Zoledronate treatment duration is linked to bisphosphonate‐related osteonecrosis of the jaw prevalence in rice rats with generalized periodontitis

Abstract Objectives To determine the extent that zoledronate (ZOL) dose and duration is associated with bisphosphonate‐related osteonecrosis of the jaw (BRONJ) prevalence in rice rats with generalized periodontitis (PD), characterize structural and tissue‐level features of BRONJ‐like lesions in this model, and examine the specific anti‐resorptive role of ZOL in BRONJ. Materials and Methods Rice rats (n = 228) consumed high sucrose‐casein diet to enhance generalized PD. Groups of rats received 0, 8, 20, 50 or 125 µg/kg IV ZOL/4 weeks encompassing osteoporosis and oncology ZOL doses. Rats from each dose group (n = 9–16) were necropsied after 12, 18, 24 and 30 weeks of treatment. BRONJ‐like lesion prevalence and tissue‐level features were assessed grossly, histopathologically and by MicroCT. ZOL bone turnover effects were assessed by femoral peripheral quantitative computed tomography, serum bone turnover marker ELISAs and osteoclast immunolabelling. Results Prevalence of BRONJ‐like lesions was significantly associated with (a) ZOL treatment duration, but plateaued at the lowest oncologic dose, and (b) there was a similar dose‐related plateau in the systemic anti‐resorptive effect of ZOL. ZOL and BRONJ‐like lesions also altered the structural and tissue‐level features of the jaw. Conclusion The relationship between BRONJ‐like lesion prevalence and ZOL dose and duration varies depending on the co‐ or pre‐existing oral risk factor. At clinically relevant doses of ZOL, BRONJ‐like lesions are associated with anti‐resorptive activity.

Discrete oral procedures involving the dentoalveolar bone (e.g., tooth extraction) are considered the most common local oral factor associated with MRONJ and are estimated to occur in 52%-86% of MRONJ cases (Aghaloo et al., 2015;Aljohani et al., 2017;Carlson & Schlott, 2014;Eleutherakis-Papaiakovou & Bamias, 2017;Filleul et al., 2010;Hamadeh et al., 2015;Khan et al., 2017;Marx, 2003;Otto et al., 2012;Ruggiero et al., 2014;Voss et al., 2017;Yoneda et al., 2017). Delaying the start of risk factor medications until surgical sites have healed, and ensuring appropriate perioperative management, including antibiotic coverage, are often effective means to prevent MRONJ associated with dentoalveolar surgery (Dimopoulos et al., 2009;Hoefert & Eufinger, 2011;Montefusco et al., 2008;Ripamonti et al., 2009;Vandone et al., 2012). However, periodontitis (PD) and periapical infection Li et al., 2016;Song et al., 2016) have also been identified as important risk factors for BRONJ/MRONJ in patients without recent dentoalveolar surgery (Carlson & Schlott, 2014;Eleutherakis-Papaiakovou & Bamias, 2017;Khan et al., 2017;Marx, 2003;Voss et al., 2017). While oral diseases involving the periodontal tissues appear to contribute to MRONJ directly, they may also contribute to cases related to dentoalveolar surgery, since teeth that require extraction often have some degree of acute or chronic infection (e.g., infection from caries, periodontal disease). Cases in which MRONJ arises without a recent dentoalveolar surgery may be more difficult to anticipate and treat, especially if these conditions are not fully resolved before beginning pAR treatment, or if they arise after pAR treatment has started. Therefore, studies that investigate the role of specific oral inflammatory diseases as local oral risk factors for MRONJ are important to develop a comprehensive understanding of the pathophysiology of the disease, and contribute evidence that might be used to develop preventative or therapeutic strategies.
Rice rats (Oryzomys palustris) develop two distinctive types of PD that require no surgical or local mechanical intervention: (a) a generalized form, which affects both maxillae and mandibles, and is accelerated by a high sucrose-casein diet (HSC) (Aguirre, Akhter, Kimmel, Pingel, Xia et al., 2012;Aguirre et al., 2016;Gotcher & Jee, 1981;; and (b) a localized form (food impaction localized PD, FILP), which overwhelmingly affects the maxillary interdental space between molar (M) 2 and M3 and is found in rice rats fed standard rodent chow (STD) (Messer et al., 2017). Importantly, rice rats fed HSC diet do not develop FILP lesions. Therefore, two different PD models can be generated in rice rats by dietary modification alone.
The pelleted HSC diet that accelerates generalized PD is based on the powdered Harvard high sucrose 700 and the ration 100 diets used in previous studies to accelerate the onset of PD in rice rats (Auskaps, Gupta, & Shaw, 1957;Gotcher & Jee, 1981;Ryder, 1980). The reason the pelleted HSC diet does not produce FILP Aguirre, Akhter, Kimmel, Pingel, Xia et al., 2012) is possibly due to the low content (3%) of insoluble fibre provided as powdered cellulose, which is in contrast to STD rodent chow (irradiated diet no. 7912, Teklad LM-485 Rodent Diet; Envigo, Tampa, FL), which contains 13.7% neutral detergent fibre (NDF) and 4.6% of crude fibre. These fibrous components of the STD chow diet appear to contribute to the initiation and persistence of FILP lesions.
The defining feature of the FILP model is direct injury to soft tissue in the maxillary interdental M2-M3 space by persistent presence of impacted plant fibre and cellular debris. Food impaction around teeth, dental implants, or embrasures in humans may also result in peri-implantitis or peri-coronitis (Du, Gao, Qi, Liu, & Lin, 2010), and is an important risk factor for localized PD in humans (Matthews & Tabesh, 2004;Nunn, 2003). Conversely, clinical and histopathologic examination of generalized PD in the rice rat shows marked accumulation of microbial plaque on molar surfaces and inflammation of periodontal tissues. This type appears to resemble human PD, in which biofilms are an important mediator and contributing factor to the pathophysiology of the disease (Genco & Borgnakke, 2013;Offenbacher, 1996;Page, Offenbacher, Schroeder, Seymour, & Kornman, 1997;Salvi, Lawrence, Offenbacher, & Beck, 1997).
Rice rats with either type of PD, and without dentoalveolar surgery, develop BRONJ-like lesions when simultaneously exposed to clinically relevant doses of the systemic pAR, zoledronate (ZOL) Messer et al., 2018). Importantly, the anatomical distribution of BRONJ-like lesions appears to be associated with the anatomical location of the PD in individual rats. Specifically, rice rats fed STD diet and injected with ZOL developed BRONJ-like lesions primarily at the same location as the FILP lesions, the M2M3 maxillary interdental space (Messer et al., 2018), whereas rats fed HSC diet and injected with oncologic ZOL developed BRONJ-like lesions equally in the four jaw quadrants .
These findings suggest that rats given clinically relevant doses of ZOL are more likely to develop BRONJ-like lesions in quadrants that are also directly affected by oral inflammation.
In humans, various local oral risk factors (e.g., oral infections, surgical/trauma events and inflammatory disease), when combined with systemic risk factor medications, can result in MRONJ. However, it is not clear if different local oral risk factors result in distinctive patterns in the relationship between MRONJ prevalence and dose and duration of systemic risk factor medications. The rice rat provides a unique opportunity to explore this hypothesis given the distinctive types of PD that exist in this species. Previous findings demonstrated that BRONJ-like lesions occur in rice rats with localized PD (FILP) in a positive dose-dependent manner, but ZOL duration did not have a significant effect on the prevalence of BRONJ-like lesions, which plateaued between 18 and 30 weeks of treatment (Messer et al., 2018). Importantly, 50%-75% of vehicle-treated STD diet rats show FILP-induced mucosal injury by age 16 weeks. This finding is in direct contrast to the age-related progression of generalized PD which does not become moderate to severe in a majority of rats until later, at age 22 weeks (Aguirre, Akhter, Kimmel, Pingel, Xia et al., 2012).
These findings suggest that the effect of ZOL dose and duration on BRONJ-like lesion prevalence in rice rats with generalized PD could differ from that in rice rats with localized PD. Examining how different oral risk factors work alongside systemic medications may provide important data that improves the current understanding of how different periodontal risk factors interact with systemic medications to initiate MRONJ.
The primary purposes of this study were to: (a) define the relationship between ZOL dose and duration and prevalence of BRONJlike lesions in rice rats with generalized PD, (b) characterize the overall structure and tissue-level features of alveolar bone in ZOLtreated rice rats with and without BRONJ-like lesions by MicroCT and histologic techniques and (c) examine the association between the anti-resorptive effect of ZOL at clinically relevant doses and the presence of BRONJ-like lesions.
We hypothesize that: (a) increased dose and duration of ZOL will increase the prevalence of BRONJ-like lesions in rice rats with generalized PD in a pattern that is distinctive from previous findings in rats with localized PD; (b) ZOL and BRONJ-like lesions will alter the structure and tissue-level features of the jaw; and (c) ZOL dose will be associated with reduced osteoclast number in the jaw and suppressed systemic bone resorption. Testing the first hypothesis will involve using data from a previously reported experiment (Messer et al., 2018) conducted simultaneously to this one.

| Animal care
A monogamous continuous-breeding system was used to generate the experimental rice rats. All rats were housed (2-5 per cage) in static filter top cages (area: 143 in 2 ) with pine shavings and continuous access to food and water. The housing rooms were maintained at 20-26ºC with humidity of 30%-70% and a 12:12 hr light:dark cycle, or 14:10 hr cycle for breeders (Aguirre et al., 2015). Experimental animals ate pelleted, semi-purified HSC diet (TestDiet 5SXA AIN-93G w/no Complex Carbohydrates; Smelt Feed and Pet Supply, Tampa, FL, USA), which can be fed for up to 30 weeks without inducing significant metabolic abnormalities (Aguirre et al., 2016). This diet is based on the formula of the powdered Harvard high sucrose 700 diet and the ration 100 diet previously used to accelerate generalized PD in rice rats (Auskaps et al., 1957;Gotcher & Jee, 1981;Ryder, 1980). Breeder rats ate rodent breeder chow (Envigo Teklad, irradiated 2919). Body weight (BW) was measured bi-weekly. If rats showed BW loss and/or reduced body condition score (BCS), saline was administered subcutaneously and gel diet was offered (Ullman-Cullere & Foltz, 1999). The Animal Care F I G U R E 1 Experimental design. Rats were weaned at age 4 weeks, placed on high sucrose-casein diet and randomized into dosing groups. Starting at age 4 weeks, rats received tail vein injections of 0, 8, 20, 50 or 125 µg/kg zoledronate (ZOL) q4wks until necropsy. One group of rats was necropsied at age 4 weeks (n = 12). Groups of rats (n = 9-16) within each dose level were necropsied at 12, 18, 24 or 30 weeks of ZOL treatment

| Study design
Two hundred and twenty-eight clinically normal, female weanling rice rats (BW ≥ 30 g; BCS ≥ 3.0) were BW-randomized into a baseline group (n = 12; age, 4 weeks) or one of five dose groups (Figure 1), with efforts made to assign littermates into different dose groups.
A previous study showed that both male and female rice rats given oncologic doses of ZOL develop BRONJ . Only females were used in the current study to avoid variation in measurements due to sexual dimorphism (Aguirre et al., 2015). All groups were fed pelleted HSC diet and received intravenous (IV) injections of 0 (n = 42), 8 (n = 44), 20 (n = 44), 50 (n = 40), or 125 µg/kg BW of ZOL (n = 46) every four weeks (q4wk), starting at 4 weeks of age. After 12, 18, 24 and 30 weeks of treatment, duration subgroups (n = 9-16) from each of the five dose groups were necropsied (Supporting Information Table S1 and Figure   S1). The minimum sample size (n = 9) was determined by power calculation (see Statistical Analysis, Section 2.11). Animals were added to later time points based on a 10% censor rate to account for rats removed from the study before reaching scheduled time points, based on attrition rates from previous long-term rice rat studies.
To compose each duration subgroup, rats with progressive BW loss that reached ≥15% were chosen to comply with humane endpoints specified in the IACUC protocol. The remaining rats for each duration subgroup were selected by BW randomization. No more than 30% of each dose/duration subgroup was composed of rats that met the BW loss endpoint criteria for euthanasia.

| ZOL dose
The dosing strategy has been previously described (Messer et al., 2018). Briefly, we used 8 μg/kg IV q4wk as the rat equivalent of the human osteoporosis (OP) dose. The dose gradient in the current study includes 0, 8, 20, 50 and 125 µg/kg, which encompasses both the OP and a range of oncologic ZOL doses. For each dose concentration, ZOL was dissolved in normal sterile saline (pH 7.2). ZOL was injected into the tail vein of rats placed in rodent restrainers at the rate of 0.4 ml/100 g BW. ZOL was kindly provided by Novartis Pharma AG (Basel, Switzerland).

| Euthanasia and tissue collection
Rats were euthanized by CO 2 followed by cardiac puncture to withdraw about 1.5 ml of whole blood. Cervical dislocation was used as a secondary confirmatory method of euthanasia. Maxillae and mandibles were excised as before (Messer et al., 2018). High-resolution (HR) photographs from the medial aspect of each hemi-mandible (Supporting Information Figure S2a) and ventral aspect of each hemi-maxillae (Supporting Information Figure S2b) were taken with a digital camera (Canon EOS 6D; Tokyo, Japan) attached to a macro lens (Canon EF 100 mm 1:2.8; Tokyo, Japan). Jaw quadrants were fixed at 4°C in freshly prepared 4% paraformaldehyde, then transferred to 70% ethanol after 48 hr.

| Gross analysis of oral lesions
Jaw quadrant photographs were independently assessed by blinded observers (JGM, DBK, JIA). Quadrants were assigned a Gross Quadrant Grade (GQG) that reflected the severity of its gross lesions (Table 1) (Messer et al., 2018). Photographs of representative maxillae and mandibles representing each GQG are shown in Supporting Information Figure S2. When at least one quadrant in a rat had an area of exposed bone, the rat was considered positive for a gross BRONJ-like lesion. Neither pocket depth nor clinical attachment level by probing were evaluated. Photographs that did not receive the same GQG from all three independent observers were re-examined by all three observers together, and a GQG was assigned when consensus was reached.

| Specimen selection for histopathologic BRONJ-like lesion analysis
Histopathologic confirmation of BRONJ-like cases was undertaken in rats that were: (a) diagnosed with gross BRONJ-like lesions; (b) not diagnosed with gross BRONJ-like lesions, but had quadrants with oral lesions of GQG ≥2; and (c) not diagnosed with gross BRONJ-like lesions, but experienced prenecropsy BW loss (see Supporting Information Figure S1). Quadrants with GQG ≥ 2 were selected because these quadrants displayed gross gingival recession and ulceration. For rats with BW loss, all quadrants were analysed because data from a previous study showed that rice rats with BRONJ-like lesions had BW loss compared to rice rats without these lesions (Messer et al., 2018). In the current study, 37% (80/218) of rats necropsied at weeks 12-30 exhibited BW loss. 92% of rats with gross BRONJ-like lesions (22/24) and 73% (22/30) of rats with BRONJ-like lesions diagnosed only using histopathology (without prior gross diagnosis) had BW loss.

| Specimen selection for PD analysis
Right maxillae and mandibles from all ZOL 0 rats, and right maxillae from baseline rats, regardless of GQG, were examined to determine periodontal status and assigned a PD Score (Table 2). When right quadrants were not available, left quadrants were used.

| Histological preparation of jaw quadrants
Maxillae and mandibles selected for sectioning were decalcified in 5% formic acid under gentle agitation for 4 weeks, with three changes per wk. Decalcified jaw quadrants were trimmed, paraffin-embedded, and sectioned at 5 µm thickness in the mesiodistal plane from the lingual to buccal aspect of each quadrant (Leica/ Jung 2265 and Accu-Cut SRM 200 Sakura microtomes, Sakura Finetek Europe B.V, Zoeterwoude, The Netherlands). Maxillae and mandibles were serially sectioned at five or six levels, starting approximately 1 mm lateral to the midline of the palate or at the lingual plate, respectively. Levels were separated by approximately 250 µm, allowing thorough examination of tissues at the lingual surface (one or two levels), around the molars (three levels), and at the buccal surface (one level). Sections were H&E stained and cover slipped.

| Histopathologic definition of BRONJ-like lesions and quantification of necrotic bone
Histopathologic BRONJ-like lesions were defined as bone that was both exposed and necrotic in at least one level within a quadrant.

| Histopathologic characterization of PD
Sections were assessed for PD Score (Table 2) and alveolar bone loss as previously described (Aguirre, Akhter, Kimmel, Pingel, Xia et al., 2012). A quadrant with PD Score ≥1 was considered positive for PD.
A rat with at least one quadrant with PD Score ≥1 was considered positive for PD. The rat PD Score is the highest quadrant PD score for that animal. Alveolar bone height (ABH) was assessed in both maxillae and mandibles by measuring the distance between the ce- Measurements were made in blinded, randomized order (JGM). ABH in baseline rats was only measured in maxillae.

| Osteoclast quantification
To assess the local effect of ZOL, PD, and BRONJ on osteoclasts in the jaws, osteoclasts were analysed on the surfaces of maxillary alveolar TA B L E 2 PD score (inflammation scoring system) bone with PD (PD Score ≥1) in rats treated with ZOL 0 (n = 11) or ZOL 8-125 (n = 4-12), ZOL-treated rats with BRONJ (n = 10), or ZOL 0 rats with no PD (n = 12  There was a positive correlation between PD Score and time on HSC diet (R = 0.496, p = 0.00124). (c) Alveolar bone height (ABH), measured as the distance between the cementoenamel junction (CEJ) to the alveolar bone crest, in mandibles was significantly lower after 24 and 30 weeks than at 12 and 18 weeks; ABH in maxillae was significantly lower after 12, 24 and 30 weeks compared to baseline; * indicates p < 0.05 compared to 12 weeks in mandibles, † indicates p < 0.05 compared to 18 weeks in mandibles, # indicates p < 0.05 compared to 0 weeks in maxillae. Mean ± SD. (d) Comparison of mandibular interdental regions (M1M2) of PD Score 0 and PD Score 3; PD Score 3 is characterized by plaque accumulation ( ‡), epithelial (epi) hyperplasia, junctional epithelial apical migration, inflammatory infiltrate (black arrow head) in periodontal ligament (PDL) and alveolar bone loss et al., 2014) and recommended (Bouxsein et al., 2010). Images were acquired using the following parameters: 80 kVP/120 µA, 0.5 mm aluminium filter, 2k camera resolution, 10 µm voxel size, 0.5° rotation step, and 360° tomographic rotation.

| MicroCT
Two analyses were completed. First, 2D images were aligned in the mesiodistal plane with all molars simultaneously visible for qualitative image analysis. Second, 2D images from the mesial aspect of M2 were aligned in the buccolingual plane and the width of the alveolar ridge was measured (Figure 9d). Mandibles from ZOL 0 rats were compared to rats treated with oncologic ZOL (20-125). To consistently measure the same area in all mandibles, the slice through the apical foramen of the mesial root of M2 was selected. Then, the vertical length of M2 between the CEJ and apex of the mesial root was measured. The buccolingual width of the alveolar ridge was then measured perpendicular to the M2 mesial root at 30% (coronal width) and 60% (apical width) the total length of the CEJ-apex distance (ImageJ Software, NIH). After scanning, quadrants underwent histopathologic analysis (See Sections 2.6 and 2.7).  Tyne & Wear, UK), in serum collected from rats treated for 24 weeks.

| Statistical analysis
Chi-square test was used to determine the relationship between BW loss and occurrence of BRONJ, between PD type (localized or generalized) and oral lesion location (maxilla or mandible), and between PD type (localized or generalized) and total number of BRONJ
There was significantly greater alveolar bone loss at 24 and 30 weeks than at 12 and 18 weeks in mandibles, and at 12, 24 and 30 weeks than at baseline in maxillae (Figure 2c). Moderate PD in rice rats (PD Score 3) rats showed bacterial plaque, apical migration of junctional epithelium, alveolar bone resorption, increased ABH, gingival hyperplasia, inflammatory infiltrate, and PDL disruption ( Figure 2d).

| Prevalence of BRONJ-like lesions
The prevalence of gross BRONJ-like lesions is shown (Figure 3a).

| Gross and histopathologic
Gross BRONJ-like lesions were found in both maxillae ( Figure 4a) and mandibles ( Figure 4d) and showed exposed bone, ulcerated mucosa, plaque and irregular gingival contours around the margins of the molars. Gross oral lesions were distributed evenly between the maxillae (46%) and mandibles (54%) (p = 0.1953).
Histopathologic all such cases, exposed necrotic bone was present in sections from adjacent levels within the same quadrant (Supporting Information Figure S3).
The percentage of empty osteocyte lacunae in BRONJ rats was significantly higher (p < 0.05) than in PD control rats in the oncologic ZOL dose groups (20-125) (Figure 4g), and similar results were found when empty osteocyte lacunae frequency was normalized to bone area (Figure 4h).

| MicroCT
Periodontitis-affected mandibles exhibited loss of ABH mesial to M1, in the M1 furcation, and in the M1M2 interdental region (Figure 5b), compared to normal mandibles (Figure 5a). Mandibles affected by BRONJ-like lesions exhibited large regions of mottled alveolar bone with numerous pores, giving the entire alveolar bone region a honeycomb-like appearance (Figure 5c,d). In severe BRONJ-like lesions,

F I G U R E 4 Features of necrotic bone in BRONJ-affected maxillae and mandibles of rats injected with zoledronate (ZOL) at gross (a, d) and microscopic levels (b, c and e, f). High-resolution photographs of gross BRONJ lesions in (a) maxillae and (d) mandible
; note heavy plaque accumulation ( ‡) and ulcerated mucosa with exposed bone (arrow). White lines indicate approximate location of sections shown in b, c and e, f. Parasagittal sections through (b, c) the palatal bone of maxillae and (e, f) the mandibular lingual plate with H&E stain; ulcerated mucosa with inflammation (black arrowheads) in the lamina propria (LP) and epithelial (epi) hyperplasia; exposed bone surface is colonized by bacteria ( ‡). Border between healthy and necrotic bone is demarcated in yellow. (c, f) Higher magnification of exposed bone tissue shows extensive field of empty osteocyte lacunae (yellow arrowheads). (g) The percentage of empty osteocyte lacunae in alveolar bone was higher in BRONJ lesions in rats treated with ZOL 20, 50 or 125 compared to BRONJ lesions in ZOL 0 (*p < 0.05) and ZOL 8 ( †p < 0.05) rats. (h) The number of empty osteocyte lacunae per bone area (mm 2 ) was significantly higher in ZOL 20, 50 or 125 doses compared to ZOL 0 (*p < 0.05). Mean ± SD (a) profound alveolar bone osteolysis, tooth migration, and spaces with apparent missing teeth were noted (Figure 5d). Histological analysis from quadrants shown in Figure 5c,d confirmed localized areas of exposed, necrotic bone at the M1M2 interdental alveolar bone crest and distal to M3.

| Alveolar bone osteoclasts
Compared to ZOL 0 rats with no PD (Figure 6a    This study supports the role of local oral inflammation as a key component in the development of BRONJ that is not directly related to dental surgery. The current study, along with studies in other rodent models, has shown that BRONJ-like lesions can arise from several types of oral inflammation processes or wounds, which is consistent with human MRONJ. BRONJ-like lesions have been observed when ZOL treatment is combined with mechanically exposed dental pulp (Kang et al., 2013;Wayama et al., 2015), experimental F I G U R E 7 Peripheral quantitative computed tomography of the femoral distal metaphysis and mid-diaphysis after treatment with zoledronate (ZOL) 0-125 for up to 30 weeks, starting at age 4 weeks. (a) Total distal metaphyseal volumetric bone mineral density (vBMD; mg/cm 3 ); (b) total volumetric distal metaphyseal bone mineral content (vBMC; mg/mm); (c) total cortical bone area (mm 2 ); and (d) cortical vBMC (mg/ mm). The main effects of dose (p < 0.001) and time (p < 0.001) were significant for all measurements, and there were some significant differences among doses within each time point. However, there were no significant differences among the ZOL oncology doses at any time. *indicates p < 0.05 compared to 0 ZOL; ‡ indicates p < 0.05 compared to 0 ZOL and 8 ZOL. Mean ± SD PD via dental ligature (Li et al., 2016) and in association with localized intraoral lesions where impacted food and hair produce periodontal inflammation (Messer et al., 2018;de Molon et al., 2014).

| D ISCUSS I ON
The requirement of local oral factors is also supported in the current study by a subpopulation of rice rats (32%) given oncologic doses of ZOL that never developed BRONJ despite receiving cumulative ZOL doses up to 1,000 µg/kg over 30 weeks. These rats had GQG of 0 (92%) or 1 (8%), and exhibited no BW loss at necropsy, suggesting healthy oral cavities. Similarly, Sprague Dawley rats treated up to 2 years with alendronate at up to 10 times the OP dose and ZOL at up to seven times the oncology dose show no BRONJ-like lesions (Merck, 2018;Novartis, 2018), possibly because of the known resistance of rodents from the genus Rattus to develop PD (Graves, Kang, Andriankaja, Wada, & Rossa, 2012;Oz & Puleo, 2011;Struillou, Boutigny, Soueidan, & Layrolle, 2010;Weinberg & Bral, 1999).
While inflammation or infection from oral diseases, or direct compromise of the gingival barrier via a wound, appears to be crucial for developing BRONJ, evidence from rice rat studies suggests that the type of oral inflammation may produce BRONJ in distinctive patterns, and with distinctive features when bone resorption is suppressed by ZOL. In the FILP model, BRONJ-like lesions developed primarily in maxillae (74%) compared to mandibles (26%) (p < 0.0001), and there was a significant increase in prevalence dependent on dose, but not duration of ZOL treatment (Messer et al., 2018). In the current study, rice rats fed the HSC diet developed generalized PD but not FILP. In this generalized PD model, BRONJ lesions were distributed evenly between maxillae and mandibles, and were associated with ZOL duration, but not dose. The FILP model also produced a significantly higher total prevalence of histopathologic BRONJ compared to the generalized PD model (p < 0.0001), at 47% (82/174) vs 31% (53/171), respectively (Messer et al., 2018). The distinctive relationships between ZOL and BRONJ prevalence may reflect differences in the initiation and progression of the two types of PD. The highest BRONJ prevalence in the generalized PD model occurred at F I G U R E 8 Serum bone turnover marker concentration in rice rats. (a) P1NP and (b) TRAcP 5b after treatment with ZOL 0-125 for 24 weeks. *p < 0.05 compared to ZOL 0; † p < 0.05 compared to ZOL 8. (c) P1NP and (d) TRAcP 5b in rats treated with ZOL 0 or oncologic ZOL doses (20-125) for 24 weeks and grouped by BRONJ status; *indicates p < 0.05 compared to non-BRONJ ZOL 0; **indicates p < 0.05 compared to non-BRONJ ZOL 0 and non-BRONJ ZOL 20-125. Non-BRONJ ZOL 0, n = 9; non-BRONJ ZOL 20-125, n = 18; BRONJ ZOL 20-125, n = 8. Mean ± SD 30 weeks, while the highest prevalence of BRONJ in the FILP model occurred at 18 weeks and plateaued at later time points (Messer et al., 2018). This pattern is congruent with the time points in which the PD lesions become moderate to severe in each model, as FILP occurs in up to 75% of rice rats by age 16 weeks (Messer et al., 2018), while destruction of tissues in the generalized PD model was more prevalent and more severe in this study after 24-30 weeks. These data suggest that the FILP model may be more severe, and produces BRONJ-like lesions more rapidly compared to generalized PD. The direct trauma to the soft tissues from the impacted materials of the FILP lesion may directly introduce microorganisms to periodontal tissues through the wound, which is in contrast to generalized PD, which is characterized by accumulation of plaque on tooth surfaces, and inflammation in the gingival and periodontal tissue. The different effects of the local oral factors on the time-to-onset of BRONJ in the two rice rat PD models may partly explain findings in humans, in which the time-to-onset of BRONJ after exposure to pARs is widely variable, ranging from 0.1 to 19.9 years (Fung et al., 2017).
The histopathological examination of BRONJ-like lesions at the tissue level was consistent with previous studies in rice rats, where exposed necrotic bone with attached bacterial colonies was observed (Aguirre, Akhter, Kimmel, Pingel, Xia et al., 2012). In the current study, osteonecrosis was not found under intact mucosa in any of the 198 examined quadrants. Additionally, four lesions showed bone that was exposed, but vital. Taken together, these novel findings suggest that bone in the periodontal lesion of pAR-treated subjects may become exposed before becoming necrotic. pARs with clinical findings of moderate/advanced PD or periapical pathology, such as bleeding on probing or intermittent tooth sensitivity, may actually have exposed, necrotic bone in periodontal pockets or in the periapical region that is only histologically detectable.
One maxilla met all criteria for a BRONJ-like lesion in a ZOL 0 rat at age 34 weeks, indicating that dead alveolar bone in rice rats with severe PD that were not treated with a pAR, can occasionally accumulate in sufficient quantity to be diagnosed as BRONJ. Bone necrosis is not generally observed in clinical examination of humans with severe PD who do not take pARs. This may be due to lack of routine tissue biopsy of severe PD lesions in humans, or because necrotic bone may be promptly removed by normal osteoclastic activity in the absence of pARs. Indeed, delayed removal of necrotic bone in humans taking pARs has been observed clinically in the treatment of femoral head osteonecrosis (Agarwala & Shah, 2011;Hong, Luo, Lin, Zhong, & Shi, 2014;Lai et al., 2005;Luo, Lin, Zhong, Yan, & Wang, 2014), suggesting that pARs may also cause persistence of necrotic bone induced by severe PD in the oral cavity, which could eventually result in clinically detectable MRONJ. Alternatively, oral ulceration with bone sequestration (Almazrooa & Woo, 2009;Farah & Savage, 2003;Khan et al., 2015;Palla, Burian, Klecker, Fliefel, & Otto, 2016;Peters, Lovas, & Wysocki, 1993;Sonnier & Horning, 1997) occurs in humans not taking pARs and may have some relevance to this finding.
MicroCT images of mandibles affected with BRONJ-like lesions revealed mottling of alveolar bone distant from the necrotic bone tissue region identified histologically. These pores may be the result of osteoclast activity far from localized sites where necrotic bone has accumulated. This finding suggests that the presence of exposed necrotic bone tissue may activate bone resorption throughout the entire alveolar process of the afflicted quadrant, bearing similarity to a regional acceleratory phenomenon (RAP) (Duncan, Frame, Frost, & Arnstein, 1969;Mueller, Schilling, Minne, & Ziegler, 1991;Verna, 2016). Indeed, the number of TRAP+ osteoclasts per mm bone perimeter was significantly higher in maxillary alveolar bone of BRONJ-affected rats, compared to maxillary alveolar bone of rats affected by PD, suggesting that exposed necrotic bone with a more extensive area of soft tissue damage and inflammation promote a greater recruitment and activation of osteoclasts.
MicroCT analysis also revealed that ZOL monotherapy can increase the width of the alveolar process in rice rats. This provides further evidence that in rats at risk for BRONJ, the entire alveolar  (Weinstein, Roberson, & Manolagas, 2009). These discrepant findings at the bone tissue level between osteoporosis patients and rice rats may be due to variation in the dose (osteoporosis vs oncologic) and type of bisphosphonate (ZOL vs alendronate), and in particular, differences in the specific bone type and microenvironments associated with PD and osteoporosis. Maxillary alveolar bone is anatomically and functionally different from the trabecular bone from the pelvis. In addition, jawbones of PD-affected rats are in close proximity to pathogenic microbes of the oral cavity. The inflammatory microenvironment of PD therefore bears limited similarity to the non-inflamed microenvironment in trabecular bone of the iliac crest.
Taken together, these findings support the idea that the unique anatomical and physiological features of alveolar bone play an important role in the specificity of osteonecrosis to the jaws.
The design of this study limits the interpretation of our findings in a few important ways. First, ZOL injections began in normal, periodontally healthy rice rats. Two injections occurred at ages 4 and 8 weeks, when the skeleton is considered juvenile. Since MRONJ is an adverse event that occurs in humans first exposed to pARs in adulthood, future rice rat studies should begin administration of MRONJ-inducing medications after the skeleton has matured (after age 12 weeks) to improve the parallel with the human condition. Second, human MRONJ also occurs in the setting of cancer or osteoporosis where patients, unlike rice rats, not only have a primary disease, but also receive other types of medications. These underlying health issues, or drug treatments, particularly in cancer patients, may elevate the risk of MRONJ. For example, corticosteroids, which negatively influence healing (Rubin & Palestine, 1989;Zitelli, 1987), are frequently used in treatment protocols for patients with cancer or autoimmune/inflammatory diseases that cause secondary osteoporosis (e.g., rheumatoid arthritis), and may suppress healing of oral wounds. This could account for the observation that humans exposed to oncology doses of ZOL have a roughly 200-fold higher prevalence of MRONJ than osteoporosis patients (Coleman, Marshall et al., 2011;Coleman, Woodward et al., 2011;Grbic et al., 2010;Rathbone et al., 2013;Ruggiero et al., 2014), which is much higher compared to rice rat data, where oncology doses produced only 2-to 10-fold higher BRONJ prevalence compared to the OP dose after 30 weeks. Third, this study is limited by our inability to fully examine PD and BRONJ in live animals. Thirty-six per cent of BRONJ cases were found in rice rats with no gross oral cavity lesions (GQG < 1), but were symptomatic in the sense that the rats experienced loss of body weight. Diagnostic criteria available in the clinical setting (e.g., periodontal probing, patient-reported pain, dental radiography, cone beam CT) cannot be achieved in a rodent study, so the exact progression of PD and BRONJ, or any spontaneous healing, could not be assessed. Finally, observations of BRONJ in rice rats at necropsy did not allow assessment of the duration of bone exposure. Longitudinal observations and oral examinations in live rice rats in future studies may address some of these limitations.
The reduction in morbidities and mortality, along with the low risk of MRONJ, strongly favours the continued use of pARs in humans.
Therefore, completely describing the aetiology and pathophysiology of MRONJ is more important than ever to patient safety, and could provide evidence that allows development of prevention and treatment strategies for both dental surgery-related and non-surgery-related MRONJ. Additionally, recent case reports describe patients with MRONJ who were not exposed to anti-resorptive or anti-angiogenic medications (Aghaloo & Tetradis, 2017), suggesting that MRONJ remains an emerging disease that may be associated with yet other medical conditions or drugs.

| CON CLUS ION
This study demonstrates a duration-dependent increase in the prevalence of BRONJ with oncologic ZOL in rice rats with progressive generalized PD. BRONJ prevalence plateaued at a dose only 2.5-fold higher than prevalence with the osteoporosis dose and was consistent with a similar plateau in the anti-resorptive effects of ZOL.
BRONJ and ZOL also produced unique alterations at the tissue level compared to PD and healthy rats treated with ZOL. The findings demonstrate that local oral risk factors play a key role in establishing an environment in which systemic medications can induce BRONJ, but that different local factors may interact with pARs in distinctive ways. Pharma AG (Basel, Switzerland) for providing zoledronate.

CO N FLI C T O F I NTE R E S T
The authors have no conflicts of interest. CVP's Institution receives research support from Bayer. CVP receives a small amount of salary support from the research support and royalties from UpToDate for writing for Bayer.