Tyrosyl motif in amelogenins binds N-acetyl-D-glucosamine.

Ameloblasts secrete amelogenins on the pre-existing enamel matrix glycoproteins at the dentine-enamel junction. The hypothesis that amelogenins may interact with enamel matrix glycoproteins is tested by hemagglutination of purified, native (porcine) and recombinant murine amelogenins (rM179 and rM166) and hemagglutination inhibition with sugars. Amelogenin agglutination of murine erythrocytes was specifically inhibited by N-acetylglucosamine (GlcNAc), chitobiose, and chitotetraose and by ovalbumin with terminal GlcNAc. The GlcNAc affinity was confirmed by dosimetric binding of rM179 with [14C]GlcNAc, specific binding in relation to varying concentrations of GlcNAc, Scatchard plot analysis and competitive inhibition with cold GlcNAc. The hemagglutination activity and [14C]GlcNAc affinity were retained by the NH2-terminal tyrosine-rich amelogenin peptide (TRAP) but not by the leucine-rich amelogenin peptide, LRAP (a polypeptide sharing 33 amino acid residues of TRAP), or by the C-terminal 13 residue polypeptide of amelogenin (rM179). Since TRAP but not the 33-residue sequence of the TRAP shared by LRAP bound to [14C]GlcNAc, we inferred that the GlcNAc binding motif was located in the 13-residue tyrosyl C-terminal domain of TRAP (PYPSYGYEPMGGW), which was absent from LRAP. [14C]GlcNAc did indeed bind to this "amelogenin tyrosyl motif peptide" but not when the tyrosyl residues were substituted with phenylalanine or when the third proline was replaced by threonine. Significantly, this latter modification mimics a point mutation identified in a case of human X-linked amelogenesis imperfecta. The amelogenin tyrosyl motif peptide sequence showed a similarity to the secondary GlcNAc-binding site of wheat germ agglutinin.

Ameloblasts, specialized cells of the inner enamel epithelium, synthesize and secrete amelogenins, the principal enamel matrix protein (Ͼ90%) on to a premineralized collagenous dentine at the dentine-enamel junction (DEJ) 1 forming a layer of extracellular matrix that eventually becomes mineralized with a calcium hydroxyapatite creating the dental enamel (1)(2)(3). Other enamel matrix components include the glycosylated and phosphorylated "non-amelogenin" proteins (enamelins, tuftelins, and ameloblastins). However, little is known about the interaction of amelogenins with these other enamel proteins or with dentine matrix proteins (collagens and sialophosphoproteins) (4 -6). It has been suggested that the EMGs 2 may bind to amelogenins in addition to interacting with the growing mineral crystals (6); however, the molecular nature of any such interactions is unclear.
Amelogenins  are phosphorylated at position Ser-16 but are not glycosylated (7,8). Three domains of amelogenin primary structure have been described: (a) a highly conserved NH 2 -terminal hydrophobic sequence of 44 -45 amino acids, containing 6 tyrosine residues (a region referred to as the "tyrosine-rich amelogenin polypeptide" or TRAP) with one of the tyrosyl-containing motifs consisting of three tyrosine residues with a spacing as: XYXXYXYX (PYPSYGYE); (b) a central hydrophobic core motif of 100 -130 residues enriched in proline, leucine, glutamine, and histidine; and (c) the amelogenin carboxyl-terminal peptide (ACP), an acidic hydrophilic sequence of 13 amino acids (see Table I).
The amino acid sequences of the amelogenins are highly conserved in the seven mammalian species so far studied (9 -13), with a near identity of sequence being observed for the NH 2 -terminal TRAP region and the C-terminal teleopeptide. The biological roles of these highly conserved sequences are little understood. During maturation of the enamel, the amelogenins are processed by proteinases within the matrix (14 -16), an event that appears to signal a change in the functional role of the intact protein.
Since amelogenins are secreted on the pre-existing organic matrix of the dentine, molecular interactions at this dentineenamel interface may be anticipated and may be functionally important to the stability of the developing tooth structure. Dentine matrix contains both collagenous and non-collagenous proteins; the latter class also includes glycoproteins and phosphoproteins (6,(17)(18)(19)(20)(21). It has been shown that the sugar residues of the glycoconjugates are capable of inhibiting hydroxyapatite crystal formation (22)(23)(24). We hypothesize that amelogenins may bind to the sugar residues of EMGs at the DEJ to facilitate biomineralization. Alternatively, such interactions may be important in amelogenin matrix structure. In this investigation, we have, for the first time, identified specific interactions between amelogenins and the GlcNAc residues of glycoconjugates. Further, we have identified the glyco-binding locus of the amelogenin structure with a conserved tyrosyl * This work was supported by National Institutes of Health NIDR Grant DE-02848. 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.

Expression, Isolation, and Purification of Recombinant Mouse
Amelogenin Proteins rM179 and rM166 -Recombinant murine amelogenin rM179 was prepared by expression in Escherichia coli and isolated and purified by high performance liquid chromatography as described previously (25). The protein was further purified by reversedphase HPLC (C4 -214TP54 column; Vydac/The Separations Group, Hesperia, CA) and the homogeneity of the protein was assessed by SDS-polyacrylamide gel electrophoresis (26) carried out in the presence of 0.1% SDS, using a 15% acrylamide gel. The gel was stained with Coomassie Brilliant Blue. The amelogenin "rM179" is identical in sequence to M180, the principal native murine amelogenin except for the lack of the amino-terminal methionine residue (27) and a phosphorylated serine at position 16 (7,8). The availability of a source of a defined and purified amelogenin protein has now permitted us to study the sugar binding properties of this molecule at a level not previously possible for amelogenin proteins isolated from in vivo sources. Porcine amelogenins (P173, P148) were extracted and purified following the protocol of Fincham et al. (9). "rM166," a construct that lacks the hydrophilic carboxyl-terminal 13-residue segment, was also purified as described for the rM179 (25). To identify the sugar binding site, we used rM179, rM166, synthetic murine amelogenin polypeptides, TRAP, LRAP (leucine-rich amelogenin polypeptide, identical to the full-length amelogenin at its two termini but lacking the center portion of the protein (28), and ACP (amelogenin C-terminal peptide) (Table I). In addition, we used "ATMP" (amelogenin trityrosyl motif peptide) and two altered ATMP peptides. In one of these, the third proline is replaced by threonine (T-ATMP), and in the other, all three tyrosine residues are replaced by phenylalanine (F-ATMP) ( Table I). All the polypeptides used in this investigation were synthesized by the USC Microchemical Core Laboratory using an Applied Biosystems model 430A one-column peptide synthesizer with the modified Merrifield procedure (29). Peptides were purified by reversed-phase HPLC (C4 -214TP54 column or C18 column; Vydac/The Separations Group, Hesperia, CA) with a shallow gradient of 40 -50% B in 60 min (buffer B was 60% v/v aqueous acetonitrile in 0.1% v/v trifluoracetic acid (TFA) and buffer A was 0.1% TFA) at a flow rate of 1.0 or 0.5 ml/min for small peptides depending on their size. Homogeneity of the recombinant amelogenins rM179 (20.16 kDa) and rM166 (18.4 kDa), and the synthetic polypeptides TRAP (5.20 kDa) and LRAP (6.82 kDa), was confirmed by SDS-polyacrylamide gel electrophoresis. The estimated molecular mass of ATMP, T-ATMP, and F-ATMP ranges from 1.45 to 1.50 kDa.
Hemagglutination (HA) Assay-The activity of the amelogenins was assayed by measuring their ability to agglutinate erythrocytes. HA assays were performed in 8 ϫ 12 microtiter plates, with U-bottomed wells, at 30°C by 2-fold dilution of the protein and visual estimation of erythrocyte agglutination on microtiter plates 1 h after adding the cells (30,31). The reagents used include Tris-buffered saline (TBS): 50 mM Tris, 100 mM NaCl, pH 6.3 and 7.2; and modified Alsevier's medium containing antibiotics (AMAB): 30 mM sodium citrate, pH 6.1, 70 mM NaCl, 114 mM glucose, 100 g/ml neomycin sulfate, 330 g/ml chloramphenicol, 0.04% TFA. Erythrocyte suspension from mouse, rat, horse, bovine, swine, and sheep (purchased from Crane Laboratories, Inc., Syracuse, NY) and human A, B, and O blood groups (from Interstate Blood Bank, Inc., Memphis, TN) were used. The volume of cells were measured by hematocrit; suspension was in AMAB for up to 1 week at 4°C. In the HA assays, the cell suspension was diluted to 1.5% in TBS, pH 7.2. All buffers and erythrocyte suspensions were warmed to 30°C for 15 min. Purified native and recombinant amelogenins and synthetic polypeptides (1 mg/ml) were suspended in 0.04% TFA, in which they were readily soluble. No difference in HA activity of amelogenins when tested in a medium containing diluted TFA (TFA (0.04%) ϩ TBS, pH 6.3) or TBS (pH 6.3 or 7.2) was observed. To each well containing 12.5 l of TBS (pH 6.3), 12.50, 6.25, 3.12, 1.56, 0.78, and 0.34 l of amelogenin was added. The final volume was adjusted to 25 l with 0.04% TFA. After tapping the wells gently, 25 l of a 1.5% suspension of erythrocytes (washed three times in TBS, pH 7.2) was added. The microtiter plate was covered with Parafilm "M" (American National Can, Greenwich, CT) and gently agitated on low speed vortex for 10 s. The degree of agglutination was scored after 1 h at 30°C. Hemagglutination titers were reported as the reciprocal of the highest dilution of protein giving complete agglutination after 60 min.
Dose-dependent Binding of [ 14 C]GlcNAc to rM179 -One hundred microliters of known amount of [ 14 C]GlcNAc (2 ϫ 10 4 cpm/100 l of TBS, pH 7.2) (purchased from Amersham International, Buckinghamshire, United Kingdom) was added to polypropylene Eppendorf tubes containing 100 l of increasing amounts of purified rM179 in 0.04% TFA, and incubated under constant agitation for 90 min at 30°C. The proteins were precipitated with 1 ml of cold ethanol (ethyl alcohol, 200 proof) or 10% trichloroacetic acid (TCA) or 150 l of 30% polyethylene glycol (w/v) (PEG 6000; purchased from Accurate Chemical Co., Westbury, NY). When the proteins were partitioned with PEG, the mixture was incubated for 15 min at 4°C to permit precipitation. The tubes were then centrifuged for 15 min at 10,000 ϫ g in a Beckman Microfuge 12, and the supernatant was removed. The tip of the tube containing the pellet was cut off and transferred to a scintillation vial and the radioactivity was measured 15 min after adding 3 ml of scintillation fluid (Safety-solve or Bio-safe 11, Research Products International Corp, Mount Prospect, IL) in a ␤ counter. The proteins precipitated with ethanol or TCA were then centrifuged for 10 min at 10,000 ϫ g, and the supernatant was removed. The unbound radioactive GlcNAc, if any, was removed completely by repeated vortex mixing and washing (three times) with ethanol or TCA. Washing with these protein precipitants did not affect the bound sugar. The final protein pellets were dissolved in 50 l of 1 N NaOH, and the bound radioactivity was measured as described above. WGA was used as a positive control and BSA as a negative control in all the experiments. All samples (in triplicates) were treated in the same way and counted for bound radioactivity.

Binding of [ 14 C]GlcNAc as a Function of Increasing Concentration of GlcNAc-
The total binding of [ 14 C]GlcNAc to rM179 was determined in duplicates using increasing concentrations of [ 14 C]GlcNAc (20 -720 pmol) to 6 nmol of rM179. The nonspecific binding of [ 14 C]GlcNAc was determined in triplicate in the presence of 10 M of cold GlcNAc, and it was subtracted from the total binding to obtain the specific binding. A Scatchard plot analysis of this specific binding was carried out.
Competitive [ 14 C]GlcNAc Binding Assay with Cold GlcNAc and NeuAc-50 l of unlabeled GlcNAc or NeuAc with increasing concentrations were prepared in duplicates in Eppendorf tubes. To each concentration of unlabeled sugar, 100 l of [ 14 C]GlcNAc (2 ϫ 10 4 cpm) was added and mixed. 100 l containing 7.5 nmol of rM179 was added to the mixture and incubated for 90 min at 30°C on a rotator. After incubation, the proteins were precipitated with ethanol and washed as described above. WGA and BSA were used as positive and negative controls, respectively.
Relative GlcNAc Binding Efficiency of Recombinant and Synthetic Amelogenin Polypeptides-The binding assays were carried out in 96-Remova-well strips in a microtiter plate (Dynatech Labs. Inc., Chantilly, VA). rM179, rM166, and synthetic polypeptides (TRAP, LRAP, ACP, ATMP, T-ATMP, and F-ATMP) were suspended in carbonatebicarbonate buffer, pH 9.6 (40). At this pH amelogenins are readily soluble (41). The polypeptides (5 nmol/100 l) were added to wells in triplicate kept at 4°C overnight. WGA and BSA were used as positive and negative controls, respectively. The plates were washed with TBS (pH 7.2). 100 l of 14 C-labeled GlcNAc (2 ϫ 10 4 cpm) in TBS (pH 7.2) was added to each well and incubated for 90 min at 30°C. After incubation the unbound radiolabeled GlcNAc was removed and washed (three times) with TBS (pH 7.2). The Remova wells were then transferred to scintillation vials. The bound [ 14 C]GlcNAc was released into 3 ml of scintillation fluid (Safety-Solve, Research Products International Corp.) by shaking the vials on a shaker for 5 min and then by vortex mixing. The solid-matrix assay was carried out in triplicates at least on two different occasions following the International Committee of Harmonization of Technical Requirements guidelines.
Statistical Analysis-Descriptive statistics were carried out throughout. In addition, Mann-Whitney sample test and regression analyses (linear and polynomial curve fit, r 2 , and test of significance of slope) were carried out using Origins software (Microcal Software Inc.). Twotailed p-values were obtained to assess the level of significance.

Homogeneity of Purified Recombinant
Amelogenins-A typical reverse phase HPLC profile of rM179 used in this study is shown in Fig. 1A. Fig. 1B shows electrophoretic homogeneity of rM179, rM166, and the synthetic peptides TRAP and LRAP used in this study. The purified preparations (rM166) were devoid of or had negligible contamination with vector proteins.
Hemagglutination of Amelogenins-Purified and homogenous native (porcine) and recombinant amelogenins agglutinated of human A, B, and O, sheep, rat, mouse, pig, and horse erythrocytes but not bovine erythrocytes. The highest HA titer of native and recombinant amelogenins (3 g/well) was 16 with mouse erythrocytes. HA titer with other erythrocytes ranged from 2 to 4. Since mouse erythrocytes were agglutinated better than other erythrocytes by rM179 and native porcine amelogenin, these erythrocytes were used for all hemagglutination inhibition experiments. Parallel experiments with the protein extracts of E. coli used for preparation of rM179/rM166 did not agglutinate any of the erythrocytes.
Inhibition of Hemagglutination of Recombinant Amelogenin by Sugars, Oligosaccharides, and Glycoproteins-The carbohydrate-binding specificity of purified rM179 was probed by sugar inhibition of the hemagglutination. Table II shows that the nature of sugars that inhibited hemagglutination of amelogenin. Of the various sugars tested, only GlcNAc and NeuAc inhibited the hemagglutination of rM179. GlcNAc inhibited the hemagglutination at 10 nM, whereas 100 nM NeuAc was required to cause such inhibition. Thus the inhibitory potency of GlcNAc is 10-fold greater than that of NeuAc. None of the other sugars (GalNAc, NeuGc, and LacNAc) tested inhibited hemag-glutination, even at concentrations as high as 100 M. Results presented in Table III confirm that amelogenin recognizes GlcNAc in chitobiose, chitotetraose, and ovalbumin.
In striking contrast to free NeuAc, sialyloligosaccharides with varying ␣-ketosidic linkages failed to inhibit the HA titer of amelogenin, suggesting that the part of NeuAc recognized by the amelogenin is not accessible when NeuAc is linked to other sugars. The above contention is also supported by HA titers observed with sialidase-treated erythrocytes. Table IV shows that HA titers of rM179 and TRAP molecules are unaffected by sialidase treatment, while hemagglutination of desialylated erythrocytes was abolished by Limax lectin. HA titers of both WGA and D. stramonium lectins also are unaffected by sialidase treatment, suggesting that hemagglutination of these lectins as well as rM179 and TRAP molecules is due to GlcNAc residues but not due to NeuAc residues.
Direct Binding of [ 14 C]GlcNAc to Amelogenin-The binding of [ 14 C]GlcNAc to amelogenin increased with the concentration of the protein in a dose-dependent manner (Fig. 2). The recovery of bound GlcNAc was higher after PEG precipitation. However, ethanolic precipitates showed a consistent polynomial fit as evidenced by significant slope (p Ͻ 0.001) and r 2 (0.83). The positive controls, WGA, a GlcNAc/NeuAc-binding lectin showed similar dose-dependent binding. However, dose-dependent binding of rM179 with PEG paralleled with that of WGA. Furthermore, there is no significant difference in the slope between WGA and rM179 in ethanol. The binding was minimal and insignificant with the negative control (BSA).
Specific Binding of [ 14  GlcNAc to rM179 indicates that the GlcNAc binding site is homogeneous with respect to the association constant.
Competitive Binding of Radiolabeled and Unlabeled GlcNAc and NeuAc to Amelogenin-Competitive binding studies confirmed the HAI observations made earlier. The unlabeled Glc-NAc dosimetrically inhibited binding of [ 14 C]GlcNAc to rM179 (Fig. 4A). The inhibition slope is significant (p Ͻ 0.001), and r 2 is 0.998. Similarly unlabeled NeuAc also inhibited dosimetrically the binding of [ 14 C]GlcNAc to rM179 (Fig. 4B) and resulted in an excellent fit (r 2 ϭ 0.999). Comparison of the slopes of unlabeled GlcNAc and NeuAc revealed somewhat similar binding affinity of the two sugars to amelogenins. The positive control WGA was used as reference. The unlabeled GlcNAc also inhibited dosimetrically the binding of [ 14 C]GlcNAc to WGA (data not presented).
Identification of the Sugar Binding Motif in Amelogenin-In order to identify the domain of the polypeptide that may recognize GlcNAc, we have used different fragments of amelogenins to assess their hemagglutination. The structural domains of amelogenin and their synthetic counterparts used for analysis are presented in Table I. The full-length rM179 agglutinated mouse erythrocytes efficiently (Table V). The recombinant amelogenin rM166, lacking the 13 C-terminal hydrophilic residues (ACP), also agglutinated equally, suggesting that the ACP may not be required for binding. The ACP alone failed to cause hemagglutination, confirming that this region is not a necessary requisite for binding. The NH 2 -terminal TRAP region of 45 amino acids also agglutinated effectively (Table V). LRAP shares the NH 2 -terminal amino acid sequence found in TRAP in addition to 26 amino acids constituting the carboxylterminal end. Despite sharing 33 NH 2 -terminal amino acids of TRAP, LRAP failed to agglutinate the erythrocytes. Since LRAP is devoid of the tyrosine-rich portion of the TRAP domain, we suspected that the ATMP consisting of 13 amino acids might be necessary for hemagglutination. The 13 amino acids constituting ATMP are PYPSYGYEPMGGW. Using radiola-

FIG. 2. Dosimetry of [ 14 C]GlcNAc binding with the purified recombinant amelogenin M179.
To polypropylene Eppendorf tubes containing 100 l of GlcNAc (2 ϫ 10 4 cpm or 300 pmol), 100 l of increasing amounts of protein (circle, rM179; solid square, WGA, positive control; open square, BSA, negative control) was added and incubated for 90 min at 30°C. After precipitation or filtration with EtOH/TCA/PEG and washing, the bound radioactivity was assessed. The mean values of triplicate analyses for each concentration were plotted on a logarithmic scale to evaluate dosimetric binding of [ 14 C]GlcNAc and assessed for polynomial regression. The polynomial curve fit (slope) is restricted to the ethanolic precipitate of rM179. r 2 and significance of the slope are indicated. The mean values of triplicate analysis at each concentration of other preparations suggest a similar trend (curve fit not shown).

FIG. 3. Specific binding of [ 14 C]Glc-NAc to rM179 and Scatchard plot.
Specific and nonspecific binding of [ 14 C]GlcNAc to purified amelogenin (6 nmol) as a function of increasing concentration of GlcNAc. The saturation binding of [ 14 C]GlcNAc to rM179 was determined using increasing concentrations of Glc-NAc (20 -720 pmol). The nonspecific binding was determined in the presence of 10 M of cold GlcNAc, and it was subtracted from the total binding to obtain specific binding. The Scatchard analysis of this specific binding is shown as an inset. Each point represents the average of triplicate determinations; Regression analysis showed that the r 2 is 0.988.
beled GlcNAc, we have further confirmed the role of ATMP in binding to GlcNAc (Fig. 5). The binding of [ 14 C]GlcNAc to both TRAP and ATMP are significantly higher than that bound to LRAP. The binding of [ 14 C]GlcNAc to the ATMP was not significantly different from the binding of [ 14 C]GlcNAc to TRAP. However, substitution of the third proline by threonine (T-ATMP) and substitution of all three tyrosine residues by phenylalanine (F-ATMP) resulted in complete loss of binding of the peptide to GlcNAc (Fig. 5). DISCUSSION GlcNAc Specificity of Amelogenin-Amelogenin, the major extracellular protein constituting the developing enamel matrix, possesses lectin-like activity. Both native and recombinant amelogenins bind to acetyl esters of GlcNAc and NeuAc but not to that of GalNAc. Amelogenin discriminates N-acetyl residue (NeuAc) from N-glycolyl residue (NeuGc) of sialic acids. These observations show that the specific sugar binding requires acetyl esters and an additional group common to GlcNAc and NeuAc but not GalNAc. The GlcNAc affinity of amelogenin was further confirmed by the dose-dependent binding of amelogenin with [ 14 C]GlcNAc, specific binding in relation to varying concentrations of GlcNAc, Scatchard plot analysis, and the competitive inhibition studies with cold GlcNAc. The minimal GlcNAc affinity constant of amelogenin (10 nM), is higher than that of other lectins, which could be due to the selfassembly properties of amelogenin (see below).
The ability of NeuAc to inhibit rM179 binding to [ 14 C]Glc-NAc suggests that both these sugars may compete for a single binding site. Yet another protein with identical dual affinity is WGA (36,37). The interaction of WGA with GlcNAc and NeuAc is attributed to their structural similarities, namely the superimposable conformation of amino and hydroxyl groups at C-5 and C-4 of the pyranose of NeuAc with C-2 and C-3 of GlcNAc (37,42). The same superimposable conformation of NeuAc and GlcNAc may be responsible for recognition of NeuAc by amelogenin. It is interesting to note that only free NeuAc, and not sialylated oligosaccharides (Table II) or sialoglycoproteins (Table III), inhibited the hemagglutination of amelogenin. Furthermore, amelogenin agglutinated desialylated erythrocytes (Table IV). Probably, the amino and hydroxyl groups required for amelogenin-sugar interaction are not accessible when sialic acid is in a bound state. However, amelogenin is able to recognize chitobiose and chitotetraose, as well as ovalbumin, in which GlcNAc also occurs as a terminal sugar (33) suggesting that C-2 and C-3 of the bound-GlcNAc is accessible to amelogenin. In this respect, amelogenin resembles a GlcNAc-specific lectin from the ascidian Didemnum ternatanum (43). The reports that GlcNAc is present in enamelins and in other EMGs (44,45) support our hypothesis that amelogenins may bind to these GlcNAc-containing glycoproteins. Based on these observations, we are undertaking a detailed immuno-and lectin histochemistry study of the DEJ-related proteins (enamelins, tuftelins, and ameloblastins) to identify and characterize the presence of such GlcNAc-containing glycoconjugates.
The ATMP Motif of TRAP Binds to GlcNAc-The GlcNAcbinding domain of amelogenins rM179 and rM166 is localized in the 45-amino acid residue TRAP sequence, but not in the LRAP sequence (see Table V and Fig. 5). Although LRAP shares the NH 2 -terminal 33 amino acid residues with TRAP (Table I), it failed to agglutinate the mouse erythrocytes or bind to [ 14 C]GlcNAc. The same is true for the ACP, which is also present in LRAP. Since [ 14 C]GlcNAc bound to the 45-mer TRAP molecule but not to the 33-residue region shared by TRAP and LRAP, we inferred that the binding motif was located in the 13 amino acid residues of ATMP domain of TRAP. This ATMP consists of 3 tyrosyl residues, spaced as XYX-XYXYX in the sequence PYPSYGYEPMGGW. Indeed, the synthetic ATMP bound to [ 14 C]GlcNAc as efficiently as the complete TRAP molecule (Fig. 5). Replacing the tyrosyl residues with phenylalanine (F-ATMP) resulted in complete loss of Glc-NAc affinity. Furthermore, and of biological significance, we found that a mutation in the ATMP sequence (substitution of proline-3 with threonine, as in T-ATMP), as has been found in some cases of human X-linked amelogenesis imperfecta (46), also resulted in failure of the peptide to bind to [ 14 C]GlcNAc. Dehydroxylation of tyrosyl residues and removal of proline-3 abolishes GlcNAc affinity, thereby confirming the importance of the tyrosyl motif in GlcNAc binding.
GlcNAc-binding Tyrosyl Motifs in Other Lectins-The Glc-NAc-binding domains of WGA have already been defined (36). The WGA has two different GlcNAc-binding sites, the secondary GlcNAc-binding site of which shows a striking similarity with the tyrosyl motif of TRAP. tyrosyl residues are also implicated in the sugar binding (36). The GlcNAc binding motif of amelogenins also showed some similarity (Ser-Tyr residues) to the GlcNAc binding motif of another lectin, UEA-II (47). Most importantly, both these motifs carried tyrosine, proline, and glycine.
Lectin-like Property of Amelogenin-In contrast to the multiple sugar-binding sites of the lectins (WGA, UEA-II), the amelogenin polypeptide has a single binding site, but its ability to bind to sugar or agglutinate erythrocytes is postulated to be dependent on the self-assembly properties of the amelogenins (48,49). The number of GlcNAc-binding sites of an amelogenin assembly may vary with the size of the nanosphere and the microenvironment in which sugar-protein interactions occur.  Remova-well plates were coated in triplicate with 100 l of 5 nmol of protein/peptide in carbonate buffer (pH 9.6) overnight at 4°C. Unbound proteins were removed, and the plates were washed with TBS/ PBS (pH 7.2). Radiolabeled GlcNAc (100 l, 2 ϫ 10 4 ) was added and incubated for 90 min at 30°C. The unbound GlcNAc was removed by washing the plates (three times) with the same buffer, and the radioactivity was measured. WGA and BSA were used as positive and negative controls, respectively. The background values (uncoated wells) are also presented in the figure. Standard deviation and the sample size for each test material are indicated in the figure. Two-tailed p-values after non-parametric analyses are indicated.

TABLE II
Inhibition of hemagglutination of rM179 by sugars Purified sugars (12.5 l) serially diluted in TBS, pH 6.3, were added to microtiter wells and mixed with rM179 (12.5 l) previously adjusted to 8 HA units; after 60 min of incubation at 30°C, 25 l of 1.5% suspension of mouse erythrocytes (TBS, pH 7.2) were added to each microtiter well and mixed. The HA titer was determined after a 2-h incubation at 30°C. The relative inhibitory potency of each sugar is indicated.

Minimal concentration required
Relative inhibitory potency There appears to be only one trityrosyl site per molecule of amelogenin. The site is restricted to the C-terminal region of the TRAP sequence. Our studies confirm that this ATMP is the GlcNAc binding domain. We postulate that the lectin-like activity of amelogenins, namely hemagglutination of erythrocytes, results from amelogenin self-assembly creating multivalent binding sites. Indeed, works of Fincham et al. (48,49) document these properties of amelogenins. GlcNAc Is the Ligand for Amelogenin at the DEJ-The molecular microenvironment at the DEJ may contain a wide array of glycosylated proteins secreted by ameloblasts or odontoblasts, although amelogenins per se are not known to be glycosylated (7,8,50). The total hexosamine content of enamel and of dentine-cementum has been determined to be 1.2 and 3.0 mol/g, respectively (51), while GlcNAc has been specifically identified in protein fractions of enamelin (44,45). During ontogeny of enamel, GlcNAc-bearing proteins are secreted earlier than amelogenin (52,53). Since the DEJ, during the pre-secretory phase, contains proteins with GlcNAc-containing oligosaccharides, the newly secreted amelogenin may be anchored to such glycoproteins. While the amelogenins constitute about 90% of the total mass of enamel matrix protein, their molecular mass (ϳ20 kDa) is smaller compared with the EMGs with masses up to 70 kDa. Recent studies of the "enamelin" protein (54) have demonstrated both O-and N-linked glycosylation sites to be present in this matrix protein (54 -56). We suggest that the GlcNAc residues of the EMGs may constitute the ligand for amelogenin, anchoring the molecules at the DEJ through the TRAP ATMP motif.
In this context, the observations of Fincham et al. (57,58) are striking in that they observed that the amino-terminal fragment of amelogenin, namely TRAP, is the principal amelogenin that persists past early enamel maturation stage. If the EMGs are implicated in the biomineralization process, then the role of the tyrosyl motifs of the TRAP region cannot be ignored because they may serve to mask sugar residues, which could be inhibitory for nucleation of calcium hydroxyapatite crystals (22)(23)(24). Indeed, Doi et al. (59) demonstrated that the seeded growth of enamel apatite crystals in vitro was inhibited by bovine "enamelin" preparations.
Significance of Amelogenin-GlcNAc Interaction in Enamel Biomineralization-It is now well established that, in vivo, the amelogenins self-assemble to form supramolecular structures (nanospheres), which comprise the bulk of the secretory-stage enamel matrix (48,49). It is presently unknown how these (ϳ20 nm diameter) structures become organized to form the three-dimensional matrix. The specific lectin-like interactions with glycosylated matrix proteins such as the enamelins or ameloblastins may be postulated to mediate this process.
If the function of the tyrosyl motif of amelogenin is to bind to a saccharide ligand at the DEJ, then one can reasonably expect that protein-carbohydrate interaction of amelogenin and EMGs in the formation of nanospheres associated with the developing apatite crystals. We envisage that the amelogenin-GlcNAc interaction may be similar to arthropodin-chitin (a polymer of GlcNAc) interaction observed in the matrix for calcification in crustacean cuticle (60,61) and molluscan shells (62).
The microenvironment of amelogenin in the developing enamel includes proteinases (7,16,63,64). Indeed, amelogenin is fragmented by these proteinases. The fragmentation commences at the COOH-terminal end (65) but not at the NH 2terminal end. During enamel maturation, fragmentation appears to extend to the ATMP region of TRAP. However, the GlcNAc-bound TRAP region may be resistant to proteinase activity. The isolated TRAP is 45 amino acid residues in length, resulting from a cleavage between Trp-45 and Leu-46 of the parent amelogenin (10,66). However, a 43-mer TRAP also exists suggesting that a cleavage may also occur between Gly-43 and Gly-44 (67). The formation of the 45-mer and 43mer TRAPs may possibly depend on whether the protein is bound to GlcNAc or not, since the COOH-terminal -MGGW may constitute the binding pocket of GlcNAc very similar to that found in WGA.
A number of investigators have observed ameloblast-derived proteins along odontoblast cell surfaces (52,68), and it is also known that the odontoblast cell processes (tubular processes in the dentine) extend to the dentine-enamel interface (69 -71). These observations suggest that amelogenins may also have interactions either with the extended odontoblast cell surface or with the product released from dentine tubular process at the dentine-enamel interface. Although the functional significance of such interactions is far from clear, it may be envisaged that amelogenins can function as cell adhesion proteins very similar to the E-, L-, and P-selectins (72)(73) in binding to TABLE IV HA titers of recombinant amelogenin (rM179) and TRAP molecule before and after sialidase treatments Details of the treatments are described under "Experimental Procedures." A reaction mixture (total 1.0 ml) containing 10% washed mouse erythrocytes in PBS, 0.05% BSA (pH 7.0) and 140 milliunits of sialidase of V. cholerae or C. perfringens or A. ureafaciens were incubated at 37°C for 4 h. The control erythrocytes were treated similarly without sialidase (Buffer-treated). Sialidase-treated cells were washed with PBS-BSA three times and pelleted by low speed centrifugation. The concentration of lectins or amelogenin or TRAP added to the first well of the titer plate is 5 g/25 l. This amount was serially diluted for testing HA titers.  HA assays were performed in a microtiter system at 30°C. All buffers and erythrocyte suspensions were warmed to 30°C for 15 min. Purified rM179 and other polypeptides (1 mg/ml) were suspended in 0.04% TFA. To each well containing 12.5 l of TBS (pH 6.3), the protein concentration was adjusted 12.5, 6.25, 3.12, 1.6, 0.8, and 0.4 g/well. The final volume was adjusted to 25 l with 0.04% TFA. 25 l of 1.5% suspension of washed erythrocytes was added. The final pH of the suspension is 6.3. The agglutination was scored after 1 h at 30°C. Bovine serum albumin Negative control 0 carbohydrate residues on the surface of odontoblasts. The specific oligosaccharide ligands expressed on odontoblasts and in EMGs deserve a more detailed investigation.
In conclusion, this study has shown a novel lectin-like function for the amelogenins, the principal proteins secreted by the ameloblasts of the developing tooth. The specificity of amelogenin for GlcNAc is significant from the point of view of the probable location of such sugars at the DEJ or within the matrix and the terminal domain of mammalian cell surface oligosaccharide-receptors. The GlcNAc-binding tyrosyl motif of amelogenin shares homology with WGA. Ser-Tyr motifs appear to be a common characteristic feature of GlcNAc-binding lectins. The ATMP, a highly conserved domain of amelogenin structure, has a potential functional role in enamel proteincarbohydrate interactions. A point mutation in the tyrosyl motif of amelogenin, as observed in human X-linked amelogenesis imperfecta, results in the loss of GlcNAc affinity of amelogenin. The significance of such amelogenin-GlcNAc interactions in enamel biomineralization is presently unknown and deserves further study.