:Hg(II) ratios.

Figure 8. Peptide titrations of aqueous HgCl2solutions. Spectra shown are difference spectra(epeptide � Hg ÿ epeptide). (a) CoilVaLd-V16C at pH 8.5.Spectra correspond to 0, 0.6, 1.2, 1.8, 2.4, 2.8, 3.0, 3.5, 4.0and 4.5 equivalents added peptide. (b) L16C at pH 5.5.Spectra correspond to 0 to 5.2 equivalents peptide (0.4increment). (c) L12C at pH 8.5. Spectra correspond to 0to 5.2 equivalents peptide (0.4 increment). The lack ofan isobestic point for the titrations in (b) and (c) is poss-ibly due to the formation of multinuclear Hg-peptidecomplexes when Hg(II) is in a large molar excess overthe peptides, as is observed for other thiolate ligands(Bowmaker et al., 1984, 1996; Busetto et al., 1990; Wang& Fackler, 1989; Wright et al., 1990a, and referencestherein).

Hg-Binding by De Novo Coiled Coils 905

The ®nding that the a-substituted peptides couldform either two or three-coordinate complexesmay be understood in terms of the known ligandexchange properties of Hg(II)

HgCl2 � 2R-Sÿ � Hg�S-R�2 � 2Clÿ

�G0a � ÿ78 kcal=mol

Hg�S-R�2 �R-Sÿ � �Hg�S-R�3�ÿ

�G0b � ÿ1:2 kcal=mol

(Wright et al., 1990a). The addition of the ®rst twothiolates to Hg(II) proceeds with an overwhel-mingly favorable free energy, which can easilyovercome the intrinsic conformational preferenceof the peptide to form trimers. Thus, at pepti-de:Hg(II) ratios of less than 2:1 there is a strong

driving force for formation of two-coordinatecomplexes. However, once the molar ratioexceeds 2:1 peptide:Hg(II), the additional equival-ent of peptide can bind to Hg(II) to form three-coordinate complexes, although the addition ofthe third ligand proceeds with a considerably lessfavorable �G0

b. For small molecule thiolateligands, �G0

b is insuf®cient to drive the bindingof a third thiolate ligand at low micromolar con-centrations of metal and ligand in aqueous sol-ution. However, under similar conditions a thirdequivalent of L9C, L16C, or CoilVaLd-V16C isable to bind in a three-coordinate manner, guidedby the conformational preference of these pep-tides to form three-helical coiled coils.

The free energy balance between the confor-mational preferences of the peptides and the chela-tion properties of Hg(II) also explains the fact thatthe peptides failed to form three-coordinate com-plexes at low pH. For instance, at pH 2.5 and 5.5the peptides are predominantly dimeric in theirapo-state, which should favor two-coordinate bind-ing modes. Further, this pH is far from the pKa

value of 8.5 for Cys thiol groups (Stryer, 1988),which attenuates their af®nity for Hg(II), assumingthey bind as thiolates. Thus, the af®nity would bedecreased by approximately 4 kcal per mole thiolat pH 5.5 versus pH 8.5, based on the difference inthe pH versus the pKa of the Cys thiolate. Clearly,the addition of the ®rst two thiolates will remainoverwhelmingly favorable at pH 5.5. However, theaddition of a third ligand would be rendered unfa-vorable, and the peptides would default to two-coordinate binding modes.

The failure of the trimeric d-substituted peptidesto form three-coordinate complexes at pH 8.5further illustrates the stringency of the geometricrequirements for three-coordinate binding. Thedifferences between the binding properties of dversus a-substituted peptides may arise from thedifferences in orientation of side-chains at thesepositions with respect to the superhelical axis incoiled coils (Harbury et al., 1994). To investigatethis possibility, Cys was introduced at the a and dpositions of the crystal structure of CoilVaLd

(Figure 10). The w1 angles for the Cys substitutionswere kept identical to those for the original resi-dues, thus making the Leu (or Val) to Cysmutations as isosteric as possible. Based on ourmodeling studies, the a position Cys residues ori-ent their side-chains towards the superhelical axis,whereas the d position Cys residues point awayfrom the center of the coiled coil. These differentorientations result in a signi®cant difference inthe distance of the Cys sulfur atom from the cen-tral axis (2.1 AÊ at a versus 3.9 AÊ at d) as well asthe average Cb-S-Hg angle (149� for the a pos-ition, 60� for the d position). The angles and dis-tances for only the a positions approach thosepreviously observed in three-coordinate Hg(II)complexes (2.43 AÊ and 120�, respectively; Wrightet al., 1990a). Further, energy minimization usingthese values as constraints led to excellent geome-

Figure 9. GuHCl denaturation titrations of L16C at(a) pH 2.5 and (b) 8.5 as a function of Hg(II) concen-tration. Mean residue ellipticities at 222 nm versusGuHCl concentration are plotted for L16C alone (opencircles), 1:2 Hg:L16C (®lled squares), 1:3 Hg:L16C (opendiamonds) and 1:4 Hg:L16C (®lled triangles). The linesdrawn through the data are shown for clarity and donot imply theoretical ®ts. Similar results were seen forL12C (data not shown).

906 Hg-Binding by De Novo Coiled Coils

tries for the a-substituted peptides, with lowenergy w angles for all amino acids and idealhelical geometries. By contrast, accommodation ofa three-coordinate site at the d position requiredthe introduction of high-energy rotamers or dis-tortions to the helical backbone. Other lowenergy rotamers for the d position Cys residueswere unable to improve these Cys-Hg inter-actions without signi®cant unfavorable stericinteractions with other hydrophobic layers.

We also modeled a two-coordinate site at the aand d positions of a two-stranded coiled coil. Incontrast to the trimer, where a clear geometric pre-ference for three-coordinate binding at the a sitewas observed, no strong preference was observedfor the two-coordinate complexes, providing arationale for the observation that two-coordinatesites may be formed at either the a or d sites of theCys-substituted peptides.

Our studies demonstrate the ability to design anenergetically challenging metal-chelation site intodesigned, water-soluble peptides. Through the useof associating amphiphilic helices, it has been poss-ible to create a three-coordinate binding site sur-rounded by hydrophobic side-chains within awater-soluble structure. The ability to control thehydrophobicity, water-accessibility and chelationgeometries of sites represents an important steptowards the overall goal of designing functionalmetalloproteins. Our results clearly show thataccurate starting geometries and careful consider-ation of the torsional preferences of the chelatingside-chains will often be essential for creating func-tional binding sites.

Materials and Methods

Modeling

The methods used for generating coiled coil modelswere similar to those described (Betz & DeGrado, 1996;Dieckmann, 1995; Shen et al., 1996). A polyalaninea-helix with 3.6 residues per turn was created in the pro-gram Insight 95 (Biosym Technologies, Inc.) using back-bone dihedral angles f � ÿ57� and c � ÿ47�. Aftercentering the helix at the origin and aligning it along the

z-axis of the Cartesian coordinate system, the a and dpositions of the helix were replaced with Leu residues intheir low energy rotamers (Dunbrack & Cohen, 1997;Dunbrack & Karplus, 1994; McGregor et al., 1987). Thesehydrophobic residues were used to form the interior ofthe coiled coil. The helix was converted to 3.5 residuesper turn, such that all residues at a given heptad positionline up when viewed down the helical axis. This conver-sion was performed using the matrix transformation inequation (1):

xnew�i�ynew�i�znew�i�

�cos�y� ÿsin�y� 0sin�y� cos�y� 0

0 0 1

24 35 x�i�y�i�z�i�

24 35 �1�

where y � ÿ2p[z(i) ÿ zmin]/p; x(i), y(i) and z(i) are theoriginal coordinates of atom i; xnew(i), ynew(i) and znew(i)are the transformed coordinates; zmin is the minimumz-coordinate of the Ca atoms; and p is the superhelicalpitch of the coiled coil. For this transformation, p isÿ189 AÊ .

Two and three-helical coiled coils were then generatedby taking the appropriate number of copies of the abovehelix and varying r (distance of the helix from the centerof the coiled coil) and e (rotation of each individual helixaround its axis to position the a and d residues towardsthe center of the coiled coil). For antiparallel models, thedirectionality of one helix was reversed. The bundle ofhelices was then given a left-handed coil using equation(1) with p � 189 AÊ (Crick, 1953; Dunker & Zaleske,1977). Appropriate values for r and e were evaluatedqualitatively by looking at the packing in the hydro-phobic interior of the coiled coil. The resulting coiledcoils with reasonable packing were subjected to in vacuoenergy minimizations using Discover (Biosym Technol-ogies, Inc.) and the CVFF force®eld (Dauber-Osguthorpeet al., 1988), with the positions of the backbone atoms ofthe helices ®xed. For a round of minimization, a combi-nation of steepest descents (500 iterations) followed by6000 iterations of conjugate gradients was utilized.Another round of minimization then followed withoutbackbone constraints. Next, the Ala residues in positionsb, c, e, f and g of the helices were replaced with theappropriate residues, and another two rounds of mini-mization followed (the ®rst with the backbone atomsconstrained, the second round without constraints).Minimizations were also performed in the presence of abox of water with periodic boundary conditions. Theonly major differences between the models re®nedin vacuo and in water were the ®nal rotamers observedfor the solvent-exposed polar residues, as would beexpected. Models of peptides containing Cys residues ata or d positions were generated by replacing the appro-priate Leu residues in the ®nal model, followed by around of energy minimization.

For the incorporation of Hg(II)-binding sites into thesemodels, the standard CVFF force®eld was modi®ed toinclude bonding parameters for Hg(II) coordinated bytwo or three thiolate ligands (Dieckmann, 1995). The par-ameters for Hg(II) were obtained from structural(EXAFS: Utschig et al., 1993) and thermodynamic data(Wright et al., 1990a) for Hg-thiolate model complexes.Addition of a Hg(II) to a model required addition of Hg-S bonds to the model, followed by a round of minimiz-ation. The parameters for Hg(II) are available from theauthors upon request.

Figure 10. MOLSCRIPT Figure showing eight residueslices through the crystal structure of CoilVaLd with Cysmutations at a (left) and d (right) positions of the hep-tad. Cys residues are shown as ball-and-stick represen-tations. The other hydrophobic residues are omitted forclarity.

Hg-Binding by De Novo Coiled Coils 907

Peptide synthesis and purification

All peptides were synthesized on a Milligen model9050 Plus continuous ¯ow peptide synthesizer usingFMOC-protected amino acids and previously publishedprocedures (Choma et al., 1994). Each peptide was acetyl-ated on the N terminus and amidated on the C terminusto stabilize helix formation. Peptides were puri®ed byreverse phase HPLC; subsequent purity checks by ana-lytical HPLC and either positive ion electrospray ioniz-ation or matrix assisted laser desorption ionization massspectrometry indicated minimum purity levels of at least95% for all peptides.

Preparation of peptide and Hg(II) solutions

Stock solutions of the peptides were prepared in dou-bly distilled water. For peptides containing Cys, Tris(2-carboxyethyl)phosphine hydrochloride (TCEP �HCl;Pierce) was added to give a 2:1 TCEP to peptide ratio toprevent oxidation. TCEP interfered with Hg(II)-bindingand therefore was not added to those experiments. Theconcentration of peptide in each stock solution wasdetermined by amino acid analysis (Protein and Carbo-hydrate Structure Facility, University of Michigan; or theProtein Chemistry Core Facility, Columbia University).For experiments requiring Hg(II), a stock solution ofHgCl2 was prepared in doubly distilled water and ana-lyzed by atomic absorption spectroscopy (MicroanalysisLaboratory, Department of Chemistry, University ofMichigan) or inductively coupled plasma (ICP; Quanti-tative Technologies Inc.).

Sedimentation equilibrium

Sedimentation equilibrium studies were performed oneither a Beckman XL-A analytical ultracentrifuge usingabsorbance optics or an XL-I analytical ultracentrifugewith interference optics. Nine data sets were ®t simul-taneously to a single ideal species model varying refer-ence signal, molecular mass and baseline offset using theBeckman Data Analysis Software. Sample data were col-lected for three sample concentrations (serial dilutionbetween 1 and 0.25 mg/ml) at three rotor speeds(20,000, 24,000 and 36,000 rpm, or else 30,000, 36,000 and50,000 rpm) for each Z-average molecular mass (Mz).Values for Mz were determined at the indicated pHusing 5 mM phosphate buffer both with and without50 mM NaF. On the XL-A, the absorbance was measuredat a wavelength between 230 nm and 240 nm (chosen tominimize the noise for each sample) at 4�C to inhibit Cysoxidation. For Cys-containing peptides, control exper-iments were also run on the XL-I in the presence andabsence of 65 mM DTT at 20�C and showed no changein sedimentation behavior. Partial speci®c volumes of0.776 and 0.767 were calculated for Tri and the cysteine-substituted derivatives of Tri, respectively, from the pep-tide composition (Cohn & Edsall, 1965). The solvent den-sity was calculated to be 1.004 (Laue et al., 1992).

Circular dichroism (CD) spectroscopy

CD spectral measurements were made at 25�C usingan AVIV 62DS spectrometer using a 1 mm pathlengthrectangular cuvette. Spectra were collected from 190 nm

to 300 nm, averaging for ten seconds at each wave-length. Samples contained 10 mM peptide and 360 mMpotassium phosphate buffer.

Guanidine-HCl denaturation titrations

To perform GuHCl titrations, two stock solutions dif-fering in GuHCl concentration were prepared and mixedto generate samples ranging from 0 M to 7.0 M GuHCl:stock solution 1 was 0 M GuHCl, 10 mM peptide (or pep-tide-Hg(II) complex) and 5 mM phosphate; stock sol-ution 2 was 7.33 M GuHCl, 10 mM peptide (or peptide-Hg(II) complex) and 5 mM phosphate. The GuHCl con-centration in stock solution 2 was determined fromrefractive index measurements at 25�C (Pace, 1986). ThepH of both stock solutions was adjusted to the desiredvalue. The titration was performed in a stirred 1 cmpathlength rectangular cell by a Microlab 500 series auto-matic titrator (Hamilton Co.) controlled by a PowerComputing computer running AVIV-supplied softwarewritten in Igor Pro (WaveMetrics). For each new point inthe titration, the sample was stirred for 90 seconds, then300 data points were collected at 25�C and averaged toobtain the signal. The data were corrected for buffer con-tributions to the CD signal by subtracting a buffer blank.The mean residue ellipticity, [y]222, was calculated withequation (2):

�y�222 � yobs=�10 � l � c � n� �2�where yobs is the ellipticity in millidegrees, l is the path-length of the cell in centimeters, c is the peptide concen-tration in mol/l, and n is the number of residues perpeptide. This gives units of deg cm2 dmolÿ1 resÿ1 for[y]222.

GuHCl-induced peptide denaturation curves were ®tto folded n-mer (dimer at pH 2.5 or trimer at pH 8.5) tounfolded monomer equilibria (Boice et al., 1996) usingthe linear extrapolation method (Santoro & Bolen, 1988).The observed mean residue ellipticity ([y]obs), composedof contributions from both monomeric and folded n-merspecies, is described by equation (3):

�y�obs � �y�mon�Cmon=Ctot� � n�y�nmer�Cnmer=Ctot� �3�where n corresponds to the aggregation state of thefolded species (2 for dimer, 3 for trimer); Cmon, Cnmer andCtot are the concentrations of monomer, folded n-merand total peptide (10 mM), respectively; [y]mon and [y]nmer

are the mean residue ellipticities of the monomer and n-mer, respectively, and depend linearly on the GuHClconcentration (equations (4) and (5)):

�y�mon � �y�0mon � bmon�GuHCl� �4�

�y�nmer � �y�0nmer � bnmer�GuHCl� �5�[y]0

mon and [y]0nmer are the monomer and n-mer mean resi-

due ellipticities, respectively, in the absence of GuHCl;bmon and bnmer represent the change in ellipticity for thefolded and unfolded peptides with respect to the GuHClconcentration, respectively. Cnmer is expressed in terms ofthe n-mer to monomer equilibrium (equation (6)):

Cnmer � �Cmon�nexp���G0 �mG�GuHCl��=�ÿRT�� �6�

in which �G0 is the free energy change for thedissociation (unfolding) reaction in the absence ofGuHCl, mG is the cosolvation term for the dissociation, Ris the ideal gas constant, and T is the temperature

908 Hg-Binding by De Novo Coiled Coils

(298 K). Cmon is solved numerically for the mass balanceequation Ctot ÿ Cmon ÿ nCnmer � 0 using the programMLAB (Civilized Software, Inc.) and introduced intoequation (3). [y]obs was ®t to the observed GuHCl dena-turation data, with �G0, mG, [y]0

mon, [y]0nmer, bmon and

bnmer being adjustable parameters. Starting values for[y]0

mon, [y]0nmer, bmon and bnmer were obtained from linear

®ts to the folded baseline of the pH 2.5 data set andunfolded baseline of the pH 8.5 data. The ®tting resultsare described in Table 2. Monomer/dimer/trimer equili-bria at pH 5.5 were ®t as described by Betz et al. (1995).To con®rm the results obtained by this treatment, the�G0 values were used to ®t the sedimentation equili-brium data to the appropriate monomer to n-mer equili-bria. The parameters from GuHCl denaturationsprovided an excellent ®t to the experimental data (datanot shown).

UV/visible titrations

UV/visible spectra were collected using a Perkin-Elmer Lambda 9 UV/visible/near-IR dual beam scan-ning spectrophotometer equipped with a Perkin-Elmer3600 data station. A scanning rate of 240 nm/minutewas used in all titrations. Peptides were added fromstock solutions in 4.4 ml aliquots (0.2 equivalent) to twosolutions: the sample, containing 2.2 ml of a 10 mMHgCl2 solution in 20 mM phosphate buffer at the desiredpH; and the reference solution containing buffer with noHg(II). The solutions were mixed and allowed to standfor approximately two minutes before a difference spec-trum was collected for that Hg(II):peptide ratio.

Peptide concentration dependence studies

Two stock solutions were prepared for concentrationdependence studies: stock solution 1 contained 500 mMpeptide and 5.5 M GuHCl in 50 mM acetate buffer at thedesired pH; stock solution 2 contained 20 mM peptidewith the same GuHCl and buffer concentrations.Titrations were performed by mixing stock solutions 1and 2. For peptide concentrations below 20 mM, stocksolution 2 was mixed with 50 mM acetate buffer contain-ing 5.5 M GuHCl. The pH of both stock solutions wasadjusted to the appropriate value. For each sample, 150data points were collected at 25�C and averaged toobtain the signal. [y]222 was calculated as describedabove.

Electrostatic calculations

For electrostatic calculations, the crystal structures ofGCN4-P1 (O'Shea et al., 1991) and CoilVaLd (Ogiharaet al., 1997) were used as templates for dimeric and tri-meric forms of coiled coils. The sequence of Tri wasthreaded onto the two structures, and the Glu and Lysside-chains were protonated to simulate pH 2.5 con-ditions. The models were subjected to in vacuo energyminimization (500 iterations steepest descents followedby 6000 iterations conjugate gradients) using Discoverwith the positions of the backbone atoms of the helices®xed. The resulting dimer and trimer models are theLys-H� structures used in the calculations. The unproto-nated Lys models were generated by deprotonating theLys side-chains without movement of any other atoms.

Using a methodology similar to that of Sharp (1996),electrostatic energies were calculated using the ®nite-difference Poisson-Boltzmann (FDPB) method im-plemented in the software package DelPhi (Gilson et al.,1988; Nicholls et al., 1991; Sharp & Honig, 1990). The fol-lowing parameters were used for the calculations: griddimensions were 65 � 65 � 65, with a scale of approxi-mately one grid per AÊ ; solutions were obtained for thenon-linear Poisson-Boltzmann equation with coulombicboundary conditions (Gilson et al., 1988) using the multi-gridding method of iteration (Holst & Saied, 1993; Sharpet al., 1995); the solvent and protein dielectric constantswere 80 and 2, respectively; the salt concentration was0.20 M. Radii and charge values were obtained from thePARSE parameter set (Sitkoff et al., 1994). Atomiccharges were placed on the atoms of the Lys side-chains(where the formal charges of the peptides are localizedat pH 2.5). For a given charged state of Lys and aggrega-tion state (dimer or trimer), FDPB calculations were per-formed for the aggregate as well as for each individualmonomer mapped to identical grid locations. The elec-trostatic energy for that aggregate and protonation stateof Lys was obtained by taking the difference betweenthe aggregate and monomer energies. The differencebetween the electrostatic energies for the protonated anddeprotonated states of Lys therefore gives the energypenalty for protonating Lys side-chains in each aggrega-tion state.

Acknowledgments

We thank Dr Joel Schneider for many stimulating dis-cussions, Zelda Wasserman for assistance with the pro-tein modeling, Scott Gibson, Tara Beers Gibson and DrJoseph Jez for preliminary modeling studies, Dr JamesBryson and Christopher Summa for critical readings ofthe manuscript, Susan Heilman and Janet Kosinski forassistance with the pH dependence studies, and theSloan Foundation for funding.

References

Arnold, P. A., Shelton, W. R. & Benson, D. R.(1997). Peptide helix induction in a self-assembling hemoprotein model. J. Am. Chem. Soc.119(13), 3181±3182.

Beasley, J. R. & Hecht, M. H. (1997). Protein design: thechoice of de novo sequences. J. Biol. Chem. 272(4),2031±2034.

Benson, D. B., Hart, B. R., Zhu, X. & Doughty, M. B.(1995). Design, synthesis, and circular dichroisminvestigation of a peptide-sandwiched mesoheme.J. Am. Chem. Soc. 117(33), 8502±8510.

Berg, J. (1990). Zinc ®ngers and other metal-bindingdomains. Elements for interactions between macro-molecules. J. Biol. Chem. 265(12), 6513±6516.

Berg, J. M. (1995). Zinc-®nger domains±from predictionsto design. Acc. Chem. Res. 28, 14±19.

Betz, S., Fairman, R., O'Neil, K., Lear, J. & DeGrado, W.(1995). Design of two-stranded and three-strandedcoiled coil peptides. Proc. Roy. Soc. London, 348, 81±88.

Betz, S. F. & DeGrado, W. F. (1996). Controlling top-ology and native-like behavior of de novo-designedpeptides: design and characterization of antiparallel

Hg-Binding by De Novo Coiled Coils 909

four-stranded coiled coils. Biochemistry, 35(21),6955±6962.

Boice, J. A., Dieckmann, G. R., DeGrado, W. F. &Fairman, R. (1996). Thermodynamic analysis of adesigned three-stranded coiled coil. Biochemistry,35(46), 14480±14485.

Bowmaker, G. A., Dance, I. G., Dobson, B. C. & Rogers,D. A. (1984). Synthesis and vibrational spectra ofsome tris(alkanethiolato)mercurate(II) complexes,and crystal structure of the hexakis-(methanethio-lato)dimercurate(II) dianion. Aust. J. Chem. 37,1607±1618.

Bowmaker, G. A., Dance, I. G., Harris, R. K.,Henderson, W., Laban, I., Scudder, M. L. & Oh, S.-W. (1996). Crystallographic, vibrational and nuclearmagnetic resonance spectroscopic characterizationof the [(PhS)2Hg(m-SPh)2Hg(SPh)2]2 ± ion. J. Chem.Soc. Dalton Trans., 2381±2388.

Brady, G. P. & Sharp, K. A. (1997). Entropy in proteinfolding and in protein-protein interactions. Curr.Opin. Struct. Biol. 7, 215±221.

Bryson, J. W., Betz, S. V., Lu, H. S., Suich, D. J., Zhou,H. X., O'Neil, K. T. & DeGrado, W. F. (1995). Pro-tein design, a heirarchic approach. Science, 270,935±941.

Busetto, L., Bordoni, S. & Zanotti, V. (1990). Reac-tions of [(Z-C5H5)(CO)FeC(S)SC(Fp)S] (Fp � (Z-C5H5)Fe(CO)2) with group IB and IIB metals;X-ray molecular structure of [(Z-C5H5)(CO)-FeC(SHgBr2)SC(Fp)S]2. J. Organomet. Chem. 389,341±350.

Cheesman, B. V., Arnold, A. P. & Rabenstein, D. L.(1988). Nuclear magnetic resonance studies of thesolution chemistry of metal complexes. 25.Hg(thiol)3 complexes and Hg(II)-thiol ligandexchange kinetics. J. Am. Chem. Soc. 110, 6359±6364.

Choma, C. T., Lear, J. D., Nelson, M. J., Dutton, P. L.,Robertson, D. E. & DeGrado, W. F. (1994). Designof a heme-binding four-helix bundle. J. Am. Chem.Soc. 116, 856±865.

Cohn, E. J. & Edsall, J. T. (1965). Proteins, Amino Acidsand Peptides as Ions and Dipolar Ions, Hafner Publish-ing Co., New York.

Coldren, C. D., Hellinga, H. W. & Caradonna, J. P.(1997). The rational design and construction of acuboidal iron-sulfur protein. Proc. Natl Acad. Sci.USA, 94, 6635±6640.

Crick, F. H. C. (1953). The Packing of alpha-helices:simple coiled-coils. Acta Crystallog. 6, 689±697.

Dauber-Osguthorpe, P., Roberts, V. A., Osguthorpe, D. J.,Wolff, J., Genest, M. & Hagler, A. T. (1988). Struc-ture and energetics of ligand binding to proteins:Escherichia coli dihydrofolate reductase-trimetho-prim, a drug-receptor system. Proteins: Struct. Funct.Genet. 4(1), 31±47.

D'Auria, G., Maglio, O., Nastri, F., Lombardi, A.,Mazzeo, M., Morelli, G., Paolillo, L., Pedone, C. &Pavone, V. (1997). Hemoprotein models based on acovalent helix-heme-helix sandwich: 2. Structuralcharacterization of Co(III) mimochrome I � and �isomers. J. Eur. Chem, 3(3), 350±362.

Dieckmann, G. R. (1995). Use of metal-binding de novo-designed a-helical peptides in the study of metallo-protein structures, Thesis, University of Michigan.

Dieckmann, G. R., McRorie, D. K., Tierney, D. L.,Utschig, L. M., Singer, C. P., O'Halloran, T. V.,Penner-Hahn, J. E., DeGrado, W. F. & Pecoraro,V. L. (1997). De novo design of mercury-binding

two and three-helical bundles. J. Am. Chem. Soc.119(26), 6195±6196.

Dunbrack, R. L., Jr & Cohen, F. E. (1997). Bayesian stat-istical analysis of protein side-chain rotamer prefer-ences. Protein Sci. 6(8), 1661±1681.

Dunbrack, R. L., Jr & Karplus, M. (1994). Conformation-al analysis of the backbone-dependent rotamer pre-ferences of protein side-chains. Nature Struct. Biol. 1,334±340.

Dunker, K. & Zaleske, D. J. (1977). Stereochemical con-siderations for constructing alpha-helical proteinswith particular application to membrane proteins.Biochem. J. 163, 45±57.

Gibney, B. R., Mulholland, S. E., Rabanal, F. & Dutton,P. L. (1995). Design of synthetic iron-sulfur proteins.In Photosynthesis: From Light to Biosphere (Mathis, P.,ed.), vol. 2, pp. 645±648, Kluwer AcademicPublishers, Dordrecht.

Gibney, B. R., Mulholland, S. E., Rabanal, F. & Dutton,P. L. (1996). Ferredoxin and ferredoxin-hememaquettes. Proc. Natl Acad. Sci. USA, 93, 15041±15046.

Gilson, M., Sharp, K. A. & Honig, B. (1988). Calculatingthe electrostatic potential of molecules in solution:method and error assessment. J. Comput. Chem. 9(4),327±335.

Handel, T. & DeGrado, W. F. (1990). De novo design ofa Zn2�-binding protein. J. Am. Chem. Soc. 112,6710±6711.

Handel, T. M., Williams, S. A. & DeGrado, W. F. (1993).Metal ion-dependent modulation of the dynamicsof a designed protein. Science, 261, 879±885.

Harbury, P. B., Kim, P. S. & Alber, T. (1994). Crystalstructure of an isoleucine-zipper trimer. Nature,371(6492), 80±83.

Hodges, R. S., Saund, A. K., Chong, P. C. S., St.-Pierre,S. A. & Reid, R. E. (1981). Synthetic model for two-stranded a-helical coiled-coils. J. Biol. Chem. 256,1214±1224.

Hodges, R. S., Zhou, N. E., Kay, C. M. & Semchuk, P. D.(1990). Synthetic model proteins: contribution ofhydrophobic residues and disul®de bonds to pro-tein stability. Pept. Res. 3, 123±137.

Hodges, R. S., Zhu, B. Y., Zhou, N. E. & Mant, C. T.(1994). Reverse-phase liquid chromatography as auseful probe of hydrophobic interactions involvedin protein folding and protein stability. J. Chromatog.676(1), 3±15.

Holst, M. & Saied, F. (1993). Multigrid solution of thePoisson-Boltzmann equation. J. Comput. Chem. 14,105±113.

Janin, J. (1996). For Guldberg and Waage, with love andcratic entropy. Proteins: Struct. Funct. Genet. 24, i± ii.

Kohn, W. D., Monera, O. D., Kay, C. M. & Hodges, R. S.(1995). The effects of interhelical electrostatic repul-sions between glutamic acid residues in controllingthe dimerization and stability of two-strandeda-helical coiled-coils. J. Biol. Chem. 270(43), 25495±25506.

Kraulis, P. J. (1991). MOLSCRIPT: a program to produceboth detailed and schematic plots of protein struc-tures. J. Appl. Crystallog. 24, 946±950.

Lau, S. Y. M., Taneja, A. K. & Hodges, R. S. (1984). Tro-pomyosin: a model protein for studying coiled-coiland a-helical stabilization. J. Biol. Chem. 259(21),13253±13261.

Laue, T. M., Shah, B. D., Ridgeway, T. M. & Pelletier,S. L. (1992). Computer-aided interpretation of ana-lytical sedimentation data for proteins. In Analytical

910 Hg-Binding by De Novo Coiled Coils

ultracentrifugation in Biochemistry and Polymer Science(Harding, S. E., Horton, J. C. & Rowe, A. J., eds),pp. 90±125, Royal Society of Chemistry,Cambridge, UK.

Lovejoy, B., Choe, S., Cascio, D., McRorie, D. K.,DeGrado, W. F. & Eisenberg, D. (1993). Crystalstructure of a synthetic triple-stranded alpha-helicalbundle. Science, 259, 1288±1293.

Lowey, S. (1965). Comparative study of the a-helicalmuscle proteins. Tyrosyl titration and effect of pHon conformation. J. Biol. Chem. 240, 2421±2427.

Lu, Y. & Valentine, J. S. (1997). Engineering metal-bind-ing sites in proteins. Curr. Opin. Struct. Biol. 7(4),495±500.

Malmstrom, B. G. (1994). Rack-induced bonding in blue-copper proteins. Eur. J. Biochem. 223(3), 711±718.

McGregor, M. J., Islam, S. A. & Sternberg, M. J. E.(1987). Analysis of the relationship between side-chain conformation and secondary structure inglobular proteins. J. Mol. Biol. 198, 295±310.

Nastri, F., Lombardi, A., Morelli, G., Maglio, O.,D'Auria, G., Pedone, C. & Pavone, V. (1997). Hemo-protein models based on a covalent helix-heme-helix sandwich: 1. Design, synthesis, and character-ization. Chem. Eur. J. 3(3), 340±349.

Nicholls, A., Sharp, K. A. & Honig, B. (1991). Proteinfolding and association: insights from the interfacialand thermodynamic properties of hydrocarbons.Proteins: Struct. Funct. Genet. 11, 281±296.

Noelken, M. & Holtzer, A. (1964). In Biochemistry ofMuscle Contraction (Gergely, J., ed.), pp. 374±378,Little, Brown and Co., Boston, MA.

Ogihara, N., Weiss, M. S., DeGrado, W. F. & Eisenberg,D. (1997). The crystal structure of the designed tri-meric coiled coil CoilVaLd: implications for engin-eering crystals and supramolecular assemblies.Protein Sci. 6, 78±86.

O'Neil, K. T. & DeGrado, W. F. (1990). A thermodyn-amic scale for the helix-forming tendencies of thecommonly occurring amino acids. Science, 250, 646±651.

O'Shea, E. K., Klemm, J. D., Kim, P. S. & Alber, T.(1991). X-ray structure of the GCN4 leucine zipper,a two-stranded, parallel coiled coil. Science,254(5031), 539±544.

O'Shea, E. K., Rutkowski, R. & Kim, P. S. (1992). Mech-anism of speci®city in the Fos-Jun oncoprotein het-erodimer. Cell, 68(4), 699±708.

Pace, C. N. (1986). Determination and analysis of ureaand guanidine hydrochloride denaturation curves.Methods Enzymol. 131, 266±280.

Pinto, A. L., Hellinga, H. W. & Caradonna, J. P. (1997).Construction of a catalytically active iron superox-ide dismutase by rational protein design. Proc. NatlAcad. Sci. USA, 94, 5562±5567.

Ralston, D. M., Frantz, B., Shin, M., Wright, J. G. &O'Halloran, T. V. (1989). In Metal Ion Homeostasis:Molecular Biology and Chemistry (Winge, D. &Hamer, D., eds), vol. 98, pp. 407±416, Liss, NewYork.

Regan, L. (1995). Protein design: novel metal-bindingsites. Trends Biochem. Sci. 20(7), 280±285.

Regan, L. & Clarke, N. D. (1990). A tetrahedral zinc(II)-binding site introduced into a designed protein.Biochemistry, 29, 10878±10883.

Robertson, D. E., Farid, R. S., Moser, C. C., Urbauer,J. L., Mulholland, S. E., Pidikiti, R., Lear, J. D.,Wand, A. J., DeGrado, W. F. & Dutton, P. L. (1994).

Design and synthesis of multi-haem proteins.Nature, 368, 425±432.

Ross, W., Park, S. J. & Summers, A. O. (1989). Geneticanalysis of transcriptional activation and repressionin the Tn21 mer operon. J. Bacteriol. 171(7), 4009±4018.

Santoro, M. M. & Bolen, D. W. (1988). Unfolding freeenergy changes determined by the linear extrapol-ation method. 1. Unfolding of phenylmethanesulfo-nyl a-chymotrypsin using different denaturants.Biochemistry, 27(21), 8063±8068.

Scott, M. P. & Biggins, J. (1997). Introduction of a [4Fe-4S (S-cys)4]

�1, � 2 iron-sulfur center into a four-ahelix protein using design parameters from thedomain of the Fx cluster in the Photosystem I reac-tion center. Protein Sci. 6(2), 340±346.

Sharp, K. & Honig, B. (1990). Electrostatic interactions inmacromolecules: theory and applications. Annu.Rev. Biophys. Biophys. Chem. 19, 301±332.

Sharp, K. A. (1996). Electrostatic interactions in hirudin-thrombin binding. Biophys. Chem. 61, 37±49.

Sharp, K. A., Friedman, R., Misra, V., Hecht, J. & Honig,B. (1995). Salt effects on polyelectrolyte-ligand bind-ing: comparison of Poisson-Boltzmann, and limitinglaw/counterion binding models. Biopolymers, 36,245±262.

Shen, L., Bruccoleri, R. E., Krystek, S. & Novotny, J.(1996). Factors in¯uencing accuracy of computer-built models: a study based on leucine zipperGCN4 structure. Biophys. J. 70, 1096±1104.

Sitkoff, D., Sharp, K. & Honig, B. (1994). Accurate calcu-lation of hydration free energies using macroscopicsolvent models. J. Phys. Chem. 98, 1978±1988.

Stryer, L. (1988). Biochemistry, 3rd edit., p. 1089, W. H.Freeman and Co., New York.

Tamura, A. & Privalov, P. L. (1997). The entropy cost ofprotein association. J. Mol Biol. 273, 1048±1060.

Utschig, L. M., Wright, J. G. & O'Halloran, T. V. (1993).Biochemical and spectroscopic probes of mercury(II)coordination environments in proteins. In Methodsin Enzymology, vol. 226, pp. 71±97, Academic Press,Inc., San Diego.

Utschig, L. M., Bryson, J. W. & O'Halloran, T. V. (1995).Hg-199 NMR of the metal receptor-site in mer Rand its protein-DNA complex. Science, 268(5209),380±385.

Vallee, B. L. & Williams, R. J. P. (1968). The entatic stateof enzyme active sites. Proc. Natl Acad. Sci. USA, 59,498±505.

Walkup, G. K. & Imperiali, B. (1997). Fluorescent che-mosensors for divalent zinc based on zinc ®ngerdomains. Enhanced oxidative stability, metal bind-ing af®nity, and structural and functional character-ization. J. Am. Chem. Soc. 119(15), 3443±3450.

Wang, S. & Fackler, J. P., Jr (1989). Mercury complexeswith one-dimensional chain structures. Synthesisand crystal structures of [Hg(C5H4NS)(CH3CO2)]n,Hg(C5H4NS)2, and Hg(CH2P(S)Ph)2. Inorg. Chem.28(13), 2615±2619.

Watton, S. P., Wright, J. G., MacDonnell, F. M., Bryson,J. W. & O'Halloran, T. V. (1990). Trigonal mercuriccomplex of an aliphatic thiolate: a spectroscopicand structural model for the receptor site in theHg(II) biosensor MerR. J. Am. Chem. Soc. 112, 2824±2826.

Winkler, J. R. & Gray, H. B. (1992). Electron transfer inruthenium-modi®ed proteins. Chem. Rev. 92(3),369±379.

Hg-Binding by De Novo Coiled Coils 911

Wright, J. G., Natan, M. J., MacDonnell, F. M., Ralston,D. M. & O'Halloran, T. V. (1990a). Mercury(II)-thio-late chemistry and the mechanism of the heavymetal biosensor MerR. In Progress in Inorganic Chem-istry (Lippard, S. J., ed.), vol. 38, pp. 323±412, JohnWiley & Sons, Inc., New York.

Wright, J. G., Tsang, H. T., Penner-Hahn, J. E. &O'Halloran, T. V. (1990b). Coordination chemistryof the Hg-MerR metalloregulatory protein: evidencefor a novel tridentate Hg-Cysteine receptor site.J. Am. Chem. Soc. 112, 2434±2435.

Zhou, N. E., Zhu, B. Y., Kay, C. M. & Hodges, R. S.(1992). The two-stranded alpha-helical coiled-coil isan ideal model for studying protein stability andsubunit interactions. Biopolymers, 32(4), 419±426.

Zhou, N. E., Kay, C. M. & Hodges, R. S. (1993). Disul-®de bond contribution to protein stability: pos-itional effects of substitution in the hydrophobiccore of the two-stranded a-helical coiled coil. Bio-chemistry, 32(12), 3178±3187.

Zhou, N. E., Kay, C. M. & Hodges, R. S. (1994). The netenergetic contribution of interhelical electrostatic

attractions to coiled-coil stability. Protein Eng. 7(11),1365±1372.

Edited by P. E. Wright

(Received 26 January 1998; received in revised form21 April 1998; accepted 21 April 1998)

http://www.hbuk.co.uk/jmb

Supplementary material for this paper is availablefrom JMB Online.

912 Hg-Binding by De Novo Coiled Coils