Construction, Rescue, and Characterization of Vectors Derived from Ovine Atadenovirus

Abstract

Gene transfer vectors derived from ovine atadenovirus type 7 (OAdV) can efficiently infect a variety of mammalian cells in vitro and in vivo to deliver and express transgenes. However, early OAdV vectors were designed on human mastadenovirus principles prior to the complete characterization of OAdV genes and transcripts. The distinctive arrangement of the OAdV genome has suggested ways to improve OAdV vector design and utility. We therefore developed a cosmid-based approach that allows efficient construction of recombinant ovine atadenovirus genomes in which the transgene is inserted at one of three sites. Viruses were rescued by transfection of viral DNA into a new ovine fetal skin fibroblast producer cell line, HVO156. The suitability of the three insertion sites was compared with respect to virus rescue efficiency, gene expression levels, and genetic stability of the vectors. We found that one vector with a transgene inserted at site 1, between the pVIII and fiber genes, was unstable. Only one vector that carried a transgene at site 2, near the right end of the genome, together with a nearby deletion was rescued. In contrast, several vectors with different transgenes inserted in site 3, between the E4 and RH transcription units, were repeatedly rescued, and these vectors were stable over at least four passages. Transgene orientation in site 3 had only little effect on expression. Finally, a vector carrying a human factor IX cDNA at site 3, when administered intravenously, produced nearly physiological levels of human factor IX in mice. The availability of an efficient method for vector construction and the identification of a new insertion site for virus rescue and gene expression substantially enhance the utility of the OAdV vector system.

The most common vectors available for gene therapy and vaccination are derived from human adenovirus type 5, partly for historic reasons, because the virus has been studied at the molecular level for many years, and partly because it is of low pathogenicity and oncogenicity. Vectors deleted in all viral genes have been developed and successfully used in animal studies (25, 26, 34). However, a large proportion of potential patients have acquired immunity to human adenovirus type 5 due to natural infection in early childhood (11), and human adenovirus type 5-mediated gene transfer will therefore be hampered in these patients, at least for some modes of vector administration.

Therefore, we and others are developing vectors that are derived from nonhuman virus species (reviewed in references 3 and 21). Ovine atadenovirus type 7 (OAdV-7), derived from isolate 287 of ovine adenovirus (5), is the prototype of the recently recognized genus of atadenoviruses (2). OAdV replicates productively in certain ovine producer cells and infects, but fails to replicate in other mammalian cells, including human cells (12, 15, 16). Infection of such cells occurs in a coxsackie adenovirus receptor (CAR)-independent manner via an unknown receptor (42). OAdV has been engineered as a vector (41, 43), and DNA up to 114% of the normal genome size can be packaged, allowing insertion of at least 4.3 kb of foreign DNA without deletion of viral sequences.

As current vectors retain all the viral genes, it is important to note that OAdV lacks transforming ability in rodent cells (43)^. After systemic or local administration OAdV vectors can efficiently deliver foreign DNA into cells and organs of small animals. Importantly, gene delivery in vivo can occur in the presence of preexisting immunity to human adenovirus type 5 (10, 18, 19, 21, 37; T. Wüest, M. Hillgenberg, U. Schneeweiss, G. W. Both, A. M. Prince, C. Hofmann, and P. Löser, submitted for publication). The characteristics of OAdV make this virus an attractive candidate as a vector for gene therapy and vaccination. However, vector development has been hindered by difficulties in the construction and rescue of OAdV-based vectors.

An important consideration in vector design is the site in the viral genome that is selected for transgene insertion because this may significantly affect vector function and utility. This was an important issue initially in the development of human adenovirus vectors. Some of the earlier adenovirus-derived vectors utilized an insertion site adjacent to the inverted terminal repeats (ITR) at the extreme right end of the adenovirus genome (33), while others carried foreign DNA within the nonessential E3 region (23, 35). However, insertion of genes into the deleted E1A/B region of adenovirus has become the most frequently used strategy for recombinant vector construction because for human adenovirus complementing cell lines such as 293 are available, and the resulting viruses are replication defective, deleted of their major oncogenes, and have an increased carrying capacity. However, OAdV lacks a homologue of the human adenovirus E1 region and does not have an E3 region between its pVIII and fiber genes (38). In addition, there are genes of undefined function at both ends of the OAdV genome and a promoter that is located within the ITR (4, 39). Thus, it is not possible to extrapolate from adenovirus to determine sites for gene insertion in OAdV.

Nevertheless, prior to the elucidation of a transcription map 13, three potential sites for insertion of foreign DNA were identified within the OAdV genome (39, 41). Site 1 (also referred to as site I) between the genes coding for the viral proteins pVIII and fiber was identified based on precedents with adenovirus type 5. Site 2 (also referred to as site II) comprises a unique SalI site located 902 bp from the 3′ end of the viral genome within the open reading frame for the RH2 gene (40). Site 3 (also referred to as site III) is located within a short noncoding region between the E4 and RH transcription units. In addition, the deletion of about 2 kb of the viral genome between sites 2 and 3 did not significantly interfere with virus growth (41). However, viruses expressing a functional transgene have been only described for sites 1 and 2 (10, 13, 16, 37), and the impact of transgene position on virus rescue and growth, stability, and transgene expression has not been examined.

The currently limited availability of OAdV-derived vectors has several reasons. First, recombinant OAdV genomes have been traditionally produced by either classical cloning methods (13, 16, 37, 39, 41), which are relatively slow and inefficient due to the complexities of working with large (≈33 kb) plasmids, or by a recombination procedure in Escherichia coli BJ5183 similar to that described for adenovirus type 5 vector construction (6, 10). For OAdV, the latter was also very inefficient, possibly due to repeat sequences in the OAdV genome which has a very high A/T content relative to mastadenoviruses (39). Second, successful rescue of virus following transfection of producer cells with recombinant genomes is a very rare event, probably due to low transfection efficiency obtained with the original producer cell line, CSL-503 (30). Third, certain recombinant viruses with site 1 transgenes appeared to be unstable (G. W. Both, unpublished data), possibly due to the site and/or transgene sequence. Therefore, the development of OAdV vectors will be strongly enhanced by the identification of an alternative transgene insertion site and the availability of an effective system for OAdV vector construction.

In this study, we describe a cosmid-based system for efficient generation of recombinant OAdV genomes carrying insertions in any of the three integration sites. We show that transfection of naked recombinant genomes that carry site 3 insertions into a new producer cell line, HVO156, results reliably in recovery of infectious viruses. We also demonstrate that viruses with the same transgene inserted into different sites behave differently with respect to vector stability, that insertions in site 3 have little impact on vector growth and that the orientation of the transgene does not influence expression levels. Lastly, we show that a recombinant with a site 3 insert is capable of high-level expression of a therapeutic gene in vivo. Together these improvements greatly enhance the utility of the OAdV vector system.

MATERIALS AND METHODS

Cosmid cloning vectors.

OAdV cloning vectors were constructed by starting with plasmid pMVKpnΔ. This contains the ampicillin resistance gene, a colE1 origin of replication, the lambda phage cos signal and a single KpnI site directly flanked by two recognition sites for I-SceI. pMVKpnΔ was generated from plamsid pMV (9) by excision of the corresponding I-SceI fragment and subsequent insertion between the I-SceI ends of a symmetric linker generated by self-hybridization of the oligonucleotide 5′-CCCTAGGTACCTAGGGATAACAG-3′. This linker restored the I-SceI sites and inserted a KpnI site between them. After deletion of an additional KpnI site at the beginning of the cos signal, pMVKpnΔ was obtained. For construction of pOAdV1-cos with unique insertion sites (XhoI, PspOMI, ApaI) for cloning into OAdV site 1, the OAdV genome was cut out from pOAdV-poly21 by digestion with KpnI and the fragment inserted into the corresponding site of pMVKpnΔ. Construction of pOAdV2-cos and pOAdV3-cos started with the insertion of a 3.7-kb noncoding fragment of human X-chromosome stuffer DNA into the cosmid backbone of pMVKpnΔ, giving rise to pMVKpnΔS. After deletion of a NotI site in the stuffer fragment plasmid pMVKpnΔNotΔS was subsequently obtained. The entire OAdV genome was excised from pOAdV600 (24) by digestion with KpnI and inserted into the unique KpnI site of pMVKpnΔNotΔS. This resulted in pOAdV3-cos with a single NotI site for transgene insertion into OAdV site 3. pOAdV2-cos was then constructed from pOAdV3-cos by deletion of the NotI/SalI fragment between sites 2 and 3 and insertion of a linker generated by annealing oligonucleotides 5′-GGCCGCTCGAAGCGCGCCG-3′ and 5′-TCGACGGCGCGCCTCGAGC-3′. The OAdV genome in pOAdV2-cos is similar to pOAdV603 (41) but has a different polylinker (SalI, AscI, XhoI, NotI) for transgene insertion.

Construction of recombinant OAdV genomes.

Plasmid pRSVpoly contains a polylinker inserted between the Rous sarcoma virus (RSV) 3′ long terminal repeat and the bovine growth hormone polyadenylation signal. These regulatory elements have a similar activity in all cell lines used in this study as determined by cotransfection experiments (data not shown). The cassette is flanked at both ends by recognition sites for XhoI, NotI, and PmeI. Insertion of the cDNAs for human α₁-antitrypsin (hAAT), firefly luciferase, human coagulation factor IX (fIX), and a C-terminally truncated hepatitis C virus core antigen (amino acids 1 to 176 of hepatitis C virus BK) into the polylinker resulted in pRSVhaat, pRSVluc, pRSVfIX, and pRSVcore, respectively. Cleavage of these plasmids with NotI released the expression cassettes from the plasmid backbone.

For pOAdV-cos-based construction of recombinant genomes, pOAdV1-, -2-, and -3-cos were linearized with NotI and ligated with the NotI fragments containing the respective expression cassettes for hAAT, luciferase, fIX, or core. The ligation products were then packaged into phage heads with Gigapack III Gold (Stratagene). Packaging and infection of Escherichia coli JM109 or DH5α were performed according to the manufacturer's protocol. The presence of the insert in plasmids isolated from bacterial colonies was confirmed by PCR with the following primer pairs for sites 1, 2 and 3, respectively: site 1, 5′-CCTCGCTCTGGAGGATTAAC-3′ and 5′-GGCCGCTCTAGAACTAGTGG-3′; site 2, 5′-GGCCGCTCTAGAACTAGTGG-3′ , and 5′-GTGGCCTGTTTAACTCATCC-3′; and site 3, 5′-CCTAACCCATTGCGTTCCTC-3′ and 5′-GCTTGGCCTGTATGTAATGC-3′. Since the ligation procedure yielded genome products in which the insert was in one of two possible orientations, they are referred to as orientation 1 (o1) when the orientation of the inserts is the same as that of the late OAdV genes and as orientation 2 (o2) in the reverse case. Orientation was determined by PCR with one primer within the respective cDNA and one primer flanking the integration site.

Generation and propagation of recombinant OAdV.

Recombinant OAdV genomes were liberated by digestion with Meganuclease I-SceI (Roche Molecular Biochemicals, Mannheim, Germany). Cleavage products were phenol-chloroform purified and ethanol precipitated. DNA (0.4 μg) was then transfected onto HVO156 cells (20) seeded 24 h before transfection into 24 well plates at a density of 2.5 × 10⁴ cells per cm². Transfection was performed with the Lipofectamine Plus reagent (InvitroGen, Karlsruhe, Germany) following the manufacturer's protocol. A cytopathic effect usually became evident between days 7 and 21 posttransfection. Recombinant vectors were passaged three times on increasing numbers of HVO156 cells and CsCl gradient-purified stocks were prepared from the A4 or A5 passage as described previously (20). The number of viral particles (vp) was determined following the method of Maizel et al. (22) by measuring the absorption at 260 nm. The number of infectious particles (ip) was determined in an end point dilution assay on HVO156 cells as described elsewhere (7). Typical virus titers were in the range of 10¹⁰ to 10¹² vp per ml and the vp/ip ratio was generally less than 20. To verify the integrity of recombinant viruses, DNA from CsCl gradient-purified virus was isolated by standard methods and subjected to restriction enzyme analysis.

Cell lines and reporter gene assays.

HVO156 cells (embryonal ovine skin fibroblast, permissive for OAdV) (20), 208F (rat fibroblast) (31), and IMR-90 (human lung fibroblast) (28) were maintained at 5% CO₂ in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 2 mM glutamine, 15% fetal calf serum (BioWest, Nuaillé, France), 100 IU of penicillin/ml and 50 μg of streptomycin/ml. HuH7 (human hepatocellular carcinoma) (27) and HepSV40 (mouse hepatocytes expressing simian virus 40 large T antigen) (29) were maintained under the same conditions but with 10% fetal bovine serum. Infection of cells with recombinant OAdV was carried out for 16 h in six-well plates (2.5 × 10⁴ cells per cm²) and was stopped by addition of fresh medium. For determination of haat and factor IX gene expression, medium was replaced 24 h after the end of infection, and samples were collected 24 h later. The enzyme-linked immunosorbent assay (ELISA) for quantification of haat has been described previously (7). Human Factor IX levels were quantified by ELISA according to the method published by Baru et al. (1) with the exception that a horseradish peroxidase (POD)-linked antibody (Pierce, Rockford, Ill.) was used for detection. Luciferase activity was determined in cell lysates 24 h after the end of infection, and the assay was performed as described earlier (17).

Western blotting.

208F cells (2.5 × 10⁴ cells/cm²) were infected overnight at an increasing multiplicity of infection with OAdV3-core.o1 or OAdV3-core.o2. At 40 h postinfection, cells were harvested, washed three times in phosphate-buffered saline, and lysed for 3 h on ice in buffer containing 50 mM HEPES, pH 7.5, 200 mM NaCl, 1 mM EDTA, 2.5 mM EGTA, 10% glycerin, 0.1% Tween 20, 10 mM beta-glycerolphosphate, 1 mM NaF, 0.1 mM Na₃VO₄, 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride, 10 μg of leupeptin/ml and 2 μg of aprotinin/ml. Samples were then centrifuged for 20 min at 15,300 rpm and 4°C, and supernatants were stored at −80°C. Proteins (40 μg) were separated on a 4 to 12% Bis-Tris gel (Invitrogen) and transferred to a nitrocellulose membrane (Amersham, Freiburg, Germany). Detection of recombinant core protein was performed with a monoclonal antibody specific for core (MA1-080; Dianova, Hamburg, Germany) at a 1:500 dilution followed by incubation with a 1:2,500-diluted POD conjugated rabbit anti-mouse immunoglobulin G1 (Dianova). Visualization of specific signals was achieved with the ECL system (Amersham).

Animal procedures.

Female BALB/c mice (6 to 8 weeks old) were obtained from Charles River, Sulzfeld, Germany. Mice were given tail vein injections of 10¹⁰ vp in a total volume of 100 μl of virus storage buffer (135 mM NaCl, 3 mM KCl, 10 mM Tris-Cl, pH 7.8, 10% glycerol). At day 3 postinjection, blood was collected from the external jugular vein and plasma was subjected to an ELISA specific for human Factor IX.

RESULTS

Construction of cosmid-based cloning vectors and virus rescue.

We first generated a set of three cosmid cloning vectors for convenient insertion of transgene cassettes into site 1, 2, or 3. The vectors contained variants of the complete OAdV genome in which unique restriction sites for transgene insertion were present at either site 1, 2, or 3 and a bacteriophage lambda cos signal that allowed application of cosmid cloning techniques. The plasmids also contained recognition sites for the super-rare cutting endonuclease I-SceI, which can release linear, infectious vector genomes with overhangs of 6 bp at each ITR after digestion. In addition, pOAdV2-cos contains an approximately 2,000-bp deletion described for pOAV603 (nucleotides 26676 to 28675 [41]) and therefore lacks most of the OAdV RH region. The principles and processes that were used for the construction and rescue of recombinant OAdV from cosmid clones are shown in Fig. 1 and are similar to those previously used for construction of helper-dependent human adenovirus vectors (9).

FIG. 1.

Principles of cosmid-based cloning of recombinant OAdV genomes. Plasmids containing the OAdV genome (pOAdV1-, -2-, and -3-cos) were cut with an appropriate restriction endonuclease and ligated to a DNA fragment containing the transgene expression cassette (insert). In vitro packaging of the ligation products into phage heads and infection of E. coli resulted in bacterial clones containing plasmids with the recombinant viral genomes. Digestion of these plasmids with I-SceI released linear recombinant OAdV genomes with free inverted terminal repeats. These were used for transfection of the producer cell line and virus rescue.

To confirm that insertion of short polylinkers at these sites did not interfere with virus growth, we first applied the method shown in Fig. 1 to pOAdV1-, -2-, or -3-cos, which lacked inserts. Transfection of I-SceI-digested plasmids onto producer cells generated infectious viruses with all three cloning vectors, each of which grew to comparable titers of 3.2 × 10¹¹ (OAdV1-cos), 2.9 × 10¹¹ (OAdV2-cos), and 4.9 × 10¹¹ vp/ml (OAdV3-cos).

Effect of insertion site on transgene expression.

To test the efficiency and utility of OAdV-cos-based plasmids for vector construction and virus rescue, recombinant OAdV genomes were generated that contained the human α₁-antitrypsin (hAAT) cDNA and bovine growth hormone polyadenylation signal under control of the RSV 3′ long terminal repeat. The same expression cassette was inserted into site 1, 2, or 3. Bacterial clones were tested for the presence of the insert by PCR, and the orientation of the transgene within the viral genome was determined by restriction enzyme analysis. Table 1 summarizes the results of the genome constructions.

TABLE 1.

Efficiency of cosmid-based construction of recombinant OAdV genomes

Cloning vector	Insert	E. coli strain	Cloning efficiency (no. of colonies per μg of DNA)	No. of clones positive in PCR/total	Insert orientation (no. positive/no. tested)
					o1	o2	Ma
pOAdV1-cos	RSV-haat	DH5α	7.6 × 103	4/6	2/4	2/4	-
		JM109	2.4 × 104	6/6	1/4	3/4	-
pOAdV2-cos	RSV-haat	DH5α	3.0 × 104	4/6	ND	ND	ND
		JM109	2.8 × 104	5/6	2/4	2/4	-
pOAdV3-cos	RSV-haat	DH5α	1.2 × 104	8/8	4/8	1/8	3/8
		JM109	7.5 × 103	8/8	2/8	2/8	4/8

^aM, multiple insertions; ND, not determined.

Recombinant OAdV were then rescued after transfection of HVO156 cells with naked viral genomes released by I-SceI digest from plasmids pOAdV1-haat, pOAdV2-haat, and pOAdV3-haat. Whereas OAdV3-haat was rescued with the insert in either orientation (o1 and o2), OAdV1-haat and OAdV2-haat yielded only viruses with the insert in one orientation (o2 for OAdV1-haat, o1 for OAdV2-haat), although the total number of viruses rescued was <5. In addition, several other recombinant OAdV2 genomes that carried different transgenes failed to rescue (P. Löser and M. Hillgenberg, unpublished data). The structure of the recombinant viral genomes is shown in Fig. 2A. Viruses carrying the haat gene in one of the three sites were then used to inoculate forty 150-mm dishes of HVO156 cells and virus was purified by CsCl gradient centrifugation. The yield of each recombinant was comparable (OAdV1-haat, 3.6 × 10¹¹ vp/ml; OAdV2-haat, 5.5 × 10¹¹ vp/ml; OAdV3-haat, 3.5 × 10¹¹ vp/ml). In comparison, typical titers of wild-type OAdV ranged between 2 and 10 × 10¹¹ vp/ml for a similar size preparation.

FIG. 2.

Structure of recombinant OAdVs expressing haat. (A) The 3′ portion of the OAdV genome with insertion sites 1, 2 and 3 is shown at the top. Insertion of the haat expression cassette resulted in the recombinants OAdV1-haat, OAdV2-haat, and OAdV3-haat. Note that OAdV2-haat lacks 2 kb of viral DNA between the insertion site and part of the RH2 gene. The positions of the XhoI sites critical for distinguishing these recombinants are indicated. (B) Restriction enzyme analysis of DNA isolated from CsCl banded virus. The bands critical for distinguishing recombinants with different insertion sites for the expression cassette are marked with asterisks. Note that the loading order is 1, 3, and 2.

To confirm viral genome structures, DNA was isolated and subjected to restriction enzyme analysis (Fig. 2B). All genomes released identical fragments of 2,026 bp after digest with XhoI which corresponds to the size of the haat expression cassette. However, the right-terminal XhoI fragments of 7,491 bp (OAdV1-haat), 2,980 bp (OAdV3-haat) and 915 bp (OAdV2-haat) differed as expected. Similarly, all genomes release identical SmaI fragments of 1,850 bp resulting from cleavage within the expression cassette but were distinguishable by the right terminal SmaI fragments of 7,444 (OAdV1-haat), 2,933 (OAdV3-haat) and 1,204 bp (OAdV2-haat). Thus, the recombinant viruses showed the expected genomic structures.

Next, recombinant viruses were compared with respect to their ability to mediate expression of the transgene (haat) in the rat fibroblast cell line 208F. Infection at multiplicities of infection (MOI) of 5 to 500 vp/cell resulted in hAAT concentrations of between 40 and 12,000 ng per ml and 24 h in the cell culture supernatant, which corresponded to a maximum production of about 60 pg of protein per cell within 24 h. The highest expression was observed with OAdV1-haat, whereas infection with OAdV2-haat and OAdV3-haat resulted in lower (25 to 35%) hAAT concentrations (Fig. 3). Since the viruses were identical with respect to their ability to infect cells, differences in transgene expression were most likely due to the location of the expression cassette within the viral genome.

FIG. 3.

haat gene expression after infection with recombinant OAdV. Rat 208F cells were infected overnight with OAdV1-haat (white bars), OAdV2-haat (light grey bars), or OAdV3-haat (black bars) at the multiplicity of infection (moi) indicated. Medium was replaced 24 h postinfection and hAAT concentrations in the cell culture supernatant were determined by ELISA 24 h later.

Recombinant OAdV with site 1 inserts can be unstable.

Since best transgene expression was obtained with OAdV1-haat, we produced a similar recombinant virus, OAdV1-f IX.o2. This virus contained the cDNA of human coagulation factor IX under the control of the RSV 3′ long terminal repeat in a right to left orientation (o2) within site 1. The virus grew to a titer of 3.6 × 10¹¹ vp/ml at passage 4. However, cleavage of the viral DNA with XhoI did not release the expected fragment of 2,103 bp characteristic of the expression cassette, but instead a smaller fragment of about 1,600 bp was observed. Analysis of DNA from a second expanded clone of this virus showed a band of about 700 bp (data not shown). This suggested that rearrangements within the vector genome had occurred resulting in the loss of part of the transgene.

To verify whether this viral genome had undergone rearrangements during amplification and that the changes were not simply due to the unexpected presence and rescue of defective plasmids, we analyzed each passage of the second clone by PCR with primers flanking insertion site 1 (Fig. 4). Whereas the band at 2,280 bp specific for the complete expression cassette was the only PCR product detected in passages A0 and A1, a second band of about 700 bp appeared from passage A2 onwards and was dominant in passages A4 and A5. Cloning and sequencing of the PCR products revealed that most of the RSV promoter as well as the 5′-terminal part of the factor IX cDNA had been deleted in both viral clones with truncated transgenes, whereas the bovine growth hormone polyadenylation signal was intact. Since vector instability was also observed for insertion of the genes for green fluorescent protein and alkaline phosphatase into site 1, we conclude that, despite the obvious stability of OAdV1-haat, certain vectors with genes inserted at site 1 might be sensitive to rearrangements.

FIG. 4.

Instability of OAdV1-fIX.o2 virus was passaged by repeated infection of HVO156 cells with 10% of the freeze/thaw lysate from the preceding passage. When cytopathic effect was complete, cell culture supernatant (2 μl) was submitted to proteinase K digest and subsequent PCR analysis. A0 is the supernatant of cells transfected with recombinant OAdV1-fIX.o2 plasmid, and A5 is the passage frequently used for preparation of CsCl banded virus. The positions of the expected 2,280-bp fragment and the unexpected approximately 700-bp product are indicated by arrows.

Rescue and characterization of OAdV with site 3 insertions.

Because of site 1 instability and because pOAdV2-cos based recombinants were more difficult to rescue from producer cells, we concentrated on recombinants with site 3 insertions. We generated viruses OAdV3-fIX, OAdV3-luc, and OAdV3-core, which contain RSV promoter-driven cDNAs of human coagulation factor IX, firefly luciferase, and hepatitis C virus core antigen, respectively, in either orientation. Restriction enzyme analysis of viral DNAs from passage A4 confirmed the integrity of viral genomes with respect to the presence and orientation of the transgenes (Fig. 5). Cleavage with XhoI, which released the transgene expression cassettes, resulted in bands of the expected sizes and no shorter fragments were visible. This indicated that viruses with truncated or deleted expression cassettes were not present in significant amounts. Additionally, no low-level contamination with rearranged virus was detected in any of four independent clones of OAdV3-fIX and OAdV3-luc containing the transgene in either orientation: PCR with DNA from passage A5 with the primer pair flanking site 3 produced only the expected products of 2,480 and 2,839 bp, respectively, whereas shorter PCR products were not detectable. The 277-bp fragment characteristic for vectors that had completely lost the transgene expression cassette was neither detected following PCR with DNA from any of the recombinants (data not shown). We therefore conclude that, in contrast to site 1, site 3 insertions are more stable with respect to maintenance of the transgene.

FIG. 5.

Genome structure of recombinant OAdVs with site 3 insertions. Viruses OAdV3-fIX (A), OAdV3-luc (B), and OAdV3-core (C) with transgenes inserted in either orientation were purified at passage A5 by double banding on CsCl gradients. DNA was purified and analyzed by restriction enzyme digestion. XhoI released the right-hand 2,927-bp fragment as well as fragments characteristic for the insert size of 2,160 bp (OAdV3-fIX), 2,520 bp (luciferase) and 1,352 bp (core). Bands indicating the insert orientation within the viral genome are indicated by asterisks.

Next, we confirmed the reproducibility of virus rescue in multiple transfection experiments. A series of 12 wells of HVO156 cells seeded in 24-well plates were transfected with pOAdV3-core.o1 and recombinant virus was rescued from five wells. Viruses were amplified until A4, and viral DNA was submitted to PCR with primers flanking site 3. In all five clones, only the expected PCR product of 1,672 bp was detected, which was identical to that obtained from the transfected plasmid (Fig. 6A). To exclude rearrangements elsewhere in the vector, a similar experiment was performed in which four independent clones of OAdV3-luc.o1 were rescued, amplified up to passage A5 and purified by banding on CsCl gradients. Again, PCR with primers flanking site 3 only produced the product of the expected size (data not shown). Restriction enzyme analysis of DNA isolated from these viruses showed the expected cleavage pattern for all four clones (Fig. 6B) proving that no other rearrangement had occurred within the vector genome. In addition, infection of 208F cells with these clones and determination of luciferase activity revealed that all four viral clones generated comparable levels of reporter gene activity (data not shown). These data and data from the previous section indicate that site 3 vectors propagate stably at least up to passage A4, while unstable vectors can already be identified at passage A2. In addition, rescue of recombinant OAdV with site 3 insertions was reproducibly achieved for a variety of transgenes.

FIG. 6.

Reproducible rescue and stability of OAdV site 3 recombinants. (A) Five clones of OAdV3-core.o1 that were rescued from independent transfection experiments were amplified up to passage A5 and viral DNA was submitted to PCR with the primer pair flanking site 3. Plasmid DNA (pOAdV3-core.o1) was used as a positive control. n.i., supernatant from uninfected HVO156 cells. (B) CsCl banded virus stock was prepared from four clones of OAdV3-luc.o1 that were rescued in four independent transfection experiments. Virus DNA was isolated and analyzed by restriction enzyme digest.

Role of transgene orientation in expression.

It was also of interest to determine whether transgene orientation would influence the level of expression from OAdV3-cos-based recombinants. Several human and rodent cell lines were infected with OAdV3-fIX.o1 or OAdV3-fIX.o2 at a multiplicity of infection of 50 or 200 vp per cell, respectively. As shown in Fig. 7A, the susceptibility of the cell lines to infection with OAdV differed, which is in agreement with an earlier report (15). However, no significant difference in factor IX gene expression was observed between viruses containing the expression cassette in a left to right (o1) or right to left (o2) orientation. A similar result was obtained with recombinant viruses that express the luciferase gene (Fig. 7B), although expression of the luciferase gene was about 30% higher with OAdV3-luc.o2 than with the luc.o1 virus. However, this small difference might be due to inaccuracies in determining in the particle number between the preparations. In addition, infection of 208F cells with OAdV3-core resulted in signals that were independent of the transgene orientation when core antigen was detected by a Western blot (Fig. 7C). We therefore conclude that the orientation of the transgene within the viral genome is not of major relevance for in vitro gene expression mediated by OAdV vectors with site 3 inserts.

FIG. 7.

Insert orientation does not influence transgene expression. (A) Several cell lines were infected with OAdV3-fIX.o1 or OAdV3-fIX.o2 at a multiplicity of infection of 20 (light grey bars) or 100 (black bars) vp. At 24 h postinfection the medium was replaced, and human Factor IX concentrations in the cell culture supernatants were determined by ELISA after an additional 24 h. The mean values of triplicate experiments are shown. (B) Cells were infected with OAdV3-luc.o1 or OAdV3-luc.o2 as in A. At 24 h postinfection cells were lysed, and luciferase activity was determined. The mean values of four independent experiments are shown. (C) 208F cells were infected with OAdV3-core.o1 or OAdV3-core.o2 at the multiplicity of infection (moi) indicated. Cell protein (30 μg) harvested 40 h postinfection was analyzed by Western blotting with an antibody specific for hepatitis C virus core protein. n.i., noninfected cells.

OAdV vector with a site 3 insert mediates therapeutic levels of human Factor IX in mice.

Finally, to confirm that OAdV site 3 recombinants are functional in vivo, BALB/c mice were injected intravenously with OAdV3-fIX.o1 or -o2, and concentrations of human Factor IX in plasma were determined by ELISA. All mice produced relatively high levels of human Factor IX. However, the average Factor IX concentration for mice injected with OAdV3-fIX.o1 (mice 1 through 3) was 3,071 ± 605.2 ng/ml, whereas injection of OAdV3-fIX.o2 (mice 4 through 6) resulted in a somewhat lower average production of 2,014 ± 340.5 ng/ml of protein (Fig. 8). Thus, OAdV recombinants with a site 3 insertion in either orientation are capable of mediating high-level transgene expression in vivo.

FIG. 8.

OAdV site 3-mediated factor IX gene expression in vivo. BALB/c mice were infected intravenously with 2 × 10¹⁰ vp of OAdV3-fIX.o1 (white bars) or OAdV3-fIX.o2 (grey bars). Concentrations of human Factor IX in plasma were determined by ELISA at 48 h postinfection. Each bar refers to an individual animal.

DISCUSSION

Vectors derived from nonhuman adenoviruses have been developed to circumvent the problem of preexisting immunity to human adenoviruses and because they may use different receptors for uptake that could extend the range of potential target tissues. OAdV-derived vectors have been shown to grow to high titers, to efficiently transfer genetic material in vitro and in vivo, and to be safe with respect to immortalization and replication profiles (3, 21). These as well as other attractive properties of OAdV vectors justify their further development. In this study we have developed a novel system that facilitates the construction of recombinant OAdV genomes in which the transgene is inserted into one of three nonessential sites. We demonstrated that the rescue of such genomes as recombinant viruses occurred with suitable efficiency. Vectors with different inserts grew to comparable titers but some vectors with a site 1 insertion were unstable during passaging. However, several vectors with a site 3 transgene were reproducibly rescued, propagated stably and generated high-level expression in vitro and in vivo that was independent of transgene orientation.

The use of cosmids to construct human adenovirus vectors has been described (24). Cosmid vectors were used to insert one of several reporter genes into either of two sites within the human adenovirus genome. Viruses were rescued by recombination between the cosmid DNA and pretreated adenovirus type 5 DNA terminal protein complexes cotransfected into permissive 293 cells. cos-based methods for vector construction were also applied to adenovirus type 5 by other groups (8, 14). With a strategy successfully exploited for the generation of helper dependent human adenovirus (9), we have constructed a set of three cosmid-based plasmids to facilitate insertion of transgenes into OAdV site 1, 2, or 3. To enable its release for virus rescue, the recombinant viral genome was flanked by I-SceI sites. The advantage of this system is that packaging of the ligation products into phage heads and subsequent infection of E. coli are much more efficient than the alternative of transforming bacteria with ligation products. This is because the phage only packages DNA molecules ranging in size from 36 to 45 kbp, thereby ensuring that bacterial colonies containing small, defective viral genomes do not arise. The occurrence of small genomes is a common problem when working with large plasmids. In addition, the super-rare cutting enzyme I-SceI is most unlikely to cut within any transgene sequence, thereby leaving the recombinant viral genome intact for transfection.

The comparison of recombinant OAdVs carrying the insert at different sites in the viral genome revealed that, at least for the haat gene, site 1 was perhaps more suitable with respect to transgene expression than were sites 2 and 3, which were substantially equivalent. However, the stability of vectors with site 1 insertions is a critical point to consider. Site 1 is located between the genes coding for protein VIII and fiber, and late viral transcription across the transgene and correct processing of the RNA must occur for fiber production. The two rearranged clones of OAdV1-fIX that carried an o2 insert had deleted the transgene promoter region and the 5′ end of the Factor IX cDNA. It therefore seems likely that an impact on fiber gene expression by high-level transcription of the site 1 transgene might have resulted in the selection of rearranged viral clones that had silenced transgene expression by promoter deletion, although this explanation does not apply to all transgenes because OAdV1-haat.o2 was clearly stable. On the other hand, other vectors with site 1 insertions in orientation 1 comprising the immediate-early promoter of human cytomegalovirus linked to the gfp or human alkaline phosphatase gene underwent transgene rearrangements (18), while OAdV220, a virus in which the RSV promoter was linked to the purine nucleoside phosphorylase (PNP) gene in orientation 1, did not (37).

As the polyadenylation signals differed between all three of these transgenes, we considered whether more efficient termination of late viral transcription at some transgene polyadenylation sites might have contributed to reduced fiber production and selection against such viruses. However, nucleotide sequencing of PCR products derived from deleted human cytomegalovirus promoter/green fluorescent protein viruses showed that two of three genomes retained the simian virus 40 polyadenylation signal of the transgene, while in all three genomes a large part or all of the human cytomegalovirus promoter had been deleted, so this hypothesis seems unlikely to be correct. The presence of cryptic splice sites in the transgene could also prevent the correct splicing of the fiber transcripts to the OAdV tripartite leader, thus reducing or preventing fiber protein production such that deleted viruses in the population may be selected. The stability of OAdV220 (37), which carries a bacterial gene sequence that is not spliced as well as the stability of a vector carrying nonexpressed human X-chromosome sequences in site 1 (P. Löser and M. Hillgenberg, unpublished data), would be consistent with this notion. This hypothesis also agrees with our finding that virus rescue is much more efficient for vectors with site 3 insertions than those with site 1 insertions. Lastly, it may be that certain transgene sequences are simply incompatible with parts of the OAdV genome and prone to intragenomic recombination events. In summary, it seems likely that the genetic instability of some vectors carrying insertions in site 1 could result from a combination of several factors that may involve both viral and transgene sequences.

In contrast to site 1, insertion site 3 of OAdV is situated between the P1 promoter for the proposed E4 region 13 and the termination point for rightward transcripts from the RH transcription unit (39). Insertion of an expression cassette into this site, particularly in orientation 2, was considered to be unlikely to interfere with E4 or RH gene expression even if temporally unregulated transcription occurred from the RSV long terminal repeat promoter. This situation is clearly different from that for site 1. The growth of viruses with site 3 insertions was unimpaired relative to the control vector lacking an insert and several viruses described in this work propagated stably. In addition, the level of gene expression was orientation independent. This may be significant because an effect of orientation was demonstrated for adenovirus type 5 vectors when the transgene was located in the E1 region, while expression became orientation independent when the transgene was inserted near the 3′ long terminal repeat (32).

The orientation dependence of gene expression was also abolished when the transgene expression cassette was flanked by insulator sequences (36). Our results provide evidence that the orientation of the transgene in OAdV site 3 was not critical for the level of gene expression in nonovine cells even though the E4 P1 and P2 promoter regions are adjacent to site 3 (13), and in orientation 1 the transgene promoter is adjacent to P1. However, at this stage we cannot be sure whether the specificity of an exogenous promoter would be affected as only the broadly active RSV promoter was used in this work. Moreover, transgene orientation at some sites may also affect the efficiency of virus rescue because we repeatedly rescued OAdV1-haat and OAdV2-haat with the insert in one orientation. However, the effect might also be sequence dependent since other recombinants with site 1 insertions were rescued in the opposite orientation (12, 37, 41). In contrast, rescue in site 3 was both reproducible and efficient in either orientation. We therefore infer that there was no interference with any viral function at this site.

A challenge now remaining is to further improve the packaging capacity of the vectors. OAdV can accommodate an insertion of about 4.3 kb of foreign DNA without deletion but sequences from the SalI sequence in site 2 to the termination point for RH transcripts (≈2 kb) are dispensable for virus growth in vitro and their replacement with foreign DNA could increase the packaging capacity to ≈6.3 kb (41). However, in this and other work only one site 2 virus (OAdV2-haat.o2) that incorporated a transgene together with this deletion was successfully rescued, although this was achieved on several occasions. This may be because the deletion has disrupted viral signals required for expression of the remaining RH1 gene. Other insertion sites, e.g., between the LH3 and IVa₂ or between fiber and E43 genes could also be considered as these would also lie between transcription units (13). Insertion adjacent to the ITRs does not appear to be an option as early promoters for the LH and RH genes lie within that region (4, 13).

Taken together, our results provide an efficient and reproducible method for the construction of recombinant ovine atadenoviruses. Our data suggest that site 3 is currently the most suitable site for transgene insertion into this novel vector. Future work will concentrate on the identification of additional transgene insertion sites and characterization of yet unknown viral functions. Replacement and transcomplementation of certain viral genes could further extend OAdV vector development.