Central Role of Pyrophosphate in Acellular Cementum Formation

Background Inorganic pyrophosphate (PPi) is a physiologic inhibitor of hydroxyapatite mineral precipitation involved in regulating mineralized tissue development and pathologic calcification. Local levels of PPi are controlled by antagonistic functions of factors that decrease PPi and promote mineralization (tissue-nonspecific alkaline phosphatase, Alpl/TNAP), and those that increase local PPi and restrict mineralization (progressive ankylosis protein, ANK; ectonucleotide pyrophosphatase phosphodiesterase-1, NPP1). The cementum enveloping the tooth root is essential for tooth function by providing attachment to the surrounding bone via the nonmineralized periodontal ligament. At present, the developmental regulation of cementum remains poorly understood, hampering efforts for regeneration. To elucidate the role of PPi in cementum formation, we analyzed root development in knock-out (−/−) mice featuring PPi dysregulation. Results Excess PPi in the Alpl−/− mouse inhibited cementum formation, causing root detachment consistent with premature tooth loss in the human condition hypophosphatasia, though cementoblast phenotype was unperturbed. Deficient PPi in both Ank and Enpp1 −/− mice significantly increased cementum apposition and overall thickness more than 12-fold vs. controls, while dentin and cellular cementum were unaltered. Though PPi regulators are widely expressed, cementoblasts selectively expressed greater ANK and NPP1 along the root surface, and dramatically increased ANK or NPP1 in models of reduced PPi output, in compensatory fashion. In vitro mechanistic studies confirmed that under low PPi mineralizing conditions, cementoblasts increased Ank (5-fold) and Enpp1 (20-fold), while increasing PPi inhibited mineralization and associated increases in Ank and Enpp1 mRNA. Conclusions Results from these studies demonstrate a novel developmental regulation of acellular cementum, wherein cementoblasts tune cementogenesis by modulating local levels of PPi, directing and regulating mineral apposition. These findings underscore developmental differences in acellular versus cellular cementum, and suggest new approaches for cementum regeneration.


Introduction
The mineralized tissues of the teeth and skeleton are subject to homeostasis of inorganic phosphate (P i ) for normal development and maintenance [1]. The hydroxyapatite (HAP) deposited to mineralize these hard tissues is a compound of P i and ionic calcium. Pyrophosphate (PP i ), composed of two molecules of P i , functions as a pivotal regulator of physiological mineralization and pathologic calcification by acting as a potent inhibitor of HAP crystal precipitation [2][3][4][5]. Though the potential for PP i to inhibit biological mineralization is clear from in vitro experiments, the in vivo role and regulation of PP i has been more difficult to elucidate. Through study of the heritable conditions such as hypophospha-tasia (HPP), spontaneous mutations, and directed gene ablations in mouse models, the key regulators of PP i have been identified, and their roles in shaping mineralized tissues have been partially defined. As measurement of PP i in vivo at mineralization fronts is not possible, the analysis of cellular proteins that manufacture, transport, or degrade PP i has served to clarify the mechanisms for PP i modulation, in conjunction with in vitro experiments.
Local tissue concentrations of PP i are controlled by a number of regulatory enzymes and transporters. Tissue nonspecific alkaline phosphatase (TNAP) is an ectoenzyme capable of hydrolyzing PP i and providing P i [6]. TNAP is expressed by mineralizing cells of bones and teeth, and is critical for proper skeletal mineralization [5,7]. Hydrolysis by alkaline phosphatase activity (ALP) thus provides a mechanism for clearance of PP i , allowing mineralization to proceed. Loss of function mutations in the TNAP gene Alpl cause hypophosphatasia (HPP), a disease marked by poor bone mineralization, rickets, and osteomalacia, as well as tooth phenotypes [8,9]. Ablation of the homologous mouse gene Alpl (formerly Akp2) produces a phenotype consistent with increased PP i and mineralization disorders of infantile HPP [10,11]. Conversely, two factors have been identified which increase local PP i in tissues. The progressive ankylosis gene (Ank; Ankh in humans) encodes a multipass transmembrane protein that regulates transport of intracellular PP i to the extracellular space [12][13][14]. Ectonucleotide pyrophosphatase phosphodiesterase 1 (NPP1; encoded by the Enpp1 gene) also works to increase extracellular PP i by hydrolysis of nucleotide triphosphates [15]. PP i removal by ALP activity thus antagonizes provision of PP i by ANK and NPP1, thereby creating a concerted regulation of P i and PP i levels ( Figure 1), and ultimately, mineralization [16,17].
The cementum covering the tooth root provides attachment for the tooth proper to surrounding alveolar bone, via the nonmineralized periodontal ligament (PDL) [18][19][20]. Cementum was first linked to PP i metabolism by the condition HPP, where premature tooth exfoliation was discovered to result from developmental cementum aplasia or hypoplasia, and thus poor periodontal attachment [7,21,22]. Intriguingly, studies to date suggest the acellular cementum (acellular extrinsic fiber cementum, AEFC) of the cervical portion of the root is severely affected by PP i dysregulation, while the apically located cellular cementum (cellular intrinsic fiber cementum, CIFC) is unaffected, or much less so [21,23]. Proper cementum formation is critical for dentoalveolar function, though cementogenesis remains poorly understood in terms of associated cells and regulatory factors involved. This is especially true in regard to differences between the acellular and cellular varieties, and how cementum differs developmentally from other hard tissues, bone and dentin [18,19].
To address how the process of cementogenesis is shaped by PP i metabolism, a set of studies was designed that employed in vivo transgenic mouse models featuring disrupted PP i regulation, as well as in vitro approaches using a cementoblast cell line for further mechanistic studies.

Results
In order to develop a comprehensive understanding of how PP i regulates tooth root development, we performed a detailed histological study of developing first mandibular molars and incisors of mice harboring homozygous knock-out ( 2/2 ) of Alpl (high PP i ), Ank, or Enpp1 (low PP i ), compared to age-matched homozygous wild-type ( +/+ ) controls. Days were selected to capture developmental time points of interest during molar root formation, i.e., during acellular cementogenesis (14 days postnatal, dpn), at completion of the root and following cellular cementogenesis (26 dpn), and after more than a month in occlusion (60 dpn). Alpl 2/2 mice were limited to a maximum age of 21 dpn because of shortened lifespan. Morphological observations on H&E stained sections were paired with in situ hybridization (ISH) and immunohistochemistry (IHC) for selected mineralized tissueassociated factors.

Acellular cementogenesis requires diminution of pyrophosphate
In the infantile form of HPP, the skeleton is properly mineralized at birth, but postnatal skeletogenesis is compromised [7]. Alpl 2/2 mice phenocopy aspects of infantile HPP, where loss of TNAP was previously reported to have little effect on bone until postnatal day 6 [10,24]. At 14 dpn, the majority of alveolar and mandibular bone in Alpl 2/2 mice was well developed, though signs of hyperosteoidosis were noted in the bone adjacent to the molar root (Figure 2A and B). In Alpl +/+ molars, acellular cementum (AEFC) covered the root dentin as a thin and uniform basophilic layer. Alpl 2/2 molars were marked by disruption of acellular cementum, visible as reduction of the basophilic layer (cementum aplasia or severe hypoplasia) and direct contact of PDL cells and tissues with dentin. By 21 dpn this cementum defect was sometimes associated with tearing at the PDL-AEFC interface, suggesting poor integration of Sharpey's fibers at the root surface (not seen at the PDL-bone interface) ( Figure 2C and D) and consistent with HPP case reports observing premature tooth exfoliation. This is not likely to be a processing artifact, as infiltrating cells were present in the tear zone. These results agree with AEFC disruption described in this Alpl 2/2 model [25], as well as a different TNAP loss-of-function mouse [23].
To further investigate the mechanism for the cementum defect in Alpl 2/2 mice, IHC was performed for two cementum markers, extracellular matrix (ECM) proteins bone sialoprotein (BSP) and osteopontin (OPN), which are present at high concentrations in acellular cementum of controls ( Figure 2E and G). Both BSP and OPN immune localization were disrupted on the Alpl 2/2 root surface ( Figure 2F and H), compared to the strong, even staining on Alpl +/+ controls. Scanning electron microscopy (SEM) provided improved resolution to explore the root surface. While Alpl +/+ molars displayed a cementum layer on the root dentin surface, this layer was absent in the Alpl 2/2 molar ( Figure 3). The disruption of cementum initiation and concomitant lack of BSP and OPN localization supports the hypothesis that high PP i in Alpl 2/2 is acting to inhibit cementogenesis and HAP apposition on the root surface.

Attenuation of pyrophosphate increases acellular cementum
Both Ank and Enpp1 2/2 mice are deficient in extracellular PP i , though by different mechanisms. In molars of both null mice at Inorganic phosphate (P i ) is a component of mineral hydroxyapatite (HAP), while pyrophosphate (PP i ) is a potent inhibitor of HAP crystal precipitation and growth. The enzyme tissue nonspecific alkaline phosphatase (TNAP) hydrolyzes PP i to release ionic P i , creating conditions conducive for mineralization. Local PP i is increased by the functions of the progressive ankylosis protein (ANK) and ectonucleotide pyrophosphatase phosphodiesterase 1 (NPP1), which act to keep the mineralization process in check. doi:10.1371/journal.pone.0038393.g001 14 dpn, the developing cervical cementum was expanded (hypercementosis) compared to Ank and Enpp1 +/+ controls ( Figure 4A-C). At the completion of root development at 26 dpn, both Ank and Enpp1 2/2 molars featured a nearly identical cementum phenotype where cervical cementum width was expanded several fold over controls ( Figure 4D-F). This thick cervical cementum included numerous cell inclusions in the matrix, in a region that is typically acellular type cementum (AEFC). Intriguingly, for both homozygous knock-out models, apical cementum (CIFC) was not morphologically different from controls ( Figure 4G-I), PDL space remained unmineralized, and dentin was not altered compared to Ank and Enpp1 +/+ mice.
The incisor in the mouse is divided into a (labial) crown analogue featuring enamel, and a (lingual) root analogue featuring strictly AEFC type cementum. Notably, histological changes in Ank and Enpp1 2/2 incisors paralleled those in molars, featuring expanded cementum ( Figure 4J-L). Sagittal sections of the mandible allowed observation of all three molars. Loss of Ank affected all molars similarly, with thickened cementum evident on all root surfaces compared to controls ( Figure 4M-R). The fact that acellular cementum on all murine teeth was similarly affected by reduced PP i supports this as a central molecular regulator of cementogenesis which is not tooth-or stage-specific in its influence.
Both Ank and Enpp1 2/2 mice featured a hypercementosis phenotype, indicating both PP i regulators function in controlling cementum formation. Comparative analysis between Ank and Enpp1 2/2 and their respective controls was accomplished by measuring the growth rate of cervical cementum over time. During early root formation between 14 and 26 dpn, Ank and Enpp1 2/2 molars featured at least 10-fold greater cementogenesis compared to controls ( Figure 5A). Ank and Enpp1 2/2 cementum continued to increase at a rate of 0.2-0.7 mm/day from 26 to 60 dpn, while over the same period, controls featured tightly controlled apposition, growing at the much slower pace of 0.01-0.05 mm/day.
While cementum was dramatically affected by loss of ANK or NPP1, dramatic changes in other tissues were not observed. Histomorphometry at age 26 dpn was performed to measure cross-sectional widths to determine if PDL and alveolar bone were affected. Cementum was significantly increased in both null models, with Ank 2/2 at 14-fold and Enpp1 2/2 at more than 13fold the width of age-matched controls ( Figure 5B). A direct comparison of the two homozygous knock-out models revealed that Ank 2/2 featured slightly, but significantly, thicker cementum at the age sampled. Histomorphometry confirmed that PDL space was maintained in both null models, even significantly larger in Ank 2/2 vs. +/+ , despite exuberant cementogenesis. Alveolar bone on the lingual aspect tended towards reduced cross sectional dimension in both Ank and Enpp 2/2 models, though the effect was not statistically significant as measured here. Tartrate resistant acid phosphatase (TRAP) staining confirmed increased numbers of osteoclast-like cells (TRAP positive, multinucleated) on the bone surface adjacent to the tooth root in Ank 2/2 molars [26]. A modeling/remodeling of bone away from the root provides a mechanism for maintenance of the PDL in the face of expanding cementum.
One of the key functional characteristics of the cervical cementum is the extrinsic nature of the collagen fibers, which serve to anchor the tooth to surrounding alveolar bone. Picrosirius red staining in association with polarized light microscopy was used to visualize the birefringent collagen fibers of the periodontia [27]. The thick cementum of Ank and Enpp1 2/2 molars featured a high concentration of extrinsic collagen fibers, which were continuous with the fibers in the PDL proper ( Figure 6B and D). As this thick cementum in the null molars features dense extrinsic collagen fibers, but also contains numerous cell inclusions, it could properly be labeled cellular extrinsic fiber cementum (CEFC), a form of cementum not typical for cervical molar roots, and furthermore, not previously described in the cementum family. Importantly, the observation of an ongoing, progressive apposition on the root surfaces of Ank and Enpp1 2/2 mice confirms this is thickening of the normally present extrinsic fiber cementum, and is not likely to be a different type of ectopic calcification on the root surface. As a comparison, Alpl 2/2 molars were examined, and confirmed tearing at the root-PDL interface, osteoid invasion of the PDL space, and poorly organized and sparsely embedded collagen fibers at the cervical root ( Figure 6F).
Cementum, bone, and dentin are also characterized by their extracellular matrix (ECM) protein composition, and these ECM proteins contribute to crystal growth and regulation, and affect mechanical properties of these tissues. Because of the dramatic changes in cementum apposition, we investigated the ECM profile in PP i deficient mice. In the low PP i environment of the Ank and Enpp1 2/2 mice, the thick cervical cementum was marked by increased OPN and dentin matrix protein 1 (DMP1), proteins of the SIBLING family ( Figure 7A-F and G-L). OPN staining strongly labeled control acellular cementum, and was intensely expressed in the corresponding Ank and Enpp1 2/2 cervical cementum and associated cementoblast cells. DMP1, a marker for osteocytes, odontoblasts, and cementocytes, was present at low or undetectable levels in acellular cementum in controls, in contrast to intense localization in expanded Ank and Enpp1 2/2 cementum. OPN and DMP1 levels were not changed in Ank or Enpp1 2/2 apical cementum, as well as in other dentoalveolar locations. The source of the increased OPN and DMP1 protein was confirmed, as cementoblast gene expression for both Opn and Dmp1 mRNA was increased in Ank and Enpp 2/2 mice ( Figure 8A-C and D-F). OPN and DMP1 expression changes were not observed in other cell populations in the dentoalveolar complex in these mice. Another characteristic marker for cementum, BSP, was present in control and null cementum ( Figure 7M-R), and where protein concentration was diluted in the larger cementum volume of the Ank and Enpp1 2/2 mice, mRNA levels in cementoblasts were unaltered ( Figure 8G-I).
Thus, increased cementogenesis in Ank and Enpp1 2/2 teeth was linked to increased OPN and DMP1 specifically in cervical cementum. It is notable that OPN was increased in cementum as a result of reduced extracellular PP i . This change is opposite to the decreased OPN that has been documented in osteoblasts and articular locations in mice lacking ANK or NPP1 [14,16].
Cementoblasts express pyrophosphate regulators in a time and space restricted manner Acellular cementum was shown to be exceptionally sensitive to regulation by PP i ; with increased PP i (as in Alpl 2/2 mice) AEFC was severely inhibited, and under reduced PP i conditions (as in Ank and Enpp1 2/2 mice) cementum thickness increased significantly, a trend not reflected in other dental hard tissues. In order to understand the sensitivity of acellular cementum to PP i metabolism, we mapped the expression of TNAP, ANK, and NPP1 during tooth root formation. We also assayed these factors in all of the null models to determine if there were compensatory or antagonistic expression changes that would contribute to phenotypes under PP i dysregulation.
We previously reported wide expression of ANK gene and protein in the tooth and supporting tissues [26], paralleling previous findings that ANK is expressed in several tissues systemwide [12]. Using a refined immunohistochemistry technique, which allowed more sensitive identification of differential ANK protein localization, we discovered that after acellular cementum formed, ANK was labeled most intensely in cementoblasts lining the molar and incisor roots ( Figure 9D). Developmental localization of NPP1 protein was similar to that of ANK, with most intense staining found in cementoblasts ( Figure 9G). Both ANK and NPP1 stained weakly in other cells, including PDL cells, osteoblasts, and odontoblasts. Immunolocalization revealed compensatory up-regulation, where NPP1 was increased in Ank 2/2 and ANK was increased in Enpp1 2/2 ( Figure 9F and H). Most interestingly, the observed increase was found only in cementoblasts, and not in other cell populations of the dentoalveolar region. These data suggested that ANK and NPP1 were differentially expressed by cementoblasts and employed to tightly regulate PP i and developmental cementum apposition. However, Figure 3. Lack of acellular cementum on Alpl 2/2 molar root surfaces. Backcattered SEM was employed to explore the cervical root surface (white boxes) in (A) Alpl +/+ control and (B) Alpl 2/2 first molars. At higher magnification, the acellular cementum layer (white arrows) in the (C) control molar can be distinguished by contrast differences due to slightly lower mineralization than underlying dentin (d). (D) No acellular cementum layer was apparent in the cervical region of the Alpl 2/2 molar. Abbreviations: d = dentin; c = acellular cementum; p = periodontal ligament. doi:10.1371/journal.pone.0038393.g003 it still remained unclear by what mechanism PP i was controlling cementum apposition and ECM composition.
Pyrophosphate controls mineralization and coupled gene expression in cementoblast cultures PP i regulators ANK and NPP1 were preferentially expressed by cementoblasts after initiation of cementogenesis, and their expression was modulated under conditions of low extracellular PP i and increased apposition. Expression levels of cementum ECM proteins OPN and DMP1 were also responsive to PP i deficiency, reflecting the altered homeostasis of P i /PP i ratio or increased cementum apposition in Ank and Enpp1 2/2 . These data together suggested that cementoblasts associated with AEFC regulate PP i as a means to tightly control the process of apposition and related gene expression. In vitro experiments were performed to determine how these genes were regulated during mineral formation, and what potential role PP i played in their regulation. Because of the technical obstacles in isolating and identifying primary cementoblasts, we opted to use an immortalized cementoblast cell line (OCCM.30) and modulate exogenously added PP i . OCCM.30 cells were cultured in control media or mineralization media where 5 mM b-glycerophosphate (BGP) was added. BGP served as an organic P i source, mimicking similar sources in vivo and commonly used for in vitro mineralization experiments [30][31][32][33]. Cells receiving control media lacking BGP failed to mineralize during the course of the experiment. While cells cultured with BGP produced mineral nodules by day 6, with increased staining and calcium incorporation at day 8 ( Figure 10A and B).
Cells were introduced to exogenous PP i to create culture conditions of low (10 mM) and high (100 mM) PP i . The lower dose of 10 mM PP i did not affect mineralization, while the higher dose of 100 mM was confirmed as an inhibitor of mineral nodule formation under these conditions. While PP i is an inhibitor of HAP crystal precipitation, it has also been reported to have cell signaling effects in osteoblasts [4,14,16]. Neither dose of PP i affected OCCM.30 cell proliferation, viability, or collagen synthesis compared to controls ( Figure 11), therefore these processes were not indirectly affecting mineralization. Cementoblast ALP enzyme activity was uniform across treatments and times, and added 100 mM PP i did not appreciably affect ALP ( Figure 10C), indicating the effect of PP i on mineralization was not by inhibition of TNAP. An enzymatic assay for 59-nucleotide phosphodiesterase I and nucleotide pyrophosphohydrolase (NTPPPH) activity demonstrated significantly increased NPP1 function with mineralization at days 4, 6, and 8, while 100 mM PP i brought activity back to basal levels of non-mineralizing cultures ( Figure 10D). PP i associated and cementoblast marker genes were assayed by quantitative PCR. Under non-mineralizing conditions, Ank, Enpp1, Figure 5. Increased cementum apposition in Ank and Enpp1 2/2 teeth. (A) During early root formation between 14 and 26 dpn, both Ank and Enpp1 2/2 molars featured at least 10-fold greater cementogenesis compared to controls. From 26 to 60 dpn, Ank and Enpp1 2/2 cementum continued to increase at a rate of 0.2-0.7 mm/day, while Ank and Enpp1 +/+ controls featured tightly controlled apposition at the pace of 0.01-0.05 mm/day. (B) Histomorphometry confirmed Ank or Enpp1 2/2 cervical cementum was significantly increased compared to controls, while PDL width was maintained and alveolar bone thickness tended towards reduction. Values with the same letter were not significantly different, while different letters indicate a statistically significant intergroup (genotype) difference (p,0.05) as tested by ANOVA followed by the Tukey test for direct pair-wise comparisons. doi:10.1371/journal.pone.0038393.g005 Opn, and Dmp1 did not change over the course of the experiment ( Figure 12). However, all four genes increased significantly under mineralizing conditions at days 3 and 5, when mineral nodules were forming. At day 3, when increases were most dramatic, Ank increased almost 10-fold, Enpp1 increased 30-fold, Opn increased more than 30-fold, and Dmp1 increased 140-fold in mineralizing cultures compared to controls. These four genes also responded in parallel fashion to PP i . While inclusion of 10 mM PP i had a mild effect on gene expression compared to ascorbic acid (AA) + BGP cultures (paralleling effects on mineralization), the higher dose of 100 mM PP i significantly depressed Ank, Enpp1, Opn, and Dmp1 expression on day 3 compared to mineralizing cells. Cells receiving the 100 mM dose also maintained significantly lower expression of Ank, Enpp1, and Opn on day 5. By day 7, expression levels of the four genes were low, and there were no differences between any of the treatment conditions. Other cementoblast marker genes assayed, including Alpl, Bsp, and Col1, did not show a coherent pattern in response to addition of PP i . Notably, the increase in Enpp1 gene expression associated with mineralization corresponds to the increase in NTPPPHase activity recorded, and inclusion of PP i decreased mineralization and correspondingly decreased Enpp1 gene expression and NPP1 enzyme activity. In contrast, PP i did not perturb cementoblast mineralization by affecting Alpl expression or ALP activity.
These results showed that PP i regulated cementoblast mineralization and associated gene expression, in vitro. In an additional experiment of similar design, the addition of 100 mM PP i was discontinued in some wells midway through the experiment. Cells with 100 mM PP i for the duration did not mineralize, while cultures relieved of PP i inhibition at day 4 showed mineralization by day 6, increased Ank, Enpp1, Opn, and Dmp1 by day 5, coincident with mineralization ( Figure 13). This experiment demonstrated that even if PP i inhibited initiation of mineralization for the first 4 days, its removal facilitated both mineralization and concomitant gene expression. These results support expression of Ank, Enpp1, Opn, and Dmp1 as being functionally coupled to matrix mineralization, i.e. linked to changes in the mineralizing matrix. Importantly, these results parallel in vivo observations, where ANK, NPP1, OPN, and DMP1 were all increased by cementoblasts under conditions of reduced PP i , i.e. Ank or Enpp1 ablation.

Discussion
These studies aimed to define the regulatory role of PP i in tooth root cementum development. We demonstrate here that PP i serves as an essential regulator of tooth root acellular cementum development, and a key determinant defining the hard-soft interface between the cementum and PDL. Dysregulation of PP i resulting from loss of any of the central PP i controlling factors explored here had profound consequences on development of acellular extrinsic fiber cementum (AEFC), a tissue essential to tooth attachment and function. To wit, loss of TNAP caused severe underdevelopment or even absence of acellular cementum. Loss of either ANK or ENPP1 resulted in loss of control of cementum apposition, causing an exceptional hypercementosis. Because these three factors, TNAP, ANK, and NPP1, primarily adjust extracellular PP i , this strongly supports PP i as the key mechanistic factor uniting the cementum phenotypes in all three of these mouse models, prompting us to propose that PP i regulates acellular cementum in a molecular ''rheostat'' fashion, i.e. acellular cementum thickness relates inversely to PP i production.
Based on these collective data, we propose a model whereby PP i plays a central and novel role in acellular cementum formation ( Figure 14). The periodontal region is extremely rich in ALP activity (reducing local PP i ) and thus a permissive milieu for cementum formation on the root surface. In the course of normal development, cementoblasts modulate PP i to curb apposition (by increasing PP i via ANK and NPP1) to maintain AEFC as a thin tissue on the root surface. When one of these PP i factors is removed from the equation, apposition cannot be fully regulated and cementoblasts attempt to compensate by increasing expression of its counterpart PP i regulator. In addition to directly controlling cementum mineral apposition, these studies suggest PP i influences ECM protein composition; in the face of rapid cementogenesis, cementoblasts increased expression of OPN and DMP1. The increase in OPN, a negative regulator of HAP crystal growth, may be an additional mechanism cementoblasts employ to limit extent of cementum apposition. In vitro experiments support this interpretation of the role of PP i in controlling both mineral accumulation and cementoblast expression profile. What emerges is a portrait of acellular cementum as a mineralized tissue heavily governed by regulation of the physical-chemical process of mineral precipitation, and the cementoblast as a cell capable of directing PP i metabolism to promote and restrain cementogenesis.
On the role of pyrophosphate as a negative regulator of acellular cementum In these studies, we confirm that loss of TNAP function in the Alpl 2/2 mouse causes aplasia or severe hypoplasia of the acellular cementum. This is in line with a previous report from this and another model of TNAP loss-of-function [23,25], as well as reports from human hypophosphatasia (HPP) subjects [21,34], who harbor a mutation in the human homologue, Alpl [7]. We extend previous analyses of the Alpl 2/2 tooth cementum phenotype with gene and protein assays. Cementoblasts express similar levels of Bsp mRNA, while protein distribution of both BSP and OPN appear disrupted. We interpret these results to mean that cementoblast phenotype is maintained in the face of loss of TNAP, but that disruption of AEFC synthesis prevents accretion of BSP and OPN proteins on the root surface. Loss of OPN protein under conditions where acellular cementum was inhibited has been reported previously [25,35], and this observation makes sense because BSP and OPN are both mineral-binding members of the SIBLING family which play a role in the mineralization process [36], and in the close relationship of AEFC cementogenesis with the act of mineralization.
We show strong TNAP localization in the developing root region, and ALP activity has been reported to be strong in the periodontium, with highest activity adjacent to the mineralizing bone and developing cementum surfaces [29]. Moreover, the same   study identified a significant correlation between measured TNAP activity and acellular cementum thickness. The critical influence of TNAP on cementum apposition is likely by clearance of mineralization inhibitor PP i , rather than by providing local P i for HAP precipitation, for several reasons. First, circulating PP i is in the micromolar range, while P i is much higher in the millimolar range, so in the highly vascular periodontal region hydrolysis of PP i is not likely to appreciably increase P i available for mineralization, though local, compartmentalized ionic dynamics in vivo are difficult to predict. Secondly, mouse models of hypophosphatemia described to date tend to feature bone, dentin, and cellular cementum disorders, while acellular cementum is less affected [37,38]. Thirdly, in studies employing a PP i analog, 1hydroxyethylidene-1, 1-bisphosphonate (HEBP), it was found that HEBP inhibited formation of acellular cementum entirely, while cellular cementum and bone matrices were produced, but remained unmineralized [35,39,40]. A parallel pattern emerged when mineralization inhibitor matrix gla protein (MGP) was ectopically expressed in bones and teeth; bone, dentin, and cellular cementum matrices were produced yet remained unmineralized, while AEFC was absent [41]; MGP and PP i may have parallel functions as mineral regulators throughout the body. These studies indicate that cementogenesis depends heavily on creation of a physicochemical environment conducive for apposition, such as by PP i clearance.

Diminished pyrophosphate relieves the negative regulation on cementogenesis
Further evidence for PP i as a central regulator of cementum thickness was garnered from studying models with deficient PP i , the Ank and Enpp1 2/2 mice. In these mice, a progressive thickening of AEFC was found during root development, which corroborated previous findings in mice harboring suspected lossof-function mutations in these genes [42,43]. By completion of root formation, these null models exhibited 12-fold or greater AEFC vs. controls, with a significantly increased rate of apposition over the developmental time period. Importantly, we have shown this expanded cervical cementum shares the same mineral and mechanical properties as WT controls [26,44,45]. This is strong evidence that PP i is a key factor controlling acellular cementum formation, for several reasons. Firstly, ANK and NPP1 are membrane-bound proteins, which have been identified as primary regulators of extracellular PP i concentrations around mineralizing cell types, as well as elsewhere in the body. However, they operate by different mechanisms, with ANK affecting PP i transport and NPP1 acting as an ectoenzyme, producing PP i through catalysis of trinucleosides. The common link in functions of both these proteins is extracellular PP i production. That nearly identical AEFC phenotypes result from ablation of either of these genes is potent evidence for the indispensable role of PP i in influencing acellular cementum formation. Though ANK and NPP1 share similarity in function by increasing extracellular PP i , loss of NPP1 causes a more severe skeletal hypermineralization phenotype in mice, a difference possibly related to inclusion of NPP1 in matrix vesicles, whereas ANK was found to be absent in matrix vesicles [16]. It is intriguing then that loss of ANK or NPP1 had nearly identical phenotypic results on acellular cementum, a tissue where there is no clear role of matrix vesicles in mineralization. Secondly, ANK and NPP1 are expressed in the dentoalveolar region during tooth formation and cementogenesis. While both ANK and NPP1 are widely expressed throughout the body, both were found to be selectively more highly expressed in cementoblasts lining the tooth root. Also, a special importance for PP i production in regulating cementum was indicated indirectly by findings that human PDL tissue expresses significantly higher basal levels of TNAP, ANK, and NPP1 than pulp [21,46]. This hypothesis is supported by the finding that cementoblasts dramatically increased either ANK or NPP1 expression in response to loss of the other factor, likely an attempt to compensate for lack of extracellular PP i output in these mice. The nature of the interaction between ANK and NPP1 in tooth formation is currently the subject of study in a series of doubledeficient mice. That expression of these PP i regulating factors is enriched in tooth root and they are inducible in each others' absence supports a central physiologic function for PP i in normal control of cementogenesis. The possible involvement of other complementary and antagonistic factors in PP i homeostasis in the root region is an intriguing question currently being studied. One candidate is CD73, a cell surface protein operating downstream of NPP1 which may regulate Alpl expression, and that has been linked to vascular calcification [47].
This essential role of PP i , however, seems to be limited to the cervical acellular cementum. ANK and NPP1 were not as consistently localized to regions of apical cementum, did not exhibit compensatory up-regulation in the apical portion of knockout molars, and loss of ANK and NPP1 did not impact the phenotype of CIFC. In this respect, the cellular cementum showed a clear difference in developmental regulation from AEFC and more similarity to alveolar bone. Similarly, loss of TNAP and the resulting increased PP i affected cellular cementum and bone in similar ways.

On the influence of pyrophosphate metabolism on cementum extracellular matrix composition
Reduced PP i not only resulted in more rapid AEFC apposition, but also led to altered cementoblast gene expression and matrix composition. OPN and DMP1, both mineral-regulating ECM proteins from the SIBLING family [36], were increased at the gene and protein level in Ank and Enpp1 2/2 cementoblasts. Expression of Bsp, a key cementoblast marker and SIBLING family member was unaltered in Ank and Enpp1 2/2 teeth. BSP immunostaining indicated a diffuse presence in the thick cementum, likely diluted in relatively greater volume of mineralized cementum. Bsp expression in cementoblasts in vitro was unaffected during mineralization or its inhibition by PP i .
OPN is a multifunctional ECM protein and a marker for cementum [48][49][50]. OPN has been shown in vitro to be an inhibitor of hydroxyapatite mineral crystal growth [51,52], an observation supported by study of the Opn 2/2 (Spp1 2/2 ) mouse [53], as well as other models where increased OPN was found to disrupt skeletal mineralization [16,54]. We suggest that increased expression of OPN by Ank and Enpp 2/2 cementoblasts represents an additional mechanism for control of apposition; like NPP1, an attempt at normalization of the cementogenesis process. While previous studies using osteoblasts have cited PP i as a signal increasing Opn expression [4,16], we found here that cementoblasts exposed to PP i in vitro significantly reduced Opn expression during the mineralizing phase of the experiment. The cementoblast reaction to increase OPN in response to mineralization under low PP i conditions, is opposite that of osteoblasts, which were found to reduce OPN in Ank and Enpp1 2/2 mice, contributing to the bone pathology. The divergent response underscores the unique mineral metabolism of cementum.
DMP1 is highly expressed in the osteocytes embedded in bone matrix, and is associated with maintenance of the lacunarcanalicular system of these cells [55][56][57]. DMP1 was increased in osteocytes in loaded bone, perhaps functioning in the mechanical response [58,59]. In the context of increased cementum apposition in Ank and Enpp1 2/2 mice, we hypothesize that induction of Dmp1 gene expression reflects rapid apposition and embedding of cervical root cementoblasts as cementocytes. The cementoblasts that direct AEFC normally remain as lining cells adjacent to this thin tissue. In vitro, mineralizing cementoblasts also increased Dmp1 gene almost 140-fold, paralleling other studies where DMP1 was induced in periodontal ligament cells in mineralizing 3-dimensional gels [60].
The mechanism for altered Opn and Dmp1 gene and protein expression is unknown, but under investigation. Based on in vivo and in vitro data, we propose an ''outside-in'' type of matrix-cell signaling mechanism whereby increased cementum apposition switches on expression of Opn and Dmp1, as well as Ank and Enpp1.

Cementoblasts as pyrophosphate sensitive cells
Localization of PP i regulators over the course of tooth development supported a role for PP i in modulating cementogenesis. The developing periodontal region shows strong immunolocalization of TNAP and high ALP activity, in effect producing a highly pro-mineralization environment. These are favorable circumstances for apposition of the cementum layer on the root dentin surface. However, a question that arises is how cementum may remain a thin and slow growing mineralized layer in such an environment permissive for mineralization. ANK and NPP1 localized most strongly to the cementoblasts of the AEFC following cementum formation, suggesting initiation of cementogenesis occurs under the influence of TNAP activity, but that after cementum deposition (usually several mm in mice), cementoblasts increase ANK and NPP1 to restrict further cementum apposition. Thus, in a scenario where either ANK or NPP1 function is lost, cementum apposition is not adequately controlled and the other is up-regulated, along with increased OPN, in an attempt to regain homeostasis of cementum.
In vitro experiments employing a cementoblast cell line provided a mechanistic platform for probing these proposed roles of ANK, NPP1, and PP i in cementoblast mineralization and gene expres- Figure 12. Pyrophosphate regulates cementoblast mineralization-coupled gene expression. Mineralizing cultures (AA + BGP) increased expression of Ank, Enpp1, Opn, and Dmp1 at days 3 and 5, concurrent with mineralization. The higher dose of 100 mM PP i significantly depressed expression of all four genes on day 3 compared to mineralizing cultures; Ank, Enpp1, and Opn suppression was maintained on day 5. The lower dose of 10 mM PP i showed milder effects on all four genes at day 3. Differences were not maintained by day 7. Unlike expression of Ank, Enpp1, Opn, and Dmp1, where PP i was able to block mineralization-associated induction, additional markers Alpl, Bsp, and Col1 were not regulated in coordinated fashion by mineralization or inclusion of either dose of PP i . Graphs show mean +/2 SD for n = 3 samples. Lowercase letters indicate treatment comparison at each time point, where different letters indicate a statistically significant intergroup difference. Uppercase letters indicate comparisons over time in the same treatment group, where different letters indicate a statistically significant intragroup difference. Values sharing the same uppercase or lowercase letter in were not significantly different. Means were compared by ANOVA (p,0.05) followed by the Tukey test for direct pairwise comparisons. doi:10.1371/journal.pone.0038393.g012 sion. The response for cementoblasts to increase ANK, NPP1, and OPN in light of rapid apposition was supported by in vitro experiments employing a cementoblast cell line. Results from the studies described here demonstrated that gene expression of PP i regulators Ank and Enpp1 and ECM proteins Opn and Dmp1 are functionally coupled to the mineralization process, increasing under mineralizing conditions and coincident with mineral nodule formation. Under conditions of higher PP i , mineral apposition was hindered, and expression of Ank, Enpp1 (and Opn and Dmp1) was blocked in dose-response fashion. These in vitro data together with Ank and Enpp1 2/2 mouse phenotypes, as well as ANK and NPP1 tissue localization, support the hypothesis that PP i modulation is employed by cementoblasts to guide the relative amount of cementum formed.
The question can be raised as to whether or not these experiments were carried out within a relevant physiological range of PP i . Circulating PP i in normal individuals has been measured in the low mM range. We carried out experiments with 10 and 100 mM added PP i , and justified these doses based on the following considerations. Firstly, our primary interest was to determine if PP i at a dose that could inhibit mineralization would also influence cementoblast gene expression. Using these cells under these culture conditions, the lower dose of 10 mM PP i was insufficient in blocking mineralization in these cells. Mineralization is affected by multiple factors including cellular activities, cell culture media, Ca 2+ and P i availability, as well as known and unknown factors present in fetal bovine serum (FBS) supplemented to the media; therefore, it is difficult to directly compare doses across studies where one or more of these variables may differ, especially cell type and time points examined [4,16]. Based on these criteria, we employed the higher dose of 100 mM PP i to create high PP i conditions for cells. Secondly, while circulating PP i levels are reported, it remains unknown how local, pericellular concentrations of PP i may vary. Biomineralization is well known to be a process dependent on compartmentalization, i.e. creation of localized, protected regions conducive to mineral precipitation. Therefore, we reasoned that cementoblasts, cells that show high in vivo and in vitro expression of PP i regulators ANK and NPP1, could potentially create localized, high PP i conditions to mediate biomineralization-related activities. This hypothesis supported using a higher exogenous dose of PP i in order to effectively regulate OCCM.30-mediated mineralization in vitro.

Insights into acellular and cellular cementum development and regeneration
In mammals, the fibrous connection of the tooth to the bony socket is classified as a gomphosis, or fibrous joint, and is unique in the body in that the periodontal ligament joins bone on one side to a non-bone substance on the other side, in this case the tooth root cementum [61]. The gomphosis attachment is unique to mammals and crocodilian reptiles, having developed from more ancient forms of tooth attachment such as direct ankylosis to bone [62][63][64][65]. This interposed ligament was made possible likely by a combination of events such as alterations in the HERS during root development, changes in the supporting bone (''bone of attachment''), and the rise of a unique tissue, the cementum, as well as diminution of mineralization in the region between bones and teeth [66]. The development of a mineral-free PDL region between mineralized bones and teeth requires localized expression of factors to establish a mineralization boundary at the hard-soft tissue interface and continue to maintain PDL space throughout the life of the tooth. In these studies, we demonstrate that maintenance of that hard-soft interface at the tooth root surface Quantitative calcium assay performed on days 4, 6, and 8 confirmed visual mineral nodule staining by von Kossa. (C) Mineralizing cultures (AA + BGP) increased expression of Ank, Enpp1, Opn, and Dmp1 at day 3, concurrent with mineralization. Inclusion of 100 mM PP i significantly depressed expression of Ank, Enpp1, and Opn on day 3 compared to mineralizing cultures. Removal of PP i on day 4 led to increased Ank, Enpp1, Opn, and Dmp1 on day 5, coincident with mineralization. Graphs in (B) and (C) show mean +/2 SD for n = 3 samples. Lowercase letters indicate treatment comparison at each time point, where different letters indicate a statistically significant intergroup difference. Uppercase letters indicate comparisons over time in the same treatment group, where different letters indicate a statistically significant intragroup difference. Values sharing the same uppercase or lowercase letter in were not significantly different. Means were compared by ANOVA (p,0.05) followed by the Tukey test for direct pair-wise comparisons. doi:10.1371/journal.pone.0038393.g013 Figure 14. Model for the hypothesized role of in formation of acellular cementum. Cementum apposition depends on precipitation of calcium (Ca 2+ ) and phosphate (P i ) ions on the root surface, and pyrophosphate (PP i ) acts as a potent inhibitor of hydroxyapatite crystal precipitation. Local pericellular PP i concentration is controlled primarily by three cellular factors: Tissue nonspecific alkaline phosphatase (TNAP; hydrolyzes PP i to P i ), progressive ankylosis protein (ANK; regulates transport of PP i from the intracellular to extracellular space), and ectonucleotide pyrophosphatase phosphodiesterase 1 (NPP1; produces PP i from hydrolysis of nucleotide triphosphates). Cementoblasts express TNAP, ANK, and NPP1 in order to regulate local PP i homeostasis and control the amount of cementum apposition. Cementoblasts are responsive to apposition by some type of outside-in feedback mechanism, and are capable of modulating expression of PP i regulators as well as production of secreted matrix proteins such as OPN and DMP1, which may influence mineralization or properties of the matrix. doi:10.1371/journal.pone.0038393.g014 depends on finely tuned PP i homeostasis. Intriguingly, unlike the ectopic calcification found in the joints, the PDL remained nonmineralized in the face of cementum expansion in Ank and Enpp1 2/2 mice, indicating the presence of additional factors that inhibit mineralization across the PDL space. These likely include multiple, redundant negative regulators of mineral growth, as well as factors that indirectly prevent mineralization by influencing cell differentiation [67].
These studies lend support to the idea that there are key differences influencing acellular versus cellular cementum development. Namely, acellular cementum is dependent on precise modulation of local PP i , whereas cellular cementum is much less sensitive to fluctuations in local PP i . The morphology, speed of formation, and ECM protein composition of acellular cementum were altered dramatically by disruption of local PP i homeostasis. Conversely, cellular cementum remained unaffected in all of these aspects. The primary cementoblasts of the AEFC adapted their expression of ANK, NPP1, and OPN in attempts to compensate for loss of control over cementum apposition, a response notably absent in cementoblasts of cellular cementum, as well as other cells of the dentoalveolar complex. However, when the regulatory influence of PP i was lessened, as in Ank and Enpp1 2/2 mice, the cementum of the cervical root grew rapidly, engulfed cells to become cementocytes, and switched on DMP1, approximating several aspects of apical cellular cementum. Thus, we propose that strict regulation by PP i is one of the major differences between acellular vs. cellular cementum types.
This finding provides insight into the origins of the two main types of cementum, and also informs clinical regenerative therapies. The profound influence of PP i metabolism on acellular cementum development immediately suggests the concept of PP i modulation for cementum regeneration. It is especially attractive to consider such a novel approach when growth and differentiation factors used to date have been limited in terms of true cementum regeneration, PDL integration, and/or predictability [68][69][70]. In a preliminary proof-of-principle study, we employed the Ank 2/2 mouse, featuring deficient extracellular PP i and increased cementogenesis, to analyze tissue repair and regeneration in a periodontal fenestration model [45]. Importantly, we found that Ank 2/2 mice featured significantly greater new cementum vs. controls, more organized mineral deposition on the root surface in the defect areas, and recapitulated expression patterns mapped during cementum development, including strong OPN and DMP1 in the cementum matrix, and elevated NPP1 in associated cementoblasts. Thus, in this pilot study in mice we found that reduced local levels of PP i promoted increased cementum regeneration. There has been concern voiced about regenerated cementum being the cellular type in a majority of studies [as summarized in [69]], and thus not optimal for PDL attachment. Our findings support that the cellular or acellular nature may be a reflection of the speed of formation, and that both can support sufficient extrinsic PDL fiber insertion.

Mouse strains
Preparation and genotyping of mouse models was previously described for Ank 2/2 [26,71], Alpl 2/2 (previously known as Akp2 2/2 ), and Enpp1 2/2 [10,14,16,54]. Ank and Alpl mice were maintained on a mixed background of 129S1/SvImJ and C57BL/ 6 strains, and Enpp1 mice were maintained on a mixed background of C57BL/66129/SvTerJ strains. Mice were housed in a specific pathogen free facility in 12 hr light-dark cycles with access to water ad libitum. Ank and Enpp1 2/2 mice were fed a standard rodent diet, while Alpl litters were provided a vitamin B6 enforced diet to reduce seizures and prolong lifespan (TestDiet, Richmond, IN). Heterozygote breeding pairs were employed to prepare homozygote 2/2 mice and age-matched +/+ controls at specific ages during tooth development. Heterozygotes were examined to determine any morphological tooth phenotype. Mice were sacrificed by cervical dislocation and mandible tissues harvested. At least three control ( +/+ ) and null ( 2/2 ) mandibles were examined for each age of interest.

Histology
Mouse mandibles were harvested and prepared for histology as previously described [26]. Briefly, mandibles were sagittally hemisected and fixed in Bouin's solution at 4uC overnight. Hemi-mandibles were demineralized (for tissues post 8 dpn) in acetic acid/formalin/sodium chloride (AFS) solution at 4uC. Tissues were paraffin embedded after standard histological processing. Five mm buccal-lingual (coronal) serial sections of the first mandibular molar or longitudinal (sagittal) sections of hemimandibles were prepared by rotary microtome and mounted on charged glass slides. Slides were deparaffinized in xylene for histological analyses, including hematoxylin and eosin (H&E) staining used for morphological characterization.

Growth measurement and histomorphometry
Cervical cementum on the lingual aspect of the mesial root of the mandibular first molar was measured at a fixed distance of 300 mm from the cementum-enamel junction (CEJ) in Ank and Enpp1 +/+ and 2/2 sections from 14-60 dpn, using two central sections from the set of serial sections. Static histomorphometry was used to measure cervical cementum, PDL, and alveolar ridge bone width on the lingual aspect of mesial roots of mandibular first molars at the age 26 dpn. Calibrated measurements were made using SPOT software (Diagnostic Instruments, Sterling Heights, MI). ANOVA followed by the Tukey test for direct comparisons was employed for statistical testing of histomorphometric measurements (PASW (SPSS) Statistics software, version 19).

Picrosirius red stain for collagen in histological sections
Tissues processed for histology were stained with a picrosirius red staining kit according to manufacturer directions (Polysciences, Inc., Warrington, PA). Deparaffinized slides were immersed in 0.2% phosphomolybdic acid hydrate, rinsed in water, incubated in direct red 80 for 60 min, then 0.01 N HCl solution for an additional 2 min. Samples were rinsed in 75% ethanol for 45 sec, then dehydrated in xylene, cleared, and mounted with coverslips. Digital images were captured with an OptiPhot-2 microscope (Nikon Instruments, Inc., Melville, NY) fitted with a light polarizer, using an EOS 5D Mark II digital camera (Canon U.S.A., Inc., Lake Success, NY).

Immunohistochemistry
Immunohistochemistry (IHC) was performed on histological sections as previously described [26]. Primary antibodies were used with biotinylated secondary antibodies (Vectastain Elite ABC, Vector Labs, Burlingame, CA) and color reactions were developed to a red product using a 3-amino-9-ethylcarbazole (AEC) substrate kit (Vector Labs). Positive controls included normal mouse tissues and negative controls were performed in the absence of primary antibody. Primary antibodies included: monoclonal rat anti-human ALPL/TNAP (R&D Systems, Minneapolis, MN); rabbit anti-mouse progressive ankylosis protein (ANK3) [12]; rabbit anti-mouse bone sialoprotein (BSP), (a gift from Dr. Renny Franceschi, University of Michigan); rabbit antirat dentin matrix protein-1 (DMP1) raised against an N-terminal (90-111) portion of DMP1 (Takara, Shiga, Japan); polyclonal goat anti-human NPP1 (Abcam, Cambridge, MA); and LF-175 rabbit anti-mouse osteopontin (OPN) (Dr. Larry Fisher, NIDCR) [72]. ALPL and ANK3 staining was performed with an additional unmasking step wherein slides were incubated overnight in 8.0 M guanidine HCl (pH 8.0) solution. IHC for each target protein was performed in sections from at least three (n = 3) animals for each age, with representative staining chosen for photographs shown in Results.

In situ hybridization
In situ hybridization was performed using a non-radioactive in situ hybridization (ISH) protocol employing a digoxigenin (DIG)labeled cRNA probe for genes of interest, as described previously [26]. Linearized probes were cleaned by phenol-chloroform precipitation. Riboprobe synthesis was performed using a digoxigenin-UTP-labeled kit (Roche Applied Science, Indianapolis, IN). Probes were fractionated at 60uC and riboprobe concentrations were checked by dot blot on Hybond N + nylon membrane (GE Healthcare, Piscataway, NJ). Messenger RNAs were labeled by incubation of deparaffinized sections with NBT/BCIP (Nitro blue tetrazolium chloride/5-Bromo-4-chloro-3-indolyl phosphate, toluidine salt). Probes used for ISH included: mouse Dmp1 plasmid (provided by Dr. Ann George, Northwestern University) [73]; mouse Opn and Bsp probes (provided by Dr. Marian Young, NIH/ NIDCR) [74]. Negative controls included sense probes.

Electron microscopy
Scanning electron microscopy (SEM) analyses were performed on hemi-mandibles from 20 dpn control and Alpl 2/2 mice as previously described [26]. Briefly, mandibles were sequentially dehydrated in aqueous ethanol solutions and mounted in roomtemperature-cure epoxy (Allied High Tech Inc, Rancho Dominguez, CA). Specimens were cut using a precision wafering saw (Buehler Ltd, Lake Bluff, IL) to expose the mesial surface of the first molar. The cut surface was then ground further distally to expose the interior of the first molar using 600 then 1500 grit SiC papers, followed by smoothening via ultramicrotoming with a 45u angle diamond knife (Diatome, Inc., Hatfield, PA) fitted onto a MT 6000-XL ultra-microtome (Bal-Tec RMC, Inc., Tucson, AZ). Specimens were mounted on SEM stubs, sputter coated with 5 nm of Pt for electron conductivity (SPI Supplies Inc, West Chester, PA), and imaged by an JSM7000F (JEOL-USA, Inc., Peabody, MA) SEM operating at 15 kV in backscattering mode.

Cell culture and in vitro assays
Isolation and characterization of OCCM.30 murine cementoblasts has been previously described [75,76]. Cells were grown in Dulbecco's Modified Eagle Medium (DMEM) with 10% v/v fetal bovine serum (FBS), 2 mM L-glutamine, 100 U/ml penicillin, and 100 mg/ml streptomycin (all reagents from Invitrogen, Carlsbad, CA). For gene expression and mineralization experiments, OCCM.30 cells were plated in standard media as described above, with media changed after 24 hrs to DMEM with 1% FBS with 50 mg/ml ascorbic acid (AA). Media were changed every 48 hrs for the remainder of the experiment. Inclusion of organic phosphate source b-glycerophosphate (bGP; 5 mM) was used to create mineralizing conditions. Inorganic PP i (10 or 100 mM) was added to assay effects on cell function. Both bGP and PP i were purchased from Sigma-Aldrich (St. Louis, MO). Cell culture experiments were performed at least three times in triplicate with representative results presented.
Cell proliferation was measured using a non-radioactive, MTSbased assay, following manufacturer's directions (CellTiter 96H AQ ueous proliferation assay, Promega, Madison, WI). Absorbance was measured at 570 nm, with reference reading at 750 nm. Absorbance is proportional to the number of living cells in culture.
Production of collagen by cells in vitro was quantified by picrosirius red staining, using methods modified from previous reports [77,78]. Briefly, cells were rinsed with PBS and fixed in Bouin's solution for 1 hr at room temperature. The fixative was removed and the plate rinsed several times in water to remove excess Bouin's solution. The collagenous matrix in plates was stained by incubation with picrosirius red dye (Direct Red 80, Polysciences, Inc., Warrington, PA) while gently shaking. Unbound dye was removed by rinsing several times with 0.01 N HCl. Bound dye was removed by incubation and shaking with 0.1 N NaOH for at least 1 hr. Picrosirius red was quantified by reading the absorbance at 550 nm. Quantity of collagen was calibrated against a standard curve created by plating and eluting known concentrations of rat tail collagen.
Von Kossa staining for mineral nodule formation was performed using standard procedures [79]. Silver stain was visualized as black, and stain intensity indicated the amount of calcium phosphate precipitation in the cell matrix (silver ions react with phosphate). Cell mineralization in vitro was quantitatively assayed by measuring calcium deposits, using a method modified from a previous report [77]. To cell culture wells, 500 ml 0.5 N HCl was added and plates were agitated for 60 min to dissolve calcium-phosphate precipitations. Eluted calcium was measured using a calcium assay (Genzyme Diagnostics, Farmingham, MA). One ml of sample was added to 99 ml Arsenazo reagent and absorbance was read at 650 nm. Standard curves were prepared using a calcium stock solution.
A modified assay for measuring in vitro alkaline phosphatase activity (ALP) was used [80]. Briefly, cell cultures were rinsed with PBS and incubated with 200 ml p-Nitrophenyl phosphate (PNPP, Sigma) in the dark at ambient room temperature for 30 min. After incubation, 10 ml supernatant for each condition was transferred to a 96 well plate containing 90 ml of 3 N NaOH (stop solution) per well. Absorbance was recorded at a wavelength of 405 nm.
An enzymatic assay was employed to measure the activity of pyrophosphate-generating ectoenzymes (nucleoside triphosphate pyrophosphohydrolase, NTPPPHase activity), based on a previously described procedure [81]. Cells were rinsed with PBS and incubated for 2 hrs with 2.0 ml of 1 mM thymidine 59 monophosphate p-nitrophenyl ester sodium (TMPNP) solution at 37uC and without CO 2 . After incubation, 20 ml supernatant for each condition was transferred to a 96 well plate containing 80 ml of 0.1 N NaOH (stop solution) per well. Absorbance was recorded at a wavelength of 410 nm.

RNA isolation and real-time quantitative RT-PCR
Isolation of RNA, synthesis of cDNA, and performance of realtime quantitative PCR was undertaken as previously described [26]. Total RNA from cells was isolated using the RNeasy Micro kit (Qiagen, Valencia, CA) and cDNA was synthesized from