Sequential structural changes upon zinc and calcium binding to metal-free concanavalin A.

The lectin concanavalin A (ConA) sequentially binds a transition metal ion in the metal-binding site S1 and a calcium ion in the metal-binding site S2 to form its saccharide-binding site. Metal-free ConA crystals soaked with either Zn2+ (apoZn-ConA) or Co2+ (apoCo-ConA) display partial binding of these ions in the proto-transition metal-binding site, but no further conformational changes are observed. These structures can represent the very first step in going from metal-free ConA toward the holoprotein. In the co-crystals of metal-free ConA with Zn2+ (Zn-ConA), the zinc ion can fully occupy the S1 site. The positions of the carboxylate ligands Asp10 and Asp19 that bridge the S1 and S2 sites are affected. The ligation to Zn2+ orients Asp10 optimally for calcium ligation and stabilizes Asp19 by a hydrogen bond to one of its water ligands. The neutralizing and stabilizing effect of the binding of Zn2+ in S1 is necessary to allow for subsequent Ca2+ binding in the S2 site. However, the S2 site of monometallized ConA is still disrupted. The co-crystals of metal-free ConA with both Zn2+ and Ca2+ contain the active holoprotein (ConA ZnCa). Ca2+ has induced large conformational changes to stabilize its hepta-coordination in the S2 site, which comprise the trans to cis isomerization of the Ala207-Asp208 peptide bond accompanied by the formation of the saccharide-binding site. The Zn2+ ligation in ConA ZnCa is similar to Mn2+, Cd2+, Co2+, or Ni2+ ligation in the S1 site, in disagreement with earlier extended x-ray absorption fine structure results that suggested a lower coordination number for Zn2+.

Concanavalin A (ConA), 1 the lectin isolated from Canavalia ensiformis (jack bean), has for decades been known and applied as a mitogen, interacting with cell surfaces by binding specific carbohydrates (Sharon and Lis, 1989). The lectin reversibly binds saccharides only when the calcium-binding site S2 is occupied (Kalb and Levitzki, 1968). Binding of calcium itself is conditional on the binding of one of several divalent metal ions (Mn 2ϩ , Co 2ϩ , Ni 2ϩ , Cd 2ϩ , or Zn 2ϩ ) (Shoham et al., 1973) or, at neutral pH, also Ca 2ϩ (Koenig et al., 1978) at the transition metal binding site S1. The metal ions in S1 and S2 are only 4.2 Å apart and are bridged by two aspartate carboxyl groups (Asp 10 and Asp 19 ) (Hardman et al., 1982). Two ligand side chains of calcium are involved in monosaccharide binding, either directly (Asn 14 ) or via a water molecule (Asp 208 ) (Derewenda et al., 1989). The binding of the metal ions causes large structural changes that pull the ligand residues of the saccharide-binding site into place and induce the carbohydrate binding capacity of the lectin. Nuclear magnetic resonance dispersion studies have allowed the successive binding to metal-free ConA of the metal ions in S1 and S2 to be followed (Brown et al., 1977). The binding of the metal ions induces a slow (from minutes to days) conformational transition in the protein, referred to as ''locking'' because it involves an increase in the affinity of both metals for the protein (Koenig et al., 1978). The high activation energy for this process is probably largely due to the isomerization of a non-proline peptide bond, Ala 207 -Asp 208 , located in a ␤-strand neighboring the S2 site. This bond has the trans-conformation in metal-free ConA  but adopts the unusual cis-conformation in the native protein.
A most interesting result of the NMRD studies was the demonstration that a significant portion of metal-free ConA, 12.5% at 25°C and pH 6.4, is locked (Brown et al., 1982). Locked metal-free ConA can bind methyl-␣-D-mannopyranoside with 7% of the affinity of the fully metallized locked conformation, which is at least a factor 10 3 higher than its binding to the metal-free unlocked protein. Even more striking is that the unlocked/locked equilibrium can be shifted from being predominantly in the unlocked conformation to more than one-half (60%) locked, by preequilibration with excess saccharide for 5 days at 25°C. The specificity of saccharide binding is preserved in the metal-free locked conformation (i.e. galactose does not bind). Brown and co-workers (Brown et al., 1977;Brown et al., 1982) concluded that it is predominantly the locked conformation of ConA that is responsible for the saccharide binding of the lectin. The contribution of the metal ions is to maintain the locked conformation. The binding of Mn 2ϩ in S1 and Ca 2ϩ in S2 shifts the unlocked/locked equilibrium completely toward the locked form.
In our work, we analyzed two structural intermediates in the pathway of the unlocked to the locked conformation, through the binding of a metal ion only in S1. A first approach was the soaking of crystals of metal-free ConA  with zinc or cobalt (apoZn-ConA and apoCo-ConA). Zn 2ϩ and Co 2ϩ can bind in S1 but not in S2 (Shoham et al., 1973), in contrast to Mn 2ϩ , Cd 2ϩ , and Ca 2ϩ that can occupy both sites (Shoham et al., 1973;Harrington and Wilkins, 1978;Koenig et al., 1978). The binding of a metal ion in the proto-S1 site does not induce significant structural changes. Because intermolecular crystal packing interactions may hinder the sequence of changes upon metal ion binding in S1, a second approach was the co-crystallization of metal-free ConA with zinc to obtain Zn-ConA. ConA ZnCa, the structure resulting from the cocrystallization of metal-free ConA with both Zn 2ϩ and Ca 2ϩ , has Zn 2ϩ bound in S1 and Ca 2ϩ in S2 and bears the structural features of the native, locked ConA.

MATERIALS AND METHODS
Purification, Demetallization, Remetallization, Crystallization, and Data Collection-ConA was purified from meal of the jack bean (Canavalia ensiformis) and demetallized as described . Throughout the crystallization experiments, systematic care was taken to prevent trace metal contamination. The solutions used for the crystallizations were batch treated with Chelex 100 beads from Bio-Rad. The pH was adjusted after equilibration with this chelator. Only metal-free pipette tips (Bio-Rad) and new plastics were used and rinsed several times with deionized water prior to use.
ApoZn-ConA and apoCo-ConA crystals originate from metal-free ConA crystals (Bouckaert et al., 1996a) soaked for 3 h in the mother liquor with 5 mM ZnCl 2 or CoCl 2 added. After soaking, the crystals were mounted and kept at 4°C for at least 2 weeks before data collection to allow any conformational equilibrium to be reached. The crystals remained in space group P2 1 2 1 2 of metal-free ConA and contain a dimer in the asymmetric unit (Table I).
Zn-ConA co-crystals (Table I) were grown by vapor diffusion against a precipitant solution of 2 M (NH 4 ) 2 SO 4 , 60 mM sodium acetate buffer, pH 5.0, and 3% poly(ethylene glycol) methylether (M r 5000). Sitting drops of 75 l were composed of one-third 1.73 mg/ml metal-free ConA in 10 mM ZnCl 2 and two-thirds of the precipitant solution. ConA ZnCa crystallized after mixing 30 l of 1.8 M (NH 4 ) 2 SO 4 and 200 mM NaH 2 PO 4 at pH 7.1, with 30 l of 6 mg/ml metal-free ConA, 30 mM ZnCl 2 , and 30 mM CaCl 2 . The metal content of the crystals was checked on a scanning electron microscope with an energy dispersive x-ray analyzer. The crystals were mounted in borosilicate capillaries. Data were collected on an Enraf Nonius FAST area detector and integrated using the MAD-NESS software (Pflugrath and Messerschmidt, 1989). Scaling and merging was done using the CCP4 programs. The statistics of the data collection are shown in Table I.
Structure Solution, Refinement, and Analysis-The program X-PLOR (Brü nger, 1992) was used for all refinement runs on a Silicon Graphics INDY workstation and on a CRAY-YMP supercomputer. Model building and graphical inspection were performed using O (Jones et al., 1991) on a Silicon Graphics Indy workstation and using FRODO (Jones, 1978) on an Evans and Sutherland PS390. Superpositions were done using the LSQMAN program package (Kleywegt and Jones, 1994). Intermolecular contacts were analyzed with X-PLOR.
In all four structures (apoZn-ConA, apoCo-ConA, Zn-ConA, and ConA ZnCa) residues Val 7 -Ile 25 , involved in metal binding, Gly 98 -Lys 101 , involved in saccharide binding, and residues Pro 206 -Ile 210 , that contain the cis-peptide linkage in the holoprotein (Hardman et al., 1982), were initially omitted from the model. Validation of the stereochemical quality was performed throughout model improvement cycles using PROCHECK (Morris et al., 1992). The refinement and geometry statistics are shown in Table II. Water molecules were included in the model when they made at least one reasonable hydrogen bonding contact, with the distance between hydrogen bond donor and acceptor atoms set to be within the range of 2.5-3.5 Å, and if they were visible in an F o Ϫ F c electron density above the 3 level. They were accepted if they reappeared at the 1 level in the 2F o Ϫ F c map after refinement.
The metal-free ConA structure (Protein Data Bank entry 1APN)  was used as the starting model for the apoZn-ConA and apoCo-ConA soaked crystal structures. Rigid body refinement could be applied directly to the model without molecular replacement. Positional and temperature factor refinement were performed at 2.7-Å resolution for apoZn-ConA (Protein Data Bank entry 1CES) and at 2.8-Å resolution for apoCo-ConA (Protein Data Bank entry 1ENS). The refinements converged rapidly, and inspection of the electron density showed that these structures share a high degree of similarity with metal-free ConA. The metal ions Zn 2ϩ and Co 2ϩ were occupancy refined, since their density did not allow full occupancy. Non-crystallographic symmetry restraints were applied to the two monomers forming the dimer of the asymmetric unit (weight, 200 kcal mol Ϫ1 Å Ϫ2 ).
Molecular replacement of Zn-ConA was performed with the program AMORE (Navaza, 1994) with the tetramer of ConA complexed with methyl ␣-D-mannopyranoside as search model (Protein Data Bank en-  try 5CNA) . A unique solution was found with a correlation coefficient of 67.6% and an R-factor of 39.2% after rigid body refinement. Individual atomic positions and temperature factors were refined using all reflections between 10 and 2.5 Å. The non-crystallographic symmetry within the dimer was strongly restrained (500 kcal mol Ϫ1 Å Ϫ2 ) for all of the residues not involved in metal binding, saccharide binding, or packing interactions. In the last two refinement cycles, the non-crystallographic symmetry restraints were somewhat relaxed (300 kcal mol Ϫ1 Å Ϫ2 ) and applied to all of the residues, except to those that could not be included in all four monomers in the asymmetric unit (Protein Data Bank entry 1ENQ). The space group of ConA ZnCa crystals is I222, and they are isomorphous to all ConA crystals solved up to now that contain both a transition metal ion and a calcium ion. Data were collected up to 1.8 Å. All of the data were used in the refinement. Rigid body refinement of the ConA CdCa model (Protein Data Bank entry 1CON) without the metal ions and the water molecules decreased the R-factor to 25.8% in the 10 -2.46-Å resolution range. Positional and individual temperature factor refinement of all the reflections between 10-Å and 1.80-Å resolution caused a further decrease to 24.59%. Inclusion of the Zn 2ϩ and Ca 2ϩ ions together with 133 water molecules reduced the R-factor to 21.5% before refinement and further to 19.04% after POWELL energy minimization and restrained B-value refinement. Repeated corrections and inclusion of multiple conformations for the side chains of the residues Ser 21 , Val 65 , Ser 72 , Thr 74 , Ser 113 , and Ser 134 and a final number of 142 water molecules, including one positioned on a 2-fold axis, resulted in an R-factor of 17.61% (Protein Data Bank entry 1ENR).

RESULTS
ApoCo-ConA and apoZn-ConA-Binding of a metal ion in the S1 site only was studied both by soaking metal-free ConA crystals and by co-crystallization in the presence of transition metal ions that, by themselves, are not capable of converting the protein to its native, locked form. In all three of these crystal structures, the Ala 207 -Asp 208 peptide bond, which adopts the unusual cis configuration in native ConA, remains trans. Cobalt or zinc binding alone is thus not sufficient to establish the cis-conformation of the Ala 207 -Asp 208 peptide bond.
Upon soaking of metal-free ConA crystals with zinc or cobalt, the structural changes around the S1 site are restricted to an adjustment of the Glu 8 side chain induced by the partial binding of the metal ion. The structures of apoZn-ConA and apoCo-ConA are highly isomorphous to that of metal-free ConA. The root mean square difference for the backbone N, C, and C␣ atoms with ConA ZnCa is about 0.7 Å for both structures, identical to the root mean square difference between metal-free ConA and ConA ZnCa. The metal-binding loop (Tyr 12 -Tyr 22 ) is too flexible to display electron density. This ⍀-loop moves around its hinges Pro 13 and Pro 23 and extends into the solvent. The resemblance to the metal-free ConA structure is apparent from the conformation of the visible edges of the loop (Pro 13 -Asn 14 and Ser 21 -Tyr 22 ) and from the clearly defined side chain conformations of the residues residing in the adjoining ␤-strands (Tyr 12 and His 24 ) . Also, the distant saccharide-binding loop (Thr 97 -Glu 102 ) has a conformation identical to that of metal-free ConA (Fig. 3). The peptide nitrogen atoms of Leu 99 and Tyr 100 , which directly take part in monosaccharide binding, move over 4 Å in concert with the whole loop upon depletion of the metal ions. The structure of a third loop, between the residues Pro 202 and Pro 206 and preceding the Ala 207 -Asp 208 peptide bond that undergoes the cis-to trans-isomerization upon demetallization is again similar, besides a small translation, to this of metal-free ConA in both apoZn-ConA and apoCo-ConA.
The subsequent steps in the metallization process are schematically represented in Fig. 1. Due to the moderate resolution of the monometallized structures (Table I), only a rudimentary sketch of their ligation can be made. The atoms Glu 8 OE2, Asp 10 OD2, and His 24 NE2 are always in their ligating position. The number of Co 2ϩ or Zn 2ϩ ligands varies from three to six (Table III), depending on the number of visible water ligands.
The S1 binding site conserves the positions of the ligand residues His 24 , Glu 8 , and Asp 10 upon demetallization (Fig. 1a) . Asp 19 , on the other hand, loses its position as a ligand residue because of its location in the metalbinding loop that detaches from the protein surface and gains high flexibility. In the holoprotein, Asp 19 anchors the metalbinding loop and helps it to fold back over the protein by its binding to both metal ions.
In apoZn-ConA and in apoCo-ConA, His 24 , Glu 8 , and Asp 10 are already in position and readily ligate the transition metal ion (Fig. 1b). Fig. 2 shows that the partially bound cobalt or zinc ion does not intrude as deeply into the S1 site as is in the S1 site of Zn-ConA or of the holoprotein. The metal-ligand distances are unusually large (Table III), probably because of averaging the position of the disordered ion that only partially occupies the site. The Glu 8 and Asp 10 side chains retain the orientation of the metal-free form. The rest of the metal coordination sphere is formed by water molecules that occupy equivalent positions as the water molecules in the empty site S1 of metal-free ConA. In monomer B of the dimer in the asymmetric unit, fewer water ligand molecules can be located, possibly because of the larger structural disorder in monomer B compared to in monomer A, reflected in higher temperature factors.
Zn-ConA-The co-crystals of metal-free ConA with zinc only (Zn-ConA) show the same overall structure in the metal-binding and the saccharide-binding region as the monometallized apoZn-ConA and apoCo-ConA. The root mean square deviation for the backbone N, C, and C␣ atoms between the four monomers in the asymmetric unit of Zn-ConA and the monomer of ConA ZnCa varies between 0.63 and 0.73 Å. The whole metalbinding loop of Zn-ConA is visible in the electron density for three of the four monomers (except D) in the asymmetric unit. The loop is extended and has large differences in dihedral angles relative to its conformation in the holoprotein Con ZnCa (Fig. 2). Zn 2ϩ is bound with full occupancy in the S1 site and in the same position as the zinc ion of Con ZnCa, in contrast to the apoCo-ConA or apoZn-ConA structures. Its binding leads to the native-like structure of the S1 site because it orients the side chains of its ligand residues Glu 8 , Asp 10 , and His 24 similarly to Con ZnCa. Despite the binding of a metal ion in S1, the S2 site is not fully formed. The ligand residues Asn 14 and Asp 19 are not in a ligating position. The side chain of Asn 14 , which serves as a calcium ligand through its amide oxygen, retains too high a mobility to be visible in the electron density.
The backbones of the saccharide-binding loops of apoZn-ConA or apoCo-ConA and Zn-ConA can be perfectly superimposed (Fig. 3), except for Tyr 100 . Tyr 100 of Zn-ConA is retracted into a position intermediate between those of metal-free and native ConA, 2.3 Å from its native position. Arg 228 , a saccharide-binding residue, has slightly regressed (1.5 Å) from the S2 site in relation to its position in the metal-free protein. In the

FIG. 2. Superposition of residues Glu 8 -His 24 of the metal-binding loop and adjacent residues.
The ribbon represents the backbone of apoZn-ConA (black, interrupted at Pro 13 and continued at Pro 23 ), Zn-ConA (dark gray), and ConA ZnCa (white). Some residues important in metal ion ligation (see also Fig. 1) and saccharide binding (see also Fig. 3) are depicted in ball-and-stick mode (MOL-SCRIPT (Kraulis, 1991)). The metal ions are denoted as spheres.
latter, its guanidinium group intrudes into the disrupted S2 site, presumably for charge compensation upon depletion of the positively charged metal ions. The loop (Pro 202 -Pro 206 ) that precedes residue Asp 208 is even more remote from its native position than in apoZn-ConA or apoCo-ConA, probably because in the latter this loop is sterically hindered to extend by contacts within 4 Å of the symmetry related residues Asp 82 , Asn 83 , Trp 182 , and Ser 184 .
The number of zinc ligands varies between four and six due to the invisibility or the absence of water molecules (Table III). Besides His 24 NE2, Glu 8 OE2, and Asp 10 OD2, three water molecules are ligands to the zinc ion (Fig. 1c). Waters OW A and OW B are in the position of the water S1 metal-ligands in ConA ZnCa and native ConA. The Asp 19 carboxylate group makes a hydrogen bond to water OW C of the zinc ion. This is most clearly seen in monomer A, hinting at an important interaction that may occur to prepare monometallized ConA for the binding of the second metal ion in S2 in monomer A of Zn-ConA. The temperature factor for this water is 31 Å 2 ; for Asp 19 , it is 56 Å 2 .
The intermolecular packing in the Zn-ConA co-crystals is distinct from any other ConA structure known. The molecules pack in a spiral staircase motif. This arrangement avoids extensive intermolecular contacts due to the crystal lattice, since there are 36 intermolecular protein-protein contacts per monomer of Zn-ConA, as compared to 93 for the monomer of ConA ZnCa. The linear ordering of tetramers gives a different amount of conformational freedom to the metal-binding and saccharide-binding regions of every of the four ConA monomers forming the tetramer in the asymmetric unit and related by non-crystallographic symmetry. In contrast to crystals of the ConA methyl ␣-D-mannopyranoside complex, not all monomers make intermolecular lattice contacts over the whole metalbinding and saccharide-binding region. The A and B monomers make contacts with residues of the metal-binding loop but not of the saccharide-binding loop. The reverse is true for the C and D monomers. In general, the loops not involved in packing interactions are more mobile and consequently not always visible in the electron density, whereas regions that take part in crystal packing interactions can be stabilized in one conformation. In monomers A and B of Zn-ConA, this leads to interpretable electron density for the metal-binding loop. There is no severe packing stress, since these "stabilized" loops still have high temperature factors.
ConA ZnCa-Co-crystallization of metal-free ConA with zinc and calcium at pH 7.1 leads to ConA ZnCa (Table I) with a structure (at 1.8-Å resolution) quasi identical to that of native ConA  and any of the locked structures reported recently to high resolution (ConA CdCa (2.0-Å resolution) (Naismith et al., 1993), ConA CoCa (1.6-Å resolution) and ConA NiCa (2.0-Å resolution) , except for having Zn 2ϩ bound in S1. Also, the zinc ligation in the S1 site of ConA ZnCa (Table III) is similar to that of Mn 2ϩ , Cd 2ϩ , Ni 2ϩ , or Co 2ϩ . This is in contradiction to reported EXAFS results (Lin et al., 1990(Lin et al., , 1991a. The zinc ion has a regular octahedral coordination sphere, and the calcium ion is hepta-coordinated (Figs. 1d and 4; Table III).

Conformational Changes in Metal-free ConA upon Metal Ion
Binding-In all three monometallized ConA structures analyzed in this work, the crystals of metal-free ConA soaked with Co 2ϩ (apoCo-ConA) or with Zn 2ϩ (apoZn-ConA) and the cocrystals of metal-free ConA with Zn 2ϩ (Zn-ConA), the lectin maintains the unlocked conformation, the predominant conformation of metal-free ConA that is incapable of binding carbohydrates. The co-crystal of metal-free ConA with Zn 2ϩ and Ca 2ϩ (ConA ZnCa) contains the locked conformation of ConA. It is thus the binding of Ca 2ϩ in the S2 site that induces the large conformational change from unlocked to locked with the creation of the saccharide-binding site, and that is critical for the trans to cis isomerization of the Ala 207 -Asp 208 peptide bond.
The Unlocked Conformation: apoCo-ConA and apoZn-ConA-The soaking of metal-free ConA crystals with Co 2ϩ or Zn 2ϩ results in partial binding of these ions in S1 without further structural change. These structures can represent the very first step in going from metal-free ConA toward the holoprotein. Metal ion binding is readily possible due to preservation of the S1 site upon demetallization, with residues Glu 8 , Asp 10 , and His 24 rigidly held in metal ligand positions. The absence of Asp 19 , located in the extended metal-binding loop, does not disable metal ion binding in the proto-transition metal binding site S1. The S1 site residues do not adopt the native conformation ( Fig. 2) but maintain the conformation of the metal-free form. Since the soaked crystals remain isomorphous to metal-free ConA crystals (Table I), the imposed crystal packing interactions may limit the accessibility of the metal-binding sites and/or hinder conformational changes upon binding of the metal ion.
The Unlocked Conformation: Zn-ConA-In the Zn-ConA cocrystal structure, Zn 2ϩ is bound with full occupancy. The binding of the positively charged transition metal ion in S1 requires not only the deprotonation of His 24 but also contributes significantly to the neutralization of the metal-binding region by binding two of its negatively charged residues (Glu 8 and Asp 10 ) and by stabilizing the third (Asp 19 ) by a hydrogen bond via its water ligand. Apparently, the binding of the zinc ion in S1 facilitates the subsequent binding of Ca 2ϩ to the S2 ligands by bringing the Asp 19 ligand in the vicinity of the would-be S2 site. In contrast to apoCo-ConA and apoZn-ConA, which maintain the metal-free like conformations for the S1 site residues, the carboxylate ligand residues Glu 8 and Asp 10 have native-like conformations, except for Asp 19 (Fig. 2). The binding of a metal ion in S1 is thus a step toward the native structure. This may explain the requirement for successive binding of the metal ions first in S1 and subsequently in S2 (Brown et al., 1977) and may also explain why a larger portion of monometallized ConA (30%) is locked compared to metal-free ConA (12.5%) (the locking of ConA with only Mn 2ϩ bound in S1 is represented by the equilibrium constant K LMP in the metal binding scheme of Brewer et al., 1983a).
Binding of a transition metal ion in S1 does not lead to the formation of the S2 site. The S2 ligand residues are still dispersed. Therefore, the S2 site is not very successfully prepared for the binding of calcium. Only three of the five protein ligand atoms that build up the hepta-coordination sphere of calcium are in ligating positions. These are the two carboxylate atoms of Asp 10 , perfectly oriented by the zinc ion, and the peptide carbonyl of Tyr 12 , 1 Å remote from its native position. The other two protein ligand atoms, Asp 19 OD1 and Asn 14 OD2, are not available. Asp 19 is too far away (Fig. 3), despite its indirect interaction with the S1 bound metal ion, and the Asn 14 side chain is too mobile. These structural features probably account for the extremely weak binding of Ca 2ϩ (C) in the S2 site to unlocked, Mn 2ϩ -bound ConA (MP), with a dissociation constant of 0.3 mM at pH 6.4 and 5°C (Brown et al., 1982) (corresponding to K CMP in the metal binding scheme of Brewer et al., 1983a), which is much weaker than for the binding of Mn 2ϩ in the better conserved S1 site (25 M at pH 6.4).
The Mechanism of Metal Ion-induced Conformational Locking-The metal-free and the monometallized ConA are in conformational equilibrium with the unlocked conformation favored. Metal ions not only bind, with high affinity, to the locked form of metal-free ConA, but moreover also bind to the unlocked conformation and induce locking. The binding of Mn 2ϩ and Ca 2ϩ decreases the intrinsic locking time constant (at 5°C) to 0.27 h (Brewer et al., 1983b) and shifts the unlocked/locked equilibrium completely toward the locked form.
The crystallographic structure of Zn-ConA indicates that the calcium ion can bind only weakly in the disrupted metal-binding site S2 of unlocked ConA after the binding of a metal ion in S1. The large flexibility of Ca 2ϩ in accepting more variable and irregular coordination geometries than similar ions (McPhalen et al., 1991) and cooperativity among the metal-binding ligands may help the Ca 2ϩ binding in S2 and lead to the formation of a fully metal-bound, but still unlocked, ConA species. To stabilize its coordination in S2, the calcium ion must induce locking and must overcome the energy barrier separating the trans and the cis isomers of the Ala 207 -Asp 208 peptide bond. The isomerization of this bond, ideally situated in the ␤-strand facing the saccharide-binding loop (Thr 97 -Glu 102 ) on the one side and the S2 site on the other side, could be the key needed for the locking. On the one side, the isomerization leads to the active conformation of the saccharide-binding loop, thereby requiring the disruption of two strong hydrogen bonds between the side chains of Asp 208 and Asn 104  see also the orientation of these side chains on Fig. 3). On the other side, the calcium ion stabilizes the cis-conformer, and one of its water ligands makes hydrogen bonds to the carbonyl and a carboxylate oxygen of Asp 208 .
The existence of an initially unlocked state of the fully metalbound ConA was demonstrated by Brown et al. (1977), who showed that EDTA readily removes the metals from unlocked, Mn 2ϩ -, and Ca 2ϩ -bound ConA (CMP). Moreover, this form binds methyl-␣-D-mannopyranoside only 1 order of magnitude better than the metal-free unlocked form (Koenig et al., 1978). The lack of saccharide binding capacity and high dissociation rates of the metal ions, particularly in the S2 site, indicate that this fully metal-bound ConA structure is not yet locked. However, it is likely that the S1 and S2 sites are already well structured by the metal ion binding in the unlocked, fully metal-bound ConA state, because the relaxivity dispersion of the water protons by the Mn ion in the unlocked and the locked species of Mn 2ϩ -and Ca 2ϩ -bound ConA have almost the same value (CMP and CMPL, respectively, in Brown et al., 1977).
The mechanism of saccharide-induced locking is different from that of metal ion-induced locking. Methyl-␣-D-mannopyranoside can bind only to the locked form of metal-free ConA and is not known to influence the intrinsic locking time of the transconformation event (Grimaldi and Sykes, 1975;Brewer et al., 1983b). Stabilization of the locked form by binding of the monosaccharide shifts the unlocked/locked equilibrium toward the locked form. It is not known whether the cis-conformer is established in locked, metal-free ConA. We presume that the locked form of metal-free ConA must have a well formed S2 site, because of its high affinity for both the S1 and S2 metal ions. Co-crystallizations of metal-free ConA with saccharides have been set up, thus far without results. FIG. 4. Omit F o ؊ F c electron densities for the metal ions Zn 2؉ in the S1 site and Ca 2؉ in the S2 site and their water ligands in ConA ZnCa.
In conclusion, a mechanism is proposed for the conformational changes upon metal ion binding to metal-free ConA that explains the results from this crystallographic study and that is in agreement with results from nuclear magnetic resonance dispersion (Brewer et al., 1983b), fluorescence (Harrington and Wilkins, 1978), and stopped-flow nuclear magnetic resonance kinetic studies (Grimaldi and Sykes, 1975). The sequence of ConA structures indicates the requirement for the sequential binding of the metal ions firstly in S1 and then in S2, followed by the initially weak binding of Ca 2ϩ in the disrupted S2 site that on its turn orders S2 and the Ca 2ϩ -induced locking that comprises the trans to cis isomerization of the Ala 207 -Asp 208 peptide bond with the ensuing formation of the saccharidebinding site and the stabilization of the whole metal-binding and saccharide-binding region.
Metal Ligation in ConA ZnCa-In the structures of the holoprotein analyzed up to now, including this ConA ZnCa derivative, the nature of the S1 metal ion (Mn 2ϩ , Cd 2ϩ , Co 2ϩ , or Ni 2ϩ ) has not been found to influence its overall coordination geometry (Hardman et al., 1982) or the saccharide binding. The only difference is the average transition metal-ligand distance, increasing from 2.18 Å for Co 2ϩ , 2.20 Å for Ni 2ϩ , 2.21 Å for mostly Mn 2ϩ bound in S1 , over 2.25 Å for Zn 2ϩ (this work), to 2.31 Å for Cd 2ϩ (Naismith et al., 1993).
The coordination number of the zinc ion in ConA ZnCa calculated from the EXAFS measurements is 6 for the solution sample but only 4 in the crystals (Lin et al., 1991a). This difference was ascribed to intermolecular crystal packing forces imposed on the residues involved in lattice contacts, transmitted to and absorbed by the zinc ion (Lin et al., 1990). The zinc ion would discharge the crystal packing stress by shortening its metal-ligand distances and the release of two of its ligands. The discrepancies that exist in coordination number of the S1 bound metal ion, not only between this crystallographic study and the EXAFS study, but also between the first interpretations of EXAFS (Kalb (Gilboa) et al., 1979;Lin et al., 1990), and its later revisions (Lin et al., 1991a;Lin et al., 1991b), can probably largely be ascribed to unobserved water ligand molecules. Zinc-coordinated water molecules are kinetically labile and typically exchange rapidly (Bertini and Luchinat, 1994).
Zinc and Cobalt Binding to ConA, Their Specificity for S1, and Their Inability to Bind in S2-Why are Co 2ϩ and Zn 2ϩ unable to bind in the S2 site of ConA? The S2 site preferentially binds calcium with a low dissociation constant (about 10 Ϫ8 M in Koenig et al., 1978). Ca 2ϩ (ionic radius, 0.99 Å) fits S2 in hepta-coordination ion. Co 2ϩ and Zn 2ϩ have a relatively small ionic radii (0.70 and 0.71 Å, respectively, for a coordination number of 6) and average coordination numbers smaller than 6 (5.7 and 5.0, respectively) (Glusker, 1991). Cobalt and zinc may be too small and make too few contacts in the large, irregular Ca 2ϩ -coordination sphere. The second zinc ion per ConA monomer found by Palmer et al. (1980) to displace Cd 2ϩ is more likely bound in the third metal-binding site distinct from the S2 site. This site is capable of binding cadmium, present as a second cadmium ion bound per ConA monomer in the crystal structure of ConA CdCa besides the S1-bound cadmium. The third metal-binding site also binds heavy metal ions like Pb 2ϩ and Sm 2ϩ (Palmer et al., 1980;Becker et al., 1975), Gd 3ϩ , Tb 3ϩ , and Eu 3ϩ (Barber et al., 1975;Sherry and Cottam, 1973).
The Zn-ConA co-crystal at pH 5 illustrates how little the binding of the S1 ion prepares the S2 site for calcium ion binding. The possibility exists that Mn 2ϩ or Cd 2ϩ bound in S1, in the absence of a metal ion in S2, leads to preferential binding of Ca 2ϩ rather than Zn 2ϩ , because they are larger (ionic radius, 0.8 and 0.91 Å, respectively, in hexa-coordination) and tend to have a higher coordination number (6.0 and 6.1, respectively) (Glusker, 1991). They might interact directly with Asp 19 instead of via a water molecule. Monometallized manganese or cadmium ConA, however, are difficult to isolate, because these ions can bind in both S1 and S2 (Bouckaert et al., 1996).