Bone Formation Is Impaired in a Model of Type 1 Diabetes

Female 8-week-old nonobese diabetic mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Mice were acclimated to a presterilized diet and water in autoclaved cages and maintained in a pathogen-free facility to maximize development of diabetes. Urine samples were monitored weekly for glucosuria using Keto-Diastix reagent strips (Bayer, Elkhart, IN). Diabetes was confirmed by the measurement of blood glucose at the time of surgery. In experiment 1, bone formation was compared in 7 untreated diabetic animals and 11 comparably aged nondiabetic mice (Fig. 1A and B and Fig. 2A and B). In experiment 2, bone formation was compared in 11 diabetic and 12 nondiabetic mice (Figs. 1C and D, 2D–F, and 3 and Tables 1 and 2). Diabetic animals in experiment 2 (n = 11) were alternately assigned to begin treatment with insulin implants (n = 6) (LinBit sustained-release insulin implant pellets; LinShin, Scarborough, Canada) or vehicle (n = 5) (blank palmitic acid micro-crystal implants; LinShin) at the time of surgery. All research protocols were approved by the Institutional Animal Care and Use Committee of the University of Arkansas for Medical Sciences.

The confirmed diabetic mice and age-matched nondiabetic NOD mice were anesthetized with sodium pentobarbital (71 mg/kg, day 1). In experiment 2 only, insulin or vehicle implants were inserted under the mid-dorsal skin using a 12-gauge trocar. A titanium ring fixator was placed on the left tibia, and a mid-diaphyseal osteotomy was created, as previously described (16). The fibula was fractured by direct lateral pressure. The periosteum and dermal tissues were closed, and buprenex (0.1 mg/kg) was given postoperatively by intramuscular injection for analgesia. Distraction was initiated on day 4 (3 days latency after surgery) at a rate of 0.075 mm twice a day (0.15 mm/day) and continued for 14 days. Mice were killed on postoperative day 17. The distracted tibiae were harvested for radiographic and histological studies, and the contralateral nondistracted tibiae were analyzed using microcomputed tomography (μCT) and biomechanical testing.

In experiment 2, serum was obtained on the day of killing for measurement of glucose, insulin, osteocalcin, and cross-linked C-telopeptides of type I collagen. Insulin was measured using Luminex xMAP technology with a multiplex assay kit (Mouse Adipokine LINCOplex Panel; Linco Research, St. Charles, MO). Osteocalcin was measured by enzyme immunosorbent assay (Mouse EIA kit; Biomedical Technologies, Stoughton, MA). Degradation products of the C-telopeptide of type I collagen were measured by enzyme immunoassay (RatLAPS; Osteometer Bio Tech, Herlev, Denmark).

After at least 48 h of fixation in 10% neutral buffered formalin, the left tibiae were removed from the fixators for high-resolution single-beam radiography: a Xerox Micro50 closed system radiography unit (Xerox, Pasadena, CA) was used at 40 kV (3 mA) for 20 s using Kodak X-OMAT film. For quantification, the radiographs were video recorded under low-power (1.25× objective) magnification, and the area and density of mineralized new bone in the distraction gaps were evaluated using National Institutes of Health Image Analysis 1.62 software. The distraction gap was outlined from the outside corners of the two proximal and the two distal cortices forming a quadrilateral region of interest (ROI). The mineralized new bone area in the gap was determined by outlining regions with radio density equivalent to or greater than the adjacent medullary bone. The percentage of new mineralized bone within the distraction gap was calculated by dividing mineralized bone area by total gap area. The distracted tibiae then were decalcified in 5% formic acid, dehydrated, and embedded in paraffin, as previously described (17,18). Longitudinal sections (5–7 microns) were stained with hematoxylin and eosin. Sections were selected for analysis to represent a central gap location. As detailed above, a quadrilateral ROI was outlined and recorded. New bone was defined as all organized osteoid/sinusoid columns. Both the proximal and distal endosteal new bone matrix was outlined and the area was recorded. The percentage of new bone area within the DO gap was calculated by dividing the new bone matrix area by the total distraction gap area. Adjacent sections were then used for immunohistochemistry.

A primary polyclonal antibody specific for insulin receptor α and normal rabbit IgG were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). The biotinylated secondary antibody was purchased from Vector Laboratories (Burlingame, CA). Sections were deparaffinized and rehydrated. Slides were then rinsed in phosphate buffered saline (pH 7.4) with 0.02% Triton-X 100 for 3 min to permeabilize the tissue and incubated for 10 min in a 1% BSA and 2% nonfat dry milk solution to block nonspecific binding. After washing, the slides were incubated with Peroxiblock for 5 min and washed twice again. The primary antibody (insulin receptor α, diluted 1:200) or rabbit IgG (for negative control, diluted 1:400) was applied and incubated at 4°C for 16 h. After washing, a biotinylated goat anti-rabbit secondary antibody (diluted 1:100) was applied for 20 min, followed by the application of horseradish peroxidase–conjugated streptavidin for 20 min. Color was developed using Nova Red. The sections were counterstained with hematoxylin, dehydrated, and mounted using permanent mounting media.

Representative specimens of distracted tibiae were determined from the 2-dimensional (2-D) radiographs and imaged by μCT using a μCT-40 (Scanco, Bassersdorff, Switzerland) and the manufacturer’s software. A total of ∼600 contiguous axial (cross-sectional) slices including the entire distraction gap and at least 0.5 mm of both proximal and distal host bone were obtained at 55 kV and 70 μA with a voxel size of 12.4 μm in all dimensions. To illustrate endosteal new bone, the endosteal surface of the host cortex at the proximal end of the distraction gap was outlined in the last cross-sectional image in which the cortex appeared intact. This outline was deleted and copied onto the slice in which ∼50% of the host cortex was missing, defining the ROI for that slice and marking the proximal end of the distraction gap. This procedure was repeated for the distal end of the distraction gap. The ROI for each of the interceding slices was defined using an automated algorithm that interpolates gradual changes in the size and position of the manually defined ROIs, commonly known as morphing. The 2-D gray scale images within each ROI were stacked before applying a gray scale threshold and a 3-dimensional (3-D) noise filter to create a binary 3-D reconstruction. To illustrate the gap dimensions, a 3-D reconstruction of the proximal and distal ends of the cortex was created and overlaid on the endosteal new bone reconstruction. An analogous procedure was used to illustrate combined endosteal and intracortical new bone in which the ROI was defined using the periosteal perimeter of the host cortex instead of the endosteal perimeter.

Trabecular and cortical bone volume and architectural properties in the contralateral tibiae from six nondiabetic, six insulin-treated diabetic, and four vehicle-treated diabetic mice were also evaluated by μCT. For trabecular analyses, the entire endosteal volume of the proximal tibia extending 1.24 mm distal to the primary spongiosa was analyzed. Cortical analyses were performed at the tibial midshaft defined as the midpoint between the proximal and distal epiphyses. The gray-scale thresholds and noise filters were optimized specifically for murine trabecular or cortical analyses. Total volume, bone volume, and architectural indexes were calculated directly from the 3-D reconstructions.

Cortical strength of the tibial midshaft was measured using 3-point bending analyses on a MTS Bionex 868 load frame operated via TestWorks version 4 software on a personal computer as described previously (19). The lower platens supported the bones with a span of ∼80% of the total bone length, and load was applied from the upper platen to the posterior aspect of the tibial midshaft. The peak load (N) and stiffness (N/mm) were recorded for each bone.

For statistical analysis, the unpaired Student’s t test was used to compare differences between groups for all serum analyses. Skeletal parameter comparisons (radiographic and μCT) were performed using one-way ANOVA and Student Newman-Keuls post hoc tests. All data are reported as means ± SE, and differences are considered statistically significant when P < 0.05.

A total of 41 mice (7 diabetic and 11 comparably aged nondiabetic animals in experiment 1 and 11 diabetic and 12 nondiabetic animals in experiment 2) underwent DO and were analyzed. At the time of surgery, the mean age of mice in experiment 1 was 15.9 ± 1.0 weeks for control animals and 16.9 ± 1.9 weeks for diabetic animals; the mean age of mice in experiment 2 was 22.2 ± 1.1 weeks for control animals, 19.3 ± 1.8 weeks in untreated diabetic animals, and 23.1 ± 1.1 weeks in insulin-treated diabetic animals (nonsignificant differences). At the time of killing, mean blood glucose measurements in the three groups were as follows: nondiabetic 187.0 ± 13.4 mg/dl, n = 23; untreated diabetic: 482.9 ± 51.6 mg/dl, n = 12; and diabetic + insulin: 261.0 ± 45.9 mg/dl, n = 6. Serum measurements for insulin, osteocalcin, and type I collagen C-telopeptide are shown in Table 1. Insulin levels confirmed relative insulinopenia in the vehicle-treated diabetic animals, with modestly higher insulin levels, compared with nondiabetic mice, noted in the insulin-treated group (∼30% above controls). Serum osteocalcin concentrations were reduced in vehicle-treated diabetic animals compared with nondiabetic mice (P ≤ 0.001) and normalized with insulin treatment, while type 1 collagen C-telopeptide levels remained unchanged.

A significant decrease in bone formation was noted in the distraction gap of diabetic mice compared with nondiabetic mice, both radiographically and histologically (Fig. 1). In experiment 1, mineralization of the DO gap, as measured from the radiographs, showed a significant decrement in bone formation in diabetic mice compared with nondiabetic mice (Fig. 1A). Analysis of histological sections revealed a similar trend in the formation of endosteal new bone columns (Fig. 1B). In experiment 2, bone formation was again impaired in diabetic mice (Fig. 1C and D). In contrast, the diabetes-induced inhibition of bone formation was prevented by treatment with insulin, as demonstrated both radiographically (Fig. 1C) and histologically (Fig. 1D). A more significant deficit in the formation of new bone columns was apparent by histological analysis than by radiographic analysis. This discrepancy is typical of DO analyses and reflects the fact that single-beam radiographic analysis will identify all sources of calcified material within the gap, hence overestimating mineralization in the gaps. (16,20)

As shown in Fig. 2, the delay in de novo bone formation in diabetic animals and the improvements in new bone volume with insulin replacement were confirmed when specimens were analyzed using μCT. Figure 2A and B depict representative μCT reconstructions of distraction gaps from nondiabetic NOD (A) or untreated diabetic (B) mice in experiment 1. Mineralized bone present within the nondiabetic gap was appreciably greater than that present within the gap of the untreated diabetic animal, consistent with the radiographic and histological data presented in Fig. 1. Figure 2C is a graphic demonstrating the microarchitecture that occurs during distraction. The bone-forming unit is described as the minimal structural element that comprises the basic biological unit of DO (20). As seen here, bone forming units can originate in parallel with the distraction force. These longitudinal bone columns can eventually bridge the gap, with new bone and mineralization emanating from the proximal and distal fracture sites. To better appreciate these phenomena, Fig. 2D–F illustrates only endosteal new bone in the context of the host cortices as described in research design and methods. In these figures, the bone shown lies within the future marrow cavity and represents intramembranous bone formation comprised of numerous bone forming units, as determined by previous histological analyses (20). In both Fig. 2D (nondiabetic NOD) and F (insulin-treated diabetic NOD), mineralization of these bone columns can be appreciated as they arise from both proximal and distal bone marrow. In contrast, the vehicle-treated NOD animal (Fig. 2E) shows minimal mineralized new bone in the distraction gap. These data together suggest that when untreated, the diabetic condition markedly impaired bridging of the distraction gap by new mineralized bone. However, insulin treatment and improved glycemic control clearly enhanced bone formation in the diabetic NOD mouse.

Contralateral tibiae (i.e., nondistracted tibiae) of animals in experiment 2 were analyzed to determine whether intact trabecular and cortical bone were altered by diabetes or insulin treatment (21). Evaluation of the contralateral tibiae by μCT demonstrated reductions in trabecular bone volume (P < 0.01) and trabecular thickness (P < 0.001) in diabetic animals, which was prevented by insulin treatment (Table 2). Furthermore, the trabecular thickness of insulin-treated diabetic mice was also significantly greater than nondiabetic mice (P < 0.01). Interestingly, insulin-treated diabetic mice displayed a significant decrease in connectivity density and trabecular number and an increase in trabecular separation compared with nondiabetic NOD mice (Table 2).

Midshaft cortical analysis of the contralateral tibiae by μCT and biomechanical testing revealed even more striking effects (Table 3). Cortical thickness was significantly less in diabetic NOD compared with nondiabetic mice, and this appeared to be due to an increase in endosteal perimeter and diameter without a significant change in periosteal measures. As a result, strength analysis by 3-point bending demonstrated that the tibial midshaft of diabetic NOD mice was significantly weaker than nondiabetic mice. All of these changes were prevented in the insulin treated group, and amazingly, peak force during 3-point bending was significantly increased above nondiabetic mice (Table 3).

Insulin may restore bone formation in diabetic animals directly via effects on osteoblast precursors or mature osteoblasts or indirectly via normalization of glycemia or other metabolic parameters. A direct effect would require the presence of insulin receptors on immature and/or mature osteoblasts within the distraction gap. Therefore, immunohistochemistry was used to search for insulin receptor α in the distraction gap. We stained serial sections from three DO gaps, one from a nondiabetic NOD mouse (shown in Fig. 3) and two from C57BL/6 mice (data not shown). The staining patterns were reproducible on all three specimens. Distraction gaps have been delineated into three zones of cellular activity (22). The fibrous interzone (FIZ) is found in the middle of the gap and contains fibroblast-like cells oriented along the tension vector. The primary matrix fronts (PMFs) border the FIZ on both sides and contain proliferating fibroblast-like cells (17). The microcolumn formation zones are juxtaposed between the PMF zones and the original cortical breaks and contain osteoblast subtypes from producers of newly deposited osteoid to osteocytes. Numerous insulin receptor+ cells were found in each of the above described zones (Fig. 3). Specifically, insulin receptor+ cells are numerous but scattered throughout the FIZ, which is postulated to contain mesenchymal osteoblastic precursors/stromal cells. Insulin receptor+ cells are also found in significant numbers in the PMF, which is believed to contain proliferating osteoblast precursors. Further, insulin receptor+ cells are characteristic of the various osteoblast stages throughout the microcolumn formation zones. These findings suggest that the insulin signaling axis could be a physiological component of each phase of osteogenesis during DO in both pre-diabetic NOD mice and C57BL/6 mice. The specificity of the insulin receptor+ staining was confirmed by the IgG controls and by comparison to immunostaining for known specific antibodies including proliferating cell nuclear antigen (data not shown).

Rodent models utilizing DO provide the opportunity to isolate and study osteoblastogenesis and de novo intramembraneous bone formation under various pathological conditions and on various genetic backgrounds (16,23). DO is a precise methodology by which long bones are severed then systematically stretched to regenerate bone de novo within the expanding gap. In rodent models, the bone created within this gap occurs principally via osteoblast-mediated events and does not require cartilaginous scaffolding or endochondral processes. Thus, DO can provide a relatively pure model in which to study the actions of systemic and local mediators on osteoblast activity and bone formation. Herein we present novel evidence that this technique can be utilized effectively in NOD mice, allowing one to study the effects of autoimmune-mediated diabetes and/or insulin administration specifically on de novo bone formation. To our knowledge, this study is the first to report the use of DO in any model of type 1 diabetes.

We provide evidence of a significant reduction in bone formation, both histologically and radiographically, in diabetic NOD mice. In addition, we noted a reduction in serum markers of bone formation in diabetic animals compared with nondiabetic animals. Deficits in bone formation in the DO gap and reductions in osteocalcin were entirely prevented with 17 days of insulin treatment, despite persistent mild hyperglycemia. Furthermore, μCT analysis of distracted tibiae demonstrated that the mineralizing new bone emanating from within the future marrow cavity of the regenerating gap was almost undetectable in the diabetic animal, yet was significantly rescued with insulin treatment. Consistent with in vitro evidence of insulin receptor expression in osteoblast-lineage cultured cells (24–26), we have demonstrated insulin receptor expression on osteoblasts in all zones of the distraction gap in vivo. Together, these findings suggest that primary insulinopenia, as opposed to metabolic derangements such as hyperglycemia, may be more detrimental to the process of intramembraneous bone formation.

Examination of bone structure and strength in the contralateral tibiae demonstrated a surprisingly strong impact of short-term type 1 diabetes and insulin on the intact skeleton. In both trabecular and cortical compartments, bone volume and thickness were reduced by diabetes, and these changes were prevented by insulin treatment. However, other parameters of trabecular microarchitecture, which may affect strength independent of mass, volume, and density, were altered only in the insulin-treated diabetic mice. Without performing strength analyses directly on the isolated spongiosa, the mechanical (strength) impact of these structural changes remains unclear. However, results of the midshaft biomechanical tests demonstrate that insulin treatment prevents and even increases cortical strength. Recent studies using μCT and finite element modeling have demonstrated striking strain-dependent trabecular microarchitecture in mice (27). However, the factors controlling architecture, independent of mass and volume, are completely unknown. Interestingly, the results presented here may have identified insulin and/or glycemic levels as the first agents known to modulate trabecular architecture.

Indirect evidence from clinical studies supports the concept that insulin deficiency, rather than chronic hyperglycemia, may contribute to the decrease in bone formation seen in diabetes. First, several studies have failed to demonstrate a consistent correlation between indicators of glycemic control and BMD in type 1 diabetes (1,2,4,5). Second, the effects of type 1 diabetes compared with type 2 diabetes on bone metabolism are discordant. While an increased incidence of osteopenia can be demonstrated among studies of type 1 diabetes (1–6), studies of type 2 diabetes, a state of hyperinsulinemia, more commonly demonstrate normal or increased bone density (3,28–30), despite the fact that years of untreated hyperglycemia could exist in individuals with type 2 diabetes. Interestingly, a direct correlation between BMD and endogenous insulin secretory capacity in type 2 diabetes has been observed (31). Similarly, Barrett-Conner et al. (32) found a positive association between bone density of the radius and spine and fasting insulin levels in nondiabetic postmenopausal women, again suggesting that hyperinsulinemia may preserve bone. Finally, several studies have demonstrated deficits in bone density, present either at the time of diagnosis (1,33) or fairly early in the course of disease, suggesting that prolonged exposure to systemic hyperglycemia alone is not a prerequisite. This is supported by the observation that diabetic osteopenia does not appear to progress faster over time than anticipated age-related decrements in bone density (34).

Several lines of evidence from in vitro bone cell cultures support the idea that insulin exerts direct anabolic effects on bone cells. For example, primary calvarial osteoblasts and multiple osteoblast-like cell lines express insulin receptors on the cell surface and have a high capacity for insulin binding (24–26). In response to physiological doses of insulin, cultured osteoblasts show increased rates of proliferation (35,36), collagen synthesis (24,37,38), alkaline phosphatase production (39,40), and glucose uptake (41,42). Insulin may also exert direct effects on osteoclasts; mature osteoclast-like cells in vitro express insulin receptors and exhibit reduced bone resorption in response to insulin stimulation (43). Taken together, these data support the idea that the actions of insulin in bone could be mediated directly via stimulation of osteoblasts in combination with inhibition of osteoclasts. In addition to the direct effects of insulin on bone cells, insulin may exert synergistic effects with other anabolic agents in bone, such as parathyroid hormone (44–46).

In summary, we have demonstrated that bone formation is clearly impaired in the NOD mouse model of diabetes and appears to be an insulin-dependent and reversible process. Specifically, impairment in bone formation can be ameliorated with insulin treatment, even in the face of persistent mild hyperglycemia. The observation that insulin receptor staining is evident throughout all cellular delineations of the distraction gap also suggests that the insulin signaling axis could be a physiological component of each phase of osteogenesis during DO. Evaluation of the contralateral tibiae, however, demonstrated some provocative findings. Specifically, while trabecular thickness and trabecular volume were enhanced in an insulin-treated diabetic model, we also noted decreases in connectivity density and trabecular number and increases in trabecular separation, findings that are indicative of osteoporosis. Together, this suggests that while insulin treatment appears to be capable of preventing the loss of cancellous bone volume and trabecular thickness, it may have other effects on trabecular microstructure. This observation is consistent with the fact that despite a discrepancy between the type 1 and type 2 diabetic populations and BMD, diabetic populations with both diseases uniformly appear to have higher risk for bone fracture (8,9,47). Consequently, the impact of insulin treatment (as well as hyperinsulinemia) on the biomechanical properties of the intact skeleton are still unclear, and further investigation is warranted.

This work was supported by a grant from the Children’s University Medical Group/Dean’s Research Development Fund of the Arkansas Children’s Hospital Research Institute (SP-030104 to K.M.T.), a National Institutes of Health grant (C06 RR16517) to Arkansas Children’s Hospital Research Institute, and a grant from the Arkansas Biosciences Institute funded by the Arkansas Tobacco Settlement Plan (35600 to C.K.L.).