Decreased Bone Strength in Skeletally Mature Female Mice with T2D is Associated with Changes in OLCN and Spatially Compromised Architecture

Abstract

Women with Type-2 diabetes (T2D) have a higher incidence of fractures and associated mortality compared to men. However, most animal models for studying diabetes-induced bone effects use males and the impact of T2D on bone quality in skeletally mature females remains unknown. We developed a mouse model to determine the effect of T2D on female bone quality. T2D was induced in 16-week-old female C57BL/6J mice using a High-Fat Diet and Streptozotocin (HFD+STZ), controls received a Low-Fat Diet and sham injections (LFD+VEH). The diabetic group displayed hyperglycemia, hypoinsulinemia, and increased body fat. T2D altered bone architecture spatially, the T2D group displayed decreased cortical and trabecular bone volume (BV) and total volume (TV) with varying magnitude at specific locations compared to the control. T2D also reduced bone yield and ultimate loads under four-point bending, without affecting tissue-level properties. Moreover, changes in TV with T2D explained up to 70% of the variance in bone strength, suggesting that the weakening effect of T2D on female bone strength is architecture-driven. Compromised architecture with T2D was associated with changes in the Osteocyte Lacuno-Canalicular Network (OLCN). T2D decreased canalicular density, the total number of nodes and increased lacunae surface area. These changes in the OLCN with T2D explained up to 37% of bone architecture variance. In summary, our novel T2D female mouse model displayed a bone phenotype with compromised OLCN associated to impaired architecture, which led to decreased bone strength. These outcomes suggest that the effect of T2D on female bone strength is architecture-driven rather than material-driven.

Graphical Abstract

INTRODUCTION

Individuals with T2D display a higher risk of fracture compared with healthy subjects, yet this increased fracture prevalence cannot be explained by changes in bone mineral density (BMD) alone, as people with T2D often exhibit normal to higher BMD levels.^1–3 The impact of T2D on bone quality remains unclear, as some human studies describe decreased trabecular bone mass accompanied by increased bone density and stiffness,⁴ while others report that T2D decreases bone quality by compromising bone architecture, mechanical properties, and composition.⁵ Comprehensive studies of the complex relationship between material traits and mechanisms by which T2D impairs bone quality in humans are not feasible due to the necessity of invasive and often destructive assays.^5–7 Thus, in order to conduct cellular and biomechanical mechanistic studies on the effect of the disease, animal models are required.

Current T2D animal models include various induction methods, such as diet manipulation, gene modification and pharmacological interventions.^{8, 9} Streptozotocin (STZ), an antibiotic that ablates pancreatic beta-cells, can be used to develop a phenotype similar to Type 1 Diabetes.^{10, 11} High-Fat Diet (HFD) models have been proposed to induce T2D, but since rodents tend to display high glucose resilience, it takes several months to induce diabetes, increasing the variability in the onset of the disease.¹² The combination of HFD and low-dose STZ develops the hallmarks of T2D, promoting the development of obesity, hypoinsulinemia, and hyperglycemia.^13–17

In male mice, T2D induced by HFD+STZ weakens bones by decreasing cortical bone mineral density (BMD) and cortical thickness, reducing whole-bone stiffness and failure load.¹⁵ Decreased load bearing capacity and compromised architecture are associated with decreased bone turnover and increased levels of advanced glycation end products (AGEs) in bone and blood.¹⁵ Additionally, T2D accelerates the accumulation of senescent osteocytes, which contributes to decreased tissue quality, increasing bone fragility and fracture risk.^{15, 18} Impaired glucose metabolism also induces changes in the osteocyte lacuno-canalicular network (OLCN), decreasing dendrite number, tortuosity, and connectivity. These changes in OLCN morphology and connectivity, correlate with decreased cortical bone volume fraction, whole-bone ultimate load and perilacunar elastic work; supporting the role of a diminished OLCN as a contributing factor to reduced whole-bone mechanical properties with diabetes.¹⁹

In humans, the effect of T2D on bone quality is sex dependent, with females displaying higher risk of fracture and associated mortality compared to males.²⁰ However, most animal models used to determine the mechanisms by which T2D impacts bone quality use male mice.^{15, 17, 21, 22} The widespread use of male mice compared to females is mostly attributed to that STZ interacts with estrogen, complicating the disease induction in females.^{10, 23}

Currently, only a few studies have successfully induced T2D in young female rodents, but there are no animal models available to study the effects of T2D on skeletally mature females, which corresponds to the majority of T2D prevalence in humans.^{24, 25} The absence of skeletally mature female animal models and mechanistic studies limit the understanding of how T2D affects female bone quality, creating a knowledge gap that precludes the development of diagnostic tools and treatments for diabetic women.

The objective of this study was to develop an animal model to determine the effect of T2D on the mature female skeleton. Specifically, this study aimed to characterize the diabetic female bone phenotype and to identify how changes in bone architecture, mechanical properties, and osteocyte lacunar-canalicular network (OLCN) contribute to the reduction in bone quality associated with T2D.

Since current studies support the role of changes in bone architecture as driving factor of the weakened bone phenotype with T2D, and previous research in our lab highlighted the contribution of changes in OCLN to decreased bone architecture,^{15, 19, 26} we hypothesize that T2D decreases female bone volume, compromising bone load bearing capacity and that the compromised bone architecture with T2D is associated with changes in the Osteocyte Lacuno-Canalicular Network (OLCN).

MATERIALS & METHODS

Animal Model

All animal procedures were approved by the University of Michigan Institutional Animal Care and Use Committee (IACUC). Ten-week-old C57BL/6J female mice were acquired from Jackson Laboratory (Bar Harbor, ME) and acclimated for two weeks (Figure 1A). After the acclimation period, mice were randomly divided into two weight-matched groups. The control group was fed a Low-Fat Diet (LFD), and the treatment group was fed a High-Fat Diet (HFD) for 4 weeks (Research Diets, LFD: D12450 and HFD: D12492). The LFD and HFD have the same sucrose content, and provide 10% and 60% of the calories from fat respectively (Supplemental Table 1). LFD was used for the control group instead of standard chow to control for source (semisynthetic) and amounts of micro and macronutrients, avoiding confounding factors.²⁷ Following 4 weeks, animals were continued on their respective diets and diabetes was induced by daily intraperitoneal (IP) administration of Streptozotocin (STZ) for seven days to the HFD group, with the following concentrations: Days 1 and 2: 50 mg/kg, days 3–5: 35 mg/kg, and days 6–7: 50 mg/kg. Mice that did not develop diabetes received one additional STZ booster shot (50 mg/kg) 14 days after the first STZ injection. Mice in the control group (LFD) received sodium citrate vehicle injections (VEH). After 2 weeks, all mice in the diabetic group developed diabetes (blood glucose>250 mg/dl). 12 weeks after establishing diabetes, mice were euthanized by carbon dioxide overdose at 30 weeks of age. Mice had ad-libitum access to food and water and were kept at a constant temperature and humidity with a 12-hour light cycle. The diabetic group (HFD+STZ) started with a higher number of mice (n=14) than the control group (n=10), accounting for an expected higher mortality associated to STZ administration. The endpoint sample size was n=13 for the diabetic group and n=8 for the control group, respectively. All death events occurred during IP administration at different time points.

Figure 1.

T2D was induced in skeletally mature female C57BL/6J mice via a combination of Streptozotocin (STZ) and high fat diet (HFD). (A) Study design schematic describing the experimental groups and timeline. Created in BioRender. Kohn, D. (2024) https://BioRender.com/f49h757 (B) Representative images of female C57BL/6J mice from each experimental group at the study endpoint (30 weeks of age). (C) Blood glucose measurements throughout the study performed by lateral tail vein puncture. (D) Endpoint blood glucose. (E) Endpoint glycated hemoglobin (HbA1c %) used as indicator of long-term blood glucose. (F) Endpoint fasting serum insulin levels. (G) Body composition analysis of adipose tissue expressed in terms of fat percentage at the study endpoint using magnetic resonance imaging (MRI). All outcomes were significantly different in the T2D group vs. controls (LFD+VEH n:8, HFD+STZ n:13).

The animal model timeline and inclusion of a diet lead-in phase were informed by previous studies to ensure comparable timepoints and mouse age.^{15, 28} Multiple low-dose injections were chosen over a single high-dose regimen to avoid excessive DNA damage in beta cells and to allow for more precise control over disease onset via periodic blood glucose monitoring.^{11, 22} The STZ dosage was increased compared to male mouse protocols to compensate for the resistance of female mice to STZ-induced diabetes, which is attributed to the interaction of STZ with estrogen.^{10, 22} Since the goal of this study was to determine the effect of T2D on female bone quality, only a control (LFD+VEH) group and a diabetic (HFD+STZ) group were included. The study of the effects of HFD and STZ alone are outside the scope of this study because neither HFD nor STZ treatment alone are sufficient to induce diabetes.¹⁵

Blood and Serum Measurements

Blood glucose was measured weekly with a glucometer (OneTouch Ultra2) by tail vein puncture. Glycated hemoglobin was measured every 4 weeks during the disease state using the A1cNOW kit (PTS Diagnostics, Whitestown, IN, USA). At the study endpoint, mice were fasted for 4 hours before euthanizing and blood glucose was recorded. Following euthanasia, serum was collected and aliquoted. Serum insulin was measured using Ultra-Sensitive Mouse Insulin ELISA Kit (CrystalChem, Elk Grove Village, IL, USA).

MRI (Body Composition)

Mouse body composition was analyzed in vivo 2 days before the study endpoint using a Nuclear Magnetic Resonance-based Bruker Minispec LF 90II (Billerica, MA, USA). The analysis was performed by the University of Michigan Animal Phenotyping Core, and measured mouse body weight, fat mass, lean mass, free water and total water.

Bone Sample Preparation and Specimen Distribution

Bone sample distribution and the assays applied on each bone are presented in Supplemental Figure 1. Following euthanasia and blood collection, left tibiae were dissected, placed in RNALater (Thermo Fisher Scientific, Waltham, MA, USA), and stored at −20°C until processing. Subsequently, femora, contralateral tibiae, radii and humeri were dissected and wrapped in gauze soaked with calcium-buffered PBS and stored at −20°C until processing.

Right femora were embedded in polymethyl methacrylate (PMMA, Koldmount, SPI Supplies) and cured at room temperature for 2 hours. Two sections from the midshaft were transversely cut with a low-speed circular diamond saw (South Bay Technology, Model 650). The first section was used for nanoindentation with an approximated thickness of 2mm and the second section was used for OLCN analysis with an approximated thickness of 700μm. The embedded blocks were wrapped in gauze soaked in calcium-buffered PBS and stored in a 1.5mL tube at −20°C until the time of testing or further processing.

Micro-Computed Tomography (μCT)

Bone architecture analysis was performed following ASBMR guidelines.²⁹ Bone microarchitecture properties were evaluated using μCT (μCT100 Scanco Medical, Bassersdorf, Switzerland) with a 12μm voxel size. Volumetric (3D) properties were analyzed using Scanco’s software. For cortical and trabecular bone the following properties were evaluated: total volume (TV), bone volume (BV), bone volume fraction (BV/ TV), bone mineral density (BMD) and tissue mineral density (TMD). In addition, trabecular thickness, number, spacing, connectivity density, and structural model index (SMI) were calculated.

Bone architecture was analyzed at standardized anatomic landmarks defined as fixed percentages of total femoral length. This approach ensures that corresponding regions are evaluated across all specimens, regardless of group or overall bone length. Specifically, we used the following anatomical references: 15% of femur length (base of femoral neck), 33% (third trochanter), 50% and 67% (diaphysis), and 85% (distal metaphysis, just proximal to the distal growth plate) (Figure 2A). Prior to μCT analysis, all selected cross-sections were reviewed in a blinded manner to confirm correspondence with these landmarks.

Figure 2.

T2D decreased cortical and trabecular bone architecture properties in a spatially dependent manner (LFD+VEH n:8, HFD+STZ n:13). (A) Regions of interest expressed as a percentage of the distance from the proximal end: base of the femoral neck (15%), third trochanter (33%), diaphyseal midpoint (50%), distal diaphysis (67%) and distal metaphysis (85%) (B) Bone length (C) Cortical Total Volume (TV), volume enclosed by the periosteal region (D) Cortical Bone Volume (BV), (E) Cortical Tissue Mineral Density (TMD) (F) Trabecular TMD (G) Trabecular Structure Model Index (Tb. SMI) (H) Trabecular TV.

Cortical bone was evaluated at 5 bone lengths, while trabecular bone was only evaluated at the metaphyses. At each area of interest, measurements were taken in a field of 480μm centered on the bone length of interest. Cross-section properties were analyzed at the same 5 bone lengths using ImageJ on the DICOM-exported scan files. The BoneJ and MomentMacro plugins were used to evaluate: cross-section area (CSA), bone marrow area, bone perimeter, cortical thickness, moments of inertia, and Feret distances.³⁰ The Feret distances provide a standardized method to measure the size and orientation of irregular objects, such as bone. They quantify the distance between two parallel lines that are tangential to the objecťs boundary at a given angle. Essentially, this represents the distance between the jaws of a caliper when used to enclose the object at a specific orientation. Differences in Feret distances and orientation between the control and T2D groups would indicate that T2D affects bone cross-sectional shape or orientation.

Four-Point Bending

After μCT scanning, left femora were kept at −20°C in calcium buffered PBS and were thawed before mechanical testing. Four-point bending tests were performed using an eXpert 450 Universal Testing Machine (Admet, Norwood, MA). The contact point geometry used was rounded, the loading span was 3.00 mm in the upper position with a support span of 9.15 mm in the lower position. The anterior surface of the midshaft was in tension and the bones were kept hydrated throughout testing. Femora were loaded to failure monotonically at 0.01 mm/sec with a 0.5N preload. The yield point was calculated using the 0.2% offset method and material properties were evaluated using beam theory and the properties of the cross-section at the region where failure occurred. Calculations were performed using custom MATLAB (Mathworks, Natick, MA) scripts.³¹ Mechanical testing outcomes were not normalized to femoral length. Since the inner and outer spans of the testing setup were fixed for both groups, the moment arm was consistent across all samples.

Furthermore, differences in bone length between groups were less than 0.5 mm, which did not affect testing, as all femora exceeded the length of the outer span of the four-point bending supports by at least 3 mm. Aspect ratios for base/span, height/span, and length/span were calculated for all femoral samples and compared between groups. There were no statistically significant differences in these ratios, and all specimens met the minimum requirements specified by ASTM standards for 4-point bending. This compliance ensures that the influence of shear stresses on failure loads was minimized across all groups

Fractography

Fractured bones were scanned via μCT and the fracture pattern was analyzed to determine whether this model and age of diabetes onset promoted a brittle phenotype. All bone scans were loaded in a blinded manner into the Dragonfly software (version 2024.1; Object Research Systems Inc, Montreal, Canada) and using the segmentation and rendering tools, bones were isolated and analyzed to determine the type of fracture. The qualitative analysis categorized the fractures as oblique (ductile) or transverse (brittle).

Nanoindentation

Sample preparation for nanoindentation testing is described in the Supplemental Material. Nanoindentation was performed using a Triboindenter Ti-950 (Hysitron, Minneapolis, MN, USA) with Berkovich tip. The loading scheme included 5 seconds to reach a 10mN load, 60 seconds of dwell to evaluate plastic behavior, and 5 seconds for unloading. Elastic modulus and hardness were calculated from the unloading segment of the curve. The integrated areas under the load-displacement curve were calculated to assess plastic and elastic work. Bone surrounding 4 osteocyte lacunae was analyzed in both anterior and lateral zones. The anterior area was chosen to match the highly organized OLCN location, while the lateral area was selected for its different strain magnitude and direction.³² Each lacuna was evaluated in two areas: the perilacunar area, within one lacunar diameter from the wall, and the intracortical area, at least two diameters away. Samples were hydrated throughout the experiment using calcium-buffered PBS and four indentations were made in each region. (Figure 5A).

Figure 5.

T2D decreased hardness and increased plastic work in a spatially dependent manner (LFD+VEH n:8, HFD+STZ n:13). (A) Areas of interest evaluated shown with optical and scanning probe microscopy (SPM). The lacunae are surrounded by a dashed white line and the indentations with the Berkovich tip are marked by the yellow arrows (B) Hardness, (C) Plastic work, and (D) Elastic modulus in the anterior region (E) Hardness, (F) Plastic work, and (G) Elastic modulus in the lateral region.

Bone Staining and Imaging

Bone staining and imaging methods were adapted from previous studies and are described in the Supplemental Material.^{19, 33, 34} Briefly, 700μm thick sections of PMMA embedded bones were fixed, cleared, and subsequently stained with 0.02% w/v rhodamine-6G (Sigma Aldrich, 83 697) in calcium-buffered PBS. After staining, samples were polished to a ~200μm thickness and 0.05μm surface finishing. Final sample clearing was performed using 2,2′-thiodiethanol (TDE, Sigma Aldrich, 166782) using serial gradients. All procedures were performed at room temperature while protecting the samples from light. Confocal microscopy imaging was performed with the samples wet mounted on glass coverslips with 97% TDE in PBS to match the refractive index of the immersion oil.³⁵ The anterior region of the femur was evaluated for all samples avoiding blood vessels or surface defects.

OLCN Topology and Lacunae Morphology Analysis

The analysis of the OLCN topology and lacunae morphology were adapted from a previous study and are described in the Supplemental Material.¹⁹ In brief, mid-diaphysis OLCN was evaluated using confocal microscope image stacks. The OLCN processed images were analyzed in MATLAB R2023a (Mathworks Natick, MA, USA) using a previously written code.^{33, 36} The outcomes of interest were the total number of nodes, number of T-nodes (branch-like nodes), number of C-nodes (cluster-like nodes), canalicular density, the space occupied by the OLCN (network fraction) and the distance to OLCN (mean minimum distance from any point in the network to the matrix) (Figure 6A). After the OLCN was captured using Rhodamine staining and confocal imaging (Figure 6B), the volume of interest was subsequently reconstructed to evaluate network connectivity (Figure 6C). The lacunae morphology was evaluated using an open-source code on MATLAB R2023a (Mathworks Natick, MA, USA).³⁷ The outcomes of interest were lacunar surface area, sphericity, span theta (alignment with respect to the femoral longitudinal axis) and closest center of mass (COM) (Figure 6H).

Figure 6.

T2D increased lacunae surface area, reduced the total number of nodes and canalicular density (LFD+VEH n:8, HFD+STZ n:13), and increased the expression of the senescence marker P16 (n:4). (A) Schematic representation of osteocyte network. Created in BioRender. Kohn, D. (2024) https://BioRender.com/b81n217 (B) Rhodamine stained cross-section image obtained through confocal imaging (C) Computational reconstruction of the OLCN for network topology analysis (D) Total number of nodes (E) Canalicular density (F) Number of T-nodes (G) Network fraction (H) Computational reconstruction of the rhodamine stained cross-section confocal image to analyze the lacunae morphology, osteocyte lacunae are shown in green and the orientation of the main diagonal of the ellipsoid is represented with the red and blue arrows (I) Lacunar surface area (J) lacunar sphericity (K) Fold-change in the expression of p16 between the diabetic and control groups.

RT-qPCR

RNA Extraction from Osteocytes for RT-qPCR is described in the Supplemental Material. After RNA extraction, RNA quality and quantity were evaluated using a NanoDrop One (Thermo Fisher Scientific, Waltham, MA, USA). cDNA synthesis was performed using Superscript II (Thermo Fisher Scientific, Waltham, MA, USA). Gene analysis was performed with TaqMan Gene expression assays (Thermo Fisher Scientific, Waltham, MA, USA). Changes in osteocyte senescence marker expression with T2D where evaluated using the P16 gene (Mm00494449_m1), which has been shown to be sufficient to identify senescent osteocytes in female mice.³⁸ The selected housekeeping gene was Gapdh (Mm99999915_g1). Final amplification was performed with Mastermix (Thermo Fisher Scientific, Waltham, MA, USA) and the QuantStudio5 Real-Time PCR system (Thermo Fisher Scientific, Waltham, MA, USA). The fold-change in gene expression was quantified using the delta-delta method.³⁹

Liquid Chromatography and Mass Spectrometry

Right ulnae and radii were analyzed for carboxymethyl lysine (CML) content via high performance liquid chromatography and tandem mass spectrometry. CML was quantified by ions of 205.118 m/z, and verified with tandem mass spectrometry (MS/MS) using the characteristic fragment of 84.08 m/z. Full details of sample processing and chromatographic techniques can be found in the Supplemental Methods.

Statistical Analysis

Statistical analyses were performed using GraphPad Prism (Version 10.3) and following the guidelines for biomedical research.⁴⁰ Significance was established at p-values (P) lower than 0.05 using 2 tailed analysis. We hypothesized that T2D decreases female bone volume, total volume, OLCN connectivity, and load bearing capacity. To test this hypothesis we evaluated the differences between the diabetic and the control groups with unpaired t-tests . Moreover, we performed 2-way ANOVA tests to determine the effect of diabetes and spatial location on bone quality, individual comparisons were assessed by Šídák's post hoc tests.

Furthermore, we tested the hypothesis that T2D-compromised bone architecture is linked to OLCN changes using Pearson’s correlation and linear model regressions (LMR). LMR was conducted on pooled groups after confirming no slope differences. The independent variables were assigned based on bone's hierarchical structure. For all LMR, the coefficients of determination (R²) and P are reported. Principal component analysis (PCA) was performed to determine the architecture variables with the highest clustering capacity. Data is presented as means ± standard deviation.

RESULTS:

T2D Was Induced in Female Mice

T2D was induced in female mice at 16 weeks of age using a combination of HFD and STZ (Figure 1). Diabetic mice displayed increased body size resembling an obese phenotype (Figure 1B) despite exhibiting no significant differences in body weight vs. controls (Supplemental Figure 2A, B). The HFD+STZ group displayed glucose levels above the diabetes threshold (250mg/dL) from the 6^th week of the study and throughout the 12 weeks of the disease phase (Figure 1C). At the study endpoint, mice were 30 weeks of age, blood glucose (Figure 1D) was significantly higher in the diabetic group (474±82 mg/dL) compared to the control group (123±16 mg/dL, P < 0.001). Consistent with chronic hyperglycemia, HbA1c increased throughout the experimental phase (Supplemental Figure 2C) and was significantly higher in the HFD+STZ group (4.4±0.3 %) compared to the control group (9.4±1.0 %, P= 0.007) at the study endpoint (Figure 1E). Furthermore, diabetic fasting insulin serum levels (0.019±0.03 ng/μL, Figure 1F) were significantly lower compared to the control group (0.056±0.03 ng/μL, P= 0.0013). In addition, the combination of HFD and STZ promoted body composition changes by significantly increasing the adipose tissue percentage (Figure 1G) in diabetic mice (19.2±3.4 %) in relation to the control mice (15.6±2.7 %, P= 0.0220). The apparent changes in body size are attributed to changes in body composition including increased fat content and decreased lean mass and body water (Supplemental Figure 2D, E). These results show that the HFD+STZ mice displayed the hallmarks of diabetes of hyperglycemia, hypoinsulinemia, and increased adipose tissue.

T2D Mice Displayed Decreased Cortical and Trabecular Bone Architecture Properties in a Spatially Dependent Manner

T2D induced changes in female bone architecture. Femora in diabetic mice (16.4±0.1 mm) were 2.98% shorter than in control mice (15.9±0.3 mm, P < 0.0001; Figure 2B) and cortical properties decreased in a spatially dependent manner. Cortical TV (Figure 2C) was significantly lower in the diabetic group at all locations compared to the control group (P< 0.0001). Cortical BV (Figure 2D) was significantly lower in the diabetic group (0.52±0.02 mm³) at the third trochanter compared to the control group (0.56±0.03 mm³, P= 0.0228). Moreover, cortical TMD (Figure 2E), was significantly lower in the diabetic group compared to the control group at the base of the femoral neck (1207±11 mgHA/cm³ vs 1190±18 mgHA/cm³, P=0.045) and the distal metaphysis (1132±17 mgHA/cm³ vs 1164±18 mgHA/cm³, P< 0.0001). T2D did not affect cortical BMD (Supplemental Figure 3A).

The spatial dependency of the changes induced by T2D in cortical bone was also present in trabecular bone with the distal region preferentially affected. At the distal metaphysis, trabecular TMD (Figure 2F), was significantly lower in the diabetic group (952±32 mgHA/cm³) compared to the control group (1001±40 mgHA/cm³, P= 0.0015). Diabetes also altered trabecular morphology by significantly decreasing SMI (1,06±0.2, Figure 2G) in relation to the control group (1.4±0.2, P< 0.0001). Trabecular TV was marginally lower in the diabetic (1.05±0.03 mm³) vs. control group (1.09±0.03mm³) (Figure 2H; p = 0.064), but a significant factor effect for diabetes was found (P= 0.0445). No changes in trabecular BV/TV were found (Supplemental Figure 3B) and no significant interaction effect between diabetes and bone location was found in the two-way ANOVA.

Mice With T2D Displayed Reduced Cortical Bone Size While Preserving Bone Shape

The diabetic group displayed a decreased bone area while preserving bone shape (Figure 3A). Diaphyseal CSA was significantly lower in the diabetic group (1.37±0.07mm² and 1.10±0.08mm² respectively) at the third trochanter and the distal diaphysis compared to the control group (1.48±0.07mm² and 1.19±0.04mm², P=0.0021 and P=0.0091, respectively) (Figure 3B). Bone marrow area was significantly lower at all diaphyseal locations measured in the diabetic group (0.77±0.07mm², 0.88±0.05mm² and 0.93±0.04 mm² respectively) in relation to the control group (0.87±0.06mm², 0.97±0.04mm² and 1.02±0.04 mm², P=0.0003, P=0.0006 and P=0.0002 respectively) (Figure 3C). Bone perimeter, mean cortical thickness and maximum cortical thickness did not change with diabetes (Supplemental Figure 3C, D,E). No significant interaction effect between diabetes and bone location was found in the two-way ANOVA.

Figure 3.

T2D reduced bone size without affecting bone shape (LFD+VEH n:8, HFD+STZ n:13). (A) cross-section view of the diaphyseal areas at bone lengths of 33%, 50% and 67% (B) Cross-section area quantified at the bone lengths of interest (C) Bone marrow area decreased with T2D (D) Ratio of the cross-section area to the bone marrow area to identify differences in the proportion of bone mass between groups (E) Feret angle (orientation of the maximum Feret distance segment) (F) Measurement of the Feret maximum distance (G) Feret minimum distance (H) Maximum moment of inertia.

The ratio of cross-sectional area to bone marrow area was unchanged with T2D, indicating consistent bone amount per volume and size-driven, not shape-related, changes. (Figure 3D). T2D did not alter Feret angle (Figure 3E) or maximum Feret distance (Figure 3F). However, T2D significantly decreased minimum Feret distance at the third trochanter, mid-diaphysis and distal diaphysis (1.43±0.04mm, 1.36±0.04mm and 1.28±0.04mm respectively) in relation to the control group (1.50±0.03mm, 1.41±0.02mm and 1.32±0.02mm, P=0.0003, P=0.0038 and P=0.0234 respectively) (Figure 3G). These outcomes suggest that T2D reduced bone volume and size, while preserving bone shape.

Furthermore, changes in bone cross-section with T2D reduced bone’s diaphyseal maximum moment of inertia, compromising the bone’s bending resistance. The maximum moment of inertia was lower in the diabetic group at the third trochanter, mid-diaphysis, and distal diaphysis (0.21±0.02mm⁴, 0.18±0.01mm⁴ and 0.16±0.02mm⁴ respectively) compared with the control group (0.24±0.02mm⁴, 0.20±0.01mm⁴ and 0.20±0.01 mm⁴, P <0.0001, P=0.0014 and P <0.0001 respectively) (Figure 3H).

T2D Decreased Whole-Bone Load Resistance

Yield and ultimate loads were significantly lower in the diabetic group (11.0±4.1N and 24.6±1.8N respectively) compared to the control group (15.6±5.0N and 26.8±1.7N, P =0.0513 and P =0.0446 respectively; Figure 4A,B). No effect on stiffness was found with T2D (Figure 4C). At the tissue-level, no changes in yield stress, ultimate stress or elastic modulus were found (Figure 4D–F). Fractographic analysis validated that T2D did not altered material behavior, with both groups displaying an oblique fracture, representative of fracture seen in healthy bone (Figure 4G). A schematic representation of ductile and brittle fractures can be found in Supplemental Figure 4.

Figure 4.

T2D decreased whole-bone load resistance (LFD+VEH n:6, HFD+STZ n:7).. (A) Yield load (B) Ultimate load (C) Stiffness (D) Yield stress (E) Ultimate stress (F) Elastic modulus (G) Micro-CT images of fractured bones. Linear regressions to describe the relationship between architecture and bone strength (H) Ultimate load vs Cross-section area at 33% (I) Ultimate load vs Total Volume at 33% (J) Ultimate load vs Total Volume at 50%.

Linear model regressions demonstrated that cross-section area and total volume with T2D explained the variance in load bearing resistance. Specifically, cross-section area at the third trochanter explained 70.88% (P:0.003) of the variance in ultimate load (Figure 4H), total volume at the third trochanter explained 66.05% (P:0.007) of the variance in ultimate load (Figure 4I), and total volume at the mid-diaphysis explained 58.32% (P:0.0024) of the variance in ultimate load (Figure 4J).

T2D Decreased Hardness and Increased Plastic Work in a Spatially Dependent Manner

The diabetic group displayed decreased hardness in the anterior perilacunar and intracortical areas (677.88±60.57MPa and 709.27±50.58MPa respectively) in relation to the control group (739.38±58.60MPa and 773.48±66.45MPa, P=0.0484 and P=0.0380 respectively) (Figure 5B). T2D increased plastic work only in the perilacunar area (2795.25±143.90J vs 2575.52±173.291, P=0.0029) (Figure 5C). No changes were found in the Elastic Modulus with T2D (Figure 5D). T2D did not affect any material properties at the nanoscale in the lateral region (Figure 5E, F G). Furhtermore, T2D bones displayed a borderline significant increase in carboxymethyllysine (CML) (Supplemental Figure 5).

T2D Increased Lacunae Surface Area, Reduced the Total Number of Nodes and Canalicular Density and, Increased the Expression of Senescence Markers

The diabetic group displayed significant decreases in total number of nodes, canalicular density and number of T-nodes (11,889±2,697 nodes, 0.154±0.06 Canaliculi/μm² and 7,159±2,236 nodes respectively) compared with the control group (14,702±1,682 nodes, 0.211±0.02 canaliculi/μm² and 9,363±1,615 nodes, P =0.0304, P =0.0373 and P =0.0457 respectively; Figure 6D, E, F). There was no change in the volume occupied by the OLCN (network fraction) with T2D (Figure 6G) . In addition, T2D did not affect distance to OLCN or number of cluster-like nodes (C-nodes), suggesting that the decreased connectivity with diabetes affects the number of connections and the number of dendrites, but not their length or reach (Supplemental Figure 6A, B).

Moreover, lacunar surface area was significantly higher in the diabetic group (354.9±63.6μm²) compared with the control group (301.0±42.96μm², P=0.0485) (Figure 6I). However, the increase in lacunar surface area did not promote changes in lacunae shape, preserving the sphericity (Figure 6J). In addition, diabetes did not affect lacunae orientation (Span theta) or lacunae density (Supplemental Figure 6C, D). Osteocyte expression analysis using RT-qPCR unveiled increased expression of the senescence associated marker P16 with T2D compared to controls (1.58±0.5 Fold, P=0.0239) (Figure 6K), demonstrating that T2D affects osteocyte activity as well as OLCN connectivity and morphology.

Compromised Bone Architecture With T2D Can Be Partly Explained By Compromised OLCN Connectivity and Lacunae Morphology

Blood glucose and HbA1c showed a significant negative correlation with bone architecture, load bearing capacity, and OLCN connectivity, highlighting hyperglycemia's adverse effects on specific properties describing bone quality (Figure 7A). Significant relationships between OLCN connectivity and morphology and compromised architecture with diabetes were explored with linear model regressions, and unveiled that lacunar surface area can explain 29.54% (P:0.0109) of the variance in TV (Figure 7B). Number of T-nodes can explain 26.61% (P:0.0408) of the variance in TV (Figure 7C) , and total number of nodes can explain 24.88% (P:0.0492) of the variance in perilacunar hardness (Figure 7D). These relationships display the highest coefficient of determination between lacunae morphology and architecture, OLCN connectivity and architecture and OLCN connectivity and material properties. Additional OLCN properties displayed significant coefficients of determination and can explain up to 38% of the variance in maximum moment of inertia and cross-section area (Supplemental Figure 7). These results suggest that impaired OLCN connectivity and lacunae morphology can partly explain the decreased architecture phenotype with T2D that ultimately promotes compromised bone strength.

Figure 7.

Hyperglycemia and hypoinsulinemia are associated with some aspects of bone quality (LFD+VEH n:8, HFD+STZ n:13). (A) Matrix of Pearson’s correlation coefficients between diabetes-associated parameters and bone quality. Linear model regressions with the highest coefficient of determination between architecture and OLCN properties at the femoral midshaft (B) Total volume vs Lacunae surface area (C) Total Volume vs T-nodes (D) Perilacunar hardness vs number of nodes.

The Main Architectural Factors Driving the Differences Between Diabetic and Control Bones are CSA at 33%, TV at 50%, and TMD at 85%

We conducted a principal component analysis on 145 architecture variables to understand their contribution to the diabetic bone phenotype (Figure 8A). Variables with similar orientation and lower loading were removed to refine the system and identify those with the highest clustering capacity. Three variables: TMD at 85%, TV at 50% and CSA at 33%, were sufficient to represent up to 91.87% of the variance of the whole dataset (Figure 8B), and allowed differentiation between diabetic and control groups (Figure 8C). The spatial heterogeneity and the variety of bone properties affected by T2D, including BV, TV, moments of inertia, and yield load, further support the need to study changes in bone quality with diabetes in more than a single location.

Figure 8.

The main architectural factors driving the differences between the diabetic and control groups are cortical cross-sectional area (CSA) at 33%, cortical total volume (TV) at 50%, and cortical tissue mineral density (TMD) at 85%. Outcomes of principal component analysis (PCA) (A) Loadings of 145 architecture variables (B) Loadings of the three architecture variables with highest loading magnitude after removing redundant factors based on loading direction and proportion of the variance that can be explained by the principal components of the analysis (C) Principal component scores used to cluster the diabetic and control groups using bone architecture properties (LFD+VEH n:8, HFD+STZ n:13).

DISCUSSION

Subjects with T2D display a higher risk of fracture compared with healthy individuals.¹ Changes in BMD alone cannot explain the increased fracture prevalence with T2D, indicating the need to address further changes in bone quality.^{2, 3} Unlike human studies, animal models of T2D display consistent results, describing a bone phenotype with compromised strength associated with decreased architecture and altered composition.^{15, 25, 26, 41} Despite that females present higher risk of fracture and associated mortality, most animal models used to study the effects of T2D on bone are male mice.^{15, 17, 21, 22} While there is a strong understanding of the T2D bone phenotype and the potential underlying mechanisms in male mice, the effect of T2D in female mice and its underlying mechanisms remain to be understood.

The absence of suitable animal models for studying the effect of T2D on skeletally mature females, who represent the majority of T2D prevalence in humans, limits the understanding of the impact of T2D on bone quality in females and its contribution to fracture risk. This gap hinders the development of diagnostic tools and treatments tailored for diabetic women.

The goal of this study was to develop an animal model to evaluate the effect of T2D on the mature female skeleton and to determine how changes in bone architecture, mechanics, and osteocyte lacunar-canalicular network (OLCN) contribute to the reduction in bone quality and increased fracture risk with T2D.

Diabetes was induced in female mice using a combination of STZ and HFD at 16 weeks of age, where mice are skeletally mature, to avoid developmental effects.^{42, 43} The HFD+STZ group displayed T2D hallmarks of hyperglycemia, increased adipose tissue and hypoinsulinemia (Figure 1). T2D increased body fat content without promoting body weight gain, potentially due to the loss in muscle mass which is also symptoms associated with T2D (Supplemental Figure 2D, E).^{44, 45}

Consistent with human studies and the diabetic bone paradox, T2D reduced some aspects bone quality and altered architecture without affecting bone mineral density (BMD) (Supplemental figure 3A).² Compared to the control group, the diabetic group displayed lower cortical TV, BV and TMD. These changes varied along bone length and were also present in trabecular bone, where TMD, SMI and TV were lower in the diabetic group compared with the control (Figure 2). While these spatially dependent changes in bone architecture with diabetes did not alter bone shape, they compromised bending resistance - the HFD+STZ group had significantly lower cross-section moment of inertia compared with the LFD+VEH group (Figure 3). The observed reduction in bone length may result from insulin deficiency-induced reduced bone formation, or from changes in bone marrow adiposity and increased cellular senescence leading to increased bone resorption.^{41, 46}

Compromised architectural properties of TV, BV and Moments of Inertia in the diabetic group led to decreased load bearing capacity. The diabetic group displayed lower yield load and ultimate load compared with the control group, but no changes in tissue-level properties were detected (Figure 4). Furthermore, changes in bone architecture - cortical TV and CSA - can explain up to 70% of the variance in the ultimate load, supporting the role of bone architecture on mechanical resistance. Although T2D-related architectural changes increased bone fragility, both control and diabetic bones showed oblique fractures typical of healthy bones. This indicates that T2D-induced changes in bone strength were due to structural rather than material properties. Nanoindentation testing further validated that the effect of T2D on bone strength was not driven by elastic material changes. No changes were found in the elastic modulus with diabetes, but the HFD+STZ groups displayed lower hardness and higher plastic work in the anterior region. This suggests that T2D did not affect elastic changes but promoted a subtle change in bone’s plastic behavior by reducing its hardness and increasing the energy that can be dissipated through deformation (Figure 5).

Increased plastic work alone could be regarded as a positive change in bone material properties, as it suggests increased energy dissipation. However, In our model, this increase is associated with decreased hardness. Both, plastic work and hardness indicate greater sample penetration and deformation by the nanoindenter. Thus, increased plastic work accompanied by decreased hardness in T2D suggests that the elevated plastic work may represent a trade-off rather than an improvement in bone material properties.

While we did not find an effect of T2D on material properties calculated from beam theory under four-point bending, T2D promoted region-specific changes in material-level properties measured via nanoindentation, such as decreased hardness and increased plastic work, in the diaphyseal anterior region. This highlights the importance of differentiating between changes in intrinsic material properties and localized matrix alterations, which may be cell-mediated.

The changes in post-yield material properties may be attributed to a borderline significant increase in carboxymethyllysine (CML) (Supplemental Figure 5), which contributes to reduced bone material properties with T2D.^{15, 47} While a substantial increase in AGEs is generally associated with increased bone brittleness in the context of diabetes, the absence of a brittle bone phenotype in our model may be due to the borderline significant increase in AGEs (CML) being insufficient to promote brittleness. These results support the hypothesis that the effect of T2D on decreased bone strength is driven by architecture changes rather than material properties.

We next evaluated the OLCN to determine whether the alterations in bone architecture could be associated with a compromised OLCN. T2D compromised OLCN connectivity - the HFD+STZ group displayed a significantly lower total number of nodes, T-nodes and canalicular density compared with the LFD+VEH group (Figure 6). T2D also led to a higher lacunae surface area, consistent with human studies, and might be associated with increased perilacunar resorption, potentially also contributing to the decreased bone mass.⁴⁸

Lacunae enlargement occurred with the preservation of lacunae shape, orientation and sphericity (Supplemental Figure 6). The volume occupied by the OLCN with diabetes also did not change, suggesting that the decreased space occupied by the dendrites in the HFD+STZ group is compensated by the space occupied by the increased lacunae size.

The changes induced by T2D on OLCN connectivity and lacunae morphology were associated with compromised bone architecture. Lacunar surface area and number of T-nodes can explain up to 30% of the variance in cortical TV. Moreover, total number of nodes could explain up to 25% of the variance of perilacunar hardness (Figure 7D). These results suggest a role of OLCN in changes induced by T2D on architecture and post-yield material properties. The relationship between compromised OLCN and bone architecture with T2D might be due to impaired cell communication and increased cell senescence, which decreases bone metabolism, contributing to the diminished architecture that ultimately led to increased bone fragility. In line with this, T2D increased the expression of a senescence associated marker (P16) in osteocytes (Figure 6K), demonstrating that in addition to compromised OLCN connectivity, T2D might impair osteocyte activity.

Linear regression modeling was conducted on pooled groups because there was no statistically significant difference between the slopes of the HFD+STZ and LFD+VEH groups. This indicates that while the diabetic bone exhibited both lower architecture and strength compared to the healthy controls, the relationship between the variables (slope) was preserved. Therefore, the effect of T2D on bone architecture and mechanics can be described as a downward shift in values, rather than an alteration in the relationship between bone architecture and mechanics. Principal components analysis further supported the relevance of the changes in architecture. TMD at the distal metaphysis, TV at the central diaphysis and CSA at the third trochanter are sufficient to cluster and differentiate the diabetic from the control group (Figure 8). These outcomes also demonstrate the importance of spatial component, describing how specific properties at specific bone locations might drive the diabetic bone phenotype.

Our results support that with T2D, decreased whole-bone load bearing capacity is primarily driven by changes in architecture, whereas local changes in material-level properties, particularly post-yield properties, may be secondary to OLCN disruption. Furthermore, we hypothesize that the spatial heterogeneity in the effects of T2D on bone properties is due to variations in mechanical loading and strain along the length of the bone. The OLCN is a mechanotransducive system that responds to strain by regulating the activity of osteoblasts and osteoclasts, thereby modulating bone formation and resorption.^{32, 49} T2D disrupts this mechanotransducive system. Thus, the spatial distribution of mechanical loading and strain may determine the extent and nature of changes in bone properties with T2D, which could explain why alterations in bone properties with T2D are not uniform.

Our mouse model mimics a late-stage T2D phenotype, rather than the natural progression of human T2D, and it reproduces several key aspects of human T2D bone traits, such as increased osteocyte lacunar size, decreased trabecular architecture, reduced hardness and load-bearing capacity with preserved BMD, and decreased trabecular total volume.^{4, 5, 48}

Compared to studies on the effect of T2D in male mice, this female mouse model did not result in body weight differences.¹⁵ While both sexes display compromised load bearing capacity with T2D, the biomechanical mechanisms of weakening are different. In male mice T2D decreases bone strength by the combined action of changes in architecture and material properties,¹⁵ while in our study in females, the reduced bone strength is attributed primarily to changes in architecture properties.

Sex comparisons present limitations due to the fact that C57BL/6J mice have sexually dimorphic glucose metabolism.^{50, 51} Furthermore, since STZ interacts with estrogen and a 2-fold higher dose has to be applied to females compared to males, the higher STZ dosage could induce different systemic effects and higher kidney load as confounding variables.^{21, 22} Thus, even if we included a male group in this study, the comparisons of outcomes to identify sex effects with diabetes would have critical limitations.

In young female rats, T2D induced during skeletal development promoted a brittle bone phenotype with decreased load bearing capacity. Reduced cortical thickness, decreased trabecular volume and lower elastic modulus, accompanied by decreased mineral to matrix ratio contribute to decreased load bearing capacity.²⁵ While our study describes similar changes in cortical and trabecular architecture, we did not find a significant effect of T2D on elastic modulus. This suggests that while early onset of T2D affects both architecture and mechanical properties, post-developmental onset of T2D impacts primarily bone architecture without promoting significant changes in material properties.

Since diabetes was induced after skeletal maturity, the effect of growth is negligible. We hypothesize that the reductions in TV and BV might be due to decreased bone formation or increased bone resorption. Future research should study whether changes in architecture with T2D are driven by osteoblast or osteoclast activity and determine what drives the spatial dependency of T2D-affected bone.

STZ causes an immediate reduction in insulin, in contrast to the initial compensatory increase in insulin levels observed in human T2D, which has anabolic effects on bone. Although we included an HFD lead-in phase to transiently increase insulin levels, we recognize that the rapid transition to hypoinsulinemia may influence the skeletal changes and may not fully recapitulate the sequence seen in humans. Also, off-target effects from chemical induction may further limit direct translation to human disease. Furthermore, the absence of insulin measurements throughout the study hinders the evaluation of the disease progression.

In addition, a limitation of this study is the absence of an LFD+STZ control group, which prevents us from fully distinguishing the independent effects of STZ and insulin deficiency from the T2D phenotype generated by combining STZ and HFD. In male mice, STZ toxicity, in the absence of diabetes, does not affect bone quality and cannot explain the skeletal phenotype,¹⁵ and insulin deficiency alone does not explain the bone phenotype in T2D.^{26, 41} But the contribution of STZ alone and insulin to the diabetic bone phenotype has not been established in females. Future studies should include this control group to help isolate the specific contributions of STZ to the skeletal phenotype in females.

In addition, some significant p-values are close to 0.05, which might be attributed to small sample size. Using greater sample sizes might reduce these marginal p-values, increasing the confidence of the results. Furthermore, while our methodology aligns anatomical landmarks as closely as possible across specimens, minor discrepancies in the analyzed regions of interest between groups may still exist and cannot be completely excluded.

This study developed a novel mouse model to determine the effect of T2D on female bone quality and used a unique approach to characterize the diabetic female bone phenotype. The integration of multiple scales and properties ranging from osteocyte activity and OLCN connectivity, to microarchitecture and the whole-bone level, unveiled a phenotype characterized by a compromised OLCN that contributed to decreased bone volume and tissue volume, which led to reduced bending resistance and ultimately decreased bone strength. These novel outcomes support the hypothesis that the effect of T2D on bone strength in skeletally mature female mice is architecture-driven. The results of this study increase the understanding of the effect of T2D on female bone quality, and serve as the foundation for future studies to determine the mechanisms that drive the architecture changes in T2D female bone. These results could enable the development of diagnostic tools for fracture risk assessment and treatment strategies for women with T2D-asssociated bone disease.

Highlights:

• Type-2 diabetes (T2D) can be induced in skeletally mature female C57BL/6J mice

• T2D compromises bone architecture in a spatially dependent manner

• Decreased bone strength with T2D can be explained by compromised architecture

• T2D affects the Osteocyte Lacuno-Canalicular Network connectivity and morphology

• The effect of T2D on female bone strength is driven by changes in architecture properties of TV, BV and Moments of Inertia