Tissue Factor as a Link Between Wounding and Tissue Repair

Four- to five-month-old female nonobese diabetic (NOD) mice (M & B, Ry, Denmark), characterized by a cumulative incidence of diabetes (38), were used in the experiments. No artificial interventions were used to induce diabetes, and thus the adversary impact of obesity on wound healing, known to occur in some diabetic animal models (39), was excluded. Blood glucose was determined using Accu-Check sensors (Roche, Mannheim, Germany). Spontaneous diabetic NOD mice were maintained at a glucose level >500 mg/dl for at least 4 weeks before experiments. Blood glucose levels were adjusted using insulin (3–5 IU s.c. semilente; Novo Nordisk, Bagsværd, Denmark). Age- and weight-matched nondiabetic NOD mice used as controls had blood glucose between 90 and 120 mg/dl. During the experiments, mice were housed one per cage, maintained under controlled environmental conditions (12-h light/dark cycle) at an average temperature of 23°C, and had free access to standard laboratory food and water.

Mice were anesthetized using 13 μl of 2.5% avertin/g body wt before hair on the skin was shaved and skin was swabbed with alcohol. Two round cutaneous wounds were generated at both sides of the lower dorsal trunk after shaving the hair and disinfecting the skin with Softasept N (B Braun Melsungen, Melsungen, Germany). The area of circular skin removed was 8 mm in diameter. Curved scissors and surgical techniques were used. The wounds were well separated by >1 cm skin. This protocol was approved by the veterinarians of the animal facilities of the University of Heidelberg and the Regierungspräsidium Karlsruhe (Az: 35-9185.81-2, experimental no. M 5/99). No increased bleeding was observed in wounds of diabetic animals compared with nondiabetic controls.

Full-length mTF cDNA (5) was cloned in the pcDNA3 expression vector (Invitrogen, Taastrup, Denmark) under the control of a cytomegalovirus promoter. Vector alone (pcDNA3) served as control. A total of 2 μg of vector or TF plasmid DNA was suspended in 38 μl of HEPES (Roth, Karlsruhe, Germany)-buffered saline (20 mmol/l HEPES, 160 mmol/l NaCl) and mixed with 12 μl of Dosper in 28 μl of HEPES-buffered saline and incubated for 15 min before being applied to the wounds on each mouse. Subsequently, the transfected wounds were covered with an adhesive surgical drape, Tegaderm (3M, Borken, Germany), for 12 h. The time of DNA application was recorded as 0 h. Transfection of vector or TF was rotated between both sides to avoid a side-dependent artifact. For all the experiments comparing nondiabetic with diabetic mice, vector was added to the wounds of the nondiabetic mice, since pilot experiments demonstrated that vector did not affect wound healing.

For PCR evaluation, the wound area was removed at the time points indicated in the figure legends using surgical scissors. The skin was cut immediately (1 mm) adjacent to the wound edge including a thin layer of tissue below the wound bed. The recovered tissue was snap-frozen in liquid nitrogen then stored at −80°C. For immunohistochemistry, a larger and deeper area of tissue was excised and denatured in neutral-buffered 4% formaldehyde solution.

Transfection with a reporter-plasmid expressing β-galactosidase (psV-β-Gal; Promega, Heidelberg, Germany) was performed as described above. X-gal immunhistologic staining was performed according to the protocol provided by the manufacturer. Briefly, the wound sample was sectioned and fixed in 37% formaldehyde fixing solution for 2 h at 4°C. After washing two times in PBS, the sample was incubated in staining solution (20 mg/ml X-Gal) for 2 h at 37°C for colorimetric development.

Transfection with a reporter-plasmid expressing luciferase (pGL-luc; Promega) was performed as described above. Wounds were harvested and immediately frozen in liquid nitrogen. The snap-frozen tissue was ground using a mortar and pestle and transferred into 200 μl lysis buffer (50 mmol/l Tris-HCl [ph 8.0], 150 mmol/l NaCl, 0.02% sodium acid, 1% Triton X-100, 1 mmol/l leupeptin, 2 mmol/l benzamidin, 20 μg/ml soybean inhibitor, 1 μg/ml aprotinin, and 0.5 mmol/l phenylmethylsulfonyl fluoride). Hair and other insoluble material were removed by centrifugation at 4,000 rpm for 2 min, before the supernatant was subjected to four freeze-thaw cycles, followed by a 20-min incubation on ice with several vortex steps (10 s each) and a final centrifugation for 5 min at 15,000 rpm. Protein concentration in the supernatant was determined using the bicinchoninic acid assay kit according to the manufacturer′s instructions (Pierce Biotechnology, Rockford, IL). A total of 20 μg of total protein extracts were separated onto 10% SDS-PAGE, followed by electroblotting to enhanced chemiluminescence–nitrocellulose membranes. Membranes were incubated with a primary antibody for luciferase (1:1,000; Promega) for 60 min at RT. After washing (2 × 10 min in TBS, 0.05% Tween), the secondary antibody (horseradish peroxidase–coupled donkey IgG, 1:5,000; Santa Cruz, Heidelberg, Germany) was added, and incubation was continued for 30 min at room temperature. Membranes were washed 4 × 10 min as above. Immunoreactive proteins were detected with the enhanced chemiluminescence–Western blot system (Amersham-Pharmacia, Freiburg, Germany) and subsequent autoradiography for 10 min.

The vertical and horizontal diameters of each wound were recorded immediately after wounding at day 0 and at subsequent days thereafter in mice anesthetized using inhalation anesthesia, preventing possible wound damage due to foreign force. The mean of the diameters was used in the calculation of each wound area. Wound size was calculated as percentage of day 0. The experiment was performed three times with similar results.

Total RNA was extracted from the excised wound tissue using RNeasy mini kit (Qiagen, Hilden, Germany), according to the manufacturer’s directions. Briefly, tissue was disrupted with mortar and pestle and homogenized by passing through a 20-G needle 20 times in 350 μl of RLT buffer. After column purification, RNA was eluted with 30 μl of RNase free water. A total of 100 ng of total RNA was annealed with 1 μg of oligo d(T) (Pharmacia, Freiburg, Germany) at 65°C for 10 min then used for reverse transcription with 15 units avian myeloblastosis reverse transcriptase (Promega, Heidelberg, Germany) at 42°C for 1 h. cDNA was used immediately or stored at -80°C. A total of 2 μl of cDNA, 20 pmol of each primer, and 1.25 units of Taq polymerase (Promega) were used in each PCR at a final volume of 50 μl. The primers used are shown in Table 1. Cycling conditions were (94°C 4 min); (94°C, 1 min; 62°C, 30 s; 72°C 1 min) × 28; (72°C, 2 min) for mglyceraldehyde-3-phosphate dehydrogenase (GAPDH); (94°C, 4 min); (94°C, 1 min; 55°C, 2 min; 72°C 2 min); (94°C, 1 min; 60°C, 30 s; 72°C, 1 min) × 32; (72°C, 2 min) for mTF and α-smooth muscle actin (α-SMA) and (94°C 4 min); (94°C, 1 min; 55°C, 2 min; 72°C, 3 min) × 34; and (72°C, 3 min) for mVEGF. The expected PCR products were 983 bp for mGAPDH, 640 bp for α-SMA, 417 bp for mTF, and 417 bp for mVEGF. After separating on a 2.0% agarose gel, PCR products were visualized and photographed (Polaroid, 667) on an ultraviolet screen. Pictures were scanned by an imaging densitometer (GS700, Bio-Rad), and the intensity was analyzed by Molecular Analyst software (Version 1.5; Bio-Rad). Intensities of target gene products were normalized by that of GAPDH to obtain the relative densities shown.

Real-time PCR was performed on a LightCycler (Roche, Mannheim, Germany), as essentially described by Galiano et al. (40). First, 3 μl of cDNA, 20 pmol of each primer, and 125 units Taq polymerase (Promega) were used in each reaction at a final volume of 50 μl. The primers used are summarized in Table 2. Cycling conditions were (95°C, 5 min); (94°C, 1 min; 50°C, 2 min; 72°C, 2 min); (94°C, 1 min; 52°C, 30 s; 72°C, 75 s) × 25; (72°C, 6 min) for mGAPDH; (95°C, 5 min); (94°C, 1 min; 55°C, 1 min; 72°C, 30 s) × 40; (72°C, 5 min) for mTF; (94°C, 4 min); (94°C, 1 min; 55°C, 20 s; 72°C, 3 min) × 34; (72°C, 3 min) for mVEGF; (94°C, 1 min); (94°C, 1 min; 55°C, 2 min; 72°C, 2 min); (94°C, 1 min; 60°C, 30 s; 72°C, 1 min) × 35; and (72°C, 2 min) for αSMA. The expected PCR products were 191 bp for mGAPDH, 230 bp for mTF, 189 bp for VEGF, and 218 bp for α-SMA. To provide the required standards for real-time PCR, these fragments were cloned using the TOPO TA cloning kit, amplified, purified using the PCR purification kit (Qiagen), and verified by sequence analysis. Serial dilutions of each purified product were made and standard curves obtained on a real-time PCR cycler (LightCycler; Roche) using the SYBR Green I kit (Roche) according to the manufacturer′s instructions. The resulting standard curves were exported into the RelQuant software (Version 1.01, Roche) to establish individual standard curves (coefficient file), allowing evaluation of the final data including normalization to GAPDH and PCR efficiency correction.

Thereafter, GAPDH was determined in the same LightCycler run as the actual target gene. Amplification was performed using the following conditions: mTF + mGAPDH: (95°C, 10 min); (94°C, 10 s; 55°C, 5 s; 72°C, 10 s) × 45; mVEGF + mGAPDH: (95°C, 10 min); (94°C, 10 s; 56°C, 5 s; 72°C, 5 s) × 40; α-SMA + mGAPDH: (95°C, 10 min); (94°C, 10 s; 57°C, 5 s; 72°C, 5 s) × 40. Data acquisition was performed at the end of each extension step. A melting curve was designed at a 0.1-degree downward slope. For each experiment, all samples were measured as duplicates, and a negative control was included by replacing cDNA with H₂O. Using the standard curves for each gene, the relative number of copies was determined using the RelQuant software and expressed as standardized units.

Immunohistochemistry was performed by indirect immunoperoxidase techniques using dewaxed 4-μm thick paraffin sections on silane-coated glass slides. Before immunostaining, the sections were treated with 10 mmol/l sodium citrate buffer (pH 6.0) in a microwave (750 W, two times for 5 min each). The slides were treated with 0.3% hydrogen peroxide and dissolved in methanol for 30 min at room temperature. Blocking was performed with 3% normal goat or horse serum for 30 min at room temperature before the affinity-purified polyclonal rabbit antibody or the monoclonal mouse antibody were used as primary antibodies. Slides were washed three times for 10 min each in PBS, pH 7.4, and detection of signals was performed using the Vectastain ABC kit (Vector Laboratories, Burlingham, CA), according to the manufacturer’s instructions. Peroxidase activity was visualized with 0.05% 3,3-diaminobenzidine-tetrahydrochloride (Vector Laboratories) before the slides were counterstained with Mayer’s hematoxylin (Vector Laboratories). Controls for immunospecificity were included in all experiments by omission of the primary antibody and its replacement by PBS and matching concentrations of normal rabbit serum (data not shown). Primary antibody of TF was raised in rabbit as described previously (2,19,25) at dilution 1:200. VEGF antibody was mab C1 (Santa Cruz, Heidelberg, Germany) at dilution 1:100. Mouse α-SMA antibody was mab, clone 1A4 (Coulter Immunotech, Krefeld, Germany) at dilution 1:50. Immunohistochemistry for fibrin/fibrinogen using a rabbit anti-murine fibrin/fibrinogen antibody (dilution 1:800; Innovative Research, Hilltop, Southield, MI) was performed without pretreatment of tissue sections.

The antibody-stained sections were scored based on the strength of the brown product present. The designations used were no staining (= −), light or weak staining (= +), strong staining (= ++), and intense staining covering most of the cellular tissue and matrix (= +++). Consecutive sections that stained without antibodies were used as negative controls for specific staining. Staining was analyzed independently by J.C., A.B., and M.K.

For the evaluation of blood vessels presented in the wounds 7 days after transfection, hematoxylin and eosin–stained sections (as described above) were analyzed under microscope at magnification of 200×. Blood vessels were counted per microscopic field at the middle of each wound section independently by J.C., L.B., and M.K. and expressed as number of vessels per microscopic field.

To visualize the vascularization of the wounds in the healing process, the animal was injected with black ink 7 days after transfection. The animal was killed in CO₂. The thorax was opened immediately, and 1–2 ml of blue ink was rapidly injected into the left ventricle while the heart was still beating. Wound and surrounding skin were isolated from the mouse when the heartbeat stopped.

For the measurement of blood flow, E-Z Trac Ultraspheres (E-Z Trac, West Los Angeles, CA) were used as previously described (41). Briefly, 7 days after transfection, the animal was anesthetized with diethylether and the chest opened. Then 1 × 10⁶ microspheres (diameter 10 μm), dissolved in 1 ml of 0.9% NaCl, were injected into the left ventricle. The mouse was killed 3 min after injection, and the skin was taken. The numbers of microspheres in each wound sample were counted under a microscope after dissolving the wound region of the skin in 5 N NaOH + 1% SDS and expressed as number of microspheres per microgram tissue.

Results were expressed as means ± SD. Student’s t test was used in statistic analysis. P values of <0.05 were considered as statistically significant.

When wound healing was studied, nondiabetic and diabetic NOD mice were subjected to a standardized procedure for cutaneous wounding. The time needed to achieve 50% closure of the wound increased from 3.6 ± 0.98 days in nondiabetic to 5.5 ± 1.92 days in diabetic mice (P = 0.002). Wound tissue was recovered at different time points, and the expression profile of TF mRNA was determined (Fig. 1A, B top row, C, and D). Rapid induction of TF (∼3.2-fold) occurred in nondiabetic mice (P = 0.046 at 1 h, 0.027 at 3 h) (Fig. 1C), while no significant induction occurred in diabetic mice (Fig. 1D).

After the increase in TF transcripts, VEGF induction occurred in nondiabetic mice (Fig. 1E), rising approximately twofold above the baseline within 12 h, however, without reaching statistical significance. Levels of VEGF transcripts were maintained for 24 h, thereafter slowly decreasing at day 7, a time when wound healing was almost complete. In contrast, tissue from diabetic mice subjected to cutaneous wounding displayed a small, but not significant, induction of VEGF transcripts not before 12 h after wounding (Fig. 1F) and further increased at day 3 when the wound in the diabetic animal was ∼47% closed. Thus, TF and subsequent VEGF gene expression in wounds from nondiabetic animals could be contrasted with lack of TF and VEGF induction in diabetic animals within the first hours of wounding.

Similar results were observed when the transcription of α-SMA mRNA was analyzed (Fig. 1A, B, G, and H). Within 1 h of wounding, nondiabetic mice displayed an apparent, but not significant, induction of α-SMA mRNA (P = 0.25), while upregulation of α-SMA transcripts was absent in diabetic mice. These data are consistent with the hypothesis that attenuation of the inflammatory response in diabetic wounds (33,34), possibly decreasing the content of proinflammatory cytokines, may contribute to the blunted induction of TF and angiogenesis.

To monitor efficacy of somatic gene transfer, wounds were transfected with human TF (hTF) plasmid DNA instead of mouse TF (mTF) plasmid to avoid cross-reactivity with endogenous mTF transcripts. These experiments demonstrated that the transfected plasmid was still transcribed in wounds 7 days after transfection (Fig. 2A). Next, wounds were transfected with plasmids overexpressing a luciferase reporter gene. Wounds were harvested on consecutive days (days 0–12), and luciferase expression levels of luciferase were determined in a course of 0–12 days by Western blot analysis. Strong luciferase gene expression was observed between day 1 and day 3 (Fig. 2B) and, to a lesser extent, was still evident until day 7. To circumvent the limitations of Western blot analysis in its sensitivity and the inability to localize the expressed antigen at the cellular level (e.g., to a subset of transfected cells), immunhistochemistry was performed after delivering a plasmid carrying LacZ into the wound. Expression of LacZ was still evident 7 days later (Fig. 2C). The above evidence confirmed that topical somatic gene transfer is an effective method to allow gene expression in wounds of mice.

Thereafter, somatic gene transfer was used to elevate TF expression in diabetic wounds, in order to assess the relationship between TF and impaired VEGF and α-SMA induction. Wounds in diabetic mice were harvested 1 day after somatic gene transfer with either vector alone or TF cDNA cloned into pcDNA3. Somatic gene transfer using the TF overexpressing plasmid resulted in increased levels of TF (Fig. 3A, P = 0.004). This change in expression of TF transcripts, following transfer of the TF cDNA, was paralleled by induction of VEGF (P = 0.014) and α-SMA mRNA (P = 0.14) in diabetic mice compared with mice treated with vector alone. Studies in nondiabetic mice displayed a much smaller increase in TF, VEGF, and α-SMA transcripts (data not shown). Real-time PCR with samples randomly chosen at time points day 1 and day 7 from vector and TF transfected wounds confirmed these results (Fig. 3B), although it failed to show a significant VEGF induction at a marginal level.

To further study the effect of TF expression on VEGF and α-SMA, levels of TF antigen expression were determined on day 2, and subsequent VEGF and α-SMA antigen were analyzed on day 7 in diabetic animals (Fig. 4). Normally, the intact granular cell layer expresses TF, whereas the basal layer of the epidermis and dermal connective tissue is weak to negative (42). In diabetic wounds treated with the TF expression vector, TF antigen was more prominent at the migrating epidermal tip compared with vector controls. This suggests successful transfection and overexpression of TF in keratinocytes. In addition, the number of fibroblasts and inflammatory cells positive for TF antigen increased in the wounded area of mice treated with the TF expression vector compared with the neighboring skin, where they were mostly negative (Fig. 4A).

Seven days after TF gene transfer, VEGF expression was higher in the diabetic wound than in neighboring tissue (Fig. 4B). A significant increase in VEGF antigen (P = 0.017) was observed in the TF-transfected wounds compared with wounds treated with vector alone (Fig. 4B). In diabetic wounds exposed to TF expression vector, VEGF-positive cells appeared to be distributed according to a gradient of antigen expression, with highest levels at the wound edge and decreasing signal intensity toward the wound center (Fig. 4B, middle). This pattern of VEGF expression was not observed in vector control–treated wounds. In the latter case, there was no significant increase in VEGF antigen either at the wound edge or in the center (Fig. 4B, top). Cells expressing VEGF in TF-transfected diabetic wounds were keratinocytes, fibroblasts, and smooth muscle cells. Increased immunostaining for VEGF antigen at the edge of TF-transfected diabetic wounds corresponded to sites of increased staining for TF antigen (Fig. 4A). Upregulation of TF, by somatic gene transfer into diabetic wounds, also increased expression of α-SMA antigen (Fig. 4C). Immunostaining of TF-transfected wounds with α-SMA antibody showed cells expressing α-SMA antigen to be mostly concentrated around vessels. About 49% more α-SMA positive vessels were seen in TF-transfected wounds (Fig. 4C, bottom) than in vector-transfected wounds (Fig. 4C, top) (P = 0.037). Furthermore, fibrin/fibrinogen-deposition was increased in wounds from mice transfected with TF (Fig. 5). Since the antibody used did not discriminate between fibrin and fibrinogen, the wounds were extensively washed before staining to remove residual fibrinogen. The reactive material represented at least in part fibrin, since we observed striking differences in fibrin/fibrinogen deposition between the vector- and TF-transfected wound that correspond to the data obtained for TF, VEGF, α-SMA, and macroscopic wound healing in the differently transfected mice. Thus, overexpression of TF results in enhanced VEGF and α-SMA transcription and expression and increased fibrin deposition in diabetic wounds.

The functional significance of the local TF delivery was shown by injecting ink into the vascular space to highlight dermal blood vessels (Fig. 6A). In both vector- and TF-treated diabetic wounds, the density of vessels was high at the wound edge, forming a vascular rim around the wound periphery. TF-transfected diabetic wounds (Fig. 6A, bottom) showed a more prominent protusion of vasculature to the wound center, as well as an apparent elevation of vessels at the center, compared with vector-transfected wounds from the same mouse (Fig. 6A, top). Hematoxylin-eosin staining of the wounds from diabetic animals treated either with vector (Fig. 6B, top) or TF (Fig. 6B, bottom) demonstrated an increase in the number of vessels per microscopic field in TF-transfected mice. Furthermore, the density of blood vessels in TF-transfected wounds (Fig. 6C) was 65% higher (P = 0.035) than in vector controls. To analyze blood flow through the wound area, microspheres were injected into the left ventricle. After dissolving the wound tissue, the TF-treated diabetic wounds displayed 92% more beads (P < 0.05) per gram of tissue than vector-treated wounds (Fig. 6D).

The effect of enhanced TF expression in wounds was not only evident at the level of induction of VEGF and α-SMA, and an increase in newly formed vessels, but also with respect to macroscopic appearance of the wound and accelerated closure (Figs. 7A–C). To evaluate the effect of TF on wound healing in situ, wound diameters in nondiabetic mice or diabetic mice either treated with vector or TF were measured daily in anesthetized mice. The boundary of the migrating tip of the neoepidermis was taken as wound edge. Wound size on a given day was compared with the original area determined immediately after skin injury. Wound closure reflected the area covered by the neoepidermis as a result of reepithelialization. Transfer of TF resulted in more rapid closure of diabetic wounds compared with vector-treated control wounds. After 3 days, TF-treated wounds displayed a significantly (P = 0.02) accelerated reduction in wound size compared with vector-treated controls (both of which were placed on the back of the same diabetic mouse) (Fig. 7A).

When the median time required to reach 50% wound closure was determined, this time was reduced from 5.5 days in vector-treated diabetic mice to 4.1 days in TF-treated animals (Fig. 7C) (P = 0.048). Wound closure in diabetic mice subjected to TF gene transfer was almost comparable to that observed in nondiabetic control mice: the time lag before 50% wound closure in diabetic TF-treated mice was 4.1 days and did not differ significantly from 3.6 days in nondiabetic animals (P = 0.321). These observations suggested that increased local expression of TF in diabetic wounds drives processes relevant to tissue repair, potentially linking early steps in the inflammatory phase to the proliferative and resolution stages.

Previous reports (33,34) have shown that the inflammatory phase of wound repair is delayed/diminished in diabetes. The results reported herein indicate that early induction of TF in wounds is an essential component of the host response and is also attenuated in a murine model of diabetes. This relative paucity of TF in diabetic wounds appears to result from decreased expression of TF in keratinocytes, fibroblasts, and smooth muscle cells rather than simply reflecting decreased influx of TF-bearing mononuclear phagocytes, leading us to hypothesize that conditions leading to TF induction in the wound are suboptimal. Elevating TF expression in diabetic wounds by local somatic gene transfer increased VEGF, α-SMA synthesis and fibrin deposition, enhanced formation of neovasculature, and, ultimately, normalized wound closure. Thus, the induction of TF in the early phase of the host response to wounding appears to link the initial inflammatory phase to later stages of the healing process. These observations add to a growing body of literature indicating that dysregulation of TF, because of excessive expression, as in septicemia; disseminated intravascular coagulation (43–45); atherosclerosis (46,47) and thrombosis (48); or deficiency, as in embryonic development in TF^−/− mice (16–18,49), has implications for pathologic and/or physiologic processes. The time course of TF induction early after wounding in order to maximally enhance reparative mechanisms provides insight into a physiologic context in which TF functions as an immediate early gene. Expression of TF must occur rapidly after wounding in order to effectively orchestrate subsequent events in repair, including expression of VEGF and α-SMA, and neovascularization.

Somatic gene transfer using TF accelerated wound healing in diabetic but not in nondiabetic (data not shown) mice despite increasing VEGF levels in each case. This suggests that there is a narrow window of TF and VEGF levels critical for triggering physiologic events, such as wound repair. However, it is important to note that the situation in the diabetic wound is complex, as it has been shown in ob/ob mice that leptin reconstituted VEGF expression in the wound but failed to restore the angiogenic response (50).

Factors underlying impairment of the inflammatory response in diabetic wounds remain to be elucidated. Several studies (33,34) have noted blunted induction of inflammatory cytokines after wounding in diabetes. This is surprising, since sustained activation of the proinflammatory transcription factor NF-κB is a characteristic feature of peripheral blood mononuclear cells derived from patients with diabetes (51). However, the situation after local wounding differs significantly from that in the systemic circulation. Namely, brisk and prominent upregulation of TF expression is required to promote wound repair, whereas long-term low-level activation of circulating mononuclear phagocytes has been observed in the intravascular space. Insight into a potentially significant negative regulator of inflammation in diabetes is derived from a recent in vitro study (52) in which high glucose induced NF-κB activation and expression of endothelial nitric oxide (NO) synthesis. Increased levels of intracellular NO apparently mediated inhibition of endothelial cell migration. Since endothelial TF expression has also been shown to be inhibited in the presence of NO (53), one might speculate that hyperglycemia-induced NF-κB activation might cause generation of NO, which would serve to inhibit TF expression.

It is likely that the effects of TF on wound healing extend beyond enhanced expression of VEGF. For example, induction of α-SMA is likely to contribute to the stability of the newly formed vessels and prevent apoptosis of the growing endothelial cell layer, as previously shown for pericytes (54). In this regard, it is noteworthy that TF has also been shown to affect migration of smooth muscle cells (55) and mononuclear phagocytes (56). Furthermore, TF might also contribute to wound repair by promoting formation of a provisional fibrin matrix facilitating epidermal cell migration (57). Consistently, the effects of TF transfection on VEGF expression, α-SMA induction, and fibrin deposition described here are minor and do not always reach statistical significance, while wound healing (Figs. 6 and 7) is significantly enhanced. Therefore, we hypothesize that the beneficial effects of TF are mediated by an interplay of different processes (VEGF expression, recruitment of α-SMA–positive cells, and fibrin deposition) and not by promoting expression of a single gene.

Although physiologic upregulation of TF in wounds is transient, its effects on the reparative process appear to be long lasting. Generation of a fibrin-rich matrix supporting the cellular response, at the level of cell migration and sequestration of growth factors, is one example of a mechanism through which an early burst of TF following wounding could modulate subsequent events important for healing. This may explain why even though relatively few cells expressed TF after somatic gene transfer (e.g., principally keratinocytes), there was a strong effect on the overall cellular response.

In vivo, the multiple contributions of TF to wound healing are difficult to separate. In addition to hemostasis, TF-mediated activation of the procoagulant cascade might also enhance wound healing through paracrine/hormonal effects of factor Xa (58,59), thrombin (60), and other products of coagulation that have been shown to impact on VEGF and platelet-derived growth factor generation (23,24). Keratinocytes form an activated epithelium that is critical for wound resurfacing (61). In vitro, factor VIIa induces TF-bearing keratinocytes to express genes associated with inflammation, proliferation, angiogenesis, cellular reorganization, and migration. All of these are central to wound healing. Our studies further suggest an important role of keratinocytes in wound healing because of their constitutive expression of TF. Furthermore, expression of VEGF by keratinocytes provides a potential situation in which TF can regulate angiogenesis at the level of the keratinocyte. In this context, TF has been linked to keratin expression in tumor cells (62). Finally, wound repair involves both dermal contraction and epithelial migration, both of which occur in the hemostatic and inflammatory responses (61). Whether delayed wound healing in diabetes involves either of these mechanisms, both of which can be modulated by TF, is not clear. The latter demands for further studies identifying the molecular mechanisms underlying the metabolic regulation of local TF expression in diabetic wounds and to answer the question, whether good glycemic control per se might influence TF expression. Our results clearly indicate that the relative lack of TF induction at the site of diabetic wounds contributes to the impaired reparative response. We propose that upregulation of TF at such local sites, potentially by somatic gene transfer, may enhance the closure of diabetic wounds, at least in settings analogous to our acute murine model.

We thank Merckle, Ulm, Germany, for support of this study. P.P.N. was supported by DFG (Na 138/4).

We thank Sven-Erik Bursell (Joslin Diabetes Center, Boston, Massachusetts), Hans A. Dresel (Progen, Heidelberg, Germany), and Udo Hartlaub (Merckle, Ulm, Germany) for helpful discussions. Part of this work was published/presented in preliminary abstract form in Thromb Hemost 464 (Suppl. 1):A1469, 1999 and Semin Thromb Hemost 24:207, 1998.