Canine Adenovirus Vectors: an Alternative for Adenovirus-Mediated Gene Transfer

Abstract

Preclinical studies have shown that gene transfer following readministration of viral vectors is often inefficient due to the presence of neutralizing antibodies. Vectors derived from ubiquitous human adenoviruses may have limited clinical use because preexisting humoral and cellular immunity is found in 90% of the population. Furthermore, risks associated with the use of human adenovirus vectors, such as the need to immunosuppress or tolerize patients to a potentially debilitating virus, are avoidable if efficient nonhuman adenovirus vectors are feasible. Plasmids containing recombinant canine adenovirus (CAV) vectors from which the E1 region had been deleted were generated and transfected into a CAV E1-transcomplementing cell line. Vector stocks, with titers greater than or equal to those obtained with human adenovirus vectors, were free of detectable levels of replication-competent CAV and had a low particle-to-transduction unit ratio. CAV vectors were replication defective in all cell lines tested, transduced human-derived cells at an efficiency similar to that of a comparable human adenovirus type 5 vector, and are amenable to in vivo use. Importantly, 49 of 50 serum samples from healthy individuals did not contain detectable levels of neutralizing CAV antibodies.

Human adenovirus types 2 and 5 were chosen as potential gene transfer vectors because of the significant amount of research performed on these serotypes. However, vectors derived from viruses that naturally infect and replicate in humans may not be the optimal candidates for therapeutic applications. Adenoviruses are ubiquitous in all populations and can be lethal in infants and immunocompromised patients (5, 18, 24). More than 90% of the adult population has detectable levels of circulating antibodies directed against antigens from human serotypes (9, 32, 33). Phase I trials using human adenovirus vectors have yielded conflicting results (8, 21, 41). A difference in humoral immunity that is directed against the vector capsid might explain, in addition to other factors, the variability between and within these studies. Furthermore, when repeat administrations were attempted (7, 42), transgene activity was not detected. Studies aimed at immunotolerization of mice, for the primary or repeat delivery of human adenovirus vectors, are interesting from the immunological standpoint but may have limited practical use in the clinic. Will immunotolerization of patients to adenovirus vectors activate latent, more virulent serotypes? Concomitantly, there are other drawbacks associated with human-derived adenovirus vectors. More than 95% of a healthy cohort had a long-lived CD4⁺ T-cell response directed against multiple human adenovirus serotypes (14). These data imply that adenovirus serotype switching (27) may have limited advantages. Furthermore, replication-competent adenoviruses (RCAs) (26) can potentially contaminate human adenovirus-derived vector stocks, including gutless adenovirus vectors (16, 22), while E1 region-positive vectors are a potential contaminant in vectors from which E1 and E4 have been deleted (ΔE1ΔE4 vectors) (40). In addition, recombination of the vector with a wild-type adenovirus, producing an RCA harboring a transgene, still remains a theoretical risk with early-generation vectors.

In order to address these issues, we previously tried to generate nonhuman adenovirus vectors from the Manhattan strain of canine adenovirus type 2 (CAV-2) (20). However, we were unable to generate a recombinant CAV vector derived from this serotype that was not significantly (>99%) contaminated with replication-competent CAV-2 particles. Replication-competent bovine, ovine, and avian adenovirus vectors have been described previously (28, 29, 37, 39) and currently appear useful as nonhuman vaccines. In order to generate vectors for gene transfer in the clinic, the potentially oncogenic CAV-2 E1 region must be deleted from the vector stock, and a CAV-2 E1-transcomplementing cell line must be generated in order to propagate the vectors. Here, we have generated nonhuman adenovirus vectors derived from the Toronto strain of CAV-2 using E1-transcomplementing cell lines derived from canine cells. These CAV vectors can be grown to high titers and are replication defective in canine cells, as well as in human cells that can transcomplement E1-deleted human adenovirus vectors. CAV vectors gave encouraging results after having been tested (i) in vitro in order to determine their ability and efficacy to transduce human-derived cell lines compared to a human adenovirus vector, (ii) for the absence of replication-competent CAV-2 contaminating the stocks, and (iii) for the particle-to-transduction unit ratio. In addition, in vivo tests show that CAV vectors can effectively transduce mouse airway epithelia when delivered intranasally. However, these CAV vectors and future derivatives will be useful only if there is no preexisting humoral immunity that can neutralize transduction. Here we show that sera from a majority of a random healthy cohort contain significant amounts of neutralizing adenovirus type 5 antibodies but not neutralizing CAV-2 antibodies.

MATERIALS AND METHODS

Cells.

DK (canine kidney cells; ATCC CRL6247), DK/E1-1 (20), DK/E1-28 (20), DK28Cre (a subclone of DK/E1-28), 911 (10), HT 1080 (ATCC CCL121), HeLa (ATCC CCL2), and A172 (ATCC CRL 1620) cells were grown in Dulbecco's modified Eagle medium (GIBCO)–10% fetal calf serum (BioWhittaker)–2 mM glutamine (GIBCO). DK/E1-1, DK/E1-28, and DK28Cre contain the CAV-2 E1 region stably integrated in the genome with the E1A region under the control of the cytomegalovirus (CMV) promoter and the E1B region under the control of its own promoter. In an effort to increase vector production, we tested two DK/E1-28 subclones for the ability to amplify the CAV vectors. One of the two, DK28Cre cells, gave a homogeneous infection pattern and a higher yield. DK/E1-1, DK/E1-28, and DK28Cre are derived from DK cells, an immortalized line.

Plasmids and viruses.

DNA preparations, restriction enzyme digests, and Southern blot analysis were performed under standard conditions (2). Details of the construction of the pretransfer and transfer plasmids, pCAVGFP and ptGFP, are available on request. Briefly, pCAVGFP contains the first 411 bp of the left end of CAV-2 and a green fluorescent protein (GFP) expression cassette containing a CMV early region enhancer/promoter, a simian virus 40 (SV40) intron with splice donor and acceptor sites, the humanized red-shifted version of the Aequorea victoria GFP (EGFP; Clontech), and an SV40 polyadenylation site followed by bp 2898 to bp 5298 of CAV-2 cloned into pSP73 (Promega). The expression cassette is transcribed from right to left in plasmids and GFP-expressing CAV vectors. pCAVβgal has been described elsewhere (20) and contains the Rous sarcoma virus promoter driving expression of lacZ. pTG5412, a generous gift from Transgene SA, contains the CAV-2 genome (strain Toronto A 26/61; GenBank accession no. J04368) flanked by NotI sites cloned in pPolyII. pTG5412 was generated by the same strategy used to generate pTG3602 (6) except that NotI linkers were used instead of PacI.

ptGFP and other transfer plasmids used to produce vectors were generated by in vivo homologous recombination in Escherichia coli BJ5183 according to the work of Chartier et al. (6) by using SwaI-linearized pTG5412 and a fragment containing the inverted terminal repeat, the GFP expression cassette, and the CAV-2 E2B regions. CAVGFPΔE1A has a deletion in the CAV-2 genome from bp 411 to bp 1024. ptGFP, and therefore the virus CAVGFP, has a deletion in the CAV-2 genome from bp 411 to bp 2898. CAV vectors were partially sequenced directly from low-molecular-weight DNA preparations from infected DK/E1-28 cells to verify their integrity. AdGFP is a first-generation ΔE1ΔE3 human adenovirus type 5 vector containing a GFP expression cassette similar to the one in the CAV vectors except that the transcription unit is oriented left to right and contains NotI sites flanking the transgene.

CAV vector preparation.

The preparation of CAVGFP is described here; other CAV vectors were prepared similarly. Transfections in DK/E1-1 cells were carried out with 5 μg of NotI-digested ptGFP and 20 μl of Lipofectamine (GIBCO) in 6-well plates containing approximately 10⁶ cells. DK/E1-1 cells were collected when a cytopathic effect was detected 1 to 2 weeks posttransfection, and the vector was freed from the cells by four freeze-thaw cycles and centrifugation to remove cellular debris. The cleared lysate was incubated with a fresh monolayer of DK/E1-28 cells and collected 48 h postinfection. This was repeated four to five times until a prestock of 10 10-cm-diameter dishes showed a complete cytopathic effect 48 h postinfection. This “prestock” was used to infect 50 15-cm-diameter plates of DK/E1-28 cells. Forty hours postinfection the cells were collected, and the vector was freed by four freeze-thaw cycles. Approximately 7 ml of cleared lysate was layered on a CsCl step gradient of 1.4 and 1.25 g/ml (2.5 ml each layer) and centrifuged for 90 min with a Beckman SW41 rotor at 35,000 rpm. The CAVGFP band was removed and further purified on a CsCl isopycnic gradient at a density of 1.32 g/ml (versus the 1.34 g/ml used for human adenovirus vectors) for 18 h by using the same speed and rotor. Both centrifuge runs were carried out at 18°C. CAV vectors banded at a density of ∼1.22 g/ml. CsCl was removed by using PD-10 columns (Pharmacia), and the virus was stored in phosphate-buffered saline (PBS) containing 10% glycerol.

Titration of CAV-2, AdGFP, and CAV vectors.

Vector concentrations were determined by the optical density at 260 nm by using two dilutions of two aliquots of each virus or vector stock as described previously (30). We have assayed the particle-to-transduction unit ratio in the most sensitive assay we could develop. DK28Cre cells are the largest of the three cell types tested (DK, DK/E1-28, and DK28Cre), are the most sensitive to CAVGFP infection, and give a homogeneous infection pattern. For the transduction unit titration of CAVGFPΔE1A and CAVGFP, DK/E1-28 or DK28Cre cells were seeded in 12-well plates and infected overnight with gentle rocking with twofold dilutions beginning with 1.25 × 10⁶ viral particles/well. Twenty-four hours postinfection, the cells were analyzed by flow cytometry (FACSCalibur; Becton Dickinson), and the percentage of GFP-positive cells was determined and used to calculate the particle-to-transduction unit ratio [(input viral particles) (GFP-positive cells)⁻¹]. Mock-infected cells and cells infected with CAVΔE1 were used as negative controls, and no background fluorescence was detected. AdGFP was similarly titrated on 911 cells, which were used because they are threefold more sensitive to human adenovirus vectors than 293 cells (10). PFU titration of AdGFP and CAV-2 was determined as follows: 0.5 ml of 10-fold dilutions of virus or vector was incubated with a confluent monolayer of 911 or DK cells in a 30-mm-diameter well overnight before a layer of agarose was used to cover the cells. The titer was determined 6 or 14 days postinfection, respectively.

In order to determine if there was background from GFP transfer (pseudotransduction), 12-well plates containing a confluent monolayer of DK28Cre cells were infected at 4°C with CAVGFP for 4 and 6 h at an input ratio of approximately 10³ particles/cell. The plate was rocked continuously and then transferred to 37°C for 30 min, the cells were trypsinized, and an aliquot was assayed by flow cytometry. The remaining cells were returned to 37°C and 6% CO₂ and were analyzed by flow cytometry 24 h postinfection.

RCA assays.

A total of 2.5 × 10¹⁰ particles of CAVβgal (divided equally into nine 15-cm dishes) and 5 × 10¹⁰ particles of CAVGFP (17 dishes), from two separate stocks, were assayed. Each dish, containing 1.3 × 10⁸ DK cells/plate, was incubated overnight with 2.7 × 10⁹ to 3.0 × 10⁹ particles of CAVGFP or CAVβgal/plate (maximum of 23 particles/cell) with gentle rocking in a humidified chamber at 37°C. The plates were removed from the shaker and placed in an incubator (6% CO₂ at 37°C) for 5 to 6 days before the cells were collected, and the cleared lysate was used to inoculate a second plate containing 5 × 10⁷ DK cells. The cleared lysate was removed from the cells 1 to 2 days later, fresh medium was added, and the cells were collected 3 to 4 days later. This was repeated until the positive controls (two 15-cm plates containing DK cells infected with 3.0 × 10⁹ particles of CAVGFP, spiked with 10² particles of CAV-2, and amplified as above) showed an extensive CAV-2-induced cytopathic effect (3 passages). The cultures transduced with CAVGFP and CAVβgal were passed an additional time and still showed no sign of cytopathic effect.

Transduction of human cells: CAVGFP versus AdGFP.

To compare infection efficiencies on human cell lines, identical 24-well plates containing monolayers of HeLa, HT 1080, or A172 cells (approximately 10⁶ cells/well) were infected with fivefold dilutions of CAVGFP or AdGFP starting with 4.3 × 10⁸ or 1 × 10⁹ particles/well, respectively. The cells were collected 48 h posttransduction and assayed for GFP expression by flow cytometry. The number of particles needed to generate 10% GFP-positive cells was calculated. Ten percent was in the range of 1 transduction unit/GFP-positive cell.

In vivo use of CAVβgal, CAVGFP, and AdGFP.

All mice were treated according to the rules governing animal care for the European Community. Eight-week-old BALB/c mice (n = 10) were lightly anesthetized with halothane (Belamont), and 10¹¹ particles diluted in PBS (total volume, 100 μl) were delivered intranasally. Mice were sacrificed on day 3, 4, or 21, and the lungs were recovered following perfusion with 2% paraformaldehyde and embedded in OCT (Tissue-Tek). To detect β-galactosidase activity, 10- to 20-μm sections were incubated overnight at room temperature in 66 mg of 5-bromo-3-indolyl-β-d-galactopyranoside/ml–2 mM MgCl₂–4 mM K₃Fe(CN)₆–4 mM K₄Fe(CN)₆ in PBS, counterstained with eosin, and analyzed for β-galactosidase activity. GFP expression was detected by using a Zeiss Axiovert fluorescent microscope with an EGFP filter (485 to 507 nm) at an original magnification of ×10.

Neutralizing adenovirus antibodies.

Fifty samples of whole blood were purchased from the Centre Transfusion de Rungis (Rungis, France). Serum was separated and complement inactivated at 56°C for 30 min. Ten microliters of serum was mixed with 100 μl of medium containing 5 × 10⁷ vector particles (CAVGFP or AdGFP) for 1 h at room temperature prior to incubation with 911 cells. The cells were tested for GFP expression by flow cytometry 24 h postinfection. For each sample, this procedure was carried out in duplicate and repeated. Results similar to those with AdGFP were obtained with an adenovirus type 2 vector expressing GFP (data not shown).

RESULTS

Isolation of CAVGFP.

Four adenovirus vectors derived from CAV-2 are described here: CAVGFPΔE1a and CAVGFP, which harbor the gene encoding GFP; CAVβgal, encoding nuclear localized β-galactosidase; and CAVΔE1, which contains a null expression cassette (see Materials and Methods and Table 1). Figure 1 shows a diagram of the plasmid used to generate CAVGFP. NotI-digested ptGFP, which places the inverted terminal repeats at the extremities of the DNA fragment and allows vector replication, was transfected into DK/E1-1 cells. In order to further characterize DK/E1-1 and DK/E1-28 cells, the CAV-2 E1 expression cassette was amplified by PCR from total genomic DNA and the PCR product was sequenced. The sequence was identical to that of the transfected plasmid and to the published CAV-2 sequence (data not shown).

TABLE 1.

Summary of virus and vectors

Virus or vector	No. of particles/ml	Particle/transduction unit ratioa	% of wild-type genomeb length	Replication competent
AdGFP	1.3 × 1012–4 × 1012	10:1–26:1	88.1	NTc
CAV-2	1.9 × 1012	NA	100	NA
CAVGFP	0.3 × 1012–2.3 × 1012	3:1–7:1	97.7	Nod
CAVΔE1	3.7 × 1011	NA	95.2	NT
CAVβgal	2.6 × 1012–5.2 × 1012	10:1	105.7	Noe

^aIn producer cell line. NA, not applicable. 

^bAdenovirus type 5 or CAV-2. 

^cNT, not tested. 

^dFewer than 1 to 2 RCAs per 5 × 10¹⁰ particles. 

^eFewer than 1 to 2 RCAs per 2.5 × 10¹⁰ particles. 

FIG. 1.

Generation of CAVGFP. ptGFP was generated by homologous recombination in E. coli BJ5183 by using SwaI-linearized pTG5412 and a 4.7-kb BglII/FspI fragment from pCAVGFP (nucleotide positions are from CAV-2; diagram not drawn to scale). NotI-digested ptGFP was transfected into DK/E1-1 cells and amplified five to six times on E1-transcomplementing cells, and CAVGFP was purified as described in Materials and Methods.

Because we were using GFP as the transgene, we were able to monitor the propagation of the vector posttransfection. One to two weeks later, the cells were collected and the cleared lysate was used to amplify the vector. CAVGFP and CAVGFPΔE1a DNAs were extracted from CsCl-purified vector stocks and digested with EcoRI (Fig. 2a). All digests gave the anticipated pattern for each vector compared to the respective transfer plasmid. No contaminating bands were detectable by ethidium bromide staining in any restriction enzyme digests (n = 6). These results demonstrated that these CAV vectors are free of gross rearrangements, deletions, or insertions. CAVβgal DNA was also analyzed by restriction enzyme digests (n = 5), and no extraneous bands were detected (data not shown). In order to further verify the integrity of CAVGFP, the digestions were assayed by Southern blot analysis using PCR-generated fragments from the CAV-2 E1 region (bp 458 to 936) (Fig. 2b) or the GFP cDNA (Fig. 2c) as the radiolabelled probe. No signal was found in pTG5412 for the GFP-derived probe, as expected, while fragments of the predicted sizes, 2.89 and 4.75 kb (Fig. 2d), were detected in the CAV vectors. The E1 region probe hybridized to the 3.6-kb band in pTG5412, as expected, but failed to hybridize specifically to CAV vector sequences. Southern blot analyses confirmed that the vectors did not acquire E1A-derived sequences during isolation or amplification.

FIG. 2.

Digestion and Southern blot analysis of CAVGFP and CAVGFPΔE1A DNAs. (a) Vector or plasmid DNA (500 ng) was digested with EcoRI and NotI (in order to remove the 2-kb pPolyII backbone from the terminal fragments seen in lanes 1, 2, and 4), electrophoresed through a 0.7% agarose gel, and stained with ethidium bromide. Lanes: 1, pTG5412; 2, ptGFPΔE1A; 3, CAVGFPΔE1A; 4, ptGFP; 5, CAVGFP; M, 1-kb DNA ladder (GIBCO). Southern blot analysis was performed by using a fragment of the E1A region (b) or GFP cDNA (c) as the radiolabelled probe. (d) Locations of the EcoRI sites and the fragment sizes in the vectors.

Vector preparation, titration, and purity.

Stocks of CAVGFP containing 2.3 × 10¹² particles/ml, with a particle-to-transduction unit ratio of less than 3:1, were generated. The CAVGFP vector yield was ∼10⁴ particles/cell, similar to the ratio found when PERC.6 cells were used to produce first-generation human adenovirus vectors (11). Due to the exceptionally low particle-to-transduction unit ratio in CAVGFP, we asked if the capsid contained GFP and therefore we were detecting protein transfer instead of gene transfer. In order to assay this, purified CAVGFP was incubated with DK28Cre cells at 4°C to allow attachment of the vector to the cellular receptor. The cells were placed at 37°C to induce internalization of the vector and were analyzed by flow cytometry. Subsequently, the cells were returned to the incubator and were assayed by flow cytometry 24 h postinfection. No GFP-positive cells were detected following the attachment-internalization step, while 34% of the cells were GFP positive 24 h postinfection, demonstrating that this assay was detecting gene transfer and not protein transfer. CAVGFP is 97.7% of the size of the wild-type CAV-2 genome (31,322 bp), CAVΔE1 (not shown) is 95.2%, and CAVβgal is 105.7%. Stocks of CAVβgal were generated at a concentration of 5.2 × 10¹² particles/ml and a particle-to-transduction unit ratio of approximately 10:1.

With human adenovirus vectors, the generation of RCAs and E1 region-containing particles during stock preparation is a significant clinical concern. With CAV vectors, the risks associated with RCAs are diminished, if not completely eliminated, because CAV-2 does not propagate in human cells. However, the E1 region of many adenoviruses encodes potentially oncogenic proteins that can transform or immortalize cells in vitro and in vivo (36), and therefore the E1 region must be deleted from an adenovirus vector if it is to be used in patients. We generated E1-transcomplementing cells to propagate these vectors and designed the cell line in order to try to reduce the likelihood of generating replication-competent CAV-2. CAVGFP and CAVβgal stocks were tested for the presence of replication-competent CAV-2 by serial amplification on permissive cells (DK cells). The sensitivity of this assay was 1 to 2 PFU/5 × 10¹⁰ particles, as 100 particles of CAV-2 (1 PFU/66 particles) were used to spike 3 × 10⁹ CAVGFP particles/plate as a positive control. We were unable to detect a CAV-2-induced cytopathic effect, demonstrating the lack of replication-competent CAV-2 in 5 × 10¹⁰ particles of CAVGFP (2.5 × 10¹⁰ particles of each stock) and 2.5 × 10¹⁰ particles of CAVβgal.

Transduction of human-derived cells: CAVGFP versus AdGFP.

We demonstrated previously, using a qualitative assay, that a CAV vector derived from the Manhattan strain of CAV-2 could transduce human-derived cells (20). However, it was impossible to determine the efficacy of transduction because the “vector stock” contained significant amounts of CAV-2 (the virus/vector ratio was >10,000:1). In order to determine the quantitative transduction efficiency with the CAV vector described here (Toronto strain), three human cell lines, HT 1080, HeLa, and A172 cells, which are derived from different cell lineages (osteosarcoma, cervical carcinoma, and glioblastoma), were quantitatively assayed for transducibility. Multiwell plates, containing equal numbers of each cell type, were incubated with a serial dilution of CAVGFP and AdGFP. Forty-eight hours posttransduction the cells were assayed for transgene expression by flow cytometry. Figure 3 shows the particle-to-cell ratio needed to generate 10% GFP-positive cells/well. In each cell line, CAVGFP was 5- to 10-fold more efficient (lower number of particles needed) than AdGFP when the particle/cell ratios were compared. However, we have found that the quality of adenovirus vector preparations can vary significantly. If the comparison between CAVGFP and AdGFP is plotted as transduction units per cell versus percent GFP-positive cells per well, the transduction efficiency of AdGFP in HeLa cells is slightly greater than that of CAVGFP (Fig. 3b).

FIG. 3.

Quantitative analysis of transduction efficiencies of CAVGFP and AdGFP in human cells. (a) A172, HeLa, and HT 1080 cells were infected with each vector and assayed for GFP expression 48 h posttransduction. Data represent the amount of vector required to generate 10% GFP-positive cells/well, expressed as the number of input particles/well. Data are means from five experiments ± standard deviations. (b) HeLa cells incubated with an increasing number of particles of CAVGFP and AdGFP and analyzed by flow cytometry 24 h posttransduction. Data are the means ± standard deviations from triplicate samples.

In vivo use of CAV vectors and comparison to AdGFP.

An in vivo study was used to assay the utility of CAV vectors. We delivered 10¹¹ particles of CAVβgal and CAVGFP intranasally in 8 week-old BALB/c mice. Figure 4a through c demonstrate lacZ expression in the airway epithelia 4 days postinoculation. We detected nuclear-localized β-galactosidase activity throughout the proximal and distal airways and in the alveoli. In several cases more than 50% of the cells in a given bronchiole were β-galactosidase positive. In some instances where expression was detected in the alveoli, thickening of the cell walls was visible (data not shown), suggesting cellular infiltration, and at 21 days posttransduction we were unable to detect β-galactosidase activity (n = 3). CAVGFP was able to transduce a slightly higher proportion of airway cells, and in several cases more than 65% of a given distal airway was GFP positive (Fig. 4d and e). Comparison of the transduction efficiency of CAVGFP versus AdGFP (Fig. 4e versus Fig. 4g) demonstrates that CAV vectors can be as efficient in vivo as those derived from human adenoviruses.

FIG. 4.

In vivo transduction of the airway epithelia in mice by using CAV vectors. A total of 10¹¹ particles of CAVβgal, CAVGFP, or AdGFP was delivered intranasally in BALB/c mice, and lung sections were assayed 3 or 4 days later. (a through c) Nuclear-localized β-galactosidase activity from CAVβgal, as demonstrated by the blue precipitate, in the proximal (a) and distal (b) airways and in the alveoli (c). (d through g) GFP expression in distal airways from CAVGFP (d and e) and AdGFP (f and g) as shown by phase contrast (d and f) and fluorescence (e and g).

Pre-existing humoral immunity.

The majority of individuals has been exposed repeatedly to adenoviruses and, not surprisingly, have detectable neutralizing adenovirus antibodies. We tested serum samples from a random healthy cohort (n = 50) for the ability to neutralize AdGFP and CAVGFP transduction. Figure 5 demonstrates that in most cases (26 of 50), as little as 10 μl of human serum contains sufficient amounts of neutralizing adenovirus type 5 (as well as adenovirus type 2 [data not shown]) antibodies to rapidly and completely inactivate 5 × 10⁷ AdGFP particles. These sera rarely (1 of 50) contain detectable neutralizing CAV-2 antibodies. In vivo, airway epithelia transport both immunoglobulin G and immunoglobulin A to the thin layer of liquid that covers the apical surface of the epithelium and thus can prevent adenovirus infection. These data are particularly significant because if one cannot circumvent this initial barrier for adenovirus-mediated gene transfer, use of human adenovirus vectors becomes limited. These CAV vectors and, importantly, more advanced versions are not inhibited at this stage.

FIG. 5.

Preexisting humoral immunity. Sera from healthy blood bank donors (n = 50) were assayed for the presence of neutralizing CAV-2 antibodies. In this assay only one sample was partially able (∼24%) to inhibit CAVGFP transduction, while 26 of 50 samples completely inactivated AdGFP transduction.

DISCUSSION

We have generated a system to produce CAV vectors for gene transfer. Several explanations for our ability to generate RCA-free CAV vectors using this strategy, as opposed to our previous attempts (20), seem plausible. The previous strategy (transfection of two linear fragments of DNA in DK/E1 cells in hopes that homologous recombination would occur and generate a recombinant vector) was similar to that used to generate first-generation human adenovirus vectors (23). Firstly, DK cells and their derivatives are difficult to transfect; normally, the efficiency is lower than 15%. Secondly, DK cells may also be less efficient at homologous recombination than 293 cells. Finally, the Manhattan strain of CAV-2 was unstable—it is able to generate at least 23 repeats of ∼120 to 150 bp in the right inverted terminal repeat (unpublished data and reference 13). All of these factors may have prevented us from isolating pure vectors. Using the strategy described here, we have eliminated the need for high transfection efficiency, homologous recombination in the cell line, and the presence of the unstable sequence in the inverted terminal repeat.

The stable packaging capacity of the human adenovirus type 5 vectors was determined to be a minimum of 75% (31) and maximum of 105% of that of the wild-type genome (4). The sizes of the GFP-expressing CAV vectors are within this range, while CAVβgal is slightly larger (see Table 1) and appears to be stable. Xu et al. (39) reported the creation of an ovine adenovirus vector that is 114% of the wild-type genome, demonstrating that the cloning capacities of this and other adenoviruses may not mimic that of the adenovirus type 5 vectors. Stocks of CAVGFP have a particle-to-transduction unit ratio as low as 3:1, while CAVβgal stocks have a particle-to-transduction unit ratio of approximately 10:1. Mittereder et al. (30) have carefully detailed the physical and biological parameters used to titer adenovirus vectors. Taking their work into account, the particle-to-transduction unit ratio may be an underestimation of the true titer, due to undetectable transgene expression from transduction occurring later during the incubation period. More CAV vectors will need to be generated to determine if the low particle-to-transduction unit ratio in these initial stocks is a general trend, an exception in these cases, or due to a more sensitive quantification assay.

All the CAV vectors, including E1A-deleted vectors, are replication defective in DK, MDCK, and, more significantly, 911 cells. This demonstrates that there is an undetectable level of transcomplementation of the adenovirus type 5 E1-derived proteins in these cells for CAV vector propagation (unpublished data). Although contamination of CAV vector stocks with RCA is certainly undesirable, it is significantly less dangerous than contaminating replication-competent human adenovirus that may be below the level of detection. The E1-transcomplementing cell lines described here do not contain the CAV-2 inverted terminal repeat or the packaging signal found at the left end of the CAV-2 genome but do contain a 55-bp overlap in the E1A promoter with the vectors described here. We are generating CAV vectors that do not contain an overlap in this region, and all subsequent CAV vectors will not be able to generate RCAs via the in vivo mechanism characterized by Hehir et al. (17).

As mentioned previously, adenovirus infections can be dangerous in infants and immunocompromised patients. RCAs have been found in patients' tonsils, adenoids, and intestines, and patients can continue to shed adenovirus intermittently for many months after a successful humoral response (18). Immunotolerization against a ubiquitous, potentially lethal virus may expose patients to unacceptable risks. If immunotolerization is an unavoidable requirement for adenovirus-mediated therapy, our data demonstrate that it may be contemplated when CAV-2-derived vectors are used. Furthermore, reducing the viral input load due to a lower particle-to-transduction unit ratio (Table 1) will diminish the induced immune response to the virus capsid.

The population as a whole is continually being exposed to wild-type adenoviruses, and the clinically relevant data presented here demonstrate that a significant proportion (98%) of this cohort has not generated neutralizing CAV-2 (Toronto strain) antibodies. Using inbred rodent strains to assay induced humoral or cellular immunity to a human adenovirus vector, followed by a challenge with CAV vectors, may also allow one to detect anti-human adenovirus antibodies that opsonize rather than neutralize CAV vectors. Cross-species barriers to adenovirus infections exist not because of a lack of infectibility, but, at least in part, due to the incompatibility of viral and cellular factors. For example, human adenoviruses grow poorly in monkey cells due to the inefficient transport or processing of the E4 and late region primary transcripts (34, 35).

We demonstrated that CAV vectors could efficiently transduce human cells and that the transduction efficiency was at least equal to that of an adenovirus type 5 vector carrying the same expression cassette. Analysis of the efficacies of various human adenovirus serotypes suggests that adenovirus types 2 and 5 may not be the optimal adenovirus serotypes for gene transfer in many tissues, and therefore our comparison using CAV vectors is useful but not all-encompassing. It will be interesting to determine if the receptors used by CAV-2 are the same as those used by adenovirus type 5 (3). However, the future of viral vectors will be with tissue-specific transduction and, in the case of adenovirus vectors, the fiber knob may be modified accordingly (38). Alternatively, efficient in vivo fiber-independent transduction using adenovirus vector-calcium phosphate precipitates (12) or polycations (19), which increase the transduction efficiency 10- to 100-fold, may be applicable.

The CAV-2 fiber appears to be a trimer, as determined by protein sequence analysis (1) and comparison to human adenovirus types 2, 40, and 41. The Toronto strain of CAV-2 has been shown to preferentially infect the upper respiratory tract of dogs but has also been found in the feces of infected animals (15). This tropism has been suggested to be due not only to the expression of the receptor but potentially to the role of the E3 region (25). We tested CAV vectors via intranasal delivery in BALB/c mice and detected effective transduction in proximal- and distal-airway cells, as well as in the alveoli. We did not detect a site preference in the lung, and the disappearance of β-galactosidase activity and GFP expression suggested that there would be little difference between E1-deleted CAV and human adenovirus vectors with respect to the inevitable immune response. It would be useful to know if viral backbone genes are expressed in human cells treated with CAV vectors, and an analysis of CAV gene expression may help determine the future course of vector development, specifically, if some or all of the backbone needs to be deleted. Improved CAV vectors are being developed to address these issues. In summary, if adenovirus vectors are to be used, CAV-derived vectors and other nonhuman adenovirus vectors could be a safe and effective alternative.