Abstract

   Bone loss and microstructural destruction in elderly men are associated
   with fractures and high mortality. While testosterone (Tes) is
   considered to be possibly protective, its regulatory mechanism in bone
   remodeling remains unclear. Here, bone microarchitectural analysis
   indicates that elderly men exhibit reduced cortical and trabecular
   thickness with elevated cortical porosity, particularly at the superior
   femoral head near the medial acetabulum. Serum profiling of 352
   individuals showed that low Tes levels (<9.415 nmol·L^−1) are
   associated with higher risk of bone loss. In vivo, tail‐suspended mice
   lacking osteoblastic androgen receptor (AR) displayed similar femoral
   deterioration, with decreased trabecular bone and increased cortical
   porosity. Mechanistically, Tes enhances osteoblastic differentiation
   via AR‐mediated upregulation of tenascin‐C (TNC). Molecular docking
   suggests the fibrinogen C‐terminal domain of TNC inhibits
   osteoclastogenesis by binding integrin αV, blocking adhesion of
   RGD‐containing proteins. A synthetic peptide (pep2) mimicking this
   domain preserved bone architecture in osteoblast‐specific Ar‐knockout,
   tail‐suspended mice. Moreover, elevated serum extracellular vesicle
   amyloid precursor protein, secondary to Tes‐AR‐TNC decline and
   osteoclast overactivation, emerged as a biomarker of bone loss when
   combined with low Tes. This study identifies the Tes‐AR‐TNC axis as a
   key regulator of male bone remodeling, offering insights into fracture
   risk assessment and targeted interventions in bone destruction.

   Keywords: bone destruction, male bones, osteoblastic androgen receptor,
   tenascin‐C
     __________________________________________________________________

   This study reveals that Testosterone–Androgen Receptor signaling delays
   elderly male bone destruction by upregulation of the osteoblastic
   extracellular tenascin‐C (TNC). The osteoprotective effect of
   fibrinogen C‐terminus of TNC is demonstrated in male osteoporotic mice
   model that osteoblast‐specific Ar‐knockout, potentially via inhibition
   of osteoclastogenesis by binding with integrin αV, which blocks the
   adhesion of pro‐osteoclastic proteins with RGD motif.

   graphic file with name ADVS-12-e01518-g007.jpg

1. Introduction

   Bone loss in elderly males is associated with elevated mortality due to
   fractures, although the overall prevalence of major fragility fractures
   in older men is approximately one‐third that in women.^[ [40]^1 ,
   [41]^2 ^] The cumulative mortality rate 1 year after hip fracture among
   men is as high as 20–40%, indicating that men are about twice as likely
   to die in hospital after a hip osteoporotic fracture as women, even
   after considering co‐morbidities, medications, and background
   mortality.^[ [42]^3 , [43]^4 , [44]^5 , [45]^6 ^] This is difficult to
   explain from the perspective of differences in the maintenance of
   skeletal homeostasis by sex hormones alone. In contrast to women, the
   majority of men do not typically experience overt hypogonadism as they
   age. It has been believed that androgen deficiency is not a significant
   factor in age‐related male diseases.^[ [46]^7 ^] However, a growing
   body of evidence indicates that low androgen levels are associated with
   bone loss in elderly men,^[ [47]^8 ^] suggesting that the regulatory
   role of androgen in aged male bones deserves more attention, and it is
   not enough to directly extrapolate to males what is known for females.
   Recent analysis with high‐resolution peripheral quantitative computed
   tomography demonstrated that bone microarchitecture was not compromised
   in trans‐men who received testosterone (Tes, a primary androgen)
   therapy and suppression of ovarian estradiol production, whereas
   trans‐women receiving 17‐β estradiol (β‐E2), with or without
   antiandrogen therapy, had deteriorated bone microarchitecture, implying
   that effects of estradiol were insufficient to counteract the reduced
   capacity of Tes to maintain bone strength,^[ [48]^9 ^] which emphasizes
   the pivotal role of Tes in protecting bone quality in men.^[ [49]^10 ^]
   Therefore, elucidating specific Tes‐mediated bone remodeling mechanisms
   in males is essential to develop precision interventions for the
   deterioration of bone mass and structure in elderly men, rather than
   relying solely on using general Tes replacement therapy (TRT).

   As the major testicular factor and best‐known androgen,^[ [50]^11 ,
   [51]^12 ^] Tes mediates signaling cascades by binding to its classical
   nuclear hormone receptor, androgen receptor (AR),^[ [52]^13 ^] which is
   widely expressed in bone cells, including osteoclasts,^[ [53]^14 ^]
   osteoblasts and their precursors.^[ [54]^15 , [55]^16 ^] Following
   orchiectomy (ORX) in mice, the rise in the RANKL/OPG mRNA ratio in bone
   tissue is counteracted by mechanical loading, suggesting that physical
   activity could mitigate bone loss in conditions of androgen deficiency
   by exerting an antiresorptive influence.^[ [56]^17 ^] While mice with
   specific deletion of AR in osteoclasts showing that androgen signaling
   within osteoclasts does not contribute to the bone‐preserving effects
   of androgens in trabecular or cortical bone compartments.^[ [57]^18 ^]
   By contrast, mice with specific deletion of AR in osteoblasts exhibited
   decreased trabecular volume and trabecular number (Tb.N) associated
   with an increase in osteoclast number; however, no effects on cortical
   bone mass and turnover were demonstrated. This evidence is consistent
   with clinical findings that the maintenance of Tes is associated with
   the protection of trabecular bone morphology and biomechanical
   properties.^[ [58]^19 , [59]^20 ^] Furthermore, the absence of changes
   in RANKL/OPG level due to AR deletion in osteoblasts and osteocytes
   implies that alternative mechanisms may account for the antiresorptive
   impact of the androgen‐AR signaling pathway on trabecular bone.^[
   [60]^21 ^] Thus, we inferred that the Tes attenuates osteoclastogenesis
   in trabecular bone indirectly via osteoblasts‐AR, while the
   spatiotemporal characteristics of bone deterioration in elderly men and
   the underlying mechanisms of Tes in the regulation of bone remodeling
   have not yet been elucidated.

   Among the four vertebrate tenascins, a role in skeletal function has
   been attributed to tenascin‐C (TNC), an extracellular matrix
   glycoprotein implicated in osteogenic differentiation and the process
   of bone mineralization.^[ [61]^22 ^] TNC typically adopts a
   characteristic hexameric structure, comprised of four distinct domains
   arranged from N‐terminus to C‐terminus as a coil‐like tenascin assembly
   domain, an epidermal growth factor‐like domain, a fibronectin type III
   (FN III)‐like domain, and a fibrinogen‐like globe (FBG) domain. This
   multimodular configuration facilitates the interaction of TNC with a
   broad array of diverse ligands, including annexin II, syndecan‐4, and
   integrins, via alternative splicing mechanisms. The precise regulation
   of distinct TNC cleavage is tightly controlled by the specific cell
   type and the microenvironment of the tissue,^[ [62]^23 ^] including
   biological factors and mechanical changes, each operating through
   unique signaling pathways, yet the reported control of TNC in bone
   remains poorly understood.^[ [63]^24 ^]

   In the present study, bone microarchitectural analysis indicated that
   age‐related bone destruction in the proximal femur of elderly males is
   mainly characterized by a decrease in the thickness of cortical and
   trabecular bone (with an increase in cortical porosity) in the superior
   portion of the femoral head close to the medial acetabulum. Serum
   detection from 352 individuals suggested that elderly men with low Tes
   (<9.415 nmol L^−1) have a high risk of bone loss. In vivo studies using
   tail‐suspended mice lacking AR in osteoblasts demonstrated similar
   structural changes to that in elderly men, that is, bone destruction at
   the top of the proximal femur with decreased amount and thickness of
   trabecular bone, and increased cortical porosity and trabecular
   separation compared with AR ^flox/Y mice. Further in vitro study showed
   that Tes activates osteoblast‐AR‐mediated upregulation of the
   extracellular TNC during osteoblast differentiation, which indirectly
   inhibits osteoclastogenesis and bone resorption. Molecular docking
   analysis using ColabFold indicated that the underlying negative
   regulatory effect of the C‐terminus of TNC on osteoclastogenesis occurs
   via binding with integrin αV (ITGαV), thereby blocking the adhesion of
   pro‐osteoclastic protein with RGD motif, such as OPN, to osteoclasts.
   Additionally, excess serum extracellular vesicle amyloid‐beta precursor
   protein (SEV‐APP), resulting from osteoclastic overactivation following
   decreased Tes‐AR‐mediated expression of the fibronectin C‐terminus of
   TNC, was identified to be a potential biomarker for early risk
   assessment of bone loss in elderly men (when combined with low Tes).
   This study reveals that the Tes‐TNC‐APP axis delays bone destruction in
   elderly men and highlights the potential value in risk assessment of
   bone loss and nonunion, in addition to precision interventions for
   deterioration of bone mass and structure.

2. Results

2.1. Bone Loss and Microstructural Destruction in the Superior Portion of the
Femoral Head in Elderly Males is Mainly Manifested by Decreases in the
Thickness of Cortical and Trabecular Bone, Accompanied by a Low Serum Tes
Level

   To understand the spatiotemporal characteristics of bone degeneration
   in elderly men, computed tomography (CT) images of the right hip of 10
   elderly men with normal bone mass and 10 elderly men with osteoporosis
   (OP) were included (both groups had equivalent baseline age and body
   size) (Figure [64]S1A, Table [65]S1, Supporting Information). The
   medial part of the trochanter and the area above the lesser trochanter
   of the proximal femur were determined as the volume of interest (VOI)
   using a standardized process as described in Methods. Compared with
   elderly men with normal bone mass, elderly men with OP showed a lower
   BV/TV in the proximal femur. For cortical bone, cortical thickness
   (Ct.Th) was significantly lower in the medial part of the femoral head
   of elderly men with OP, simultaneously appearing with higher cortical
   porosity (Ct.Po), representing the thinning and discontinuity of the
   cortical bone, which determines mechanical strength and resistance to
   external impact. For trabecular bone, a significantly lower trabecular
   thickness (Tb.Th) was observed at the top position of the femoral head
   for elderly men with OP compared with elderly men with normal bone mass
   (almost 2.5 times), while the trabecular number (Tb.N) and trabecular
   separation (Tb.Sp) showed no significant difference between elderly men
   with and without OP. Contrary to our hypothesis, the cortical area
   fraction (Ct.Ar/Tt.Ar) of elderly men with OP was slightly higher than
   those of elderly men with normal bone mass. Perhaps this result is
   related to our limited resolution, but it indicates that the bone loss
   and microstructural destruction in the proximal femur of elderly men
   are likely concentrated at the superior portion of femoral head, close
   to the contact site with the medial part of the acetabulum (the socket
   of the hip joint), the specific “load zone” on the surface of the
   femoral head that experiences the highest compressive forces and shear
   stresses during daily activities, affecting the fixation effect and
   strength of internal fixation for femoral neck fractures (Figure
   [66]1A; Figure [67]S1B,C, Supporting Information).

Figure 1.

   Figure 1
   [68]Open in a new tab

   The thickness of cortical and trabecular bone in the superior portion
   of the femoral head decreases in elderly men with low serum Tes levels.
   A) 3D reconstruction and bone analysis of CT images of the right hip
   from elderly men, n = 10 per group; β‐CTx and PINP in serum of 97
   elderly men with osteoporosis and 86 elderly men with normal BMD. B)
   Comparisons of Tes level in serum among young men with normal BMD
   (Y‐NC, n = 105), elderly men with normal BMD (E‐NC, n = 86), elderly
   men with osteopenia (E‐ON, n = 64), and elderly men with osteoporosis
   (E‐OP, n = 97). C) Spearman's correlation analysis between serum Tes
   level and DXA‐BMD of one‐third distal radius in elderly men (n = 247).
   (D) ROC curve analysis of serum Tes level (n = 86 for normal BMD;
   n = 161 for bone loss). Student's t‐test was used for two groups of
   comparisons, one‐way ANOVA with Tukey's multiple comparisons test was
   used for multiple comparisons. All tests were two‐sided; **p < 0.01;
   ***p < 0.001; ns = no significance. Tes = testosterone; DXA =
   dual‐energy X‐ray absorptiometry; BMD = bone mineral density; AUC =
   area under the curve.

   Serum samples were collected from 352 individuals to investigate the
   association between bone mineral density (BMD) and serum Tes, including
   105 young men with normal BMD (Y‐NC), 86 elderly men with normal BMD
   (E‐NC), 64 elderly men with osteopenia (E‐ON), and 97 elderly men with
   OP (E‐OP) (Table [69]S1, Supporting Information). Serum levels of Tes,
   the bone resorption marker type I collagen C‐telopeptide‐related
   fraction (β‐CTx), and the bone formation marker procollagen type 1
   amino‐terminal propeptide (PINP) were measured using Elisa. BMD at the
   one‐third distal radius was assessed using Dual‐energy X‐ray
   absorptiometry (DXA). The results showed that β‐CTx levels increased
   with aging. Compared with elderly men with normal BMD, β‐CTx were
   significantly elevated in both the E‐ON and E‐OP groups, while no
   significant difference was observed between the E‐ON and E‐OP. In
   contrast, PINP levels remained relatively stable (Figure [70]1A; Figure
   [71]S1D, Supporting Information). Compared with elderly men with normal
   BMD, serum Tes declined significantly in the E‐ON and E‐OP groups.
   There was no significant difference between serum Tes levels in E‐ON
   and E‐OP. The serum Tes level showed no significant difference between
   young men and elderly healthy controls (Figure [72]1B). Linear
   regression analysis revealed a positive correlation between BMD of the
   one‐third distal radius and serum Tes levels in elderly men (n = 247,
   Spearman's correlation, r = 0.3, p < 0.001), whereas no significant
   correlation was found in young men (n = 105, Spearman's correlation,
   r = 0.1, p = 0.48), indicating that the association between serum Tes
   level and bone loss is more significant in elderly men (Figure [73]1C;
   Figure [74]S1E, Supporting Information). Receiver operating
   characteristic (ROC) curve analysis demonstrated an area under the
   curve (AUC) of 0.65 (95% CI 0.57–0.74, n = 86 for healthy controls,
   n = 161 for bone loss) (Figure [75]1D), suggesting that serum Tes
   levels may have potential for predicting the risk of bone loss in
   elderly men. In addition, we analyzed the correlation between serum Tes
   levels and heel BMD measured by quantitative ultrasound (QUS) in 3,097
   individuals from the UK biobank database. Although QUS is less accurate
   than DXA for BMD assessment, a positive correlation was observed
   between QUS‐BMD and serum Tes levels (Figure [76]S1F, Supporting
   Information), consistent with the findings in Figure [77]1C.

   These results highlight the significance of elucidating the mechanism
   of Tes‐regulated bone remodeling for the early diagnosis and timely
   treatment for male bone loss.

2.2. Tes Stimulates Osteoblast Differentiation and Inhibits Osteoclast
Formation, Mainly through Indirect Effects of Osteoblast‐Derived
Extracellular Secretions

   To investigate the effects of Tes on osteoblast and osteoclast
   differentiation, an adult male physiological serum concentration of Tes
   (10^−8 m) was added to MC3T3E1 (osteoblast precursor, pre‐OB) or
   RAW264.7 (osteoclast precursor, a monocyte) cells during their
   differentiation process.^[ [78]^25 , [79]^26 , [80]^27 ^] The OD[405]
   value of Alizarin Red‐positive osteoblasts in cells cultured with Tes
   was significantly higher than that in controls cultured with DMSO on
   both Day 15 (metaphase stage) and Day 18 (mature stage) (Figure
   [81]2A). These results indicate that Tes continuously promotes
   osteoblast differentiation with increased production of collagen.

Figure 2.

   Figure 2
   [82]Open in a new tab

   Tes stimulates osteoblast differentiation and inhibits osteoclast
   formation, mainly through indirect effects of osteoblast‐derived
   extracellular secretions. A) Representative images and quantification
   of Alizarin Red staining of MC3T3E1 (osteoblast precursor) cells after
   15 and 18 days of osteogenic differentiation in the presence of Tes.
   B,C) Representative images and quantification of TRAP staining (B) and
   bone resorption (C) of osteoclasts treated with Tes compared with
   controls (DMSO). (D) Schematic illustration of the experimental design
   for Figure [83]2E,F. E,F) Representative images and quantification of
   TRAP staining (E) and bone resorption (F) of osteoclasts treated with
   MC3T3E1‐derived GM or DM supernatant. All results are representative of
   data from at least three independent experiments. Student's t‐test was
   used for two groups of comparisons, one‐way ANOVA with Tukey's multiple
   comparisons test was used for multiple comparisons. All tests were
   two‐sided; *p < 0.05; **p < 0.01; ***p < 0.001; ns = no significance.
   Tes = testosterone; β‐GP = β‐glycerophosphate; LAA = L‐ascorbic acid;
   OC = osteoclast; GM = general medium derived from MC3T3E1 culture
   supernatant; DM = differentiation medium derived from MC3T3E1 culture
   supernatant. Scale bar, 200 µm.

   After 5 days of RANKL induction, the formation of mature osteoclasts
   (>3 nuclei) was confirmed by TRAP staining. The number and
   rearrangement area of TRAP‐positive (TRAP^+) multinucleated osteoclasts
   cultured in the presence of Tes were lower than those of cells cultured
   with DMSO (Figure [84]2B; Figure [85]S2A, Supporting Information).
   Additionally, the area of the pits resorbed on Osteo Assay Surface
   Plates by TRAP^+ osteoclasts cultured with Tes was lower than that for
   cells cultured with DMSO (Figure [86]2C). Tes exerted moderate
   inhibitory effects on the phenotypes and bone‐resorptive function of
   RANKL‐induced osteoclasts.

   The interaction between osteoblasts and osteoclasts plays a significant
   regulatory role in bone remodeling. Osteoblasts can regulate the
   formation of osteoclasts by secreting various cytokines.^[ [87]^28 ^]
   The direct inhibitory effect of Tes on osteoclasts was relatively mild
   (Figure [88]2B,C). We speculate whether Tes regulates osteoclast
   formation mainly through osteoblasts. The Tes‐free cell supernatant of
   Tes‐stimulated MC3T3E1 was collected, defined as GM. The Tes‐free cell
   supernatant of Tes‐stimulated osteoblasts (differentiated from MC3T3E1
   cells induced by β‐GP and LAA) at Days 15 and 18 of the differentiation
   process were separately collected and mixed, defined as DM. To study
   the effect of Tes through osteoblasts on osteoclasts, GM and DM were
   used as a conditional medium to cultivate osteoclasts (differentiated
   from RAW264.7 cells induced by RANKL) (Figure [89]2D). Under the
   condition that GM was used as the conditional medium, Tes‐stimulated
   MC3T3E1 (per‐OB) had no significant impact on the formation of
   osteoclasts. Under the condition that DM was used as the conditional
   medium, Tes‐stimulated osteoblasts significantly suppressed the
   formation of osteoclasts, as the numbers of TRAP^+ osteoclasts and the
   rearrangement area were lower (Figure [90]2E; Figure [91]S2B,
   Supporting Information). The area of the pits resorbed on Osteo Assay
   Surface Plates was smaller (Figure [92]2F). Taken together, our data
   support a dual role of Tes in protecting bone quality. On one hand, Tes
   promotes osteoblastic differentiation, and on the other hand, it
   primarily exerts an inhibitory effect on osteoclasts formation through
   the indirect effects of osteoblast‐derived extracellular secretions.

2.3. Male Mice Lacking AR in Osteoblasts Show Bone Loss and Microstructure
Destruction

   Based on our in vitro findings regarding the indirect effect of
   Tes‐stimulated osteoblastic inhibition on osteoclasts, Ocn‐Cre‐driver
   (targeting osteoblasts) Ar‐knockout (Ocn‐Ar^−/Y ) male mice were
   generated to simulate a low Tes‐AR effect on osteoblast in vivo (Figure
   [93]3A). PCR verified the presence of the floxed allele and the Cre
   transgene in the mouse genotype. Western blot verified Ar knockout
   specifically in osteoblasts (Figure [94]S3A,B, Supporting Information).
   The body weight increased with age and showed no significant difference
   between Ocn‐Ar^−/Y male mice and Ar^flox/Y controls (Figure [95]S3C,
   Supporting Information). However, the BV/TV significantly declined in
   12‐month‐old Ocn‐Ar^−/Y male mice compared with Ar^flox/Y controls
   (declined 4.4%), and the difference was greater at 16 months of age
   (declined 13.5%). No significant difference was shown between younger
   (aged 2 and 6 months old) Ocn‐Ar^−/Y male mice and Ar^flox/Y controls
   (Figure [96]S3D, Supporting Information). These results demonstrate Tes
   plays a more important role in older mice, which is consistent with the
   conclusion that Tes is positively related to BMD in elderly men
   (Figure [97]1C).

Figure 3.

   Figure 3
   [98]Open in a new tab

   Androgen receptor (Ar)‐deficiency in osteoblasts induces bone loss and
   accelerates bone destruction in mice. A) Schematic showing the strategy
   to generate male osteoblast‐specific Ar‐knockout mice (Ocn‐Ar^−/Y ). B)
   3D reconstruction and bone analysis of micro‐CT images of the proximal
   femur, cortical bone, and trabecular bone from femurs of Ocn‐Ar^−/Y
   mice (male, 12‐month‐old) with or without tail suspension; n = 5 per
   group. C) Three‐point bending test, performed to assess the bone
   strength of the proximal femur of Ocn‐Ar^−/Y mice (male, 12‐month‐old)
   with or without tail suspension; n = 5 per group. One‐way ANOVA with
   Tukey's multiple comparisons test was used for multiple comparisons.
   All tests were two‐sided; *p < 0.05; **p < 0.01; ***p < 0.001.
   N = without tail suspension; TS = with tail suspension.

   A microgravity OP model was constructed with mice tail suspended for
   4 weeks. To clarify the role of Tes through osteoblast‐AR in bone
   remodeling and bone loss, we not only performed bone analysis on the
   femurs of Ocn‐Ar^−/Y male mice but also analyzed the femurs of male
   mice from a new constructed OP model by subjecting the Ocn‐Ar^−/Y male
   mice to tail suspension for 4 weeks, which resulted in significant bone
   loss and microstructure destruction. The mice were divided into four
   groups: Ar^flox/Y mice (N‐Ar^flox/Y ), Ocn‐Ar^−/Y mice (N‐Ocn‐Ar^−/Y ),
   Ar^flox/Y mice with tail suspension (TS‐ Ar^flox/Y ), and Ocn‐Ar^−/Y
   mice with tail suspension (TS‐Ocn‐Ar^−/Y ). All mice were 12‐month‐old
   male mice (n = 5 per group) and had equivalent baselines of body length
   and weight (Figure [99]S3E, Supporting Information). The same standard
   for VOI determination (i.e., the medial part of the trochanter and the
   area above the lesser trochanter of the femur) as for the hip of
   elderly men was used in micro‐CT images from the proximal femur of the
   mice. Dragonfly bone analysis showed a significant decrease in BV/TV,
   Ct.Ar/Tt.Ar, Ct.Th, Tb.N and Tb.Th, while showing an increase in Ct.Po
   and Tb.Sp in Ocn‐Ar^−/Y mice. This revealed that Ar‐knockout in
   osteoblasts induced bone loss, decreased the number and thickness of
   trabecular bone, and increased cortical porosity and trabecular
   separation at the top position of the proximal femur, which was
   consistent with the results of Notini^[ [100]^29 ^] And Chiang.^[
   [101]^30 ^] Furthermore, these differences were more significant in
   TS‐Ocn‐Ar^−/Y mice (Figure [102]3B), suggesting that
   osteoblast‐specific Ar knockout combined with tail‐suspension may be an
   optimized animal model for studying OP in elderly men. The three‐point
   bending test was performed to evaluate the mechanical characteristics
   of the proximal femur based on structural changes. The ultimate force,
   stiffness, elastic modulus, and ultimate stress of TS‐Ocn‐Ar^−/Y mice
   were significantly reduced compared with N‐Ar^flox/Y controls, and
   their differences were greater than that between N‐Ocn‐Ar^−/Y mice with
   N‐Ar^flox/Y controls (Figure [103]3C).

2.4. Extracellular Secreted Proteins TNC is Significantly Upregulated during
Tes‐Stimulated Osteoblast Differentiation

   As the above findings showed that an indirect inhibitory effect on
   osteoclast formation through osteoblast‐derived extracellular
   secretions, tandem mass tag (TMT)‐based quantitative mass spectrometry
   (MS) was conducted to compare protein changes in pre‐OB, OB, and
   Tes‐stimulated OB (OB‐T), screening key extracellular proteins secreted
   from osteoblasts. Two proteomics experiments were performed separately
   using the OB and OB‐T cells collected on Day 15 and Day 18 of the
   osteoblast differentiation process (Figure [104]4A). A total of 8166
   proteins were identified in the Day 15 experiment. Among these
   proteins, 374 differentially expressed proteins (DEPs) were filtered
   (Figure [105]S4A, Table [106]S2, Supporting Information). A total of
   9246 proteins were identified in the Day 18 experiment, and 368 DEPs
   were filtered (Figure [107]S4A, Table [108]S3, Supporting Information).
   Gene Ontology analysis was conducted on the DEPs of Day 15 and Day 18.
   The DEPs were predominantly enriched in the extracellular region, and
   responsible for identical protein binding and cell adhesion
   (Figure [109]4B; Figure [110]S4B, Supporting Information). Hierarchical
   clustering heatmaps were also constructed and information on proteins
   in each cluster is shown in Tables [111]S2–S3 (Supporting Information).
   Protein expression trends in clusters 6 and 7 of Day 15 and cluster 4
   of Day 18 showed a sequential increase from pre‐OB to OB, and then to
   OB‐T, which implies these proteins play important roles in OB
   differentiation mediated by Tes (Figure [112]4C; Figure [113]S4C,
   Supporting Information). These proteins in Day 15 were predominantly
   enriched in extracellular regions (Figure [114]4D) and had tight
   interactions, with five upregulated extracellular proteins identified
   as central proteins in the interaction network using the cytoHubba
   plugin of Cytoscape (BottleNeck algorithm, Figure [115]4E). Among the
   five central proteins, TNC was also identified in the protein
   interaction network of DEPs in cluster 4 of Day 18 (Figure [116]S4C,
   Supporting Information). 61 DEPs overlapped between Tes‐stimulated
   osteogenesis of Day 15 and 18 (Figure [117]4F), which were mainly
   enriched in the negative regulation of megakaryocyte differentiation
   (Figure [118]4G, Table [119]S4, Supporting Information). Further
   hierarchical clustering heatmaps indicated that the DEPs in cluster 3
   of overlapped 61 DEPs were consistently upregulated (Figure [120]4H),
   with TNC characterized by a high score and a large number of unique
   peptides (Figure [121]4I). These results suggest that TNC is likely a
   key regulatory factor secreted by Tes‐stimulated osteoblasts.

Figure 4.

   Figure 4
   [122]Open in a new tab

   Extracellular secreted protein TNC is significantly upregulated during
   Tes‐stimulated osteoblast differentiation. A) Flowchart of proteomic
   experiments. B,C) GO analysis (B) and hierarchical clustering heatmaps
   (C) of DEPs at day 15 in the process of osteoblast differentiation and
   Tes‐stimulated osteoblast differentiation. D) Cellular component
   analysis of Clusters 6 and 7 from panel C. E) STRING analysis of the
   DEPs from the extracellular region, extracellular space, and
   extracellular matrix. The center five proteins are hub proteins in the
   network using the BottleNeck algorithm of the Hubba plug‐in in
   Cytoscape software. F) Venn diagram showing overlapping DEPs for
   Tes‐mediated osteogenesis on days 15 and 18. G) Kyoto Encyclopedia of
   Genes and Genomes enrichment analysis of the 61 overlapping DEPs. H)
   Hierarchical clustering heatmaps of the 61 overlapping DEPs. I) Mass
   spectroscopy information for Cluster 3 proteins from panel (H). DEPs =
   differentially expressed proteins.

2.5. Tes Stimulates Upregulation Of TNC through AR of Osteoblasts

   The upregulation in TNC expression in OB and OB‐T was verified from the
   peptide abundances, which were indicated by the intensities of reporter
   ions (Figure [123]5A). The serum TNC in elderly men with OP was
   significantly lower than that in elderly men with normal BMD
   (Figure [124]5B) and positively correlated with serum Tes
   (Figure [125]5C). The serum TNC in Ocn‐Ar^−/Y male mice was lower
   compared with Ar^flox/Y controls (Figure [126]5D), indicating the Tes
   stimulates upregulation of extracellular TNC through AR of osteoblasts.
   To verify whether Tes plays a role in the upregulation of extracellular
   TNC in osteoblasts, supernatants from Days 9, 12, and 15 in the process
   of osteogenic differentiation were collected. The ALP level of
   osteoblasts gradually increased beginning on Day 9, and showing that
   osteoblast differentiation with Tes was higher than that in the control
   group (Figure [127]S5A, Supporting Information). ELISA identified that
   TNC in the cell supernatant increased significantly after Tes treatment
   of osteoblasts, reaching its peak on Day 12; the secretion level
   decreased on Day 15, but it was still higher in the Tes‐stimulated
   group than in the control group (Figure [128]5E). Western blot with an
   antibody targeting the C‐terminus of TNC showed that expression of TNC
   in the supernatant of Tes‐stimulated osteoblasts was increased, which
   was mainly manifested by a band of 53 kDa (Figure [129]5F; Figure
   [130]S5B, Supporting Information). To verify that the primary source of
   TNC is osteoblasts in bone marrow, we performed Immunofluorescence (IF)
   co‐staining for TNC and OCN in mice femoral head. Co‐staining analysis
   revealed a significant co‐localization of TNC with OCN^+ osteoblasts
   and overlapped with bone trabeculae area. TNC was similarly highly
   expressed and localized at the bone marrow cavity, suggesting that TNC
   may act as a secreted protein that is produced by osteoblasts and
   secreted into the bone marrow (Figure [131]S5C, Supporting
   Information). ScRNA‐seq data from the NCBI Gene expression omnibus
   database ([132]GSE253355)^[ [133]^31 ^] analysis confirmed single
   “OCN^+ Osteoblast” cell had the highest level of TNC expression among
   35 clusters in the bone marrow (Osteo‐MSC, Fibro‐MSC, etc.), supporting
   that “OCN^+ Osteoblast” is one of the primary source of TNC in bone
   marrow (Figure [134]S5D,E, Supporting Information).

Figure 5.

   Figure 5
   [135]Open in a new tab

   Tes stimulates upregulation of TNC through AR of osteoblast. A)
   Representative mass spectrum data of TNC. The intensity of TMT
   precursor ions was used for protein quantification. B) TNC levels in
   the serum of elderly men with normal BMD (n = 19) and elderly men with
   OP (n = 43) by ELISA. C) Correlation analysis between TNC and Tes in
   the serum in elderly men (n = 62). D) TNC levels in the serum of
   Ar^flox/Y (n = 10) and Ocn‐Ar^−/Y (n = 12) male mice aged 12 months old
   by ELISA. E) TNC levels in the cell supernatant of MC3T3E1 on days 9,
   12, and 15 in the process of osteogenic differentiation, determined by
   ELISA to explore the effect of Tes on TNC. F) TNC levels in the cell
   supernatant of MC3T3E1 on day 15 in the process of osteogenic
   differentiation detected by Western blot with an antibody targeting the
   C‐terminus of TNC. The Ponceau S staining was used to verify the sample
   loading was equal. G) Predicted AR binding sequence in TNC (from the
   JASPAR website). H) ChIP‐qPCR detection of the binding of AR of
   different regions of TNC promoter in osteoblast. I) Luciferase reporter
   assay was performed to determine the transcriptional activation of TNC
   regulated by AR. J) ChIP‐qPCR detection of the binding of AR in
   Tes‐stimulated osteoblasts and vehicle‐treated control. K) TNC levels
   in the cell supernatant of MC3T3E1 in the process of osteogenic
   differentiation were determined by ELISA. Student's t‐test was used for
   two groups of comparisons, all tests were two‐sided; *p < 0.05;
   **p < 0.01; ***p < 0.001. E‐NC = elderly men with normal BMD; E‐OP =
   elderly men with osteoporosis; Tes = testosterone.

   We explored the mechanism of Tes regulation of TNC. QPCR showed that
   TNC mRNA levels in Tes‐stimulated osteoblasts were higher than the
   control group (DMSO), indicating Tes regulated the expression of TNC at
   the transcriptional level (Figure [136]S5F, Supporting Information).
   The Jaspar website predicted that AR binding to the TNC promoter
   utilizes an AAGTACAGGCAGTCC motif with a score of 14.7
   (Figure [137]5G). Using ChIP‐qPCR and dual‐luciferase assay, we found
   AR bound to region 1 of the TNC promoter (Figure [138]5H, Text [139]S1,
   Supporting Information) and increased the TNC promoter activity
   (Figure [140]5I). In order to verify the recruitment of AR to region 1
   of the TNC promoter was regulated by Tes, we compared the enrichment of
   region 1 in Tes‐stimulated osteoblasts with vehicle‐treated control.
   The result showed that AR in Tes‐stimulated osteoblasts bound more DNA
   fragments derived from region 1 of the TNC promoter (Figure [141]5J).
   By silencing of AR in osteoblasts using siRNA (Figure [142]S5G,
   Supporting Information), the expression of extracellular TNC on Days 9,
   12, and 15 of osteogenic differentiation significantly decreased in the
   cell supernatant (Figure [143]5K). These results suggest that the
   Tes‐stimulated upregulation of TNC is through binding with its receptor
   AR.

2.6. The Fibronectin C‐Terminus of TNC Inhibits the Differentiation of
Osteoclasts by Binding to Integrin αV and has Therapeutic Effect on
Osteoporotic Male Mice

   TNC has previously been reported to bind to ITGαV‐ITGβ3, which is a
   transmembrane protein with elevated expression in mature osteoclast
   differentiation and is believed to bind to non‐collagenous
   extracellular matrix proteins with RGD motif, such as OPN, to promote
   the osteoclast differentiation by enhancing adhesion and migration of
   osteoclast precursor cells on bone absorption site.^[ [144]^32 ,
   [145]^33 ^] The RGD motif is replaced by RVD in mouse TNC.^[ [146]^34
   ^] There are two RVD motifs in the mouse TNC amino acid sequence,
   located in the fibronectin type‐III 3 (FIII‐3, TNC ΔA805‐D894) domain
   and the fibrinogen C‐terminal (TNC‐C, TNC ΔG1884‐N2099) domain
   respectively (Figure [147]6A). To determine the effects of these two
   domains on osteoclast formation, we purified recombinant FIII‐3
   fragment and TNC‐C fragment from Escherichia coli and used them to
   treat RAW264.7 during the process of osteoclast differentiation. TRAP
   staining showed that the TNC‐C fragment significantly inhibited the
   fusion area and number of osteoclasts compared with the FIII‐3 fragment
   (Figure [148]6B). In addition, we used an antibody targeting the
   C‐terminus of TNC to block the TNC‐C in the supernatant from
   Tes‐stimulated osteoblast. The ability of osteoclast differentiation
   was restored, though it remained lower than that of the group without
   TES‐stimulated osteoblast supernatant, suggesting that other factors in
   the osteoblast supernatant also play a role in inhibiting osteoclast
   differentiation, which may require further exploration in the future
   (Figure [149]S6A, Supporting Information). Immunoprecipitation assays
   showed a decrease in the binding of OPN to integrin αV after
   stimulation with the TNC‐C at various stages of osteoclast
   differentiation, including RAW264.7 cells, Pre‐OCs (40‐h RANKL‐induced
   RAW264.7 cells), and OCs, which support the C‐terminus of TNC competes
   with OPN for binding to integrin αV on osteoclasts, thus blocking the
   adhesion of pro‐osteoclastic proteins (Figure [150]S6B, Supporting
   Information). To identify the mechanism by which the TNC‐C inhibits
   osteoclast fusion and key amino acids, ColabFold was used to simulate
   the docking of the TNC‐C domain and the extracellular region of
   ITGαV‐ITGβ3. Two docking models were obtained. One model showed the
   residue Ser2040 of TNC bound to Gln244 of ITGαV and Asn2057 of TNC
   bound to Ala243 of ITGαV. The two binding sites, residue Ser2040 and
   Asn2057 of TNC were conserved (Figure [151]6C). The other model showed
   the residue Arg2007 of TNC bound to Asp180 of ITGαV, Ala2041, and
   Ile2042 of TNC bound to Ala245 of ITGαV (Figure [152]S6C, Supporting
   Information). Based on the results of docking, two peptides including
   the binding sites were synthesized. Pep1 contains 43 amino acids,
   located from D2003 to C2045 of TNC protein. Pep2 contains 30 amino
   acids, located from D2035 to M2064 of TNC protein. TRAP staining
   demonstrated pep2 is more effective than pep1 as an inhibitor of
   osteoclast differentiation (Figure [153]6D). IF staining results showed
   the colocalization of FITC‐pep2 peptide and ITGαVβ3 (Pearson's
   correlation coefficient was 0.56 in RAW264.7 and 0.34 in OC),
   indicating FITC‐pep2 peptide could bound to the ITGαVβ3
   (Figure [154]6E).

Figure 6.

   Figure 6
   [155]Open in a new tab

   The fibronectin C‐terminus (residues Gly1884‐Asn2099) of TNC (TNC‐C)
   binds to integrin αV and significantly inhibits the differentiation of
   osteoclasts. A) Domain structures for TNC. B) TRAP staining was used to
   detect the inhibition effect of FIII‐3 and TNC‐C on the osteoclast
   formation. C) Overview of ColabFold docking results for TNC‐C
   (Gly1884‐Asn2099, orange) with integrins αV (Phe31‐Pro983, blue) and β3
   (Asn28‐Asp717, grey) and conservation analysis of the predicted binding
   sites. D) TRAP staining was used to detect the inhibition effect of
   peptides from two predicted docking models on osteoclast formation. E)
   Detection of the localization of FITC‐pep2 and ITGαVβ3 by IF.
   Student's t‐test was used for two groups of comparisons, one‐way ANOVA
   with Tukey's multiple comparisons test was used for multiple
   comparisons. All tests were two‐sided; *p < 0.05; ***p < 0.001; ns = no
   significance. OC = osteoclast; BF = bright field.

   Mutations were introduced at the two binding sites in pep2 predicted by
   docking. The mutant peptides showed about a 20% reduction in the
   ability to inhibit osteoclast fusion and about a 50% reduction in the
   ability to reduce osteoclast formation (Figure [156]7A). This suggests
   Ser2040 and Asn2057 of TNC blocks the binding of other proteins with
   RGD sequences, such as OPN, to ITGαV‐ITGβ3, thereby inhibiting
   osteoclast differentiation. In order to explore the feasibility of pep2
   in treating OP with low Tes‐AR signaling, 12‐month‐old mice were
   injected subcutaneously with 80 µg kg^−1 day^−1 pep2 or PBS, 5 days per
   week for 6 weeks (Figure [157]S7, Supporting Information). Micro‐CT and
   bone analysis were performed to evaluate the therapeutic efficacy of
   pep2. For both the osteoblast‐specific Ar‐knockout OP mice model and
   the osteoblast‐specific Ar‐knockout combined with tail suspension OP
   mice model, the femur of the pep2‐treated mice had greater BV/TV,
   Ct.Ar/Tt.Ar, Ct.Th, Tb.N and Tb.Th than those of the Ocn‐Ar^−/Y
   controls (Figure [158]7B,C; Figure [159]S7B,C, Supporting Information).
   To assess the therapeutic effect of pep2 on bone turnover, serum
   markers of bone resorption and bone formation were analyzed. The
   results showed a significant reduction in β‐CTx serum level after pep2
   treatment. Additionally, serum levels of the bone formation marker PINP
   were significantly increased (Figure [160]7D,E). These data
   demonstrating the pep2 therapeutic effect on OP, further investigation
   of the pharmacokinetics of the TNC peptides would help to further
   clarify the dose and course of polypeptide therapy.

Figure 7.

   Figure 7
   [161]Open in a new tab

   The therapeutic effect of TNC peptides on male elderly osteoporotic
   mice. A) Evaluation of the importance of the binding sites by detecting
   the inhibitory effects of mutant peptides on osteoclast formation. B,C)
   3D reconstruction and bone analysis of micro‐CT images of the femur of
   Ocn‐Ar^−/Y mice (male, 14‐month‐old) with or without tail suspension to
   determine the therapeutic effect of TNC peptides. The serum levels of
   β‐CTx and PINP in osteoblast‐specific Ar‐knockout OP mice model (D) and
   osteoblast‐specific Ar‐knockout combined with tail suspension OP mice
   model (E); n = 6 per group. Student's t‐test was used for two groups of
   comparisons, one‐way ANOVA with Tukey's multiple comparisons test was
   used for multiple comparisons. All tests were two‐sided; *p < 0.05;
   **p < 0.01; ***p < 0.001.

2.7. SEV‐APP Combined with Tes Level is a Potential Biomarker in Serum for
Early Risk Assessment of OP in Elderly Men

   In addition to the potential treatment of TNC in OP, could TNC also be
   useful as a biomarker for the auxiliary diagnosis of OP? The ROC curve
   analysis showed that TNC was not effective in diagnosing OP (Figure
   [162]S8A, Supporting Information, p = 0.05). In order to identified
   effective biomarkers, the cell lysates of osteoclasts (NC), 10 ng mL^−1
   (con1), and 20 ng mL^−1 (con2) pep2 peptide‐stimulated osteoclasts were
   collected to performed the proteomics. A total of 5371 high‐confidence
   proteins were identified (Table [163]S5, Supporting Information). By
   defining proteins with a ratio greater than 1 as an upregulated trend,
   and conversely, proteins with a ratio less than 1 as a downregulated
   trend, the identified proteins were divided into 4 Clusters based on
   the change trend (Figure [164]8A). Cluster 1 comprised 972 proteins
   that exhibited a pep2 concentration‐dependent downregulation pattern.
   This consistent dose‐response relationship suggests these proteins are
   likely regulated by TNC and may participate in osteoclast
   differentiation and function. Our previous proteomics study found that
   208 proteins in serum‐derived extracellular vesicles (SEV) were
   increased in elderly men with OP.^[ [165]^35 ^] The Venn diagram showed
   there were 6 proteins overlapped (Figure [166]8B). Among the 6
   proteins, APP (protein identification confident score is 194) and Hmgb2
   (protein identification confident score is 15) were enriched in the
   RAGE receptor binding network. Several studies demonstrated that RAGE
   (abbreviation of the receptor for advanced glycation end products)
   plays a role in bone homeostasis and can contribute to the onset of
   several bone‐related diseases, including OP.^[ [167]^36 ^] The
   protein–protein interaction analyzed by STRING showed APP occupied a
   central position in the network (Figure [168]8C). The result of the
   western blot indicated the expression of APP in pep2‐stimulated
   osteoclasts was significant reduced (Figure [169]8D). We also used
   ELISA to detect the APP in serum EV of mice after pep2 treatment. The
   results showed that regardless of whether combined with tail
   suspension, the APP in serum EV of osteoblast‐specific AR knockout mice
   were increased and decreased significantly after treatment with pep2
   peptide (Figure [170]S8B,C, Supporting Information). The result of
   ELISA showed the SEV‐APP level gradually increased in the elderly
   healthy men, elderly ON men, and elderly OP men (Figure [171]8E). These
   results were consistent with the mass spectrometry data. Correlation
   analysis showed a negative correlation between SEV‐APP and BMD in
   elderly men (Figure [172]8F). We further analyzed the potential
   clinical significance of SEV‐APP. ROC curves were performed to compare
   in detail the effectiveness of Tes, TNC, APP, in combination of two
   markers and in combination of all three markers in diagnosing OP. The
   results demonstrated the AUC of Tes combined with SEV‐APP was 0.77, and
   the AUC of the combination of all three markers was also 0.77, which
   was higher than that of others (Figure [173]8G; Figure [174]S8D–F,
   Supporting Information). Based on these results, the Tes combined with
   SEV‐APP thought to have potential clinical value in assisting the early
   risk assessment and diagnosis of OP. A graphical abstract of this study
   is shown in Figure [175]8H, demonstrating that Tes suppresses the
   overactivation of osteoclast‐mediated bone degeneration in elderly men
   via osteoblast‐AR‐mediated upregulation of the fibronectin C‐terminus
   of TNC.

Figure 8.

   Figure 8
   [176]Open in a new tab

   SEV‐APP combined with Tes level is a potential biomarker in serum for
   early risk assessment of OP in elderly men. A) Hierarchical clustering
   heatmaps of proteins of cell lysates of osteoclasts and osteoclasts
   stimulated by 10 or 20 ng mL^−1 pep2. B) Venn diagram showing
   overlapping proteins. C) STRING analysis of the proteins in RAGE
   receptor binding. D) Verification of APP expression in the cell lysates
   of osteoclasts stimulated by pep2 by Western blot. E) Comparisons of
   SEV‐APP level in elderly men with normal BMD (E‐NC, n = 19), elderly
   men with osteopenia (E‐ON, n = 21), and elderly men with osteoporosis
   (E‐OP, n = 21) by ELISA. F) Correlation analysis between SEV‐APP and
   DXA‐BMD of one‐third distal radius in elderly men (n = 61). G) ROC
   curve analysis to evaluate the effectiveness of diagnosing OP (n = 19
   for normal BMD; n = 21 for OP). H) Graphical abstract of the study.
   One‐way ANOVA with Tukey's multiple comparisons test was used for
   multiple comparisons. All tests were two‐sided; *p < 0.05. OC =
   osteoclasts; EV = extracellular vesicles; DXA = dual‐energy X‐ray
   absorptiometry; BMD = bone mineral density; Tes = testosterone.

3. Discussion

   Although the beneficial effects of sex hormones on bone have been
   reported widely, the elucidation of the molecular mechanisms underlying
   the roles of androgen in bone structure and strength is still
   incompletely understood.^[ [177]^37 ^] Estrogen deficiency is commonly
   accepted to be a critical factor in the degeneration of the skeleton in
   both sexes; Vanderschueren et al. reported that lack of bioavailable
   β‐E2 could potentially play a role in the development of OP in aging
   men.^[ [178]^38 ^] However, the estrogen‐centric view of bone
   remodeling is increasingly being challenged by a greater understanding
   of the pathogenesis of bone loss. Although a legitimate argument can be
   made that the effects of androgen on bone are mediated principally by
   the aromatization of estrogen,^[ [179]^10 ^] accumulating evidence
   indicates that androgens exert direct effects in the maintenance of
   body composition and trabecular parameters.^[ [180]^39 ^]

   It remains likely that Tes impacts bone size by influencing periosteal
   apposition and, potentially indirectly, bone mass by augmenting muscle
   mass.^[ [181]^19 ^] As individuals age, it is typical for Tes levels to
   decline gradually. However, there should not be an abrupt drop in Tes
   levels in men that parallels the drastic decrease in estrogen observed
   during menopause in women.^[ [182]^7 ^] There is an immediate decline
   in Tes levels due to androgen deprivation therapy (ADT) for prostate
   cancer, leading to bone loss. Typically, BMD decreases by ≈2–8% within
   the first year after starting ADT. While TRT is thought to help
   preserve and even increase BMD in hypogonadal men, there is currently
   no evidence to support the idea that TRT can prevent bone fractures.
   Presently, TRT is not recommended solely for the purpose of enhancing
   and maintaining BMD in hypogonadal men. Rather, it should be considered
   as one of the treatment options for improving both hypogonadal symptoms
   and BMD simultaneously in symptomatic hypogonadal men with ON.^[
   [183]^8 ^] Our study identifies that BMD is positively correlated with
   the serum Tes level in elderly men, and the cut‐off value of the ROC
   curve for evaluation of bone loss in elderly men was 9.415 nmol L^−1
   (normal range in men > 60‐years‐old 3.02–27.07 nmol L^−1). That is, a
   serum Tes level < 9.415 nmol L^−1 in elderly men is a negative factor
   for bone health, and there is a risk of bone loss. Close monitoring of
   bone health and active intervention in terms of lifestyle, exercise,
   and fall prevention are needed. However, the lack of deep understanding
   of the mechanisms by which Tes protects bone quality and bone strength
   still restricts early diagnosis and precise intervention in bone loss
   in elderly men.

   To investigate the role of Tes in OP among elderly men, we employed a
   mouse model for in vivo studies. OP mouse models are broadly divided
   into two types: reduced bone formation models, such as the senile
   osteoporotic model for aging‐related OP, and increased bone resorption
   models, such as ovariectomized osteoporotic model, the gold standard
   for postmenopausal OP research, and the disuse model for OP due to
   limited or impaired physical activity. Additionally, genetic
   engineering‐based OP models are in the exploratory stage. Current
   researches show that knockout of genes such as OPG, RUNX2, and TGF‐β1
   can induce OP models. Our study focuses on the effects of Tes on the
   development and progression of OP in elderly men. Existing animal
   models do not sufficiently simulate the effect of low Tes on OP in
   elderly men. Based on findings that Tes inhibits osteoclast formation
   through osteoblasts, we constructed osteoblast‐specific AR knockout
   mice. Analysis of the femur in 12‐month‐old mice indicates that
   osteoblast‐specific AR knockout in elderly male mice can induce OP,
   representing a new modeling approach. Furthermore, considering the
   decreased mobility in the elderly, we combined the osteoblast‐specific
   AR knockout with the disuse osteoporotic tail suspension model. This
   combination results in a more significant reduction in bone mass. The
   use of old male mice with osteoblast‐specific AR knockout combined with
   the tail suspension serves as an animal model to investigate the
   pathogenesis of OP associated with age‐related Tes decline in elderly
   males and to identify effective therapeutic strategies. However, humans
   are the only primate species in nature that rely primarily on fully
   upright walking as their main mode of locomotion. The load‐bearing
   mechanisms in existing animal models cannot fully replicate the
   biomechanical forces exerted on the human femoral head during upright
   walking.

   Our data from experiments in mice demonstrate that Tes action via the
   AR regulates trabecular bone remodeling, which is consistent with
   research in humans showing that maintenance of serum Tes levels but
   loss of estrogen production led to significant decreases in cortical
   area and thickness (with increases in Ct.Po). Khosla et al. estimated
   that of the total effect of sex steroids on bone resorption in men,
   estrogen accounted for 70% and Tes for at most 30%.^[ [184]^40 ^]
   However, estrogen levels in men are consistently low
   (73.42–172.837 pmol L^−1) and are associated with estrogen‐only when
   they are <40 pmol L^−1.^[ [185]^41 ^] This parsing of the resorption
   data to reflect estrogen effects on cortical bone versus Tes effects on
   trabecular bone (70:30) is remarkably consistent with the fact that the
   skeleton is ≈80% cortical bone and ≈20% trabecular bone.^[ [186]^42 ^]
   Therefore, we speculate that bone loss in elderly men, especially the
   destruction of bone microstructure, is closely related to the weakening
   of the regulatory Tes‐AR signal of osteoblasts, and the effect of
   enhancement of osteoclast activity on trabecular bone, and its
   regulatory mechanism needs to be elucidated. In the present study,
   analysis of bone microarchitecture indicated that bone destruction in
   the proximal femur of elderly males is mainly manifested as a decrease
   in the thickness of cortical bone and trabecular bone and an increase
   in the Ct.Po at the superior portion of the femoral head, accompanied
   by a reduction in the serum Tes level. In vivo studies in mice lacking
   AR in osteoblasts identified that Ct.Po and Tb.Sp were increased at the
   top of the proximal femur after tail suspension.

   For clinical significance, the medial cortex at the junction area of
   the femoral head and femoral neck, in addition to the trabecula in the
   subchondral bone of the femoral head, are of great account for the
   biomechanical stability of internal fixation (either the femoral neck
   system or cannulated cancellous screws) for femoral neck fractures.^[
   [187]^43 ^] It is true that the femoral neck is much more important as
   a fracture site, while in view of the poor prognosis and easy failure
   of surgery, the present study is more focused on the mechanical loading
   area for internal fixation, and the underlying molecular regulatory
   path of Tes in bone remodeling of aged men with OP. Hip fractures in
   the elderly include two major diseases, femoral neck fracture and
   intertrochanteric fracture, and one of the commonly used hip‐preserving
   surgeries is hollow nail fixation and the other is proximal femoral
   nail antirotation. Both procedures require strong bony support in the
   subchondral bone of the femoral head and the Adam's arch of the femoral
   neck, otherwise, serious complications such as surgical failure,
   delayed fracture healing, and re‐fracture are likely to occur. Thus,
   the VOI was determined as which includes the complete coronal position
   of the femoral head, and the medial part of the trochanter, and the
   area above the lesser trochanter of the femur. As we previously
   reported, the integrity of these structures determines the
   effectiveness of surgical treatment,^[ [188]^44 , [189]^45 ^]
   suggesting that identification of the underlying mechanisms in bone
   loss and microstructural destruction is conducive to improve the
   prognosis and reduce the failure rate of surgery of those elderly men
   with low Tes levels.

   Research findings indicate that β‐E2 significantly contributes (≈70%)
   to the inhibition of osteoclast formation through the induction of Fas
   ligand expression in osteoblasts.^[ [190]^46 ^] While Tes inhibits
   osteoclastogenesis by ≈30%, the specific mechanism through which it
   does so remains unidentified.^[ [191]^15 ^] There is limited evidence
   to support the hypothesis that Tes directly affects osteoclasts.^[
   [192]^8 ^] Our in vivo studies suggested that the loss of Tes‐AR
   signaling in osteoblasts leads to accelerated bone destruction under
   microgravity‐induced osteoclast overactivation. Before the in vitro
   studies, we searched the literature and found the Tes concentration in
   male blood is equivalently to the concentration level 10^−8 m.^[
   [193]^25 , [194]^26 , [195]^27 ^] The concentrations of 10^−8 m Tes and
   10^−9 m DHT had similar significant promoting effects on osteoblasts
   (Figure [196]S9, Supporting Information). In vitro study identified
   that Tes activates osteoblast‐AR‐mediated upregulation of the
   fibrinogen C‐terminus of TNC during the differentiation of osteoblasts,
   which inhibits osteoclast differentiation and the area of osteoclast
   fusion. ITGαV‐ITGβ3 is known to be highly expressed in fully
   differentiated and mature osteoclasts;^[ [197]^47 ^] it engages in
   osteoclast migration, adhesion, and fusion by binding to the
   colony‐stimulating factor‐1 receptor, thereby initiating cytoskeletal
   formation and activating the ERK/c‐Fos signaling pathway.^[ [198]^48 ^]
   Antagonists targeting ITGαV‐ITGβ3 are regarded as effective
   pharmaceuticals for postmenopausal OP.^[ [199]^49 ^] A binding site
   within the FN III repeat of TNC has been identified for ITGαV‐ITGβ3,
   which contains a cryptic RGD sequence.^[ [200]^32 ^] This sequence is
   typically obscured by the adjacent second FN III repeat.^[ [201]^50 ^]
   Another binding site for ITGαV‐ITGβ3 has been located within the
   C‐terminal FBG region of TNC. This site is responsible for the
   diminished adhesion of endothelial cells to TNC substrates.^[ [202]^32
   , [203]^51 ^] However, the precise binding site and mechanism of action
   in ITGαV‐ITGβ3‐expressing cells of the C‐terminal FBG of TNC have not
   yet been explored in detail.^[ [204]^51 ^] We noticed that the focal
   binding of FITC‐pep2 with ITGαV‐ITGβ3 demonstrating a sign of
   polarization, which is not uniformly consistent. On one hand, there are
   different kinds of integrins (including but not limited to αVβ3, α2β1 ,
   and αVβ5) are expressed in the process of osteoclast differentiation
   and maturation, and their expression on the osteoclast cell membrane is
   dynamically changing. On the other hand, the highly selective binding
   domain in TNC‐C pep2 can probably change the distributions of
   ITGαV‐ITGβ3.^[ [205]^48 , [206]^52 , [207]^53 ^]

   In these in vitro studies, we used murine‐derived osteoblasts and
   osteoclasts. The conclusions may be somewhat inadequate to reflect the
   real situation in the human body. Both human and mouse cell lines have
   been artificially modified and cannot fully reflect the real situation
   in the body. While the DNA binding domain of AR is the most highly
   conserved region between the different members of the steroid hormone
   nuclear receptor family.^[ [208]^54 , [209]^55 ^] Perhaps we need to
   develop models that are closer to what the human body really looks
   like, such as bone organoids. However, at present, the construction of
   bone organoids still faces challenges, such as the types, sources,
   identification, and separation methods of bone stem cells, and how to
   select and optimize the extracellular matrix suitable for the
   construction of bone organoids still needs further research.^[ [210]^56
   , [211]^57 ^]

   ColabFold was used in our study to simulate the docking of the
   fibrinogen C‐terminus of TNC with the extracellular regions of
   ITGαV‐ITGβ3. The fibronectin C‐terminus thus blocks the binding of
   other proteins with RGD sequences to ITGαV, thereby inhibiting
   osteoclast fusion and bone resorption. On the contrary, full‐length TNC
   has been noted to facilitate cartilage repair and deter the onset of
   osteoarthritis without worsening synovitis.^[ [212]^58 ^] Meanwhile,
   alternative research identified TNC as an endogenous danger signal
   responsible for driving TLR4‐dependent fibroblast activation. Moreover,
   its persistence impedes the resolution of fibrosis.^[ [213]^59 ^]
   Hence, considering the structural, expression, and functional traits of
   TNC, the alternative splicing variants of TNC assume diverse roles in
   various physiological and pathological conditions.^[ [214]^23 ^]

   Research findings also demonstrate a negative relationship between
   serum Tes and TNC among young men diagnosed with major depressive
   disorder (average age: 36.48 ± 15.73),^[ [215]^60 , [216]^61 ^] which
   indirectly confirms our findings that Tes and TNC pose multiple
   correlations. Notably, young men exhibit elevated Tes levels compared
   to elderly men, with both control and major depressive disorder groups
   displaying higher levels than those observed in the elderly men
   included in our study. While major depressive disorder and OP represent
   distinct conditions, they may share common underlying mechanisms. The
   mouse tail suspension model serves not only as a model for OP but also
   as a model for depressive disorder. Combining our findings investigated
   from human samples and in vivo mouse models, Tes and TNC might exert
   regulatory influences on both OP and major depressive disorder.
   However, a concentration‐dependent shift is very likely existing in the
   regulatory effect of Tes on TNC, resulting from the neuro‐bone tissue
   specificity and distribution of androgen receptors, which play a
   distinct role in the mechanisms underlying various diseases.

   In terms of our investigation, lack of AR‐Tes signaling would lead to
   downregulation of the extracellular fibrinogen C‐terminus of TNC,
   indirectly inhibiting osteoclastogenesis and bone resorption. Thereby,
   the osteoclastic overactivation resulting from physical inactivity is
   disinhibited. Combined with the decreased osteoblastic bone formation,
   a net decrease in cortical and trabecular thickness occurs, decreasing
   bone strength and increasing fracture risk.^[ [217]^62 ^] This can
   explain why some elderly men lose bone mass rapidly when they are
   bedridden or have limited activities. Although most men will not suffer
   from extreme hypogonadism, as they age, the Tes levels of some men drop
   to lower than normal levels.^[ [218]^63 ^] When this group of people
   undertake limited activities, significantly reduce weight‐bearing
   exercises, or even become bedridden due to advanced age or injury,
   rapid osteoclast activation and bone resorption cannot be limited by
   the fibrinogen C‐terminus of TNC secreted by osteoblasts, resulting in
   a reduction in the thickness of the bone cortex and trabecular bone,
   and a reduction in the density of trabecular bone connections, thereby
   lowering bone strength. Our data demonstrating the TNC‐C pep2
   therapeutic effect on bone loss and destruction of those elderly men
   with low Tes, while numerous investigations including chemistry,
   manufacturing and control and pharmacokinetics of the peptides would to
   be done before it processes to clinical application.

   Osteoclast activity results in the breakdown of the bone matrix,
   leading to the release of specific biochemical markers into the
   bloodstream.^[ [219]^64 ^] These include C‐terminal telopeptide (CTX)
   and N‐terminal telopeptide (NTX), products of type I collagen, which
   are indicative of osteoclastic activity and bone resorption.^[ [220]^65
   ^] They are commonly used to assess the risk of OP and to monitor the
   efficacy of OP treatment, but cannot be used alone to diagnose or
   differentiate OP from normal bone.^[ [221]^66 ^] In the present study,
   we combined the SEV‐APP and serum Tes levels in elderly males to
   determine senile bone destruction—the results significantly improved
   the accuracy of identifying the elderly population at high risk of bone
   destruction. This is a new strategy for advancing the diagnostic window
   of bone destruction assessment in elderly men, realizing early and
   accurate intervention, and evaluating treatment prognosis. In addition,
   we also paid attention to the bone nonunion (a fracture that persists
   for a minimum of 9 months without signs of healing for 3 months), which
   means the body's inability to heal a fracture and causes devastating
   outcomes for patients. The ability to promptly identify patients at
   high risk for nonunion will allow for the success of early appropriate
   targeted therapeutic intervention. Research and progress have been made
   in understanding the pathophysiological processes of bone therapy to
   deal with markers of increased nonunion susceptibility. Moghaddam
   et al. found that patients with atrophic non‐union exhibited
   significantly lower serum levels of CTX in the first week and reduced
   TRACP 5b at weeks 4 and 8.^[ [222]^67 ^] Zimmermann et al. found that
   while TGF‐β1 levels initially rise after a fracture, they decline
   earlier in patients with delayed bone healing.^[ [223]^68 ^] In our
   study, we evaluated the ability of Tes combined with SEV‐APP to predict
   the risk of bone nonunion following femoral neck fractures. According
   to the cutoff value of Tes (9.985 nmol L^−1) and SEV‐APP
   (4.868 pg µg^−1), elderly (>60 years) men with OP were divided into
   four groups (Figure [224]S10A, Supporting Information, n = 20 per
   group). The occurrence of bone nonunion in elderly OP men with Tes
   level below 9.985 nmol L^−1 and SEV‐APP level above 4.868 pg µg^−1
   (Figure [225]S10B, Supporting Information), indicating Tes combined
   SEV‐APP might be a potential biomarker for predicting the bone
   nonunion. Recent studies have focused on how the APP and its
   metabolites affect bone cell function. APP is a transmembrane protein
   that can be cleaved to produce amyloid‐beta peptides. It has been found
   that APP and its derivatives can regulate the differentiation and
   activity of osteoblasts and osteoclasts. APP can either enhance or
   inhibit Wnt‐mediated signaling in osteoblasts, depending on the
   specific context and the presence of other regulatory factors.^[
   [226]^69 ^] The emerging concept of the “bone–brain axis” highlights
   the communication between the skeletal system and the nervous system.
   APP, given its role in neurodegenerative diseases and its potential
   impact on bone cells, is being investigated as a key player in this
   crosstalk.^[ [227]^70 ^] However, the presented available data could
   only serve as preliminary findings toward this goal. Further
   large‐scale clinical studies are required. For example, quantitative
   analysis of specific APP‐derived peptides in the blood or cerebrospinal
   fluid may provide insights into the risk of bone loss and the
   progression of OP.

   In this study, neither a single‐center validation using the primary
   index of DXA‐BMD of one‐third distal radius nor the secondary
   validation with QUS‐BMD of heel bone collected from UK Biobank could
   fully satisfy our ultimate goal of studying the DXA‐BMD of the proximal
   femur. Compared with DXA, QUS has the advantages of portability, low
   cost, and no radiation,^[ [228]^71 ^] which can obtain a larger sample
   size. However, due to the impact of soft tissue and bone redundancy,
   and the diagnostic accuracy of QUS in the heel bone is lower than that
   of DXA and the long‐term follow‐up capability is still unclear.^[
   [229]^72 ^] Although DXA‐BMD of the one‐third distal radius may provide
   useful clinically information, its predominantly cortical bone
   component may not fully reflect metabolic changes in trabecular bone.
   Compared with the proximal femoral site, one‐third distal radius
   measurements may be less sensitive for early osteoporosis detection^[
   [230]^73 ^] and can only meet our needs to a certain extent. In
   addition, the cortical bone of mice is quite different from that of
   humans, because humans have extensive intracortical osteonal remodeling
   that is absent from mice.^[ [231]^19 ^] Thus, the study of the
   regulation of bone microstructure by Tes, particularly in the
   trabecular bone of elderly men, requires sufficient quantity and
   higher‐resolution CT images combined with prudent interpretations from
   mouse models.

4. Conclusion

   In conclusion, this study reveals that Tes‐AR signaling delays elderly
   male bone microarchitectural destruction by upregulation of the
   osteoblastic TNC, particularly the fibrinogen C‐terminus, exerts a
   suppressive effect on osteoclastogenesis by competitively binding to
   integrin αV, thereby interfering with pro‐osteoclastic signaling. The
   Tes‐AR‐TNC axis emerges as a central regulatory pathway in preserving
   bone microarchitecture under age‐related or mechanical unloading
   conditions. Moreover, the identification of APP in extracellular
   vesicles as a biomarker linked to AR‐TNC axis deficiency provides a
   potential diagnostic biomarker for early detection of bone loss. These
   findings offer mechanistic insights and highlight novel targets for
   precision prevention and treatment of bone loss and nonunion.

5. Experimental Section

Reagents

   Reagents and kits were purchased from commercial sources. Iodoacetamide
   (IAA), dithiothreitol (DTT), urea, and cell culture medium were from GE
   Healthcare (Little Chalfont, Bucks, UK). The BCA protein assay kit, TMT
   labeling kits, and fetal bovine serum (FBS) were from Thermo Fisher
   Scientific (Waltham, MA, USA), and the protease inhibitor cocktail was
   from Roche (Basel, Switzerland). Sequencing‐grade endoproteinase
   Trypsin/Lys‐C was from Promega (Madison, WI, USA). Ascorbic
   acid‐2‐phosphate (LAA), β‐glycerol phosphate (β‐GP), β‐E2, Tes, the
   TRAP, and the Alizarin red staining kits were purchased from Sigma (St.
   Louis, MO, USA). An Xbridge BEH300 C18 column (4.6 × 250 mm, 5 µm) was
   obtained from Waters (Milford, MA, USA). A fused silica capillary
   column (75 µm ID, 150 mm length) was purchased from Upchurch (Oak
   Harbor, WA, USA), and C18 resin (300 A, 5 µm) was from Varian (Palo
   Alto, CA, USA).

Serum Samples and Computed Tomography (CT) Images of Recruited Individuals

   Elderly (aged above 60 years) and young (aged below 60 years) men
   diagnosed with osteoporosis (OP)/osteopenia (ON) and male volunteers
   with normal bone were recruited to this study at the Chinese PLA
   General Hospital (PLAGH, Beijing, China) from December 2021 to
   September 2023 (Table [232]S1, Supporting Information). All these
   patients and volunteers involved in the study were without diabetes,
   thyroid diseases, autoimmune diseases, or tumors, and had not received
   treatment for ON or OP prior to this study. Serum samples were
   collected at the same time of day and under fasting conditions. CT
   images of the right hip from 10 elderly men with normal bone mass vs 10
   elderly men with OP were included (Both groups had the equivalent
   baselines of age and body size, Figure [233]S1A, Supporting
   Information). BMD‐based bone assessments were measured by DXA (OsteoSys
   EXA 3000 Bone Density Device) at the one‐third distal radius site. In
   men aged ≥60, OP and ON were diagnosed according to the application of
   the WHO diagnostic T‐score criteria to BMD measurements (with T‐scores
   at −1 or above defined as “normal”; between −1 and −2.5 defined as
   “ON”; and at −2.5 or below defined as “OP”). However, in men <60 years
   of age, the ethnicity or race‐adjusted Z‐scores were recommended by the
   International Society for Clinical Densitometry (ISCD) instead of
   T‐scores in the diagnostic classification (with Z‐scores of −2 or lower
   defined as either “low bone mineral density for chronological age” or
   “below the expected range for age” and those above −2 defined as
   “within the expected range for age”). Informed consent was obtained
   from all patients and normal volunteers. The study was performed with
   the approval of the Ethics Committee of the PLAGH.

Animal Studies

   The animal study was approved by the Ethics Committee of the Chinese
   Academy of Medical Sciences (No: ACUC‐A01‐2020‐014). All mice were bred
   and housed under specific‐pathogen‐free (SPF) conditions with a 12‐hour
   light/dark cycle. The wild‐type (WT) C57BL/6N mice were purchased from
   Beijing Vital River Laboratory Animal Technology. Osteocalcin‐Cre
   (Ocn‐Cre) transgenic mice express cre‐recombinase in osteoblasts were
   from The Jackson Laboratory (JAX:019509).

   Briefly, Ar^flox/flox mice were created from Cyagen Biosciences
   (Suzhou, China) by CRISPR/Cas‐mediated genome engineering, with two
   loxP sequences flanked exon 2 of mouse Ar gene. Then, a homozygous
   female Ar^flox/flox mouse was bred with a male Ocn‐Cre mouse to
   generate mice that were heterozygous for the Ar^flox/Y allele and
   heterozygous for the Cre transgene. The male mice expressing both the
   Ar^flox/Y allele and Cre recombinase were the targeted Ocn‐Ar^−/Y mice.
   Approximately 25% of the progeny from this mating would be the targeted
   mice. Twelve‐month‐old mice, maintained in a specific pathogen–free
   environment, were used for experiments. More details can be found in
   the Supporting Information.

Tail Suspension Test

   The male mice were raised to 12 weeks. Then the male mice were
   submitted to hind limb unloading through tail suspension or had the
   tail suspension device but were not suspended. Four‐week periods were
   performed. Each animal was single‐housed in a polycarbonate cage
   (25 × 18 × 20 cm) with a bearing supporting the suspension hanging
   system. The bearing had very low friction so that mice could easily
   move around on the cage surface, but they could not lean on walls with
   their hind limbs. The mice axis was adjusted to 30.^[ [234]^74 ^] Each
   mouse had free access to water and food.

Cell Culture

   The RAW 264.7 and MC3T3E1 cell lines were obtained from the China
   Infrastructure of Cell Line Resources (Beijing, China). Osteoclast
   formation was performed as previously described by Xiong et al.,^[
   [235]^75 ^] with minor modifications. Briefly, to generate mature
   multinucleated osteoclasts, RAW 264.7 cells were cultured in α‐minimum
   essential medium (α‐MEM) with 10% FBS in 6‐well plates and incubated at
   37 °C under 5% CO[2]. Cells were cultured by adding 10 ng mL^−1 RANK
   with or without 10^−8 m β‐E2 and testosterone to the culture system; an
   equal volume of DMSO was added to the control and all groups were
   analyzed in triplicate. The culture medium was refreshed every other
   day. On day 5, osteoclast cells were confirmed by TRAP staining before
   being harvested, and the triplicates of each group were mixed and
   prepared for TMT labeling. For osteoblast differentiation, MC3T3E1
   cells were induced with a differentiation medium, comprising the growth
   medium supplemented with 50 µg mL^−1 LAA and 10 mm β‐GP. The
   differentiation medium was replaced every 2 days. β‐E2 (10^−3 m) and
   Tes (10^−5 m) were dissolved in DMSO and subsequently diluted 1000‐fold
   in the medium for cell culture at final working concentrations 10^−6 m
   for β‐E2 and 10^−8 m for Tes. On day 15 or day 18, osteoblast cells
   were confirmed by Alizarin red staining before harvesting, and the
   triplicates of each group were mixed and prepared for TMT labeling.

Micro‐Computed Tomography (Micro‐CT) Assessment of Bone Structure

   Before fixation with 4% paraformaldehyde for 24 h, all soft tissue
   surrounding the femur was carefully removed. The femur bones were then
   securely positioned within a plastic tube to minimize any potential
   movement during scanning. Micro‐CT analysis was conducted utilizing a
   Scanco micro‐CT‐100 machine, employing the following settings: a
   resolution of 10 µm, X‐ray tube potential set at 70 kV, intensity at
   200 µA, and an integration time of 300 ms. The raw data was processed
   using Scanco evaluation program v6.6 (Scanco Medical) to obtain the
   final results.

Dragonfly Bone Analysis

   3D reconstruction and analysis of the CT images of humans and micro‐CT
   images of mice were performed with the Dragonfly 2022.2 (Comet
   Technologies Canada, Inc.) software with DICOM files. The standardized
   processes were followed to determine the VOI for all bones (complete
   coronal position of the femoral head, and the medial part of the
   trochanter, and the area above the lesser trochanter of the femur). The
   Bone Analysis module in Dragonfly was used to segment cortical and
   trabecular bone^[ [236]^76 , [237]^77 ^] and to compute bone
   morphometric measurements, following the guidelines described by Neral
   M and Mologne TS et al.^[ [238]^78 , [239]^79 ^] The bone trabecular
   parameters were set as 0.5 mm for the proximal femur of humans and
   0.1 mm for the mice. Using the Bone Analysis module, bone volume
   fraction (BV/TV), specific bone surface (BS/BV, mm⁻¹), and bone surface
   density (BS/TV, mm⁻¹) were determined for the overall description of
   the VOI in the proximal femur. Cortical area fraction (Ct.Ar/Tt.Ar, %),
   cortical porosity (Ct.Po), and cortical thickness (Ct.Th, mm) were
   determined for cortical bone. Trabecular thickness (Tb.Th, mm) and
   trabecular separation (Tb.Sp, mm) were determined for trabecular bone;
   trabecular number (Tb.N, mm⁻¹) was calculated based on 3D measurements
   for the spacing of trabeculae. In addition, the other parameters
   reflecting the structural and mechanical characteristics of bone
   tissues including endocortical perimeter (Ec.Pm, mm), periosteal
   perimeter (Ps.Pm, mm), structure model index, connectivity density
   (Conn.D, mm⁻^3), anisotropy (MIL) and anisotropy (SVD), were determined
   for assessment.

Mechanical Testing

   A three‐point bending test was performed until failure of the proximal
   femur using an [240]E10000 All‐Electric Dynamic Test Instrument
   (Instron Corporation, Canton, MA). The specimens were positioned
   between two supporting transverse bars (distance = 3 mm), the loading
   rod precisely situated at the midpoint. Utilizing a 500‐N load cell,
   all tests were conducted at a consistent loading rate of 3 mm min^−1
   until fracture occurred. Data acquisition was performed at intervals of
   5 milliseconds. The ultimate force and stiffness were directly derived
   from the load and deflection curves. Elastic modulus and ultimate
   stress were subsequently computed using the following equations:
   elastic modulus = KL3/(48 Imin); ultimate stress = FultLc/(4 lmin). K =
   stiffness; L = Gauge‐length; I = moment of inertia; Fult = ultimate
   force; Lc = bone radius.^[ [241]^80 ^]

TRAP Staining

   Mature osteoclasts were defined as TRAP‐positive cells containing three
   or more nuclei. TRAP staining was performed using the TRAP staining kit
   according to the manufacturer's instructions. Briefly, cells were
   washed three times with PBS and fixed with a fixative solution (65%
   acetone and 3.7% formaldehyde in citrate solution). Fixative solution
   was completely removed and cells were washed three times with ddH[2]O.
   Cells were stained with TRAP solution for 1 h at 37 °C in the dark.

Alizarin Red Staining

   Mineralization in osteogenic cultures was determined by Alizarin red
   staining. Cells were washed with PBS, fixed with 3.7% formaldehyde for
   10 min, washed three times with PBS, and then stained with 40 mm
   Alizarin red (pH 4.1) for 15 min at room temperature. The cultures were
   then washed three times with ddH[2]O and photographed under a light
   microscope. For matrix mineralization quantification, calcium‐bound
   Alizarin red was solubilized with 10% acetic acid for 30 min. The
   supernatant was then collected and ammonium hydroxide was added to pH
   4.1. The degree of mineralization was indicated by supernatant
   absorbance at 405 nm.

Assay of Alkaline Phosphatase (ALP) Activity

   Using an Alkaline Phosphatase Assay Kit (Beyotime, P0321), ALP activity
   was determined on days 3, 7, 9, 15, and 18 after MC3T3‐E1 cells were
   cultured in DM. The cells were trypsinized and harvested before
   resolution in RIPA lysis buffer. After sonication, the cell lysates
   were centrifuged at 12 000 rpm for 5 min, and the supernatants were
   added to the 96‐well plate ALP assays. All the samples and controls
   were assayed in triplicate. The absorbance at 405 nm was measured based
   on the conversion of colorless pnitrophenyl phosphate to colored
   p‐nitrophenol. The enzyme activity was normalized against the protein
   content quantified with the BCA protein assay kit.

Bone Resorption Activity

   The bone resorption capacity of osteoclasts was evaluated using the
   Osteo Assay Surface Plate (3987, Corning). RAW264.7 cells were seeded
   on the Osteo Assay Surface and stimulated with RANKL for 5 days. On day
   5, cells were removed with pipetting. The remaining matrix was stained
   with 1% toluidine blue. The bone resorption pit image was photographed
   under a light microscope.

Sample Preparation and TMT Labeling

   For TMT labeling, sample preparation and labeling were performed as
   previously described.^[ [242]^75 ^] Cells were washed with cold PBS
   three times, followed by the addition of 8 m urea. After sonication,
   cell lysates were clarified by centrifugation. Supernatants were
   obtained and protein concentration was quantified with the Nano Drop
   2000. Labeling was performed using the TMT kit following the
   manufacturer's protocol with slight modifications. Equal amounts of
   protein (100 µg) from each of the four groups were reduced with DTT and
   alkylated with IAA. Protein digestion was completed by incubation with
   Trypsin/Lys‐C at a mass ratio of 1:25 (enzyme: protein) for 12 h at
   37°. The digestion was terminated by heating at 60 °C for 30 min.
   Digested proteins were then desalted, dried, and finally dissolved in
   200 mm triethylammonium bicarbonate buffer (pH 8.5). Each group was
   distinguished by using different TMT labels: For osteoblast formation
   on day 15, TMT‐128 was used for the pre‐OB group; TMT‐129 was used for
   the OB group and TMT‐131 for the OB‐T group. For osteoblast formation
   on day 18, TMT‐127 was used for the pre‐OB group; TMT‐128 was used for
   the OB group, and TMT‐130 for the OB‐T group. After labeling, samples
   were pooled, desalted, and dissolved in 0.1% trifluoroacetic acid
   (TFA).

High‐Performance Liquid Chromatography (HPLC) Separation

   The TMT labeled peptides (100 µL in 0.1% TFA) were fractionated by HPLC
   (UltiMate 3000 UHPLC, Thermo Scientific) equipped with an Xbridge
   BEH300 C18 column maintained at 45 °C. The peptides were eluted by a
   gradient acetonitrile elution buffer (pH 10) at a flow‐rate of
   1 mL min^−1. Each 1.5 mL elution was collected as one fraction. A total
   of 47 fractions were collected, dried, and combined into 20 samples
   according to the peptide abundance. The samples were dissolved in 20 µL
   of 0.1% TFA for subsequent liquid chromatography (LC)‐MS/MS analysis.

Liquid Chromatography (LC)‐MS/MS Analysis

   The LC‐MS/MS analysis was performed as previously described.^[ [243]^81
   , [244]^82 ^] Briefly, peptides were first passed through a homemade
   fused silica capillary column (75 µm ID, 150 mm length; Upchurch)
   packed with C‐18 resin (300 A, 5 µm; Varian, Lexington, MA, USA) column
   on an EASY‐nLC 1000 system using gradient TFA elution buffer (pH 1–2)
   at a flow‐rate of 0.3 µL min^−1. The peptide eluate was analyzed with a
   directly interfaced Thermo Orbitrap Fusion mass spectrometer (Thermo
   Scientific) in positive‐ion mode. Xcalibur 3.0 software was used for
   mass data acquisition in a data‐dependent mode.

Data Analysis

   Proteins were identified using Proteome Discoverer 3.0 software (Thermo
   Scientific). In detail, the raw MS data files were searched against the
   SwissProt‐reviewed mouse proteome database. The following parameters
   were set as the search criteria: precursor mass tolerance, 20 ppm;
   fragment mass tolerance, 0.02 Da; total intensity threshold, 20 000;
   minimum number of peaks, 200; and a maximum of two missed cleavages
   were allowed. Carbamidomethylation (on C) and TMT 6‐plex (on K and the
   peptide N‐terminus) were set as static modifications, and oxidation (on
   M) was specified as a dynamic modification. Protein identification was
   considered valid when Score SEQUEST HT > 0 and unique peptide > 1 (if
   at least one peptide was statistically significant, P < 0.05 with a
   false discovery rate of 5). Reporter monoisotopic m/z was tuned
   according to the raw spectral data. Proteins were quantified using
   unique peptides. JMP software was used to analyze the distribution of
   protein ratios in four different groups, including log2 (OB/pre‐OB) of
   day 15, log2 (OB‐T/OB) of day 15, log2 (OB/pre‐OB) of day 18, and log2
   (OB‐T/OB) of day 18. The differential expression cutoffs were defined
   as the 95% prediction interval of these four groups. Representative
   MS/MS spectral data of identified peptides and the intensity of TMT
   precursor ions were used for protein quantification.

Bioinformatics Analysis

   Gene Ontology (GO) analysis and KEGG pathway enrichment analysis were
   performed with DAVID ([245]https://david.ncifcrf.gov/tools.jsp).
   Pathways with at least three target genes and p < 0.05 were considered
   to be statistically significant. Enriched pathways were visualized by
   Cytoscape (3.9.0) software. For Hierarchical Clustering analysis, the
   MS proteomics data presented as the ratios were input into JMP
   software. The STRING (Search Tool for the Retrieval of Interacting
   Genes/Proteins) database ([246]http://string‐db.org) was used for
   predicting protein networks. All STRING network analyses were performed
   using UniProt accession numbers as the input and with “Experimental”
   and “Database” evidence at a medium (0.4) confidence level. The
   networks were visualized with Cytoscape to give a better idea of the
   potential relationships between the proteins. The networks were
   downloaded as tab‐delimited text files and visualized and reorganized
   using Cytoscape (3.9.0) software. JASPAR website
   ([247]https://jaspar.elixir.no/) was used to predict the AR binding
   motif.

Western Blot Analysis

   For Western blot analysis, the lysed proteins (20 µg total protein)
   were separated by SDS‐PAGE. Gels were run according to standard methods
   and proteins were transferred electrophoretically to nitrocellulose
   membranes. Membranes were blocked with 5% non‐fat dried milk in TBS‐T
   (TBS plus 0.05% Tween‐20) for 30 min. The membranes were then incubated
   with primary antibodies overnight at 4 °C. Membranes were washed and
   incubated with horseradish peroxidase‐conjugated secondary antibodies
   and ECL reagents.^[ [248]^82 ^]

   For the Western blot of cell supernatants, the same amount of protein
   was loaded as the internal reference, which was validated by Ponceau S
   staining. In detail, the cell supernatant was collected and
   concentrated using a 3 kDa ultrafiltration tube. The protein
   concentration of the concentrated cell supernatant was determined by
   BCA, and a total of 25 µg protein per group was loaded for WB. After
   electrophoretic transfer, the membrane was stained with Ponceau S to
   verify that the sample loading of different groups was the same. After
   staining, the Ponceau S was washed off with PBS and the experiment was
   continued according to the WB procedure.

Effects of Tes‐Induced Osteoblast Secretions on Osteoclastogenesis

   MC3T3E1 cells were cultured in four groups (GM with DMSO, GM with
   10^−8 m testosterone, DM with DMSO, and DM with 10^−8 m testosterone).
   On day 17, DMSO and testosterone were removed from the culture
   conditions. On day 21, culture supernatants of the osteoblasts in the
   four groups were collected and prepared for the subsequent assays. On
   day 21, osteoblasts were confirmed by Alizarin red staining. RAW 264.7
   cells were cultured and supplemented with osteoblast culture
   supernatants; both groups were prepared in triplicate. The culture
   medium was refreshed every other day. On day 5, cells were confirmed by
   TRAP staining and photographed for calculation of the cell number and
   rearrangement area.

ChIP‐qPCR

   Tes‐stimulated osteoblasts were washed once in PBS and then
   cross‐linked with 1% formaldehyde for 15 min at room temperature. The
   reaction was stopped by 0.125 m glycine addition. Cross‐linked cells
   were then scraped for pellets and resuspended in lysis buffer (50 mm
   Tris⋅HCl, pH 8.1, 10 mm EDTA, 1% SDS, 1× protease inhibitor). Nuclear
   lysates were sonicated using a sonicator (QSONICA, Q800R, Newtown USA)
   to shear the DNA to a length under ≈500 bp in size. The supernatants
   were precleared with agarose beads for 1 h at 4 °C to and then the
   supernatants were incubated with antibodies against AR (Santa Cruz,
   sc‐7305) or isotype controls (mouse IgG, cell signaling technology,
   #5415). Immunoprecipitates were captured using Protein G agarose and
   eluted in TE buffer at 37 °C. After reverse‐crossing, the
   immunoprecipitated (or no IP input) DNA was analyzed via qRT‐PCR. The
   following qPCR primers were used for the ChIP‐qPCR analysis: region 1:
   F, ACTCCTATGCCCACCCTTTTAC; R, TAAACAAGGCCAGGGTTGTGGATT; region 2: F,
   TGAATGTCTGTGCATATGAATGGC; R, GTCACATCCTCAATTGCTTCCTT; region 3: F,
   TACCTGATTCCGAAGTGCATTG; R, CGCAAGAGCCTCATGACTAGAAA.

ELISA Assay

   The Mouse tenascin (TNC) ELISA Kit (CUSABIO, CSB‐EL023954MO) and Human
   tenascin (TNC) ELISA kit (CUSABIO, CSB‐E13125h) were used to detection
   of TNC in serum. The Human Amyloid Precursor Protein ELISA Kit (abcam,
   ab216944) and Mouse amyloid precursor protein (APP) ELISA Kit
   (MyBioSource, MBS2020954) were used to detect EV‐APP in serum. Mouse
   procollagen I N‐terminal peptide ELISA Kit (CUSABIO, CSB‐E12775m) was
   used to detect PINP in mouse serum. Mouse β‐CTx ELISA Kit (MyBioSource,
   MBS9901663) was used to detect β‐CTx in mouse serum. The ELISA assays
   were carried out according to the instruction manual.

Dual Luciferase Assays

   The full‐length mouse TNC promoter cDNA (−2000–+100 bp; UCSC mm39 DNA
   range = 63877922–63880022) was ligated into the promoter‐driven
   luciferase reporter plasmid pGL3‐basic. 293T cells were plated in
   24‐well plates and incubated for 24 h. An expression vector encoding AR
   and a pRL‐TK plasmid were transiently co‐transfected with pGL3‐TNC
   plasmid for 48 h. A dual‐luciferase reporter assay system (Promega,
   Madison, WI, USA) was used to calculate luciferase activity. The
   Renilla luciferase gene was used as the internal reference. Relative
   luciferase activity = firefly luciferase activity/renilla luciferase
   activity.

Molecular Docking

   The Colabfold was used to predict the structures of the TNC_C
   terminal‐ITGαV‐ITGβ3 protein complex. The input protein sequence of
   TNC_C terminal was the fibrinogen C‐terminal domain (Gly1884‐Asn2099)
   of mouse TNC (Uniprot: [249]Q80YX1), the extracellular segment
   (Phe31‐Pro983) of mouse ITGαV (Uniprot: [250]P43406) and the
   extracellular segment (Asn28‐Asp717) of mouse ITGβ3 (Uniprot:
   [251]O54890). The docking results were visualized using PyMOL.

Statistical Analysis

   Unless indicated otherwise, the experiments were performed in
   triplicate (for statistics) and repeated three times (for
   reproducibility). Data were presented as mean values ± SD. Statistical
   analysis was performed using GraphPad Prism 9 software. For two groups,
   the Student's t‐test was used. For multiple groups, ordinary one‐way
   analysis of variance (ANOVA) with Tukey's multiple comparisons test was
   used. Receiver operating characteristics (ROC) curves were computed
   using the Wilson/Brown method. The correlation was analyzed using
   Pearson's correlation coefficient. For most experiments, unless
   indicated otherwise, differences between groups were evaluated by
   Student's t‐test and p < 0.05 was considered to indicate statistical
   significance.

Data Availability

   The mass spectrometry proteomics data have been deposited to the
   ProteomeXchange Consortium
   ([252]https://proteomecentral.proteomexchange.org) via the iProX
   partner repository^[ [253]^83 , [254]^84 ^] with the dataset identifier
   PXD052247. Quantitative data that support the findings of this study
   are available within this article and its supplementary files. All
   other data supporting the findings of this study are available from the
   corresponding authors upon reasonable request.

Ethics Approval Statement

   IRB protocol for human studies was approved by the Ethics Committee of
   the Chinese PLA General Hospital (No: SB2021056). All animal
   experiments were approved by the Ethics Committee of the Chinese
   Academy of Medical Sciences (No: ACUC‐A01‐2020‐014).

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   Y.X., M.P., and Z.Z. contributed equally to this work. W.G. and P.T.
   conceived and designed the experiments. Y.X., M.P., Z.Z., H.L., and
   X.W. performed the experiments. P.T. and W.G. contributed to the
   materials. Y.X., M.P., Z.Z., L.Z., W.L., and P.T. analyzed the data.
   Y.X., M.P., Z.Z., H.L., and X.W. contributed to the discussion. Y.X.,
   M.P., Z.Z., and W.G. wrote the paper.

Supporting information

   Supporting Information
   [255]ADVS-12-e01518-s002.docx^ (5.1MB, docx)

   Supplemental Table 1
   [256]ADVS-12-e01518-s005.xlsx^ (33.5KB, xlsx)

   Supplemental Table 2
   [257]ADVS-12-e01518-s004.xlsx^ (61KB, xlsx)

   Supplemental Table 3
   [258]ADVS-12-e01518-s006.xlsx^ (61.1KB, xlsx)

   Supplemental Table 4
   [259]ADVS-12-e01518-s003.xlsx^ (12.8KB, xlsx)

   Supplemental Table 5
   [260]ADVS-12-e01518-s001.xlsx^ (762.6KB, xlsx)

Acknowledgements