Effect of rapamycin on bone mass and strength in the α2(I)‐G610C mouse model of osteogenesis imperfecta

Abstract Osteogenesis imperfecta (OI) is commonly caused by heterozygous type I collagen structural mutations that disturb triple helix folding and integrity. This mutant‐containing misfolded collagen accumulates in the endoplasmic reticulum (ER) and induces a form of ER stress associated with negative effects on osteoblast differentiation and maturation. Therapeutic induction of autophagy to degrade the mutant collagens could therefore be useful in ameliorating the ER stress and deleterious downstream consequences. To test this, we treated a mouse model of mild to moderate OI (α2(I) G610C) with dietary rapamycin from 3 to 8 weeks of age and effects on bone mass and mechanical properties were determined. OI bone mass and mechanics were, as previously reported, compromised compared to WT. While rapamycin treatment improved the trabecular parameters of WT and OI bones, the biomechanical deficits of OI bones were not rescued. Importantly, we show that rapamycin treatment suppressed the longitudinal and transverse growth of OI, but not WT, long bones. Our work demonstrates that dietary rapamycin offers no clinical benefit in this OI model and furthermore, the impact of rapamycin on OI bone growth could exacerbate the clinical consequences during periods of active bone growth in patients with OI caused by collagen misfolding mutations.

mediated mRNA decay and haploinsufficiency, result in a milder clinical phenotype (OI type I). Mutations that introduce structural missense mutations into the collagen α-chains result in more severe OI phenotypes (OI type II, III, IV). Of the many hundred missense mutations described, the most common mutations are glycine substitutions in triple helical domains of the α1(I) and α2(I) chains (>80%). 1,2 These substitutions interrupt the obligatory Gly-X-Y repeat sequence of the collagen helix, causing misfolding and a structurally abnormal helix.
The molecular mechanism of how these structural mutations cause the bone phenotype has long been thought to be because of a dominant negative effect of incorporating mutant collagen into the collagen trimers. This can affect collagen folding and reduce secretion, with even small amounts of secreted mutant-containing trimers adversely affecting collagen fibril assembly, stability, and crucial collagen-ECM interactions. 2,3 While the central tenets of this model are correct, recent studies on collagen I mutations in OI and collagen II and collagen X mutations in other skeletal dysplasias have added complexity to this model. Notably, the endoplasmic reticulum (ER) stress response to the misfolded collagens has been identified as a significant component of the disease pathology. 3,4 In the case of collagen I helix Gly substitutions, studies have shown that the mutations destabilize and delay helix formation and cause increased posttranslational modification of lysine and intracellular retention. 5 However, the mutant collagens do not up-regulate or bind to BiP, the sentinel chaperone recognizing misfolded proteins in the ER and the initiator of the canonical ER stress response. 6 Subsequent studies on cells transfected with collagen I containing helix Gly mutations demonstrated that mutant-containing collagen trimers aggregated in the ER and were degraded via autophagy. 7 The precise nature of the ER stress response to collagen containing Gly substitutions is not yet clear and may involve mutation and chain (α1(I) or α2(I)) specificity.
Recent studies on a mouse model of mild to moderate OI (OI type IV) has yielded important information on disease mechanisms and possible therapeutic approaches. 8 This model features a Col1a2 triple helical codon 610 Gly to Cys substitution (α2(I) G610C), corresponding to a mutation first identified in an Amish family. 9 The α2(I) G610C mutation disturbs the collagen triple helix and results, as expected, in ER accumulation of the mutant-containing misfolded collagen trimers. This results in an unusual form of ER stress which does not involve the canonical unfolded protein response (UPR). 8 This unconventional ER stress response involves modest up-regulation of CHOP, eIF2α phosphorylation, and chaperones αβ crystalline and HSP47. It is also associated with striking negative effects on osteoblast function, affecting cell differentiation and maturation and an abnormal response to key signalling pathways. Importantly, this study showed that the ER retention of the mutant collagen stimulated autophagy and this was a key cellular adaptive response which modified the severity of the cell stress. 8 Furthermore, they found that stimulating autophagy in osteoblasts in vitro with rapamycin reduced intracellular levels of mutant collagen and improved collagen secretion and extracellular matrix deposition. 8 Thus, it is reasonable to hypothesize that therapeutic induction of autophagic degradation of the intracellularly retained mutant misfolded collagens should ameliorate the ER stress and deleterious downstream consequences on osteoblast differentiation and function in vivo. Preliminary testing of this approach used autophagy stimulation by a low-protein diet in the OI mice. 10 While this resulted in some beneficial effect on osteoblast differentiation and mineralization, the diet significantly reduced both WT and OI mouse growth, preventing any conclusions on the therapeutic effects on OI mouse bone.
To further explore the potential therapeutic benefit of autophagy induction in this α2(I) G610C OI mouse model, we treated mice with rapamycin, a well-characterized autophagy stimulator. 11,12 Rapamycin is a commonly used immunosuppressant and chemotherapeutic drug which inhibits mTOR, a key nutrient sensitive serine-threonine kinase. The mTOR pathway is involved promoting anabolic processes, ribosome biogenesis, protein synthesis and many cellular pathways, inhibiting cell stress responsive pathways, and protein degradation by autophagy. 13 Inhibiting mTOR with agents such as rapamycin retards protein synthesis and enhances cell stress responsive pathways, such as autophagy. 11 Treatment with rapamycin and rapalogs (rapamycin analogues), while mostly applied in cancer therapeutics, improves ER stress-induced diabetes in mice 14 and is generally beneficial in several mouse models of protein misfolding/ aggregation neurodegenerative diseases such as Alzheimer's disease, Huntington's disease, Parkinson's disease, and amyotrophic lateral sclerosis. 11 Rapamycin has also been shown to increase longevity of wild-type mice. 15 However in other studies, rapamycin was shown to exacerbate motor neuron degeneration in a SOD1 mouse model of amyotrophic lateral sclerosis, 16 suggesting caution must be taken when employing rapamycin (and rapalogs) and take into account the possible role of mTOR inhibition at the tissue level, disease, and mutation context. The role of mTOR signalling via the mTORC1 complex on osteoclast, osteoblast, and osteocyte differentiation and function is not fully understood with both positive and negative effects on bone formation reported. 17 However, rapamycin is generally considered to be a largely bone-sparing drug which may improve compromised bone quality. 18,19 In a recent study, mice with unloading induced bone loss were treated with rapamycin which restored osteoblast differentiation and bone volume. 20 Similarly, rapamycin reduces the severity of senile osteoporosis in rats 21 and osteopenia in mice with systemic sclerosis. 22 Here we treat the α2(I)-G610C OI mice with rapamycin for 5 weeks from weaning and measure the effect of this treatment on the structure and mechanical properties of the long bones and vertebrae to determine if rapamycin stimulation of autophagy is a viable new clinical approach in the treatment of OI caused by collagen I misfolding/aggregation mutations.

| Animals
Heterozygous α2(I)-G610C osteogenesis imperfecta mice 9  Louis, MO, USA) as previously described. 15 The control 5LG6 diet contained 154 ppm Eudragrit. This rapamycin diet has been previously shown to result in detectable rapamycin in the blood and various mouse tissues 23,24 along with reduced mTORC signalling in several mouse tissues. 23,25 Male WT and α2(I)-G610C mice were fed the rapamycin or control diets from 3 weeks of age until 8 weeks (n = 8 per experimental group). Mice were selected for the study by random allocation of male mice to control or rapamycin groups.
There were no heterozygous runts and no mice were excluded. Regular serological assessment of sentinel mice was conducted to ensure the colonies were free of infection. Body weight was determined regularly during the 35 days of rapamycin treatment (daily determinations during the first week of treatment and three times a week for the remaining treatment period). At 8 weeks, mice were killed and tissues harvested for analysis.

| Micro-CT analysis
Ex vivo micro-CT was performed on tibiae using the SkyScan 1076 system (Bruker-micro-CT) as previously described. 26 Image acquisition settings were as follows: 9 μm voxel resolution, 0.5 mm alu-

| Mechanical testing
Right femora, right tibiae, and L4 vertebrae were frozen in salinesoaked gauze at −80°C. After thawing, destructive testing was carried out by four-point bending on an Instron 5944 mechanical testing machine (Canton, MA, USA) with data collected using Bluehill 3 software (Instron). Tibiae were positioned so that the medial side was resting across the supporting bottom span. Femora were positioned so that the anterior side was resting across the supporting bottom span.
Bones were tested using a support span of 10 mm, with an upper span of 5 mm. Samples were pre-loaded at 0.25 mm/min until a load of 1N was reached, at which point the loading rate increased to 0.5 mm/min until failure. Testing of vertebrae was performed on a custom jig featuring a support pin that threaded the neural canal to provide stability, and an upper plate with a corresponding female fit. The lower plate was covered with a fine grit sandpaper to minimize slippage. Prior to testing, vertebrae had their processes removed using scissors. Compression testing was carried out at 3 mm/min until failure with vertebrae oriented along their cephalocaudal axis, with the superior end facing up. 27 Biomechanical properties were adjusted for body weight and bone length as described in the recent guidelines for mouse biomechanical analysis. 28

| Calvarial cell isolation and culture
Osteoblasts were obtained from the parietal bones of 10-day-old mice by sequential digestion with PBS containing 2 mg/mL collagenase (Worthington Biochemical Corporation; Collagenase II; Thermo-Fisher) and containing 0.1% (w/v) trypsin. 29    Previous studies have shown that OI mice have a slightly smaller body weight than WT mice at day 21 to day 60, 9 and this trend was apparent in our OI mice ( Figure S2) although the difference was not statistically significant at 3 weeks or 8 weeks of age ( Figure 1A). Likewise, the length of tibiae ( Figure 1B) was indistinguishable between OI and WT mice at 8 weeks. To test the effect of rapamycin on OI and WT bone parameters in vivo, 3-week-old male mice were administered rapamycin via their diet for 5 weeks until 8 weeks of age. Rapamycin treatment reduced the increase in body weight normally observed between 3 and 8 weeks in OI mice, but not in WT mice ( Figure 1A).

| Osteoblast apoptosis
Similarly, OI but not WT, tibiae also showed impaired longitudinal growth after rapamycin treatment ( Figure 1B). These data demonstrate that rapamycin treatment had a selective deleterious effect on the growth and skeletal development of OI mice.

| Rapamycin increased trabecular bone mass in both WT and OI mice
In WT mice, rapamycin led to a greater trabecular bone mass ( Figure 2D) than controls, as previously reported. 9,30 All these parameters were significantly improved in OI mice by rapamycin treatment (Figure 2). Moreover, they reached levels of trabecular bone mass (Figure 2A), thickness ( Figure 2B), and number that were not significantly different from untreated WT trabecular bone. Rapamycin-treated trabecular spacing was reduced compared to WT but similar to that of rapamycin-treated WT mice. These data indicate that rapamycin treatment over this 5-week developmental window may restore OI trabecular bone mass to levels comparable to WT trabecular bone.

| Rapamycin treatment selectively reduced OI cortical bone mass
In contrast to its effect on trabecular bone, rapamycin treatment for 5 weeks had no significant influence on the cortical bone micro-CT parameters of WT mice ( Figure 3A-E) except the polar moment of inertia ( Figure 3F). This showed a small but statistically significant reduction. Cortical bone from untreated OI mice showed significant deficits compared to WT bone in all cortical micro-CT parameters ( Figure 3A,B,D,F) except endocortical perimeter ( Figure 3C) and cortical thickness ( Figure 3E). Comparable data demonstrating the osteopenic phenotype of these OI mice have been previously reported. 9,30 Rapamycin treatment did not improve any of these OI cortical bone parameters, but to the contrary specifically further negatively impacted the OI cortical bone. The cortical area ( Figure 3B), cortical thickness ( Figure 3E), and polar moment of inertia ( Figure 3F) were all significantly lower in rapamycin-treated OI bones compared to untreated OI bones.

| Mechanical deficiencies in OI long bone and vertebrae are not rescued by rapamycin
Biomechanical testing of WT and OI femora confirmed the biomechanical weakness of the OI long bones; they demonstrated reduced maximum load to failure and energy to maximum load ( Figure 4A-C). The most dramatic mechanical deficiency of untreated OI femora was revealed by analysis of energy after maximum load ( Figure 5D) confirming the "brittleness" of the OI bones compared to untreated WT femora. 27 Likewise, OI tibiae showed a comparable reduction in these parameters ( Figure 4E-H).
Consistent with the negative impact of rapamycin on OI cortical bone parameters, the maximum load to failure for tibiae from rapamycin-treated OI mice was further reduced ( Figure 4E).
Femora showed a similar trend ( Figure 4E). In both femora and tibiae, the OI long bone brittleness ( Figure 4D,H) was not rescued by rapamycin treatment. An important novel finding of our study was that rapamycin treatment of WT mice resulted in a reduced energy after maximum load of both femora and tibiae ( Figure 4D, H) suggesting that this rapamycin treatment regime increases the brittleness of normal bones. • WT, ■ OI. Data were analysed using a two-way ANOVA with Tukey's post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 age. Despite its many potential cellular effects, rapamycin is clinically well-tolerated and has been used effectively in several studies on protein misfolding/aggregation disorders. 11 Rapamycin-treated WT mice had normal body weight and long bone length, and the improvement in trabecular bone structure with rapamycin treatment is consistent with the concept that an autophagy-inducing low-protein diet may improve osteoblast differentiation and/or bone mineralization. 10 Recent studies have similarly shown that autophagy promotes osteogenic differentiation of human bone marrow mesenchymal stem cells 32 and osteoblast-specific ablation of autophagy in mice results in reduced bone mass. 33 Moreover, in several mouse or rat models of osteopenia, rapamycin treatment was beneficial, 20,22 suggesting possible therapeutic applications in some bone loss conditions. In addition, it has been reported that rapamycin-induced autophagy improves bone fracture healing. 34,35 In contrast to the beneficial effects on the trabecular compart-   • WT, ■ OI. Data were analysed using a two-way ANOVA with Tukey's post hoc test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 The molecular basis of this selective impact on the growth of the α2(I)-G610C mutant mouse bone is not known. Rapamycin acts by associating with FK506-binding protein 12 which then binds the mTOR Complex 1 (mTORC1) preventing downstream signalling.
While autophagy stimulation is a key downstream consequence, it is important to appreciate that mTORC1 is a signal integrator that is involved in numerous cellular processes involved in cell growth and energy metabolism, including protein synthesis. 13 The role of mTOR signalling in bone formation and homeostasis is still controversial with evidence for both stimulation and inhibition of osteogenic differentiation and bone formation. 17 To correct for the reduced body weight and bone length of rapamycin-treated OI mice the data were plotted against individual mouse weight × bone length previously described. 28 For each plot, the significance (P-value) of the difference between the slopes of untreated OI (■, OI Control) and rapamycin-treated OI (•, OI Rapa) bones are indicated (ANCOVA). The r 2 value and P-value for each plot is given below to describe if the slope is significantly non-zero. either enriched on the exposed surfaces of WT/mutant composite collagen fibers, or polymerized into mutant-only fibrils prone to proteolytic attack. 45 Thus, it is possible that the amount of mutant, its organization, and structural consequences may be affected by differential bone remodelling in trabecular vs cortical bone. Likewise, the extent of ER stress and its downstream consequences may be subtly different in trabecular and cortical bone compartments.
This study cannot distinguish between effects of altered mTOR signalling and autophagy. To evaluate the true therapeutic value of autophagy stimulation in OI resulting from collagen misfolding mutations, it will be important to test other autophagy modulators with increased specificity, 12  We conclude that the interaction of rapamycin with bone in the context of OI is complex and may be fundamentally altered compared to healthy bone. We show that rapamycin treatment offered no functional benefit in this OI preclinical model, and suggest that the impact of rapamycin on OI bone growth could potentially exacerbate the clinical consequences in OI patients during periods of active bone growth. These data underline the need for clinical caution about using mTOR inhibitors, rapamycin, and rapalogs, in patients with OI caused by collagen mutations.