The effect of carbamazepine on bone structure and strength in control and osteogenesis imperfecta (Col1a2 +/p.G610C ) mice

Abstract The inherited brittle bone disease osteogenesis imperfecta (OI) is commonly caused by COL1A1 and COL1A2 mutations that disrupt the collagen I triple helix. This causes intracellular endoplasmic reticulum (ER) retention of the misfolded collagen and can result in a pathological ER stress response. A therapeutic approach to reduce this toxic mutant load could be to stimulate mutant collagen degradation by manipulating autophagy and/or ER‐associated degradation. Since carbamazepine (CBZ) both stimulates autophagy of misfolded collagen X and improves skeletal pathology in a metaphyseal chondrodysplasia model, we tested the effect of CBZ on bone structure and strength in 3‐week‐old male OI Col1a2 +/p.G610C and control mice. Treatment for 3 or 6 weeks with CBZ, at the dose effective in metaphyseal chondrodysplasia, provided no therapeutic benefit to Col1a2 +/p.G610C mouse bone structure, strength or composition, measured by micro‐computed tomography, three point bending tests and Fourier‐transform infrared microspectroscopy. In control mice, however, CBZ treatment for 6 weeks impaired femur growth and led to lower femoral cortical and trabecular bone mass. These data, showing the negative impact of CBZ treatment on the developing mouse bones, raise important issues which must be considered in any human clinical applications of CBZ in growing individuals.

expression, such as nonsense mutations which lead to mRNA decay, result in milder clinical phenotypes (e.g. OI type I), whereas mutations introducing missense changes causing structural disruptions in collagen proα-chains lead to more severe OI phenotypes (e.g. OI types II, III, IV). These severe phenotypes most commonly exhibit impaired collagen I triple helix structure and stability, usually caused by glycine substitutions that interrupt the obligatory Gly-X-Y amino acid repeat sequence required for collagen helix folding. A common consequence is dysregulated collagen proteostasis, with intracellular retention and aggregation of the misfolded collagen. 3 Its intracellular accumulation can induce an endoplasmic reticulum (ER) stress response which might contribute to the pathology. 2,4 A mouse model of OI type IV (Col1a2 +/p.G610C ) with a Gly to Cys Col1a2 missense mutation in the triple helix, based on an Old Order Amish OI mutation, 5 provides us with an experimental tool for pre-clinical therapeutic testing. Previous studies on this mouse model have shown that, as expected, the glycine substitution mutation leads to intracellular mutant collagen accumulation, a poorly characterized non-canonical ER stress response and downstream impairment in osteoblast differentiation and function 6 with clinically relevant changes to bone structure and fragility. 7 While the molecular detail of how helix-disrupting OI mutations cause intracellular ER stress and bone pathology are not fully resolved, 8 a promising therapeutic approach would be to reduce the load of toxic misfolded collagen and thereby reduce the OI bone pathology to a milder phenotype. There are few FDA-approved drugs that could be re-purposed to achieve this. Rapamycin, which stimulates autophagy via the mTOR pathway, showed promise on osteoblasts from the Col1a2 +/p.G610C mouse in vitro. 6 However, in vivo testing in the Col1a2 +/p.G610C mouse demonstrated that while some OI bone trabecular properties were improved, off-target effects suppressing longitudinal and transverse bone growth in OI mice precluded rapamycin as a viable therapeutic option. 9 Another drug stimulating mTOR-independent autophagy, carbamazepine (CBZ), has been effective in a mouse model of metaphyseal chondrodysplasia, type Schmid (MCDS), where intracellular accumulation of misfolded mutant collagen X causes ER stress and downstream pathological signalling. 10 CBZ-stimulated degradation reduced the toxic mutant collagen load and ER stress, and improved the MCDS cartilage and bone pathology. 11,12 As a result, CBZ is now in clinical trials with MCDS patients (mcds-therapy.eu). Because of the broad similarity of the molecular phenotypes in OI and MCDS (collagen misfolding and intracellular retention) and the predicted therapeutic effect of CBZ in stimulating autophagic clearance of misfolded collagen I, there is considerable interest in CBZ as a potential OI therapy.
Here, we treated 3-week-old Col1a2 +/p.G610C and control mice with CBZ for 3 and 6 weeks from weaning and assessed structure and strength of the long bones. While CBZ treatment for 3 weeks does not affect control bone, longer term treatment (6 weeks) during this period of active bone growth reduces bone length, width and strength. Furthermore, CBZ treatment for 3 or 6 weeks does not rescue the OI bone phenotype in Col1a2 +/p.G610C mice. Male control and Col1a2 +/p.G610C mice were treated by oral gavage with CBZ (250 mg/kg/day) from 3 to 6 weeks. 11,12 A second cohort was additionally treated from 6 to 9 weeks by subcutaneous implantation of a slow-release pellet of carbamazepine (250 mg/kg/day) (Innovative Research of America USA C-113). 11 At 6 and 9 weeks, mice were euthanased and tissues harvested.

| Micro-computed tomography (μ CT)
Femora were assessed on a Skyscan 1276 micro-CT system (Bruker) at 9 micron voxel resolution as previously described, 9 using NRecon (version 1.7.1.0), DataViewer (version 1.5.4) and CT Analyser (version 1.16.4.1). Trabecular bone structure and multi-level thresholding were carried out in the metaphysis, starting at 7.5% of bone length from the distal femoral growth plate and along 15% of total bone length. Cortical structure was assessed in the diaphysis, starting at 30% of bone length from the distal femoral growth plate and along 15% of total bone length. Medio-lateral widths and craniocaudal widths were determined at the femoral midpoint. Single thresholds for cortical bone were 0.487 and 0.786 g/cm 3 CaHA for 6-and 9-week-old mice, respectively, and for trabecular bone 0.219 and 0.435 g/cm 3 CaHA for 6-and 9-week-old mice, respectively. To segregate bone into multiple density levels, nonparametric unsupervised 4 level Otsu-thresholding algorithm was used, as previously described. 13 Each density area was normalized to total bone area, and thresholds of 0.301-0.613, 0.614-1.003 and >1.003 mg/ cm 2 CaHA were used for low-, mid-and high-density bone areas, respectively; the background level (0-0.3 mg.cm 2 CaHA) was discarded.

| Three point bending strength tests
After μCT, femora were stored in 70% ethanol. Three point bending was performed with a Bose Biodynamic 5500 Test instrument (Bose, DE, USA) and WinTest 7 software. Femora were centred onto supporting pins with a span width of 6 mm and loaded with 0.5 mm/s. As described previously, 14 the load/deformation data were collected over 5 s with a sampling rate of 250 Hz. For material properties, the data were normalized to medio-lateral and cranio-caudal widths and average cortical thickness of the diaphysis of each bone.

| Synchrotron-based Fourier-transform infrared microspectroscopy (sFTIRM)
sFTIRM was conducted on a Bruker Hyperion 2000 IR microscope coupled to a V80v FTIR spectrometer at the Australian Synchrotron IR Microspectroscopy beamline, as previously described. 15

| Six-week-old Col1a2 +/p.G610C mice have defective bone structure that is not modified by CBZ treatment from 3 weeks of age
Previous studies have shown that from 8 weeks of age (but not at 10 days of age), Col1a2 +/p.G610C mice have less cortical and trabecular bone and lower bone strength than controls. 5,7,9,18 We showed that the phenotype can also be detected at 6 weeks of age. Femora from 6-week-old vehicle-treated Col1a2 +/p.G610C mice were narrower with significantly lower cortical area, periosteal and endocortical perimeters compared to controls ( Figure 1A-D). In addition, the marrow area, mean polar moment of inertia and medio-lateral and craniocaudal widths were significantly lower than controls (Table S1).
Col1a2 +/p.G610C femora from 6-week-old mice also had significantly lower trabecular bone volume, number, separation and shorter femora than controls ( Figure 1E-I). Treatment with 250 mg/kg/day CBZ for 3 weeks, a dose equivalent to that used in treating epilepsy in human subjects taking into account species differences 19 and to correct cartilage defects in the MCDS mouse, 11,12 did not significantly modify cortical or trabecular bone mass in either control or Col1a2 +/p.G610C femora (Figure 1A-I; Table S1).
3.2 | CBZ from 3 to 9 weeks of age had no effect in Col1a2 +/p.G610C mice but impaired bone growth in control mice The lower trabecular bone mass, and smaller femoral length and width of 6-week-old vehicle-treated Col1a2 +/p.G610C mice compared to controls were also detected at 9 weeks of age ( Figure 2; Table S2). The extended period of CBZ treatment (from 3 to 9 weeks of age) had no significant effect on trabecular or cortical structure in Col1a2 +/p.G610C femora ( Figure 2; Table S2). Although short-term CBZ treatment in control mice for 3 weeks had no effect on bone structure ( Figure 1; Table S1), a longer duration of CBZ treatment for 6 weeks led to significantly lower cortical area, cortical thickness, periosteal and endocortical perimeters ( Figure 2A-D), marrow area, mean polar moment of inertia, mediolateral and cranio-caudal widths (Table S2), indicating suppressed transverse bone growth. Six weeks of CBZ treatment in control mice lowered trabecular bone volume, thickness and number ( Figure 2E,F,H). In these 9-week-old mice, trabecular thickness and number were lower than in the younger vehicle-treated control mice ( Figure 1E,F,H), suggesting loss of trabecular bone mass, rather than impaired trabecular bone accrual. In addition, while femoral length was greater at 9 weeks than 6 weeks in vehicle-treated control mice, CBZ-treated control mice had shorter femora than vehicle-treated controls ( Figures 1I and 2I), indicating that a longer duration of CBZ treatment suppressed both transverse and longitudinal bone growth.

| Col1a2 +/p.G610C mice have disrupted cortical bone consolidation, but this is modified by CBZ only in control mice
Cortical bone consolidates by changing from a material with a high proportion of low-density bone to a material with mainly highdensity bone. 16 This transition can be measured along the length of the growing bone from the growth plate (more low-density bone) to the diaphysis (more high-density bone). Since cortical diaphyseal bone of Col1a2 +/p.G610C mice has greater average tissue mineral density, 18 we used multi-level thresholding to assess whether highdensity material accrued more rapidly in less mature regions of the femoral metaphysis of Col1a2 +/p.G610C mice.
When total proportions of low-, mid-and high-density bone were assessed in vehicle-treated animals, Col1a2 +/p.G610C femora had a higher proportion of mid-density bone material than controls, with no significant difference in proportions of low-or highdensity bone ( Figure 3). This was not modified by CBZ treatment ( Figure 3).
When assessed along the length of the bone, vehicle-treated control and Col1a2 +/p.G610C bones both showed the expected reduction in low-density and increase in high-density bone with increasing distance from the growth plate ( Figure S1A). However, Col1a2 +/p.G610C bones had less high-density bone in the least mature region (close to the growth plate), more low-density bone in the most mature region (near the diaphysis) and a higher proportion of mid-density bone along the full length of the metaphysis, indicating region-specific differences in cortical bone maturation in this model of OI. CBZ treatment from 3 to 9 weeks of age increased the proportion of high-density bone and reduced the proportion of low-density bone in control mice (Figure 3). This was significant along the full extent of the metaphysis, indicating that it existed at all stages of cortical bone consolidation and that CBZ could change the type of bone deposited ( Figure S1B). CBZ treatment did not affect these parameters in Col1a2 +/p.G610C femora at any point along the metaphysis ( Figure S1C).

| sFTIRM shows mineral: matrix content of Col1a2 +/p.G610C bone progressively increases in deeper sub-periosteal bone
Eight-week-old, but not 10 day old, male Col1a2 +/p.G610C mice exhibit high mineral: matrix ratio, 7 consistent with an imbalance between F I G U R E 1 Femoral bone structure of 6-week-old male control and Col1a2 +/p.G610C mice treated with vehicle or carbamazepine (CBZ) for 3 weeks assessed by micro-computed tomography (μCT). Shown are (A) cortical area, (B) cortical thickness, (C) periosteal perimeter and (D) endocortical perimeter, (E) trabecular bone volume, (F) trabecular thickness, (G) trabecular separation, (H) trabecular number and (I) femur length from male control and Col1a2 +/p.G610C mice. Data shown are mean ± SEM, n = 6-13 mice/group. *p < 0.05, **p < 0.01, ***p < 0.001 vs. treatment-matched controls collagen production and mineral accrual. 20 Since 12-week-old tibial cortex exhibits a gradient of matrix maturation from the periosteal edge moving inwards, including mineral accrual, collagen compaction and carbonate substitution, 15,17 we assessed whether the OI phenotype or CBZ treatment modified bone matrix composition and maturation.
Both 6-and 9-week-old vehicle-treated Col1a2 +/p.G610C tibial cortex had greater mineral: matrix ratio (higher phosphate: amide I ratio) than controls ( Figure 4A,C), indicating that this matrix defect is already significant at 6 weeks of age. In vehicletreated animals, the mineral: matrix ratio increased in both control and Col1a2 +/p.G610C bone at 10 μm and 20 μm inwards from the periosteal edge compared to the periosteum, at both 6 and 9 weeks of age; the slope of this increase was notably greater in the more mature 9-week-old bone ( Figure 4B,D). Additionally, 6-week-old Col1a2 +/p.G610C bones had a higher mineral: matrix ratio at 10 μm and, in 9-week-old mice, at both 10 μm and 20 μm from the periosteum compared to the same regions in control bones ( Figure 4B,D). The mineral: matrix ratio using the amide II reference showed a similar response (Table 1; Figure S2A,E). This indicates that although mineral is initially deposited at normal levels on the growing periosteum of Col1a2 +/p.G610C bones, it accrues mineral more rapidly in Col1a2 +/p.G610C mice than in controls. were not significantly different between vehicle-treated control and Col1a2 +/p.G610C bone at 6 or 9 weeks of age (Table 1). Although carbonate: phosphate ratio was lower in vehicle-treated Col1a2 +/p.G610C bones than control at 6 weeks of age (Table 1), this genotype-dependent effect was not observed at 9 weeks of age. While average carbonate: phosphate ratio was lower in vehicle-treated Col1a2 +/p.G610C than controls at 6 weeks of age, region-specific analysis showed no change in carbonate: phosphate ratios with matrix maturation in either controls or Col1a2 +/p.G610C bones at any age ( Figure S2B,F). However, average crystallinity increased and amide I: II ratio decreased with increasing distance from the periosteum in both genotypes at 9 weeks of age ( Figure S2G,H), suggesting that Col1a2 +/p.G610C bone undergoes normal mineral maturation at 9 weeks of age, apart from the abnormal increase in mineral relative to collagen. No significant effects of CBZ were detected by sFTIRM in either control or Col1a2 +/p.G610C mice ( Figure 4; Figure S2A-H; Table 1).

| CBZ does not improve the Col1a2 +/p.G610C bone strength defect but reduces bone strength in control mice
Three point bending tests in vehicle-treated mice showed that femora from 6-and 9-week-old Col1a2 +/p.G610C mice were weaker than control femora ( Figure 5). This difference was greater in 9-week-old mice, which showed a ~60% lower maximum load than controls ( Figure 5E) while 6-week-old Col1a2 +/p.G610C femora were only ~30% weaker than their age-matched controls ( Figure 5A).
Col1a2 +/p.G610C femora were also more brittle compared to control femora, and again, this was more pronounced in 9-week-old mice, shown by a ~90% lower post yield displacement ( Figure 5F). There were also greater differences between the genotypes in yield load (Tables S1 and S2) in 9-week-old than 6-week-old mice.
Since Col1a2 +/p.G610C femurs were narrower than controls, strength testing data were corrected for bone widths and cortical thickness. When corrected for bone size, vehicle-treated Col1a2 +/p.G610C femora were still more brittle than controls at both ages, shown by a lower post yield strain ( Figure 5D,H), and lower elastic deformation, shown by a lower yield strain (Tables S1 and   S2). The stiffness of Col1a2 +/p.G610C femora was double that of controls, shown by a higher elastic modulus (Table S2). There was no difference in maximum stress in Col1a2 +/p.G610C femora compared to control femora ( Figure 5C) at 6 weeks of age but a significant reduction at 9 weeks ( Figure 5G). This suggests that the genotype-dependent reduction in bone strength at 6 weeks can be explained by the smaller size but the bone strength defect persists in 9-week-old mice due to the progressing material defect. This indicates that the OI phenotype progressively worsens between 6 and 9 weeks of age.
CBZ had no significant effect on any parameter measured by three point bending tests in Col1a2 +/p.G610C mice at 6 or 9 weeks of age ( Figure 5; Table S1 and S2). However, in control mice, although 3 weeks of CBZ treatment did not modify bone strength, after 6 weeks of treatment maximum load, yield load and displacement were significantly lower than in vehicle-treated controls ( Figure 5A,E; Tables S1 and S2). When corrected for bone width, no strength parameters were significantly altered by either 3 or 6 weeks of CBZ treatment (Figure 5C,D,G,H; Tables S1 and S2).
This indicates that CBZ treatment of 3-week-old control mice for 6 weeks suppresses bone growth sufficiently to reduce functional strength.

| DISCUSS ION
This study was based on the premise that stimulation of autophagic mutant collagen I degradation by carbamazepine (CBZ) treatment might provide therapeutic benefit in OI. In collagen misfolding disorders, the best evidence for this approach comes from a mouse model of Metaphyseal Chondrodysplasia, Type Schmid (MCDS), where CBZ treatment stimulated autophagic degradation of the mutant misfolded collagen X, reducing ER stress and improving the cartilage and bone disease severity. 11,12 This provided a strong rationale for testing the possible therapeutic effect of carbamazepine in the Col1a2 +/p.G610C mouse model of OI.
However, we found that treatment of 3-week-old Col1a2 +/p.G610C mice for either 3 or 6 weeks with CBZ, at the dose effective in the MCDS model, had no therapeutic benefit for the deficiencies in F I G U R E 3 Multi-level-thresholding analysis of the average femoral metaphysis segregated by low-(light-grey), mid-(dark-grey) and high-density (black) levels of the femoral cortex from 9-weekold male control and Col1a2 +/p.G610C mice treated with vehicle or carbamazepine (CBZ) for 6 weeks by micro-computed tomography (μCT). Data are mean ± SEM; n = 5-8 mice/group. **p < 0.01 vs. treatment-matched controls. #q < 0.05, ##q < 0.01 vs. treatmentand region-matched controls bone structure, strength or composition in Col1a2 +/p.G610C mice.
The study also provides new information about the Col1a2 +/p.G610C mouse bone phenotype and reveals that CBZ treatment in mice with healthy skeletons can be detrimental.
The lack of beneficial effect of CBZ on Col1a2 +/p.G610C bone is not due to unavailability of CBZ in the OI bone with this treatment regime, since the impact of CBZ on the control bone (discussed below) clearly indicates that the CBZ was bio-available in these growing bones. Furthermore, this dose of CBZ was effective in modifying the ER stress response of the MCDS mouse growth plate cartilage. 11,12 This suggests that CBZ does not stimulate sufficient intracellular mouse. Other approaches based on clearing mutant collagen may still be effective, since rapamycin 9 and CBZ have known 'off-target' effects.
An alternative approach that has gained traction in several diseases is to reduce the mutant misfolded protein load, not by degradation, but by chemical chaperones such as 4-phenylbutyric aid (4-PBA) to assist F I G U R E 4 Analysis of average (A,C) and regional (B,D) mineral: matrix ratios of tibial cortex from 6-week-old (A,B) and 9-week-old (C,D) male control and Col1a2 +/p.G610C mice treated with vehicle or carbamazepine (CBZ) for 3 or 6 weeks, respectively, by synchrotron Fouriertransform infrared microspectroscopy (sFTIRM). Data are shown as the average phosphate: amide I ratio and regional measurements, which show data at 10 micron intervals with increasing distance from the periosteum. Ratios were calculated from integrated areas of phosphate (1180-916 cm −1 ) and amide I (1712-1588 cm −1 ) curves. Data are mean ± SEM; n = 7-11 mice/group for 6-week-old mice and n = 5-8 mice/ group for 9-week-old mice. *p < 0.05, **p < 0.01 vs. treatment-matched controls. +q < 0.05, ++q < 0.01, +++q < 0.001 vs. genotype-and treatment-matched region 0 μm. #q < 0.05, ##q < 0.01 vs. treatment-and region-matched controls in refolding the mutant proteins. 21

4-PBA reduces ER stress in vitro in
fibroblasts with a range of mutations, including COL1A1 and COL1A2 misfolding mutations. 22,23 In vivo, 4-PBA treatment improved bone abnormalities in a zebrafish collagen helix misfolding mutation (Col1a1a p.G736D) OI model, 24 and in the Aga2 +/− mouse model, where a Col1a1 frameshift mutation in the C-propeptide results in failed procollagen I assembly. 25 In contrast, in the Col1a2 +/p.G610C mouse 4-PBA reduced the OI bone growth deficiency by enhancing the transition of the OI hypertrophic chondrocytes to osteoblasts, a process suppressed with this OI mutation. 26 4-PBA had little, or no, impact on the OI osteoblasts and did not rescue the bone fragility phenotype. 27 Thus, 4-PBA may be effective for some OI mutations, less so for others, most likely due to the nature of the collagen folding defects caused by different mutations and the resulting cellular response to the misfolded collagen. The cellular response to the disturbed proteostasis resulting from collagen I misfolding, and whether a canonical or 'non-canonical' UPR results remains unclear with many mutations, especially those disturbing the triple helix. 8 Those impacting the C-propeptide may be more likely to elicit a canonical UPR 8 and it is tempting to speculate that CBZ may be more effective with these given their similarity to the CBZ-responsive collagen X mutations, 11 and the positive effect of 4-PBA on the Aga2 +/− mice. 25 While CBZ treatment could show mutation specificity in its effect, the use of CBZ as a OI therapeutic is, however, moot since we found that the 6-week CBZ treatment of 3-week-old growing bone in control mice was deleterious. CBZ impaired bone growth, in both longitudinal and transverse directions, and reduced the cortical and trabecular bone mass of healthy mice leading to a significant reduction in functional bone strength. Although CBZ treatment changed the morphology of the control bone to reach similar proportions to the OI bone, the strength defect caused by CBZ was less severe than the OI phenotype. This was because CBZ changed bone morphology without accompanying changes in bone composition, measured by sFTIRM. The reason why CBZ has a negative impact on control bone growth, but not the mouse OI bone, is not known. CBZ has been shown in vitro to inhibit both osteoclastogenesis in human peripheral blood mononuclear cell-derived osteoclast precursors and osteoblastogenesis in human mesenchymal stem cell osteoblast precursors. 28 Disturbance to the delicate balance of these processes during bone growth could explain the CBZ effect on control bones.
In the Col1a2 +/p.G610C OI mouse model, bone marrow stromal cells had a much-reduced ability to form mature osteoblasts compared to controls 6 and in vivo, recent studies have shown the maturation and trans-differentiation of hypertrophic chondrocytes to osteoblasts is significantly reduced, resulting in the reduced growth of the OI bones. 26,27 Given the already severely deficient osteoblast differentiation and reduced bone growth in the OI mouse, it seems likely that the OI bones could be less sensitive to any additional impact on osteoblast differentiation and bone growth resulting from CBZ treatment.
CBZ is a widely used anti-epileptic drug (AED), and there have been multiple reports that short-and long-term CBZ treatment in children and adults with epilepsy leads to low serum 25-hydroxyvitamin D, and high serum parathyroid hormone and alkaline phosphatase. [29][30][31] This suggests CBZ, like other anti-convulsant therapies, causes osteomalacia and secondary hyperparathyroidism and may increase fracture risk. 32 However, analyses of the effect of CBZ on bone structure in children with epilepsy have been limited to small observational studies, with some reporting reduced areal bone mineral density (BMD) 30,33 but others indicating no significant effect. [34][35][36][37] No studies have reported TA B L E 1 Additional bone composition parameters of tibial cortex from 6-week and 9-week-old male control and Col1a2 +/p.G610C mice treated with vehicle or carbamazepine (CBZ) for three and 6 weeks, respectively, by synchrotron Fourier-transform infrared microspectroscopy (sFTIRM). Ratios were calculated from integrated areas of phosphate (1180-916 cm −1 ), carbonate (890-852 cm −1 ), amide I (1712-1588 cm −1 ) and amide II (1600-1500 cm −1 ) curves. Crystallinity sub-peak was calculated by the integrated area from 1030-1020 cm −1 . Data shown are mean ± SEM; n = 7-11 mice/group for 6-week-old mice and n = 5-8 mice/group for 9-week-old mice. *p < 0.05, **p < 0.01, ***p < 0.01 vs. controls lengthen, but cortical bone thickens, and the material changes from woven bone to an increasingly dense lamellar structure 40,41 ; this is reflected in the micro-CT scans showing a transition with increasing distance from the growth plate from a high proportion of lowdensity bone to a material containing more high-density bone. 13,16 Col1a2 +/p.G610C femora did exhibit cortical bone maturation, but the presence of less high-density bone in the least mature region (close to the growth plate), and more low-density bone in the most mature region (near the diaphysis), could indicate an initial more rapid process of cortical maturation and an early plateau, such that there is a higher proportion of mid-density bone along the full metaphyseal length. Our sFTIRM analysis also indicates that the greater proportion of mineral relative to collagen in Col1a2 +/p.G610C bone is already significant at 6 weeks of age and is only detected in more mature regions of bone. This indicates that although mineral is initially deposited at normal levels on the periosteum of Col1a2 +/p.G610C bone matrix, it accrues more rapidly in these mice than controls, probably driven by the defective collagen structure.
Our data raise two important considerations for any future applications of CBZ in genetic skeletal disease therapy. Firstly, given the lack of effect with this Col1a2 +/p.G610C mutant, it seems unlikely that CBZ will be useful for broad-based OI therapy. It remains possible that a beneficial CBZ effect could be OI mutant-specific, but this would require extensive mutation-specific testing in patients.
Secondly, and more importantly, the negative impact of CBZ on normal bone growth and structure is worrying in terms of any proposed therapies. If experimental evidence emerges that CBZ is beneficial with specific protein misfolding mutations, it will be critically important to closely monitor bone growth in clinical trials to ensure that any potential benefits outweigh the possible side effects. This may be particularly important when treatment coincides with bone growth.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data available on request from the authors