A Role for Advanced Glycation End Products in Diminished Bone Healing in Type 1 Diabetes

In vivo experiments were performed in BALB/cByJ mice (Jackson Laboratories, Bar Harbor, ME). Male animals, age 8 weeks, were maintained according to approved protocols (Boston University Institutional Animal Care and Use Committee), and were given free access to tap water and NIH31M mouse diet (5K52; Purina Mills). Generation of diabetic and control nondiabetic animals was accomplished using the multiple low-dosage streptozotocin (STZ) methodology (35). The number of animals per experimental group was six, unless otherwise indicated.

The diabetic condition was characterized. Blood glucose (Accu-Check Advantage; Roche Diagnostics), and urine glucose levels (Multistix 10SG reagent strips; Bayer) were monitored twice weekly throughout the typical 33-day experimental periods, and diabetes onset on experimental day 12 was confirmed in all diabetic animals (blood glucose levels ≥250 mg/dl). Blood and urine glucose values for control animals were normal (100–110 mg/dl). Protein and ketones in the urine were assayed twice weekly and were not detected (Multisix 10SG reagent strips; Bayer). Levels of GHb in blood (Glyc-Affin GHb; Isolab, Akron, OH) and levels of insulin in serum (Linco Research, St. Charles, MO) were measured when animals were killed.

Uniform craniotomy defects were created on day 19, 7 days after confirmation of diabetes onset. General anesthesia of the animals was accomplished by intraperitoneal injection of 0.12 ml/100 g body wt. of 62.5 mg/ml ketamine hydrochloride (Ketalar; Parke-Davis, Morris Plains, NJ) and 6.25 mg/ml xylazine (Rompum; Mobay, Shawnee, KS). Initial experiments investigating the effect of diabetes on bone healing used circular craniotomy defects of 1.0-, 1.6-, or 2.1-mm diameter prepared in parietal bones of diabetic and nondiabetic animals with cylindrical low-speed carbide burs (Nos. 2011, 2016, 2021; Meissinger, Dusseldorf, Germany) and saline irrigation. Subsequent experiments used 1-mm defects. As an analgesic after surgery, animals were given buprenorphine (Temgesic 0.3 mg/ml, i.p.; Reckitt & Collman, Hull, U.K.) 3 mg/kg twice a day for 3 days. In all, <2% of animals showed clinical signs of infection; the infected animals were excluded from the study. Animals were killed by CO₂ asphyxiation 14 days after the surgical procedures. Heads were processed for histology and quantitative histomorphometric analyses. This time point was chosen based on time studies of healing craniotomy defects in rats (36,37).

Four groups of six nondiabetic animals each (n = 6) were used to determine whether CML-albumin applied by surgically implanted osmotic mini-pumps inhibited bone formation compared to unmodified albumin. The four groups of animals received 90 or 900 μg of control mouse serum albumin (MSA) or CML-modified MSA continuously applied over the 2-week experimental period. Circular craniotomy defects (1.0-mm diameter) were prepared in the parietal bones. This defect size was chosen because bone healing is normally complete in 14 days in nondiabetic animals (Fig. 1), and it allowed the determination of inhibitory effects of locally applied substances. Alzet mini-pumps were placed in dermal pouches created in the backs of animals and did not remain exposed after surgical procedures. The pump catheter end was positioned with the aperture being located slightly over the edge of the bone defect, and glued in place with cyanoacrylate. Animals were killed 14 days after the surgical procedure.

CML-MSA was prepared by chemical modification from highly purified MSA (Calbiochem), as previously described (20). Products were chromatographed through a Detoxi-Gel affinity column (Pierce). The final concentration of the CML-MSA and control MSA was 9 mg/ml. Endotoxin levels were <2.5 pg/ml (E-Toxate kit; Sigma). In total, 30% of lysine residues in CML-MSA were converted to CML, as determined by the TNBS (trinitrobenzenesulfonic acid) assay and consistent with previous studies (20,38). CML-MSA was highly reactive on Western blots with anti-CML monoclonal antibody 6D12 (Waako) (32,39), whereas control MSA was not reactive.

Block biopsies were prepared and specimens were fixed in 10% buffered formalin at 4°C, decalcified in EDTA for 5–7 days, dehydrated, and embedded in paraffin. Serial sections (4 μm) were stained with hematoxylin and eosin. The three most central sections of each defect were analyzed. Linear measurements obtained with an image analysis system (Image-Pro Plus 4.0; Media Cybernetics, Silver Spring, MD). Bone bridging was expressed as a percentage of the total defect width. Measurements included 1) the distance between the rims of the initial bone defect and 2) the distance between the rims of the remaining defect, and the percent bone bridging was calculated. The area of regenerated bone was measured from digitally captured images using ImagePro 4.0 software.

The expression of RAGE was determined in diabetic and nondiabetic healing defects by quantitative immunohistochemical analyses of six animals per group in a separate experiment. Defects (1-mm diameter) were created as described above on day 19 in diabetic and nondiabetic control animals. Block biopsies were taken 10 days after the surgical procedure. A time point of 10 days of healing was selected because we wished to evaluate tissues during active healing in both diabetic and nondiabetic animals. Specimens were fixed in 4% paraformaldehyde for 18 h, decalcified in EDTA for 5–7 days, and placed in 10% sucrose overnight at 4°C. Samples were then stored in 2-methylbutane at −80°C. The specimens were embedded in Tissue Prep (Fisher Scientific), after which 6-μm cryosections were mounted on ProbeOn Plus slides (Fisher Scientific) and stored at −80°C until staining. Immunostaining was performed using the immunoperoxidase/diaminobenzidine technique (avidin-biotin-horseradish peroxidase complex−Vectastain complex kit Elite; Vector Laboratories), according to the manufacturer’s protocols. RAGE antiserum was generously provided by Merck (40). Slides were counterstained with hematoxylin. Quantitative histomorphometric analyses were carried out, with three central sections being measured for each bone defect; five nonoverlapping fields were measured per slide at 400× magnification. The total surface area per field was 420 μm². The total number of RAGE positive cells was counted in each of the three anatomical areas of interest: the periosteum, the dura mater, and the granulation tissue filling the healing bone defect.

Total RNA was isolated from MC3T3-E1 cells cultured in Dulbecco’s modified Eagle’s medium containing 10% FCS and antibiotics using TRIzol (Gibco/BRL). polyA + RNA was isolated (Qiagen Oligotex mRNA spin column). RT was performed on 6 μg polyA + mRNA using a RAGE-specific anti-sense primer (GTATCAAATGTTACTCAG) and a Superscript II kit (Promega). PCR was then performed on aliquots with primer sets A and B. Primer set A flanks intron 2 of RAGE and will produce a 194-bp cDNA from RAGE mRNA; primer set B flanks intron 7 and should produce a 159-bp cDNA from RAGE mRNA or a 326-bp product from RAGE genomic DNA. The sequences of primer set A are TTGGAGAGCCACTTGTGCTA (forward) and CCCTCATCGACAATTCCAGT (reverse); for primer set B, they are TCCACTGGATAAAGGATGGTG (forward) and GACCCTGATGCTGACAGGAG (reverse). PCR cycling parameters were 30 cycles—94°C, 30 s; 60°C, 45 s; and 72°C 1 min—and buffer contained 1.5 mmol/l MgCl₂.

Western blots for RAGE were performed on SDS-PAGE sample buffer extracts of MC3T3-E1 cells and cultures of primary rat osteoblasts (41). Goat anti-RAGE antiserum (1:1,000) (40) or nonimmune goat antiserum were used as primary antibodies, and bands were visualized with alkaline phosphatase-coupled rabbit anti-goat IgG and Western Blue substrate (Promega).

Biochemical measurements were analyzed with the Student’s t test for ordinal variables and with the Wilcoxon’s signed-rank test for cardinal variables. Histomorphometric measurements for bone area determinations were analyzed with Student’s t test, whereas those for bone bridging were analyzed with Wilcoxon’s signed-rank test. Analyses of immunohistochemistry studies of positively stained cells for RAGE were performed with a t test for independent samples. Histomorphometric measurements for bone area and bone bridging on defects treated with exogenous application of MSA or CML-MSA were analyzed with one-way ANOVA. Post hoc statistical testing was performed using a t test for independent groups and Wilcoxon’s signed-rank test. A value of 95% or higher for α was used to declare statistical significance for the studies reported.

Experiments were performed to determine if type 1 diabetes results in diminished bone healing. The multiple low-dosage STZ mouse model was chosen because the level of hyperglycemia is comparable to that in human type 1 diabetes (35). Three different calvaria defect sizes were evaluated in diabetic and nondiabetic animals to investigate whether diabetes influences healing in bone lesions normally capable of full healing in 14 days and in larger defects. For these studies, 36 animals were divided into six groups, three diabetic and three nondiabetic, with 6 animals in each group.

Twice weekly monitoring of blood and urine glucose levels indicated that diabetes developed by day 12, reaching the threshold level of 250 mg glucose/dl, in all animals receiving STZ, and did not develop in control animals, as expected. The metabolic measurements summarized in Table 1 demonstrate that the diabetic state of animals occurred without unexpected complications. Glucose measurements in urine confirmed blood glucose measurements. Twice weekly assays for the presence of ketones and protein were below detectable levels as expected for diabetes without ketoacidosis and kidney complications. GHb levels measured on day 33 (Table 1) were elevated in diabetic animals (42). These data indicated that hyperglycemia was consistent and not transient. Food intake was elevated in diabetic animals, whereas animal weight was reduced by ∼10%; this indicated the presence of hyperphagia accompanied by modest weight loss, as expected (43), and not unexpected metabolic dysregulation. Weight loss in diabetic animals occurred during diabetes onset until day 11, and weight gain then resumed at a normal rate for the rest of the experimental period (days 12–33). Insulin levels in serum were reduced in diabetic animals, as expected (Table 1) (42).

On experimental day 19, 7 days after the onset of diabetes, circular craniotomy defects with diameters of 1.0, 1.6, or 2.1 mm were created in diabetic and nondiabetic animals as described above, resulting in six groups of animals. Animals were killed after 14 days of healing (day 33) (36,37). Histologic evaluation of healed lesions revealed that bone formation was more abundant in nondiabetic animals than in the diabetic animals for all three lesions sizes (Figs. 1A–C [normal] versus D–F [diabetic]). Intense osteoblastic activity was seen in the growing osteogenic fronts, in both the periosteal and dural surfaces. Woven bone was observed within the lesion extending toward most of the linear dimension of the defects in normal animals. Marrow cavities had formed and numerous osteocytes were embedded in the bone matrix. By contrast, more limited bone formation was seen in diabetic animals. Marrow spaces appeared less developed and fibrous connective tissue with low amounts of cellular and vascular structures filled the bone defect.

Linear and surface area histomorphometric evaluations demonstrated that bone healing was inhibited in diabetic animals (Fig. 2). Both bone bridging and the area of bone formed were significantly inhibited compared to normal controls (Fig. 2). In normal animals, defects of 1.0-mm diameter were healed almost completely (92% bone bridging), and defects of 1.6- and 2.1-mm diameter exhibited partial healing (Fig. 2A). The healing in larger defects was characterized by discrete osteogenesis and predominant soft tissue closure. Interestingly, bone bridging in diabetic animals was inhibited by ∼40% compared with that of normal animals for all the defect sizes tested. These results demonstrated that bone healing is significantly and consistently diminished in diabetic animals.

An experiment was performed to determine whether the observed diminished bone healing was attributable to insulin deficiency. A control and test group were formed with diabetic animals (n = 7 in each group). The test animals received daily insulin injections and the control animals received daily saline injections. Daily insulin injections (10 IU/kg, i.p.) began immediately after the onset of hyperglycemic diabetic status (day 12). Then 7 days later, 1-mm circular craniotomy defects were created in the parietal bones. The efficacy of insulin treatment and dosage was determined by daily urine glucose measurements and by determination of GHb levels when the animals were killed on day 33. GHb levels in blood from insulin-treated animals were lower than those in untreated animals (8.36 ± 1.95 vs. 11.35 ± 1.03%, respectively; P < 0.0001), supporting the observation that insulin administration consistently reduced blood glucose levels, as expected.

Histology of healed lesions after 14 days of healing (day 33) revealed that bone formation was more abundant in diabetic animals treated with insulin than in untreated diabetic animals (Fig. 3B). Osteoblastic activity was seen in both the periosteal and dural surfaces. Woven-like bone was present within the lesion extending through most of the linear dimension of the defects. Marrow cavities were formed and numerous osteocytes were embedded in the bone matrix. Significant cellular and vascular activity occurred in the connective tissue surrounding the newly formed bone. Less bone formation was seen in diabetic animals not treated with insulin (Fig. 3A).

Linear and surface area histomorphometric evaluations of histologic slides demonstrated that bone healing was rescued in diabetic animals treated with insulin. Defects from insulin-treated diabetic animals exhibited significantly more bone bridging than the defects from control diabetic animals (84.81 ± 12.24 vs. 62.26 ± 24.76%; P = 0.001) (Fig. 3C). Results from bone area measurements paralleled those from bone bridging measurements. Defects from control diabetic animals exhibited significantly less regenerated bone than defects from insulin-treated diabetic animals (0.06 ± 0.04 vs. 0.14 ± 0.04 mm²; P = 0.001) (Fig. 3D). These data provide direct evidence that the alterations in bone healing observed in our model resulted from the consequences of insulin deficiency and hyperglycemia rather than from potential secondary effects of STZ.

RAGE mediates many of the effects of AGEs in different cell types. To our knowledge, RAGE expression has not been linked to altered bone formation or healing. The localization and quantitation of RAGE was, therefore, determined in healing calvaria defects in diabetic and nondiabetic animals by immunoperoxidase methodology with a specific goat anti-RAGE primary antibody. The data demonstrated higher levels of RAGE in calvarial tissues of diabetic animals (Figs. 4B and E), with low, but still detectable levels of RAGE staining observed in nondiabetic tissues (Figs. 4A and E). In diabetic tissues, positive-stained cells with morphology consistent with that of mesenchymal cells were observed in the dura mater, periosteum, and nonmineralized granulation tissue filling the bone defect (Fig. 4B). Mononuclear cuboidal and fusiform mesenchymal cells associated with the bone on the growing osteogenic front were also positively stained (Fig. 4B), suggesting that osteoblasts and pre-osteoblasts may overexpress RAGE. RAGE staining was observed in cells distant from calvaria defects, suggesting that wounding is not required for increased RAGE expression in diabetic calvaria. Multinucleated cells adjacent to bone staining for RAGE were not observed, suggesting that increased osteoclast activation in diabetes probably was not occurring after 14 days of healing. Taken together, the results demonstrated that RAGE is upregulated in fibrogenic cells in calvaria of diabetic animals.

Because of the presence of RAGE in healing calvaria, it was hypothesized that AGEs could contribute to diminished bone healing in diabetes via AGE-RAGE interactions. Experiments were performed to determine whether a locally applied RAGE ligand could inhibit intramembranous bone healing in normal (nondiabetic) animals. MSA was modified to contain biologically active levels of CML, a known RAGE ligand (20). Continuous local delivery of CML-MSA or MSA over 14 days to calvaria defects was accomplished with surgically implanted Alzet mini-pumps. Unilateral craniotomy defects were created in nondiabetic animals for this experiment. All the animals were killed 14 days after the surgical procedure.

Histologic evaluation of control MSA-treated lesions demonstrated woven bone within the lesion extending through most of the linear dimension of the defects (Figs. 5A and B). Osteoblastic activity was seen in the growing osteogenic fronts in both the periosteal and dural surfaces. Well-developed marrow cavities with numerous osteocytes were embedded in the bone matrix. Significant cellular and vascular activity was seen in the connective tissue surrounding the newly formed bone. By contrast, bone formation was significantly inhibited in defects treated with CML-MSA and resembled diabetic lesions in histologic appearance (Figs. 5C and D).

Histomorphometric analyses revealed that bone defects receiving MSA healed almost completely. Application of CML-MSA to 1.0-mm craniotomy defects in normal animals resulted in inhibition of bone formation (Fig. 5E). A dosage-response effect was noted for the two dosages of CML-MSA tested; the higher dosage resulted in significantly more inhibition of bone healing than the lower dosage. These alterations were independent of sustained hyperglycemia, as expected, as levels of GHb were normal and essentially the same as in the nondiabetic blood samples shown in Table 1.

It has been suggested that there is a pathway by which AGEs interact with cell surface receptors, leading to altered osteoblast function (44–46). Our data suggest that RAGE in particular may contribute to AGE modulation of osteoblast function. It is notable that RAGE expression in osteoblasts has not been previously reported. We wished to determine whether phenotypically normal murine osteoblastic MC3T3 cells expressed RAGE mRNA and protein. Murine RAGE cDNA and genomic sequences were obtained to permit design of RT-PCR primers for assessment of RAGE mRNA. BLAST searches of murine Expressed Sequence Tag databases were performed against the published rat RAGE cDNA sequence identified murine cDNAs, with 98% similarity to the rat sequence (e.g., accession number AA882247). BLAST searches of the murine cDNA sequence against murine genomic databases identified the RAGE gene sequence on mouse chromosome 17 (accession numbers AC006289 and MMHC29N7). Exon/intron sequences and organization were identified and two RT-PCR primer pairs were designed that flank introns 2 and 7 (primer sets A and B, respectively; see research design and methods). As is shown in Fig. 6A, both primer sets in RT-PCR reactions of total RNA from murine MC3T3 cells produced PCR products consistent with the presence of RAGE mRNA. RT-PCR products were cloned and sequenced and found to be identical to the expected sequences for murine RAGE cDNAs. Thus murine MC3T3 osteoblasts express RAGE mRNA.

To determine the presence of RAGE protein in osteoblasts, SDS-PAGE sample buffer extracts of MC3T3 cells and primary rat osteoblast cultures were assayed by Western blotting using anti-RAGE antiserum. As shown in Fig. 6B, extracts made from MC3T3 cells and primary rat osteoblasts contained a specific immunoreactive band at ∼50 kDa, consistent with the presence of full-length RAGE protein (47).

Osteopenia is a complication of diabetes in humans (4–6,11–14). Evidence for reduced osteoblastic activity and osteopenia has been reported in type 1 diabetes (11–14,48,49). The results of the present studies indicate that type 1 diabetes impairs intramembranous bone healing, as demonstrated by our finding that the degree of healing of circular osteotomies was reduced by 40% in diabetic animals versus nondiabetic controls. To rule out other complications that could potentially arise from STZ treatment, a diabetic group was treated with insulin. Insulin treatment largely restored healing, indicating that the impaired healing was strictly related to the diabetic state.

Biochemical and immunohistochemical studies indicate that CML-modified proteins are AGEs that accumulate in vivo in diabetes (20,30–34). The identification of these modifications as unique AGE species with the capacity to bind RAGE suggests a mechanism by which the engagement of RAGE by CML-proteins results in diabetic complications (20). In vitro studies have shown that murine osteoblastic cells possess receptors that bind AGEs, although RAGE itself has not been identified (50). We have reported here that osteoblasts do express RAGE mRNA and protein. To investigate the role of diabetes in RAGE expression in bone, immunohistochemical studies were carried out in healing osseous wounds in normal and diabetic mice. RAGE expression was detected in the nondiabetic mice and was significantly upregulated in diabetic animals. RAGE expression was noted in cells associated with bone-forming activity, suggesting that enhanced expression could affect osseous healing.

Further evidence for a pathologic role of AGE interactions in the inhibited bone healing observed was obtained by determining whether controlled local application to defects of a known RAGE ligand, CML-MSA, could inhibit healing in normal animals and mimic deficient healing observed in diabetic animals. The application of CML-MSA to craniotomy defects significantly impaired healing by at least 40% compared to animals treated with unmodified control MSA, which is not a ligand for RAGE. In future studies, it will be interesting to determine the effects on bone healing of CML extracellular matrix proteins modified to contain varying amounts of CML; this will allow us to assess the effects of more physiologically important bone proteins modified with more physiological levels of CML. It should be noted that the 30% of the lysine residues in the modified albumin used in the present study were CML residues.

Systemic injections of AGEs result in pathologic alterations, similar to the typical consequences of diabetes (51,52), and AGE accumulation in target tissues is pathogenic (30–34). CML residues are AGE antigens (30–34) and are linked to the pathogenesis of functional tissue damage in complications of diabetes (20). The finding that locally applied CML-MSA impairs osseous healing is consistent with the concept that formation of AGEs during chronic hyperglycemia represents a critical event that contributes to defective bone healing in type 1 diabetes. Moreover, this interpretation is supported by recent findings that AGE adducts inhibit osteoblast differentiation and may contribute to bone disease in part by altering the response to parathyroid hormone (50,53). The observations in the present study directly identify the increased presence of RAGE in healing diabetic bone and RAGE expression in osteoblast cultures. Moreover, the finding that biological activity of CML-MSA in inhibiting bone healing and formation in vivo mimics the effect of diabetes further supports the notion that AGEs inhibit osteoblast function and contribute to diabetic bone disease. Our results may have relevance to the mechanistic basis for increased alveolar bone loss in diabetic mice with periodontal disease (19). We speculate that AGE-RAGE interactions on osteoblastic cells inhibit osteoblast function and contribute to diminished bone formation. Further studies linking diabetic osteopenia to AGEs and osteoblast function are envisioned.

This research was supported by National Institutes of Health Grants DE-12209 and DE-14066 (to P.C.T.), DE-12482 and DE-14079 (to S.A.), and DE-13191 (to D.T.G.), and a CAPES Fellowship (Brazil) to R.B.S.

The authors are grateful to Drs. Michael Neeper, Allan Shaw, and Emilio Emini of Merck & Co., West Point, PA for the gift of goat anti-RAGE antiserum. We thank Dr. Yasuo Ido of Boston University School of Medicine for valuable advice regarding management of diabetic mice, and Dr. Wayne Gonnerman of Boston University for expert technical assistance.