Mechanism of Calcite Crystal Growth Inhibition by the N-terminal Undecapeptide of Lithostathine*

Pancreatic juice is supersaturated with calcium carbonate. Calcite crystals therefore may occur, obstruct pancreatic ducts, and finally cause a lithiasis. Human lithostathine, a protein synthesized by the pancreas, inhibits the growth of calcite crystals by inducing a habit modification: the rhombohedral {10 1̄4} usual habit is transformed into a needle-like habit through the {112̄0} crystal form. A similar observation was made with the N-terminal undecapeptide (pE1R11) of lithostathine. We therefore aimed at discovering how peptides inhibit calcium salt crystal growth. We solved the complete x-ray structure of lithostathine, including the flexible N-terminal domain, at 1.3 Å. Docking studies of pE1R11 with the (101̄4) and (11 2̄0) faces through molecular dynamics simulation resulted in three successive steps. First, the undecapeptide progressively unfolded as it approached the calcite surface. Second, mobile lateral chains of amino acids made hydrogen bonds with the calcite surface. Last, electrostatic bonds between calcium ions and peptide bonds stabilized and anchored pE1R11 on the crystal surface. pE1R11-calcite interaction was stronger with the (11 2̄0) face than with the (10 1̄4) face, confirming earlier experimental observations. Energy contributions showed that the peptide backbone governed the binding more than did the lateral chains. The ability of peptides to inhibit crystal growth is therefore essentially based on backbone flexibility.

Biomineralization has occurred for millions of years (1). Half of these biogenic minerals contain calcium (2); for example, the teeth of the sea urchin contain calcium carbonate (CaCO 3 ) in the form of calcite (the most common polymorph existing in nature), and primitive mollusks have aragonite spicules. In humans, biomineralization is observed not only during skeletal formation, but also in biological fluids generally supersaturated with a calcium salt, such as oxalate in urine, phosphate in saliva, or carbonate in pancreatic juice. As a result, calcium salt crystals form spontaneously. While beneficial for tooth or bone mineralization, precipitation of calcium salts can be extremely harmful in fluids because it leads to the formation of stones and to the development of a lithiasis.
Various structural motifs emerge at calcite surfaces. Adsorption of additives, i.e. foreign substances, takes place either on growth sites or along the steps that spread over the faces or even onto the flat area of the faces. Since growth proceeds mostly through kinks, i.e. defects, blocking these sites is sufficient to hinder crystal growth. In some cases, adsorption is irreversible. In other cases, however, adsorption of additives is temporary and reversible: the oncoming growing units continuously repulse the additive molecules, which try to incorporate into the crystal lattice, in front of faces in growth. Mineralization must therefore be controlled at the molecular level (3). This control is ensured mainly by macromolecules, essentially proteins. Since the discovery of dentin (4), an acidic protein in vertebrate teeth, many other proteins involved in crystal growth control in humans have been described (for a review, see Ref. 5).
Among these proteins, lithostathine has been well documented. It is a protein of 144 amino acids that is produced by acinar cells of the pancreas and secreted into pancreatic juice. The structure of human lithostathine has been reported at 1.5-Å resolution (6); no structural information was obtained on the N-terminal domain (residues 1-13), whereas the C-terminal part, which belongs to the C-type lectin superfamily, was fully characterized. Yet, lithostathine is very susceptible to proteolysis. It produces a soluble N-terminal undecapeptide (pE 1 R 11 ) and a C-terminal form of 133 amino acids that precipitate and form fibrils (7). The specific association of lithostathine with CaCO 3 suggested that it could be involved in the control of crystal growth (8). This suggestion was confirmed by Bernard et al. (9). However, several other groups published conflicting results depending on how crystal growth measurements in solution had been made (10,11). Looking at the calcite crystal morphology in the presence of lithostathine, we previously observed that lithostathine changes the growth of calcite crystal by inducing a habit modification in vitro: the rhombohedral {10 1 4} 1 usual habit is first transformed into a subcubic * This work was supported by Grant ACC-SV 5 from the Ministère de l'Enseignement Supérieur et de la Recherche. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The habit and finally into a needle-like habit with the main {11 2 0} form (12). The synthetic form of pE 1 R 11 also gave the same calcite morphology changes. It required a concentration of only ϳ100 times higher than that of the whole protein, in the same range as for CBP1 (0.1-0.5 mM) (13).
The mechanisms of protein-crystal interactions are still poorly understood. Nothing is known about the protein groups involved in the binding to the crystal surface or about the relative energy of these interactions. The purpose of this study was therefore to examine the main characteristics of a good protein inhibitor of calcium salt crystal growth.

X-ray Structure Determination
Crystallization and Heavy Atom Search-The purification and crystallization of human lithostathine have been previously described (14). Crystals belong to the hexagonal P6 5 space group with a ϭ b ϭ 48 Å and c ϭ 111 Å and contain one molecule of lithostathine/asymmetric unit. Soaking crystals for 24 h in a solution containing 0.5 mM p-chloromercuribenzenesulfonic acid provided the usable heavy atom derivative.
Structure Determination-A 1.5-Å resolution native data set and a 1.6-Å resolution p-chloromercuribenzenesulfonic acid derivative data set were collected from single crystals on a MarResearch imaging plate system mounted on the wiggler line W32 at the Laboratoire pour l'Utilisation du Rayonnement Electromagnétique synchrotron radiation facility (Orsay, France). The structure was solved by using SIRAS phases and refined to 1.55-Å resolution with X-PLOR (15), applying the parameter set of Engh and Huber (16). The final model includes 138 water molecules and 131 out of the 144 protein residues corresponding to the C-type lectin domain of lithostathine (6).
Model Building and 1.3-Å Refinement-As no interpretable density data for N-terminal residues 1-13 were obtained, a 1.3-Å resolution data set was subsequently collected on a MarResearch imaging plate system at the Hamburg synchrotron radiation facility (Germany). Data were processed with DENZO and SCALEPACK (17). The atomic coordinates resulting from the 1.55-Å resolution refinement were used as the starting model. Energy was first minimized with REFMAC (18) using individual isotropic B-factors. The (2mFo Ϫ DFc) and (mFo Ϫ DFc) electron density maps, calculated with SIGMAA (18), clearly showed an extended density around residue 14, into which it was possible to build the N-terminal part of the protein (residues 1-13). Residues 9 -11 were built as alanine residues since the density was not interpretable for building the side chains. Finally, a new refinement procedure followed by electron density map calculation was carried out with the model including the complete protein chain. After energy minimization using REFMAC, the R-factor and the free R-factor were only 17.4 and 18.6%, respectively.

Peptide Synthesis, Purification, and Characterization
The N-terminal undecapeptide pyro-EEAQTELPQAR (pE 1 R 11 ) was assembled according to the method of Barany and Merrifield (19) on 4-hydroxymethylphenoxymethylcopolystyrene and 1% divinylbenzenepreloaded resin (0.5-0.65 mmol; Perkin-Elmer) on an automated synthesizer (ABI 433A, Perkin-Elmer). To avoid derivatives with deletion, the N-terminal extremities without Fmoc 2 were capped with a mixture of 4.75% acetic anhydride (Merck), 6.25% 2.0 M n,N-diisopropylethylamine, 1.5% 1 M 1-hydroxybenzotriazole (Perkin-Elmer), and 87.5% N-methylpyrrolidone (Perkin-Elmer). The peptides were deprotected and removed from the resin with trifluoroacetic acid complemented with 10% methylphenylsulfide (Merck) and 5% ethanedithiol (Merck) as scavenger and with Fmoc on the N-terminal residue. Each deprotection step was monitored with a conductivity device. Purification was done on a Beckman HPLC apparatus (Gold system) equipped with a C8 reverse phase column (10 ϫ 250 mm; Merck). Buffer A was water with 0.1% trifluoroacetic acid, and buffer B was acetonitrile with 0.1% trifluoroacetic acid. The elution gradient consisted of increasing the buffer B ratio from 20 to 40% in 40 min with a 2-ml/min flow rate. HPLC analysis was done with a C8 reverse phase column (4 ϫ 125 mm; Merck) with same buffers, but with a gradient of 10 -50% buffer B in 40 min and with a 0.8-ml/min flow rate. Electrospray mass spectrometry was carried out with a single quad PE-SCIEX API 150ex (Perkin-Elmer).
Amino acid composition was analyzed on a Beckman Model 6300 amino acid analyzer.

Circular Dichroism Measurements
CD spectra were recorded at 37°C with a 50-m path length cell from 260 to 178 nm with a MARK VI UV CD spectrophotometer (Jobin-Yvon, Longjumeau, France). The instrument was calibrated with (ϩ)-10-camphorsulfonic acid. A ratio of 2.1 was found between the positive CD band at 290.5 nm and the negative band at 192.5 nm. Data were collected at 0.5-nm intervals with a scan rate of 1 nm/min. CD spectra are reported as ⌬⑀/amide. The samples were in 20 mM phosphate buffer (pH 7). Protein concentrations (ϳ1 mg/ml) were determined by the Beckman Model 6300 amino acid analyzer.

Molecular Dynamics Simulation Experiments
Initial Conditions-Calcite unit cell parameters were taken from the literature (20): space group R3 c (rhombohedral system), Z ϭ 6, a ϭ 4.990 Å, and c ϭ 17.002 Å. For partial charges on calcite atoms, the values of Catti et al. (21) obtained from ab initio calculations were used: ϩ1.865 e Ϫ on calcium, ϩ1.075 e Ϫ on carbon, and Ϫ0.980 e Ϫ on oxygen. As the overall surface charge of calcite is neutral at physiological pH (22), no specific surface charge or surface ionic species like -CO 2 Ϫ or -CO 2 H groups were taken into account. Periodic boundary conditions were applied on a simulation box to get rid of boundary effects. The simulation box size was chosen to allow the undecapeptide to move freely on the calcite surface. To avoid pE 1 R 11 adsorption under the top of the box where periodic boundary conditions replicate the calcite interface, a 100-Å box height was used. This height gave enough space for this undecapeptide to move over the calcite surface. Under these conditions, the (10 1 4) and (11 2 0) surface areas were 1411.52 Å 2 (35 unit cells) and 2057.20 Å 2 (42 unit cells), respectively. Finally, a slice depth of three unit cells was necessary to properly evaluate the interactions of the undecapeptide with calcite faces.
Force Field Description-DREIDING 2.21 (23) was used to determine the crystal lattice energy of calcite. We also employed DREIDING 2.21 to describe bonding (stretching, bending, torsion, and inversion potentials) and nonbonding (6 -12 Lennard-Jones-like van der Waals, coulombic, and 10 -12 Lennard-Jones-like hydrogen bond potentials) interactions. Long-range electrostatic interactions were computed using Ewald summation. Partial charges were computed using the qEq equilibration method, a geometry-sensitive charge assignment method implemented in the Cerius2 program package (MSI, San Diego, CA). To mimic the presence of the C-terminal part of lithostathine attached to the N-terminal undecapeptide through the COO Ϫ group of Arg 11 , this functional group was replaced by a -(CϭO)C(CH 3 ) 3 group.
Docking Studies-Molecular dynamics simulations were carried out in the NVT ensemble (molecule number, volume, and temperature are kept constant during the simulation). They were performed with Ce-rius2 running on a Silicon Graphics R5000PC workstation. The simulation box contained a fixed calcite interface, representing either the (10 1 4) or the (11 2 0) face, and the pE 1 R 11 peptide, whose coordinates were taken from x-ray data, without the carbohydrate moiety. A 0.5-fs integration step was used. Temperature was kept constant at 310 K using the Nosé-Hoover equation of motion with a 0.3 relaxation factor value that guaranteed a proper convergence of the dynamics simulation. No solvent molecules were added in our model. The undecapeptide was then introduced into the box and set at a 20-Å distance from the calcite interface. A typical dynamics simulation included a 20-ps relaxation period of the system during which the undecapeptide moved toward the calcite surface and reached rather stable location and conformation. It was followed by a simulation period of several tens of ps. Snapshots taken from the dynamics simulation trajectory files were then analyzed. The interaction energy between pE 1 R 11 and the calcite surface was computed as the difference between the energy of the whole system and the energy of pE 1 R 11 alone plus the energy of the calcite crystal. Ewald parameters were kept constant for all the calculations. As pE 1 R 11 coverage of the calcite surface could vary from one simulation snapshot to another, we expressed the undecapeptide-calcite interaction energy E i (J/mol of unit cell) in terms of microscopic adhesion work W i (J/m 2 ) as proposed by Lin et al. (24): W i ϭ ϪE i n 2 /N A S hkl , where S hkl is the interface area, N A is the Avogadro number, and n 2 is the number of molecules each covering a contact area S c (S hkl ϭ n 2 S c ).

Quality of the X-ray Model and Overall Protein Structure-
The final refinement process was carried out with SHELX (25). Anisotropic temperature factor refinement led to subsequent R-factor and free R-factor values of 13.2 and 15.9, respectively, for all 32,058 reflections in the resolution range of 8.0 to 1.3 Å. X-ray data are provided in Table I. A Ramachandran plot (26) showed that 88.1% of the residues were in the most favored regions, and none of the non-glycine residues were in the disallowed regions (data not shown). The final model included 135 water molecules and the entire protein (144 residues) (Fig. 1A). Surprisingly for such a small protein, the complete high resolution structure of human lithostathine showed two domains: a well organized globular C-terminal domain (residues 14 -144) (6) and a flexible N-terminal region (residues 1-13) that do not interact. The C-terminal domain belongs to the C-type lectin superfamily, although it does not bind carbohydrate (6). It is separated from the N-terminal domain by the C-14 -C-25 disulfide bridge. The N-terminal domain is a 13-residue peptide chain that stretches out of the heart shape of the C-type lectin domain. Three residues (positions 9 -11) are involved in a helix turn motif that represented the only secondary element of this domain. As a result, the N-terminal domain was much more agitated in the crystal (averaged B-factor of 36 Å 2 ) than the rest of the protein (averaged B-factor of 22 Å 2 ). This is consistent with the fact that we needed x-ray data at 1.3-Å resolution to build this domain.
Earlier studies have shown that lithostathine is O-glycosylated on Thr 5 (27). During structure refinement, we observed a large electron density around Thr 5 , clearly indicating the presence of O-linked sugars to the Thr 5 side chain. This enhanced density of synthetic pE 1 R 11 was confirmed by semi-empirical quantum chemical method calculations using the program MOPAC from the Cerius2 package (data not shown). This carbohydrate moiety was divided into two carbohydrate chains perpendicular to the protein backbone (Fig. 1A). Furthermore, because of the high quality of our electron density maps (Fig.  1B), it was possible to assign our isoform to isoform K (27) and to build three sugar residues (NeuAc(␣2-6)GalNAc(␤1-3)Gal) out of the four forming the carbohydrate chain. In addition, mass spectroscopy results for the crystallized protein (17,199 Da) agreed very well with the theoretical weight of this isoform (17,194 Da) (data not shown).
CD Experiments-That the N-terminal domain remained disordered in the crystal prompted us to test the behavior of the synthetic counterpart, pE 1 R 11 , which is also an inhibitor of calcite crystal growth, under physiological conditions (pH 7, 37°C). The CD spectrum (Fig. 2) was typical of a random coil structure (28) and strengthened our x-ray data. The negative band near 200 nm had an intensity in the range of those observed with model peptides with no steric constraints or internal hydrogen bonds. Random coil CD spectra were characterized by an intense negative band at 200 nm due to -* transition and low intensity bands near 210 nm due to n-* transitions (28). No typical secondary structures could be deduced from this spectrum. Similar results were obtained in trifluoroethanol or SDS (data not shown). Both x-ray and CD experiments therefore suggest that pE 1 R 11 is a highly flexible molecule that can display many configurations depending on the environment.   ing oxygen atoms. In addition, row spacing is different: 6.358 Å on the (10 1 4) face and 4.99 Å on the (11 2 0) face.
Docking Studies-A qualitative analysis of the dynamics simulation trajectories showed that, as pE 1 R 11 approached the calcite face (10 1 4) or (11 2 0), it gradually unfolded (data not shown). Simultaneously, it attached to the calcite surface through electrostatic bonds between its oxygen atoms from CϭO groups and calcium ions regularly emerging from the surface and by means of hydrogen bonds between its hydrogen atoms from N-H or O-H groups and oxygen atoms from the carbonate ions. This dual bonding strengthened the pE 1 R 11calcite surface interaction. Both types of bonds participated in the global interaction. Interestingly, we noticed that electrostatic bonds do not break during the course of simulation studies, which indicates very strong binding. Indeed, once the Ca surface -O undecapeptide bonds were established, the undecapeptide stayed rather still, only making or breaking hydrogen bonds. With such an inertia, the undecapeptide is a good crystal growth inhibitor candidate: it sticks to the calcite surface long enough to impede the attachment of crystal growth units and thus reduces the face growth rate. Furthermore, dynamics simulation assays with pE 1 R 11 surrounded by water molecules showed a similar tendency, but convergence time was beyond our computer time capacity (data not shown).
Interaction of the pE 1 Table  II shows the energy values obtained after adsorption of pE 1 R 11 on the (10 1 4) face. Four features could be deduced from these results. First, the nature of the pE 1 R 11 -calcite interaction was mostly coulombic (Ϫ175.4 kcal⅐mol Ϫ1 ), whereas van der Waals and hydrogen bonding contribution played a minor role (Ϫ26 and Ϫ12.8 kcal⅐mol Ϫ1 , respectively). Second, these 15 strong bonds represented little more than 60% of the total energy interaction, although neither the Ca surface -OϭC undecapeptide bond nor the hydrogen bond contributions to the total energy prevailed. The remaining amount was due to weak attractive van der Waals and coulombic interactions. Third, Ca calcite surface -OϭC undecapeptide bond stability was due to their coulombic nature (Ϫ105.6 kcal⅐mol Ϫ1 ), whereas the stability of the five hydrogen bonds was due to both their coulombic and The two spectra were measured simultaneously from 260 to 178 nm. The absorption spectrum shows a positive band at 190 nm due to the amide chromophore. The intensity of this band was similar to that of any peptide at 1 mg/ml. The CD spectrum is typical of a random coil structure.
hydrogen nature (Ϫ37.5 and Ϫ11.4 kcal⅐mol Ϫ1 , respectively). Last, for all those bonds, the van der Waals contribution to the interaction energy with the (10 1 4) calcite interface was positive because the repulsive term of the van der Waals potential dominates its attractive term at the short distances involved.
Interaction of the pE 1 R 11 Peptide with the (11 2 0) Face- Fig.  4B shows a top view of the pE 1 R 11 peptide adsorbed on the (11 2 0) calcite interface. Anchoring of the peptide was ensured by 18 bonds that represent 70% of the total interaction energy: 9 hydrogen bonds (H undecapeptide -O calcite surface ) and 9 electrostatic bonds (Ca surface -O undecapeptide ). Hydrogen bonds were divided as follows: three with the peptide bond (H-N), Glu 6 , Ala 10 , and Arg 11 ; six with lateral chains, pyro-Glu 1 , Gln 4 , and Gln 9 ; and three with Arg 11 . For the electrostatic bonds, four came from the peptide bond (OϭC), Gln 4 , Glu 6 , Pro 8 , and Arg 11 ; and five came from lateral chains, pyro-Glu 1 , Glu 2 , Gln 4 , Thr 5 , and Gln 9 . In addition, five oxygen atoms settled between two calcium ions (Ca surface -O undecapeptide -Ca surface bonds). As Ca-O distances were similar (2.4 or 2.5 Å) whether one or two calcium atoms were involved, the electrostatic attraction energy of the Ca-O-Ca bonds was roughly double the energy of the Ca-O bond. Bond distances were similar to those observed with the (10 1 4) face. Table III summarizes the energy values obtained. The same observations as for the interaction of pE 1 R 11 with (10 1 4) applied (see Table II). However, the total energy measured was greater with the (11 2 0) face than with the (10 1 4) face (Ϫ302.1 versus Ϫ214.2 kcal⅐mol Ϫ1 ) for two reasons. First, the calcium row spacing is larger on the (11 2 0) face than on the (10 1 4) face, Ca surface -O undecapeptide -Ca surface bonds are more numerous on the (11 2 0) face. Second, as this type of bond has a larger electrostatic interaction energy than do Ca surface -O undecapeptide bonds and as the total interaction energy is mostly of electrostatic nature, the pE 1 R 11 peptide binds more strongly on the (11 2 0) face than on the (10 1 4) face.
Coulombic Interactions and Microscopic Adhesion Work-Qualitatively, the interactions of the undecapeptide with the (10 1 4) and (11 2 0) faces were almost equivalent. Quantitatively, however, they differed. Indeed, because of row spacing, calcium ions are more accessible on the (11 2 0) face than on the (10 1 4) face. Consequently, coulombic interactions between a negatively charged atom of pE 1 R 11 and a calcium ion were greater on the (11 2 0) face (Ϫ213.4 versus Ϫ105.6 kcal⅐mol Ϫ1 ). By performing many simulations and analyzing several snapshots (data not shown), we calculated the mean microscopic adhesion work. It amounted to 350 mJ/m 2 for the interaction of pE 1 R 11 with the (10 1 4) face and 550 mJ/m 2 with the (11 2 0) face, i.e. 1.5 times greater. This result clearly shows a stronger attachment of pE 1 R 11 to the (11 2 0) face than to the (10 1 4) face. According to theories of crystal growth with additives (35,36), strongly attachment would lead to the appearance of the (11 2 0) face when pE 1 R 11 is present, thus confirming our previous experimental observations (12).

DISCUSSION
Our results show that the ability of peptides to inhibit crystal growth is essentially due to their backbone flexibility. Yet, until now, the theoretical model explaining the inhibition of crystal growth by proteins was based on lattice matching. Indeed, a neutral but ionic crystal surface like calcite shows a periodic array of localized binding sites for ions of opposite charges. This periodicity prompted several authors to speculate that proteins tying down such layers should adopt a regular ␤-sheet conformation. For instance, this is the case with acidic glycoproteins from mollusk shells (37,38) or poly-Asp (39), which adsorb strongly onto calcite surfaces. Other studies have suggested there might be bonds between calcium atoms and carboxylate groups belonging to small synthetic ␣,-dicarboxylic acids (40). CBP1, an artificial regular ␣-helical peptide, has been shown to bind to calcite surfaces, inducing morphological changes (13). It was therefore concluded that regular secondary structures, either of the ␤-sheet or ␣-helix type, are a prerequisite for the peptide to bind to crystal surfaces and to disturb their growth.
All proteins display dicarboxylic acids and regular secondary structures such as ␤-sheet or ␣-helix; but only a few are good inhibitors of crystal growth, thus ruling out the lattice matching model. It has become obvious that other factors such as stereochemical, charge-dependent, or electrical polarization ones are also required (41,42). The growth of crystals proceeds by a two-step mechanism: growth unit diffusion toward the surface and subsequent integration into the crystal lattice. We believe that inhibition of calcite growth by pE 1 R 11 , and more generally by any protein inhibitor, proceeds by a similar mechanism. First, the undecapeptide pE 1 R 11 unfolds as it approaches and then diffuses over and binds to the calcite surface. This diffusion means that interpeptide and water-calcite hydrogen bonds are progressively broken and replaced by peptide-calcite interactions. Then, through electrostatic bonds, pE 1 R 11 progressively integrates the crystal lattice, so that it cannot totally desorb. Flexibility is therefore the key point for permanent trapping into the crystal. This trapping hinders the smooth flow of growing steps, which are somehow forced to flow between additives, e.g. the pE 1 R 11 peptide. Similar conclusions have been drawn by Sicheri and Yang (43) with the antifreeze protein from winter flounder. They showed that the current ice-binding models (44,45), also based on ice-lattice matching, were no longer adequate.
Interestingly, human serum albumin, an excellent binder of calcium oxalate crystals, does not inhibit their growth. Conversely, it is a strong nucleation enhancer of calcium oxalate dihydrate crystals by ionotropic effect (46). The structure of albumin shows a repetitive configuration of three helical homologous domains, each formed by two smaller subdomains. This is a highly organized structure (47,48) conferring a notable rigidity that prevents albumin from disrupting some of its own hydrogen, van der Waals, or electrostatic bonds. Consequently, albumin is easily rejected from the crystal surface by oncoming growing units. This rejection explains why albumin is not an inhibitor of crystal growth and confirms why flexibility is the key point in the inhibition of calcium salt crystal growth.
This "unfolding-binding" mechanism mirrors our crystallographic data and the deduced resulting structural organization of lithostathine. Indeed, in previous experiments, no structural information was obtained for the first 13 amino acids, including pE 1 R 11 . Consequently, as this domain was disordered in our crystals, it was assumed to be unstructured and highly flexible. Refinement of the structure at 1.3 Å allowed us to build up the complete protein chain. In addition, the much higher temperature factor value than for the C-terminal domain confirmed this extreme flexibility. This is consistent with the structural organization of both domains as two independent entities: the globular C-terminal domain interacts solely with the elongated flexible N-terminal peptide through the Arg 11 -Ile 12 peptide bond. In addition to secondary structure elements, the C-terminal domain is well stabilized in this region by the C-14 -C-25 disulfide bridge. In opposition, the flexible N-terminal domain is stabilized in the crystal by only two ion pairings with sym- Atoms that participate in the bonding between the two entities are highlighted using a colored ball representation. Calcium and oxygen atoms from the calcite surface are displayed in light blue and orange, respectively. pE 1 R 11 oxygen and hydrogen atoms are in red and white, respectively. Each of the 11 amino acids forming the backbone is labeled by a specific color: purple for pyro-Glu 1 , yellow for Glu 2 and Glu 6 , blue for Ala 3 and Ala 10 , ocher for Gln 4 and Gln 9 , light green for Thr 5 , gray for Leu 7 , light brown for Pro 8 , and green for Arg 11 . The white and gray C(CH 3 ) 3 group on Arg 11 stands for the remaining C-terminal part of lithostathine. metry-related molecules. As a consequence, the peptide conformation, as revealed by the crystal structure and CD experiments, may be only one of several energetically stable conformations for this highly flexible domain in solution.
Since the structures of N-and C-terminal domains are independent, both domains could be independently denatured during the calcite growth inhibition. We believe that, since the stable C-terminal domain is highly polarized with a unique distribution of charges on its surface (6), the resulting large dipole moment may play a role in the orientation of the whole lithostathine as it approaches the calcite, whereas the undecapeptide alone gradually unfolds and attaches to the crystal surface. This would contribute to the protein trapping into the crystal. Also, it explains why the whole protein is a better inhibitor than the undecapeptide alone (12).
Refinement at 1.3-Å resolution of the lithostathine crystal structure allowed us to model the first three residues of the Thr 5 -linked carbohydrate moiety. However, we did not include this moiety in our simulations for two reasons. First, it has been shown that the synthetic undecapeptide, which is devoid of carbohydrate, has the same effect as the native glycosylated one on calcite crystal growth (9). Second, the sugar moiety of lithostathine purified from different patients presents a high degree of heterogeneity both in length and sequence (27,49). 3 Consequently, the nature of the carbohydrate motif does not seem to be important for activity, although research on the role of carbohydrates has gained interest recently, in particular in signaling pathways. For instance, recent studies have suggested that carbohydrate units linked to proteins and lipids on cell surfaces are key participants in cell-cell recognition during development, cell-cell adhesion, or molecular targeting. However, so far, it has not been established that the lithostathine carbohydrate moiety plays a role, although lithostathine is a secretory protein. A reasonable possibility is that the sugar moiety protects, to some extent, the very labile Arg 11 -Ile 12 bond from in vivo proteolysis (7). Another possibility is that lithostathine bears different oligosaccharides when expressed in tissues other than the exocrine pancreas or that the carbo-hydrate moiety is involved in specific localization of the protein.
In conclusion, a prototypic protein inhibits calcium salt crystal growth through three basic rules. First, it must be flexible enough to spread onto the crystal surface. Second, it must adopt a conformation that allows it, through its backbone, to develop strong coulombic interactions with the crystal surface. Last, it must further stabilize these interactions through interactions of the crystal surface and lateral chains. Since calcium oxalate and calcium phosphate crystals are related to calcite, mechanisms of growth inhibition in other biological fluids should proceed in the same way.  (10 1 4)

interface interaction
The 15 bonds identified, as well as the splitting of the total interaction energy in terms of interaction type (6 -12 Lennard-Jones-like van der Waals type, coulombic type, and 10 -12 Lennard-Jones-like hydrogen bond type), are shown. In this particular snapshot, the undecapeptide surface coverage represents 27% of the (10 1 4)  a Computed using a dielectric constant () value equal to 1.

TABLE III
Energy values related to the pE 1 R 11 -(11 2 0) interface interaction The 18 bonds identified, as well as the splitting of the total interaction energy in terms of interaction type (6 -12 Lennard-Jones-like van der Waals type, coulombic type, and 10 -12 Lennard-Jones-like hydrogen bond type), are shown. In this particular snapshot, the undecapeptide surface coverage represents 20.6% of the (11 2 0) interface area. a Computed using a dielectric constant () value equal to 1.