The biological response to traumatic injury in higher organisms can be subdivided into two general classes: regenerative and nonregenerative types of wound healing (Gross, 1996; Stocum, 1996). In the mammalian adult, the healing process is normally accomplished by the replacement of mature cells, but not organs, and the formation of scar tissue (Clark, 1996). In studying through-and-through ear hole wounds and their closure, we found both phenotypic (Clark et al., 1998; Heber-Katz, 1999; Samulewicz et al., 2002) and genotypic (McBrearty et al., 1998; Blankenhorn et al., in press) differences in the wound healing response of MRL vs. C57BL/6 (B6) adult mice. Healing in the MRL appeared fetal-like (Hopkinson-Woolley et al., 1994; Armstrong and Ferguson, 1995) with the formation of a blastema, scarless healing, and with the replacement of lost tissue such as cartilage by functionally and architecturally normal tissue. This finding was not true for the B6 strain, which scarred at the margins of the wound with minimal hole closure. Thus, this experimental model is regenerative rather than repair and is one more of the rare examples of regeneration among mammals including the replacement of antlers (Goss, 1970), digits (Douglas, 1972; Illingworth, 1974; Borgens, 1982; Neufield, 1989), and closure of rabbit ear-holes (Goss and Grimes, 1975).
The breakdown of the extracellular matrix (ECM) is a critical event in wound healing (Clark, 1996; Werb, 1997; Murphy and Gavrilovic, 1999). Matrix metalloproteinases (MMPs) are part of a general protease system that includes serine and metallo- and cysteine proteinases and play a key role in ECM breakdown and in a variety of normal and pathologic processes (Matrisan, 1992; Parks and Mecham, 1998). They are involved in development (Hay, 1984; Chin and Werb, 1997) as well as in regenerative processes (Grillo et al., 1968; Yang and Bryant, 1994; Stocum, 1995; Chernoff et al., 2000).
MMP-2 and MMP-9 (known as type IV and V collagenases or the 72-kDa gelatinase A and 92-kDa gelatinase B, respectively) are secreted by several cells during ECM remodeling (Parks, 1999; Steffensen et al., 2001). The level of MMP-2 and MMP-9 expression is up-regulated in the wound stroma due to migratory fibroblasts. These cells have the potential to cleave and degrade type IV collagen in the basement membrane and collagens type I and III in the ECM. MMPs are not expressed, however, by stationary tissue fibroblasts (Hakkinen et al., 2000). Neutrophils (Schwartz et al., 1998) and macrophages infiltrating through vessel walls and vascular basement membrane are key producers of MMPs in wound healing with these molecules also being important in phagocytosis (Yong et al., 1998).
The MMPs are secreted as inactive zymogens (pro-form) and activated by proteolytic cleavage of their N-terminus (Parks and Mecham, 1998). This process is carried out by the protease MT1-MMP (Okada et al., 1997). A family of specific endogenous tissue inhibitors of MMP (TIMP) control MMP cleavage leading to enzymatic activity (Woessner, 1991; Baker and Leaper, 2000). This family has four members: TIMP-1, TIMP-2, TIMP-3, and TIMP-4. TIMP-1 is a strong inhibitor of all MMPs, can bind to MMP-9, and be secreted from the cell as a complex (Clark, 1996). TIMP-2 preferentially interacts with MMP-2 (Boone et al., 1990; Howard et al., 1991) and TIMP-3 with MMP-9 (Suomela et al., 2001). In both of these cases, TIMP interactions block MMP catalytic activity but have no effect against other types of proteinases. In addition, TIMPs have been reported to have an ability to promote growth in a variety of human and bovine cell types (Hayakawa et al., 1992).
In the studies presented here, we have shown the disappearance of the ECM and basement membrane in the healing MRL ear wounds before blastema formation and have examined MMP-2 and MMP-9 expression, activity, and cellular location during ear hole closure. We have similarly examined TIMP-2 and TIMP-3 expression. We have shown that there are clear differences between the MRL and B6 response to injury which could explain the differences in healing and blastema formation in the MRL and lack of it in the B6.
Breakdown of the ECM After Ear Punching
Ear hole punches were made in the ear pinnae. At various times after injury, ears were removed, fixed in 10% formalin, sectioned through the ear hole, and stained with hematoxylin and eosin. Because eosin binds to protein and is fluorescent, changes in protein levels and distribution in the ECM and basement membrane can be visualized.
The rate of ear hole closure in the MRL accelerates starting at approximately day 5 (Clark et al., 1998; Heber-Katz, 1999). As seen in Figure 1A–H, epifluorescence microscopy showed ECM and basement membrane buildup and breakdown that was different between the MRL and B6 healing ear. The amount of fluorescent material was greater on day 4 in the B6 (Fig. 1A) than the MRL (Fig. 1B) at this same time point, although in both cases, the basement membrane dividing the epidermis and dermis is clear. By day 5, much of the thick ECM fluorescence is gone in the B6 (Fig. 1C), although the basement membrane is still present. On the other hand, the basement membrane in the MRL can no longer be seen (Fig. 1D). On days 6 and 7, the B6 still shows a strong basement membrane (Fig. 1E,G), whereas the MRL has little or none (Fig. 1F,H). This is an intriguing result which might be accounted for by the action of proteases.
Protease Expression During Healing
Our initial studies examined RNA extracted from tissue around the circumference of the healing ear hole from the B6 and MRL mouse on days 0 and 5 after wounding. Microarray analysis (Clontech arrays) indicated that MMP-2 was up-regulated in the MRL healing ear hole tissue compared with the B6 on day 5 after injury and TIMP-3, an inhibitor of MMP-9, was lower in the MRL ear hole tissue compared with B6 on both days 0 and 5 (data not shown), suggesting a role for MMP-9 as well as for MMP-2. Because MMP-2 and MMP-9 have similar functions and have been shown to work in combination in many diverse systems (Mackay et al., 1990; Matsubara et al., 1991; Fini et al., 1995; Trengove et al., 1999; Itoh et al., 2002), we focused on both MMP-2 and MMP-9 expression during healing.
Enzymatic Activity by Using Zymography
Zymography is an electrophoretic technique that allows the separation of molecules by their molecular weights and their ability to break down given substrates that are incorporated into the running gel. This method requires the use of nondenaturing conditions for the display of enzymatic activity.
Protein extracted from the healing tissue of the ear holes on days 0, 1, 3, 5, and 7 after ear punching was examined for both MMP-2 and MMP-9 by using a gelatin (denatured collagen) substrate. Zymograms showed multiple enzymatic bands consistent with the pro-form (92–105 kDa) and the active form (87–89 kDa) of MMP-9 (Fig. 2A). On day 0, the pro-form of MMP-9 is seen in both strains. After injury, the MRL expressed more pro- and active form of MMP-9 than B6 at all time points.
Proteolytic bands for MMP-2 were seen on gelatin substrate gels at 72 kDa (pro-form) and 66–68 kDa (active form). Pro- and active forms of MMP-2 were seen on day 0 in both strains. After injury, MMP-2 activity was significantly less than MMP-9 activity in general; however, the MRL expressed a higher level of the active form of MMP-2. Relative levels of activity as determined by the amount of protein loaded, and the zymogram band density (see Fig. 2A) can be seen (Fig. 2D). Similar results were seen by using collagen type IV as a substrate (Fig. 2C), although only MMP-9–specific bands could be seen. Other bands were also seen on gelatin gels, and in all cases, MRL expressed higher levels than B6. Such bands were found at approximately 115 to 130 kDa (data not shown). With higher concentrations of protein run on a gelatin substrate, bands were also seen between 30 and 45 kDa (Fig. 2C).
Cellular Localization of Proteases
The same time points as those above were used to detect specific molecules in paraffin-embedded tissue sections. Antibodies shown to be specific for MMP-2 (Fig. 3A,B) and MMP-9 (data not shown) as well as antibodies to TIMP-2 (Fig. 3C,D) and TIMP-3 (data not shown) were used. Little staining in normal ear sections was seen. After wounding, stained cells were found near the wound site with a dramatic increase in cell number 1 to 3 days after wounding and then a continual decrease until day 7 after wounding (Figs. 3, 4). The staining was localized mainly to a population of small cells. As can be seen in Figure 4, the number of cells expressing MMP-2 was lower in the B6 than in the MRL at all time points, although the expression of TIMP-2 was nearly identical. MMP-9–positive cells were greater in number in the MRL, and TIMP-3–positive cells were greater in the B6.
The data in Figure 4 are re-expressed in Figure 5 to allow a direct comparison of the number of MMP- and TIMP-positive cells in each strain. In the B6 ears, the ratio of MMP-2– to TIMP-2–positive cells and MMP-9– to TIMP-3–positive cells was less than 1 from days 1 to 7 after wounding. On the other hand, in the MRL, the ratios of MMP-2– to TIMP-2–positive cells and MMP-9– to TIMP-3–positive cells were always greater than 1 from days 1 to 7. If these ratios (Table 1) represent functional TIMP and MMP activity in wounds, then this finding would indicate that the MRL should have significantly more active form of MMP compared with the B6 (see Fig. 2D). This would explain the more rapid breakdown of the ECM and basement membrane observed in the MRL.
Table 1. MMP/TIMP Ratios
No. of cells positive for MMP/no. of cells positive for TIMP
aMMP, matrix metelloproteinase; TIMP, tissue inhibitor of metelloproteinase.
Further examination of the cell types expressing these molecules (Fig. 6A,B) showed that the ratios shown above held true for small cells (5.6 microns ± 0.8; Fig. 3B, upper inset). Immunospecific double labeling of these cells found in the ear after wounding with anti–MMP-2 and anti–Ly-6G showed that approximately 80% of these cells were Ly-6G positive, a neutrophil-specific antibody (Fig. 7), supporting the notion that these were inflammatory cells and that almost all of the neutrophils were MMP-2 positive. This finding is consistent with the time of arrival, the specific morphology, and the localization in the wound.
A minority population of cells that were considered large cells (14.4 microns ± 2.1; Fig. 3B, lower inset) were shown to be approximately 30% mast cells by staining with toluidine blue, and approximately 70% were macrophages as stained with F4/80 (data not shown). However, the large cells made up only approximately 10% of the total cells stained. This large cell population was both greater in number in the MRL and was higher in MMP-2–positive cells and lower in TIMP-2–positive cells from days 5 to 7 compared with the B6.
In addition to infiltrating cells, anti–MMP-9 antibody stained keratinocytes in the epithelial cell layer as well as fibroblasts and the ECM (data not shown). MMP-2 expression was seen mainly in fibroblasts and ECM. MMP-2 (see Fig. 3B, arrows) and MMP-9 expression is highest at the apex of the closing wound and peaks on day 1, similar to the kinetic pattern seen with the neutrophil infiltrates.
RNA Expression Levels
RNA from healing ear hole tissue was extracted and real-time reverse transcription-polymerase chain reaction (RT-PCR) was carried out on these samples. MMP levels were normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) levels for each sample. MMP-9 expression levels follow that seen in protein expression levels in the zymograms and also that seen from the immunohistochemistry and the number of small cells appearing in the ear after injury (Fig. 8B). MRL levels are at most twofold higher than B6 levels. This finding correlates well with the neutrophils infiltrating the wound site that are synthesizing MMP-9.
On the other hand, MMP-2 RNA expression levels are highest in unwounded ears and especially in the MRL (Fig. 8A). This finding follows the results from the zymograms, which show the presence of active MMP-2 proteinase in uninjured ears. However, it does not follow the protein expression levels in the early inflammatory cells as determined by immunohistochemistry and suggests that these cells express protein before they arrive.
DISCUSSION AND CONCLUSION
Vertebrates exhibit several types of progressively more complex repair and regenerative processes: wound repair, tissue regeneration, and regeneration of entire organs (Gross, 1996; Stocum, 1995, 1996). Wound healing involves inflammation, re-epithelialization, matrix deposition, and tissue remodeling along with angiogenesis and the formation of granular tissue, and all of these processes require the concerted action of various proteolytic enzymes, including matrix metalloproteinases (MMP). It has been shown that MMPs play an important role in the breakdown of the provisional matrix formed after wounding by breaking down several components of the ECM, including type IV collagen, fibronectin, laminin, entactin, and elastin, as well as clarifying the cellular and fibril debris that surround the wound bed (Baker and Leaper, 2000). This helps keratinocytes and fibroblasts to migrate through the basement membrane as well. There is a significant amount of literature showing the role of MMPs in the breakdown of the ECM in regeneration models. They include the formation of the newt limb blastema (Grillo et al., 1968; Dresden and Gross, 1970; Miyazaki et al., 1996) as well as the axolotl limb regenerate (Yang and Bryant, 1994). MMPs have been shown to play a role in tail resorption during metamorphosis of tadpoles (Gross, 1966) and are regulated by thyroid hormone (Oofusa and Yoshizato, 1991; Damjanovski et al., 2000; Jung et al., 2002). After injury, planaria release MMPs, which are probably involved in remodeling and regrowth of their body parts (Sawada et al., 1999).
All enzymes of the MMP family have several common characteristics. They have a highly conserved zinc-binding region (sequence-HEXGHXXGLXH) that associates with Zn2+ upon activation and a TIMP-binding regulatory domain. Another domain is the N-terminus (PGCGVPD) pro-domain that is responsible for maintaining the enzyme in an inactive (zymogene) form. Among the family of zinc-containing proteinases, different classes can be distinguished according to their substrate specificity. All enzymes of the MMP family also contain a C-terminal 22-kDa domain that has structural homology to hemopexin. This domain is involved in the modulation of MMP-TIMP interaction and determines substrate specificity (Clark, 1996). MMP-2 and MMP-9 are the only two proteases that degrade gelatin (denatured collagen), basal lamina collagens (type IV, V, VII, X, XI), and the proteoglycan core protein. MMP-2 and -9, therefore, can breakdown the components of the basement membrane, including laminin, collagen type IV, and fibronectin (Yong et al., 1998).
The results presented here show first that both ECM and basement membrane breakdown in the healing ear hole precedes the formation of the blastema. This timing is true for the MRL mouse, which goes on to close its ear hole, to heal without scarring, and to regenerate lost cartilage (Clark et al., 1998; Heber-Katz, 1999). The B6 mouse ear, on the other hand, displays an intact basement membrane, does not develop a blastema, and shows scar formation without ear hole closure (Clark et al., 1998; Heber-Katz, 1999). Both the MRL and B6 mouse generate an inflammatory response after wounding with cells that are MMP-2– and MMP-9–positive. These same cells also express TIMP-2 and TIMP-3. However, the MRL shows more infiltrates and also more MMP-positive cells. Perhaps more important, however, is that the number of TIMP-positive cells is greater in the B6 than in the MRL, providing a situation in which inhibition of the MMPs is more likely in the B6. That this is in fact the case is suggested by the zymography of healing ear tissue showing that MMP-9 and to a lesser extent MMP-2 in the MRL is far more active. Why MMPs express at a higher level and the TIMPs at a lower level in the MRL is not clear but has been attributed to the function of both TNFα and IL-1 (Ito et al., 1990; Murphy et al., 1994). It has been reported that MRL macrophages are dysregulated in these two cytokine responses (Hartwell et al., 1995; Alleva et al., 1997).
The predominant population of MMP-positive cells that appear in the ear after wounding are small cells, which are mainly neutrophils and appear early after injury. These cells appear to be actively synthesizing MMP-9 but not MMP-2 as seen in the RT-PCR results (Fig. 8), although there is clearly MMP-2 protein in these cells as shown by double-labeling experiments (see Fig. 7). This finding is unusual, as it is generally thought that neutrophils are not MMP-2–positive. Furthermore, these cells also make TIMPs.
Because TIMPs are inhibitors of MMPs and both groups of molecules are secreted into the ECM, the MMP to TIMP ratio as measured in the extracellular wound bed (Ladwig et al., 2002) has been considered a proxy for MMP effectiveness. We analyzed our data similarly, but instead of examining secreted molecules, we analyzed immunohistochemically positive cells, and specifically neutrophils. This analysis was consistent with the level of active MMPs found through zymography. Furthermore, this supports the cell surface activation of MMPs and their interaction with TIMPs (Brooks et al., 1996; Butler et al., 1998; Steffensen et al., 1998; Miyamori et al., 2001).
Recent data from our laboratory shows that not only is ECM breakdown important in the regenerative response in the MRL ear but also in the MRL heart (Leferovich et al., 2001). Thus, hydroxyproline as a measure of scar formation is increased in the B6 heart where the scar response is predominant and is decreased in the MRL heart where cardiomyocyte proliferation and little scarring is seen. In a third model system, lack of scar formation correlates with axonal regeneration in a spinal cord injury model in B6 mice (Seitz et al., 2002). In the case of the ear hole, we propose that the MRL healing response is similar to what is seen in the amphibian limb. Initial events after wounding resemble wound repair such as the formation of the provisional matrix and epithelial migration over the wound site and is similar in the MRL and B6 mouse (Clark, 1996) as well as the amphibian (Repesh and Oberpriller, 1978, 1980; Stocum, 1995). After this, differences are seen. In the mouse, a basement membrane is formed, whereas in the amphibian, it is not (Repesh and Oberpriller, 1980). However, the data shown here suggest that the basement membrane at approximately 5 days after injury is selectively broken down in the MRL mouse by the two basement membrane–degrading MMPs, MMP-2 and -9 (Mackay et al., 1990). This process should then allow for epithelial–mesenchymal interactions leading to the growth of the blastema as seen in the amphibian limb (Stocum and Dearlove, 1972; Globus et al., 1980; Stocum and Crawford, 1987; Brockes, 1997; Christensen et al., 2002) as well as during development (Sun et al., 2002).
Finally, a genetic analysis of the wound healing trait described in this study has shown that ear hole closure is a highly complex trait with approximately 20 loci currently described (McBrearty et al., 1998; Masinde et al., 2001; Blankenhorn et al., in press; Heber-Katz et al., manuscript submitted for publication). It is of interest that both MMP-2 (on chromosome 8 at 42 cM) and TIMP-2 (on chromosome 11 at 72 cM) are both candidate genes for this wound healing phenotype.
The MRL/MpJ mice were obtained from the Jackson Laboratories (Bar Harbor, ME) and C57BL/6 (B6) were obtained from Taconic Laboratories (Germantown, NY). Animals were maintained under standard conditions at the Wistar Institute Animal Facility (Philadelphia, PA).
Mouse ears were cleaned with isopropyl alcohol and punched with a 2.1-mm-diameter surgical punch. Punches were made in each ear on the midline. On days 1, 3, 5, and 7 after wounding, the entire area encompassing the healing wounds was excised. For protein sample extraction to be used in the zymography analysis, wound tissues were immediately frozen in liquid nitrogen. For the histopathologic examination, the samples were incubated overnight in Prefer fixative (Anatech, Ltd., Battle Creek, MI) and embedded in paraffin, and 5-micron-thick sections were cut. Samples were collected from at least three separate animals for each time point/strain.
For general pathologic analysis, the Prefer-fixed and paraffin-embedded ear sections were stained with hematoxylin (Surgipath, catalog no. C.I. 75290) and eosin (Surgipath, catalog no. C.I. 45380) with a 5-min and 1-min exposure, respectively. For the staining of mast cells, a 0.5% (pH 5) solution of toluidine blue-O (Hartman-Leddon, catalog no. 364) was used for 20 min followed by a 10-min wash with H2O.
For immunohistologic staining, we used rabbit anti-human antibodies to MMP-2 (Sigma, catalog no. M6302), MMP-9 (Sigma, catalog no. M5302), TIMP-2 (Sigma, catalog no. T8062), and TIMP-3 (Sigma, catalog no. T7812). As a secondary antibody, we used a biotinylated goat anti-rabbit antibody (Jackson ImmunoResearch, catalog no. 111-065-144). For labeling neutrophils, a rat anti-mouse antibody against Ly-6G and Ly-6C was used (Ly-6G, rat anti-mouse, PharMingen, catalog no. 553126).
For nonfluorescent labeling, the sections were exposed to ABC by using the Vectastain Elite ABC Kit (Vector, catalog no. PK-6100) in PBST (0.1 M sodium phosphate buffer [pH 7.4], 0.005% Triton X-100, 0.9% NaCl) for 15 min, washed four times with PBST (5 min) and developed with diaminobenzidine (DAB) using the DAB kit (Vector, catalog no. SK-4100).
For fluorescent double labeling, donkey anti-rabbit, CyTM3-conjugated (Jackson ImmunoResearch, catalog no. 711-165-152) was used for anti–MMP-2 and goat anti-rat, fluorescein isothiocyanate–conjugated (Santa Cruz, catalog no. 2011) was used for anti–Ly-6G.
A standard Lowry procedure was used with detection by ultraviolet light at 750 nm. The concentration of total protein was determined by using an Ultrospec 3000 spectrophotometer (Pharmacia Biotech) using reagents from Bio-Rad Laboratories-DC Protein Assay Kit (catalog no. 500-0116).
RNA was isolated by homogenizing the ear tissue in Trizol (Gibco BRL) following the manufacturer's instructions. RNA was treated briefly with DNase to remove any DNA that copurified. RNA concentration was calculated by OD260 readings and samples pooled (1 μg total RNA/per animal). RNA was reverse transcribed in a 20-μl of cocktail that included 1 U of RNaseIN, 0.2 μM dNTPs, 1 μM oligo (dT), and 200 U of SuperScript reverse transcriptase (Gibco-BRL, catalog no. 18053-017) in (1×) RT buffer. Control reactions omitted the RT enzyme. Quantitative real-time PCR was carried out on a Lightcycler (Roche) to determine the amount of MMP-2 message (for primer sequences, see Table 2, line 1) and MMP-9 message (see Table 2, line 2) (Wittwer et al., 1997). The reaction was set up by using the Fast-start DNA Master Syber Green I Kit (Roche, catalog no. 3003230) according to instructions in a 15-μl cocktail (2 mM). The samples were then normalized to GAPDH message (see Table 2, line 3). A period of 10 min at 95°C was used to denature and inactivate the Taq-start antibody. Cycling was carried out 45 times by using 95°C for 10 sec, 61°C for 10 sec, 72°C for 20 sec, and the acquisition temperature was set at 85°C for 10 sec. The readout is based on the fluorescence of intercalated SYBER Green I dye.
Protease activity was determined through zymography using different substrate casting gels adapted from published methods (Hamaguchi et al., 1995). Briefly, healing tissue from around the circumference of the ear holes was collected, milled in a glass-to-glass grinder and liquid nitrogen, and the homogenized tissue mixed with proteinase inhibitor cocktail (Complete from Boehringer Mannheim) in 1× PBST (1 μg of tissue dissolved in 15 μl of Complete cocktail). The sample extraction continued on the rocker platform (4°C) for 20 hr, after which the sample was clarified by centrifugation. After sedimentation, samples were normalized by the Lowry procedure. Those samples were subjected to electrophoresis with copolymerized 10% sodium dodecyl sulfate–polyacrylamide gels with 1 mg/ml gelatin (from swine skin; Sigma, catalog no. G-2500) and 0.5 mg/ml of collagen type IV (from human placenta; Sigma, catalog no. C-5533). The gels were washed with H2O for 1 hr and incubated in development buffer (50 μM Tris OH pH 7.5, 0.2 M NaCl, 10 μM CaCl2) for 20 hr at 37°C. Gel staining was performed by using Coomassie blue for at least 3 hr and briefly destained in destaining solution (50 ml of methanol, 40 ml of H2O, and 10 ml of acetic acid). Zymography data are shown as color inverted for better visualization.
We thank the F.M. Kirby Foundation and The G. Harold and Leila Y. Mathers Foundation for their very generous support. E.H.K. and L.C. received funding from the NIH, and D.G. received funding from an NIH training grant.