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

Processing of Retroviral RNA

Following transcription of the provirus, retroviral RNA transcripts are subject to the same processing events as cellular RNAs, including cap addition at the 5′ end, cleavage and polyadenylation at the 3′ end, and splicing to form subgenomic-sized RNA molecules (see Fig. 1). The cap addition and 3′-end processing precede splicing events; thus, all viral RNAs are capped and polyadenylated. Full-length retroviral RNA transcripts serve two functions: They encode the gag and pol gene products, and they are packaged into progeny virion particles as genomic RNA. Subgenomic-sized RNA molecules provide mRNAs for the remainder of the viral gene products. Simple retroviruses splice a subset of the genomic RNA into a transcript that encodes the env gene product. Complex retroviruses generate both singly and multiply spliced transcripts that encode not only the env gene products, but also the sets of regulatory and accessory proteins unique to these viruses. The relative proportions of full-length RNA and sub-genomic-sized RNAs vary in infected cells, and modulation of the levels of these transcripts is a potential control step during retroviral gene expression. The processing of retroviral transcripts is performed by host-cell machinery; to exploit this machinery, the virus contains the necessary cis-acting regulatory elements.

Termination of Transcription, 3′-end Cleavage, and Polyadenylation

The Basic Events

Transcription of a provirus can be terminated at various positions outside of the viral template, and the retrovirus appears to have no control over this process. The precise positioning of the 3′ ends of retroviral RNAs is generated by posttranscriptional processing. Like most mRNAs, retroviral RNAs have a poly(A) tail at their 3′ end. The size of the poly(A) tail has not been measured in most viral RNAs; however, the range is expected to be similar to that for cellular mRNAs (i.e., 200–300 nucleotides). Transcription of the provirus proceeds past the site of poly(A) addition, and then the primary transcript is processed by endonucleases and polyadenylated. The only site of 3′-end processing is the R/U5 border that is encoded by the 3′ LTR; thus, all viral RNAs have an identical 3′ end. Similar sequence-dependent signals mediate the 3′-end processing of viral and cellular transcripts; therefore, it is expected that cellular processing machinery modifies viral transcripts.

It is thought that polyadenylation affects RNA stability and is important for the transport of transcripts out of the nucleus (Atwater et al. 1990). Recent evidence from developmental systems in which the length of poly(A) tails is regulated indicates that polyadenylation also affects the translational efficiency of mRNA (Wickens 1992). Thus, to enter the cytoplasm and to be translated, retroviral RNA must have a poly(A) tail (see also Chapter 7.

In the case of ASLV, it has been demonstrated that a significant number (15%) of viral transcripts retain cellular sequences (Herman and Coffin 1987). These “readthrough” transcripts are polyadenylated via cellular poly(A) signals and can be packaged into virions as functional genomic RNA (Swain and Coffin 1989). In other viruses (e.g., SNV), the proportion of readthrough transcripts is reported to be much lower (Iwasaki and Temin 1990). Nevertheless, the major biological significance of these readthrough transcripts is likely to be their role in the acquisition of cellular proto-oncogenes (see Chapter 4.

Despite the similarities between viral and cellular RNAs, retroviral RNA processing requires distinct regulatory mechanisms and may provide unique insights into cellular processing machinery. Specifically, there are two R/U5 borders in the primary retroviral transcript that could be processed. For RNA to span the viral coding region, the second, rather than the first, R/U5 border must be used. Our understanding of 3′-end processing, particularly alternative site usage, is incomplete in both viral and cellular systems.

Host-cell Machinery

The use of host-cell machinery for 3′-end processing and polyadenylation of cellular and viral transcripts has been reviewed extensively (Wickens 1990; Proudfoot 1991; Wahle and Keller 1992; Guntaka 1993). The processing of the 3′ ends of cellular RNAs is mechanistically separable from the termination of transcription. In contrast, 3′-end cleavage and polyadenylation are intimately linked. The chemical reactions involved in processing are simple: Endonucleolytic cleavage of the phosphodiester backbone generates a free 3′-OH group that serves as a substrate for the addition of adenosine nucleosides by a poly(A) RNA polymerase. These events are regulated by a highly conserved cis-acting element (the sequence AAUAAA) that is usually found 20–30 bp upstream of the site of polyadenylation in the primary transcript. This poly(A) signal functions in RNA as a binding site for a multisubunit RNA-binding protein termed the cleavage and polyadenylation specificity factor (CPSF). The AAUAAA sequence and CPSF are necessary for both endonucleolytic cleavage and poly(A) addition. After an initial slow phase of polyadenylation, in which a ten-residue tail is synthesized, poly(A) polymerase cooperates with the poly(A)-binding protein to proceed with efficient tail lengthening.

Although the poly(A) signal AAUAAA appears to be sufficient for recruiting processing machinery to primary transcripts, additional cis-acting signals have been identified. The site of cleavage shows only a moderately conserved CA dinucleotide; however, mutation of these residues can affect processing efficiency. In many mRNAs, GU-rich and U-rich elements have been identified downstream from the polyadenylation signal. Poor conservation and possible redundancies have hampered analyses of these elements. Nevertheless, some evidence exists that these elements function as binding sites for regulatory proteins. Regulatory sequences upstream of the poly(A) signal have been identified in several retroviruses (see below), but only in a few cellular genes. Alternative usage of polyadenylation signals has been reported for cellular genes, but the mechanism of selection is not clear. Retroviral RNAs exhibit specific use of one of two potential polyadenylation sites, as described below.

Retroviral 3′-end Processing

Retroviral primary transcripts contain the basic regulatory elements for the direction of cleavage and polyadenylation events by host-cell machinery (see Fig. 16). All retroviruses contain an identifiable polyadenylation signal. The consensus sequence AAUAAA is most common; however, MMTV uses the sequence AGUAAA. The position of the polyadenylation signal varies with respect to the transcription start site. In a subset of viruses (e.g., HIV, Mo-MLV, and SNV), the signal is in the R region, whereas in another group (e.g., HTLV-1, MMTV, and ASLV), the signal is in the U3 region. Most retroviruses have the commonly observed CA sequence at the site of polyadenylation such that the first A of the poly(A) tail is templated. Additional upstream and downstream signals have been identified. A comparison of G+T content in the U5 and R regions of several retroviruses shows that the U5 region is rich in G+T, which is consistent with the ubiquitous presence of a downstream GU-rich element (Guntaka 1993). The function of this region has been confirmed experimentally in only a few viruses, including ASLV, HIV, and HTLV-1.

Retroviral RNAs always contain two copies of the site for polyadenylation because the LTRs exhibit terminal redundancy; there is an R/U5 boundary in both the 5′ and 3′ends of the primary transcript (see Fig. 1). The production of full-length genomic RNA requires that the 3′ site (not the 5′ site) be used. In retroviruses in which the AAUAAA sequence is templated by U3 (upstream of R), the use of the second R/U5 border is explained simply by the incompleteness of the cis-acting elements upstream of the 5′ site (e.g., HTLV-1, MMTV, and ASLV; see Fig. 16). The key regulatory signal, the AAUAAA sequence, is present only at the 3′ end of the transcript; thus, the 5′ R/U5 border is easily bypassed. The situation is more complex in retroviruses in which there is a poly(A) signal upstream of each 5′ and 3′ R/U5 border. In HIV, both poly(A) signals have been shown to be fairly efficient (Proudfoot 1991; Cherrington and Ganem 1992), although the 3′ poly(A) signal is dominant. Additional upstream signals that map between the TATA box and the cap site in U3 have been implicated in dictating poly(A) signal use (Brown et al. 1991; DeZazzo et al. 1992; Gilmartin et al. 1992; Valsamakis et al. 1992).

Figure 16. Retroviral 3′-end processing and polyadenylation signals: Selected retroviral LTRs with positions of transcription start (horizontal arrow) noted.

Figure 16

Retroviral 3′-end processing and polyadenylation signals: Selected retroviral LTRs with positions of transcription start (horizontal arrow) noted. (AAA) Site of 3′-end processing and polyadenylation in RNA transcript. (Colored box) Site (more...)

Dominance by the 3′ poly(A) site also can be explained by negative regulation of 5′ poly(A) site usage. Studies of SNV and HIV provide support for inhibition by promoter proximity. Artificial constructs in which the poly(A) site is placed within 500 bp of the cap site are not processed efficiently (Iwasaki and Temin 1990; Weichs an der Glon et al. 1991). A similar negative control mechanism has been implicated for retroid elements such as cauliflower mosaic virus (Sanfacon and Hohn 1990) and hepatitis virus (Russnak and Ganem 1990; Cherrington et al. 1992). Recently, inhibition of the 5′ poly(A) site in HIV has been linked to RNA splicing (Ashe et al. 1995). The presence of the splice donor site in close proximity to the poly(A) site appears to lead to assembly of splicing machinery, rather than 3′-end processing machinery at this site in the RNA transcript. This mechanism could explain previous observations of inhibition by promoter proximity.

Viruses in the HTLV-BLV subgroup present an interesting dilemma in the control of 3′-end processing. The poly(A) signal is in the U3 region, but the processing of viral transcripts must accommodate an unusual spacing between the AAUAAA signal and the cleavage site with its downstream GU-rich element (Fig. 16). In HTLV-1, the unusually long R region which contains the Rex responsive element (RexRE) separates these two control sites by 250 bp. Rex regulates HTLV-1 gene maintenance of full-length and singly spliced RNA molecules via this element (see below). The predicted folding pattern of the RexRE displays a highly stable RNA stem-loop structure of 255 nucleotides that would closely appose the distant poly(A) site and the polyadenylation signal. The role of this structure in 3′-end processing has been tested (Ahmed et al. 1991; Bar-Shira et al. 1991). Deletion of the RexRE was shown to facilitate 3′-end processing, whereas mutations that disrupted the predicted structure inhibited the use of the viral processing site. Importantly, the ability of cellular machinery to associate with this structured RNA has been tested. The binding of CPSF to the upstream AAUAAA and the utilization of the GU-rich signal have been demonstrated in in vitro reconstitution reactions. There is no evidence for involvement of the Rex protein in the role of RexRE in facilitating polyadenylation. It is possible that a similar RNA secondary structure might explain the situation in cellular genes in which an obvious AAUAAA signal has not been noted close to the poly(A) site.

In conclusion, the appropriate site for retroviral 3′-end processing and polyadenylation is selected by (1) suppression of the 5′-site as a result of promoter and/or splice site proximity and (2) enhancement of the 3′-site usage as a result of signals in the U3 sequences (either the AAUAAA signal or other control elements). The molecular basis for this positive and negative regulation involving cellular processing machinery remains to be elucidated.

Regulation of Retroviral RNA Splicing and Transport

The Basic Events

A fraction of retroviral primary transcripts are spliced to generate subgenomic-sized mRNAs (Fig. 17). All spliced transcripts bear the same first exon, which spans the U5 region of the 5′ LTR. The final exon always retains the U3 and R regions encoded by the 3′ LTR, although it varies in size. The composition of the remainder of the RNA structure depends on the number of splicing events and the choice of alternative splice sites. For example, complex retroviruses and certain oncogenic retroviruses display alternative splicing patterns, and well over one dozen different species of RNA have been detected for HIV (Muesing et al. 1985; Schwartz et al. 1990; Purcell and Martin 1993).

Figure 17. Alternative splicing patterns in retroviruses.

Figure 17

Alternative splicing patterns in retroviruses. Examples of different patterns of retroviral splicing include the single splicing event that generates the MLV env RNA, alternative splicing of RSV responsible for env and src RNAs, and the multiple alternative (more...)

For retroviral gene expression, both unspliced and spliced RNAs must be transported to the cytoplasm, and different classes of retroviruses have evolved distinct solutions to this problem. The simple retroviruses, which use only full-length and singly spliced RNAs, regulate the cytoplasmic ratios of these species through cis-acting elements within the RNAs. The complex viruses encode proteins that regulate the transport and possibly splicing of the different genomic and sub-genomic-sized RNA species (Table 3).

Table 3. Retroviral Regulators of RNA Processing.

Table 3

Retroviral Regulators of RNA Processing.

Host-cell Machinery for Splicing

RNA splicing is the process by which intervening or “intronic” RNA sequences are removed and the remaining “exonic” sequences are ligated to provide continuous reading frames for translation. A molecular understanding of this process has emerged over the last decade and is only briefly summarized here (for review, see Green 1991; Moore et al. 1993; Nilsen 1994; Sharp 1994). The events between transcription and translation of mRNA are orchestrated in an assembly-line fashion by a set of regulatory RNAs and proteins. Following transcription, pre-mRNA is associated with heterogeneous nuclear ribonucleoproteins (hnRNPs), components of the spliceosome assemble onto the mRNA precursor, and the introns are removed. Spliced mRNAs are then transported to the cytoplasm. The best understood parts of this pathway are the events of the spliceosome cycle (Fig. 18). The spliceosome contains six snRNAs, each with a defined set of associated proteins (termed snRNPs). Additional protein components of the spliceosome also have been identified. The function of the spliceosome requires a network of RNA-RNA interactions both among the snRNAs and between these RNAs and the pre-mRNA. Thus, cis-acting elements in the pre-mRNA mediate both assembly of the machinery and the chemistry of the splicing events.

Figure 18. Eukaryotic mRNA splicing by spliceosome pathway.

Figure 18

Eukaryotic mRNA splicing by spliceosome pathway. The splicing of a single intron is illustrated. The sequences GU, A, and AG represent highly conserved cis-acting elements that are functional at the splice donor, branchpoint, and splice acceptor sites. (more...)

The chemistry of splicing involves two transesterification reactions. In the first, a nucleophilic attack by the 2′-OH of an adenosine nucleoside in the intron (the branchpoint) cleaves the 5′ splice site. This results in the formation of a free 5′ exon and a lariat structure composed of the intron attached to the 3′ exon. In the second, the free 3′-OH of the 5′ exon attacks and cleaves the 3′ splice site. Figure 18 illustrates the roles of snRNAs during the spliceosome cycle. Dramatic changes in the base pairing between the different snRNAs and the mRNA appear to direct the proper positioning of mRNA bases and drive catalysis. Many proteins including the snRNPs and other non-snRNPs (e.g., U2AF and ASF/SF2) help coordinate the function of the RNA:RNA network of the spliceosome. Some of these proteins are likely to play a part in alternative splice site selection (see below).

Splice Site Selection

Spliceosomes are directed to their site of action by the interaction between cis-acting signals in the pre-mRNAs and in the U1 and U2 snRNAs (Fig. 18). However, due to the size of many introns, the 5′ and 3′ splice sites in cellular RNAs are often separated by great distances. A solution to this logistical problem is found in a model of splice site selection that evokes an exon definition mechanism (Robberson et al. 1990; Berget 1995). Exons that are to be retained would be selected by a bridging complex that requires the 3′ splice site of the upstream intron and the 5′ splice site of the downstream intron. This model is directly relevant to the selection of the many possible exons within the genome of complex retroviruses (see Fig. 1). A bridging complex would define each exon. A corollary of this model proposes that the retention of the most 5′ exon would be modulated by the presence of a special bridging complex involving recognition of the 5′ cap of the mRNA. Likewise, processing of the most 3′ exons would involve the polyadenylated 3′ end of the mRNA. Singly spliced retroviral transcripts would use only these special exon-definition pathways.

All retroviral genomic-length transcripts contain signals for the removal of at least one intron for the generation of mRNA: a 5′ splice site and a distal 3′ splice site within the polypyrimidine stretch (see Figs. 1 and 18). These elements match the consensus sites derived from cellular mRNAs with a GU just 3′ of the 5′ splice site and an AG 5′ to the 3′ splice site. The A nucleoside branchpoint is presumed to be present but has been identified functionally in only a few retroviruses. In the case of the multiply spliced transcripts of the complex viruses, additional splice sites and branch sites are present. Production of the necessary spliced products requires alternative splice site selection. Alternative splice site selection also is seen in cellular mRNAs (for review, see McKeown 1992; Horowitz and Krainer 1994). In the most simplistic model, splice site selection is mediated by the relative binding affinities of the pre-mRNA cis-acting elements for components of the basic splicing machinery. However, it is likely that the elements that regulate alternative splice site selection may not be present in the well-studied basic machinery. Analyses of alternatively spliced cellular transcripts have implicated additional cis-acting elements in both introns and exons that affect splicing frequency. Regulatory proteins that interact with these elements also have been identified. In the well-studied case in which alternative splicing mediates sex determination in Drosophila, there is both genetic and biochemical evidence that specific factors interface with basic splicing machinery to both repress and activate splicing events (Mattox et al. 1992).

Retroviral RNA Transport

The control of transport of unspliced and spliced RNAs from the nucleus to the cytoplasm is clearly an important component of posttranscriptional regulation of retroviral gene expression. Relatively little is known about the mechanisms by which cellular mRNAs are exported from the nucleus (for review, see Krug 1993; Zapp 1995). It has been established that mRNAs exit the nucleus through nuclear pores in association with hnRNP proteins such as A1 hnRNP; these may act as chaperones, escorting RNA out of the nucleus and then shuttling back into the nucleus. As another example of RNA export from the nucleus, the binding of 5S ribosomal RNA to the TFIIIA transcription factor facilitates its export out of the nucleus (Guddat et al. 1990). Proper 5′- and 3′-end RNA processing is required for mRNA transport; however, the role of the proteins that associate with these structures is unknown. Viral proteins such as Rev in HIV (see below) and NS1 in influenza virus provide new tools for dissecting the cellular mechanisms of control of mRNA transport.

The control of RNA splicing and transport has been investigated in only a few retroviruses. In most studies, the focus has been on the key control step of maintaining unspliced RNAs for packaging as genomic RNA. Different retroviruses appear to have adopted different strategies to ensure export of sufficient quantities of unspliced RNA to the cytoplasm for expression of Gag and Pol proteins and for packaging of genomic RNAs. These different strategies involve the use of cis-acting viral RNA sequences and the trans-effects of viral or host cellular proteins (Fig. 19). The following sections contrast these controls in simple and complex retroviruses.

Figure 19. Mechanisms for nuclear export of unspliced or partially spliced retroviral RNAs.

Figure 19

Mechanisms for nuclear export of unspliced or partially spliced retroviral RNAs. Retroviruses have evolved several different mechanisms for ensuring that sufficient levels of unspliced RNAs are transported to the cytoplasm for translation of gag and (more...)

Maintenance of Unspliced RNA in Simple Retroviruses

The regulation of splicing and RNA processing in simple retroviruses is mediated through the interaction of cis-acting viral RNA sequences with cellular factors. No viral gene products have been identified that influence RNA processing in these viruses (Stoltzfus 1988). Cis-acting RNA sequences may promote the cytoplasmic expression of unspliced RNAs by affecting (1) splicing efficiency or (2) RNA transport (Fig. 19). Although the former mechanism likely has a role, recent studies have suggested that inefficient splicing may not be sufficient to explain the observed levels of unspliced cytoplasmic RNA. Cis-acting RNA elements that may have direct effects on transport have recently been identified in several simple retroviruses.

Most of the studies on RNA processing in simple retroviruses have been conducted with avian retroviruses. RSV, for example, exhibits a steady-state ratio of unspliced RNA to spliced RNA of approximately 2:1 (Katz et al. 1988). This may be due to inefficient utilization of retroviral splice sites (Fig. 19) (Katz and Skalka 1990), even though the splice recognition sites differ little from the published splice acceptor consensus sequence (Stoltzfus 1988). Other discrete RNA segments, which are independent of the splicing signals, have also been shown to inhibit splicing efficiency (Arrigo and Beemon 1988). Host cellular factors can also influence the efficiency of retroviral splicing. For example, although RSV can infect and transform both avian and mammalian cells, productive infection is not observed in mammalian cells (Altaner and Temin 1970). More recent studies have shown that this restriction is related to enhanced efficiency of RSV splicing, resulting in low levels of unspliced RNA (Deng et al. 1977) and production of aberrant double-spliced RNAs (Berberich et al. 1990) (see below).

Mechanisms for the generation of unspliced ASLV RNAs have been investigated using a mutant virus with a 24-nucleotide linker inserted in the vicinity of the env splice acceptor (Katz et al. 1988). This mutation resulted in a replication-defective virus that was characterized by markedly increased splicing of the full-length genomic RNA to form env RNA. A series of second-site reversion mutations were derived that restored viral replication (Katz et al. 1988; Katz and Skalka 1990). These mutations decreased splicing efficiency and restored the balance of unspliced and spliced RNAs. An analysis of the effect of these different mutations on in vitro splicing efficiency (Fu et al. 1991) confirmed that wild-type RNA was inefficiently spliced. The insertion mutation resulted in the creation of a new branchpoint site for lariat formation upstream of the 3′ splice site and increased utilization of this site. The second-site suppressor mutations reduced the efficiency of in vitro splicing. Some disrupted the new branch site, others in the 3′ exon apparently acted by interfering with formation of the 60S spliceosome complex, and still others produced splicing intermediates that were not efficiently ligated (Bouck et al. 1995). The location of these second-site mutations identified elements within the 3′ exon that could decrease efficiency of splicing of the full-length ASLV messages. Neither viral protein synthesis nor the 5′ splice site was required for these effects. These studies suggest that the efficiency of the env splice acceptor site function directly regulates the cytoplasmic expression of unspliced ASLV RNAs.

Other RNA sequences in addition to the actual splice signals may also modulate retroviral splicing efficiency. For example, a segment of gag acts as a negative regulator of splicing (NRS) for ASLV RNAs (Arrigo and Beemon 1988; Stoltzfus and Fogarty 1989) and heterologous RNAs (Arrigo and Beemon 1988; McNally et al. 1991). The element appears to function at the RNA level and to be most active in proximity to the 5′ splice site (Arrigo and Beemon 1988). Analysis of the effects on in vitro splicing reactions of insertion of the NRS into an adenoviral RNA splicing substrate revealed that the NRS inhibited splicing before the first transesterification step (Gontarek et al. 1993). The NRS contains splice-site-like sequences and was found to interact with two nonconventional snRNPs, U11 and U12, found in the aberrant spliceosome. Mutations in the NRS that eliminated the inhibitory effect were also accompanied by dissociation of U11 and U12, suggesting that these snRNPs may have a role in negative regulation. More recently, the NRS has also been shown to interact with splicing regulatory factors such as SF2/ASF (see below), suggesting the hypothesis that the NRS may act as a nonfunctional “mini-exon” whose recognition by the splicing machinery may result in a complex that inhibits splicing at the authentic 5′ and 3′ splice sites (McNally and McNally 1996). A second inhibitory sequence has been identified in RSV in the 3′ segment of the src intron (Stoltzfus and Fogarty 1989; McNally and Beemon 1992) and has been mapped to 70–148 nucleotides upstream of the 3′ splice acceptor for the src gene (McNally and Beemon 1992). Several other important examples of the contextual effects of retroviral sequences in suppressing splicing efficiency have been seen in the expression of transduced viral oncogenes and in the characterization of retroviral vectors. Efficient expression of the v-rel oncogene in Rev-T virus is dependent on the removal of gag and pol sequences that inhibit production of the spliced v-rel mRNA (Miller et al. 1988). Relatively inefficient splicing of cellular genomic sequences was observed following the insertion of the human chorionic gonadotrophin gene into a replication-competent vector derived from RSV (Sorge and Hughes 1982), suggesting that cis-acting sequences influence the utilization of the normally efficient cellular splice sites.

Recent studies have suggested that some simple retroviruses also contain elements that may directly affect RNA transport (Fig. 19), thus suggesting an additional mechanism for regulation of the cytoplasmic ratios of spliced and unspliced RNAs. A small (219-nucleotide) element in the 3′-untranslated segment of the genome of the type-D Mason-Pfizer monkey virus (M-PMV) (located between the env and the 3′ LTR) can functionally substitute for the Rev protein and its target RRE sequence (see below) in promoting transport of unspliced HIV RNAs (Bray et al. 1994). It is proposed that this element, called the constitutive transport element (CTE), interacts with cellular factors to promote RNA export from the nucleus. A CTE element has also been identified in the M-PMV-related simian retrovirus-1 (SRV-1) type-D retrovirus (Zolotukhin et al. 1994) and shown to be composed of a region of extensive RNA secondary structure (Tabernero et al. 1996). Similar cis-acting elements may also be present in other simple retroviruses. Inhibition of splicing by the ASLV NRS element in the context of a heterologous plasmid (Arrigo and Beemon 1988) is not sufficient to induce nuclear export of unspliced RNA, suggesting that ASLV RNA may also contain an element to promote RNA export from the nucleus. More recent studies have in fact identified such an element in ASLV (Ogert et al. 1996). A 100-nucleotide segment of the 3′-nontranslated region of ASLV was shown to substitute for Rev-RRE interactions in promoting HIV gag expression. This region contained a direct repeat present at both ends of the RSV src gene. Deletion of both direct repeat elements reduced both unspliced RNA in the cytoplasm and viral replication; however, restoration of a single direct repeat allowed efficient cytoplasmic transport of unspliced RNA. Sequences similar to the direct repeat are present in other avian retroviruses, suggesting that these viruses may also contain functional CTEs. Studies are actively being pursued in a number of laboratories to identify CTE-like elements in other retroviruses.

Maintenance of Unspliced RNA in Complex Retroviruses

The HIV Rev Protein

The Rev proteins of HIV and other lentiviruses (Table 3) are responsible for the efficient transport of unspliced and singly spliced viral RNAs to the cytoplasm of infected cells (Fig. 19). HIV Rev is an essential viral protein encoded by a small, multiply spliced mRNA synthesized at early times after infection (for review, see Cullen 1992; Rosen 1992). In the absence of Rev, only very low levels of unspliced, full-length HIV RNAs and singly spliced env, vpu, vif, and vpr mRNAs are found in the cytoplasm; however, high cytoplasmic levels of small, multiply spliced RNAs are present (Feinberg et al. 1986; Sodroski et al. 1986; Malim et al. 1988; Felber et al. 1989; Hadzopoulou-Cladaras et al. 1989). When nuclear and cytoplasmic HIV RNAs are fractionated, both spliced and unspliced HIV mRNAs can be identified in nuclei even in the absence of Rev. These experiments demonstrated that Rev has a critical role in the processing and transport of unspliced or singly spliced HIV RNAs encoding structural genes and enzymatic activities, as well as HIV genomic RNA. Thus, on the basis of the presence of Rev, the HIV life cycle can be divided into an “early” regulatory phase and a “late” productive phase (Kim et al. 1989). The small mRNAs encoding Tat, Rev, and Nef predominate during early times after infection. Apparently, when a threshold level of Rev is produced, unspliced and singly spliced RNAs begin to accumulate in the cytoplasm, allowing productive infection to proceed. Failure to generate this threshold level of Rev may contribute to latent HIV infection (Pomerantz et al. 1990; Malim and Cullen 1991).

HIV-1 Rev is a 116-amino-acid protein that is predominately nucleolar, but it is also found in the nucleoplasm and cytoplasm. Rev functions through binding to a highly structured segment of HIV RNA called the Rev response element (RRE) (Fig. 20, top) (Rosen et al. 1988; Daly et al. 1989; Hadzopoulou-Cladaras et al. 1989; Malim et al. 1989b, 1990; Zapp and Green 1989; Heaphy et al. 1990). The minimal RRE is a 234-nucleotide RNA sequence within the env-coding region (Malim et al. 1989b); more recent studies have suggested that an elongated 351-nucleotide RRE provides optimal Rev function by allowing binding of more Rev molecules (Mann et al. 1994). The RRE is present in all of the RNAs dependent on Rev for their cytoplasmic expression; conversely, it is spliced out of the small Rev-independent RNAs. Not surprisingly, mutations in the RRE can interfere with Rev function, resulting in a phenotype identical to that seen for deletion of the rev-coding sequence (Hadzopoulou-Cladaras et al. 1989; Malim et al. 1990).

Figure 20. Structure of the HIV-1 Rev response element and Rev-binding peptide.

Figure 20

Structure of the HIV-1 Rev response element and Rev-binding peptide. (Top) Predicted secondary structure of the RRE, showing the location of the primary Rev-binding site, indicated in color. Arrows indicate G residues involved in G:G base pairing. (Bottom (more...)

Rev binds as a multimer to RRE RNA through a basic, arginine-rich region present in the amino-terminal half of the protein (Malim et al. 1989a; Kjems et al. 1991; Zapp et al. 1991). One of the several stem-loop structures in the RRE (Fig. 20, top) seems to be the principal RNA target for Rev binding and may serve as the nucleation site for binding of Rev oligomers. A bubble within this region of secondary structure serves as the high-affinity Rev-binding site (Bartel et al. 1991; Heaphy et al. 1991; Holland et al. 1992; Kjems and Sharp 1992; Tiley et al. 1992; for review, see Gait and Karn 1993). NMR structural analysis demonstrated that binding of a Rev peptide to RRE stabilizes non-Watson-Crick, purine-purine base pairing between residues that form the bubble (G:A and G:G) (Battiste et al. 1994; Peterson et al. 1994), as predicted from previous studies (Bartel et al. 1991; Heaphy et al. 1991). A kink in the region of a bulged U residue allows formation of an anti-anti G:G base pair resulting in opening of the major groove allowing basic residues in Rev to bind (Battiste et al. 1995). The α-helical Rev peptide binds deeply within the widened major groove, making base-specific contacts through interactions of arginine and asparagine residues with bases on both sides of the groove (Fig. 20, bottom) (Battiste et al. 1996). The amino acid residues shown by NMR to be involved in these base-specific contacts had previously been shown by genetic (Bartel et al. 1991) and biochemical approaches (Kjems and Sharp 1992) to be critical for Rev-RRE interactions. Additional arginines in the RNA-binding peptide of Rev also interact with the RRE phosphate backbone. A detailed understanding of the molecular interactions between Rev and the RRE could provide interesting therapeutic approaches for HIV infection; for example, aminoglycosides have been identified that specifically block Rev-RRE binding and inhibit HIV infection (Zapp et al. 1993).

For Rev to be functional, it is essential that it binds to its RNA target. In experiments analogous to those performed with Tat and TAR, Rev has been found to be active even if it binds to RRE RNA through heterologous protein-RNA interactions. For example, a chimeric Rev protein containing the RNA-binding segment of the MS2 phage coat protein functions on an HIV RNA target containing the MS2 coat protein-binding site (McDonald et al. 1992; Venkatesan et al. 1992). Cellular proteins also bind to RRE (Vaishnav et al. 1991; Park et al. 1994), although their role in Rev function and the specificity of their interactions with RRE are unclear. Some RRE-binding proteins also bind to the HIV TAR region (Park et al. 1994) and thus may simply be cellular proteins that recognize extensive RNA secondary structure.

In addition to the domains that specify RNA binding and multimerization, Rev also contains a leucine-rich effector domain through which Rev function is mediated, presumably via interactions with cellular proteins involved in RNA processing and/or transport (Malim et al. 1989a, 1990). Mutations in this effector domain result in the generation of a Rev protein that exhibits a transdominant inhibitory phenotype (Malim et al. 1989a). These mutants retain the ability to bind to the RRE but inhibit the function of wild-type Rev protein, apparently through inhibiting interactions with cellular proteins. Interestingly, these transdominant inhibitory HIV Rev proteins are being used in efforts to develop gene therapy for HIV infection through the induction of “intracellular immunization” (Baltimore 1988) to block viral replication in target CD4+ T cells (Malim et al. 1992), and their expression seems to be protective against HIV infection in vivo (Woffendin et al. 1996).

Cellular proteins that bind to Rev have also been identified, and these represent potential targets for Rev action. One such protein exhibits homology with splicing regulatory proteins (Luo et al. 1994): This observation is intriguing in view of the potential mechanisms of Rev function. The protein, YL2, has been shown to potentiate Rev function in transfection assays; however, its significance in HIV-infected cells has not yet been established. Recently, a human nuclear protein related to the nucleoporins (proteins that form the nuclear pores) was identified as a putative cellular partner for Rev by the yeast two-hybrid assay for protein-protein interactions (Bogerd et al. 1995; Fritz et al. 1995). This protein, referred to as Rab (Bogerd et al. 1995) or hRIP (Fritz et al. 1995), is particularly interesting as its interaction with Rev suggests a model of Rev function (see below). Rab/hRIP was shown to synergize with Rev in the activation of a Rev-deficient HIV provirus (Bogerd et al. 1995). Furthermore, Rab/ hRIP clearly binds to the Rev effector domain, and mutations in this domain that block Rev function also result in loss of Rab binding.

It is clear that Rev acts to increase the steady-state levels of unspliced and singly spliced RNAs in the cytoplasm of infected cells. This results in efficient synthesis of Gag, Pol, and Env proteins, as well as in accumulation of genomic RNA for packaging. A number of models have been proposed to explain how Rev acts, including effects on splicing, stability, transport, and translation of RRE-containing RNAs. Recent advances in the understanding of mechanisms regulating the export of proteins and RNAs from the nucleus, coupled with the identification of Rev-interacting proteins, have provided significant evidence in favor of an important direct effect of Rev on the nuclear export of RNAs to which it is bound. It is likely, however, that Rev-mediated effects on different aspects of RNA processing are not mutually exclusive. Different models of Rev action are illustrated in Figure 21 and are briefly summarized below.

Figure 21. Models of HIV-1 Rev action.

Figure 21

Models of HIV-1 Rev action. Possible models of Rev function are illustrated (see text for further discussion). Recent observations strongly suggest an important effect of Rev in directly mediating nuclear export of RRE-containing RNAs through interaction (more...)

Much of the initial data on the effects of Rev mutations strongly suggested that Rev has an effect on the transport of RRE-containing RNA from the nucleus to the cytoplasm. In the absence of Rev, RRE-containing RNAs are retained in the nucleus (Emerman et al. 1989; Felber et al. 1989; Malim et al. 1989b). Similar levels of unspliced and spliced RNAs are found in the nucleus in the presence or absence of Rev, but significant levels of unspliced RNAs appear in the cytoplasm only in the presence of Rev. On the basis of these results, it was proposed that Rev mediates efficient export of RRE-containing RNAs, which causes them to bypass the splicing machinery. Further evidence in support of this model came from the observation that Rev protein shuttles between the nucleus and the cytoplasm and is exported from the nucleus to the cytoplasm in the presence of RRE-containing RNAs (Kalland et al. 1994; Meyer and Malim 1994). Enhancement of nuclear-cytoplasmic export of RRE-containing RNAs by Rev was also demonstrated in microinjected Xenopus oocytes (Fischer et al. 1994). More definitive evidence that Rev promotes export of RNAs to which it is bound came from the identification of a nuclear export signal in Rev, corresponding to the Rev effector domain (Fischer et al. 1995; Wen et al. 1995). This sequence, when attached to heterologous proteins, promotes their export from the nucleus to the cytoplasm following microinjection into nuclei. The Rev nuclear export sequence is related to other peptides involved in nuclear export of either RNAs (the TFIIIA protein involved in the export of 5S rRNA; Fischer et al. 1995) or proteins (PKI: protein kinase inhibitor α; Wen et al. 1995). The fact that an apparent target of the Rev nuclear export sequence is a nuclear protein(s) related to the nucleoporins also suggests that the primary function of Rev is to mediate export of RRE-containing RNAs through a general nuclear export pathway. Recent studies have confirmed this hypothesis. The nuclear export sequences of both TFIIIA and PKI can functionally replace the activation domain of Rev in promoting nuclear export of HIV RNAs (Fridell et al. 1996a,b), and the PKI sequence also binds to Rab/hRIP (Fridell et al. 1996a).

Although it is now clear that the primary effect of Rev is in promoting the nuclear export of RRE-containing HIV RNAs, considerable evidence suggests that Rev may also affect other aspects of the processing of RRE-containing RNAs, including splicing, RNA stability, and even mRNA translation. mRNA splicing efficiency may be linked to the ability of Rev to promote transport of unspliced RNAs. The relatively low efficiency of HIV splicing may allow Rev-RRE interactions to disrupt splicing and target HIV RNAs for nuclear export (Chang and Sharp 1989). Mutation of the 5′ env splice site to alter binding to the U1 snRNA (involved in spliceosome formation) inhibits Rev-dependent env expression; however, cotransfection with U1 RNA containing a compensatory mutation that restores binding to the altered 5′ splice site reestablishes env expression, suggesting that the Rev effect may be linked to association of RRE-containing RNAs with at least some components of the splicing apparatus (Lu et al. 1990). Direct inhibition by Rev peptides of the in vitro splicing of RRE-containing RNAs has also been demonstrated (Kjems et al. 1991). Rev may exert this effect by blocking entry of the U4/U6 and U5 snRNPs into the spliceosome (Kjems and Sharp 1993). It is clear that if Rev does exert a physiologically important inhibitory effect on splicing, specific mechanisms must exist such that the multiple splicing events that generate the small HIV RNAs are not completely blocked (Malim and Cullen 1993).

Another role of Rev appears to be to overcome the effects of cis-acting inhibitory elements present in HIV RNAs (Rosen et al. 1988; Cochrane et al. 1991; Maldarelli et al. 1991; Schwartz et al. 1992a), which have been found in gag, pol, and env. These elements are known as CRS (cis-acting regulatory sequences) or INS (inhibitory sequences). In the absence of Rev-RRE interactions, they inhibit cytoplasmic transport and promote degradation of HIV RNAs. Extensive mutagenesis of a gag INS element led to increased RNA stability and the induction of Rev independence of expression of the HIV gag gene (Schwartz et al. 1992b). These results suggest that cellular factors interacting with INS elements might lead to RNA nuclear retention and degradation. Rev binding to the RRE could shunt INS-containing RNAs toward export from the nucleus to the cytoplasm. In the absence of such export, nuclear degradation could represent a default pathway, mediated by the cis-acting instability elements.

Rev has also been proposed to have effects on the translation of RRE-containing RNAs, by increasing the efficiency of polysome loading (Arrigo and Chen 1991; D'Agostino et al. 1992). In this regard, Rev has been shown to enhance the association of the poly(A)+ binding protein-1 with RRE-containing HIV mRNAs (Campbell et al. 1994), and the Rev protein has been found to associate with eukaryotic translation initiation factor eIF5A (Ruhl et al. 1993). Furthermore, at least one cellular RRE-binding protein may influence translational initiation by acting as an inhibitor of the interferon-induced protein kinase, PKR (Park et al. 1994).

From these various studies, it appears that Rev interactions with RRE may have effects on several levels of HIV RNA processing and function. Rev indeed may act as a “chaperone” to guide RRE-containing HIV RNAs through various RNA transport and processing steps. Studies of Rev have helped to identify a general pathway mediating the export of proteins and RNAs from the nucleus. Further studies of Rev function and Rev interacting proteins will undoubtedly provide important insights into the basic mechanisms responsible for the processing of HIV RNA and the normal metabolism of cellular RNAs.

HTLV Rex Protein

The HTLVs (and the related BLV) encode a protein, Rex (or p27rex), that is quite similar in function to the lentiviral Rev proteins (for review, see Greene and Cullen 1990; Cullen 1992). Like Rev, Rex is required for the cytoplasmic expression of large unspliced or singly spliced HTLV RNAs (Inoue et al. 1987; Hidaka et al. 1988; Hanly et al. 1989). Rex also contains an arginine-rich nucleolar localization signal that mediates binding to a segment of HTLV RNA, the Rex response element (RexRE) (Bogerd et al. 1991; Grassmann et al. 1991; Unge et al. 1991).

HTLV-1 RexRE is formed by a 254-nucleotide element overlapping the U3 and R regions in the 3′ LTR (Seiki et al. 1988; Hanly et al. 1989; Toyoshima et al. 1990). Like the RRE, RexRE RNA is highly ordered, containing multiple stem-loop structures. In addition to mediating the binding and function of Rex, RexRE secondary structure also specifies RNA sequences involved in the proper polyadenylation of HTLV RNAs (see above) by bringing together distant AAUAAA and GU-rich sequences required for poly(A) addition. Unlike the RRE, the positioning of RexRE in the 3′ LTR means that all HTLV RNAs will contain the Rex-binding element. Thus, binding of Rex to RexRE in and of itself will not provide the discrimination between spliced and unspliced RNAs required to ensure proper transport of unspliced RNAs out of the nucleus. Rex function in HTLVs also depends on sequences in the vicinity of the 5′ splice donor (Seiki et al. 1988; Black et al. 1991) in addition to the RexRE. Deletion of 5′ splice donor sequences results in constitutive cytoplasmic expression of unspliced HTLV-1 RNAs, suggesting that sequences in this region may serve as cis-acting repressive elements, similar to those reported for HIV (Seiki et al. 1988). Although additional regulatory signals involving the 5′ splice site could account for differential effects of Rex on spliced and unspliced RNAs, it is still not clear how Rex would distinguish singly spliced from multiply spliced HTLV RNAs.

Evidence for a close functional relationship between Rev and Rex comes from the observation that HTLV-1 Rex can functionally substitute for Rev in promoting the cytoplasmic expression of full-length and singly spliced HIV RNAs (Rimsky et al. 1988). In fact, Rex can restore the replicative ability of an HIV-1 provirus containing a Rev mutation. Rex binds to the HIV RRE at a position distinct from that bound by Rev (Ahmed et al. 1990; Solomin et al. 1990). However, Rev is not able to replace Rex functionally in regulating expression of unspliced HTLV RNAs, possibly because it does not bind to RexRE efficiently (Hanly et al. 1989). Transdominant inhibitory mutants of Rex have been identified that are analogous to those described for HIV (Rimsky et al. 1989; Böhnlein et al. 1991). These Rex mutants still bind to RexRE and RRE, but they inhibit both the effect of Rex on HTLV RNAs and the effect of Rev on HIV RNA. These results suggest that at least some of the cellular targets for these two RNA regulatory proteins are shared by both proteins; in fact, the recently cloned Rev-interacting protein, Rab/hRIP, also interacts with Rex (Bogerd et al. 1995).

Recent evidence suggests that Rex may also promote the stability and/or expression of cellular RNAs. In particular, the expression of IL-2 seems to be synergistically activated by the expression of both HTLV-1 Rex and Tax (McGuire et al. 1993), and Rex has been shown to stabilize IL-2 receptor α RNA (Kanamori et al. 1990). Therefore, transformation by HTLV-1 may involve both transcriptional and posttranscriptional effects encoded by its two regulatory proteins.

Alternative Splice Site Selection in Simple and Complex Retroviruses

In addition to preserving sufficient levels of unspliced RNAs, complex retroviruses, as well as some acutely transforming viruses, must employ mechanisms to regulate the relative amounts of synthesis and/or transport of alternatively spliced RNAs. As noted above, little is known about the cellular mechanisms governing alternative splicing in avian and mammalian cells, mechanisms that must also have a role in alternative retroviral splicing. There has been very little work on alternative splicing in retroviral systems; most available information has come from studies on ASLV and HIV. Two different singly spliced RSV RNAs encode the env and src mRNAs, respectively. Not only is there a balance between unspliced and spliced RNAs in this system, but there is also alternative splicing regulating the relative levels of the env and src RNAs. Mutations in the 3′ splice acceptor sites for either of these RNAs alters the levels of the other, although not always in predictable ways (Berberich and Stoltzfus 1991). A direct role for cellular splicing factors in regulation of these alternative splice sites has been suggested by the observation that mammalian cells infected with RSV or transfected with RSV DNA exhibit very little env RNA. Instead, a doubly spliced src RNA that uses a cryptic splice donor site in env is produced. Infected or transfected chicken cells do not make this doubly spliced RNA (Berberich et al. 1990). A cis-acting RNA element (suppressor of src splicing, SSS) located between the env and src genes seems to inhibit src mRNA splicing in avian cells but not in mammalian cells, apparently through binding to a cellular factor present in chicken embryo fibroblasts (Amendt et al. 1995b). The SSS seems to act in concert with a suboptimal 3′ splice site to limit the efficiency of src splicing (Zhang et al. 1996).

The lentiviruses, especially HIV, provide striking examples of the complexity of alternative splicing that can occur during retroviral infection. This selection process appears to be regulated. Mutations that disrupt competing splice acceptors can cause shifts in the splicing patterns and can affect viral infectivity (Purcell and Martin 1993). In vitro splicing assays have been used to identify elements within HIV that regulate splice site selection. A 20-nucleotide splicing regulatory element, called an exon splicing silencer (ESS), was identified within the second exon (first coding exon) of tat (Amendt et al. 1995a). This element inhibits splicing at the upstream 3′ splice acceptor for the first exon of tat, resulting in increased use of downstream 3′ splice sites that serve as acceptors for rev and env mRNAs (Amendt et al. 1994). Mutations in this element increase production of tat RNAs in vivo. Similarly, complex regulation exists for utilization of the shared 3′ splice site between the exons of tat and rev, which is responsible for removal of the env intron to generate tat and rev mRNAs (Amendt et al. 1995a; Staffa and Cochrane 1995). Regulation of the splicing efficiency of this intron is due both to suboptimal splice signals and to the presence of exon splicing enhancers and silencers. Thus, although detailed mechanisms regulating HIV splice site usage remain unknown, regulation at this level has effects on HIV Tat and Rev expression, on viral infectivity, and thus ultimately on the pathogenesis of HIV disease.

Copyright © 1997, Cold Spring Harbor Laboratory Press.
Bookshelf ID: NBK19413