Altered Amelogenin Self-assembly Based on Mutations Observed in Human X-linked Amelogenesis Imperfecta (AIH1)*

A hallmark of biological systems is a reliance on protein assemblies to perform complex functions. We have focused attention on mammalian enamel formation because it relies on a self-assembling protein complex to direct mineral habit. The principle protein of enamel is amelogenin, a 180-amino acid hydrophobic protein that self-assembles to form nanospheres. We have used independent technical methods, consisting of the yeast two-hybrid (Y2H) assay and surface plasmon resonance (SPR), to demonstrate the importance of amelogenin self-assembly domains. In addition, we have analyzed mutations in amelogenin observed in patients with amelogenesis imperfecta who demonstrate defects in enamel formation. Assessments of self-assembly of these mutant amelogenins by either SPR or Y2H assay yield concordant data. These data support the conclusion that the amelogenin amino-terminal self-assembly domain is essential to the creation of an enamel extracellular organic matrix capable of directing mineral formation. It also suggests that a pathway through which point mutations in the amelogenin protein can adversely impact on the formation of the enamel organ is by disturbing self-assembly of the organic matrix. These data support the utilization of the Y2H assay to search for protein interactions among extracellular matrix proteins that contribute to biomineralization and provide functional information on protein-protein and protein-mineral interactions.

The self-assembly of a protein complex is a hallmark of biological systems. From processive enzymatic reactions to extracellular matrices, cells achieve control over their microenvironment through the assembly of a mixture of proteins. Protein assemblies are created with a specific stoichiometry. Such assembly precision is achieved through the interaction of protein motifs contained within the constituent protein members. The yeast two-hybrid (Y2H) 1 assay has been used successfully to study protein assembly and offers distinct advantages over many biochemical techniques used previously. An example of the value of the Y2H assay can be seen in the assembly properties of type X collagen. Type X collagen is a homotrimer of ␣1(X) chains encoded by the COL10A1 gene, which is expressed in hypertrophic chondrocytes during the process of endochondral ossification. Dwarfism in domestic pigs is a result of a single change (Gly 590 3 Arg) to the ␣1(X) chain of type X collagen (1). In humans, a similar phenotype has been described for a mutation at the equivalent position (Gly 590 3 Glu) of type X collagen. These patients have Schmid metaphyseal chondrodysplasia, which is a mild skeletal disorder associated with dwarfism and growth plate abnormality. The Y2H assay and in vitro assembly experiments demonstrated that the amino acid substitution interfered with the ability of the mutated collagen molecules to engage in trimerization (1). In this example, confirmation of a valid animal model for a human genetic disease has been possible in part by the Y2H assay. At the same time, the Y2H assay has given unique information about assembly dynamics that add to the genetic, radiologic, histologic, and biochemical data previously available. Although biochemical methods can be used to detect dimerization, they often lack specificity, they may not be quantitative, they are not easy to perform, and they do not facilitate the genetic analysis of protein-protein interactions under physiological conditions.
We have focused attention on mammalian enamel as an example of the self-assembly of an extracellular matrix protein complex. In the case of enamel, a complex mixture of proteins self-assemble to form supramolecular complexes that are capable of guiding the crystal habit of hydroxyapatite crystallites (2)(3)(4). The crystals organize in such a way that the final product resists wear from repeated use during mastication, resisting non-catastrophic failure in a wet and bacterial laden environment (5). The principal protein of enamel is amelogenin, a hydrophobic protein that has been shown to undergo self-assembly to form nanospheres (6). Two domains in amelogenin (an A domain consisting of amino-terminal residues 1-42 and a B domain consisting of carboxyl-terminal residues 157-173) were subsequently identified as mediating amelogenin selfassembly by using the Y2H assay (7). In the present study, we corroborate the importance of these two self-assembly domains to the process of amelogenin self-assembly using surface plasmon resonance (SPR). We further define altered assembly dynamics for changes in single residues in the A domain, which are altered in individuals affected with X-linked amelogenesis imperfecta (AIH1). These mutations result in a Thr 21 3 Ile alteration (8) and a Pro 41 3 Thr alteration (9). Amelogenesis imperfecta is a hereditary disease of enamel and, when linked to the amelogenin locus, severely reduces the ability of amelogenin to self-assembly as measured by either the Y2H assay or by SPR.
We have used the Y2H system and SPR to perform a genetic analysis of the determinants of amelogenin self-assembly. The favorable correlation between assembly parameters for amelogenin as measured by either the Y2H assay or SPR support the utility of these techniques to further define and understand the contribution of protein assembly to direct specific mineralized architecture. Enhanced understanding of the physicochemical guidelines governing protein self-assembly and the ability to manipulate these interactions will contribute to the studies of nanotechnology and biomimetics (10).

Expression and Purification of Recombinant Proteins for Surface
Plasmon Resonance Studies-Preparation of the recombinant mouse amelogenin rM179 has been described elsewhere (11). Preparation of the recombinant histidine-tagged wild-type mouse amelogenin (rp(H)M180) and the recombinant, mutated amelogenin constructs (rp(H)M180⌬A, rp(H)M180⌬B, rp(H)M180T 21 3 I, rp(H)M180P 41 3 T, and rp(H)M180T 21 3 I,P 41 3 T) have been described elsewhere (12). Briefly, polyhistidine-containing recombinant proteins were prepared in the expression vector pQE30 (Qiagen Inc., Valencia, CA) and recovered using nickel-nitrilotriacetic acid metal affinity chromatography using the protocol supplied by Qiagen Inc. The rp(H)M180⌬A contains the FLAG (Eastman Kodak Co., New Haven, CT) epitope in place of domain A (2, 12), and rp(H)M180⌬B contains three repeats of the hemagglutinin epitope in place of domain B (12). The recombinant amelogenin constructs with single or dual point mutations did not contain any marker peptides besides the polyhistidine cartridge at the amino terminus (Qiagen Inc.). The A domain and B domain of mouse amelogenin and the relative positions of the Thr 21 3 Ile and the Pro 41 3 Thr mutations for human amelogenesis imperfecta phenotypes are illustrated in Fig. 1. The amino acid sequences for each construct prepared for SPR have been published previously (12).
Surface Plasmon Resonance-The Biacore ® 3000 system, CM5 (carboxymethylated dextran surface) sensor chips, software, and reagents were obtained from the Biacore AB company (Uppsala, Sweden). There are four flow cells in a CM5 chip and among these four cells, three were used for the assay. Flow cell 1 was used as a control surface, whereas flow cells 2 and 3 were used as test surfaces. The carboxyl groups on the surface of each flow cell were activated with an injection of a solution containing 0.2 M N-ethyl-NЈ-(3-diethylaminopropyl)carbodiimide and 0.05 M N-hydroxysuccinimide. Recombinant mouse amelogenin rM179, suspended in 10 mM acetate, pH 4.5, was injected over flow cells 2 and 3 at a flow rate of 10 l/min until ϳ2000 response units of protein was coupled to the sensor surface (data not included). One response unit corresponds to an immobilized protein density of 1 pg/mm 2 (13). Flow cell 1 was treated in an identical manner but without any protein. The immobilization was completed by a 7-min injection of 1 M ethanolamine hydrochloride to block any remaining activated carboxyl groups.
Protein purity and concentrations were determined by SDS-PAGE analysis (12) using laser densitometry to compare the recombinant amelogenin to bovine albumin (molecular mass 69.3 kDa) of a known concentration and subsequently extrapolating molar concentrations for each of the amelogenin samples. To calculate the concentration of each amelogenin, individual molecular weights were taken into account (12). Each test protein was diluted from stock solution in HBS-EP buffer (10 mM Hepes with 0.15 M NaCl, 3.4 mM EDTA, and 0.005% Surfactant P-20, pH 7.4) to a desired concentration. The sample was then injected at 10 l/min at 25°C with an ϳ5-min contact period, followed by a 2-min dissociation period. The sensor chip was regenerated with 1 M MgCl 2 after each experiment.
Data transformation and overlay plots for all experimental interactions were prepared with BIAevaluation software 3.0 (Biacore AB). To correct for refractive index changes and nonspecific binding, the binding responses generated from flow cell 1 were subtracted from the responses generated in flow cells 2 and 3.
Plasmid Constructs Prepared for the Yeast Two-hybrid Assay-Wildtype and mutated amelogenin cDNAs were engineered into the GAL4 DNA-binding hybrid vector pPC97 (7). The signal peptide of amelogenin has been excluded from all constructs in this study. Briefly, the plasmids prepared above (rM180, rp(H)M180, rp(H)M180⌬B, rp(H)M180T 21 3 I, rp(H)M180P 41 3 T, and rp(H)M180T 21 3 I,P 41 3 T M180) were used as template DNA for PCR using the forward primer TTCGGATCCTATGCCCCTACCACCT and the reverse primer TAGG-GAGCTCTTAATCCACTTCTTCCCG. These two primers included a BamHI (forward primer) and SacI (reverse primer) restriction endonuclease restriction sites (underlined) to allow for efficient, in-frame cloning of the PCR product into pPC97. The construct rp(H)M180⌬B includes the hemagglutinin epitope (12). Respectively, these amelogenincontaining, GAL4 DNA-binding hybrid vectors have been called p97-M180, p97-M180⌬B, p97-M180T 21 3 I, p97-M180P 41 3 T and p97-M180T 21 3 I,P 41 3 T M180. DNA nucleotide sequences for each plasmid were obtained to confirm correct orientation, framing, and sequence of the cDNA insert. The domain A-deleted amelogenin in the GAL4 binding domain plasmid had been prepared for a previous study (pMa97/3 (7)) and will be referred to in this work as p97-M180⌬A. Construct p97-M180⌬A does not contain the FLAG epitope. The wildtype amelogenin-containing GAL4 DNA transcriptional activating hybrid vector had also been prepared for a previous study (pMa86/N (7)) and will be referred to in this communication as p86-M180. Positive hybrid plasmids for p53 and SV40 large T antigen and hybrid plasmids for H-ras and CDC25 have been used and reported previously (7).
Yeast Two-hybrid Assay-The liquid assay used to quantitate ␤-galactosidase activity has been reported previously (7).

Mutations to Amelogenin Domain A Interfere with
Amelogenin Self-assembly Using the Yeast Two-hybrid Assay-The results from the yeast liquid culture assay are presented (Table  I). Yeast hybrid assay was used to compare relative interaction strengths (including standard deviations) between wild-type and mutated amelogenin proteins. For each double-transformant combination eight yeast clones were grown until an A 600 reading of 0.6 -0.8 was reached and assayed for ␤-galactosidase activity. Wild-type amelogenin interacting with itself is calculated at unity; and other readings are reported relative to this including calculated p values. Each of the mutated amelogenin proteins studied showed a statistically significant decreased ability to interact with wild-type amelogenin; the most significant was the domain A-deleted construct followed by the Thr 21 3 Ile point mutation construct. Positive controls were the tumor suppressor protein p53 cotransformed with the SV40 large T antigen and H-ras (wild type) cotransformed with the CDC25 protein of Saccharomyces cerevisiae (14). Negative controls involve each of the amelogenin protein-hybrid constructs cotransformed with either pPC86 (for the pPC97 hybrids) or pPC97 (for the pPC86 hybrids). All negative control combinations had no ascertainable ␤-galactosidase activity as measured by the liquid culture assay.
Mutations to Amelogenin Interfere with Amelogenin Selfassembly Using Surface Plasmon Resonance-The real-time interaction of wild-type and mutant amelogenins were determined using SPR at concentrations of 0.25 M (test molecule) to the rM179 ligand. The first sample tested (rp(H)M180) was tested a second time following the completion of the experiment and gave consistent data for both passes. In addition, this experiment in its entirety was repeated in two different laboratories by different individuals and gave identical results. A representative overlay of a sensorgram generated for each interaction is presented (Fig. 2). It is apparent from these data that the chip surface remains fully active between each sample. It is also apparent that the greatest disruption to the amelogenin, as measured by response units, is seen for the ⌬A (rp(H)M180⌬A) test construct.
The real-time interaction of the various test molecules was determined at concentrations of 0.25 and 0.50 M to the rM179 ligand. For each analysis at these two concentrations, the sensorgram was calculated at the same selected time points; threshold, response, and final response. Threshold was measured 5 s before the end of the injection of the sample, the response was measured 10 s after the end of the injection, and the final response point was measured after an additional 10 s (an example is illustrated in Fig. 3A). The data for both these concentrations are presented in Fig. 3B. At a 0.25 and 0.50 M concentration, rp(H)M180 produced the highest binding response whereas rp(H)M180⌬A yielded the smallest binding response (Fig. 3B). DISCUSSION Amelogenin is an extracellular matrix protein, the amino acid sequence of which is highly conserved across species. Following the initial description of amelogenin proteins in the early 1960s, scientists have attempted to gain information about its secondary and tertiary structure. Amelogenin CD spectrum suggests that the amelogenin amino terminus contains ␤-sheet structure, whereas the central region and carboxyl-terminal region exhibits a random-coil conformation (15). It has been suggested that that amelogenin protein fails to retain stable secondary or tertiary structure (16); however, stable supermolecular structures are observed and take the form of 10 -15-nm diameter "nanospheres" (6). How amelogenin as-sembly into nanospheres is achieved is today being addressed using a number of technologies.
The Y2H system was used to define the "A" and "B" domains of amelogenin (7). Domain A is essential for amelogenin to amelogenin interactions. In isolation, domain A has the capability to direct interactions with another amelogenin protein. In contrast, domain B cannot be separated as an interacting domain in isolation, suggesting that conformation is of significance. The previous observation that a truncated amelogenin lacking domain B can no longer interact with domain A (7) could explain in part matrix disassembly following proteolytic cleavage at the carboxyl end of M180 amelogenin by a specific enamel protease (17).
Atomic force microscopy (AFM) and dynamic light scattering (DLS) have recently been used to study the assembly properties of recombinant amelogenin proteins that have been engineered with gross alterations to either domain A or to domain B or point mutations within domain A (12). By measuring the parameters of nanosphere size and assembly kinetics, it was concluded that domain A and domain B (of amelogenin) have significant and different roles to play in the nature and dynamics of amelogenin self-assembly (12). The study here blends SPR and Y2H methodologies to investigate amelogenin assembly characteristics and presents data that help to explain previous AFM and DLS data in a quantitative manner. For both SPR and Y2H methodologies we provide  wild-type and mutated amelogenin proteins Positive controls were p53 cotransformed with SV40 large T antigen and H-ras cotransformed with CDC25. Negative controls involve each of the amelogenin protein-hybrid constructs cotransformed with either pPC86 or pPC97. The p values for each of the mutated amelogenins interacting with wild-type amelogenin are given; p values were determined by comparing each amelogenin combination to the self-assembly of wild-type amelogenin (line 1). GAL4  relative quantitative data for wild-type amelogenin as well as for mutated amelogenins to interact with wild-type amelogenin. Previously AFM and DLS studies gave data related to amelogenin multimolecular assemblies (i.e. nanospheres), and these data included assembly dynamics and size distributions (12). The Y2H assay provides data related to the ability of two molecules to interact and is both qualitative and quantitative (7). Surface plasmon resonance provides additional information relating to how proteins react, in real time, to other proteins or protein complexes. The aqueous environment chosen to dilute the test protein for the SPR studies is similar to the aqueous environment of the enamel matrix as determined previously, i.e. HBS-EP buffer at pH 7.4 (including150 mM NaCl) approximates closely these ionic values previously determined for maturing enamel (pH 7.26, with 140 mM Na ϩ and 150 mM Cl Ϫ ) (18). Although the recombinant proteins prepared for the SPR studies are non-phosphorylated, a characteristic apparent in some animal species (19), the use of recombinant proteins to study amelogenin biochemistry is widely accepted and well documented (4,11,12,17). To simplify our data, we present the following relationship among the studied amelogenins. For the Y2H we determined the strength of interaction for wild-type amelogenin (p86-M180) interacting with the wild-type or altered amelogenins in the pPC97 cassette to be: M180 Ͼ (M180T 21 3 I,P 41 3 T Ն M180⌬B Ն M180P 41 3 T Ն M180T 21 3 I) Ͼ M180⌬A. The ratio of magnitude of interaction recorded for the two extreme readings (M180:M180⌬A) is ϳ10:1. In a similar fashion, we can present that the SPR response activity for wildtype amelogenin (rM179) interacting with wild-type and altered amelogenin complexes. In the case of SPR the relationship of response activity (for the histidine-tagged recombinant proteins) would be M180 Ͼ (M180T 21 3 I Ն M180P 41 3 T) Ͼ (M180T 21 3 I,P 41 3 T Ն M180⌬B) Ն M180⌬A where the ratio of response units of the two extreme readings (M180:M180⌬A) is ϳ10:1. Thus, data from the Y2H and SPR techniques are largely concordant. The one anomalous result appears to relate to the M180T 21 3 I,P 41 3 T protein. The data from the SPR assay indicate that amelogenin selfassembly is severely affected with this double mutation whereas the Y2H data suggest that this double mutation is minor when compared with other combinations. The double mutation studied in this recombinant protein does not exist in nature and was created to explore another aspect of amelogenin assembly, that being if multiple point mutations have an additive adverse effect on enamel matrix assembly capabilities. We do not know if teeth created with such a double mutation would exhibit a phenotype.
The study here has demonstrated that the most dramatic mutation, with respect to interaction and nanosphere assemblies, occurs with the removal of the A domain (amino-terminal 42 amino acids). This domain is roughly equivalent to the tyrosine-rich amelogenin peptide (amino-terminal 45 amino acids of mouse amelogenin). This is not a surprising result and emphasizes previous DLS data in which we found that the A domain is required for protein-to-protein interactions leading to nanosphere self-assembly whereas the absence of the A domain led to inhibition of the self-assembly process (12). Dynamic light scattering showed that the A domain-deleted construct (rp(H)M180⌬A) in solution produced a very heterogeneous size distribution (12). To equate this DLS conclusion to our SPR data, it is clear that the A domain-deleted construct reacted weakly with the sensor-bound wild-type amelogenin. Presumably this interaction was mediated through the intact B domain of the sensor-bound protein. Although some self-assembly of the mutated protein may occur prior to passing over the sensor, there appears to be a significant population of monomeric (mutated) amelogenins. This has been determined by generating a series of sensorgrams for selected protein combinations (wild-type interacting with serial dilutions of the A domain-deleted construct) for which kinetic data were determined and matched that of the 1:1 Langmuir binding model (data not shown) (13).
Removal of the domain B (p97-M180⌬B) resulted in a loss of 75% ␤-galactosidase activity, compared with wild-type amelogenin, as determined by the yeast assay. Surface plasmon resonance spectroscopy data showed that the interaction of rp(H)M180⌬B with rM179 produced a significantly reduced response when compared with the rp(H)M180 to rM179 amelogenin interaction using identical molar concentrations (Figs. 2 and 3B). The removal of the hydrophilic carboxyl terminus may promote hydrophobic interaction between amelogenin molecules. It has previously been postulated that the amelogenin hydrophilic carboxyl terminus is exposed on the surface of the amelogenin nanosphere, making the nanosphere soluble in an aqueous environment and thus preventing the fusion of neighboring nanospheres (12). It has been shown that disruption of the carboxyl terminus would then encourage interactions with neighboring molecules and molecular assemblies. Indeed, DLS data have demonstrated that amelogenin assemblies lacking domain B are unstable and progressively fuse with neighboring assembled amelogenin nanospheres (12).
Finally, there is a question of correlating the in vitro data derived here to in vivo data already available. Biomaterials inspired by designs of nature comprise the newly emergent discipline of biomimetics. We anticipate that identifying and understanding the role for the various domains within amelogenin to direct its self-assembly into nanospheres will improve our understanding of the structural hierarchy of the native material and also provide the knowledge base for an enamel biomimetic. With the goal of predicting protein-to-protein and protein-to-mineral interactions using in vitro approaches, we can then correlate such outcomes to the data derived from in vivo experiments using transgenic animals bearing engineered amelogenin protein constructs. Ultimately such a correlation will allow the construction of a biomimetic matrix that can direct the formation of hydroxyapatite into a highly ordered material mimicking the unique properties of the original enamel composite ceramic and do so at physiologic parameters of temperature and ion concentration.