Genetic footprinting of a retroviral Gag gene suggests an important role in virus replication

Aretrovirus particle is a membrane-bounded sphere ≈120 nm in diameter. The principal constituent of such a particle is the Gag protein, and expression of this protein in a eukaryotic cell, even in the absence of other viral components, leads to assembly and release of virus-like particles. After release of the particle, the Gag protein molecules are cleaved by the viral protease (PR) into a series of cleavage products; this set of cleavages, collectively termed “maturation” of the particle, must occur before the particle becomes infectious. Gag proteins are always cleaved into at least three products, termed (from N to C terminus) matrix (MA), capsid (CA), and nucleocapsid (NC); additional cleavage products are frequently formed, but these are not conserved between different retroviral genera.

Although it is obvious that Gag plays a critical “structural” role in particle formation, it is now becoming clear that its cleavage products also perform important functions when a mature particle infects a new host cell. This is one of the principal implications of a new mutational analysis by Auerbach et al. (ref. 1, which appeared in a recent issue of PNAS) of an ≈600-nt stretch of the Gag coding region of Moloney murine leukemia virus (MLV), a prototype member of the Gammaretrovirus genus of retroviruses. The article presents a technical tour de force: a scan of hundreds of positions within this stretch, testing whether insertion of a 12-aa sequence at any given position interferes with the ability of the virus to replicate.

The basic strategy used in this study was developed by Singh et al. (2) several years ago. A large pool of mutants is first prepared by incubating the target DNA with an appropriate double-stranded oligonucleotide and MLV integrase. The integrase inserts the oligonucleotide, leading to the addition of 36 bases to the target DNA, at positions that are, to a first approximation, random. Positions that contain inserts are identified by PCR, using one primer from the oligonucleotide and one from the target DNA. In the present experiments, a high degree of “coverage” of the target DNA was ensured by preparing a pool with an average of 400 insertions at each position. The mutants in the pool are then subjected to a test, in this case, the ability to infect and replicate in permissive cells. DNA is then prepared en masse from those viruses that have successfully replicated, and positions with inserts in this population of replication-competent viruses are identified by PCR as above. Any position at which there is an insert in the starting pool of mutants, but no insert in the population prepared from the replicating viruses, is inferred to be a site at which the insertion interfered with viral replication. Blocks of such sites are referred to as “genetic footprints.”

The great power of this approach is, of course, the ability to study hundreds of mutants at once, rather than producing and characterizing them one at a time. The authors performed a number of controls to validate their analysis. Thus, 45 of their mutants were chosen at random and individually subjected to more quantitative scrutiny. In 15 of these mutants, the insertion had generated a stop codon in the Gag reading frame; these result in truncation of the protein, and none was capable of assembling into virus particles. All of the remaining 30 mutants could assemble into particles, and among these mutants there was perfect concordance between the footprinting assays for replication competence and the more systematic individual analyses.

When an MLV particle matures, the Gag protein is cleaved into four fragments: MA, p12, CA, and NC. In the current study, Auerbach et al. (1) scanned ≈35% of the MLV Gag gene. The stretch encodes the C-terminal 31 residues of MA, all of p12, and the N-terminal 77 residues of CA. Fig. 1 shows the genetic footprints they identified. There are two very short footprints within MA; one footprint spanning the MA-p12 junction; one longer footprint and three short footprints within p12; and finally, a very long footprint consisting of the p12-CA junction and the entire segment of CA analyzed here. In other words, the authors have identified eight regions within this stretch of Gag in which insertion of 12 amino acids is lethal for the virus.

Fig. 1.

Locations of footprints in a ≈600-nt stretch of the MLV Gag gene. The upper line indicates MLV Gag protein; the lower line indicates the segment analyzed by Auerbach et al. (1). Vertical arrows indicate sites of maturation cleavages of Gag. The eight horizontal bars (A–H) represent blocks of insertion-sensitive sites (footprints) identified.

There are several remarkable aspects of these results. First, they bespeak the sheer power of this elegant genetic approach for the identification of insertion-sensitive regions. Some of these footprints are as short as 5–7 bp, and there is no a priori reason to select these regions for mutagenesis; thus, it would take many years to find them by conventional mutational analysis. Second, as the authors point out, one might have expected retroviruses to have been selected for maximal efficiency in their use of genetic information. However, the data here show that, within p12 and the portion of MA analyzed here, ≈2/3 of the positions in the proteins are tolerant to the 12-residue insertions. Third, by contrast, every insertion within the portion of CA studied here, representing ≈30% of the CA protein, was lethal for the virus.

What is the functional significance of the critical regions that were pinpointed by their sensitivity to insertion? In several cases, we simply do not know. One of the striking results reported here is that all of the 30 insertion mutants that were analyzed in some detail were able to generate virus particles when the mutant DNA was transfected into mammalian cells. Thus, none of these insertions blocks assembly per se. The only well known function of the MA domain of MLV Gag is that it directs the association of Gag with the plasma membrane, where the particle assembles. In turn, this function is attributable to a fatty-acid modification at the extreme N terminus of Gag (3) and, possibly, by analogy with the HIV type 1 (HIV-1) MA domain, to basic residues within MA (4). Auerbach et al. now report that two regions near the C terminus of the MA domain, each only 2–3 aa long, are sensitive to the insertions. The explanation for the sensitivity of these specific sites must await further analysis.

p12, the Gag cleavage product between MA and CA in MLV, is only 84 aa long. Little is known about its role(s) in viral replication; MLVs in which 5-aa blocks in the central portion of p12 are replaced by runs of alanine are still viable (5). However, p12 is known to contain the “late” domain of MLV Gag, i.e., P-P-X-Y (6). This short amino acid motif interacts with the cellular machinery normally responsible for directing specific proteins into multivesicular bodies, and the interaction is evidently critical for the release of the budding virus particle from the cell (for reviews of this important topic, see refs. 7 and 8). Other regions of p12 apparently function early in the infectious process, promoting the incorporation or retention of newly synthesized viral DNA in the nucleus after mitosis (9). The footprints within p12 contain both these regions and the P-P-X-Y motif, as might be expected.

One of the most remarkable results to emerge from the analysis by Auerbach et al. is the uniform sensitivity of the N-terminal region of CA to insertions: unlike the data with MA and p12, the authors found no sites within the first 77 residues of CA that could tolerate insertions. Maturation of a retrovirus particle entails a dramatic morphological change, involving the appearance of a condensed core within the interior of the particle. Several kinds of studies, principally with HIV-1, support the following model of this morphological change. When the N terminus of CA is formed by cleavage of Gag during maturation, its N-terminal region undergoes a conformational rearrangement. The N-terminal proline is folded into the interior of the CA, forming a buried salt bridge with an aspartate or glutamate residue. This rearrangement permits the formation of new interfaces between the N-terminal domains of neighboring CA molecules (10, 11). In turn, the new interfaces lead to the formation of hexameric lattices by the mature CA molecules. Within these lattices, the N-terminal domains of six CA molecules form a hexameric ring, and the rings are joined with each other by dimeric contacts between the C-terminal domains of CA molecules in adjacent hexamers (12, 13). The presence of 12 pentameric “defects” in the lattice leads to the closure of the sheet of CA molecules; this closed shell, surrounding the complex of NC and viral RNA, is the visible core in the mature particle. The shape of the core (which differs in retroviruses of different genera) is determined by the distribution of the pentameric positions within the lattice (14).

Disassembly of the normal mature core is an essential, carefully controlled step in infection.

Mature CA, and/or the structure of the mature core of the particle, appears to play a crucial role in the infectious process. One fact suggesting a function for CA during infection is its presence in the MLV “preintegration complex,” i.e., the complex of integrase and viral DNA that is formed in the cytoplasm of the newly infected cell, before the newly synthesized viral DNA is integrated into the cell's chromosomal DNA (15). MLV CA is also the target of the mouse Fv-1 restriction system, which blocks infection by sensitive MLVs at a step between reverse transcription and integration (16–20). In HIV-1, mutations in CA that alter the stability of the core in viral lysates affect the kinetics of DNA synthesis during infection (21). Similarly, mutations in HIV-1 that replace either the N-terminal proline or its aspartate partner in the buried salt bridge not only block the morphological maturation of the particle (10, 22, 23) but also prevent the synthesis of viral DNA when the particles infect new host cells (despite the presence of normal levels of reverse transcriptase in the particles) (22, 23). Recent data show that the reverse transcriptase in the aspartate mutant particles is not in its normal site in the particle: although cores isolated from wild-type mature particles contain some reverse transcriptase, the enzyme is completely absent from the corresponding fractions from mutant particles (24). Taken together, these results suggest the possibility that disassembly of the normal mature core is an essential, carefully controlled step in infection; the disassembling core may form a kind of scaffolding for the steps between entry of the particle and integration of the viral DNA into cellular DNA.

It is in this light that the results of Auerbach et al., implicating the entire N-terminal region of CA in the infectious process, are so intriguing. It will be fascinating to learn more about the phenotypes of their mutants. Also, it will be important to determine the nature of the defects introduced by those few insertions in MA and p12 that proved lethal for the virus. Their experimental approach is bound to offer rich new insights into molecular mechanisms in retrovirus replication.