Distribution of the Human Intracellular Serpin Protease Inhibitor 8 in Human Tissues

Abstract

Ovalbumin-like serine protease inhibitors are mainly localized intracellularly and their in vivo functions are largely unknown. To elucidate their physiological role(s), we studied the expression of one of these inhibitors, protease inhibitor 8 (PI-8), in normal human tissues by immunohistochemistry using a PI-8-specific monoclonal antibody. PI-8 was strongly expressed in the nuclei of squamous epithelium of mouth, pharynx, esophagus, and epidermis, and by the epithelial layer of skin appendages, particularly by more differentiated epithelial cells. PI-8 was also expressed by monocytes and by neuroendocrine cells in the pituitary gland, pancreas, and digestive tract. Monocytes showed nuclear and cytoplasmic localization of PI-8, whereas neuroendocrine cells showed only cytoplasmic staining. In vitro nuclear localization of PI-8 was confirmed by confocal analysis using serpin-transfected HeLa cells. Furthermore, mutation of the P₁ residue did not affect the subcellular distribution pattern of PI-8, indicating that its nuclear localization is independent of the interaction with its target protease. We conclude that PI-8 has a unique distribution pattern in human tissues compared to the distribution patterns of other intracellular serpins. Additional studies must be performed to elucidate its physiological role.

Serine protease inhibitors (serpins) are a large superfamily of structurally related proteins that occur in viruses, insects, plants and higher organisms, but not in bacteria or yeast (Silverman et al. 2001). Serpins regulate the activity of proteases involved in such diverse processes as coagulation, fibrinolysis, inflammation, cell migration, and tumorigenesis (Potempa et al. 1994). Serpins share about 30% homology and have similar 3D structure and inhibition mechanisms (Potempa et al. 1994; Silverman et al. 2001). They contain a surface-exposed reactive site loop (RSL), which acts as a “bait” for proteases by mimicking a substrate sequence. The P₁ residue within the RSL largely determines the specificity of the serpin. On binding of the target protease to the serpin, the RSL is cleaved at its P₁ residue, after which the active site serine of the protease is covalently linked to the P₁ amino acid residue due to a rapid insertion of the RSL into the β-sheet of the serpin. The protease in the complex thus formed is inactive (Huntington et al. 2000).

Recently, a subfamily of serpins, of which chicken ovalbumin is the archetype, with substantial higher homology (around 50%) has been identified (Remold-O'Donnell 1993). Serpins belonging to this subfamily lack a typical cleavable N-terminal signal sequence and reside mainly intracellularly. Members of this human ovalbumin-like subfamily include plasminogen activator inhibitor type 2 (PAI-2) (Ye et al. 1987), monocyte neutrophil elastase inhibitor (MNEI) (Remold-O'Donnell et al. 1992), squamous cell carcinoma antigen (SCCA)-1 (Suminami et al. 1991), leupin (SCCA-2) (Barnes and Worrall 1995), maspin (PI-5) (Zou et al. 1994), protease inhibitor (PI)-6 (Coughlin et al. 1993), PI-8 (Sprecher et al. 1995), PI-9 (Sprecher et al. 1995), and bomapin (PI-10) (Riewald and Schleef 1995). Within this family the serpins PI-6, PI-8, and PI-9 show the highest structural homology (up to 68% amino acid identity) (Sprecher et al. 1995).

The physiological role of intracellular serpins remains largely unknown, in part because their cognate substrates (serine proteases) are not yet identified. Although in vitro studies have been performed to identify target proteases, most of the proteases identified reside extracellularly and are therefore probably not the physiological target proteases in vivo. In addition, knowledge about the site of expression of these serpins in vivo is of importance to understand their function. Although ovalbumin-like serpins are widely expressed among human tissues, each has its own restricted expression pattern. For example, the PI-8-related serpin PI-9 is mainly expressed by dendritic cells, endothelial cells, cytotoxic lymphocytes, and cells of immune-privileged sites (Bladergroen et al. 2001). In vitro, PI-9 inhibits the cytotoxic lymphocyte-specific serine protease granzyme B that induces apoptosis when released on target cell recognition. Hence, the in vivo PI-9 distribution supports the hypothesis that PI-9 serves to scavenge misdirected granzyme B in cytotoxic cells and to protect other cells from granzyme B released in their neighborhood, as is the case during the immune response (Spaeny-Dekking et al. 1998; Buzza et al. 2001). Thus, the tissue distribution patterns of intracellular serpins can provide clues about their physiological function.

PI-8 is a 45-kD serpin with arginine at the P₁ position in its RSL, indicating that it probably inhibits trypsin-like proteases. PI-8 inhibits trypsin, thrombin, factor Xa, subtilisin A, furin, and also chymotrypsin in vitro (Dahlen et al. 1998). Yet the function of this serpin is still unknown. As a first step to resolve the (patho-)physiological functions of PI-8, we studied the expression of this inhibitor in various normal human tissues. PI-8 appears to be mainly expressed by cells of the differentiated layers of squamous epithelium, by certain neuroendocrine cells, and by monocytes. Remarkably, PI-8 showed prominent nuclear localization in some but not all of these cells.

Materials and Methods

Materials and Antibodies

Dulbecco's modified Eagle's medium (DMEM), RPMI 1640, and Iscove's modified Dulbecco's medium (IMDM) were obtained from Bio-Whittaker Europe (Verviers, Belgium). Fetal bovine serum (FBS) was purchased from Life Technologies (Rockville, MD). All primers were synthesized by Eurogentec (Brussels, Belgium). Qiaex II Gel Extraction Kit and Qiagen Plasmid Maxi Kit were obtained from Qiagen (Hilden, Germany). JM109 high-efficiency competent cells were obtained from Promega (Madison, WI). The ABI prism sequence kit Thermo Sequenase DNA sequencing kit was purchased from Amersham (Arlington Heights, IL). The following reagents were obtained from DAKO (Glostrup, Denmark): normal rabbit serum, normal swine serum, rabbit anti-human glucagon antibody, mouse anti-human thyroid-stimulating hormone (TSH) antibody, mouse anti-human chromogranin antibody, biotinylated rabbit anti-mouse F(ab)₂ Ig, biotinylated swine anti-rabbit F(ab)₂, HRP-conjugated rabbit anti-mouse Ig, HRP-conjugated swine anti-rabbit Ig, FITC-conjugated rabbit anti-mouse Ig, avidin–biotin–HRP complex (sABC), biotinylated tyramine (BT), and streptavidin–FITC. Biotin-labeled goat anti-mouse IgG1 and HRP-labeled goat anti-mouse IgG_2a Abs were obtained from Southern Biotechnology Associates (Birmingham, AL) and tyramine–rhodamine from DuPont Pharmaceuticals (Wilmington, DE). FuGENE 6 transfection reagent was obtained from Roche Molecular Biochemicals (Indianapolis, IN). 4-(2-aminoethyl)-benzenesulfonylfluoride-HCl was obtained from AG Scientific (San Diego, CA), propidium iodide from Alexis Biochemicals (Lausanne, Switzerland), and Nonidet P-40 and poly-l-lysine from Sigma (St Louis, MO). The Micro BCA Protein Assay was obtained from Pierce (Rockford, IL).

The anti-PI-9 monoclonal antibodies MAb PI-9-17 and MAb PI-9-1 were produced as described previously (Blader-groen et al. 2001). The anti-PAI-2 MAb MAI-21 was purchased from Biopool International (Umea, Sweden). The MAb anti-PI-8k was raised against Pichia pastoris-expressed human PI-8, produced and purified as described previously (Dahlen et al. 1997).

Plasmids and Cloning

The constructs PI-6-pcDNA3, PI-8-pcDNA3, and PI-9-pcDNA3.1, comprising the cDNA sequence coding for the full-length proteins, were prepared as previously described (Sprecher et al. 1995; Bladergroen et al. 2001). PAI-2-pCI was kindly provided by Dr. E.K. Kruithof (Division of Angiology and Hemostasis, University Hospital Geneva; Geneva, Switzerland). Each of these expression vectors contains an SV40 origin of replication and a cytomegalovirus promotor for protein overexpression in eukaryotic cells. Plasmid DNA was amplified by transformation to JM109 cells and isolated with a Qiagen plasmid maxi kit.

The point mutation of the P₁ residue (Arg³²¹→Ala³²¹) in the RSL of PI-8 was induced by the PCR method using the PI-8-pcDNA3 plasmid as a template. The PI-8-P_1(arg–ala) construct was amplified with a Kozak sequence (GCC ACC) in front of the ATG and cloned as a BamHI/XhoI fragment into PCR-3 (Invitrogen) plasmid and sequenced on both strands to confirm the mutation. PI-8 without the first ATG (amino acids 2–374) was cloned into a modified PCR-3 vector containing a FLAG sequence as a SalI/XhoI fragment. From the N-terminal side the amino acid sequence is M(DYKDDDDK)EFCRYPSHWRPLDDDL with the FLAG sequence between brackets and the PI-8 sequence in bold.

Cell Culture and Transfection

Cos-1 cells (CRL-1650, American Type Culture Collection, Manassas, VA) were cultured in IMDM supplemented with 5% (v/v) heat-inactivated FBS, 2 mM l-glutamine, and penicillin/streptomycin (fc. 50 IU/ml and 50 μg/ml, respectively). HeLa cells (CCL-2; ATCC) were cultured in RPMI 1640 medium supplemented with 5% (v/v) heat-inactivated FBS, 2 mM l-glutamine, and antibiotics. All cells were maintained at 5% CO₂/95% air in a humidified incubator at 37C.

Cos-1 or HeLa cells were grown in 25-cm² culture flasks or on sterile glass coverslips in 6-well plates for 24 hr. Medium was refreshed just before transfection with plasmid DNA. Cells were transfected with either PI-6-pcDNA3, PI-8-pcDNA3, PI-9-pcDNA3.1, PAI-2-pCI, PI-8-P_1(arg–ala)-PCR3, or an empty control plasmid. Transfection was performed with FuGENE 6 transfection reagent according to the manufacturer's instructions. FuGENE 6 transfection reagent (microliters) to plasmid DNA (micrograms) was used in a ratio of 3:2 and subsequently cultured for at least 48 hr. The transfected cells on coverslips were further analyzed by immunofluorescence staining (see below). For production of cell lysates, cells in the plates were washed twice with PBS and lysed by adding lysis buffer [PBS with 0.2% (w/v) Nonidet P-40] directly in the culture plate. Cells were incubated for 15 min on ice, after which the lysate was harvested and centrifuged for 10 min at 3000 rpm to remove cell debris and DNA. The supernatant (cell lysate) was stored at −80C until further use.

Immunohisto/cytochemistry and Immunofluorescence Staining

Slides with sections of formalin-fixed, paraffin-embedded normal human tissues were obtained from the tissue bank of the Department of Pathology, VU University Medical Center (Amsterdam, The Netherlands). All tissues were sampled from surgical specimens within 2 hr after resection. Formalin-fixed [10% (v/v) for 18 hr], paraffin-embedded tissue was used. Sections (3 μm thick) were mounted on poly-l-lysine-coated tissue slides and deparaffinized. Cytospins of serpin-transfected cells were fixed in 10% formalin for 10 min. Endogenous peroxidase activity was blocked by incubation in 0.3% (v/v) H₂O₂ in methanol for 30 min. Unless stated otherwise, tissue sections and cytospins were subjected to antigen retrieval by boiling in 10 mM, pH 6.0, sodium-citrate buffer for 10 min in a microwave oven. All antibodies and normal serum were diluted in PBS containing 1% (w/v) bovine serum albumin (BSA). Tissue sections were pre-incubated for 10 min with normal rabbit or swine serum, followed by incubation for 1 hr with purified MAb anti-PI-8k (at 1.4 μg/ml). To identify specific cell populations, sequential sections of several tissues were incubated for 1 hr with the appropriate Abs against various cell markers. Figure 2 shows staining with the following Abs: rabbit anti-human glucagon (1:100), mouse anti-human TSH (1:100), and mouse anti-human chromogranin (1:100). After washing with PBS, slides were incubated with a biotin-conjugated secondary antibody (rabbit anti-mouse F(ab')₂ Ig, 1:500 dilution; swine anti rabbit F(ab')₂, 1:300 dilution) for 30 min. Slides were incubated with streptavidin–biotin–HRP complex (sABC; 1:1000 dilution) for 1 hr, followed by incubation with biotinylated tyramine (BT) for 10 min. After a second incubation with sABC (1:200), PI-8 or cell markers were visualized with 3-amino-9-ethylcarbazole (AEC) or 3,3′-diaminobenzidine (DAB; 0.1 mg/ml, 0.02% H₂O₂). Cytospins were incubated with AEC directly after the first incubation with sABC. Slides were counterstained with hematoxylin and mounted with Depex. Negative control slides were stained with mouse IgG of the appropriate subclass.

For immunofluorescence staining of coverslips containing transfected Cos-1 or HeLa cells, slides were washed twice in PBS, fixed in 10% formalin for 10 min at RT, and washed again in PBS. Cells were then permeabilized in PBS with 0.5% (w/v) Nonidet P-40 for 10 min at RT. Incubation for 5 min in PBS with 0.75% (w/v) glycin, followed by two incubations for 10 min in PBS with 0.5% BSA, were performed as blocking steps before the immune fluorescence labeling. All antibodies were diluted in PBS with 0.5% BSA. Fixed cells were stained for PI-8 with purified MAb anti-PI-8k (at 1.1 μg/ml). PI-9 was detected using MAb PI-9-17 at a concentration of 2 μg/ml. On washing with PBS with 0.5% BSA, the cells were incubated in the dark with the secondary antibody, FITC-labeled rabbit anti-mouse F(ab')₂ Ig diluted 1:500. Cells were counterstained with propidium iodide (20 μg/ml), diluted 1:100 in PBS with 0.5% BSA for 5 min and mounted in Vectashield (Vector Laboratories; Burlingame, CA). Images were recorded using a Zeiss LSM-510 confocal laser scanning microscope equipped with argon and helium/neon lasers. Excitation was at 488 nm for FITC and 568 nm for propidium iodide.

Western Blotting

About 20 μg of tissue section lysate protein or 10 μl of cell lysate (4.10⁵ cells) from serpin transfected Cos-1 or HeLa cells was resolved by electrophoresis on a 10% (w/v) SDS-polyacrylamide gel under reducing conditions. After electrophoresis, proteins were transferred to nitrocellulose membranes by electrophoretic blotting. After transfer, the membranes were blocked for 1 hr in blocking buffer [5% (w/v) skim milk powder, 0.5% (w/v) BSA, 0.1% (w/v) Tween-20 in PBS]. Membranes were then incubated overnight at 4C with MAb anti-PI-8k at a concentration of 0.6 μg/ml in blocking buffer and washed in PBS containing 0.1% (w/v) Tween. The membranes were incubated for 1 hr with HRP-conjugated rabbit anti-mouse Ig (1:1000 dilution in blocking buffer), followed by another washing step. Bound Abs were visualized with a chemiluminescence development reagent (ECL system; Amersham) according to the manufacturer's instructions.

Results

Characterization of MAb Anti-PI-8k Directed Against Human PI-8

A panel of MAbs was obtained from a fusion experiment of a mouse immunized with recombinant human PI-8 produced in a Pichia pastoris expression system (Dahlen et al. 1997). In screening experiments one of these MAbs, PI-8k (subtype IgG₁), specifically recognized the 45-kD PI-8 protein but not the two most homologous serpins PI-6 and PI-9 or the less related serpin PAI-2 by Western blotting (Figure 1A), immunocytochemistry (Figure 1B), and immunofluorescence staining (not shown) of serpin-transfected Cos-1 cells. Expression of the other serpins was confirmed when the Western blot was probed with the serpin-specific antibodies MAb PI-9-17 (anti-PI-9), MAI-21 (anti-PAI-2), or MAb PI-9-1 (anti-PI-9, crossreactive with PI-6 and PI-8) (Bladergroen et al. 2001) (results not shown).

Figure 1

Anti-PI-8 MAb anti-PI-8k is specific for PI-8. (A) Cos-1 cells were transfected with PI-6-pcDNA3, PI-8-pcDNA3, PI-9-pcDNA3.1, or PAI-2-pCI. Control cells were transfected with empty plasmid. A total of 10 μl of the cell lysate (4.10⁵ cells) was analyzed by 10% SDS-PAGE and Western blotting. Blots were incubated overnight with MAb anti-PI-8k. The molecular mass of marker bands is indicated at left. (B) Cytospins of Cos-1 cells transfected with the same constructs as mentioned in A were analyzed by immunocytochemistry using MAb anti-PI-8k as described in Materials and Methods. Nuclei were counterstained with hematoxylin.

On the basis of these data, MAb anti-PI-8k was considered to be specific for PI-8 and was used for further studies.

PI-8 Is Mainly Expressed in Nuclei of Epithelial Cells

Analysis of the expression of PI-8 in normal human tissues by immunohistochemistry (Figure 2) showed that PI-8 was prominently expressed in the nuclei of squamous epithelium. Figure 2a shows a tonsil lined with squamous epithelium in which the nuclei express PI-8. The lymphoid tissue lying beneath the epithelium (Figure 2a, asterisk) does not contain PI-8. A higher magnification of the squamous epithelial layer shown in Figure 2a is depicted in Figure 2b. The basal cells do not express PI-8 (Figure 2b, arrow), while PI-8 expression in the nuclei of keratinocytes strongly increases during epithelial differentiation (Figures 2b and 2d, arrowheads). PI-8 protein occurs mainly in the nuclei, whereas the cytoplasm of these cells does not contain measurable amounts of the serpin. The same pattern was observed in squamous epithelia from other sites, such as the esophagus (Figure 2c) and skin (Figure 2d). The nuclei of epithelial cells present in skin appendages, such as hair follicles (Figure 2e), also expressed PI-8. Interestingly, nuclei from epithelial cells present in the Hassall's corpuscles in the thymus were also positive for PI-8 (results not shown).

The other non-squamous epithelial tissue types showed a more heterogeneous pattern. Epithelium from colon (Figure 2g), lung (Figure 2k), and exocrine pancreas (Figure 2q, asterisk), salivary glands (except for the ducts), endometrium, prostate, and breast (not shown) were in general negative or sporadically positive for PI-8. In contrast, the epithelium of the antral part of the stomach (Figure 2f), the small intestine, tubes, and endocervix (not shown) showed moderate to strong nuclear staining of PI-8, whereas the cytoplasm of these cells hardly contained the serpin. In most other organs tested, such as kidney (Figure 2i), heart (Figure 2j), brain (Figure 2m), placenta (Figure 2n), and liver (not shown), the epithelial and/or mesenchymal structures did not express PI-8. As an internal positive control, PI-8-positive monocytes (see below) were detected in all cases.

PI-8 Expression by Monocytes and Neuroendocrine Cells

Monocytes detected in the blood vessels of most organs showed strong PI-8 expression. Figure 2n shows a high magnification of a monocyte (arrow) in which cytoplasmic as well as nuclear staining is observed. PI-8-positive monocytes were used as a positive internal control for the immunohistochemical staining in PI-8-negative organs (Figure 2n, arrow). Although alveolar macrophages in the lung showed weak staining (Figure 2k, arrow), other macrophage or dendritic subsets present in the lymphoid (follicular dendritic cells, sinus macrophages) or other organs (Kupffer cells in the liver) did not express PI-8 (not shown). Cells from the lymphoid system such as T-, B-, and plasma cells, did not contain detectable amounts of PI-8, as can be observed in Figures 2a and 2l in which the lymphoid tissue from the tonsil or spleen (asterisk) is shown, respectively. Neutrophilic granulocytes were mainly PI-8-negative, although sometimes weak staining was observed. Endothelial cells were negative for PI-8, whereas mesothelial cells were positive (not shown).

Interestingly, high PI-8 protein levels were detected in certain neuroendocrine cells of the pituitary gland (Figure 2o), endocrine pancreas (islets of Langerhans, Figure 2q), paraganglia (not shown), and gastrointestinal tract (Figure 2g). In the pituitary gland, only the TSH-producing cells expressed PI-8 (Figure 2o), as confirmed by sequential staining for TSH (Figure 2p). The other hormone-producing cells of the pituitary gland were mainly PI-8-negative. In the islets of Langerhans of the pancreas, the cells arranged around the periphery expressed PI-8 (Figure 2q). Sequential staining identified these cells as the glucagon-producing or α-cells (Figure 2r), whereas the insulin-producing or β-cells in the central region of the islet were PI-8-negative (not shown). Neuroendocrine cells in the gastrointestinal tract also contained PI-8. Figure 2g shows several scattered PI-8-positive cells in the crypts of the colon mucosa (arrowhead). The cells were lying between the PI-8-negative epithelial cells, and sequential staining for the neuroendocrine marker chromogranin (Figure 2h, arrowhead) showed that these cells belong to the neuroendocrine system. The parathyroid gland and the adrenal gland showed no PI-8 staining (results not shown). In contrast to the nuclear expression in the epithelial cells, PI-8 appears to be expressed mainly in the cytoplasm of the neuroendocrine cells, as clearly shown in Figure 2o. As negative control, all tissues were also stained using an isotype-matched irrelevant monoclonal antibody that showed no background staining (results not shown).

Western Blotting Analysis of Tissues Expressing PI-8

The distribution pattern of PI-8 among the various tissues as seen by immunohistochemistry, as well as the identity of the protein, was broadly confirmed by Western blotting (Figure 3). Consistent with the immunohistochemical data, no or hardly any PI-8 was detected in brain, heart, and renal tissues. The endocrine (pituitary gland, pancreas) and epithelial tissues were positive for PI-8. The weak expression in the pancreas was probably due to the low percentage of α-cells in the pancreatic tissue. The spleen, liver, and lung also showed strong staining, presumably due to presence of high amounts of monocytes and alveolar macrophages, respectively. The PI-8 protein detected has the same M_r (45 kD) as that of the positive control consisting of HeLa cells transfected with PI-8/pcDNA3 plasmid (arrow) and represents the uncomplexed, uncleaved PI-8 protein. Interestingly, in the pituitary gland and liver there was also a higher band of 60 kD (arrowhead), and in the oropharyngeal mucosa yet another band, with an M_r of approximately 70 kD, was observed. These HMW bands may represent SDS-resistant complexes of PI-8 covalently linked to as yet unidentified target serine proteases.

Subcellular Distribution of PI-8 In Vitro

During immunohistochemical survey of normal human tissues, PI-8 was observed in the nuclei of epithelial cells and in both the nuclei and cytoplasm of monocytes. However, no nuclear localization was observed in neuroendocrine cells.

To confirm the subcellular distribution in vitro, we analyzed the presence of PI-8 in the human epithelial cell line HeLa by using MAb anti-PI-8k in confocal microscopy. Western blotting analysis showed low levels of endogenous PI-8 in HeLa cells (Figure 3), which could not be detected by confocal microscopy analysis (Figure 4m, right). Therefore, we analyzed HeLa cells transiently transfected with the full-length PI-8 gene. Confocal microscopic analysis using MAb anti-PI-8k showed a strong nuclear staining of PI-8 in these transfected cells (Figure 4m). Interestingly, within the strong PI-8 positive nuclei, the nucleoli consistently showed a lower signal for PI-8 (Figures 4k–4m and 4q). As a control, HeLa cells were also transfected with a FLAG-tagged PI-8 construct, after which PI-8 was visualized using an MAb directed against the FLAG sequence. Using this approach, a similar subcellular localization of PI-8 was found, i.e., nuclear localization with lower amounts in the nucleoli (results not shown). If the primary antibody (anti-PI-8k) was omitted, no fluorescent labeling was observed, excluding nonspecific labeling with the FITC-conjugated secondary antibody (not shown). PI-8 was not or only faintly present in the cytoplasm (Figure 4). Cytoplasmic localization was mainly seen in cells with the highest PI-8 expression levels (compare cell at the left with the cell at the right in Figure 4m) and is probably due to overexpression of the protein. These in vitro data therefore support the in vivo observation that PI-8 is transported to the nucleus in epithelial cells.

The Nuclear Localization Is Independent of the RSLP₁ Residue

The presence of PI-8 in the nucleus suggests that either PI-8 itself contains a nuclear localization signal (NLS) or that PI-8 interacts with another protein in the cytoplasm that contains the appropriate signal. However, based on primary amino acid composition, PI-8, in contrast to PI-10 (Chuang and Schleef 1999), lacks a conventional NLS. Therefore, we investigated whether the nuclear targeting of PI-8 is dependent on binding to its (unknown) target protease. Binding to a cytoplasmic protease could facilitate transport to the nucleus or, on the other hand, binding to a nuclear target could prevent the translocation back to the cytoplasm. The protease specificity and inhibitory function of serpins is determined mainly by the amino acid residue at the P₁ position in the RSL (Huntington et al. 2000). Hence, we constructed a PI-8 mutant, PI-8-PM_1(arg–ala) in which the arginine at the P₁ position (amino acid residue 321) was replaced by alanine. As shown in Figure 4, HeLa cells transfected with this mutant (Figures 4n–4p and 4r) showed a similar nuclear localization compared to the non-mutated PI-8 gene (Figures 4k–4m and 4q). These data rule out the possibility that the interaction of the P₁ residue of PI-8 with its target protease might be responsible for nuclear localization.

Figure 2

Immunohistochemical detection of PI-8 in normal human tissues. The following human tissues were stained for PI-8, unless mentioned otherwise: (a) tonsil at the surface squamous stratified epithelium, ^∗; lymphoid tissue, (b) detail of the squamous stratified epithelium of a, arrow; basal cells, arrowhead; differentiated cells (c) esophagus, (d) epidermis, arrowhead; differentiated cells (e) hair follicle, (f) stomach antrum, (g) colon, (h) colon stained for chromogranin, (i) kidney, arrow; monocyte, (j) heart, (k) lung, arrow; alveolar macrophage (l) spleen, asterisk; lymphoid tissue (m) brain, (n) placenta, arrow; monocyte (o) pituitary gland, (p) pituitary gland stained for TSH, (q) Langerhans islet in the pancreas, asterisk; exocrine pancreas, and (r) Langerhans islet in the pancreas stained for glucagon. Tissue sections were stained with MAb anti-PI-8k (a-o,q), anti-chromogranin (h), anti-TSH (p), or with anti-glucagon (r) as described in Materials and Methods. The following slides were stained with DAB: a-c,e,f,h-n. Slides d,g,o-r were stained with AEC. All slides were counterstained with hematoxylin. Original magnifications: (a,e,g,h,k,l,q,r) X400; (c,d,f,i,j,n,-p) X630; b X1000.

Figure 3

Western blotting analysis of PI-8 expression by normal human tissues. Five 5-μm-thick sections of the indicated tissues were dissolved in 50 μl PBS containing 1% Nonidet P-40. Twenty μg of total lysate protein was separated by 10% SDS-PAGE and analyzed by Western blotting with MAb anti-PI-8k (0.6 μg/ml). As a positive control, lysate was included of untransfected HeLa cells and HeLa cells transfected with PI-8 cDNA. The position of the molecular mass marker bands is indicated at left. Arrow indicates the PI-8 protein and arrowhead indicates possible complexes between PI-8 and a serine protease.

Discussion

The (patho)physiological role for intracellular serpins is far from clear. Identification of the tissue or cell types expressing these serpins in vivo will provide clues to the processes regulated by these inhibitors and may help to identify their physiological target proteases. In the present study we investigated the distribution of the intracellular serpin PI-8 in normal human tissues and performed in vitro experiments to verify key findings of these studies. Although a previous study using Northern blotting analysis on human tissues showed a widespread distribution of PI-8 mRNA (Sprecher et al. 1995), the present study shows a more restricted pattern of PI-8 protein expression, in that PI-8 is expressed mainly by cells in the differentiated layer of squamous epithelium, by certain non-squamous epithelial cells, by several (neuro)endocrine cells, and by monocytes. A remarkable finding was the prominent nuclear localization of PI-8, particularly in epithelial cells. Although nuclear localization of intracellular serpins has recently been described (Chuang and Schleef 1999; Bird et al. 2001), our study is, as far as we know, the first to describe nuclear localization of an endogenous serpin in human tissues.

Intracellular serpins show substantial higher structural homology (about 50%) than other members of the serpin superfamily. As a consequence, antibodies raised against one of these serpins may crossreact with the other serpins. The MAb anti-PI-8k, raised against human PI-8 and used in this study, was checked for crossreactivity against the most homologous family members of PI-8, which are PI-6 and PI-9, sharing 68% and 63% amino acid identity with PI-8, respectively (Sprecher et al. 1995). PAI-2 was included as a more distant serpin, sharing 53% amino acid identity to PI-8. The MAb specifically recognized PI-8 and did not crossreact with these other serpins, as determined by transfection studies of Cos-1 cells (Figure 1). Furthermore, the observed tissue distribution pattern of PI-8 was completely different from that of PI-9 (Bladergroen et al. 2001) and more distant intracellular serpins such as maspin (breast) and SCCA-1, the latter only showing very weak and cytoplasmic (!) staining of keratinocytes (Pemberton et al. 1997; Cataltepe et al. 2000). Thus, these results show that it is very unlikely that MAb anti-PI-8k cross reacts with other intracellular serpins and therefore is specific for PI-8.

In this study we have found PI-8 expression in the nucleus of differentiated keratinocytes, whereas basal epithelial cells were PI-8 negative. These results suggest that PI-8 is related to the differentiation of these cells. The notion that intracellular serpins are involved in the cellular differentiation of keratinocytes is supported by the finding that other intracellular serpins are also expressed by differentiating epithelial cells. Epithelial cells from the epidermis also express the serpin PI-6 and, like PI-8, protein levels of PI-6 increase upon differentiation. (Scott et al. 1998). PAI-2, another intracellular serpin, is also expressed in epithelial cells in the epidermis and is believed to be involved in keratinocyte differentiation (Jensen et al. 1995; Risse et al. 1998). Remarkably, PAI-2 and PI-6 share with PI-8 an arginine at the P₁ position in their RSL. Thus, at least three intracellular serpins with specificity for trypsin-like proteases are preferentially expressed in the epidermis, suggesting a role for trypsin-like serine proteases in the differentiation of keratinocytes. Although a PI-6/protease complex has been observed in keratinocytes (Scott et al. 1998), the targets for these serpins in the skin remain to be identified.

Figure 4

Subcellular localization of PI-8 in vitro is independent of the RSL P₁ residue. HeLa cells were grown on coverslips and transfected with either wild-type PI-8 (PI-8wt) (a-m,q) or the P₁-Arg³²¹→ Ala PI-8 mutant (PI-8-P_1(arg-ala)) (n-p,r). Cells were stained for PI-8 protein using MAb anti-PI-8k and FITC-conjugated rabbit anti-mouse Ab (green in a-e,k,n). Nuclei were counterstained with propidium iodide (PI: red in f-j,l,o). Co-localization appears in yellow (m,p). In a-j, a series of CLSM sections through a cluster of cells is shown, demonstrating the subcellular localization of PI-8. In q and r, the intensity of the green and red fluorescent signals along the indicated arrow in m and p, respectively, is plotted. (q) A transfected (left) and a non-transfected (right) HeLa cell. In the non-transfected HeLa cell no PI-8 was detected, whereas in the transfected cell PI-8 is strongly expressed in the nucleus and not in the cytoplasm. The intensity of PI-8 in the nucleus is inversely correlated with the propidium iodide content, indicating that PI-8 levels are lower in the nucleoli. The same distribution was observed with PI-8-P_1(arg-ala) (h).

Of all cells of hematopoietic origin, PI-8 was expressed only by monocytes, although sporadic staining of granulocytes was observed. PI-6, PI-10, PAI-2, and monocyte neutrophil elastase inhibitor are all expressed at different stages of monocyte differentiation and activation. (Genton et al. 1987; Remold-O'Donnell 1985; Schleuning et al. 1987; Wohlwend et al. 1987; Riewald et al. 1998; Scott et al. 1999). Therefore, different intracellular serpins may be involved in the various stages of differentiation and could therefore serve as differentiation markers. Moreover, these cells produce proteases in different stages of cellular differentiation and their activity has to be tightly controlled. PI-6, for example, inhibits cathepsin G and protects the cell from inadvertent exposure to this protease (Scott et al. 1999). In agreement, PI-8 was found in part in the cytoplasm of these cells. It is not known which monocytic protease is inhibited by PI-8, and further studies are required to identify the role of PI-8 in monocytes.

Surprisingly, PI-8 was expressed selectively by certain cells from the neuroendocrine system, such as TSH-producing cells in the pituitary gland, glucagon-producing α-cells in the pancreas, and neuroendocrine cells present in the gastrointestinal tract. Recently, a novel intracellular serpin, endopin 1, has been identified in the neurosecretory vesicles of chromaffin cells from the adrenal medulla. Endopin regulates the chromaffin granule prohormone thiol protease, which is involved in proenkephalin processing (Hwang et al. 1999). The specific expression of PI-8 by various neuroendocrine cells leads us to postulate that PI-8, like endopin, may be involved in the regulation of prohormone processing, although in different cell types than endopin-1.

The most striking feature in the distribution of PI-8 in tissues was its predominant, if not exclusive, nuclear localization in epithelial cells and, to a lesser extent, in monocytes. This nuclear localization in tissues was confirmed in vitro by transient expression of the PI-8 gene in the human epithelial cell line HeLa. In these experiments, PI-8 was present in the cytoplasm as well, but only in cells that expressed the highest amounts of protein. Therefore, cytoplasmic localization in these cells may be a result of saturation of the nuclear routing. Until recently, intracellular serpins were considered to reside only in the cytoplasm, although some, such as PAI-2, are partially secreted. As a first exception, PI-10 was shown to have an insertion between the helices C and D, which contains an NLS (Chuang and Schleef 1999). PI-6, PI-8, and PI-9 do not contain such an insertion and also do not contain a conventional NLS.

Bird and co-workers, using cell lines, showed that various intracellular serpins (PI-6, PI-8, PI-9, and PAI-2) are localized in both the nucleus and cytoplasm. Although these serpins lack conventional NLS sequences, the nuclear localization was found to be an active process requiring an as yet unidentified non-conventional nuclear import pathway (Bird et al. 2001). These authors postulated that this nucleocytoplasmic distribution is a common feature of intracellular ovalbumin-like serpins. However, our in vivo studies show a different and more complicated picture concerning this nucleocytoplasmic expression. The tissue distribution of PI-9 showed that this serpin was mainly detected in the cytoplasm of dendritic cells, lymphocytes, endothelial cells, and cells in immune-privileged sites (Bladergroen et al. 2001). Although we indeed have observed faint nuclear localization of certain PI-9-expressing cells, most of the cells showed only cytoplasmic staining. In sharp contrast, PI-8 showed exclusive nuclear localization in epithelial cells, a pattern distinct from that of PI-9, whereas in (neuro)endocrine cells PI-8 was found only in the cytoplasm and not in the nucleus. In contrast, confocal analysis using HeLa cells transfected with the PI-9 gene showed localization of this serpin in both the nucleus and the cytoplasm, as compared to previous observations (data not shown) (Bird et al. 2001). These results show the additional value of immunohistochemical studies using human tissues, which provide a more physiological picture than in vitro studies. Hence, we postulate that the mechanism governing nuclear localization may be different for various intracellular serpins and may depend on the type and stage of differentiation of the cell.

To what extent interaction with the cognate proteases modulates nuclear localization of intracellular serpins is not clear. The localization pattern of the P₁ mutant of PI-8 was similar to that of wild-type PI-8 in the transfected HeLa cells, suggesting that such an interaction does not provide a positive signal for nuclear localization. On the other hand, by immunoblotting analysis we found, for the first time, evidence for PI-8/protease complexes in endocrine tissues (see Figure 3). Therefore, one could speculate that the interaction of a serpin with a cytoplasmic target proteinase inhibits nuclear translocation.

In conclusion, we show that the distribution of PI-8 in normal human tissues is mainly restricted to the nuclei of epithelial cells and that its expression is strongly related with the stage of cell differentiation. PI-8 is also detected in monocytes and in the cytoplasm of certain (neuro)endocrine cells. This distribution pattern of PI-8 in vivo suggests a relation of this serpin with differentiation of keratinocytes, processing of prohormones in neuroendocrine cells, and protection of monocytes against spillover of proteases from their lysosomal granules. Additional studies are required to elucidate the function of PI-8 in these processes.