Recombinant, Replication-Defective Adenovirus Gene Transfer Vectors Induce Cell Cycle Dysregulation and Inappropriate Expression of Cyclin Proteins

Abstract

First-generation adenovirus (Ad) vectors that had been rendered replication defective by removal of the E1 region of the viral genome (ΔE1) or lacking the Ad E3 region in addition to E1 sequences (ΔE1ΔE3) induced G₂ cell cycle arrest and inhibited traverse across G₁/S in primary and immortalized human bronchial epithelial cells. Cell cycle arrest was independent of the cDNA contained in the expression cassette and was associated with the inappropriate expression and increase in cyclin A, cyclin B1, cyclin D, and cyclin-dependent kinase p34^cdc2 protein levels. In some instances, infection with ΔE1 or ΔE1ΔE3 Ad vectors produced aneuploid DNA histogram patterns and induced polyploidization as a result of successive rounds of cell division without mitosis. Cell cycle arrest was absent in cells infected with a second-generation ΔE1Ad vector in which all of the early region E4 except the sixth open reading frame was also deleted. Consequently, E4 viral gene products present in ΔE1 or ΔE1ΔE3 Ad vectors induce G₂ growth arrest, which may pose new and unintended consequences for human gene transfer and gene therapy.

Based on the tropism of wild-type adenovirus (Ad) for the respiratory epithelia (30) and its ability to infect nonreplicating cells (40), replication-defective Ad vectors were thought to be the ideal approach by gene therapy to correct the physiological defects in the airways of individuals having the inherited human disease cystic fibrosis (CF). Numerous studies have established the feasibility of Ad vectors to transfer the cDNA encoding the human cystic fibrosis transmembrane conductance regulator (CFTR) to cells in vitro and in vivo in animal models (4). Culminating in human clinical trials (4), these studies have become the prototype for other Ad-mediated gene therapy protocols targeting cancers, inherited metabolic deficiencies, and cardiovascular disease (1, 20, 73).

Although Ad vectors achieve high levels of transgene expression compared to other viral and nonviral gene transfer strategies (57), several obstacles have hindered the success of human trials for CF gene therapy. Gene transfer to differentiated columnar ciliated cells lining the human airway epithelium is poor (24); by contrast, the preferred targets for infection by Ad vectors are the underlying basal cells exposed to the airway lumen following mechanical abrasion (53) or regenerating epithelial cells (16). Ad-mediated gene transfer is epichromosomal, limiting the duration of the CFTR cDNA expression as the airway epithelium undergoes cellular turnover (49), thus requiring periodic exposure to Ad vectors (82). Moreover, contrary to early expectations, the inflammatory and host immune responses evoked by replication-defective Ad vectors (36, 39, 80) have led to concerns regarding their potential safety for gene therapy (4) and suggestions that immunosuppressive therapy be given during Ad-mediated human gene transfer (33, 37).

Little information is available regarding the impact of the transcription of viral genes remaining in ΔE1 and ΔE1ΔE3 Ad vectors on host cells (38). In the context of CF gene therapy, deletion of the E2a Ad gene along with E1 sequences has been reported to diminish deleterious cytotoxic T-lymphocyte responses (80). The Ad E4 region contains seven open reading frames (ORFs) encoding regulatory proteins involved in the viral life cycle (43). Deletion of most E4 ORF sequences from ΔE1 Ad vectors is thought to minimize the generation of replication-competent Ad and the induction of cellular immune responses and cytopathic effects (74). At the level of host cell DNA synthesis, infection with a ΔE1ΔE3 Ad vector has been reported to decrease cell proliferation (S phase) and induce apoptosis in human primary airway cells in vitro (70). An observation among adenovirologists has been the occurrence of large nonadherent cells present in Ad-infected cultures. Because cell size increases as cells traverse the cell cycle, we hypothesized that they might represent cells arrested in G₂-M. This study demonstrates that first-generation ΔE1 and ΔE1ΔE3 Ad vectors perturb normal cell cycle progression and affect cyclin protein expression. Based on the comparison with a second-generation ΔE1 Ad vector containing only the E4 ORF6 region, we postulate that this mechanism involves proteins encoded by E4 sequences that remain in first-generation ΔE1 and ΔE1ΔE3 Ad vectors. Because ΔE1 and ΔE1ΔE3 Ad vectors lacking cDNA in the expression cassette or containing a marker transgene such as the bacterial lacZ gene are often used as control vectors, their effects on cell proliferation add new variables to gene transfer studies using first-generation Ad vectors.

MATERIALS AND METHODS

Ad vectors.

ΔE1 and ΔE1ΔE3 Ad vectors were propagated in 293 cells, purified by cesium chloride density centrifugation, and titrated by plaque assay as previously described (77). Recombinant virus was stored in vehicle buffer (10% glycerol, 10 mM Tris HCl, 1 mM MgCl₂ [pH 7.4]) at −70°C. Viral infections are given as the multiplicity of infection (MOI) expressed as the number of PFU per cell. The AdCFTR, AdCL, Adα1AT, AdLacZ, and AV1Null vectors (described further in Results) are based on the Ad type 5 (Ad5) genome and lack all of the E1a, 69.5% of the left-hand portion of the E1b, and 66% of the middle section of the E3 regions. The absence of Ad E1a sequences was verified in aliquots of all ΔE1 and ΔE1ΔE3 vectors by using a PCR-based assay (17), and virus replication was assessed as previously described (65). The deletion mutant Addl312 lacks the E1a region (32) and Addl366 lacks the majority of the E4 region (26); both were gifts from T. Shenk (Princeton University, Princeton, N.J.). The Ad2/CMVβgal-5 vector, modified in the ΔE4 region to contain only ORF6 (ΔE1E4ORF6), uses a cytomegalovirus (CMV) promoter to drive the expression of the bacterial lacZ gene (2) and was obtained from Genzyme Corp. (Framingham, Mass.). In all experiments, uninfected control cells were treated with the largest volume of vehicle buffer used for virus dilution. Particle-to-infectious unit ratios (47) were below 50:1 for all vectors (8:1 for AdCFTR; 22:1 for AdCL; 32:1 for AV1Null; 35:1 for Ad2/CMVβgal-5; 40:1 for Adα1AT; 42:1 for AdLacZ; and 45:1 for Addl312).

Cells and cell culture.

IB3-1 cells, derived from the bronchial epithelium of a CF individual (84), were obtained from P. Zeitlin (Johns Hopkins University, Baltimore, Md.) and grown in modified LHC-8 medium. The normal human tracheal epithelial cell line HTE-80 (25), a gift from W. Guggino (Johns Hopkins University), was grown in Iscove’s modified Dulbecco’s medium supplemented with 20% fetal calf serum and 25 mM HEPES. Both IB3-1 and HTE-80 cells were propagated on fibronectin-coated substrates. CFPAC-1 cells, derived from a pancreatic adenocarcinoma of a CF individual (63), were a gift from R. Frizzell (University of Alabama, Birmingham, Ala.) and grown in Iscove’s modified Dulbecco’s medium with 10% fetal calf serum, as were the T84 and HT29 colonic adenocarcinoma cell lines (American Type Culture Collection, Rockville, Md.). Normal human bronchial epithelial (NHBE) cells were obtained from Clonetics (San Diego, Calif.) and propagated in SAGM medium according to the manufacturer’s directions. NHBE cells were used between four and six cell divisions after recovery from cryopreservation.

Flow cytometric analysis.

DNA cell cycle analysis was measured on propidium iodide (PI)-stained nuclei by using either a FACScan (Becton Dickinson Immunocytometry Systems, San Jose, Calif.) or EPICS Elite (Coulter, Hialeah, Fla.). Cell cycle compartments were deconvoluted from single-parameter DNA histograms of 10,000 cells by using Multicycle (Phoenix Flow Systems, San Diego, Calif.), and debris and doublets were removed via software algorithms (78). In all experiments, control and Ad-infected cells were harvested by trypsinization and pooled with the supernatant media from the corresponding culture to reflect accurately changes in the entire cell population.

Correlated measurements of CFTR protein expression across the cell cycle were obtained by indirect immunofluorescence of acetone-fixed (15 min at −20°C) cells incubated (1 h) with a goat anti-human antibody to the C terminus of the CFTR protein (Genzyme) followed by fluorescein isothiocyanate (FITC)-labeled goat anti-mouse immunoglobulin G (IgG) (Southern Biotech, Birmingham, Ala.). Labeled cells were stained with PI (10 μg/ml) and treated with RNAse (100 μg/ml) prior to flow cytometric analysis. Protease inhibitors (46) were included in all buffers, as was 10% goat serum. Isotype-matched mouse IgG was used in place of the CFTR antibody for negative controls.

5-Bromodeoxyuridine (BrdU) incorporation was measured in cells pulsed (30 min) with 10 μM BrdU (Sigma). BrdU content was analyzed as described by Schutte et al. (64), using an FITC-labeled monoclonal antibody to BrdU (clone B44; Becton Dickinson). Labeled cells were washed and resuspended in phosphate-buffered saline (PBS) containing 10 μg of PI per ml for 30 min prior to flow cytometric analysis.

For evaluation of cyclin protein, p34^cdc2 protein kinase, and MPM-2 antibody expression across the cell cycle, trypsinized cells were fixed for 30 min at −20°C in a 1:1 acetone-methanol mixture (cyclins A and B1) or 10 min with 1% paraformaldehyde in PBS (cyclin D), followed by washing in 70% ethanol and PBS, permeabilized for 5 min with 0.25% Triton X-100 in PBS containing 1% bovine serum albumin, and incubated overnight (4°C) with monoclonal antibodies to cyclin A, B1, or D (Pharmingen, San Diego, Calif.) (13) or p34^cdc2 (clone HCDC1; ICN Biomedicals, Costa Mesa, Calif.) (3). Cells were subsequently incubated (90 min) with FITC-labeled goat-anti-mouse Ig antibody (Caltag Laboratories, Burlingame, Calif.) and resuspended in PBS containing PI (10 μg/ml) and RNase (1 mg/ml) for 30 min prior to analysis. Mouse IgG (cyclin B1, cyclin D, and p34) or IgE (cyclin A) was used at the same antibody concentrations as negative controls. In all dual-parameter flow cytometric measurements, green FITC fluorescence was measured at 525 nm and red PI fluorescence was measured at >650 nm. When measurements were obtained with a FACScan, photomultiplier FL3 was reconfigured for doublet discrimination. All bivariate distributions were gated on PI fluorescence area versus fluorescence height to exclude G_0/1 and G₂ doublets from the analyses. Each experiment was repeated two to three times unless indicated otherwise.

Additional procedures.

Cell volumes were measured with a Coulter Counter (model ZBI) connected to a C256 channel analyzer (Coulter). Cyclin kinase activity was measured by immunoprecipitation with either an anti-Cdk2 or an anti-cyclin B1 antibody (Santa Cruz Biotechnology, Santa Cruz, Calif.); [³²P]ATP (Amersham, Arlington Heights, Ill.) and histone H1 (Gibco BRL, Gaithersburg, Md.) were used as substrates (11). lacZ activity was quantitated by using C₁₂ fluorescein di-β-d-galactopyranoside as the substrate (ImaGene Green lacZ kit; Molecular Probes, Eugene, Oreg.) as specified by the manufacturer. Endogenous β-d-galactosidase activity was inhibited by treatment with chloroquine.

RESULTS

Ad vectors induce G₂ cell cycle arrest.

In epithelial cell lines commonly used to assess the efficiency of Ad-mediated gene transfer for CF gene therapy (CFPAC-1, IB3-1, and HTE-80), infection with ΔE1ΔE3 Ad vectors conveying the human CFTR (AdCFTR), the human catalase (AdCL), or human α-1-antitrypsin (Adα1AT) cDNA, or control vectors lacking human transgenic cDNA (AV1Null or the ΔE1 vector Addl312), caused an Ad dose-dependent increase in the number of nonadherent cells in the culture medium. Microscopically, these rounded cells were large and upon staining with a DNA fluorochrome (Hoechst 33342) possessed none of the morphologic attributes of apoptotic cells (i.e., nuclear condensation or fragmentation and decreased area).

Because the cell volume of the nonadherent cells was roughly double that of the attached cells, we hypothesized that Ad infection might induce G₂-M arrest. DNA cell cycle analysis verified a correlation between Ad dose and an increase in the G₂-M fraction (Table 1). Flow cytometric features of apoptosis (12), such as the presence of a subdiploid peak to the left of the G_0/1 peak in the DNA histograms (Fig. 1, for example), a decrease in forward angle light scatter, a decrease in the integrity of the plasma membrane as assessed by supravital staining with PI or Hoechst 33342, or detection of apoptotic cells by using BrdU to label DNA strand breaks (44), were absent in Ad-infected cells (data not shown). Although CFTR protein overexpression has been reported to cause G₂-M arrest in nonhuman primate kidney cells (62), cells infected with non-CFTR-containing vectors (AdCL and Adα1AT) or lacking human transgene cDNA (AV1Null and Addl312) became G₂-M arrested (for example, a >150% increase above control levels for cells infected with AV1Null [Table 1]), indicating that this effect was independent of the cDNA contained in the expression cassettes.

TABLE 1.

Replication-defective Ad vectors increase the G₂-M fraction

Cell type	Mean % of G2-M cells ± SEMa
	AdCFTR dose MOI of:
	0	5	20	50	100	200	400	800
IB3-1	9.4 ± 0.7	10.3 ± 1.5	15.8 ± 3.4 (b)	22.0 ± 1.8 (b)	32.2 ± 3.1 (b)	35.2 ± 3.4 (b)	43.8 ± 5.1 (b)	50.3 ± 18.1 (b)
CFPAC	11.3 ± 0.9	10.4 ± 2.4	—	13.6 ± 1.0 (a)	—	15.5 ± 0.4 (a)	—	27.3 ± 2.0 (b)
HTE-80	18.3 ± 2.6	—	—	22.8 ± 2.0	—	30.4 ± 1.8 (b)	42.3 ± 1.6 (b)	47.1 ± 0.9 (b)
NHBE	8.6 ± 0.8	15.2 ± 2 (b)	—	24.1 ± 1.6 (b)	—	36.9 ± 1.6 (b)	44.7 ± 2.5 (b)	—

Mean % of G2-M cells ± SEMa
Infection withb:
No vector	AdCFTR	AdCL	Adα1AT	AV1Null	Addl312
8.2 ± 0.5 (14)c	37.5 ± 3.5 (11)	34.8 ± 1.5 (6)	43.5 ± 0.6 (2)	18.3 ± 0.3 (2)	23.2 ± 1.25 (3)
10.5 ± 0.6 (9)	16.6 ± 1.7 (5)	15.2 ± 1.4 (5)	—	—	—
18.2 ± 1.8 (6)	30.4 ± 1.8 (3)	28.1 ± 4.7 (3)	—	—	—
8.6 ± 0.8 (13)	36.9 ± 1.6 (7)	28.2 ± 1.5 (7)	—	—	66.5 ± 4.3 (2)

^aDerived from single-parameter DNA histograms of cells infected with Ad vectors for 72 h. The coefficient of variation of the G_0/1 peak, a measure of resolution of DNA cell cycle analysis, was <2.5%. The ratio of the G₂-M to the G_0/1 peak was 2.0 (range, 1.92 to 2.02), indicating stoichiometric binding of PI. —, data not obtained. t-test P values for comparisons of uninfected and AdCFTR-infected cells are <0.05 (a) and <0.01 (b) (n = 3 to 12). 

^bCells were infected with Ad vectors at an MOI of 200. Uninfected cells were treated with virus vehicle. P values are <0.01 for cells infected with different Ad vectors. Infection with laboratory-grade and clinical-grade AdCFTR vectors (IB3-1; 5 days) produced essentially equivalent increases in the percentage of G₂-M cells (56.3% ± 2.7% [n = 6] versus 62.5% ± 6.4% [n = 3], respectively). DNA histograms from Ad vectors (AdCFTR and Addl312) inactivated by UV irradiation (254 nm at 6 cm for 2 min [66]) were superimposable on histograms from cells treated with vehicle alone (control). 

^cNumber of determinations. 

FIG. 1.

Ad vectors induce G₂ cell cycle arrest. IB3-1 cells were exposed to either 5 μM SKF 96365, a reversible and specific inhibitor that arrests cells in M phase (50), for 24 h, virus vehicle (control), or AdCFTR (200 PFU/cell) for 72 h, fixed, and incubated with MPM-2 antibody. The trapezoidal window represents the level of immunofluorescence staining of cells stained with isotype IgG control, indicating that only M-phase cells reacted with this antibody. The corresponding DNA histogram from each bivariate display is shown projected above the dual-parameter dot plots. The G₂- and M-phase populations are indicated by solid boxes. Data represent measurements from 25,000 cells. An arrow represents the position of G₂-M cells in single-parameter DNA histograms. There is no peak to the left of the G_0/1 peak, denoting the absence of apoptotic cells.

To exclude the possibility that Ad-induced G₂-M arrest was specifically restricted to human bronchial epithelial cells immortalized by simian virus 40 (IB3-1 and HTE-80) or epithelial cells derived from tumors (CFPAC-1), primary cultures of NHBE cells were infected with Ad vectors and subjected to DNA cell cycle analysis. NHBE cells arrested in G₂-M in an Ad dose-dependent manner (Table 1), comparable to the levels observed in Ad-infected IB3-1 cells. Infection at low MOIs (5 PFU/cell in NHBE cells and 20 PFU/cell in IB3-1 cells for 72 h [Table 1]) nearly doubled the percentage of G₂-M cells. Generally, the magnitude of G₂-M growth arrest was proportional to Ad dose, culture time after infection, and S-phase transit rate. In cells with slow rates of growth (HTE-80 or normal human endothelial vein cells) infected with Addl312, arrest occurred at either higher Ad dose (200 PFU/cell) or at an MOI of 25 following increased culture time after infection (6 to 8 days [data not shown]).

Because single-parameter DNA histograms cannot discriminate G₂ from M-phase cells, immunofluorescence staining of M-phase-specific proteins identified by the MPM-2 antibody (14, 23) was used to delineate which of these compartments was increased in ΔE1 or ΔE1ΔE3 Ad-infected cells. Ad infection specifically arrested cells in the G₂ phase of the cell cycle (Fig. 1). Similar results were obtained when total nuclear protein was correlated with DNA content (54), another flow cytometric technique to distinguish G₂ from M cells; in these experiments, M cells constituted <5% of the total number of cells in Ad-infected cultures.

Ad vectors affect S-phase entry.

ΔE1ΔE3 Ad vectors reportedly decrease cell proliferation (70). To assess directly their effect on cellular DNA synthesis, we exposed AdCFTR-infected cells to BrdU and quantitated incorporation of this thymidine analog by flow cytometry (30-min pulse) (Fig. 2A). The percentages of BrdU-labeled cells decreased slightly from 73 ± 1 (control) to 67 ± 1 (MOI of 25) and 55 ± 2 (MOI of 200; P < .05). Diminished BrdU incorporation was also observed with other vectors (AdCL and Addl312 [data not shown]).

FIG. 2.

(A) Ad vectors inhibit progression from G₂ and delay the entry of cells into S phase. IB3-1 cells were treated with virus vehicle or AdCFTR (25 and 200 PFU/cell) for 72 h, pulse-labeled with BrdU, and monitored for an additional 24 or 48 h. Data represent measurements from 25,000 cells. (B) Cartoon of the positions of the BrdU⁺ and BrdU⁻ populations across the cell cycle.

These observations suggested that infection with the ΔE1 or ΔE1ΔE3 Ad vector may affect multiple cell cycle checkpoints. To confirm this hypothesis, the progression of control and Ad-infected cells through the cell cycle was analyzed by BrdU pulse-chase analysis (Fig. 2A). Following a 30-min BrdU pulse, cells were chased for 24 or 48 h in fresh growth medium. During the first cell division cycle after the BrdU pulse (24-h chase), BrdU-labeled S-phase cells traverse the cell cycle into G₂, divide, and give rise to BrdU-expressing (BrdU⁺) G₁ cells. Cells in G₂-M during the BrdU pulse are unlabeled and after mitosis result in G₁ cells lacking BrdU incorporation (BrdU⁻ G₁ cells). Likewise, cells in G₁ during the BrdU pulse are also unlabeled and appear as BrdU⁻ S-phase cells after the first division cycle. Following a second cell division (48-h chase), BrdU⁺ S-phase cells subsequently reappear. In control, uninfected IB3-1 cells, BrdU⁺ G₁ and BrdU⁺ S-phase cells were present at the 24- and 48-h chase time points, respectively (Fig. 2A, vehicle). In contrast, by the 24-h chase, the majority of IB3-1 cells infected with AdCFTR at an MOI of 200 PFU/cell appeared as a BrdU⁺ G₂-arrested population. Only BrdU⁺ G₁ cells from IB3-1 cells infected with AdCFTR at a low MOI (25 PFU/cell) partially reinitiated DNA synthesis (BrdU⁺ S-phase cells [Fig. 2A, 48-h chase). At an MOI of 200 PFU/cell, BrdU⁺ G₁ cells failed to reenter S phase at 48 h.

Some cells escape from Ad vector-induced cell cycle arrest.

In Ad-infected IB3-1 cultures, a small number of cells escaped G₂ arrest and, because BrdU was absent in the chase medium, subsequently appeared by the 48-h chase as BrdU⁻ S-phase cells (Fig. 2B, S). This subpopulation could have arisen from cells resistant to infection by the ΔE1ΔE3 Ad vectors. To explore this possibility, we infected IB3-1 cells with an ΔE1ΔE3 Ad vector coding for bacterial β-galactosidase and quantitated lacZ expression by flow cytometry. At 3 and 5 days after infection, only 0.3 to 0.4% of the cells did not express lacZ (at an MOI of either 5, 25, or 200 PFU/cell; correspondingly, the G₂-M fraction at these MOIs increased to 11, 18, and 38%, respectively). This contrasts with the 3 to 10% BrdU⁻ S-phase cells in the pulse-chase experiments and excludes the conclusion that this subpopulation arose from uninfected cells. An alternate explanation is that the small number of cells in G₂-M at the time of Ad infection were able to overcome growth arrest and undergo subsequent cell division.

Ad vectors affect cyclin protein distribution and expression.

Since progression through the eukaryotic cell cycle is regulated by cyclin-dependent kinase holoenzyme complexes (68), it was possible that the cell cycle perturbations observed in ΔE1 or ΔE1ΔE3 Ad-infected cells involved alterations in cyclin protein expression. From studies using synchronized cells, flow cytometric assessment of cyclin protein expression is comparable to conventional blotting techniques; however, in the case of asynchronous populations, as studied here, this technique permits the detection of cyclin protein expression with high sensitivity between cell cycle compartments (see reference 13 for a review).

Differential expression of cyclin A protein begins at the G₁/S transition and reaches maximal levels in the late S and G₂ phases (13). Flow cytometric analysis of cyclin A protein expression across the cell cycle confirmed that this observation held true for uninfected IB3-1 cells (Fig. 3). However, in AdCFTR-infected cells, cyclin A protein was overexpressed in late S and G₂ (Fig. 3). Similarly, cyclin B1 protein expression, differentially expressed in late S and G₂ (34), also increased with Ad dose and was inappropriately expressed to G₁ cells (Fig. 3). Ad-infected NHBE cells showed similar aberrant patterns of cyclin A and B protein expression, beginning at an MOI of 5 (Fig. 4).

FIG. 3.

Overexpression and unscheduled expression of cyclins A, B1, and D in IB3-1 cells exposed to Ad vectors. Control (vehicle) or AdCFTR-infected cells at the viral MOIs indicated were harvested 72 h after viral infection. Similar results were obtained when cells were infected with AdCL. The trapezoidal window represents the immunofluorescence levels of cells stained with an isotype IgG control (cyclin B1 or D) or IgE (cyclin A). In the cyclin B panel, inappropriate expression of cyclin B protein to G₁ cells is indicated by the dashed box. At 25 and 200 PFU/cell, cyclin B protein immunofluorescence is present in the portion of this box above background levels (trapezoidal window). Bivariate histograms were obtained from 25,000 cells.

FIG. 4.

Overexpression and unscheduled expression of cyclins A and B1 in NHBE cells exposed to Ad vectors. Control (vehicle) or AdCFTR-infected cells at the viral MOIs indicated were harvested 72 h after viral infection. Similar results were obtained when cells were infected with AdCL. The positions of the cell cycle compartments (G₁, G₂-M, G₂, and M) are indicated by either broken lines or arrows and are based on DNA content. The trapezoidal window represents the immunofluorescence levels of cells stained with an isotype IgG control (cyclin B1 or D) or IgE (cyclin A). There is inappropriate expression of both cyclin A and B1 proteins to G₁-phase cells. Bivariate histograms were obtained from 25,000 cells.

Because D-type cyclins are rate limiting for progression from G₁ into S phase (13), the decreased entry of ΔE1 or ΔE1ΔE3 Ad-infected cells into S phase might be due to a reduction in cyclin D protein expression. This was not observed (Fig. 3). Actually, in Ad-infected cells, cyclin D protein expression was noticeably increased, in an Ad dose-dependent manner (Fig. 3) in both G₁ and G₂-M phases.

The inability of ΔE1 or ΔE1ΔE3 Ad-infected cells to progress from G₂ to M may be the result of a decrease in cyclin-dependent kinase levels. To explore this possibility, we analyzed cyclin-dependent kinase p34^cdc2 protein levels and kinase activities in control and Ad-infected cells. Normally, p34^cdc2 protein is expressed across all cell cycle phases and is absent only in noncycling G₁ cells (Fig. 5A, vehicle) (3). In Ad-infected cells however, p34^cdc2 protein was increased in an Ad dose-dependent manner (Fig. 5A). Cdc2 and Cdk2 kinase (cyclin A also associates with Cdk2) activities appeared to be unaffected (Fig. 5B).

FIG. 5.

p34^cdc2 kinase expression and kinase activity in AdCFTR-infected IB3-1 cells. (A) Analysis of Cdc2 protein expression across the cell cycle. The trapezoidal window represents the immunofluorescence levels of cells stained with isotype IgG controls. The positions of the G₁ and G₂-M populations, denoted by dotted boxes, are based on DNA content. Histograms depict measurements from 25,000 cells. (B) Histone H1 kinase activity of Cdk2 (cyclin A) and Cdc2 (cyclin B1) in IB3-1 cells exposed to vehicle buffer (VHC) or infected with AdCFTR at the indicated MOIs. Cells were harvested 72 h after infection.

Ad gene transfer is not limited to specific cell cycle compartments.

G₂ arrest by Ad vectors raised the possibility that gene transfer and subsequent reconstitution of protein and/or function are restricted to a specific cell cycle phase. To address this question, CFTR protein expression following infection with the AdCFTR vector was measured by immunofluorescence and correlated with DNA content. CFTR protein was expressed throughout the cell cycle in asynchronously growing T84 cells, which endogenously express abundant levels of CFTR protein (5) (Fig. 6). By contrast, CFTR protein was absent in the control (Fig. 3, vehicle) or AdCL-infected IB3-1 cells (data not shown). In AdCFTR-infected IB3-1 cells, CFTR protein expression increased in an Ad dose-dependent manner across all cell cycle compartments (Fig. 6), thus confirming that Ad transgene expression was not limited to a specific cell cycle compartment.

FIG. 6.

Immunofluorescence measurement of CFTR protein expression across the cell cycle. Control (vehicle) or cells infected with AdCFTR at the MOIs indicated were labeled by indirect immunofluorescence with an anti-CFTR antibody (vertical axis) and counterstained with PI for DNA content (horizontal axis). The trapezoidal window represents the immunofluorescence levels of cells stained with isotype IgG controls. The G₁ and G₂-M populations are indicated by broken lines. The solid line in the T84 bivariate distribution (arrow) shows the level of background immunofluorescence. In this experiment, >98% of T84 reacted with the CFTR antibody. Bivariate histograms were obtained from 25,000 cells.

Aneuploid DNA histogram patterns following ΔE1ΔE3 Ad infection.

In the case of 3T3 mouse fibroblasts, commonly used as a host cell to express foreign genes, infection with the ΔE1 or ΔE1ΔE3 Ad vector induced an additional G_0/1 peak (Fig. 7), consistent with the presence of aneuploidy in DNA histograms (76). This DNA aneuploid peak began to appear after infection at an MOI of 20 and became more clearly discernible with increasing Ad dose. Peak position remained constant, denoting a stable population with abnormal DNA content. Interestingly, when Ad-infected IB3-1 cells were passaged by trypsinization and recultured for 72 h, a proliferating DNA aneuploid population was also observed (Fig. 7).

FIG. 7.

Aneuploidy in Ad-infected cells. (A) 3T3 cells infected with vehicle buffer (VHC) or AdCL at the MOIs indicated for 48 h. A solid arrow shows the position of the aneuploid population in the Ad-infected cells. (B) Appearance of aneuploid subpopulations in recultured Ad-infected IB3-1 cells. Cells were exposed to vehicle buffer or Ad vector for 72 h, trypsinized, and recultured for an additional 72 h in fresh growth medium before DNA cell cycle analysis. The positions of the G₁ diploid and G₁ aneuploid peaks are labeled. The solid arrow depicts the position of the diploid G₂-M peak, while the open arrow indicates the position of aneuploid G₂-M cells.

Infection of HT-29 cells, a model for colonic cell differentiation (58) and the in vivo assessment of the response of human colon cancer cells to therapeutic modalities in immunodeficient animals (51), with an ΔE1ΔE3 Ad vector induced polyploidy (cells having >4N DNA content), where G₂-arrested cells undergo subsequent rounds of DNA synthesis without proceeding through mitosis. Polyploidization occurred in an Ad dose-dependent manner independent of vector (AdCL, AdCFTR, or Addl312) at an MOI of 5 or greater and was accompanied by the elevated levels of cyclin A and B1 proteins (Fig. 8).

FIG. 8.

Ad vectors induce polyploidization in HT-29 cells. Cells were treated with either vehicle buffer or AdCFTR at the indicated MOIs and harvested after 72 h for BrdU analysis or for analysis of cyclin A or B1 protein expression. Prior to harvesting, cells were pulsed with 10 μM BrdU for 30 min. The trapezoidal window represents the immunofluorescence levels of cells stained with isotype IgG controls. Data represent measurements from 10,000 cells.

G₂ growth arrest is absent in an ΔE1E4ORF6 Ad vector.

To address the possibility that the E4 gene region was responsible for G₂ growth arrest caused by ΔE1 and ΔE1ΔE3 Ad vectors, IB3-1 and HT-29 cells were infected with a second-generation Ad2-based vector, Ad2/CMVβgal-5, that had been modified in the E4 region to contain only ORF6 (ΔE1E4ORF6). In contrast to the results observed with the ΔE1-only or ΔE1ΔE3 Ad vector, the percentage of G₂-M cells remained essentially unchanged compared to control cells over a wide range of MOIs (5 to 1,000 PFU/cell) in IB3-1 cells infected with ΔE1E4ORF6 (Fig. 9). Only at ≥1,500 PFU/cell did the G₂-M fraction significantly increase. A similar situation was observed for NHBE cells (data not shown). While polyploid HT-29 cells were detected after infection with the ΔE1 or ΔE1ΔE3 Ad vector at 5 PFU/cell (Fig. 7), infection with the ΔE1E4ORF6 required an MOI of >1,000. At 400 PFU/cell, DNA histograms showed increased G₂-M and a small number of cells having >4N DNA content (Fig. 9). DNA histograms from HT29 or IB3-1 cells infected with an ΔE4 vector (Addl366) at an MOI estimated at >1,000 PFU/cell likewise did not show increased levels of G₂-M cells.

FIG. 9.

Cell cycle arrest and polyploidization is absent in cells infected with Ad2/CMVβgal-5, an Ad vector that contains only the E4 region ORF6. In IB3-1 cells, the G₂-M fraction remained unchanged up to 1000 PFU/cell (12% versus 16%, P > 0.8). Between 50 and 200 PFU/cell, the percentage of G₂-M cells remained essentially unchanged in HT-29 cells, and polyploid cells having DNA content of >4N were absent.

DISCUSSION

G₂ cell cycle arrest was a direct consequence of infection with a replication-defective ΔE1 or ΔE1ΔE3 Ad vector. Growth arrest was absent in cells exposed to UV-irradiated Ad vectors (Table 1), indicating the requirement for Ad gene expression in this process. Moreover, cells infected with a clinical-grade AdCFTR vector, required for use in human gene therapy trials, became G₂ arrested (Table 1), thus eliminating the possibility that contaminants in the laboratory-grade constructs were responsible for growth arrest. (Clinical-grade vector is prepared under Good Laboratory Practices and Good Manufacturing Practices for administration in human trials. Vector prepared to these specifications has a frequency of replication-competent viral particles of <1 in 10⁹ [6a]). Interestingly, sporadic examples found in the literature show decreased [³H]thymidine uptake in primary vascular smooth muscle cells (8), approximately a 50% decrease in cell proliferation in U373MG cells (9), a fourfold increase in the G₂-M fraction in human non-small-cell lung cancer cells lines (34), and elevated numbers of G₂-M cells in numerous tumor cell lines (60) infected with control ΔE1ΔE3 Ad vectors (expressing lacZ) compared to uninfected cells. These changes, although not addressed by the authors, further support our conclusion that cell cycle arrest in primary and immortalized cells is an inherent characteristic of infection with ΔE1 and ΔE1ΔE3 Ad vectors.

The ability of the ΔE1 or ΔE1ΔE3 and not the ΔE1E4ORF6 Ad vector to induce G₂ cell cycle arrest suggests that gene products from the Ad E4 region other than ORF6, present in the vectors tested here and demonstrated to have oncogenic properties (48), may affect eukaryotic cell cycle regulation. While the E4 ORF1 encodes a transforming protein (75), the E4 ORF4 protein induces apoptosis in a p53-independent manner in transformed (69) and rodent (41) cells. Biochemically, the product of E4 ORF4 interacts with protein phosphatase 2A (43). Progression through the G₂-M phase of the cell cycle requires dephosphorylation of the inactive mitosis-promoting factor complex, consisting of cyclin B and its cyclin-dependent kinase p34^cdc2. Passage through mitosis and subsequent division requires MPF inactivation by degradation of cyclin B via the ubiquitin pathway (21). In this context, the elevation of cyclin B protein and p34^cdc2 kinase protein levels in Ad-infected cells suggests the possibility that Ad E4 gene products remaining in the ΔE1 and ΔE1ΔE3 Ad vectors can interfere with the degradation of cyclin B by ubiquination or affect the activity of upstream regulators of the cyclin B-cyclin kinase complex, notably Cdc25C, a phosphatase which activates p34^cdc2 (10).

The E1a region of the wild-type Ad genome is potentially oncogenic (79) and in this respect was considered to be the most likely candidate to affect the host cell cycle. Screening of our Ad vectors by PCR confirmed the absence of E1a transcripts. While Addl312, defective in E1a, has been reported to replicate in HeLa cells at MOIs of ≥80 PFU/cell (67), viral replication could not be detected in AdCFTR- or AdCL-infected IB3-1 or HT-29 cells (24 or 72 h at MOIs of 5, 25, and 200 [data not shown]), confirming the absence of wild-type E1a virus or cell-derived E1a-like function. Since G₂ arrest was observed in cells infected with clinical-grade AdCFTR vector, Ad replication could be further excluded. Ad5 E1a induces cellular DNA synthesis in quiescent cells (35), while infection of asynchronous populations of A549, HeLa, or KB cells with Ad12 E1A induces S-phase arrest (22). However, the S-phase fraction was unchanged in serum-starved quiescent IB3-1 cells infected with the ΔE1 or ΔE1ΔE3 Ad vector (data not shown), and Ad-infected cells did not arrest in S phase (Fig. 1 and 2), further confirming the absence of Ad E1a proteins.

Arrest at the G₂ stage of the cell cycle is not a phenomenon unique to adenoviruses. G₂-M arrest is observed in human foreskin fibroblasts infected with human CMV (31, 59) and in HeLa (55), 293 cells (28), and monocytes (56) infected with human immunodeficiency virus type 1. While the specific portion of the CMV genome responsible for cell cycle arrest is unknown, the human immunodeficiency virus vpr gene product appears to prevent activation of the p34^cdc2-cyclin B complex, possibly by interfering with upstream regulation of this complex (28). Consequently, the highly conserved mechanism regulating the G₂/M transition of eukaryotic cells appears to be a common target for a diverse group of viral gene products.

The overexpression of cyclin proteins and their inappropriate appearance in cell cycle compartments are important hallmarks in the oncogenic progression of cells (13, 29). Cyclin A overexpression has been implicated in the neoplastic transformation of alveolar epithelial cells (6), and cyclin D1, overexpressed in a number of diverse human cancers, may have a role in the development of non-small-cell lung cancer (61). Interestingly, hepatitis B virus integrates into the human genome at the cyclin A locus, resulting in the constitutive overexpression of cyclin A in hepatocarcinomas (72).

Aneuploid peaks have been observed in DNA histograms from human tissues infected by papillomavirus (7), herpes simplex virus types 1 and 2 (71), and Epstein-Barr virus (52). Several lines of experimental evidence support the conclusion that the increase in G₂-M cells (Fig. 1 and Table 1), aneuploidy in recultured IB3-1 cells (Fig. 7), and polyploidy in HT-29 cells (Fig. 8) results from the interaction of viral regions remaining in ΔE1 and ΔE1ΔE3 Ad vectors with cellular elements and does not reflect spurious histogram peaks arising from the contribution of Ad vector DNA itself upon PI staining. AdCFTR infections at 5 PFU/cell (G₂-M arrest in NHBE cells [Table 1]) correspond to 40 particles/cell (approximately 1/2,500 of genomic DNA content) and, at 200 PFU/cell, range from 1,600 (AdCFTR) to 4,400 particles/cell (AdCL), or 1/60 and 1/25 of genomic DNA, respectively, levels far lower than the DNA content of 96 human chromosomes present in G₂-M cells. More importantly, infection with the Ad2/CMVβgal-5 vector at 200 (HT-29) or 1,000 (IB3-1) PFU/cell corresponding to 1/14 or ∼1/3 of genomic DNA, respectively, had no effect on cell cycle distributions (Fig. 9). In the case of the recultured IB3-1 cells, the most probable explanation is that the aneuploid population represents proliferating Ad-infected cells while the diploid fraction corresponds to cells that escaped cell cycle arrest. By contrast, the single aneuploid DNA stemline observed in Ad-infected 3T3 cells that increases in an Ad dose-dependent manner could be the result of Ad infection limited to a specific cell cycle phase.

In a typical Ad clinical protocol, a single dose of 10⁹ virus particles is administered (19), and infection with 10,000 PFU/cell may be required to impart CFTR function to CF epithelia in vivo (24). The degree of cell cycle inhibition induced by ΔE1 and ΔE1ΔE3 Ad vectors varied between cell types (i.e., abnormalities in cyclin protein expression at an MOI of 5 in primary NHBE cells; polyploidization in HT29 cells at an MOI of 5; nearly a 100% increase in G₂-M cells in IB3-1 cells at an MOI of 20). In this context, it is conceivable that the cellular target which interacts with the gene product(s) from the Ad E4 region to induce G₂ growth arrest may be present at different levels in a variety of cells. The broad range of MOIs used to infect cells and deliver transgenes (0.5 to 1,000 PFU/cell and >10,000 PFU/cell) in various gene therapy studies reflects both the lack of consensus among investigators for standardization of Ad vector concentrations (47) and the fact that MOIs, especially in ex vivo and in vivo experiments, are estimated values based on arbitrary assumptions (for example, the density of cells in the nasal epithelium has been estimated at 2 × 10⁶ [45] or 3 × 10⁶ [83] cells/cm²). Consequently, the levels used here (i.e., an MOI of 5 to roughly double the percentage of G₂-M cells in NHBE cells) are within the range (MOI of 66 to 200) used clinically in an attempt to correct the defect in Cl⁻ transport in the nasal epithelium of CF individuals with an Ad vector containing the CFTR cDNA (83).

In the context of gene therapy for inherited diseases, G₂ arrest could conceivably prolong gene expression, normally reduced as cells undergo successive cycles of cell division. Cell differentiation usually occurs in the G_0/1 phase; however, in the case of CF, where nasal polyps have been shown to have increased proliferative activity (27) and it is estimated that nearly 20% of the airway cells are proliferating, G₂ growth arrest by ΔE1 and ΔE1ΔE3 Ad vectors may potentially slow remodeling of the respiratory epithelium (42). Although speculative, this hypothesis is consistent with the observation that the majority of the regenerating cells in explant cultures of human nasal polyp tissue, which were targeted by an AdLacZ vector, lacked reactivity with the Ki-67 antibody, which recognizes proliferating cells and is absent in only quiescent cell cycle compartments (16). In human bronchial xenograft models, Ad-mediated transgene expression was present in all cell types of the surface epithelium except basal cells (18). While it is presumed that these transfected cells arose from Ad-infected basal cells that subsequently underwent division and differentiation, these results are at odds with the concept of differentiation as a stochastic event, implying that all infected basal cells would have differentiated within 72 h of Ad infection.

Based on their ability to induce G₂ growth arrest, first-generation ΔE1 and ΔE1ΔE3 Ad vectors may be more appropriate for gene therapy of human cancers (15). Coadministration of genotoxic agents with Ad vectors conveying the cDNAs for G₁-phase growth regulatory genes (8, 11, 60), along with the cellular and humoral immune responses evoked by ΔE1 and ΔE1ΔE3 Ad vectors (81), might provide a multifaceted approach to target selectively proliferating tumor cells. In this respect, G₂ growth arrest was observed in a wide variety of colonic and mammary adenocarcinoma cell lines commonly used to assess the efficacy of chemotherapeutic agents in vitro and in vivo (77a). By contrast, second-generation vectors, having deletions in the E1 and E4 regions, may alleviate the problems reported here and minimize potential safety concerns regarding use of Ad vectors for gene therapy of inherited diseases of humans.

ADDENDUM

While this report was under review, Brand and Strauss (4a) reported that first-generation ΔE1 and ΔE1ΔE3 Ad vectors induce growth retardation and prolongation of the G₂-M phase in p53 knockout murine hepatocytes.