Establishment and assessment of rodent models of medication-related osteonecrosis of the jaw (MRONJ)

Medication-related osteonecrosis of the jaw (MRONJ) is primarily associated with administering antiresorptive or antiangiogenic drugs. Despite significant research on MRONJ, its pathogenesis and effective treatments are still not fully understood. Animal models can be used to simulate the pathophysiological features of MRONJ, serving as standardized in vivo experimental platforms to explore the pathogenesis and therapies of MRONJ. Rodent models exhibit excellent effectiveness and high reproducibility in mimicking human MRONJ, but classical methods cannot achieve a complete replica of the pathogenesis of MRONJ. Modified rodent models have been reported with improvements for better mimicking of MRONJ onset in clinic. This review summarizes representative classical and modified rodent models of MRONJ created through various combinations of systemic drug induction and local stimulation and discusses their effectiveness and efficiency. Currently, there is a lack of a unified assessment system for MRONJ models, which hinders a standard definition of MRONJ-like lesions in rodents. Therefore, this review comprehensively summarizes assessment systems based on published peer-review articles, including new approaches in gross observation, histological assessments, radiographic assessments, and serological assessments. This review can serve as a reference for model establishment and evaluation in future preclinical studies on MRONJ.


INTRODUCTION
Medication-related osteonecrosis of the jaw (MRONJ) is an adverse side effect of antiresorptive and antiangiogenic medications widely used to treat bone metastasis and osteoporosis 1 . The clinical manifestation of MRONJ is bone exposure with or without intraoral/extraoral fistula in the maxillofacial region, lasting over 8 weeks 1,2 , which is staged 1 to 3 (stage 0 represents the prodromal period without specific clinical or radiographic symptoms) according to the MRONJ staging system updated by the American Association of Oral and Maxillofacial Surgeons (AAOMS) in 2014 1,3 . Although keeping new ossified periosteum for the neo-mandible has been utilized as a therapeutic approach, massive resection of the jaw bone caused by MRONJ still affects the physiological and mental health of patients 4 . Furthermore, the pathogenesis of MRONJ has not been fully illuminated. Previous clinical and preclinical studies have indicated that systemic risk factors (e.g., drug administration [5][6][7][8][9] , patients' medical conditions 10,11 ) and local oral risk factors (e.g., tooth extraction 12,13 , pulpitis 14 , and periodontitis 15 ) participate jointly in the development of MRONJ.
The animal model is an essential in vivo experimental platform for exploring the pathogenesis of and interventions for various diseases. Compared with large vertebrates, rodents have universal benefits of easy feeding and operation and relatively low cost. Among all kinds of MRONJ animal models from 2003 to date, most were rodents, as 60% were rats, 27% were mice, and 2% were rice rats 16 . Large vertebrates such as dogs (~4%), pigs (~3%), and sheep (~2%) have also been applied to establish MRONJ models 16 . However, rodents boast additional advantages in MRONJ research because the presence of Haversian remodeling in large vertebrates is not involved in the development of MRONJ. Exploration into the pathogenesis of and interventions for MRONJ in rodents has been ongoing since the initial case report of MRONJ in 2003 (Fig. 1). Sonis et al. first reported an MRONJ model in rats in 2009, which was established by systemic administration of antiresorptive drug following extraction of molars 17 . In their study, rats treated with zoledronic acid (ZA), an antiresorptive drug, plus dexamethasone (DEX) presented unhealed extraction sites characterized by a base of exposed bone, some erythema, necrotic bone, and areas of inflammatory infiltration. These macroscopic and histologic findings are consistent with established criteria of MRONJ diagnosis in clinic, indicating the occurrence of MRONJ in rats. Since then, drug administration as a systemic risk factor combined with tooth extraction as a local risk factor has become the classical method of MRONJ model establishment.
Over time, modified rodent models with local inflammation and infection have been established to complement the simulation of infectious tooth extraction or spontaneous MRONJ without invasive operation in clinic. These modified MRONJ rodents were established by local risk factors of pulpitis or aggressive periodontitis 18,19 . Another powerful model for spontaneous periodontitis, the rice rat, has also been utilized in MRONJ research 20,21 . Apart from different local risk factors adopted in establishing models, to further simulate abnormal bone tissue metabolism conditions exhibited by MRONJ patients, such as osteoporosis, researchers have created MRONJ models based on ovariectomized (OVX) rodents 8,16,22-28 . In general, the approach of establishing MRONJ rodent models mainly constitutes two steps: drug administration as a systemic risk factor and oral stimulation as a local risk factor (Fig. 2). The ideal MRONJ model should imitate the pathophysiological characteristics of MRONJ, including bone necrosis, inflammatory cell infiltration, and angiogenesis inhibition. Current MRONJ models only mimic some of the pathophysiology with the systemic and local risk factors used for model establishment. At present, specific questions, such as which rodent 29,30 , drug dosage and frequency 20,[31][32][33][34][35][36] , and local risk factor 18,[37][38][39][40][41][42] should be selected, have not received unitive answers yet. Thus, in this review, we summarize the selections made by existing classical and modified rodent models of MRONJ.
Furthermore, the assessment of the established MRONJ model plays a critical role in proving the effectiveness and efficiency of the methods and in verifying the reliability of the models. As various techniques have been introduced for the assessment of MRONJ models, we comprehensively overview current techniques in gross observation, histological assessments, radiographic assessments, and serological assessments, which will contribute to the establishment of a unified MRONJ model assessment system. To reduce variation in the criteria used to define MRONJ in current studies, we also summarize highlights in the assessment of MRONJ.

APPROACHES OF ESTABLISHING MRONJ MODELS
Classical method Systemic drug induction plus tooth extraction is the most popular approach adopted for MRONJ rodent model establishment, as tooth extraction is the most commonly reported local risk factor. In fact, 52%-61% of MRONJ patients have a history of tooth extraction, and an individual's risk of MRONJ is 16 times higher after tooth extraction 43 . In previous rodent animal studies, the protocol used to establish MRONJ models includes the selection of animal, drug type, dosage, duration, co-medications, route of administration, tooth extraction site, and the time interval between tooth extraction and drug induction.
Considering those discrepancies in each part may cause diverse protocols for MRONJ models, we carefully analyzed and elucidated the most commonly used protocols, taking into consideration the animal selection, drug type, dose, administration route and time, and tooth extraction sites and intervals. Given the cost and effectiveness, we also sorted out the differences in administration times, total duration, and success rates between classical methods to provide references for subsequent researchers in establishing their protocols (Table 1).
Regarding species selection, approximately 60% of relevant studies chose rats rather than mice for research, usually 8-12 weeks old and female [44][45][46][47][48] . Rats were selected due to their low cost, rapid and easy reproduction, and simple maintenance conditions 49 . They have a larger size and can live longer in such a long experimental period compared with mice. Adult rodents were usually selected for studies based on clinical observation, suggesting osteonecrosis risk increased with age 17 . Yet mice are more advantageous for exploring biological mechanisms, especially transgenic mice 50 . 2021 Mine et al. Experimental hyperocclusion was adopted for induction of MRONJ-like lesions.
Used an immunocompromised mouse model.
Anti-RANKL antibody-treated mice were built for denosumab-related osteonecrosis.
Periapical disease and BPs induced osteonecrosis of the jaws in mice.
Rice rats with periodontitis were first adopted for induction of MRONJ-like lesions.
Combined the oral bacteria with tooth extraction in a mouse model of MRONJ.
Report of a rat model of MRONJ constructed with a sequence of ZA and DEX following extraction of mandibular or maxillary molars. When it comes to drug administration for the systemic risk factor, antiresorptive drugs, mainly bisphosphonates (BPs) and anti-receptor activators of NF-κB ligand (RANKL) antibody (Ab) (denosumab in clinic), are the most common choice, especially BPs. In contrast, antiangiogenic drugs have not been widely used for general MRONJ models. RANKL monoclonal antibody (mAb) is usually given in combination with cyclophosphamide (CY) at a dose of mAb (5 mg·kg −1 ) once every three weeks plus CY (150 mg·kg −1 ) twice a week before tooth extraction and once a week after tooth extraction for five or seven weeks 51,52 . Among various BPs, ZA is the most potent induction drug because it is associated with the highest risk of MRONJ onset in clinic 53,54 . The induction dose of BPs on rodents is related to the dosage used in osteoporosis patients or bone tumor patients 17 . Kim et al. pointed out that the drug induction dose should refer to four aspects: the relevant ZA doses used in humans (oncologically, 67 µg·kg −1 for 4 weeks), the relatively rapid bone metabolism of rodents, the route of drug administration, and maximizing drug exposure during the relatively short experimental period 26 . Considering the routes of administration applied in MRONJ rodents, intraperitoneal administration is the simplest technique, but it requires a high drug dosage. Intravenous injection requires greater precision, but it possesses the fastest onset and the highest bioavailabilityplasma drug concentrations of the subcutaneous and intramuscular routes are lower than those of the intraperitoneal and intravenous routes. In 2013, Kang et al. chose to use a higher dose of ZA, namely, 200 µg·kg −1 , which is approximately three times the oncologic ZA dose, to increase the incidence of osteonecrosis in mice. They also elected to inject mice three times per week to mimic the monthly injections in humans, given the estimation that 17 days of a rodent's life corresponds to one human year 41 .
In summary, the dosage of ZA ranges from 0.0075 to 2.25 mg·kg −1 , with the most commonly used drug regimen being 0.1 mg·kg −1 ZA thrice a week for 6-9 weeks through intraperitoneal injection 47,[55][56][57][58] . In addition, some studies used a corticosteroid drug such as DEX with BPs to increase the prevalence of osteonecrosis. Sanda et al. and other researchers created MRONJ models by subcutaneously or percutaneously administering ZA (0.066 mg·kg −1 ) and DEX (5 mg·kg −1 ) thrice a week for 4-6 weeks, and the success rate reached as high as 100% 46,59,60 .
The choice of extraction site varies less compared to drug induction. Most studies chose to remove the right or left maxillary first molar 59,61-67 . Maxillary teeth are easy to see during extraction, and compared with simultaneous extraction of bilateral maxillary first molars or unilateral three molars, the extraction of only one maxillary first molar is more straightforward, and it reduces the possibility of root fracture and causes minor trauma to the rodents, which is conducive to the rat's recovery of feeding ability.
The time interval between tooth extraction and drug induction depends on the overall time arrangement of the model establishment. The priority of the time arrangement is to maximize drug exposure while relatively shortening the experimental period. The duration of drug induction varies in previous studies, but 4-6 weeks is most common. Almost all studies chose to administer ZA with or without DEX or other drugs for systemic induction lasting 2-4 weeks before tooth extraction. After tooth extraction, about half of the researchers administered medication until the end of the experiment, whereas the other half kept the rodents under persistent observation until MRONJ developed. Considering the clinical definition of MRONJ, bone exposure in the human maxillofacial region should persist for more than eight weeks, which corresponds to approximately one week of a rat's life 55 . Thus, the MRONJ-like lesions should exist for at least one week in rodents before verifying the successful establishment of the MRONJ model. The total duration ranges from 2 weeks 68 to 17 weeks 56,58 , although it exceeds eight weeks in most studies 31,44,45,[69][70][71][72] , indicating a relatively long induction period.
In general, establishing the classical MRONJ rodent model is relatively simple, involving drug injection and tooth extraction. This approach has been widely used over the past two decades, suggesting good reliability and repeatability. the first hit is a long-term medication history, the second is pathologic inflammatory conditions, and the third is structural defects in soft tissue integrity caused by dentoalveolar trauma 42 . This model attaches importance to associations between MRONJ occurrence and surgical interventions under pathologic inflammatory conditions. Thus, some researchers suggested inducing inflammation in the extraction site to better mimic tooth extraction based onset of MRONJ in clinic 19,37,42,76,77 . Numerous rodent models under local dental infection (mainly pulpitis and periodontitis) have been established to complement classical methods by extracting infectious teeth. These modified methods adopting infection as a local risk factor can be summarized in three steps: injecting systemic drugs, inducing dental infection to create an inflammatory environment, and extracting the tooth in the inflamed area after a specified number of weeks 19,37,42,76,77 . Compared to the classical method, modified methods with infectious tooth extraction result in more severe MRONJ 19,42 . The presence of bone necrosis increases in infectious tooth extraction groups, with the larger necrotic bone areas and more empty osteocyte lacunae. Yet the overall time of establishing MRONJ is not prolonged compared with classical tooth extraction methods 19,42 .
Generally, MRONJ induction by pulpitis involves drilling a hole in the target tooth (first and second molars) to expose pulp for at least three weeks. The time of systemic drug induction is more than eight weeks. MRONJ induction by periodontitis, as another popular infection method, typically involves ligaturing 28,32,42,76,77,85,86 or a high sugar diet without ligaturing in rice rat models 33,[82][83][84] . Ligaturing is the conventional method for periodontitis induction; as shown in Table 2, more than half of the studies using MRONJ induction by periodontitis adopted ligaturing. The general process of ligaturing is similar to pulp exposure: silk ligatures are wrapped around the neck of the target tooth for at least 3 weeks of periodontitis infection, and the overall time of systemic drug induction is over four weeks.
The eventual assessment of pure infection-induced MRONJ rodent models without tooth extraction is generally regarded as MRONJ stage 0. There is no visible epithelial damage or necrotic bone but there is pathological necrotic bone and empty bone lacunae. This could be due to the lack of the third hit in the MRONJ development model 42 , because structural defects in soft tissue integrity are caused by tooth extraction. Several studies have tried to improve induction methods to generate more obvious MRONJ symptoms. As MRONJ development is associated with induction time, Hadaya et al. extended the ligaturing time to 10 weeks and the overall administration time to 22 weeks; the MRONJ model tissue sections showed continuous epithelial damage and necrotic bone exposure 32 .
As a standard animal periodontitis model, rice rats have also been used as MRONJ models with generalized periodontitis induced by a high sugar diet without ligaturing 33,[82][83][84] . The occurrence of gross MRONJ with exposure to the alveolar or palatal bone in rice rats is 13%-18%, and histological MRONJ is around 70% 33,84 . The systemic drug induction of rice rats is different from rats and mice. The most commonly used dose of ZA on rice rats is 80 μg·kg −1 , injected every 4 weeks for 24 weeks 82,83 . Compared to rat or mouse models, the induction time of rice rats is significantly prolonged. But the induction method is simplified as a standard diet with reduced ZA administration frequency.
Pulpitis and periodontitis are bacterial infections, and bone exposure to the oral cavity provides access to oral bacterial invasion in MRONJ development. As some studies pay attention to the function of oral microbiota colonized on the bone surface of MRONJ 28,87 , it is worth noting that to better control the baseline, amoxicillin, metronidazole, and other antibiotics are often used to remove possible oral pathogens ahead of exerting stimulation. Furthermore, a wash-away period of about 3 days is used to eliminate the impact of indigenous antibiotics 76 .
Mechanical stimuli-inducing method Mechanical stimuli besides tooth extraction have also been adopted as local risk factors in establishing MRONJ models, such as implantation, because the clinical risk of MRONJ onset after implantation is comparable to that after tooth extraction 1 . Inoue et al. placed an implant in rat maxilla after 12 weeks of drug injection to mimic the development of MRONJ around implants 88 . Bone grinding by drilling has also been used for more significant bone defects in establishing MRONJ models 89 . In addition to invasive dental procedures, other factors which exert sustained and micromechanical stimuli may also induce MRONJ. Previous studies have reported that occlusal overload on the prosthesis or caused by rheumatoid arthritis might have contributed to MRONJ 90,91 , but methods for establishing MRONJ models based on sustained mechanical stimuli are still in the preliminary stages. Mine et al. developed a novel mouse model with experimental hyperocclusion to investigate the potential role of occlusal/ mechanical trauma in MRONJ 92 . This model provides reasonable evidence for the feasibility of using the mechanical load in MRONJ models. However, the specific role of mechanical stimuli in MRONJ development has not been verified.
In short, classical methods, which combine systemic drug injection with healthy tooth extraction, currently offer the most versatility for MRONJ research. Modified methods adopting infection induction are expected to improve upon classical methods by extracting infectious teeth, thereby mimicking tooth extraction resulting from dental infections, which is much closer to MRONJ onset in clinic. Some MRONJ models are established by implantation, trauma, hyperocclusion, and other mechanical stimuli. For long-term MRONJ progression, we suggest that for research on the pathogenesis of MRONJ, various time points can be set in the pretest study to collect samples and examine the occurrence and development of MRONJ no matter which local risk factors are adopted. As an in vivo experimental platform, the timing of exerting preventive and therapeutic interventions on MRONJ rodent models should also be considered. For research on MRONJ prevention, interventions can be exerted immediately after applying local risk factors to observe whether the experimental manipulation can delay or inhibit the occurrence of MRONJ. Before testing the efficacy of interventions on treatments, it is recommended to wait for the completion of MRONJ models with a systematic evaluation, which verifies the selected animals have been successfully modeled for follow-up experiments, generally including tests on newly developed drugs and diverse applications of existing drugs, biological materials, and cell delivery (Fig. 3).

ASSESSMENT OF MRONJ-LIKE LESIONS IN RODENT MODELS
Assessment of MRONJ-like lesions in rodent models provides evidence presenting the effectiveness and efficiency of adopted methods. The assessment is generally composed of two parts: the occurrence rate of MRONJ and the characteristics of MRONJ-like lesions. Judgment on the occurrence of MRONJ includes the onset of gross MRONJ and histological MRONJ. The gross MRONJ presents unhealing mucosa and exposed bone observable by the naked eye. The definition of histological MRONJ varies among previous studies, but necrotic bone is the gold standard. It is thus clear that assessment of occurrence is based on evaluating characteristics of MRONJ-like lesions in rodents. As various techniques have been introduced to assess MRONJ-like lesions, we comprehensively overviewed current practices in gross observation, histological assessments, radiographic assessments, and serological assessments in Table 3. Generally, gross observation and histological and radiographic assessments have three essential aspects for assessing MRONJ rodents' lesions. First, all the studies presented histological and pathological findings. Second, more than half showed gross observation by the naked eye or radiographic assessment. Third, indicators for quantitative analysis have also been widely used. In addition to the comprehensive assessment system, we also highlighted the critical aspects of assessment. However, the heterogeneity of methods for establishing MRONJ rodent models in terms of species, drug type, and local risk factors brings difficulty to finding general characteristics in assessing MRONJ-like lesions. To reduce the bias in the evaluation, we screened the past 3 years of studies on establishing rat models by classical methods from Table 1. Also, we took into consideration other studies that utilized novel detection approaches.

Gross observation
Mucosal healing and bone exposure are two important indexes in gross observation. Most studies mainly describe the gross MRONJ as "incomplete and delayed mucosal healing and bone exposure" 30,44,45,47,62,64,67,72,[93][94][95][96] . However, in infection-induced MRONJ rodent models, the gross observation may only present mild to moderate inflammation with soft tissue swelling and erythema without exposure to bone tissue on probing 45,85 . For pulpexposed MRONJ rodent models, there may even be no oral mucosal lesion or exposed bone observed 39 .
Some studies conducted quantitative analysis; the whole unhealed mucosal area and ratios of the exposed area were calculated to assess the wound healing conditions 97,98 . Different scoring systems of mucosal healing were also put forward by researchers 67,72 . Merloni et al. defined three stages of healing conditions by the exposed area ratio as grade 1: socket surface more comprehensive than the occlusal area of the second maxillary molar and dark, rough, and irregular appearance of the wound surface; grade 2: intermediate features between grades 1 and 3; and grade 3: socket surface more minor than the occlusal area of the second maxillary molar 72 . Gao et al. established a blind healing score with ten levels based on the degree of mucosal closure in gross observation and the detection of sequestration in radiographic images 67 . The details of the healing score are as follows: 1-3: exposed necrotic bone plus evidence of sequestration in the micro-computed tomography (μCT); 4-6: mixed granulation tissue and mucosal coverage, some exposure, and no evidence of sequestration in μCT; 7-9: mixed granulation tissue and mucosal coverage with complete coverage, no sequestration; and 10: normal mucosal coverage, no sequestration, evidence of the regular bone filling of the defect 67 . For rice rats, the assessment of gross MRONJ depends on gross quadrant grade (GQG) (0-4), showing the severity of soft tissue lesions. Gross MRONJ of rice rat is defined as a lesion of GQG = 3 or 4 with exposure of alveolar or palatal bone 33,84 . It should be noted that the same calibrated examiner should perform the assignment of the score and measurement in a blind fashion.

Histological assessments
The indicators in histological sections commonly used to assess the healing conditions are epithelial integrity, inflammation, bone necrosis, osteoclast, osteoblast, and blood vessels. In histological assessment, the epithelial integrity usually refers to the length of necrotic bone exposed toward the oral cavity and the distance between the edges of the epithelial surfaces 59,60,63,98 . The distance between the edges of the epithelia is defined as the shortest end-to-end distance of the ripped epithelium 60 . The length of the necrotic bone exposed toward the oral cavity is defined as the longest distance of the exposed necrotic bone 60,63 . This assessment should note the consistency of the placement of samples when embedded in paraffin. The method of Soundia et al., useful as a reference, involves making cross-sections perpendicular to the long axis of the alveolar ridge at the area of the mucosal defect or the area between the first and second molars to determine whether the mucosa is complete healed 97 .  Main approaches for MRONJ treatment tested in rodent models involve novel drugs, biomaterials, and gene-engineered cells for delivery. A typical MRONJ mandible presents a prominent gingival ulcer with exposed necrotic bone (blue), osteolysis (black), and abscess formation (yellow) Inflammatory conditions are also assessed. The simplest method for inflammation evaluation is to use the number of polymorphonuclear cells under a fixed area 49,67,99 . Some studies utilized a simple scoring system based on the intensity of inflammatory cells in the defect area as 0: no inflammation; 1: mild inflammation; 2: moderate inflammation; and 3: severe inflammation 49,57,93 . A more complex system further considers the infiltration and bone sequestra for a comprehensive assessment 63 .
The formation of necrotic bone is the most important hallmark of MRONJ occurrence in rodent models. However, the definition of necrotic bone differs among various studies 49,[57][58][59]64,69,72,96,100 . Generally, necrotic bone is identified as an area of at least 5-10 confluent empty or karyolytic osteocyte lacunae 49,[57][58][59]69,72,96,100 . The area of necrotic bone was defined as the bone area with empty osteocyte lacunae, and the necrotic bone ratio was defined as the ratio of empty occupied osteocyte lacunae 59,60 . Instead of measuring necrotic bone, some studies directly used the average number of empty lacunae per area to indicate bone necrosis 30,45,95,96,98 . According to the histology images and statistics presented in different studies, it should be more objective and persuasive to calculate the empty lacunae and analyze the necrotic bone area to reflect the overall situation of bone defects.

Histopathological assessments
Healing conditions 59,60,63,98 Histological sections show the soft tissue healing with the distance between the edges of the epithelia, and bone defects with the length of the necrotic bone exposed towards the oral cavity.
Empty bone lacunae 30,45,95,96,98 The proportion or the number of empty bone lacunae in a certain area indicates the degree of osteocyte loss, which present the bone necrosis.
Blood vessels 65,67,69 The extent of angiogenic inhibition is assessed by the density of blood vessels generally marked by CD31. PET/CT 108 To present bone metabolism and inflammation with specificity and higher resolution.
Portable X-ray devices 94 To present bone quality by drawing the Regions of interest (ROI) to obtain the attenuation coefficient (similar to BV/TV), the ratio between the average ROI values on the surgery side and the control side.
SEM 101 To present osteocytes in bone lacunae.
TEM 106 To illustrate osteoclasts with ruffled border adjacent to the alveolar wall.
Raman spectroscopy 47 To calculate mineral/matrix ratio and carbonate/phosphate.
ICG-based NIF imaging 89 To mark affected bone tissues with pathological examination with quantification detection of fluorescence intensity.
A cross-modality imaging pipeline 107 To combine Atomic Force Microscopy and Scanning Electron Microscopy to acquire complementary hallmarks of MRONJ.
MRONJ medication-related osteonecrosis of the jaw; TRAP tartrate-resistant acid phosphatase; VEGF vascular endothelial growth factor; GluOC uncarboxylated osteocalcin; CTX-1 C-terminal peptide of type I collagen; TRAcP-5b tartrate-resistant acid phosphatase 5b; P1NP N-terminal propeptide of type I procollagen; PET/CT Positron emission tomography/computed tomography; μCT micro-computed tomography; SEM scanning electron microscope; TEM transmission electron microscope; ICG indocyanine green; NIF near-infrared fluorescence; PET/CT positron emission tomography/computed tomography osteoclasts and attached osteoclasts increases in MRONJ-like lesions 30,62 . For standardization comparison, it is recommended to calculate the number of osteoclasts or osteoblasts with the length of the bone surface. Disturbance in blood vessels is also one of the pathological characteristics of MRONJ. Studies have found that sustained ZA treatment causes a microcirculatory inflammatory reaction in the mandibular periosteum 102 . Thus, vessel assessment was performed mainly in studies exploring angiogenic effects in MRONJ models 65,72 . CD31 is the most commonly used marker, indicating vessels in histological sections 65,67,69 . Tamari et al. creatively utilized Dil molecules directly incorporated into the cell membrane, labeling functional vessels in red 103 .
Radiographic assessments Radiographic methods contribute to evaluating bone quality and formation of bone sequestra in MRONJ defects. Micro-CT (µCT) imaging was the most commonly used method. The description of MRONJ sites is usually based on sectional images or 3D stereoscopic images constructed from μCT data. For quantitative statistics, bone morphometric indices of bone volume fraction and bone volume/tissue volume (BV/TV) were utilized most frequently 30,48,59,60,69,[95][96][97][98]100,101,103,104 , which generally showed a significant decrease in most rat models 30,48,60,69,[95][96][97]100,101,103,104 . Trabecular separation (Tb.Sp) is another parameter from µCT, presenting thickness of space, in which a higher value indicates reduced connectivity of trabecular bone. Tb.Sp of MRONJ rodents shows an increasing trend 69,96,98,100,104 . Bone mineral density (BMD) exhibiting the bone mineral mass per bone volume was also calculated to assess the newly formed bone in MRONJ models used for testing therapeutic interventions 69,95,104 . 3D images constructed by μCT were also adopted to evaluate the size of bone defects comprehensively 105 .
In addition to µCT examination, other radiographic methods have been applied for assessing the bone quality of MRONJ-like lesions in rodents. Paulo et al. used a portable X-ray device for radiographic evaluation and analyzed the images through ImageJ 94 . To obtain the attenuation coefficient (similar to BV/TV), the ratio between the average values on the surgery side of the mandible and the control side was calculated. To assess changes in bone composition, De Sousa Ferreira et al. adopted Raman spectroscopy to calculate mineral/matrix ratio in bone tissue 47 . Besides, a scanning electron microscope (SEM) and a transmission electron microscope (TEM) were utilized for optical imaging to assess the cell morphology of osteoblasts 101 and osteoclasts 106 . Reier et al. established a crossmodality imaging pipeline combining µCT with atomic force microscopy and SEM to acquire complementary hallmarks of MRONJ 107 . These radiographic methods present changes in bone quality and cell morphology in MRONJ-like lesions, which is beneficial for further exploration of MRONJ pathophysiology.
Novel radiographic methods for MRONJ diagnosis and treatment have been tested in MRONJ rodent models to explore their feasibility in clinic. Positron emission tomography/computed tomography (PET/CT) appeared to be a sensitive imaging modality for identifying markers of inflammation and bone metabolism to diagnose MRONJ in a rat model, including a ZA/DEX group 108 . The decreased bone remodeling tendency highlighted by PET/CT may indicate a possible risk of MRONJ before the onset of clinical signs and symptoms. Xia et al. utilized indocyanine green (ICG), a molecular probe applied in bio-imaging for many years, to mark MRONJ-affected bone for removal and preserve normal tissue as much as possible for the first time in a rat model 89 . Applying these radiographic methods in MRONJ rodent models has generated preclinical evidence in support of their feasibility for diagnosis and treatment.
Serological assessments MRONJ presents disorders in bone turnover, which results in changes in bone formation and resorption products. Detection of these characteristic products in serum can be a promising approach to the prediction and treatment of MRONJ. Although serological assessment is not a general examination in rodent models, attention to serum markers for MRONJ has increased 64,67,98,103 . A decrease in the statistical significance of serum vascular endothelial growth factor (VEGF) was found in an MRONJ group 67,98,103 , which indicated the inhibition of angiogenesis. Serological bone turnover indicators such as uncarboxylated osteocalcin (GluOC) 64 , C-terminal peptide of type I collagen (CTX-1) 98 , tartrate-resistant acid phosphatase 5b (TRAcP-5b) 64 , and N-terminal propeptide of type I procollagen (P1NP) 64 were also detected in the MRONJ assessment. However, these bone turnover indicators are still controversial as biomarkers of MRONJ 109,110 . The significance of bone metabolism markers of MRONJ still needs further preclinical and clinical evidence.

CONCLUSION
Methods for establishing MRONJ rodent models have evolved as the understanding of MRONJ pathogenesis, especially local risk factors, has deepened. Tooth extraction is the most commonly used local risk factor, which assumes a central role in exploring pathogenesis and testing novel interventions. Various approaches to infection induction have also been developed to better mimic MRONJ onset following patients' clinical status. Mechanical stimuli have emerged in model establishment, including implantation and other invasive procedures and sustained stress from hyperocclusion. Modified methods with the extraction of infectious teeth are expected to become superior alternatives to classical methods as they present more obvious MRONJ-like lesions that conform more closely to lesions encountered in clinical practice.
Because there is still a lack of a standard assessment system for MRONJ rodent models, we summarized current techniques for assessing MRONJ-like lesions. The histological assessment is the most effective method, mainly characterized by empty osteocyte lacunae. Meanwhile, gross observation, radiographic assessment, and serum indicators also contribute to the comprehensive MRONJlike lesion examination. Although MRONJ rodent models are gradually becoming more mature and reliable with more comprehensive assessment criteria, establishing models simulating natural MRONJ pathogenesis is still challenging due to the long induction time and onerous induction procedures. Shortening the induction time with an improved success rate is critical for future research on MRONJ rodent models because of its lower cost and higher efficacy. Emerging approaches such as new-found risk factors and distinctive drug combinations have sprung up in MRONJ-related studies, which are expected to improve the establishment of MRONJ rodent models. Researchers still need to continue exploring how to more comprehensively simulate the clinical pathogenesis of MRONJ to make rodent models more reliable for preclinical research.