Diabetes Enhances mRNA Levels of Proapoptotic Genes and Caspase Activity, Which Contribute to Impaired Healing

Genetically type 2 diabetic C57BL/KsJ-lepr^−/− (db/db) mice and their normoglycemic littermates C57BL/KsJ-lepr^−/+ (db/+) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). db/db mice developed diabetes at ∼7–8 weeks of age and had fasting serum glucose levels >250 mg/dl for a minimum 20 days before the experiments were started. The glucose levels during the experimental period were typically 400–450 mg/dl in the db/db mice and 100–150 mg/dl in the normoglycemic controls. All animal procedures were approved by the institutional animal care and use committee at Boston University Medical Center.

Porphyromonas gingivalis was used to induce injury of connective tissue and bone based on its capacity to stimulate tissue loss in humans and in animal models (24,25). Broth-grown P. gingivalis strain 381 at log phase growth was fixed with 1% paraformaldehyde for 6 h. After mice were anesthetized with an injection of ketamine (80 mg/kg) and xylazine (10 mg/kg), 5 × 10⁸ bacteria were inoculated at the midline of the scalp between the ears. This produces an inflammatory response, destruction of connective tissue, bone resorption, and a healing response that peaks in both diabetic and normoglycemic mice on day 8, as we have previously described (22,23). Mice were killed at 5 and 8 days after inoculation, and at no point did mice die from inoculation. There were six mice for each group at each data point (n = 6). The scalp and associated calvaria were prepared for histological sections. In some experiments the scalp was dissected free, snap-frozen in liquid nitrogen, pulverized, and protein or RNA extracted.

Total RNA was isolated by Trizol reagent followed by RNeasy clean up (Qiagen, Valencia, CA). For each group, db/db or db/+, there were eight mice. The integrity, purity, and quantity of the samples were examined by agarose gel electrophoresis and by optical density readings (A₂₆₀/A₂₈₀), which were within the range of 1.9–2.1. The RNA levels were examined using a Mouse Genome 430 2.0 array (Affymetrix, Santa Clara, CA). Apoptosis-related genes were selected from the array data, using PathwayAssist software (Iobion, La Jolla, CA). All steps in microarray probe preparation, hybridization, and reading of fluorescent intensity was carried out by the Harvard-Forsyth Collaborative Microarray Core (Cambridge, MA). Two separate arrays were carried out each for normoglycemic and diabetic mice. The values for each gene were established and normalized using a probe logarithmic intensity error estimate (PLIER; Affymetrix). To be considered as modulated by diabetes, the intensity values were at least 1.7-fold different for each of the diabetic arrays compared with the mean of the normoglycemic group, and the mean of the diabetic group had to be at least 2.0-fold different compared with the mean of the normoglycemic group. For selected genes RNA was isolated from a separate set of eight db/db or eight db/+ mice, and the results were confirmed by real-time PCR. Taqman primer and probe sets for murine tumor necrosis factor-α (TNF-α), caspase-3, caspase-9, FADD (Fas-associating protein with death domain), Foxo1 (forkhead box 1), and Fas were purchased from Applied Biosystems (Foster City, CA). Results were normalized with an 18S ribosomal primer and probe set (Applied Biosystems). Each amplification was performed three times with duplicate specimens and the results from the replicates combined to establish statistical significance with Student’s t test.

After inoculation of bacteria, animals were treated with the pancaspase inhibitor Z-VAD-FMK (Kamiya Biomedical). The inhibitor was applied starting 3 days after inoculation of bacteria (so as not to interfere with the early inflammatory phase) by intraperitoneal injection (2 mg/kg). Additional injections were performed daily thereafter until death. Control mice received the same volume of vehicle alone (0.2% DMSO in PBS).

After euthanasia, the head of each mouse was fixed for 72 h in cold 4% paraformaldehyde. Specimens were decalcified by incubation in cold Immunocal (Decal, Congers, NY) for ∼2 weeks and then washed with Cal-Arrest (Decal). The scalp and underlying calvarial bone were kept intact, and paraffin-embedded sagittal sections were prepared at a thickness of 5 μm. All histological counts and measurements were from one examiner and were confirmed by an independent examiner. In all histological analysis, there were six mice per data point (n = 6). Student’s t test was used to determine significant differences between the experimental and control groups.

Apoptotic cells were detected by an in situ transferase-mediated dUTP nick-end labeling (TUNEL) assay by means of a TACS 2 TdT-Blue label kit purchased from Trevigen (Gaithersburg, MD), following the manufacturer’s instructions. This kit detects double-strand breaks in genomic DNA and identifies most stages of apoptosis. The number of fibroblastic apoptotic cells was counted at high magnification (1,000×) and were identified by their characteristic appearance in the connective tissue between the coronal and occipital sutures.

Van Gieson–stained sections were prepared as described previously (26) and were used to assess the area of new collagen formation at 400× magnification. Newly formed connective tissue matrix stains blue and was measured with computer-assisted image analysis. The total fibroblast number was counted at 1,000× magnification in the connective tissue between the coronal and occipital sutures in sections stained with hematoxylin and eosin. Van Gieson–stained sections described above were also used to measure newly formed bone, which is stained blue, whereas previously formed bone is stained red. The area of new bone formation area per bone length (mm²/mm) was determined at 400× magnification by image analysis software. The length of eroded bone surface on day 8 was measured at 400× magnification in sections stained with hematoxylin and eosin as previously described (27), using an image analysis system. As an indication of the degree of coupling between bone formation and resorption, the amount of newly formed bone was divided by the percent eroded bone surface.

Total RNA was extracted from the scalps of mice using Trizol (Invitrogen, Rockville, MD). RNA from six animals was pooled, and gene expression was measured by a RNase protection assay. ³²P-labeled riboprobes specific for murine procollagen I and III were incubated with 4 μg total RNA. After hybridization, specimens were subjected to RNase digestion using a kit from BD Pharmingen (Franklin Lakes, NJ), following the manufacturer’s instructions. After electrophoresis on a 6% polyacrylamide gel, radiolabeled bands were visualized using a PhosphorImager (Bio-Rad Laboratories, Hercules, CA). The optical density of the protected bands was measured with Image ProPlus software (Media Cybernetics, Silver Spring, MD), which was then normalized by the value of GAPDH (glyceraldehyde-3-phosphate dehydrogenase) in the same line. The results of three separate RNase protection assays were combined. Significance was determined by Student’s t test.

Caspase activity was assayed by a fluorometric kit purchased from R&D Systems (Minneapolis, MN). Frozen tissues from six specimens were pooled per data point, pulverized, and incubated in cell lysis buffer provided by R&D Systems. Total protein was determined using a BCA protein assay kit (Pierce, Rockford, IL). Then, ∼300 μg of total protein was assayed per data point. Caspase-3 activity was detected, using the specific caspase-3 fluorogenic substrate DEVD [Z-Asp(OMe)-Glu(OMe)-Val-dl-Asp(OMe)] peptide conjugated to 7-amino-4-trifluoromethyl coumarin (AFC). Caspase-8 activity was detected by using the fluorogenic substrate, IETD [Z-Ile-Glu(OMe)-Thr-Asp(OMe)]-AFC and caspase-9 activity with LEHD [Z-Leu-Glu(OMe)-His-Asp(OMe)]-AFC. Measurements were made on a fluorescent microplate reader using filters for excitation (400 nm) and detection of emitted light (505 nm). The results of three separate assays were combined. Statistical difference between samples was determined by Student’s t test.

To measure the impact of diabetes on healing of a bacteria-induced injury, bacteria were inoculated at the midline of the scalp between the ears of the mouse. This model facilitates quantitative analysis of both soft and hard tissue repair, apoptosis, and apoptosis pathways (22,23,28). To establish how diabetes modulates mRNA levels of apoptotic genes, RNA profiling was undertaken and the data analyzed by PathwayAssist software (Tables 1,TABLE 2–3). A fairly conservative approach was taken in defining an increase as a minimum of 1.7-fold in each of two microarrays and at least 2.0-fold for the mean of the two arrays. Of the 276 total number of apoptotic genes identified by the software, mRNA levels of 71 were enhanced by diabetes, and 3 were downregulated. Of the 71 genes that increased twofold in diabetic specimens, 63 were proapoptotic, and 8 were antiapoptotic. Thus, diabetes had a global effect that was markedly proapoptotic. Apoptotic genes were subclassified into groups based on functional criteria: 1) directly pro- or antiapoptotic, which includes mitochondrial factors and caspases; 2) intercellular mediators, such as death-inducing ligands; 3) receptors and intracellular mediators that include death receptors and accessory molecules; and 4) transcription factors. Diabetes increased the RNA levels of genes in each of these categories, further supporting the impact of diabetes as altering the global pattern of apoptotic gene expression. Supplemental data in an online appendix (available at http://diabetes.diabetesjournals.org) provides a list of genes that were not modulated twofold.

Real-time PCR was carried out to verify diabetes-modulated RNA levels of TNF-α, caspase-3, caspase-9, FAS, FADD, and FOXO1 (Fig. 1). In every case where expression was enhanced by diabetes more than twofold in the microarray, real-time PCR confirmed the increase. For FOXO1, where the microarray data indicated less than a twofold increase, a similar result was obtained with real-time PCR.

To establish whether caspase activity was increased in diabetic mice at the functional level, protein was isolated from the healing tissue on days 5 and 8. We previously reported that initial healing takes place on day 5 and peaks on day 8, and the highest levels of apoptosis are found on day 8 (23). On day 5 the level of caspase-3, -8, and -9 activity was 2.2-, 2.5-, and 2.7-fold higher, respectively, in the diabetic compared with normoglycemic mice, all of which were significant (P < 0.05) (Fig. 2). On day 8 there was a 3-fold increase in caspase-3 activity, a 1.3-fold increase in caspase-8 activity, and a 1.9-fold higher level of caspase-9 activity in the diabetic mice. The difference between control and diabetic mice was statistically significant for each caspase tested (P < 0.05).

To assess the impact of enhanced caspase activity, diabetic mice were treated with a pancaspase inhibitor. The inhibitor was started 3 days after inoculation of bacteria to minimize the impact on the inflammatory response, which peaks 1 day after inoculation (23). The effect on fibroblast apoptosis was assessed in histological sections and identified by the characteristic appearance of fibroblasts in the TUNEL assay (Fig. 3). Quantitative analysis indicated that after the administration of caspase inhibitor, the number of apoptotic fibroblasts in the diabetic mice decreased only slightly on day 5 but was reduced by more than half on day 8, which was significant (P < 0.05) (Fig. 3A). The number of apoptotic fibroblasts was approximately sevenfold higher than that observed for leukocytes on days 5 and 8, indicating that the leukocytes represent a small percentage of apoptotic cells during healing (data not shown). Moreover, there was no difference in the number of apoptotic leukocytes in diabetic and normoglycemic mice (data not shown).

The potential impact of apoptosis on fibroblast numbers was measured in diabetic mice treated with caspase inhibitor compared with vehicle alone. Treatment with inhibitor substantially increased the number of fibroblasts (Fig. 3B). Quantitative analysis indicated that treatment with the caspase inhibitor did not increase fibroblast density on day 5 (P > 0.05). On day 8 there was a 1.6-fold increase, which was significant (P < 0.05). Moreover, fibroblast density in the diabetic group reached a level similar to that of normoglycemic animals (P > 0.05).

The formation of new matrix is a critical step in wound repair. The functional effect of diabetes-enhanced caspase-3 activity was measured by assessing the expression of procollagen I and III mRNA (Fig. 4). Treatment with caspase inhibitor increased expression of procollagen I by 2-fold and procollagen III by 1.7-fold, both of which were statistically significant (P < 0.01). Thus, treatment of diabetic mice with caspase inhibitor significantly enhanced collagen I and III mRNA to a level that was ∼60% of the normoglycemic mice.

The formation of new connective tissue matrix was identified by VanGieson staining of histological sections (Fig. 5A). Treatment with caspase inhibitor improved the formation of extracellular matrix. The amount of new matrix produced in mice treated with caspase inhibitor was twofold higher than in the untreated mice, which was statistically significant and which agreed well with the increase in collagen expression (P < 0.01) (Fig. 5B). With caspase inhibitor the formation of new matrix was 69% of the value for normoglycemic controls, compared with 33% for the untreated diabetic mice.

In addition to examining changes in soft connective tissue, the calvarial model can be used to measure the impact of diabetes and apoptosis on new bone formation. In this model, a stimulus induces a cycle of bone resorption followed by repair (22,29). Recently formed bone was identified by VanGieson staining (Fig. 6A). Histomorphometric analysis demonstrated that there was a 1.4-fold increase in bone formation in the diabetic mice treated with caspase blocker compared those treated with vehicle alone (P < 0.01) (Fig. 6B).

Both physiological and pathological bone remodeling involves a cycle of bone loss followed by bone formation. To assess this coupling phenomenon, the amount of newly formed bone was divided by a measure of bone loss, the percent eroded bone surface (Fig. 7). Diabetes caused a significant degree of uncoupling because the amount of bone formed per eroded bone surface was threefold less in the diabetic compared with normoglycemic mice (P < 0.01). Treatment with the caspase inhibitor significantly improved coupling by 1.4-fold, consistent with the degree of improvement in new bone formation (P < 0.01).

In the scalp/calvarial model, fixed bacteria are inoculated that induce an inflammatory injury, which is similar in normal and diabetic mice (23). We used fixed bacteria so that differences between diabetic and normoglycemic mice reflect differences in the host healing response rather than a diminished capacity of diabetic mice to kill bacteria. In this model soft and hard tissue healing occurs in a similar time frame, and the wound is not exposed to the outside environment where healing can be affected by external factors (22,23). Using this model we previously reported that diabetes causes an increase in apoptosis of critical matrix-producing cells (22,23). The current studies established that diabetes-enhanced apoptosis represents a physiologically important event by using a caspase inhibitor. Several parameters of healing were improved by blocking apoptosis. These included fibroblast density, level of collagen expression, and the amount of new connective tissue matrix formed. Treatment with caspase inhibitor significantly increased new bone formation as well, although the magnitude of the effect was less than that observed for soft connective tissue formation. Thus, the higher levels of diabetes-associated apoptosis of fibroblastic and osteoblastic cells previously reported (22,23) significantly contributes to a deficient healing response in diabetic individuals.

mRNA profiling experiments provided insight into mechanisms by establishing that diabetes caused a global induction of proapoptotic genes during the repair process. Of a total of 276 apoptotic genes examined, 71 genes increased twofold or more. Of these genes, 63 were proapoptotic and 8 were antiapoptotic genes. Only three genes were downregulated, and all of these were antiapoptotic. Although there have been no publications on the impact of diabetes on the global pattern of apoptotic gene expression during healing or in response to infection, it has been reported that diabetes enhances proapoptotic gene expression in the liver and retina of diabetic animals compared with normoglycemic animals (30,31). Thus, one of the potential mechanisms whereby diabetes negatively affects the function of an organ is to enhance the expression of proapoptotic genes. This, in turn, may promote apoptosis by tipping the intracellular balance in a proapoptotic direction.

One of the apoptotic genes upregulated at the RNA level was caspase-3. Functional studies indicated that caspase-3 activity was also enhanced in the diabetic group, along with both caspase-8 and caspase-9, suggesting that both extrinsic and intrinsic apoptotic pathways are involved. Activation of caspase-3 is associated with hyperglycemia-induced myocardial apoptosis (32) and apoptosis of neuronal cells in diabetic neuropathy (33–35). It is also associated with apoptosis of mesangial cells that occurs in diabetic nephropathy (36,37). Activation of caspase-3 typically occurs via the cytosolic and/or mitochondrial pathways (38). Data reported here indicate that both of these pathways appear to contribute to diabetes-enhanced apoptosis during dermal healing.

Previous studies have supported the concept that repopulation of wounds by fibroblasts is caused by a decrease in growth factor production and cellular proliferation (7,9). Thus, the failure to achieve a sufficient number of fibroblasts could potentially come from two different mechanisms: a failure to stimulate sufficient proliferation or a significantly enhanced rate of programmed cell death. Consistent with this principle are findings that conditions that enhance apoptosis are associated with impaired healing (39), whereas those that reduce apoptosis are correlated with qualitative and quantitative improvements in wound repair (40). Similarly, enhanced osteoblast apoptosis is thought to play a role in deficient bone formation. For example, glucocorticoid-induced osteoporosis occurs under conditions that enhance osteoblast apoptosis (41). Increased osteoblast apoptosis also impairs bone formation in myeloma bone disease (42). Conversely, reduced osteoblast apoptosis may enhance bone formation (43,44).

Therefore, studies presented here provide the first direct evidence that increased apoptosis of fibroblasts and osteoblasts caused by diabetes functionally contributes to impaired healing.

This work was supported by National Institute of Dental and Craniofacial Research Grants DE07559 and DE11254.

We thank Renee Cabral for technical assistance and Alicia Ruff for help in preparing the manuscript.