Ascorbic Acid Modulates Collagen Properties in Glucocorticoid‐Induced Osteoporotic Bone: Insights into Chemical, Mechanical, and Biological Regulation

Abstract

Osteoporosis is a prevalent skeletal disorder characterized by decreased bone mass and structural deterioration, leading to an increased risk of fractures. This study focuses on the regulatory role of Vitamin C (ascorbic acid; AA) in the context of glucocorticoid‐induced osteoporosis (GIOP), which results from long‐term glucocorticoid (GC) therapy. The data showed that GCs impair AA metabolism in osteoblasts, thereby disrupting collagen synthesis and compromising extracellular matrix (ECM) integrity. Notably, AA integration in the collagen matrix improved its biochemical and mechanical properties. Additionally, it has been shown that the presence of AA restored osteoblast and endothelial function, enhanced collagen production, and improved endothelial barrier function under GC exposure. These results underscore the critical role of Vitamin C in bone matrix maintenance and homeostasis. Collectively, this work highlights the therapeutic potential of Vitamin C as a supportive treatment to counteract the deleterious skeletal effects of long‐term GC therapy.

Osteoporosis from long‐term glucocorticoid (GIOP) use elevates susceptibility to fracture. This study shows GCs impair ascorbic acid (AA) metabolism in osteoblasts, collagen synthesis and extracellular matrix integrity. AA enhanced collagen biochemical and mechanical properties and restored osteoblast and endothelial function. These findings underscore the essential role of AA in bone homeostasis and highlight its therapeutic potential against GIOP.

1. Introduction

Osteoporosis is a common skeletal disorder characterized by reduced bone mass and structural deterioration, leading to a significantly increased risk of fractures.^{[ 1 ]} Aging and age‐related physiological changes contribute to the progressive decline in bone health,^{[ 2 ]} with one in three women and one in five men expected to develop primary osteoporosis after the age of 50.^{[ 3} , ^{4 ]} In addition to primary osteoporosis, secondary osteoporosis may arise from different factors such as long‐term glucocorticoid (GC) therapy, which is classified as GC‐induced osteoporosis (GIOP), and is the most prevalent form of secondary osteoporosis.^{[ 5} , ⁶ , ^{7 ]} There are multiple causes of secondary osteoporosis, collectively affecting more than 50% of premenopausal women, up to 80% of men, and ≈30% of postmenopausal women.^{[ 8} , ⁹ , ¹⁰ , ^{11 ]}

GC therapy disrupts several critical metabolic pathways in bone tissue, resulting in impaired bone formation and increased resorption. GCs perturb bone metabolism and remodeling, driving an imbalance between osteoblasts and osteoclasts by upregulating Receptor Activator of Nuclear Factor‐ kappa B Ligand (RANKL) expression and downregulating osteoprotegerin (OPG).^{[ 12} , ¹³ , ^{14 ]} Furthermore, GC signaling is crucial for maintaining normal bone formation and architecture in mature osteoblasts via 11β‐Hydroxysteroid dehydrogenase type 2 (11β‐HSD2) signaling, which its overexpression leads to reduced collagen synthesis and bone volume.^{[ 15} , ^{16 ]} The 11β‐HSD2 enzyme converts active cortisol hormone into inactive cortisone hormone, protecting the mineralocorticoid receptor. Additionally, GCs suppress osteoblast survival and function through the AKT/MAPK (Mitogen‐activated protein kinase) signaling pathways, leading to a reduction in bone‐forming cells.^{[ 17} , ^{18 ]} GCs also impair osteoblast anabolic activity by inhibiting Insulin‐like Growth Factor 1 (IGF‐1) production^{[ 19} , ^{20 ]} and disrupting mitochondrial function in bone cells.^{[ 21 ]} This suppression of osteoblast activity results in decreased production of extracellular matrix (ECM) components, such as collagen, and bone‐related proteins such as alkaline phosphatase, osteocalcin, and osteonectin, which are all crucial for bone matrix integrity.^{[ 22} , ²³ , ²⁴ , ^{25 ]} Furthermore, GCs stimulate expression of matrix metalloproteinases (MMPs) in both osteoblasts and osteoclasts, promoting the breakdown of collagen and other ECM components, thereby impairing mineralization and weakening the bone.^{[ 26} , ²⁷ , ^{28 ]} Recent studies by Lee et al. further revealed that GCs impair bone microvascular barrier function and interfere with osteoblast‐endothelial cell interactions via the Connexin 43/MAPK pathway.^{[ 29 ]} Taken together, these effects undermine bone homeostasis between osteoblasts, osteoclasts, endothelial cells, and ECM, resulting in reduced bone mass, microstructural deterioration, and an increased risk of fractures.^{[ 30 ]}

GCs disrupt multiple metabolic pathways involved in the uptake and utilization of essential vitamins and minerals by bone cells, leading to deficiencies that impair bone homeostasis. It has been reported that GCs inhibit calcium metabolism by decreasing intestinal calcium absorption and increasing renal calcium excretion, which results in negative calcium balance in the bone tissue.^{[ 31} , ^{32 ]} In addition, GCs suppress the activation of Vitamin D, diminishing levels of calcitriol—the hormonally active form—thereby weakening its role in calcium regulation and bone mineralization. This is further exacerbated by reduced expression of the vitamin D receptor (VDR), which attenuates calcitriol's anti‐resorptive and bone‐protective effects, contributing to impaired mineralization and increased bone fragility.^{[ 33} , ³⁴ , ^{35 ]} As a result, calcium and Vitamin D supplementation has been widely recommended for patients undergoing GC therapy to support bone health and restore mineral balance.^{[ 36} , ^{37 ]}

While calcium and Vitamin D have been studied in GIOP, there is little to no in‐depth quantitative knowledge regarding the role of ascorbic acid (AA) as a critical factor in regulating bone homeostasis. AA deficiency results in epigenetic changes in the bone, with physiological effects such as severe defects in the mechanical strength and structure of mouse bone, as well as causing an inability for ECM deposition and mineralization of bone marrow stromal cells in vitro.^{[ 38 ]} Other in vivo studies revealed AA can reverse the effects of bone loss in an osteoporotic rat model and mechanistically activates osteoblastogenesis while inhibiting osteoclastogenesis via the Wnt/b‐Catenin/ATF4 pathways.^{[ 39} , ^{40 ]} Despite its recognized importance in bone physiology,^{[ 38} , ^{41 ]} there is limited quantitative and mechanistic understanding of how GC therapy alters AA metabolism and its downstream effects on the bone ECM. In our study, we demonstrated that GIOP is associated with a marked reduction in osteoblast activity, collagen type I production, and ascorbate metabolism in bone tissues. Notably, the integration of AA into the collagen I matrix, an important component of bone ECM, enhanced both the mechanical and biochemical properties, demonstrating the significance of AA in the bone ECM homeostasis. Furthermore, in the presence of GCs, AA was able to recover osteoblast function, restore collagen synthesis, and improve endothelial barrier integrity. These findings highlight the emerging therapeutic potential of AA in preserving bone matrix homeostasis and mitigating the negative skeletal effects of GCs.

2. Results

2.1. Glucocorticoids (GCs) Negatively Impact Osteoblast Function and Collagen Matrix Deposition

To assess the effects of glucocorticoids (GCs) on bone cell function and the key ECM protein COL1A1, we conducted in vivo studies. Consistent with our previous findings,^{[ 29 ]} Prednisolone‐treated mice (Psl) exhibited significant trabecular bone loss compared to untreated controls (Veh; vehicle), as shown by H&E staining (Figure 1 A). Furthermore, bone tissue from Psl‐treated mice displayed a 1.2‐fold reduction (p‐value<0.05) in alkaline phosphatase (ALP) activity, indicating impaired osteogenic function (Figure 1B,C). Since osteoblasts are primarily responsible for COL1A1 synthesis and deposition,^{[ 29} , ^{42 ]} we examined whether GCs control collagen content in bone. Verhoeff–van Gieson staining of collagen fibers (red) in cortical bone revealed a 1.3‐fold decrease (p‐value<0.05) in collagen fiber formation in Psl‐treated samples (Figure 1D,E). In line with these findings, immunostaining showed a 2.6‐fold reduction (p‐value<0.05) in COL1A1‐specific protein expression in bone tissue from Psl mice (Figure 1F,G). To further explore the regulation of COL1A1 by GCs in human cells in vitro, we conducted untargeted metabolomic analysis of osteo‐spheroids (Figure S1, Supporting Information). Pathway analysis identified that with GC treatment, there is an increase in putative metabolite levels, such as cotinine glucuronide, that hinder collagen synthesis, and a reduction in key putative metabolites essential for collagen biosynthesis,^{[ 43} , ⁴⁴ , ⁴⁵ , ^{46 ]} mainly involved with amino acid synthesis, such as Histidinyl‐Proline and Glutaminylhistidine, in the presence of GCs (Figure 1H–J). With many metabolomic putative hits suggesting altered levels of amino acids and metabolites associated with collagen synthesis, in conjunction with immunostaining of COL1A1, our results demonstrated that GCs impair COL1A1 metabolism, likely by disrupting key pathways involved in collagen synthesis.

Figure 1.

Glucocorticoids impair osteoblast function and collagen expression in vivo. A) H&E staining of mouse bone after 60 days without (Veh) or with the GC, prednisolone (Psl); Scale bar: 1000 µm. B) Representative images for Alkaline phosphatase (ALP) staining of mouse bones; Scale bar, 400 µm; Magnification: Scale bar, 100 µm. C) Quantitative analysis of the ALP‐positive surface intensity in bone tissue sections for n = 5 mice. D) Representative images for Verhoeff van Geison staining showing general elastin‐ (black) and collagen‐specific proteins (red) of mouse bones; Scale bar, 400 µm; Magnification: Scale bar: 100 µm. E) Semi‐quantitative analysis of the collagen fibers‐positive surface intensity in bone tissue sections; n = 5 mice. F) Representative image of collagen type I alpha I (COL1A1) staining of bone sections. Yellow arrows indicate COL1A1 protein expression (red). Green diamonds indicate the interior of cortical bone, near the bone marrow. Scale bar, 50 µm. G) Graph showing the COL1A1 protein expression levels (artificial units; A.U.) in control and Psl‐treated bone tissues. H) Metabolomics KEGG visualization of in vitro osteo‐spheroids treated without (Veh) and with Psl at 100 µm for 7 d. I,J) Analysis of collagen‐related putative metabolites after significance analysis using Metaboanalyst only reported fold‐changes with p‐value < 0.05. Data are expressed as mean ± SD. N = 5, n = 3; significant differences: * p‐value < 0.05.

2.2. GCs Alter Ascorbic Acid (AA) In Vivo and In Vitro

Building on those findings, we further investigate the signaling mechanisms by which GCs regulate COL1A1 expression. Untargeted metabolomic analysis revealed decreased putative metabolites from the ascorbate pathway, previously associated with COL1A1 biosynthesis,^{[ 47 ]} in response to GC treatment (Figure 2 A). Notably, key metabolites involved in steroid hormone metabolism, including those that can negatively impact AA metabolic pathways such as 4a‐Methylzymosterol, are increased, whereas Xestoaminol C (Vitamin C; AA) and related putative metabolites such as propionylcarnitine are decreased (Figure 2B,C).^{[ 48} , ⁴⁹ , ^{50 ]} To determine whether GCs also affect AA uptake at the cellular level, we assessed the expression of two critical AA transporters in bone, SVCT2 and GLUT1,^{[ 50} , ⁵¹ , ⁵² , ⁵³ , ^{54 ]} due to their involvement in AA cellular uptake in the form of AA or oxidized AA, dehydroascorbic acid (DHA) (Figure 2D). Protein expression levels of SVCT2 and GLUT1 were reduced by 4‐fold (p‐value<0.0001) (Figure 2E,F) and 9‐fold (p‐value<0.0001) (Figure 2G,H), respectively, in bone tissues from Psl‐treated mice compared to control (Veh). Collectively, these results highlight the significant impact of GCs on AA metabolism and cellular uptake, potentially contributing to impaired collagen synthesis.

Figure 2.

Glucocorticoids decrease ascorbic acid (AA)‐related marker expression in vitro and in vivo. A) Top putative metabolic signaling pathway hits related to collagen synthesis from untargeted metabolomics analysis of in vitro osteo‐spheroids treated with and without Psl. Semi‐quantitative analysis of top collagen‐related metabolites B) increased and C) decreased in osteo‐spheroids from metabolic signaling pathways in response to Psl‐treatment for 7 d. D) Schematic of the pathway for AA uptake into cells utilizing transporters SVCT2 and GLUT1. E) Representative immunostained images of SVCT2 (red) and nuclei (DAPI) in mouse cortical bone tissues after 60 days without (Veh) or with (Psl) treatment. Yellow arrows indicate SVCT2 protein expression (red). Green diamonds indicate interior of cortical bone, near bone marrow. Scale bar, 50 µm. F) Graph showing the SVCT2 immunostaining mean intensity (artificial units; A.U.) in control and Psl‐treated bone tissues. G) Representative immunostained images of GLUT1 (red) and nuclei (DAPI) in mouse cortical bone tissues after 60 days without (Veh) or with (Psl) treatment. Yellow arrows indicate GLUT1 protein expression (red). Green diamonds indicate interior of cortical bone, near bone marrow. Scale bar, 50 µm. H) Graph showing the GLUT1 immunostaining mean intensity (artificial units; A.U.) in control and Psl‐treated bone tissues. The data are expressed as mean ± SD. N = 5, n = 3; significant differences: **** p‐value < 0.0001.

2.3. AA Integration in Collagen Matrix Alters its Material Properties

To further clarify the role of AA in the bone extracellular matrix, particularly its involvement in collagen, the primary component of the bone matrix (Figure 3A), we first investigated whether the integration of AA in collagen matrix would alter its chemical properties (Figure 3B; Schematic created with BioRender). Three AA concentrations of AA were tested, with No Additional AA (No AA; Control): 0.28 mmol L⁻¹, Moderate AA (+AA): 18.28 mmol L⁻¹, and High AA: 400.28 mmol L⁻¹. High AA. Interestingly, SEM and rheological analysis revealed a severe impairment in gelation at the high AA concentration (Figure S2A,B, Supporting Information), leading to its exclusion from subsequent experiments. Elemental analysis using EDS/X revealed changes in collagen composition in the presence of AA, including a marked increase in sodium content, suggesting the potential increased availability for cellular uptake of AA (Figure 3C,D). Complementary FTIR analysis demonstrated significant alterations in amide bond stretching, with increased peak intensities at characteristic wavenumbers corresponding to amide I (≈1640 cm⁻¹), amide II (≈1540 cm⁻¹), and amide III (≈1240 cm⁻¹), indicating enhanced protein structural organization (Figure 3E; Bond “stretching” schematic created with BioRender). Additionally, the intensified broad peak between 3650–3200 cm⁻¹, corresponding to hydroxyl groups, suggests increased proline hydroxylation—supporting AA's known role in collagen stabilization and polymerization.^{[ 55 ]} To further investigate how AA contributes to acellular collagen gelation, we examined AA effects on the pH and ionic strength of the collagen gel, as it has been established that these parameters promote collagen fibrillogenesis.^{[ 56} , ^{57 ]} pH titration results showed that AA‐containing gels required a higher volume of NaOH (≈340 µL) to reach pH 7.4, compared to gels without AA (25 µL) (Figure 3F, Table S4, Supporting Information). Since the pH was similar in both gels, we next investigated whether other factors, such as ionic strength, could influence gel properties. Measurements revealed that AA increased the ionic strength of collagen gels from 118.9 mM to 159.7 mM (p‐value<0.05) (Figure 3G) and molar conductivity (Table S5, Supporting Information), suggesting that this effect results from the presence of AA on the collagen gelation process.

Figure 3.

AA enhances the chemical composition and helical polymerization of COL1A1 matrix. A) Schematic of chemical composition of human bone matrix. B) BioRender schematic of molecular mechanism by which ascorbic acid functions as a cofactor in collagen (blue helices) polymerization; hydroxylating proline and crosslinking of basic amino acids between collagen fibers. C) Elemental analysis (EDAX) representative images from Scanning Electron Microscopy (SEM) images under low vacuum mode. Scale bar: 10 µm; Representative elemental expression plots underneath from three randomized spots. D) Quantitative analysis of the elements as identified by EDAX. E) Fourier‐Transform Infrared Spectroscopy (FTIR) of COL1A1 matrix without AA (no AA) and with AA (+AA) with BioRender schematic of the bond “stretching” in the collagen molecules during polymerization reactions. G) Titration volumes of 1N NaOH required for pH 7.4 in collagen matrices and F) Conductivity (mS/cm) to assess ionic strength. The data are expressed as mean ± SD. N = 3, n = 3; significant difference: * p‐value < 0.05.

2.4. AA Implementation to the Collagen Matrix Enhances its Mechanical Properties

We further extended our study to determine whether incorporation of AA into the collagen matrix alters not only its gelation properties but also its mechanical properties. SEM analysis (Figure 4 A) revealed that AA integration significantly increased the number of collagen fibers by 1.5‐fold (p‐value<0.05) (Figure 4B), without altering the average fiber diameter of each respective diameter < or > 60 nm (Figure 4C). These biochemical and structural changes were closely linked to the mechanical behavior of the gels.^{[ 58} , ⁵⁹ , ^{60 ]} Specifically, AFM (Figure 4D) analysis showed that the presence of AA led to a 1.9‐fold increase (p‐value<0.01) in surface roughness (Figure 4E) and a corresponding average ≈10‐fold increase (p‐value<0.05) in Young's modulus (Figure 4F). In addition, we assess the viscoelastic properties by conducting rheological measurements (Figure 4G; Schematic created with BioRender). Our results demonstrated a 1.4‐fold increase (p‐value<0.05) in complex shear modulus after 50 min of polymerization with AA supplementation (Figure 4H,I). Collectively, these findings highlight the positive impact of AA integration in the collagen matrix on enhancing the biochemical and mechanical properties of the collagen matrix.

Figure 4.

AA improves the mechanical properties of COL1A1 matrix. A) Scanning Electron Microscopy (SEM) imaging: Scale bar, 5 µm. Histograms demonstrate B) total number of fibers for COL1A1 matrix without AA (no AA) and with AA C) the number of fibers with diameter less than 60 nm and more than 60 nm D) Representative 2D and 3D rending images of COL1A1 matrix by Atomic Force Microscopy (AFM). Quantitative analysis of E) Surface Roughness [Root Mean Square (RMS)] and F) Young's Modulus. G) BioRender schematic of process for rheological measurements H) Measurements of Complex Shear Modulus (Pa) over the period of 50 min COL1A1 polymerization period. I) Complex Shear Modulus (Pa) at 50 min for COL1A1 in absence of AA (no AA) and presence of AA (+AA). The data are expressed as mean ± SD; N = 3, n = 3; significant differences: * p‐value < 0.05; ** p‐value < 0.01; ****, p‐value < 0.0001.

2.5. Integration of AA in Collagen Matrix Improves its Biological Properties

Building upon the observed biochemical and mechanical enhancements imparted to the collagen matrix by AA, we next investigated whether AA also reinforces its biological functionality. To this end, we embedded osteo‐spheroids into collagen matrices with or without AA supplementation. Brightfield images (Figure S3A, Supporting Information) and nuclei staining (blue; DAPI) of osteo‐spheroids (Figure S3B, Supporting Information) revealed that AA significantly promoted osteoblast displacement from the edge of the spheroid, with a 3.5‐fold increase (p‐value<0.005) compared to control (Figure S3C, Supporting Information). Osteogenic differentiation was elevated in AA‐containing matrices treated with Psl, as indicated by a 1.5‐fold increase in ALP activity (p‐value<0.01; Figure 5A,B). Functional assays revealed enhanced matrix maturation, with osteo‐spheroids showing a 2‐fold increase in calcification by Alzarin Red staining (p‐value<0.0001; Figure 5C,D) in the presence of AA and a 4‐fold increase in mineralization by OsteoImaging (p‐value<0.005; Figure 5E). Furthermore, lysyl oxidase (LOX) activity was increased more than 3‐fold after 7 days with additional AA compared to matrices lacking supplemental AA in the presence of Psl (p‐value<0.0001; Figure S4A, Supporting Information). Additionally, qRT‐PCR data showed that AA rescued osteogenic activity under GCs conditions based on the upregulation in ALP activity and increase in key osteogenenic markers, such as BGLAP (osteocalcin) (1.4‐fold; p‐value<0.05), DMP1 (1.8‐fold; p‐value<0.05), DLX3 (1.5‐fold; p‐value<0.05), while RUNX2 expression showed a non‐significant change (1.3‐fold, p‐value>0.05) (Figure 5F). To further validate the biological protective effects of AA, we performed RT‐qPCR analysis to quantify the transcriptional levels of key collagen biosynthesis‐related markers in the presence of AA with and without Psl treatment. With Psl treatment, there is an upregulation or rescue of collagen‐synthesis‐related genes in the presence of AA, including DLX3 (1.7‐fold; p‐value<0.05), P3H3 (1.2‐fold; p‐value<0.05), LOX (2.4‐fold; p‐value<0.05), COL22A1 (1.5‐fold; p‐value<0.05), and IFITM5 (1.4‐fold; p‐value<0.05), suggesting a potential role of AA in matrix stabilization and maturation (Figure 5G). Finally, to further explore the protective effects of AA against GCs in human osteoblasts in vitro, we conducted untargeted metabolomic analysis from the secretome of osteo‐spheroids after 7 d of Psl treatment (Figure S5, and Table S2,S3, Supporting Information). Untargeted metabolomics analysis from the osteo‐spheroids with or without AA treated with Psl suggests the protective effect of AA is via improving the ECM properties against GC treatment via increased putative metabolites related to collagen synthesis,^{[ 48} , ⁴⁹ , ^{50 ]} such as Histidinyl‐Proline, Inosinic Acid, and Xestaminol C (Figure 5H).

Figure 5.

AA enhances COL1A1 biological activity and restores osteoblast function and COL1A1 regulation in the presence of Psl. A) ALP staining (red) of osteo‐spheroids embedded in COL1A1 matrices, cultured for 7 d with (+AA), and in the Psl presence (+Psl) or absence (–Psl); Scale bar: 300 µm. B) Quantification of Mean Intensity of ALP (A.U.). C) Alzarin Red (AR) activity assay of osteo‐spheroids after 7 d with or without additional AA. Scale bar, 400 µm. D) Graph showing quantification of Mean Intensity of AR (A.U.). E) OsteoImaging mineralization assay quantifications of Mean Intensity (A.U.) of osteo‐spheroids incubated with or without additional AA for 7 d. F) Gene expression analysis of osteogenic markers (BGLAP, DMP1, DLX3, RUNX2) of osteo‐spheroids embedded in COL1A1 matrices, cultured for 7 d with (+AA), and in the Psl presence (+Psl) or absence (–Psl); Heatmap demonstrates the ∆CT averages. G) Gene expression analysis of COL1A1‐related genes (PLOD1, PLOD3, DLX3, P3H1, P3H2, P3H3, LOX, SVCT2, COL1A2, COL22A1, IBSP, P4HA2, P4HA3, IFITM5) in osteo‐spheroids embedded in COL1A1 matrices, cultured for 7 d with or without ascorbic acid (AA), and in the presence (+Psl) or absence (–Psl) of prednisolone (Psl); Heatmap demonstrates the ∆CT averages H) Fold change of top collagen‐related putative metabolites elevated in the presence of endogenous ascorbic acid (+AA) within the 3D matrix of osteo‐spheroids after 7 d of Psl (+Psl) treatment, based on untargeted metabolomics analysis; fold change threshold > 2 with significance; FDR p‐value < 0.05. The data are expressed as mean ± SD; N = 3, n = 3; significant differences: ** p‐value < 0.01; *** p‐value < 0.005; **** p‐value < 0.0001.

2.6. Endogenous AA Improves Endothelial Barrier Function Following GC Treatment

We next investigated the broader impact of AA on GC‐induced effects, extending our focus beyond osteoblasts to other cell types. Our previous work demonstrated that GCs impair endothelial barrier function.^{[ 29 ]} Based on these findings, we sought to determine whether AA also exerts a protective effect on bone microvascular function under GC exposure. To address this, we developed and employed a novel 3D bicellular microfluidic model that recapitulates the structural and functional features of the bone microvasculature (Figure 6 A; Schematic created with BioRender). This model incorporates two key cell types, human osteoblasts (HOBs) spheroids (osteo‐spheroids) and human endothelial cells (ECs), within a 3D collagen matrix with or without AA (Figure 6B; Schematic created with BioRender). The system features a cylindrical, perfusable vascular channel lined with ECs, enabling dynamic analysis of vascular behavior (Figure 6C). Using this platform, we assessed the diffusive permeability coefficient (P_d) for vessel leakiness by tracking the diffusion of 70 kDa fluorescent dextran via real‐time microscopy.^{[ 29 ]} Our results demonstrated that AA significantly enhanced microvascular barrier function, showing a 1.8‐fold (p‐value<0.05) improvement to decrease vessel leakiness under normal conditions. Notably, AA also exerted a protective effect under GC exposure, rescuing barrier function by 2.25‐fold decrease (p‐value<0.01) in vessel leakiness compared to GC‐treated controls (Figure 6D,E). These findings suggest a broader role for AA in maintaining the integrity of the bone microenvironment.

Figure 6.

AA improves microvascular barrier function and decreases endothelial leakiness in the presence of Psl. A) BioRender schematic of PDMS 3D microfluidic device. B) BioRender schematic showing the cross‐sectional area of the 3D platform. The channel of endothelial cells (ECs) is formed in a 3D extracellular matrix containing osteo‐spheroids (green) within a microfabricated PDMS gasket. C) Representative confocal immunofluorescence image capturing the formed endothelium (ECs; Red) surrounding by osteo‐spheroids (green), and nuclei for DAPI (blue) (Scale bar: 100 µm). D) Representative endothelial integrity images on the devices using fluorescent‐labeled 70 kDa Texas red dextran. Scale bar, 200 µm. E) The graph demonstrates endothelial leakiness by diffusive permeability coefficient (P_d ) in the presence of AA and Psl. The data are expressed as mean ± SD. N = 3, n = 3; significant differences: * p‐value < 0.05; ** p‐value < 0.01.

3. Discussion

The widespread use of GCs continues to pose a significant clinical challenge due to their harmful effects on bone homeostasis, including disruptions in bone metabolism, reduced bone mineral density (BMD), and compromised bone strength—factors that collectively increase the risk of fractures.^{[ 30} , ⁶¹ , ^{62 ]} A key contributor to GIOP is the disruption of COL1A1 biosynthesis and metabolic pathways, which is a vital structural protein in bone.^{[ 29} , ⁶³ , ^{64 ]} This disruption arises primarily from an imbalance between collagen synthesis by osteoblasts and its degradation by osteoclasts,^{[ 22} , ^{65 ]} leading to decreased collagen content and altered collagen quality, ultimately weakening the bone matrix. Consistent with those previous findings, our data show that GCs reduce osteoblast function and COL1A1 expression. Specifically, there is a significant decrease in ALP after treatment with Psl and a decrease in collagen based on Verhoeff van Geison and immunofluorescence analysis. Furthermore, our untargeted metabolomic analysis revealed significant changes in putative metabolites involved in COL1A1 metabolism, such as Histidinyl‐Proline, N‐Methyltryptamine, and L‐Cysteine in response to GC exposure.^{[ 43} , ⁴⁴ , ⁴⁵ , ^{46 ]}

Importantly, our study identifies a previously underexplored axis involving AA (Vitamin C) metabolism as a key mechanistic pathway underpinning ECM integrity in the context of GC‐induced bone damage. While the role of AA in collagen biosynthesis and bone health is established,^{[ 41} , ⁶⁶ , ⁶⁷ , ^{68 ]} our findings offer new insights into how GCs disrupt AA‐dependent processes essential for osteoblast function and COL1A1 synthesis. As generated scaffolds mimic extracellular additional AA, the “+ AA” condition makes the final AA concentration approximately three times higher than the dietary AA ingested by an osteoporotic rat model, showing rescue effects of AA against osteoporosis induction.^{[ 40 ]} In vivo data showed that Psl downregulates critical AA and dehydroascorbic acid (DHA) transporters, named SVCT2 and GLUT1, respectively,^{[ 51} , ^{69 ]} indicating the impairment of cellular AA uptake from the cells. In addition, our untargeted metabolomic data suggested that Psl directly perturbs the ascorbate metabolic pathway, altering the levels of key ascorbate‐intake metabolic pathways, including amino acids propionylcarnitine and proline,^{[ 48} , ⁴⁹ , ^{50 ]} demonstrating that GCs induced AA intracellular deficiency.

In addition, supplementation of AA, particularly via endogenous incorporation into the collagen matrix, demonstrated enhancement of biochemical and mechanical properties of the matrix, providing functional evidence of its protective role against GC‐induced damage. Biochemical changes in collagen composition were pointed out by relevant peaks, mainly evaluated for stretching in the amide I (≈1650 cm⁻¹), II (≈1550 cm⁻¹), and III (≈1240 cm⁻¹) bonds that contribute to tighter helical structures.^{[ 55} , ^{70 ]} These changes reflect a more structurally organized matrix with improved mechanical resilience and durability by stabilizing molecular interactions, optimizing molecular geometry, and promoting dense packing^{[ 60} , ^{71 ]} based on the SEM, AFM, and rheological data. These results align with existing literature linking AA deficiency to impaired bone formation and increased fracture risk^{[ 38} , ^{41 ]} but go further by illustrating the interaction between AA availability and collagen matrix assembly in the presence of GCs. Elevated ionic strength is known to promote collagen fibril aggregation,^{[ 72 ]} resulting in denser networks and potentially enhanced crosslinking. In our study, adding AA increased the ionic strength, which may strengthen electrostatic interactions between collagen fibers, facilitating fibrillogenesis and improving the mechanical integrity of the gels.^{[ 73} , ^{74 ]} These effects are independent of pH, which was carefully maintained throughout all experiments, and are instead attributable to the rise in ionic strength from AA and/or its specific biochemical interactions with collagen. Our findings reveal a novel modification of collagen matrix properties, demonstrating that the addition of AA alone—without altering pH, polymerization temperature, or final volume—significantly increases gel ionic strength, highlighting the critical role of biochemical and mechanical factors in collagen gel behavior.

The effects of AA on acellular collagen gels arise from indirect mechanisms, such as changes in ionic strength or stabilization of non‐covalent fibrillar interactions, rather than post‐translational modification. While AA is essential for intracellular collagen hydroxylation during biosynthesis,^{[ 64} , ^{75 ]} in the absence of active prolyl or lysyl hydroxylases, it cannot induce covalent changes in extracellular, post‐gelation collagen, making direct modification biochemically implausible. Although extracellular AA can enhance collagen gelation and matrix properties in vitro, direct comparisons to clinical AA deficiency are limited, as in vivo effects on bone formation and fracture risk are driven primarily by intracellular AA's role in collagen synthesis and stabilization. Nevertheless, AA‐supplemented collagen tissue‐engineered constructs may serve as a promising strategy for developing new bone scaffolds for therapeutic applications.

Our data demonstrated that the presence of AA under GCs conditions revealed the rescue of AA‐dependent enzyme that catalyzes hydroxylation of lysine residues within collagen alpha chains, such as Procollagen Lysyl Hydroxylase 1 (PLOD1). Additionally, PLOD1 operates in a complex with P3H3 and P3H4, indicating proper collagen fibril cross‐linking and stability. The hydroxylated lysine residues generated by this complex serve as glycosylation sites and are critical for the formation of stable intermolecular cross‐links within collagen fibers.^{[ 38} , ⁴⁴ , ⁷⁶ , ^{77 ]} Collectively, those data collectively support the central role of AA not only in enzymatic collagen maturation but also in maintaining the structural and functional integrity of the collagen under GC‐induced conditions.

In addition to changes in COL1A1 metabolism, the presence of AA under GC conditions rescued osteoblast and endothelial function, highlighting the significance of endogenous AA in overall bone homeostasis. Specifically, to explore the AA effect on GCs treated osteoblasts, we used a 3D in vitro system, which incorporates osteoblast spheroids into a collagen matrix. The generated 3D osteo‐spheroids enabled the evaluation of the dynamic activity and migration of osteoblasts from their origin within the bone matrix in the presence of AA. Additionally, our data showed that AA restored osteogenic activity under GCs conditions based on the upregulation in ALP and AR activity and key osteogenic markers, such as BGLAP (osteocalcin), DMP1, and DLX3.^{[ 29} , ⁷⁸ , ^{79 ]} In addition to osteogenesis, our untargeted metabolomics revealed that the presence of AA rescued the compromised levels of collagen metabolites, such as Histidinyl‐Proline^{[ 43} , ⁴⁵ , ^{46 ]} indicating the AA protective role in the collagen bone matrix.

Furthermore, this study aimed to evaluate whether AA can attenuate the adverse effects of GC treatment on the bone microvasculature. Our previous work, employing a 3D osteoblast–endothelial microfluidic model that recapitulates the structural and functional attributes of the bone microenvironment, demonstrated that GCs impair osteoblast–endothelial interactions via the MAPK/Cx43 signaling pathway and increase endothelial barrier permeability.^{[ 29 ]} Here, building upon these findings, we integrated 3D osteo‐spheroids embedded in a collagen–ascorbic acid matrix into a vascularized microfluidic platform. Using this system, we observed that AA restored endothelial barrier integrity under GC exposure, underscoring its broader role in preserving the structural and functional integrity of the bone microenvironment.

Future studies will explore the role of AA in regulating osteoclast activity, as well as its interactions with other key bone components such as hydroxyapatite. A current limitation of this work is the absence of osteoclasts in the 3D bone microphysiological system. We anticipate that incorporating osteoclasts, together with hydroxyapatite mineral, in future models will enable a more comprehensive evaluation of the bone microenvironment. Additionally, extracellular AA can enhance collagen gelation and matrix properties in vitro, direct comparisons to clinical AA deficiency are limited, as in vivo effects on bone formation and fracture risk are driven primarily by intracellular AA's role in collagen synthesis and stabilization. Nevertheless, AA‐supplemented collagen tissue‐engineered constructs may serve as a promising strategy for developing new bone scaffolds for therapeutic applications.^{[ 80 ]} The effects of AA on acellular collagen gels arise from indirect mechanisms, such as changes in ionic strength or stabilization of non‐covalent fibrillar interactions, rather than post‐translational modification. While AA is essential for intracellular collagen hydroxylation during biosynthesis,^{[ 45 ]} in the absence of active prolyl or lysyl hydroxylases, it cannot induce covalent changes in extracellular, post‐gelation collagen, making direct modification biochemically implausible.

Overall, these findings demonstrate that AA supports bone extracellular matrix production and function by regulating collagen synthesis, osteoblast activity, and endothelial function, underscoring its multifaceted role in bone homeostasis and disease. Building on this work—and on prior evidence from a vitamin C–deficient rat model showing that dietary AA supplementation enhances osteoblastogenesis, suppresses osteoclastogenesis, and promotes bone formation^{[ 39 ]} – future studies will aim to evaluate the incorporation of AA into a GIOP mouse model to assess its potential protective or restorative effects against drug‐induced osteoporosis. This benefit highlights the therapeutic potential of AA in bone disorders and suggests its value as an adjunct therapy for patients with GIOP.

4. Experimental Section

Cell Culture

Human osteoblasts (HOBs) (PromoCell) were cultured in osteoblast growth medium (GM) (PromoCell) and differentiated in osteoblast mineralization medium (MM) (PromoCell). Human umbilical vein endothelial cells (ECs) (Lonza) were cultured in EC growth medium EGM‐2 (Lonza). All experiments were performed with HOBs and ECs in passages 3 to 6. To achieve HOBs labeled with a green fluorescent protein and ECs labeled red with mApple fluorescent protein, the cells were transduced with a lentiviral construct pCSCG‐EGFP and pCSCG‐mApple (Addgene), respectively, as described previously.^{[ 29 ]}

Collagen I Matrix Formation

Collagen I (COL1A1) matrices were generated by combining 5 µg mL⁻¹ fibronectin (Corning) with 3 mg mL⁻¹ rat tail telo‐collagen Type I (BD, Corning), enabling fibril formation and subsequent polymerization following matrix preparation. The mixture was supplemented with or without an additional 18 mmol L⁻¹ L‐ascorbic acid (AA, ThermoFisher), 1× M199 medium (ThermoFisher), 1 mmol L⁻¹ HEPES, 0.1 mol L⁻¹ NaOH, and NaHCO₃ (0.052% per volume fraction). Polymerization was then carried out at pH 7.2 for 45 min at 37°C. Preliminary experiments testing three concentrations of AA (No Additional AA: 0.28 mmol L⁻¹, Moderate AA: 18.28 mmol L⁻¹, and High AA: 400.28 mmol L⁻¹) revealed the “High AA” condition to be disruptive to collagen polymerization and was therefore excluded from future experiments. No AA (Control; Ctrl; 0.28 mmol L⁻¹) indicates “no additional AA” outside the baseline media and reagents. With (+) AA (18.28 mmol L⁻¹) indicating “18 mmol L⁻¹ additional AA” on top of baseline media and reagents. See Table S4 (Supporting Information) for details on reagent breakdown for “No AA” and “+ AA” matrices.

Generation of Osteo‐Spheroids

HOB cells seeded at density of 10,000 cells/well in 24‐well low‐adhesive surface plates (Corning). After 48 h, spheroids were gently collected and added directly to the COL1A1 matrix with or without AA (ThermoFisher), followed by polymerization for 45 min at 37°C. Next, the embedded osteo‐spheroids were treated with 100 nmol L⁻¹ prednisolone (Psl) for 7 d.

ALP Staining

The osteo‐spheroids were assayed for osteogenesis by staining for ALP with Leukocyte Alkaline Phosphatase kit (Abcam) based on the manufacturer's instructions. Briefly, the spheroids were fixed in in 4% (by volume fraction) PFA (Sigma–Aldrich) for 20 min at 37°C, washed twice in PBS. They were permeabilized with 0.1% (by volume fraction) of Triton‐X 100 in PBS for 1 h and treated with blocking solution of 0.01% (by volume fraction) Triton‐X 100, 5 g/100 mL goat serum (Sigma–Aldrich) in PBS overnight at 4°C. Samples were stained with ALP for 1 h Imaging was performed the following day using a microscope (Olympus) or Nikon CSU‐W1 Spinning Disk Confocal with SjuhyoRa super‐resolution and TIRF microscope, and image analysis was conducted in ImageJ using maximum intensity Z‐projection and channel merging.^{[ 81 ]}

OsteoImaging Mineralization Assay

The osteo‐spheroids were assayed for mineralization using the OsteoImaging Mineralization Assay (Lonza) based on the manufacturer's instructions. Briefly, after 7 d of growth, the spheroids were fixed in in 4% (by volume fraction) PFA (Sigma–Aldrich) for 20 min at 37°C, washed twice in PBS. The OsteoImage solution was prepared fresh and added to each well, after which the samples were incubated at RT, protected from light, for 30 min. Mineralization was captured using a Plate Reader (Agilent) with excitation/emission of 492/520 nm.

Alzarin Red Calcification Assay

The osteo‐spheroids were assayed for calcification using Alzarin Red (AR; Sigma–Aldrich) based on the manufacturer's instructions. Briefly, after 7 d of growth, the spheroids were fixed in 4% (by volume fraction) PFA (Sigma–Aldrich) for 20 min at 37°C, washed twice in PBS. AR staining solution was added to each well, after which samples were incubated at RT, protected from light, for 30 min. Each well was washed thoroughly, and images were captured using a microscope (Olympus), and image quantification was conducted in ImageJ.

Lysyl Oxidase (LOX) Activity

LOX activity in osteo‐spheroids was measured using the Lysyl Oxidase Activity Assay Kit (Fluorescent; Abcam) following the manufacturer's instructions. Briefly, after 7 d of growth, spheroids were fixed in 4% (per volume fraction) PFA (Sigma–Aldrich) for 20 min at 37°C and washed twice with PBS. Fresh LOX solution was then added to each well, and samples were incubated at 37°C for 30 min in the dark. LOX activity was measured using a plate reader (Agilent) with excitation/emission at 540/590 nm.

Fabrication of Microfluidic Platform

Scaffolds for the microfluidic devices were created as described previously.^{[ 29 ]} Briefly, PDMS devices were treated with 0.01% (by mass fraction) poly‐L‐lysine (PLL; Sigma–Aldrich) and 0.5% (by mass fraction) glutaraldehyde (Sigma–Aldrich). After washing overnight in water, steel acupuncture needles (diameter = 300 µm Seirin, Kyoto, Japan) were introduced into the devices and ultraviolet (UV)‐sterilized for 30 min before adding COL1A1± AA with or without osteo‐spheroids. After gel polymerization, acupuncture needles were carefully removed to create hollow channels. ECs were then seeded into the channels at a concentration of 10⁶ cells/mL. For GCs conditions, ECs were pretreated with 100 nmol/L Psl (Sigma–Aldrich). Devices were inverted for 4 min to allow cell adhesion to the upper channel surface, then returned to an upright position for an additional 4 min to promote adhesion to the lower surface. Cells adhering to reservoir regions were removed with a pipette tip, and non‐adherent cells were flushed out with fresh EGM‐2 medium. Devices were placed on a platform rocker (BenchRocker BR2000) and incubated at 37°C with 5% (by volume fraction) CO₂.

Endothelial Barrier Integrity Measurement

Since osteoblasts support vessel maturation and barrier function,^{[ 29 ]} endothelial leakiness (P_d) was measured near osteo‐spheroids placed within 100 µm of the endothelial tube. The P_d was measured by adding red fluorescent dextran (70 kDa Texas Red, ThermoFisher) into EGM‐2 medium at a concentration of 12.5 µg mL⁻¹. Dextran diffusion was imaged in real‐time with an Olympus IX71 at 10 × magnification. A time sequence image was analyzed by evaluating the mean intensity (artificial units; A.U.) over a large region next to the cellular vessel in successive images for more than 30 s. Analysis was started at t = 0 when the vessel was fully saturated and fluorescence was measured via ImageJ for the following 10 frames collected at 3 s/frame. The time derivative of the intensity was determined by linear regression for each region. The time derivative of the intensity, the mean intensity (I), and the capillary radius (r) were used to determine the diffusive permeability coefficient (P_d ) by the equation $=\frac{dI}{dt}\times \frac{r}{2I}$.

Real‐Time Quantitative PCR Analysis

Total RNA was isolated after 7 d of osteo‐spheroids in COL1A1 with indicated media or treatments using Trizol Reagent (Invitrogen) following the manufacturer's instruction. SuperScript III cDNA Synthesis kit (ThermoFisher) was used to synthesize cDNA from RNA in Applied Biosystems Veriti 96‐Well Thermal Cycler. The real‐time quantitative PCR (RT‐qPCR) was performed using a PowerTrack SYBR Green Master Mix and ViiA 7 Real‐Time PCR System with 384‐Well Block (Applied Biosystems) in accordance with the manufacturer's instructions. The primers used in this study are described in Table S1 (Supporting Information). The expression level of each gene was normalized to the expression level of 18S ribosomal RNA (18S rRNA) internal control. The relative gene expression was calculated and reported as ∆CT by the standard curve method using the target Cq values and the Cq value for 18S rRNA in each lane.

In Vivo Studies

All procedures and experiments involving live animals were approved by the Georgetown University Institutional Animal Care and Use Committee (IACUC Protocol Number: 2022‐0058. Title: Effect of glucocorticoids on bone vascularization).^{[ 29 ]} Four‐ to six‐month‐old Swiss Webster mice were purchased from Charles River Laboratories. The mice were randomized into two groups with 10 animals in each, and an equal number of males and females were used to begin with. The GC‐treated group had slow‐release pellets subcutaneously implanted in the dorsal scapular region with Psl at the dose of 3.8 mg/60 d = 2.1 mg/kg/d, assuming 30 g as the average weight of these mice (Innovative Research of America). The control group received placebo pellets subcutaneously in the same area containing the corresponding vehicle. Throughout the experiment, mice were monitored for body weight loss. After 60 d post pellet implantation, the mice were euthanized, the long bones harvested, fixed in 10% (by volume fraction) neutral buffered formalin, decalcified in 14% (by volume fraction) EDTA, paraffin‐embedded, and subjected to histopathological analysis.

In Vivo Immunostaining

The tissue sections were deparaffinized at 60°C then rehydrated in a series of 5 min per reagent at RT: two times in 100% (by volume fraction) xylenes, two times 100% (by volume fraction) EtOH, then 95%, 80%, and 75% EtOH (by volume fraction) followed by tap water rinse before the antigen retrieval step. For antigen retrieval, 1x citrate buffer of pH 6.0 was used and brought to quick boil for roughly 10 s and placed on 90°C hot plate for 10 min, followed by 30 min cool down on benchtop at RT. Slides were then gently washed two times with distilled water and 1x with Tris‐buffered saline with Tween 20 (TBST) before blocking for 1 h in TBST + 5% (by volume fraction) goat serum. Next the slides were incubated with the following primary antibodies overnight at 4°C: rabbit anti‐human Sodium Vitamin C co‐transporter 2 (SVCT2) (1:100 dilution, ThermoFisher, AST‐022); rabbit anti‐human Glucose transporter 1 (GLUT1) (1:100 dilution, ThermoFisher, SA0377); rabbit anti‐human COL1A1 (1:100 dilution, Proteintech). The next day, tissue slides were gently washed twice in buffer solution (PBS or TBS), followed by the addition of secondary antibody mixture, consisting of DAPI (1:1000, ThermoFisher), goat‐anti‐rabbit IgG Alexa Fluor 568 (1:100, Invitrogen, A‐11036) for detection of primary antibody for 1.5 h and then washing with PBS. Slides were briefly examined under a fluorescence microscope to confirm staining quality before undergoing a graded EtOH dehydration series at RT with 70%, 80%, two changes of 95%, and two changes of 100% EtOH (by volume fraction) every 3 min. Slides were then mounted using mounting medium (Fisher Scientific, 23–245691), coverslipped, and allowed to dry for a minimum of 12 h. Imaging for in vitro and in vivo studies was performed using a microscope (Olympus) and Nikon CSU‐W1 Spinning Disk Confocal with SoRa super‐resolution and TIRF microscope, and image analysis was conducted in ImageJ using maximum intensity Z‐projection and channel merging.^{[ 81 ]}

Assessment of Roughness and Stiffness

COLIAI matrix samples with and without AA were prepared as described above. The following day, samples were fixed in 4% (per volume fraction) paraformaldehyde (PFA; Sigma–Aldrich) in PBS for 30 min at 37°C, washed three times with PBS, dried with ethanol (EtOH) 100% (by volume fraction) two times for 10 min, then dried in a 65°C oven for up to 2 h. Subsequently, the samples were allowed to equilibrate to roughly 25°C, or room temperature (RT), in a partially covered container in a room with 55% humidity for 1 to 3 d to prevent rapid drying and detachment (curling) from the glass substrate. Controlled humidity during drying was critical to maintain sample integrity and ensure consistent, representative force curve measurements. Next, Atomic Force Microscopy (AFM) was conducted using the NTEGRA Prima system (NT‐MDT). Surface topography was evaluated in semi‐contact mode using HA‐NC probes (K‐TEK Nanotechnology), enabling high‐resolution imaging of the dried matrix surface to visualize 3D fiber structures and assess surface roughness. For stiffness measurements, force curves were obtained in contact mode using a NSG03 silicon probe (K‐TEK Nanotechnology). Young's modulus (E) values were calculated using NT‐MDT Image Analysis P9 software.

Assessment of Rheological Properties

Shear modulus data were obtained through oscillatory and rotational rheological measurements using a rheometer with a thermal enclosure (MCR302, Anton Paar), equipped with a 50 mm parallel‐plate measuring system and a 1 mm gap height. Approximately 3 mL of COL1A1 solution with and without AA, as described above, was loaded onto the rheometer stage at 37°C immediately after bringing the pH to 7.2. After setting the gap, excess material at the edges was trimmed, and water was added around the sample area to minimize evaporation and control humidity before sealing the thermal enclosure. Measurements were conducted in triplicate (N = 3, n = 3) at a constant temperature of 37°C, with small‐amplitude oscillations (5%, well within the linear viscoelastic range) recorded every 10 s over a 50 min period (300 total data points per run). Storage modulus (G′), loss modulus (G″), and corresponding raw stress‐strain signals were monitored. The complex shear modulus was defined as G(ω) = G′(ω) + iG″(ω)**, and the magnitude |G| = √(G′² + G″²) at ω = 1 rad/s.

Scanning Electron Microscopy (SEM) and Energy‐Dispersive X‐ray Spectroscopy (EDS/X)

Collagen containers were fabricated by bonding polydimethylsiloxane (PDMS; Sylgard 184, Dow‐Corning; Krayden) blocks to microscope cover glasses. Briefly, PDMS prepolymer and curing agent (10:1 ratio) were mixed, degassed in a desiccator, and poured into Petri dishes. After curing at 80°C for 6 h, the PDMS was cut into 10 mm × 10 mm × 4 mm blocks, and 5 mm diameter central wells were created using a 5 mm biopsy punch. The PDMS blocks were bonded to cover glasses via oxygen plasma treatment. Each PDMS well was filled with 200 µL of COL1A1 solution and polymerized at 37°C for 1 h. Medium was then added to avoid COL1A1 drying and the containers were incubated overnight at 37°C. A serial dehydration protocol adapted from previous studies^{[ 82 ]} was performed to prepare the gels for SEM and EDS. Collagen gels with or without AA were fixed in 4% (per volume fraction) PFA (Sigma–Aldrich) in PBS for 1 h at RT, washed three times with PBS, and twice with distilled water. Samples were then incubated at RT with 1% (by volume fraction) of Osmium Tetroxide (OsTe) for 1 h then washed three times with distilled water. Dehydration was performed through a series of graded EtOH solutions (30%, 50%, 70%, 90%, and 100%, EtOH by volume fraction in distilled water), followed by graded EtOH/hexamethyldisilazane (HMDS) mixtures (33%, 50%, 66%, and 100% HDMS by volume fraction). Samples were air‐dried for 24 h on glass slides. Dried gels were mounted on aluminum stubs using conductive carbon tape and imaged using a Field Emission Scanning Electron Microscope (FESEM) Teneo LV (FEI, GWNIC) in high vacuum mode at 1–2 kV, with magnifications of 20,000 x. Elemental analysis was performed using the EDAX FEI APEX TEAM stage‐control software. Samples were imaged using low vacuum mode at a higher voltage and scanned each spot (n = 3) on each sample (N = 3) for comprehensive elemental analysis over 10,000 x. Analysis for fiber diameter and total fibers was performed with DiameterJ, an ImageJ plugin, and results were presented as histograms.

Fourier‐Transformed Infrared Spectroscopy (FTIR)

FTIR spectra (400‐4000 cm⁻¹) were acquired using a Nicolet iS50 FT‐IR Spectrometer (ThermoScientific). All spectra from wet collagen samples were background corrected on the OMNIC software and collected in triplicates per biological replicate with 64 scans per measurement at RT. For visualization purposes, spectra from each condition were averaged across replicates.

Assessment of Ionic Strength

Conductivity was measured automatically during zeta potential analysis using the Particle Lite instrument (Litesizer 500, Anton Paar Kalliope), which had been pre‐calibrated with a standard conductivity solution. 1 mL of each collagen suspension was transferred into a clean zeta potential/conductivity cuvette. Measurements were performed at 25°C, and results were reported in milliSiemens per centimeter (mS/cm).

Untargeted Metabolomics

The conditioned media (CM) was collected after 7 d of in vitro culture of osteo‐spheroids treated and untreated with Psl, with or without AA, and prepared for metabolomics.^{[ 75 ]} The collected CM was spun down at 1,000 x g for 5 min at 4°C and added to medium extraction solution (MES) at a 1:6 ratio of CM:MES. MES consisted of 75% (by volume fraction) acetonitrile, 25% (by volume fraction) methanol, 0.2% (by volume fraction) formic acid. After vortexing, samples were centrifuged at 18,000 x g at 4°C for 10 min. Supernatants were then filtered in Pall Nanosep 0.2 µm bioinert centrifugal filters (ODM02C34) at 14,000 x g at 4°C for 5 min. An aliquot was placed in a glass liquid chromatography (LC) vial and equal volumes from each sample were pooled as a quality control (QC). Samples were injected into a Waters Corp. Ultra Performance Liquid Chromatography (UPLC) system coupled to a Xevo G2 time‐of‐flight mass spectrometry (MS) system. Data were collected in both electrospray ionization (ESI) positive (POS) and negative (NEG) modes with data independent acquisition in MSE mode.^{[ 75} , ^{83 ]} Untargeted putative metabolites were reported for discovery purposes of global changes between conditions. Analysis: data were deconvoluted and reverted to.csv files using Progenesis QI (NonLinear Dynamics, Newcastle, UK). POS and NEG mode putative IDs were assigned based on searches through databases (empirical METLIN, LIPID MAPS, LipidBlast, and HMDB) in Progenesis QI with a ppm error of 10. These updated files with the descriptions matched for peak intensities were analyzed with the online software MetaboAnalyst 6.0 for Statistical Analysis [one factor].^{[ 84 ]} Using the Variance filter and interquartile range, 25% of the lowest values were filtered out. With data already normalized to QC before uploading, no further normalization or data scaling was conducted. Data were log₁₀ transformed. More information on MS collection parameters and specific putative metabolites pointed out are found in Tables S2,S3 (Supporting Information).

Statistical Analysis

Statistical analysis was conducted using GraphPad Prism 10. Sample sizes for each experimental group are provided in the figure legends (N). All in vitro experiments were performed in triplicate and independently biologically repeated three times for N = 3, n = 3. Data was assessed for normality, and no significant variation between groups was detected. For comparisons between two groups, a two‐tailed Student's t‐test was applied. For experiments involving multiple time points or treatment conditions, one‐way or two‐way ANOVA was used, followed by Tukey's post hoc test where appropriate.

For metabolomic assays, the reported putative metabolites and pathway results were analyzed for significance using FDR: p‐value<0.05, and significant hits were exported to a.csv file with fold changes and significance, as well as figures exported of heatmaps, volcano plots, biplots, and KEGG Global Network Analysis. MetaboAnalyst 6.0 software was used for pathway analysis, then using the KEGG pathways and significance ratios, R Studio software was used to generate the pathway dot plot for more comprehensive visualization of the exported results, as seen. R Studio code for the dot plot was accessible on GitHub per request.

For in vivo studies, the unit of measurement was the individual animal. A two‐way ANOVA was used to assess differences between groups at each time point, as well as changes over time within groups. Sex was included as a biological variable in the experimental design. A priori power analysis, assuming a type I error rate of 5% and a statistical power of 90%, indicated that a sample size of at least n = 5 animals per group was sufficient to detect statistically significant differences. All in vivo experiments were conducted and analyzed in a blind manner to ensure unbiased interpretation. Results are presented as mean ± standard deviation (SD), and p‐value<0.05 was considered statistically significant.

Conflict of Interest

The authors declare no conflict of interest.

Author Contributions

Conceptualization, MC and SA; methodology, MC, YK, XZ, JS; software, MC and SA; validation, MC and SA; resources SA; writing‐original draft preparation, MC and SA; writing‐review and editing, MC, YK, JS, EL, JK, SA. All authors have read and agreed to the published version of the manuscript.

Supporting information

Supporting Information

Supplemental Movie 1

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.