Mutational Analysis of the Avian Adenovirus CELO, Which Provides a Basis for Gene Delivery Vectors

Abstract

The avian adenovirus CELO is being developed as a gene transfer tool. Using homologous recombination in Escherichia coli, the CELO genome was screened for regions that could be deleted and would tolerate the insertion of a marker gene (luciferase or enhanced green fluorescent protein). For each mutant genome, the production of viable virus able to deliver the transgene to target cells was monitored. A series of mutants in the genome identified a set of open reading frames that could be deleted but which must be supplied in trans for virus replication. A region of the genome which is dispensable for viral replication and allows the insertion of an expression cassette was identified and a vector based on this mutation was evaluated as a gene delivery reagent. Transduction of avian cells occurs at 10- to 100-fold greater efficiency (per virus particle) than with an adenovirus type 5 (Ad5)-based vector carrying the same expression cassette. Most important for gene transfer applications, the CELO vector transduced mammalian cells as efficiently as an Ad5 vector. The CELO vector is exceptionally stable, can be grown inexpensively in chicken embryos, and provides a useful alternative to Ad5-based vectors.

Adenovirus has been studied for its role in human disease (25), as a model for many important discoveries in molecular biology, including mRNA splicing, DNA replication, transcription, and cell transformation (reviewed in reference 44), and more recently as a powerful reagent for transient gene expression (12, 46). A detailed description of the adenovirus life cycle is well established (reviewed in reference 50). Since the initial efforts to use adenovirus as a gene transfer vector (18, 27, 52), the virus has gained in popularity as a vector and a number of methods of generating alterations in the viral genome to carry novel genes have been developed (see references 2, 5, 11, 15, 21, 23, 26, 32, 38, 41, 43, and 48 and the reviews in references 31 and 49). Because of the ease of vector construction and purification and because these vectors have a potent ability to transiently transduce novel genetic material into a variety of mammalian cell types in vivo, adenovirus vectors were used extensively in early efforts at clinical gene therapy. Unfortunately, several features of the adenovirus type 5 (Ad5)-based vectors initially used have limited the success in the initial applications. These included both the host immune response to adenovirus (reviewed in reference 55) as well as the failure of the virus to efficiently enter certain target cell types (20, 58, 59). Thus, there is now an interest in adenovirus types that could provoke less aggressive host immune responses and could enter target cells with greater efficiency.

A large number of alternate adenovirus serotypes are known and may provide advantages in some applications over Ad5-based vectors. Additional adenoviruses that have recently been modified as vectors include the ovine adenovirus 287 (53, 56), the bovine adenovirus type 3 (40, 60), and the canine adenovirus (30). It is considered that these alternate serotypes would provide a novel vector backbone to which there is no preexisting immune response in the target host. Furthermore, because adenoviruses are extremely species specific in their replication capacity (50), a degree of security against inappropriate vector replication is gained by using an vector derived from a distant species of adenovirus.

CELO (chicken embryo lethal orphan or fowl adenovirus type 1; reviewed in reference 39) was characterized as an infectious agent in 1957 (57). However, there are few serious health or economic consequences of CELO virus infection. CELO can be isolated from healthy chickens and, in general, does not cause disease when experimentally reintroduced into chickens (10).

CELO virus is structurally similar to the mammalian adenoviruses, with an icosahedral capsid of 70 to 80 nm made up of hexon and penton structures (33); the CELO virus genome is a linear, double-stranded DNA molecule with the DNA condensed within the virion by virus-encoded core proteins (33, 36). CELO virus has a larger genome than Ad5 (44 kb versus ca. 36 kb [6]). The CELO virion has been reported to have two fibers of different lengths at each vertex (24, 33, 35) rather than the single fiber of most other serotypes (reviewed in reference 50). CELO virus is not able to complement the E1A functions of Ad5, and CELO virus replication is not facilitated by Ad5 E1 activity (37). The complete DNA sequence of CELO (6) revealed additional differences between CELO virus and the mastadenoviruses, including the absence of sequences corresponding to the Ad5 early regions E1A, E1B, E3, and E4. The CELO genome contains approximately 5 kb of sequence at the left end and 12 kb at the right end, rich in open reading frames, which have no sequence homology to Ad5 but probably encode the early functions of the virus.

We are developing CELO into a gene delivery vector. The virus is naturally defective in mammalian cells, and this property should limit the possibility of complementation by wild-type mammalian adenovirus. The CELO virion has increased DNA packaging capacity and much greater physical stability than the Ad5 virion. One practical feature of CELO is the ability to grow the virus in chicken embryos, a system of low cost and high convenience (9, 33).

This article reports our efforts to characterize the frontier sequences of CELO, i.e., the leftmost 5 kb and rightmost 13 kb of the CELO genome that are largely unexplored and are not common to other adenoviruses. A series of altered CELO genomes were constructed, bearing specific deletions combined with the insertion of a luciferase expression cassette. The modified CELO genomes were engineered as bacterial plasmids using homologous recombination in Escherichia coli (5). Subsequently, after their release from the plasmid backbone by enzymatic digestion, the viral genomes were transfected into a chicken cell line supporting wild-type CELO virus replication. Monitoring the production of luciferase in cells treated with lysates of the initial transfectants allowed us to determine if virus replication and production of transducing viral particles had occurred. These strategies were used to determine the essential portions of both left and right frontier sequence. As anticipated, some of the sequences were required in cis and presumably these contain packaging signals, transcriptional promoters, or other transcription signals. In addition, this study has defined a number of virus sequences that were essential for virus replication yet could be supplied in trans. This information will be useful in the design of subsequent replication-defective viruses and complementing cell lines.

A major practical outcome of this study was the development of a CELO vector, CELO AIM46, that carries a deletion of CELO sequence near the right end of the genome, allows the insertion of an expression cassette for novel genes, and replicates without complementation. Cell tropism studies with CELO AIM46 demonstrated that the CELO derivative delivered genes into avian cells with efficiencies 10- to 100-fold better than an Ad5 vector. Surprisingly, in a variety of mammalian cell types, CELO AIM46 functions with efficiencies that are comparable to those of an Ad5 vector, demonstrating the utility of CELO vectors for mammalian gene transfer applications.

MATERIALS AND METHODS

Cloning the terminal fragments of the CELO genome.

The two terminal HindIII fragments of CELO were cloned. CELO genomic DNA (CsCl purified) was digested with HindIII; the 1,601-bp left end fragment and the 959-bp right end fragment were purified from a low-melting-point agarose gel. The 5′ ends of adenovirus genomes are linked by a phosphodiester bond to a serine residue on the viral terminal protein (22, 47). The peptide remaining after proteinase digestion must be removed to allow ligation and cloning. Accordingly, the terminal peptides were removed from the DNA fragments by adding NaOH to 0.3 N and heating at 37°C for 90 min (22). The solutions were then cooled to room temperature; Tris pH 7.4 was added to 0.1 M and HCl was added to 0.3 M to neutralize the NaOH. The fragments were heated to 56°C for 20 min and slowly cooled to room temperature (1 h) to facilitate reannealing. The DNA was then purified (Qiaquick column; Qiagen), SpeI linkers (New England Biolabs; catalog no. 1085) were added, and each fragment was cloned via SpeI and HindIII sites into a pBR327 (GenBank accession no. L08856 [51]) derivative containing an SpeI site in a destroyed EcoRI site (Fig. 1A). Both the left and the right terminal HindIII fragments were cloned in this manner, and DNA sequence analysis was performed to verify the terminal 300 bp of both fragments.

FIG. 1.

Construction of plasmids carrying wild-type and mutant CELO genomes. (A) Construction of a plasmid bearing the termini of the CELO genome. CELO genomic DNA was digested with HindIII, and the two terminal fragments were isolated, treated to remove the terminal peptides, and cloned as SpeI-HindIII fragments after the addition of SpeI linkers to generate the plasmid pWü-H35. (B) Cloning of full-length CELO genome as a bacterial plasmid. pWü-H35 was linearized with HindIII and recombined with CELO genomic DNA to generate the full-length plasmid clones of the CELO genome. The natural terminal repeats were flanked by SpeI sites to allow excision of the viral genome from the bacterial plasmid. There are no SpeI sites within the CELO genome. (C) Cloning of modified versions of the CELO genome. Transfer vectors were produced by manipulating subfragments of the CELO genome, as either pWü-H35 (with the terminal HindIII fragments of CELO) or pWüHpa (with the terminal HpaI fragments of CELO), using standard ligation cloning methods in order to delete portions of the CELO genome and insert a luciferase cassette. The linearized transfer vector was recombined with wild-type genomic CELO DNA. Recombination occurs in two ways, either to include the deletion/luciferase cassette or to exclude the deletion/luciferase cassette to generate a wild-type CELO plasmid. Plasmids bearing the desired mutation were identified and used to initiate virus infection.

Subsequently, the two CELO end fragments were cloned into a pBR327 derivative containing an SpeI site, a destroyed EcoRI site, and a ClaI/BamHI excision to remove the second HindIII site, creating the plasmid pWü-H35 (Fig. 1A). Details and full sequence information of these intermediate plasmids are available upon request.

Cloning the entire CELO genome.

HindIII-linearized, alkaline phosphatase-treated pWü-H35 vector was mixed with purified CELO virus DNA and introduced into electrocompetent E. coli JC8679 (17, 42) by electroporation. Recombination between the CELO terminal sequences on pWü-H35 and the termini of the CELO genomic DNA generated a plasmid containing a full-length CELO genome (pCELO) flanked by SpeI sites (Fig. 1B).

Modifications in the left end of the CELO genome.

The luciferase cassette containing the cytomegalovirus (CMV) immediate-early enhancer/promoter and luciferase cDNA (14) followed by a rabbit β-globin intron/polyadenylation signal was derived from pCLuc (45), modified by PCR to add flanking BamHI sites, and cloned into pBlueScript II (SK) to generate pBlueLuc. For most of the CELO insertions, the luciferase cassette was isolated from pBlueLuc by BamHI digestion and the termini were made blunt by treatment with Klenow enzyme.

Modifications of the left end CELO region were made by using pAIM3, which contains, on a pBR327 backbone, the CELO left end (nucleotide [nt] 1 to 5501) and a portion of the right end (nt 30500 to 30639 and nt 40065 to 43804; derived by removing an Asp718 fragment from pWüHpa). The deletions in the CELO genome involved digestion of pAIM3 at two enzyme sites in the CELO left end sequence, excision of the subregion, and then insertion of a luciferase cassette. All manipulations were confirmed by restriction analysis and sequence analysis. This strategy was used to generate the transfer plasmids pAIM7, -16, -22, -23, and -24 (Table 1). Note that the CELO nucleotide sequence numbering is derived from reference 6 and GenBank accession no. U46933.

TABLE 1.

Plasmids used in CELO recombinant construction

Plasmid name	CELO left end sequences (nt)	CELO right end sequences (nt)	Comments
pWü-H35	1–1601	42845–43804	Left and right terminal CELO HindIII fragments cloned with flanking SpeI linkers in pBR327
pB5.5	1–5503		Left terminal CELO HpaI fragment in pBluescript
pB13.3		30502–43804	Right terminal CELO HpaI fragment in pBluescript
pWüHpa	1–5503	30502–43804	Left and right terminal CELO HpaI fragments, with flanking SpeI linkers in pBR327
pAIM3	1–5503	30502–30639 40064–43804	pWüHpa derivative (removal of Asp718 fragment)
pAIM7 (luciferase)	1–1064 4335–5503	30502–30639 40064–43804	pWüHpa derivative transfer vector for pAIM11
pAIM16 (luciferase)	1–2980 4335–5503	30502–30639 40064–43804	pWüHpa derivative transfer vector for pAIM21
pAIM22 (luciferase)	1–937 2301–5503	30502–30639 40064–43804	pWüHpa derivative transfer vector for pAIM25
pAIM23 (luciferase)	1–1064 2681–5503	30502–30639 40064–43804	pWüHpa derivative transfer vector for pAIM26
pAIM24 (luciferase)	1–1686 2901–5503	30502–30639 40064–43804	pWüHpa derivative transfer vector for pAIM27
pAIM43 (luciferase)	1–5503	30502–30639 40064–43804	pWüHpa derivative transfer vector for pAIM45
pAIM44 (luciferase)	1–5503	30502–30639 40064–43804	pWüHpa derivative transfer vector for pAIM46
pAIM52 (EGFP)			Transfer vector for homologous recombination with pAIM46

Generation of recombinant CELO genomes.

The plasmids pAIM7, -16, -22, -23, and -24 were linearized by double digestion, using Asp718 and HpaI, and recombined with purified CELO DNA by homologous recombination in E. coli BJ5183 (5, 13) to generate the CELO genome plasmids pAIM11, -21, -25, -26, and -27 (Fig. 1C and Tables 1 and 2). Restriction enzyme digestions were performed to identify the correct recombinants. Furthermore, all constructs were sequenced across the deleted regions to verify the constructs (Table 1).

TABLE 2.

Plasmids containing CELO variants

CELO genome construct	CELO sequences deleted (nt)	Replication in LMH cells
CELO left end
pAIM11	1065–4334 (3,270 bp), AatII + NcoI	Defective, cannot be complemented
pAIM25	938–2300 (1,363 bp), Eco47-3	Defective, can be complemented
pAIM26	1065–2680 (1,612 bp), AatII + SphI	Defective, can be complemented
pAIM27	1687–2900 (1,214 bp), PmaCI	Defective, can be complemented
pAIM21	2981–4334 (1,358 bp), StyI	Defective, cannot be complemented
CELO right end
pAIM45	33358–43684 (10,327 bp), EcoRV	Defective, cannot be complemented
pAIM46 pAIM53	41731–43684 (1,954 bp), EcoRV	Replication competent
pAIM69	41523–43684 (2,161 bp), PvuII-EcoRV	Replication competent
pAIM70	40065–43684 (3,619 bp), Asp718-EcoRV	Replication competent

Modifications in the right end of the CELO genome.

Using methods similar to those described above, plasmids containing both the left and right HpaI fragments of CELO were generated and manipulated to insert the luciferase cassette and to remove an EcoRV fragment from either nt 33358 to 43684 (pAIM43) or nt 41731 to 43684 (pAIM44). These plasmids were linearized at the unique HpaI site and recombined in BJ5183 cells with wild-type CELO DNA to generate either pAIM45 or pAIM46.

Evaluation of the recombinant CELO genomes on LMH cells and preparation of viral stocks.

The recombinant CELO plasmids were digested with SpeI to release the viral genome from the plasmid, extracted with phenol and chloroform, and then purified by gel filtration (Pharmacia Nick Column) equilibrated with TE (10 mM Tris [pH 7.5], 0.1 mM EDTA).

Transfection complexes were prepared using a modification of the PEI technique (1, 3). The DNA was condensed with PEI in two steps as follows: PEI, M_r, 2,000 (2.5 μl of 10 mM PEI in 125 μl of HBS [150 mM NaCl, 20 mM HEPES, pH 7.4]), was added dropwise to 3 μg of DNA diluted in 125 μl of HBS. The sample was incubated at room temperature for 20 min. Subsequently, PEI, M_r, 25,000 (3.5 μl of 10 mM PEI in 125 μl of HBS) was added dropwise to the sample, and the complex was incubated at room temperature for an additional 20 min.

Leghorn male hepatoma (LMH) cells (28) were seeded the day before transfection in 24-well plates at 7 × 10⁴ cells per well (24-well dish). For transfection, the cell culture medium was replaced by 400 μl of Dulbecco modified Eagle medium (DMEM) supplemented with 10 μg of polymyxin B (no serum)/ml. The transfection complex (90 μl per well) was added to the cells for 4 h at 37°C, after which the medium was replaced with fresh, serum-containing medium. Transfection efficiency was monitored by measuring luciferase activity in cell lysates at 24 h posttransfection.

To test for amplification of virus, cleared lysates from transfected or transduced cells were prepared as follows. Cells plus supernatant were harvested and collected by centrifugation, and the cell pellets were resuspended in 2 ml of processed supernatant. The material was frozen and thawed three times and sonicated in a bath sonicator to release viral particles, the cell debris was removed by centrifugation, and the cleared lysate was used for further amplification on fresh cultures of LMH cells. CELO purification by CsCl gradient was performed as previously described (9). Virus was quantified based on protein content, with a conversion factor of 1 mg protein/ml equal to 3.4 × 10¹² virus particles/ml (34).

Construction of EGFP expressing CELO AIM53.

The luciferase cDNA in pAIM46 was replaced by a enhanced green fluorescent protein (EGFP) cDNA to generate pAIM53. The replacement was obtained by homologous recombination in E. coli between pAIM46, linearized at the unique PacI site in the luciferase cDNA, and pAIM52, a transfer plasmid carrying an EGFP cDNA under the control of the same CMV promoter, and the β-globin intron and polyadenylation signal used in the luciferase cassette of pAIM46, thus providing homologies for recombination.

Generation of recombinant type 5 adenoviruses. (i) AdLuc.

The luciferase cassette was cloned via the flanking BamHI sites into pDE1sp1B (2), to produce pDE1sp1BLuc, with the luciferase cDNA in the same orientation as E1 transcription. Recombinant virus was generated by using recombination after cotransfecting pDE1sp1BLuc with pJM17 (2) into 293 cells (19). At 10 days posttransfection, cell lysates were prepared and used to infect fresh 293 monolayers and virus was amplified from a single plaque. The virus stock used here was prepared from material that was subsequently passed through two additional rounds of plaque purification, amplified, purified by banding in CsCl, and quantified by protein content (1 mg/ml protein = 3.4 × 10¹² virus particles/ml) (34).

(ii) AdGFP.

A fragment containing the CMV promoter, EGFP coding region, and simian virus 40 poly(A) sequences was excised from pEGFP-C1 (Clontech), using AseI/MluI. Overhanging ends were filled in by Klenow and cloned into the EcoRV site of pDE1sp1B (2), with the EGFP cassette in the same orientation as E1 transcription. Recombinant virus was generated as described above by using recombination with pJM17 in 293 cells.

Analysis of heat stability of viruses.

CELO AIM46 and AdLuc were diluted to 4 × 10⁹ particles/100 μl concentration in HBS (final glycerol concentration was 2.4% [vol/vol]) and exposed for 30 min to temperatures ranging from 48 to 68°C. Subsequently, aliquots of the virus were tested for the ability to transduce luciferase activity into either A549 or CEF38 cells.

Immunofluorescence.

LMH cells were plated on gelatin-coated glass slides (Labtek; Nunc) at 10⁵ cells/chamber and infected the next day with CELO virus ranging from 10 to 1,000 viral particles/cell in DMEM medium containing 2% fetal calf serum (FCS). At the indicated times after the infection, cells were fixed in cold methanol-acetone (1:1) at room temperature and CELO proteins were visualized by immunofluorescence as follows. Nonspecific binding sites were blocked by using phosphate-buffered saline (PBS) + 1% bovine serum albumin (BSA) at room temperature for 1 h. Polyclonal anti-CELO antibody was diluted 1:1,000 in PBS + 1% BSA and incubated for 1 h. After three 5-min washes in PBS at room temperature, a goat-anti-rabbit (Boehringer-Mannheim) detection antibody coupled to fluorescein isothiocyanate (1:400 dilution) was added in PBS + 1% BSA. The slides were again washed, 4′,6-diamidino-2-phenylindole (DAPI) was included in the last wash for visualization of the nuclei, and the slides were mounted in Mowiol for examination by fluorescence microscopy.

Generation of anti-CELO virion polyclonal serum.

Rabbits were injected with 100 μg of CsCl-purified, heat-inactivated (70°C for 60 min) CELO virions in complete Freund’s adjuvant, they were boosted at 2, 4, and 5 weeks with 100 μg of CELO in incomplete Freund’s adjuvant, and serum was collected subsequently. Western analysis demonstrated that the pooled sera used here reacted specifically with all major CELO capsid proteins but not with lysates of noninfected avian cells.

Additional reagents.

Wild-type CELO (FAV-1 and Phelps) virus was originally obtained from G. Monreal (Free University of Berlin) and purified from infected chicken embryos as previously described (9).

The LMH cell line (28) was obtained from Ulla Protzer (ZMBH, Heidelberg, Germany), the A549 cell line was from the ATCC, the chicken fibroblast CEF38 cell line was obtained from Martin Zenke (MDC, Berlin, Germany), and healthy human dermal fibroblasts were obtained from Clonetics and were used between passages 5 and 15; all four cell types were cultured in DMEM–10% FCS. The 293 cell line (19) was from the ATCC and was cultured in MEMalpha with 10% newborn calf serum.

RESULTS

Construction of a plasmid copy of the CELO genome.

Initially, the terminal HindIII fragments of CELO were purified from CELO virus DNA, treated with base to remove the terminal peptides, linkers encoding SpeI restriction sites were added, and the two terminal fragments were cloned in the correct orientation into a low-copy-number plasmid (Fig. 1A and Materials and Methods). This plasmid, encoding the two ends of the virus (pWü-H35), was linearized with the unique HindIII site and recombined with CELO genomic DNA to generate a full-length CELO genome flanked by SpeI sites on a bacterial plasmid (Fig. 1B). Several independent clones of the viral genome were obtained, and the correct structure was verified by restriction analysis and by the production of virus upon transfection (results not shown).

Analysis of unique sequences required for virus replication.

We developed a screening method for determining the requirement of CELO sequences for virus replication. Deletions were first introduced into bacterial plasmid copies of the viral genome by using homologous recombination in bacteria. In all cases, the deleted viral sequences were replaced with a luciferase cassette to allow monitoring of both the initial transfection efficiency into cells that support wild-type virus replication and the replication and transduction potential of the mutant virus in subsequent passages. As will be shown below, the CELO genome allows the insertion of at least 2 kb of sequence beyond the wild-type genome size; thus, the concern that introducing the luciferase cassette itself might impair replication was not realized. The mutant viral genomes were excised from the plasmid and transfected into LMH cells either alone, to determine if the deletion removed essential DNA sequences, or with a plasmid bearing the region of the CELO genome that spans the deletion, to determine if complementation of the deletion could occur. Five days after transfection, the cells were lysed, a portion of the lysate was assayed for luciferase activity to monitor transfection efficiency, and a second portion was used to infect a fresh monolayer of LMH cells. After another 5-day period, the cells were monitored for cytopathic effect and lysates were prepared and assayed for luciferase and a portion was again used to infect fresh LMH cells.

Analysis of CELO left end.

Using the strategy described above, the unique left-end CELO sequences were analyzed for replication function. The map of the left-end open reading frames (ORFs) of 99 amino acids and larger is shown in the upper portion of Fig. 2. An ORF encoding a functional dUTPase is found at position 784 (54). An ORF beginning at position 1991 encodes a protein with significant homology to the parvovirus rep gene. An additional five ORFs are also indicated. Mutant genomes were constructed that removed first the entire region (pAIM11) or deleted single or small groups of ORFs (pAIM21, -25, -26, and -27). The genome plasmids were introduced into LMH cells by transfection. pAIM11, which has a deletion removing the entire region, was positive for luciferase in the first lysate but unable to transfer luciferase gene expression in subsequent passaging attempts, either in the absence or presence of a complementing left-end fragment (Fig. 2). A more discrete mutant (pAIM21) disrupts only three of the unknown ORFs but leaves intact the dUTPase and the rep-like ORFs. However, similar to pAIM11, the pAIM21 genome was also unable to transfer luciferase gene expression in subsequent passaging attempts either in the absence or presence of a complementing left-end fragment (Fig. 2). Thus, both of the mutations alter sequences that must be present in cis for virus replication. pAIM27 deletes only rep, while pAIM25 and -26 delete the dUTPase, rep, and one unknown ORF. These three genomes all produced luciferase activity in the first lysate. Subsequent passage of the material revealed that CELO AIM25, -26, and -27 were not capable of replicating in the absence of complementation. However, unlike pAIM11 and -21, passageable luciferase activity was observed if the initial transfection contained the complementing left-end plasmid (Fig. 2). These three complemented viruses (CELO AIM25+, -26+, and -27+) were amplified for an additional six passages in LMH cells with, surprisingly, only modest declines in their ability to transduce luciferase activity (results not shown). PCR and Southern analysis revealed a substantial contribution of apparently wild-type CELO in the passage three material, demonstrating that recombination had occurred that reintroduced the sequences deleted in the original mutants. Thus, an apparently wild-type CELO was produced which provided complementation functions for the luciferase-bearing mutant CELO.

FIG. 2.

Analysis of CELO left-end mutations. (Top) ORFs of greater than 99 amino acids in the left ∼5,000 nt of the CELO genome are indicated in either black (rightward transcription) or grey (leftward transcription). The ORFs coding for a dUTPase and a protein with parvovirus rep homologies are indicated. (Bottom) Analysis of replication. The nucleotide numbers of deletions introduced into the CELO genome are listed. The modified CELO genomes were linearized with SpeI to release the genome from the bacterial plasmid and transfected into LMH cells either alone or in the presence of a plasmid (pB5.5) bearing wild-type CELO sequences from nt 1 to 5501 (plus left end). At 5 days posttransfection, cells were harvested and lysed by freezing-thawing and sonication and the lysates were applied to a fresh LMH culture. This amplification was repeated twice, and equal aliquots from the third passage of virus were tested for their ability to transduce luciferase activity in LMH cells. The averages of three transductions with the standard deviations are indicated.

In conclusion, of the left-end sequences, a region was identified (between nt 2981 and 4334) that is essential in cis for virus replication. This requirement could be due to the presence of cis-acting signals (e.g., polyadenylation sites) in the deleted region or it could be due to inappropriate or insufficient protein expression from ORFs carried on the plasmid rather than within the virus genome. A second region was identified (between nt 940 and 2900) which is essential for virus replication but could be supplied in trans. Formally, it is possible that a series of recombination events generated a viral genome that contained both the originally deleted sequence and the luciferase cassette; however, the simplest explanation of this pattern is that a simple recombination occurred between pB5.5 and the mutant CELO genome to generate a wild-type CELO genome, which in subsequent passages provided complementation activity for a small number of the mutant (luciferase positive) viruses in the mixture. Some of these viral genomes contain net insertions of sequence over the wild-type size, with the largest containing a 1,612-bp deletion combined with a 3.3-kb luciferase cassette insertion. Thus CELO virus, which in the wild-type form is already 8 kb larger than Ad5, can package, at least, an additional 1,700 bp of sequence.

A portion of the CELO right end is dispensable.

A similar mutational strategy was used for an analysis of the right-end sequences of CELO. The genome plasmid pAIM45 contains a large deletion from 33358 to 43684, deleting 10 ORFs of 99 amino acids or larger, including the previously characterized GAM1 gene (Fig. 3) (7). The plasmids pAIM46, pAIM69, and pAIM70 contain more discrete deletions and disrupts either two ORFs (pAIM46 and -69) or three ORFs (pAIM70; Fig. 3). All of these mutant genomes included a luciferase cassette in place of the deleted sequences.

FIG. 3.

Analysis of CELO right-end mutations. Analysis was performed as outlined in the legend to Fig. 2 except that the right-end complementing plasmid pB13.3 was used in place of the pB5.5 (plus right end).

The genomes were transfected into LMH cells either alone or with a plasmid bearing the wild-type CELO right-end sequences (pB13.3). Luciferase activity was determined for lysates of the transfected cells demonstrating successful transfection (results not shown). Subsequent passage of the material on fresh LMH cells revealed that pAIM45, with the extensive right-end deletion, was not capable of generating infectious CELO particles, either in the absence or in the presence of the intact right-end sequences (Fig. 3). Not surprisingly, this extensive deletion removed sequences that were essential and most likely some of these are required in cis, as evidenced by the absence of complementation by the wild-type right-end sequences. In contrast, the pAIM46, -69, and -70 genomes were found to generate infectious and passageable virus in both the presence or the absence of the complementing genome fragment (Fig. 3). The disrupted ORFs in pAIM46, -69, and -70 are thus dispensable for cell culture growth of CELO as well as for growth in chicken embryos (see below).

To verify the structure of pAIM46 and of the genome carried by CELO AIM46, a PCR analysis was performed to demonstrate that the deletion/insertion constructed in the plasmid was maintained in the genome of the amplified CELO AIM46 virus. As shown in Fig. 4, both the plasmid pAIM46 and DNA isolated from the CELO virus AIM46 produced the expected PCR products. Primers that span the deletion/insertion site generate the predicted PCR product of 3,422 bp with pAIM46 target DNA (Fig. 4, lane 4) and with DNA derived from two CELO AIM46 preparations (Fig. 4, lanes 5 and 6), while PCR with wild-type CELO virus DNA produces the predicted DNA molecule of 2,039 bp (Fig. 4, lane 3). Furthermore, primers that recognize the luciferase insert produce the predicted 958-bp product from DNA derived from pAIM46 or from two isolates of CELO AIM46 (Fig. 4, lanes 9 to 11) but not from DNA derived from wild-type CELO (Fig. 4, lane 8).

FIG. 4.

PCR analysis of wild-type CELO versus CELO AIM46. Lanes: M, marker DNA, (EcoRI/HindIII-cut lambda DNA); 2 to 6, primers OAIM24+OAIM26 were used with irrelevant target DNA (lane 2), wild-type CELO DNA (lane 3), plasmid pAIM46 DNA (lane 4), DNA isolated from CELO AIM46 (lanes 5 and 6); 7 to 11, primers OAIM24 and OAIM26 were used with irrelevant DNA (lane 7), wild-type CELO DNA (lane 8), plasmid pAIM46 DNA (lane 9), DNA isolated from CELO AIM46 DNA (lanes 10 and 11). DNA sizes (in base pairs) are indicated for some of the marker molecules (left side) and for the expected PCR products (right side). The primers used for PCR are the following: OAIM24 (CCGAGAATCCACCAATCGTA) is a sense oligonucleotide hybridizing in the CELO virus right end (nt 41699). OAIM25 (CAGCGTGTCGCTATACGCAA) is an antisense oligonucleotide hybridizing in the CELO virus right end (nt 43752). OAIM26: (GCGATGACGAAATTCTTAGC) is a sense oligonucleotide hybridizing in the luciferase expression cassette. PCR with OAIM24 and OAIM25 should give a 2,053-bp product with a wild-type CELO template and a 3,422-bp product with the AIM46 template. PCR with OAIM24 and OAIM26 should give a 958-bp product with an AIM46 template and no product, with the wild-type CELO template.

Immunofluorescence analysis of CELO AIM46 versus wild-type CELO replication.

Luciferase data showed that CELO AIM46 can replicate in LMH cells in the absence of complementation. To analyze more directly the replication of CELO AIM46 in comparison to wild-type CELO, the two virus types were used to infect LMH cells and the progression of virus infection was monitored by immunofluorescence microscopy using a polyclonal antiserum directed against CELO capsid proteins (Fig. 5). For both wild-type CELO and CELO AIM46, replication is first detectable at 10 h postinfection and the signal increases over the next 30 h until cytopathic effect results in detachment of cells and a decline in the fluorescence signal. Thus, in a cell culture infection, CELO AIM46 appears to replicate with kinetics that are similar to those of wild-type CELO.

FIG. 5.

Immunofluorescence, monitoring the replication of wild-type CELO versus CELO AIM46. LMH cultures were infected at 500 particles per cell with either CELO AIM46 or wild-type CELO. Cell samples were fixed at the indicated times postinfection, and production of CELO virion proteins was monitored by immunofluorescence, using an antiserum against CELO virion.

Growth of CELO AIM46 in chicken embryos.

In the initial stages of this work, LMH cells were used for cell culture propagation of CELO AIM46. Because the nature of the transformation event that established this cell line is not clear, it remains possible that the LMH cells provide some helper functions for CELO AIM46 that wild-type chicken embryonic cells might lack. We were also interested to determine if CELO AIM46 was capable of growing in chicken embryos for practical considerations: the low cost and ease of handling of embryos would facilitate production of these viruses. Equal quantities of either wild-type CELO or CELO AIM46 were injected into the allantoic cavities of 9-day-old chicken embryos. After incubation at 37°C for 3 days, the allantoic fluid was harvested and virus was purified by banding in CsCl density gradients. Yields of purified wild-type CELO ranged from 0.149 to 0.9 mg per egg (average, 0.427 mg/egg), while CELO AIM46 yields were from 0.119 to 0.828 mg per egg (average, 0.301 mg/egg; Table 3). The modifications introduced in CELO AIM46 appear to affect the growth of AIM46 in chicken embryos to only a modest extent.

TABLE 3.

Yield of CELO virus from eggs^a

Virus type	Prepara-tion no.	Yield of puri-fied virus (mg)b	No. of eggs	Virus/egg (mg)	Avg yield/ egg (mg)
Wild-type CELO	1	1.80	2	0.90	0.427
Wild-type CELO	2	0.496	2	0.248
Wild-type CELO	3	0.345	2	0.173
Wild-type CELO	4	1.33	2	0.665
Wild-type CELO	5	5.96	40	0.149
CELO AIM46	1	1.66	2	0.828	0.301
CELO AIM46	2	0.133	1	0.133
CELO AIM46	3	0.247	1	0.247
CELO AIM46	4	0.618	2	0.309
CELO AIM46	5	0.340	2	0.170
CELO AIM46	6	0.237	2	0.119

^aWild-type CELO or CELO AIM46 (8 × 10⁸ particles in 100 μl of HBS) was injected into the allantoic cavities of 9-day-old chicken embryos. After 3 days of incubation at 37°C, allantoic fluid was harvested and virus was purified by banding in CsCl density gradients as previously described (9). 

^bThe virus yield is expressed as purified virus protein. Protein was measured by Bradford assay using BSA as a standard. 

Physical stability of CELO AIM46.

A distinctive feature of the CELO virion is physical stability, most readily measured by resistance of the virion to elevated temperatures. While mastadenoviruses such as Ad5 are inactivated by exposure to temperatures of 48°C and higher (4, 8, 16), CELO was originally reported to be stable at 56°C (57) and subsequent isolates of the virus have been reported with stability at higher temperatures as well as to other harsh treatments (reviewed in reference 39). The molecular nature of CELO virus stability has not been determined. A major component of Ad5 capsid stability, pIX, has not been identified in CELO virus. Perhaps hexon or other capsid components have altered sequences which allow more stable protein-protein interactions. It is likely that this stability is important in the wild for CELO virus survival in the harsh avian environment. In any case, it was of interest to determine if the CELO recombinant vector retains the stability of the wild-type CELO virion.

A recombinant Ad5 bearing a luciferase expression unit (AdLuc) and CELO AIM45 were exposed to heat titration (30-min exposure to defined temperatures from 42 to 68°C). Subsequently, each sample was tested for its ability to transfer luciferase activity to either human A549 or avian CEF cells (Fig. 6). As previously demonstrated for Ad5, the virus capsid, and thus the transduction ability of the virus, is sensitive to heat (4, 8, 16). Ad5 transduction of human cells declines by a factor of more than 100 when exposed to 48°C for 30 min and is inactivated at 52°C and higher temperatures (Fig. 6). In strong contrast, CELO AIM46 transduction ability is not affected by heating at 56°C and the virus only begins to lose activity when exposed to 60°C for 30 min (Fig. 6). We found that transduction with wild-type CELO displays similar heat stability (results not shown), indicating that the alterations introduced in CELO AIM46 do not significantly alter the virion’s stability.

FIG. 6.

Heat stability of AIM46 versus AdLuc. Aliquots of CELO AIM46 or AdLuc (4 × 10⁹ virus particles in 100 μl of HBS–2.4% glycerol) were exposed to the indicated temperatures for 30 min. The treated virus was then used to transduce either human A549 cells or chicken CEF cells (10³ virus particles/cell) in triplicate. Luciferase activity was measured 24 h later. The values are the averages of three transductions with the standard deviations indicated.

CELO can transduce a variety of cell types.

In considering future applications, it is of interest to determine the types of cells that can be transduced by a CELO-based vector. We tested a panel of commonly used mammalian and chicken cell types for their transducibility by CELO AIM46. For comparison, we used the Ad5 derivative Adluc carrying the same luciferase expression cassette. The results for four of these cell types are presented in Fig. 7. Cells of avian origin (e.g., the chicken fibroblast line CEF38) were transduced with nearly 100-fold greater efficiency with the CELO vector than with the human AdLuc (Fig. 7). Note that CEF38 cells do not support virus replication, so the difference between the Ad5 vector and the CELO vector cannot be ascribed to virus replication and must be due to primary transduction or gene expression effects. In the human cell types tested, CELO virus acted comparably to the Ad5 vector. The human cell types include the hepatoma line HepG2, the lung epithelial carcinoma line A549, and primary human dermal fibroblasts (Fig. 7). Similar results were obtained with the human carcinoma line HeLa, the murine myoblast line C2C12, and the canine epithelial line MDCK (results not shown).

FIG. 7.

Tropism of CELO virus versus Ad5. The indicated cell types were exposed to aliquots of AdLuc or CELO AIM46 of from 1000, 300, 100 or 30 particles per cell (see Materials and Methods for the protocol for a 24-well plate). At 24 h postinfection, luciferase activity was determined. The values are the averages of three transductions with the standard deviations indicated.

In conclusion, we find that CELO AIM46 is capable of transducing avian cells approximately 100-fold more efficiently than a human Ad5 vector. Surprisingly, CELO AIM46 also transduces mammalian cell types with efficiencies comparable to those of an Ad5-based vector.

GFP expression from adenovirus and CELO vectors.

GFP has emerged as a useful marker for gene transfer studies. Accordingly, we prepared a CELO vector (CELO AIM53) expressing EGFP (Clontech) in the CELO AIM46 background. The activity of this vector was compared to that of an Ad5 vector bearing the same CMV/EGFP/β-globin expression unit. We find that both vectors function to transfer a GFP gene in human A549 cells (Fig. 8). Although immunofluorescence with GFP is not quantitative in this format, it appears that, similar to the luciferase recombinants, there are not large differences in transduction capacity between the CELO and the Ad5 EGFP viruses when transducing human A549 cells.

FIG. 8.

Transduction of EGFP using recombinant Ad5 or CELO vectors. The EGFP-expressing adenovirus AdGFP and CELO AIM53 were used to infect human A549 cells over a range of virus/cell ratios (10 to 1,000 particles per cell). At 24 h postinfection, cells were fixed and GFP expression was monitored by fluorescence microscopy.

DISCUSSION

We have identified CELO genome regions that are essential for virus replication. Most importantly, we have identified a region in the left end of the virus genome that can be deleted and supplied in trans. This region is thus an appropriate candidate for establishing a complementing cell line. We have also identified a region in the right end of the genome that can be deleted with no detectable effects on virus replication in either cell culture or in embryos. We have shown that an expression cassette for foreign genes can be inserted in this region to generate a gene delivery vector. We have demonstrated that CELO vectors can package an additional 1.7 kb of DNA sequence over the wild-type genome size, which is already 8 kb larger than Ad5 vectors. Replication-competent CELO vectors bearing either a luciferase expression unit or an EGFP expression unit were developed. These vectors were monitored for their ability to transduce a variety of avian and mammalian cell types. As expected, the CELO vectors work much better than an Ad5 vector in avian cells. However, in all mammalian cell types tested, the CELO vectors functioned with transduction efficiencies comparable to those of Ad5 vectors, suggesting that CELO vectors based on this genome will be useful for gene transfer applications in mammalian systems. Furthermore, these vectors have obvious vaccine applications in avian species for which viral replication can promote immune responses. The ability to propagate the virus in inexpensive chicken embryos will certainly facilitate production of large quantities of the vector for any of these applications.

A variety of nonhuman adenovirus vectors have been developed in recent times, including a vector based on the bovine adenovirus type 3 (40), on the ovine adenovirus OAV287 (29, 53), and on the canine adenovirus type 2 (30). There are several justifications for pursuing these alternate viral subtypes. For vaccine applications in nonhuman hosts, these viruses, if properly modified, may provoke more effective immune responses than a human adenovirus-based vector. Furthermore, a more robust immune response might be expected from a replication-competent virus; thus, a vector is most useful in a host in which replication is partially or fully permissive. This is not the case with human adenovirus-based vectors in nearly all nonhuman hosts. In this regard, the CELO vector described here, AIM46, is ideally suited for avian vaccine applications.

An additional argument for pursuing a nonhuman adenovirus comes from human gene therapy applications. Preexisting immune responses to human adenovirus can impair the initial transduction by human adenovirus-based vectors or might exacerbate the cellular immune response to transduced cells. A patient may have no immune experience with an adenovirus from a distant species (although two of seven patients had neutralizing antibodies to the canine adenovirus vector [30]), and initial transductions will not be compromised by the host response to viral antigens. Except for certain agricultural workers, a previous immune exposure to CELO antigens would not be expected in most of the human population. CELO vectors might therefore have an advantage over vectors based on more common human adenovirus serotypes.

An additional conceptual advantage of CELO-based vectors is that CELO, like the bovine, ovine, and canine adenoviruses, is naturally replication defective in human cells. Thus, rampant replication of these vectors is not going to occur in human patients even in the presence of a wild-type human adenovirus infection. A more likely problem however, may be caused by expression of viral proteins in human cells. The expression of viral genes in human cells has only been carefully examined in OAV287 vectors (29), and a similar analysis will have to be performed with CELO vectors.

Our future efforts will focus on several aspects of CELO. The transforming genes of this virus are not yet defined. Using functional assays, we have identified GAM-1, an E1B 19K homologue (7), and a protein that interacts with Rb and stimulates the E2F pathway (33a). CELO functions that disrupt p53 signalling are expected to exist and these are being sought. This information is of biological interest and, similar to the human adenovirus examples, the application of CELO vectors will be dependent on the clear identification and removal of all transforming genes. The ability to generate a replication-defective CELO vector will be facilitated by the deletion analysis performed here. Efforts to construct cell lines expressing CELO complementing functions are under way.