The cleavage of Gag, Pol and Env precursors forms the products in the mature infectious virions. These proteins are named by convention by a two-letter code: MA for matrix or membrane-associated protein; CA for capsid; NC for nucleocapsid; PR for protease; DU for dUTPase; RT for reverse transcriptase; IN for integrase; SU for surface protein; and TM for transmembrane protein.³²³ The localization of these proteins in the mature virion is not known with great precision, but a highly schematic version of the generic retrovirion can be drawn (Fig. 47.3).

The genomic RNA is highly condensed in the virion by its association with the nucleocapsid protein, NC. The complex is contained within a protein core largely composed of the capsid protein CA, another Gag gene product. The shape of the core is different among the various retroviral genera, and is a distinguishing feature of the genera. In most of the viruses the core is roughly spherical, but in some cases can be either conical or cylindrical. In all the viruses the core is surrounded by a roughly spherical shell consisting of MA, which in turn is surrounded by the lipid bilayer of the virion envelope. The virion membrane contains the envelope glycoprotein, with the TM subunit present as a single-pass transmembrane protein anchor, and the SU subunit as an entirely extravirion protein bound to TM. The envelope proteins for those viruses examined closely have been found to reside in the membrane as trimers.

The process of reverse transcription is initiated from the paired 3′ OH of a primer tRNA annealed to the viral RNA genome at a complementary sequence termed the primer binding site (pbs). DNA is first synthesized from this primer, using the plus strand RNA genome as template, to form minus strand DNA sequences. Synthesis occurs toward the 5′ end of the RNA to generate U5 and R sequences. The intermediate formed in this step is termed minus-strand strong-stop DNA. The primer tRNA remains attached to its 5′ end.

The next step involves the translocation, or “jump,” of the strong-stop DNA from the 5′ to the 3′ end of the genome. This translocation requires the degradation of those 5′ RNA sequences that were placed in RNA:DNA hybrid form by the formation of strong-stop DNA. The degradation is mediated
by the RNase H activity of RT; mutants with altered RNase H activity do not mediate the translocation. This step exposes the ssDNA and facilitates its annealing to the r sequences at the 3′ end of the genome.⁹⁵ Normally a full-length strong-stop DNA, synthesized by copying to the 5′ cap of the RNA, performs the translocation, though incomplete molecules can jump at low efficiency. The NC protein may facilitate the transfer step. Although there have been reports that jumping is always in trans, from one RNA template to the other RNA in the virion, the best evidence is that minus-strand strong-stop jumping goes randomly to either RNA.

The annealing of minus-strand strong-stop DNA recreates a suitable primer-template structure for DNA synthesis, and RT can now continue to elongate the minus-strand strong-stop DNA to form long minus-strand products. Synthesis ends in the vicinity of the pbs. As the genome enters RNA:DNA hybrid form, the RNA becomes susceptible to RNase H action and is degraded.

The primer for plus-strand synthesis is created by the digestion of the genomic RNA by RNase H. A particular short purine-rich sequence near the 3′ end of the genome, the polypurine tract or ppt, is relatively resistant to the activity of RNase H. The oligonucleotide remains hybridized to minus strand DNA and serves as the primer for synthesis of plus strand DNA, using minus strand DNA as template. The sequence of the PPT, an unusual structure of the nucleic acid at the PPT, and residues of the RNase H domain of RT have all been implicated in defining the cleavages that form the primer. Sequences upstream of the polypurine tract, an AT-rich region called the T-box, are also important for proper priming. The primer, once it has served to initiate DNA synthesis, is removed from the DNA. Synthesis proceeds toward the 5′ end of the minus strand, first copying the U3, R, and U5 sequences, then extending further to copy a portion of the primer tRNA still present at its 5′ end. Elongation stops at a modified base normally found at position 19 of the tRNA. The resulting intermediate is termed plus-strand strong-stop DNA.

In some viruses, secondary plus-strand initiation sites are used. There may be multiple RNA primers generated from the RNA genome by the nuclease action of RNase H that can initiate DNA synthesis at dispersed heterogeneous sites. In the case of the lentiviruses and spumaviruses, a second copy of the ppt sequence near the center of the genome is used at high efficiency, and is important for proper completion of reverse transcription.⁹¹

In the next step, the primer tRNA at the 5′ end of the minus strand DNA is removed by RNase H. Its removal may occur in two stages: with an initial cleavage near the RNA–DNA junction and a second one within the tRNA. The cleavage need not occur exactly at the RNA–DNA junction, and a single ribonucleotide base (A) is normally left on the 5′ terminus of the HIV-1 minus strand without affecting subsequent processes. The posttranscriptional modifications present in natural tRNA are probably important for proper recognition by RT and for plus-strand strong-stop translocation.

The removal of tRNA exposes the 3′ end of the plus-strand strong-stop DNA to permit its pairing with the 3′ end of the msDNA. The sequences anneal via the shared pbs sequences. This annealing forms a circular intermediate, with both 3′ termini in a suitable structure for elongation.

Both strands are now elongated. The final extension of minus strand DNA is coupled to displacement of the plus-strand strong-stop DNA from the 5′ end of the minus strand; as minus-strand elongation occurs, the plus-strand strong-stop is peeled away and transferred to the 3′ end of the minus
strand. At the end of this elongation, the circle is opened up into a linear DNA. The plus strands are then extended. When multiple plus-strand initiation events have occurred, the completed plus strand will consist of adjacent fragments and contain nicks or discontinuities. Displacement synthesis by an upstream fragment can slowly displace downstream RNAs and DNAs, leading to longer plus strands. However, some nicks or gaps may persist in the final double-stranded product. These breaks may be at heterogeneous positions, though strong sites of plus-strand initiation, such as the one at the central ppt of lentiviruses, can lead to specific sites for such discontinuities. Sequences near the central ppt of the lentiviruses cause termination of synthesis during elongation from upstream primers, ensuring the maintenance of a discontinuity at this site.⁹² This site retains a partially displaced sequence or overlap of a few nucleotides: 99 nt in the case of HIV-1. The structure has been shown to persist even to the time of integration of the DNA into the cell. Host DNA repair processes ultimately correct all such discontinuities.

Although most of the viral DNA is made in the cytoplasm, it may not always be completed in the cytoplasm. For some viruses, completion of the two DNA strands may occur only after entry into the nucleus. Specific mutants with alterations in the Cys-His residues of the NC protein show an interesting phenotype: the formation of linear DNA with heterogeneous and truncated ends.²⁰⁸ These experiments suggest that NC plays a role in the completion, or the stabilization of the ends, of the viral DNA.

A key consequence of the two translocation events that occur during reverse transcription is the duplication of sequences: duplication of U5 during minus-strand strong-stop DNA translocation and of U3 during plus-strand strong-stop DNA translocation. The resulting DNA thus contains two LTRs that have been assembled during reverse transcription. Each LTR consists of the sequence blocks U3-R-U5. The positions of the LTR edges—the left edge of U3, and the right edge of U5—are determined by the sites of initiation of DNA synthesis for the two DNA strands. Thus, the terminal sequences of the complete DNA molecule are also determined by these sites of initiation. These sequences for most viruses are perfect or imperfect inverted repeats, and serve an important role during integration of the DNA (see the Viral att sites section).

DNA polymerase activity is similar to that of all host and viral polymerases in requiring a primer, which can be either RNA or DNA, and a template, which can also be either RNA or DNA. RTs incorporate dXTPs to a growing 3′OH end with release of PPi, and require divalent cations, usually Mg++. The primer must contain a 3′OH end that is paired with the template. RTs cannot perform nick-translation reactions, but they can efficiently perform strand displacement synthesis. The only fundamental way in which RTs are unusual among the DNA polymerases is that they exhibit comparable specific activity on either DNA or RNA templates.

RTs are readily isolated from purified virion particles, and can be even more easily prepared as recombinant proteins expressed in bacteria. RTs are relatively slow DNA polymerases, under standard conditions only incorporating 1 to 100 nucleotides per second, depending on the template. Further, they exhibit poor processivity, and tend to release primer-template frequently in vitro. The enzyme must then rebind to the substrate to continue synthesis. Secondary structures in RNA templates can strongly enhance the pausing of RT and its tendency to release from the template.²²⁶ The enzyme also exhibits low fidelity, and though the values of the error rate vary widely with the primer, template and type of assay, the misincorporation rate of most RTs under physiologic conditions is on the order of 10⁻⁴ errors per base incorporated. This rate suggests that during replication there would be approximately one mutation per genome per reverse transcription cycle. The mutation rate observed in vivo is roughly consistent with this high error rate, though fidelity in vivo may be somewhat better than in vitro. Drug-resistant variants that do not incorporate chain-terminating analogs are often found to exhibit higher fidelity, perhaps because they require a more precise fit for the correct incoming triphosphate to allow for discrimination against the analog. A wide range of types of mutations are created by RT errors, and both the type and the frequency of appearance of each type of mutation exhibit a complex dependence on sequences and structures in the template.

RTs do not generally exhibit a proofreading nuclease activity,³⁵ and misincorporated bases are not removed as efficiently by most RTs as they are by host DNA polymerases. However, mutants of the HIV-1 RT resistant to AZT have been shown to exhibit an enhanced ability to remove the incorporated AZT moiety at the 3′ end through a pyrophosphorolysis reaction.³⁸² Thus, it is possible for RT to remove some such analogs and rescue a terminated chain for continued elongation.

The RNase H activity of RT is an endonuclease that releases oligonucleotides with a 3′OH and a 5′PO₄. This property allows the products of RNase H action to serve as primers for initiation of DNA synthesis by the DNA polymerase function of RT. There is an obligate requirement that the RNA be in duplex form, normally an RNA–DNA hybrid. However, retroviral RTs are also able to degrade RNA–RNA duplexes, an activity termed RNaseH*.²⁴³ The RNase H enzyme is capable of acting on the RNA of a template in concert with the polymerase as it moves along a nucleic acid, and as it does so its active site is located about 17 to 18 bp behind the growing 3′ end.²⁰⁶ RNase H can also act independently of polymerization. All RNase H activity requires a divalent cation.

RT is incorporated into the virion particle during assembly in the form of a large Gag-Pol precursor (see below), and is released by proteolytic processing of the precursor during virion maturation. Different viruses make somewhat different cleavages in the precursor, and thus the RTs exhibit several different subunit structures (see below). In the gammaretroviruses, RT is a simple monomer in solution, corresponding only to the aminoterminal DNA polymerase and the carboxyterminal RNase H domains. These two domains can be expressed separately, and the isolated proteins exhibit their respective activities,⁵⁸⁷ though the specificity of the RNase H is affected by this separation. In the avian viruses, the RT is present as an αβ heterodimer, comprised of a smaller subunit containing the DNA polymerase and RNase H domains; and a larger β subunit containing these two domains but also retaining the integrase domain. In the lentiviruses, RT is again a heterodimer with a larger subunit (p66) containing the DNA polymerase and RNase H domains, and a smaller subunit (p51) lacking RNase H. The properties of the different enzymes as DNA polymerases are very similar in spite of these different subunit structures, and thus the significance of these various compositions for RT function is unclear. A curious observation was made that some RT inhibitors—the so-called nonnucleoside RT inhibitors—can potently enhance the association of p66 and p51, locking them into an inactive dimer.⁵⁸⁰

The three-dimensional structure of a number of RTs have been determined by X-ray crystallographic studies. Structures of the unliganded HIV-1 RT,²⁴⁶,⁵¹⁵ RT bound to nonnucleoside RT inhibitors,¹²⁷,¹³⁵,³⁰⁰,⁵⁰⁴ RT bound to an RNA pseudoknot inhibitor,²⁶⁰ RT bound to a duplex oligonucleotide,¹⁷,²⁴⁸,²⁵⁸,²⁵⁹ and RT bound to a polypurine tract RNA:DNA hybrid,⁵³¹ as well as the isolated RNase H domain,¹²⁸ have all been reported. The two subunits are folded very differently so that the overall structure is highly asymmetric. The structure of the p66 is similar to that of a right hand, with fingers, palm, and thumb domains named on the basis of their position in the structure (Fig. 47.8). The nucleic acid lies in the grip of the hand, held by the fingers and thumb. The YXDD motif present at the active site for the DNA polymerase lies at the base of the palm. The RNase H domain is attached to the hand at the wrist. The p51 subunit, while made up of the same domains as the aminoterminal part of p66, is folded differently and lies under the hand, not making direct contact to the nucleic acid and thus not thought to participate in chemistry. The structure of p66 with and without a liganded nucleic acid is very different, with the thumb domain flexing to allow substrate binding. A surprising aspect of the structures is that the nucleic acid helix can be highly bent, perhaps accounting for the enzyme’s ability to sense conformationally strained substrates.⁵³¹ Theoretical considerations suggest that the thumb may move during elongation.

Structures of the fingers and palm subdomain and of the complete Moloney MuLV RT at very high resolution have also been determined.¹²⁶,²⁰⁰ The monomeric protein is broadly similar to the p66 subunit of HIV-1 RT.

RT is a major target of antiviral drugs useful in the treatment of retroviral diseases such as AIDS. All such drugs used to date are inhibitors of the DNA polymerase activity of RT, and fall into two classes: nucleoside analog inhibitors (chain terminators), and nonnucleoside RT inhibitors (NNRTIs). The nucleoside analogs are typically prodrugs, and need to be activated by phosphorylation to the triphosphate form. These are then incorporated by RT into the growing chain, and serve to block further elongation. Examples include AZT, ddC, ddI, d4T, and 3TC. The NNRTIs are a group of compounds that are structurally diverse, but nevertheless interact with a common binding pocket in RT to prevent its normal activity.⁶⁰⁰ There are indications that the binding may inhibit the enyzme’s flexibility. For both classes of inhibitors, monotherapy with a single drug selects for drug-resistant variants that quickly predominate in the virus population, and for each drug, a pattern of mutations has been identified that serves to indicate the appearance of drug resistance.³¹⁵ In many cases these mutations alter the binding side for the nucleoside or NNRTI such that the drug cannot bind and therefore cannot inhibit the enzyme. In the case of AZT, however, the mutations do
not prevent the binding and incorporation of AZTTP into the growing chain, but rather seem to activate a reverse reaction in which the AZT nucleotide is removed from the chain, subsequently permitting normal elongation.³⁸² Combination therapy, typically involving the simultaneous treatment with three different drugs, can suppress virus replication to such an extent that variants resistant to all the drugs do not appear, at least for months or years.

In the first step, the two terminal nucleotides at the 3′ ends of the blunt-ended linear DNA are removed by the integrase to produce recessed 3′ ends and correspondingly protruding 5′ ends. This cleavage occurs endonucleolytically at a highly conserved CA sequence, and releases a dinucleotide. For most viruses the terminal sequence is such that a TT dinucleotide is released, though this rule has exceptions. The ends do not remain covalently bound to protein, and the energy of the hydrolyzed phosphodiester bond is not retained.

In the second step, the 3′OH ends created by processing are used in a strand transfer reaction to attack the phosphodiester bonds of the target DNA.¹⁸⁶ The attack occurs by an Sn2-type reaction, with inversion of the phosphorus center as detected with chiral labeling of the phosphate.¹⁵⁸ The formation of the new phosphodiester bond between the viral end and host DNA displaces one of the phosphodiester bonds in the host DNA, leaving a nick. The protruding 5′ end of the viral DNA is not joined to the host DNA by IN. The reaction is a direct transesterification, and thus no ATP or other energy source is required. Mutational studies strongly suggest that the two activities—processing and joining—utilize the same active site residues. In fact, the two steps involve similar chemistry: 3′ end processing is an attack on DNA by a hydroxyl residue of water, while joining is an attack on DNA by a 3′ hydroxyl residue of another DNA. It should be noted that other hydroxyl residues can participate; alcohols such as glycerol can be utilized, and
the 3′OH of a DNA can even attack a phosphodiester bond on the same DNA, forming a cyclic product.¹⁵⁸

The IN protein exhibits a third enzymatic activity in vitro: a reversal of the integration reaction known as disintegration.⁹⁸ This activity releases DNA from a branched structure and seals the nick at the site of the branch. The significance of the activity in vivo is uncertain.

In a wild-type virus, when the two ends are joined to the two strands of the target DNA, the two sites of attack are staggered by a few base pairs. After the joining, the resulting structure contains short gaps in the host DNA and unpaired bases from each 5′ end of the viral DNA. The 5′ ends of the viral DNA are not joined to host DNA by any known activities of IN. However, the 5′ ends are very quickly repaired in vivo, almost as quickly as the initial integration reaction.⁵¹⁶ These discontinuities are presumed to be repaired by the host repair enzymes, though it is possible that the viral RT or IN could participate. The processing and filling in the gaps creates a short duplication of sequence that was present only once at the target site; these duplications flank the integrated provirus. The number of bases duplicated is characteristic of each virus. Thus, the murine and feline viruses cause a 4-bp duplication, HIV-1 causes a 5-bp duplication, and the avian viruses cause a 6-bp duplication.

The transcriptional elements in the U3 region of the simple viruses contain both core promoter sequences and enhancers. The core promoters contain a TATA box, bound by TFIIB; a CCAAT box, bound by CEBP⁵²⁵; and sometimes an initiator sequence near the U3-R border. The U3 regions of even closely related retroviruses are very diverse, and can evolve rapidly during viral replication. The enhancers are similar to those found
in many host promoters in containing multiple short-sequence motifs, arranged in very close packing; often there are tandemly repeated copies of some of these motifs. These short sequences are the binding sites for a large number of host factors that regulate transcription (e.g., see 562). Different cells and cell types will make use of distinct arrays of these factors to mediate transcription from a given viral LTR.²¹³ The factors are not simply additive but may interact in complex ways on particular viral sequences. A partial list of these factors used by various retrovirus LTRs includes: Sp1; USF-1; the Ets family of factors, which include more than 20 members in vertebrates; the core-binding factor (CBF), consisting of an a-b heterodimer; nuclear factor 1 (NF1); and a mammalian type C retrovirus enhancer factor (MCREF-1). Specific viruses may often contain recognition sites for other more specific positive regulatory factors. Major examples of such factors include the glucocorticoid receptors, driving expression of the MMTV genome, and to a much lesser extent, other MuLVs; NF-κB, important for expression from the HIV-1 LTR in certain cell types; the GATA factors for Cas-BR-E and other viruses; and the myb protein. Evidence has been obtained that the STAT factors, DNA binding proteins normally activated the Janus kinases (Jaks) may also be important for MMTV transcription.⁴⁹³

A number of negative regulatory factors that reduce viral expression have been identified. Embryonic carcinoma cells, and true embryonic cells, are the best-characterized examples of cell types that strongly repress LTR-mediated transcription through expression of negative regulatory proteins. The MuLVs are silenced via a stem-cell specific repressor that binds to a site, curiously, overlapping almost perfectly with the proline tRNA primer binding site.²⁸⁹,⁴⁷³ The proteins responsible for this silencing in mouse embryonic stem cells have recently been identified as TRIM28 (Kap-1) and the zinc-finger protein ZFP809.⁶³⁷,⁶³⁸ Viruses that use an alternate primer tRNA and thus lack the pbs recognition site for these proteins can escape the repression.²³⁹ Other negative factors include one known variously as UCRBP, NF-E1, or YY1,¹⁷⁶ and a cellular embryonal LTR-binding protein (ELP; 598).

The complex retroviruses encode an array of small regulatory proteins that can activate transcription from the viral LTR in trans. Examples of these transactivators include the HTLV-1 Tax protein¹³² and the HIV-1 Tat protein.¹²⁰ The Tax protein acts in concert with a complex of host proteins, the activating transcription factor/CRE-binding protein (ATF/CREB), and binds to three cAMP response elements in the viral LTR. Tax thus sets up a positive feedback loop that results in high levels of viral transcripts. The Tat protein is unusual among transcriptional activators in that it binds to a structure in the 5′ end of nascent viral RNA, rather than to DNA.¹³⁶,⁵⁴⁸ Tat binds to a bulged hairpin structure, the TAR element, and recruits a pair of host proteins, cyclinT/cdk9, to the RNA. These proteins enhance the ability of RNA polymerase to elongate beyond the LTR and down the genome with high processivity, probably by phosphorylation of the C-terminal repeat domain (CTD) of the polymerase. Again, the result is a strong positive feedback loop that results in high levels of viral RNA. (For more detailed discussion of tat function, see 120, and Chapter 49 of this book.)

In the gammaretroviruses and epsilonretroviruses, the Gag and Pro-Pol ORFs are in the same reading frame, separated by a single UAG stop codon at the boundary between Gag and Pro-Pol. The translation of Gag-Pro-Pol in these viruses occurs by translational readthrough—that is, by suppression of termination—at the UAG stop codon.⁶⁵¹ Most of the time, translation of the RNA results simply in the formation of the Gag protein. But approximately 5% to 10% of the time, ribosomes translating the RNA do not terminate at the UAG codon, and instead utilize a normal aminoacyl tRNA, usually a glutamine tRNA, to insert an amino acid at the position of the stop codon. Translation then continues, in frame, through the entire long pro-pol ORF, resulting in the formation of a long Gag-Pro-Pol precursor protein.

The high-level suppression of termination is specified by a specific structure in the RNA immediately downstream of the UAG stop codon.²⁴¹,⁴⁴⁹ The precise features of this structure that are required for suppression are not completely known, but they include a purine-rich sequence immediately downstream of the stop codon, and a pseudoknot formed from the next 60 or so nucleotides.¹⁷² The structure may slow translation, and it may also in some other way alter the balance between termination, which requires binding of termination factors eRF1 and eRF3 by the ribosome, versus incorporation of an amino acid, which requires misreading of the codon by an aminoacyl tRNA. No changes in the tRNA pool occur during infection. The signals in the RNA can operate to mediate suppression of both UAA and UGA termination codons as well as UAG.

A screen for proteins interacting with the MuLV RT resulted in the identification of the eukaryotic termination factor eRF1, and subsequent studies showed that overexpression of RT could inhibit termination and promote translational readthrough of the Gag stop codon in vivo.⁴³⁴ Mutant viruses with point mutations in RT blocking the interaction with eRF1 were unable to express normal levels of Gag-Pol and failed to replicate. These results suggest that RT, likely in the context of the nascent
Gag-Pol protein, can bind and inhibit eRF1, increasing the level of readthrough to increase its own synthesis. The final level of Gag-Pol produced in this positive feedback loop presumably is ultimately limited by other factors.

In the alpharetroviruses and lentiviruses, the gag and pol ORFs lie in different reading frames, and the formation of the Gag-Pro-Pol fusion is mediated by a translational frameshift mechanism.²⁵⁷ Most of the time, translation again results in the simple formation of the Gag protein. But approximately 10% of the time, as the translation approaches a specific site near the end of the gag ORF, the ribosome slips back one nucleotide (a –1 frameshift) and proceeds onward in the new reading frame. The ribosome passes through the stop codon out of frame and continues to synthesize protein using the codons of the pol ORF. As for readthrough, the determinants of frameshifting lie in the RNA sequence and structure near the site of the event. The requirements for frameshifting include a “slippery site,” a string of homopolymeric bases where the frameshift occurs; these are oligo U or oligo A in different viruses. In addition, the frameshifting requires either a very large and near-perfect hairpin or stem-loop structure (as for HIV-1 group M viruses); or a large pseudoknot structure (as for HIV-1 group O viruses), similar to those used in readthrough, though apparently containing a distinctive bend at the junction of the two paired sequences. As for readthrough, the proper frameshifting efficiency is crucial for normal virus replication.

In the betaretroviruses (e.g., MMTV) and deltaretroviruses (e.g., BLV, HTLV-1), the pro gene is present as a separate ORF, in a different reading frame from that of gag or pol. Two successive frameshifts are utilized to make the long Gag-Pro-Pol fusion protein. Near the 3′ end of the gag ORF, ribosomes carry out a first (–1) frameshift and continue into the pro ORF; near the 3′ end of the pro ORF, they perform a second (–1) frameshift and continue on into the pol ORF. These two frameshifts occur at extremely high frequencies—as much as 30% of the time that the ribosome transits through each site—so that the overall frequency of formation of the Gag-Pro-Pol protein is perhaps 10% that of formation of Gag.

The spumaviruses are unique among the retroviruses in that the synthesis of the Pol protein is not mediated by the formation of a Gag-Pol fusion protein. Instead, a subgenomic spliced mRNA is translated directly to form a separate Pro-Pol protein.¹⁵⁹,³⁴⁰ This protein must be directed to the assembling virion by distinct domains rather than by the Gag portion of a Gag-Pol fusion.