Two-Metal Binding Motifs in Protein Crystal Structures

Structural Chemistry, Vol. 12, Nos. 3/4, 2001

1040-0400/ 01/ 0800-0323$19.50/ 0 2001 Plenum Publishing Corporation

323

Two-Metal Binding Motifs in Protein Crystal Structures1

Jenny P. Glusker,2,4 Amy K. Katz,2 and Charles W. Bock2,3

Received August 29, 2000; revised December 28, 2000; accepted January 12, 2001

The binding of two metal ions that are in close proximity in proteins is examined using acombination of (1) crystallographic structural database analyses and (2) density functional theorycalculations on model complexes. Divalent magnesium and manganese ions are the focus ofthe present study. It is found that in all proteins in the Protein Databank that have two closelypositioned magnesium or manganese ions, these metal ions are generally bridged by at leastone negatively charged oxygen-containing group—carboxylate, phosphate, or sulfate. This grouptransfers (negative) electron density to the metal ions and this helps to reduce electrostatic repulsionin the region. The geometry of the two-metal complex appears to depend on the nature of thenegatively charged group between them. When a single oxygen atom is also in a bridging position,the two metal ions are found to be closer together than when only a carboxylate group binds themtogether. This suggests that this bridging oxygen atom may be negatively charged, e.g., a hydroxideion rather than a water molecule. Details of the geometry of such bridges and the relevant motifsthat are found in crystal structures are described.

KEY WORDS: Two-metal protein sites; protein active sites; density functional theory; crystallographicdatabases.

INTRODUCTION

Proteins, both structural and catalytic, often bindmetal ions in order to maintain a required folding shapeor to aid in the chemistry of their enzymatic mecha-nism. Metal–ligand interactions are geometrically moreprecise and rigid than are hydrogen-bonding interactions[1]. Therefore, the use of metal ions may assure the preciseorientation of a functional group that is required for a par-ticular enzymatic reaction. In addition, the unique chemi-cal properties of individual metal ions may be a highly sig-nificant factor in the catalytic mechanism [2, 3].

When a metalloprotein folds, presumably it firstemploys hydrogen bonds and hydrophobic interactionsto form the overall shape that the protein needs for func-

1Issue in honor of Bill Watson: Structural Chemistry.2 The Institute for Cancer Research, Fox Chase Cancer Center, 7701Burholme Avenue, Philadelphia, Pennsylvania 19111.

3 Department of Chemistry, Philadelphia University, Henry Avenue andSchool House Lane, Philadelphia, Pennsylvania 19144.

4 To whom all correspondence should be addressed. email: jp [email protected]

tion. Certain amino acid residues within this folding mayprovide a cluster of functional groups that constitute areceptive (negatively) charged region for metal ion bind-ing. Sometimes such a binding site is naturally designedto bind a specific metal ion. This is achieved if the natureand arrangement of amino acid residues accommodatesthe unique chemistry of the metal ion. In other cases thesite may be designed to bind a selection of metal ionswith the correct size and charge; this type of binding siteis nonspecific.

A major chemical effect of a metal ion in proteinsis the polarization by it of any functional group to whichit is bound. This means that electronic charge is redis-tributed. For example, an O—H bond in a metal ion-bound water molecule may be weakened when a metalion is bound to it, and this water molecule may yielda hydroxyl ion if neighboring groups can abstract thehydrogen ion that is now only weakly held. Thus a metalion may change the pKa of a water molecule so that ahydroxyl ion may become more readily available for abiochemical reaction.

Glusker, Katz, and Bock324

Fig. 1. Syn, anti and direct (bidentate) nomenclature for metal–car-boxylate complexes and its relationship to the conformers of carboxylicacid from which these complexes are derived.

The most likely ligands for metal ions in proteinsare the carboxylate groups in glutamate and aspartateresidues. These have a single negative charge at nor-mal cellular pH values and, if several such carboxylategroups cluster closely, they can help neutralize the pos-itive charge of one or more metal ions. We have pre-viously studied the relative orientations with which iso-lated carboxylate groups bind various metal ions. Thiswas done by an analysis of high-resolution structuresof small molecules in the crystalline state [4] that arelisted in the Cambridge Structural Database (CSD) [5].This study involved a large number of crystal structurescontaining many of the metal ions that are of biologicalinterest [4, 6]. It was concluded that a metal ion maybind at either of the two sites occupied by lone pairsof electrons on each oxygen atom of the carboxylategroup; these sites are labeled syn and anti (see Fig. 1[7, 8]). In the syn (Z-) form of a carboxylic acid theproton is on the same side of the C—O bond as theC——O bond (Fig. 1a), while in the anti conformation(E-form) it is on the opposite side (Fig. 1b). Gandour[8] has estimated that for a carboxylate group, syn pro-tonation is 104 times more favorable than anti protona-tion. The most common type of carboxylate–metal ionbinding involving divalent transition metal ions was alsofound to be syn (Fig. 1c). Those cations that were mostlikely to bind in the anti conformation are the smallestcations with metal–oxygen distances up to about 1.98

A, and the largest of the cations, such as cesium. Somecations, particularly those of the alkali metal and alka-line earth series, bind out of the carboxylate plane as wellas in it [4]. Other metal ions, which have metal–oxygendistances in the range 2.3–2.6 A may bind both oxygenatoms of the carboxylate ion equally to give a biden-tate or “direct” complex, as shown in Fig. 1d. Cationswith a strong tendency to do this are Cd(II), Tl(III),U(VI), and, to a somewhat lesser extent, Ca(II) [4]. Themetal–oxygen distances, however, are not the sole crite-ria for “direct” bidentate binding, since it is found thatsome cations with metal–oxygen distances in the 2.3–2.6A range bind mainly in the bidentate mode, while othersdo not.

To summarize, our studies have indicated that thereare three categories of metal ion–carboxylate motifs: (1)in-plane interactions of metal ions involving the oxygenatom lone pairs (syn and anti), (2) out-of-plane interac-tions of metal ions, particularly alkali metal ions, and (3)“direct” or bidentate binding, as found with calcium ions,for example. The bidentate motif is found in sites in pro-teins, such as those that contain the EF hand [9] (a motifthat is specific for the binding of calcium ions). Thebinding between a metal ion and a carboxylate group ismuch more geometrically rigid than are hydrogen bonds[12] and, therefore, if there is a need to keep a functionalgroup in a rigid position, a metal ion is an excellent can-didate for this.

Similar results to those for carboxylates were foundwhen the geometries of metal ion–histidine interactionswere investigated [1, 10]. Such interactions in proteinswere examined by Chakrabarti [10], who showed thatcations lie approximately in the imidazole plane, alongthe lone-pair direction of the nitrogen atom, with anaverage deviation of 5.88 from the imidazole plane. Ouranalysis of crystal structures in the CSD containing thisinteraction showed that when metal ions coordinate topyridine or imidazole rings, there is much less out-of-plane binding of a metal ion than of the hydrogen atomof a hydrogen-bonding group [1]. The energetic conse-quences of distorting a metal-histidine interaction out ofthe plane of the histidine residue, estimated by ab ini-tio molecular orbital calculations, support this conclu-sion (see Table I).

Proteins appear to bind metal ions with certain pat-terns, possibly indicative of the function of the proteinor enzyme. We identified a simple motif, which we call“motif I,” in protein crystal structures in the ProteinDatabank (PDB) [11]. It consists of a divalent metal ion

Two-Metal Binding Motifs in Protein Crystal Structures 325

Table I. Energetic Consequences of Distorting an ImidazoleInteraction with a Metal Ion or with a Hydrogen-Bonding Hydrogen

(kcal/ mol)a

a. Energies (kcal/ mol) involved in distorting a bond Mn+ orH· · ·:N from the imidazole ring plane (containing the N:) byab initio molecular orbital calculations

Deviation from plane

Interaction 208 408

Zn2+ · · ·:N 1.45 3.19Mg2+ · · ·:N 1.21 2.78X–H· · ·:N 0.26 1.03

b. Average observed angular deviations from nitrogen-containingheterocyclic rings in crystal structures in the CSD and PDB

Average deviationInteraction (8) Structures examined

Mn+ · · ·:N 3.9 184 Imidazoles (CSD)Mn+ · · ·:N 5.8 47 Proteins (PDB)b

X–H· · ·:N 10.1 57 Entries with hydrogenbonds (CSD)

a Carrell (Ref. [1]).b Chakrabarti (Ref. [10]).

that has bound two oxygen atoms, one in water, the otherin a carboxylate group, and in which the other oxygenatom of the carboxylate group forms a hydrogen bondto the bound water molecule (see Fig. 2a). This motifis also found in many small-molecule crystal structures,e.g., some metal formates [12]. Protein structures in thePDB that contain motif I include enzymes such as cate-chol O-methyltransferase [13], mandelate racemase [14],avian sarcoma virus integrase [15] and the chemotaxisprotein CheY [16]. Since the formation of this motif didnot appear to enhance or reduce the ability of the metalion to polarize a water molecule, it was surmised thatits function is to orient the coordination sphere of thosemetal ions with a rigid coordination polyhedron (suchas magnesium or manganese, not zinc). We named thiseffect “coordination clamping” [17]. For example, mag-nesium ions bind oxygen atoms in functional groups inan approximately octahedral arrangement. If the orien-tation of two oxygen atoms with respect to the magne-sium ion are rigidly constrained (as in the formation ofmotif I), then the orientations of the other four coordi-nating oxygen atoms are rigidly defined. This “coordina-tion clamping” may aid the stereochemical requirementsof the enzymatic mechanism.

There are several metalloenzymes that contain morethan one metal ion in their active site, 3–5 A apart

[18]; these are called “dinuclear metalloenzymes.” Elec-trostatic repulsion prevents the two positively chargedmetal ions in these enzymes from approaching each othertoo closely. In this article we examine some of thesedinuclear sites in protein crystal structures, concentrat-ing primarily on the simple divalent magnesium or man-ganese ions. Magnesium has only one common valencestate and it provided a good starting point for our studiesof dinuclear metalloenzymes. Manganese has a higheratomic number than magnesium and, therefore, it is eas-ier to locate with precision in protein electron-densitymaps. Therefore, both have been studied here. Thereare, however, many other metals that adopt this motif;iron in hemerythrin provides a well-known example.We elected, however, to study magnesium complexesbecause no metal oxidation state change is possible forit; we then proceeded to study manganese.

METHODS

To investigate the assumption that gas-phase calcu-lations might mimic motifs found in proteins, we haveperformed calculations on a variety of model two-metalcomplexes and then have compared the results withstructural data from X-ray diffraction studies of bothsmall and large molecules.

Computational Methods

Density functional theory (DFT) calculations wereperformed at the BP/ DN** computational level withSPARTAN v5.0 on Silicon Graphics computers [19].This level uses the nonlocal Becke-Perdew (BP) 86functional and employs the numerically defined DN**

basis set, which includes polarization functions on allthe atoms [20]. Complete optimizations for a varietyof singly and doubly bridged dimagnesium complexeswere carried out; water molecules, hydroxide ions, andformate ions were used as ligands. No symmetry con-straints were employed in the calculations in order tominimize the likelihood of optimizing to a transitionstate. We performed frequency analyses for a few of thesmaller complexes in order to ensure that the optimizedstructures were local minima on the potential energysurfaces (PESs). For the larger complexes, particularlythose including water and/ or formate ligands, it was notpractical to study all possible conformations. In thesecases, we usually chose an unsymmetric starting geom-


etry for the optimization and considered only a few pos-sible conformers. For many of the complexes we studied,the graphics utilities of SPARTAN [19] were used exten-sively to examine the electron density, the electrostaticpotential, and the Kohn–Sham highest occupied molec-ular orbital (HOMO) and lowest unoccupied molecu-lar orbital (LUMO) [21]. Atomic charges for all thecomplexes were calculated from the electrostatic poten-tials. For comparison, we also performed a few opti-mizations using DFT at the B3LYP/ 6-31+G* computa-tional level [22, 23] and molecular orbital calculationsusing second-order Møller–Plesset perturbation theory[24] at the MP2(FC)/ 6-31+G* computational level usingthe GAUSSIAN 98 series of programs [25].

To simplify the discussion, we will consistently usethe results from the BP/ DN** computational level. Forthose complexes where we also carried out B3LYP/ 6-31+G* and/ or MP2(FC)/ 6-31+G* calculations, the re-sults from these levels were in good agreement withthose from the BP/ DN** level.

Database Analyses

The PDB [11] was searched for protein crystal struc-tures with two or more magnesium or manganese metalions. All structures with a resolution better than 2.2 Awere selected. Each protein structure was examined usingRASMOL [26] to determine which ligands lay within3.5 A of the metal ions. This visual analysis also showedwhether the metal ions were independent of each other orwhether they were bound together by means of a sharedligand. A computer file was prepared containing all rel-evant structures, and the metal ion and its ligands wereexamined using ICRVIEW [27]. Structural diagrams weredrawn and further metal interactions were sought in orderto establish if there were any common binding motifs. Acomparison of analyses of metal binding for all the pro-teins in our list and for those obtained from our DFT cal-culations was made using the scaling and planarity capa-bilities of ICRVIEW [27].

Nomenclature

In order to describe the bridging group betweentwo metal ions, we have used curly brackets “{” and“}”, and, if there are several bridging groups we haveput a comma between each in the list. For example,Mg2+{HCO−

2 , OH−}Mg2+ denotes two divalent magne-sium ions bridged by both a formate ion and an hydro-

Fig. 2. Motifs described in the text: (a) Motifs with one metal ion(motifs I and II); (b) motifs with two metal ions (motifs 1–5); and (c)combinations of motifs (motifs 6–9).

xyl ion. Many of the complexes we studied compu-tationally involve water molecules liganded to magne-sium ions and in some of these complexes a two-water


hydrogen-bonded bridge (M2+ · · ·O—H· · ·O· · ·M2+) wasformed during the optimization. This particular bridgingcomponent is not explicitly denoted in the curly brack-ets. We have also used “aq” for water in order to simplifythe formulae given. The terms “motif I,” “motif II,” etc.,are used to denote motifs that involve only a single metalion (Fig. 2a), whereas those motifs with two metal ionsin close proximity are referred to as “motif 1,” “motif2,” etc. (Fig. 2b and c).

RESULTS

Structural Analyses

A list of protein crystal structures that contain twomanganese or magnesium ions in close proximity isgiven in Table II. The data were obtained by usingRASMOL [26] to examine all protein crystal structuresin the PDB that contain more than one divalent man-ganese or magnesium ion. We also searched for shortmetal–metal distances for symmetry-related metal ionsin different subunits, but none were found. This analy-sis only involved structures with a resolution higher than2.2 A. The resolution of each structure, the metal–metaldistance, and any motifs that we identified are listed inTable II. Interestingly, nearly all of the structures withtwo close metal ions contained motif 2, which involvesa negatively charged carboxylate, sulfate, or phosphategroup spanning the two positively charged metal ions.This suggests that such a negatively charged group isused by proteins to bring two metal ions into closeproximity, and when this occurs with carboxylate, sul-fate or phosphate groups, the metal–metal distance isabout 5.2 A. Apparently two oxygen atoms (presumablyhydrogen-bonded water molecules, motif 3 or motif 9)can also span the two metal ions and, in these cases, themetal–metal distance is also near 5.2 A (see Fig. 2).

In addition to the spanning carboxylate ion, manyenzymes have an oxygen atom between the two metalions (motif 1, in Fig. 2b). In these enzymes the dis-tance between the two metal ions is significantly shorter,3.3–3.6 A. This probably indicates that the source of thebridging oxygen atom is an ion, i.e., a hydroxyl (or evenoxide) ion.

Phosphate groups, with their high negative charge,are ideal for facilitating the approach of positivelycharged metal ions near to each other. Some examplesof the geometries of motifs that have been identifiedin crystal structures in the PDB are shown for magne-

sium and manganese complexes in Fig. 3. These aremotifs that have been identified in Fig. 2 and the metalion–metal ion distances are shown.

Several examples of the dimetallic motifs found inprotein structures in the PDB are illustrated in Fig. 4for manganese enzymes; these motifs are tabulated forboth magnesium and manganese enzymes in Table II,where references to the individual structures are listed.The motifs 2 and 3, which combine to give motif 9,are shown in Fig. 4a (1EO4) in which two carboxylategroups span the two metal ions and two water molecules(hydrogen bounded to each other) form a single bridge;the metal ions are 5.2 A apart. Figure 4b (1AZ9) con-tains a similar grouping that lacks one water molecule sothat it is a combination of motifs 1 and 2 to give motif7. The intermetal ion distance is reduced to 3.3 A andpossibly the “water” peak is actually a hydroxide group.A similar motif is found in Fig. 4c (117E), but the singlebridging oxygen atoms comes from a phosphate group;the distance between metal ions here is 3.7 A. Distancesbetween other pairs of metal ions in this crystal struc-ture are also listed in Table II. The bridging carboxy-late in Fig. 4(d) (in concanavalin A, 1NLS, which con-tains one manganese and one calcium ion) also formsa direct (bidentate) linkage to the calcium ion. Figures4(e) (1ATP) and (f) (1PHK) show two types of interac-tion involving ATP with a two-metal site (motifs 1, 2a,2b, 4, and 5). Magnesium-binding enzymes appear to beparticularly prone to bind phosphate groups.

Density Functional Theory Analyses: SinglyBridged Magnesium Complexes

We initially studied singly bridged dimagnesiumcomplexes of the form Mg2+{X}Mg2+, where X is for-mate, hydroxide, or water (HCO−

2 , OH− , or H2O) byDFT calculations. When a neutral water molecule wasused to form an oxygen bridge between the two magne-sium ions, the structure Mg2+{H2O}Mg2+ (net charge c

+4e) was not found to be bound at any of the compu-tational levels we employed; the structure disintegratedon computation. On the other hand, when an hydroxideion was used to form the oxygen bridge, the complexMg2+{OH−}Mg2+ (net charge c +3e) remained bound(intact) during the calculation (see structure 2 in Fig.5). In this complex the distance between the two mag-nesium ions is quite short, 3.882 A, while the bridg-ing Mg2+ · · ·O distances are 2.000 A; the O—H dis-tance (0.989 A) is 0.01 A longer than the value we

Glusker, K

atz, and Bock

328

Table II. Motifs in Dimanganese- and Dimagnesium-Containing Proteins

PDB Numbera Metal–metalcode Resolution of metals distance (A) Motifsb Ref.

A. Mn-containing enzymes1B8A 1.90 3 3.1 2a, 1, 1, 4 (c)1AZ9 2.00 2 3.3 2a, 2a, 1 (d)1QH3 1.90 2(Zn) 3.3 2a, 1, 1 (e)1RLA 2.10 2 3.4 2a, 1, 1 (f)117E 2.15 4 3.6 2a, 2a, 1 (g)1QMG 1.60 2 3.6 2a, 1 (h)117E 2.15 4 3.7 2a, 1 (g)1C30 2.00 3 3.7 2a, 1, 4 (i)1CDK 2.00 2 3.8 2a, 1, 2b, 4 (j)1PHK 2.20 2 3.9 2a, 2a, 2a, 1, 4 (k)1ATP 2.20 2 4.0 2a, 2a, 4, 4 (l)1A6Q 2.00 2 4.0 2a, 1 (m)1NLS 0.94 2(Mn + Ca) 4.2 (n)8XIA 1.90 2 4.7 2a, 3 (o, p)1EO4 1.90 4 5.2 2a, 2a, 3 (q)117E 2.15 4 5.4 2a (g)1B8A 1.90 3 5.4 2a, 2a (c)117E 2.15 4 6.2 2a (g)1B8A 1.90 3 6.4 2a, 4 (c)

B. Mg-containing enzymes1DAK 1.60 2 3.6 2b, 2b, 1 (r)1IR3 1.90 2 3.6 1, 1, 4 (s)1GSA 2.00 2 3.6 2c, 2b, 1 (t)1IOW 1.90 2 3.6 2b, 2b, 1 (u)1T7P 2.20 3 3.7 2a, 2a, 1, 5 (v)1EBG 2.10 2 4.2 1 (w)2NSY 2.00 2 4.8 2b, 2b, 2b, 4, 4 (x)1QK5 1.60 2 4.8 2b (y)1RVC 2.10 2 5.8 2b (z)1TFR 2.06 2 6.3 3 (aa)

C. Mg–Mn-containing enzymes1AQ2 1,1 2b, 4 (bb)

Tw

o-Metal B

inding Motifs in P

rotein Crystal Structures

329a Per subunit.b See Fig. 2: (1) M — O — M; (2a) M — O — C — O — M; (2b) M — O — P — O — M; (2c) M — O — S — O — M; (3) M — O — O — M; (4)M — O — P — O — P — O — M; (5) M — O — P — O — P — O — P — O — M.

c Schmitt, E.; Moulinier, L., Fujiwara, S.; Imanaka, T.; Thierry, J.-C.; Moras, D. EMBO J. 1998, 17, 5227–5237. (PDB refcode: 1B8A.)d Wilce, M. C. J.; Bond, C. S.; Dixon, N. E.; Freeman, H. C.; Guss, J. M.; Lilley, P. E.; Wilce, J. A. Proc. Natl. Acad. Sci. USA 1998, 95, 3472–3477.(PDB refcode: 1AZ9.)

e Cameron, A. D.; Ridderstrom, M.; Olin, B.; Mannervik, B. Struct. Fold. Des. 1999, 7, 1067–1078. (PDB refcode: 1QH3.)f Kanyo Z. F.; Scolnick, L. R.; Ash, D. E.; Christianson, D. W. Nature (London), 1996, 383, 554–557. (PDB refcode: 1RLA.)g Tuominen, V.; Heikinheimo, P.; Kajander, T.; Torkkel, T.; Hyytia, T.; Kapyla, J.; Lahti, R.; Cooperman, B. S.; Goldman, A. J. Mol. Biol. 1998,

284, 1565–1580. (PDB refcode: 117E.)h Thomazeau, K.; Dumas, R.; Halgand, F.; Forest, E.; Douce, R.; Biou, V. Acta Crystallogr. 2000, D56, 389–397. (PDB refcode: 1QMG.)i Thoden, J. B.; Huang, X.; Raushel, F. M.; Holden, H. M. Biochemistry 1999, 38, 16158–16166. (PDB refcode: 1C30.)j Bossemeyer, D.; Engh, R. A.; Kinzel, V.; Ponstingl, H.; Huber, R. EMBO J. 1993, 12, 849–859. (PDB refcode: 1CDK.)k Owen, D. J.; Noble, M. E. M.; Garman, E. F.; Papageorgiou, A. C.; Johnson, L. N. Structure 1995, 3, 467–482. (PDB refcode: 1PHK.)l Zheng, J.; Trafny, E. A.; Knighton, D. R.; Xuong, N.-H.; Taylor, S. S.; Ten Eyck, L. F.; Sowadski, J. M. Acta Crystallogr. 1993, D49, 362–365.(PDB refcode: 1ATP.)

m Das, A. K.; Helps, N. R.; Cohen, P. T. W.; Barford, D. EMBO J. 1996, 15, 6798–6809. (PDB refcode: 1A6Q.)n Deacon, A.; Gleichmann, T.; Kalb (Gilboa), A. J.; Price, H.; Raftery, J.; Bradbrook, G.; Yariv, J.; Helliwell, J. R. J. Chem. Soc. Faraday Trans.

1997, 93, 4305–4312. (PDB refcode: 1NLS.)o Carrell, H. L.; Glusker, J. P.; Burger, V.; Manfre, F.; Tritsch, D.; Biellmann, J.-F. Proc. Natl. Acad. Sci. USA 1989, 86, 4440–4444. (PDB refcode:

8XIA.)p Whitlow, M.; Howard, A. J.; Finzel, B. C.; Poulos, T. L.; Winborne, E.; Gilliland, G. L. Proteins: Struct. Function Genet. 1991, 9, 153–173. (PDBrefcode: 2XIS.)

q Horton, N. C.; Connolly, B. A.; and Perona, J. J. J. Amer. Chem. Soc. 2000, 122, 3314–3324. (PDB refcode: 1EO4.)r Kack, H.; Gibson, K. J.; Lindqvist, Y.; Schneider, G. Proc. Natl. Acad. Sci. USA 1998, 95, 5495–5500. (PDB refcode: 1DAK.)s Hubbard, S. R. EMBO J. 1997, 16, 5572–5581. (PDB refcode: 1IR3.)t Hara, T.; Kato, H.; Katsube, Y.; Oda, J. Biochemistry 1996, 35, 11967–11974. (PDB refcode: 1GSA.)u Fan, C.; Park, I.-S. Walsh, C. T.; Knox, J. R. Biochemistry 1997, 36, 2531–2538. (PDB refcode: 1IOW.)v Doublie, S.; Tabor, S.; Long, A. M.; Richardson, C. C.; Ellenberger, T. Nature (London), 1998, 391, 251–258. (PDB refcode: 1T7P.)w Wedekind, J. E.; Poyner, R. R.; Reed, G. H.; Rayment, I. Biochemistry 1994, 33, 9333–9342. (PDB refcode: 1EBG.)x Rizzi, M.; Bolognesi, M.; Coda, A. Structure 1998, 6, 1129–1140. (PDB refcode: 2NSY.)y Heroux, A.; White, E. L.; Ross, L. J.; Davis, R. L.; Borhani, D. W. Biochemistry 1999, 38, 14495–14506. (PDB refcode: 1QK5.)z Kostrewa, D.; Winkler, F. K. Biochemistry 1995, 34, 683–696. (PDB refcode: 1RVC.)aa Mueser, T. C.; Nossal, N. G.; Hyde, C. C. Cell 1996 85, 1101–1112. (PDB refcode: 1TFR.)bb Tari, L. W.; Matte, A.; Hughes, G.; Delbaere, L. T. J. Nature Struct. Biol. 1997, 4, 990–994 (PDB refcode: 1AQ2).


Fig. 3. Motifs found in magnesium- and manganese-containing proteincrystal structures in the PDB. The metal–metal ion distance is shownfor each.

found for an isolated hydroxide ion at the same com-putational level, 0.979 A. A frequency analysis con-firms that this structure is a local minimum on the PES.Atomic charges calculated from the electrostatic poten-tial for this complex show that only about 0.08e of neg-ative charge is transferred from the hydroxide ion toeach of the magnesium ions, reducing the net chargeon each magnesium ion to +1.92e. When a formateion is used to form a carboxylate bridge, the complexMg2+{HCO−

2 }Mg2+ (net charge c +3e) (1 in Fig. 5) isalso found to be stable and remain bound on compu-tation. The initial structure of this complex was suchthat the magnesium ions were in a syn orientation, butthey moved to an anti orientation during the optimizationto reduce the electrostatic repulsion between the mag-nesium ions. A frequency analysis confirmed that theanti form of Mg2+{HCO−

2 }Mg2+ is a local minimum onthe PES. Approximately 0.17e of negative charge was

Fig. 4. Examples of enzyme structures in the PDB that contain twoclosely positioned manganese ions. (a) Type I restriction enzymeEcoR/ V (1EO4). In each case the view is directly onto the plane ofthe oxygen atoms of two of the metal-bridging groups (designated withfilled bonds).

found to be transferred from the bridging formate ionin Mg2+{HCO−

2 }Mg2+ to each of the two magnesiumions; this is approximately twice the charge transferredto each of the cations in Mg2+{OH−}Mg2+. It is interest-ing to note that /–OCO is only 122.08 in this complex,which is nearly 98 less than the value of this angle foran isolated formate ion at the same computational level,130.98. The calculated charge on each of the oxygenatoms of the carboxylate group in Mg2+{HCO−

2 }Mg2+

is the same as it is in an isolated formate ion, but thecharges on the carbon and hydrogen atoms are more pos-itive. The distance between the divalent magnesium ionsin Mg2+{HCO−

2 }Mg2+ is 5.930 A, which is more than2.0 A longer than that found in the correspondinghydroxide complex (3.882 A), whereas the Mg2+ · · ·Odistances, 1.874 A, are slightly shorter. We note thatthe electrostatic energy between point charges, q c

+2e, separated by 3.882 A, is nearly 120 kcal/ molhigher than when these charges are farther apart,separated by 5.903 A. Thus the adverse interaction


Fig. 4. (b) proline peptidase (1AZ9) and (c) inorganic phosphatase(117E).

between the two positively charged magnesium ions inMg2+{OH−}Mg2+ (at 3.882 A) is substantially larger

Fig. 4. (d) concanavalin A, Mn, and Ca (INLS) and (e) cAMP-dependent protein kinase (1ATP).

than that in Mg2+{HCO−

2 }Mg2+ (at 5.903 A). In spite ofthis, we found that the energy change for the displace-ment reaction


Fig. 4. (f ) phosphorylase kinase (1PHK).

Mg2+{OH−}Mg2++HCO−

2 r Mg2+{HCO−

2 }Mg2++OH−

(1)

is only +12.2 kcal/ mol.We also optimized one conformer of the singly

bridged 5aq • Mg2+{HCO−

2 }Mg2+ • 5aq complex (netcharge c +3e), in which each magnesium ion hasbeen hydrated with five water molecules (see struc-ture 5 in Fig. 5). The starting geometry of 5aq• Mg2+{HCO−

2 }Mg2+ • 5aq that we employed wasunsymmetric, and in the final optimized structure thetwo bridging Mg2+ · · ·O distances are slightly differ-ent, 2.071 and 2.091 A. These distances are some0.2 A longer than those (1.874 and 1.873 A) wefound in Mg2+{HCO−

2 }Mg2+; in addition, the angle/–OCO for the bridging formate ion in 5 aq •Mg2+{HCO−

2 }Mg2+ • 5aq is calculated to be 3.78 largerthan in Mg2+{HCO−

2 }Mg2+ (which has a value of122.08). These differences in structural parameters areconsistent with a longer distance between the magnesiumions in 5aq • Mg2+{HCO−

2 }Mg2+ • 5aq, 6.456 A, thanin Mg2+{HCO−

2 }Mg2+, 5.930 A. The calculated atomiccharges from the electrostatic potential show that the for-mate ion has lost a total of 0.36e of negative charge tothe two Mg2+ • 5aq entities. The charges on the twomagnesium ions are only +1.41e and +1.34e, however,showing that considerable electron density has also beentransferred from the water ligands to the magnesiumcations.

Density Functional Theory Analyses: DoublyBridged Magnesium Complexes

We next studied the doubly bridged magnesiumcomplexes Mg2+{X, Y}Mg2+, where X, Y c HCO−

2 ,OH− , and/ or H2O. Our attempts to find a complex of theform Mg2+{X, H2O}Mg2+ (net charge c +3e), in whicha single water molecule is one of the two bridging units,consistently resulted in singly bridged complexes of theform Mg2+{X}Mg2+ • aq, i.e., the water molecule hadmigrated to one of the magnesium ions during the opti-mization process (see structures 3 and 4 in Figure 5). Thedistance between the magnesium ions in each of thesemonohydrated complexes is similar to the distance inthe corresponding complexes without an additional waterligand. As might be expected for Mg2+{HCO−

2 }Mg2+ •aq, the formate ions eventually end up in an anti ori-entation. The formate ion in this complex loses 0.42eof negative charge and the water molecule loses 0.17e,leaving the two magnesium ions with rather different netcharges, +1.83e and +1.58e (the latter with the additionalwater bound to it). Thus, a single water bridge could notbe adequately modeled at this computational level.

On the other hand, when additional charged specieswere introduced, it was possible to obtain stable models.The complex Mg2+{HCO−

2 , HCO−

2 }Mg2+ (net charge c

+2e) with two carboxylate bridges is shown as struc-ture (9) in Fig. 6. A frequency analysis confirms thatthis complex is a local minimum on the PES. Themagnesium ions in this conformer are essentially in asyn orientation with respect to each of the carboxy-late groups (the average of the angles < COMg is147.38) and the complex is nearly planar. The distancebetween the magnesium ions is 4.105 A, more than 1.8 Ashorter than in the singly bridged anti conformer ofMg2+{HCO−

2 }Mg2+. The bridging Mg2+ · · ·O distancesare not all the same, but are centrosymmetrically related;the average M2+ · · ·O distance is 1.880 A. Each formateion in Mg2+{HCO−

2 , HCO−

2 }Mg2+ has lost a total ofabout 0.27e of negative charge, leading to a net chargeof +1.73e on each magnesium ion.

The complex Mg2+{OH− , OH−}Mg2+ (net charge c

+2e) with two oxygen bridges is shown as structure (11)in Fig. 6. A frequency analysis indicates that this com-plex also is a local minimum on the PES. This conformerof Mg2+{OH− , OH−}Mg2+ is essentially planar and itis quite compact. For example, the distance between themagnesium ions is only 2.871 A, more than 1.0 A shorterthan in Mg2+{OH−}Mg2+, and the bridging Mg2+ · · ·Odistances are also shorter by about 0.08 A. Further-


Fig. 5. Singly bridged dimagnesium complexes Mg2+{X}Mg2+, Mg2+{X}Mg2+ • aq, 4aq • Mg2+{X}Mg2+

• 5aq and 5aq • Mg2+{X}Mg2+ • 5aq (X c HCO−

2 and OH− ). Results of density functional theory calcula-tions. Distances were calculated at the BP/ DN** computational level and the atomic charges were obtainedfrom the electrostatic potential. Calculated Mg2+ charges are given. The total charge of each complex isindicated at the lower right of each diagram.


Fig. 6. Doubly bridged dimagnesium complexes Mg2+{X, Y}Mg2+, 4aq • Mg2+{X, Y}Mg2+ • 4aq and 3aq • HCO−

2 • Mg2+{X, Y}Mg2+ • HCO−

2• 3aq (X, Y c HCO−

2 and OH− ). Results of density functional theory calculations. Distances were calculated at the BP/ DN** computational leveland the atomic charges were obtained from the electrostatic potential.

more, the /–MgOMg angle in this complex is only 96.88,some 55.38 smaller than in the singly bridged complexMg2+{OH−}Mg2+. Each hydroxide unit has transferreda total of 0.20e of negative charge to the cations so thateach magnesium ion carries a charge of +1.80e. In themixed complex Mg2+{HCO−

2 , OH−}Mg2+ (net charge c

+2e) (see structure (10) in Fig. 6), with one carboxylateand one oxygen bridge, the distance between the mag-nesium ions is 3.277 A. This distance is about 0.2 Ashorter than the average of the magnesium–magnesiumdistances in Mg2+{HCO−

2 , HCO−

2 }Mg2+ and Mg2+{OH− ,OH−}Mg2+. In this conformer, the magnesium ions are


essentially in a syn orientation with respect to the car-boxylate group. The hydroxide ion gives up a total of0.13e of negative charge, but the formate ion gives upconsiderably more, 0.35e; the charges on the magnesiumions are nearly the same, +1.75e and +1.76e. Interest-ingly, the energy change for the displacement reaction

Mg2+{OH− , OH−}Mg2+ + HCO−

2

r Mg2+{OH− , HCO−

2 }Mg2+ + OH− (2)

which gives this mixed complex, is +48.4 kcal/ mol.This reaction energy is nearly four times more endother-mic than the energy change for the corresponding singlybridged reaction (1) and may indicate that certain mixedligands are disfavored.

We also investigated the structures of the double-bridged complexes 4aq • Mg2+{X, Y}Mg2+ • 4aq, X,Y c HCO−

2 and OH− (net charge c +2e) in which eachof the magnesium ions has been hydrated with 4 watermolecules. These complexes are shown as structures(12–14) in Fig. 6. The water ligands increase the dis-tance between the magnesium ions in these structuresby 0.334, 0.237, and 0.157 A for the {HCO−

2 , HCO−

2 },{HCO−

2 , OH−}, and {OH− , OH−} bridging combina-tions, respectively, compared to the corresponding struc-tures without the water ligands; the Mg2+ · · ·O bridge dis-tances are also found to be somewhat larger. Interest-ingly, the optimized structure of each of these hydratedcomplexes has a two-water hydrogen-bonded bridge inaddition to the oxygen and/ or carboxylate bridges; thistwo-water bridge was not present in the starting geom-etry. This type of bridge has been found in protein crystalstructures such as that of the type I restriction enzymeEcoRV shown in Fig. 4a. This additional bridge pro-vides an interesting asymmetry to these complexes thatappears to make the surface of the hydroxide and/ or for-mate groups on the side opposite this bridge more acces-sible. For example, in the complex 4aq • Mg2+{HCO−

2 ,HCO−

2 }Mg2+ • 4aq, which is shown in more detail inFig. 7i, the distance between the oxygen atoms of the twowater molecules involved in the hydrogen-bonded bridgeis quite short—only 3.114 A. The distance between themagnesium ions is larger (4.440 A), while the distancebetween the oxygen atoms of the water molecules oppo-site the bridge is even larger (5.974 A). It is also evidentfrom this figure that in order to make room for the two-water bridge, the Mg2+{HCO−

2 , HCO−

2 }Mg2+ unit in 4aq• Mg2+{HCO−

2 , HCO−

2 }Mg2+ • 4aq is no longer nearlyplanar, as it was without the additional water ligands.

Fig. 7. BP/ DN** optimized structures of (i) and (ii) 4aq • Mg2+

{HCO−

2 , HCO−

2 }Mg2+ • 4aq. The view for (i), (iii), (v), and (vii) is ontothe plane of the two metal ions and the two bridging water molecules.The view for (ii), (iv), (vi), and (viii) is onto the plane of the metal ion-bound oxygen atoms of two bridging groups (represented with filledbonds). Phosphate groups are also illustrated with filled bonds.

In this complex the two formate ions lose only 0.31eand 0.26e of negative charge, similar to what is lost bythe formate ions in the corresponding complex withoutthe water molecules. On the other hand, the net charges


Fig. 7. (iii) and (iv) 4aq • Mg2+{HCO−

2 , OH−}Mg2+ • 4aq.

on the two magnesium ions in 4aq • Mg2+{HCO−

2 ,HCO−

2 }Mg2+ • 4aq, +1.39e and +1.51e, are much lesspositive than in Mg2+{HCO−

2 , HCO−

2 }Mg2+, showingthat charge density has also been transferred from thewater ligands to the magnesium ions. In the related com-plex 4aq • Mg2+{OH− , OH−}Mg2+ • 4aq, the two hydro-xide groups have been displaced somewhat and the cen-tral Mg2+{OH− , OH−}Mg2+ unit is no longer planar as itwas without the water ligands. The hydroxide ions lose0.34e and 0.31e of negative charge, but the net charge

Fig. 7. (v) and (vi) 3aq • HCO−

2 • Mg2+{HCO−

2 , HCO−

2 }Mg2+ •HCO−

2 • 3aq (two carboxylate bridges).

on the magnesium ions are +1.54e and +1.28e. Thus, themagnesium ions have gained some electron density fromthe surrounding water ligands; a typical water molecule(not involved in the hydrogen-bonded bridge) has a netelectrostatic charge of about +0.07e. The mixed-bridgecomplex 4aq • Mg2+{HCO−

2 , OH−}Mg2+ • 4aq is shownin Fig. 7iii and iv. Comparing the structures in Fig. 7i and7iii it is evident that they are quite similar, but that 4aq •Mg2+{HCO−

2 , OH−}Mg2+ • 4aq in Fig. 7iii is more com-


Fig. 7. (vii) and (viii) 3aq • HCO−

2 • Mg2+{HCO−

2 , HCO−

2 }Mg2+ •HCO−

2 • 3aq (one carboxylate and one oxygen bridge).

pact than 4aq • Mg2+{HCO−

2 , HCO−

2 }Mg2+ • 4aq in Fig.7i. We note that the energy change for the displacementreaction

4aq • Mg2+{OH− , OH−}Mg2+ • 4aq + HCO−

2

r 4aq • Mg2+{OH− , HCO−

2 }Mg2+ • 4aq + OH−

(3)

is 48.5 kcal/ mol. This reaction energy is nearly the same

as that for reaction (2) above, suggesting that the waterligands have relatively little effect on the thermodynam-ics of this type of displacement process.

We also attempted to find a doubly bridged com-plex of the form 4aq • Mg2+{HCO−

2 , H2O}Mg2+ • 4aqcontaining a carboxylate and a water oxygen bridge; ourinitial structure of this complex also included a two-water hydrogen-bonded bridge. During the course of theoptimization, however, the structure reverted to a singlybridged complex of the form 4aq • Mg2+{HCO−

2 }Mg2+

• 5aq with the magnesium ions in an anti orientation.The distance between the magnesium ions was found tobe 6.246 A, about 0.2 A shorter than that found in 5aq• Mg2+{HCO−

2 }Mg2+ • 5aq (6.456 A). A more compre-hensive search of the PES will be needed to ascertainwhether or not a doubly bridged structure of the form4aq • Mg2+{HCO−

2 , H2O}Mg2+ • 4aq is a local mini-mum.

The optimized dimagnesium complexes that wehave described so far are charged species. We also stud-ied one conformer of each of the doubly bridged com-plexes 3aq • HCO−

2 • Mg2+{X, Y}Mg2+ • 3aq • HCO−

2 ,X, Y c HCO−

2 , and/ or OH− , which are net neutral. Thesecomplexes are shown as structures (15–17) in Fig. 6.In the optimized structures of the specific conformerswe investigated, each formate ion that is not bridgingthe two magnesium ions is hydrogen bonded to one ofthe liganding water molecules; these hydrogen bondswere not present in the initial geometry we employed.All of these complexes also have the two-water hydro-gen bonded bridge we observed previously; for thesecomplexes our starting geometry for the optimizationsincluded this bridging structure.

Considering first the complex 3aq • HCO−

2 •Mg2+{HCO−

2 , HCO−

2 }Mg2+ • HCO−

2 • 3aq, which isshown in Fig. 7v, we note that the distance between themagnesium ions is about 0.1 A smaller than in 4aq •Mg2+{HCO−

2 , HCO−

2 }Mg2+ • 4aq and the H—O· · ·Hdistance in the two-water bridge is approximately 0.2 Asmaller. The bridging carboxylates in this structure aretwisted to a greater extent relative to each other than weobserved in our other doubly bridged carboxylate struc-tures (cf. Fig. 7i and v). The average of the four Mg· · ·Obridging distances in 3aq • HCO−

2 • Mg2+{HCO−

2 ,HCO−

2 }Mg2+ • HCO−

2 • 3aq, 2.041 A, is about 0.04 Alarger than the average Mg· · ·O distances in 4aq •Mg2+{HCO−

2 , HCO−

2 }Mg2+ • 4aq Despite these struc-tural differences, the net charges on the magnesium ionsin these two complexes are quite similar (see Fig. 6).For comparison, we studied a second form of 3aq •


HCO−

3 • Mg2+{HCO−

2 , HCO−

2 }Mg2+ • HCO−

2 • 3aq inwhich one of the bridging formate ions acts as an oxy-gen bridge, while the other acts as a carboxylate bridge(see Fig. 7iv). Interestingly, this conformer is about 0.6kcal/ mol lower in energy than the conformer in whichboth formate ions act as carboxylate bridges. The dis-tance between the magnesium ions, 3.846 A, is about0.5 A shorter than in the two-carboxylate form, and theelectrostatic charges on these magnesium ions are +1.61eand +1.41e. As can be seen from Fig. 7vii, all the watermolecules are involved in hydrogen bonds with otherwater molecules or formate ions; in the other conformerof 3aq • HCO−

2 • Mg2+{HCO−

2 , HCO−

2 }Mg2+ • HCO−

2• 3 aq, shown in Fig. 7v, two water molecules are notinvolved in hydrogen bonding.

Turning next to the complex 3aq • HCO−

2 •Mg2+{OH− , OH−}Mg2+ • 3aq • HCO−

2 , we note thatthe distance between the magnesium ions is 3.022 A,which is nearly the same as it is in 4aq • Mg2+{OH− ,OH−}Mg2+ • 4aq (see structure 17 in Fig. 6). The netcharges on the magnesium ions are the smallest weobserved in this study, +1.25e and +1.26e. Since thetwo hydroxide ions lose a total of only 0.78e of neg-ative charge, it is clear that charge density has also beentransferred from the carboxylate and water ligands to themagnesium ions. In the mixed complex 3aq • HCO−

2 •Mg2+{HCO−

2 , OH−}Mg2+ • 3aq • HCO−

2 , the distancebetween the magnesium ions is 3.568 A, which is about0.1 A below the average for the corresponding com-plexes with two carboxylate bridges, {HCO−

2 , HCO−

2 },(4.338 A), and two oxygen bridges, {OH− , OH−} (2.871A). The hydroxide ion loses 0.21e of negative chargewhile the bridging formate ion loses only 0.18e of nega-tive charge. In this complex the two nonbridging formateions lose more charge than the bridging formate ion. Thecharges on the magnesium ions are +1.59e and +1.50e.The energy change for the displacement reaction

3aq • HCO−

2 • Mg2+{OH− , OH−}Mg2+ • 3aq

• HCO−

2 + HCO−

2

r 3aq • HCO−

2 • Mg2+{OH− , HCO−

2 }Mg2+

• 3aq • HCO−

2 + OH− (4)

that forms the mixed complex is +66.1 kcal/ mol. Thisreaction is some 20 kcal/ mol higher than that of the cor-responding reactions, where each magnesium ion is lig-ated to four neutral water molecules.

Finally, we studied one form of the doubly bridged

complex 3aq • HCO−

2 • Mg2+{HCO−

2 , H2O}Mg2+

• HCO−

2 • 3aq (net charge c +1e) with the for-mate ion acting as a carboxylate bridge and thewater molecule acting as an oxygen bridge; the ini-tial geometry of this complex also included a two-water hydrogen-bonded bridge. Although our attemptsat finding doubly bridged complexes of the formMg2+{HCO−

2 , H2O}Mg2+, Mg2+{OH− , H2O}Mg2+, or4aq • Mg2+{HCO−

2 , H2O}Mg2+ • 4aq all resulted insingle-bridged complexes with the water molecule thatwas initially in a bridging position migrating to one ofthe magnesium ions during the optimization, the bridg-ing water molecule in 3aq • HCO−

2 • Mg2+{HCO−

2 ,H2O}Mg2+ • HCO−

2 • 3aq definitely remained betweenthe two magnesium ions (see Fig. 8); the two-waterbridge also remained intact. Interestingly, in the opti-mized geometry of this complex, one of the originalnonbridging formate ions has become formic acid byabstracting a proton from one of the water molecules.It should be noted that the starting geometry for theoptimization had each of the nonbridging formate ionsforming a hydrogen bond with a water molecule, sim-ilar to those shown in Fig. 7iii. Additional calculationswill be required to see if it is possible to find a con-former of 3aq • HCO−

2 • Mg2+{HCO−

2 , H2O}Mg2+ •HCO−

2 • 3aq in which both hydrogen bonds are betweenformate ions and water, and to determine the energet-ics of proton transfers in this kind of environment. Thedistance between the magnesium ions, 4.090 A (in Fig.8), is relatively long for a doubly bridged complex (seeFig. 6). The bridging Mg2+ · · ·O distances involving thewater molecule are 2.459 and 2.261 A, significantlylonger than the bridging Mg2+ · · ·O distances involvingthe formate ion, 2.026 and 2.001 A. The bridging watermolecule loses 0.14e of negative charge and the bridgingformate ion loses 0.34e of negative charge.

DISCUSSION

In this study, we have investigated the manner inwhich two positively charged divalent magnesium ormanganese ions are brought into proximity in a proteincrystal structure. Our structural database analysis, whichconcentrated mainly on proteins, showed that when twomagnesium or two manganese ions are found to be closein proteins, they are bridged by at least one carboxy-late group; this effectively reduces the overall charge inthe vicinity of the two metals. When the bridging groupis a single carboxylate group, the distance between thetwo metal ions is about 5.2 A. This value is not signifi-


Fig. 8. BP/ DN** optimized structure of 3aq • HCO−

2 • Mg2+{HCO−

2 ,H2O}Mg2+ • 2aq • OH− • HCOOH. This view is onto the plane ofthe oxygen atoms of the bridging formate ion and water molecule. Itcan be compared with similar diagrams in Fig. 7. Note the ionizationstate of the uppermost formate ion and its adjacent hydroxyl group.

cantly changed when other carboxylate groups also spanthe two cations. If, however, as is often found, a singleoxygen atom (from an hydroxide group or one oxygenatom from a carboxylate or phosphate group) bridges thetwo magnesium or manganese ions, the distance betweenthem is shortened to 3.3–3.6 A, a decrease of 1.6–1.9 A.

The structural data from the PDB have been cate-gorized by the identification of certain structural motifs(see Fig. 2) in each protein that contained two diva-lent magnesium or manganese ions in close proximity. Inthe most commonly found motif, motif 2, a carboxylategroup bridges the two metals. Another common motif ismotif 1 in which the metal ions are bridged by just oneoxygen atom. Often in enzymes this oxygen atom is animportant part of the catalytic mechanism. Other morecomplicated motifs, shown in Fig. 2c are common, par-ticularly motifs 6, 7, 8, and 9. Motif 7 provides a method

for affecting one oxygen atom since charge in its vicin-ity will be somewhat neutralized if the oxygen atom hasor acquires a negative charge.

Here, we also used DFT calculations to inves-tigate a variety of isolated dimagnesium complexes.When there is only one bridging carboxylate group, wefind that the two magnesium ions are in anti orien-tations with respect to the carboxylate group and thatthe distance between the magnesium ions is relativelylong, 5.9–6.5 A (see Fig. 5). When there is a sin-gle bridging hydroxide group in place of the carboxy-late group, the distance between the magnesium ions isreduced by about 65%. For doubly bridged complexes,when both groups are carboxylate bridges, the magne-sium ions were found to be effectively in syn orien-tations and the distance between the magnesium ionswas in the range 4.1–4.4 A (see Fig. 6). In the cor-responding doubly bridged hydroxide ion complexes,the distance between the magnesium ions is about69% shorter. For comparison, we note that the calcu-lated distance between the magnesium ions in the com-plexes Mg2+{HCO−

2 }Mg2+, Mg2+{HCO−

2 , HCO−}Mg2+,Mg2+{HCO−

2 , HCO−

2 , HCO−

2 }Mg2+ and Mg2+{HCO−

2 ,HCO−

2 , HCO−

2 , HCO−

2 }Mg2+ is 5.903, 4.105, 3.316, and2.819 A, respectively. Atomic charges from electrostaticpotential calculations suggest that more electron densityis usually transferred to the magnesium ions by a car-boxylate than by a hydroxide bridge. Nonbridging wateror formate ligands also transfer electron density to themagnesium ions.

A scatterplot of the locations of the magnesiumions in the structures we studied using DFT calcula-tions is shown in Fig. 9a. This represents the arrange-ment around the carboxylate group in ten of the struc-tures that were investigated. The majority of the mag-nesium ions are located near, but not at, the syn site;if these divalent metal ions were precisely in the synorientation, they would be only about 2.2 A apart, i.e.,at the calculated O· · ·O distance for an isolated formateion; it would be difficult to overcome the large elec-trostatic repulsion associated with two magnesium ionsat such a short distance. Bimetallic sites around a car-boxylate group (aspartate or glutamate) in proteins con-taining divalent manganese are illustrated as a scatter-plot in Fig. 9b. Here there appear to be two main typesof interactions. One involves the major scatter locationin Fig. 9a, while the other involves a bidentate metalion–carboxylate interaction and an anti location for thesecond metal ion (Fig. 9c). The arrangement for a bridg-ing phosphate group is shown in Fig. 9d.


Fig. 9. Scatterplots of pairs of metal ions around carboxylate groups.(a) Density functional theory (DFT) metal ion scatterplot; (b) ProteinDatabank (PDB) protein-2Mn2+ ion scatterplots; (c) schematic dia-grams of the results in (b); and (d) PDB protein-2Mn2+-phosphateinteractions.

CONCLUSION

Our analysis has revealed a consistency betweenthe theoretical calculations on model complexes and theexperimental results found by X-ray diffraction studiesof crystalline metal ion-containing proteins. This impliesthat motifs that we identify in protein structures are theresult of electrostatic interactions between groups withthe same characteristics as seen in our model complexes.Since the geometries are so similar in both cases, we alsoassume that our conclusions about charge distributionsare valid. We have only examined two-metal ion systems

Fig. 9. Continued.

so far, but plan to extend this to models of complexesthat contain more metal ions.

Two metal ions, each with double positive chargesand only 3–6 A apart, bring a large amount of positivecharge together in one place in a protein. The strongtendency of the two metal ions to repel each other canonly be counteracted by the introduction of a nega-tively charged ligand that will then hold the two metalions together. In the examples we have described here,this negatively charged group is generally a carboxylate,phosphate or sulfate group.

Certain metal ions have specific coordinationgeometries. For octahedral complexes, which are com-mon for divalent cations, if two Mn+ · · ·O directions at908 to each other are fixed as a result of binding inthe complex, the other four will also be defined, as ifa clamp had been introduced by the first two interac-tions. This is what we called “coordination clamping.”Examples are given in Fig. 7 where the geometry of thetwo-water bridge distorts the rest of the metal coordi-nation sphere: the hydrogen-bonded O· · ·O distance isabout 3 A, while the O· · ·O distance on the other sideof the carboxylate bridge becomes 5–6 A. Both mag-nesium and divalent manganese ions have a strong ten-


dency to form fairly symmetrical octahedral complexes.Some other metal ions, such as zinc, are less likely to dothis.

The result of such complex formation is a redis-tribution of electronic charge; negative charge is with-drawn from the ligand to the metal ion, making the metalion less positively charged, thus partially neutralizing theimpact of the high positive charge originally present.On the other hand, this means that the ligand has lostsome of its negative charge. Presumably the effect oftwo metal ions that share a water molecule is to facil-itate the formation of a negatively charged species, thatis, a hydroxide ion.

We have described here a preliminary analysis oftwo nearby magnesium or divalent manganese ions inproteins. We are now proceeding to examine similareffects with other divalent metal ions.

ACKNOWLEDGMENTS

This work was supported by grants CA-10925 andCA-06927 from the National Institutes of Health and byan appropriation from the Commonwealth of Pennsyl-vania. Its contents do not reflect the official views ofthe National Cancer Institute. We thank Carol Afshar fortechnical assistance and Dave Zacharias for helpful sug-gestions.

REFERENCES

1. Carrell, A. B.; Shimoni, L.; Carrell, C. J.; Bock, C. W.; Murray-Rust, P.; Glusker, J. P. Receptor 1993, 3, 57–76.

2. Mildvan, A. S. In The Enzymes, Boyer, P. D., Ed.; Academic Press:New York, 1970; vol. 2, pp. 446–536.

3. Frausto da Silva, J. J. R.; Williams, R. J. P. The Biological Chem-istry of the Elements. The Inorganic Chemistry of Life; ClarendonPress: Oxford, 1991.

4. Carrell, C. J.; Carrell, H. L.; Erlebacher, J.; Glusker, J. P. J. Amer.Chem. Soc. 1988, 110, 8651–8656.

5. Allen, F. H.; Bellard, S.; Brice, M. D.; Cartwright, B. A.; Dou-bleday, A.; Higgs, H.; Hummelink, T.; Hummelink-Peters, B. G.;Kennard, O.; Motherwell, W. D. S.; Rodgers, J. R.; Watson, D. G.Acta Crystallogr. 1979, B35, 2331–2339.

6. Rosenfield, R. E., Jr.; Swanson, S. M.; Meyer, E. F., Jr.; Carrell,H. L.; Murray-Rust, P. J. Mol. Graphics 1984, 2, 43–46.

7. Rebek, J., Jr.; Duff, R. J.; Gordon, W. E.; Parris, K. J. Amer. Chem.Soc. 1979, 101, 1076–1079.

8. Gandour, R. Bioorg. Chem. 1981, 10, 169–176.9. Kretsinger, R. H.; Nockolds, C. E. J. Biol. Chem. 1973, 248,

3313–3326.10. Chakrabarti, P. Prot. Eng. 1990, 4, 57–63.11. Bernstein, F. C.; Koetzle, T. F.; Williams, G. J. B.; Meyer, E. F.,

Jr.; Brice, M. D.; Rodgers, J. R.; Kennard, O.; Shimanouchi, T.;Tasumi, M. J. Mol. Biol. 1977, 112, 535–542.

12. Kaufman, A.; Afshar, C.; Rossi, M.; Zacharias, D. E.; Glusker, J.P. Structural Chem. 1993, 4, 191–198.

13. Vidgren, J.; Svensson, L. A.; Liljas, A. Nature (London) 1994,368, 354–358. (PDB refcode: 1 VID)

14. Landro, J. A.; Gerlt, J. A.; Kozarich, J. W.; Koo, C. W.; Shah,V. J.; Kenyon, G. L.; Neidhart, D. J.; Fujita, S.; Petsko, G. A.Biochemistry 1994, 33, 635–643. (PDB code: 1MDR)

15. Bujacz, G.; Jaskolski, M.; Alexandratos, J.; Wlodawer, A.; Merkel,G.; Katz, R. A.; Skalka, A. M. Structure 1996, 4, 89–96.

16. Bellsolell, L.; Prieto, J.; Serrano, L.; Coll, M. J. Mol. Biol. 1994,238, 489–495.

17. Bock, C. W.; Katz, A. K.; Markham, G. D.; Glusker, J. P. J. Amer.Chem. Soc. 1999, 121, 7360–7372.

18. Strater, N.; Lipscomb, W. N.; Klabunde, T.; Krebs, B. Angew.Chem. Intern. Ed. Engl. 1996, 35, 2024–2055.

19. SPARTAN V.5.0, Wavefunction, Inc. 18401 Von Karman Avenue,Suite 370, Irvine, CA 92612.

20. Perdew, J. P. Phys. Rev. 1986, B33, 8822–8824; Erratum 1986,B34, 7406.

21. Kohn, W.; Sham, L. J. Phys. Rev. 1965, 140, A1133-A1138.22. Becke, A. D. J. Chem. Phys. 1993, 98, 5648–5652.23. Gill, P. M. W.; Johnson, B. G.; Pople, J. A.; Frisch, M. J. Chem.

Phys. Lett. 1992, 197, 499–505.24. Møllet, C.; Plesset, M. S. Phys. Rev. 1934, 46, 618–622.25. Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.;

Robb, M. A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery,J. A.; Stratmann, R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.;Daniels, A. D.; Kudin, K. N.; Strain, M. C.; Farkas, O.; Tomasi,J.; Barone, V.; Cossi, M.; Cammi, R.; Mennucci, B.; Pomelli, C.;Adamo, C.; Clifford, S.; Ochterski, J.; Petersson, G. A.; Ayala,P. Y.; Cui, Q.; Morokuma, K.; Malick, D. K.; Rabuck, A. D.;Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz, J. V.;Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; Komaromi, I.;Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M.A.; Peng, C. Y.; Nanayakkara, A.; Gonzalez, C.; Challacombe, M.;Gill, P. M. W.; Johnson, B. G.; Chen, W.; Wong, M. W.; Andres,J. L.; Head-Gordon, M.; Replogle, E. S.; Pople, J. A. GAUSSIAN98, Gaussian, Inc., Pittsburgh, PA, 1998.

26. Sayle, R.; Milner-White, E. J. Trends Biochem. Sci. 1995, 20,374–376.

27. Erlebacher, J.; Carrell, H. L. ICRVIEW, program from the Institutefor Cancer Research, Philadelphia, PA, 1992.

Documents

Two-Metal Binding Motifs in Protein Crystal Structures