The roles of individual gamma-carboxyglutamate residues in the solution structure and cation-dependent properties of conantokin-T.

The solution structure of the Ca2+-loaded conantokin-T (con-T), a gamma-carboxyglutamate (Gla)-containing 21-residue peptide (NH2-G1EgammagammaY5QKMLgamma10NLRgammaA15EVKKN20A-CONH2,gam ma = Gla), has been elucidated by use of distance geometry calculations with experimental distances derived from two-dimensional 1H NMR spectroscopy. An end-to-end alpha-helix was the dominant conformation in solution, similar to that of apo-con-T, except that reorientation of several side chains occurred in the Ca2+-coordinated complex. The most notable examples of this were those of Gla10 and Gla14, which were more optimally positioned for complexation with Ca2+. In addition to the stabilization offered to the alpha-helix by Ca2+ binding, hydrophobic clustering of the side chains of Tyr5, Met8, Leu9, and Leu12, and ionic interactions between Lys7 and Gla3/Gla10 and between Arg13 and Gla14, along with hydrogen bonding between Gln6 and Gla10, were among the side chain interactions likely playing a significant role in maintenance of the alpha-helical conformation. Docking of Ca2+ in the con-T structure was accomplished using genetic algorithm-molecular dynamics simulation approaches. The results showed that one Ca2+ ion is most likely coordinated by four side chain oxygen atoms, two each from Gla10 and Gla14. Another bound Ca2+ ion has as its donor sites three oxygen atoms, two from Gla3 and one from Gln6. To examine the functional roles of the individual Gla residues, a series of variant peptides have been synthesized with Ala substituted for each Gla residue, and several properties of the resulting variants have been examined. The data obtained demonstrated the importance of Gla10 and Gla14 in stabilizing binding of the highest affinity Ca2+ site and in governing the conformational change induced by Ca2+. The critical nature of Gla3 and Gla4 in inhibition of the spermine-induced potentiation of the binding of MK-801 to open ion channels of the N-methyl-D-aspartate receptor was established, as well as the role of Gla4 in stabilizing the apo-con-T alpha-helical conformation.

con-T 1 is a neuroactive peptide found in the venom of Conus tulipa (1). It is among a vast number of small peptides, typically 10 -30 amino acids in length, employed by these and other species of cone snails to immobilize their prey and their predators (2,3). The targets of this array of peptides are neuroreceptors and skeletal muscle receptors, and the remarkable selectivity shown by this general class of peptides has also encouraged their use as laboratory reagents to block specific classes of these receptors (4,5). Specifically, a general class of Conus peptides, the conotoxins, which contain a relatively high number of disulfide bonds, interact with nicotinic acetylcholine receptors (␣-conotoxins), voltage-sensitive Na ϩ channels (conotoxins), and voltage-sensitive Ca 2ϩ channels (-conotoxins) (6). Conantokins, such as the 17-amino acid residue peptide con-G (7) and the homologous 21-residue con-T (1), are Gla-containing peptides without disulfide bonds. They have been shown to function as noncompetitive inhibitors of the enhancement by spermine and spermidine of Ca 2ϩ flow into neurons, by targeted action on glutamate/glycine receptors of the NMDA subclass (8 -11).
Certain structural features of these small peptides are relevant to their functions. con-G and con-T contain five and four residues of Gla/mol of peptide, respectively (1,7). Because of this feature, both peptides interact with divalent cations that are important to the functions of neuronal cells, such as Ca 2ϩ and Mg 2ϩ (3,12), as well as other metal ions (12,13). The binding of Ca 2ϩ and similar cations induces a large conformational change in con-G (12) from an essentially random structure to an ␣-helix. On the other hand, con-T, although it also interacts with these cations, does not undergo as dramatic a conformational change, because apo-con-T is already highly organized in an ␣-helical conformation (12,14).
The conantokins possess the potential to serve as agents that inhibit the flow of Ca 2ϩ into neurons via the NMDA receptor route and thus eliminate the harmful effects of entry of this cation into neuronal cells. However, these peptides cannot readily be directly employed as pharmaceutical agents for this purpose in humans because they do not cross the blood-brain barrier. Thus, as part of a rational drug design program, elucidation of the structure-function relationships of these peptides is essential. A recent emphasis on this topic is witnessed by publications on elaboration of the NMR-derived, three-dimensional structures of apo-con-G (13,15), Ca 2ϩ ⅐con-G (13), and apo-con-T (13,14,16). Most of these investigations have focused on backbone conformations, but one has rigorously defined side chain orientations of apo con-T (14). In the present phase of our efforts in this area, we have defined the threedimensional conformation of the Ca 2ϩ ⅐con-T complex and have employed mutant con-T peptides to attempt to understand the role of individual Gla residues in defining its metal binding properties and its bioactivity. The results of this study are summarized in this report.
Peptide Synthesis, Purification, and Characterization-The peptides were synthesized on a 0.1 mmol scale on a PAL resin support (PerSeptive Biosystems, Framingham, MA) using an Applied Biosystems (Foster City, CA) model 433A peptide synthesizer, as described previously (12). Purification of con-T and mutant con-T peptides was accomplished by fast protein liquid chromatography on a Bioscale DEAE-20 column (Bio-Rad) equilibrated with 10 mM NaBO 3 , pH 8.0. A 500-ml linear gradient of NaCl, from 10 mM NaBO 3 , pH 8.0 (start solution), to 10 mM NaBO 3 , 500 mM NaCl, pH 8.0 (limit solution), was employed. The target material was pooled, lyophilized, and then desalted on a Sephadex G-15 (Pharmacia Biotech Inc.) column that was equilibrated and eluted with 0.1% NH 4 OH. The peptides were characterized by reverse-phase high performance liquid chromatography and delayed extraction-matrix-assisted laser desorption ionization-time of flight mass spectrometry, as described in an earlier communication (12).
Calcium Binding-The Ca 2ϩ binding isotherms for each peptide were determined by potentiometry at 25°C using a semi-micro Ca 2ϩselective electrode (Orion, Los Angeles, CA). A fused double-walled titration vessel with inlet and outlet ports was attached to a circulating water bath for precise temperature control in these titrations. Chelex-100 (Bio-Rad)-treated peptides (1-2 mM peptide in 10 mM NaBO 3 , 100 mM NaCl, pH 6.5) were titrated with CaCl 2 . Data analyses were conducted as described previously (19).
Circular Dichroism-CD spectra were recorded between 190 and 260 nm on an Aviv model 62DS spectrometer. Peptides were dissolved in 10 mM NaBO 3 , 100 mM NaCl, pH 6.5, to a final concentration of 35 M. A 1-cm-path-length cell was employed. Spectra representing the average of five scans were collected at a 1.0-nm bandwidth at 1.0-nm intervals. The ␣-helical contents at various Ca 2ϩ concentrations were calculated according to a previously published method (20).
Two-dimensional 1 H NMR-The peptide was dissolved in 10 mM NaBO 3 , 100 mM NaCl, pH 6.5, to a final concentration of approximately 2 mM. A volume of 50 l of 2 H 2 O was added to 450 l of the peptide solution to provide the NMR deuterium lock signal. The internal reference was 4,4-dimethyl-4-silapentane-L-sulfonate. Dilute HCl to was added to adjust the pH of the sample to 6.5 to minimize the loss of NH signals that would occur at higher pH values. One-and two-dimensional NMR experiments were carried out on a Bruker AMX-500 spectrometer. The data were collected, processed, and analyzed as described previously for structure determination of apo-con-T (14). Briefly, sequence-specific assignments of the proton resonances for con-T in the presence of Ca 2ϩ were achieved by combining the procedures of spin system identification using TOCSY and DQF-COSY, followed by sequential assignments through NOE connectivities (21). Assigned NOE cross-peaks were characterized as strong, medium, or weak as determined from the number of contours and converted to distance upper bounds of 2.7, 3.7, and 5.0 Å, respectively. Spectral processing was carried out using FELIX (Biosym Technologies, San Diego, CA) and an in-house program, nmrDSP, on Silicon Graphics workstations. Spectral contrast-enhancement methods were applied to resolve severely overlapped proton resonances (22,23). The Sybyl software package (Tripos, Inc., St. Louis, MO) was used for spectral visualization.
Peptide conformations, free from steric overlaps and consistent with the NMR data, were generated by DG calculations using the fixed bond lengths and bond angles provided in the ECEPP/3 data base (24). All side chain-side chain or side chain-main chain medium range (i ϩ 2 and i ϩ 4) constraints were set to an upper bound of 5 Å and a lower bound of 2.0 Å. In the absence of metal coordination sites, we chose not to employ energy minimization as a structural refinement because of the potential bias that this could introduce in the side chain orientations. Because the major backbone NOE contacts in apo-con-T and in Ca 2ϩ ⅐con-T were essentially the same, our approach to modeling the Ca 2ϩ ⅐con-T NMR-derived structure was to start with the 10 energyminimized convergent structures of apo-con-T (14) as templates for the DG calculations, with the view that the generated structures that satisfied the distance constraints would be the best approximations of metal-loaded, energy-minimized structures. Initial structures that were generated from well resolved NOEs allowed several degenerate resonances to be assigned. Remaining NOEs were inputted as ambiguous constraints that take into account all of the possible interactions in an unbiased manner.
Identification of the Positions of the Metal Ions in con-T-Docking of the metal ions in con-T and further energy minimization-based refinement of the metal-bound structure was accomplished by the genetic algorithm-molecular dynamics simulation approach described previously (25). The initial coordinates for con-T were those determined by NMR in this study for the Ca 2ϩ ⅐con-T complex. The genetic algorithm was employed to determine the initial positions of the Ca 2ϩ ions by searching through the O-O midpoints that were within 6 Å of each other, using all oxygen atoms in the midpoint calculation. A total of 150 midpoints was found. The lowest Amber (26) energy structure, verified also by a systematic search, was subjected to a Particle Mesh Ewald (27) molecular dynamics simulation. For this simulation, the peptide was solvated in a 9.0-Å box of TIP3P water. The H 2 O, Na ϩ , and Ca 2ϩ were then energy-minimized at constant volume as described (25). Next, a Particle Mesh Ewald molecular dynamics simulation was performed on Na ϩ , Ca 2ϩ , and H 2 O for 100 ps, followed by energy minimization. Another Particle Mesh Ewald molecular dynamics simulation was then performed for 150 ps on side chain residues, the ions, and H 2 O for 150 ps, with fixed backbone, followed by data collection over 300 ps.
[ 3 H]MK-801 Binding Assays-Adult Sprague-Dawley rats were sacrificed by decapitation after being administered isoflurane vapor. Forebrains were removed and processed as described previously (9). Control experiments showed that use of this anesthetic did not affect the con-T inhibition isotherms.
Inhibition assays were performed in triplicate in a total volume of 500 l in 5 mM Na ϩ -Hepes, 4.5 mM Tris-Cl, pH 7.4, in the nominal absence of glycine and glutamate, with varying concentrations of peptide. The final concentrations of [ 3 H]MK-801 (New England Nuclear, 23.9 Ci/mmol) and spermine were 5 nM and 50 M, respectively. Binding was initiated by the addition of 300 l of membrane suspension containing 100 -200 g of protein, and incubation was carried out at room temperature for 2 h. Assays were terminated by rapid filtration over Whatman GF/B filter strips (presoaked in assay buffer containing 0.03% polyethyleneimine) using a Brandel (Gaithersburg, MD) 24-well cell harvester. Two 5-ml washes with cold buffer followed. Basal

RESULTS
NMR of Ca 2ϩ ⅐con-T-In a previous study, it was demonstrated that the CD spectrum of con-T underwent a small change to a more ␣-helical conformation consequent to the binding of Ca 2ϩ to this peptide (12). In the present study, the nature of this conformational alteration was further explored, with the ultimate goal of elucidating the Ca 2ϩ -bound structure of con-T. The one-dimensional 1 H NMR spectrum of Ca 2ϩ ⅐con-T showed well resolved resonances, similar to those observed earlier for apo-con-T (14). This allowed ready assignment of all of its proton resonances using a combination of TOCSY, DFQ-COSY, and NOESY experiments. An example of the NOESY spectrum of the ␣NH-␣CH fingerprint region of Ca 2ϩ ⅐con-T illustrating some of the sequential assignments is provided in Fig. 1. The complete proton resonance assignments of Ca 2ϩloaded con-T, obtained by use of this combination of methods, are listed in Table I.
Several of the proton resonances of con-T in the presence of Ca 2ϩ were shifted from those previously obtained for apo-con-T, and this is revealing in terms of possible differences in the structures of these peptides. A graphical summary of the chemical shift differences between Ca 2ϩ ⅐con-T and apo-con-T of the ␣NH and ␣CH backbone protons for each of the residues of this peptide are illustrated in Fig. 2. Also present in this figure are representative differences in the terminal (farthest sequentially removed from the ␤-carbons) nonexchangeable side chain protons (e.g. ⑀CH 2 of Lys, ␦CH 3 of Leu, ␦CH 2 of Arg, and so forth) between these two peptides. The largest chemical shift changes observed in the backbone protons occur in the residue 10 -18 region for the amide protons. Smaller differences are present in the ␣CH protons, except for that of Gla 10 , which undergoes a particularly large alteration. Of the terminal side chain protons, Gln 6 and Arg 13 show the most dramatic shifts.
Differences between the observed ␣CH chemical shifts and their random coil values provide valuable secondary structural information (28). Specifically, upfield shifts of these proton spins of Ն0.1 ppm from their random coil values indicate the presence of helical structure, if these values occur uninterrupted over four or more consecutive residues. The random coil values of each con-T residue were determined experimentally by obtaining the spectrum in 6 M urea. The differences between the ␣CH proton chemical shifts for Ca 2ϩ ⅐con-T from their same values in 6 M urea are illustrated in Fig. 3. Also plotted in Fig.  3, for comparison, are the values similarly obtained for apocon-T (14). In the case of Ca 2ϩ ⅐con-T, as with apo-con-T, residues 2-21 possess relative chemical shifts Ͼ0.1 ppm, indicating a high content of ␣-helical structure. This is especially evident in the region of residues 5-17, where chemical shift differences Ͼ0.6 ppm are found, indicating the presence of a significant population of ␣-helices. The ␣CH proton chemical shift of Gla 10 , however, is opposite in sign from that expected for a residue in an ␣-helical conformation. This was also observed in a previous work (13), in which it was attributed to the unknown random coil chemical shift value of the ␣CH proton of Gla 10 . However, our data, which show that this chemical shift is not anomalous, allow this interpretation to be abandoned. Because neither CD nor NOESY data demonstrate any serious disruption in the continuum of ␣-helix of this peptide, we attribute the observation of the positive chemical shift of the ␣CH proton of Gla 10 to the previously unknown consequence of direct Ca 2ϩ coordination to this residue. This is also a very likely explanation for the slightly smaller relative ␣CH proton chemical shifts for several other Ca 2ϩ ⅐con-T residues as compared with those of apo-con-T (Fig. 3).
Structure of Ca 2ϩ ⅐con-T-The structure of con-T in complex with Ca 2ϩ was primarily derived from DG calculations utilizing NOE connectivities present in this peptide. A summary of the NOE connectivities for intraresidue sequential (i ϩ 1) and medium (i ϩ 2, i ϩ 3, and i ϩ 4) NOEs used in the DG calculations are shown in Fig. 4. The high ␣-helical content of this peptide is readily noted from the strong backbone ␣NH-␣NH NOEs (nn(i ϩ 1)), the stronger (nn(i ϩ 1)) cross-peaks as compared with those of ␣CH-␣NH (␣n(i ϩ 1)) connectivities, and the presence of many (␣n (i ϩ 3)) and (␣n(i ϩ 4)) NOEs. In addition, three-bond ␣CH-␣NH coupling constants ( 3 J ␣N ) were approximately 5 Hz for all residues with well resolved ␣CH-␣NH correlation peaks; these data are also strongly suggestive of the presence of ␣-helical structure.
In addition to ␣n(i ϩ 3) NOE contacts observed for most pairs of residues, backbone-side chain NOE connectivities were found between many proton pairs (Table II), such as those present between Lys 18 -␣CH and Ala 21 -␤CH 3 , between Gln 6 -␣CH and both Leu 9 -␥CH and Leu 9 -␦CH 3 , and between Leu 9 -␣CH and Leu 12 -␦CH 3 , among many others. Furthermore, side chain-side chain NOE connectivities were found between a number of residues, including Gln 6 -␥CH 2 and Leu 9 -␦CH 3 , Arg 13 -NH and Val 17 -␥CH 3 , and Glu 16 -␥CH 2 and Asn 20 -␦CH 2 . All of this information strongly suggests a high structural order for con-T in complex with Ca 2ϩ , consistent with an ␣-helical nature.
The Ca 2ϩ -loaded structure of con-T was generated from 160 distance constraints, including 4 intraresidue, 70 sequential, 86 of medium range, and 5 group NOEs (Table II). The convergent structures from the apo-con-T study (14) were used as starting templates for the DG calculations to generate the 10 metal-loaded structures with the lowest distance violations this structure indicates that additional stabilization of the ␣-helix can originate from an extensive electrostatic network on one face of the helix, involving Gla 3 , Lys 7 , Gla 10 , Arg 13 , Gla 14 , and Lys 18 , and an appropriately spaced hydrophobic cluster of side chains, comprising Tyr 5 -Met 8 -Leu 9 -Leu 12 , on the opposite face. This situation also occurs in the apo-con-T structure (14).
The structure of Ca 2ϩ ⅐con-T is compared with the apocon-T (14) in Fig. 7. The structures are very similar in backbone conformations, and changes were observed in only a few of the backbone residues, viz, Gln 6 , Gla 10 , Arg 13 , and Gla 14 . These changes are noted in NOE data of Table II and from similar data for apo-con-T (14). For example, in apo-con-T, there are no side chain-side chain NOE connectivities between Gla 10 and Gla 14 (14), whereas in Ca 2ϩ ⅐con-T, NOEs are seen between Gla 10 HG and Gla 14 HG and between Gla 10 HG and Gla 14 HN.
Locating the Ca 2ϩ Ions in con-T-With the establishment of the Ca 2ϩ -loaded con-T structure, it became of interest to attempt to locate Ca 2ϩ ions in their binding site(s). For this purpose, a genetic algorithm-molecular dynamics simulation procedure was employed as described (25), using a FIG. 2. Proton chemical shift (␣CH, ␣NH, and side chain) differences between apo-con-T and Ca 2؉ -loaded con-T. The chemical shifts plotted were obtained by subtraction of the apo-con-T spectrum from that of Ca 2ϩ ⅐con-T for each of the proton groups illustrated. The particular side chain protons chosen were those that were at terminal and nonexchangeable locations. The buffer was 10 mM NaBO 3 , 100 mM NaCl, pH 6.5, in the absence of Ca 2ϩ or in the presence of 40 mM Ca 2ϩ at 5°C. The final con-T concentration was 2 mM. Ⅺ (black), ␣NH protons; छ (red), ␣CH protons; E (green), side chain protons.  The chemical shifts were determined in 10 mM NaBO 3 , 100 mM NaCl, pH 6.5, at a peptide concentration of 2 mM and a Ca 2ϩ concentration of 40 mM. The chemical shifts cited are relative to 4,4-dimethyl-4-silapentane-L-sulfonate, which was set to 0 ppm. The Ca 2ϩ ⅐Conantokin-T Complex 2Ca 2ϩ ⅐con-T binding model with 1 Na ϩ ion as the counterion to balance charge. An average structure was generated from the resulting coordinates and is illustrated in Fig. 8. This structure shows Ca 2ϩ ions located at one site, containing the side chains of Gla 10 and Gla 14 , with another at Gla 3 and Gln 6 . Gla 4 is not suitably located to provide such a donor site for Ca 2ϩ . In this model, two oxygen atoms from both Gla 10 and Gla 14 coordinate one Ca 2ϩ ion, and two oxygen atoms from the ␥-carboxylate of Gla 3 and one from the side chain carbonyl atom from Gln 6 are donor groups for binding of the other Ca 2ϩ . During the course of these calculations, energy minimization was conducted to derive the final structure with Ca 2ϩ ions in their appropriate locations. In comparing the two models, viz, the NMR-derived structure (Fig. 7) and that further refined through Ca 2ϩ docking and energy minimizations (Fig. 8), it is clear that no major differences in backbone or side chain conformations exit. The most notable changes occur in the ring orientation of Tyr 5 and the side chain positioning of Lys 7 , along with other minor differences. None of these, however, alter the face of the helix on which these residues reside, and none change the conclusions regarding the ␣-helix stabilizing forces. On the whole, the differences observed in the peptide structures revealed by the two models are surprisingly small. These are not further elaborated upon herein because the uncertain- FIG. 4. The sequential and interresidue backbone NOEs observed for Ca 2؉ ⅐con-T. The thickness of the bars are a qualitative measure of the strengths of the NOE cross-peaks. An asterisk (*) indicates that the assignment of the crosspeak was ambiguous due to resonance overlaps. The buffer was 10 mM NaBO 3 , 100 mM NaCl, pH 6.5, at 5°C. The final con-T concentration was 2 mM. The Ca 2ϩ ⅐Conantokin-T Complex ties in deriving the models themselves from these disparate approaches are likely as large as the small differences in positions of flexible side chains between them. Instead, it is stressed that the two models are very similar and provide a very good set of models with which to explore structure-function relationships of this peptide. con-T Variants-To test the importance of side chains found to be critical in stabilization of Ca 2ϩ binding to con-T, several variant peptides were constructed, and their properties were assessed. This group consisted of changes of all Gla residues and one other residue, Gln 6 , that was identified as a possible contributor to stabilization of Ca 2ϩ in the molecular dynamics simulation. In all cases, Ala was the amino acid that was substituted, because Ala would favor ␣-helix stability (29) but would not provide side chain atoms that could coordinate Ca 2ϩ .
The effects of these substitutions on the stability of the apo-con-T helix and that of the Ca 2ϩ -loaded peptide, as well as the quantitative ability of Ca 2ϩ (C 50 values) to induce the ␣-helical conformation in the variant peptides, are summarized in Table III. Representative CD titrations with Ca 2ϩ of con-T, con-T[Gla 3 -Ala] and con-T[Gla 14 -Ala] are provided in Fig. 9. The only variation in this series that resulted in significant destabilization of the apo-con-T ␣-helix was that of Gla 4 -Ala, where approximately one-half of the ␣-helical content was observed as compared with the wild-type peptide (Table III). However, the C 50 values characterizing the Ca 2ϩ -induced conformational transition in this peptide were similar for all peptides, except for those of con-T[Gla 10 -Ala], con-T[Gla 14 -Ala], and con-T[Gla 10 -Ala/Gla 14 -Ala], where substantially higher concentrations of Ca 2ϩ were required to induce their Ca 2ϩ -dependent conformations. Furthermore, in these latter cases, the conformational change appeared to be very small and of marginal significance, and it was opposite in character to that a Backbone-backbone, backbone-side chain, and side chain-side chain.
The Ca 2ϩ ⅐Conantokin-T Complex induced in the other mutant peptides, the final extents of which were similar to those of con-T and con-T[Gla 3 -Ala] (Fig. 9). The observation that the ␣-helicity of con-T[Gla 10 -Ala], con-T[Gla 14 -Ala], and con-T[Gla 10 -Ala/Gla 14 -Ala] decreases as the peptide is titrated with Ca 2ϩ can be due to two possible factors: 1) electrostatic screening of favorable side chain-side chain or backbone-side chain interactions, and/or 2) the coordination of metal to individual Gla residues that leads to a specific disruption of critical charge-charge interactions. These possibilities were examined by investigating the ionic strength dependence on ␣-helicity for the peptides con-T[Gla 10 -Ala], con-T[Gla 14 -Ala], and con-T[Gla 10 -Ala/Gla 14 -Ala] using NaCl. The results revealed that nonspecific ionic strength changes alone cannot account for the decrease in ␣-helicity for these peptides. Therefore, the Ca 2ϩ -induced decreases are primarily due to specific coordination with the remaining Gla residues.
The binding isotherms of Ca 2ϩ to each of the con-T mutant peptides were established by Ca 2ϩ -specific electrode titrations. Examples of the titration data obtained are illustrated in Fig.  10. In most cases, the binding data were fit to a model with two independent sites, one strong site with a K d within the range of 0.2-0.5 mM and one weaker site of K d approximately 10-fold higher. Exceptions have been observed in the cases of con-T[Gla 10 -Ala], con-T[Gla 14 -Ala], and con-T[Gla 10 -Ala/Gla 14 -Ala], wherein the data could be fit several different models of very weak Ca 2ϩ binding. For these peptides, the binding characteristics were determined through Michaelis-Menten fits, which showed that the amount of free Ca 2ϩ required to reduce the binding to 50% was approximately 1-4 mM. These results indicate that only the strong Ca 2ϩ sites have been eliminated by mutations at Gla 10 and Gla 14 and/or that these two mutations led to peptides that required different Ca 2ϩ binding models to satisfy the experimental data. Lastly, the effects of the mutations in con-T on the ability  of the resulting peptides to inhibit MK-801 binding to open rat brain NMDA receptor channels in the presence of exogenous spermine has been examined. The neuronal membranes were washed and thus contained only low levels of glutamate and glycine, but these ligands were present at sufficient concentrations to lead to significantly increased ion channel opening upon addition of spermine. The amount of [ 3 ]MK-801 binding to the membranes, as a function of the concentration of con-T-derived peptides, is shown for con-T and for con-T[Gla 10 -Ala/Gla 14 -Ala] in Fig. 11. Values of the IC 50 for peptide inhibition have been calculated from the concentration midpoints of the differences between the basal level of MK-801 binding to the peptide and that induced by spermine at a peptide concentration of zero. The values obtained are summarized in Table III. Very large increases in the IC 50 value were seen for con-T[Gla 3 -Ala] and con-T[Gla 4 -Ala], and a more modest increase was observed for con-T[Gla 10 -Ala] and con-T[Gla 10 -Ala/Gla 14 -Ala]. Thus, major binding determinants for the peptide to its site on the membrane are provided by Gla 3 and Gla 4 of con-T. DISCUSSION A variety of divalent cations, including Ca 2ϩ and Mg 2ϩ , alter the CD properties of con-T (12) in a manner suggesting that the relative ␣-helical content of the peptide is increased. The Gla residues of this peptide, which serve as Ca 2ϩ binding loci when present in blood coagulation proteins and in proteins and peptides present in bone, are the strongest candidate residues for metal ion coordination in con-T. Titrations by ion-specific electrode methods reveal that con-T has one strong Ca 2ϩ binding site, and possibly a second weaker site of interaction (12). A previous study has detailed the backbone FIG. 7. Comparison of the average apo-con-T and Ca 2؉ ⅐con-T conformations. The backbones of the peptides have been superimposed. The structure of apo-con-T was taken from Ref. 14.
FIG. 8. The Ca 2؉ -bound refined structure of con-T. A total of two Ca 2ϩ ions (black) were docked in this structure using the genetic algorithm, beginning with the NMR-derived coordinates of con-T loaded with Ca 2ϩ . The structure was then refined by molecular dynamics simulation. In this illustration, the side chains (from the ␤-carbons) of all amino acid side chains of con-T are illustrated, and certain key residues are labeled. All side chain carbon atoms are colored orange, nitrogen atoms are blue, oxygen atoms are red, and the sulfur atom of Met 8 is green. A, space-filling representation. B, relationships between side chain residues implicated in binding to Ca 2ϩ and the Ca 2ϩ ions. This view also includes Gla 4 , which does not appear to be involved in Ca 2ϩ binding. The orientation is as in A. The representations in A and B are not on the same scale. c Determined from a Michaelis-Menten type plot, which provides the Ca 2ϩ concentration at 50% binding of Ca 2ϩ , independent of the binding model. and side chain conformations of con-T and concluded that a significant population of molecules were end-to-end ␣-helices (14). Despite this, changes occurred in both the NMR and in the CD spectra of con-T as a result of addition of Ca 2ϩ that indicated that a higher content of ␣-helix resulted from this interaction (Figs. 2-4 and 9 and Table II). As shown in Fig. 2, these changes did not appear to be in backbone conformations but rather in side chain orientations, and such types of changes should not have such dramatic influence on the ␣-helical content. It is most likely that the population of molecules in the ␣-helical conformation increases in the presence of Ca 2ϩ , and the side chain reorientations reflect the optimization of Ca 2ϩ binding sites.
The NMR-derived solution conformation of Ca 2ϩ -loaded con-T has been determined by methods similar to those published for the apo-con-T structure (14). However, in this case, only DG calculations of backbone-backbone NOEs and of backbone-side chain and side chain-side chain NOE constraints were employed to calculate the structure at this stage. Because the exact position of the bound Ca 2ϩ ion(s) was unknown, full energy minimization was not performed in these structure calculations. This is because the exclusion of the bound ions from the energy minimizations would have resulted in nonoptimal definition of metal-induced conformation. However, the major differences in side chain conformation that result from Ca 2ϩ binding to con-T are centered at Gln 6 , Gla 10 , Arg 13 , and Gla 14 . Such changes are consistent with the binding of Ca 2ϩ to Gla 10 and Gla 14 . In apo-con-T, the guanidino group of Arg 13 appears to structurally mimic the role of Ca 2ϩ in the metal-loaded form, reducing electrostatic repulsion between the Gla 10 and Gla 14 side chains. This model is predicted from our previous structural work (14) and is consistent with the observation that con-T[Arg 13 -Ala] reduces the apo-␣-helicity by Ͼ50% (data not shown). Therefore, binding of a metal ion to the tight site concomitant with displacement of the Arg 13 side chain is consistent with a significant chemical shift in this residue. Gla 3 and that of Gln 6 are in spatial proximity in the metalpeptide complex, either as the result of hydrogen bonding or because they are directly coordinated with the metal. The chemical shift changes of the Gla 6 -⑀NH 2 side chain protons in the Ca 2ϩ ⅐con-T complex can be perhaps explained by the amide carbonyl serving as a ligand for Ca 2ϩ . In addition, disruption of the Gln 6 /Gla 10 apo-con-T hydrogen bond may also contribute to the chemical shift changes.
The roles of the individual Gla residues of con-T as donors for binding of Ca 2ϩ and their abilities to function in promotion of the Ca 2ϩ -induced conformational change in con-T have been directly assessed through study of variant con-T-based peptides. The Ca 2ϩ binding isotherm displays one clear, tight binding site of approximately 0.25 mM (12), and additional evidence suggests that at least one additional, weaker site is present, with a K d Ͼ10-fold higher than the strong site (Table  III). Alteration of Gla 3 or Gln 6 to Ala residues preserves the binding at the tight Ca 2ϩ site, but diminishes Ca 2ϩ binding at the weaker site (Table III). This result is in concert with the model of the Ca 2ϩ -docked con-T (Fig. 8), which shows that these two residues are capable of coordinating Ca 2ϩ . Alteration of Gla 4 to Ala did not affect Ca 2ϩ binding, again in agreement with this model, which predicts that Gla 4 would not coordinate Ca 2ϩ . Both this model and that derived from NMR measurements (Fig. 7) place Gla 4 on the opposite side of the face of the ␣-helix that appears to house the Ca 2ϩ binding loci. On the other hand, alteration of Gla 10 and/or Gla 14 to an Ala residue resulted in Ca 2ϩ binding that was so weak that a unique binding model could not be fit. Therefore, for these peptides, an average binding constant was estimated from a Michaelis-Menten fit of the binding data (Table III). The value obtained was at least 15-fold larger than that displayed for the strong Ca 2ϩ binding site of native con-T. Thus, we propose that Gla 10 and Gla 14 serve as coordination sites for the tightly bound cation.
The only mutant that affected the stability of the apo-con-T ␣-helix was con-T[Gla 4 -Ala], which reduced the population of ␣-helical molecules to one-half of their values in wild-type con-T and the other mutants tested (Table III). Thus, despite the change of Gla 4 to a residue that in itself should not be disruptive to the ␣-helical character of this peptide, destabilization of the ␣-helix nonetheless occurred. This conclusion is in agreement with our analysis of the apo-con-T structure, in which Gla 4 was predicted to be an important capping residue at the amino terminus and to interact favorably with the ␣-helix macrodipole (14). However, the population of con-T[Gla 4 -Ala] molecules in the ␣-helical conformation substantially increased upon addition of Ca 2ϩ , as was the case for wild-type con-T and most of the other mutants. The notable exceptions to this were con-T[Gla 10 -Ala], con-T[Gla 14 -Ala], and con-T[Gla 10 -Ala/Gla 14 -Ala], wherein addition of Ca 2ϩ did not increase the population of molecules in the ␣-helical conformation and, in fact, somewhat decreased this distribution. More quantitative comparisons of the C 50 values for Ca 2ϩ required to induce overall ␣-helix transitions were similar for wild-type con-T, con-T[Gla 3 -Ala], con-T[Gla 4 -Ala], and con-T[Gln 6 -Ala]. On the other hand, the same values for ␣-helix destabilization in con-T[Gla 10 -Ala], con-T[Gla 14 -Ala], and con-T[Gla 10 -Ala/Gla 14 -Ala] were approximately 10-fold higher. Thus, the Ca 2ϩ site localized at Gla 10 and Gla 14 is essential for stabilization of the ␣-helix in con-T.
Finally, the effects of these mutations on the bioactivity of con-T have been examined through the effects of these peptides on the spermine-induced [ 3 H]MK-801 binding to washed rat neuronal membranes. The efficacy data of Table  III are in general agreement with previous results using con-G mutants (11) in that mutations at Gla 3 and Gla 4 greatly reduced the ability of these peptides to inhibit the spermine-induced potentiation of MK-801 binding to open membrane channels. Mutations at Gla 10 and Gla 14 in con-T, as was the case with con-G, did not affect this property to the same extent, although the effect of the Gla 10 mutation was significant (Table III). Thus, it appears that the nature of the amino acid residues at the amino terminus of these peptides may be a critical property in its bioactivity, with Ca 2ϩ binding at other loci perhaps stabilizing critical conformations. This could explain the reasons for the tight homology of Gly 1 -Glu-Gla-Gla at the amino terminus of the currently known conantokin structures. Alternatively, the cation-induced ␣-helices might simply be a reflection that such structures can be formed from other peptide-ligand interactions, such as might occur when these peptides interact with groups at or on the membrane surface.
In conclusion, we have provided experimentally based models of the solution structure of Ca 2ϩ -loaded con-T and identified some groups most likely involved in coordination of the metal cations. The major Ca 2ϩ binding site is coordinated by side chain carboxylate oxygen atoms from Gla 10 and Gla 14 , and Ca 2ϩ binding at this location serves to dramatically stabilize the ␣-helical conformation of con-T. The Gla 3 ␥-carboxylate group in cooperation with the Gln 6 side chain carbonyl moiety likely coordinate a second weakly bound Ca 2ϩ site. Gla 4 does not appear to function in this regard. The major roles of Gla 3 and Gla 4 appear to reside in their abilities to serve as major determinants of the bioactivity of con-T on NMDA receptor ion channels. These results have provided major insights into structure-function relationships of this neuroactive naturally occurring polypeptide and will serve as a focal point in design of new con-T-based active molecules.