Differential expression of the ed sequence-containing form of cellular fibronectin in embryonic and adult human tissues

Fibronectins are adhesive high molecular weight glycoproteins found in basement membranes, in interstitial connective tissue matrix and in soluble form in plasma and other body fluids (Mosher, 1980; Ruoslahti et al. 1981; Vartio & Vaheri, 1983; Furcht, 1983; Hynes, 1985; Yamada et al. 1985). Plasma fibronectin (pFn) is composed of two very similar disulphide-linked subunit polypeptides of M_r 220000–250000 which consist of a series of three different types of repeating, fairly homologous structural domains (Petersen et al. 1983). The term cellular fibronectin (cFn) is generally used when the insoluble tissue or extracellular matrix form of the protein or that secreted by cultured cells is referred to.

Differences in Fn polypeptides seem to be due to multiple mRNAs produced by alternative splicing of the primary transcript encoded by a single gene (Schwarzbauer et al. 1983; Kornblihtt et al. 1984, 1985; Hynes, 1985). Thus, depending on the mRNA used as a template, Fn polypeptides containing different domain compositions may be generated. Among these alternatives there is an extra domain (ED) which forms one of type III homologies and resides between the cell-binding site and the more distal heparin-binding site of Fn. The ED sequence is specific for at least certain forms of cFns (Kornblihtt et al. 1984, 1985; Vibe-Pedersen et al. 1984; Hynes, 1985). In the present study this form of cFn is designated as EcFn. We now report on a monoclonal mouse hybridoma antibody which has a specificity for the ED sequence. Immunohistochemical studies showed that the ED-containing Fn is characteristic in embryonic tissues but is re-expressed in reactive connective tissues of adults.

Human embryonic (F84–75) and adult (F84–67) skin fibroblasts were established from skin biopsies by using standard techniques. The A8387 human fibrosarcoma cells were kindly provided by Dr J. Keski-Oja (Department of Virology, University of Helsinki). HT-1080 human fibrosarcoma cells (ATCC, CCL 121) were from American Type Culture Collection, Rockville, MD. Human umbilical vein endothelial cells were isolated by collagenase perfusion (Hormia et al. 1983). The cells were grown in RPMI 1640 medium, supplemented with 10% foetal calf serum (Flow Laboratories, Irvine, Scotland) and antibiotics. For SDS-PAGE and immunoblotting (see below) the cells were trypsinized, seeded and grown in serum-free RPMI medium to avoid contamination of the cellular samples by serum Fn.

Human embryonic tissues were obtained from foetuses aborted on medical grounds because of severe malformations. Adult human tissues and samples of tumours were obtained freshly from surgical operations. Endometrial tissues were obtained by gynaecological scrapings and decidual tissues from chorion villus biopsies. In this study specimens from breast carcinomas, renal adenocarcinomas, thvroid carcinomas, lung adeno-, squamous cell, and small cell carcinomas were studied. Granulation tissue was obtained from a pancreatic pseudocyst. The tissue samples were frozen in liquid nitrogen and stored at −70°C.

Horse, calf and newborn calf sera were obtained from GIBCO (Paisley, Scotland). Human plasma was obtained from the Finnish Blood Transfusion Service (Helsinki, Finland). Fn from the sera or human plasma was isolated by affinity chromatography on gelatin-Sepharose 4B (Pharmacia, Uppsala, Sweden), as described (Engvall & Ruoslahti, 1977). Cellular Fns (cFns), produced by A8387 fibrosarcoma cells or embryonic fibroblasts, were isolated by using the same method from collected spent growth medium of the cells, seeded after trypsinization in serum-free medium to avoid co-purification of serum Fn. cFn from HT-1080 fibrosarcoma cells was as described (Borsi et al. 1987). Protein concentrations were measured according to Lowry et al. (1951) using bovine serum albumin (BSA) as a standard.

Polyclonal rabbit antiserum against human pFn was as described (Vartio et al. 1981a). Monoclonal antibody (N295, IgG₁) against the cell-binding domain of Fn was from Mallinckrodt (St Louis, MO). Two other monoclonal antibodies, 52BF12 (BF) and 52DH1 (DH) were made by immunizing Balb/c mice with fibronectin produced by A8387 cells. Three days after the third immunization the spleens were removed, the cells fused with myeloma cells (NS-1) and hybrid selection was initiated 3 days later by standard techniques (Kohler & Milstein, 1976; see Virtanen et al. 1985). Hybridomas were screened by using a solid-phase immunoassay with pFn and cFn as well as immunofluorescence microscopy with cultured human fibroblasts. Cloning of the hybridomas was done manually by collecting single cells with a micropipette and the clones were initially propagated by using mouse peritoneal macrophages as feeder cells. The BF and DH antibodies were of lgG₁ type, as determined by radial immunodiffusion by using a commercial kit (Miles, Elkhart, IN) according to the instructions of the manufacturer. Ascites fluid production in Balb/c mice with the hvbridoma cells was initiated by standard techniques (Kühler & Milstein, 1976; Virtanen et al. 1985). Immunoglobulins from the ascites fluids were isolated by affinity chromatography on Sepharose-coupled sheep anti-mouse Ig (Cappel Laboratories, Cochranville, PA). The immunoglobulins were eluted by 0·2M-glycine buffer, pH 2·8. The eluate was dialysed against PBS buffer (140 mM-NaCl, 10 mM-sodium phosphate, pH 7·4).

Human neutrophil leucocyte cathepsin G was a gift from Dr Jeremy Saklatvala (Strangeways Research Laboratory, Cambridge, UK). Digestion of purified human pFn (1·5–3·0 mgml⁻¹) with cathepsin G was performed in 50mM-Tris HC1, pH7·5, at 37°C at an enzyme/substrate ratio of 1:200 (w/w) for 30 s or 1 h and the digestion was terminated by the addition of phenylmethanesulphonvl fluoride to give a final concentration of 1 mM (Vartio, 1982). Digestion of HT-1080-cFn with thermolysin (protease type X; Sigma Chemical Co., St Louis, MO) and purification of defined fragments containing the ED sequence of cFn was as described (Borsi et al. 1987).

β-galactosidase-Fn fusion proteins were a kind gift from B. F. Baralle (Sir William Dunn School of Pathology, Oxford, UK). As previously reported (Carnemolla el al. 1987) pXFN-111 contains the ED sequence plus 129 amino acids at its amino-terminal and 158 amino acids at its carboxyterminal end. The fusion protein pXFN-154 is identical with pXFN-111 except that it lacks the entire ED sequence. The fusion protein pXFN-5 contains a part of the gelatin-binding domain of Fn.

Subconfluent human adult and embryonic fibroblasts and A8387 fibrosarcoma cells were supplied with methione-free MEM for 1 h before L-[³⁵S]methionine was added (20 μCiml⁻¹; 1000 Ci mmol⁻¹ ; Radiochemical Centre, Amersham, UK) for 6h. Samples of 1ml of the clarified supernatants were preabsorbed with sheep anti-mouse IgG-Sepharose (Cappel Laboratories) to avoid unspecific binding. Then, the samples were incubated with 10 μl of the BF or DH culture supernatants overnight at 4°C. 40 μl of sheep anti-mouse IgG-Sepharose was added and the incubation was continued for 2h. The Sepharoses were then washed with PBS, the bound proteins were eluted with Laemmli’s sample buffer and subjected to SDS-PAGE. Total proteins were precipitated with ammonium sulphate (176 mg ml⁻¹).

Sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE; Laemmli, 1970) was performed by using 6 or 9% or 4% to 18% gradient vertical slab gels under reducing conditions. After electrophoresis, the gels were either protein stained (Fairbanks et al. 1971), fluorographed (Bonner & Laskey, 1974) or immunoblotted. Immunoblotting (Towbin et al. 1979) was performed by transferring SDS-PAGE-separated polypeptides onto nitrocellulose sheets (type I HAWP filter, Millipore, Bedford, MA). Immunoreactions were detected by using peroxidase-coupled sheep anti-mouse IgG antiserum (Dakopatts, Glostrup, Denmark).

Disposable polystyrene microtitre wells (Labsystems, Helsinki, Finland) received serially diluted concentrations (up to 8 μgml⁻¹) of purified humanpFn or A8387-cFn, diluted in 50 mM-sodium carbonate buffer, pH 9·6. 75 μl volumes of dilutions represented an active immobilization area of 0·65 cm². The wells were covered with adhesive plastic sheet during the overnight incubation (16—20 h) at +4°C. The immobilization was terminated by a 10 min incubation with 200 μl of PBS containing 0·02% Tween 20 (Z-sorbitan monododecanoate poly(oxythane-l,-2-diyl) derivative; PBS/Tween) at room temperature, followed by 1% BSA in PBS.

Solid-phase-bound fibronectins received 75 μl of supernatant monoclonal antibodies (BF and DH) diluted 1:10 in PBS supplemented with 0·05% Tween 20 and 1% BSA for 2h at 37°C. After five washes the wells received 75 μl of 1:800-diluted alkaline-phosphatase-labelled rabbit antimouse IgG antiserum (Orion Diagnostica, Helsinki, Finland) in PBS/Tween/BSA buffer at 37°C for 1 h. After five washes, 75 μl of 0·2% (w/v) disodium p-nitrophenyl phosphate in sodium carbonate buffer was added and the enzyme reaction was terminated after 30 min at 37°C by the addition of 75μl of 1 M-NaOH. A405 was measured with a verticalpathway spectrophotometer.

Cultured cells were fixed in methanol and frozen sections of tissue samples in acetone, both cooled to −20°C, for 10 min. The cells on coverslips or frozen sections were reacted with monoclonal anti-Fn antibodies for 30 min. After washes, they were reacted with FITC-coupled goat anti-mouse IgG antiserum (Cappel Laboratories). For double IIF the tissue sections were first exposed to rabbit anti-laminin antibodies (Bethesda Research Laboratories), followed by FITC-coupled sheep anti-rabbit IgG antiserum (Cappel Laboratories) and were then exposed to the BF or DH antibodies, followed by TRITC-coupled sheep anti-mouse IgG antiserum (Cappel Laboratories). Double immunostaining with TRITC-coupled sheep anti-human Fn-antiscrum (Bethesda Research Laboratories) was done by first exposing the coverslips to the polyclonal Fn antibodies and then to the DH antibody followed by the FITC-conjugate. The specificity of the DH and BF antibodies was also tested in immunofluorescence by absorption experiments: preabsorption of the culture supernatants with fibroblast cFn (50 μg ml⁻¹; 2h at room temperature) or A8387-cFn (50 μg ml⁻¹) completely abolished the reactivity of both monoclonal antibodies on tissue sections and cultured cells whereas preabsorption with pFn only abolished the reactivity of the BF antibody and did not have any effect on the activity of the DH antibody. Different fluorochrome-coupled conjugates used in the study did not give any positivity when applied without the primary antibodies. A Zeiss Universal microscope, equipped with an epi-illuminator III RS and filters for FITC and TRITC fluorescence, was used. For peroxidase-immunostaining, the specimens were first reacted with culture supernatants, and after washing were processed by the Vectastain^® procedure (Vector Laboratories, Burlingame, CA). The peroxidase reactions were developed by using the diamine-pvrocatechol chromogen (Sigma; Hanker et al. 1977).

Fns from various sources were immunoblotted by the 52BF12 (BF) and 52DH1 (DH) monoclonal antibodies. The BF antibody reacted with plasma and serum Fns at the positions of about M_r 220 000 and with both cFns which had a slightly slower electrophoretic mobility (Fig. 1A). On the contrary, the DH antibody reacted only with cFns (Fig. IB). Immunoblotting of a larger quantity (20–40 μg) of human pFn at the same dilution of the DH antibody showed a distinct narrow immunoperoxidase reaction band at the top of the bulk of the protein (Fig. 1C).

Proteins from different cultured cell lines, grown under serum-free conditions, were also immunoblotted with the BF and DH antibodies. Both antibodies had similar reactions at the position of reduced cFns (M_r about 220000; not shown) from both adult and embryonic fibroblasts as well as in A8387 fibrosarcoma cells and precipitated cFn from [³⁵S]methionine-labelled culture media of these cells (Fig. IF).

In enzyme immunoassay the BF antibody reacted with both humanpFn (Fig. ID) and A8387-cFn (Fig. IE) in a dose-dependent way, while the DH antibody reacted only with the latter (Fig. IE). No reaction occurred if the monoclonal antibodies were omitted or were replaced by an irrelevant hybridoma supernatant (PKK1 cytokeratin antibody; see Virtanen et al. 1985).

Human pFn was digested by cathepsin G and the fragments were immunoblotted by the N295, BF or DH antibodies. The immunoperoxidase reactions of the BF antibody (Fig. 2A) were identical with those of the N295 antibody against the cell-binding site of Fn (Fig. 2B). Polyclonal anti-Fn gave a different reaction pattern (Fig. 2C) and the DH antibody, as expected, gave no reaction at all. Isolated IgG of the BF antibody inhibited attachment of human fibroblasts on substrata coated with pFn similarly as described previously (e.g. see Pierschbacher et al. 1981; data not presented).

In immunoblotting the DH antibody reacted with the fusion protein pXFN-111, which contained the ED sequence (Fig. 3B, lane 3). The antibody did not react with the pXFN-154 that is otherwise identical but lacks the ED sequence or with the pXFN-5 used as a negative control. Furthermore, the DH antibody recognized the thermolysin cFn fragments which also carried the ED sequence (Fig. 3B, lanes 2 and 6). Corresponding protein-stained samples are shown in Fig. 3A. When the ED sequence was removed by thermolysin digestion (see Borsi et al. 1987) from the purified fragments, no reactivity with the DH antibody was seen (not shown).

In indirect immunofluorescence (HF) microscopy of human embryonic and adult fibroblasts both the BF (adult; Fig. 4A) and DH antibodies (adult; Fig. 4B) gave a typical fibrillar pericellular Fn-like reaction, in addition to perinuclear cytoplasmic staining. The BF (Fig. 4C) and DH antibodies (Fig. 4D) showed a reminiscent positivity in tiny pericellular fibres in A8387 fibrosarcoma cells. Primary cultures of human umbilical vein endothelial cells showed bright positivity with the BF (Fig. 4E) and DH antibodies (Fig. 4F). Double immunostaining with polyclonal Fn antibodies and the DH antibody gave complete codistribution (not shown).

In immunoperoxidase staining of frozen sections of adult human tissues, the BF antibody showed positivity widely in the perimysium of the muscle (Fig. 5A), and diffusely in the intertubular and perivascular areas of kidney (Fig. 5C) as well as in glomerular mesangium. With the DI I antibody the endothelial cells of larger blood vessels showed EcFn abundantly in both muscle (Fig. 5B) and kidney (Fig. 5D). Additionally, the glomerular mesangium was positive for EcFn (Fig. 5D). In the embrv-onic kidney, a prominent peroxidase reaction was seen with the DF1 antibody in tubular basement membranes (Fig. 5E) and in the mesangial area of developing glomeruli (Fig. 5F).

Double IIF stainings with the DH and anti-laminin antibodies showed codistribution of EcFn-and lam-inin-positivities in tubular and capillary basement membranes of embryonic kidney (Fig. 6A,B). In adult kidney, only occasional tiny positive fibrils could be revealed in the intertubular area with the DH antibody (Fig. 6C) and the laminin-positive basement membranes lacked EcFn (compare Fig. 6C and D). A similar double-staining using the BF antibody revealed, however, a strong, diffuse interstitial Fn-positivity and only faint staining of laminin-positive basement membranes (compare Fig. 6E and F). In early embryonic muscle (14 weeks) EcFn was revealed around the developing myoblasts (Fig. 7A) whereas in older embryos (22 weeks), with more mature mvo-tubes, it was confined to endothelial cells of larger blood vessels (Fig. 7B). In embryonic (Fig. 7C) but not in adult (Fig. 7D) cerebellum the DH antibody reacted brightly with the capillaries. In embryonic liver EcFn could be revealed in the sinusoidal areas (Fig. 7E) whereas in adult liver the DH antibodypositivity was only seen in larger blood vessels (Fig. 7F). Similarly, in normal human endometrium only the blood vessel endothelia contained EcFn (not shown) while in decidual tissue, i.e. transformed endometrium, all the rounded stromal cells were invested by this form of the protein (Fig. 8A). Also the granulation tissue associated with a pancreatic pseudocyst was prominently positive for EcFn (Fig. 8B).

In various human carcinomas studied, both the Dll and BF antibodies revealed abundant Fn-positivity in the stroma (DH; renal carcinoma, Fig. 8D; breast carcinoma, Fig. 8F), the tumour cells being virtually negative in all cases (see the corresponding phasecontrast views, Fig. 8C and E).

As it has been established that Fns may be composed of alternative subunit polypeptides (see Introduction), it is evident that different forms of the protein are expressed in different situations. Structural and functional differences have already been shown between pFns and cFns (Yamada & Akiyama, 1984) although there has not been a means of distinguishing between them accurately, e.g. at the tissue level. We have now established a monoclonal antibody which recognizes the ED sequence of cFn and use it to study the distribution of this domain-containing subset of the protein in various tissues.

Both cultured embryonic and adult fibroblasts as well as malignant mesenchymal cells secreted and deposited Fn, recognized equally with the DH and BF antibodies. The DH antibody reacted, however, exclusively with cFn as judged by immunoblotting. When larger quantities of pFn were analysed, the DEI-antibody-positivc polvpeptide band could be seen at the top of the bulk of the protein, indicating nonidentity with pFn. Apart from selective reactivity towards at least partially denatured proteins in immunoblotting, the DH and BF antibodies reacted selectively also in El A. This indicates that the differences in the reactivities are not due to conformational changes between denatured or native proteins.

The BF antibody had an antigenic determinant at or close to the cell-binding site of human pFn as judged by immunoblotting of fragments of the protein and comparison of the immunoperoxidase reaction pattern with that obtained by another monoclonal antibody (N295) towards this site. Furthermore, the BF antibody inhibited cell attachment and spreading on pFn-coated substrata.

The DH antibody reacted with the fusion protein and thermolvsin cFn fragments which contained the ED sequence but not with those lacking it. 1’his indicates that the antigenic determinant of the DH antibody is located within the ED sequence. Thus, the DEI antibody is very similar to a monoclonal antibody IST-9 that has been reported to have the same specificity (Borsi et al. 1987; Carnemolla et al. 1987).

Fn is lost when primitive mesenchvmal cells differentiate into more specialized cells (Wartiovaara et al. 1976) but is again abundant in developing tissues (Dessau et al. 1978; Hassel et al. 1979) as well as in reactive rheumatoid synovium (Vartio et al. 1981b) and connective tissue stroma of carcinomas (Stcnman & Vaheri, 1981). It has been shown that Fn in adult basement membranes is only an extrinsic component and does not contribute to its main structure (Obcrley et al. 1979; Martinez-Hernandez & Amenta, 1983). Furthermore, Oh et al. (1981) and Hayman & Ruos-lahti (1979) have shown that pFn becomes deposited both in vitro in pericellular matrices and in vivo in tissues. Thus, Fn-immunoreactivitv found in tissues by using polyclonal or monoclonal antibodies recognizing both pFn and cFn may, in fact, mainly reveal a deposited plasma form of the protein and our immunostaining studies with the BF and DH antibodies provide further evidence for this.

Accordingly, the BF antibody, which reacted with both pFn and cFn, showed immunoreactivity in adult tissues, including the interstitial areas of kidney and muscle tissue. However, in line with many earlier studies (Oberley et al. 1979; Boselli et al. 1981; Martinez-Hernandez & Amenta, 1983), no clear enrichment of Fn could be seen in basement membranes of adult tissues. The DH antibody, lacking affinity to pFn, failed to reveal positivity in most domains of adult tissues which were positive for the BF antibody. It appeared that in adult human tissues studied, only the endothelium of larger blood vessels contained abundant EcFn as defined with the DH antibody. Previous studies have shown that pFn is most probably produced by hepatocytes (Tamkun & Hynes, 1983), which were negative for EcFn. Thus, the present results suggest that the small amount of EcFn in the circulation may be produced by endothelial ceils.

In embryonic tissues the DH antibody showed EcFn-positivity in co-localization with laminin in the basement membranes of kidney tubules and glomeruli at various developmental stages. Similarly, embryonic myotubes showed basement membrane-like EcFn-positivity. These results suggest that EcFn may be an integral, but transient, component of basement membranes during embryonic development and may thus have a role in tissue modelling as suggested by Hay (1983; see also Brownell et al. 1981), possible via a direct influence on cellular cytoskeleton (see Virtancn et al. 1982; Woods et al. 1986). Accordingly, the present results showed that capillaries in developing, but not in adult, tissues arc strongly positive for EcFn. In this respect, it is of interest that in chicken embryos the receptor complex for Fn appears to be very similarly expressed, being abundant in developing tissues and declining upon tissue maturation (Chen et al. 1986; Duband et al. 1986). The distribution of EcFn also indicates that at least this subset of cFn is different from tenascin/cytotactin (Chiquet-Ehris-mann et al. 1986): the absence of cFn in embryonic cerebellum (cf. Grumet et al. 1985; Crossin et al. 1986), adult muscle (cf. Chiquet & Fambrough, 1984a,b) and the highly specific expression of EcFn in adult larger blood vessels and glomerular mesangium (cf. Chiquet & Fambrough, 1984b; Grumet et al. 1985; Crossin et al. 1986).

Results with the DH antibody indicated that Fn-positi vity, typically found in the stroma of the malignant tumours, represents mainly EcFn and not deposited pFn, as suggested recently (Lorke & Moller, 1985). This result suggests that malignant cells may induce synthesis and deposition of EcFn by their stromal cells (see Liotta, 1982). Thus, adult mesenchymal cells can be induced to deposit EcFn as also implicated by the expression of the protein in granulation tissue and decidual stromal cells. These results provide further evidence for the similarities between fibroblasts in granulation and embryonic tissues and tumour cell stroma (cf. Gabbiani et al. 1976; Lagace et al. 1985; Schor et al. 1986).

The present results indicate that EcFn may have a specific role in connective tissue formation in developing and reactive tissues as well as in the initiation of the basement membrane formation. On the other hand, the integrity of mature connective tissue may in most cases involve pFn. Thus, pFn and EcFn seem to be separate functional entities. It would be of interest to find out the factors which lead to differential expression of different forms of Fn in various physiologic and pathologic situations.

The skilful technical assistance of Ms Tuija Järvinen, Ms Pipsa Kaipainen, Ms Saija Roine, Ms Anna Maria Siivonen and Ms Raili Taavela is kindly acknowledged. This study was supported by a research contract with the Academy of Finland and by grants from the Sigrid Jusélius Foundation, the Finnish Cancer Foundation, the Emil Aaltonen Foundation and the University of Helsinki. The work was also partially supported by a grant from the Italian Research Council Progetto Finalizzato Oncologia (to L.Z.).