Introduction

Top Abstract Introduction Materials & Methods Results Discussion References

Recombinant adenoviruses are currently being evaluated as gene transfer vectors for the treatment of inherited pulmonary diseases such as cystic fibrosis (CF) and alpha-1-antitrypsin deficiency (reviewed in 1 and 2). Extensive preclinical and initial clinical trials have demonstrated that recombinant adenoviruses mediate high efficiency gene transfer to cultured airway epithelial cells and to the airway of cotton rats (3), but a markedly lower efficiency to the conducting airway of primates (4), CFTR-deficient mice (5), human airway xenografts (6), and the nasal epithelium of patients with CF (7, 8). Thus, in addition to overcoming problems posed by the host immune response and the need for repeated vector delivery because of transient expression (1, 2), it appears that the development of adenoviruses for gene therapy will require improvements in gene transfer efficiency.

Recently, attention has focused on the molecular mechanisms that mediate adenovirus binding and its entry into cells, specifically the role of integrin cell adhesion receptors in promoting adenovirus internalization. The v3 and v5 adhesion molecules have been reported to mediate viral internalization through interaction with an RGD sequence in the penton base protein (9). Although recent data suggests that the distribution of integrins on airway epithelia may contribute to gene transfer efficiency with recombinant adenoviruses (10, 11), little has been done to address the hypothesis that normal airway epithelium presents one or more barriers to adenovirus infection, and by implication, to gene transfer with recombinant adenoviruses. The unique properties of differentiated airway epithelial cells that may act as barriers to viral infection include cilia, tight junctions and polarized receptor expression, apical membrane glycoproteins, and mucins.

The transmembrane mucin MUC1 is a heavily sialylated glycoprotein that is expressed in human airway epithelial cells. In contrast to most of the other known mucins, MUC1 is associated with the plasma membrane via a hydrophobic transmembrane domain that has a 69-amino-acid-long cytosolic tail (12, 13). In common with other mucins, the protein backbone of MUC1 consists of a variable number of proline-rich tandem repeats, with several serine and threonine residues that are O-glycosylated (13). During post-translational processing and recycling (18), sialic acid is added as a terminal residue to the O-linked oligosaccharide chains, and thereby contributes to the net negative charge of MUC1.

The function(s) of MUC1 in human airway remain unknown, but investigations in a variety of cell culture systems suggest that it influences cell-matrix and cell-cell interactions. MUC1 has been shown to inhibit aggregation of melanoma and SV40-immortalized normal human mammary cells (19). This effect was partially reversed when cells were pretreated with neuraminidase, suggesting that the terminal sialic acid residues on MUC1 may at least in part mediate its anti-adhesive effects. More recently, overexpression of MUC1 was shown to inhibit cell binding to extracellular matrix proteins (20), suggesting that MUC1 may shield integrins from their interactions with binding sites in extracellular matrix (ECM).

The physical characteristics and reported inhibitory effects of MUC1 on cell aggregation and binding to ECM proteins led us to hypothesize that MUC1 may limit the efficiency of recombinant adenovirus-mediated gene transfer by altering the interaction of the adenovirus with epithelial cells. Using cultured human airway epithelial cells, Madin Darby canine kidney (MDCK) cells that were retrovirally transduced to express human MUC1, and a xenograft model of human airway, we demonstrate that MUC1 and other sialylated proteins reduce the efficiency of adenovirus-mediated gene transfer, and that enzymatic removal of sialic acid from epithelial cells in vitro and in vivo increases the efficiency of gene transfer with this vector.

	Materials and Methods

Top Abstract Introduction Materials & Methods Results Discussion References

Production of Recombinant Adenoviruses

Serotypes 5 and 2, E1-deleted recombinant adenoviruses containing the Escherichia coli lacZ reporter gene driven by the cytomegalovirus promoter (AdCMVlacZ and Ad2gal2), were constructed, purified, and assayed for the number of plaque forming units (pfu) per ml as previously described (21). Viral stocks were stored in 10% glycerol or 5% sucrose and kept frozen at 80°C until use.

Isolation of Human Bronchial Epithelial (HBE) Cells

HBE cells were isolated from native lungs of transplant recipients, including patients with chronic obstructive pulmonary disease, interstitial lung disease, pulmonary hypertension, and CF, or unused sections of the donor lungs. Airways were dissected from surrounding adventitial tissue, and placed in ice-cold Hepes-buffered minimum essential medium containing penicillin, streptomycin, and amphoterecin B. For CF lung tissue, media was supplemented with ciprofloxacin, piperacillin or ceftazidime, and tobramycin. After multiple washes with cold Hanks' balanced salt solution (HBSS), cartilaginous airway segments were cut longitudinally, and incubated overnight at 4°C in 0.1% Protease XIV (Sigma, St. Louis, MO). Airway epithelial cells were obtained by gently scraping the epithelium with the blunt end of forceps. Supernatant from the washed tissue was spun, and the cell pellet was plated on type IV human placental collagen (Sigma) coated tissue culture plates in bronchial epithelial growth media (BEGM; Clonetics Corp., San Diego, CA). When seeded with this protocol and examined by phase microscopy, isolated undifferentiated (flat, non-ciliated) cells and patches of differentiated (columnar, ciliated) airway epithelium adhere and spread. Over 7 to 10 days, undifferentiated cells proliferate and migrate to form a monolayer. The sheets of attached differentiated epithelial cells persist in culture, thereby allowing both differentiated and undifferentiated airway epithelial cells to be exposed to the same viral solution.

Retroviral Transduction of MUC1 and Neomycin Resistance Genes into MDCK Cells

To develop an in vitro polarized cell culture system in which the impact of MUC1 on adenovirus infection could be determined, MDCK I cells were transduced with a MFG-based retroviral vector encoding a cDNA for human MUC1 (24). As a control, MDCK I cells were transduced with a MFG-based retroviral vector encoding a cDNA for two irrelevant genes, neomycin (G418) resistance gene and herpes virus 4 cellular interleukin 10 homologue (vIL-10) gene (25). Subconfluent MDCK cells were infected with retrovirus supernatant. MUC1-expressing cells were selected using a FACScan cell sorter (Becton Dickinson, Rutherford, NJ) and monoclonal anti-human MUC1 antibody BC3 (see below). Individual clones of MUC1-positive cells were obtained by dilution. MDCK cells that were transduced with the neomycin resistance gene were selected by growth in medium containing G418 at 750 µg/ml. MDCK cells that did not express human MUC1 by flow cytometry (designated MDCKv1) and retrovirally transduced MDCK cells containing irrelevant genes (designated MDCK-neo) served as control cells. Two clones of MUC1-expressing cells were found by Western blot analysis of cell extracts (see below) to express a MUC1 that contained approximately 4 (MDCKv2) or 18 (MDCKv3) internal tandem repeats. Only these two MUC1-expressing clones were utilized for gene transfer experiments. MDCK cells were grown in Dulbecco's minimum essential media (DMEM)/Ham's F12 with 3% fetal bovine serum (FBS) and were used from passage 5 to 25.

Antibodies and Lectins

Monoclonal antibody BC3, which binds to an epitope in the internal tandem repeat region of MUC1 (26), was a kind gift from Dr. Ian McKenzie (University of Melbourne, Parkville, Victoria, Australia). Fluoroscein isothiocyanate (FITC) and tetramethylrhodamine isothiocyanate (TRITC)-conjugated or biotinylated lectins from Sambucus nigra (SNA, elderberry bark), Arachis hypogaea (PNA, peanut lectin), and Maackia amurensis (MAL2, Maackia seeds), and biotinylated, FITC- and horseradish peroxidase-conjugated anti-mouse antibodies were obtained from Vector Laboratory (Burlingame, CA). Monoclonal antibody L230 against the v integrin receptor subunit was provided by Dr. Dean Sheppard (University of California at San Francisco, San Francisco, CA). Polyclonal antibody to E. coli -galactosidase was obtained from 5'-3', Inc. (Boulder, CO).

Detection of MUC1 in Cultured Cells Using Flow Cytometry, Western Blotting and Immunocytochemistry

Expression of MUC1 in MDCK cells was demonstrated by flow cytometry. Briefly, cells were trypsinized to a single cell suspension, washed with phosphate-buffered saline (PBS), and then incubated with primary antibody BC3 for 1 h. After washing with PBS, the cells were incubated with FITC-conjugated goat anti-mouse secondary antibody for 30 min. Fluorescence intensity was determined using a Becton-Dickinson FACScan cell sorter.

Western blot analysis of protein extracts from MDCK and HBE cells was performed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) according to Laemmli (27). After washing the culture flasks with ice-cold PBS, cell proteins were extracted at 4°C in PBS containing 60 mM n-octyl -glucopyranoside (Sigma) by scraping flasks with a rubber policeman. Protein concentrations were determined using Bradford analysis (28). Equal quantities of protein were loaded in a 2 to 7.5% gradient gel with 4% stack. Following electrophoresis, samples were transferred onto nitrocellulose and immunoblotting was performed using a standard protocol (29). Briefly, nonspecific binding was blocked by preincubation with 5% nonfat dry milk for 30 min at room temperature. The blot was incubated with primary antibody BC3 for 2 h; after washing with PBS, the blot was incubated with horseradish-peroxidase-conjugated anti-mouse secondary antibody. Bound secondary antibody was detected using an epichemiluminescence detection kit (Amersham Life Sciences, Arlington Heights, IL).

Immunocytochemistry and lectin-binding were performed on cells that were grown on collagen-coated glass coverslips. Cells were fixed with 2% paraformaldehyde, blocked with 5% goat serum, washed with PBS, and incubated with BC3 for 1 h. Cells were then washed with PBS, and incubated with secondary antibody prior to washing in PBS and mounting with Cytofluor (University of Kent at Canterbury, Canterbury, UK). For lectin binding, cells were incubated with FITC or TRITC-conjugated lectins for 1 h and washed in PBS prior to imaging with a Nikon microscope (Nikon Microphot-FXL; NikonUSA, Chicago, IL). Differential interference contrast (DIC) imaging was used to determine cell morphology and the presence or absence of cilia. Image analysis was performed using Optimas software (Optimas BioScan, Inc., Edmonds, WA) to determine relative fluorescence intensities. Confocal microscopy was performed using a Molecular Dynamics 2007 CSLM microscope (Sunnyvale, CA) to determine the precise location of MUC1 expression. Epithelial cell monolayers were labeled with BC3 and an FITC-conjugated secondary antibody, and mounted in Cytofluor. Using the confocal microscope, serial 1,024 × 1,024 pixel scans of the monolayers were collected using a 488-nm laser line, 510-nm primary dichroic, and 510-nm final barrier filter. The microscope pinhole was set to the minimum size to ensure maximal X, Y, and Z resolution. Sections were taken at 0.5-µm intervals over the entire depth of the monolayer. To reconstruct both the localization of the protein on the cell surface and the protein within the cell, the three dimensional reconstruction engine within Imagespace (Molecular Dynamics) was used; this allows topographic reconstruction of the cell surface, as well as rendered slices along the Z axis.

Neuraminidase Treatment of Cultured Cells

To determine the importance of sialic acid on membrane glycoproteins, cells were treated with neuraminidase III (chromatographically purified from Vibrio cholera, from Sigma), which nonspecifically cleaves terminal sialic acid residues. For Western blotting, cell cultures were incubated with 100 mU/ml of neuraminidase at 4°C for 1 h prior to protein extraction. For immunocytochemical detection of MUC1 and sialic acids, MDCK and HBE cells grown on collagen-coated glass coverslips were stained with BC3 antibody, or SNA and PNA lectins, and the staining intensity with and without neuraminidase pretreatment at a concentration of 100 mU/ml for 1 h at 4°C was evaluated by fluorescence microscopy. Fluorescent images were captured using Optimas software, and converted to greyscale. The average fluorescence intensity of differentiated cells, identified by DIC imaging, was determined using Optimas software, with fluorescence on a relative linear scale. The mean fluorescence of five representative areas was recorded for statistical analysis.

Recombinant Adenovirus Infection of MDCK and HBE Cells

To determine whether retroviral transduction itself alters adenovirus-mediated gene transfer, parental MDCK cells and MDCK-neo cells were infected with recombinant adenovirus while in suspension or adherent. After trypsinization of MDCK cells to a single cell suspension, recombinant adenovirus was added at approximate multiplicities of infection (moi; number of infectious particles per cell) of 5, 15, or 50. To adherent cells, adenovirus was added at an moi of 500. Virus was adsorbed at 37°C for 1 h. The cells were washed with HBSS, plated in fresh growth medium and assayed for -galactosidase expression the following day.

Subsequent experiments were designed to determine the importance of MUC1 and sialic acid residues for adenovirus infection. Each of the MDCK cell lines were grown in 24-well culture plates and used 3 d after reaching confluency, as previous work demonstrated that tight junctions are fully developed three days after confluency is reached (30). The growth medium was removed, and replaced with either fresh DMEM/F12 medium (control wells), or DMEM/ F12 containing 100 mU/ml of neuraminidase (neuraminidase wells), with or without 0.3 µM PMSF to abolish residual bacterial protease activity. The plates were incubated at 4°C for 30 min and 1 h. The medium and neuraminidase were then removed, wells were rinsed with HBSS, and recombinant adenovirus was added at approximate moi of 10, 50, 100, or 1,000. After viral adsorption at 4° or 37°C for 1 h, the wells were washed three times with HBSS, and fresh growth medium was added. Similarly, confluent HBE cells grown in BEGM on collagen coated 24-well plates and coverslips were incubated with AdCMVlacZ or Ad2gal2 at approximate moi of 5, 25, 100, and 1,000. Each set of experiments was performed in duplicate on MDCK cells and on HBE cells harvested from at least two different samples of non-CF and CF lung tissue.

In addition, in separate experiments, viral adsorption was performed on MDCK cells that were adherent or in suspension to determine the importance of cell polarization for gene transfer to control and MUC1-expressing MDCK cells. Confluent cells in 24-well plates were brought to a single cell suspension by incubation in trypsin/EDTA. After neutralization with growth medium, the cells were pelleted and resuspended in media containing adenovirus at moi of 2 and 50, and incubated for 1 h at 4°C. Cells were then pelleted, washed and plated. Cells in other wells of the same 24-well plate were infected with virus at the same moi while adherent for comparison to cells infected in suspension.

Assay for -galactosidase Expression

Forty-eight to 72 h after viral infection, expression of the -galactosidase transgene was determined by either staining the cells with 1 mg/ml of 5-bromo-4-chloro-3-indoyl--galactopyranoside (X-gal; Boehringer Mannheim Corp., Indianapolis, IN) solution for 4-5 h (6), or by double immunofluorescent staining using antibodies against MUC1 and -galactosidase. After X-gal staining, the percentage of lacZ-expressing (blue) cells in three representative fields (> 1,000 cells/field) was determined by counting under inverted phase microscopy. The colocalization experiments were performed by counting -galactosidase and MUC1 positive cells in five to seven randomly chosen fields. The percentage of -galactosidase-positive cells among MUC1 positive and MUC1 negative cells was recorded for each field. In addition, double negative cells were counted but not included in the data analysis. For each experiment, duplicate wells or coverslips were used for each condition; and negative controls included neuraminidase pretreatment but no adenovirus, and no neuraminidase or adenovirus.

A fluoReporter lacZ Flow Cytometry Kit (FACSgal; Molecular Probes, Eugene, OR) was used to confirm the results obtained by direct counting. Briefly, cells were trypsinized to a single cell suspension, warmed in a 37°C water bath for 10 min, and incubated with warm 2 mM fluorescein di--D-galactopyranoside (FDG) for 1 min at 37°C. FDG loading was terminated by adding ice-cold staining medium containing 1 µg/ml propidium iodide. FACScan analysis was performed to quantitate -galactosidase expression by comparing the fluorescence intensity of cells that were or were not infected with recombinant adenovirus and exposed concomitantly to FDG.

Creation of Human Airway Xenografts

Third to sixth generation bronchi were dissected from the remaining lung tissue of transplant recipients and silicone tubing was ligated to one end of a 2-cm bronchial segment. One xenograft was implanted subcutaneously into each flank of severe combined immunodeficient (SCID) mice, as previously described (31). Approval for use of human tissue and animals was obtained from the Institutional Review Board and Institutional Animal Use and Care Committees of the University of Pittsburgh.

Recombinant Adenovirus Infection of Human Airway Xenografts

Xenografts were flushed with saline and then air to remove accumulated mucus prior to instillation of neuraminidase or viral solutions. Three xenografts were intralumenally treated with 50 µl of neuraminidase in BEGM at a concentration of 200 mU/ml for 30 min. After flushing the xenograft lumen, 50 µl of AdCMVlacZ at a concentration of 10¹⁰ pfu/ml was instilled via silicone tubing to xenografts. Contralateral xenografts on the same mice were rinsed with BEGM alone for 30 min and then injected with 50 µl of the same viral solution. Mice were killed 72 h later, and the xenografts were harvested. Alternating rings were stained en bloc with X-gal (6), or frozen in OCT compound for later immunostaining. After macroscopic evaluation of en face stained xenograft rings, the tissue was embedded in paraffin and sectioned for microscopic examination.

Adsorption of Adenovirus in the Presence of Polybrene

To address a potential role for cell surface charge in altering the efficiency of adenovirus-mediated gene transfer, HBE cells grown to confluency in 24-well plates were incubated with recombinant adenovirus at moi of 10, 25, and 100 for 1 h at 4° or 37°C in the presence or absence of polybrene (Sigma), heparin (Wyeth-Ayerst Lab, Philadelphia, PA), or both polybrene and heparin.

Statistical Analyses

For the colocalization and lectin binding experiments, differences between groups were determined using unpaired two-tailed t tests. For experiments comparing the percentage of X-gal positive cells between cell lines or under various conditions, the mean and standard error of the mean of duplicate wells were determined and analyzed for statistical difference by analysis of variance (ANOVA) using Statview software (Abacus Concepts, Inc., Berkeley, CA). When statistical differences were found, individual comparisons were made using Fisher's PLSD post-hoc analysis.

	Results

Top Abstract Introduction Materials & Methods Results Discussion References

MUC1 Expression Is Associated with Inefficient Gene Transfer to Cultured Human Bronchial Epithelial Cells

Comparison of MUC1 and -galactosidase expression was performed to begin to address a potential influence of MUC1 on adenovirus-mediated gene transfer. Immunostaining of unpassaged HBE cells with an antibody against MUC1 revealed punctate staining predominantly on the apical surface of a subset of epithelial cells (see Figure 1A). Confocal laser microscopy of HBE cells revealed the label for MUC1 to be predominantly localized to the apical surface of the cells examined (see Figure 1B). A surface reconstruction of the label within cells and the relative intensity of staining are shown using a pseudocolor scale (see inset, Figure 1B). From this image it is clear that the staining of cells, while sometimes quite intense, is variable, and several cells appear to have little or no staining at all. When a vertical, or Z section, is taken through the section series as delineated by the red outline, it can be seen that the majority of the label is restricted to the apex of the cells.

View larger version (86K): [in this window] [in a new window]   Figure 1.   Expression of MUC1 in cultured human bronchial epithelial cells. (A) HBE cells that were grown on coverslips were immunostained with an antibody against human MUC1, which was detected with an FITC-conjugated secondary antibody. Fluorescent photomicrograph reveals cells with punctate fluorescence (arrows in A) that under DIC imaging are ciliated cells (see inset, showing a single ciliated cell). Arrowheads show cells that do not express MUC1. Magnification, ×200 (inset ×600). (B) Fluorescently labeled monolayers were scanned with a confocal microscope, and the image series reconstructed. Labeling is seen to be heterogeneous across the monolayer, as reflected by the differential labeling of individual cells when viewed as an en face reconstruction. The membrane localization was determined by reconstruction and viewing of selected areas in the Z-axis. As can be seen from the reconstruction of the area delineated in red, and shown in the bottom panel of this image set, labeling is almost entirely apical.

Immunofluorescence staining of HBE cells following AdCMVlacZ infection at moi of 100 and 1,000 revealed significant differences in the percentage of cells positive for -galactosidase staining in relation to the expression of MUC1. As shown in Figure 2, and summarized in Figure 3, after AdCMVlacZ infection at an moi of 100, -galactosidase expression in MUC1-positive cells was approximately five-fold less than those that did not express this mucin (15  ± 3% versus 69 ± 4%, P < 0.0001 by two-tailed unpaired t test). Similar results were obtained at moi of 1,000 (18 ± 2% in MUC1-expressing cells versus 81 ± 3% in cells that do not express MUC1, P < 0.0001). The lower transgene expression by MUC1-expressing HBE cells suggested that MUC1 is either a phenotypic marker of a cell that is relatively refractory to adenovirus-mediated gene transfer, or that it interferes with gene transfer.

View larger version (83K): [in this window] [in a new window]   Figure 2.   Colocalization of MUC1 and -galactosidase in cultured human bronchial epithelial cells following recombinant adenovirus infection. HBE cells on coverslips were incubated with a recombinant adenovirus containing the lacZ reporter gene. Two days later, cells were fixed and double immunofluorescent staining was performed with antibodies against human MUC1, detected with an FITC-conjugated secondary antibody, and -galactosidase, detected with a TRITC-conjugated secondary antibody. Panels A, B, C and D, E, F are two different fields imaged at different fluorescent wavelengths. Panels A and D are TRITC images, B and E are FITC images, and C and F show colocalization. The majority of -galactosidase positive cells do not express MUC1 (arrowheads), and the majority of MUC1-expressing cells are negative for -galactosidase staining (arrows); note that only a few cells express both MUC1 and -galactosidase (yellow arrows in A-C). Magnification, ×400.

View larger version (16K): [in this window] [in a new window]   Figure 3.   Percentage of -galactosidase positive cells in relation to MUC1 expression. HBE cells were infected with recombinant adenovirus at an approximate moi of 100 and 1,000 as described in Figure 2, and the percentage of cells expressing MUC1 and/or -galactosidase was determined by counting six representative fields. Shown are the mean ± SEM percentage of -galactosidase positive cells among HBE cells that do or do not express MUC1. At each moi, the percentage of cells that express both -galactosidase and MUC1 is significantly less than the percentage that express -galactosidase alone. The comparison between groups was evaluated using an unpaired t test. *P < 0.0001.

Adenovirus-mediated Gene Transfer to MDCK Cells Expressing MUC1 Is Less Efficient than to Retrovirally Transduced Control MDCK Cells

To further evaluate the hypothesis that MUC1 presents a barrier to adenovirus-mediated gene transfer, experiments were performed in MDCK cells that were transduced to express human MUC1 or neomycin resistance, and parental MDCK cells (designated MDCKv1). After retroviral transduction, cell sorting with antibody BC3, and clonal dilution, two MUC1-expressing clones, and only these clones, were chosen for further analysis. Flow cytometry revealed that over 95% of cells expanded from the two clones, designated MDCKv2 and MDCKv3, expressed human MUC1, whereas the fluorescence intensity histogram of MDCKv1 cells was not different from background (data not shown). In addition, the distribution of MUC1 expression on MDCK cells was determined by immunofluorescent staining. As was observed on HBE cells, MUC1 expression was restricted to the apical membrane in MDCK cells, and the relative fluorescence intensity was greater in the HBE cells than in the MDCK cells (data not shown). MDCK cells transduced to express neomycin resistance (designated MDCK-neo) were selected by growth in medium containing G418 at 750 µg/ml. These cells proliferated normally in G418-containing media at rates similar to parental MDCK cells, whereas 100% of parental MDCK cells died in G418-containing medium.

Parental MDCKv1 cells were compared with MDCK-neo cells to determine the effect of retroviral transduction on adenoviral infection. As shown by FACSgal histograms in Figure 4a, there was no difference in lacZ-expression between the parental MDCKv1 cells and MDCK-neo cells containing a retroviral construct, indicating that retroviral transduction itself does not inhibit adenovirus-mediated gene transfer. No differences were observed at a higher moi of 50 which resulted in 100% infection efficiency, and at moi of 500 in adherent cells that resulted in approximately 9 to 10% infection efficiency in both cell lines (data not shown). Subsequent experiments were performed using parental MDCKv1 cells as controls.

View larger version (22K): [in this window] [in a new window]   View larger version (17K): [in this window] [in a new window]   View larger version (13K): [in this window] [in a new window]   Figure 4.   (a) Flow cytometric analysis of -galactosidase expression in parental MDCK cells (v1) and MDCK cells that were transduced with a neomycin resistance gene (neo). Cells were brought to suspension and infected with Ad2gal2 at moi of 5 and 15 for 1 h. After 48 h FACSgal analysis was performed to quantitate -galactosidase expression. Shown are fluorescence intensity histograms from a representative experiment using Ad2gal2 at an moi of 15. Vertical axis denotes cell number, and horizontal axis denotes log fluorescence intensity. The histograms for cells incubated with staining solution but not infected with adenovirus are shaded gray, and the histograms for Ad2gal2-infected cells are unshaded. The numbers on each unshaded histogram indicate the percentage of cells expressing -galactosidase. Note that the shift in fluorescence intensity and percentage of -galactosidase expressing cells in these two cell lines are almost identical. (b) FACSgal analysis of -galactosidase expression in parental MDCK cells (v1) and MDCK cells that were retrovirally transduced to express MUC1 containing approximately 4 (v2) or 18 (v3) internal tandem repeats. Each of the adherent MDCK cell lines was infected with Ad2gal2 at an approximate moi of 1,000. After 48 h, FACSgal was performed to determine the percentage of positive cells. Shown are fluorescence histograms from a representative experiment. Note that 17% of control cells have a fluorescence above background, whereas less than 1.5% of MUC1-expressing cells exceed background fluorescence. (c) Comparison of -galactosidase expression in adherent MDCK cell lines. MDCKv1, -v2, and -v3 cells were infected with recombinant adenovirus at moi of 10, 100, or 1,000. After 48 h, X-gal staining was performed and the percentage of blue cells determined by counting. Shown are the mean ± SEM of duplicate wells from a representative experiment. Statistical comparison by ANOVA revealed a significant difference (P  < 0.0001) in the percentage of transgene-expressing cells among the three lines at moi of 100 and 1,000. Individual comparisons were made using Fisher's posthoc analysis. *P < 0.0001 versus v1 cells at same moi. (d) Inhibitory effect of MUC1 on gene transfer was independent of cell polarization. Each of the three MDCK cell lines (parental MDCKv1 cells, and MUC1-expressing -v2, and -v3 cells) were incubated with recombinant adenovirus while adherent (C) or in suspension (S) for 1 h, and lacZ expression was quantitated as described in Figure 4c. Shown are mean ± SEM for each cell line under each condition. *P < 0.0001 versus (C) for each cell line.

Comparison of gene transfer to MDCKv1 and MUC1-expressing cell lines was performed using both FACSgal and direct counting. As shown in Figure 4b, FACSgal analysis of -galactosidase expression in these cell lines that were infected with Ad2gal2 at moi of 1,000 while adherent revealed a significant shift of the mean fluorescence intensity only in parental MDCK cells. Note that the percentage of cells above background fluorescence is 17% in MDCKv1 cells versus 1.5% and < 1% in MDCKv2 and MDCKv3 cells, respectively. Direct counting of cells confirmed these results. As summarized in Figure 4c, the percentage of lacZ expressing MDCKv1 cells at an moi of 100 was approximately eight-fold higher than the MDCKv2 and MDCKv3 cells that express human MUC1 (8 ± 2% versus 1 ± 0.6% and 0.5 ± 0.2%, P < 0.0033 by ANOVA). At an moi of 1,000, the difference was three-fold (12 ± 3% versus 4 ± 1% and 4 ± 1% for MDCKv1 versus MDCKv2 and MDCKv3 cells, respectively, P < 0.0033). Notably, the reduced efficiency of adenovirus-mediated gene transfer was observed in MUC1 transfectants expressing either 4 or 18 tandem repeats, suggesting that a MUC1 with as few as 4 tandem repeats (hence 12 O-linked oligosaccharide side chains) is sufficient to influence the efficiency of adenovirus-mediated gene transfer.

As shown in Figures 3 and 4c, the efficiency of gene transfer at an equal moi was significantly lower in MDCK cells than in undifferentiated HBE cells. For example, at an moi of 100, 70% of MUC1-negative HBE cells, compared with no more than 10% of MDCKv1 cells, express lacZ. This could reflect a lack of necessary receptors for human adenovirus on MDCK cells, altered susceptibility to adenovirus infection after retrovirus-mediated transduction, or other factors. To address this and to determine the importance of polarized expression of MUC1, MDCK cells were infected in suspension at an moi of 2 and 50 and assayed for lacZ expression 24 h after plating. When infected at moi of 50, 100% of cells expressed lacZ (data not shown), demonstrating that this cell line is in fact susceptible to adenovirus-mediated gene transfer. The results of adenoviral adsorption to cells in suspension as compared with adherent cells at moi of 2 are shown in Figure 4d. The percentage of lacZ-expressing cells was significantly greater in MDCKv1 than in MDCKv2 and MDCKv3 cells (P < 0.001 by ANOVA), irrespective of cell polarization. In addition, within each of the three cell lines, infection of cells in suspension resulted in a significantly higher percentage of lacZ-expressing cells than adherent cells infected at the same moi. These data indicate that disruption of cell polarization increases susceptibility of a cell to adenovirus-mediated gene transfer, and that polarized expression of MUC1 is not necessary for its effect on gene transfer efficiency.

Human Airway Epithelial Cells Express (2,6) but Not (2,3)-linked Sialic Acid Residues on Their Apical Membrane

Since MUC1 is heavily sialylated, we next sought to determine the role of sialic acid residues in reducing the efficiency of adenovirus-mediated gene transfer. To accomplish this, we initially characterized the expression of sialic acid containing glycoconjugates on HBE cells, and then determined the conditions necessary to enzymatically remove sialic acid residues. As shown in Figures 5A and 5B, localization studies using SNA lectin, which recognizes terminal sialic acid residues linked (2,6) to galactose (Gal) or N-acetyl-galactosamine (GalNAc) (32), revealed that SNA binds to HBE cells, and that, like the BC3 antibody against MUC1, SNA binds at the apical membrane of ciliated cells, with additional binding to a small subset of nonciliated cells. In contrast, MAL2 lectin, which binds to terminal sialic acid residues linked (2,3) to galactose (33), does not bind to HBE cells or to frozen tissue sections of human airway (data not shown). These data demonstrate that the apical membrane glycoproteins of HBE cells contain (2,6)- but not (2,3)-linked sialic acid. In contrast, MDCK cells bind both SNA and MAL2 lectin (data not shown), indicating that these cells contain both (2,6)- and (2,3)-linked sialic acid.

View larger version (115K): [in this window] [in a new window]   Figure 5.   Elderberry bark (SNA) lectin binding to HBE cells without (A, B) and with (C, D) neuraminidase pretreatment. Panels A, B and C, D show the same fields with immunofluorescent and DIC imaging, respectively. Cells grown on coverslips were incubated with FITC-conjugated SNA lectin and subsequently imaged using immunofluorescence (A, C) and DIC (B, D) microscopy. In untreated HBE cells (A, B), there is punctate apical membrane staining of differentiated ciliated cells (arrows). In cells stained after neuraminidase treatment (C, D), there is markedly diminished SNA lectin binding (arrows). Magnification, ×400.

Neuraminidase Treatment of MDCK and HBE Cells Reduces Sialylation of MUC1

In preliminary experiments, the neuraminidase concentration and time of exposure necessary to reduce the sialylation of MUC1 was determined by lectin binding and Western blot analysis. As shown in Figures 5C and 5D, SNA lectin binding was grossly diminished after incubation of cells with neuraminidase. The mean ± SEM fluorescence intensity of untreated cells was 45 ± 9; this was significantly greater than the fluorescence intensity of cells treated with neuraminidase (10 ± 2, P < 0.0001 by unpaired t test). In addition, PNA lectin, which recognizes galactosyl-(1,3)-GalNAc (34) that is masked by terminal sialic acid residues, bound to less than 1% of untreated differentiated HBE cells but over 70% of differentiated HBE cells that had been incubated with neuraminidase for 1 h prior to immunostaining (data not shown). Immunofluorescence staining of HBE cells with antibody BC3 after treatment with neuraminidase revealed no difference in the intensity or distribution of expression (data not shown).

To specifically address the sialylation of MUC1, MDCK cells expressing MUC1 were treated with neuraminidase at concentrations of 100 or 200 mU/ml for 30 or 60 min at 4°C before protein extraction. Western blot analysis of MDCKv1 cell extracts with the BC3 antibody revealed no immunoreactivity for human MUC1. As shown in Figure 6, Western blot of extracts from untreated MDCKv2 and MDCKv3 cells revealed single bands at approximately 160 kD (lane 1, MDCKv2) and 270 kD (lane 3, MDCKv3) that represent mature (fully glycosylated) MUC1. Protein extracts from MDCKv2 and MDCKv3 cells that were treated with neuraminidase had BC3 reactive bands with reduced electrophoretic mobility compared to untreated cells, consistent with removal of negatively charged terminal sialic acid residues from the oligosaccharide side chains of MUC1 (35, 36). Western blot analysis of HBE cell extracts revealed a band at approximately 600-700 kD that represents mature MUC1, and lower molecular weight bands that represent unglycosylated forms. As was observed in the MDCK cells, neuraminidase pretreatment of HBE cells from the same donor revealed a broad band with diminished electrophoretic mobility (see lane 6). The effect of neuraminidase on electrophoretic mobility of MUC1 was maximal at the 100 mU/ml concentration and 60 min incubation; therefore, for subsequent in vitro experiments, cells were treated with 100 mU/ml of neuraminidase for 1 h prior to recombinant adenovirus infection.

View larger version (33K): [in this window] [in a new window]   Figure 6.   Western blot analysis of MUC1 in MDCK and HBE cell extracts with and without neuraminidase pretreatment. Cell extracts were separated by reduced SDS-PAGE on a 2-7.5% gradient gel, and Western blot analysis was performed using a monoclonal antibody against MUC1. Lanes 1 and 2 are extracts from MDCKv2, lanes 3 and 4 from MDCKv3, and lanes 5 and 6 from HBE cells. In lane 1, there is a band with approximate molecular weight (MW) of 160 kD that represents glycosylated MUC1 with approximately 4 tandem repeats (see arrowhead). In lane 3, there is a major band at approximately 270 kD that represents glycosylated MUC1 with approximately 18 tandem repeats (arrowhead). Lane 5 reveals glycosylated MUC1 from HBE cells with an MD of approximately 600-700 kD (arrowhead). Lanes 2, 4, and 6 are extracts of MDCKv2, -v3, and HBE cells following neuraminidase treatment. Compared with the corresponding untreated extracts, there is a shift in electrophoretic mobility due to removal of sialic acid residues (the region shown by the arrowheads and brackets adjacent to lanes 2, 4, and 6 indicate the undersialylated form of MUC1).

Neuraminidase Treatment of HBE and MDCK Cells Increases the Efficiency of Adenovirus-mediated Gene Transfer

To determine whether sialic acid residues on MUC1 and other apical membrane glycoproteins contribute to the efficiency of gene transfer, HBE and MDCK cells were treated with neuraminidase prior to or after incubation with recombinant adenovirus. The efficiency of reporter gene transfer to HBE cells increased significantly (two-fold) when the cells were pretreated with neuraminidase (see Figure 7a). This finding was evident at moi of 5 and 25 (3 ± 0.4% versus 9 ± 0.4% at moi of 5, and 36 ± 2% versus 60  ± 2% at moi of 25, P < 0.0083 for each group by ANOVA). The increased percentage of transgene expressing cells was observed both in morphologically undifferentiated (flat, nonciliated) and differentiated (columnar, ciliated) cells. For cells with a differentiated cell phenotype, the number of X-gal positive cells per group of differentiated cells increased approximately six-fold (1.2 ± 0.4 in control wells versus 6.2 ± 1.1 in neuraminidase treated wells, P < 0.05 by unpaired t test). Similar results were obtained when PMSF was added to the neuraminidase, and treatment of HBE cells with neuraminidase after viral adsorption did not increase gene transfer efficiency (data not shown). There were no significant differences in gene transfer efficiency among the HBE cells harvested from different samples of non-CF lung tissue, and the magnitude of the neuraminidase effect was similar in non-CF and CF airway cells (data not shown).

View larger version (12K): [in this window] [in a new window]   View larger version (16K): [in this window] [in a new window]   Figure 7.   Effect of neuraminidase on the efficiency of gene transfer with recombinant adenovirus. Shown are results of representative experiments in which HBE (a) and MDCK (b) cells were infected with (+) or without () neuraminidase treatment before viral adsorption. HBE cells were infected at moi of 5 and 25, and MDCK cells at an moi of 50, and transgene expression determined as described in Figure 4c. The mean ± SEM of blue cells in duplicate wells are shown. Statistical analysis with ANOVA and Fisher's posthoc analysis revealed a significant difference in the groups. *P < 0.0001 by ANOVA.

As shown in Figure 7b, treatment of MDCK cells with neuraminidase prior to viral adsorption also resulted in a significant increase in the percentage of transgene expressing cells. This effect was most pronounced in MDCKv2 and MDCKv3 cells (five-fold increase), and was of lesser magnitude in MDCKv1 cells (two-fold increase). In contrast, when cells were treated with neuraminidase after adsorption of AdCMVlacZ or Ad2gal2, there was no significant change in the percentage of transgene expressing cells (data not shown). There was no transgene expression in cells that were incubated in media alone or media with neuraminidase (data not shown).

The Efficiency of Adenovirus-mediated Gene Transfer to Human Bronchial Epithelium in Airway Xenografts Increases with Neuraminidase Pretreatment

Human airway xenografts were used to determine whether neuraminidase would alter the efficiency of adenovirus-mediated gene transfer to human conducting airway epithelium in vivo. As shown in Figures 8A and 8B, there was a significant increase in the number of transgene expressing areas in xenografts treated with neuraminidase prior to instillation of AdCMVlacZ. Examination of representative sections revealed that the percentage of X-gal positive cells was less than 1% in the control xenografts but approximately 2-5% in xenografts that had been treated with neuraminidase prior to instillation of virus. As shown in Figures 8C and 8D, the transgene expressing cells appeared morphologically normal. Moreover, the basolateral expression of the v integrin subunit receptor was unaltered after neuraminidase treatment (data not shown).

View larger version (150K): [in this window] [in a new window]   Figure 8.   Effect of neuraminidase on adenovirus-mediated gene transfer to human airway epithelium in vivo. Human airway xenografts were infected with media containing 1010 pfu/ml of recombinant adenovirus, without (A, C) or with neuraminidase (B, D) pretreatment. Xenografts were removed 3 d later, stained with X-gal, and viewed with a dissecting microscope prior to obtaining microscopic sections. Photographs of a representative area from each xenograft reveal significantly higher gene transfer efficiency in xenografts following neuraminidase treatment (B) compared with control (A). Arrows indicate patches of -galactosidase positive cells. Histologic sections from A and B demonstrate transgene expression (arrows) in a higher percentage of epithelial cells in xenografts treated with neuraminidase (D) than control xenografts (C). Magnification (A, B), ×2.5; (C, D), ×200.

The Polycation Polybrene Increases the Efficiency of Adenovirus-mediated Gene Transfer

Viral adsorption was performed in the presence of polybrene to address a potential influence of cell surface charge on gene transfer efficiency. Polybrene is an approximately 3,600-d polycationic polymer that has been previously used to neutralize anionic charges at the cell surface (37). As shown in Figure 9, adsorption of HBE cells at 4°C with 4 µg/ml of polybrene resulted in a four-fold increase in the percentage of cells expressing the transgene (12 ± 1% versus 48 ± 3%, P < 0.0001 by ANOVA). Incubation in the presence of 50 U/ml of heparin, as a representative polyanion, had no effect, whereas incubation with both heparin and polybrene revealed that heparin completely negated the effect of polybrene (48 ± 3% versus 13 ± 1%, P  < 0.0001). As an additional control, polybrene was added after viral adsorption, and found not to significantly increase the percentage of X-gal positive cells (P > 0.05). Viral adsorption to cells at 37°C revealed quantitatively similar effects (data not shown).

View larger version (12K): [in this window] [in a new window]   Figure 9.   Effect of polyions on adenovirus-mediated gene transfer. HBE cells were infected with recombinant adenovirus at an moi of 10 for one hour at 4°C in the presence or absence of 4 µg/ ml of polybrene. Cells were stained with X-gal three days later and the percentage of positive cells determined as described in Figure 4c. Results from a representative experiment, expressed as mean ± SEM, are shown. (A) adenovirus alone, (B) adenovirus adsorbed with polybrene, (C) adenovirus adsorption followed by incubation with polybrene, (D) neuraminidase treatment followed by adenovirus adsorption, (E) neuraminidase pretreatment followed by adenovirus adsorbed with polybrene, (F) adenovirus adsorbed with heparin, (G) adenovirus adsorbed with heparin and polybrene. Comparison between groups was made using ANOVA and Fisher's posthoc analysis. *P < 0.01 versus adenovirus alone.

To determine whether the effects of polybrene are additive to those of neuraminidase, cells were treated with neuraminidase prior to viral adsorption in the presence or absence of polybrene. Adsorption with polybrene further increased the efficiency of gene transfer (25 ± 1% versus 35 ± 2%, P = 0.004), implying that negatively charged moieties other than sialic acid may alter adenovirus binding. Similar results were obtained using the MDCK cell lines (data not shown). There were no lacZ-expressing cells in wells that received polybrene alone.

	Discussion

Top Abstract Introduction Materials & Methods Results Discussion References

Previous studies of recombinant adenovirus vectors have shown efficient gene transfer to epithelial cells in vitro and to some rodent airways in vivo (2), but gene transfer is inefficient in other models of normal airway epithelium (5, 6). Moreover, recent work has demonstrated increased gene transfer efficiency to injured or undifferentiated epithelium (5, 10, 11, 38). These observations suggest that in humans, the normal differentiated airway epithelium constitutes a barrier to recombinant adenovirus-mediated gene transfer, and that disruption of the normal architecture alters the barrier and thereby increases gene transfer efficiency. In the current study, we evaluated whether the mucin MUC1, a sialylated apical membrane glycoprotein that has previously been shown to alter cell-cell (19, 39) and cell-matrix (20) interactions, presents a significant barrier to adenovirus-mediated gene transfer.

Using human airway epithelial and MDCK cells, we have determined that MUC1 and other sialylated glycoconjugates reduce the efficiency of adenovirus-mediated gene transfer. In human airway epithelial cells, gene transfer was approximately five-fold less efficient to MUC1- expressing cells compared with cells that do not express MUC1. As MUC1 may have been merely a phenotypic marker of differentiated HBE cells that are refractory to infection with adenovirus, the colocalization experiments provide indirect evidence that MUC1 alters gene transfer efficiency. To control for potential confounding variables such as tight junctions, restricted distribution of integrin adhesion receptors, and glycosylated apical membrane proteins, gene transfer efficiency was determined in MDCK cells that were transduced to express human MUC1. Gene transfer to MDCK cells expressing human MUC1 with approximately 4 or 18 tandem repeats was three- to eight-fold less efficient than to control cells that do not express MUC1, providing direct evidence that MUC1 reduces the efficiency of adenovirus-mediated gene transfer. Similar results were obtained when cells were infected in suspension, demonstrating that polarized expression of MUC1 is not necessary for this effect. Subsequent studies determined that enzymatic removal of sialic acid residues prior to viral adsorption significantly increased the efficiency of gene transfer. Taken together, these data provide one explanation for the observation that adenovirus-mediated gene transfer to differentiated epithelial cells is inefficient, and demonstrate for the first time that sialic acid residues on epithelial cells influence the susceptibility of a cell to gene transfer with adenoviruses.

A likely mechanism for this observation is that MUC1 impairs adenovirus binding to the exposed apical membrane of differentiated epithelial cells. Electron micrographs demonstrate that the glycosylated form of MUC1 assumes an extended conformation and therefore projects 250-500 nm above the apical membrane (40), well beyond the glycocalyx (14). As such, this glycoprotein is uniquely situated to sterically hinder the binding of bacteria, viruses, and other particulates in the airway lumen to the apical membrane. In addition, the negative charge imposed by terminal sialylation of the three O-glycosylation sites in each tandem repeat of MUC1 (16, 17) may impair binding or limit access to the apical membrane. The finding of increased gene transfer efficiency after removal of sialic acid residues or after addition of a polycation provides evidence for important charge interactions in adenovirus binding. Moreover, the observation that the effects of neuraminidase were seen in cells that do not express human MUC1 mucin (that is, untransduced MDCK cells and undifferentiated HBE cells) suggests that sialic acid residues on other glycoconjugates (41) may interfere with adenovirus binding.

Adenoviruses appear to enter epithelial cells in two stages, with initial binding followed by internalization that appears to be mediated in part by v3 and v5 integrins (9). Once internalized, adenoviruses efficiently escape from endosomes, and the viral DNA is rapidly delivered to the nucleus (42). In our studies, the effect of neuraminidase was similar in magnitude when viral adsorption was performed at 4° or 37°C, suggesting that removal of sialic acid increases adenovirus binding. Although the cell binding site within the fiber knob has not been identified, the C-terminal sequence of fiber from serotypes 2 and 5 includes five and six negatively charged amino acids, respectively (43). Further studies will be necessary to determine the role of fiber knob and other capsid proteins in the important charge interactions suggested by our data. It is conceivable that alteration of the fiber knob to optimize virus binding to a target cell could be a useful strategy to increase the gene transfer efficiency of recombinant adenoviruses.

There are several important implications of our findings for the clinical application of the current adenoviral vectors to diseases such as cystic fibrosis. First, despite a significant increase in the efficiency of gene transfer after neuraminidase treatment, gene transfer to differentiated HBE cells in vitro and in human airway xenografts in vivo is still inefficient. This is most probably related to the presence of other cellular barriers to adenoviral infection, as suggested by the finding that in MDCK cells, the efficiency of gene transfer increases significantly when the virus is adsorbed to cells in suspension. This implies that tight junctions and other factors related to cell polarization also influence the susceptibility of a cell to adenovirus infection. Second, for patients with cystic fibrosis and advanced lung disease who are chronically infected with Pseudomonas species, production and release of pseudomonal neuraminidase (44, 45) might increase gene transfer efficiency. However, for those patients with earlier stages of disease who are not chronically colonized with Pseudomonas and for whom gene therapy may prevent the development of lung disease, adenovirus-mediated gene transfer may be significantly less efficient. Since the majority of CF patients recruited into clinical trials are adults with mild to moderate lung disease, the prevalence of Pseudomonas colonization is likely high. Therefore, the finding of gene transfer in cells recovered from patients colonized with Pseudomonas may not be generalizable to all patients and should be interpreted cautiously. Third, the augmentation in gene transfer efficiency by neuraminidase or polybrene raises the possibility that alteration of cell surface charge may enhance the clinical utility of adenoviral vectors for airway diseases like CF. Further studies in intact animals will be necessary to address this. Lastly, the presence of negatively charged moieties on apical membrane proteins such as MUC1 may have important implications for the clinical application of other viral and nonviral vectors, and should be assessed for each vector.

In addition to the implications of our findings for gene transfer, the data presented suggest that the impact of sialylated glycoproteins on adenovirus infection are very different from other studied viruses. For example, the infectivity of animal rotaviruses (46, 47), bovine coronaviruses (41), and influenza virus (48) is dependent on the presence of sialic acid on the cell surface. To our knowledge, our finding that removal of sialic acid increases the efficiency of gene transfer with adenoviruses represents the first documentation that sialylated glycoconjugates inhibit infection of a cell with a virus.

In summary, MUC1 and other apical membrane sialoglycoconjugates present a barrier to adenovirus-mediated gene transfer. This effect is mediated in part by sialic acid residues, and charge interactions appear to alter the attachment of adenoviruses to the cell surface. Further delineation of these and other cellular barriers to gene transfer may ultimately improve the prospects for gene therapy for human airway diseases.