Root resorption and ensuing cementum repair by Wnt/β-catenin dependent mechanism

Abstract

Introduction:

Physiological root resorption is a common occurrence in mammalian teeth, which suggests that there must be a corollary consisting of physiological cementum repair. The mechanism(s) responsible for this physiological repair process is unknown and was the focus of this study.

Methods:

Using a rat model, we explored first the prevalence of physiological root resorption and then asked whether this prevalence changed as a result of an osteoporotic phenotype. The cellular mechanisms of resorption were characterized using a combination of finite element modeling coupled with in-vivo histologic, molecular, and cellular analyses in rats. A potential molecular mechanism for cementum repair was uncovered using a strain of transgenic mice in which Wnt-responsive cells could be labeled and followed over time.

Results:

In rats, most resorption lacunae were concentrated on the distal surfaces of the roots. Rat molars undergo a physiological tooth drift distally, and using finite element modeling, we calculated the magnitude of the compressive strains that accumulated on these surfaces in response to mastication. Although the overall strain magnitudes were low, they were constant and coincided with the presence of resorption lacunae. Where resorption lacunae were present, progeny from a Wnt-responsive population of stem cells, embedded in the periodontal ligament, directly contributed to the repair of the lacunae.

Conclusions:

Despite the fact that both are clastic conditions, an osteoporotic phenotype in rats was not associated with an increase in the prevalence of physiological root resorption. The location of the resorption lacunae corresponded to sites of low but constant compressive strains produced by physiological distal drift. At least 1 mechanism responsible for physiological cementum repair involved the contribution of Wnt-responsive stem or progenitor cells originating in the periodontal ligament. These data point toward a potential Wnt-based strategy to regenerate cementum in subjects with disease or damage.

Most dentists view root resorption as a pathologic process, the untoward and alarming consequences of excessive and continuous orthodontic forces¹ or trauma.² Rare conditions including idiopathic cervical root resorption and Paget disease are also characterized by an uncontrolled resorptive activity on the roots of teeth.^3,4 Given these kinds of clinical observations, one might legitimately assume that root resorption is solely a pathologic process. This assumption, however, is incorrect. For example, the roots of the primary dentition resorb as a part of normal tooth replacement.⁵ Greater than 90% of human teeth exhibit evidence of root resorption and repair.⁶ This finding demonstrates that cementum remodeling is a continual process. What then is the etiologic basis for this process?

In their 1951 histologic examination of cadaveric specimens, Henry and Weinmann⁶ reported that most resorption lacunae were found on the mesial surfaces of the teeth. Human teeth undergo physiological drifting mesially,^7–9 which has led Henry and Weinmann⁶ to speculate that the 2 events are related. This speculation was based on an established correlation between the duration and magnitude of tooth movement and the severity of root resorption.^10,11 The current dogma states that compressive strains, produced by tooth movement, trigger osteocyte apoptosis and death, followed by osteoclast-mediated resorption of the damaged bone.^12,13 The same relationship between strains and cell death may also take place within the cementum.

We sought to better understand the relationship between force and root resorption and opted to make use of a rodent model. Rat teeth undergo a physiological drift of their molars, only this movement is distally.^14–19 To calculate the magnitude of the compressive strains produced by this physiological drift, we had to first amass a detailed understanding of the density of the alveolar bone²⁰; the rate of the physiological drift¹⁷; the magnitude of masticatory forces²¹; and the direction of those forces based on molar cuspal geometry.^22,23

We also considered the possible relationship between root resorption and the disease state of osteoporosis. There is scant clinical data addressing whether root resorption is more prevalent in osteoporotic patients compared with patients with normal bone turnover.²⁴ There is, however, an abundance of clinical and preclinical data demonstrating that both osteoporosis and root resorption are catabolic activities characterized by the resorption of mineralized matrix.²⁵ We, along with others,^26,27 hypothesized that root resorption may be increased in individuals with an osteoporotic skeleton because the underlying mechanisms of resorption appear to be shared.

Finally, we considered that if root resorption is so common, then there must be a physiological mechanism of repair. For example, in their analysis of cadaveric specimens, 85% of the resorption sites on human teeth showed evidence of repair.⁶ Because the repair occurred on the external root surface, the repair mechanism likely involved the activation of cementoblasts. Many molecules have been identified as playing a role in cementum formation or resorption including, among others, fibronectins, insulin-like growth factor 1, fibroblast growth factors, epidermal growth factor, bone morphogenetic proteins, interleukins, and transforming growth factor–β.^28–31 We were interested in any molecular mechanism involved in cementogenesis, but here we focused our efforts on a detailed investigation of a single pathway. Recent reports implicated the Wnt pathway as having a critical role in cementogenesis.³² Also available were an established (Axin2^LacZ/+) and a novel (Axin2^CreERT2/+;R26R^mTmG/+) strain of Wnt reporter mice, which allowed us to identify and follow Wnt-responsive cells and their progeny in the periodontal ligament (PDL) and on the root surfaces. Finally, we had access to mice carrying a conditional loss-of-Wnt function (Ocn-Cre;Wls^fl/fl) mutation, which permitted us to determine the consequences of reduced Wnt signaling on the formation or repair of resorption lacunae of the tooth roots. We reasoned that if it were possible to reinitiate the formation of cementum on a tooth surface, then potentially one could devise new treatments for diseases where this tissue is lost.³³ Therefore, we set out to determine if rodents, similar to humans, showed evidence of widespread root resorption and repair.

MATERIAL AND METHODS

Stanford Committee on Animal Research approved all protocols (#13146), which conformed to the Animal Research: Reporting of In Vivo Experiments guidelines. The analyses were conducted on 37 ovariectomized (OVX) and 13 female Wistar rats (control group). They were aged 3-4 months, and their average weight was 150-200 g. In the case of osteoporotic rats, animals underwent ovariectomy (OVX group) at 7 weeks old. A dorsal midline incision was made between the midback and tail base. The peritoneal cavity was accessed through incisions in the bilateral muscle layer, the ovary was identified, the connection between the fallopian tube and the uterine horn was suture-ligated, and the ovary was removed. Eight weeks later, when their osteoporotic phenotypes had developed,³⁴ rats were analyzed for the incidence of root resorption.

Axin2^CreERT2/+;R26R^mTmG/+ mice (#018867 and #007576), Axin2^LacZ/+ mice (#11809809), and Ocn-Cre;Wls^fl/fl (#01288 and #019509) were purchased from Jackson Laboratories (Bar Harbor, ME) and housed in a temperature-controlled environment with 12-hour light and dark cycles. The littermates of OcnCre;Wls^fl/fl, including both OcnCre;Wls^fl/+ and Wls^fl/fl, were indistinguishable from wild-type mice,^35,36 therefore, they were used as controls in this study. In the Axin2^LacZ/+ strain, the Axin2 gene is replaced bya LacZ gene, and the expression of LacZ is under control of the endogenous Axin2 promoter and enhancer regions. Because Axin2 is a direct target of Wnt/β-catenin signaling,³⁷ cells in the Axin2^LacZ/+ strain that express LacZ (which is detectable by β-galactosidase staining³⁸) do so in response to a Wnt signal.³⁹ To induce Cre expression in Axin2^CreERT2/+;R26R^mTmG/+ mice, tamoxifen (4 mg/25 g body weight) was delivered intraperitoneally for 3 consecutive days; animals were killed 5, 30, and 90 days later.

Hemimaxillae were harvested and fixed in 4% paraformaldehyde, decalcified in ethylene diamine tetra-acetic acid, dehydrated using an ascending graded ethanol series, and embedded into paraffin blocks for sectioning. Tissue sections were generated at an 8-μm thickness. For Axin2^LacZ/+ mice, after complete demineralization, specimens were incubated in 30% sucrose in phosphate-buffered saline overnight and then embedded with Tissue-Tek (Sakura Finetek USA, Inc, Torrance, Calif) optimal cutting temperature embedding medium for cryosections.

To evaluate the frequency and distribution of root resorption, 1 in every 10 slides of the transverse maxillary sections were selected for histologic analysis. Slides were deparaffinized in Citrisolv (#1601; Decon Labs, Inc, King of Prussia, Pa) and hydrated by means of a descending graded ethanol series. After staining, sections were dehydrated in a graded series of ethanol and Citrisolv and subsequently cover-slipped with Permount (#SP15; Fisher Scientific, Pittsburgh, Pa) mounting media.

Aniline blue, a biological dye in fluorescence microscopy, which is used to stain collagen, was used to detect and locate the resorptive lacunae. Slides were treated with a saturated solution of picric acid, followed by a 5% phosphotungstic acid solution and staining in 1% aniline blue.

Pentachrome staining was performed as described.⁴⁰ In brief, after dehydration, slides were stained with 1% alcian blue (#A5268; Sigma-Aldrich, Saint Louis, Mo), Verhoeffs hematoxylin (#S71299; Fisher Scientific), sodium thiosulfate (#14518; Alfa Aesar, Ward Hill, Mass), crocein-scarlet-acid fuchsin solution (#22914; Chem-Impex International, Wood Dale, Ill; #F8129; Sigma-Aldrich), 5% phosphotungstic acid (#P4006; Sigma-Aldrich), and saffron (#3801; HARLECO, MilliporeSigma, Bedford, Mass), with washing steps between each stain using ethanol, acetic acid, and distilled water. In pentachrome-stained tissues, the nuclei stain blue to black, cytoplasm stains red, collagen stains yellow to greenish yellow, and fibrous tissue stains intense red.

Picrosirius red staining was used to detect collagen. Tissues were stained with picrosirius red solution (0.5 g sirius red [#35780, Pfaltz & Bauer, Inc, Waterbury, Conn] dissolved in 500 mL saturated picric acid solution) and then viewed under polarized light. Tightly aligned fibrillary collagen molecules appear red compared with fewer organized collagen fibrils that show colors of shorter (green-yellow) wavelengths.

Tissue sections were deparaffinized and blocked with 5% goat serum. Immunohistochemistry was performed⁴¹ with primary antibodies: rabbit anti–cathepsin K (ab19027; Abcam, Cambridge, UK), rabbit anti–matrix metalloproteinase 13 (MMP13) (ab39012; Abcam), and secondary antibodies including biotinylated goat anti-rabbit IgG antibody (BA-1000;Vector Laboratories, Burlingame, Calif). The staining was visualized by ABC Peroxidase Standard Staining Kit (32020; Thermo Fisher Scientific, Rockford, Ill). Tissue sections were photographed using a Leica digital imaging system (Leica Microsystems, Wetzlar, Germany).

Tartrate-resistant acid phosphatase (TRAP) and alkaline phosphatase (ALP) activity analyses were performed as previously described.⁴² To detect ALP activity, tissue sections were treated with ALP-detection solution containing 5-bromo-4-chloro-3-indolyl phosphate (#11383221001; Roche, Basel, Switzerland) and nitro blue tetrazolium chloride (#11383213001; Roche) at 37°C for 30 minutes according to the manufacturer’s instructions. TRAP activity was observed using a Leukocyte Acid Phosphatase Staining Kit (#386A-1KT; Sigma-Aldrich). Tissue sections were processed according to the manufacturer’s instructions.

To detect β-galactosidase activity, sections were stained overnight at 37°C in Xgal solution buffered with 5-mM potassium ferricyanide, 5-mM potassium ferrocyanide, 2-mM magnesium chloride. Sections were rinsed with phosphate-buffered saline, dehydrated in a graded ethanol series, and cleared in Citrisolv then mounted with Permount.

Vital dye labeling was performed as previously described.⁴³ Briefly, mice were intraperitoneally injected with 20 mg/kg of calcein (Sigma-Aldrich) and 30 mg/kg of alizarin red (Sigma-Aldrich) with a 12-day interval. Mice were killed 2 days after alizarin red injection. Samples were fixed in 4% paraformaldehyde overnight and dehydrated with 30% sucrose for 24 hours and then were processed for hard tissue embedding and sectioning using Kawamoto’s method.^44,45

We developed a novel grading system to target deficits in sensitivity and practicality present in the comparable grading system. To our knowledge, this is the first grading system of its kind to assess root resorption on the basis of histologic data. This grading system allows for rapid classification of craters, partly owing to the ease of visualization for each grade. Once all the roots exhibiting root resorption were identified, they were all individually graded by 2 individuals according to criteria indicated in Table I and Figure 1, A. A unanimous decision between the raters was achieved for each root.

Table I.

Description of the grading system to assess the extent of root resorption

Grades	Description
Grade 1	One resorption crater in the cementum. No dentin involvement.
Grade 2	Multiple resorption craters in the cementum. No dentin involvement.
Grade 3	One resorption crater extended to the dentin.
Grade 4	Multiple resorption craters extended to the dentin.
Grade 5	Multiple resorption craters extended to the dentin and involving at least 1/4 of the root perimeter.

Fig 1.

Root resorption is a common finding in rats and the distribution of resorption lacunae appears to be influenced by the physiological drift of the dentition. (A) Schematic presentation of the grading system. (B) Representative μCT analyses of bone density in the distal femur in a control and (C) in an OVX rat. (D) Statistical comparison of (BV/TV) ratios in the distal femur of control and OVX groups. (E) Prevalence of root resorption on sample basis. (F) Prevalence of root resorption on root basis. (G) Side distribution of the resorption lacunae. (H) Aniline blue staining showing resorption lacunae on distal surfaces of the roots. (I) μCT sagittal section showing the distal inclination of the rat’s molar teeth. (J) FE model presenting the first and third principal strains and displacements. BV/TV, bone volume/total volume; M, molar.

A μCT tomography system (VivaCT 40; Scanco, Bruüttisellen, Switzerland) at 10.5 μm voxel size (70 kV, 115 μA, 300-ms integration time) was used for scanning and reconstruction of the molar and its associated alveolar bone. Bone morphometry was performed using acquisition system’s analysis software (Scanco). Multiplanar reconstruction and volume rendering were carried out using Avizo (FEI, Hillsboro, Ore) and ImageJ (National Institutes of Health, Bethesda, Md) software. The Digital Imaging and Communications in Medicine files were imported into ScanIP (Synopsys Inc, Mountain View, Calif), a software program for comprehensively processing 3-dimensional image data. In this program, restricted threshold values were used to select the molar and distinguish it from its associated alveolar bone structure. Boolean operations were performed to eliminate unnecessary fine details of the bone, the tooth structure, and the socket. Recursive Gaussian and mean filtering were used to reduce surface complexity, which aided computations in the next software program. From the ScanIP program, separate .stl files were generated for the molar and for the alveolar socket. These .stl files were then imported into COMSOL 5.4 (COMSOL, Burlington, Mass) for subsequent finite element (FE) analysis.

In COMSOL, Boolean operations were used to fill the region between the molar and the wall of the alveolar socket with a solid material assigned to be the PDL. Because of the thresholding limitations in ScanIP, ~ one half of the PDL height was replicated. The molar and the PDL were modeled as linear elastic materials, with Young’s elastic moduli of 20 GPa and 20 MPa and Poisson’s ratios of 0.3 and 0.45, respectively.^46,47 The FE model was then run with geometric nonlinearity; 63,298 tetrahedral elements; and 301,014 degrees of freedom. In addition to the degree of angulation of the rat molars and their cusps,²² additional inputs into the model included a value of an apically directed masticatory force on the first molar of approximately 1 N. This value was estimated as a midpoint of the range of values suggested in studies of mastication in mice eating hard vs soft food diets.²¹ The 1 N force was distributed over the molar’s 3 cusps; the inclination of each cusp was measured to make approximately a 32.6° angle with the z-axis, which was approximately parallel to the occlusal plane.²³

Statistical analysis

A nonparametric, Pearson chi-square test and odds ratio calculation were used to determine any significant difference between the groups. A parametric 2-sample t test was used to compare bone density measurements of the groups. Significance was attained at P <0.05 (*), P <0.01 (**), and P <0.001 (***).

RESULTS

Osteoporosis is primarily a bone resorptive disease, characterized by excessive osteoclast activity and decreased osteoblast activity. Consequently, we hypothesized that root resorption would be increased in animal models of osteoporosis. Eight weeks after OVX, μCT analyses confirmed that the rats had developed low–bone density in the distal femur (P = 0.00038) (Fig 1, B–D).⁴²

The rodent dentition was then evaluated for evidence of root resorption. Skeletally mature, age-matched (aged ~ 4 months) female rats served as controls. The prevalence of root resorption in the OVX rats (78%) was equivalent to the control group (77%, P <0.05; see Table II and Fig 1, E). The calculated odds ratio (1.0875) also did not reveal any increased risk of root resorption in the osteoporotic rats. The severity of root resorption was then evaluated by a semiquantitative scale that was developed on the basis of the number, depth, and area of each resorption lacuna. The frequency of severe resorption lacunae was significantly higher in the control than the OVX group, but the overall prevalence was equivalent (Table III).

Table II.

A 2 × 2 contingency table for risk factor (osteoporosis)/outcome (root resorption) comparison and the results of Pearson chi-square test and odds ratio calculation

	Outcome present	Outcome absent	Total	P value	Odds ratio
Risk factor present	29	8	37
Risk factor absent	10	3	13	0.913231	1.0875
Total	39	11	50

Note. Significance level was predetermined as P <0.05.

Table III.

Severity score frequencies, percentages, and the results of the Pearson chi-square test

Group	Grade 1, mild	Grade 2, mild-moderate	Grade 3, moderate	Grade 4, moderate-severe	Grade 5, severe	P value
OVX	12 (8.57)	27 (19.29)	16 (11.43)	63 (45.00)	22 (15.71)
Control	5 (5.05)	4 (4.04)	15 (15.15)	47 (47.47)	28 (28.28)	0.0024
Total	17 (7.11)	31 (12.97)	31 (12.97)	110 (46.03)	50 (20.92)

Note. Values are given as severity score frequency (%).

We considered the distribution of the resorption lacunae as a function of the involved root (Fig 1, F). In both the controls and the OVX animals, a minority of the roots (between 33% and 23% of the roots, respectively) showed evidence of resorption lacunae (Fig 1, F). We then focused on the distribution of resorption lacunae as a function of the root surface. Here, we discovered that 95% of control and 86% of OVX samples had resorption craters on the distal surfaces of the roots (quantification in Fig 1, G and examples in Fig 1, H). This striking surface preference prompted us to consider whether a particular mechanical environment promoted root resorption.

An FE model was developed to determine the displacement, stresses, and strains ofa distally angulated molar (Fig 1, I) under apically directed masticatory force. The masticatory force on the first molar was applied as a net inward-normal component of 0.842 N plus a tangential component of 0.538 N on a plane cut parallel to the mesial cusp at approximately 32° to the z-axis (Fig 1, J). The outer surfaces of the PDL were constrained to represent the surrounding alveolar bone. Loading resulted in a z-displacement of the molar equal to ~ 6 μm. The compressive strains were, on average, 6%-7% on the distal root surfaces compared with 4% on the mesial root surfaces (Fig 1, J).

On any given root surface, resorption (arrow) and repair (asterisks) occurred simultaneously (Fig 2, A and A’). To identify the cell type(s) responsible for resorption, we first mapped the distribution of TRAP activity. Only infrequently did we find evidence of TRAP activity on dentin surfaces (Fig 2, B and B’). At the same time, however, TRAP activity was uniformly detected on alveolar bone surfaces (Fig 2, C and C’).

Fig 2.

Cementum and dentin were vulnerable to degradation and resorption. (A) Pentachrome staining identifies 2 repaired (asterisks) and 1 active resorption (arrow) lacunae at the distal surface of a molar root. (A’) High-magnification image of the resorption lacuna. (B) TRAP-positive clastic cells on the root surface. (B’) High-magnification image of B. (C) TRAP-positive osteoclasts on the alveolar bone surface. (C’) High-magnification image of C. (D) Cathepsin K–positive cells on the alveolar bone and root surfaces. (D’) High-magnification image of D. (E) MMP13-positive cells on the alveolar bone and root surfaces. (E’) High-magnification image of E. (F) Sost-expressing cells in the resorption lacunae. ab, alveolar bone; d, dentin. Scale bars = 20 μm.

We performed an additional immunostaining for cathepsin K, a cysteine protease secreted by clastic cells that is essential for the degradation of matrix collagen and the activation of TRAP.⁴⁸ Cathepsin K expression was occasionally detected at or near resorption lacunae but similar to TRAP activity was predominantly expressed by cells located on the alveolar bone surfaces (Fig 2, D and D’). We considered other molecular markers of clastic cells and turned to MMP13, a matrix metalloproteinase, which is involved in matrix remodel- ing⁴⁹ and known to stimulate osteoclast activity.⁵⁰ However, we failed to detect any immunostaining for the protein on the root surfaces (Fig 2, E and E’). Robust MMP13 immunostaining, however, was evident on the alveolar bone surfaces (Fig 2, E and E’). Finally, we decided to perform immunostaining for sclerostin (Sost), a glycoprotein that inhibits bone formation by inhibiting Wnt/β-catenin signaling and activation of osteoblasts.⁵¹ Sost expression precisely correlated with the sites of root resorption (Fig 2, F).

We considered the repair process in cementum. We found evidence of resorption lacunae, occupied by PDL cells and bounded by a hypomineralized layer of cementum (Fig 3, A; arrow). Within the resorption lacunae the presumptive PDL cells showed no evidence of Sharpey’s fiber attachments (Fig 3, B). Some resorption lacunae were covered by a thin layer of acellular cementum (Fig 3, C; arrow), and in these subjects, picrosirius red staining clearly demonstrated a reattachment of Sharpey’s fibers to the acellular cementum (Fig 3, D). Thus, provided there was a thin layer of cementum lining the crater, some resorption lacunae appeared to reestablish a connection between the PDL and the tooth.

Fig 3.

Resorption lacunae can undergo functional and anatomic regeneration. (A) Pentachrome staining identifies an active resorption lacuna bounded by hypomineralized layer of cementum (arrow). (B) Picrosirius red staining of the same resorption lacuna showing no evidence of Sharpey’s fiber attachment. (C) Pentachrome staining showing resorption lacuna lined by a thin layer of acellular cementum (arrow). (D) Picrosirius red staining of the same lacuna showing evidence of Sharpey’s fiber reattachment. (E) Pentachrome staining identifies fully repaired resorption lacunae (asterisks) with cellular cementum. (F) ALP staining showing 2 fully repaired resorption lacunae (asterisks). (G) Picrosirius red staining showing the reattachment of the PDL fibers to the repaired surfaces of the resorption lacunae (asterisks). (H) Aniline blue staining showing a fully repaired resorption lacuna and reattachment of Sharpey’s fibers (asterisk). (I) Double dye vital staining showing asymmetrical new cementum formation on the mesial surface of the root (asterisk). cc, cellular cementum; d, dentin; ac, acellular cementum; p, pulp; ab, alveolar bone; D, distal; c, cementum; M, mesial. Scale bars = 20 μm.

Some lacunae were filled with cellular cementum (Fig 3, E; asterisks). Resorption lacunae that were filled with cementum were able to support the reattachment of the PDL, as shown by PDL fibers that stained positively for ALP activity (Fig 3, F; asterisks) and picrosirius red (Fig 3, G; asterisks). Only slight discontinuities in the cementum distinguished these repaired lacunae from neighboring intact cementum (Fig 3, H; asterisk).

To identify regions of the cementum undergoing repair, we used vital dye labeling. Calcein green dye was injected, followed by alizarin red dye 12 days later. Analyses of the tooth surfaces by fluorescence microscopy revealed asymmetrical new cementum formation, which was especially pronounced on the mesial surface of the roots (Fig 3, I; asterisk).

Thus far, our data demonstrated that resorption lacunae occur frequently and in rodents whose teeth drift distally, most of the lacunae are found on the root surface that is exposed to low but continuous compressive strains. Histologic and cellular analyses showed that some lacunae were repaired with cementum, and in some subjects, this cementum had PDL fibers embedded in it, demonstrating that a reparative process was underway. To identify the cells responsible for this repair, we first focused on their distribution along the root surface. For example, most of the resorption lacunae were observed in acellular cementum; therefore, the repair cells had to originate from the PDL.

We first established that mice, similar to rats, exhibited resorption lacunae on distal root surfaces and that their general appearance was similar to those in rats (Fig 4, A and B). In pathologic conditions, these resorption lacunae can occupy up to two thirds of the root surfaces.^35,52,53 Second, we used Axin2^LacZ/+ mice^37,39 to identify β-galactosidase–expressing, Wnt-responsive cells on the root surfaces (Fig 5, A). Axin2 is a universal target of Wnt/β-catenin signaling.³⁷ In Axin2^LacZ/+ mice, the expression of LacZ is under the control of the endogenous promotor or enhancer regions of Axin2. When the cells receive the Wnt signaling, the cells respond to it by producing LacZ. The LacZ gene encodes β-galactosidase, which can be visualized by Xgal staining.³⁸ We focused on resorption lacunae that fulfilled 2 criteria: they had to penetrate the dentin (eg, grade 3-5) and they had to be occupied by cells (Fig 5, B). In most of these subjects (n = 13), we found Xgal^+ve cells surrounding and inside of the resorption lacunae (Fig 5, C). Thus, Wnt-responsive cells were present at the right time and in the right place (eg, the PDL) to be involved in cementum repair.

Fig 4.

The distribution and general appearance of resorption lacunae are similar between mice and rats. (A) Aniline blue staining showing the resorption lacuna on the root surface of a mouse molar. (B) Pentachrome staining showing the resorption lacuna. cc, cellular cementum; d, dentin. Scale bars = 50 μm.

Fig 5.

Wnt-responsive cells and their progeny in the PDL are responsible for cementum regeneration. (A) Schematic of Axin2^LacZ/+ mice, in which Wnt-responsive cells were visualized by Xgal staining (blue). (B) Pentachrome staining showing cementum resorption in the maxillary molar root of adult mice. (C) Xgal^+ve Wnt-responsive cells around the root of molar. (D) Schematic of Axin2^CreERT2/+; R26R^mTmG/+ mice, in which Wnt-responsive cells transition from expressing membrane Tomato (red) to membrane GFP (green) in the presence of tamoxifen. (E) GFP^+ve Wnt-responsive cells populate the PDL and the root surfaces, visualized 5 days after tamoxifen injection. On posttamoxifen injection (F) day 30 and (G) day 90, progeny of the initial Wnt-responsive population was analyzed. (H) Pentachrome staining showing the periodontium of control mice in which Wnt signaling is undisturbed. (I) Picrosirius red staining showing the PDL fiber alignment in the PDL space of control mice. (J) Pentachrome staining showing the cementum resorption in OcnCre;Wls^fl/fl mice, where Wnt signaling is abrogated. (K) Picrosirius red staining showing the discontinued PDL fiber in the cementum resorption site in mice, where Wnt signaling is disturbed. d, dentin; ac, acellular cementum; GFP, green fluorescent protein; TAM, tamoxifen; M, month; D, day; ab, alveolar bone. Scale bars = 50 μm.

We then used a tamoxifen-inducible Wnt lineage tracer strain of mice to identify and follow Wnt-responsive progeny.⁵⁴ Tamoxifen was given to Axin2^CreERT2/+;R26R^mTmG/+ mice at age of 1 month The initial Wnt-responsive population expresses green fluorescent protein, and as a consequence, the ultimate fate of the labeled cells was verified over time (Fig 5, D). Five days after tamoxifen exposure, a few Wnt-responsive PDL cells lined the acellular cementum (Fig 5, E). Thirty days later, the initial Wnt-responsive population had expanded considerably, and most of the green fluorescent protein–expressing cells lined the acellular cementum (Fig 5, F). In subjects where a resorption lacuna was evident, the progeny of the initial Wnt-responsive population occupied the crater (Fig 5, F; arrow). Our third analysis was conducted 3 months after the start of the experiment: the progeny of the initial Wnt-responsive population persisted in the PDL closely associated with the acellular cementum surface (Fig 5, G). Thus, Wnt-responsive cells populated the PDL space, and when resorption lacunae were evident, the repair cells arose from a Wnt-responsive population.⁵⁴

Is Wnt signaling essential for cementum repair? Analyses of a Wntless mutant strain of transgenic mouse demonstrate that the answer is yes. In OcnCre;Wls^fl/fl mice, the transporter protein Wntless has been deleted in osteocalcin expressing cells,³⁶ and the result is an abrogation in Wnt signaling in the periodontium.⁵⁵ In comparison with control mice (Fig 5, H and I), Ocn-Cre;Wls^fl/fl mice have resorption lacunae covering their entire root surfaces (Fig 5, J) to which PDL fibers cannot attach (Fig 5, K). Thus, we conclude that cementum regeneration is instigated by a population of Wnt-responsive PDL cells and that this population is essential for new cementum deposition that repairs the lesions.

DISCUSSION

Rather than solely being a pathologic activity, our data suggest that root resorption and repair are a physiological process. Root resorption is observed in humans,⁶ cats,⁵⁶ rabbits,⁵⁷ and rats,¹⁴ suggesting that it is part of constant remodeling process.⁵⁸ Data shown here support a model in which the instigation of root resorption is related to the physiological drift of the molars (Fig 1). Assuming that the initial biological environment on the distal and mesial root surfaces is similar, then a reasonable explanation for the distal surface preference for resorption lacunae in rodents is the mechanical environment created by physiological distal drift. We showed that the angulation of the molars, coupled with masticatory forces, collectively contribute to distal physiological drift, which in turn creates a compressive strain environment on the distal root surfaces (Fig 1). The calculated magnitude of the principal tensile and compressive strains are of the same order of magnitude (eg, on the order of a few percent) as the stresses produced in the PDL during orthodontic treatment (Fig 1).⁵⁹ Therefore, although the magnitude of distal compression strain was low, it was nonetheless concentrated on the distal root surfaces, consistent with the primary location of the resorption lacunae.

If a compressive strain environment is responsible for root resorption then how does this mechanical environment shift to become permissive for cementum repair? We speculate that it is related to the cyclic nature of tooth movement. Teeth undergo intrusion in response to masticatory loading during the day, which is reversed during the nighttime by eruption.⁶⁰ A similar mechanism may explain why a compressive environment may be relieved when masticatory forces are largely suspended during the nighttime.

Some investigators have speculated that cementum is more resistant to resorption than is alveolar bone.^61,62 Alveolar bone certainly has the capacity for remodeling,¹⁷ but in cementum, we speculate that there may be a disconnect between the resorptive and reparative phases. Indirect support for this hypothesis comes from the observation that despite the rampant presence of resorption lacunae in wild-type and mutant strains of mice, TRAP, cathepsin K, and MMP13 positive cells were rarely found on the root surfaces (Fig 2).⁵² We did note, however, that Sost expression precisely correlated with sites of root resorption (Fig 2, F). The failure to detect other clastic proteins on the cementum was not because of technical problems (we always found expression on the alveolar bone surfaces; Fig 2) but rather because of their scarcity on the root surfaces. The presence of Sost-expressing cells in the resorption lacunae, however, strongly suggests that these are sites where endogenous Wnt signaling is inhibited.

We found evidence of resorption craters that were fully or partially repaired with new cementum (Fig 3). The first histologic evidence of repair appeared to be the presence of a thin, uncalcified cementoid matrix lining the lacunae (Fig 3). This acellular tissue in rats contrasts with the cellular intrinsic fiber cementum reported in humans.²⁸ After this layer of acellular cementum is deposited, it appears that reattachment of the PDL apparatus can be established (Fig 3). Other investigators have posited that cementum repair is executed by differentiated PDL cells.³⁰ Our analyses support this interpretation and add to it an underlying molecular mechanism whereby an endogenous Wnt signal is involved in the recruitment of these PDL cells to the repair process (Fig 5). If this finding holds true, then one might consider a means by which Wnt signaling could be locally amplified in the PDL to stimulate cementum regeneration after periodontal breakdown.

In this study, we failed to find a correlation between the prevalence of root resorption and either the age of the animal or the osteoporotic status of the animal. This was a surprise to us because we had initially hypothesized that the OVX model, in which osteoclastogenesis is amplified,^63,64 would also be associated with the clastic activity of root resorption. Our analyses, however, failed to demonstrate such a correlation. The literature has little to offer regarding this important point. For example, in a rat model of experimental tooth movement, young and aged rats had the same incidence of root resorption.⁶⁵ In patients, age was not identified as a risk factor for orthodontically induced root resorption.⁶⁶ When we understand the pathoetiology of root resorption, then we may gain critical insights into why there does not seem to be an increased likelihood of developing resorption lacunae in an osteoporotic animal.

Our study provides new insights into the cellular and molecular mechanisms of physiological root resorption and repair, but similar to all studies, it also had limitations. The study limitations originated from its retrospective, cross-sectional, and observational design. Future prospective studies will help to validate the preliminary data shown here. For example, in ongoing work, we are genetically testing the direct contribution of Wnt-responsive cells to cementum repair.

CONCLUSIONS

1. An osteoporotic phenotype in rats was not associated with an increase in the prevalence of physiological root resorption.

2. The location of the resorption lacunae corresponded to sites of low but constant compressive strains produced by physiological distal drift.

3. Wnt/β-catenin dependent mechanism, involving the contribution of Wnt-responsive stem or progenitor cells originating in the periodontal ligament, was found to be responsible for physiological cementum repair.