INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

Retroviral integrase (IN) mediates the covalent insertion (integration) of a DNA copy of the viral genome into the host cell DNA, an obligate step in the retroviral life cycle (17). Following virus entry into a cell, reverse transcriptase completes the synthesis of a DNA copy of the viral genome within a viral nucleoprotein complex, also called the preintegration complex (PIC). In first acts within the nucleoprotein complex by mediating an endonucleolytic cleavage at the 3' end of each strand of viral DNA immediately beyond a conserved subterminal CA dinucleotide. This step, called 3'-processing, occurs in the cytoplasm and leaves a terminal hydroxyl group at the 3' end of each strand of viral DNA. After the nucleoprotein complex migrates to the nucleus, IN mediates a concerted nucleophilic attack involving the viral 3' hydroxyl residues and phosphate residues on either side of the major groove in the target DNA, a step termed strand transfer (13, 34). The two viral ends attack the target DNA in a coordinated, 5'-staggered fashion, the extent of the stagger determining the length of the virus-specific direct repeat of host DNA that flanks the integrated provirus. In the initial product of the strand transfer reaction, the gapped intermediate, each 3' end of the viral DNA is attached to the target DNA. The 5' ends of the viral DNA are joined subsequently to host cell DNA through undefined mechanisms. IN can mediate a reversal of the strand transfer event in vitro when supplied with a synthetic gapped intermediate substrate, an activity called disintegration (8), but a role for this function in viral replication has not yet been identified.

Attachment (att) sites, virus-specific sequences located at each end of viral DNA, and IN, the protein encoded by the 3' end of the viral pol gene, are the only viral factors known to be essential for integration (3, 17). In vitro assays, using purified wild-type or mutant HIV-1 IN and synthetic DNA substrates that mimic the viral att sites, have provided much of the information for the currently accepted mechanism of IN activity (5, 21, 24, 41, 47). Coupled with amino acid sequence alignment (16, 20), the in vitro activity data for wild-type and mutant IN proteins have led to a working model of IN with three domains: the amino-terminal or HHCC domain, the core or catalytic domain, and the carboxy-terminal or DNA binding domain (1, 37). The functions of the amino-terminal and carboxy-terminal domains remain unclear, but the amino-terminal domain has been shown to bind zinc (4, 6, 7) and to be involved in IN oligomerization (10, 50), and the carboxy-terminal domain is thought to be involved in sequence-independent DNA binding (12, 26, 39, 45, 48, 49). The function of the core domain is the best understood of the three domains and is the site of IN catalytic activity.

Critical to the catalytic activity of the core domain is the highly conserved D,D(35)E motif found in all retroviral IN proteins and numerous transposable elements. The D,D(35)E motif refers to three absolutely conserved acidic amino acids (two aspartic acids and one glutamic acid) in the order indicated, with a conserved spacing of 35 amino acids between the second and third residues (11, 16, 18). Mutating any of these three conserved residues produces a loss of all three IN activities in vitro, leading to speculation that the triad is essential for a functional catalytic site of IN (6, 11, 20, 25, 43, 46). Consistent with these observations, the core domain has recently been shown to interact with the viral att site and the target DNA (15).

The structure-function data for HIV-1 IN, as outlined above, have been generated by using in vitro assays employing synthetic oligonucleotides that mimic the viral att sites and purified IN protein. To test the current model of IN function in the context of viral replication, we recently characterized the phenotype of a number of HIV-1 virions with point mutations in the IN coding sequence (23). We used a viral construct in which the env gene is largely replaced by a hygromycin resistance gene. This allows for only a single round of infection and for the quantification of infectious titer by selection in hygromycin-containing media. Virus encoding IN protein with an alteration at any one of the three residues of the D,D(35)E motif had a 10³- to 10⁴-fold reduction in infectious titer compared with virus encoding wild-type IN, but each mutant virus remained able to generate low numbers of hygromycin-resistant colonies (23). While the colonies are a measure of stable integration of the incoming viral genome, hygromycin resistance could arise from stable integration of only a part of the viral genome. To determine if the integrated viral DNA resulted from characteristic IN-mediated processes, or from an alternate mechanism, we characterized the integrated viral DNA from a number of hygromycin-resistant colonies generated by virus encoding wild-type IN and by virus encoding mutations at any one of the residues of the D,D(35)E motif, mutants D64V, D116I, and E152G. The results are presented here.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Cell lines. HOS (human osteosarcoma) cells and 293T cells were grown at 37°C in 5% CO₂ in high-glucose Dulbecco's modified Eagle medium (DMEM-21) supplemented with 10% fetal calf serum (Gibco), 100 U of penicillin G sodium per ml, and 0.1 mg of streptomycin sulfate per ml (complete medium).

Virus stocks. All virus consisted of an envelope-deleted HIV-1 genome pseudotyped with amphotropic murine leukemia virus (MLV) envelope, using a previously described system (22, 23, 32). In brief, a hygromycin resistance gene replaces much of the HIV-1 envelope coding sequence, allowing for the clonal expansion of infected cells by selection in hygromycin, thereby providing a means to quantify the number of integration events per unit of virus stock. A functional env gene is supplied in trans by cotransfection of the genome-containing vector with a vector expressing the MLV amphotropic envelope protein. The HIV-1 genome of the virus stocks encodes either a wild-type IN protein or IN with a mutation at one of the three residues of the D,D(35)E motif, D64V, D116I, or E152G. Virus stocks were generated by calcium phosphate transfection of 293T cells at 50% confluence, using 10 µg of envelope-expressing plasmid DNA and 10 µg of HIV-1 genome-containing plasmid DNA. Transfected DNA was removed from the plates after 8 to 12 h, and cells were fed with 20 ml of complete medium. Culture supernatant was collected 48 h later and filtered through a 0.2-µm Millex-GV syringe filter (Millipore) to generate virus stock. Virus was stored frozen at 70°C for later use. Construction of the proviral clones with each of the IN mutations was described previously (23).

Southern blots. HOS cells were infected at very low multiplicity of infection, 10 to 20 infectious particles per 100-mm-diameter plate, using virus pseudotyped with the MLV amphotropic envelope protein and expressing either a wild-type IN protein or IN mutant D64V, D116I, or E152G. Clones of infected cells were generated by selection in complete medium containing 200 µg of hygromycin (Boehringer Mannheim) per ml, isolated by using cloning cylinders, and expanded in hygromycin selection. DNA was isolated from a confluent monolayer of cells in a 100-mm-diameter dish by using DNAzol (Gibco BRL), with one modification to the manufacturer's recommendations: we added 1 µl of RNase (5 µg/µl) to the lysate and incubated the mixture at 37°C for 15 min prior to precipitation with ethanol. The DNA pellet was resuspended in water. For each clone, 10 µg of DNA was digested for 3 h at 37°C in a 40-µl reaction using buffers recommended by the manufacturer (New England Biolabs), followed by electrophoresis on a 0.9% Tris-borate-EDTA: agarose gel at 35 V for 12 h. The Southern blots were prepared by using alkaline denaturation of the DNA prior to transfer and Hybond-N⁺ (Amersham) filters during the transfer, and hybridizations were performed overnight in Church buffer (2). The probe for the central region of the viral genome was made from a 534-bp fragment of the hygromycin resistance gene (Fig. 1A). Southern blots for determining the integrity of the left and right ends of the integrated viral genomes were performed with DNA from each clone digested with EcoRV and ClaI and with BamHI and HindIII, respectively (Fig. 1B). All probes were generated by using the Ready To Go labeling beads (Pharmacia) random primer method and [-³²P]CTP (Amersham).

View larger version (16K): [in this window] [in a new window]   FIG. 1.   Southern blot analysis of integrated HIV-1 DNA demonstrates intact viral genomes. The integrated proviral HIV-1 DNA is depicted with the LTRs flanking the gag, pol, and hygromycin (hygro) genes. The numbers 1 and 9718 indicate the terminal nucleotide residues of the HXB2 genome. Numbers below the restriction enzymes (KasI, XhoI, EcoRV, ClaI, BamHI, and HindIII) indicate the nucleotide positions of the respective cleavage sites in the HXB2 genome. Probes are indicated by a boxed region and correspond to a 534-bp sequence from the middle of the hygro gene (A) and HXB2 sequence from positions 641 to 828 (left end) and 8604 to 8895 (right end) (B). Horizontal arrows and their associated numbers indicated the expected sizes of restriction fragments for an intact proviral genome. The Southern blot analysis proceeded in two steps, first using a hygromycin probe to detect an intact genome between the two LTRs (A). The second step used individual probes for each end (B) to check for intact sequence to within 112 bp of the left end and to within 103 bp of the right end.

Inverse PCR and DNA sequencing. Independent digestions were performed for each viral end, using restriction enzymes with known restriction sites internal to each end (Fig. 2). Twelve micrograms of DNA from each clone was digested with NsiI, PstI, SpeI, or SphI for the left end and with XhoI, BamHI, or BlpI for the right end. Ten micrograms of the digested DNA was used for Southern blot hybridization, and the remaining 2 µg was saved for ligations in a later step (Fig. 2). Samples that generated a fragment size of less than 3 kb following Southern blot hybridization were used for inverse PCR. First, 500 ng of the digested DNA was ligated in a 50-µl volume, using 1 U of T4 DNA ligase (Gibco-BRL), buffer provided by the manufacturer, and incubation at 16°C for 12 h. Ligations were heated at 65°C; the DNA was precipitated, washed twice with 70% ethanol, and resuspended in 10 µl of water. The resuspended, ligated DNA (500 ng) was amplified in a 25-µl reaction mixture containing 60 mM Tris HCl (pH 9.5), 15 mM (NH₄)₂SO₄, 2.5 mM MgCl₂, 10 pmol of each divergently placed oligonucleotide (one within the long terminal repeat [LTR] and another outside the LTR; Fig. 2), 160 µmol of each deoxynucleoside triphosphate, and 2.5 U of Taq polymerase (Perkin-Elmer). Primers corresponded to HXB2 nucleotide positions 163 to 146 and 641 to 659 for the left end and nucleotide positions 9055 to 9031 and 9580 to 9600 for the right end. Forty cycles of PCR were performed (94°C for 1 min, 55°C for 1 min, and 72°C for 3 min), followed by an extension at 72°C for 8 min, using an M.J. Research thermocycler. Then 2 µl of a 1/10 dilution of the product of the inverse PCR was used as a template for second (nested) PCR, using a second set of oligonucleotides located internal to the first set of oligonucleotides (Fig. 2). Oligonucleotides for the nested PCR corresponded to HXB2 nucleotide positions 93 to 73 and 945 to 965 for the left end and nucleotide positions 8908 to 8888 and 9646 to 9663 for the right end. When XhoI was used to digest the DNA for use in inverse PCR, the nested PCR step used a primer corresponding to nucleotide positions 8978 to 8956 instead of the one corresponding to nucleotide positions 8908 to 8888. Nested PCR conditions were identical to those described above, and 15 µl of the nested PCR product was electrophoresed on a 0.9% Tris-borate-EDTA-agarose gel. For DNA that yielded a clean, discrete PCR product of a size predicted from the Southern blot, the PCR product was sequenced by using a Sequenase PCR product sequence kit (U.S. Biochemical; Amersham) according to the manufacturer's protocol. In brief, 5 µl of nested PCR product from each clone was treated with 1 µl each of exonuclease 1 (10 U) and shrimp alkaline phosphatase (2.0 U). The mix (7 µl) was incubated at 37°C for 15 min. The enzymes were inactivated by heating to 80°C for 15 min. To the treated PCR products, 7.5 pmol of primer was added in a final volume of 10 µl. The reaction mixture was heated at 100°C for 4 min and immediately cooled by placing the vial in ice. The labeling reaction was performed as recommended by the manufacturer, using [³⁵S]ATP. The samples were electrophoresed on 6% denaturing gels. Autoradiographs were developed after 72 h at 70°C.

View larger version (23K): [in this window] [in a new window]   FIG. 2.   Inverse PCR approach to clone and sequence the host-virus junctions. DNA isolated from each clone was digested with restriction enzymes selected for their proximity to one or the other LTR. Restriction enzyme digestions and PCR were specific for each viral end, allowing each viral end and its flanking DNA to be sequenced independently. Digested DNA was diluted to favor intramolecular ligations. A set of divergently oriented primers (primers A and B), unique for each viral end, was used in inverse PCR to amplify the left and right virus-host junctions. Primer B is oriented with its 3' end directed away from the virus-host junction, and primer A is oriented with its 3' end oriented toward the virus-host junction. Inverse PCR was followed by nested PCR using a second set of primers (A' and B') unique for each end, to produce a well-defined band for DNA sequencing. Southern blotting performed after the digestion step was used to predict the final PCR fragment size, providing an internal control for the validity of the final PCR product.

IN sequence determination. For each mutant, the DNA sequence of the IN coding region from two hygromycin-resistant clones was determined by using Sequenase as instructed by the manufacturer (U.S. Biochemical). DNA for sequencing was obtained by using primers that anneal to the amino- and carboxy-terminal sequences of wild-type IN, using methods identical to those used above. Five microliters of the nested PCR product was used for sequencing.

Native target site sequence determination. For clones in which we successfully used inverse PCR to determine both the right and the left virus-host DNA junctions, we selected primers that would anneal to the host DNA flanking the provirus. We chose sequences that were 100 to 150 bp from the host-virus junction to generate an expected PCR product of 200 to 300 bp in length. The amplifications were performed in 25-µl reaction volumes containing 60 mM Tris HCl (pH 9.0), 15 mM (NH₄)₂SO₄, 2.0 mM MgCl₂, 10 pmol of each oligonucleotide, 200 µmol of each deoxynucleoside triphosphate, 2.5 U of Taq Polymerase (Perkin-Elmer), and 50 ng of genomic DNA isolated from uninfected HOS cells. Cycling parameters included three cycles at 95°C for 5 min, 55°C for 45 s, and 72°C for 1 min, followed by 30 cycles at 95°C for 45 s, 55°C for 30 s, and 72°C for 30 s, with a final 8-min extension at 72°C, using an M.J. Research thermocycler. Two microliters of a 1/10 dilution of the PCR product was used as a template for a second PCR using the conditions described above except that the initial three cycles with a 5-min melting was not performed. Fifteen microliters of the PCR product was electrophoresed on a 0.9% Tris-acetate-EDTA-agarose gel; 5 µl was used for sequencing as described above.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Integrated viral DNA is intact and colinear with the incoming viral genome. We previously used replication-defective HIV-1 to study the effects of IN mutations on viral integration (23). Each viral genome contains a hygromycin resistance gene, allowing for the selection of cells with a stably integrated provirus. Mutations at any of the three residues in the D,D(35)E motif of the IN core domain resulted in a 3- to 4-log reduction in infectious titer relative to virus encoding wild-type IN. None of the mutations, however, prevented entirely the formation of hygromycin-resistant colonies: wild-type virus yielded 8.5 × 10⁵ hygromycin-resistant colonies/µg of p24; virus containing the single IN mutation D64V, D116I, or E152G yielded 90, 260, or 60 hygromycin-resistant colonies/µg of p24, respectively (23). While it is possible that each hygromycin-resistant colony represents an IN-mediated integration event, hygromycin resistance requires only the stable insertion of the 1,582-bp hygromycin resistance gene cassette that has its own simian virus 40 promoter. The latter could occur through mechanisms other than IN-mediated integration.

View larger version (78K): [in this window] [in a new window]   FIG. 3.   DNA sequence at the host-virus junction for proviral clones derived from virus encoding wild-type or mutant IN. The line drawing shows a provirus with its LTRs flanked by host chromosomal DNA (wavy lines). Sequence for the top strand of viral DNA (HXB2), flanked by arrows representing host DNA, is depicted immediately below the line drawings. DNA sequences from clones derived from wild-type IN and from each of the three mutant IN sequences (D64V, D116I, and E152G) are grouped, separated by a bold horizontal line. Since sequence for only the top strand of the viral DNA is presented, the terminal CA dinucleotide (underlined) is seen at the right (3') end of the viral sequence, while a complementary GT dinucleotide (underlined) is seen at the left (5') end of the viral DNA. The dashed vertical lines delineate the 5 bp of flanking host DNA, and those sequences are in an enlarged font. For each viral end, 50 to 150 bp of flanking DNA sequence was determined, but only 18 bp from each end are shown. Viral sequences that diverge from the original (HXB2) DNA sequence have a thick horizontal line above the mutant sequence. Missing viral sequence is indicated by a dash.

Integrated viral genomes are rarely flanked by a 5-bp direct repeat of host DNA. In addition to providing viral DNA sequence, the inverse PCRs allowed us to determine the sequence of host DNA flanking the integrated proviruses. HIV-1 proviruses are characteristically flanked by a 5-bp direct repeat of host cell DNA, the consequence of a 5-bp staggered cleavage of host DNA during IN-mediated strand transfer. As expected, we found perfect 5-bp direct repeats of host DNA flanking six of seven wild-type clones for which we have DNA sequence at both junctions (Fig. 3). The seventh clone, WT-15, has a 5-bp direct repeat of flanking host DNA except for a single nucleotide change (Fig. 3). In contrast, a 5-bp direct repeat of host DNA is found in only 1 of the 11 mutant clones for which we have DNA sequence for both the right and the left virus-host DNA junctions (Fig. 3, E152G-4). Two other clones show alternate-length direct repeats of flanking host DNA: D116I-14 has a 13-bp direct repeat flanking the provirus, and E152G-12 has a near-perfect 17-bp direct repeat flanking the provirus (Fig. 3). Eight of the 11 mutant clones for which we have DNA sequence at both the right and the left virus-host DNA junctions lack any direct repeat in the flanking DNA.

Strand transfer often produces a small deletion of target DNA. To further investigate the nature of the strand transfer reaction, we cloned and sequenced the original target sites for four of the eight clones for which we have DNA sequence at both the right and left virus-host DNA junctions and that lack a direct repeat of flanking DNA (Fig. 4). The DNA sequence at each insertion site matched perfectly with the flanking sequences identified by inverse PCR cloning for each provirus, independently validating the host DNA sequences in Fig. 3. All four host target sites surprisingly underwent a deletion during the viral integration process, with deletions ranging from 6 to 11 bp in length (Fig. 4). The 5-bp direct repeat of host DNA normally flanking HIV-1 proviruses is due to the left and right ends of the incoming virus ligating to the host DNA in a 5'-staggered fashion. A deletion of target DNA could occur if an integration event involved a 3'-staggered cleavage of the target DNA, as opposed to the normal 5'-staggered cleavage, as explained in the discussion section. Target-site DNA sequence for D116I-9 revealed that 7 bp immediately flanking the left viral end, and 5 bp immediately flanking the right viral end, are not of host site origin (Fig. 3 and 4). These bases are indicated above the target site DNA sequence in Fig. 4. Possible origins of this extra DNA are addressed below.

View larger version (24K): [in this window] [in a new window]   FIG. 4.   Host DNA sequence at the integration sites of virus encoding a mutant HIV-1 IN. DNA sequence of the host cell genome at the integration site for each indicated clone is presented. Boxed sequence indicates host DNA deleted during the integration event, and underlined sequence to each side of the box is the host DNA immediately flanking the integrated viral DNA for each clone, matching the flanking DNA in Fig. 3. The deletion at the integration site of each of these clones, ranging from 6 to 11 bp in length, contrasts with the 5-bp direct repeat of host DNA that flanks proviruses generated with wild-type IN (Fig. 3). DNA immediately flanking the D116I-9 integrated provirus, 5 bp on the right and 7 bp on the left, are not of target site DNA origin. These bases are in italics above the target site DNA.

Target site selection may not be random for virus with mutations in the D,D(35)E motif. While the location of wild-type retroviral integration within the host cell genome is not a random event, DNA structural characteristics, and not nucleotide sequence, appear to be the primary determinants that influence target site selection (19, 28, 29, 35, 36, 38, 40, 42, 44). Consistent with this, the flanking host DNA from our wild-type clones shows no sequence preference (Fig. 3). That is, there is no nucleotide sequence preference immediately adjacent to the conserved viral CA dinucleotide at each end of the integrated proviral DNA. The mutant clones, however, demonstrate a striking preference (17 of 19 clones [89%]) for a G residue immediately adjacent to the conserved viral CA dinucleotide at the left viral end (Fig. 5). Twelve of the 19 clones (63%) also have a T residue 2 bp from the CA dinucleotide at the left viral DNA end (Fig. 5). Similarly, a G was found immediately adjacent to the right viral end in 10 of 15 (67%) clones. If we exclude the clones with truncated right viral ends, because they may integrate via mechanism different than that used by clones with intact ends, a G is present immediately adjacent to the viral CA in 9 of 11 (82%) of the clones. Of these nine clones, six (67%) have an adjacent T residue (Fig. 3 and 5).

View larger version (25K): [in this window] [in a new window]   FIG. 5.   Host DNA sequence flanking integrated viral DNA from virus encoding mutant IN protein is not random. Shown is the frequency of each nucleotide in the host DNA at the five positions immediately flanking the 3' end of each strand of DNA. That is, the host DNA sequence is adjacent to the top strand of viral DNA on the right but from the bottom strand of viral DNA on the left. This contrasts with Fig. 3, in which both the right and the left flanking host DNA sequences are adjacent to the top strand of the integrated viral DNA. For sequence to the left of the provirus, host DNA at 1 is immediately adjacent to the provirus and 5 is located 5 bp from the provirus; for the right side, host DNA at +1 is immediately adjacent to the provirus and +5 is located 5 bp from the provirus. A host G residue is disproportionately represented immediately flanking each end of the provirus, and there is a tendency for a T at positions 2 and +2.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

We previously reported that mutations at any one of the three phylogenetically conserved acidic residues of the D,D(35)E motif of the HIV-1 IN catalytic domain produce a 3- to 4-log reduction in infectious titer (23). Here we characterize the integrated viral DNA and flanking host DNA to better understand the nature of the integration events mediated by these mutants. In all cases, an intact or nearly intact provirus was found in the target cell DNA. However, one or more aspects of the integration processes is aberrant in each clone, including a failure to maintain the characteristic 5-bp direct repeat of flanking host DNA, the production of small deletions in the target DNA, the integration of viral DNA that has not undergone 3'-processing, and/or a failure to maintain a viral CA dinucleotide at the 3' end of each strand of viral DNA. D,D(35)E motif mutations also demonstrate sequence-dependent target site selection, having a bias toward integration at a host G residue. We conclude that the integrated viral DNA from D,D(35)E mutants does not arise via IN-mediated strand transfer. The residual infectious titer of the mutant virions is therefore due to processes that are not mediated by the retroviral IN protein.

Integrated viral DNA from D,D(35)E mutants is largely intact. The series of Southern blots shown in Fig. 1 and the sequence data in Fig. 3 show that mutations in the D,D(35)E motif do not prevent the integration of an intact HIV-1 genome. There is an occasional loss of five to eight terminal nucleotides, at the right end more often than at the left end. The intact structure of the integrated DNA suggests that it is able to serve as a template for the production of new viral particles. We are currently addressing this possibility.

Some proviruses are flanked by a direct repeat of target DNA. All clones derived from virus encoding wild-type IN have a perfect or near-perfect 5-bp direct repeat of host DNA flanking the provirus. The 5-bp direct repeat is an obligate consequence of the HIV-1 IN-mediated strand transfer step in which the 3' end of the viral DNA is involved in a 5'-staggered cleavage of target DNA. Only three proviruses generated from virions encoding a mutation at any of the three residues of the D,D(35)E demonstrate a direct repeat of host flanking DNA, with only one, clone E152G-4, having the 5-bp direct repeat seen with HIV-1 integration events. Notably, E152G-4 has a mutation in the highly conserved CA dinucleotide at one viral end. The extensive in vitro data supporting the critical importance of an intact terminal CA dinucleotide (5, 21, 24, 41, 47), and the observation that a mutation in one att site prevents IN-mediated activity at both att sites (31), lead us to conclude that this clone did not arise via an IN-mediated process. The two other clones with a flanking direct repeat are D116I-14, with a 13-bp direct repeat of host DNA, and E152G-12, with a near-perfect 17-bp direct repeat of flanking host DNA. The lengths of the direct repeats do not support an HIV-1 IN-mediated recombination mechanism for these two clones. The exact mechanism responsible for the flanking direct repeats in these three clones cannot be determined from our data.

Strand transfer frequently produces a deletion of target DNA at the site of integration. We sequenced the target site DNA for four of the eight clones lacking a flanking direct repeat of host DNA and found that each integration event produced a 6- to 11-bp deletion of target DNA (Fig. 4). Target DNA deletion occurring during the recombination event is incompatible with an IN-mediated process. If the mechanism involves the primary linkage of the viral 3' ends with the target DNA, it must have occurred with a 3'-staggered cleavage of the target DNA. The 3' overhang of target DNA produced by such an event would presumably be excised by host enzymes during repair of the gapped intermediate, in contrast to the filling-in reaction that normally repairs the gapped intermediate following the 5'-staggered cleavage of target DNA mediated by wild-type IN. Alternatively, the viral DNA could be randomly ligated to target DNA by host enzymes at sites of DNA nicks or breaks, with successful integration requiring only that the two break sites be close enough to accommodate physical limitations of viral end movement imposed by the PIC. Such a process could explain both the variable deletions and the variable direct repeats of host DNA at the integration sites.

Target site selection is not sequence independent. Target site selection by retroviruses is not random (19, 28, 38, 40, 42, 44), but structural aspects of the target DNA, and not DNA sequence, appear to be most critical for determining if a given stretch of DNA will be the site of an integration event (30, 35, 36). All three of our mutant viruses demonstrated a marked preference for a G residue immediately adjacent to the conserved CA dinucleotide found at the 3' end of each strand of viral DNA, along with a tendency for the G to be followed by a T residue (Fig. 5).