Biochemical Characterization of the Cystic Fibrosis Transmembrane Conductance Regulator in Normal and Cystic Fibrosis Epithelial Cells*

Affinity-purified polyclonal antibodies, raised against two synthetic peptides corresponding to the R domain and the C terminus of the human cystic fibrosis transmembrane conductance regulator (CFTR), were used to characterize and localize the protein in human epithelial cells. Employing an immunoblotting tech- nique that ensures efficient detection of large hydrophobic proteins, both antibodies recognized an approx- imately 180-kDa protein in cell lysates and isolated membranes of airway epithelial cells from normal and cystic fibrosis (CF) patients and of T84 colon carcinoma cells. Reactivity with the a n t i 4 terminus anti- body, but not with the anti-R domain antibody, was eliminated by limited carboxypeptidase Y digestion. When normal CFTR cDNA was overexpressed via a retroviral vector in CF or normal airway epithelial cells or in mouse fibroblasts, the protein produced had an apparent molecular mass of about 180 kDa. The CFTR expressed in insect (Sf9) cells by a baculovirus vector had a molecular mass of about 140 kDa, prob- ably representing a nonglycosylated form. The CFTR in epithelial cells appears to exist in several forms. N - glycosidase treatment of T84 cell membranes reduces the apparent molecular mass of the major CFTR band

Affinity-purified polyclonal antibodies, raised against two synthetic peptides corresponding to the R domain and the C terminus of the human cystic fibrosis transmembrane conductance regulator (CFTR), were used to characterize and localize the protein in human epithelial cells. Employing an immunoblotting technique that ensures efficient detection of large hydrophobic proteins, both antibodies recognized an approximately 180-kDa protein in cell lysates and isolated membranes of airway epithelial cells from normal and cystic fibrosis (CF) patients and of T84 colon carcinoma cells. Reactivity with the a n t i 4 terminus antibody, but not with the anti-R domain antibody, was eliminated by limited carboxypeptidase Y digestion.
When normal CFTR cDNA was overexpressed via a retroviral vector in CF or normal airway epithelial cells or in mouse fibroblasts, the protein produced had an apparent molecular mass of about 180 kDa. The CFTR expressed in insect (Sf9) cells by a baculovirus vector had a molecular mass of about 140 kDa, probably representing a nonglycosylated form. The CFTR in epithelial cells appears to exist in several forms. Nglycosidase treatment of T84 cell membranes reduces the apparent molecular mass of the major CFTR band from 180 kDa to 140 kDa, but a fraction of the T84 cell CFTR could not be deglycosylated, and the CFTR in airway epithelial cell membranes could not be deglycosylated either. Moreover, wheat germ agglutinin absorbs the majority of the CFTR from detergent-solubilized T84 cell membranes but not from airway cell membranes. The CFTR in all epithelial cell types was found to be an integral membrane protein not solubilized by high salt or lithium diiodosalicylate treatment. Sucrose density gradient fractionation of crude membranes prepared from the airway epithelial cells, previously surface-labeled by enzymatic galactosidation, showed a plasma membrane localization for both the normal CFTR and the CFTR carrying the Phe608 deletion (AF 508). The CFTR in all cases co-localized with the Na+,K+-ATPase and the plasma membrane calcium ATPase, while the endoplasmic reticulum calcium ATPase and mitochondrial membrane markers were enriched at higher sucrose densities. Thus, the CFTR appears to be localized in the plasma membrane both in normal and AF 508 CF epithelial cells.
3 To whom correspondence and reprint requests should be addressed.
The gene containing the mutations responsible for the development of the lethal genetic disease, cystic fibrosis (CF),' has been cloned (1)(2)(3), and the hypothetical protein product has been named the Cystic Fibrosis Transmembrane Conductance Regulator (CFTR) (2). Based on hydropathy analysis and homologies with specific domains of other proteins, the gene product has been predicted to contain 12 membranespanning regions, two nucleotide binding domains, a large regulatory (R) domain with numerous sites for phosphorylation by protein kinases, and two sites for N-linked glycosylation in an extracellular segment between the seventh and eighth proposed membrane-spanning helices (2). In most CF cases, the mutation responsible for the disease, which probably stems from a defect in CAMP-dependent regulation of epithelial chloride channels (4-9), has been shown to be a deletion of the phenylalanine residue at position 508 (AF 508) (3). This information prompted the suggestion that the normal CFTR is a plasma membrane-localized nucleotide binding protein, which either regulates or forms protein kinase Adependent chloride channels (2,10,11).
Recent reports have shown that the expression of a normal CFTR in AF 508 mutant cells corrects the abnormal chloride channel behavior (12)(13)(14)(15). Moreover, by expressing the CFTR in insect Sf9 cells, the appearance of CAMP-dependent, protein kinase A-regulated chloride channels was demonstrated (16). Regarding the pathological effect of the PheSo8 deletion, one report (17) indicated that the AF 508 CFTR expressed in heterologous cells had a smaller apparent molecular mass than the normal protein, and the suggestion was made that the mutant protein may not be folded and then glycosylated correctly, resulting in a failure to reach the plasma membrane to form or regulate chloride channels. However, no evidence was provided for the maturation and the localization of the endogenous CFTR protein.
In spite of the rapid progress in the expression of the normal In the present report, using antibodies specific for two different regions of the CFTR and an immunodetection technique with high efficiency and sensitivity, we describe a variety of biochemical features of this protein in normal and CF (AF 508 homozygote) airway epithelial cells as well as T84 colon carcinoma cells which are of epithelial origin. We also compare the characteristics of these proteins to those overexpressed via viral vectors in human airway epithelial cells, mouse NIH 3T3 fibroblasts, and insect Sf9 cells. In addition to establishing several fundamental properties of the CFTR in normal, CF, and overexpressing cell systems, these studies may provide useful tools for future studies of the structure, localization, and function of this newly recognized protein so essential for normal epithelial cell function.
Antibodes-Polyclonal antibodies against synthetic peptides deduced from the CFTR gene sequence (2) as likely to be antigenic (18) were raised in rabbits. Antibodies " R (K770 and K50) were raised against a KLH-conjugate of the peptide identical with residues 776-795 (SVNQGQNIHRKTTASTRKVS), while antibodies "C" (K858 and K82) were raised against a KLH-peptide identical with residues 1458-1480 (CKSKPQIAALKEETEEEVQDTRL). The C terminus peptide, containing a cysteine at its N terminus, was directly coupled to KLH by the sulfhydryl reactive reagent, m-maleimidobenzy1-Nhydroxysuccinimide ester, while the R domain peptide was synthesized with a cysteine added to its N terminus and then similarly coupled to KLH. Rabbits were immunized with 500 pg of the peptide-KLH complexes in complete Freund's adjuvant, by intradermal injections. One month later, intradermal injections were repeated using 200 pg of peptide in Freund's incomplete adjuvant. Subsequently, the rabbits were boosted with monthly intraperitoneal injections of the peptide-KLH complexes adsorbed to aluminum hydroxide. Ten days after each boost, rabbits were bled, and the sera were isolated. Sera (1:lOO dilutions) were screened for the presence of anti-peptide antibodies by incubation with 2 fig of peptide immobilized on nitrocellulose. Bound antibodies were detected subsequently by incubation with alkaline phosphatase-conjugated anti-rabbit IgG followed by color development with NBT + BCIP (see below). The antibodies were affinity-purified by using the relevant peptides covalently linked to Sulfolink resin. Bound antibodies were eluted with 0.1 M glycine-HCl, pH 2.8, and were immediately neutralized by the addition of 1 M Tris base.
A monoclonal antibody specific for the human erythrocyte membrane calcium ATPase (PMCA), 5F10, was kindly provided by Dr. J. T. Penniston. A monoclonal antibody raised against the dog cardiac sarcoendoplasmic reticulum calcium ATPase (SERCA), IId8, was a gift of Dr. Kevin Campbell. A polyclonal antibody raised against the bovine heart inner mitochondrial membrane Complex 111 was kindly provided by Dr. C. Hackenbrock. A polyclonal antibody against the purified CY + @subunits of the dog kidney Na,K-ATPase was generated by Dr. E. Price.
Cell Cell extracts and crude membrane fractions were prepared from the various cultured cells as follows: confluent cell cultures were washed in phosphate-buffered saline and the cells were scraped into Tris mannitol buffer (50 mM Tris, pH 7.0, with HCl, containing 300 mM mannitol and 0.5 mM PMSF). When the proteins of whole cells were to be assayed, the cell suspensions were precipitated with trichloroacetic acid (4% w/v final concentration) and centrifuged for 5 min at 5,000 X g, and the pellet was dissolved in SDS-PAGE disaggregation buffer (see below). For membrane preparations, the cells were lysed and homogenized using a glass-Teflon tissue homogenizer in TMEP (50 mM Tris, pH 7.0, with HC1, containing 50 mM mannitol, 2 mM EGTA, 10 pg/ml leupeptin, 8 pg/ml aprotinin, 0.5 mM PMSF, and 2 mM 8-mercaptoethanol), and the undisrupted cells and nuclear debris were removed by centrifugation at 500 X g for 10 min. The supernatant fluid was then centrifuged for 60 min at 100,000 X g, and the pellet containing the membranes was resuspended in TMEP at a protein concentration of 5-10 mg/ml. All procedures were carried out at 4 "C, and the membranes were stored at -70 "C. Human red cell membrane vesicles and lymphoblast (Jurkat) membranes were prepared as described previously (21,22).
SDS-PAGE and Immunoblotting-Membrane suspensions were either directly mixed with disaggregation buffer (50 mM Tris, pH 6.8, with H3P04, containing 2% SDS (w/v), 15% glycerol (w/v), 2% @mercaptoethanol (v/v), 1 mM EDTA, and 0.02% (w/v) bromphenol blue) or precipitated with trichloroacetic acid (4% (w/v) final concentration), centrifuged for 5 min at 5,000 X g, and then suspended in the disaggregation buffer. Samples were disaggregated at room temperature for 20 min at a protein concentration of 1-3 mg/ml. Electrophoresis of the samples was usually carried out using Bio-Rad Mini-Protean ready gels (4-15% acrylamide gradient, 0.375 M Tris-HC1, pH SA), in a Mini-Protean cell. The maximum amount of membrane protein which could be run in these gels with good resolution was about 50 pglwell. For increased resolution of the high molecular mass proteins or for loading greater amounts of membrane proteins, electrophoresis was carried out in Laemmli-type gels (15-X 15-cm 6% acrylamide resolving gel with 5% acrylamide stacking gel) in a Hoefer electrophoresis cell. The running buffer contained 25 mM Tris glycine, pH 8.3, and 0.2% (w/v) SDS. In the Mini-Protean cells, gels were run for 60 min at 110 V, the large gels were run for 4 h at 100 V.
Electroblotting of the proteins was carried out in a Tris buffer containing a high concentration of glycine and no methanol (0.7 M glycine, 25 mM Tris, pH 7.7), which provided an efficient transfer of large hydrophobic proteins as described in Ref. 23, in Bio-Rad Trans-Blot cells (35 V, 12 h, 15 "C) or in Mini-Trans-Blot cells (60 V, 1.5 h, with cooling unit). In order to obtain efficient binding of the proteins, PVDF membranes (Bio-Rad, 0.2-p pore size) were used. When nitrocellulose membranes were used as in Ref. 23, most of the proteins migrated through the membrane. For molecular mass estimation, Amersham Rainbow prestained standards and Bio-Rad high molecular mass standards were used.
For immunodetection of the proteins, PVDF membranes with the blotted proteins were incubated in TBS-Tween solution (200 mM NaCl and 0.1% (v/v) Tween 20 in 50 mM Tris, pH 7.4, with HCl) containing 5% (w/v) Carnation nonfat dry milk, for 60 min at room temperature. Polyclonal anti-CFTR antibodies were diluted lOO-fold, while other polyclonal and monoclonal antibodies were diluted 500fold in the TBS-Tween-milk buffer and incubated with the blots for 60 min. The blots were then washed twice for 15 min in TBS-Tween solution and incubated with the appropriate second antibody solu-tions (peroxidase-conjugated antibodies diluted 20,000-fold, alkaline phosphatase-conjugated antibodies diluted 750-fold) for 60 min. The blots were then washed three times (15 min each) in TBS-Tween solution. The alkaline phosphatase-labeled blots were developed in color reaction buffer (100 mM NaCl and 5 mM MgC1, in 100 mM Tris, pH 9.5, with HCl), containing 0.3 mg/ml NBT and 0.15 mg/ml BCIP (24). Peroxidase-labeled blots were developed by the enhanced chemiluminescence method, using the Amersham kit. Protein staining on the PVDF blots was carried out by immersing the membrane in methanol for 10 s and then staining for 3 min in a mixture of methano1:acetic acidwater (5:1:5) containing 0.2% (w/v) Coomassie Brilliant Blue R-250. Destaining was achieved by shaking in several changes of the above methanol/acetic acid/water mixture over a period of 20 min. Quantitative densitometry of the x-ray films from the luminograms of the alkaline phosphatase reaction-stained bands on the PVDF membranes was performed with a Hoefer model GS-300 scanning densitometer interfaced with an Apple I1 computer. A density scan program was used to integrate the relevant peak areas.
Carboxypeptidase and Glycosidase Treatment of Isolated Membranes-Carboxypeptidase Y digestion of the isolated membranes was carried out by incubating the crude membrane suspensions (1.5 mg of membrane protein/ml) in 25 mM Tris, pH 7.0, with HCl, containing 100 mM KC1 and 2 mM EDTA, with 75 units/ml of carboxypeptidase Y at room temperature for the indicated time periods. In some experiments, the incubation solutions were supplemented with 0.5% (v/v) Triton X-100. The reaction was terminated, and the membranes were collected by trichloroacetic acid precipitation (4% (w/v) final concentration) and then dissolved in SDS-PAGE disaggregation buffer. N-Glycosidase digestion of the cell membranes was carried out as described (16). In brief, crude membranes were disaggregated for 15 min at room temperature in 2% (w/v) SDS, and the mixture was then diluted 10-fold with a solution containing 25 mM EDTA and 1% (v/ v) Triton X-100 and incubated with 2 units of N-glycosidase/300 pg of membrane protein at 37 "C for 15 h. The membranes were then precipitated by ice-cold ethanol, and the pellet was dried in vacuo and dissolved in SDS-PAGE disaggregation buffer. 0-Glycosidase and neuraminidase treatment of the membranes were carried out under similar conditions. In certain experiments, membranes suspended in SDS-free or EDTA-free buffers were used.
Expression of the CFTR in Airway, NIH 3T3, and Sf9 Cells-The LCFSN retrovirus vector was constructed by insertion of a 4.5kilobase pair AvaI-SstI fragment containing the normal CFTR cDNA (15) into the EcoRI site of the plasmid pLXSN (26). All ends were made blunt using the large fragment of DNA polymerase I. A stable retrovirus packaging cell line producing LCFSN was derived from the amphotropic cell line PA317 (American Type Tissue Culture Collection, Rockville, MD), essentially as described previously (26). CFTl and HBE airway epithelial cells or NIH 3T3 fibroblasts were infected with LCFSN in the presence of 8 pg/ml Polybrene for 2 h. Infected cells were selected in the presence of 150 pg/ml (CFT1 and HBE) or 250 pg/ml (NIH 3T3) G418. Mock infection of the cells was carried out with a similar vector containing the IL-2 receptor cDNA.
In order to express the CFTR in Sf9 cells, site-directed mutagenesis was used to introduce NotI restriction endonuclease sites at position 89 and 4594 in the CFTR cDNA. The engineered cDNA was sequenced in its entirety to verify the absence of any additional mutations. This analysis detected one nucleotide difference between the mutagenized cDNA and the reported sequence (2). This is a C instead of an A at position 1990, identical with the sequence difference reported by Gregory et al. (13). The full length cDNA was cloned into the NotI site of the baculovims transfer vector pVL 1392 (In Vitrogen). Recombinant virus containing CFTR was generated according to established procedures (27). Three independent occlusion negative plaques were isolated and subjected to a second round of plaque purification. When the recombinant viral preparation was used to infect Sf9 cells, similar levels of CFTR expression were observed for each of the three samples cultured for 4 days. Cells infected with one of the viral stocks were further characterized as described under "Results." Binding of Solubilized Glycoproteins to Lectin Agarose-Crude membranes were dissolved in solubilization (S) buffer, containing 2% (w/v) CHAPS, 1% (v/v) Triton X-100, 0.5 mM PMSF, and 2 mM pmercaptoethanol in 25 mM Tris, pH 7.0, with HC1, at a final concentration of about 1 mg of membrane protein/ml. The suspension was incubated for 15 min at 4 "C and then diluted with 10 volumes of the same buffer without detergents. The solution was then centrifuged for 10 min at 1,000 X g, and the supernatant fluid was used for lectin binding studies. Lectin agarose beads, preincubated and washed with buffer 0.1 X S (containing the same ingradients as buffer S but only 0.2% (w/v) CHAPS and 0.1% (v/v) Triton X-100) in a final packed volume of 0.2 ml, were mixed with 0.8 ml of the 1000 X g supernatant fluid. After 2 h of incubation (4 "C, with gentle shaking), the lectin beads were sedimented by centrifugation for 1 min at 200 X g, and the supernatant fluid was collected and analyzed for the presence of CFTR. After three washes of the beads with 5 ml of buffer 0.1 X S, the bound glycoproteins were eluted by incubating the beads for 20 min with 1 ml of buffer 0.1 X S, containing a 0.5 M concentration of the appropriate sugar. The mixture was centrifuged again for 1 min at 200 X g, and the proteins in the supernatant fluid were precipitated by adjusting the solutions to 4% (w/v) trichloroacetic acid. Protein precipitation was facilitated by the addition of 100 pg/ml (final concentration) bovine serum albumin. The mixture was centrifuged for 5 min at 5,000 X g, and the pellet was dissolved in disaggregation buffer for SDS-PAGE analysis. The pH of the samples was neutralized by the addition of small amounts of Tris base.
Surface Labeling of Cells with [3H]UDP-galacto~e-Labeling of cell surface glycoproteins by tritiated galactose was carried out according to the method described in Ref. 25. In brief, approximately 10' airway cells or 5 X 10' T84 cells, collected as described above, were resuspended in 5 ml of a buffer containing 160 mM NaC1,25 mM Tris, pH 7.0, with HCl, 5 mM MnC12, 0.5 units of galactosyltransferase, and 5 pCi of UDP-[3H]galacto~e and incubated at 4 "C for 40 min. The labeled cells were then washed with Tris mannitol buffer, and membranes were prepared as described above. Sucrose Density Gradient Separation of Membranes-Crude membranes were resuspended in Tris KC1 buffer (50 mM Tris, pH 7.0, with HCl, containing 0.1 M KC1, 0.5 mM €"SF, 2 mM P-mercaptoethanol) at a protein concentration of about 0.5 mg/ml, and 2-ml aliquots were layered on top of a discontinuous sucrose gradient consisting of 2-ml aliquots of 50%, 45%, 35%, and 25% (w/v) sucrose in Tris KC1 buffer in a 12-ml centrifuge tube. The gradients were centrifuged in a Beckman SW 40 Ti rotor (150,000 X g average) at 4 "C for 3 h, and the supernatant fluids and the membrane layers at the top of each sucrose solution (0.6 ml) were collected and assayed for protein content and tritium content when appropriate. Additionally, the proteins in 0.5-ml aliquots of the collected samples were mixed with 100 pg of bovine serum albumin (final concentration), precipitated with trichloroacetic acid (4% (w/v)) final concentration, and the precipitates were collected by centrifugation. The trichloroacetic acid pellets were then dissolved in the amounts of disaggregation buffer necessary to obtain equal membrane protein concentrations, and the samples were analyzed by SDS-PAGE as described above.
Other Methods-Membrane protein concentrations were determined by a modified Lowry method (28). 3H measurements were performed in PG 2:l (29) scintillation fluid in a Packard Tri-Carb 1500 liquid scintillation counter. Fig. 1 shows Western blots of several whole epithelial cell and membrane preparations immunostained by two different polyclonal antibodies (K770 and K858) generated against polypeptide regions of the R domain (anti-CFTR-R) and the C terminus (anti-CFTR-C) of the CFTR, and affinity-purified by binding to the respective synthetic peptides as described under "Experimental Procedures." In the trichloroacetic acidprecipitated, SDS-solubilized epithelial cells (panel A ) or crude membrane preparations thereof (panels B and C ) , both antibodies recognized several proteins including an approximately 180-kDa protein, which, as will be elaborated upon below, corresponds to the CFTR. This recognition was absent when the antibodies were preincubated with the relevant peptides (panels A and C, lane 5). Although shown only for the T84 cells, immunostaining of all of the bands in the other cells was likewise eliminated by preincubation of the antibodies with their specific peptides.

RESULTS
Consistent with the results shown in Fig. 1, in experiments carried out with more than 10 different cell and membrane preparations, the 180-kDa bands were present with about equal densities in both normal (Beas, HBE) and CF (AF 508 homozygote CF-T43 and CFT1) cells or in their isolated membranes. In T84 cells or membranes, the band appeared to be broader and more diffuse (see below), and the intensity of immunostaining in the 180-kDa region was always 20-50 times greater than in the airway cells, whether compared on the basis of cell number or membrane protein. As demonstrated in Fig. lA, lane 4, the 180-kDa band was also recognized by the anti-CFTR antibody in cells of a primary culture of an airway tissue sample obtained from a AF 508 CF patient. A similar recognition of the 180-kDa band was observed in nasal scrapes of normal individuals and the CF (AF 508 homozygote) patients (data not shown). In HeLa cells and in Jurkat lymphoblast membrane preparations, a 180-kDa band was also observed, albeit with less intensity than in the airway cells, while no protein staining by either antibody could be detected in isolated human red blood cell membranes (not shown). Fig. 1 also clearly shows that the two anti-peptide antibodies recognized several higher and lower molecular mass proteins in addition to the 180-kDa protein. In epithelial cells and membranes, the K770 anti-R antibody recognized proteins at about 350 kDa, 140 kDa, 105 kDa, and an intensive triplet between 45 and 55 kDa. The anti-C antibody K858 recognized bands at 140 kDa, 105 kDa, 65 kDa, and 40 kDa. The fact that all of these bands disappeared in the peptide competition experiments (Fig. 1, lanes 5 ) indicates the presence of similar epitopes in these proteins. Two other antibodies, generated against the R domain and the C terminus of the CFTR (K50 and K82), recognized the same 180-and 140-kDa bands and weakly the 105-kDa bands in all the epithelial cell types, while the recognition pattern for the lower molecular weight bands was different (see below, Fig. 3). Although a protein with about the expected molecular mass of the CFTR (168,000 + possible glycosylation) is clearly recognized by both types of our anti-CFTR polyclonal antibodies in the epithelial cells and their isolated membranes, the obvious presence of several other antibody-reactive proteins in these cells necessitated a further investigation of the identity of the 180-kDa band. In order to determine whether or not the epitope recognized by the anti-C terminus antibody is indeed at the C terminus of the various proteins, a limited carboxypeptidase Y digestion of the isolated cell membranes was performed, and the products were analyzed in Western blots. Incubations were carried out at room temperature with or without 0.5% Triton X-100 in the medium. As shown in Fig.   2, panel B,  Lanes 1, membranes incubated for 60 min without carboxypeptidase Y in the presence of 0.5% Triton X-100; lanes 2, membranes incubated for 60 min with carboxypeptidase Y in the absence of 0.5% Triton X-100; lanes 3, membranes incubated for 5 min with carboxypeptidase Y in the presence of 0.5% Triton X-100. was no change in the recognition of the immunoreactive bands over a 60-min period, irrespective of the presence or absence of detergent. Identical results were obtained with isolated membranes of Beas and CF-T43 airway epithelial cells (data not shown). These findings support the identity of the CFTR with the 180-kDa band recognized by the anti-C terminus antibody.
To further explore the identity of the 180-kDa protein, we examined the expression of the CFTR by introducing its cDNA into several host cells via viral vectors. For these experiments, the antibody-reactive proteins were separated in longer, 6% Laemmli-type gels, which provide a better resolution for the 180-kDa region than the minigels. As shown in Fig. 3A, lanes 1, in T84 cells, both the R domain (K770 and K50) and the C terminus (K858) antibodies recognize a broad, multiple, and/or fuzzy band between 165-and 185 kDa, and a sharp band at the top of it (about 185 kDa), slightly overlapping the fuzzy band. In CFTl (AF 508 homozygote) airway cells (lanes 2 ) , this latter sharp band and minor bands corresponding to the broad T84 band are apparent. In all the cells, a band at 140 kDa is also recognized by both the R domain and the C terminus antibodies. A similar picture was observed in the case of all the airway epithelial cell types examined (not shown).
Retroviral introduction of the normal CFTR into CFTl (Fig. 3A, lanes 3 ) airway epithelial cells induced a dramatic increase in the immunoreactive bands in the 180-kDa region, including increases in the intensities of the sharp upper band, the broad fuzzy band, and the 140-kDa band. Importantly, these increases were seen with all three anti-CFTR antibodies. Although not shown, similar results were obtained with control and CFTR cDNA-infected normal (HBE) airway epithelial cells. Collectively, these data strongly suggest that the immunoreactive bands in the 180-kDa and 140-kDa regions represent various forms of the CFTR protein. It is important to note that in the CFTR-overexpressing epithelial cells or fibroblasts there was no major change in the amount of another characteristic membrane protein, the plasma membrane calcium ATPase, and no significant differences could be observed in the protein pattern of the Coomassie blue-stained blots (not shown). These latter data indicate that the CFTR, even when overexpressed, is a minor protein component of the mammalian cell membranes. In CFTR-expressing Sf9 cells, a visible, although low intensity protein band, at 140 kDa could be detected by Coomassie Blue staining.
Taken together, all of the above data suggest that the approximately 180-kDa proteins recognized in the epithelial cells by the anti-CFTR-peptide polyclonal antibodies, indeed represent the CFTR. They also point to a polydispersity of the CFTR, probably caused by glycosylation and/or other post-translational modifications.
In the following experiments we intended to examine the nature of the possible glycosylation of the CFTR in the epithelial cells. Again using 6% gels for higher resolution, we found that, when isolated, detergent-treated T84 membranes were incubated with N-glycosidase, the molecular mass of the major, broad CFTR band decreased from about 180 kDa to 140 kDa (Fig. 4A, lanes 1 and 2 ) . However, a sharp band near the top (about 185 kDa), and another band, near to the lower  edge (about 160-165 kDa) of the major CFTR band, were not, or only partially, degraded. By using different deglycosylation periods up to 72 h, no further degradation of these bands was detected, and, although shown here only for the anti Cterminus antibody, K858, the same changes were reported by the anti-R antibodies. The molecular mass of the 14O-kDa, deglycosylated CFTR was identical with that seen in baculovirus CFTR vector-infected Sf9 cells, in which the 140-kDa band was unaffected by N-glycosidase digestion (Fig. 4A, lanes  3 and 4 ) . In similar experiments, N-glycosidase had no visible effect on the endogenous CFTR of detergent-treated airway epithelial (Beas, HBE, CF-T43, and CFT1) cell membranes, and this was true even when the N-glycosidase was supple-mented with neuraminidase and/or 0-glycosidase (not shown). It should be mentioned that if EDTA was not included in the glycosidase incubation medium, a complete loss of the CFTR was observed, probably due to the action of endogenous proteinases.
For a further characterization of the glycoprotein nature of the CFTR, detergent-dissolved epithelial cell membranes were incubated with a variety of lectin-agarose beads. No significant binding of the CFTR to concanavalin A, pokeweed (Phytolacca americana), lentil ( L e n s culinaris), or Ricinus communis RCAso and RCA,zo lectin agarose beads was detected, while incubation of T84 membranes with wheat germ agglutinin (WGA, Triticum uulgaris lectin) agarose, followed by removal of the lectin beads resulted in a significant (60-80%) reduction of the CFTR in the supernatant fluid. Elution of the WGA agarose-bound material with a solution of Nacetylglucosamine, which specifically binds to this lectin, yielded a solution with significant (8-10-fold) enrichment of the CFTR relative to other membrane proteins, with a recovery of about 20-30% (Fig. 4B, I ) . Moreover, this purification procedure resulted in the loss of most of the lower molecular mass anti-CFTR (K858) antibody-binding proteins. The nonglycosylated plasma membrane calcium ATPase (PMCA) and sarcoendoplasmic reticulum calcium ATPase (SERCA) were not detected in the WGA-agarose eluted material (Fig. 4B, ZZ and HZ), providing evidence that the enrichment of the CFTR was not caused by a nonspecific binding of membrane proteins to the beads. In the case of detergent extracts of Beas and CF-T43 membranes, WGA-agarose did not bind a significant amount of the 180-kDa CFTR.
The CFTR is postulated to be a highly hydrophobic, integral membrane protein (2). To determine whether the 180-kDa, anti-CFTR antibody-reactive band represents an integral or a peripheral membrane protein, we subjected the isolated epithelial cell membranes to washes with a high ionic strength medium or with lithium diiodosalicylate, which dissociate peripheral proteins from membranes (30). These treatments removed some of the low molecular mass anti-CFTR antibody-reactive proteins from T84 cell membranes and correspondingly enriched the membrane pellets with respect to the 180-kDa band. The distribution of a known integral membrane protein, the plasma membrane calcium ATPase (PMCA) was also determined. The 140-kDa PMCA partitioned in the same manner as the 180-kDa CFTR band, indicating that the 180-kDa protein is also an integral membrane component. High salt or lithium diiodosalicylate washes of airway epithelial cell membranes yielded identical results (data not shown).
In order to obtain information as to the subcellular localization of the CFTR, crude membrane preparations from several cell lines were fractionated on a discontinuous sucrose density gradient, and the resulting fractions were analyzed for the presence of a variety of membrane markers. The cell surface of the intact epithelial cells was labeled at 4 "C by enzymatic galactosidation, using the UDP-['HH]galactose galactosyltransferase system. This labeling results in the radioactive galactosidation of the cell surface glycoproteins that have glucose molecules at the end position (25). For specific plasma membrane markers, antibodies to the Na+,K+-ATPase and the plasma membrane calcium ATPase (PMCA) were used. Antibodies to the sarcoendoplasmic reticulum calcium ATPase (SERCA) were used to detect the presence of endoplasmic reticulum, and an antibody to the inner mitochondrial membrane Complex I11 was used as a mitochondrial marker. The polyclonal anti-Na,K-ATPase antibody recognized both the (Y (about 100-kDa) and the p (about 60-kDa) subunits of the enzyme, and the monoclonal anti-PMCA antibody detected a 140-kDa protein, which is the monomer size of this ATPase. The anti-SERCA monoclonal antibody reacted with the 100-kDa calcium ATPase present in the endoplasmic reticulum, while the polyclonal antibody against the mitochondrial inner membrane complex I11 detected five prominent bands (two core proteins, two cytochromes, and ironsulfur protein) in the range of about 45-24 kDa.
Pilot experiments with various epithelial cells indicated that the surface and intracellular organelle membranes in our crude membrane preparations could be effectively separated in sucrose step gradients at sucrose concentrations between 25 and 50%. The plasma membrane markers had the highest intensities at the 35% sucrose shelf, while the endoplasmic reticulum marker was greatly enriched at 50% sucrose. The mitochondrial markers were found to be present in each density gradient fraction, but the highest levels were found at 50% sucrose. The subcellular distribution of the various markers is presented below as part of Fig. 6. Fig. 5 demonstrates the distribution of the anti-CFTR-C reactive bands of the Beas (normal) and the CF-T43 (AF 508 homozygote) cell membranes in the sucrose gradient fractions. For both membrane preparations, the 180-kDa CFTR band had the highest intensity at the 35% sucrose shelf, where the plasma membrane markers are enriched (Fig. 6), while most of the lower molecular mass antibody-reactive proteins were concentrated in the lower density fractions. This is in good agreement with the fact that these proteins are easily removed from the membranes by high salt or lithium diiodosalicylate treatments. When detected with the anti-R domain antibodies, the CFTR showed the same localization in the sucrose fractions, and the lower molecular mass bands were likewise enriched in the lower density fractions (data not shown). Fig. 6 shows a quantitative compilation of the data obtained by densitometry of the CFTR, the PMCA, the SERCA, the Na',K'-ATPase, and the Complex I11 immunoreactive materials, and by radioactivity measurements for the distribution of the [3H]galactose surface label, in membrane fractions from Beas (normal) and CF-T43 (AF 508 CFTR) airway epithelial cells after sucrose density gradient centrifugation. Clearly, in both cell membranes, the 180-kDa CFTR co-localizes with the plasma membrane marker enzymes and with the peak of

B. CF-T43 CELLS
FIG. 5. Distribution of the CFTR of Beas (normal) and CF-T43 (AF 508) airway epithelial cells in the various membrane fractions separated by sucrose density gradient centrifugation. Membranes were fractionated on a sucrose step gradient, and aliquots of different fractions were precipitated with trichloroacetic acid and dissolved in SDS-PAGE disaggregation buffer as described under "Experimental Procedures." Samples were electrophoresed and immunoblotted, and the immunostaining was visualized as described in the legend for Fig. 1. Immunodetection was carried out with the anti-CFTR C terminus antibody. 10 pg of membrane protein was deposited in each well. Panel A, Beas cell membranes; panel B, CF-T43 cell membranes. Lanes 1, 0% sucrose supernatant fluid; lanes 2, membranes just above the 25% sucrose layer; lanes 3, membranes just above the 35% sucrose layer; lanes 4, membranes just above the 50% sucrose layer. the 3H activity. The CFTR is much less abundant in the lighter fractions and at the higher sucrose densities, where the endoplasmic reticulum and mitochondrial membranes are enriched. As shown in panels C and D, in contrast to the integral plasma membrane markers, some ['H]galactose label is also present near to the top of the gradient, presumably reflecting the surface labeling of nonintegral membrane glycoproteins or glycolipids. This interpretation is supported by the fact that about 30% of the ['Hlgalactose label is removed from the membranes by extraction with 1 M KC1 (data not shown). Although not shown, in T84 cell membrane preparations, the CFTR and the different membrane markers exhibited the same distribution pattern as demonstrated for the airway cells. Fig. 7 shows subcellular fractionation data for the CFTR in the sibling airway epithelial cell lines, HBE (normal) and CFTl (AF 508 homozygote), as well as in the CFTl cells infected with CFTR cDNA-containing retroviral vector and thus expressing large amounts of the normal CFTR. Similar to the results obtained with the Beas, CF-T43, and T84 cells, for all these cells the CFTR was enriched in the sucrose density fractions corresponding to the plasma membrane markers and not in the heavier fractions enriched in endoplasmic reticulum and mitochondrial membranes. Membranes were fractionated on a sucrose step gradient, and aliquots of different fractions were precipitated with trichloroacetic acid and dissolved in SDS-PAGE disaggregation buffer as described under "Experimental Procedures." Samples were electrophoresed and immunoblotted, and the immunostaining by anti-CFTR C terminus antibody (K858) was visualized as described in the legend for Fig. 1. Equal amounts of protein (10 pg) were deposited in each well. The intensity of the immunostained bands was quantitated by densitometry and computer-assisted integration of peak areas. Data are from a representative experiment. ., intensity of the 180-kDa CFTR band in CFTl cell membrane fractions; ., intensity of the 180-kDa CFTR band in HBE cell membrane fractions; H, intensity of the 180-kDa CFTR band in membrane fractions of CFTl cells infected with a retroviral vector containing the normal CFTR cDNA.

DISCUSSION
The cloning of the gene responsible for cystic fibrosis and the determination of the gene sequence (1-3) opened up many new avenues of approach to the investigation of this severe genetic disease. In the present study, we have used this sequence information to prepare immunoreagents to be used in experiments to detect the CF gene product and determine its biochemical features and cellular localization, with the future aim to purify it, and eventually ascertain its physiological function in a reconstituted system.
Although not elaborated upon under "Results," several obstacles encountered in these studies are worth noting briefly. In order to avoid nonspecific reactions of the antisera on Western blots, the antibodies had to be affinity-purified using the appropriate peptides immobilized on a Sulfolink column. We also found that standard SDS-PAGE and blotting techniques did not efficiently transfer the high molecular mass integral membrane proteins, rendering a quantitative evaluation of the amounts of these proteins on Western blots impossible. However, the use of a high glycine, no methanol transfer buffer (23) and high binding capacity PVDF membranes, and transfer from low percentage, preferably 6% or 4-15% gradient acrylamide gels during electroblotting, greatly improved the transfer efficiency. Moreover, the visualization of the peroxidase-stained blots by enhanced chemiluminesce vastly improved the sensitivity of the technique, allowing the detection of CFTR in whole cell or tissue samples, in crude membranes, and in sucrose density gradient fractions. By allowing different development times for the luminograms, a linear range for the protein concentrationband intensity ratio could be selected and thus the intensities quantitatively assessed by computerized densitometry.
The results presented demonstrate that polyclonal antibodies prepared against sequences of the R domain and C terminus of the CFTR recognize a heterogeneous protein band in the region of 180 kDa. Consistent with relative CFTR messenger RNA levels (2,31), the amount of the 180-kDa protein was found to be much greater in the T84 tumor cells than in the airway epithelial cells. Moreover, carboxypeptidase Y digestion of the membranes eliminated the binding of the anti-C terminus antibody to the 180-kDa band, leaving its recognition by the anti-R domain antibody unchanged. This result indicates that the anti-R and anti-C antibody epitopes in the 180-kDa protein are in different locations and that the anti-C antibody reacted with an epitope located at the C terminus as expected from the CFTR gene sequence.
The most compelling evidence that the approximately 180-kDa band is the CFTR came from the CFTR expression experiments. The CFTR overexpressed in human epithelial cells and in mouse fibroblasts via viral vectors, albeit with a somewhat different appearance in the different cell types, was identically recognized by both types of the anti-CFTR antibodies at about 180 kDa. In the airway cells, overexpression increased the intensity of the sharp CFTR bands and also produced a broad, fuzzy band similar to that seen in the T84 tumor cells (see Fig. 3). In NIH 3T3 mouse fibroblasts, the CFTR overexpressed by the human CFTR cDNA showed a broad fuzzy band in the 180-kDa region (Fig. 3B). When the CFTR was expressed in insect Sf9 cells, similar to an earlier report (16), a 140-kDa, probably un-or underglycosylated form of the protein, was detected by both the anti-C and anti-R CFTR antibodies. This 140-kDa immunoreactive band, with relatively low intensity in the control cells, but increasing with overexpression of the CFTR, was also observed in the human epithelial and in the NIH 3T3 cells. The smaller molecular mass proteins in various cells, recognized differently by the different antibodies, are most probably not related to the CFTR, inasmuch as the amounts of these proteins do not change detectably when cells are overexpressing the CFTR.
All the above findings strongly support the conclusion that our polyclonal antibodies recognize the CFTR and that the endogenous, mature CFTR in human epithelial cells has an apparent molecular mass of about 180 kDa. Nevertheless, a molecular heterogeneity of the mature CFTR is suggested by the present experiments. In the 165-185-kDa region of the immunoblots, multiple immunoreactive bands can be resolved which, although similarly recognized by the different antibodies, are clearly different. N-Glycosidase treatment of the CFTR in T84 cell membranes reduced the molecular mass of the major CFTR band from about 180 kDa to 140 kDa, equivalent to the molecular mass of the underglycosylated CFTR in Sf9 cells (Fig. 4A). However, certain T84 cell CFTR bands and most of the airway epithelial cell CFTR could not be enzymatically deglycosylated. Detergent-solubilized CFTR of T84 cell membranes was shown to be specifically bound to wheat germ agglutinin agarose and then eluted with the leutin-specific sugar N-acetylglucosamine (Fig. 4B), while no such lectin agarose binding was found in the case of the airway epithelial cell CFTR. These findings suggest a different pattern of glycosylation and/or other post-translational modifications of this protein in the different cells. The biosynthesis of heterogeneous forms and various glycosylation patterns of the multidrug resistance protein (P-glycoprotein), a protein apparently similar to the CFTR, has already been described (32). An alternative explanation for the differences observed here could be an incomplete solubilization of the epithelial cell membranes in the detergents used and the formation of membrane protein clusters which may prevent glycosidase and/or specific lectin interactions. Thus, the nature of the glycosylation and post-translational modification of the CFTR in the various cell types requires further investigation.
Three other important biochemical features of the CFTR are also apparent from these studies. First, treatment of isolated membranes by high salt concentrations and lithium diiodosalicylate, agents which remove loosely attached proteins (30), did not solubilize the CFTR, indicating that it is an integral membrane protein. Second, the fully mature, 180-kDa form of the CFTR is present both in normal and in AF 508 mutant airway cells, indicating that the Phe5OS deletion does not interfere in any major way with the normal biosynthesis of the endogenous CFTR. Third, in all of our experiments, it was observed that the CFTR is a minor membrane component, unrecognizable by Coomassie Blue protein staining even in the CFTR-overexpressing mammalian cells or in lectin-isolated membrane glycoprotein fractions.
A major question related to the function of the CFTR is its subcellular localization, as in CF a possible error in the cellular processing and transport of this protein has been suggested (17). Immunocytochemical techniques could be decisive in this respect. However, such studies have been made uncertain by the presence of lower molecular mass immunoreactive proteins, possibly unrelated to the CFTR, in the epithelial cells. We therefore employed a biochemical approach to establish the subcellular localization of the CFTR. Because culturing of the airway epithelial cells in large quantities is time-consuming, laborious, and expensive, only submilligram quantities of membranes could be used for the fractionation studies. We therefore developed a relatively simple sucrose density step gradient technique which proved to be useful in differentiating between cell surface and internal membranes in small samples of crude membranes. Moreover, the use of specific antibodies against the relevant membrane marker enzymes significantly improved the sensitivity of the detection over most conventional marker enzyme assays. The significant enrichment of the plasma membrane markers at the 35% sucrose shelf and the enrichment of the endoplasmic reticulum and mitochondrial markers at higher sucrose densities indicates a reasonably efficient separation of the different cellular membranes by this technique.
The most important findings in the subcellular localization experiments are that the CFTR is a plasma membrane protein and that it is localized in the plasma membranes of both normal and AF 508 homozygote airway epithelial cells. The experiments were carried out with two different normal (Beas and HBE) and two different AF 508 mutant (CF-T43 and CFT1) cell lines, and identical results were obtained in all cases (Figs. 6 and 7). A similar localization of the CFTR was found for the airway cells expressing large amounts of the CFTR (Fig. 7) and for the T84 cells as well. These results render it unlikely that the major problem in the AF 508 CFTR cells is defective maturation and/or transport of the mutant endogenous CFTR to the plasma membrane. Recent work by Gregory et al. (33), involving expression of CFTR mutants which lack glycosylation sites, clearly demonstrates that carbohydrate addition per se is not necessary for CFTR to reach its site of action or to generate CAMP-activated C1-channels. Thus, although in a heterologous expression system the maturation of AF 508 CFTR seems to be impaired (17), our results suggest that explanations for the CF pathogenesis other than the absence of the CFTR in the plasma membrane are more likely to be correct.