Episomal Segregation of the Adenovirus Enhancer Sequence by Conditional Genome Rearrangement Abrogates Late Viral Gene Expression

Abstract

We have constructed a recombinant adenovirus gene delivery system that is capable of undergoing growth phase-dependent site-specific recombination. When propagated in 293 producer cells, the vector retains its linear double-stranded form and can be propagated to high titer and purified by conventional procedures. Upon introduction into target cells, the viral chromosome undergoes cyclization to generate an autonomously replicating circular episome and a detached linear fragment. The viral enhancer and reporter gene segregate with the circular episome, which contains no adenovirus open reading frames. The effect of rearrangement of adenovirus gene expression was assessed by quantitative reverse transcription-PCR measurement of the abundance of transcripts encoding the tripartite leader sequence (TPL) of the major late promoter. Whereas nonrearranging viruses produced approximately 10⁴ TPL transcripts per 10⁶ infecting genomes in the HepG2 liver cell line, no transcripts were detectable in the same cells infected with comparable levels of circularizing vector. Because no helper virus is required to propagate these vectors, the problems of recombination with and contamination by helper virus are eliminated. We also present an efficient and reliable method for generating recombinant adenoviruses.

Adenovirus vectors (AdV) have a recognized potential for gene delivery, founded on their broad host range, robust growth in culture, and capacity to infect mitotically quiescent cells (13, 45). AdV can be propagated in a helper cell line, 293, a human embryonic kidney cell line transformed by adenovirus type 5 (14). 293 cells express the viral E1 gene products (E1a and E1b) that are the master regulatory proteins for subsequent viral gene expression. E1-deleted viruses can propagate in 293 cells but not in other cells. Although it would be expected that E1-deleted viruses lack the machinery to express viral genes, several studies have demonstrated that cellular E1-like components can stimulate viral gene expression (24, 39, 43). The expression of these viral genes results in the relatively rapid elimination of transduced cells in vivo as a result of a cytotoxic T-cell responses (46–48). Thus, attention has focused on eliminating the remaining vestiges of viral expression. Viral genes that have been deleted for this purpose include the gene for E4 proteins (4, 29, 51), DNA-binding protein (9, 12), DNA polymerase (3), and the preterminal protein (42). The most aggressive approach has been the creation of helper virus-dependent vectors that lack all viral genes (18, 28, 32, 34, 40). These vectors have high capacity, evoke reduced cellular immune responses, and show prolonged expression in vivo (37). However to deploy these viruses on the scale required for human clinical application presents major challenges because a CsCl gradient is needed to remove the helper virus.

We have developed a different approach to the problem of creating a gene expression unit devoid of vector-derived open reading frames. In this strategy, the vector is propagated as a linear, helper-independent virus in the conventional way on 293 cells. Upon introduction into the host cell, the viral chromosome undergoes a site-specific rearrangement to form a circular episome expressing the gene of interest. The viral enhancer is excised as a result of the rearrangement, leaving a deleted linear DNA without enhancer function.

The adenoviral enhancer and packaging signals are found as a mosaic structure at the left end of the viral genome and are essential for viral gene expression and encapsidation of the viral genome (19, 20). In principle, viral gene expression in target cells should be decreased by disconnecting this sole enhancer element from the genome. This approach should also allow more sophisticated rearrangements to be incorporated, in which the linear remnant is subjected to additional genoclastic remodeling. Like the vector systems recently described by Berk, Perricaudet, and coworkers, the rearrangement is carried out by Cre recombinase acting on loxP sites (31, 44). However the present system does not rely on a helper virus expressing Cre and, perhaps because of this, is not subject to the unexplained toxicity seen in the two-virus system (44).

We also introduce here a convenient and effective strategy for producing recombinant adenoviruses that is based on the use of intron endonucleases and bacteriophage lambda in vitro packaging. Due to its simplicity and reliability, this approach has substantial advantages over traditional methods, which rely on homologous recombination between plasmids cotransfected into 293 cells, and over various recently described strategies (7, 18, 27, 35, 36, 38) that aim to circumvent the rate-limiting in vivo recombination step. In this report we present a two-cosmid system for generating AdV, based on λ phage packaging and the use of highly specific intron-encoded endonucleases, I-CeuI and PI-PspI. Inclusion of intron endonuclease cleavage sites at the ends of the viral DNA allows the viral genome to be liberated from its cosmid context, resulting in improved (10- to 100-fold) generation of recombinant viruses compared to a related cosmid approach (10).

MATERIALS AND METHODS

Adenovirus type 2 (Ad2) DNA was purchased from Sigma. Restriction enzymes, intron endonucleases I-CeuI and PI-PspI, and Moloney murine leukemia virus reverse transcriptase (RT) were purchased from New England Biolabs, Stratagene, or Promega. DNA fragments amplified by PCR were confirmed by DNA sequencing. Nucleotide (nt) positions refer to the wild-type Ad2 sequence in GenBank (J019017).

Two-cosmid system.

The EcoRI-BsaI fragment that spans the ampicillin resistance gene in pBR322 was deleted and replaced by a synthetic adapter, and the bacteriophage λ cos site was inserted between the unique StyI and BsmI sites. A PCR-amplified Ad2 fragment containing the left-end inverted terminal repeat (ITR) enhancer elements, and the encapsidation signal (nt 1 to 376) was created and inserted into the adapter (Fig. 1) to yield the tetracycline-resistant left-end plasmid pLEP. The right end of Ad2 from the AflII site to the right end (nt 3527 to 35937) was assembled into an ampicillin-resistant cosmid vector, pACKrr3, by multiple steps of PCR amplification and fragment interchange. The resultant cosmid was termed pREP7. To expand vector capacity, two deletions were incorporated into the pREP7 cosmid, an E3 gene deletion (nt 27901 to 30841, 2,840 bp; cosmid pREP8) and a 1.3-kb deletion (nt 34121 to 35469) in the E4 region of the Ad2 region (pREP12). Complete vector sequences are available from the authors.

FIG. 1.

Schematic structure of the AdV system. (A) Diagram of the pLEP cosmid polylinker region and its position relative to the adenoviral left ITR. The adenovirus enhancer (Enh)/packaging sequence (ψ) is boxed. (B) Generation of a single cosmid encoding the AdV genome by the direct ligation of two smaller plasmids. A gene expression unit (CMV-GFP) was inserted into the pLEP cosmid at the polylinker region. pLEP and pREP cosmids were digested with an intron endonuclease (PI-PspI), ligated, and packaged in vitro to generate pAd2CMVGFP. This DNA was then digested with another intron endonuclease (I-CeuI) to expose the ITRs at both ends of the viral genome (L., left; R., right). Finally, cosmid digestion mixtures were transfected into 293 cells. Plaques generated by recombinant viruses are detected in 7 to 10 days. MCS, multiple cloning site; WT, wild type; TP, terminal protein.

Preparation of recombinant AdV.

One microgram of PI-PspI-digested pLEP plasmid was dephosphorylated and ligated to 1 μg of a PI-PspI-cleaved pREP plasmid in a 20-μl reaction for 2 h at room temperature, and 4 μl of the ligated DNA was packaged in a λ phage extract (MaxPlax lambda packaging extracts; Epicentre Technologies). One-tenth of the packaged material was used to transduce Escherichia coli DH5α or DK1 cells. Transductants containing pLEP fused to pREP were selected on agar containing ampicillin (25 μg/ml) and tetracycline (12.5 μg/ml) (Amp/Tet). Colonies were selected and DNA was isolated (Qiagen). DNA was used either for restriction analysis or for transfection of 293 cells as described below.

Cultured cells and primary human hepatocytes.

293 (human embryonic kidney) cells were obtained from Microbix Biosystems (Ontario, Canada). HepG2 (human hepatocellular carcinoma), HeLa (human cervical epitheloid carcinoma), A-431 (human epidermoid carcinoma), and HT29 (human colon adenocarcinoma) cells were all obtained from the American Type Culture Collection. All cells were cultured in complete Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 2 mM glutamine, and penicillin-streptomycin (Gibco-BRL) and maintained at 37°C and 5% CO₂ atmosphere in an incubator. Primary human hepatocytes, generously provided by Albert Edge (Diacrin, Inc., Charlestown, Mass.), were isolated and cultured as described (15).

Generation of AdV by transfecting linearized cosmid DNA.

293 cells were cultured in 10-cm dishes in DMEM-FBS. Cells were grown to ∼50% confluence on the day of transfection. Ten micrograms of cosmid DNA was digested with I-CeuI in a volume of 50 μl. The reaction mixture was transfected into 293 cells by calcium phosphate precipitation (13) without purification. After transfection, cells were cultured and examined daily for the appearance of cytopathic effects. Virus propagation and purification, plaque assay, and viral DNA isolation were performed by established protocols (13).

To compare the efficiency of AdV generation by homologous recombination, 20 μg of pREP7 was cotransfected into 293 cells with 10 μg of a plasmid encoding the left end of the adenoviral genome and a green fluorescent protein (GFP) reporter gene (pLITREF1αGFP). pLITREF1αGFP contains the Ad2 left end (nt 1 to 376), an EF1α promoter-GFP expression unit, and Ad2 sequence (from 3525 to 8120) that overlaps the same sequence in pREP7. This overlap fragment served as the region for homologous recombination. Each cotransfection was performed in duplicate.

Liver-specific promoter.

Human hepatic control regions 1 and 2 (HCR1 and -2) of the ApoE/C gene locus (2, 8) were amplified by PCR using 293 cell genomic DNA as the template. The following primers were used to amplify both HCR1 and HCR2 fragments: HCRtop (5′GCGGAATTCGGCTTGGTGACTTAGAGAACAGAG3′) and HCRbot (5′GCGGGATCCTTGAACCCGGACCCTCTCACACTA3′). The amplified PCR fragments (∼0.39 kb) were cloned into pUC19. The HCR1 and HCR2 sequences were confirmed by dideoxy DNA sequencing. The two fragments were assembled in a head-to-tail orientation, fused with a synthetic basal TATA element (B. Seed, unpublished), and cloned in a parental pLEP vector containing a GFP reporter gene. The resultant plasmid was named pLEPHCR12GFP.

Self-resolving AdV.

The self-resolving AdV was generated using the two-cosmid system. A plasmid, pAd2237 (Eunchung Park, personal communication), containing loxP sites, a GFP reporter gene (16), EBNA-1/OriP, and sequences encoding Cre recombinase was used to incorporate these elements into the pLEP plasmid to yield pLEP1BHCRGFP. To generate the self-resolving AdV, 1 μg of pLEP1BHCRGFP DNA and 1 μg of pREP8 (ΔE1ΔE3) were digested with PI-PspI, ligated, packaged, and assembled into one cosmid named pAdVEBV as described above.

Assessment of GFP expression and titering.

GFP expression was assessed by fluorescence microscopy (Olympus IX70) or microtiter plate reader (PerSeptive Biosystem CytoFluor II). For the latter, 5 × 10⁴ 293 cells were seeded into each well of a 96-well plate. The cells were infected with serially diluted virus stock and cultured for 48 h, and GFP intensity was measured. Titers were estimated by interpolation into the linear region of a standard curve prepared from virus of known (plaque-derived) titer. All titers were determined in triplicate.

Virus infection, DNA, RNA isolation, and blot analysis.

HepG2 and HeLa cells were seeded in 35-mm dishes and cultured to approximately 80% confluence in DMEM-FBS. Cells were infected with the desired multiplicity of virus in a volume of 1 ml at 37°C for 2 h. At the end of the incubation, cells were washed with phosphate-buffered saline (PBS) twice and cultured in 2 ml of medium. Cells were collected in parallel at desired points for low-molecular-weight DNA and RNA extraction. Low-molecular-weight DNA was extracted from the infected cells as described (22), and total RNA was prepared using RNAzol solution (Tel-Test, Inc.).

For Southern blot analysis, 5 μg of DNA was digested with BglII, fractionated on a 1% agarose gel, and subjected to blot analysis using a labeled EBNA-1 gene fragment as the probe.

RT-PCR, PCR, and quantitative PCR.

Four micrograms of total RNA was reverse transcribed into cDNA using Moloney murine leukemia virus RT by a standard protocol (Promega). The cDNA from each sample (1 μl) was used in subsequent PCRs. PCR primers were designed to amplify the tripartite leader sequence of the adenovirus late genes: TPL1 (5′ ACT CTC TTC CGC ATC GCT GT 3′) and TPL2 (5′ CTT GCG ACT GTG ACT GGT TAG 3′). For detection of the AdV genome in the Hirt DNA samples, 1 μg of DNA was employed in the PCR amplification using the following primers, which are specific for the adenovirus DNA in the fiber gene: Fiber1 (5′ CCG CAC CCA CTA TCT TCA TA 3′) and Fiber2 (5′ GGT GTC CAA AGG TTC GGA GA 3′). PCRs were performed as 95°C for 30 s, 54°C for 30 s, and 72°C for 30 s for 30 cycles. All amplified products were analyzed on a 2% agarose gel.

For quantitative PCR, a molecular beacon-based universal amplification and detection system was used (Intergen). A common leading sequence (Z sequence; 5′ ACT GAA CCT GAC CGT ACA 3′) was added to the TPL1 and Fiber1 primers. The TPL2 and Fiber2 primers, described above, were used in the quantitative PCRs. cDNA (1 μl) and 1 μg of Hirt DNA from each sample were used in the assay. The PCRs were carried out in a 96-well spectrofluorometric thermal cycler (Applied Biosystems Prism 7700). The number of template molecules in the PCR was calculated from the standard curve using linearized plasmid as the templates.

RESULTS

Development of a two-cosmid system for efficient construction of recombinant AdV.

To simplify and facilitate the generation of recombinant AdV, we established a system to assemble the desired AdV genome in a single plasmid by ligation (Fig. 1). The system consists of two component vectors, a left-end plasmid, pLEP, and a right-end plasmid, pREP. The left-end adenovirus sequences (nt 1 to 376) in pLEP include the viral inverted terminal repeat, the cis-acting packaging sequences, and the viral enhancer. The adenovirus sequences are followed by the gene expression unit intended for delivery and an intron endonuclease (PI-PspI) cleavage site. The right-end plasmid contains a PI-PspI site followed by the Ad2 genome from the end of the E1 locus rightward (nt 3527 to 35937).

pLEP is a small tractable vector for cloning, whereas pREP is much larger and contains less frequently manipulated genes. Both pLEP and pREP contain a bacteriophage λ cos site, oriented to generate a single cosmid of appropriate length for in vitro packaging following ligation of the two plasmids at the PI-PspI cleavage site. pLEP is tetracycline resistant (Tet^r), and pREP is ampicillin resistant (Amp^r), allowing the recombinants to be selectively isolated by coselection for both markers. In the resulting assembled cosmid, the adenoviral sequences are closely flanked by cleavage sites for the intron endonuclease I-CeuI. Digestion with I-CeuI liberates the entire recombinant AdV genome from the parent cosmid (Fig. 1).

Three classes of pREP have been constructed to allow the preparation of AdV bearing E1 (pREP7), E1 and E3 (pREP8), or E1, E3, and E4 (pREP12) deletions. pREP7 contains nt 3527 to 35937 of the Ad2 genome; pREP8 carries an additional deletion in the E3 region (Δ nt 27901 to 30841). pREP12 has deleted open reading frames 1 to 4 of the E4 region (Δ nt 34121 to 35469, 1,348 bp). AdV generated with these cosmids should be able to accommodate inserts of 5, 8, and 10 kb, respectively.

An example of the construction of an AdV carrying a cytomegalovirus major immediate-early promoter–GFP expression unit (CMV-GFP) is outlined in Fig. 1. pLEPCMVGFP (Tet^r) was digested with PI-PspI and ligated to pREP7 (ΔE1, Amp^r) digested with the same enzyme. The ligation mixture was packaged with λ phage extracts, and a fraction of the packaged phage was used to infect a recombination-deficient E. coli host, with selection for the assembled plasmid on Amp/Tet plates. Figure 2A shows typical results for the BglII digestion pattern of a pLEP3CMVGFP/pREP7 hybrid cosmid, pAd2-7CMVGFP. Because of the size minimum (∼40 kbp) for λ phage in vitro packaging and the double antibiotic selection, most of the colonies growing on Amp/Tet plates are the desired hybrid cosmids, and undesired rearrangements are rarely seen. In the present example, all four pAd2-7CMVGFP clones exhibited the digestion pattern predicted from the inferred sequence. The entire recombinant AdV genome was then released from the cosmid by I-CeuI digestion (Fig. 2B). I-CeuI digestion leaves 10 nt to the left of the left ITR and 8 nt to the right of the right ITR. Short flanking sequences have been reported to be eliminated during replication of recombinant viruses after transfecting the DNA into 293 cells (17).

FIG. 2.

Confirmation of the structure of recombinant adenoviruses. (A) Restriction analysis of cosmids carrying the full-length AdV DNA, showing uniform generation of the desired vector DNA. DNA samples (2 μg) from four pAd2-7CMVGFP colonies were digested with BglII, resolved on a 1% agarose gel, and stained with ethidium bromide. The predicted sizes of the DNA fragments are 13,261, 7,684, 5,228, 5,088, 2,284, 1,757, 1,549, 1,270, 351, and 275 bp. The 5,228- and 5,088-bp fragments appear as a doublet, and the 351- and 275-bp fragments are too small to be seen on the gel. (B) Release of the recombinant DNA from cosmids by I-CeuI digestion. pAd2-7CMV DNA (2 μg) from two clones was digested with I-CeuI. Arrows indicated the position of the released recombinant AdV DNA and the vector fragments of approximately 35 and 5 kb.

The digestion reaction can be transfected into 293 cells without purification. At day 6 posttransfection, 5 to 30 viral plaques per 10-cm dish per 10 μg of DNA are usually apparent, which compares favorably with the 30 to 50 plaques found for 293 cells transfected with purified wild-type Ad2 DNA. To compare the efficiency of recombinant virus production, similar viruses were also generated by homologous recombination; 10 μg of pLITREF1αGFP and 20 μg of pREP7 were cotransfected into 293 cells. Initial plaques took longer to appear (14 days posttransfection) and were less abundant (0 to 3 plaques per plate).

Data in the literature suggest that exposed ITR ends favor efficient virus production (17). To assess the importance of this effect, we constructed an AdV cosmid, pIAdEF1αGFPB, in which the AdV ITRs were flanked with a different restriction site at each end. pIAdEF1αGFPB DNA was digested with BsaBI to expose the right ITR, I-CeuI to expose the left ITR, or both together to expose both ends. Digested cosmid DNA samples were transfected into 293 cells, and plaques were allowed to develop. Ten days after transfection, the viruses were harvested, and viral titers were determined. The average titer for the viral stocks (Fig. 3) was 1.3 × 10⁴ PFU/ml from transfection with undigested DNA; 2.4 × 10⁵ PFU/ml from BsaBI-linearized DNA (free right ITR); 1.1 × 10⁵ PFU/ml from I-CeuI-linearized DNA (free left ITR); and 2.7 × 10⁶ PFU/ml for the BsaBI/I-CeuI doubly digested DNA (both ITRs free). Thus, liberation of each end results in an approximately 10-fold increase in the efficiency of generating virus (Fig. 3).

FIG. 3.

Exposure of ITRs enhances the efficiency of AdV generation. (A) The appearance of plaques in 293 cells transfected with 10 μg of pIAdGFPB with no ITR exposed (undigested), one ITR exposed (BsaBI or I-CeuI), or both ITRs exposed (BsaBI plus I-CeuI) was determined. Values represent the mean plaque counts per dish (CPEs) and the time required for plaque development in 293 cells from three separate experiments. I, I-CeuI; B, BsaBI. (B) Plaques were allowed to grow over 10 days after transfection. Viruses were harvested, and the titer of each virus stock was determined by a GFP-based semiquantitative titration procedure described in the text. Values represent the mean ± standard error for three independent determinations.

Construction of an AdV capable of self-rearrangement.

One approach to attenuating adenoviral gene expression and improving transgene persistence is the creation of viruses capable of undergoing internal, self-directed rearrangement upon delivery to the target tissue. In principle, this objective can be achieved through the regulated expression of site-specific recombinases in vectors that contain the cis-acting target of recombinase action. To allow such vectors to be created, the recombinase activity must be suppressed during propagation in the packaging cell line. We have explored a number of strategies to achieve this end; one of the more effective to date has been the use of a lineage-specific promoter to control recombinase expression.

An example of this is shown in Fig. 4. The expression of Cre recombinase is controlled by a liver-specific promoter. In 293 cells, this promoter is silent, allowing the viral chromosome to be propagated with minimal rearrangement. Any rearranged viruses that are formed lack packaging signals and so disappear from the pool of propagating vectors. In liver cells, the Cre recombinase is induced by the action of the tissue-specific promoter. The resulting Cre-induced recombination excises a circular episome and redirects the transcriptional output of the liver-specific promoter so that it directs the synthesis of the transgene of interest. The remaining linear fragment consists of an adenoviral genome lacking the enhancer and packaging signals and a Cre expression unit devoid of promoter sequences.

FIG. 4.

Linear AdV that resolves into a circular episome. The elements involved in the self-directed rearrangement of the vector are shown schematically in pLEP1BHCRGFP/EBV and in the corresponding AdV. Starting from the left ITR (L.ITR), the elements are shown in the following sequence: L.ITR, 147 bp; first 34-bp loxP site; 185-bp enhancer/packaging signal; 64-bp splicing acceptor (SA) from EF1α gene first intron; 720-bp GFP cDNA; 230-bp simian virus 40 (SV40) poly(A) site; 1.7-kb TK-EBNA-1/OriP; 970-bp HCR12 promoter; 1-kb EF1α gene first intron containing splicing donor (SD) and acceptor (SA) sites with the second loxP site inserted at 64 bp upstream of the 3′ end; 1.2-kb Cre gene tagged with AU1 and a nuclear localization signal; ∼120-bp poly(A) signal and PI-PspI site. After infection of liver cells, the HCR12 promoter drives the expression of Cre, which results in cleavage of the two loxP sites. This results in circularization of the fragment containing the EBV replicon. The excision severs the connection between the enhancer/packaging signals and the remainder of the AdV genome. The Cre gene becomes promoterless and is left on the AdV genome fragment. After excision, the HCR12 promoter drives the expression of the GFP reporter gene. The EBV replicon maintains the excised circle as an episome in host cells.

In the form discussed here, one loxP site is located at nt 147 of the Ad2 genome, between the left ITR and the enhancer-packaging sequences, and the second loxP site is placed inside an intron a few bases upstream of the splice acceptor sequence. Hence, the loxP site does not appear in the resulting mature transcript. The Cre coding sequence that remains on the right-end linear fragment after rearrangement lies downstream from a splice acceptor that lacks a splice donor or upstream promoter sequences. This effectively terminates the expression of Cre following excision.

Prior to recombination, the Cre recombinase gene is under the control of a synthetic promoter (HCR12), consisting of hepatic locus control elements from the human ApoE/C locus fused to the first intron of the human EF1α gene. After cyclization, the HCR12 promoter lies upstream of the transgene (in this case, GFP), and the distal segment of the intron (beyond the loxP site) contains the adenoviral enhancer. To facilitate manipulation of the plasmids in E. coli, the human immunoglobulin G1 (IgG1) hinge-CH2 intron (118 bp) was inserted in the Cre coding sequence at nt 237, suppressing Cre expression in bacteria. The circularized episome contains the latent origin of replication (OriP) and trans-acting DNA replication protein (EBNA-1) of Epstein-Barr virus (EBV) and hence is capable of autonomous replication in synchrony with the host mitotic cycle (50).

Using the system described above, the pLEP plasmid containing the self-resolving components, pLEP1BHCR12, was ligated with pREP8 (ΔE1ΔE3) to create pAdVHCRGFP/EBV. The latter was digested with I-CeuI and transfected into 293 cells. Appearance of plaques from AdVHCRGFP/EBV was retarded (by 8 days) compared to nonrearranging viruses, perhaps as a result of basal expression of the liver-specific promoter in 293 cells. However, high-titered viral stocks of 10¹² nominal (absorbance-determined) particles/ml could be achieved.

Rearrangement in target and nontarget cells.

To test excision efficiency, HepG2 (hepatocellular carcinoma) and HeLa (cervical carcinoma) cells were infected with virus at a multiplicity of infection (MOI) of 1,000 nominal particles/cell. This titer corresponds to approximately 10 PFU per cell. Cells were examined for GFP expression before extraction of DNA for analysis of chromosomal rearrangement. Hirt DNA (5 μg) was digested with BglII and analyzed by DNA blot techniques (Fig. 5). The BglII fragment from the noncircularized AdV is 3,162 bp, generated from the 5′ end of the AdV to the first BglII site in the AdV. The circularized fragment, created from the two loxP sites, has a size of 4,915 bp (Fig. 5A). Densitometry revealed that at 72 h postinfection, 95% or more of the input genomes had undergone circularization in HepG2 cells. In contrast, low but detectable levels of circularized fragment could be visualized in HeLa cells infected at the same time and at the same MOI used for the HepG2 cells (Fig. 5B).

FIG. 5.

Blot analyses of the self-resolving AdV genome, documenting DNA self-rearrangement in HepG2 and HeLa cells. Cells were infected with Ad2HCRGFP/EBV viruses for 2 h at 37°C. Hirt DNA samples were extracted from the cells, and ∼5 μg of Hirt DNA samples were digested with BglII and analyzed by Southern blot techniques using a ³²P-labeled EBNA-1 fragment as the hybridization probe. (A) Schematic representation of the loxP sites and EBNA-1 locations in the AdV genome. The relevant BglII site is also shown. (B) Time course from 0 to 72 h of rearrangement in HepG2 and HeLa cells at an equal MOI of 1,000 particles per cell. (C) DNA blot results obtained from HeLa cells infected at an MOI of 10,000 and HepG2 cells at an MOI of 1,000. The upper bands (4,915 bp) represent the circularized DNA fragments, whereas the lower bands (3,162 bp) represent the noncircularized AdV.

At the time of infection (Fig. 5B), the amount of input viral DNA detected by DNA blot was higher for HepG2 cells than for HeLa cells when similar virus multiplicities were applied (MOI 1,000). This may reflect differences in AdV adsorption or infection efficiency between the two cell types, possibly as a result of the lower levels of coxsackievirus-adenovirus receptor on the HeLa cell surface (X. Wang and B. Seed, unpublished data). To achieve similar viral genome input into HepG2 and HeLa cells, HeLa cells were infected with 10-fold more virus (MOI ∼10,000) than HepG2 cells (MOI 1,000). Episomal DNA samples were extracted and analyzed by blotting. The results (Fig. 5C) indicate that when comparable amounts of viral genome are present in the nucleus, the cyclization rate in both cell types is similar. Because the level of subsequent GFP expression is much higher in HepG2 cells than in HeLa cells (Fig. 6A), it is likely that very small amounts of Cre recombinase suffice to promote rearrangement and that recombinase expression is not limiting for rearrangement in either HepG2 or HeLa cells.

FIG. 6.

GFP expression in liver and nonliver cells infected with the Ad2HCRGFP/EBV viruses. Cells were cultured in 35-mm dishes and infected with the Ad2HCRGFP/EBV virus at the desired MOI. (A) HepG2 cells were infected with 1,000 particles per cell, whereas HeLa cells were infected with 10,000. GFP expression was examined at the indicated time points after infection. Fluorescent cells were photographed using an Olympus SC 35-mm camera mounted on an Olympus IX70 fluorescent microscope, at ×200 magnification, using a filter with peak excitation and emission wavelengths of 450 and 510 nm, respectively. HepG2, HeLa, A431, and HT29 cells (B) and human primary hepatocytes (C) were seeded in 35-mm dishes and infected with the Ad2HCRGFP/EBV virus at an MOI of 10,000 particles per cell. GFP expression was examined at 72 h after infection. The human primary hepatocytes were photographed under bright-field (left) and fluorescent (right) conditions.

GFP expression cannot be detected until rearrangement has taken place, so the measurement of the fraction of GFP-positive cells provides a simple alternate method for assessing the degree of productive rearrangement. Figure 6A shows that GFP expression develops quickly in transduced HepG2 cells but that only a few GFP-positive cells can be detected in Hela cells infected with a 10-fold-higher MOI, conditions that allow circularization to an extent comparable to that seen in HepG2 cells (Fig. 5C). The HCR12 promoter specificity was also tested by infecting two additional nonhepatic cell lines, A431 (human epidermoid carcinoma) and HT29 (human colon adenocarcinoma), with the Ad2HCRGFP/EBV vector. A few cells with a weak GFP signal were detected at 72 h after infection in these cells (Fig. 6B). In contrast, these nonhepatic cells could be infected efficiently with a first-generation AdV, Ad2CMVGFP virus (data not shown), indicating that the low GFP signal was not due to the low infectivity of these cells by AdV. To further assess the utility of the AdV genome rearrangement, primary human hepatocytes were infected with the Ad2HCRGFP/EBV vector. GFP expression was readily detected 72 h after infection (Fig. 6C).

Diminished viral gene expression in rearranged AdV.

After excision, the adenovirus major enhancer/packaging signal segregates with the episomal DNA, yielding a linear fragment containing the remainder of the AdV genome without this important cis element (Fig. 4). To assess the impact of enhancer deletion, quantitative RT-PCR measurement of late viral gene expression was performed. As most late adenoviral gene transcripts have a common ∼200-bp tripartite leader (TPL) sequence (1), the TPL sequence was chosen as a marker of viral gene expression. HepG2 cells were infected with the first-generation vectors Ad2CMVGFP and Ad2HCRGFP or the self-resolving vector Ad2HCRGFP/EBV using increasing MOIs. Total cellular RNA and low-molecular-weight DNA were isolated in parallel. RT-PCR was performed to quantitate the amount of RNA encoding the TPL in the cDNA samples. PCR amplification of a 201-bp fiber gene fragment from the AdV genome was used to detect the amount of viral genome in the DNA samples. A representative result of three experiments is shown in Fig. 7A. TPL sequences were detected, 72 h postinfection, with either 100 or 1,000 viruses infected per cell using both the first-generation adenoviruses (upper panel). In contrast, no TPL signal was detected in the self-resolving Ad2HCRGFP/EBV-infected cells, even at an MOI of 100,000/cell. PCR amplification of the AdV fiber gene revealed comparable levels of AdV genomic DNA in cells infected at comparable MOIs (Fig. 7A, lower panel). The cDNA samples in which the TPL signals were detected were further analyzed by real-time fluorescence PCR. The corresponding genomic DNA samples were also analyzed to determine the number of AdV genomes present in each sample. The results are summarized in Fig. 7B. Approximately 10⁴ TPL per 10⁶ AdV genomes were detected in the Ad2HCRGFP-infected cells, but no detectable TPL was found in the self-resolving Ad2HCRGFP/EBV-infected cells. These results suggest that adenoviral gene expression is dramatically reduced by the separation of the viral enhancer sequences occasioned by the rearrangement of the self-resolving vector.

FIG. 7.

PCR analyses of adenovirus late gene expression in cells infected with the first-generation AdVs or the self-resolving Ad2HCRGFP/EBV. HepG2 cells were cultured in 35-mm dishes and infected with increasing MOIs (0, 10, 100, 1000, 10,000, and 100,000) of AdV. RNA and DNA were isolated in parallel from the cells at 72 h after infection. (A) RT-PCR was performed to detect the TPL sequence (upper panel) for virus late gene expression; and PCR was performed in the DNA samples for detection of the AdV genomes. The specific target sequences are described in detail in Materials and Methods. (B) Summary of quantitative RT-PCR and PCR results. Each determination is the average of three experiments.

DISCUSSION

Viral gene delivery vehicles are attractive vectors for human gene therapy. They transduce foreign genes efficiently and deliver large amounts of genetic material without incurring the high mutation rate that naked DNA undergoes upon transfection (5, 6, 30, 41). Large-scale production of recombinant viruses appears feasible, and clinically acceptable formulations could be deployed therapeutically without investment in a substantial infrastructure for local production and use.

Evading the genetically honed ability of the immune system to recognize and destroy viruses and virus-infected cells poses a substantial challenge. Both the humoral and cellular arms of the immune system participate in viral rejection, the humoral arm through antibody recognition and neutralization of the viral particle (11, 48), and the cellular arm through recognition of intracellular peptides either synthesized de novo by viral genes or borne into the cell during infection (25, 48). This paper reports the development of a vector designed to enhance the persistence of virally delivered genes and evade the cellular immune response by severing the connection between the sole adenoviral enhancer and the sequences encoding potentially antigenic viral proteins. Our data indicate that segregation of the enhancer from the multiple adenoviral genes that utilize the common TPL sequence results in a dramatic reduction in viral gene expression.

Previous studies have demonstrated that excision of the enhancer element could affect late gene transcription either directly (21), by diminishing the expression of other early gene products (19, 20, 33), or through a combination of these factors. Conversely, early viral genes can be upregulated by insertion of the adenovirus enhancer element, even in the absence of E1 function (23). Although the relative potency of the different possible contributions to diminished expression is not known, the magnitude of the observed effect is substantial and supports the utility of regulated rearrangement as a strategy for the creation of scaleable AdV for clinical uses. Because of the intermingling of the adenovirus enhancer and packaging sequences, it has been difficult to assess enhancer function by generating an enhancerless virus without compromising viral growth or viability (20). The self-resolving AdV system provides a tool to analyze the roles of the enhancer in viral gene regulation and virus growth.

The circularization of the vector via the action of Cre recombinase has the additional benefit of placing the therapeutic gene on a self-replicating episome. Vector circularization occurs in a tissue-targeted manner, in this case as a result of the activation of a synthetic liver-specific promoter upstream of Cre. Once circularized, the EBV replicon in the episome confers improved persistence on the therapeutic gene, as detected by reporter gene expression and direct assay for the presence of vector DNA sequences (26, 31, 49, 50).

The self-resolving adenovirus/EBV vector used in the present study differs from two previously reported adenovirus/EBV hybrid vectors (31, 44), both by removal of the enhancer and by the method by which Cre recombinase is delivered to the host cell to induce vector rearrangement. In these Cre helper virus-dependent systems, coinfection with an AdV expressing Cre recombinase is required to initiate the cyclization process. The present system places the Cre recombinase and loxP recombination sites on a single vector. Because manipulation of plasmids containing both Cre and loxP sites in this study was occasionally frustrated by the expression of low levels of Cre in E. coli, we introduced an intron into the Cre cDNA. Constructs bearing the intron were easily manipulated in E. coli and gave robust expression in mammalian cells. The use of hepatic locus control elements from the apoE/C gene region allows adequate suppression of expression in 293 cells while permitting recombination and subsequent gene expression in the target tissue. This approach eliminates the requirement for a helper virus, thus avoiding two potential limitations of that system. First, the continuous expression of Cre recombinase may lead to toxicity in host cells, either as a direct consequence of the protein's activity or via its immunogenicity. Second, the Cre helper virus may itself produce antigenic viral proteins that contribute to the immunologic elimination of infected host cells. In contrast, the self-resolving adenovirus/EBV vector system provides no alternative source of viral proteins, and Cre expression is terminated upon rearrangement.

The synthetic liver-specific promoter used in this work provided a means to control Cre recombinase expression during propagation of the vector in 293 cells and allowed us to test the consequences of abstracting the enhancer from the linear vector DNA upon delivery of the DNA to the target cells. However, promoter activity was not completely extinguished in some nonhepatic cells, as revealed by cyclization of the input DNA in HeLa cells. Despite this, expression of the transduced gene was minimal in HeLa cells compared to that seen in HepG2 cells or primary hepatocytes. Future efforts will be directed at creating more efficient molecular switches to improve control of circularization and gene expression.

We have also reported here a convenient general system for creating recombinant adenoviruses, which may increase their attractiveness as gene transduction tools for basic research. The system employs two conventional plasmid vectors and a λ phage packaging step. The entire recombinant AdV genome is assembled into a single cosmid that is easily amplified in E. coli. A similar approach using a three-plasmid ligation system (10) appears to be less effective, probably due to the inefficiency of initiating viral replication from DNA embedded within a larger plasmid structure. The use of intron endonuclease recognition sequences flanking the ITRs enhances virus production while simplifying insertion of therapeutic gene sequences into the pLEP shuttle plasmid. The convenience of this vector system has facilitated the construction of over 200 recombinant viruses to date.