INTRODUCTION

Graft-versus-host disease is the major cause of morbidity and mortality after bone marrow transplantation and is sometimes, though more rarely, threatening patients after liver transplantation (1, 2). In graft-versus-host disease, immunocompetent donor lymphocytes from transplanted bone marrow or liver recognize and react against recipient's tissues, leading to cutaneous and intestinal lesions. Commonly, the first manifestation of acute graft-versus-host disease is pruritic maculopapular skin rash characterized by typical histopathological findings such as vacuolar degeneration of the basal cell layer, basal cell necrosis, Langerhans cell depletion, intercellular edema, and acantholysis (1, 2). Intestinal manifestations of acute graft-versus-host disease include diarrhea, abdominal cramping, intestinal bleeding, and vomiting. Endoscopic findings range from normal to extensive edema, mucosal sloughing and bleeding, and histology reveals, for example, crypt-cell necrosis (1, 2).

Matrix metalloproteinases (MMPs) are a family of Ca2+-dependent zinc-ion containing endopeptidases. Since the early sixties, 21 human members of this enzyme family have been cloned. According to their structure and substrate specificity, MMPs can be grouped into collagenases (MMP-1, -8, and 13), gelatinases (MMP-2 and -9), stromelysins (MMP-3, -10, -11, and -12), membrane-type metalloproteinases (MMP-14, -15, -16, -17, and -24), matrilysins (MMP-7 and -26), and other MMPs (MMP-19, -20, -23, -25, and -28; 3, 4). MMPs are regulated mainly at the transcriptional level, and their enzymatic activity can be inhibited by tissue inhibitors of metalloproteinases (TIMP), a protein group that currently comprises four members (TIMP-1, -2, -3, and -4). TIMP-1 inhibits the activity of most MMPs except MMP-2 and MT1-MMP (5, 6). TIMP-3 inhibits the activity of MMPs -1, -2, -3, -9, and -13 (7). In addition, it inhibits the endopeptidase TACE (ADAM-17) (8), which can shed TNFR1 and TNF-α from the cell surface. MMPs play an important role in several physiological and pathological processes such as fetal development; wound healing; inflammatory bowel disease; rheumatoid arthritis; and cancer growth, invasion, and metastasis (3).

Matrix metalloproteinases and TIMPs have previously been implicated in the pathomechanism of several inflammatory disorders of the skin and intestine, such as celiac disease, ulcerative colitis, and Crohn's disease (9, 10), as well as granulomatous skin disorders, lichen planus, and scleroderma (11, 12, 13, 14). Several studies have shown the expression of various MMPs, particularly MMPs -1, -3, -12 and -14, in intestinal ulcers of patients with inflammatory bowel disease (9, 10, 15, 16). Functional evidence for the critical role of MMPs, particularly that of MMP-3, in mucosal injury comes from studies using a fetal model for T-cell–mediated injury of the gut (17, 18). In this model, the synthetic MMP inhibitor prevents mucosal injury after T cell activation.

The synthetic MMP inhibitor (KB-R7785), with activity against MMPs-1, -3, and -9, prevents acute graft-versus-host disease but preserves graft-versus-leukemia effect in mice undergoing allogeneic bone marrow transplantation (19, 20). TNF-α and FasL, both released and activated by MMPs (3, 21), are implicated in the pathogenesis of acute graft-versus-host disease (22). In addition to the effects of MMPs on the activation of cytokines or on the shedding of their receptors (23), they might function as end-stage effector molecules in intestinal and cutaneous lesions also in graft-versus-host disease. Therefore, we wanted to investigate whether MMPs 1, 3, 7, 9, 10, 12, and 19 and TIMPs 1 and 3 are expressed in graft-versus-host disease–related lesions of the skin and intestine. Our results indicate that TIMP-3 might contribute to basal cell apoptosis in cutaneous lesions, whereas none of the MMPs was overexpressed. In contrast, expression of MMPs-1, and -19 and TIMPs 1 and 3 was observed in intestinal graft-versus-host lesions.

MATERIALS AND METHODS

Samples

Formalin-fixed, paraffin-embedded specimens of intestine (n = 13) and skin (n = 9) from 16 patients were obtained from the Department of Pathology, Helsinki University Central Hospital. Intestinal and skin specimens were from different patients. Fifteen patients had undergone bone marrow transplantation and one patient, liver transplantation. Each sample was investigated by an experienced pathologist, and the diagnosis of cutaneous graft-versus-host disease was verified by a dermatopathologist. Cutaneous and intestinal samples were graded (Tables 1 and 2) according to previously published criteria (24, 25).

TABLE 1 Clinicopathological Information and In Situ Hybridization or Immunohistochemistry Results for Patients with Cutaneous GVHD after Bone Marrow Transplantation
Full size table
TABLE 2 Clinocopathological Information and In Situ Hybridization or Immunohistochemistry Results for Patients with Intestinal GVHD after Bone Marrow Transplantation
Full size table

In Situ Hybridization

The production and specificity of the anti-sense human MMP-1, -3, -7, -10, and -12 and TIMPs-1 and -3 cRNA probes have been described (10, 26, 27, 28, 29). After treatment with proteinase K, 4-μm sections were hybridized with 35S-labeled cRNA probes overnight in a humidified chamber at 50–55° C, washed under stringent conditions, and treated with RNAse A to minimize nonspecific signal as described (11, 30). Photographic emulsion was developed after 20–45 days of autoradiography, and the slides were stained with hematoxylin and eosin. Samples previously known positive for each probe were used as positive controls: ulcerative colitis for MMPs 1, 3, 7, and 12; wounds for MMP-10 and for TIMPs 1 and 3. Each tissue was stained twice in separate experiments and analyzed by two investigators independently.

To detect fragmented DNA of apoptotic cells in the paraffin sections, a TUNEL (terminal deoxytransferase–mediated digoxin-dUTP nick end labeling) in situ technique (Apoptag; Oncor Inc., Gaithersburg, MD) was used as described elsewhere (31).

Immunohistochemistry

Immunostainings were performed by the avidin–biotin–peroxidase complex technique (Vectastain ABC kit; Vector Laboratories, Burlingame, CA). Antibodies were incubated on tissue sections overnight at +4° C. Diaminobenzidine was used as chromogenic substrate and Harris hematoxylin as counterstain, as described previously in detail (32). For TIMP-3, sections were pretreated by antigen retrieval in 95° C for 25 minutes (DAKO, antigen retrieval solution), and for Apoptag and MMP-9, sections were pretreated with 10 mg/mL trypsin. MMP-19 and TGF-β stainings did not require pretreatment. Antibody dilutions were as follows: for anti-TIMP-3 (IM43T, Oncogene, Boston, MA), 1:1500; for anti-MMP-19 (RDI-MMP19abR, Research Diagnostics Inc., Flanders, NJ), 1:70; for TGF-β (MAB1032, Chemicon), 1:500; and for anti-MMP-9 (GE213, Diabor, Oulu, Finland), 1:400.

RESULTS

Skin

TIMP-3 mRNA was detected in 7/7 samples (Table 1). In situ hybridization signal was strong in lesional keratinocytes (Fig. 1A) and appeared to concentrate in areas with vacuolizing and apoptotic keratinocytes in the basal cell layer (Fig. 1A–B). There was no evidence of vacuolization in the basal cell layer when TIMP-3 was absent (Fig. 1A–B). TIMP-3 protein co-localized with corresponding mRNA (Fig. 1D–E). Because TIMP-3 has recently been implicated in apoptosis (33, 34), the Apoptag kit was used to identify apoptotic keratinocytes: large amounts of apoptotic cells were detected in affected epidermis (Fig. 1C). Regions with apoptotic keratinocytes were also positive for TIMP-3 mRNA (Fig. 1A–C). TIMP-3 was expressed also in the dermis by vascular endothelial cells and perivascular mononuclear cells (Fig. 1D). There was a tendency for elevated stromal expression of TIMP-3 with increasing grade (Table 1). Keratinocytes of normal intact epidermis do not express TIMP-3 (29, 35).

FIGURE 1
figure 1

Expression of TIMPs 1 and 3 in cutaneous lesions of graft-versus-host disease. A, dark-field image of TIMP-3 mRNA–positive lesional keratinocytes. B, corresponding bright-field image. Arrowheads depict the vacuolizing basal cell layer. Arrow depicts intact epidermis. The line points out the area between affected (on right-hand side) and unaffected epidermis (on left-hand side). C, apoptotic keratinocytes with fragmented DNA identified by TUNEL in situ technique in a serial section. D, dark-field image of TIMP-3 mRNA expression in another sample of cutaneous graft-versus-host disease. E, immunohistochemical staining of a nearby section, showing TIMP-3 protein. F, immunohistochemical staining on a serial section to E using antibodies against TGF-β1 protein. Arrows depict corresponding spots in D–F. G, TIMP-1 mRNA expression in a cutaneous lesion. H, higher magnification of stromal TIMP-1–expressing cells. Scale bars: 50 μm (D–G); 25 μm (A–C, H).

Full size image

TGF-β is known to induce TIMP-3 expression in primary keratinocytes (29). Furthermore, TGF-β antibodies prevented lung and skin fibrosis in a murine model for sclerodermatous graft-versus-host disease (36). Thus, we performed TGF-β immunostainings on nearby sections for those used for TIMP-3 in situ hybridization experiments. TGF-β–expressing keratinocytes were found in TIMP-3 positive areas (Fig. 1D, F).

TIMP-1 mRNA was detected in 5/7 samples. In contrast to TIMP-3, which was abundantly expressed by epidermal keratinocytes, TIMP-1 was expressed mainly by dermal endothelial cells and perivascular mononuclear cells (Fig. 1G–H) beneath epidermis affected by graft-versus-host disease. Epidermal expression was detected in a few cells in 2/7 cases.

Signal for MMPs -1, -3, -9, -10, or -12 was not seen in any sample (Table 1). Sweat glands were positive for MMP-7 in most cases (data not shown), which is in accordance with our previous data (37). Positive staining for MMP-19 was found in occasional fibroblasts and endothelial cells in 4/9 samples (Table 1), in accordance with previous studies (38). MMP-19 was not detected in areas of keratinocyte apoptosis or vacuolar degeneration (data not shown).

Intestine

Expression of MMPs and TIMPs was studied also in gut affected by graft-versus-host disease. The only MMPs detected in most cases were MMPs-1 and -19 (Table 2). Both TIMP-1 and -3 were abundantly expressed by large amounts of stromal cells.

Expression of MMP-1 mRNA was detected in 9/13 samples. Epithelial expression of MMP-1 was detected in 7/13 samples. In addition to epithelial cells, MMP-1–positive cells included stromal plump, activated fibroblast/macrophage-like cells around destructed crypts (data not shown). In some cases, shedding epithelium expressed MMP-1 mRNA (Fig. 2A–B).

FIGURE 2
figure 2

Expression of collagenase-1 and TIMPs 1 and 3 in intestinal lesions of graft-versus-host disease. A, expression of MMP-1 mRNA by shedding epithelial cells in an intestinal lesion. B, higher magnification showing MMP-1–positive cells. C, MMP-19 is expressed in epithelial cells of crypts. D, higher magnification showing stromal macrophage/activated fibroblast–like cells positive for MMP-19. E, TIMP-1 expression in subepithelial cells after the detachment of epithelium. F, higher magnification showing TIMP-1–positive cells. G, TIMP-3 expression by stromal cells in a serial section. H, higher magnification showing TIMP-3–positive cells. Scale bars: 50 μm (C); 25 μm (A, E, G); 12.5 μm (B, D, F, H).

Full size image

The other MMP found in intestinal lesions was MMP-19. Positive signal was found in 10/13 samples. Cells expressing MMP-19 were mainly epithelial cells located in the crypts (Fig. 2C), but also occasional macrophage/activated fibroblast-like cells in the stroma (Fig. 2D). MMP-19 expression did not associate with apoptosis as judged by TUNEL staining (data not shown). In normal intestine, epithelial cells in the crypts were negative for MMP-19 (data not shown).

In contrast to previous data on inflammatory bowel disease (10, 15, 16), only occasional cells expressed mRNA for MMP–3 in 5/13 samples (Table 2). In one sample, the number of MMP-3–positive cells was comparable to the number of MMP-1–expressing cells. MMP-3–positive cells had the appearance of fibroblast/macrophage-like cells and were located around destructed crypts.

MMPs 7 and 10, previously associated with enterocyte migration (10, 15), were found in 5/13 and 1/13 samples, respectively (Table 2). Surprisingly, we did not detect MMP-12 mRNA in any biopsy from graft-versus-host disease affected intestine. Immunohistochemistry revealed MMP-9 in occasional stromal cells in 5/13 samples (data not shown).

Signal for TIMP-1 and TIMP-3 was detected in 12/13 and 10/13 cases, respectively. TIMP-1 was expressed mainly in superficial regions of injured mucosa (Fig. 2E–F), whereas TIMP-3 expression concentrated in deeper parts of mucosa (Fig. 2G–H), particularly in stromal areas with diminished crypt density. TIMP-1 expression concentrated in surface parts of mucosa and was particularly strong in subepithelial cells at sites where epithelium had been shed away. Both TIMP-1 and -3 were expressed by fibroblast/macrophage-like cells, whereas epithelial cells remained negative, which is in accordance with our previous data (10, 15). The number of cells expressing TIMP-1 exceeded the number of TIMP-3 expressing cells in all cases.

DISCUSSION

Experimental studies with mice and rats have demonstrated the crucial role for TNF-α and Fas-FasL in the pathogenesis of acute graft-versus-host disease (22). Because MMP inhibitor prevented lethal acute graft-versus-host disease in mice (19, 20), we hypothesized that in addition to the cytokine-modulating effect of MMPs, they might also be lesionally expressed and contribute to destruction on the tissue level. In this work, we studied the expression of MMPs and TIMPs previously shown to be abundantly expressed in intestinal ulcers related to inflammatory bowel disease. We did not detect marked up-regulation of any MMP studied in stromal cells of lesional gut, but MMP-19 was expressed in the crypt epithelium. In situ hybridization or immunohistochemistry did not reveal MMP-1, -3, -7, -9, -10, or -12 expression in skin lesions. However, TIMPs 1 and 3 were up-regulated in both intestinal and skin lesions, suggesting that altered MMP–TIMP balance would favor matrix synthesis rather than breakdown.

In cutaneous lesions, we did not detect mRNA for any MMP studied, but TIMPs-1 and -3, particularly TIMP-3 mRNA and protein, were clearly up-regulated (Table 1; Fig. 1). Interestingly, vacuolar degeneration, a typical histopathological finding for graft-versus-host disease, was detected in epidermal areas positive for TIMP-3 (Fig. 1), whereas in TIMP-3–negative areas, these findings were absent. TIMP-3 has been implicated in apoptosis (33, 34), and our TUNEL staining demonstrated large numbers of apoptotic keratinocytes in areas affected with graft-versus-host disease. Therefore, TIMP-3 might function as a proapoptotic agent in this disease. To our knowledge, there are no data on TIMP-3 in other cutaneous disorders associated with apoptosis, such as lichen planus or erythema multiforme. The finding of TIMP-3 up-regulation in the stroma of high-grade and late graft-versus-host disease is in accordance with our earlier results that TIMP-3 is also overexpressed in other skin disorders associated with severe dermal fibrosis, such as scleroderma (12).

Antibodies for TGF-β prevented the development of skin and lung fibrosis in a murine model of sclerodermatous graft-versus-host disease (36). The same cytokine induces expression of TIMP-3 in keratinocytes (29), whereas TNF-α has no effect. As TGF-β protein was produced by keratinocytes in TIMP-3 positive areas in cutaneous graft-versus-host disease, TIMP-3 might be up-regulated by TGF-β.

TIMP-3 was expressed by epidermal keratinocytes, but also by dermal endothelial cells, analogously to our previous data on endothelial TIMP-3 expression in the neoangiogenic vasculature of intestinal ulcers (10). TIMP-1 was expressed only by dermal macrophage/fibroblast-like cells. Lack of MMP expression and up-regulation of dermal MMP inhibitors may lead to the accumulation of extracellular matrix and fibrosis, both important histological findings in chronic cutaneous graft-versus-host disease. In fact, we have previously demonstrated TGF-β–mediated activation of TIMP-3 in sclerodermatous skin fibroblasts in culture (12).

MMP-1 and MMP-3 are abundantly expressed in inflammatory bowel disease–related intestinal ulcers and duodenal epithelium of patients with celiac disease (10, 15, 16, 39). MMPs have been implicated in the degradation of mucosal extracellular matrix that leads to ulcer formation. In this study, we found only occasional MMP-3–expressing cells, whereas MMP-1 was expressed in the majority of intestinal biopsies. MMP-1 expression was detected in occasional epithelial cells in regions with epithelial injury, but extensive stromal expression, comparable to that in inflammatory bowel disease, was absent. Epithelial MMP-1 expression might be related to epithelial shedding. In an experimental inflammatory bowel disease model, the gut injury has been found to depend on MMP-3 production (17, 40). However, our current results indicate that in graft-versus-host disease, MMPs do not function in tissue destruction, or at least destructive metalloenzymes differ from those considered to be important in inflammatory bowel disease.

Both TIMPs were abundant in intestinal biopsies. TIMP-1 expression concentrated in superficial regions, particularly in areas where epithelium was absent, giving the impression that lack of epithelial cells induces TIMP-1 expression by subepithelial cells. TIMPs inhibit the enzymatic activity of several MMPs and thereby protect extracellular matrix from degradation. Thus, TIMP-1 might stabilize newly synthesized basement membrane and contribute to the wound-healing process by helping to maintain an optimal environment for migrating epithelial cells. TIMP-3 expression was detected in deeper parts of the mucosa, particularly in areas with total crypt destruction, possibly participating in the scarring process. In normal intestine, only occasional cells expressing TIMP-3 have been reported (10).

Because the expression of MMPs-7 and -10 seem to be restricted to wound healing, epithelial cell migration, and malignancies of the gastrointestinal tract (10, 15, 41), it is not surprising that we did not detect mRNA for these MMPs in gut affected by graft-versus-host disease. MMP-12 is usually produced by macrophages and overexpressed by duodenal subepithelial macrophages in celiac disease as well as in ulcers of patients with inflammatory bowel disease (9, 10, 15). In this study, we failed to detect mRNA for MMP-12 in any intestinal or cutaneous sample, although our positive controls worked. Obviously, graft-versus-host disease leads to an inflammatory reaction of the gut mucosa, including the activation of macrophages. However, after bone marrow transplantation, the patients are strongly immunosuppressed, leading to a dysfunction of the immunological defense system, including macrophages. This might explain the lack of MMP-12 expression in inflammatory reaction seen in intestinal graft-versus-host disease.

Our results indicate that the MMP expression profile of intestinal graft-versus-host disease differs from that of other T-cell–mediated destructive conditions, namely inflammatory bowel disease and celiac disease. Whereas MMP expression was differentially regulated in lesions of skin and gut graft-versus-host disease, TIMPs were up-regulated in both organs. TIMP-3, in particular, may contribute to keratinocyte apoptosis.