Fluorescent risedronate analogue 800CW-pRIS improves tooth extraction-associated abnormal wound healing in zoledronate-treated mice.

Background: Bisphosphonate-related osteonecrosis of the jaw (BRONJ) is a rare but serious side effect of nitrogen-containing bisphosphonate drugs (N-BPs) frequently prescribed to reduce skeletal-related events in bone malignancies and osteoporosis. BRONJ is associated with abnormal oral wound healing after dentoalveolar surgery and tooth extraction. We previously found that N-BP chemisorbed to bone mineral hydroxyapatite was dissociated by secondary applied N-BP. This study investigated the effect of the surface equilibrium-based removal of N-BP from jawbone on tooth extraction wound healing of zoledronate (ZOL)-treated mice.
Methods: A pharmacologically inactive N-BP derivative (the 4-pyridyl isomer of risedronate equipped with a near-infrared 800CW fluorescent imaging dye, 800CW-pRIS) was designed and synthesized. 800CW-pRIS was intra-orally injected or topically applied in a deformable nano-scale vesicle formulation (DNV) to the palatal tissue of mice pretreated with ZOL, a potent N-BP. The female C56BL6/J mice were subjected to maxillary molar extraction and oral wound healing was compared for 800CW-pRIS/ZOL, ZOL and untreated control groups.
Results: 800CW-pRIS is confirmed to be inactive in inhibiting prenylation in cultured osteoclasts while retaining high affinity for hydroxyapatite. ZOL-injected mice exhibit delayed tooth extraction wound healing with osteonecrosis relative to the untreated controls. 800CW-pRIS applied topically to the jaw one week before tooth extraction significantly reduces gingival oral barrier inflammation, improves extraction socket bone regeneration, and prevents development of osteonecrosis in ZOL-injected mice. Conclusions: Topical pre-treatment with 800CW-RIS in DNV is a promising approach to prevent the complication of abnormal oral wound healing associated with BRONJ while retaining the anti-resorptive benefit of legacy N-BP in appendicular or vertebrate bones.

Introduction

Nitrogen-containing bisphosphonates (N-BPs) are prototypical antiresorptive agents and effective drugs for treating abnormal osteolysis associated with tumors residing in and metastasizing to bone marrow, such as multiple myeloma and breast cancer, respectively[1-3]. At a lower dose, N-BPs have also been widely used to reduce the risk of osteoporotic bone fractures[4]. N-BPs are synthetic analogues of pyrophosphate, which plays an important role in bone metabolism and cholesterol biosynthesis. Through the mevalonate pathway, isopentenyl pyrophosphate (IPP) is converted to farnesyl pyrophosphate (FPP) by FPP synthase (FPPS). Pharmacologically active N-BPs such as risedronate (RIS) and zoledronate (ZOL) occupy the allylic substrate binding pocket of FPPS, effectively preventing the synthesis of FPP from IPP[5-7]. As a result, osteoclasts that internalize N-BP exhibit poor prenylation of the downstream proteins, leading to abnormal cytoskeleton formation and premature detachment from the bone resorption lacunae[8].

N-BPs have been shown to prevent skeletal-related events (SRE) and decrease bone pain in cancer patients, with few serious side effects[9,10]. However, the FDA adverse event reporting system (FAERS) has posted clinical reports of oral complications collectively known as medication-related osteonecrosis of the jaw (MRONJ)[11]. The American Association of Oral & Maxillofacial Surgeons (AAOMS) defined MRONJ as an unhealed oral wound with exposed jawbone or fistula reaching to the jawbone surface in patients with a history of antiresorptive medications, including N-BPs and humanized anti-receptor activator of nuclear factor-κB ligand (RANKL) monoclonal antibody (Denosumab), as well as of angiogenesis inhibitors[12]. The oral side effect of anti-resorptive agents in the FAERS database peaked from the first quarter of 2010 to the first quarter of 2014 with ~30,000 cases during this period; N-BP comprised the largest segment of reported adverse events[11]. Widespread concern about the rare but often severe oral complications is believed to have contributed to markedly decreased acceptance of N-BPs by osteoporosis patients, despite the well-established benefit of these drugs for this indication[13]. Currently, there is no established and effective preventative treatment or therapy for N-BP-related oral complications.

We have previously reported that N-BP adsorption to bone minerals is not permanent but rather establishes an equilibrium[14,15]. Thus, we hypothesized that the legacy N-BP adsorbed on the jawbone might be displaced and replaced by a pharmacologically inactive N-BP through the equilibrium-based molecular competitive binding, leading to the attenuation of abnormal oral wound healing. The chemical structure of N-BPs acutely determines their antiresorptive potency. Ca2+ ion chelation by the bidentate phosphonate moieties of the BP structure abetted, if present, by an α-hydroxy side chain (Fig. 1a) facilitates a high affinity to bone mineral, which is primarily hydroxyapatite (HAp)[16]. After adsorption to the bone surface, N-BPs are thought to remain quiescent until osteoclastic bone resorption. Steady-state release of adsorbed N-BP drugs from bone is slow, as shown by studies determining their half-lives in the skeleton, which range from months to several years[17]. However, we have shown that N-BPs can be rapidly displaced from synthetic HAp or bone mineral surfaces by the application of competing BPs in solution[14,15]. This led us to explore the idea that displacement of legacy N-BP from the jawbone by local treatment with a pharmacologically inactive N-BP prior to the high-risk dentoalveolar surgical manipulations would abate the development of abnormal wound healing.

Fig. 1

Chemical structure and synthesis of 800CW-pRIS.

a General structure of bisphosphonates. b Chemical structure of risedronate (RIS) and para-pyridyl-risedronate (pRIS). c Chemical structure of 800CW-pRIS. d Synthesis scheme of 800CW-pRIS (anionic counterions are omitted for clarity).

Holtmann et al. reviewed the literature and reported that the majority of animal models for bisphosphonate-related ONJ (BRONJ) were laboratory rodents: rats (52.3%) and mice (34.1%)[18]. Those rodent models from different laboratories follow a similar protocol: N-BP systemic injection and tooth extraction. The complications associated with dentoalveolar surgery and tooth extraction have been reported as a characteristic disease phenotype of BRONJ[19], which was similarly observed in the rodent tooth extraction models[20-22]. Human patients show various confounding factors that may influence the development of MRONJ/BRONJ. Recently, modified rodent models have been developed to evaluate the role of these confounding factors such as autoimmune diseases[23] and periodontitis[24] on the severity of BRONJ lesions. However, the pathological mechanism of BRONJ has not yet been fully established.

The present study hypothesized that the presence of N-BP on the jawbone played an important role in oral wound complications. Here we describe the evaluation of this hypothesis in the tooth extraction wound healing of an N-BP-treated murine model, using as the displacing agent: a novel fluorescent dye-conjugated RIS analogue, 800CW-pRIS, modified to remove its inherent anti-resorptive activity. The newly synthesized 800CW-pRIS lacked pharmacological function while maintaining a strong affinity to hydroxyapatite and displaced pretreated ZOL. In ZOL-pretreated mice, topical application of 800CW-pRIS to oral mucosa attenuated abnormal tooth extraction socket wound healing and jawbone osteonecrosis development.

Methods

Chemical reagents

IRDye® 800CW-carboxylate was purchased from Licor Biosciences. N,N,N,N-tetramethyl-O-(N-succinimidyl)uronium tetrafluoroborate (TSTU) was purchased from Oakwood Chemical. All other reagents were purchased from either Sigma-Aldrich or Alfa Aesar. Allylamine was distilled under N2; CH2Cl2 was distilled from P2O5. All other reagents were used as supplied by the manufacturer. Anhydrous N,N-dimethylformamide (DMF), tetrahydrofuran (THF), dichloromethane (DCM), chlorobenzene, acetonitrile, methanol, ethanol, ether, and diisopropylethylamine (DIPEA) were purchased from EMD Millipore Corporation or VWR. Sodium carbonate, sodium bicarbonate, phosphoric acid, magnesium sulfate, sodium hydroxide, and hydrochloric acid were purchased from Fisher Scientific. ZOL was obtained from UCLA Pharmacy (Reclast®, Novartis, Basel, Switzerland). DOTAP (1,2-dioleoyloxy-3-(trimethylammonium)propane sulfate), DPPC (diphosphatidylcholine), CH (cholesterol) and a nonionic surfactant (Span 80) were acquired from Sigma-Aldrich (St. Louis, MO).

Ethics statement

All experimental protocols using animals were reviewed and approved by the UCLA Animal Research Committee (ARC#1997-136) and followed the PHS policy for Humane Care and use of Laboratory Animals and UCLA Animal Care Use Training Manual guidelines.

Design of 800CW-conjugated para-pyridyl-risedronate (800CW-pRIS)

The position of nitrogen in the pyridine ring of RIS is an important determinant of its antiresorptive activity[25]. When the 3-nitrogen of RIS is relocated to the 4-position (i.e., from meta to para, pRIS; Fig. 1b), antiresorptive activity is reduced while the high affinity of the BP moiety to HAp is retained[26,27]. We further modified pRIS by conjugation to a near-infrared dye, IRDye 800CW (Fig. 1c).

Synthesis of 800CW-pRIS (Fig. 1d)

tert-Butyl allylcarbamate[28] (1)

A solution of di-tert-butyl dicarbonate (2.18 g, 10 mmol) in dry DCM (7 mL) was added dropwise to an ice-cold solution of allylamine (0.57 g, 10 mmol) in dry DCM (3.5 mL). The solution was brought to room temperature and stirred for 4 h. The reaction mixture was then diluted with additional DCM (10 mL) and extracted with 5% citric acid solution, followed by brine. The organic layer was dried over Na2SO4 and concentrated under vacuum to provide 1.4 g (89%) of tert-butyl allylcarbamate (1) as a yellow oil which was used in next step without further purification. 1H NMR (300 MHz, CDCl): δ 5.92−5.75 (m, 1H), 5.24−5.06 (m, 2H), 4.59 (br, 1H), 3.74 (br, 2H), 1.44 (s, 9H).

tert-Butyl (oxiran-2-ylmethyl)carbamate[29] (2)

To an ice-cold solution of 1 (1.4 g, 8.90 mmol) in DCM (50 mL) was gradually added meta-chloroperoxybenzoic acid (mCPBA) (3.07 g, 17.81 mmol). The solution was brought to rt and stirred for 20 h, then diluted with DCM (50 mL) and sequentially extracted with 75 mL each of 10% Na2SO3, saturated NaHCO3 (3x), water, and brine. The DCM layer was dried over MgSO4 and concentrated under vacuum to provide 1.27 g (82%) of 2 as a light-yellow oil which was used in next step without further purification. 1H NMR (300 MHz, CDCl): δ 4.74 (br, 1H), 3.60−3.42 (m, 1H), 3.25−3.13 (m, 1H), 3.12−3.01 (m, 1H), 2.76 (t, J = 4.8 Hz, 1H), 2.53 (dd, J = 4.7, 2.7 Hz, 1H), 1.44 (s, 9H).

(1-Hydroxy-2-(pyridin-4-yl)ethane-1,1-diyl)bis(phosphonic acid)[30] (3)

A 500 ml round bottom flask was charged with 4-pyridylacetic acid hydrochloride (2.60 g, 15 mmol), phosphorous acid (3.68 g, 45 mmol), and chlorobenzene (75 mL). The mixture was heated to 100 °C and phosphorus trichloride (6.18 g, 45 mmol) was added along the sides of the flask using a syringe. A reflux condenser was attached and heating continued at 100 °C for 3 h. A yellow gummy oil formed during the course of the reaction. After cooling to rt and then 0 °C, excess chlorobenzene was decanted and aq. 3 N HCl (125 mL) was added to the residual yellow oil. The flask was refitted with a reflux condenser and heated at 100 °C overnight. The solvent was then evaporated under vacuum, the residue was suspended in warm water (75 mL) and methanol was added until the solution became turbid. After reflux for 1 h, the precipitated product was filtered, washed with EtOH, and dried under high vacuum to provide 2.58 g (58%) of 3 as a white solid. 1H NMR (300 MHz, D2O + Na2CO3): δ 8.31 (d, J = 5.88 Hz, 2H), 7.72 (d, J = 6.12, 2H), 3.26 (t, J = 12.42, 1H). 31P NMR (121 MHz, D2O + Na2CO3): δ 16.01 (s).

1-(3-((tert-Butoxycarbonyl)amino)−2-hydroxypropyl)−4-(2-hydroxy-2,2-diphosphonoethyl)pyridin-1-ium (4)

To (1-hydroxy-2-(pyridin-4-yl)ethane-1,1-diyl)bis(phosphonic acid) (3) (142 mg, 0.5 mmol) was added H2O (2 mL) and the pH of the mixture was adjusted to 6.1 using 1 M NaOH. Next, 2 (0.086 g, 0.5 mmol) was dissolved in a minimum volume of MeOH (200 mL) and added and the reaction mixture was heated at 40 °C. The progress of the reaction was monitored by 31P NMR. After 16 h, conversion to product was 70% complete and an additional 1/3 equivalent of 2 (0.03 mg, 0.16 mmol) was added with continued stirring at 40 °C. This step was repeated several times, until all of the starting material (3) was consumed (62 h). The reaction mixture was filtered and the filtrate was evaporated, leaving a viscous yellow oil. Ether (3 mL) was added, the mixture was sonicated, and the ether was decanted. The residue was then placed under high vacuum to provide 0.269 g (>99%) of 4 as an off-white foam which was used in next step without further purification. 1H NMR (300 MHz, D2O): δ 8.42 (d, J = 6.32 Hz, 2H), 7.92 (d, J = 6.32 Hz, 2H), 4.19 (dd, J = 14.16, 9.55 Hz, 1H), 4.02−3.91 (m, 1H), 3.39−3.08 (m, 4H), 1.29 (s, 9H). 31P NMR (121 MHz, D2O): δ 16.12 (s).

1-(3-Amino-2-hydroxypropyl)−4-(2-hydroxy-2,2-diphosphonoethyl)pyridin-1-ium (5)

Intermediate 4 (0.269 g, 0.51 mmol) was refluxed in 6 N HCl (8 mL) for 6 h. The solvent was evaporated, leaving a light-yellow gum which was sonicated in acetonitrile—MeOH to provide 275 mg (>99%) of the product 5 as an off-white solid which was used in next step without further purification. 1H NMR (300 MHz, D2O): δ 8.50 (d, J = 5.45 Hz, 2H), 7.93 (d, J = 5.81 Hz, 2H), 4.36−4.13 (m, 2H), 3.36 (t, J = 12.10 Hz, 2H), 3.26−3.16 (m, 2H, obscured by MeOH peak), 3.05−2.81 (m, 1H). 31P NMR (121 MHz, D2O): δ 16.44 (s).

Characterization of intermediate compounds 4 and 5 (Fig. S1)

For the synthesis of 4: 0.5 mmol of 3 is converted to 4 with a >99% yield. The stated weight of the obtained product 4 is 0.269 g. The pH at which the reaction is performed is stated to be 6.1. At this pH, MarvinSketch calculations suggest the compound should exist as a trisodium salt. The MW of 4 as a trisodium salt is 522.05 g/mol, which is close to the MW of the compound obtained (0.269 g/0.5 mmol = 538 g/mol). It could be stated that this intermediate was obtained as a trisodium salt. In the following conversion to 5, 0.269 g, 0.51 mmol of starting material (4) is used. If 4 is a trisodium salt, then this number is valid.

The conversion to 5 is an HCl mediated Boc-deprotection. It is reported that 275 mg of compound is obtained in ~99% yield. The MW of 5 as a dichloride salt is 428.01 g/mol. If we obtained 0.51 mmol of 5 as a dichloride salt, then this would result in a mass of 218.3 mg. Therefore, we believe that this compound contains NaCl impurities. Two NaCl per product molecule would result in a pseudo MW of 543.92 g/mol which is close to the calculated MW of the compound at 539.2 g/mol. The presence of NaCl impurities should be properly assessed. As such, we have estimated that 4 is a trisodium salt and also contains one H2O, while 5 is a dichloride salt with some sodium chloride impurities.

800CW-pRIS

To an ice-cold mixture of IRDye® 800CW-carboxylate (6) (Licor Biosciences, Lincoln, NE) (0.05 g, 0.045 mmol) and DIPEA (0.071 mL, 0.41 mmol) in dry DMF (2 mL), kept under N2 and in the dark, was added a solution of TSTU (0.02 g, 0.068 mmol) in DMF (200 mL). After 3 h, TLC (DCM/MeOH, 3:2) showed disappearance of the 6 spot and formation of the 800CW-NHS ester (7). The solvent was removed under vacuum and the residue was dissolved in DMF (750 mL), then added to a solution of 5 (0.11 g, 0.22 mmol) in H2O (2.5 mL) which was adjusted to pH 8.33 using Na2CO3 while constantly maintained in the dark at rt. The addition of the 800CW-NHS ester decreased the pH of the reaction mixture to 7.46. The pH was readjusted to 8.05 using Na2CO3 and the reaction mixture was stirred at rt overnight. HPLC analysis of the reaction mixture after 18 h confirmed formation of the desired product (800CW-pRIS), which was isolated by prep. RP-HPLC using a Phenomenex Luna® 5 µm C18(2) 100 Å, 250 × 21.2 mm, AXIA™ Packed LC Column using A = 0.1 M TEAAc/20%MeOH (pH = 5.0− 5.3) and B = 0.1 M TEAAc/70% MeOH (pH = 5.0−5.3). HPLC Method: 0% B (0–7 min), 0–100% B (7–25 min), 100% B (25–30 min). 800CW-pRIS eluted at ~21 min and was obtained as a triethylammonium salt in 46% yield (free acid basis), product yields were determined by absorbance at 775 nm (PBS buffer, pH 7.4, ε = 242,000 M−1cm−1)[31]. 1H NMR (500 MHz, D2O): δ 8.46 (d, J = 6.86 Hz, 2H), 7.93 (d, J = 6.82 Hz, 2H), 7.69 (d, J = 9.01 Hz, 2H), 7.67−7.56 (m, 6H), 7.12−7.04 (m, 4H), 5.96 (t, J = 14.03 Hz, 2H), 4.59−4.54 (m, 1H), 4.18 (dd, J = 14.21, 9.86 Hz, 1H), 4.00−3.95 (m, 1H), 3.83 (dt, J = 24.61, 7.27 Hz, 3H), 3.35 (t, J = 12.28 Hz, 2H), 3.23−3.14 (m, 2H), 2.91 (q, J = 14.54, 7.11 Hz, 1H), 2.78 (t, J = 7.43 Hz, 2H), 2.50 (br t, J = 6.15 Hz, 3H), 2.09 (t, J = 6.94 Hz, 2H), 1.80 (s, 21H), 1.75−1.63 (m, 4H), 1.59−1.51 (m, 2H), 1.47−1.39 (2H). 31P NMR (121 MHz, D2O): δ 15.82 (s). HRMS (ESI-TOF) m/z calcd for C56H70N4O22P2S4 [M-2H]−2 669.1342; found: 669.1351. The characterization of 800CW-pRIS was presented in Fig. S2 (HPLC purity); Fig. S3 (HRMS analysis); Fig. S4 (1H NMR spectrum); Fig. S5 (31P NMR spectrum); Fig. S6 (UV-VIS absorption and fluorescent emission spectra).

Displacement of 5-FAM-ZOL by 800CW-pRIS in vitro

Synthetic apatite (carbonate apatite) coated cell culture wells (Bone resorption assay plate 24, Cosmo Bio Co. Ltd, Tokyo, Japan) were incubated with fluorescent-tagged ZOL (10 µM 5-FAM-ZOL in 500 µl MilliQ-treated water: MQW) overnight at 37 °C, 2% CO2. After three washes with Milli-Q treated pure water (MQW) for 10 min each, the 5-FAM-ZOL coated wells were then incubated with 10 µM 800CW-pRIS in 500 µl MQW for 2 h at 37 °C, 2% CO2 followed by three washes with MQW. One group of wells (n = 3) were further incubated by the second application of 10 µM 800CW-pRIS in 500 µl MQW followed by 3 washes using 500 µl MQW for 10 min each. The control wells were incubated with MQW. The fluorescent signal of each well (apatite discs) and recovered solutions after each treatment (either incubating with 5-FAM-ZOL, wash or challenged by 800CW-pRIS) were evaluated (IVIS Lumina II, PerkinElmer, Waltham, MA) using the preset conditions for GFP and 800CW. The experiment was triplicated, and Student’s t test was used to compare the 5-FAM-ZOL treated group to each of other groups.

In silico modeling

In silico docking experiments were performed using the AutoDock Vina software, and ligands were prepared using AutoDockTools 1.5.6[32]. The crystal structure of human FPPS (PDB 1YV5) was used. To verify the validity of our docking method, we compared the RMSD between the docked RIS ligand with the crystallographic RIS pose. We calculated a RMSD of 1.67 Å which is below the commonly used RMSD ≤ 2 cutoff[33]. To assess the pharmacological efficacy of RIS and its analogues we compared their calculated binding energies and to investigate the reasons for the observed differences in binding energies we determined the hydrogen bonding distance and angle between the protonated pyridyl protonated N hydrogen and the T201 oxygen and  the K200 carbonyl oxygen. To force bulky 800CW-pRIS into the relevant FPPS active site, we built the molecule from within the site using ICM-Pro 3.2.

In vitro osteoclast resorption pit assay

Synthetic apatite (carbonate apatite) coated cell culture wells were incubated with 10 µM ZOL in 500 µl MQW or 10 µM 800CW-pRIS in 500 µl MQW for overnight at 37 °C, 2% CO2 followed by extensive washes using 500 µl MQW. Some ZOL-preincubated wells were further incubated with 10 µM 800CW-pRIS in 500 µl MQW once or twice. Control wells were treated with 500 µl MQW. To all wells, RAW274.1 cells (2.5 × 104 cells per well) were inoculated in culture medium supplemented by mouse recombinant receptor activator of nuclear kappa-B ligand (RANKL; Sigma-Aldrich) (100 ng/ml) and incubated at 37 °C, 2% CO2. The culture medium was changed after 3 days. After 6 days of incubation the cells were removed with 0.25% Trypsin and the resorption pit generated in the synthetic apatite was photographed and measured using a Java-based image processing program (ImageJ, NIH, Bethesda, MD). This experiment was performed in triplicated wells per group.

Tooth extraction of zoledronate (ZOL)-pretreated mice

The ZOL-pretreated mouse model has been previously reported in the literature[21,22]. Female C57Bl6/J(B6) mice with an age of eight to ten weeks were purchased from Jackson Laboratory (Bar Harbor, ME) and maintained in the vivarium of the Division of Laboratory Animal Medicine at UCLA with free access to food and water. The mice were lightly anesthetized by isoflurane and ZOL (500 µg/Kg in medical grade saline: 400 µM, 100 µl) was administered as a bolus IV injection through the retro-orbital venous plexus. Seven days after ZOL injection, the maxillary left first molar was extracted using a dental explorer and analgesics (2 mg/Kg Carprofen) was provided every 12 h for 2 days[21,22]. The only mice that tooth extracted without remaining root were included in the study. To avoid food impaction, the mice were fed a gel diet (DietGel® Recovery, ClearH2O, Westbrook, ME) for the first 7 days. The mice were euthanized after 14 days of tooth extraction by 100% CO2 inhalation and the maxillae were harvested.

800CW-pRIS administration by local intra-oral injection

The intra-oral injection of 800CW-pRIS (100 µM, 2 µl) to mouse palatal tissue using Hamilton syringe with a 33-gauge needle under a surgical microscope[14] was first examined for 800CW fluorescent signal 3 days after the injection by the in vivo imaging system with 740 nm emission (IVIS Lumina III, Perkin Elmer, Waltham, MA).

We have previously used the mouse model described above to examine the effect of intra-oral injection of non-nitrogen containing BP for the attenuation of abnormal or delayed tooth extraction wound healing of ZOL-treated mice[14], based on which we determined the number of mice per group to be n = 8 by the preliminary results. The primary outcome measure was not used to determine the sample size.

Four days after the ZOL IV injection, ZOL-pretreated mice were randomly divided into two groups. One group received an intra-oral injection of 2 µl of 100 µM 800CW-pRIS and the other group received an intra-oral injection of 2 µl of 0.9% saline as a negative control (n = 8 per group, n = 16 in total). After the mice were anesthetized by isoflurane, intra-oral injection to the palatal gingiva adjacent to the left first molar was performed. The maxillary left first molar was extracted 3 days after the 800CW-pRIS intra-oral injection and the mice were euthanized after 14 days of tooth extraction by 100% CO2 inhalation. The maxilla containing the tooth extraction wound was harvested and photographed. The maxilla tissue was then fixed in 10% buffered formalin and subjected to micro-CT imaging followed by EDTA-decalcification and conventional paraffin-embedded histological section preparation for outcome assessments (see below).

800CW-pRIS-DNV synthesis

800CW-pRIS-DNV were synthesized using established protocols[34]. Briefly, a lipid mixture containing diphosphatidylcholine (DPC), cholesterol (CH), 1,2-dioleoyloxy-3-(trimethylammonium)propane-sulfate (DOTAP) for the positive surface charge or 1,2-dipalmitoyl-sn-glycero-3-phosphocholine (DMPC) for the negative surface charge, was dissolved in isopropyl alcohol and Span80 (15% v/v) was added. 800CW-pRIS was dissolved in pure water. The aqueous and organic solutions were injected into the microfluidic reactor and the positive charge-DNV and negative charge-DNV were synthesized.

Characterization of 800CW-pRIS-DNV

After the synthesis of 800CW-pRIS-DNV, the size was assessed by dynamic light scattering and the surface charge was assessed by zeta potential measurements using a Zeta sizer instrument (Nano-ZS, Malvern, Worcestershire, UK). Then, samples were dialyzed overnight to remove encapsulated 800CW-pRIS. To calculate encapsulation efficiency of 800CW-pRIS, DNVs were disintegrated by ice-cold acetonitrile followed by water bath sonication. The solution was centrifuged (15 min, 20,000 × g) and the supernatant was collected. The absorption of 800CW was determined at 770 nm (SpectraMax M5, Molecular Devices, San Jose, CA). The encapsulation efficiency (EE) was calculated:

Trans-gingival epithelium administration test

To assess the efficacy of trans-gingival administration of 800CW-pRIS-DNV, we used a commercially available cultured gingival tissue model (EpiGingival™, MatTek Corporation, Ashland, MA). This in vitro model consists of human oral epithelial cells cultured on cell culture insert. The cell insert was filled with 300 µl of assay medium. Negatively or positively charged 800CW-pRIS-DNV powder was suspended in MiliQ water to make a 25 µM solution. 100 µl of such 800CW-pRIS-DNV solution was added on the oral epithelial cells in the cell insert and incubated at 37 °C in a 5% CO2 humidified incubator, according to the manufacture’s protocol. As a control without DNV, we used 25 µM 800CW-pRIS in MiliQ water. The assay medium was changed every 30 min and measured the absorbance in a spectrophotometer at 774 nm with a micro plate reader (SYHNERGY H1 plate reader). The optical density (OD) measurement per well was described as the cumulative OD data, in which the OD measurement was sequentially added to the OD measurement of the previous time point. From this study, positively charged 800CW-pRIS-DNV was selected for the following in vivo studies.

800CW-pRIS-DNV topical application to mouse palatal gingiva

For topical application to mouse gingival tissue, 800CW-pRIS-DNV powder was suspended in MiliQ water to make 100 µM solution. To protect the topical treatment site, a custom-made oral appliance was used to cover the whole palatal tissue for 1 h and then the oral appliance was removed. The delivery of 800CW-pRIS was confirmed by in vivo fluorescent imaging as described above 3 days after the application of DNV topical formulation. Specifically, 20 h after ZOL-IV injection, the mice were randomly divided into three groups (n = 8 per group, n = 24 in total) for topical treatment with 800CW-pRIS-DNV solution once or twice a week, as well as topical treatment with MiliQ water as a negative control. The mice were anesthetized by isoflurane inhalation and 3 µl of 100 µM 800CW-pRIS-DNV was applied on the palatal gingival tissue for each time application in the 800CW-pRIS-DNV treated groups. After the topical application of 800CW-pRIS-DNV or MiliQ water, the maxillary left first molar was extracted for each mouse as above. Fourteen days after the tooth extraction, mice were euthanized and the maxillary tissues were harvested for intra-oral photographs, micro-CT imaging and histological preparation for the assessments as described below.

Outcome assessments of tooth extraction wound healing

In this study, all mice received a ZOL IV injection followed by tooth extraction. The experimental groups were further treated with 800CW-pRIS by intra-oral injection or as a formulation in DNV applied topically. The outcome for each experimental group was compared to the appropriate control groups.

Gingival swelling (Fig. S7)

Using the intra-oral photographs, the swelling area of oral mucosa around the tooth extraction socket was measured by ImageJ (NIH, Bethesda, MD) and normalized by normal mucosa surrounding the maxillary right first molar.

Micro CT analysis (Fig. S8)

The extracted maxilla was fixed in neutral buffered formalin. The fixed maxilla was scanned with micro-CT machine (SkyScan 1172 scanner, Skyscan, Kontich, Belgium) at source voltage of 60 kV and source current of 166 µA with the scanning resolution of 10 µm per pixel. New bone formation in the socket was analyzed by CTAn version 1.11 using the following protocol:

Step 1: A horizontal image of micro-CT containing the extraction sockets and the corresponding contralateral remaining mesial root, palatal root and distal root were identified.

Step 2: The cross-sectional area of the remaining mesial root, palatal root and distal root was determined.

Step 3: The mirror image of the cross-sectional area of remaining roots was transferred to the tooth extraction side, identifying the “original” tooth extraction socket area.

Step 4: A total of 35 layers covering from the apex of the identified tooth extraction cross-sectional area were stacked up. The bone volume over tissue volume (BV/TV) in each extraction socket was determined and combined.

Histological osteonecrosis analysis

After scanning the maxilla with micro-CT, the maxillary samples were fixed with 70% ethanol and decalcified with 10% EDTA in tris buffer. The conventional paraffin-embedded cross-sections were made through the medial root (mesial section) or the palatal/distal root (distal section) of the remaining right first molar and stained with hematoxylin and eosin. The sections containing palatal bone of the middle and distal area of tooth extraction site were used for measuring the osteonecrosis area. The osteonecrotic bone area, which was defined as the presence of a cluster of four or more empty osteocytic lacunae, was measured by ImageJ (NIH, Bethesda, MD) and standardized by the occlusal half cross-section area of the palatal bone.

Histological inflammation index analysis (Fig. S9)

All sections with H-E staining were reviewed and assessed with visual inflammatory index system (0 = normal, 1 = mild, 2 = moderate, 3 = severe) by three blinded investigators.

Inflammation Index 0: No inflammatory cell infiltration in the palatal gingiva

Inflammation Index 1: Mild inflammatory cell infiltration, localized on the maxillary bone surface.

Inflammation Index 2: Moderate inflammatory cell infiltration occupying <50% area of the palatal gingiva

Inflammation Index 3: Severe inflammatory cell infiltration occupying more than 50% area of the palatal gingiva.

Immunohistochemistry of Th17 and Th1 cells in the BRONJ lesion

EDTA-decalcified paraffin-embedded histological sections from the control mice and 800CW-pRIS-DNV treated mice were used for immunohistochemistry. The histological cross-sections were first subjected to the antigen retrieval process using microwave irradiation. The sections were incubated with rabbit recombinant anti-RORγt antibody or rabbit recombinant anti-T-bet antibody (EPR20006 and EPR9302, respectively, Abcam, Waltham, MA) at 1/50 dilution followed by anti-rabbit secondary antibody incubation and diaminobenzidine substrate treatment. The sections were counter stained with hematoxylin.

800CW fluorescent signal in mouse femurs

Two days after the intra-oral injection of 100 µM 800CW-pRIS in 2 µl saline solution or after topical application of 100 µM 800CW-pRIS-DNV in 3 µl MQW, mouse femurs were harvested and the 800CW fluorescent signal was examined by the in vivo imaging system. The region of interest was selected at the femur head (n = 3 per group). For the imaging control, one mouse received 100 µM 800CW-pRIS IV in 100 µl saline solution via the retro-orbital plexus.

Statistics and reproducibility

For the statistical analysis, Student’s t test or one-way analysis of variance (ANOVA) with Dunnett post hoc test was used for comparisons of in vitro and in vivo data. A significant difference was defined by p < 0.05. Figures of numerical data contain the raw measurements, means and standard deviations. The data were showed as mean and standard error of the mean.