MMP20-generated amelogenin cleavage products prevent formation of fan-shaped enamel malformations

Dental enamel forms extracellularly as thin ribbons of amorphous calcium phosphate (ACP) that initiate on dentin mineral in close proximity to the ameloblast distal membrane. Secreted proteins are critical for this process. Enam−/− and Ambn−/− mice fail to form enamel. We characterize enamel ribbon formation in wild-type (WT), Amelx−/− and Mmp20−/− mouse mandibular incisors using focused ion beam scanning electron microscopy (FIB-SEM) in inverted backscatter mode. In Amelx−/− mice, initial enamel mineral ribbons extending from dentin are similar in form to those of WT mice. As early enamel development progresses, the Amelx−/− mineral ribbons develop multiple branches, resembling the staves of a Japanese fan. These striking fan-shaped structures cease growing after attaining ~ 20 µm of enamel thickness (WT is ~ 120 µm). The initial enamel mineral ribbons in Mmp20−/− mice, like those of the Amelx−/− and WT, extend from the dentin surface to the ameloblast membrane, but appear to be fewer in number and coated on their sides with organic material. Remarkably, Mmp20−/− mineral ribbons also form fan-like structures that extend to ~ 20 µm from the dentin surface. However, these fans are subsequently capped with a hard, disorganized outer mineral layer. Amelogenin cleavage products are the only matrix components absent in both Amelx−/− and Mmp20−/− mice. We conclude that MMP20 and amelogenin are not critical for enamel mineral ribbon initiation, orientation, or initial shape. The pathological fan-like plates in these mice may form from the lack of amelogenin cleavage products, which appear necessary to form ordered hydroxyapatite.

Enamel formation is characterized by defined stages of development, which are arranged in the continuously growing rodent incisor from earliest (apical/posterior) to latest (incisal/anterior) 1 . In the anterior portion of the apical loop a basement membrane separates actively proliferating epithelial and mesenchymal layers 2 . Moving anterior from the apical loop, differentiating odontoblasts initiate secretion of mantle predentin composed primarily of type-I collagen 3 . Predentin is secreted adjacent to less-differentiated ameloblasts that become increasingly columnar and polarized 4 . As ameloblasts continue to differentiate, they begin secreting amelogenin 5 and extend finger-like projections through the thinning basement membrane and into the predentin between collagen fibers 6 . As this process proceeds, an increasing proportion of ameloblasts extend cell processes through the thinning basement membrane. The remaining membrane fragments are endocytosed by ameloblasts and subsequently degraded by lysosomes 7 . Small foci of mineralization occur in predentin within a few micrometers from the ameloblasts 8,9 . These islands of mineral expand and merge into a continuous layer of dentin that encases the predentin 9 . Ameloblast finger-like projections initiate the formation of enamel mineral ribbons primarily on the sides and tips of mineralized collagen fibrils to form the DEJ 8 . These initial ribbons are part of the interrod component of enamel 10 . The mineral ribbons elongate as the ameloblast finger-like projections retreat back from the mineralized collagen to the distal surface of the ameloblast. Interrod enamel ribbons continue growing from

Materials and methods
Mice used in study. All animals used in this study were housed in Association for Assessment and Accreditation of Laboratory Animal Care International (AAALAC)-accredited facilities and were treated humanely based on protocols approved by the Ohio State University and the University of Michigan Institutional Animal Care and Use Committees and were carried out in compliance with ARRIVE guidelines. Experimental protocols were designed along University and National Institutes of Health (NIH) guidelines for the humane use of animals. Amelx −/− and Mmp20 −/− mice 28,29 were separately backcrossed into C57BL/6 (wild-type, WT) mice for at least eight generations to obtain a homogeneous genetic background so that analyses would not be affected by genetics other than the assessed respective deleted genes.
Real-time quantitative PCR (qPCR) analyses. Total RNA extracted from 5-day old mouse first molars was isolated using Direct-zol RNA Mini Preps (Zymo Research). Total RNA was reverse-transcribed and cDNA was quantified by qPCR with mRNA specific primers. Relative expression levels of Amelx, Ambn and Enam were calculated by the 2 −ΔΔCt method using Gapdh cycle threshold (Ct) value as the housekeeping gene control 30 . In situ hybridization. Continuously erupting maxillary incisors were from adult mice and first molars were from postnatal day 5 pups. Both were from wild-type C57BL/6 mice. Teeth were formalin fixed and decalcified in 150 mM NaCl/10% acetic acid and were paraffin embedded, sectioned and deparaffinized in xylene. The antisense mouse Mmp20 probe and assay kit was from Advanced Cell Diagnostics (Newark, CA). In situ hybridization was performed according to the manufacturer's protocol.
Mouse incisor sample preparation for microscopy. As  www.nature.com/scientificreports/ bles were dissected, soft tissue removed, and the labial bone over the incisors was removed. Postfixation in the described perfusion buffer was for 4-6 h at 4 °C. The samples were washed with 0.1 M sodium cacodylate buffer overnight. Whole hemi-mandibles were then lipid stained with 1% reduced osmium tetroxide for 2 h, washed in several changes of distilled water, dehydrated in an acetone gradient, embedded with epoxy resin, and cured in a 60 °C oven for 48 h. Hemi-mandibles were cross sectioned at 1 mm increments at levels 1-8 and the slices were glued onto plastic stubs. The "levels" described herein represent the linear distance starting at the tip of the apical loop and following the initially curved (~ 0.2 mm) path of the basement membrane, then the ameloblast predentin/dentin interface, and finally the DEJ to the incisal tip.
Definition of levels corresponding to stages of enamel development. Level 1 (starting approximately 1 mm curve distance from the tip of the apical loop) corresponds to the beginning of the secretory stage, Level 3 is around the transition between the secretory and maturation stages, and Level 8 is fully mature enamel about to erupt into the oral cavity at the gingival margin along the labial side of the incisor 31 . Blocks from right incisors were faced and aligned across the transverse plane using glass knives on a Leica Ultracut E ultramicrotome. Blocks from left mandibular incisors were sawed from the mounting stubs and reoriented for cutting in the sagittal (longitudinal) plane. These blocks were also faced and oriented using glass knives.
Focused ion beam scanning electron microscopy (FIB-SEM). This method was described previously in detail 8 . Briefly, the glass-faced blocks were attached to 45° chamfered mounting stubs with conductive silver paste and placed into a Helios Nanolab 660 FIB-SEM (FEI, Systems for Research, Longueuil, QC, Canada). A sampling area 100 µm × 100 µm in size was selected and milled with gallium ions at rough (45 nA) followed by fine (9.4 nA) settings. Imaging was done with the through lens detector (TLD) and where possible with the in column detector (ICD). All FIB-SEM imaging was done using blocks prepared for sagittal (longitudinal) views of the enamel layer and associated enamel organ cells. Where possible, blocks from 3 different mice per genotype were examined. Charging on the surfaces of block faces was reduced by coating with a thin layer of platinum (3 nm) where required. Curtaining defects in TLD images were reduced using python language software modified from Schwartz et al. 32 .
Small area electron diffraction (SAED) imaging and analyses. SAED imaging of the enamel layer from wild type, Mmp20 −/− and Amelx −/− mice was done in the postsecretory transition to early maturation region (between Levels 3-4) using blocks prepared in the transverse plane of the incisor for preparing thin sections cut at 85 nm on a ultramicrotome and mounted on 300 mesh grids, and blocks prepared in the sagittal plane for FIB-SEM milling at 90° into the block faces thereby producing thin lamella of about 100 nm thickness that were also then viewed in a transverse plane by TEM 33 . Electron diffraction imaging was done in a Tecnai F20 transmission electron microscope (FEI, Systems for Research, Longueuil, QC, Canada) equipped with TVIPS XF416 camera (Tietz Video and Image Processing Systems, Eremitenweg 1, D-82131, Gauting, Germany) using Seri-alEM protocol for acquisition 34 . Diffraction patterns were collected by locating and centering a feature of interest at × 7,800, inserting a 40 µm selected area aperture and collecting SAED patterns at a nominal camera length of 520 mm. Imaging in wild-type mice was done over dentin located near the DEJ and over single spots within the inner, middle, and outer thirds of the enamel layer. Imaging in Mmp20 −/− enamel was done near the DEJ, within the core of the mineral fans and in the outer disorganized layer at the surface. Imaging in Amelx −/− enamel was done near the DEJ, within the stalk of the fans and within the body of the fans. These sites of sampling are illustrated in Fig. 8

Results
Amelx, Ambn, and Enam transcript levels are normal in Mmp20 −/− mice. We confirmed by in situ hybridization that wild-type Mmp20 is expressed in ameloblasts and odontoblasts (Fig. 1A). We also demonstrate by quantitative real-time PCR (qPCR) that mRNA transcript levels for Ambn, Amelx, and Enam are not significantly different between WT and Mmp20 −/− mouse molars (Fig. 1B). Since this study addresses the relative contribution of MMP20 and its enamel matrix cleavage products to enamel formation, we confirm that in the absence of MMP20 other enamel matrix proteins are still transcribed and secreted into the enamel matrix. Both protein gels 28 and immunoblots 35 have demonstrated that enamel matrix proteins are secreted at similar levels in wild-type (WT) and Mmp20 −/− mice. Therefore, the enamel phenotype in Mmp20 −/− mice is not due to secondary effects from altered expression enamel matrix proteins.
Ameloblasts extend finger-like projections to penetrate the basement membrane and this occurs in Amelx −/− and Mmp20 −/− mice. In wild-type mice, disruption of the basement membrane between pre-ameloblasts and pre-odontoblasts occurs near the apical end of the incisor (Level 0.73, see "Materials and methods" for level definitions). Ameloblasts extend finger-like processes from their distal surface that breach the thinning basement membrane. As development proceeds, increasing numbers of these ameloblast www.nature.com/scientificreports/ processes penetrate the basement membrane and extend a short distance into predentin ( Fig. 2A, yellow arrows). A similar sequence of events occurs in Amelx −/− ameloblasts (Fig. 2B). The basement membrane contains Type IV collagen 7,36 and this collagen is an MMP20 substrate in vitro 37 . Therefore, we asked if MMP20 plays a role in ameloblast basement membrane penetration. We discovered that in the absence of MMP20, the basement membrane thins and is penetrated by ameloblast finger-like projections  3C) mice, dentin mineralization starts with small foci of mineral that generally initiate within several micrometers of the ameloblast layer near the eventual DEJ 8 . As these foci form, they grow in size and begin to merge with adjacent areas of mineralization. As this process proceeds, the area nearest the ameloblasts is the first to have a continuously mineralized dentin matrix. Periodic mineralized extensions also form that project toward the odontoblasts, and these extensions are typically associated with large collagen fiber bundles (Fig. 3A, green arrow). This same dentin mineralization process also occurs in Amelx −/− and Mmp20 −/− mice (Fig. 3B,C). Although odontoblasts express MMP20 (Fig. 1), the steps leading to dentin mineralization in Mmp20 −/− mice also occur in an apparently normal fashion (Fig. 3C).
A comparison of the initial enamel in WT, Amelx −/− , and Mmp20 −/− mice. WT enamel mineral ribbons initiate on preexisting dentin mineral, predominantly on the sides and tips of mineralized collagen fibers at the DEJ, and are associated with ameloblast finger-like extensions 8 . These ribbons extend from their origins on mineralized collagen along the path that the finger-like processes take when they retract back to the ameloblast cell membrane (Fig. 4A) 8 . Oriented enamel mineral ribbons also form on the dentin surface of Amelx −/− (Fig. 4B), and Mmp20 −/− (Fig. 4C) mice, but do not form on mice lacking enamelin 8 or ameloblastin 26 .
Subtle differences in the initial mineral ribbons of WT, Amelx −/− , and Mmp20 −/− mice become more obvious as mineralization progresses. In contrast to WT ribbon growth (Fig. 4A), the Amelx −/− enamel ribbons increase in density near the dentin surface when the ribbons are less than 3 µm in length (Fig. 4B,D). This increase in mineralization temporarily blurs the dentin-enamel boundary (Fig. 4B,D), and this results from the premature thickening of the mineral ribbons.
The Mmp20 −/− mineral ribbons appear as closely packed thin dark lines when seen on-edge or as broad, gray rectangular areas when seen in side view (Fig. 4C), presumably because they are coated with uncleaved amelogenins that adsorb onto mineral through their charged C-termini 38 , which MMP20 ordinarily cleaves soon after amelogenin secretion 39,40 . The binding of uncleaved amelogenins to the sides of the Mmp20 -/initial mineral After forming the initial enamel, Amelx −/− ameloblasts appear to form distal processes (Fig. 4D) and thereafter enamel mineralization deviates significantly from WT because the succeeding dysplastic enamel mineral layer lacks rod/interrod architecture and spaces of Weber. From the initial Amelx −/− enamel layer emerges an unevenly mineralized second layer comprised at its base of mineralized stems and branching structures that resemble the staves of Japanese fans (Fig. 5B). These fan-like structures elongate and then thicken into mineral plates (Figs. 6B, 7A-D). The fans appear to arise through the formation of multiple mineral branches from a relatively thin, mineral stem (Fig. 6B).
Previously the crystalline phase associated with the fans in Amelx −/− mice was shown to be octacalcium phosphate (OCP) 27 and here we confirm this result (Fig. 8 diffraction patterns at top, Fig. 9, and Supplemental data Fig. S6). The fans are disorganized, and tilt at angles toward the ameloblast layer and overlap. As development proceeds, the tips of the OCP fan staves are in close proximity with the distal membrane of the ameloblasts and thicken and merge only a short distance away (Fig. 7, top).
Dysplastic enamel formation in Mmp20 −/− mice also leads to the development of poor-quality enamel that contains large fan-like structures (Fig. 5C). However, in contrast to Amelx −/− mice, the mineral in Mmp20 −/− mice begins organizing almost immediately into "micro fans", some of which branch at or near the DEJ (Fig. 6C). As in Amelx −/− mice, the enamel mineral ribbons eventually thicken and apparently crystalize to form fan-like branches that grow to contact the distal membranes of ameloblasts (Fig. 7). At Level 3 the Amelx −/− and Mmp20 −/− bilayers (~ 20-24 µm in thickness) are both comprised of a forest-like array of fan-shaped plates ( Fig. 4A-C). SAED analyses indicate that the diffraction patterns for the fans in Mmp20 −/− mice are similar to diffraction patterns in Amelx −/− mice and the crystal lattice spacing for both patterns matches that of OCP (Fig. 9, Supplemental data Fig. S6). These findings are consistent with results from a recent study which showed that forming enamel in Mmp20 −/− mice contains a mixture of hydroxyapatite and OCP 44 .
From Levels 2.8 through 3.0 a third layer of dysplastic enamel is formed in Mmp20 −/− mouse incisors (Fig. 5D-G; Appendix Fig. S1). This pathological layer does not form by the extension of mineral ribbons, but by the deposition of amorphous mineral possibly facilitated by rapid secretion and surface pooling of kallikreinrelated peptidase-4 (KLK4) cleaved enamel proteins deposited from secretory granules (Fig. 5E,F). This mineralized layer does not form in Amelx −/− mice (Supplemental data Fig. S2), suggesting that amelogenins play an important role during the formation of this third mineralized layer in Mmp20 −/− mice. SAED analyses indicated  (Fig. 10A). The surface of the tips of the Mmp20 −/− fans are similarly covered by a thin granular mineralized layer (Fig. 8C,E). Subsequently, a third layer forms on top of the OCP fans ( Fig. 10B-D) that is poorly organized and increases the overall dysplastic enamel thickness by ~ 15-20 µm to a total thickness of ~ 40 µm (Figs. 5C,D, 8C, 10C,D). This outer layer is not comprised of oriented mineral crystals, but is poorly structured and its pathological formation is highlighted by its abnormal ion composition 45 . Nevertheless, SAED analyses indicate that the mineral phase present is not OCP, as present in the fans, but likely hydroxyapatite (Fig. 8, top row box 5 for fans and box 6 for outer layer; (see Supplemental data on SAED pattern matching). An interesting observation in the Mmp20 −/− outer layer is the occasional appearance of a less-dense area that that is shaped and oriented like an enamel rod (Fig. 8C), and appears to be deposited during the retrograde movement of a single ameloblast (Fig. 8C). The mineral in the outer layer does not form by the elongation of mineral ribbons at the ameloblast membrane, but by the secretion of an amelogenin-rich matrix that hardens into mineral. When these secretions do not mix with those secreted by adjacent cells, the resulting mineral tracks the path of the retreating ameloblast, but is not comprised of oriented enamel mineral ribbons, like an enamel rod.
Ultrastructure of Mmp20 −/− secretory stage mouse incisor ameloblasts. The most significant difference between ameloblasts from WT and Mmp20 −/− mice is the absence of a functional Tomes' processes and the presence of anomalous structures in ameloblasts. Many secretory stage ameloblasts from Mmp20 −/− mouse incisors contain enlarged lysosomes linearly arrayed between the nucleus and the distal surface, with others at supranuclear and infranuclear locations (Fig. 11). Some ameloblasts show enlarged Golgi (Fig. 12A) and abundant, large mitochondria (Fig. 12B), consistent with significant levels of protein synthesis (Fig. 12). Many ameloblasts contain swollen endoplasmic reticulum 46,47 (Fig. 12C,E-G), and occasionally there is a granulated ameloblast 48 (Fig. 12E). At times, secreted material that contributes to the formation of the outer layer appears to accumulate laterally or in the matrix along the cell borders (Figs. 8F, 12D). The Mmp20 −/− transition ended at Level 3.2-3.3, after which the ameloblasts showed clear vacuoles that appeared to be depleted of spherical bodies (Fig. 12H).

Discussion
We conducted in-depth ultrastructural analyses of dental enamel formation in mouse mandibular incisors from Amelx −/− and Mmp20 −/− mice. Montages of magnified images (1500 or 2500x) show 13 (~ 90 µm) segments of Amelx −/− and 24 segments of Mmp20 −/− from continuously erupting mandibular incisors that were analyzed by FIB-SEM (Supplemental data Figs. S1 and S2). All early events occurred normally in the Amelx −/− and Mmp20 −/− mice, including deposition of collagen-rich predentin, basement membrane degradation, extension of ameloblast fingerlike processes into predentin, the appearance of mineral foci in predentin near the ameloblast layer, and the coalescing of mineral islands into a continuous layer of mineralized dentin prior to enamel ribbon formation 8,9 . www.nature.com/scientificreports/ An important finding was the formation of initial enamel mineral ribbons extending from the dentin surface to the ameloblast distal membrane in both Amelx −/− and Mmp20 −/− mice. The formation of oriented enamel mineral ribbons initiating on the dentin surface and extending to the ameloblast membrane is a defining feature of dental enamel formation. As previously stated, enamel mineral ribbons do not form in mice lacking enamelin 8 or ameloblastin 26 . Interestingly, the gar makes enamel that starts as mineral ribbons and has Ambn and Enam genes, but has no amelogenin gene 49,50 . Therefore, neither amelogenin nor MMP20 likely play a role in initial enamel ribbon nucleation, shaping or orientation in vivo, as well as when this process continues a short distance away from the ameloblast membrane to provide overall lengthening of enamel crystallites through appositional growth of the enamel layer. www.nature.com/scientificreports/ Even so, amelogenin comprises ~ 90% of enamel matrix protein 51 and is a critical enamel protein. Without amelogenins, the initially oriented murine mineral ribbons prematurely increase in density and thickness when they are only a few micrometers long. By mid-secretory stage (Level 2) the Amelx −/− initial enamel is already more highly mineralized than dentin. The similarities between the dysplastic enamel formed in the second mineral layer of the Amelx −/− and Mmp20 −/− mice, which both feature the formation of distinctive OCP fans terminating 20-24 µm above the dentin surface, is truly extraordinary. Since Mmp20 −/− mouse molars express normal levels of enamel matrix proteins, the only common deficit between the Amelx −/− and Mmp20 −/− mice is the absence of amelogenin cleavage products.
Previously it was shown that when the most abundant mouse amelogenin splice product (M180) was expressed in the Amelx −/− background, removing MMP20 reduced enamel hardness and quality of the enamel architecture, suggesting that amelogenin cleavage products might be necessary for proper enamel formation 52 . Furthermore, physical and electron microscopy studies have demonstrated that enamel protein cleavage products are different from their original secretory forms and may show distinct spatial distribution patterns within the enamel layer 53 . Since the present FIB-SEM study demonstrated that OCP fans form in the absence of amelogenin and in the absence of MMP20 (neither have amelogenin cleavage products), this is the first study to specifically implicate amelogenin cleavage products as important for proper hydroxyapatite formation and presents in vivo evidence that MMP20 is not simply degrading enamel proteins, but serves a major role by processing amelogenins into cleavage products that serve important secondary functions.
After formation of the initial enamel fan-like stems in Amelx −/− and Mmp20 −/− mice, it appeared as though the ameloblasts lost their ability to transport secreted ions onto the tips of the existing mineral ribbons so that they would mostly elongate rather than thicken into fan-like structures. Or perhaps the environmental conditions favored the direct formation of OCP fan structures. Regardless, the fan structures appeared to occur at about the same time in Amelx −/− and Mmp20 −/− mice. This anomalous mineralization continued in the same way in Amelx −/− and Mmp20 −/− mice and was therefore independent of enamelin and ameloblastin proteolytic processing. Although this occurred in an atypical environment, it may suggest that the most significant function of MMP20 is the processing of amelogenins so that the enamel mineral ribbons can reach their full extension and covert into HAP at an appropriate time.
A notable difference between the Amelx −/− and Mmp20 −/− mice was the formation of an outer disordered HAP layer exclusively in Mmp20 −/− enamel. The absence of this layer in Amelx −/− mice indicates that the expression of amelogenin, and perhaps its KLK4 cleavage products, are critical for its formation in Mmp20 −/− mice. The outer layer was previously evident in Mmp20 −/− mice 28,45 , but was comprised mostly of surface nodules 54 . Before this study, we bred the Amelx −/− and Mmp20 −/− mice back into the C57BL/6 background, which may account for this difference. Our ultrastructural analyses demonstrate that the Mmp20 −/− outer HAP layer does not form by the elongation of mineral ribbons along the secretory surface of the ameloblast membrane. Although it is unstructured, it can become highly mineralized.
From interpretation of these data a new theory of dental enamel formation can be deduced. First, a feedback mechanism to downregulate enamel matrix protein transcription when the forming enamel is flush with proteins likely does not exist. Second, amelogenin MMP20-mediated cleavage products, but not uncleaved amelogenins, are likely responsible for preventing development of OCP-based fan structures. Cleaved amelogenins may also be directly or indirectly responsible for successful Tomes' process formation necessary for the development of rod and interrod areas within the developing enamel. Third, enamel mineral ribbons initiate and elongate to form the initial enamel in the total absence of amelogenin or MMP20, and as a consequence, the absence of enamelin and ameloblastin cleavage products. Overall, this study brings to light several mechanisms of enamel formation, such as the important role of amelogenin cleavage products, that provide new insights into the molecular mechanisms of dental enamel formation.