INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Little information is available on the mechanism of selective and polar DNA encapsidation of adenovirus (Ad) genomes into preformed, empty capsids late in the infectious viral life cycle. It is thought that cis-acting packaging sequences and trans-acting protein components act in conjunction to mediate DNA packaging, similar to a number of bacteriophages like lambda or 29 (reviewed in references 7, 11, and 25). Ad DNA encapsidation occurs in a polar manner from left to right and relies on a cis-acting packaging domain located between approximately nucleotides (nt) 200 and 380 (5, 14, 16, 22, 28). The location of the Ad type 5 (Ad5) packaging domain is schematically depicted in Fig. 1A. The Ad5 packaging domain consists of at least seven redundant, albeit not functionally equivalent, elements termed A repeats I through VII (AI through AVII) (12, 13). We have previously shown that the four most dominant A repeats, I, II, V, and VI, exhibit a bipartite consensus motif 5'-TTTGN₈CG-3', which is conserved across a number of Ad serotypes (26). These four repeats are aligned relative to each other in Fig. 1B and C. There are spacing constraints between the two conserved parts of the bipartite consensus motif rather than between different A repeats. Multimerized packaging elements can restore viability to a mutant virus lacking a packaging domain, thus supporting the notion of an independent character of the each element.

View larger version (24K): [in this window] [in a new window]   FIG. 1.   The Ad5 packaging domain. (A) Schematic representation of the left end of the Ad5 genome. Nucleotide positions, relative to the left terminus, are indicated by numbers. The ITR is represented by a grey box. Viral packaging repeats AI to AVII (arrows) are located between nt 194 and 380. The E1A transcriptional start site is indicated by an arrow at nt 499. Viral enhancer elements I and II are designated E1A enhancer. (B) Alignment of AI and AII with AV and AVI. The packaging repeat consensus motif is boxed for each A repeat. Invariant nucleotides are indicated by lines between the sequences. (C) The packaging repeat consensus motif. Shown is an alignment of AI, AII, AV, and AVI. Nucleotides comprising the bipartite consensus motif for AI, AII, AV, and AVI are boxed and enlarged. The consensus motif is shown at the bottom.

It was demonstrated in early mutational studies that neither inversion of the Ad packaging domain nor its relocation to the right-end terminus affect its function (14, 16). Despite its positional flexibility, however, it must be located within 600 bp of the genomic terminus for proper function. In addition, spacing constraints exist between the packaging domain and sequences to the left, as shown by insertional mutagenesis (13). These observations raise the possibility that sequences in the inverted terminal repeat (ITR) play a direct role in viral DNA packaging. If this proves to be the case, two lines of evidence delimit left-end sequences with a potential function in Ad DNA packaging. Deletion analysis revealed that sequences between nt 53 and 194 are entirely dispensable for viral packaging in vivo. Additionally, a mutant virus lacking the region between nt 44 and 195 (dl309-44/195) exhibits defects in DNA replication, whereas DNA packaging is not impaired in a mixed infection with wild-type virus (15). Taken together, these results implicate a possible involvement of sequence elements between nt 1 and 44 in viral DNA packaging. This region contains the viral core origin of replication, and by extension, there may be a physical linkeither cis or trans actingbetween Ad DNA replication and packaging.

Nothing is known about the identity of trans-acting packaging components involved in Ad packaging, but several lines of evidence support their existence in limiting concentrations in the infected cell. Most notably, cotransfection of an excess of unlinked packaging domain sequences with wild-type Ad genomes dramatically decreases virus yield without a significant effect on DNA replication and late transcription (13). This decrease in virus growth is thought to reflect the competition of limiting, trans-acting packaging components from viral genomes by the unlinked packaging domain fragments, resulting in their inability to be encapsidated. Additionally, in coinfection experiments, viruses carrying more A repeats package with greater efficiency than viruses with fewer A repeats even though comparable levels of nuclear viral DNA are available for encapsidation with both viral genomes (26). The same viruses package with equal efficiency in individual, single-virus infections. Similar results are observed in coinfection experiments using wild-type Ad5 and packaging domain mutants: a greater reduction in packaging efficiency of mutant viruses is found in coinfections than in single infections with the same viruses (12, 13, 26). These observations are consistent with the model that competition for a limiting trans-acting component occurs within an infected cell. Viruses with a wild-type packaging domain or a greater number of A repeats package more efficiently than mutants with fewer A repeats due to more successful competition for a limiting packaging component(s).

We have analyzed the ability of individual A-repeat elements, or a combination of A repeats, to direct the packaging of Ad5 DNA in the absence of an intact packaging domain. Multimers of different A repeats are able to direct packaging of viral DNA but at different efficiencies, which confirms the existence of a functional hierarchy of packaging repeats and rates the importance of individual elements directly. AV to VII (collectively referred to as AV-VII) and AI, when multimerized, serve as the most efficient packaging domains in vivo, followed by AII as an element with moderate activity and AVI as the weakest packaging element. A repeats are believed to constitute binding sites for trans-acting packaging components. Minimal packaging elements were used as probes for in vitro protein binding assays in order to identify trans-acting cellular and/or viral components of the Ad packaging machinery. The binding of a cellular protein(s) (termed P-complex) with various packaging repeat probes was observed, suggesting that cellular factors might be participants of the viral packaging machinery. A direct correlation was seen between the binding affinity of P-complex for different A repeats in vitro with the ability of the fragments to support DNA packaging in vivo. The TTTG, but not the CG, packaging consensus half site is critical for P-complex interaction. In addition, the P-complex was found to bind to core replication sequences in the ITR. The cellular P-complex activity, by virtue of its ability to interact with both packaging and core replication sequences, might constitute a trans-acting link between viral DNA replication and encapsidation. We also detected the binding of a cellular transcription factor, chicken ovalbumin upstream promoter transcription factor (COUP-TF), to minimal segments of the viral packaging domain. Its binding affinity does not correlate with viral DNA packaging in vivo, the implications of which will be discussed.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Virus constructions. Ad5 dl309, the parent for all viruses described in this report, is a phenotypically wild-type virus that contains a unique XbaI cleavage site at 3.8 map units (19). Plasmid pE1A-194/814 contains the left-end Ad5 XbaI fragment (nt 1 to 1339) with a deletion between nt 194 and 814 and a unique XhoI restriction site at the junction of the deletion. A head-to-tail hexamer of an oligonucleotide containing AVI (5'-TCGACCGCGGGGACTTTGACC-3':5'-TCGAGGTCAAAGTCCCCGCGG-3') was cloned into the 194/814 deletion. Similarly, head-to-tail hexamers of oligonucleotides containing AI (5'-TCGAGTTGTAGTAAATTTGGG-3':5'-TCGACCCAAATTTACTACAAC-3') or AII (5'-TCGACCGAGTAAGATTTGGCC-3':5'-TCGAGGCCAAATCTTACTC GG-3') were cloned into the pE1A-194/814 background. The recombinant plasmids were subsequently rebuilt into intact viruses by the method of Stow (27). Viruses were amplified and titers were determined on 293 cells. Mutant viruses were screened by restriction analysis of viral DNA obtained from infected 293 cells by the Hirt procedure (17), and all insertions were verified by nucleotide sequence analysis of viral DNA by using PCR-based sequencing.

Cultured cells and infections. Monolayer HeLa and 293 cells were maintained in Dulbecco's modified minimal essential medium containing 10% calf serum (HyClone). Suspension cultures of HeLa cells were grown to a density of 4 × 10⁵ cells per ml in suspension-modified minimal essential medium containing 7% calf serum. Virus stocks were generated by three freeze-thaw cycles of infected cell lysates, and titers were determined by plaque assays on 293 cells. Virus infections were performed at a multiplicity of infection of 3 PFU per cell for 1 h at 37°C. Cells were then washed twice with Tris-buffered saline solution and overlaid with fresh medium.

Determination of virus yield and packaging efficiency. Both assays were performed as described previously (26). For the determination of virus yield in a single infection, infected cell lysates were prepared 48 h postinfection and the amount of infectious virus was determined by plaque assays on 293 cells. Packaging efficiency of the mutant viruses was tested in a coinfection of 293 cells with both mutant and wild-type dl309 virus. Forty-eight hours postinfection, one half of the cells was used to isolate total nuclear DNA; the other half was used for the preparation of viral DNA from purified virions. Both DNA preparations were digested with XbaI to distinguish between mutant and wild-type DNA and quantitated by Southern blot hybridization using pE1A-WT, P labeled by the random primer method (8), as a probe. The relative intensities of the bands in autoradiograms were determined by densitometric scanning. The data presented for virus yield in the single infections and the data for packaging efficiency based on coinfection experiments represent the averages of at least three independent experiments.

Extract preparation and gel mobility shift assays. Nuclear extracts were prepared by the method of Dignam et al. (6) and dialyzed overnight against 20 mM HEPES (pH 7.5)-100 mM NaCl-10% glycerol-5 mM MgCl₂-0.2 mM EDTA-0.5 mM dithiothreitol-0.5 mM phenylmethylsulfonyl fluoride (DB-100). The dialysate was cleared by centrifugation at 25,000 × g for 20 min. Two to 5 µg of nuclear extract was incubated with 0.5 µg of poly(dI-dC) and 20,000 cpm of ³²P-labeled probe DNA (2.5 to 5 fmol of DNA) per in vitro binding reaction. The binding reaction was carried out in a total volume of 20 µl for 1 to 2 h at room temperature in 40 mM HEPES (pH 7.5)-70 mM NaCl-0.1 mM EDTA-0.5 mM DTT-0.5 mM phenylmethylsulfonyl fluoride, 10 µg of bovine serum albumin per ml-4% Ficoll. The complexes were resolved electrophoretically at 10 V/cm on a 3.5% 30:1 (acrylamide/bisacrylamide) polyacrylamide gel in 0.5× TBE (25 mM Tris [pH 8.3], 25 mM boric acid, 0.5 mM EDTA) at 4°C. For gel mobility shift assays performed with in vitro-translated COUP-TFI protein, 0.25 to 1.5 µl of rabbit reticulocyte extract programmed with in vitro-synthesized RNA transcript encoding COUP-TFI was assayed by using the binding conditions described above. In vitro transcription and translation was performed as recommended by the manufacturer (Promega). For gel mobility supershift experiments, 0.5 µl of a rabbit polyclonal anti-COUP-TF antiserum (a gift from Alonzo D. Garcia) was added after a 1-h binding reaction, and incubation was then continued for an additional 30 min.

Plasmids, probes, and competitor fragments. Head-to-tail hexamers of AI and AVI, individually, and a dimer of AV-VII were cloned into pUC9. The sequences of monomers are as follows: AI, 5'-TCGAGTTGTAGTAAATTTGGG-3':5'-TCGACCCAAATTTACTACAAC-3'; AVI, 5'-TCGACCGCGGGGACTTTGACC-3':5'-TCGAGGTCAAAGTCCCCGCGG-3'; and AV-VII, 5'-TCGACCGCGTAATATTTGTCTAGGGCCGCGGGGACTTTGACCGTTTACGTGGAG ACTCC-3':5'-TCGAGGAGTCTCCACGTAAACGGTCAAAGTCCCCGCG GCCCTAGACAAATATTACGCGG-3'. The fragments were liberated from the vector by digestion with EcoRI and HindIII, gel purified, and ³²P end labeled with Klenow DNA polymerase and [-³²P]dATP. For the preparation of the ITR 1-13 probe, a monomeric oligonucleotide representing the left-end 13 nt flanked by Xho/Sal linkers (5'-TCGACATCATCAATAATC-3':5'-TCGAGATTATTGATGATG was end labeled in the same way, using [-³²P]dCTP.

Heparin agarose fractionation. HeLa cell nuclear extract (~100 mg of total protein) was subjected to heparin agarose chromatography with a bed volume of 1 ml/10 mg of loaded protein. The column was loaded in buffer DB-100 and washed with 3 bed volumes of buffer. Bound proteins were eluted by using a linear gradient between 0.1 and 1 M NaCl in DB. Subsequent dialysis was through a hollow fiber bundle against DB-100. P-complex activity was assayed by gel shift analysis using different packaging and ITR sequences as probes. Generally, a volume of 0.1 to 0.5 µl per fraction was used for gel mobility shift assays using the same binding conditions as described above.

Western blot analysis. Western analysis was performed with 10-µl aliquots of each heparin agarose fraction, boiled in 2× sodium dodecyl sulfate (SDS) sample buffer, and run on an SDS-15% polyacrylamide gel. Proteins were transferred to nitrocellulose and probed with a 1:1,000 dilution of polyclonal anti-COUP-TF antiserum. Proteins were visualized by using a secondary horseradish peroxidase-conjugated antibody and chemiluminescence as recommended by the manufacturer (Amersham).

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Minimal Ad packaging domains. Ad packaging elements are functionally redundant, but in spite of this redundancy, different elements are not functionally equivalent with respect to each other. Elements I, II, V, and VI constitute the functionally most dominant A repeats (12, 13, 26). The selection of revertant Ad from a packaging-deficient parent virus has previously defined AVI as an independent cis-acting unit (26). A hexamer of AVI in place of the packaging domain yields a viable virus, although the mutant is reduced >100-fold in growth compared to wild type. Such a mutant is under strong evolutionary pressure for the amplification of packaging elements, since revertants with significantly improved growth were found to evolve by amplification of preexisting copies of AVI. In contrast, a fragment containing AV-VII functions efficiently to direct packaging, and these A repeats did not amplify upon virus propagation (Fig. 2A) (26). Sequences flanking the packaging domain are identical in both of these mutant viruses (a deletion of sequences between nt 194 and 814). This raises the question of whether there is a hierarchy of importance among the four most dominant A repeats, with AVI as a functionally less dominant element, or alternatively, whether a combination of different elements supports packaging better than only one type of A repeat.

View larger version (24K): [in this window] [in a new window]   FIG. 2.   Functional hierarchy among different packaging repeats. (A) A schematic representation of left-end sequences of dl309 wild-type (WT) virus is shown at the top. The ITR is represented by a shaded box. AI, AII, AV, and AVI are represented by boxes of distinct shading. Numbers represent nucleotide positions relative to the left terminus. The mutant viruses contain a deletion between nt 194 and 814 and the insertion of six copies each of AVI (194/814:AVI6), AII (194/814:AII6), and AI (194/814:AI6) or a dimerized copy of AV, AVI, and AVII (194/811:AV-AVII2). Mutant virus yields in the single infections (Yield) are expressed as fold reduction in yield relative to that of the wild-type virus. The results from the coinfection experiments and Southern blot analysis (Coinf.) are expressed as fold reduction in packaged mutant DNA relative to packaged wild-type DNA. These data were normalized to the levels of viral DNA (mutant and wild type) present in total nuclear DNA. NV, virus was not recovered in three independent attempts to reconstruct the mutant virus from plasmid DNA; ND, packaged viral DNA was below the level of accurate quantitation. (B) Southern blot analysis of total nuclear and virion DNA isolated from 293 cells coinfected with wild-type and mutant viruses depicted in panel A. Total nuclear DNA (lanes 1 to 3) or virion DNA (lanes 4 to 6) was digested with XbaI and subjected to Southern blot analysis using an Ad5 left-end fragment as a 32P-labeled probe. The corresponding left-end fragments of wild-type (WT) and mutant (M) genomes are indicated on the left. The mutant viruses analyzed were 194/814:AI6 (lanes 1 and 4), 194/814:AII6 (lanes 2 and 5), and 194/814:AVI6 (lanes 3 and 6).

A cellular complex (P-complex) interacts with Ad packaging elements. Minimal packaging domains defined in vivo were used as probes for gel mobility shift assays for the detection of trans-acting packaging components. Since such components could be viral and/or cellular in origin, we initially carried out binding studies with both uninfected and Ad-infected 293 cell nuclear extracts. Infections were performed with either wild-type Ad dl309 or a temperature-sensitive virus, ts19, defective for virus assembly when grown at the restrictive temperature (33). Extracts from ts19-infected cells were tested in view of the fact that packaging factors may be encapsidated with wild-type Ad and consequently not present in nuclear extracts used for in vitro binding studies. At no point did we detect any difference between complex formation using nuclear extracts from infected or uninfected cells; therefore, all experiments presented below were performed with extracts from uninfected cells.

View larger version (59K): [in this window] [in a new window]   FIG. 3.   A cellular complex interacts with Ad packaging elements. A gel mobility shift competition experiment using nuclear extract prepared from uninfected 293 cells and a dimeric AV-VII probe is presented. Radiolabeled probe, nuclear extract, and nonspecific competitor DNA [poly(dI-dC)] were incubated in the absence (lanes 1 and 24) or presence (lanes 2 to 23) of unlabeled oligonucleotides corresponding to packaging elements in the wild-type or mutated configuration. Protein-DNA complexes formed were resolved on a 4% nondenaturing polyacrylamide gel. P-complex DNA binding activity is indicated by an arrow. Increasing amounts of specific competitor oligonucleotides (see Materials and Methods for the sequences) are indicated and represent 40- and 200-fold molar excesses of A repeats relative to the probe. The competitors are named according to the A repeats they represent. The suffix "LS" was appended when the TTTG consensus motif in the oligonucleotide was mutated; the suffix "CG" was appended when the CG consensus dinucleotide was mutated.

COUP-TF interacts with adenovirus packaging elements. Uninfected HeLa cell nuclear extract was subjected to heparin agarose chromatography. Eluate fractions were subsequently assayed for the presence of P-complex in a gel mobility shift assay using the hexameric AI probe. The results are shown in Fig. 4A; only fractions 16 to 31 are depicted. The peak of P-complex binding activity was detected in fractions 19 to 24. Several prominent faster-migrating complexes were observed in fractions 24 to 31. The same eluate fractions were tested in gel mobility shift assays using as a probe a hexamer of AVI, (Fig. 4B), which serves as a weak minimal packaging domain in vivo. P-complex binding activity was again detected in fractions 19 to 24. In addition, we observed a striking ladder of protein-DNA complexes in fractions 24 to 31. 

View larger version (53K): [in this window] [in a new window]   FIG. 4.   The peak of binding activity to AVI corresponds to COUP-TF protein distribution. (A) Uninfected HeLa cell nuclear extract was fractionated by heparin agarose chromatography. Gel mobility shift experiments were performed with 0.5 µl of each fraction (only fractions 16 to 31 are represented), using the hexameric AI probe, as described in the legend to Fig. 3. (B) Fractions were assayed for binding activity using the hexameric AVI probe as described above. (C) A volume of 10 µl per fraction was used for protein separation on SDS-polyacrylamide gels. Proteins were transferred to nitrocellulose and subjected to Western blotting analysis using anti-COUP-TF antiserum and chemiluminescence. Individual fractions are specified by the numbers above the data. The low-molecular-weight form of COUP-TF protein is detected.

View larger version (88K): [in this window] [in a new window]   FIG. 5.   Binding of COUP-TFI to minimal packaging domains. Gel mobility shift assays were performed, as described in the legend to Fig. 3, using COUP-TFI synthesized by in vitro translation. A hexamer of AVI (lanes 1 to 9) and a hexamer of AI (lanes 10 to 18) were used as radiolabeled probes. A volume of 2 µl of unprogrammed reticulocyte lysate (Unprog) or increasing amounts of COUP-TFI-programmed lysate (COUP) was used per binding reaction (0.25, 0.5, 1, and 1.5 µl, respectively). The addition of preimmune serum (preimm.) or anti-COUP antiserum (-COUP) is indicated above the lanes; 0.5 µl of polyclonal antiserum was added per antibody supershift experiment. The amount of COUP-TFI-programmed lysate used for supershift experiments was 1.5 µl of lysate.

P-complex interacts with viral core origin sequences. Figure 6A shows data from a gel mobility shift experiment in which a heparin agarose fraction containing peak levels of P-complex (Fig. 4A, fraction 21) was incubated with hexameric AI probe and specific competitor DNAs. P-complex formed in the absence of specific competitor DNA is shown in lane 1. In lanes 2 and 3, a 40-fold molar excess of competitor oligonucleotides representing wild-type or mutated A repeat I was added to the binding reaction. Efficient competition was observed with wild-type (lane 2) but not mutated AI (lane 3), as seen previously with 293 cell nuclear extract (Fig. 3). In experiments not shown, we had observed that P-complex binding activity was specifically competed not only by various packaging repeats but also by sequences derived from the left terminus of the Ad genome. We further delineated these left-end sequences to the terminal 13 nt, using competition experiments (Fig. 6A). A 10-, 50-, or 250-fold molar excess of competitor oligonucleotide representing Ad5 left-end sequences was included in the binding reactions (lanes 4 to 9). Specific competition for P-complex activity was observed with sequences between nt 1 and 13 (lanes 4 to 6) but not with sequences between nt 10 and 22 (lanes 7 to 9).

View larger version (104K): [in this window] [in a new window]   FIG. 6.   P-complex interacts with ITR nt 1 to 13. (A) Fraction 21 from the heparin agarose chromatography was subjected to gel mobility shift competition assays using hexameric AI as a probe. No specific competitor (+; lane 1) or a 40-fold molar excess of multimeric AI was added, either in its wild-type configuration (lane 2) or containing a mutation in the TTTG consensus half site (lane 3). Increasing molar amounts (10-, 50-, and 250-fold relative to binding sites in the probe) of unlabeled monomeric left-end sequences were added as specific competitors in lanes 4 to 9. These sequences correspond to ITR nt 1 to 13 (lanes 4 to 6) and ITR nt 10 to 22 (lanes 7 to 9). (B) Fraction 21 from the heparin agarose chromatography was subjected to gel mobility shift competition assays using monomeric Ad5 sequences from nt 1 to 13 as a probe. No specific competitor was added in lane 1 (+). Unlabeled monomeric ITR fragments representing nt 1 to 13 and 10 to 22, respectively, were added in lanes 2 and 3 and lanes 4 and 5 at 50- and 200-fold molar excesses relative to the probe. Multimeric packaging element competitors were added in lanes 6 and 7 (AI, wild type) and lanes 8 and 9 (AI, mutated in the TTTG consensus half site) at 50- and 200-fold molar excesses relative to the probe.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

We previously defined minimal Ad packaging sequences that exhibit packaging activity in vivo (26). A dimer of a fragment containing AV, AVI, and AVII rescued packaging efficiency in a coinfection experiment to wild-type levels in the background of a virus lacking left-end sequences between nt 194 and 811 and to near wild-type levels in a mutant background containing a deletion of sequences between nt 53 and 811. Therefore, this fragment constituted a minimal packaging domain that is necessary and sufficient for viral DNA packaging in vivo. The fact that multimers of AVI could not reconstitute viral packaging to the extent that a combination of AV, AVI, and AVII did (Fig. 2A) suggests that a combination of different A repeats may support viral packaging better than only one type of element. Consistent with this idea, a functional hierarchy of different A repeats was found when tested for the ability to direct viral DNA packaging. An AVI hexamer in place of the packaging domain resulted in poor virus growth in a single infection, and packaged DNA was not detectable in a coinfection experiment. In addition, we found sequences that most likely represent AVI itself amplified upon propagation of the virus (Fig. 2B). This was observed previously with revertants selected from a packaging-deficient Ad and with a virus carrying AVI multimers in a 194/814 deletion background (26). In contrast, neither of the viruses carrying multimerized AI or AII packaging elements exhibited amplification of packaging sequences. In comparison to the virus with AVI repeats, a virus with multimerized AII elements showed a significant improvement in growth in a single infection as well as packaging efficiency in the coinfection. A virus with multimerized AI repeats grew and packaged its DNA to near wild-type levels. This rates AI as the functionally most dominant packaging element when tested in isolation of other A repeats. We do not know yet how an AV hexamer compares with this group. In light of the fact that a dimer of a fragment comprising AV to AVII rescues the nonviable parent virus to near wild-type levels (26), while a hexamer of AVI appears to be a relatively weak packaging element (Fig. 2), the ability of AV multimers to direct packaging should be informative since only minor importance was assigned to AVII in previous studies (13).

We detected two cellular binding activities which interact in a specific manner with Ad5 packaging elements. One activity is cellular transcription factor COUP-TF, whereas the other activity, P-complex, is of unknown identity. Although we do not yet know the biological function of either activity in the context of viral infection, gel mobility shift analyses suggest that Ad may reprogram cellular factors for the encapsidation of viral DNA just like it does for many other aspects of its life cycle, including viral transcription and replication. Little information is generally available on the mechanism of DNA encapsidation of eucaryotic DNA viruses or the cis- and trans-acting components involved. Evidence exists, however, for the binding of packaging signals by cellular proteins. In the case of human cytomegalovirus, an approach similar to ours was used to identify trans-acting components that interact with a minimal, highly conserved, cis-acting element (a sequence) which is sufficient to serve as a cleavage/packaging and recombination signal in vivo. A cellular protein, pac2B, which binds this minimal element in a sequence-specific manner was partially purified and characterized (20). The related a sequence in the herpes simplex virus type 1 genome has also been used as a probe for in vitro protein binding studies and has yielded both virus infected- and mock-infected cell-specific protein-DNA complexes (3). The biological significance of these findings, however, remains unknown.

Several lines of evidence demonstrate sequence-specific binding of a known cellular transcription factor, COUP-TF, to viral packaging elements in vitro. First, COUP-TFI bound A repeats when synthesized in vitro (Fig. 5) or when expressed by using baculovirus (data not shown). Highest affinity was observed for AVI multimers. Second, heparin agarose chromatography correlated the peak of binding activity interacting with AVI multimers with peak levels of COUP-TF protein (Fig. 4B and C). Third and not shown in this report, gel mobility supershift experiments using minimal packaging domains as probes revealed the presence of a COUP-TF-related binding activity in uninfected nuclear 293 and HeLa cell extracts.

Binding studies with COUP-TF were originally initiated because database searches with fragments comprising AI and AII as well as AV and AVI revealed the presence of putative COUP-TF binding sites within viral packaging repeats. COUP-TF has been shown to interact with DNA in homo- or heterodimeric form, and the recognition sequences for homodimers have been characterized extensively by Cooney et al. (4). In solution, COUP-TF exists as a stable homodimer (24). An alignment of natural COUP-TF binding sites in a number of cellular promoters, and binding studies using synthetic recognition sites demonstrated that there is remarkable flexibility in the interaction of COUP-TF homodimers with perfect or imperfect direct or palindromic 5'-GGTCA-3' or 5'-TGACC-3' half sites separated by variable spacings. Despite its semipromiscuous DNA binding, homodimeric COUP-TF requires the presence of two consensus half sites, which correlates with the fact that COUP-TF dimers contact both half sites equally as tested by methylation interference assays (4, 18).

It is unclear which nucleotides within Ad packaging repeats are contacted by COUP-TF, but a number of sequence motifs resemble consensus binding sites. AVI is the only packaging element that contains a perfect COUP-TF half site as part of an inverted consensus motif (5'-GGACTTTGACC-3'; the COUP-TF inverted repeat is underlined, and AVI is in boldface). In correlation with this observation is the high-affinity interaction between AVI and COUP-TF (Fig. 5). AVI as a relatively weak packaging element in vivo (Fig. 2), and the amplification of AVI in the course of virus propagation previously was observed (26). The affinity of COUP-TF for AVI multimers may result in its recruitment to the packaging domain, where it might inhibiteither passively or activelysubsequent events in viral DNA encapsidation. The observed amplification of AVI in the course of virus propagation could allow for the titration of such a packaging antagonist. We do not believe that selection pressure for the observed amplification is solely due to the need to titrate out COUP-TF for the following reason. Coinfection experiments demonstrated that viruses with more copies of AVI have a competitive advantage over viruses with fewer copies when challenged to compete with each other in coinfection experiments (26). If the observed amplification of A repeats were to merely remove a suppressor of DNA packaging from viral packaging domains, genomes containing more copies of AVI would help compete the antagonist away from genomes containing fewer copies. In a coinfection experiment, the presence of more copies of A repeats would therefore not be expected to provide a selective advantage over the presence of fewer copies. Rather, these data support the idea that AVI constitutes an intrinsically weak binding site for the bona fide packaging factor and that amplification of AVI increases the chance of packaging factor binding.

Several arguments support the authenticity of the cellular P-complex as a bona fide Ad packaging component. Complex formation is observed on all minimal packaging domains that exhibit functional activity in vivo. The affinity of P-complex for the different multimeric A repeats in vitro correlates well with the ability of the respective cis-acting sequences to support viral DNA packaging in vivo. Specifically, AI and AV-VII constitute the strongest P-complex binding sites, and they confer maximal packaging activity in vivo. On the other hand, AVI is noted as the weakest binding site for P complex in vitro, and it serves as a particularly weak packaging domain in vivo. We have previously defined the Ad packaging consensus motif as a bipartite sequence with a conserved AT-rich and a GC-rich half site (5'-TTTGN₈CG-3') (26). Mutational studies have shown that the TTTG as well as the CG nucleotides are critical determinants for DNA packaging in vivo. In an attempt to correlate these results with in vitro binding data, we performed gel mobility shift competition experiments using specific competitors representing packaging elements in their wild-type configuration as well as harboring mutations within the TTTG and CG consensus half sites (Fig. 3). The integrity of the TTTG is necessary for P-complex interaction in vitro. In contrast, the CG dinucleotide, while critical for virus packaging in vivo, is not required for P-complex binding. At this point we do not have an indication as to what the mechanistic role of the CG-rich half site may be.

From a comparison of P-complex formation using extracts from uninfected and Ad-infected cells, and from the appearance of identical complexes in both cases, we infer that P-complex constituents are of cellular origin. Preliminary experiments also suggest that neither the on-rate nor the off-rate of complex formation is changed upon incubation of extracts from infected versus uninfected cells with packaging sequences (data not shown). However, we cannot exclude the possibility that viral components are contained within P-complex but escape detection in our assays. We do not believe that COUP-TF is a bona fide packaging factor since COUP-TF binding in vitro does not correlate with data using minimal packaging domain mutant viruses in vivo. Further, COUP-TF does not appear to be a part of P-complex. Polyclonal COUP-TF antiserum added before and after probe addition did not affect complex formation (data not shown), and the peak of COUP-TF protein present in heparin agarose fractions overlapped, but did not coincide with, the appearance of P-complex (Fig. 4).

The cellular P-complex interacts not only with packaging elements but also with sequences derived from the left terminus (nt 1 to 13) of the Ad genome. Complex formation on packaging elements is specifically inhibited by the addition of left-end fragments. Similarly, complexes formed on a probe corresponding to the left-end nt 1 to 13 sequences were specifically competed by packaging elements (Fig. 6). The ITR binding site for P-complex encompasses nt 1 to 13 and thus does not contain binding sites for cellular factors NFI and NFIII, known to interact with other motifs within the ITR (30). On the other hand, a cellular activity, ORP-A, that interacts with ITR nt 1 to 12 in a gel mobility shift assay has been identified (23). This activity has not been cloned, and binding to the core origin in vitro has not been correlated with a physiological role in vivo. It remains to be seen whether ORP-A is contained within or identical to P-complex activity. Dual specificity of a trans-acting component for both packaging and ITR sequences may be biologically significant in light of the fact that physical closeness between the viral packaging domain and ITR sequences is required for DNA encapsidation in vivo. A direct role for the genomic left terminus in viral DNA packaging, however, has not been demonstrated.

Our working model, shown in Fig. 7, is based on the data from protein binding studies presented here. We envision a coordinate interaction of packaging factors with viral A repeats. In vivo, at least three copies of A repeats are required for efficient DNA encapsidation (12, 13), which likely reflects the need for the presence of multiple protein binding sites. Either the same or a different trans-acting component may bind the left-end 13 nt of the Ad genome. Physical association between the components bound to ITR and packaging sequences may result in the formation of a nucleoprotein complex within the viral left end, marking the respective molecule as a bona fide packaging substrate. This complex is presumed to correspond to the P-complex detected in our gel mobility shift assays since it exhibits binding specificity for both packaging and ITR sequences. Subsequent capsid recognition may involve recognition of this complex on a structural level, or alternatively, through direct protein-protein interactions between components of P-complex and capsid subunits. Our data implicate the AT-rich packaging consensus half site in the initial recognition of A repeats by packaging factors. Perhaps the CG-rich half site and proteins bound to it are involved in secondary events like capsid recognition or insertion of the viral DNA into the capsid. It is noteworthy that the spacing of 11 bp, or one helical turn of the DNA, which separates the AT-rich and the CG-rich consensus half site is critical for DNA encapsidation in vivo. This may reflect the need for a physical interaction between components of P-complex and CG-bound unidentified components, to allow for the timing and/or coordination of successive steps in Ad DNA packaging.

View larger version (13K): [in this window] [in a new window]   FIG. 7.   P-complex and Ad DNA packaging: a working model. The left terminus of the Ad genome is schematically represented with ITR and packaging (PACK) domains denoted by boxes. trans-acting components binding ITR and packaging sequences are identical in the model on the left, whereas different factors interact with the respective sequences in the model on the right, as indicated by circles of different color.

What may be the functional significance of a trans-acting link between Ad DNA replication and DNA packaging? We can only speculate on this issue, but several possibilities come to mind. Compaction of the viral DNA into a capsid is energetically unfavorable and appears to universally require ATP hydrolysis, as has been clearly demonstrated in numerous bacteriophage systems including phage lambda, 29, and X174 (reviewed in references 7 and 11). In these systems, DNA packaging has been extensively studied in vitro by using purified components. If Ad replication and assembly were to be physically coupled, as suggested by Weber et al. (32), concomitant DNA replication might provide energy required for the tight packing of the DNA within the prohead. Alternatively, a link between DNA replication and DNA packaging may function as a timing device for the switch from replication to packaging at late times in infection. For bacteriophage X174, such a component has been identified. This bacteriophage is known to replicate its DNA, present in the capsid as a circular, single-stranded molecule, via different successive mechanisms, before DNA encapsidation can be initiated. In vitro experiments suggest that the viral gene C protein inhibits rolling-circle DNA synthesis and subsequent semiconservative replication of replicative-form DNA by binding to the replication initiation complex. The gene C-associated initiation complex is then able to synthesize single-stranded DNA in tight association with its packaging. In the case of X174, gene C protein not only synchronizes the different stages of DNA replication with DNA packaging but also physically couples the two processes (1). The coupling of events linked to Ad replication with viral DNA packaging is a reasonable possibility in light of studies using mutants in Ad TP (10), the viral terminal protein which is covalently linked to the 5' ends of the genome and directs the priming of viral DNA replication (30). Certain Ad TP mutants are defective for virus growth in vivo (10) yet display wild-type levels of initiation and elongation of viral DNA replication in vitro (9). Such mutants may reveal distinct functions of Ad TP involved in DNA replication and another activity, such as viral DNA packaging.