Depletion of skeletal TGFβ signaling with age

To elucidate the molecular mechanisms responsible for the age-related decline in bone quality, we first performed RNA sequencing on cortical bone from aged male C57/Bl6 (WT) mice. Unbiased hierarchical clustering of differential gene expression with age in bone (Fig. 1a, b) revealed several important profiles. Pathways that are downregulated with age include the TGFβ, Wnt, parathyroid hormone, PI3K-Akt, and calcium signaling pathways, as well as processes associated with ECM remodeling and adhesion (Fig. 1a). Because of the known role of TGFβ signaling in the control of bone quality,14,15,17,26 we further examined the effect of age on genes involved in the TGFβ signaling pathway. The levels of mRNA encoding TGFβ ligands, receptors, and effectors changed with age; some were induced, and others were repressed (Fig. 1c). Transcriptional readouts for the TGFβ inducible target gene Serpine1 are significantly repressed in 2 ½-year-old WT bone. The upregulated pathways include those involved in osteoclast differentiation and NF-κΒ signaling, as well as markers of osteocyte senescence or the senescence-associated secretory phenotype (SASP), including Cdkn2a (p16), Fasl, and Tnf, which were dramatically upregulated with age, as previously reported (Fig. 1d).12,44 Together, these results suggest a repression of TGFβ signaling at the transcriptional level with age in normally aging bone.

Fig. 1

Bone aging is associated with reduced TGFβ signaling and increased cellular senescence. Unbiased hierarchical pathway clustering of KEGG pathways for the whole transcriptome of male C57BL/6 (WT) mouse bone showed several pathways that were downregulated (a) or upregulated (b) with age, including the downregulated TGFβ signaling pathway. Analysis of individual genes at the transcriptional level showed reduced levels of several genes in the TGFβ signaling pathway (c) and elevated levels of markers of osteocyte senescence and the senescence-associated secretory phenotype (SASP) (d) in bone with age. Multiplexed ELISA of homogenized bone lysate (blue) showed age-related reductions in TGFβ ligand levels in the bone of both males and females (e–h). The circulating TGFβ ligand levels in blood serum (red) remained stable over time in males but showed decreases in females. P < 0.008 using Student’s t test with the Bonferroni correction with respect to (wrt.) 2 months (*), 12 months (#), and 24 months (☨), n = 3-4 per group. e–h *P < 0.008 wrt. 4 months from the same source (bone lysate or serum), ☨P < 0.008 wrt. The age-matched opposite tissue source (lysate vs. serum) in post hoc pairwise Fischer’s LSD test after two-way ANOVA, n = 12–15 per group

To determine whether posttranscriptional mechanisms contribute to an age-dependent loss in TGFβ signaling, we evaluated the levels of TGFβ1, 2, and 3 in cortical bone lysate and serum at 4, 12, and 15 months of age. The concentrations of both TGFβ1 and TGFβ2 decreased with age in bone in both males and females (Fig. 1e–h). The only significant change in serum with age was a decrease in TGFβ2 levels in female mice. Overall, the most significant predictor of ligand abundance in serum and bone over time was age. TGFβ1 levels were higher in serum than in bone, but the TGFβ2 levels were equivalent in both compartments. Although TGFβ levels in bone did not differ between the sexes, females had 25% more circulating ligand levels than males and showed a greater percentage loss of circulating TGFβ with age than males. Thus, evidence of TGFβ signaling in bone declines with age, with transcriptional and posttranscriptional changes at multiple levels of the signaling pathway.

Osteocyte-intrinsic TGFβ signaling impacts tissue-scale bone aging

In addition to the decline in TGFβ signaling in bone with age, bones from young mice with impaired osteocytic TGFβ signaling show signs of altered aging. Specifically, both aged WT mice and TβRIIocy−/− mice, in which TGFβ receptor type II is deleted in osteocytes, exhibit poor bone quality and dysregulated osteocyte lacunocanalicular networks relative to those of young or Cre-negative controls.26,34 However, the role of TGFβ signaling in osteocytes and their precursors in skeletal aging is unknown. Also unknown are the molecular and material mechanisms by which changes to cellular function, with age or impaired TGFβ signaling, compromise bone quality. While 10 kb DMP1-Cre may cause excision in osteoblasts, previous works with TβRIIocy−/− mice have not demonstrated alterations in bone mass or shape, osteoblast numbers, or osteocyte numbers, which might be anticipated if its effects were due primarily to the excision of TβRII from osteoblasts.26,27,34 The predominance of effects attributed to osteocyte-intrinsic interruptions to TGFβ signaling in these mice led to their use as a model to dissect the osteocyte-intrinsic role of TGFβ signaling in the mechanical and biological manifestations of bone aging.

Microcomputed tomography revealed genotype-dependent differences in the effect of age on cortical and trabecular bone. In males (Fig. 2a) at 4 months, TβRIIocy−/− cortical bone was thicker, larger, and more irregularly shaped than that of age-matched controls, with increased cortical thickness, total volume, and moment of inertia (MOI) (Fig. 2b–d). Both the control and TβRIIocy−/−, mice showed the anticipated progressive expansion of cortical bone with age. Other important genotype-dependent differences in cortical bone were observed in cortical porosity and tissue mineral density (TMD) (Fig. 2e, f). Intracortical pores were observed as early as 4 months in male TβRIIocy−/− bone, even though no such pores were detected until 15 months of age in Cre-negative controls (Fig. 2a, e). Finally, although both groups show the expected age-related increase in TMD,45TβRIIocy−/− males displayed lower TMD at all ages, which was consistent with previous findings.26,27 Although all μCT parameters reveal an altered aging trajectory for male TβRIIocy−/− bone only the cortical thickness presented a significant interaction term between aging and genotype, implying that the change in cortical thickness with age is dependent on the genotype. All other parameters, except cortical porosity, presented significant main effects for both age and genotype by two-way ANOVA with no interactions. Cortical porosity was the only μCT parameter to display only genotype main effects. Interestingly, direct comparison of Cre-negative control males at 15 months of age and TβRIIocy−/− males at 4 months showed no significant differences in post hoc pairwise comparisons for cortical thickness, MOI, total volume, or cortical porosity, while direct tests for statistical similarity (overlapping 95% confidence intervals) also showed these two groups to be statistically indistinguishable from one another (Fig. S1).

Fig. 2

Accelerated bone structure changes in TβRIIocy−/− cortical bone. μCT of male TβRIIocy−/− bone (a) revealed several signs of an accelerated age-like phenotype, including increased cortical thickness (b), overall bone volume (c) and MOI (d). Male TβRIIocy−/− bone also displayed the development of intracortical porosities (a, e), resulting in trabecularization of the cortex. TβRIIocy−/− males also exhibited lower bone mineral density (TMD) (f) than Cre-negative controls at all ages but did show expected age-related increases in bone density. Females (g) did not show any genotype-specific differences in pairwise comparisons to Cre-negative controls until 1 year, when they begin to display small, nonsignificant increases in thickness (significant main effect of genotype and age) (h) and significant increases in bone volume and MOI (i, j). Similar to males, female TβRIIocy−/− bone developed cortical porosity (significant main effect of genotype); however, these differences were not significant in pairwise comparisons at the corrected alpha levels (k). Similar to that in males, aged female TβRIIocy−/− bone displayed lower tissue mineral density than the Cre-negative controls (l). *P < 0.007 wrt. 4 months Cre-negative control, ☨P < 0.007 wrt. Age-matched within-genotype group in the 7-way Bonferroni corrected post hoc pairwise Fischer’s LSD test after two-way ANOVA. n = 8–12 per group

As previously reported, young TβRIIocy−/− female bone was indistinguishable from controls by μCT (Fig. 2g).27 However, at 12 and 15 months, TβRIIocy−/− female bones exhibited slight increases in cortical thickness and significantly elevated MOI and total volume compared to their age-matched controls (Fig. 2h–j), similar to the enlargement observed in males. In fact, although post hoc analysis falls short in showing certain pairwise significant changes, for instance, in cortical thickness, the significant main effects of genotype, along with the significant main effect of age, imply that with age, even female TβRIIocy−/− bones show meaningful changes in bone morphology that are dependent on TGFβ signaling.

With age, TβRIIocy−/− female bone shows a trend toward the accumulation of intracortical pores (Fig. 2k), which was not detected in female control bone at these times. TβRIIocy−/− female bone showed similar changes to aging TβRIIocy−/− males but to a lesser degree. Similar to males, this is the only parameter observed by μCT to exhibit a significant main effect of genotype alone without an accompanying significant effect of age. TβRIIocy−/− females resembled TβRIIocy−/− males in TMD (Fig. 2l) to a lesser degree and were not significantly different from Cre-negative controls until 15 months of age in post hoc pairwise comparisons at each age.

Within the trabecular compartment (Table 1), the Cre-negative controls showed anticipated losses of trabecular bone with age.46 As previously reported, young male TβRIIocy−/− bone had increased BV/TV and mineral density (TB. TMD) compared to the Cre-negative controls,16,26,27 but no differences were observed in trabecular number, thickness, or spacing. The elevated BV/TV in young TβRIIocy−/− males tempered the age-related trabecular bone losses seen in the controls. The effect of age on female trabecular bone was more pronounced than that on male trabecular bone. In contrast to the persistence of male TβRIIocy−/− trabecular bone with age, female trabecular bone for both genotypes exhibited similar age-related losses. In summary, the effect of osteocytic TGFβ deficiency on bone structure is more severe in males than in females and is consistent with an altered aging phenotype. While some bone structure parameters are unaffected by age in TβRIIocy−/− mice (e.g., male trabecular BV/TV), others are more sensitive to aging than controls (e.g., male cortical thickness).

Table 1 Male TβRIIocy−/− trabecular bone displays resistance to aging in μCT parametersWhole bone stiffness and strength are sensitive to age and limited TGFβ signaling

Three-point bending mechanical tests were performed to determine alterations in bone strength and quality in TβRIIocy−/− mice with age. Load‒displacement curves and bone fracture surfaces (insets) (Fig. 3a) revealed striking differences in failure behavior between male Cre-negative controls and TβRIIocy−/− bones. Observation of the load‒displacement curves shows that young male Cre-negative controls display, on the whole bone scale, extended postyield deflection and jagged or oblique fracture surfaces, evidence that crack deflection mechanisms were utilized to dissipate stress and increase toughness.47 These behaviors are less prominent in the 12-month-old Cre-negative controls and in the young TβRIIocy−/− males, as indicated by the truncation of the load‒displacement curves and the occurrence of more perpendicular fracture surfaces. These altered failure behaviors are further exaggerated in 12-month-old TβRIIocy−/− male bones.

Fig. 3

Aging and TβRIIocy−/− bone demonstrates reduced bone toughness despite increases in tissue stiffness and strength. a Observation of load displacement curves and fracture surfaces via SEM from three-point bend mechanical testing of femurs from male TβRIIocy−/− bone and age-matched controls showed similar fracture and failure mechanisms in TβRIIocy−/− bone and aged controls compared to young control bone. The quantification of mechanical properties showed increases in tissue stiffness (b) and fracture force (c) with both age and the loss of TGFβ signaling in males. Despite increases in tissue strength, the aged controls and TβRIIocy−/− bone displayed a decrease in postyield displacement (d), resulting in no improvement in tissue-level toughness in the work of fracture (e). Female TβRIIocy−/− (f–i) bones did not show significant changes in three-point bend parameters in age-matched pairwise comparisons to Cre-negative controls but did show age-related changes. The ultimate force and fracture work did display an important main effect of genotype. Both Cre-negative controls and TβRIIocy−/− females showed age-related decreases in the work of fracture (i). *P < 0.007 wrt. 4 months Cre-negative control, ☨P < 0.007 wrt. age-matched-within-genotype group in a 7-way Bonferroni-corrected post hoc pairwise Fischer’s LSD test after two-way ANOVA. n = 8–12 per group

Quantitative analysis confirmed the alterations in whole bone strength with age and diminished osteocytic TGFβ signaling. Consistent with prior reports,46,48,49 we qualitatively observed anticipated age-dependent changes, including increased stiffness and ultimate force with age (Fig. 3b, c) and decreased postyield displacement (Fig. 3d). The same trends were apparent in aging TβRIIocy−/− males, but they were accelerated so that stiffness and strength, measured by ultimate force, were increased, while 4-month-old male TβRIIocy−/− bones exhibited reduced postyield displacement relative to Cre-negative controls of the same age. However, postyield displacement in male TβRIIocy−/− mice did not decline further with age, showing no significant differences in this parameter across 4, 12, or 15 months. The increases in stiffness and strength in male TβRIIocy−/− bones offered no advantage to their material quality, as assessed by my fracture work (Fig. 3e), which did not differ significantly between genotypes at any age. Other measures of material quality from three-point bend analysis, calculated from the structural parameters normalized by the bone cross-section, including the bending modulus, yield, and ultimate stress, also showed no differences between Cre-negative controls and TβRIIocy−/− mice (Fig. S2).

The mechanical testing of young female bones produced results that matched those in previous reports,27 with no side effect of impaired osteocytic TGFβ signaling. Female Cre-negative controls showed the anticipated decreases in postyield displacement and work-to-fracture with age.11,50,51 Cre-negative control and TβRIIocy−/− bones did not significantly differ in whole bone stiffness, ultimate force, postyield displacement, or work of fracture at any age (Fig. 3f–i). For ultimate force and work of fracture, two-way ANOVA showed a significant main effect of genotype between TβRIIocy−/− females and Cre-negative controls regardless of age, but subsequent pairwise post hoc comparisons for each independent age group failed to reach the corrected significance level. The results of these complex macromechanical and μCT analyses reflect longitudinal responses to altered osteocytic TGFβ signaling with age.

Relationship of age and osteocytic TGFβ signaling in nanoscale material mechanisms

Since the complexity of tissue-scale analyses and the gross alterations to bone morphology obscure the role of osteocytic TGFβ signaling in regulating bone material properties with age, we pursued analyses at smaller scales to interrogate changes occurring in the principal components of bone: collagen and mineral. Leveraging differences in the nanoscale periodic structure of collagen and minerals, high-intensity monochromatic synchrotron-generated X-rays can discriminate their material behavior in in situ tensile mechanical tests (Fig. 4a).11,52,53 When incident X-rays pass through bone, the nanoscale molecular repeats of collagen fibrils scatter X-rays at a shallow angle, while the Angstrom scale spacings of the mineral lattice diffract X-rays at a wider angle. The resulting separation between small-angle and wide-angle X-ray (SAXS and WAXD) signals was captured, and changes in the signals during tensile testing allowed the strain to be calculated independently for the collagen and mineral phases. Whole tissue strain was calculated using visible light CCD images of the bone. Ideally, the sums of the strain slope fits (Table 2) from bone constituent materials, mainly collagen and mineral, would be approximately 1, indicating that 100% of the tissue strain is captured by the nanomaterials observed. This relationship is apparent in young control male and both control and TβRIIocy−/− female bone (Fig. 4b), where collagen and mineral slopes sum to approximately 1, indicating a stable mechanical composite in these cases. The appearance of young male Cre-negative bone to have a summed slope >1 likely results from the collection of data on separate contralateral bones instead of simultaneous collection on a single sample. However, a slope <1, as in young, male TβRIIocy−/− bone, can indicate losses to collagen, mineral, or other unobserved material-associated strains and a disconnect between material-scale and tissue-scale strain. The relative distribution of strain between the organic and mineralized components shows sex and genotype-dependent differences with age (Fig. S3).

Fig. 4

Synchrotron SAXS/WAXD shows altered collagen and mineral behavior in TβRIIocy−/− bone during in situ tensile testing. a Synchrotron-generated X-ray exposure during tensile testing of the forearm bones of TβRIIocy−/− and Cre-negative controls at multiple ages generates signals unique to organic collagen (SAXS) and the inorganic mineral components (WAXD) of bone based on their hierarchical size, order, and structure. The summation of composite strains (b) reveals deficiencies in the ability of collagen and minerals to carry tissue-level strains in young, male TβRIIocy−/− and aged, female Cre-negative controls. Compared to Cre-negative controls, young TβRIIocy−/− males showed a deficiency in collagen strain capacity (c) that recovered with age (d). Female TβRIIocy−/− mice did not show this weakness at young ages (e). With age, female Cre-negative controls showed a decrease in collagen strain, while TβRIIocy−/− did not, resulting in significant differences in collagen strain between these two groups (f). The mineral strains in young male bones showed genotype-dependent differences (g) that did not change with age (h), while females did not show genotype differences at either age studied (i, j). *P < 0.012 5 for 4-way Bonferroni correction in comparisons of regression slopes in an extra sum-of-squares F test. n = 5-8 per group

Table 2 Slopes of composite material strain vs. tissue strain

Analysis of the collagen and mineral material behavior yielded insight into the complex role of osteocytic TGFβ signaling in bone aging. SAXS revealed that collagen in young male TβRIIocy−/− bone is less able to carry strain than that in age-matched control bone (Fig. 4c). By 15 months, the collagen strain in male TβRIIocy−/− bone was statistically indistinguishable from that in the control bone (Fig. 4d), indicating a role for osteocyte-intrinsic TGFβ signaling in the age-related decline in the material quality of collagen.

The sexually dimorphic control of bone quality was also apparent in the SAXS analyses of both Cre-negative control and TβRIIocy−/− bone. Collagen in young female TβRIIocy−/− bone did not differ from that in the controls, which was consistent with the lack of other genotype-dependent differences in young females (Fig. 4e). Collagen strains in control aged female bone were almost half that of their younger counterparts, and therefore, the age-related loss of collagen strain is more severe in female than in male bone (Fig. S3a, c). In contrast, the collagen strain remained constant in TβRIIocy−/− female bone with age, resulting in a significant difference in collagen strain between control and TβRIIocy−/− female bone with age (Fig. 4f).

An evaluation of the spread of the scattering and diffraction peaks from X-ray scans by the second Legendre coefficient (P2) showed that prior to testing, collagen organization was lower by both age and TGFβ deficiency (Fig. S4A, B). Additionally, the change in P2 throughout mechanical testing revealed a small but significant inability for collagen in young TβRIIocy−/− males to realign (increase in P2) with applied strain, suggesting that collagen fibrils may slide against each other instead of engaging in straining mechanisms (Fig. S4C–F), a behavior not observed in other comparisons.

The WAXD results for contralateral bones showed genotype-dependent but not time-dependent changes in mineral strains in males, while the mineral material in female bone was unaffected by either time or osteocytic deficiency in TGFβ signaling. Young male TβRIIocy−/− bones once again showed significant losses to material strain capacity compared to that of Cre-negative controls (Fig. 4g), and these differences persisted with age (Fig. 4h). Female bone did not show significant genotype-dependent differences in mineral strains at either age (Fig. 4i, j). Together, the SAXS and WAXD results highlight distinct roles in male and female bone for osteocytic TGFβ signaling in the regulation of bone’s constituent materials and their behavior.

Loss of TGFβ signaling in young male osteocytes regulates cellular aging

Given that the losses to bone quality extend to the nanoscale, we evaluated several parameters in osteocytes, the cells responsible for the upkeep and management of bone material. We previously established that the suppression of osteocytic PLR in male TβRIIocy−/− bone was responsible for the loss of bone quality.26 Therefore, we examined the effect of aging on the expression of genes implicated in PLR26 and the dependence of age-related LCN degeneration on osteocytic TGFβ signaling. The expression of several genes implicated in osteocyte PLR showed age-related changes in WT control cortical bone (Fig. 5a). The expression of the collagenases Mmp2 and Mmp14 showed significant and lasting downregulation from 2 months through 2 ½ years of age, while Mmp13 showed a repression of transcriptional activity by 1 year, with some amount of recovery by 2 and 2 ½ years. Additionally, the expression of Ctsk, Acp5, and genes implicated in acidification followed similar patterns to Mmp13 in expression, showing repression at 1 year of age but full recovery by 2 ½ years.

Fig. 5

Osteocyte LCN integrity shows coordinated control by age and osteocytic TGFβ signaling. Osteocyte aging in WT C57/Bl6 control bone shows suppressed PLR, with an early and prolonged decline in the expression of MMP2 and MMP14, and dynamic expression of other factors implicated in PLR (a) analyzed with Fischer’s LSD test after one-way ANOVA. Visualization of the osteocyte LCN via fluorescence confocal microscopy in TβRIIocy−/− and Cre-negative controls with age (b, c) shows a genotype-specific degeneration of the LCN in males, with losses to canaliculi/osteocyte (d) and increased canalicular tortuosity (e) in TβRIIocy−/− bone that does not worsen with age. Cell death in males increased only with age (f). Females showed age-related changes in these parameters (h–i) in both genotypes. *P < 0.012 5 for 4-way Bonferroni corrected post hoc pairwise Fischer’s LSD test after two-way ANOVA. n = 4–5 per group

The age-related suppression of PLR protease expression and TGFβ signaling (Fig. 1c) corresponds to the previously reported degeneration of the LCN with age and osteocytic TGFβ deficiency.26,34 To determine the role of osteocyte-intrinsic TGFβ signaling, we examined the integrity of the osteocyte LCN within the cortical bone of young and aged TβRIIocy−/− mice and their age-matched Cre-negative controls in 3D with fluorescence confocal microscopy (Fig. 5b, c). Similar to prior reports of younger TβRIIocy−/− bone or bone from wild-type mice treated with a TβRI inhibitor,26,34 4-month-old male TβRIIocy−/− bone displayed significantly fewer canaliculi per osteocyte and increased canalicular tortuosity compared to age-matched Cre-negative controls (Fig. 5d, e). Male Cre-negative control bones displayed similar qualitative LCN changes over time, with a significant loss of canaliculi and an increase in canalicular tortuosity, which indicated LCN degeneration with age. The interaction term between aging and genotype was significant for the canalicular number (P = 0.000 9), highlighting the different effects of aging on the male genotypes and LCN integrity. Interestingly, the interaction for canalicular tortuosity was not significant (P = 0.221 7), but canalicular tortuosity did display important significant changes in age and genotype independently, indicating significant changes in both age and genotype on LCN structure. Importantly, however, TβRIIocy−/− male bone showed no pairwise differences between 4 months and 15 months in either canalicular number or tortuosity, emphasizing the similarity between the early LCN dysregulation in TβRIIocy−/− male bone and the time-dependent LCN degeneration in Cre-negative controls. In both Cre-negative controls and TβRIIocy−/− males, osteocyte death assessed by TUNEL staining (Fig. 5f) only showed age-related increases and did not differ by genotype at either age. Combined, these observations implicate the loss of TGFβ in age-related LCN degeneration in Cre-negative control bone.

With age, both control and TβRIIocy−/− females showed losses to canalicular number and increased canalicular tortuosity compared to those of their young Cre-negative controls, but no differences in these values was observed between the genotypes at the same ages (Fig. 5g, h). These differences were not reliant on osteocyte death, since the increase in osteocyte TUNEL staining with age was unaffected by genotype in either male or female bone (Fig. 5f, i). Therefore, the age-related decline in LCN integrity in male, but not female, bone relies on osteocytic TGFβ signaling.

Posttranslational modifications in TβRII ocy−/− bone collagen occur with altered enzymatic function

Given the role of osteocyte-intrinsic TGFβ signaling in the age-related decline in LCN and collagen integrity, we took several approaches to examine collagen posttranslational modifications (PTMs) and the osteocytic mechanisms controlling them. Collagen PTMs include those formed by lysyl oxidase (LOX) and prolyl 3 hydroxylases (P3H), help to direct collagen self-assembly, crosslinking, and fibril structure.54,55,56,57,58 Accordingly, collagen PTMs underpin bone quality and contribute to fragility in aging, diabetes, and osteogeneses imperfecta, where specific PTMs are suppressed.55,57,59

An RNA-seq analysis of WT male cortical bone showed great repression of several LOX and P3H family members with age (Fig. 6a, b). Relative to young bone, transcript levels for Lox, Loxl1, and Loxl2 displayed sustained repression up to 2 ½ years, whereas Loxl3 and Loxl4 levels were decreased at the 1-year timepoint. Of the P3H enzymes responsible for proline oxidation that enables tropocollagen helix formation, only P3H2 did not show significant and prolonged age-related transcriptional repression. Using RT‒qPCR, we investigated whether the expression of candidate collagen crosslinking enzymes was sensitive to changes in osteocytic TGFβ signaling with age. The Loxl3 mRNA levels were repressed at 4 and 15 months in male TβRIIocy−/− bone compared to age-matched Cre-negative controls (Fig. 6c), whereas females displayed only an age-related repression of Loxl3 expression in Cre-negative control mice (Fig. 6d), similar to the regulation of Loxl2 and Periostin (Fig. S5).

Fig. 6

Collagen quality is impacted by both age and osteocytic TGFβ signaling. RNA sequencing of C57/Bl6 (WT) mouse cortical bone showed age-dependent transcriptional repression of several lysyl oxidase (a) and prolyl 3 hydroxylase (b) isoforms responsible for posttranslational enzymatic collagen modifications *P < 0.008 wrt. 2 months, #P < 0.008 wrt. 1 yr. in 6-way Bonferroni corrected post hoc pairwise Fischer’s LSD test after one-way ANOVA, n = 3–4 per group. RT‒qPCR from TβRIIocy−/− and Cre-negative control bone revealed significant repression of Loxl3 mRNA in male (c) TβRIIocy−/− regardless of age, while females showed only an age-related decrease in Cre-negative controls (d). Quantification of nonenzymatic collagen crosslinking in the form of fluorescent advanced glycation end products (AGEs) shows an accelerated rate of accumulation in male and female TβRIIocy−/− bone with age compared to that