In 1964, Howard Temin proposed an explanation for the stable, heritable phenotypic changes in cells following retroviral infection of cells, based on results of his experiments comparing hybridization of RSV RNA to DNA from infected versus uninfected cells. “. . . It seems the simplest hypothesis to explain these data is that the provirus of Rous sarcoma is a region of DNA homologous with viral RNA stably integrated into the molecules of cellular DNA in the nucleus” (Temin 1964). Evidence supporting this bold hypothesis accumulated during the years that followed, culminating in the molecular characterization of the integrated proviral DNA joined to target DNA sequences (Hughes et al. 1978, 1981; Dhar et al. 1980; Ju and Skalka 1980; Shimotohno et al. 1980; Hishinuma et al. 1981; Majors and Varmus 1981; Majors et al. 1981).
Integration is an essential step in the life cycle of most, and probably all, retroviruses. Mutations that appear to interfere specifically with integration block replication of murine leukemia virus (MLV), human immunodeficiency virus type 1 (HIV-1), avian sarcoma/leukemia virus (ASLV), and spleen necrosis virus (SNV) (Donehower and Varmus 1984; Panganiban and Temin 1984b; Schwartzberg et al. 1984; Colicelli and Goff 1985; Donehower 1988; Quinn and Grandgenett 1988; Goff 1992; LaFemina et al. 1992; Sakai et al. 1993; Cannon et al. 1994; Vicenzi et al. 1994; Engelman et al. 1995; Englund et al. 1995; Wiskerchen and Muesing 1995). There is no well-documented example of retroviral replication in the absence of integration. Isolated reports suggesting that replication of SNV, simian immunodeficiency virus (SIV), or visna virus might occur in the absence of integration have not been reproduced (Panganiban and Temin 1983; Harris et al. 1984; Prakash et al. 1992; Panganiban and Talbot 1993).
Integration contributes to viral replication in two important ways. First, since retroviral DNA molecules are not ordinarily able to replicate autonomously as episomes, they depend on integration for stable maintenance in dividing cells. Once integrated, however, the provirus is replicated along with host-cell DNA and genetically transmitted as an integral element of the host genome. Integration also stabilizes the viral DNA against degradation: Unintegrated DNA molecules in an acutely infected cell are degraded by unknown processes within hours to days (Donehower and Varmus 1984; Kim et al. 1989; Pauza 1990; Roe et al. 1993; Barbosa et al. 1994; Pauza et al. 1994). Second, integration is important for efficient transcription of viral DNA into new copies of the viral genome and mRNAs that encode viral proteins. Thus, integration defines a turning point in the life cycle at which the virus can begin to multiply.
The structure of the integrated provirus is precisely defined and uniform for each viral strain (Fig. 1). First, the integrated provirus is colinear with the final linear DNA product of viral DNA synthesis, except for the loss of two base pairs from each end (Hughes et al. 1978, 1981; Dhar et al. 1980; Shimotohno and Temin 1980; Shimotohno et al. 1980; Majors and Varmus 1981). Second, the junctions between viral and host-cell DNA are always precise with respect to viral DNA sequences—a phylogenetically conserved CA dinucleotide, typically located two bases internal to each 3′ end of the unintegrated linear precursor, defines the 3′ ends of the integrated provirus (Dhar et al. 1980; Shimotohno et al. 1980; Shoemaker et al. 1980; Van Beveren et al. 1980; Hughes et al. 1981; Lerner et al. 1981; Majors and Varmus 1981). Third, integration is accompanied by duplication of a short sequence from the target site, which flanks the integrated provirus as a direct repeat of 4–6 bp (Dhar et al. 1980; Ju and Skalka 1980; Shimotohno et al. 1980; Van Beveren et al. 1980; Hishinuma et al. 1981; Hughes et al. 1981; Majors and Varmus 1981; Ju et al. 1982). The length of the direct repeat is determined by the virus, not the host cell, an early clue that the integration process was mediated by a viral enzyme. These structural features strongly suggested that integration occurred by the general mechanism for transposition that was outlined by Shapiro (1979). The key features of such a mechanism are that a specific site on one strand from each end of the transposable element is joined to each strand of the target DNA, at sites that are staggered in the 5′ direction by a number of bases that correspond to the length of the short flanking repeats. This results in an intermediate with gaps flanking the inserted element. Repair of the gaps generates the flanking repeats. Direct analysis of integration intermediates, described below, has established the validity of this general model. The ability to study integration in vitro using viral DNA-protein complexes from infected cells (Brown et al. 1987) or, in a more limited way, purified components (Craigie et al. 1990) has filled in many mechanistic details.
Once the integration process is completed, it is irreversible. Although proviruses can undergo partial deletion by homologous recombination between the two long terminal repeats (LTRs) (Varmus et al. 1981; Seperack et al. 1988; Stoye et al. 1988) and could presumably undergo complete elimination by gene conversion in a cell hemizygous for the provirus, these events are rare, and they do not appear to occur at a rate higher than the rate observed for nonviral sequences with the same salient characteristics. There is no evidence for any specific mechanism for excision of the provirus. However, the occurrence of precise excision of a Drosophila retrotransposon, apparently not by a gene-conversion mechanism (Kuzin et al. 1994), argues that the rare occurrence of a similar excision process for retroviruses, although unlikely, cannot be dismissed out of hand.
Retroviral integration is inherently mutagenic. Tumors resulting from dominant insertional mutations that altered the expression pattern of a proto-oncogene were among the earliest-recognized pathological sequelae of a retroviral infection (Fig. 2A) (Hayward et al. 1981; Neel et al. 1981; Nusse and Varmus 1982; Fung et al. 1983; Varmus 1983; Nusse et al. 1984) (see Chapter 10. Recessive mutations caused by natural or experimental retroviral integration events are less readily observed, yet many have been reported (Jenkins et al. 1981; Varmus et al. 1981; Schnieke et al. 1983; Frankel et al. 1985; Jaenisch et al. 1985; King et al. 1985; Soriano et al. 1987; Stoye et al. 1988; Ben-David et al. 1990; Chang et al. 1993; Friedrich and Soriano 1993; Hubbard et al. 1994) (see Chapter 8. Several familiar mutations in the mouse have been shown to be the result of retroviral insertions (Fig. 2B) (Jenkins et al. 1981; Stoye et al. 1988; Cachon-Gonzalez et al. 1994). Experimentally, retroviral insertion mutations have been identified and isolated using heterozygous or hemizygous target genes (Varmus et al. 1981; Frankel et al. 1985; King et al. 1985; Hubbard et al. 1994); by using enhancer or promoter trap vectors, which can allow the expression pattern of the mutagenized gene to be recognized by its effects on expression of a marker gene in the vector that gives rise to a visible product (Gossler et al. 1989; Friedrich and Soriano 1991, 1993; Reddy et al. 1992; von Melchner et al. 1992; Chang et al. 1993); or by crossing mice heterozygous for a particular proviral insertion to obtain homozygotes (Schnieke et al. 1983; Jaenisch et al. 1985; Soriano et al. 1987; Friedrich and Soriano 1991, 1993).
The mutations that result from proviral insertions can have consequences for evolution. An amusing example is the insertion of a gibbon ape leukemia virus (GALV)/MLV-related provirus upstream of a duplicated pancreatic amylase gene in the genome of a common progenitor of the great apes and Old World monkeys, giving rise in its descendants to expression of amylase in the parotid gland (Samuelson et al. 1990; Ting et al. 1992; Meisler and Ting 1993). By introducing amylase into saliva, this retroviral insertional mutation may have influenced the dietary preferences of apes, perhaps even accounting in part for the importance of starchy foods in the human diet (Fig. 2C) (see Chapter 8.
It is unclear at present whether the limited target specificity observed for retroviral integration serves to minimize the potential for deleterious genetic consequences, perhaps by directing integration preferentially to nonessential sequences (King et al. 1985; Shih et al. 1988; Sandmeyer et al. 1990; Craigie 1992; Kitamura et al. 1992; Pryciak et al. 1992a; Pryciak and Varmus 1992b; Withers-Ward et al. 1994). There is no systematic evidence for preferential integration into nonessential sequences by any retrovirus, however, and anecdotal evidence suggests that some important genes are used more frequently (Hubbard et al. 1994) while other genes are used less frequently than expected as integration targets (Frankel et al. 1985; King et al. 1985). The integration specificity of Ty1, Ty2, and Ty3 in yeast clearly shows a considerable bias toward nonessential sequences (Chalker and Sandmeyer 1990, 1992, 1993; Sandmeyer et al. 1990; Xu and Boeke 1990; Craigie et al. 1991; Garfinkel and Strathern 1991; Ji et al. 1993; Menees and Sandmeyer 1994; Smith et al. 1995). The absence of an efficient mechanism for horizontal transmission of these retroelements, in contrast to retroviruses, and the ability of these Ty elements to transpose in haploid cells, presumably led to stronger selective pressure on these transposable elements toward selection of innocuous integration sites (Craigie 1992). Because, at any rate, retroviral integration can occur at substantial frequency into important or essential genes, it provides a valuable experimental tool. Indeed, the potential for indiscriminate mutagenicity is a major consideration in the risk-benefit analysis of retroviruses as vectors for gene therapy (Miller 1992; Morgan and Anderson 1993; Mulligan 1993), providing impetus for efforts to confer target specificity on retroviral integrases (Bushman 1994; Goulaouic and Chow 1996; Bushman and Miller 1997).
Retroviruses are the only animal viruses that depend for replication on integration of the viral genome into the DNA of the host cell, or encode an enzyme whose specific function is to mediate integration, or integrate their genomes into host DNA with high efficiency. Integration of viral DNA is not unique to retroviruses, however; other animal virus genomes are occasionally integrated (Botchan et al. 1976; Handa et al. 1977; Cheung et al. 1980; Yaginuma et al. 1985). For most of them, integration appears to be an infrequent aberration with no important role in the viral life cycle. Parvoviruses may be a partial exception to this generalization. For adeno-associated virus (AAV), integration may represent a significant alternative replication pathway in vivo, used when the helper virus required for its replication is unavailable at the time of initial infection (Handa et al. 1977; Cheung et al. 1980; Berns and Linden 1995). Although it is not strictly required for integration, a protein encoded by the AAV rep gene appears to increase the efficiency of AAV integration and confers upon it a striking target site specificity (Kotin et al. 1990; Giraud et al. 1994; Weitzman et al. 1994).
Integration of viral DNA into the host-cell chromosome is also a frequent feature of the replication strategies of bacteriophages. For some bacteriophages, like bacteriophage Mu, integration is essential for replication (Ljungquist and Bukhari 1977; Ljungquist et al. 1979). The well-studied bacteriophage systems, especially bacteriophage Mu, as well as nonviral transposons, have provided useful models for thinking about the mechanism and biology of retroviral integration (Shapiro 1979; Howe and Berg 1989; Mizuuchi 1992b; Polard and Chandler 1995).
Cold Spring Harbor Laboratory Press, Cold Spring Harbor (NY)
Brown PO. Integration. In: Coffin JM, Hughes SH, Varmus HE, editors. Retroviruses. Cold Spring Harbor (NY): Cold Spring Harbor Laboratory Press; 1997.