NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Coffin JM, Hughes SH, Varmus HE, editors. Retroviruses. Cold Spring Harbor (NY): Cold Spring Harbor Laboratory Press; 1997.

Cover of Retroviruses


Show details

Outline of the Integration Process

The integration process will be considered, for the purposes of this chapter, to encompass all the events between completion of viral DNA synthesis and initiation of the expression of the newly integrated provirus. The process is illustrated schematically in Figure 3.

Figure 3. The integration process.

Figure 3

The integration process. The protein components of the preintegration complex are represented by the gray circle surrounding the viral DNA molecule (color). Integrase is represented by the ovoids at the ends of the viral DNA. The host DNA is shown assembled (more...)


The viral DNA molecule at the completion of its synthesis is a blunt-ended linear molecule whose termini, corresponding to the boundaries of the long terminal repeats, are specified by the primers for plus- and minus-strand DNA synthesis. Viral DNA synthesis begins in the cytoplasm of the infected cell and may be completed before (typically, in the case of MLV) or after (typically, in the case of Rous sarcoma virus [RSV]) entry into the nucleus. The linear viral DNA is the proximal precursor to the integrated provirus and is contained in a specific nucleoprotein complex. This preintegration complex is derived in part from the virion core particle and retains a subset of the virion proteins. The preintegration complex probably also contains specific cellular proteins.


Soon after completion of viral DNA synthesis, usually while still in the cytoplasm, a viral enzyme, integrase, cleaves the 3′termini of the viral DNA, eliminating the terminal two (or, rarely, three) bases from each 3′end, The resulting recessed 3′-OH groups provide the sites of attachment of the provirus to host DNA and thus ultimately define the ends of the integrated provirus.


The viral nucleoprotein complex enters the nucleus. This step probably precedes 3′-end processing in the RSV life cycle, and usually follows the end-processing step for MLV. Oncoretroviruses gain access to the nucleus during mitosis, when the nuclear membrane is disassembled. HIV, and probably other lentiviruses, can likewise enter the nucleus during mitosis, but in addition they can enter the nucleus during interphase, by active transport through the nuclear pore, probably mediated by signals in the viral MA protein and Vpr.


Upon entry into the nucleus, the preintegration complex encounters the host DNA. Although specific target sequences are not required for integration, the host genome is not uniformly used as a target. Highly bent DNA sites, such as are found at specific positions in nucleosomes, are strongly preferred. Host-cell DNA-binding proteins may occlude potential target sites, preventing their use. In some cases, cellular proteins that bind to host DNA may be recognized by the viral integration machinery, directing integration to specific sites. Ongoing cellular DNA synthesis or transcription of the target DNA sequences are not required.


Binding of host DNA by the integrase-viral DNA complex is followed by a concerted, integrase-catalyzed reaction in which the 3′-OH groups at the viral DNA ends are used to attack phosphodiester bonds on opposite strands of the target DNA, at positions staggered by four to six bases in the 5′direction, and therefore on the same face of the double helix, separated by the major groove (inset in Fig. 3). In this direct transesterification reaction, the energy of the broken phosphodiester bonds in the target DNA is used for formation of new bonds joining the viral 3′ends to the target DNA.


DNA synthesis, perhaps guided by viral proteins or carried out by the viral reverse transcriptase, extends from the host DNA 3′-OH groups that flank the host-viral DNA junctions, filling in the gaps that flank the viral DNA and displacing the (usually) mismatched viral 5′ends. Following a ligation step, proviral integration is complete. At some point after the initial joining of viral DNA to host DNA, the complex between integrase and the viral DNA ends needs to be dissassembled. Apart from the fact that it occurs rapidly, within minutes after the viral DNA 3′ends are joined to host DNA, nothing is known of the details of this final, gap repair step in the integration process or the mechanism by which the preintegration complex is dissasembled.

Viral proteins that entered the cell in the infecting virion might also have additional roles after completion of the integration process per se (e.g., they might direct assembly of transcription factors or other cellular proteins onto the newly integrated provirus). This conjecture remains to be investigated.

The Precursor to the Integrated Provirus Is in a Nucleoprotein Complex

The viral proteins required for delivery into the nucleus and integration are carried into the cell by the virion, and they appear to remain associated with the intracellular replication intermediates through the steps leading to integration. The linear viral DNA that is the precursor to the integrated provirus can be isolated from infected cells in a specific nucleoprotein complex that is capable of integrating the viral DNA in vitro (Brown et al. 1987; Bowerman et al. 1989; Ellison et al. 1990; Farnet and Haseltine 1991b). Similar complexes are implicated in the integration of the yeast retrotransposon Ty1 (Eichinger and Boeke 1988; Burns et al. 1992). The preintegration complex isolated from cells infected with MLV is large, sedimenting at 160S (Bowerman et al. 1989). Its components include the capsid and the integrase proteins and probably other viral and cellular proteins (Bowerman et al. 1989; Lee and Craigie 1994; Bushman and Miller 1997). The defining characteristic of the preintegration complex, its ability to mediate integration of the viral DNA, implicitly requires integrase. Biochemical evidence for the presence of the capsid protein in the MLV preintegration complex is reinforced by the observation that the sequence of the viral capsid gene determines whether an MLV strain can integrate efficiently in specific mouse strains (see below). In contrast to the MLV preintegration complexes, the corresponding HIV intermediates do not appear to contain the capsid protein (Farnet and Haseltine 1991b; Bukrinsky et al. 1993b; Karageorgos et al. 1993). The HIV preintegration complexes are also large (80–320S) and have been reported to contain the matrix protein, Vpr, and reverse transcriptase, and perhaps other viral and cellular proteins, in addition to integrase (Farnet and Haseltine 1991b; Bukrinsky et al. 1993b; Karageorgos et al. 1993; Kalpana et al. 1994; Farnet and Bushman 1997). Differences in the experimental methods used for isolation and assay of putative HIV preintegration complexes, as well as their inherent limitations, may account for the significant differences in the composition and physical properties of these complexes as reported by different groups. Biochemical evidence for the presence of the matrix and Vpr proteins in the HIV preintegration complex is supported by genetic evidence that these proteins can play a part in mediating entry of a replication intermediate into the nucleus of nondividing cells (Bukrinsky et al. 1993c; Heinzinger et al. 1994; von Schwedler et al. 1994; Gallay et al. 1995). The composition, structure, and metamorphosis of the intracellular forms of an infecting retrovirus remain the least well characterized aspects of the viral life cycle.

Transport of Viral Intermediates from the Cell Periphery toward the Nucleus

Once a retrovirus has entered a cell, it must traverse the cytoplasm, moving from the cell periphery to the nucleus. The mechanism by which this process occurs is completely obscure at present. In principle, this movement could occur by passive diffusion, but biophysical studies of the viscosity and permeability of the cytoplasm suggest that such a mechanism is unlikely (Mastro and Keith 1984; Luby-Phelps et al. 1986, 1987; Luby-Phelps 1994). Macromolecular structures in the expected size range of viral replication intermediates—for example, Ficoll beads 25 nm in diameter—show negligible diffusion in the cytoplasm, presumably reflecting the high vicosity and small effective pore size of the cytoskeletal meshwork (Luby-Phelps et al. 1987). Most intracellular organelles, even vesicles of less than 100 nm in diameter, use cytoskeletal motors to move within the cell (Kelly 1990). There is weak evidence that herpes simplex virus type 1 (HSV-1) nucleocapsids, and perhaps adenoviruses, use tubulin-based motor systems for vectorial transport through the cytoplasm (Dales and Chardonnet 1973; Luftig and Weihing 1975; Kristensson et al. 1986; Topp et al. 1994). The evidence is much stronger for the use, by vaccinia virus and several bacterial intracellular pathogens, of an actin-based motility mechanism for intracytoplasmic transport (Theriot et al. 1992; Cudmore et al. 1995; Sanders and Theriot 1996). Comparable studies have not been reported for any retrovirus, and neither the viral nor cellular components that would presumably mediate intracytoplasmic transport of a retroviral intermediate have been identified. Nevertheless, the distinctive alterations in subcellular localization of viral intermediates that can result from alterations in the viral gag genes suggest that these genes can play a part in normal intracytoplasmic targeting and perhaps active transport of these intermediates (Rhee and Hunter 1987, 1990, 1991; Rhee et al. 1990; Pryciak and Varmus 1992a; Bukrinsky et al. 1993c; Yuan et al. 1993; von Schwedler et al. 1994; Gallay et al. 1995; see Chapter 7).

Nuclear Entry

The target DNA for integration is confined, for most of the cell cycle, to the nucleus of the host cell. Since entry of the infecting retrovirus occurs initially into the cytoplasmic compartment, the virus faces the challenge of delivering the replication intermediate into the nucleus for integration. RSV and HIV, and presumably other viruses, appear able to enter the nucleus before viral DNA synthesis is complete (Lee and Coffin 1991; Miller et al. 1995), although substantial RSV DNA synthesis clearly occurs in the cytoplasm (Varmus et al. 1974). The incompletely synthesized state of most RSV DNA in the cytoplasm, and the linkage between cell division and the completion of viral DNA, raises the possibility that completion of RSV DNA synthesis may be linked to nuclear entry (Lee and Coffin 1991). For MLV, nuclear entry can, and usually does, follow the completion of viral DNA synthesis and 3′-end processing by integrase (Brown et al. 1987, 1989; Fujiwara and Mizuuchi 1988; Roth et al. 1989; Roe et al. 1993, 1997). Nucleoprotein complexes that are competent to carry out integration when assayed in vitro are readily isolated from the cytoplasm of MLV- or HIV-infected cells (Brown et al. 1987; Bowerman et al. 1989; Ellison et al. 1990; Farnet and Haseltine 1990). Similar complexes competent for completion of DNA synthesis and subsequent integration in vitro can be recovered from RSV-infected cells (Lee and Coffin 1990, 1991).

How do these large complexes enter the nucleus? At least two different strategies appear to be employed by different retroviruses. It has been recognized since the earliest studies of retroviral replication that many, and perhaps all, oncoretroviruses depend on an actively dividing host cell for replication (Temin and Rubin 1958; Rubin and Temin 1959; Temin 1967; Humphries and Temin 1972; Fritsch and Temin 1977b; Varmus et al. 1977; Harel et al. 1981; Humphries et al. 1981; Chen and Temin 1982; Hsu and Taylor 1982; Springett et al. 1989; Miller et al. 1990). At least two phases of the viral life cycle, viral DNA synthesis and integration, are restricted by the host cell cycle (see Chapter 4. A detailed investigation of the cell-division requirement for MLV integration has established that the requirement is specifically for mitosis (Hajihosseini et al. 1993; Roe et al. 1993; Lewis and Emerman 1994). In actively dividing cells, the earliest MLV integration events can be detected within 4–7 hours of infection (Pryciak et al. 1992a; Roe et al. 1993). In cells arrested in S phase or at the G2/M boundary, viral entry into cells, uncoating, and DNA synthesis can be completed at a normal rate—within 8 hours—and with essentially normal efficiency (Roe et al. 1993). Moreover, the nucleoprotein complexes that contain the resulting full-length viral DNA molecules are competent for integration when assayed in vitro. But integration fails to occur in the arrested cells, even after more than 20 hours (Roe et al. 1993; Lewis and Emerman 1994). Access to the nuclear DNA that serves as the target for integration appears to be the missing requirement in the arrested cells. In situ hybridization experiments demonstrate that entry of the unintegrated MLV DNA into the nucleus is mitosis-dependent (Roe et al. 1993). Thus, for MLV, and perhaps all oncoretroviruses, it appears that rather than traversing the interphase nuclear membrane, viral replication intermediates gain access to the nuclear DNA when the nuclear membrane is disassembled at mitosis. Structures up to about 26 nM in diameter can be transported through the nuclear pore into the interphase nucleus, in an energy-dependent process, provided they carry nuclear localization signals at a sufficient density on their surface (Dingwall and Laskey 1992). The sedimentation coefficient of the preintegration complexes from either MLV or RSV, roughly 160S, is consistent with a diameter of approximately 30 nM for a spherical nucleoprotein complex (Bowerman et al. 1989; J. Coffin, pers. comm.). This may simply be too large for transit through the nuclear pore complex, forcing the virus to use an alternative strategy for entry. MLV integration does not appear to occur during mitosis, but rather after mitosis is completed (Roe et al. 1993). Thus, access to the nuclear DNA, provided by the breakdown of the nuclear membrane at mitosis, is not in itself sufficient to account for localization of the viral genome to the nucleus after the nuclear membrane reassembles (Swanson and McNeil 1987; Benavente et al. 1989). The mechanism by which the preintegration complex ensures that it is retained in the nucleus when the nuclear envelope is reassembled is not yet known.

For HIV-1, and perhaps other lentiviruses, nuclear membrane disassembly during mitosis provides one route of entry into the nucleus of a dividing host cell. Unlike the oncoretroviruses, however, HIV-1 is able to infect nondividing cells (Weinberg et al. 1991; Lewis et al. 1992). Indeed, such cells (especially macrophages) may provide an important component of the target cell population in lentiviral infections (Hirsch et al. 1991; Meltzer and Gendelman 1992). HIV-1 preintegration complexes can enter the interphase nucleus, apparently via nuclear localization signal-mediated, energy-dependent import through the nuclear pore (Bukrinsky et al. 1992, 1993c; Gulizia et al. 1994; Heinzinger et al. 1994; von Schwedler et al. 1994). Remarkably, HIV appears to have at least two distinct, partially redundant signals, either of which alone can serve to direct transport of replication intermediates through the nuclear pore (Bukrinsky et al. 1993c; Heinzinger et al. 1994; von Schwedler et al. 1994; Freed et al. 1995; Gallay et al. 1995). This redundancy could have made identification of the signal for nuclear entry problematic. Fortuitously, however, initial experiments to test the hypothesis that a putative nuclear localization signal in the matrix protein was essential for infection of nondividing cells were carried out using a viral strain with a truncated Vpr-coding region (Bukrinsky et al. 1993c). Attempts to repeat these experiments using a Vpr+ virus showed that, in the Vpr+ background, the NLS matrix mutations conferred only a partial deficiency in infection of nondividing cells (Heinzinger et al. 1994). Thus, it appears that the necessary signal for entry of the preintegration complex into the interphase nucleus can be provided either by the matrix protein or by the Vpr protein. Even viruses lacking both Vpr and the putative NLS in the matrix protein can still infect dividing cells efficiently (see Bukrinsky et al. 1993c; Heinzinger et al. 1994) and appear to retain residual ability to infect nondividing cells (see Freed et al. 1995).

The ability of the HIV-1 matrix protein to direct nuclear entry of preintegration complexes may be regulated by phosphorylation of its carboxy-terminal tyrosine residue, which becomes phosphorylated in a small fraction of matrix molecules in acutely infected cells. Replacement of this terminal tyrosine with phenylalanine, which cannot be phosphorylated, appears to abolish the ability of matrix to promote nuclear entry in nondividing host cells under some (Gallay et al. 1995), but perhaps not all, conditions (Freed et al. 1997). The role of this potential control mechanism in viral replication remains to be determined.

Many of the key target cells for gene therapy divide infrequently or not at all. The requirement for host-cell division has therefore been a significant obstacle to the use of oncoretroviral vectors for many potential applications of gene therapy (Miller 1992; Morgan and Anderson 1993; Mulligan 1993). Recognition of the key role that nuclear entry has in this cell cycle restriction, and of the apparent ability of the HIV-1 matrix protein to direct HIV intermediates into the nucleus, has prompted experiments aimed at conferring this same ability on oncoretroviral vectors. Although chimeric viruses in which the MLV matrix domain has been replaced by the HIV matrix domain retain vestigial ability to replicate, the chimeras are unable to infect nondividing cells (Deminie and Emerman 1994). It remains to be seen whether the provision of a nuclear targeting signal to an oncoretroviral replication intermediate will be sufficient to overcome this barrier. Efforts to develop lentiviral vectors that can transduce nondividing cells appear to be more promising (Naldini et al. 1996a,b).

Unintegrated Retroviral DNA Is Found in Linear and Circular Forms, but Only the Linear Form Is a Precursor to the Provirus

In addition to DNA synthesis intermediates, four classes of extrachromosomal viral DNA molecules are observed in acutely infected cells (Fig. 4). One, the actual intermediate in replication, is the full-length, linear product of reverse transcription (Brown et al. 1987, 1989; Fujiwara and Mizuuchi 1988). This is the most abundant form of viral DNA in acutely infected cells, particularly at early times after infection, and it is the only form that accumulates in appreciable quantities in the cytoplasm. The other three are circular forms, found in detectable quantities only in the nucleus, and they generally comprise an increasing fraction of the unintegrated DNA molecules at later times after infection (Gianni et al. 1975; Guntaka et al. 1976; Fritsch and Temin 1977a; Shank and Varmus 1978; Yoshimura and Weinberg 1979; Yang et al. 1980b; Roth et al. 1990). The three circular classes are (1) 1-LTR circles, whose structure is consistent with their having been formed by homologous recombination between the two LTRs of a linear viral DNA molecule (Shank and Varmus 1978; Ju and Skalka 1980; Shoemaker et al. 1980; Swanstrom et al. 1981); (2) 2-LTR circles, whose structure is consistent with their having been formed by ligation of the two ends of the linear precursor, often with deletions of a few nucleotides from one or both ends, and occasionally with larger deletions or sequences inserted between the two joined ends (Shank and Varmus 1978; Ju and Skalka 1980; Shoemaker et al. 1980; Donehower et al. 1981; Swanstrom et al. 1981); and (3) “autointegration” products whose structure is consistent with their having been formed by intramolecular integration of the viral DNA ends using the viral DNA molecule itself as a target, resulting in one or two rearranged circular products, depending on the orientation with which the viral DNA ends integrate (Shoemaker et al. 1980, 1981).

Figure 4. Unintegrated viral DNA structures.

Figure 4

Unintegrated viral DNA structures. (A) The linear product of viral DNA synthesis (see Chapter 4 is the precursor to the integrated provirus. (B) 1-LTR circle. This structure is consistent with one that could be formed by homologous recombination between (more...)

None of these circular forms serve as a precursor to the integrated provirus, and none appear to contribute significantly to viral replication. Rather, they all appear to be dead-end by-products of aborted infections. It was once widely believed that the 2-LTR circles were key intermediates in replication and that they served as the proximal precursors to integrated proviruses. The “circle-junction” sequence, formed by the juxtaposed ends of the linear precursor, was proposed to serve as the attachment site for joining the provirus to cellular DNA (Panganiban and Temin 1984a; Duyk et al. 1985). This belief was initially based on circumstantial evidence, including the similar time courses of circularization and integration and the concordant responses of circularization and integration to a variety of experimental manipulations (Guntaka et al. 1975, 1976; Sveda and Soeiro 1976; Shank and Varmus 1978; Jolicoeur and Rassart 1980; Yang et al. 1980a,b). Seemingly decisive support was provided by a report claiming that a synthetic circle junction sequence inserted at an inter- nal site in a modified SNV genome could serve efficiently as an attachment site for SNV integration (Panganiban and Temin 1984a). That dramatic result has never been reproduced, and indeed it has been convincingly discredited (Brown et al. 1987, 1989; Fujiwara and Mizuuchi 1988; Ellis and Bernstein 1989; Lobel et al. 1989). Similar experiments employing viruses carrying internal circle-junction sequences have led to the opposite conclusion—that these sequences cannot serve as sites for efficient joining to cellular DNA (Ellis and Bernstein 1989; Lobel et al. 1989; Panganiban and Talbot 1993). Moreover, when an intermediate in the MLV integration process was isolated from an in vitro integration reaction, its structure precisely matched that of an intermediate predicted to occur during the direct integration of a linear precursor and not during integration of a 2-LTR circle (Fujiwara and Mizuuchi 1988; Brown et al. 1989). Biochemical studies of substrates and intermediates in the integration reaction, using purified components in vitro, have since provided overwhelming evidence that the linear molecule is the direct precursor to the integrated provirus, whereas circle junction sequences appear to be utterly inert as substrates for integration (see below).

With the exception of the autointegration products, the circular molecules do not require integrase for their formation. They are therefore presumed to result from the action of cellular enzymes on viral DNA molecules, perhaps derived from nucleoprotein complexes that are defective for integration. Two observations suggest that these molecules arise as by-products from failures of the integration process: (1) They accumulate in elevated numbers in infections with integration-defective mutant viruses (Donehower and Varmus 1984; Colicelli and Goff 1985, 1988; Roth et al. 1990) and (2) the ter- minal viral DNA sequences comprising the circle junctions of 2-LTR circles frequently show alterations that would be expected to impair their use as substrates for integration (Van Beveren et al. 1980; Kulkosky et al. 1990; Pullen and Champoux 1990; Smith et al. 1990; Whitcomb et al. 1990; Hong et al. 1991; Whitcomb and Hughes 1991; Randolph and Champoux 1993).

The exclusively nuclear location of the 1-LTR and 2-LTR circular forms has been a virtually universal observation among diverse combinations of retroviruses and host cells, so much so that viral DNA circularization has often been used as a surrogate marker for nuclear entry. The basis for this consistent correlation has not been determined. It may reflect a structural change in the viral nucleoprotein complex that occurs only in the nucleus, exposing the viral DNA to the cellular enzymes that mediate circularization, or it may reflect a strictly nuclear localization of the cellular enzymes that mediate circularization (Bowerman et al. 1989; Farnet and Haseltine 1991a,b; Bukrinsky et al. 1993a; Karageorgos et al. 1993).

The failure to observe autointegration products in the cytoplasm is more difficult to rationalize. Viral nucleoprotein complexes found in the cytoplasm of acutely infected cells are active for autointegration as well as intermolecular integration, when tested in vitro (Farnet and Haseltine 1990, 1991a; Lee and Coffin 1990; Pryciak and Varmus 1992a; Lee and Craigie 1994). Access to nuclear DNA has no prima facie role in the autointegration reaction. The lack of autointegration during the often prolonged interval spent by the preintegration intermediates in the cytoplasm therefore presumably points to an active repression of the autointegration activity (and perhaps integration activity generally) that is relieved upon entry into the nucleus, or by the conditions used in in vitro assays. In vitro experiments suggest that a component of the preintegration complex can repress autointegration (Lee and Craigie 1994). The component can be removed or inactivated by treatment with high salt concentrations and restored by fractions from uninfected cells. The mechanism by which autointegration is suppressed in the cytoplasm, and its relief upon nuclear entry, remains an important unsolved problem that may have implications for antiviral therapy. The phenomenon is superficially reminiscent of the “transposition immunity” shown by many prokaryotic transposable elements (Mizuuchi 1992b). These transposons specifically avoid integrating into DNA molecules that contain sequences from their own ends. For bacteriophage Mu, there has been progress toward understanding the biochemical basis of this phenomenon (Adzuma and Mizuuchi 1989).

For ASLV, most of the viral DNA molecules found in the cytoplasm are replication intermediates with incomplete plus strands and, especially at early times after infection, incomplete minus strands as well (Shank and Varmus 1978; Lee and Coffin 1991). Synthesis is completed rapidly when extracts are incubated in vitro. This observation has prompted the suggestion that, in vivo, the completion of ASLV DNA synthesis may require entry into the nucleus (Lee and Coffin 1991). Such a requirement might serve as an alternative mechanism for limiting suicidal autointegration, by avoiding the accumulation of integration-competant viral DNA molecules in the cytoplasm.

Estimates of the efficiency with which newly synthesized viral DNA molecules complete the subsequent steps leading to integration are technically difficult to obtain. However, several studies have suggested that under favorable conditions, approximately 10–50% of the viral DNA molecules synthesized in an acutely infected, permissive cell are ultimately integrated (Varmus et al. 1973; Guntaka et al. 1976; Yang et al. 1980b; Roe et al. 1993; Barbosa et al. 1994).

The frequent observation of high levels of unintegrated DNA in cytopathic infections by HIV (Shaw et al. 1984), feline leukemia virus (FeLV) (Donahue et al. 1991), SIV (Hirsch et al. 1991), SNV (Keshet and Temin 1979), avian leukemia virus (ALV) (Weller et al. 1980), and spumaviruses has led to the suggestion that these DNA molecules might have a role in pathogenesis, perhaps by supporting high levels of viral gene expression. Indeed, it is clear that unintegrated HIV-1 DNA molecules can be transiently transcribed (Engelman et al. 1995; Wiskerchen and Muesing 1995). The weight of the evidence, however (see, e.g., Bergeron and Sodroski 1992), favors the model that the abundant unintegrated DNA is a manifestation of high-multiplicity superinfection, but does not itself play an essential direct part in pathogenesis.

Integration Requires the Virally Encoded Integrase Protein and Specific Sequences at the Ends of the Viral DNA Molecule

To date, all the mutations that lead to a specific block in integration have fallen into two classes: (1) Mutations that alter sequences near the very ends of the linear viral DNA molecule—defining the “attachment sites” (Panganiban and Temin 1983; Colicelli and Goff 1985, 1988; Cobrinik et al. 1987, 1991; Vicenzi et al. 1994) and (2) mutations in the 3′-terminal portion of the pol-coding region, which encodes the integrase protein (Donehower and Varmus 1984; Schwartzberg et al. 1984; Quinn and Grandgenett 1988; Stevenson et al. 1990; LaFemina et al. 1992; Sakai et al. 1993). These genetic results led to the simple model that the protein proteolytically cleaved from the carboxy-terminal portion of Pol polyprotein recognizes specific sequences present at the ends of the viral DNA molecule and promotes their joining to target DNA. This model has been decisively confirmed by direct biochemical evidence. Indeed, as discussed below, these two viral components—the terminal approximately 10–15 nucleotides of the viral DNA ends and integrase—are the only specific viral components that are required to carry out model integration reactions in vitro. The possibility remains open that other viral or cellular proteins, or viral DNA sequences, participate in regulating or enhancing the efficiency of integration in vivo. Indeed, the results of in vitro reconstitution experiments with partially purified MLV or HIV-1 preintegration complexes suggest that, for both viruses, a cellular factor contributes to efficient integration (Lee and Craigie 1994; Farnet and Bushman 1997). Precedents exist for distant DNA sequences having accessory roles in integration of bacteriophages, notably Mu, and in other site-specific recombination reactions (Mizuuchi 1992a). However, most of the retroviral genome can be ruled out as a participant in integration in cis, since defective retroviral genomes used as vectors can be integrated with apparently normal efficiency despite massive deletions of internal viral sequences and substantial deletions from the LTRs (see Chapter 9).

The possible involvement of other viral or cellular proteins in integration is more ambiguous. Despite unequivocal evidence that integrase alone is sufficient to mediate proper integration of model DNA substrates, several observations suggest the possibility that other cellular or viral proteins may have important roles in the in vivo process. First, the efficiency and fidelity of model integration reactions in vitro using purifed integrase have yet to match that of preintegration complexes isolated from infected cells. Under conditions reported to date, the purified enzymes appear to be remarkably inept at coordinating the joint integration of the two ends of the viral DNA (Bushman et al. 1990; Craigie et al. 1990; Bushman and Craigie 1991; Fitzgerald et al. 1992; Vora et al. 1994), whereas under similar reaction conditions, the preintegration complexes assembled in vivo rarely fail to coordinate the integration of the two ends into precisely spaced positions on opposite strands of the target DNA (Fujiwara and Mizuuchi 1988; Brown et al. 1989; Farnet and Haseltine 1990). Second, the optimal conditions for the integration reaction mediated by preintegration complexes differ from those for the reaction carried out by purified integrase. For example, the activity of purified integrase is greatest in the presence of Mn++, whereas the preintegration complexes have equal or better activity in the presence of Mg++. Third, in vivo, the integration of the viral DNA ends into internal sites in the viral genome, a process that can be remarkably efficient in vitro, is completely repressed while the preintegration complexes are in the cytoplasm, and a similar repression can be reproduced in vitro using MLV preintegration complexes and extracts from uninfected cells (Lee and Craigie 1994). Fourth, uninfected cell extracts, or a purified cellular protein, HMG I(Y), can markedly enhance integration activity of HIV-1 preintegration complexes (Farnet and Bushman 1997). Fifth, a functional interaction between the product of a mouse gene, Fv1, and the capsid protein of MLVs, can block integration or a closely linked step (see below).

There is presently no genetic evidence for a direct role in integration for any viral protein other than integrase. The absence of such evidence is inconclusive, however, because all other viral gene products appear to have essential functions that precede integration, in a replication cycle that begins with a provirus. Nonconditional mutations in these genes could therefore arrest the viral life cycle prior to integration, preventing observation of their effects on integration.

Copyright © 1997, Cold Spring Harbor Laboratory Press.
Bookshelf ID: NBK19395