INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Members of the family Adenoviridae are nonenveloped viruses with double-stranded linear DNA genomes that are 25 to 45 kb in length (13). Human adenoviruses (type 2 and 5) have been for some time developed as gene delivery vectors for vaccination or gene therapy (2, 9, 17, 24, 35, 36) and more recently, adenoviruses from various species (bovine, ovine, porcine, canine, and avian) have been studied for similar applications (11, 16, 23, 25, 28, 29, 30, 32). Among avian adenoviruses, which include at least 10 serotypes (3, 4, 8, 12, 27, 32), chicken embryo lethal orphan (CELO), which represents serotype 1, is the best characterized (6, 7, 19, 20, 22, 23, 26). The CELO virus genome is 43,804 bp long and has been completely sequenced (7), and its transcriptional organization has been established (26). The central region of the viral genome is strongly homologous with other adenoviruses: the lower strand encodes replication functions (DNA polymerase, DNA-binding protein, pTP), and the upper strand, which is transcribed under the control of a single major late promoter, encodes capsid structural proteins. On either side of this central part there are two regions encoding at least 22 unassigned open reading frames (ORFs) that have no sequence homology with the E1, E3, and E4 regions of mammalian adenoviruses (mastadenoviruses). Only 2 of these 22 genes have been studied: ORF8 encodes GAM-1 protein, which was identified as a functional homolog to human adenovirus E1B 19K protein (6), and ORF22 encodes a protein that interacts with the retinoblastoma protein, which is similar to human adenovirus E1A protein, and cooperates with GAM-1 to activate the E2F pathway (20).

In order to use CELO virus as a viral vector for gene transfer in general and in particular as a vaccine vector for aviculture, we had to define the functions of the above ORFs. It was necessary to establish which ORF is essential for viral replication and has to be conserved to obtain a replication-competent vector and which ORF could be deleted or inactivated for insertion of foreign DNA into the viral genome. In the first stage, we inactivated each ORF by introducing termination codons in the first third of the coding sequence, in order to determine the effect of the mutation on the viral cycle for each mutant. From these results, together with data from our laboratory on transcriptional organization of the viral genome (26) and also the deletion experiments published by Michou et al. (23), we selected possible sites for deletions and insertions into the viral genome.

There are three commonly used methods for construction of recombinant adenoviruses, all of which are based on homologous recombination. The first method uses homologous recombination between two overlapping viral DNA fragments in eucaryotic cells: this process is arduous and gives a poor yield of recombinant viral DNA. The second method, described by Graham et al. (10), uses homologous recombination between two plasmids in eucaryotic cells, one containing the entire viral genome with enough exogenous DNA to prevent it from packaging and the other carrying the desired modification (insertion, deletion, etc.) flanked by sequences homologous to the region of the viral genome that has to be modified. The third method, described by Chartier et al. (5), is based on the same principle but uses homologous recombination in Escherichia coli. The last two methods give better results, but the amplification of plasmids carrying the whole viral genome often gives poor yields of recombinant viral DNA.

Our method is based on the use of cosmids (31), which allows us to obtain recombinant viral DNA very easily and efficiently by taking advantage of the similar size of bacteriophage  and CELO virus DNA and of the ability of phage  capsid to package DNA molecules of 40 to 50 kb, provided that they are flanked by two COS sites. In this system, the recombinant CELO genome is constructed by in vitro ligation followed by phage amplification, which allows recovery of only ligation products that reconstitute a complete viral genome. Each step of the construction can be precisely controlled, and high yields of recombinant viral DNA are produced. This method was first applied to the mutagenesis of the ORFs with unknown functions during the CELO viral cycle. We produced 16 independent mutant viruses that all replicated in cell culture. We then constructed recombinant vectors carrying various deletions in the CELO genome and insertions of foreign genes. We easily obtained positive results in vitro, first with vectors carrying the enhanced green fluorescent protein (eGFP) reporter gene and second with vectors carrying genes of interest for applications in aviculture. Evaluation of these vectors in vivo in chickens is now in progress.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Cloning the CELO genome into a cosmid vector. The complete CELO genome was first cloned into plasmid pPolyII (2,057 bp, derived from pBR322) (Fig. 1). The CELO terminal sequences were cloned between the PacI and AscI sites of this plasmid, and then a plasmid containing the full-length CELO genome (pPolyII/CELO, 45,800 bp) was generated by homologous recombination between the cloned terminal sequences and the purified CELO virus DNA in E. coli strain BJ5183. The cos/CELO cosmid was then obtained by cloning the complete CELO genome into the 5,944-bp cos1 cosmid vector (gift of A. Epstein, Centre National de la Recherche Scientifique, Lyon-Villeurbanne, France) derived from the SuperCos vector (Stratagene). The 43,804-bp CELO genome was excised from pPolyII/CELO by PacI and AscI digestion, purified from a low-melting-point agarose gel, and cloned between the PacI and AscI sites of cos1 by in vitro ligation. The ligation product was then packaged in vitro using bacteriophage  extracts and amplified as described by the supplier (Expand cloning kit, Boehringer Mannheim) by infecting E. coli strain DH5 cells cultured in Luria-Bertani medium containing 10 mM MgSO₄ and 0.2% maltose to induce pilus formation. The cos/CELO cosmid DNA (49,644 bp) was then purified by DNA minipreparations (High Pure Plasmid preparation kit, Boehringer Mannheim) from ampicillin-resistant colonies, and positive clones were selected by restriction endonuclease analysis.

View larger version (14K): [in this window] [in a new window]   FIG. 1.   Cloning the CELO genome into a cosmid vector. (a) The complete CELO genome was first cloned into plasmid pPolyII by homologous recombination in E. coli to generate a viral genome with single restriction sites (i.e., PacI and AscI sites) at both ends. (b) The cloned CELO genome was excised from the plasmid and ligated with the cosI cosmid vector, which was linearized by PacI and AscI digestion. The ligation reaction generated concatenated DNA molecules, with COS sites separated by 50 kb from each other, which allowed for packaging into  bacteriophage heads. (c) The ligation product was then packaged and amplified by infecting E. coli cells, resulting in the cos/CELO cosmid that contained the full-length CELO genome.

Generation of modified CELO genomes by the cosmid method. The cloned CELO genome can be divided into four single restriction fragments: A (PacI-PmeI 7,430-bp fragment), B (PmeI-NotI 9,956-bp fragment), C (NotI-FseI 18,298-bp fragment), and D (FseI-AscI 8,116-bp fragment) (Fig. 2). To modify, for example, the left end of the CELO genome (the A fragment), cos/CELO was digested by PacI and PmeI and subjected to electrophoresis in a 0.7% low-melting-point agarose gel (Seakem) in 1× Tris-acetate-EDTA buffer. The larger fragment (cos/CELO without the A fragment, 42,214 bp) was isolated from agarose by gelase treatment (Epicentre) and stored at 4°C until use. The A fragment was purified by using the QiaexII gel extraction kit (Qiagen) and subcloned into the 2,860-bp plasmid pMECA (provided by J. M. Thomson and W. A. Parrot, University of Georgia, Athens) (33). After modification, the A fragment was excised from plasmid pMECA by PacI and PmeI digestion and reintroduced into cos/CELO by in vitro ligation. The ligation product was then packaged, amplified, and purified as described above, and positive clones were selected by restriction endonuclease analysis. The same principle was applied for modifying the other regions of the CELO genome, except that other restriction endonucleases were used for cos/CELO digestion.

View larger version (17K): [in this window] [in a new window]   FIG. 2.   Construction of recombinant CELO vectors using the cosmid-based method. Modification of the right end of the viral genome (D fragment) is shown as an example. (a) The CELO D fragment was released from cosmid cos/CELO by AscI and FseI digestion and subcloned into plasmid pMECA for modification. The deleted cosmid was isolated by agarose gel electrophoresis, purified by gelase treatment, and stored at 4°C until use. (b) Modification was achieved in the D fragment using standard molecular biology methods (directed mutagenesis, deletion and/or insertion of foreign DNA), represented by a star (). The modified D fragment is released from pMECA and ligated back to the deleted cosmid. (c) The resulting recombinant CELO genome, flanked by two COS sites, is packaged into  bacteriophage heads and amplified by infecting E. coli cells. The cosmid DNA containing the modified genome can be used directly for transfection of LMH cells.

Directed mutagenesis of CELO ORFs. Only ORFs located at the ends of the CELO genome were mutated, i.e, the 16 ORFs located on the A fragment (ORFs 1, 2, 3, 4, 12, 13, 14, and 15) and the D fragment (ORFs 7, 8, 9, 10, 11, 16, 17, and 18) (Fig. 3). The last six ORFs (ORFs 5, 6, 19, 20, 21, and 22) are located on the C fragment, which is more complex to manipulate due to its large size (18.3 kb). For each of the 16 mutants, a restriction fragment containing the ORF was subcloned into plasmid pAlter (Promega), and a termination codon (TAA, TAG, or TGA) was introduced into the coding sequence by PCR using the QuickChange site-directed mutagenesis kit (Stratagene). The restriction fragment and PCR primer used for the mutagenesis and the position of the mutation are given in Table 1 for each mutated ORF. The mutated fragment was then reintroduced into the A or D fragment subcloned into pMECA, and the presence of the mutation was verified by sequencing before reintroducing the mutated A or D fragment into the CELO genome.

View larger version (14K): [in this window] [in a new window]   FIG. 3.   Positions of the 22 ORFs and major late promoter (MLP) on CELO genome. ORFs are represented by hatched squares and designated by their number according to the notation of Chiocca et al. (7). Arrows indicate the direction of transcription for each DNA strand. Relative positions of the single restriction sites and restriction fragments used for construction of recombinant viral genomes are indicated.

Construction of packaging-defective CELO genomes. Two viral genomes with deletion of the region carrying the putative packaging signal () were constructed. These genomes could be used as helper viruses for replication-deficient vectors if necessary. The first genome (₁) was constructed by deletion of 462 bp between the EcoRI (nucleotide [nt] 80) and the PstI (nt 542) sites. For this construction, a 378-bp EcoRI-PstI fragment, corresponding to sequences between PstI (nt 542) and PstI (nt 920), was synthesized by PCR using Pfu DNA polymerase (Stratagene) and reintroduced in the right orientation into the A fragment. The second genome (₂) was constructed by deletion of 841 bp between the EcoRI (nt 80) and the PstI (nt 920) sites. Note that ₁ and ₂ were both deleted from the ORF1 promoter and that ₂ was also deleted from the 5' leader sequence and part of the coding sequence of ORF1.

Construction of GFP-expressing CELO vectors. The first recombinant CELO-GFP vector was constructed by simple insertion into the CELO genome (Fig. 4). The eGFP cDNA was put under the control of the human cytomegalovirus (CMV) immediate early promoter (P_CMV), and this 1,411-bp cassette was inserted into the A fragment at the SphI nt (2681) site in the 3' part of CELO ORF2. This probably inactivated the ORF2 gene but allowed the GFP to be expressed by using the ORF2 polyadenylation signal (ORF2_polyA) at nt 2830. From this first construct, a 1,560-bp GFP expression cassette, carrying P_CMV, eGFP cDNA, and ORF2_polyA, was amplified by PCR using Pfu DNA polymerase (Stratagene). Different primers were used to create appropriate restriction sites at both ends of the cassette in order to use it for the construction of other CELO-GFP recombinants (Fig. 5; see Table 3).

View larger version (20K): [in this window] [in a new window]   FIG. 4.   First step for the construction of CELO GFP expression vectors. (a) The CELO A fragment was subcloned into plasmid pMECA. A 1,411-bp cassette, carrying the GFP cDNA under the control of the CMV promoter, was inserted at a SphI site in the 3' part of CELO ORF2, upstream from the ORF2 polyadenylation signal (AATAAA). The resulting modified A fragment carrying a functional GFP expression cassette was reintroduced into the CELO genome to obtain our first GFP-expressing CELO vector. (b) From this construction, a 1,560-bp expression cassette, carrying the CMV promoter, GFP cDNA, and ORF2 polyA, was amplified using Pfu DNA polymerase. This expression cassette was used for the construction of the other GFP-expressing CELO vectors.

View larger version (14K): [in this window] [in a new window]   FIG. 5.   Positions of the deletions and insertions carried out in the CELO genome for the construction of GFP expression vectors. (a) At the left end of the viral genome, a simple insertion of the GFP expression cassette (at the SphI site) and three distinct deletions were performed in the A fragment (PacI-PmeI). (b) At the right end, four distinct deletions were performed in the D fragment (FseI-AscI). All eight resulting recombinant genomes produced infectious viral particles after transfection of LMH cells.

Construction of CELO vectors carrying genes for vaccine application. For this purpose, we chose to use the vector backbone carrying the largest deletion, i.e., the 3,620-bp KpnI-EcoRV deletion in the CELO D fragment. In order to facilitate the insertion of various genes, we constructed two intermediate pMECA/deleted D fragment plasmids (pMECA/D).

View larger version (18K): [in this window] [in a new window]   FIG. 6.   Construction of a CELO vector carrying an expression cassette with a gene of interest. (a) A transfer plasmid was constructed by cloning the CELO D fragment into plasmid pMECA. A 3,620-bp KpnI-EcoRV fragment carrying CELO ORFs 9, 10, and 11 was deleted and replaced by an MCS in which the expression cassette was inserted. (b) The vector genome was obtained after ligation of the deleted D fragment carrying the expression cassette into the cos/CELO cosmid.

Generation and amplification of recombinant CELO vectors on LMH cells. Leghorn male hepatoma (LMH) cells (ATCC CRL-2117) (15) were directly transfected with the recombinant cos/CELO DNA preparations, without prior release of the recombinant CELO genomes by PacI-AscI digestion. Transfections were performed by the calcium phosphate method using the Calphos maximizer kit (Clontech), essentially as described by the supplier. Cosmid DNA (4 µg) was combined with CaCl₂ to a final volume of 100 µl, the DNA-calcium mixture was added dropwise to 100 µl of 2× HEPES-buffered saline during bubbling with a pipette, and the precipitate was allowed to form for 20 min. LMH cells were seeded the day before transfection into 6-well plates at 6 × 10⁵ cells per well in 2 ml of William's E medium containing 100 U of penicillin/ml and 100 mg of streptomycin/ml and supplemented with 6% fetal calf serum (FCS). Medium was changed 1 h before transfection, and the DNA precipitate was added (200 µl per well) and incubated with cells for 6 h. Cells were then washed with phosphate-buffered saline (PBS) and incubated in fresh medium at 37°C. Transfection efficiency was checked by transfecting cells of one well with a GFP expression plasmid (consisting of the eGFP cDNA cloned at the BamHI site of pcDNA3.1/zeo(+) plasmid [Invitrogen]), and GFP expression was verified 12 h posttransfection. After 5 to 15 days, depending on the time of appearance of a cytopathic effect (CPE), cells and supernatant were harvested and lysates were subjected to three cycles of freezing at 80°C and thawing at 37°C to release viral particles from cells. Cell debris was removed by centrifugation and viruses were amplified by infecting fresh cultures of LMH cells with the cleared lysates. Viral stocks were prepared by three successive amplifications on LMH cells and stored at 80°C.

Titration of viral stocks. Viral stocks were titrated by the plaque assay method as follows: LMH cells were seeded 2 days before infection into 6-well plates at 6 × 10⁵ cells per well in 2 ml of William's E medium supplemented with 6% FCS. Serial dilutions of the viral stocks were prepared in William's E medium supplemented with only 2% FCS, and infection was achieved using 1 ml of each dilution per well. Cells were incubated with virus preparation for 2 h at 37°C, washed in PBS, and covered with 3 ml of William's E medium containing 6% FCS and 1% low-melting-point agarose (SeaPlaque agarose; FMC Bioproducts). After 2 days, 2 ml of agarose-containing medium was added to each well. Four days after infection, living cells were stained with a 0.05% neutral red solution (Sigma), and plaques were counted on the next 2 days.

Analysis of viral genomes. For each CELO vector preparation, genome stability was checked by analyzing the viral DNA. LMH cells were seeded the day before infection into 100-mm-diameter culture dishes at 4 × 10⁶ cells per dish. For infection, culture medium was replaced with fresh medium containing 10⁷ infectious particles, and cells were incubated for 3 to 5 days at 37°C. When the maximum CPE was reached, cells and supernatant were harvested and cell debris was removed by 10 min of centrifugation at 300 × g. The supernatant was then subjected to 1 h of centrifugation at 100,000 × g, and the resulting pellet was resuspended in 400 µl of lysis buffer (100 mM Tris-HCl [pH 7.5], 150 mM NaCl, 12.5 mM EDTA, 1% sodium dodecyl sulfate). Cellular RNA was digested by addition of 100 µg of RNase A per ml and incubation for 1 h at 37°C. Proteins were eliminated by addition of 500 µg of proteinase K per ml and 2 h of incubation at 37°C, followed by two successive phenol-chloroform extractions. DNA was then precipitated with 20 µl of 5 M NaCl and 1 ml of ethanol, pelleted by 15 min of centrifugation at 11,000 × g, and resuspended in 100 µl of TE buffer (10 mM Tris-HCl [pH7.5], 0.1 mM EDTA). Viral DNA was then subjected to restriction endonuclease analysis, using 10 µl of the DNA solution for each digestion.

Flow cytometry analysis of GFP expression. LMH cells plated in P6 wells were infected with CELO-GFP vectors or with wild-type CELO virus (used as a control) at a multiplicity of infection (MOI) of 5 or 10 PFU per cell. Cells were harvested at 16 h postinfection by trypsin-EDTA treatment and resuspended in PBS containing 10 µg of propidium iodide (PI) per ml to color apoptosing cells in red (18). Cells were then submitted to flow cytometry analysis of green and red fluorescence (FACSort flow cytometry system; Becton Dickinson).

Radiolabeling and immunoprecipitation of CELO and IBDV antigens. Monolayers of LMH cells plated in P12 wells were infected with recombinant CELO-IBDV vectors at a multiplicity of 4 PFU per cell. At 5 h postinfection, 10 µl of Pro-mix 35S (Amersham) was added to the cell medium (1.5 ml) for an overnight incubation. Next, cells were solubilized in 0.5 ml of immunoprecipitation buffer (50 mM Tris [pH8], 150 mM NaCl, 2% Triton X-100). After clarification by centrifugation at 13,000 × g for 30 min, aliquots of lysates were incubated with 1 ml of either monoclonal antibody against VP2 or VP3 of IBDV (undiluted hybridoma ascites fluid) or polyclonal antibody against CELO virus (undiluted rabbit antiserum) and 35 ml of a 1:1 slurry of protein A-Sepharose 4B (Pharmacia) under gentle agitation for 2 h at room temperature. The beads were then washed four times with 1 ml of immunoprecipitation buffer, treated for 2 min at 100°C in Laemmli's denaturing buffer plus 5% 2-mercaptoethanol, and centrifuged. Resulting supernatants were subjected to electrophoresis on a 10% polyacrylamide gel which was then processed for autoradiography.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Construction of recombinant CELO vectors (rCELO) in cosmids is very efficient. This method takes advantage of the CELO genome size (43.8 kb), which is close to the size of bacteriophage  DNA (~52 kb). We cloned the CELO genome into the 5.8-kb cos1 cosmid vector to obtain the 49.6-kb cos/CELO cosmid (Fig. 1), which is the perfect length to be packaged into bacteriophage  heads, which can accommodate DNA molecules from 38 to 52 kb in length when flanked by two COS sites. This system allows the construction of recombinant CELO genomes using only in vitro ligation (avoiding difficulties caused by the use of homologous recombination) and the selection of ligation products only of the expected length. Moreover, E. coli infection with cosmid-containing bacteriophage  heads is much more efficient than transformation, particularly for plasmids of about 50 kb. It gives high yields of ampicillin-resistant colonies and large amounts of recombinant viral DNA.

View larger version (108K): [in this window] [in a new window]   FIG. 7.   Restriction pattern of cosmid and viral DNA preparations of three CELO vectors digested with HindIII. Lanes b to d show wild-type CELO DNA, cloned in cosmid cos1 (b), extracted from CsC1 gradient-purified CELO virus (c), and extracted from infected LMH cells (d). Lanes e and f show DNA from the GFP-expressing CELO vector carrying the largest (3.6 kb) deletion at the right end of the viral genome, cloned in cosmid cos1 (e) and extracted from infected cells (f). Lanes g and h show DNA from the rCELO-IBDA vector, cloned in cosmid cos1 (g) and extracted from infected cells (h). Raoul DNA marker (Quantum Appligene) (lane a) and  DNA digested with BstEII (lane i) were used as size markers.

Sixteen of the 22 CELO unassigned ORFs seem to be dispensable for virus replication. We used directed mutagenesis to introduce translation termination codons (TAA, TAG, or TGA) into each of the ORFs located at the left end (the A fragment, which includes ORFs 1, 2, 3, 4, 12, 13, 14, and 15) and at the right end (the D fragment, which includes ORFs 7, 8, 9, 10, 11, 16, 17, and 18) of the CELO genome (Fig. 3 and Table 1). When one ORF was overlapping another, on the same or on the opposite strand, we managed to introduce a mutation that did not change the amino acid sequence of the second ORF, taking advantage of the degeneration of the genetic code. For all mutant viruses obtained after transfection of LMH cells with a cos/CELO cosmid carrying a mutation, we verified conservation of the mutation by amplification of the mutated region with the high-fidelity Pfu DNA polymerase and sequencing of both strands of the purified PCR product. The results obtained on LMH cells with the 16 point-mutated CELO genomes are given in Table 2. We noted the time of appearance of a CPE after transfection and then after inoculation of fresh cells with aliquots of the transfection lysates. Only 2 of the 16 mutants (ORF1- and ORF16-deficient mutants) did not produce a CPE after transfection, but we harvested transfected cells within a period of 15 days because after that time untransfected cells also became degraded. However, all of the 16 mutants tested produced a CPE between 2 and 4 days after inoculation of cells with the transfection lysates. All of them could be amplified, giving titers comparable to those obtained with the wild-type CELO. The restriction profile of viral DNA was identical for mutants and wild-type CELO, and the mutation was maintained in all cases, as shown by unambiguous sequencing results. As none of the 16 ORFs seems to be essential for viral replication, at least in vitro in LMH cells, strategies could be developed for deletion and insertion of foreign DNA into the A and D fragments of the CELO genome without affecting the replication functions of the vector.

Deletion of CELO virus putative packaging signal. The packaging signal sequences () are localized between 100 and 500 bp in all known adenoviruses, near the 5' inverted terminal repeat. We supposed that this would be the case for CELO virus, although no sequence homology was found by comparing this region with packaging signals that have been clearly identified in other adenoviruses. In order to confirm this hypothesis, and also to dispose of packaging-defective CELO genomes that could be used as helper viruses for production of nonreplicating vectors, we constructed two genomes (CELO 1 and CELO 2) with various deletions between nt 80 and 920. Neither construction produced infectious particles when transfected into LMH cells. No CPE occurred even 15 days after transfection or after inoculation of fresh cells with the transfection lysates (Table 2). Although these two deletions also inactivated the ORF1 gene, the lack of virus production was more probably due to deletion of the packaging signal, as expected, since we demonstrated that inactivation of ORF1 by directed mutagenesis did not affect viral replication.

Evaluation of various CELO vector constructions by using the GFP reporter gene. Using our results in mutagenesis experiments and the data published by Michou et al. (23) on deletion experiments and those published on the transcriptional organization of the CELO genome (26), we estimated the possibilities of deletion and insertion of foreign DNA into the vector backbone. In order to evaluate our constructions on LMH cells, we used eGFP cDNA under control of the CMV promoter as a reporter system. This system allowed us to monitor transfection efficiency and propagation of the vector through the cell monolayer in real time, without affecting the cells (Fig. 8). As the CMV promoter is activated immediately upon entering the nucleus, we could observe the formation of plaques due to propagation of the vector, appearing as green fluorescent areas, earlier than the time necessary for appearance of a CPE.

View larger version (140K): [in this window] [in a new window]   FIG. 8.   Fluorescence microscopy observation of the generation of a replication-competent GFP-expressing CELO vector. (a) GFP expression 12 h after transfection of LMH cells with cosmid DNA carrying the recombinant CELO-GFP genome. (b) Appearance of comet-like fluorescent plaques due to propagation of the recombinant virus from primary transfected cells to the adjacent cells. (c) Single fluorescent plaque showing disappearance of part of the fluorescent cells due to CPE induced by replication of the vector. (d) Same field observed in white light, confirming that the fluorescent plaque corresponded to a lysis plaque similar to that observed with wild-type CELO virus.

View larger version (47K): [in this window] [in a new window]   FIG. 9.   Flow cytometry analysis of CELO-GFP-infected cells. LMH cells were infected with the eight distinct CELO-GFP vectors (#1 to #8, see Table 3) at an MOI of 5 or 10 PFU/cell. Cells infected with wild-type CELO virus in the same conditions and uninfected LMH cells were used as controls. After 16 h, cells were incubated with PI to color apoptosing cells in red and submitted to flow cytometry. (a) Percentages of green fluorescent (GFP) and red fluorescent (PI) cells were determined for each vector to assess transduction efficiency and cell death, respectively. (b) Average intensity of the green fluorescence per cell was measured to compare the level of GFP expressed by the various vectors.

Generation of CELO vectors for vaccine applications. As shown with our previous experiments, it is possible to delete up to 3,620 bp at the right end of the viral genome (D fragment) and to preserve the ability of the vector to replicate (Table 4). This large deletion between KpnI (nt 40065) and EcoRV (nt 43685) eliminated ORF9, -10, and -11 coding sequences and probably the promoter for ORF22 but allowed for insertion of up to 5 kb of foreign DNA, since the CELO capsid can package at least 1.4 kb of additional DNA. We initially decided to use this vector backbone, which was carrying the largest deletion, to construct CELO vectors for vaccine applications, especially as it was one of the most effective in term of GFP expression (Fig. 9, vector no. 8). In order to simplify the insertion of various genes into this vector, we constructed an intermediate plasmid containing the deleted D fragment with either a large MCS (Fig. 6) or a smaller MCS flanked by the CMV promoter and BGH polyadenylation signal in place of the deletion.

View larger version (70K): [in this window] [in a new window]   FIG. 10.   Immunoprecipitation of radiolabeled proteins expressed in rCELO-IBDV-infected LMH cells. Proteins were extracted from LMH cells alone (lanes b, g, m) or cells infected at an MOI of 4 PFU/cell with wild-type CELO (lanes c, h, n), rCELO-IBDA (lanes d, i, o), rCELO-IBDA (lanes e, j, p), and rCELO-IBDA-GFP (lanes f, k, q). Lanes b to f show proteins recognized by polyclonal anti-CELO antibody. Lanes g to k show proteins recognized by anti-VP2 monoclonal antibody. Lanes m to q show proteins recognized by anti-VP3 monoclonal antibody. Molecular size markers were present in lanes a and l.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Our method for construction of recombinant adenoviruses based on the use of cosmids is very efficient, at least when applied to the avian adenovirus CELO. When starting vector construction from a plasmid such as pMECA/D/MCS, which was developed to simplify the cloning of foreign DNA into the CELO genome, it is possible to obtain up to 10 recombinant cosmids within 2 weeks (with only one or two operators) and then the recombinant virus vectors within 2 or 3 further weeks using standard biological methods. In order to simplify the procedure, cos/CELO circular DNA was directly transfected into LMH cells without release of the viral DNA from the cosmid vector, despite the possibility that viral particles may have been obtained more efficiently by transfecting linearized viral DNA. Homogenous viral stocks were obtained without need for plaque purification.

Using this method, we were able to generate mutant viruses by introduction of knockout point mutations into 16 of the 22 CELO virus unassigned ORFs. In order to avoid any wild-type virus contamination, handling of the mutated CELO viruses (i.e., transfections and infections) was performed in separate laminar flow cabinets where wild-type CELO had never been handled. For each mutant virus, presence of the introduced mutation was verified by sequencing of the mutated region, and the expected nucleotide was always found at the expected position on both DNA strands. As sequence determination of the mutation showed no ambiguity, it confirmed the absence of the wild-type sequence. Absence of wild-type virus contamination was also confirmed by the fact that we never obtained infectious particles from the packaging-defective CELO genomes 1 and 2, which were handled under the same conditions. We established that none of these 16 ORFs were essential to achieve a complete viral cycle in LMH cells. Even when no CPE was seen within 15 days after transfection, which was the case for the ORF1- and ORF16-deficient mutants, infectious viruses could be rescued and propagated from the transfected cells. The lack of CPE might therefore have been simply due to low transfection efficiency.

These surprising results may be due to the fact that the mutated ORFs have little or no function at the intracellular level but act only in a whole body context, e.g., to interact with the host immune response. That is probably not true for the only two ORFs which have already been identified. The first one, ORF8, encodes the GAM-1 protein, which is functionally homologous to the human adenovirus E1B 19K protein with antiapoptotic properties (6, 20). This protein, together with ORF22 (which encodes a protein functionally homologous to the E1A protein [20]), interacts with the retinoblastoma protein. It activates the E2F pathway and induces entry of the host cell into the S phase, probably to make possible the efficient replication of the viral genome. The second one, ORF1, encodes a protein that has been demonstrated to be a dUTPase (34) and could therefore help viral DNA replication by keeping a pool of dUMP for dTTP synthesis. Apart from these two genes, only ORF2 is significantly homologous to a known gene, the parvovirus REP gene (7), but its function remains unknown.

Another possibility may be that each of the 16 ORFs has one or more functional homologues among the other viral genes (duplicate genes), so that suppression of one viral function involves inactivation of several ORFs. The last hypothesis would be that the LMH cell line, which is descended from a chemically induced hepatocarcinoma (15), provides an environment so favorable to CELO replication that effects of mutations are masked, or that CELO ORFs are functional homologues to cellular factors that counteract the effects of the mutations. Thus, although LMH cells do not seem to be suitable for study of the function of CELO genes during the viral cycle, they provide a powerful tool for the generation and amplification of recombinant CELO vectors. In order to identify the functions of the mutated CELO genes, we plan to evaluate the replication capacity of our mutant CELO viruses in vitro on cultured chicken primary cells or embryo cells from various sources (e.g., liver, kidney, and lymphocytes), as well as in vivo in specific-pathogen-free (SPF) chickens. This may also indicate the route CELO virus uses to invade the body and the cell type or organ where it naturally replicates in vivo.

From the results obtained in vitro with the 16 point-mutated CELO viruses, together with our previous results on the transcriptional organization of the viral genome, we were able to theoretically define regions of the genome where deletions and insertions of expression cassettes will not impair the ability of the vector to replicate, at least in vitro. We therefore constructed several recombinant CELO viruses carrying a GFP expression cassette inserted at various sites of the genome.

As we did not detect any transcription products for ORF2 and ORF14 by reverse transcriptase PCR or by PCR screening of cDNA libraries (26), and as we showed that the ORF2-deficient and ORF14-deficient mutants replicated efficiently, we chose the region of these two ORFs (between nt 1500 and 3500) for our first constructs. The first rCELO-GFP vector was obtained by simply inserting a P_CMV-GFP cassette at a SphI site localized in ORF2. This resulted in addition of 1.4 kb of foreign DNA to the viral genome and disruption of ORF2 on the upper strand, but no coding sequence was altered on the lower strand. A second vector was then obtained by insertion of a GFP expression cassette in place of a 1.4-kb deletion that eliminated ORF2 and ORF14. Those two constructions produced infectious viruses on LMH cells, confirming our previous results, and GFP was efficiently expressed.

We also constructed two vectors with a GFP expression cassette inserted in place of short deletions in the region of ORF1, between nt 700 and 2000 at the left end of the genome: a 404-bp deletion that eliminated 65% of the coding sequence in the 5' part of ORF1, and a 329-bp deletion in the 3' noncoding region that eliminated the ORF1 polyadenylation signal (AATAAA at nt 1945). This region was chosen as a site for deletion and insertion because only ORF1 was impaired. This gene encodes a dUTPase and its early transcription has been clearly demonstrated during the viral cycle (26), but its exact function remains unknown. A sequence homologous to CELO ORF1 has been found in other adenoviruses, in particular human adenovirus E4 ORF1 (34) and avian adenovirus serotype 8 (FAV-8) long terminal repeat-1 genes (4). Conservation of this sequence among various adenovirus types may imply that the dUTPase function is of importance for virus multiplication. This could explain the poor results obtained from CELO-GFP no. 4 (with ORF1 deleted), which gave poor yields of transduced cells as well as low GFP expression levels. Despite these findings, inactivation of CELO ORF1, either by point mutation or by deletion or insertion, did not preclude viral replication. Nevertheless, this observation was made on a cell line that was actively replicating, but this may not be true for other cell types, in particular for quiescent cells. Thus, ORF1 may have an essential function for completion of the viral cycle in vivo, as GFP was expressed efficiently.

The other rCELO-GFP vectors were constructed by insertion of the GFP expression cassette in the region of ORFs 9, 10, and 11, between nt 40000 and 43700 at the right end of the genome. Four different deletions were achieved: a 1.6-kb deletion that eliminated ORFs 9 and 10, 1.4-kb and 1.9-kb deletions that both eliminated ORFs 10 and 11, and finally a 3.6-kb deletion that eliminated all three ORFs. These two last deletions had been previously described by Michou et al. (23), who established that the deletions were not impairing CELO replication in LMH cells and chicken embryos. The three genes were located on the upper strand and no overlapping coding sequence was present on the lower strand, so this region was a good candidate for insertion of large DNA fragments. Amino acid sequence comparison of CELO ORFs and databases showed some homology between ORFs 9 and 10 and the receptor for interleukin-3. Interestingly, these two ORFs share significant homologies and could be a duplicate gene. ORF11 showed homology with the CD4 precursor (a cell surface glycoprotein) of T lymphocytes. These homologies were weak and no extrapolation can be made on the functions of CELO ORFs, but it is possible that ORFs 9, 10, and 11 are involved in escaping or modulating the host immune response against the virus. This could explain why inactivation of these ORFs had no visible effect on virus multiplication in vitro, but could also indicate that functions associated with these ORFs are necessary for virus replication in vivo. One argument for this hypothesis is that sequences homologous to ORF9 and -11 have been found in FAV-8, another avian adenovirus (4). The FAV-8 ORF RTR4 shares homology with CELO ORF11 and is located in the same region on the upper strand. FAV-8 RTL1 shares homology with CELO ORF9 but is located on the lower strand. Transcription of these ORFs has been demonstrated to appear early in the viral cycle, for both FAV-8 and CELO viruses. Thus, the consequences of the deletion of these genes in our constructions must be investigated in vivo in SPF and conventional chickens. In the case of the two constructions with the largest deletions, the deletion went up to the nt 43685 EcoRV site and eliminated the putative ORF22 promoter (with a TATA box at nt 43,450). This had no visible effect on virus multiplication in LMH cells, but it could prevent efficient viral replication in cells that do not proliferate so actively. Although we didn't demonstrate that ORF22 transcription was really disrupted by the deletion, the effect on ORF22 expression may have to be investigated as it could be of crucial importance for CELO replication in vivo.

Since our results with directed mutagenesis experiments and those obtained with the various rCELO-GFP vectors were in agreement, we created recombinant CELO vectors carrying genes for vaccine applications. The first vectors we constructed and evaluated in vitro were all obtained by insertion of an expression cassette into the right end of the CELO genome, in place of the largest 3.6-kb deletion that includes ORFs 9, 10, and 11 and the putative ORF22 promoter. For the vector carrying the complete A segment of IBDV (1), we showed that the polyprotein was expressed efficiently and was processed into the three expected polypeptides, VP2, VP3, and VP4 (21), after infection of LMH cells. This indicated that the CELO virus can be used as an efficient expression vector, at least in vitro. Moreover, the restriction profile of this vector did not change after 10 serial passages on LMH cells.

All these findings offer interesting prospects for various in vitro applications, but many questions remain to be resolved concerning the use of CELO virus as a gene transfer vector in vivo. Although in vitro replication and stability of CELO virus seemed not to be affected by modification of its genome, little is known about the in vivo virus cycle in the chicken. Some preliminary experiments have been performed in our laboratories on SPF chickens (devoid of preexisting antibodies). Inoculation of wild-type CELO virus into 1-day-old chickens through various routes, with doses of up to 10⁷ PFU/chicken, did not induce any morbidity or pathology within 6 weeks after inoculation. Viral DNA could be detected by PCR in various organs between 3 and 7 days. Anti-CELO antibodies were detected by enzyme-linked immunosorbent assay in blood samples from all animals 3 weeks after inoculation, but neutralizing antibodies were found only in blood samples of 20 to 25% of the animals, within 3 to 4 weeks after inoculation.

Thus, the innocuousness of CELO virus on SPF chickens, its early spread into the whole body, and the late appearance of antibodies directed against viral proteins are all arguments supporting the utilization of CELO virus as a vector for vaccination of poultry. Although our CELO vector carrying genes for vaccine application replicated efficiently in vitro, we do not know at the present time if it will replicate in vivo, despite the deletion of several viral genes (particularly as the functions of those genes are unknown), and if the inserted gene will be sufficiently expressed to induce protective immunity. Moreover, we do not know how the virus will behave when inoculated into conventional farm chickens that might already possess anti-CELO antibodies, and we also do not know the doses necessary for efficient vaccination.

Another potential application that could be investigated is the use of CELO-based vectors for gene transfer in mammalian species. It has been reported that CELO virus can transduce a variety of mammalian cell lines but does not replicate in such cells (23). We ourselves verified these observations on adherent cells of porcine, equine, simian, and human origins by using a GFP-expressing CELO vector. Thus, CELO virus could be used either for vaccination of nonavian species (e.g., pigs, horses, dogs) or for gene therapy applications in humans.

Whatever the future applications for CELO vectors, much is unknown about the biology of this virus and has to be investigated, particularly its oncogenic properties, its natural target cells, and the route for cell entry. Studying genome stability and the versatility of the recombinant viruses is also of primary concern before such a vector can be used for therapeutic or research applications.

Our method, which uses cosmids for construction of recombinant adenoviruses, is theoretically applicable to any adenovirus type, provided that the viral genome possesses single restriction sites. This allows for cloning of subfragments into a standard plasmid, where it is possible to introduce any modification by standard methods. Reintroduction of the modified fragment into the viral genome by simple ligation is then very easy, and packaging of the ligation product into bacteriophage  heads provides efficient recovery of the expected recombinant genome, avoiding the laborious screening step needed when using homologous recombination methods. Using this method, we are now able to modify any region of the CELO genome, and so we can easily delete or mutate viral genes with pathogenic or oncogenic properties to make the vector safer. We can also engineer the viral structural components for retargeting of the vector or for structural study of virion assembly.