INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The great interest in human adenovirus type 5 (Ad5) as a gene delivery platform is due in part to its ability to efficiently infect many cell types. Its wide tropism is mediated by a primary interaction between the Ad5 capsid protein, fiber, and its high-affinity cellular receptor, the coxsackievirus-adenovirus receptor (CAR). The fiber is a homotrimeric protein present 12 times on the viral capsid (32, 33). It has three domains: an N-terminal tail that interacts with the penton base in the viral capsid (24), a rod-like shaft containing 22 copies of a 15-amino-acid -sheet structure (35), and a globular knob domain (4, 34). It is the knob domain that mediates binding to CAR during cell attachment (14, 31) (Fig. 1). After the initial binding event, a second, multivalent interaction takes place between the penton base and _v integrins on the cell surface (3, 39). This step promotes virus internalization and subsequent gene transfer.

View larger version (59K): [in this window] [in a new window]   FIG. 1.   Structure of the Ad5 fiber knob monomer. -Strands are indicated in orange and are labeled A through I. Intervening coils and loops are indicated in blue. The regions mutated in this report are in yellow and are labeled KO1 through KO10 (see Table 2 for the specific residues involved). The coordinates were taken from Xia et al. (42), and the image was produced with Cn3D version 3.0 (National Center for Biotechnology Information, Bethesda, Md.).

There are many cases in which it is desirable to deliver therapeutic genes to a subset of cell types. Reducing the undesired virus-tissue interactions and increasing the intended interaction would allow lower viral doses to be used and thereby minimize possible toxicities and host immune responses. For these reasons, there has been much effort to specifically target Ad5 vectors (12, 19). This capability involves the detargeting away of a vector from its natural receptor and the simultaneous retargeting toward a new receptor that is preferentially expressed on a given cell type. The resulting vector would represent an important step in the development of this gene therapy platform, from the standpoints of both efficacy and safety.

Several strategies have been used to alter the receptor tropism and binding specificity of adenoviral vectors. These strategies include replacement of the fiber knob domain with a knob from another adenovirus serotype with a different receptor specificity (20, 30, 31), insertion of peptides into the C terminus of fiber (22, 38, 41) or the exposed HI loop (18), and use of bifunctional antibodies (40). The result of these efforts has been an expansion of viral tropism, which is suitable for some gene therapy applications, such as vascular gene therapy, in which the aim is to improve the gene transfer efficiency of adenovirus vectors that are delivered locally. However, to specifically transduce certain cell types with systemically delivered adenoviral vectors for indications such as cancer, rheumatoid arthritis, and genetic diseases, it will be necessary to combine ablation of the natural receptor tropism with introduction of a high-affinity targeting ligand.

Until recently, success in blocking the adenovirus-CAR interaction has been limited to the multicomponent systems that simultaneously block the natural receptor tropism and redirect receptor specificity toward specific cell types by using bifunctional antibodies (10), antibody-ligand conjugates (8, 26, 40), or soluble CAR-ligand fusions (5). A number of groups have now reported genetic modifications of fiber itself that reduce its ability to bind CAR (16, 28, 29). The structure of the Ad5 fiber monomer is shown in Fig. 1. Amino acids that are involved in CAR binding have been localized on the side of the fiber knob, involving residues in the AB loop, the B -sheet, the DG loop (28), and the E and F -sheets (16, 17). Roelvink et al. have described a mutant fiber protein containing a deletion of amino acids 489 to 492 in the FG loop of the fiber knob (28). Viruses encoding this mutant fiber have a reduction in transduction efficiency relative to virus containing a wild-type fiber protein. With this one exception, all of the data on fibers containing CAR-binding mutations have been generated by using purified mutant fiber proteins and measuring their ability to bind soluble CAR or to compete for fiber-mediated adenovirus transduction. Analysis of a large number of fiber mutations in the context of viral particles remains a tedious process that involves the genetic incorporation of modified fiber genes into the adenoviral DNA, rescue of the resulting adenoviral genome, and virus purification. Furthermore, since the incorporation of mutated fiber genes into the adenovirus genome may affect its efficient growth and propagation, the generation and evaluation of adenoviral vectors containing mutated fiber proteins may require alternative means of growing the vectors that will allow for efficient production of high-titer viral stocks (7, 9).

We report here a novel system to rapidly analyze modified fiber proteins in the context of the viral particle. This "transient transfection/infection" system is based on the ability to pseudotype a fiberless Ad5 vector, Ad5.gal.F, with fiber proteins expressed transiently from an episomal plasmid (Fig. 2). Ad5.gal.F is an E1-, E3-, and fiber gene-deleted Ad5 that expresses -galactosidase under the control of the simian virus 40 (SV40) early promoter (36). The modified fiber proteins for pseudotyping are produced from expression plasmids designed for high-level fiber protein expression (37). The primary advantage of this system is that modified fiber proteins can be quickly incorporated into virions and functionally analyzed in their most relevant contextthe viral particle. We used this system to analyze a panel of fiber mutants for their ability to mediate adenoviral gene transfer to HeLa cells, a CAR-expressing cell line (2). We show that the transient transfection/infection system can efficiently pseudotype a fiberless viral capsid with levels of wild-type fiber protein indistinguishable from those seen on wild-type virions. Finally, we describe Ad5 fiber amino acid residues in the AB, CD, FG, and HI loops that, when mutated, significantly reduce the ability of Ad5 to transduce cells and mediate gene transfer.

View larger version (18K): [in this window] [in a new window]   FIG. 2.   Production of adenoviral vectors pseudotyped with transiently expressed fiber proteins by using the transient transfection/infection system. (A) The fiber-deleted adenoviral vector Ad5.gal.F can be grown in packaging cell lines transiently or stably expressing different fiber proteins to generate Ad5.gal.F/F+ adenoviral particles containing fiber. The adenoviral vector is used to infect 293T cells that have been transfected with a fiber expression plasmid. The resulting particles will have new receptor tropisms dependent on the fiber protein expressed or on the peptide ligand that was inserted into the HI loop of the fiber protein. (B) Western immunoblot analysis of the Ad5 fiber and penton capsid proteins. An equivalent number of adenovirus particles for the indicated vectors were subjected to denaturing SDS-PAGE and Western immunoblot analysis. For detection of the fiber protein incorporated onto the adenoviral capsid, the membrane was developed with the rabbit anti-fiber polyclonal antibody and an anti-rabbit IgG horseradish peroxidase-conjugated secondary antibody by chemiluminescence. For detection of the penton capsid protein, the same membrane was developed with the rabbit anti-penton polyclonal antibody and an anti-rabbit IgG horseradish peroxidase-conjugated secondary antibody by chemiluminescence.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Plasmids and fiber gene mutagenesis. The Ad5 fiber cDNA was cloned into pcDNA3.1 to generate pDV60, as previously described (37). The pDV60 plasmid contains the cytomegalovirus (CMV) immediate-early promoter, the first Ad5 tripartite leader (TPL) exon, the natural first intron, and the fused second and third TPL exons upstream of the Ad5 fiber gene, followed by the bovine growth hormone gene polyadenylation signal (PA). The pDV55 control plasmid is similar to pDV60, except that it lacks the fiber gene (37). All individual amino acid changes described in this report were incorporated into the fiber cDNA by using the pDV60 plasmid as the template. Individual amino acid residues were mutagenized by the QuickChange Site-Directed Mutagenesis system (Stratagene, La Jolla, Calif.). The oligonucleotide primers used for the incorporation of amino acid changes are listed in Table 1 for each single- or double-amino-acid modification. The thermal cycler protocol was 95°C for 30 s, followed by 18 cycles of 95°C for 30 s, 55°C for 1 min, and 68°C for 20 min.

Viruses. Ad5.gal.wt is an E1-, E3-deleted Ad5 containing a lacZ reporter cassette in the E1 region (36). Ad5.gal.F is identical to Ad5.gal.wt, except that the fiber gene is deleted (36).

Cells. Human 293T cells were obtained from American Type Culture Collection and were cultured in growth medium consisting of Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS). The 293T cells stably express the SV40 large T antigen that allows for the amplification of plasmids from the SV40 origin of replication. The 633 cells stably express the Ad5 fiber protein (37) and are derived from AE1-2a, a cell line that complements E1- and E2a-deleted adenoviral vectors (11). 633 cells were grown in Richter's complete medium and 10% FBS. HeLa cells (ATCC CCL-2) were grown in DMEM supplemented with 10% FBS.

Transient transfection/infection. Mutated fiber proteins were incorporated into adenoviral particles by using the transient transfection/infection system. For each virus preparation, four 15-cm-diameter dishes of 70% confluent 293T cells were used. For fiber expression plasmid transfections, 100 µg of the appropriate plasmid DNA (Table 2), 400 µl of Lipofectamine (Life Technologies, Rockville, Md.), and 3.6 ml of Opti-MEM 1 medium (Life Technologies, Rockville, Md.) were combined and incubated at room temperature for 30 min. Sixty milliliters of Opti-MEM 1 medium was added, and a 16-ml aliquot of this transfection mixture was added to each of the four dishes of 293T cells. The plates were incubated at 37°C in 5% CO₂ for 5 h, after which the transfection medium was replaced with 20 ml of growth medium. The dishes were then incubated at 37°C in 5% CO₂ for 24 h to allow expression of the fiber protein. The transfected 293T cells were then infected with Ad5.Gal.F/F⁺ virus at a particle/cell ratio of 350. The Ad5.Gal.F/F⁺ virus is Ad5.Gal.F that was propagated in the fiber-complementing cell line 633 such that the capsid contains wild-type Ad5 fiber protein (37). The growth medium was replaced with 2.5 ml of infection medium (DMEM, 2% FBS, and Ad5.Gal.F/F⁺), and the dishes were slowly rocked at 37°C in 5% CO₂ for 2 h. Twenty milliliters of growth medium was then added, and the plates were incubated at 37°C in 5% CO₂ overnight. The medium was replaced the next day, and the incubation was continued until a complete cytopathic effect was observed, typically in 3 to 4 days. The transfected/infected 293T cells were harvested by gently dislodging the cells, pelleting them by centrifugation, and resuspending them in 1 ml of phosphate-buffered saline (PBS). A crude viral lysate was prepared by five freeze-thaw cycles. The virus was purified by centrifugation through a 1.25- to 1.4-g/ml discontinuous CsCl gradient for 1 h at 140,000 × g. The infectious virus band was isolated, placed on a 1.33-g/ml continuous CsCl gradient, and centrifuged for 18 h at 350,000 × g. The purified virus band was dialyzed against a buffer containing 200 mM Tris, 50 mM HEPES, and 10% glycerol (pH 8.0). The virus particle titer was determined spectrophotometrically as described previously (23). Yields of Ad5.Gal.F/F⁺ virus pseudotyped with modified fiber protein were typically in the range of 10¹¹ to 10¹² particles.

View this table: [in this window] [in a new window]   TABLE 2.   Transduction efficiency of pseudotyped Ad5.gal.F/F+ on HeLa cells

Production of anti-Ad5 fiber- and anti-Ad5 penton-specific antiserum. Both of the primary antibodies used in the antifiber and antipenton Western immunoblot analysis were generated by immunizations of New Zealand White rabbits (Loftstrand Laboratories, Ltd., Gaithersburg, Md.). The Ad5 fiber and penton proteins were expressed with a baculovirus expression system as previously described for expression of the Ad5 fiber protein (31). The purified Ad5 fiber protein and partially purified penton base proteins were used for immunizations according to standard protocols. The antiserum obtained was tested for immunoreactivity against the Ad5 fiber and penton proteins by Western immunoblot analysis.

Western immunoblot analysis. The expression and incorporation of each fiber protein onto adenoviral particles were verified by both nondenaturing and denaturing sodium dodecyl sulfate (SDS)-4 to 12% polyacrylamide gel electrophoresis (PAGE) and Western immunoblot analysis. An aliquot of each adenoviral vector preparation corresponding to 5 × 10⁹ particles per lane was analyzed. The proteins were transferred to a nitrocellulose membrane with a mini-TransBlot apparatus for 90 min at 30 V. The membrane was blocked for at least 1 h at room temperature in 10 mM Tris (pH 7.4) containing 150 mM NaCl, 2 mM EDTA, 0.04% Tween 20, and 5% dried milk. The blocked membrane was incubated for 1 h with a 1:1,000 dilution of the primary rabbit anti-Ad5 fiber polyclonal antiserum. The membrane was then developed with a 1:5,000 dilution of the secondary donkey anti-rabbit immunoglobulin G (IgG) horseradish peroxidase-conjugated antibody (Amersham Life Sciences, Arlington Heights, Ill.) by using the ECL enhanced chemiluminescence system (Amersham Life Sciences). The membrane was exposed to film for approximately 1 to 20 s. The membrane was then used to reprobe for detection of the adenoviral penton protein to ensure equivalent loading of viral particles. Briefly, the membrane was incubated for 1 h with a 1:5,000 dilution of the primary rabbit anti-Ad5 penton polyclonal antiserum. The membrane was then redeveloped with a 1:5,000 dilution of the secondary donkey anti-rabbit IgG horseradish peroxidase-conjugated antibody as described above.

Adenoviral transduction. HeLa cells were infected with the adenoviral particles pseudotyped with different recombinant fiber proteins to evaluate the effects of fiber amino acid mutations on CAR interaction and subsequent gene expression. Monolayers of HeLa cells in 12-well dishes were infected with 1,000 particles per cell for 2 h at 37°C in a total volume of 0.35 ml of DMEM containing 2% FBS. The infection medium was then replaced with 1 ml of growth medium per well. The cells were incubated for an additional 24 h to allow for -galactosidase expression. -Galactosidase expression was measured by a chemiluminescence reporter assay and by histochemical staining with a chromogenic substrate. The relative levels of -galactosidase activity were determined by using the Galacto-Light chemiluminescence reporter assay system (Tropix, Bedford, Mass.). Briefly, the cell monolayers were washed with PBS and processed according to the manufacturer's protocol. The cell homogenate was transferred to a microcentrifuge tube and centrifuged to remove cellular debris. The total protein concentration was determined by using the bicinchoninic acid (BCA) protein assay (Pierce, Inc., Rockford, Ill.) with bovine serum albumin as the assay standard. An aliquot of each sample was then incubated with the Tropix -galactosidase substrate for 45 min in a 96-well plate. A luminometer was used to determine the number of relative light units (RLU) emitted per sample and then normalized for the amount of total protein in each sample (RLU per microgram of total protein). For the histochemical staining procedure, the cell monolayers were fixed with 0.5% glutaraldehyde in PBS and then were incubated with a mixture of 1 mg of 5-bromo-4-chloro-3-indolyl--D-galactoside (X-Gal) per ml, 5 mM potassium ferrocyanide, 5 mM potassium ferricyanide, and 2 mM MgCl₂ in 0.5 ml of PBS. The monolayers were washed with PBS, and the blue cells were visualized by light microscopy.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Transient transfection/infection system. To rapidly evaluate a panel of potential CAR-binding fiber mutants in the context of viral particles, we have developed a transient transfection/infection system. This system, which is based on pseudotyping a fiberless virus with the mutant fiber proteins, consists of two components. The first is an E1-, E3-, and fiber-deleted adenovirus, Ad5.gal.F (36). This virus, when grown on a non-fiber-complementing cell line such as 293T, yields viral particles lacking fiber protein. For our purposes here, these fiberless virions are designated Ad5.gal.F/F. When Ad5.gal.F is produced on the fiber-complementing cell line 633 (37), the virions contain a full complement of wild-type fiber protein on the surface. These virions are referred to as Ad5.gal.F/F⁺. The second component of the system is an expression plasmid that supplies fiber protein to the assembling virus in trans. This plasmid, pDV60, is designed to express high levels of fiber protein (37).

Fiber mutation analysis. The transient transfection/infection system was then used to evaluate a series of mutations in the Ad5 fiber knob for their effect on CAR-mediated gene transfer of viral particles. A panel of expression plasmids encoding 14 mutant fiber proteins was constructed (Table 2). The regions of the fiber monomer that are mutated are shown in Fig. 1. The residues we chose to mutate were based on several criteria: amino acid sequence conservation among the fiber genes of CAR-binding adenovirus serotypes, surface exposure of the residues, and identified contact points between the fiber of the Ad12 serotype and CAR (1). For amino acid substitutions, the homologous residue in the Ad3 fiber, which does not bind CAR, was used. As controls, plasmids encoding wild-type Ad5 fiber (pDV60) and a null construct (pDV55) were used (37). Expression plasmids were transfected into 293T cells, followed by infection with Ad5.gal.F/F⁺. The resulting virions were pseudotyped with the plasmid-encoded fibers. The expression and assembly of each fiber protein into the adenoviral capsid were examined by Western immunoblot analysis of all of the CsCl-purified virus stocks. A representative immunoblot of virus particles containing 12 of the 14 recombinant fibers is shown in Fig. 3. The relative levels of fiber protein on the capsid (Fig. 3A) were compared with the amount of penton protein (Fig. 3B) to control for equal loading of viral particles in each lane. All fiber mutants were expressed and incorporated into the Ad5.gal.F viral particles. Trimerization of the expressed fiber mutants on these viral particles was confirmed by nondenaturing SDS-PAGE and Western immunoblot analysis (data not shown).

View larger version (44K): [in this window] [in a new window]   FIG. 3.   Western immunoblot analysis of capsid proteins in pseudotyped adenoviral virions. Pseudotyped Ad5.gal.F/F+ virus preparations containing the following fiber protein mutants (by lane) were subjected to SDS-PAGE and Western immunoblot analysis under denaturing conditions: 1, F; 2, KO11; 3, wild type; 4, KO1; 5, KO2; 6, KO1+2; 7, KO2a; 8, KO2b; 9, KO2c; 10, KO3; 11, KO4; 12, KO5; 13, KO4+5; and 14, KO10. A total of 5 × 109 viral particles were applied per lane. Levels of fiber were evaluated with a rabbit anti-Ad5 fiber polyclonal antiserum (A). To control for loading differences, levels of penton were also analyzed with a rabbit anti-Ad5 penton polyclonal antiserum (B). Chemiluminescence detection of the bound antibodies was then performed. The positions of fiber and penton monomer are indicated.

View larger version (14K): [in this window] [in a new window]   FIG. 4.   Differential fiber-dependent adenoviral transduction properties of HeLa cells with pseudotyped Ad5 vectors. HeLa cells were transduced with 1,000 total particles of Ad5.gal.F/F+ per cell pseudotyped with the indicated fiber or were mock transduced (). After 24 h, the cells were analyzed for -galactosidase activity by using a chemiluminescence reporter assay. The -galactosidase activity of each pseudotyped adenoviral vector containing a mutated fiber protein is expressed as a percentage of the activity measured for Ad5.gal.F/F+ pseudotyped with a wild-type fiber protein (wt). All values are derived from five to six separate transductions, each of which was carried out in triplicate, and are presented as the mean percentage of that of the wild type ± standard deviation. *, significantly different from wild-type fiber according to an unpaired, two-tailed t test analysis (P < 0.001).

View larger version (106K): [in this window] [in a new window]   FIG. 5.   Differential fiber-dependent adenoviral transduction properties of HeLa cells. HeLa cells were transduced with 1,000 total particles per cell with the indicated pseudotyped adenoviral vectors: F, wild type (wt), KO1, KO2a, KO2, KO2c, KO4, and KO10. After 24 h, the cells were analyzed for -galactosidase expression by staining the monolayers with X-Gal as described in Materials and Methods. Representative photomicrographs are shown.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

In this article, we have described the development of a novel system to analyze mutant fiber proteins in the context of the viral particle. Using this system, we identified a number of mutant Ad5 fiber proteins that retained the ability to incorporate into viral particles, but had a reduction in fiber-mediated gene transfer, presumably due to a diminished interaction with CAR. The most dramatic mutations in this respect were the knob domain deletion of KO11 (404-581) and the mutations localized to the fiber AB loop (KO1, S408E P409E) and the CD loop (KO2c, V441A K442A). We also observed effects on gene transfer efficiencies when mutating the FG loop (KO4, G509 K510) and HI loop (KO5, G538 T539).

We have identified novel residues in the CD loop of the Ad5 fiber that are involved in mediating viral transduction. All mutants that incorporated amino acid changes at either V441 or K442 (KO1+2, KO2, KO2a, KO2b, and KO2) displayed levels of fiber-mediated gene transfer of 10% or less than levels seen with wild-type fiber. The effect of the KO2b and KO2c mutations showed that residue K442, and perhaps V441, was critical for functional transduction. Notably, mutant KO2c had wild-type levels of fiber incorporated into the capsid, but only 10% of wild-type levels of gene transfer. Several other mutations in this region (KO2, KO1+2, and KO2a) reduced transduction greatly, but also reduced the level of fiber incorporated into viral particles, thus complicating any quantitative assessment of the roles of the individual amino acids. Because of the nature of this transient transfection/infection system and the mutations introduced into the fiber gene, the levels of fiber on the capsid surface can vary. This system provides a rapid screen for the effects of fiber modifications on gene transfer efficiencies. However a more quantitative analysis of any fiber mutants requires the use of genetically modified viruses which contain a full complement of fiber on the capsid. In the case of the KO2 mutant, we have genetically introduced the KO2 mutation into a viral genome and shown that the resulting virus has a full complement of the mutant fiber protein on the capsid. This vector construct still showed a reduction in adenovirus-mediated gene transfer and expression (data not shown). These studies demonstrate the importance of the V441 and K442 residues in Ad5 fiber-dependent adenoviral gene transfer.

Our data for KO1 are similar to those of Roelvink et al., in which purified fiber knob proteins containing either S408E or P409E substitutions showed reduced competition for adenoviral transduction and no detectable binding to immobilized soluble CAR (28). In addition, the homologous residues in the Ad12 fiber (P417 and P418) were also identified as contact points in the Ad12 fiber-CAR complex (1). Our study extends these analyses by generating pseudotyped viral particles containing Ad5 fiber proteins that incorporate both of these mutations and then directly demonstrating an effect of these fiber mutations on viral gene transfer. These data suggest that the S408 and P409 residues in the AB loop of Ad5 fiber are directly involved in CAR interaction.

In the Ad12 fiber-CAR crystal structure, Bewley et al. defined a number of distinct contact points between fiber and CAR that were located on two adjacent monomers (1). Amino acid residues in the Ad12 fiber AB loop (D415, P417, and I426), CD loop (V450 and K451), and E and F strands (Q487, Q494, S497, and V498) of one monomer and the FG loop (P517, P519, N520, and E523) on the adjacent monomer formed the binding site with CAR. However, analysis of the Ad5 fiber-CAR interaction (16, 17, 28) has suggested that not all of the homologous residues in the Ad5 fiber affect CAR binding. Specifically, residues in the FG loop on the adjacent monomer did not appear to play a significant role in CAR interaction, suggesting that the binding site on the fiber was localized to a single monomer. We report here, however, that fiber proteins containing amino acid mutations in the FG loop (KO4: 509-510) had significantly reduced capability for transducing HeLa cells compared to virus particles pseudotyped with wild-type fiber protein. This suggests that in the Ad5 fiber, residues in the FG loop are also involved in CAR binding and implies that each binding site on the fiber trimer spans two adjacent monomers. Our method of assessing the effect of mutations in the context of transduction by intact virions instead of with purified proteins may provide a more relevant probe of fiber-CAR interactions.

Interestingly, we also saw a smaller but significant reduction in transduction efficiency with virions pseudotyped with the KO5 mutant in the HI loop (G538-T539). It remains unclear whether this represents an Ad5 fiber contact point with CAR or a secondary effect on the overall structure of the fiber. Bewley et al. have described, in the native structure of the Ad12 fiber-CAR complex, effects on CAR binding with the deletion of residues G550 and I551 in the HI loop, even though they were not identified as contact points with CAR (1). G550 and I551 in the Ad12 fiber are homologous to G538 and T539 in the HI loop in the Ad5 fiber. In their model, they describe the presence of hydrogen bonds between G550 in the HI loop and R518 and A521 in the FG loop that stabilize the FG loop in a conformation that allows it to interact with CAR. They proposed that deletion of G550 and I551 in the HI loop results in the loss of these hydrogen bonds and destabilization of the FG loop. We propose that we are observing a similar phenomenon with our KO5 mutant. We see a significant reduction in Ad5 gene transfer efficiency with the deletion of G538 and T539 in the HI loop. Deletion of these residues may disrupt the stabilizing interactions between the Ad5 FG and HI loops, thereby reducing the ability of the FG loop to interact with CAR. When we combined the FG and HI loop mutations (KO4+5), the result was to restore much of the loss in transduction efficiency displayed individually by the KO4 and KO5 mutants. Perhaps simultaneously shortening the FG and HI loops restores some stabilizing interactions between the FG and HI loops. This may have the effect of improving fiber-CAR interactions, stabilizing fiber trimers, or restoring the efficiency of fiber assembly into virions.

The second requirement for an adenovirus that transduces in a cell-type-specific manner is the introduction of a novel tropism. The most efficient means is by genetic modification of the fiber gene. Krasnykh et al. (18) have shown that the HI loop is an appropriate location in the fiber protein in which to insert peptides with novel receptor specificities. For example, the cRGD ligand (25) inserted into the HI loop has been shown to expand the tropism of adenovirus both in vitro (6) and in vivo (27). For some applications, it may be sufficient to improve adenovirus transduction efficiency, without diminishing the native tropism, to achieve the desired therapeutic effect. Examples include ex vivo gene therapy applications, such as in transplantation, and in situ applications, such as local gene delivery to blood vessels or intratumoral adenoviral administrations. However, in cases in which systemic delivery is required, a combination of a novel tropism using high-affinity ligands and the ablation of the natural receptor specificity by fiber protein mutagenesis will be necessary. The next step in the development of a highly specific adenovirus is to combine these two components into a single fiber protein.

One distinct advantage of the transient transfection/infection system described here is that there is no need for a pseudoreceptor system to propagate virions that do not bind CAR. CAR binding is an essential function of fiber needed for efficient viral production. The production of high-titer vector stocks with which to test fiber-CAR interactions in the context of the virus particle is difficult without an alternative cell-binding pathway (7, 9). Virus production in the transient transfection/infection system involves infection with the fiber-deleted virus, transfection with a fiber expression plasmid, and a single round of replication that results in a viral capsid pseudotyped with the fiber mutants expressed in trans. It should be possible therefore to more easily test combinations of CAR-binding mutations and targeting ligands for their ability to mediate transduction.

A "transient transcomplementation" system has recently been described in which fiber expression plasmids were cotransfected with fiber-deleted adenoviral genomic plasmids in order to rescue infectious virus (21). The transient transfection/infection system we describe here differs in a number of respects. For one thing, we are introducing the fiber gene-deleted adenoviral genome by infection with a virus previously pseudotyped on a complementing cell line. In addition, we are applying this system in a novel way to perform structure-function studies of the fiber protein. In this report, we describe the use of this system to analyze mutant fiber proteins for their CAR-binding properties. However, the system can also be adapted to rapidly assess the effect of novel ligands on fiber-mediated transduction and gene transfer, as well as combinations of ligands and CAR-binding mutations. The transient transfection/infection system gives us the ability to quickly evaluate promising candidates in the context of a fiber on the viral particle for their ability to transduce the desired cell types.