Abstract
We used a specific monoclonal antibody to human hepatocyte growth factor activator inhibitor type 1 (HAI-1) in immunohistochemical procedures to determine the distribution and localization of HAI-1 in human tissues. In normal adult tissues, HAI-1 was predominantly expressed in the simple columnar epithelium of the ducts, tubules, and mucosal surface of various organs. In all cases, HAI-1 was localized predominantly on the cellular lateral (or basolateral) surface. By contrast, hepatocytes, acinar cells, endocrine cells, stromal mesenchymal cells, and inflammatory cells were hardly stainable with the antibody, and stratified squamous epithelium showed only faint immunoreactivity on the surface of cells of the basal layer. In the gastrointestinal tract, the surface epithelium was strongly stained. RNA blot analysis confirmed the presence of specific mRNA transcript in the gastrointestinal mucosa, and in situ hybridization revealed that HAI-1 mRNA showed a similar cellular distribution pattern. Although HAI-1 was not expressed in normal hepatocytes, strong immunoreactivity was observed on the epithelium of pseudo-bile ducts and on the surface of scattered hepatocytes in fulminant hepatitis. The enhanced expression was also noted in regenerating tubule epithelial cells of the kidney after infarction. We conclude that HAI-1 is preferentially expressed in the simple columnar epithelium of the mucosal surface and duct, that the predominant localization of HAI-1 is the cell surface, and that the expression of HAI-1 can be modulated by tissue injury and regeneration.
H
HGFA is a blood coagulation factor XIIa-like serine proteinase and is secreted by the liver as an inactive zymogen, circulating in the blood in this form (Miyazawa et al. 1993,1996; Shimomura et al. 1995). The zymogen can be activated by limited proteolysis brought about by thrombin in an injured tissue, and the activity of HGFA is not inhibited by major plasma proteinase inhibitors (Shimomura et al. 1993,1995; Miyazawa et al. 1996). Recently, two novel Kunitztype serine proteinase inhibitors with efficient inhibitory activity against HGFA were purified from the serum-free cultured conditioned medium of a human gastric carcinoma cell line MKN-45 and their cDNAs were cloned. These inhibitors were designated as HGF activator inhibitor Type 1 and Type 2 (HAI-1 and HAI-2) (Kawaguchi et al. 1997; Shimomura et al. 1997). At the same time, HAI-2 was also independently discovered by two other groups as a novel placental bikunin and as a protein overexpressed in pancreatic cancer (Marlor et al. 1997; Müller-Pillasch et al. 1998).
Mature HAI-1 has two well-defined Kunitz domains. The first domain appears to be responsible for the inhibition of HGFA (Shimomura et al. 1997). Interestingly, this inhibitor has a presumed transmembrane domain in the C-terminal region, although precise subcellular localization of this inhibitor remains to be determined. Because the molecular weight of mature HAI-1 extracted from MKN-45 cells was around 66 kD and those of the secreted forms were 58 kD and 40 kD, the secreted HAI-1 proteins appeared to be processing products probably cleaved at the C-terminal region (Shimomura et al. 1997; and unpublished observations). Previous RNA blot analysis indicated that notable HAI-1 mRNA signals were present in multiple organs such as the placenta, kidneys, prostate, pancreas, small intestine, and colon, but were hardly detectable in the liver, lungs, skeletal muscle, and brain (Shimomura et al. 1997; Kataoka et al. 1998). However, the distribution and cellular localization of this protein in tissues expressing the corresponding mRNA are still undetermined, and its physiological roles in vivo are uncertain at present. Because activation of proHGF/SF in the extracellular environment is a crucial limiting step of the HGF/SF signaling pathway and because proteolytic activation of HGF/SF by HGFA is observed in injured tissue (Miyazawa et al. 1994,1996), HAI-1 may have an important function in regulating regeneration induced by HGF/SF in injured tissue. The principal aim of the present study was to evaluate the distribution and localization of HAI-1 protein in human normal tissues to help to construct a hypothesis about its function. To some extent, this work provides evidence for the possible involvement of HAI-1 protein in epithelial regeneration that occurs in response to tissue injury.
Materials and Methods
Monoclonal Antibody
A Balb/c mouse was immunized with purified HAI-1 protein prepared from conditioned medium of MKN-45 cells as described (Shimomura et al. 1997). Monoclonal antibodies (MAbs) were generated according to established procedures, and the resulting hybridoma supernatants were screened by enzyme immunoassay. Briefly, mouse spleen cells were removed and fused with P3U1 myeloma. Hybridomas were screened for secretion of antibodies against HAI-1, and several hybridomas were obtained. Among the antibodies obtained, 1N7 (IgG1k was suitable to stain formalin-fixed, paraffin-embedded tissue specimen and was used for this study. The epitope of this antibody was around the second Kunitz domain of HAI-1 protein (Shimomura et al. 1997; and unpublished observations).
Immunohistochemistry
All human tissues were obtained from surgical specimens or at autopsy. Formalin-fixed, paraffin-embedded tissue specimens were prepared according to the routine procedure. Sections were dewaxed in xylene and rehydrated in decreasing ethanol solutions and water. After antigen retrieval (5 min of autoclaving in 10 mM citrate buffer, pH 6.0), the sections were treated with 3% H2O2 in PBS for 10 min and washed in PBS twice, followed by blocking in 3% bovine serum albumin (BSA) in PBS for 1 hr at room temperature (RT). Then the sections were incubated with the primary antibody (1N7; 10 μg/ml of PBS containing 1% BSA) for 16 hr at 4C. Negative controls consisted of omission of the primary antibody, and as positive control a colon mucosal tissue in which HAI-1 mRNA expression was confirmed previously was used (Kataoka et al. 1998). For the adsorption test, the antibody was pretreated with a 10-fold excess of recombinant HAI-1 protein. The sections were then washed in PBS and incubated with Envision-labeled polymer reagent (DAKO; Carpinteria, CA) for 45 min at 37C. The reaction was revealed with nickel, cobalt-3,3′-diaminobenzidine (ImmunoPure Metal Enhanced DAB Substrate Kit; Pierce, Rockford, IL) and counterstained with Mayer's hematoxylin. Immunoreaction stronger than the positive control was judged as strongly positive (++) and that comparable with the positive control was judged as positive (+).
Immunoelectron Microscopy
Human stomach surgical specimens fixed with 4% paraformaldehyde in 0.1 M phosphate buffer were carried through 5%, 10%, 15%, and 20% sucrose in PBS for 2 hr each and finally immersed in 20% sucrose containing 50% OCT compound (Lab Tek; Naperville, IL) in PBS overnight, and then rapidly frozen in liquid nitrogen. The frozen tissues were sectioned at 6 μm with a cryostat and dried on glass slides. Before the immunostaining, we treated sections with 50 mM Tris-HCl buffer, pH 9.0 or pH 10.0, for 48 hr at RT. We confirmed that pretreatment with a higher pH solution influenced the degree of unmasking of the epitope (Shi et al. 1995). After washing with PBS, the sections were sequentially treated with 50 mM ammonium chloride in PBS for 30 min, rinsed in PBS containing 10% normal goat serum (NGS) and 1% BSA (NGS-BSA-PBS) for 1 hr, and then incubated with MAb 1N7 (20 μg/ml of NGS-BSA-PBS) for 24 hr at 4C. After rinsing with PBS, the sections were immersed in horseradish peroxidase (HRP)-conjugated anti-mouse IgG goat Fab' (1:50) (MBL; Nagoya, Japan) in NGS-BSA-PBS overnight at 4C. After rinsing with PBS, the sections were fixed in 1% glutaraldehyde in PBS for 20 min. After rinsing with PBS, the sections were treated with Graham-Karnovsky medium (Graham and Karnovsky 1966) without H2O2 for 30 min and then with the complete medium containing 0.03% H2O2 for 20 min at RT. After rinsing with PBS, the sections were postfixed with 1% OsO4 in 0.1 M cacodylate buffer, pH 7.4, containing 1% potassium ferrocyanide (Brown and Farquhar 1989), dehydrated through a graded series of ethanol, and then embedded in Epon.
Immunoblot Analysis
Subconfluent MKN-28 gastric adenocarcinoma and CCK-81 colon adenocarcinoma cells on 100-mm dishes were washed three times with PBS and immediately scraped into 2 ml of 10% trichroloacetic acid on ice. MKN-28 and CCK-81 were obtained from IBL (Fujioka, Japan) and Japanese Cancer Research Resources Bank, respectively. The precipitated proteins were harvested by centrifugation (14,000 rpm, 3 min) and the pellet was extracted with 200 μl of 7 M urea/2% Triton X-100/5% 2-mercaptoethanol, followed by centrifugation (14,000 rpm, 3 min). To prepare a tissue extract specimen, fresh human nonneoplastic gastric mucosa (0.16 g), which was obtained from the surgically resected stomach of a gastric cancer patient, was immediately frozen in liquid nitrogen, crushed, and homogenized in an extraction buffer containing 50 mM Tris-HCl (pH 7.5)/150 mM NaCl/5 mM EDTA/2 mM phenylmethyl sulfonyl fluoride/5 μg/ml leupeptin/2% Triton X-100/2% Nonidet P-40, followed by centrifugation (14,000 rpm, 3 min). The resultant supernatants were mixed with SDS sample buffer and boiled for 3 min. SDS-PAGE was performed under reducing conditions using a 4-12% gradient gel. After electrophoresis, the proteins were transferred electrophoretically onto Immobilon membrane (Millipore; Bedford, MA). After blocking the nonspecific binding sites with 5% nonfat dry milk in 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.05% Tween 20 (TBS-T), the membrane was incubated with MAb 1N7 (1 μg/ml in TBS-T containing 1% BSA) at 4C overnight, followed by washing in TBS-T four times and incubation with peroxidaseconjugated swine anti-rabbit immunoglobulin IgG (Bio-Rad; Hercules, CA), diluted 1:5000 in TBS-T with 1% BSA for 1 hr at RT. The labeled proteins were visualized with a chemiluminescence reagent (NEN Life Science; Boston, MA).
In Situ Hybridization Analysis of HAI-1 mRNA
For the probe, a 622-BP cDNA fragment of HAI-1 corresponding to bases 144-765 of HAI-1 cDNA sequence was generated by the polymerase chain reaction (PCR), using a plasmid harboring full-length HAI-1 cDNA as a template (Shimomura et al. 1997). The PCR product was subcloned into plasmid pCR II (Invitrogen; San Diego, CA), sequenced to confirm the correct sequence and direction of the inserted cDNA, and was used as the transcription template. In vitro transcription to generate sense and antisense hybridization probes incorporating digoxigenin-labeled UTP was done according to the manufacturer's instructions (Boehringer; Mannheim, Germany). Transcription products were checked with agarose gel electrophoresis, and the same amount of each sense or antisense probe was used for hybridization. For hybridization, a small piece of fresh nonneoplastic gastric mucosa was fixed in a fixation solution (freshly prepared 4% paraformaldehyde in PBS) for 1 hr on ice. Then frozen sections 4 μm thick were cut, placed on 3-aminopropyltriethoxysilane-coated slides, fixed again with the fixation solution, and rinsed in nuclease-free water. Sections were digested for 10 min at RT with 20 μg/ml nuclease-free proteinase K (Gibco BRL; Gaithersburg, MD) in 10 mM Tris-HCl, 1 mM EDTA, pH 8.0. Then the sections were postfixed with 4% paraformaldehyde in phosphate buffer (PB) for 10 min, rinsed in PB once, incubated in 0.2 N HCl for 10 min at RT, and rinsed once in PB. Tissues were acetylated in 0.25% acetic anhydride containing 0.1 M triethanolamine for 10 min to reduce potential nonspecific binding sites, followed by washing in PB and dehydration with ethanol. Hybridization was done in 10 mM Tris-HCl (pH 7.6) containing 50% deionized formamide, 1 mM EDTA, 0.25% SDS, 600 mM NaCl, 10% dextran sulfate, 1 × Denhardt's solution, 200 μg/ml yeast tRNA, and 50 ng per slide of digoxigenin-labeled probe at 42C overnight in a humid chamber. After the hybridization, the sections were washed in 2 × standard saline citrate (SSC) containing 50% deionized formamide at 50C for 30 min, followed by a single washing in 10 mM Tris-HCl (pH 7.6), 0.5 M NaCl, 1 mM EDTA (TNE) at 37C for 10 min. Then the sections were treated with 30 μg/ml of ribonuclease A in TNE at 37C for 30 min, followed by serial washing. After blocking nonspecific binding sites with 3% BSA in TBS, signals were detected by alkaline phosphatase-conjugated anti-digoxigenin sheep IgG Fab fragment (Boehringer) and chromogenic substrate (BM Purple; Boehringer) according to the manufacturer's instructions.
RNA Blot Analysis
Total RNA was extracted using Trisol reagent (Gibco BRL) from gastrointestinal mucosa and from cultured cells. Twenty μg of total RNA was electrophoresed on a 1% formaldehyde agarose gel and transblotted onto Hybond-N+ nylon membrane (Amersham; Poole, UK) and RNA was UV-crosslinked onto the membrane. Hybridization was performed in a mixed solution of 50% formamide, 5 × Denhardt's solution, 25 mM PB (pH 6.5), 0.1% SDS, 100 μg/ml of sonicated and heat-denatured salmon sperm DNA, and 5 × SSC at 42C for 16 hr. The blots were washed as follows: three times in 0.1% SDS in 1 × SSC for 15 min at RT and twice in the same solution for 20 min at 65C. The membranes were autoradiographed with Kodak XR-5 film at −80C for 6 hr or 24 hr. Human HAI-1 cDNA (Shimomura et al. 1997) was used as a probe. For internal control of loading, the blots were subsequently hybridized to a glyceraldehyde-3-phosphate dehydrogenase (G3PDH) probe (Clontech; Palo Alto, CA). Both probes were radiolabeled by random priming with [α-32 P]-CTP.
Results
Specificity of MAb 1N7
Before we started the immunohistochemical study, the specificity of MAb 1N7 was assessed by immunoblot analysis using whole extracts of cultured MKN-28 and CCK-81 cells and normal human gastric mucosa. As shown in Figure 1A, a single major band of MW around 66 kD and 64 kD was detected in the cultured cells and gastric mucosa, respectively, indicating the specificity of MAb 1N7. The tumor cell-derived HAI-1 showed a slightly higher MW than that derived from normal mucosa, and this difference may reflect an altered glycosylation of HAI-1 in tumor cells. RNA blot analysis revealed that HAI-1 mRNA was in fact present in MKN-28 and CCK-81 cells and in normal gastrointestinal mucosa (Figures 1B and 1C). The levels of HAI-1 mRNA were correlated with the amounts of immunoreactive cellular HAI-1 protein (Figures 1A and 1B). Furthermore, the immunohistochemical reaction was adsorbed by an excess amount of recombinant HAI-1. This was performed by using a normal colon mucosal specimen (Figures 2A and 2B) in which the presence of HAI-1-specific mRNA and of proteins had been previously confirmed (Kataoka et al. 1998).
(A) Immunoblot analysis of HAI-1 protein using MAb 1N7. Whole extract sample of cultured cells (equivalent to 2 × 105 cells) was applied in a 4-12% SDS-PAGE gradient gel. Lane 1, CCK-81 cells; Lane 2, MKN-28 cells; Lane 3, gastric mucosa. (B) RNA blot analysis of HAI-1 mRNA. Lane 1, CCK-81 cells; Lane 2, MKN-28 cells. Corresponding G3PDH signals are also shown. Twenty μg of total RNA was applied in each lane. (C) RNA blot analysis of HAI-1 mRNA in the gastrointestinal mucosa. Lane 1, gastric mucosa; Lane 2, colon mucosa. Twenty μg of total RNA was applied in each lane.
Tissue and Cellular Distribution of HAI-1
Subsequent immunohistochemical studies showed that HAI-1 immunoreactivity was evident in the surface epithelium of the gastrointestinal tracts. In the stomach, the immunoreactivity was largely restricted to the surface epithelial and pit-lining cells, and both fundic and pyloric glandular epithelia were hardly stained (Figure 2C). In the positive cells, the cell surface was predominantly stained (Figure 2D). The stromal components, such as fibroblasts, smooth muscle cells, and inflammatory cells, were negative. In situ hybridization revealed a similar distribution of HAI-1 mRNA in the gastric mucosa (Figure 2F), supporting the specificity of the immunostaining. In contrast, the esophageal stratified squamous epithelium was poorly stained, and only very faint immunoreactivity was detectable, preferentially in the cells of the basal layer (Figures 2G and 2H).
Examples of immunostaining using MAb 1N7 and in situ hybridization for HAI-1 on representative human tissues. (A,B) Adsorption test of immunostaining using colon mucosa. In A, the tissue was stained with MAb 1N7 after adsorption with excess amounts of recombinant HAI-1 proteins. The immunoreactivity observed in B disappeared in A. (C,D) HAI-1 immunostaining in gastric mucosa. (E,F) In situ hybridization for HAI-1 mRNA in gastric mucosa. Sense (E) or antisense (F) riboprobe was used, and specific signals were observed only with the antisense probe. (G,H) HAI-1 immunostaining at junction of esophagus with stomach. Hematoxylin and eosin stain (G) and immunostaining of a serial section (H) are shown. (I,J) HAI-1 immunostaining in liver. Epithelium of intrahepatic bile duct was positive (I), whereas normal hepatocytes were negative (J). (K,L) HAI-1 immunostaining in pancreas. (M) HAI-1 immunostaining in prostate. (N,0) HAI-1 immunostaining in endocervix of uterus. The immunostainings and in situ hybridization were counterstained with hematoxylin and Safranin O, respectively. Bars = 50 μm.
In the liver, the bile duct epithelium was stained, whereas the hepatocytes and hepatic sinusoidal cells were negative (Figures 2I and 2J). The epithelia of the extrahepatic bile ducts and the gallbladder were also stained. In the pancreas, the epithelium of the duct system was stained. In particular, the cells of centroacini, intercalated ducts, and intralobular ducts were strongly stained. By contrast, the acinar cells and cells of the islets were negative (Figures 2K and 2L).
In the kidney, HAI-1 immunoreactivity was observed in the tubule epithelium. The immunoreactivity was strong in the distal tubules and collecting ducts. On the other hand, the proximal tubules were only faintly stained. The glomerular cells were largely negative, and weak signals were occasionally detectable in the glomerular epithelial cells. The transitional epithelial cells of the renal pelvis, ureter, and urinary bladder were hardly stainable. The prostate glandular epithelium was also positively stained, and the immunoreactivity was predominantly observed in the secretory compartment of the epithelium compared with the basal compartment (Figure 2M). The endocervical epithelium of the uterus was strongly stained and the endometrial epithelium was also positive, whereas myometrial tissues were completely negative. In the endocervix, the surface epithelium was stained more strongly than the glandular epithelium (Figures 2N and 2O). The cervical stratified squamous epithelium was only faintly focally stained.
The results of immunostaining are summarized in Table 1. In general, HAI-1 was detectable in the simple columnar epithelium covering the mucosal surface or ducts. Predominant cellular surface localization was suggested, and the lateral surfaces of the cells were preferentially stained (Figures 2D and 2O). It should be noted that the organ distribution shown in Table 1 was in accordance with the previous RNA blot studies (Shimomura et al. 1997). In addition, Schwann cells were positively stained. Staining for the endothelial cells was disputable, and occasionally the endothelia of the capillaries and lymphatics were stained positively.
Immunolocalization of HAI-1 in normal human tissues a
+ +, strongly positive in more than 90% of cells; +, positive; − to ±, faintly positive in less than 20% of cells; −, negative.
Immunoelectron Microscopy
We confirmed that pretreatment of cryostat sections with Tris buffer, pH 9.0 or pH 10.0, remarkably enhanced HAI-1 immunostaining by MAb 1N7, although the ultrastructures were damaged to a lesser extent by the alkaline treatment. In the body portions of human stomach, the immunostaining for HAI-1 was observed on the surface of the basolateral plasma membrane of the surface epithelial cells. The lateral plasma membrane generally showed intenser immunoreaction than that of the basal plasma membrane (Figure 3). These results were quite compatible with the light microscopic findings above. The control experiments did not exhibit any immunoreactions (not shown).
Immunoelectron microscopy for HAI-1 in gastric mucosal surface epithelium. The section was pretreated at pH 9.0. Immunoreactivity was observed along with the lateral surface of the cells (arrowheads). (A) Low magnification. Bar = 1 μm. (B) High magnification. Bar = 500 nm.
Enhanced Expression of HAI-1 Protein in Injured Tissues
Because the generation of active HGFA is believed to occur exclusively in injured tissue through proteolytic activation of pro-HGFA by thrombin (Shimomura et al. 1993; Miyazawa et al. 1994,1996), it is reasonable to postulate that the expression of HAI-1, an endogenous inhibitor of HGFA, might be modulated in injured tissues. To examine this hypothesis, two severely injured tissues were immunostained for HAI-1. Figure 4A shows a specimen of infarcted renal parenchymal tissue after embolization therapy for renal cell carcinoma. In the infarcted area (right two thirds of Figure 4A), regenerating renal tubules were also noted. Of interest was the observation that the regenerating epithelium displayed significant expression of HAI-1 in comparison to the epithelium in the noninfarcted area (Figures 4B-4D). The second case, a specimen of liver tissue, was obtained at autopsy from a patient who had died of fulminant hepatitis (Figure 4E). Although hepatocytes in the normal liver were negative for HAI-1 (Figure 2J), in the fulminant hepatitis tissue, pseudo-bile canaliculi and scattered hepatocytes were strongly stained (Figure 4F). Again, predominant cell surface staining was evident.
HAI-1 immunostaining in injured tissues. (A) Kidney of postinfarction state. Right two thirds represents the infarcted area, showing hemorrhage, tubule necrosis, and regeneration. Hematoxylin-eosin. (B,D) HAI-1 immunostaining of a serial section to A. Regenerating epithelial cells in the infarcted area are strongly stained. (C) HAI-1 immunostaining of noninjured renal tubule tissue. Immunoreactivity is observed on the lateral surface of distal tubule epithelial cells. This immunoreactivity was significantly upregulated in the regenerating epithelium of injured tissue shown in D. (E,F) HAI-1 immunostaining in liver of fulminant hepatitis. Hematoxylin-eosin stain is shown in E, and strong cellular surface immunoreactivity is noted in pseudo-bile ductules and in hepatocytes (F). Bar = 50 μm.
Discussion
This work presents the first systematic immunohistochemical study using an MAb directed against HAI-1 protein in human tissues. Specificity of the immunohistochemical reaction was verified in the following manner. (a) Staining of positive and negative control tissue sections in each assay was prepared. (b) The immunoreactivity was adsorbed by a recombinant HAI-1 protein. (c) In situ hybridization using specific anti-sense riboprobe for HAI-1 mRNA was followed by a stringent ribonuclease A treatment, resulting in a distribution pattern of the specific signal similar to the immunohistochemistry results. (d) The antibody recognized specifically HAI-1 protein in the immunoblot analysis. Finally (e), the immunostaining pattern of HAI-1 in paraffin sections using antigen retrieval is comparable with that obtained in frozen section (Kataoka et al. 1998). Furthermore, the results of immunohistochemistry were in accordance with previous RNA blot analysis (Shimomura et al. 1997). Therefore, it is reasonable to assume that the signals we detected were specific for HAI-1 protein.
HAI-1 protein was predominantly expressed on the surface of the simple columnar epithelium covering the mucosal surface and duct lumen. In contrast, stratified squamous epithelia of various organs, transitional epithelia of the urinary tract, ciliated bronchial epithelium, and hepatocytes were largely negative or were stained only very weakly. In addition to the columnar epithelium, Schwann cells were also stained for HAI-1. The predominant cellular surface localization of HAI-1 confirmed in this report is consistent with the molecular structure of HAI-1 protein deduced by the mRNA sequence, i.e., the presence of a presumed transmembrane domain in the HAI-1 molecule (Shimomura et al. 1997). Although this unique structure has been reported in other Kunitz-type serine proteinase inhibitors, such as amyloid-β protein precursor (APP) and its homologues (APP-like proteins), the cell surface localization of APP has been disputable and it may be preceded by a secretory form intracellularly (Kuentzel et al. 1993; Haass et al. 1995). In contrast, our study suggests that HAI-1 is consistently present on the cell surface and thus may be acting on the cell surface. In fact, there is evidence that exogenously added active HGFA proteins may bind to cell surface HAI-1 in in vitro culture conditions (unpublished observations).
At present, the exact biological role of HAI-1 in vivo is uncertain. Because HAI-1 is an efficient endogenous inhibitor of HGFA, the expression of HAI-1 may have an important role in regulation of the activation of HGF/SF in local tissues. This hypothesis may agree with the fact that the cells which expressed HAT 1 protein in this study are susceptible to HGF/SF. For example, HGF/SF is an important mitogen for the gastric, bile duct, and renal tubule epithelium (Kawaida et al. 1994; Takahashi et al. 1995; Matsumoto and Nakamura 1996; Kinoshita et al. 1997) and Schwann cells (Kranoselsky et al. 1994). However, another important target of HGF/SF, hepatocytes, did not express HAI-1 protein, and bronchial and alveolar epithelia, which are also susceptible to HGF/SF (Ohmichi et al. 1996), expressed a very low level of HAI-1 protein. Several explanations can be offered for these discrepancies. (a) Another system controlling HGFA activity might exist in organs such as liver and lungs. Another recently identified HGFA inhibitor, HAI-2, may be one of the candidates. However, the levels of HAI-2 mRNA were also low in the liver and lungs (Kawaguchi et al. 1997; Marlor et al. 1997). (b) The expression of HAI-1 might be highly situational, depending on the cell type and the presence of other constituents in the intra- or extracellular milieu. The fact that HAI-1 protein was stained in hepatocytes of a severely inflamed liver but not in a normal liver argue in favor of this hypothesis. (c) In organs such as the liver and lungs, HAI-1 expression might be downregulated to ensure efficient activation of proHGF/SF by HGFA. (d) In specific cell types, HAI-1 might be very rapidly secreted so that the protein is not detectable in the cells immunohistochemically. However, this explanation is unlikely because mRNA levels of HAI-1 in the liver and lungs were also very low. An alternative hypothesis is that HAI-1 might have an undefined important function in the columnar epithelium covering the mucosal surface, possibly acting against excess proteinase(s) to maintain the architecture and integrity of the mucosal epithelium. This hypothesis may be supported by the fact that HAI-1 has two Kunitz domains and the second domain appeared to be not involved in the inhibition of HGFA (Shimomura et al. 1997). Therefore, HAI-1 could have an unknown target proteinase(s) in addition to HGFA.
Of particular interest was the observation that immunoreactivity to HAI-1 protein was upregulated in injured and regenerating tissues. In the injured tissues, proteolytic activation of zymogen of HGFA would occur through thrombin generation, generating a considerable amount of active HGFA (Shimomura et al. 1993; Miyazawa et al. 1994,1996). In this context, there may exist two possible explanations for the upregulation of HAI-1 in the injured tissues. First, in response to the elevated HGFA activity, HAI-1 was upregulated to regulate the excess HGFA activity. Second, HAI-1 may serve as a specific cell surface acceptor for the active HGFA to localize the enzyme efficiently on the surface of cells that are going to enter the regenerative process. In fact, the binding between HAI-1 and active HGFA appeared to be reversible (unpublished observations). Alternatively, in addition to HAI-l's role as an HGFA inhibitor, it is possible that increased HAI-1 production in injured tissues may, in certain situations, stimulate cell survival and proliferation. Indeed, a number of studies have shown that serine proteinase inhibitors can act as cell survival and growth factors. Serine proteinase inhibitors increase the survival of hepatocytes in serum-free culture by inhibiting trypsin-like proteinase associated with plasma membrane (Nakamura et al. 1984), and exhibit growth stimulatory activity on endothelial cells (McKeehan et al. 1986). In addition, the overexpression of HAI-1 in pseudo-ductules of fulminant hepatitis raise the possibility that elevated HAI-1 activity is involved in defective regeneration in the liver that has undergone fulminant hepatitis. The only proliferating compartment in fulminant hepatitis is the pseudo-ductules, which overexpress HAI-1. Moreover, a high level of HGF/SF in plasma during fulminant hepatitis has been reported (Tsubouchi et al. 1991), and we have observed that HGFA protein is in fact expressed in hepatocytes (unpublished observations). Therefore, it can be postulated that the ineffectiveness of HGF/SF on hepatocyte regeneration in fulminant hepatitis may be caused by inhibition of HGF/SF activation by overexpression of the inhibitor in the proliferating cell compartment.
In summary, the distribution and cellular localization of HAI-1 protein were examined in normal adult tissues. Although the exact function and role of HAI-1 in vivo remain to be clarified, its distribution suggests that this protein may have an important function in the surface columnar epithelium of various organs. We also hypothesize that HAI-1 may be somehow involved in tissue responses to injury and in regenerative processes, both of which require further investigation.
Footnotes
Acknowledgements
We are grateful to Dr K. Denda (Department of Life Science, Tokyo Institute of Technology) for recombinant HAI-1, to Dr H. Tsubouchi (Second Department of Internal Medicine, Miyazaki Medical College) for helpful suggestions, and to Mr S. Ide and Ms N. Iwakiri for excellent technical assistance.