Integrin subunits β1C-1 and β1C-2 expressed in GD25T cells are retained and degraded intracellularly rather than localised to the cell surface

Integrins are heterodimeric transmembrane glycoproteins composed of α and β subunits that form cell surface receptors, which bind a wide range of ligands such as extracellular matrix and cell surface proteins. The β1 subunit can dimerise with at least 12 α subunits and form receptors for different ligands such as fibronectin (FN), collagens, vitronectin and laminins. The pre-mRNA of the integrin β1 subunit can be alternatively spliced resulting in five variant cytoplasmic domains denoted β1A, β1B (Altruda et al., 1990), β1C-1 (Languino and Ruoslahti, 1992), β1C-2 (Svineng et al., 1998) and β1D (van der Flier et al., 1995; Zhidkova et al., 1995). These variants have been suggested to generate different intracellular signals. The β1D variant has been found to be conserved in species such as human, rat and mouse and is expressed only in striated muscle cells where it replaces the β1A variant upon differentiation from myoblasts to myotubes. In vitro studies have shown that the intracellular protein talin, which is involved in linkage of the integrin cytoplasmic tails to actin filaments, binds more strongly to β1D than to β1A cytoplasmic tails (Belkin et al., 1997; Pfaff et al., 1998), suggesting a function for the shift from β1A to β1D as a way to improve the strength by which the integrin binds the cytoskeleton in muscle cells. Both the β1B and the β1C variants have only been found in humans and are not encoded by the mouse β1 gene, raising questions as to their functional importance (Baudoin et al., 1996; Svineng et al., 1998). Integrin heterodimers composed of β1B are unable to bind any ligand due to an inactive conformation of the extracellular domain (Balzac et al., 1994; Retta et al., 1998). These integrins have been suggested to regulate negatively the functions of β1A integrins. The β1C variant has been reported to be expressed at the protein level in HEL cells, endothelial cells, liver and epithelial cells of the prostate. Expression of β1C has been found to be down-regulated in prostate carcinoma compared with the surrounding normal tissue, suggesting that there is a correlation between loss of β1C expression and neoplasia (Fornaro et al., 1996). Immunohistochemistry has also been used to demonstrate expression of β1C in nonproliferative, differentiated epithelia in normal prostate glands and liver bile ducts, indicating that undifferentiated and proliferating epithelial cells down-regulate the expression of β1C (Fornaro et al., 1998). The anti-β1C antibodies used in these studies were raised against peptides that are common for both β1C-1 and β1C-2, and therefore cannot distinguish between the two β1C forms. Transfection of several cell lines with β1C-1 cDNA resulted in inhibition of entry into the S-phase of the cell cycle and thereby inhibition of cell proliferation (Fornaro et al., 1995, 1998, 1999; Meredith et al., 1995, 1999). The specific peptide sequence responsible for this effect has been mapped to a stretch of 18 amino acids present in both β1C-1 and β1C-2. Interestingly, the growth inhibitory effects of the β1C-1 cytoplasmic sequence were observed even if it was expressed as a cytosolic protein (Meredith et al., 1999).

We have previously identified a splice variant of the integrin β1 subunit, β1C-2, that differs from β1C-1 by a deletion of 6 amino acids in the middle of the cytoplasmic tail as a result of utilisation of a more 3′ splice acceptor site in exon C. By using RT-PCR we could identify the mRNA for both β1C-1 and β1C-2 together with β1A in all human primary cells and cell lines that we investigated. We also identified exon C as part of a complete Alu element located within the human integrin β1 gene (Svineng et al., 1998). Alu elements have a copy number of approximately 500,000 in a haploid human genome and are a primate specific retroposon, explaining the lack of exon C and β1C-1 and β1C-2 in the mouse.

In order to investigate the specific functions of β1C-1 and β1C-2 we have performed expression studies in β1 deficient GD25T cells. The results presented here show that only small amounts of the synthesised β1C-1 and β1C-2 were localised to the cell surface, whereas most of the β1C-1 and β1C-2 proteins were found to be retained intracellularly. The vast majority of these immaturely glycosylated β1C-1 and β1C-2 chains were not associated with any α-subunits and were intracellularly degraded.

Hamster anti-mouse integrin β1 monoclonal antibody (mAb) HMβ1-1, fluorescein isothiocyanate (FITC)-conjugated hamster anti-rat integrin β1 mAb Ha2/5, rat anti-mouse integrin β1 mAb 9EG7, hamster anti-mouse integrin β3 mAb 2C9.G2, hamster anti-mouse integrin αV mAb H9.2B8, and rat anti-mouse integrin α5 mAb 5H10-27 were all purchased from PharMingen (San Diego, CA, USA). The polyclonal rabbit anti-rat integrin β1 antiserum was prepared in our laboratory as described previously (Bottger et al., 1989; Johansson and Hook, 1984). Secondary antibodies used in flow cytometry were FITC-conjugated goat anti-rabbit immunoglobulin (IgG), FITC-conjugated goat anti-Armenian hamster IgG, and FITC-conjugated anti-rat IgG, and were all purchased from Jackson ImmunoReseach Laboratories (West Grove, PA, USA). Secondary antibodies used in immunostaining were FITC- or Cy3-conjugated goat anti-rabbit IgG, and Cy3-conjugated goat anti-Armenian hamster IgG, and were all of multiple labelling quality and were purchased from Jackson ImmunoReseach Laboratories (West Grove, PA, USA). GRGDS peptide was purchased from Bachem Feinchemikalien AG (Bubendorf, Switzerland). Protein A Sepharose CL-4B was obtained from Pharmacia Biotech (Uppsala, Sweden), and endo-β-N-acetylglucosaminidase H (EndoH, EC 3.2.1.96) was obtained from Boehringer Mannheim Biochemica (Mannheim, Germany).

The GD25 cell line and its subclone GD25T, which was established by stable transfection with the Tet repressor-encoding vector (pUHD15-1hyg), have been described previously (Fässler and Meyer, 1995; Fässler et al., 1995; Svineng et al., 1998; Wennerberg et al., 1996). The GD25 cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% (v/v) fetal calf serum, 2 mM L-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin and 2.5 μg/ml fungizone (GibcoBRL). GD25T cells were cultured in the same medium with the addition of 200 μg/ml hygromycinB (CLONTECH Laboratories Inc., Palo Alto, CA, USA). The GD25T cells were further stably transfected with the pUHD10-3 expression vector containing a tetracycline and doxycycline hydrochloride (Dox) regulated CMV promoter upstream of cDNAs for β1A (pTetβ1A), β1C-1 (pTetβ1C-1) and β1C-2 (pTetβ1C-2). The cloning of these expression vectors has been described previously (Svineng et al., 1998). Briefly, pTetβ1A contains a complete mouse cDNA, while both pTetβ1C-1 and pTetβ1C-2 contain a hybrid cDNA where the extracellular domain is encoded by the mouse cDNA, while the transmembrane and cytoplasmic domains are encoded by the human β1 cDNA (identical to mouse at the amino acid level). The expression vectors were transfected into the GD25T cells either by using the Tfx-20 transfection reagent (Promega Corporation, Madison, WI, USA) or by electroporation using a Bio-Rad Gene Pulser (Bio-Rad Laboratories, Hercules, CA, USA). The Tfx-20 transfection reagent was used according to the manufacturer’s recommendations and the cells were co-transfected with 10 μg pTetβ1 and 0.5 μg pPGKpuro-vectors using 4.5 μl Tfx-20 per μg DNA. The cells were grown in medium containing 0.1 μg/ml Dox (Sigma Chemical Co., St Louis, MO, USA) in order to keep the expression of the integrin β1 turned off (Gossen et al., 1994; Gossen and Bujard, 1992). GD25T cells that were transfected by electroporation were detached by trypsin/EDTA treatment and resuspended in 10 mM Hepes in phosphate buffered saline (PBS) at a concentration of 1×10⁷cells/ml. 100 μl of the cell suspension were mixed with a total of 40 μg pTetβ1 and 2 μg pPGKpuro-vectors electroporated at 25 μF and 800 V and seeded in medium containing 0.1 μg/ml Dox. In both cases, medium containing 10 μg/ml puromycin (Sigma Chemical Co., St Louis, MO, USA) was added to the cells 2 days post-transfection in order to select for positively transfected cells.

GD25T cells were seeded one day prior to transfection. Equal amounts of expression-vectors were mixed with 7 μl Superfect (QIAGEN GmbH, Germany) transfection reagent per μg DNA according to the manufacturer’s instructions. The Superfect:DNA mixtures were removed after 4 hours and the cells grown for a total of 48 hours before analysis by immunoprecipitation and immunoblotting (see below).

Integrin expression on the cell surface was determined by FACScan analysis (Becton Dickinson and Co., Mountain View, CA, USA). The stably transfected β1 clones were harvested by trypsin-EDTA treatment and resuspended in PBS containing 10% goat serum for 30 minutes at 4°C. Internalisation of the cell surface integrins was prevented by keeping the cells on ice in the presence of 0.001% azide during the entire procedure. Approximately 1×10⁶cells were incubated for 1 hour with primary antibody (5 μg/ml) in PBS containing 2% goat serum and 0.001% NaN₃(FACS-PBS). In experiments including RGD containing peptides, the cells were incubated with 1 mg/ml GRGDS peptide in FACS-PBS for 45 minutes before the addition of primary antibody. After being washed twice with ice-cold FACS-PBS, the cells were incubated for 30 minutes with secondary antibody (50 μg/ml) in FACS-PBS. The cells were then washed twice and resuspended in 500 μl FACS-PBS containing a final concentration of 1 μg/ml propidium iodine (Sigma Chemical Co., St Louis, MO, USA). The cells (10000 cells/sample) were analysed in a FACScan (Becton Dickinson and Co., Mountain View, CA, USA) equipped with 5-W argon laser at 488 nm. The propidium iodine positive cells were presumed to be dead or damaged and excluded from the analysis.

Total RNA was purified from cell lines by the RNeasy kit (QIAGEN GmbH, Germany) and 10 μg of each sample was mixed with 3 volumes of NorthernMax Gel Loading solution (Ambion, Inc., Austin, TX, USA) and 0.5 μg ethidium bromide. After denaturing by heating at 65°C for 15 minutes, the RNA samples and a 0.24-9.5 kb RNA Ladder (GibcoBRL) were separated in a formaldehyde/MOPS denaturing agarose gel. The RNA was partially hydrolysed by soaking the gel in 0.05 M NaOH for 25 minutes and neutralised by 0.1 M Tris, pH 7.5, 0.15 M NaCl for 10 minutes and subsequently transferred to a Hybond-N⁺nylon membrane (Amersham Pharmacia Biotech UK Limited, Buckinghamshire, UK) in 20× SSC by vacuumblotting (VacuGene, LKB/Pharmacia, Sweden). After transfer, the RNA was cross-linked to the membrane by UV radiation followed by pre-hybridisation in ExpressHyb Hybridisation solution (CLONTECH Laboratories Inc., Palo Alto, CA, USA) at 68°C for 30 minutes. A mixture of a mouse integrin β1 specific probe and a β-actin probe, both labelled with [α-³²P]dCTP using the Random Primed DNA Labelling Kit (Boehringer Mannheim Biochemica, Mannheim, Germany), was denatured by heating to 95°C for 5 minutes before it was added to the ExpressHyb Hybridisation solution and incubated for 1 hour at 68°C. The membrane was washed extensively in 2× SSC, 0.05% SDS at room temperature for 40 minutes followed by a 40 minute wash in 0.1× SSC, 0.1% SDS at 50°C. The membrane was sealed in a plastic bag and exposed to a Kodak X-Omat film at –70°C.

Cells were seeded in 25-75 cm²flasks one day prior to labelling and were approximately 80% confluent at the day of labelling. The cells were washed twice with PBS and incubated with DMEM supplemented with 25 mM Hepes, 2 mM L-glutamine, 1 mg/ml bovine serum albumin, and 2 μM L-methionine (DMEM-2μM Met) for 30 minutes before the medium was replaced with 2-4 ml DMEM-2μM Met containing 100 μCi [³⁵S]methionine (Tran³⁵S-label, ICN Pharmaceuticals, Irvine, CA, USA) and further incubated for various times. The cells were subsequently lysed in 2% Triton X-100 in 10 mM Tris-HCl, pH 7.4, containing a mix of protease inhibitors (PI) consisting of 2 mM phenylmethylsulfonyl fluoride, 2 mM N-ethylmaleimide and 1 μg/ml pepstatin A. The material was scraped off the plastic and insoluble material was removed by centrifugation at 15,000 gfor 5 minutes before the supernatant was pre-cleared by incubation with 40 μl of a 50% slurry of Protein A Sepharose CL-4B (Pharmacia, Uppsala, Sweden) in Tris buffered saline (TBS) containing 0.1% Triton X-100. After 2 hours of end-over-end incubation at 4°C the Protein A Sepharose was collected and the supernatant incubated with approximately 2.5 μg anti-β1 mAb (HMβ1-1) for a minimum of 2 hours at 4°C. The immuno-complex was then precipitated by addition of 40 μl of a 50% slurry of Protein A Sepharose CL-4B. After an additional 30 minute incubation, the Protein A Sepharose was washed three times with 0.5 M NaCl, 0.1% Triton X-100 and PI in TBS, three times with 0.1% Triton X-100 and PI in TBS and three times with TBS containing PI. Samples that were deglycosylated by EndoH treatment were resuspended in 40 μl 50 mM sodium acetate, pH 5.6, containing PI and 0.16 mU EndoH and incubated at 37°C overnight. Immunoprecipitated material was released by boiling in sample buffer containing 4% SDS and separated by electrophoresis in a 7% polyacrylamide gel. After SDS-PAGE the gel was fixed in isopropanol:water:acetic acid (25:65:10) for 30 minutes, stained with Coomassie brilliant blue, and destained in metanol:water:acetic acid (25:67.5:7.5). The gel was then incubated in Amplify (Amersham Pharmacia Biotech UK Limited, Buckinghamshire, UK) for 30 minutes before vacuum drying at 80°C and exposure to Fujifilm RX at –70°C for 2-14 days.

After SDS-PAGE, the gel was soaked in blotting buffer for 30 minutes before transfer to a Protran nitrocellulose membrane (Schleicher & Schuell, Dassel, Germany). The membrane was blocked in 5% fat-free dried milk in TBS containing 0.1% Tween-20 (TBS-Tween) for 1 hour at room temperature or at 4°C overnight, followed by incubation with a rabbit polyclonal anti-β1 antiserum diluted 1:100 in TBS-Tween for 1 hour at room temperature. After washing, HRP-conjugated anti-rabbit antibody (Amersham Pharmacia Biotech UK Limited, Buckinghamshire, UK) diluted 1:5000 in TBS-Tween was incubated with the membrane for 45 minutes followed by detection using the ECL-kit (Amersham Pharmacia Biotech UK Limited, Buckinghamshire, UK).

Cells were seeded on coverslips coated with 10 μg/ml human FN in PBS or in DMEM containing 10% fetal calf serum on non-treated coverslips. The following day the cells were washed twice with PBS and fixed in freshly made 2% paraformaldehyde in PBS for 8 minutes. The cells were washed in PBS containing 0.2% NaN₃(PBS-azide) and permeabilised with 0.5% Triton X-100 in PBS-azide for 20 minutes at room temperature and then subjected to three 10 minute washes with 0.05% Tween-20 in PBS-azide (Tween-PBS-azide). The cells were blocked by incubation in 10% goat serum in PBS-azide for 1 hour and further incubated with primary antibody (rabbit anti-β1 IgG (25 μg/ml) and hamster anti-β3 mAb (5 μg/ml)) in 10% goat serum in PBS-azide for 1 hour. The cells were then subjected to three 10 minute washes with Tween-PBS-azide and subsequently incubated with secondary antibody (FITC anti-hamster (50 μg/ml) and Cy3 anti-rabbit or -hamster (0.5 μg/ml)) in 10% goat serum in PBS-azide for\ 30 minutes. Finally, the cells were subjected to three 10 minute washes with Tween-PBS-azide, one 10 minute wash with PBS-azide and then mounted on slides with Fluoromount-G (Southern Biotechnology Associates, Inc., Birmingham, AL, USA) and sealed with nail polish. Staining of non-permeabilised cells followed the same protocol except that Tween-20 was omitted from the solutions.

In order to investigate the function of the β1C-1 and β1C-2 splice-variants compared to the β1A variant of the integrin β1 subunit, the various cDNAs were expressed in the mouse GD25T cell line. These cells are deficient in β1 expression due to a knock-out of the integrin β1 gene (Fässler et al., 1995; Wennerberg et al., 1996). Previous attempts to express the β1C-1 and β1C-2 isoforms in GD25 cells had not been successful using a conventional constitutively active promoter (Svineng et al., 1998). Since we suspected that this was due to the presumable growth inhibitory effect reported for the β1C-1 variant (Fornaro et al., 1995; Meredith et al., 1995), the inducible expression system developed by Gossen and Bujard was used in this study (Gossen et al., 1994; Gossen and Bujard, 1992). The GD25 cells were stably transfected with an expression vector for the Tet-repressor protein and named GD25T (Svineng et al., 1998). In the presence of Dox, the Tet repressor protein is unable to bind and activate transcription from the CMV promoter located upstream of the inserted β1 cDNAs. Therefore, when the GD25T cells were transfected with the various β1 expression-constructs, the cells were grown in the presence of 0.1-5 μg/ml of Dox. Puromycin resistant clones, named GD25Tβ1A, GD25Tβ1C-1 and GD25Tβ1C-2, were isolated and the β1 expression induced by removal of Dox from the medium. A total of 12 β1A, 20 β1C-1, and 13 β1C-2 clones were tested for β1 cell surface expression by flow cytometry. All GD25Tβ1A clones were found to express β1A protein on the cell surface, whereas all GD25Tβ1C-1 and surface was present in an active conformation and that a similar number of molecules could be activated by the GRGDS peptide. Due to the fact that the Ha2/5 mAb was directly conjugated to FITC while the 9EG7 mAb was used together with a FITC-conjugated secondary antibody, the MFI values obtained with the two antibodies can not be directly compared with each other. Nevertheless, the MFI values for the Ha2/5 and 9EG7 epitopes on the β1A clone (165 and 266, respectively) and on the β1C-1 clone (16 and 17, respectively), indicate that a smaller fraction of the total cell surface β1C-1 molecules were in an active confirmation after GRGDS stimulation than the corresponding fraction of β1A.

In order to investigate if the lower cell surface expression was due to lower mRNA levels in the GD25Tβ1C-1 and GD25Tβ1C-2 clones, northern blot analysis was performed. The probe used was derived from the β1 cDNA that encodes the extracellular domain, and the amount of loaded total RNA was determined by hybridisation with a β-actin probe (Fig. 4). As expected, the GD25T cells contained no β1-transcript (data not shown), while a band of approximately 3.5 kb could be identified in the GD25Tβ1-clones (Fig. 4). Typically, the tested GD25Tβ1C-1 and GD25Tβ1C-2 clones contained equal or greater amounts of β1-transcript compared to the tested GD25Tβ1C-2 clones showed similar or only slightly higher mean fluorescence intensity (MFI) values compared to the GD25T cells (Fig. 1). Even after several weeks of growth in the absence of Dox, the cell surface levels of β1C-1 and β1C-2 did not increase significantly (data not shown). Similar results were obtained using one FITC-conjugated rat anti-β1 mAb, one hamster anti-β1 mAb and one rabbit polyclonal anti-β1 antibody and two different secondary FITC-conjugated antibodies (data not shown). Nevertheless, GD25Tβ1C-1 and GD25Tβ1C-2 clones with higher MFI values than the GD25T cells were chosen for further studies. Addition of Dox reduced the MFI value close to that for the GD25T cells, proving that the increased MFI value indeed correlated with expression from the transfected β1-cDNAs (Fig. 2).

In order to investigate the activation status of the expressed β1, the anti-β1 mAb 9EG7 (Bazzoni et al., 1995; Lenter et al., 1993), which recognises an epitope on β1 in an active confirmation, was employed (Fig. 3). Using the 9EG7 mAb, we found that a fraction of the β1C-1 localised to the cell GD25Tβ1A clones. This excludes the possibility that the difference in cell surface expression was due to lower β1-mRNA production in the GD25Tβ1C-1 and GD25Tβ1C-2 clones compared to the GD25Tβ1A clones.

The results of the northern analysis prompted us to investigate the biosynthesis of the β1 chains in the various clones. This was performed using metabolic labelling followed by immunoprecipitation with an anti-β1 antibody. As expected, no β1 protein was immunoprecipitated from the GD25T cells, whereas the tested GD25Tβ1C-1 andGD25Tβ1C-2 clones synthesised similar amounts of β1 protein compared to the GD25Tβ1A clones (Fig. 5). The β1 chain contains 11 potential N-linked glycosylation sites (Argraves et al., 1987), and the transition from the pre-β1 form to the maturely glycosylated form increases the size from approximately 110 to 120 kDa (Fig. 5A). Removal of immature N-linked glycan chains by digestion with the endoglucosidase EndoH (Kobata, 1979) reduces the size to approximately 80 kDa, while the maturely glycosylated β1 band of approximately 120 kDa is not cleaved by EndoH (Fig. 5A). Using continuous metabolic labelling for 18 hours, it was shown that approximately half of the β1A chains were in the EndoH-sensitive pre-β1 form of approximately 110 kDa, while the other half of approximately 120 kDa were resistant to EndoH treatment and therefore maturely glycosylated (Fig. 5A). The GD25Tβ1C-1 and GD25Tβ1C-2 clones contained as much or more pre-β1 chains as the GD25Tβ1A clones, but no or very little maturely glycosylated β1 chains (Fig. 5A). In order to determine if the bands of approximately 120 kDa in the EndoH treated β1C-1 and β1C-2 samples were mature β1 or EndoH-cleaved α subunits, anti-β1 mAb immunoprecipitated material was cleaved by EndoH, immunoblotted and probed with an anti-β1 antibody (Fig. 5B). These results support the latter explanation since only very small amounts of the 120 kDa β1 band was detected in the blot of the β1C clones. In order to determine if the produced β1C-1 and β1C-2 chains were accumulated or degraded, pulse chase experiments were performed. Metabolic labelling for 2 hours and chase for up to 20 hours showed that the β1A chains became maturely glycosylated and co-immunoprecipitated several α-subunits, while the β1C-1 chains were degraded without becoming maturely glycosylated and completely disappeared after 20 hours of chase (Fig. 6). Continuous metabolic labelling for 18 hours of several different β1C-1 and β1C-2 clones verified that none expressed maturely glycosylated β1 chains to the same extent as the β1A clones (data not shown). Similar results were also obtained in short-term expression studies. Equal amounts of pTetβ1A, pTetβ1C-1, pTetβ1C-2 and control expression vectors were transiently transfected into GD25T cells. After 48 hours, β1 expression was analysed by β1 immunoprecipitation and immunoblotting (data not shown). These results confirmed that the 120 kDa mature β1 could only be seen in the β1A transfected cells, while the pre-β1 form of 110 kDa could be identified in all cells, except in cells transfected with the control vector. In several independent transient expression experiments, the amount of β1A protein was repeatedly found to be higher than both β1C-1 and β1C-2 despite the transfection of equal amounts of cDNA. This further indicates that the β1C chains are more rapidly degraded than the β1A chains. Thus, the β1C chains are produced in large amounts, but remain in the immaturely glycosylated form of approximately 110 kDa, which according to the pulse-chase experiment is degraded without becoming maturely glycosylated. Conversely, β1A chains become maturely glycosylated and the total amount is not markedly altered after 20 hours of chase.

Previous studies have revealed that GD25 cells transfected with β1A expressed α3β1, α5β1, and α6β1 on the cell surface (Wennerberg et al., 1996). Analysis of the α5 cell surface expression was done by flow cytometry, and showed that the relative amount of α5 correlated with the amount of β1 on the cell surface of all GD25Tβ1 clones (Fig. 2). The MFI value for α5 on the cell surface of both GD25Tβ1C-1 (MFI=5.5) and GD25Tβ1C-2 (MFI=7.0) cells was slightly higher than the background level measured on the GD25T cells (MFI=2.7), suggesting that at least some of the β1C-1 and β1C-2 subunits on the cell surface were associated with the α5 subunit. The MFI values measured for α5 were also reduced in the presence of Dox in accordance with the reduced levels of β1 expression under these conditions. As a control, the cells were also tested for the cell surface expression of the αV subunit (Fig. 2), which was found to have high expression levels in all clones compared with the background levels in samples without primary antibody.

Using metabolic labelling followed by immunoprecipitation with an anti-β1-antibody, bands of approximately 140-160 kDa could be co-immunoprecipitated from the GD25Tβ1A cells (Fig. 5A). These bands most likely represent α-subunits, and bands migrating at the same distance as α3, α5 and α6 could be identified (data not shown). The relative intensity of the α- band was approximately the same as the mature EndoH-resistant β1 band after continuos metabolic labelling (Fig. 5A). Only faint bands in the area corresponding to where α-subunits would be expected to be seen was identified in the GD25Tβ1C-1 and GD25Tβ1C-2 lanes (Fig. 5A). The bands that co-immunoprecipitated with the β1C-2 chains were cleaved by EndoH and therefore represents immaturely glycosylated α-subunits. After longer exposure times (data not shown), α-bands in the GD25Tβ1C-1 lane resistant to EndoH could be identified; however, several different GD25Tβ1C-1 and GD25Tβ1C-2 clones have been subjected to β1-immunopreciptiation, and the amount and size of the co-immunoprecipitated α-subunits varied (data not shown). These results suggest that only a very small portion of the produced β1C-1 and β1C-2 chains are dimerised with any α-subunit after 18 hours of continuous metabolic labelling.

Additional bands of both lower and higher molecular masses could be detected in β1-immunoprecipitates from all clones. In particular a band of approximately 90 kDa, which correlated in intensity with the amount of immaturely glycosylated β1, was co-immunoprecipitated from all clones. A possible candidate molecule is calnexin, a chaperone located in the endoplasmic reticulum. Calnexin has a molecular mass of 90 kDa and has previously been demonstrated to bind immaturely glycosylated β1 integrin chains (Lenter and Vestweber, 1994). A prominent band of approximately 200 kDa could be co-immunoprecipitated with β1 from the GD25Tβ1A cells. This band was only observed in lanes were immaturely glycosylated β1A chains was present. Previous studies have indicated that talin, with an apparent molecular mass of 225 kDa on SDS-PAGE, is involved in the intracellular processing of β1 integrins (Albiges-Rizo et al., 1995; Priddle et al., 1998). However, immunoprecipitation of talin from our clones, using the same conditions as used in Figs 5and 6, showed that talin and the 200 kDa band did not co-migrate (data not shown). Thus, the identity of the 200 kDa band remains unresolved.

Immunostaining was performed on non-permeabilised and permeabilised cells in order to investigate the cellular localisation of the expressed β1A, β1C-1 and β1C-2 chains. When seeded on FN in medium without serum, the β1A expressing cells showed the expected staining at the focal adhesion sites (data not shown). When seeded without any pre-coating in medium containing serum, the GD25Tβ1A cells showed the typical striped staining of the β1 chains along FN fibres, but essentially no staining of focal adhesion sites (Fig. 7C). Both the GD25Tβ1C-1 and the GD25Tβ1C-2 cells adhered poorly to FN coated surfaces, and no typical focal adhesion staining could be seen in the few cells that did adhere (data not shown). GD25Tβ1C-1 and GD25Tβ1C-2 cells seeded in the presence of serum adhered well but showed no immunostaining for β1 along FN fibrils such as that in the β1A expressing cells. Instead, typically 10-40% of the cells in the various GD25Tβ1C-1 and GD25Tβ1C-2 clones showed intense β1 staining of the nuclear membrane and other intracellular membranes (Fig. 7E,G). This staining pattern, which indicates an ER-localisation, was not seen in the presence of Dox or in non-permeabilised cells (data not shown), proving that the antibody staining was specific for integrin β1 and that most of the antigen was located intracellularly. The remaining cells in the β1C-1 and β1C-2 cultures showed diffuse staining (data not shown). Identical results were obtained using a rabbit anti-β1 antiserum and a hamster anti-β1 mAb. As a control, staining for the integrin β3 subunit showed indistinguishable staining of focal adhesion sites in GD25T cells and β1A, β1C-1 and β1C-2 expressing clones, with (data not shown) and without Dox (Fig. 7). In clones with very high β1A or β1C-1 expression, up to 90% of the cells showed intracellular staining (Fig. 8A,C). Turning off transcription from the β1-cDNA constructs by addition of Dox resulted in sustained β1A cell surface staining and reduced intracellular staining. However in the β1C-1 clones, no peripheral staining was observed; rather the total β1 staining was reduced after addition of Dox (Fig. 8B,D).

Again this suggests that the β1A chains are transported to the cell membrane, while most of the β1C chains are intracellularly degraded. These results further support the finding that expression of the integrin β1C variants results in an intracellular localisation, which is consistent with immature glycosylation and lack of cell surface localisation.

In a previous study, we attempted to express both β1C-1 and β1C-2 in GD25 cells, but after testing a large number of antibiotic resistant clones, we were unable to identify any clones with significant cell surface expression of either β1C-1 or β1C-2. However, transient expression of β1A and β1C-2 cDNA in GD25T cells, revealed that β1C-2 could be expressed, but at a very low level compared with β1A, suggesting either a negative selection for cells expressing β1C or degradation of the expressed β1C proteins (Svineng et al., 1998). To answer these questions, we investigated further the synthesis and processing of the expressed β1A, β1C-1, and β1C-2 proteins in GD25T cells. Here, we have demonstrated that the expression of the integrin subunits β1C-1 and β1C-2 in GD25T cells result in impaired cell surface localisation of these two variants compared with the β1A variant. Our results also indicate that the β1C-1 and β1C-2 proteins are intracellularly degraded. These conclusions are based on the following observations:

1. GD25T cells stably transfected with the cDNAs for β1C-1 and β1C-2 showed very little or no surface expression whereas expression of β1A was found to result in high surface expression of the β1 protein in all antibiotic resistant clones.

2. Analysis of the β1 mRNA levels in the stable clones revealed that the β1-mRNA level in the analysed β1C-1 and β1C-2 clones was equal or higher than in the β1A expressing clones.

3. Metabolic labelling showed that the β1C-1 and β1C-2 mRNAs were translated into protein. However, the amount of EndoH resistant maturely glycosylated β1C-1 and β1C-2 was much lower than that of EndoH resistant β1A.

4. Pulse chase experiments demonstrated that the β1C-1 and β1C-2 chains were degraded without going through the stage of being maturely glycosylated, while after 20 hours of chase, almost all β1A chains were maturely glycosylated. Transient expression of equal amounts of cDNA constructs showed that after 48 hours, much more β1A protein than β1C-1 and β1C-2 protein could be immunoprecipitated from the transfected cells.

5. Immunostaining of the β1A transfected cells showed β1 chains along fibronectin fibres or in focal contacts, while the β1C-1 and β1C-2 chains were localised to intracellular membranes or diffusely at the plasma membrane. Turning off the β1-transcription resulted in a re-localisation of intracellular β1A to the plasma membrane, while the intracellular β1C-1 disappeared without first moving out to the plasma membrane. Previous studies on GD25 cells have revealed that αVβ3, αVβ5, and α6β4 are expressed on the cell surface, while after transfection with β1A, α3β1A, α5β1A, and α6β1A were also exposed. The subunits α1, α2, α4 and α9 were analysed and found not to be present (Wennerberg et al., 1996). The same set of β1-containing heterodimers were expressed after transfection with β1B (Retta et al., 1998). In this study we have used an inducible expression system (Gossen et al., 1994; Gossen and Bujard, 1992); however, we were unable to find any clones with high β1C-1 or β1C-2 cell surface expression, while all the β1A transfected clones had high cell surface expression as determined by flow cytometry. The β1C-1 and β1C-2 cDNAs were both expressed using two different 3′ untranslated regions in order to exclude any influence by transcript stability, but no difference in cell surface expression could be identified (data not shown). Expression of β1B, which was cloned into the same construct in a corresponding manner, was expressed at similar levels as β1A (Armulik et al., 1999). Thus, the inefficient expression of the β1C-1 and β1C-2 chains was solely due to the β1C-1 and β1C-2 specific amino acid sequences.

Fornaro et al. observed β1C-1 cell surface expression comparable to ours when the human cDNA was transiently expressed in Chinese Hamster Ovary (CHO) cells (Fornaro et al., 1995) or stably transfected into rat NRP152 or CHO cells (Fornaro et al., 1999). Microinjection of mouse 10T1/2 cells with human β1A and β1C-1 cDNAs showed that while β1A localised to focal adhesion sites, β1C-1 showed a diffuse staining pattern on permeabilised cells (Meredith et al., 1995). In addition, twice as much β1C-1 cDNA compared to β1A cDNA had to be injected in order to get equal cell surface staining after 24 hours as measured by microfluorometry. This suggests that the inefficient cell surface expression of β1C may not be unique to the GD25T cells. It has also been convincingly shown that expression of β1C-1, or the chimeric protein of β1C-1 specific domain and the IL2-receptor, results in decreased cell proliferation in several cell lines (Fornaro et al., 1995, 1998, 1999; Meredith et al., 1995, 1999). The primary inhibitory domain has been mapped to an 18 amino acid long region present in both β1C-1 and β1C-2. Even cytosolic β1C-1 domains were found to reduce cell growth (Meredith et al., 1999), suggesting that β1C localised in intracellular membranes is able to inhibit growth. However, our clones expressing β1C-1 and β1C-2 are able to grow for prolonged times without any radical reduction in cell number. The reason for the discrepancy between our data and that of those mentioned above is unknown. However, it should be noted that the GD25 cells are immortalised by transfection with SV40 large T-antigen, which could possibly override the negative effects that β1C has on cell proliferation. SV40 large T-antigen has been shown to abolish growth arrest in terminally differentiated epithelial cells of the intestine (Chandrasekaran et al., 1996) and in mouse fibroblasts (Doherty and Freund, 1997). The only difference between β1A and β1C expressing GD25T cells we could notice was that GD25Tβ1C-1 clones with the highest production of β1C-1 protein were more rounded and adhered less well to extracellular matrix molecules than β1A expressing cells (our unpublished results). The rounded morphology of the GD25β1C-1 cells also correlates with heavy intracellular staining by the anti-β1-antibodies, suggesting that a very high level of β1C protein production is the reason for this morphological change.

It has been shown that the integrin α and β subunits form heterodimers when they are still located in the ER compartment (Heino et al., 1989). In several cell lines the expression of β1 has been reported to be in excess of synthesised α-subunits (Akiyama and Yamada, 1987; Heino et al., 1989; Koivisto et al., 1994; Lenter and Vestweber, 1994). In this study, cell surface localisation of α5 and αV was investigated by flow cytometry and the α5 expression was found to correlate with the β1 cell surface expression in different clones. Expression of β1A in GD25T showed that these cells have the capacity to expose high levels of α5 on the cell surface. The low, but detectable, levels of α5 on the GD25Tβ1C-1 and GD25Tβ1C-2 cells indicate that some of the β1C-1 and β1C-2 chains localised on the cell surface are heterodimers with α5 chains. Thus, the β1C subunits are able to form stable heterodimers with α-subunits that can be transported to the cell surface, but for some reason, only a very few such heterodimers are formed.

By investigating the synthesised β1-chains in the GD25Tβ1C clones, it was clear that they were produced to the same extent as the β1A chains. However, the majority of the β1C chains never became maturely glycosylated. Normally, N-linked glycan chains transferred to the β1 chain in the endoplasmic reticulum (ER) are modified in both the ER and Golgi compartments before reaching the cell surface. The attached glycan chains can be cleaved off by the endoglycosidase EndoH until they have been modified by the Golgi mannosidase II in the late Golgi network. Hence, sensitivity to the EndoH enzyme is a marker for immaturely glycosylated forms of α and β integrin subunits, which normally do not localise to the cell surface (Akiyama and Yamada, 1987; De Melker et al., 1997; Heino et al., 1989). The small amount of detected EndoH resistant β1C-1 and β1C-2 chains correlates well with the low number of β1 molecules expressed on the cell surface. It was also shown by pulse chase experiments that the β1C-1 and β1C-2 chains had been degraded after 20 hours of chase, while at the same time-point all the β1A chains had been further processed into the maturely glycosylated form and were present at high levels. The increased degradation of the β1C-1 and β1C-2 chains, compared with the β1A chains, was also shown by transient expression. Many mutated proteins have been shown to be degraded in an ER-associated manner mediated by cytoplasmic proteasomes (Brodsky et al., 1999; de Virgilio et al., 1998; Werner et al., 1996; Xiong et al., 1999). Several chaperone molecules, both cytoplasmic and ER-localised, have been implicated in this process. Binding of the ER-chaperone calnexin to both α and β integrin subunits has been previously demonstrated, and proper removal of the three glucose residues from the oligosaccharide, initially transferred to the proteins, has to be completed before the chains are released from calnexin during heterodimer formation (Lenter and Vestweber, 1994). It is unclear if calnexin or any other ER-localised chaperone has anything to do with the abnormally increased retention and degradation of β1C-1 and β1C-2 in the ER compared to that of β1A, since the difference between the C and the A variants is located in the cytoplasmic tail and not in the part present in the ER lumen. Still, the possibility that cytoplasmic tails can influence on the extracellular conformation of the β1 chain, which could be recognised by chaperones as mis-folded, has to be considered. Interestingly, the cytoplasmic chaperones Hsp90 and Hsp70 have been demonstrated to be involved in ER-associated degradation of cystic fibrosis transmembrane conductance regulator (CFTR) and apolipoprotein B, respectively (Fisher et al., 1997; Loo et al., 1998). Thus, it is tempting to speculate that some cytoplasmic chaperone is able to recognise the β1C-encoded domain as ‘wrong’, and thereby destine the β1C proteins to degradation. We have also found that the cytoplasmic tails of both β1C-1 and β1C-2, in contrast to the β1A cytoplasmic tail, have a strong tendency to aggregate when produced as fusion proteins in E. coli(G. Svineng, B. Wärmegård and S. Johansson, unpublished results). Using computer programs, the estimated hydrophobicity of the β1C-1 and β1C-2 cytoplasmic domains are higher than for both β1A and β1D, which may explain why the β1A cytoplasmic fusion protein is soluble while β1C-1 and β1C-2 are not unless dissolved in 6 M urea (G. Svineng, B. Wärmegård and S. Johansson, unpublished results). This might suggest that the retention of the β1C-1 and β1C-2 chains in the ER is due to aggregation of the β1C cytoplasmic tails, which could lead to recognition by chaperones and also possibly prevent association with α-subunits. However, at present, our results do not allow us to determine whether protein aggregation, ER luminal chaperones or cytoplasmic chaperones are involved in the mechanism for recognition and degradation of the β1C-1 and β1C-2 chains.

In non-transfected cell lines the expression level of β1C is generally very low. Both β1C-1 and β1C-2 transcripts have been identified in several human primary cells and cell lines (Svineng et al., 1998), but re-amplification using nested primers was necessary to identify β1C-1 and β1C-2 transcripts in all tested samples. Several immunohistochemical studies have been conducted in order to study the in vivo distribution of β1C expression (Fornaro et al., 1996, 1998). Consistently, staining for β1C has been found to correlate with non-proliferating cells, but it is not clear whether the staining is intracellular or not. Furthermore, the antiserum was directed against the part of β1C that is encoded by the Alu element (Svineng et al., 1998), hence further studies are needed in order to exclude that the β1C-peptide antibodies used do not cross-react with other proteins containing a similar Alu-encoded region. Surface localisation of β1C proteins has been demonstrated after surface iodination and immunoprecipitation of human erythroleukemia cells and human umbilical cord endothelial cells stimulated with TNF-α (Fornaro et al., 1995; Languino and Ruoslahti, 1992). In both cases, the amount of precipitated β1C was low compared with β1A. Thus, while the expression levels in transfected cell lines represent a situation that normally does not occur, it is conceivable that abnormal overproduction of β1C-1 or β1C-2 in vivo may cause pathophysiological conditions including reduced cell proliferation.

Exchange of the C-terminal cytoplasmic part of β1A with the β1C-1 or β1C-2 C-terminal parts increases its retardation and degradation in the ER-compartment in the GD25T cells. This is a property which is not shared by the β1B and β1D splice variants that have been successfully expressed on the surface of GD25 cells (Armulik et al., 1999) (Belkin et al., 1997; Retta et al., 1998). Further studies are required to understand the role of integrin cytoplasmic tails in the interaction of integrins with chaperones and in intracellular processing and degradation.

We thank Dr R. Fässler for providing the GD25 cells, Drs M. Gossen and H. Bujard for providing the inducible tetracycline/ doxycycline regulated expression system, Drs E. Fries and M. Thuvesson for advice about metabolic labelling, and I. Eriksson for advice about northern blotting. We also thank B. Wärmegård for technical assistance and Dr P. McCourt for English revision of this manuscript. This work was supported by the Swedish Medical Research Council (grant no. 7147) and the King Gustaf V 80-årsfond.