Osteoporotic TgRANKL mice (Tg5519 line) (17, 18), and WT control mice of C57BL/6J background were maintained and bred in the animal facility of Biomedical Sciences Research Center “Alexander Fleming” under specific pathogen-free conditions. All animal work was approved by the Institutional Protocol Evaluation Committee and was licensed by the Veterinary Authorities of Attica Prefecture (registered code: 203222/11-03-2020) in compliance with the PD 56/2013 and the European Directive 2010/63/EU.

A schematic of the study design is shown in Figure 1. 6-week-old WT and TgRANKL mice with equal average body weights were divided into 6 treatment groups. Group I: WT vehicle group injected subcutaneously (s.c.) with phosphate-buffer saline (PBS) twice per week for 18 weeks (n=10, both sexes), Group II: TgRANKL vehicle group injected s.c. with PBS twice per week for 18 weeks (n=10, both sexes), Group III: TgRANKL+Dmab group injected s.c. with 3 mg/kg denosumab (Prolia^®) once per week for 18 weeks (n=10, both sexes), Group IV: TgRANKL+Dmab/Discontinuation group injected s.c. with 3 mg/kg denosumab once per week for 6 weeks followed by PBS injections s.c. for 12 weeks (n=11, both sexes), Group V: TgRANKL+Dmab/ZOL group injected s.c. with 3 mg/kg denosumab once per week for 6 weeks followed by intraperitoneal (i.p.) injections of 150 μg/kg zoledronate (ZOL) (Aclasta^®) twice per week for 12 weeks (n=7, both sexes) and Group VI: TgRANKL+Dmab/ZOL/Discontinuation group injected s.c with 3 mg/kg denosumab once per week for 6 weeks then switch to i.p. injections of 150 μg/kg zoledronate twice per week for 12 weeks and cessation of zoledronate for another 18 weeks (n=5, 3 males and 2 females). Body weight was recorded weekly for each mouse. All mice were euthanized 12 weeks after denosumab discontinuation except the mice from Group VI experimental group which were euthanized 18 weeks after ZOL cessation. Upon sacrifice, both femurs were harvested from all animals. Right femurs underwent initial ex vivo microCT scanning, followed by either random allocation for osmium tetroxide staining and bone marrow adipose tissue (BMAT) quantification (n=5-6) or decalcification and embedding for histological evaluation (n=5) with balanced sex representation. Left femurs were snap-frozen and stored at -80°C for RNA extraction and gene expression analysis (n=4/group, sex-balanced) or reserved for additional assessments. In addition, blood was collected to measure serum Ctx, P1NP and free calcium. Mice were not fasted prior to blood collection. Euthanasia was performed with CO₂ with a 30% per minute displacement of the chamber volume, followed by cervical dislocation (Directive 2010/63/EU).

Planar X-ray images of the animals were acquired using the In vivo Xtreme imaging system (Bruker). Mice were anesthetized with isoflurane in an induction chamber ( 2 % isoflurane in O2 at 5 L/min), transferred to the In-Vivo Xtreme imaging cabinet, and maintained under anesthesia during acquisition ( 2% isoflurane in O2 at 1 L/min), with dexpanthenol ophthalmic gel applied for corneal protection. The Bone Density Software (Bruker Molecular Imaging Software) was used to measure the bone density (BD) (g/cm³) in order to monitor disease development. Bone density software Module utilizes low-dose “soft” X-rays to accurately measure bone density at the midshaft region of the distal femur.

Femurs were fixed in 10% formalin (Carlo Erba) for 16 hours, decalcified in 13% EDTA, and embedded in paraffin. Sections of 5‐μm thickness were stained with hematoxylin/eosin. Osteoclasts were visualized by staining for TRAP activity using a leukocyte acid phosphatase kit (Sigma‐Aldrich). TRAP staining was quantified as total osteoclast number in total bone surface and as an osteoclast surface fraction (percentage of osteoclast surface in total bone surface, Oc.S/BS, %). Analysis was focused on the endocortical bone surfaces at the metaphyseal region of distal femur and was performed using the open source software for bone histomorphometry “TrapHisto” (19).

Femurs and lumbar spine were fixed in 10% formalin overnight at 4°C and then washed and stored in PBS. Microarchitecture of the distal femurs and lumbar spine (L5) from all experimental mice was evaluated using a high-resolution SkyScan1172 microtomographic (microCT) imaging system (Bruker). Images were acquired at 50 KeV, 100 µA with a 0.5 mm aluminum filter. Three-dimensional reconstructions (8.8 mm cubic resolution) were generated using NRecon software (Version 1.7.4.2, Bruker) as previously described (17, 18). The trabecular area of the distal femur and lumbar spine was assessed through bone volume fraction (BV/TV, %), bone volume (BV, mm³), trabecular number (Tb.N, mm^-1), trabecular separation (Tb.S, mm), connectivity density (Conn.Dn, mm^-3) and bone mineral density (BMD, g/cm³). Femoral trabecular geometry was assessed using 300 continuous CT slides (1800 µm) located at the trabecular area 50 slides underneath the growth plate and the lumbar spine’s trabecular geometry was assessed using 250 continuous CT slides (1487 μm). Femoral cortical geometry was assessed using 100 continuous CT slides (600 µm) located at the femoral midshaft as previously described (17, 20). Cortical bone was assessed through cortical bone volume fraction (Ct.BV/TV, %), cortical thickness (Ct.Th, mm), bone marrow volume (BMV, mm³), total cortical porosity (Po(tot), %) and tissue mineral density (TMD, g/cm³).

Femurs were stained with osmium tetroxide for analysis of BMAT, as previously reported (21, 22) with optimized modifications. Briefly, upon microCT scanning of bone samples, femurs were decalcified in 13% EDTA, pH 7.4, for 14 days. After washing with tap water for 2 hours, 500 μl 1% osmium tetroxide solution (Electron Microscopy Services, Hatfield, PA) was added to each femur placed in a 1.5-mL microtube. Bones were stained in the fume hood for 5 days at room temperature with two intermediate changes of freshly made osmium tetroxide solution. Osmium solution was carefully removed and decontaminated into a small liquid waste container that had been filled with corn oil. Bones were washed under tap water for 3 h at room temperature, transferred to a new 1.5 mL microtube containing 1 mL PBS buffer and proceeded for microCT scanning.

Total RNA was extracted from the whole femur tissue (no flushing) using a monophasic solution of guanidine isothiocyanate and phenol according to the manufacturer’s instructions (TRI Reagent, MRC, Cincinnati, OH, USA). After removal of DNA remnants with DNase I treatment (Sigma‐Aldrich, St. Louis, MO, USA), first‐strand cDNA was synthesized using 2 μg of total RNA and M‐MLV (Invitrogen). Templates were amplified with HOT FIREPol^® EvaGreen Master Mix (Solis Biodyne) on the CFX96 Connect real time PCR instrument (Bio-Rad Laboratories). Data analysis was performed following the 2^-ΔΔCT method. The primers were purchased from Eurofins Genomics and are listed in Supplementary Table 1. For each experiment at least three biological and two technical replicates were used.

The following serum parameters were measured in stored frozen (-80°C) serum samples while samples of individual mice were assayed in the same run: total calcium and albumin (Dimension Intergrated Chemistry System Analyzer, Siemens Healthcare Diagnostics Inc., Newark (DE),U.S.A.), [Calcium intraassay coefficient of variation (CV) ≤3.0%, interassay CV ≤4.5%], [Albumin intraassay coefficient of variation (CV) ≤ 1.6%, interassay CV ≤ 2.4%]; N-terminal propeptide of type I procollagen (PINP) [Rat/Mouse PINP EIA, ELISA, Immunodiagnostics Systems Inc., Boldon, UK], [intraassay coefficient of variation (CV) ≤11.8%, interassay CV ≤11.3%]; and C-terminal telopeptides of type I collagen (CTX-I) [Rat-Laps (CTX-I) EIA, ELISA, Immunodiagnostics Systems Inc., Boldon, UK], [intraassay coefficient of variation (CV) ≤13.0%, interassay CV ≤14.0%].

All results are represented as box-plots showing each data point, the median and the interquartile range. Bone histomorphometry data are present as mean values ± standard deviation (SD). Statistical significance was assessed by one-way analysis of variance (ANOVA) and Tukey post-hoc test was performed to compare means of multiple groups. Two-way ANOVA and Tukey post-hoc was performed to compare means of BD values acquiring from imaging analysis among multiple groups. For datasets with small sizes (n=4-5) the assumption of normality was evaluated using the Shapiro–Wilk test and no significant deviations from normality were detected (all p>0.05). Therefore, parametric ANOVA with Tukey’s post-hoc test was applied to these datasets. For all tests, p<0.05 was considered statistically significant, *p< 0.05, **p< 0.01, ***p< 0.001, ****p< 0.0001.

We investigated the development of rebound bone loss following denosumab discontinuation in the osteoporosis mouse model TgRANKL, characterized by progressive trabecular bone loss, cortical porosity and BMAT formation (17). Treatment with denosumab initiated at 6-week-old TgRANKL mice, that already developed bone loss in order to validate its therapeutic benefits. Thus, we implemented a therapeutic protocol scheme (Figure 1A), including experimental group I: WT-vehicle mice, II: TgRANKL-vehicle mice, III: TgRANKL mice continuously treated with denosumab for 18 weeks (TgRANKL+Dmab), IV: TgRANKL mice treated with denosumab for 6 weeks and then denosumab discontinuation for 12 weeks (TgRANKL+Dmab/Discontinuation). To track the rebound bone loss phenomenon after denosumab discontinuation, we performed in vivo X-ray imaging every 6 weeks in all mice for monitoring the progression of bone density in distal femurs. Treatment was well tolerated by all experimental animal groups during the intervention period. We did not observe any acute toxicities nor any significant alterations in body weights among the treated groups during the experimental study (Supplementary Figures 1, 2).

X-ray in vivo imaging verified a 23% lower femoral bone density on average in 6-week-old TgRANKL mice compared to WT littermates in both sexes (Figure 1B, Supplementary Table 2). However, a 6 week-treatment period with denosumab was sufficient to fully rescue the osteoporotic phenotype of TgRANKL mice as shown by a 28% increase in femoral bone density, reaching the baseline levels of WT-vehicle mice (mean bone density (BD); WT: 2.11, TgRANKL-vehicle:1.52 vs TgRANKL+Dmab: 2.08, p<0.05) (Figure 1B, Supplementary Table 2). As expected based on the known denosumab pharmacokinetics, discontinuation after 6 weeks of treatment in TgRANKL mice, resulted in a continued rise in bone density for the following 6 weeks, followed by a decline to TgRANKL-vehicle levels over a subsequent 6-week period by 24 weeks of age (mean BD; female: 40%, male: 30%). However, continuous treatment of TgRANKL mice with denosumab over an 18 week-period between week 6 to 24 progressively increased bone density in both sexes. These findings indicate that discontinuation of denosumab for 12 weeks led to a pronounced osteoporotic phenotype, comparable to that observed in the TgRANKL-vehicle group, (mean BD-female; TgRANKL-vehicle: 1.21 vs TgRANKL+Dmab/Discontinuation: 1.17, p>0.05) closely mirroring the rebound effect typically observed following denosumab cessation in osteoporotic patients (Figure 1B).

The femoral bone structure was further examined with histological analysis both at the trabecular and cortical areas. As expected, 24-week-old mice of the TgRANKL-vehicle group displayed a severely osteoporotic phenotype as shown by absence of trabecular bone, cortical porosity, increased numbers of osteoclasts and BMAT expansion, that was rescued by continuous denosumab treatment (Figure 2). However, denosumab discontinuation led to a rapid trabecular bone loss, cortical porosity and increased osteoclast number (Figures 2A, B), resulting in a phenotype reminiscent of the TgRANKL-vehicle mice, confirming the rebound effect of denosumab discontinuation (Figure 2).

High-resolution imaging of femoral bone microarchitecture with quantitative microCT analysis confirmed that continuous denosumab treatment rescued trabecular bone loss displayed in TgRANKL-vehicle mice while denosumab discontinuation led to a rapid trabecular bone loss over 93% compared to TgRANKL+Dmab group and recurrence of the osteoporotic phenotype (Figure 3, Supplementary Table 3). Similarly, microCT analysis of cortical bone in mid-diaphysis of femurs revealed significant cortical bone loss and increased cortical porosity in TgRANKL-vehicle mice compared to WT controls, that was rescued with denosumab treatment (Figure 4, Supplementary Tables 3, S4). Denosumab discontinuation resulted in rebound cortical loss and porosity similarly to TgRANKL-vehicle group, as shown by the dramatic decrease of the cortical bone fraction (44%), the cortical thickness (62%) and the bone mineral density (65%) (Figure 4C). In addition to the femurs, rebound bone resorption was also observed in the lumbar vertebrae following denosumab discontinuation (Supplementary Figure 3, Supplementary Table 5).

Our previous work showed that TgRANKL mice of both sexes exhibit a marked expansion of BMAT compared to WT littermates, which is prevented by denosumab treatment (17). Therefore, we examined whether discontinuation of denosumab influences BMAT accumulation by performing histological analysis and microCT quantification using osmium tetroxide staining. Our analysis demonstrated that continuous denosumab treatment suppressed BMAT expansion in the distal femurs, reducing the adipocyte volume fraction by over 93% compared to the TgRANKL-vehicle group (Figure 5). Interestingly, the denosumab discontinuation group showed a moderate rebound in BMAT formation (11-fold on average), which was less pronounced than in the TgRANKL-vehicle group (71-fold on average) relative to the TgRANKL+Dmab group (Figure 5, Supplementary Table 6).

Furthermore, we investigated the efficacy of sequential therapy with zoledronate following treatment with denosumab to sustain the anti-resorptive effects of RANKL inhibition. Thus, in the present study we included group V: TgRANKL mice treated with denosumab for 6 weeks and then switched to zoledronate treatment for 12 weeks (TgRANKL+Dmab/ZOL) (Figure 1A). Imaging analysis revealed a progressive increase of bone density during the treatment period at TgRANKL+Dmab/ZOL, similarly to continuous denosumab treatment (mean BD; TgRANKL+Dmab: 2.45 vs TgRANKL+Dmab/ZOL: 2.68, p>0.05) (Figure 1B, Supplementary Table 2). Histological analysis at the 24-week endpoint demonstrated that zoledronate treatment following denosumab therapy effectively sustained the protective effects of denosumab, by reducing osteoclast numbers and preserving both trabecular and cortical bone structures (Figure 2). MicroCT analysis confirmed that switching to zoledronate treatment sustained the effects of denosumab in the trabecular bone leading to higher bone volume fraction and BMD compared to WT, TgRANKL and TgRANKL+Dmab/Discont group, displaying similar values with TgRANKL+Dmab group mice (BV/TV (%); TgRANKL+Dmab: 22.63, TgRANKL+Dmab/ZOL: 24.72, p>0.05) (Figure 3, Supplementary Table 3). Moreover, sequential zoledronate treatment fully rescued the cortical bone structure displaying a similar bone profile with the continuous denosumab treatment as shown by 3D microCT images and cortical bone parameters (Figure 4, Supplementary Tables 3, S4).

In our experimental design, all mice were sacrificed at week 24, except for a small subset from the Dmab/ZOL group (n=5), which were retained to monitor bone turnover following zoledronate discontinuation (TgRANKL+Dmab/ZOL/Discont group). Imaging showed that mice in the zoledronate discontinuation group preserved the peak bone density attained with Dmab/ZOL treatment for a duration of 12 weeks, while a sharp decline in bone density occurred over a 6-week period at 42 weeks of age (BD reduction; female: 13%, male: 29%) (Figure 1B, Supplementary Table 2). Histological evaluation of TgRANKL+Dmab/ZOL/Discont mice at week 42 demonstrated a pronounced loss of trabecular bone, increased cortical porosity, and elevated numbers of osteoclasts (Figures 2A–C). MicroCT analysis confirmed the bone loss developed in the TgRANKL+Dmab/ZOL/Discont group (Figures 3A–C).

The gene expression profile regulating bone remodeling was analyzed in femurs from all experimental groups, with or without therapeutic intervention, using qPCR. This included assessment of RANKL/RANK/OPG signaling components, osteoclast markers (Dcstamp and Ctsk), and osteoblast markers (Runx2 and Alp). Furthermore, blood serum levels of bone turnover markers (BTMs), CTx and P1NP as well as calcium were measured at the endpoint (Figure 6).

Continuous denosumab treatment had the most dramatic reduction in all above markers compared to TgRANKL-vehicle group, indicating profound suppression of bone turnover to baseline levels. Discontinuation led to the loss of denosumab’s beneficial effects, as indicated by bone turnover markers rising to levels observed in the TgRANKL-vehicle group. Switch to zoledronate led to a moderate reduction in the expression of genes associated with bone remodeling compared to the TgRANKL-vehicle group, while serum bone markers decreased to levels comparable to those observed in the TgRANKL+Dmab group. Zoledronate discontinuation produced gene expression patterns similar to those seen in the TgRANKL/Dmab/Zol group (Figures 6A–D). However, serum levels of CTx and P1NP were elevated compared to this group, indicating increased bone turnover. At last, the circulating levels of free calcium in blood serum remained unchanged among experimental intervention schemes and vehicle treated groups, respectively (Figure 6E).