Translational Control of Viral Gene Expression in Eukaryotes

Abstract

As obligate intracellular parasites, viruses rely exclusively on the translational machinery of the host cell for the synthesis of viral proteins. This relationship has imposed numerous challenges on both the infecting virus and the host cell. Importantly, viruses must compete with the endogenous transcripts of the host cell for the translation of viral mRNA. Eukaryotic viruses have thus evolved diverse mechanisms to ensure translational efficiency of viral mRNA above and beyond that of cellular mRNA. Mechanisms that facilitate the efficient and selective translation of viral mRNA may be inherent in the structure of the viral nucleic acid itself and can involve the recruitment and/or modification of specific host factors. These processes serve to redirect the translation apparatus to favor viral transcripts, and they often come at the expense of the host cell. Accordingly, eukaryotic cells have developed antiviral countermeasures to target the translational machinery and disrupt protein synthesis during the course of virus infection. Not to be outdone, many viruses have answered these countermeasures with their own mechanisms to disrupt cellular antiviral pathways, thereby ensuring the uncompromised translation of virion proteins. Here we review the varied and complex translational programs employed by eukaryotic viruses. We discuss how these translational strategies have been incorporated into the virus life cycle and examine how such programming contributes to the pathogenesis of the host cell.

Perhaps nowhere in nature is a parasitic relationship as well defined as that which occurs between a virus and its host cell. Viruses rely on the host cell for propagation, utilizing cellular machinery for the replication and assembly of viral components and the release of progeny virions. Whether possessing a DNA or RNA genome, the eukaryotic virus exhibits a general life cycle that is initiated through interaction with its cognate receptor(s) on the surface of the host cell (117) (Fig. 1). After virion adsorption and internalization, uncoating exposes the viral genome and associated proteins to the host milieu, whereupon genome replication and transcription take place. The translation of viral RNA is followed by the assembly of structural proteins, packaging of the viral genome, and eventual release of progeny virions. While some viruses encode or carry the enzymatic machinery required for autonomous genome replication and/or transcription, others recruit host polymerases to carryout this task (117). In contrast, viruses do not encode or carry the machinery for mRNA translation. Thus, the ensuing stage of viral protein synthesis is completely dependent on the translational machinery of the host cell (Fig. 1). Not surprisingly, viruses have devoted much attention to this dependency and have evolved strategies that reduce the impact of translational dependence on viral replication. As discussed in this review, these strategies are themselves limited by the nature of the viral RNA, the cellular translation machinery, and the translation regulatory pathways of the host cell.

FIG. 1.

General model of eukaryotic viral replication. Viral particles are shown in black. Viruses recognize their target cell through interaction with specific receptors and/or other components on the cell membrane. Interaction with the host cell induces cell membrane penetration and virion internalization. Virion uncoating releases the viral genome, whereupon it is available for transcription and translation. Poxviruses and the RNA viruses (with the exception of retroviruses) replicate in the cytoplasm. Transcription of all other DNA viruses takes place in the nucleus. Transcription and genome replication are followed by the cytoplasmic stages of mRNA translation and virion assembly. Release of mature virions may include membrane lysis and death of the host cell. Adapted with modification from reference 322.

This treatise presents an overview of translation strategies used by viruses that infect the cells of higher eukaryotes. Where appropriate, we have focused on specific virus systems to present examples of the diverse mechanisms by which viruses overcome the problems of translational dependence. For complementary material on mRNA translation, virus-host interactions, host antiviral pathways, and the virus-host interactions of lower eukaryotes and bacteria, we direct the reader to several fine texts and reviews (1, 10, 117, 137, 179, 199, 303, 324, 353, 415, 416, 428, 455, 480, 484). We begin this review with a brief overview of the current models for eukaryotic mRNA translation, including points of translational control, and effects on host translation due to virus infection. This is followed by examples of translation strategies that are dependent on the structure of the viral mRNA and those that are directed at recruitment and modification of the translation machinery and other host factors. Given the recent emphasis in the translational control field on identifying and characterizing cellular signaling pathways that govern mRNA translation (43, 120, 433), we have included discussion of how viruses might exploit these pathways to facilitate completion of their translational programs. Attention is directed to the ways in which disruption of host translational control pathways may contribute to viral pathogenesis and disease progression. Finally, we conclude with a section describing the prospects of targeting viral mRNA translation for antiviral therapy, as well as perspectives for future research in the increasingly overlapping disciplines of virology, viral pathogenesis, and translation control.

Overview of Eukaryotic mRNA Translation and Sites of Viral Regulation

Translation in eukaryotes is a complex multistep, multiprotein process (198, 335). As with most complex biochemical reactions, it is subject to strict regulatory controls, and is extremely sensitive to both the intracellular and extracellular environments (43, 163, 232, 281, 324, 433). In general, the translation of a given mRNA can be modulated in response to nutrient availability, mitogenic stimulation and cell cycle regulation, stress, and, as described herein, viral infection (reviewed in detail in reference 199). It is also increasingly clear from research spanning the past several years that regulation of mRNA translation is critical for maintaining control of cell growth (96, 287, 429). As presented in the “Viral persistence and translational control” section (below), disruption of the major translation checkpoints and signaling cascades renders cells unable to respond to translation-modulatory signals and may constitute a mechanism of oncogenic transformation (67). The following sections provide a general overview of viral translational programming and eukaryotic mRNA translation. Major sites for virus regulation of translation are noted, and they are discussed in detail in this review.

Viral translational programming.

Viruses face enormous pressures to maintain a “functional” genome size, which greatly influences the rate and efficiency of viral replication. Thus, host translational dependence may in part reflect the limitations placed on viral replication due to the enormous genome capacity that would be needed to encode the components for autonomous viral protein synthesis (335). This idea is supported by the highly specialized nature of the protein synthetic machinery, which encompasses well over 30 different gene products and yet remains highly conserved between the prokaryotic and eukaryotic kingdoms (335, 470). Eukaryotic viruses have evolved effective means of exploiting their innate translational dependence through mechanisms of translational programming. This is the process in which eukaryotic viruses (i) redirect the host translation machinery to favor viral protein synthesis and (ii) control the expression of their own gene products. The latter is especially important for the RNA viruses, which have limited transcriptional control and rely heavily on translational control strategies to modulate viral gene expression.

Translational programming, such as the use of regulatory upstream open reading frame(s) (uORFs), overlapping reading frames, multicistronic transcripts, and termination control, allows viruses to conserve the functional genome size by making efficient use of genome coding capacity. In general, the mechanisms of translational programming are intrinsic to the structure of the viral mRNA itself. As summarized in Table 1, structural elements within a viral mRNA that affect translational efficiency or impart translational control include the length and structural complexity of the 5′ and 3′ untranslated regions (UTR), the position and context of the initiator AUG codon, the stability and accessibility of the of the m⁷G cap and the cap-binding complex, and the presence of uORF(s) preceding the major cistron (137, 148, 149, 269, 270, 322, 329, 430, 431). In addition, cis-acting sequence elements that recruit or bind trans-acting factors can impart an additional level of translational control to viral mRNA by facilitating translational selectivity (23, 76, 187, 244, 361, 362, 394, 411). As described in the following sections, virus translational programming affects all levels of the translation process, including translation initiation, elongation, termination, and host translational control signaling pathways.

TABLE 1.

mRNA structural features that confer translational control.

Structural feature	Effect on translation	References
Length of 5′ UTR	Influences scanning. Long 5′ UTR may impede initiation.	11, 107, 432
Secondary structure of 5′ UTR and coding region	Complex 5′ UTR structures may impede scanning. IRES structures promote cap independence and allow the ribosome to largely avoid scanning. Pseudoknot structures promote frameshifting and recoding. May mediate binding with trans-acting factors.	12, 23, 107, 118, 221, 223, 286, 396, 430
Sequence context of the initiation codon	Imparts ribosome selectivity for first AUG codon. “Weak” AUG codon promotes leaky scanning.	221, 271, 272
M7G cap	Promotes mRNA stability and interaction with eIF4F. Facilitates the translation of most cellular mRNAs. Accessibility to initiation factors may influence translational efficiency.	17, 136, 300, 335, 432, 442, 457
uORF	May impede ribosome scanning to downstream cistron(s). Multiple uORFs promote the selective translation of GCN4 in yeast.	146, 148, 202, 342
Poly(A) tail	Length imparts stability and translational efficiency to mRNA. Interaction with PABP mediates association with cap-binding complex on the mRNA.	27, 136, 225, 289, 393, 401
3′ UTR	Mediates closed-loop translation complex via PABP interaction. Structural complexity may influence translational efficiency and interaction with trans-acting factors.	63, 75, 225, 276, 401, 430, 452, 466, 479, 492
Codon usage	Use of nonabundant tRNAs may impede elongation. Influences frameshifting and recoding.	12, 40, 152
trans-acting factors	Specific RNA sequence and/or structural motifs promote interaction with RNA-binding proteins, which may influence translational efficiency.	63, 335, 426, 492

Translation initiation.

The majority of control over cellular mRNA translation occurs during initiation. Translation initiation is the process in which the mRNA assembles into a macromolecular complex with the components required for protein synthesis, including the eukaryotic initiation factors (eIF) and elongation factors (EF). Figure 2 shows the major steps in the cap-dependent translation initiation process and important sites of virus regulation (for comprehensive reviews of translation initiation, the reader is referred to references 163, 221, 335, and 359). Initiation begins with the binding of initiator methionyl-tRNA (Met-tRNA_i) to the 40S ribosomal subunit. This step is facilitated through the formation of an eIF2–GTP–Met-tRNA_i ternary complex (462). The recent discovery and functional analyses of eukaryotic homologues of prokaryotic initiation factor 2 (IF2) indicates that Met-tRNA_i delivery also proceeds via a more general, universally conserved mechanism (60, 291). In this case, IF2 does not participate in the formation of a ternary complex but, rather, may bind directly to the ribosomal A site and facilitate binding of the Met-tRNA_i to the ribosomal P site during translation initiation. IF2 activity is not subject to direct regulation and therefore may not contribute to the control of mRNA translation. In contrast, formation of the eIF2-dependent ternary complex and its delivery of Met-tRNA_i to the 40S ribosomal subunit can constitute a rate-limiting step when the alpha subunit of eIF2 (eIF2α) is phosphorylated by specific protein kinases (see below) (65, 335). Phosphorylation of eIF2α thus represents a major point of control over the translation initiation process. eIF2α phosphorylation dramatically alters the efficiency and rate of mRNA translation and is a critical component of antiviral and cell growth control pathways (243, 321, 322, 335). eIF2 directs the ternary complex to the 40S ribosomal subunit to form a 43S pre-initiation complex that includes eIF3 (Fig. 2). eIF3 facilitates binding of the 43S pre-initiation complex to the mRNA via the cap-binding complex, eIF4F, that has been assembled around the mRNA m⁷G cap structure (335, 359).

FIG. 2.

Schematic illustration of eukaryotic mRNA translation and major sites of viral regulation. Details of the translation process are described in the text. Translation initiation factors are shown by their letter and number designation (335). 40S and 80S denote the small ribosomal subunit and the elongating ribosome, respectively. 1, Ternary-complex formation and assembly of the 43S pre-initiation complex; 2, assembly of the cap-binding complex and ribosome loading onto the mRNA; 3, Ribosome scanning to the first AUG codon, recycling of eIF2-GDP, and joining of the 60S ribosomal subunit. TER denotes a translation termination codon. Major sites for viral control of translation and mechanisms of translation control are shown in the surrounding boxes. Not shown is the mRNA 3′ UTR, which can also influence translational efficiency. aa, amino acid.

Cap-binding reaction.

Assembly of the eIF4F complex on the mRNA is dependent on the eIF4E component of this complex, which recognizes and binds the m⁷G cap (173). The affinity of eIF4E for the m⁷G cap constitutes a second major control point in the translation initiation pathway and is subject to variation through eIF4E phosphorylation (237, 315, 433). In addition, the cap-binding activity of eIF4E can be blocked through the formation of an eIF4E-eIF4E binding protein (4E-BP) complex, resulting in inhibition of cap-dependent translation (367, 433). Formation of the eIF4E/4E-BP complex itself is subject to regulation through 4E-BP phosphorylation and dramatically affects cell growth control by altering the efficiency and selectivity of mRNA translation (reviewed by Sonenberg and Gingras [433]). As discussed in detail below, these regulatory steps are targeted by a group of viruses, which are best defined by the family of picornaviruses and includes poliovirus and encephalomyocarditis virus (EMCV) (155, 174, 322). These viruses initiate translation through a cap-independent mechanism that involves internal ribosome entry through use of the internal ribosome entry site (IRES); virus-mediated cleavage of the 220-kDa cap-binding protein, eIF4G (161, 174, 283); and dephosphorylation of the 4E-BPs to sequester eIF4E in an inactive eIF4E/4E-BP complex (Fig. 2) (155). IRES-mediated translation requires specific cis-acting sequences within the viral RNA that mediate the interaction with trans-acting host factors. Thus, the global process of IRES-mediated translation essentially eliminates the competition for host factors from cap-dependent cellular mRNA translation, favoring the translation of viral mRNA.

Ribosome scanning and AUG site selection.

Following association with the mRNA, the 43S preinitiation complex begins scanning from the 5′ end of the mRNA or the site of ribosome entry (as in the case of cap-independent translation) and continues scanning until the Met-tRNA_i interacts with the initiator AUG codon. Ribosomal scanning is not always compatible with mRNAs that possess a long and/or structured 5′ UTR. As described in “Mechanism and control of viral mRNA translation” (below), viral mechanisms to overcome the inefficiency of scanning the 5′ UTR and to bypass the host shutoff phenomenon include the use of internal ribosomal entry on the mRNA and the ribosomal shunt (93, 223, 494). These mechanisms allow the preinitiation complex to effectively avoid a large part of the 5′ UTR and begin scanning within the region of the initiator AUG codon on the viral mRNA (Fig. 2).

Once the Met-tRNA_i associates with the initiator AUG codon, GTP is hydrolyzed from the ternary complex, bound initiation factors are released, and the 60S ribosomal subunit joins the preinitiation complex. The resulting 80S initiation complex then mediates the elongation phase of translation. In this model, initiation begins at the 5′-proximal AUG codon (221, 335). However, the AUG site selection for translation initiation is dependent in part on the context of the AUG codon, where the canonical accAUGg sequence (initiation codon in capitals) exerts the highest preference for initiation (221, 270). Departure from this sequence is associated with leaky scanning, in which the preinitiation complex will recognize a noncanonical or weak AUG only at a low frequency and scans past to initiate translation at a downstream codon more closely matching the canonical initiator AUG (Fig. 2) (221). Leaky-scanning initiation of translation is popular among viruses, and in retroviruses it can provide a mechanism for achieving defined stoichiometric ratios of translation products (412, 413).

Elongation.

During the elongation phase of translation, the mRNA is associated with multiple 80S ribosomes, or polyribosomes, as amino acid residues are sequentially placed on the carboxyl end of the growing peptide chain. In many virus systems the replicative cycle is demarked by early- and late-stage events that can be distinguished by the differential recruitment of viral mRNA into polyribosomal complexes at specific times after infection. As with herpes simplex virus type 1 (HSV-1), this often coincides with the synthesis of latency factors and determinants of virulence (114, 159, 279). The process of translation elongation itself is subject to viral regulatory control (Fig. 2). Elongation control mechanisms include ribosomal frameshifting (108), functional recoding (151), and virus-directed modification of EF-1 (257); the first of these is prevalent in retroviruses and reveals otherwise cryptic ORFs within the viral mRNA (78, 108, 109, 151).

Termination.

The process of translation termination occurs when the translating 80S ribosome encounters an in-frame termination codon within the template mRNA. The termination codon is recognized by a release factor, which mediates the hydrolysis of the peptide chain from the bound tRNA (335, 501). This results in the release of the nascent polypeptide from the 80S ribosome and leads to the eventual dissociation of the ribosomal subunits. Once termination has occurred, the 40S subunit is free to continue scanning the mRNA (Fig. 2). In multicistronic transcripts, termination can be followed by reinitiation at the downstream cistron, subject to ternary-complex availability (221). However, reinitiation is usually very inefficient, and the presence of a uORF can confer limitations to the translational efficiency of the major, downstream ORF. As described in “Frameshifting” (below), this termination-reinitiation translational control mechanism is prevalent among viruses and is used to control the synthesis of specific viral gene products (148).

Improving translation efficiency: the closed-loop model of mRNA translation.

Since the discovery of 3′ polyadenylation in eukaryotic mRNA, it has become quite clear that the poly(A) tail imparts stimulation of mRNA translation in eukaryotes (reviewed by Jacobson [225]). More recent analyses indicated that the translation-stimulatory function of the poly(A) tail was due, in part, to the actions of the poly(A)-binding protein (PABP). In mammalian cells, PABP interacts with elements of the cap-binding complex assembled on the 5′ end of the mRNA, thus rendering a “closed-loop” translation complex (Fig. 3) (138, 225, 401). PABP promotes the closed loop by binding to eIF4G and to PABP-interacting protein 1 (Paip-1) (75). Paip-1 interacts with components of the mRNA cap-binding complex, including eIF4G and the eIF4A helicase (306). Analyses of translation initiation in yeast and plants indicate that the interaction between PABP and eIF4G stimulates mRNA translation (289, 452, 476). The proximity of the mRNA ends provided by the closed-loop translation complex is thought to contribute to the stability of the mRNA and the 5′ cap complex and to provide for the efficient recruitment and recycling of ribosomal subunits (225). Thus, the overall effect of the closed loop is to increase translation efficiency. Viruses exploit the closed-loop translation complex as a means of redirecting the host translation machinery to favor viral mRNA translation. As described below, viruses accomplish this by targeting PABP and disrupting the interaction of the mRNA ends, resulting in attenuation of host mRNA translation (233, 259, 376).

FIG. 3.

Model of the closed-loop mRNA translation complex. The mRNA-bound eIF4F initiation complex interacts with the 3′ end of the mRNA via PABP. Poly(A) sequences within the 3′ UTR direct PABP binding to the mRNA. PABP mediates interaction with the cap-binding complex either directly through eIF4G (4G) (452) or indirectly through an eIF4G, eIF4A (4A)-dependent interaction with Paip-1 (75). Assembly of the closed-loop complex may stabilize the interaction of the 40S ribosomal subunit with the mRNA (225, 401).

TRANSLATIONAL CONTROL OF VIRAL GENE EXPRESSION

Translational dependence has driven viruses to adopt translational programming that maximizes efficiency and facilitates the selective translation of viral mRNA over the endogenous host transcripts. Viral translation strategies have evolved to utilize both the advantages and the limitations inherent within the cap-dependent host translation process. These range from cap-dependent translation competition strategies to cap-independent strategies of IRES-mediated translation initiation. As discussed below, such strategies allow viral mRNA translation to persist, even under the extreme conditions imposed by the host shutoff phenomenon, which severely limits cellular metabolism. This section describes the various translation strategies utilized by eukaryotic viruses to overcome the problems associated with translational dependence and concludes with a discussion of how viral translational programming may present novel targets for the development of anti-viral therapeutics.

Advantages and liabilities of cap-dependent host translation.

The majority of mRNA translation within eukaryotic cells is dependent on the m⁷G cap, a unique structure present at the 5′ terminus of the mRNA (335). The 5′ cap promotes mRNA stability and nuclear export and provides for various levels of control over the translation initiation process. Cap-dependent control of mRNA translation confers several advantages to the cell. First, and perhaps most importantly, cap dependency allows the cell an immediate mechanism through which to control gene expression by modulating the assembly and activity of cap-binding complex components. Second, cap-dependency provides selectivity of translation by combining the translational regulatory properties inherent within a specific mRNA with those due to modification of the cap-binding complex. Translational control thereby allows the cell to fine-tune gene expression by stimulating or repressing the translation of specific mRNAs, usually through the reversible phosphorylation of translation factors (335).

While cap-dependent translation clearly affords several advantages to the host cell, it also presents liabilities that are effectively exploited by viruses. Cap dependency requires an intact pool of specific initiation factors, namely, the components of the eIF4F cap-binding complex (100, 292, 335, 457, 468). Moreover, it necessitates the capping and nuclear export of mRNAs. Viruses have learned to disrupt these processes in order to reprogram the host cell toward the synthesis of viral proteins. Viral disruption of cap-dependent host translation contributes to the host shutoff that is often observed during productive infections (10).

Host Shutoff and Selective Translation of Viral mRNA

Host shutoff is the process in which cellular macromolecular synthesis is suppressed due to viral domination of host metabolism that occurs during infection (reviewed in reference 10). Host shutoff is not an absolute; not all virus infections exhibit host shutoff, and shutoff is not always required to facilitate viral replication. Within the many viral systems in which host shutoff is known to occur, shutoff ultimately favors the translation of viral mRNA over endogenous host transcripts, although host shutoff itself may not be directly attributed to viral disruption of host mRNA translation (1, 10). The selective translation of viral mRNA during the host shutoff is clearly a multicomponent process that has been attributed to a variety of factors. These include viral perturbation of intracellular ion concentration (144) and nucleotide metabolism (215, 244, 279), alterations in RNA stability, processing, and export (119, 245, 311, 352, 497), and the recruitment of specific host factors (249, 294). From a simpler perspective, the selective translation of viral mRNA during host shutoff may reflect a general competition between viral and host mRNA for the translational machinery. For example, host shutoff in cells infected with vesicular stomatitis virus (VSV) is coupled to the selective translation of viral mRNA (Fig. 4, lanes 3 and 4). Interestingly, however, the abundance and stability of cellular mRNAs and their efficiency of translation initiation remain unaltered (54). Examination of VSV-infected cells revealed that the preferential translation of viral mRNAs was a result of ribosome competition from an overwhelming abundance of viral mRNA (309). At the other end of the spectrum is the host shutoff that occurs during picornavirus infection. In this case, the shutoff of host protein synthesis and selectivity for viral mRNA translation is clearly a virus-directed event mediated, in part, through cleavage of eIF4G by the virus-encoded 2A protease (2A-pro) (174, 273, 468). Cleavage of eIF4G by 2A-pro disrupts cap-dependent translation initiation to favor the IRES-mediated translation of the picornavirus mRNA (Fig. 4; also see Fig. 14) (23). Similarly, the host shutoff in cells infected with influenza virus features a strong selection for viral mRNA translation (Fig. 4, lanes 5 and 6). However, unlike the picornaviruses, influenza virus mRNA translation is cap dependent (251). In this case, the predominance of viral protein synthesis is facilitated, in part, by virus-mediated endonucleolytic cleavage of the host mRNA m⁷G cap, subsequent mRNA destabilization, and the dephosphorylation of eIF4E (244). The selectivity of viral mRNA translation is then mediated through the recruitment of the cellular G-rich sequence factor 1 (GRSF-1) protein and other host factors to the 5′ leader sequence of the influenza virus mRNAs (361, 362).

FIG. 4.

Virus-induced shutoff of host cell protein synthesis. Murine NIH 3T3 cells (lanes 1 to 4) or Madin-Darby bovine kidney cells (lanes 5 and 6) were mock infected (U) or infected (I), respectively, with EMCV (lanes 1 and 2), VSV (lanes 3 and 4), or influenza virus (lanes 5 and 6). To visualize the virus-induced host shutoff of protein synthesis and the concomitant shift to viral protein synthesis, proteins were biosynthetically labeled by the addition of [³⁵S]methionine to the culture medium. Protein equivalents from mock-infected and virus-infected cells were separated by gel electrophoresis and visualized by autoradiography of the dried gel. Arrows denote the positions of viral proteins. The positions of molecular mass standards are indicated in kilodaltons.

FIG. 14.

Domain structure of eIF4G. The arrow points to the site of cleavage by poliovirus 2A-Pro. Shaded areas indicate the eIF4E-, eIF3-, and eIF4A-binding domains. The bar indicates the region responsible for binding to Mnk1. Numbering refers to the prototypic eIF4GI (153).

Host shutoff can be seen as beneficial for viruses, since it places cellular resources largely at their disposal. However, the shutoff phenomenon also presents several challenges to the virus, not least of which is maintaining the integrity of the host cell long enough to complete the virus replicative cycle. This is especially important from the standpoints of translational dependence and viral persistence, in which the virus must ensure that the host translation machinery remains competent for the synthesis of viral proteins. Problematically, however, the metabolic repression and stress of host shutoff are potent inducers of cellular apoptosis and translational control programs that function within the cellular antiviral response to block viral infection (179). Viruses have taken a two-pronged approach to these problems of host shutoff, and they encode mechanisms to (i) disrupt host apoptotic programs (353) and (ii) control the antiviral translational response imposed through the phosphorylation of eIF2α (65, 66). As described in detail in “Viral modification of cellular factors” (below), eIF2α phosphorylation by the cellular serine/threonine protein kinase (PKR) presents a translational blockade to viral replication (68, 131). Disruption of host apoptosis and the phosphorylation of eIF2α therefore facilitates viral replication by maintaining host cell integrity and ensuring translational competence during host shutoff.

The relationship between translational control, apoptosis, and viral infection has been an intense area of study in recent years. The emerging picture now suggests that translational suppression through eIF2α phosphorylation is an important component of apoptotic programming (450). Thus, disruption of eIF2α phosphorylation may serve the dual purpose of maintaining the translational competence of the host cell and preventing apoptosis during host shutoff. This idea is supported by the many studies of vaccinia virus replication in which the viral K3L and E3L gene products have been implicated in disrupting eIF2α phosphorylation and blocking apoptosis (52, 55, 82, 83, 135, 255, 420). Moreover, studies by Roizman and colleagues have demonstrated that disruption of eIF2α phosphorylation by the HSV-1 γ₁34.5 gene product was a requisite for sustained translational competence and viral persistence during the host shutoff induced by HSV-1 infection (188, 190). Influenza virus similarly ensures that eIF2α phosphorylation is blocked and translational competence is maintained during host shutoff (244, 294). However, rather than preventing shutoff-induced apoptosis, influenza virus may delay or reprogram apoptosis to facilitate cell lysis and virion release during late-stage infection (116, 445, 446). In closing this section, it is important to note that maintenance of translational competence during host shutoff may ultimately contribute of viral pathogenesis. As described in “Viral persistence and translational control” (below), the ability to suppress mRNA translation is a key component for the control of cell growth. In persistent viral infections, such as those by hepatitis C virus (HCV) or the DNA tumor viruses, constitutive modulation of host translational control pathways and release of translational suppression may make important contributions to viral oncogenesis (125, 134, 248, 444).

Mechanisms and Control of Viral mRNA Translation

Viruses utilize the canonical translation factors and machinery of the host cell to facilitate completion of their translational programming. Figure 5 depicts various means by which viruses implement their translational programming toward the common end of synthesizing viral proteins and completing the virus life cycle. Reflecting the nature of the virus-host relationship itself, the host cell has evolved countermeasures that impose blockades upon viral protein synthesis. As described below, viral translational programming often includes mechanisms to manipulate the host translational machinery and overcome these antiviral blockades.

FIG. 5.

Viral mechanisms of translational programming. The top diagram shows structural features of a representative mRNA containing a m⁷G cap and consisting of a series of overlapping and nonoverlapping reading frames (denoted by rectangles). The first reading frame is indicated by an AUG initiation codon and is preceded by a 5′ UTR. Upright arrows indicate translation initiation of the corresponding reading frame(s), resulting from the mechanisms listed at left. Specific viruses examples presented within the text are listed at right. Depiction of the IRES and ribosome shunt includes the relevant stem-loop structures within the 5′ UTR. Arrow shows ribosome bypass, or shunting, around the stem-loop. Figure adapted with modification from reference 322.

Internal ribosome entry.

Translation initiation, mediated through the internal entry of ribosomes onto the substrate mRNA, was first found in 1988 during studies of poliovirus and EMCV replication (227, 370). Examination of the nucleotide sequence of the picornavirus 5′ UTR has revealed a region of significant secondary structure spanning approximately 500 nucleotides (nt) and punctuated by multiple AUG codons (222, 228). This region was initially known as the ribosome landing pad and later termed the internal ribosome entry site (IRES) (for detailed reviews of the IRES, see references 193, 223, and 329). Translation studies performed in vitro and in vivo demonstrated that the IRES could confer internal ribosome entry to a downstream ORF when placed between the cistrons of a multicistronic mRNA (226, 227, 370). Moreover, incorporation of the IRES to precede the ORF of a circular mRNA facilitated ribosome entry and translation of the circular cistron (57). These studies concluded that the IRES is a genetic element that facilitates internal ribosome entry and mRNA translation independent of the m⁷cap structure. Since then, the observation of IRES-mediated translation has been extended to include other virus families, most notably the other picornaviruses and the members of the genera Pestivirus and Hepacivirus of the family Flaviviridae (which include bovine diarrhea virus and HCV, respectively [Table 2]) (223). We also note that IRES-mediated translation of certain cellular mRNAs has also been identified and may constitute a minor proportion of total cellular mRNA translation within the cell (223, 234, 235, 347).

TABLE 2.

Representative viruses that utilize IRES-mediated translation

Group	Representative members
Enteroviruses and rhinovirusesa
Human rhinoviruses	Common cold virus/many serotypes
Enteroviruses	Poliovirus, coxsackieviruses, enterovirus 70, echoviruses
Cardioviruses and aphthovirusesa
Cardioviruses	Encephalomyocarditis virus, mengovirus
Aphthoviruses	Foot-and-mouth disease virus
Hepatovirusesa	HAV
Pestiviruses	Bovine diarrhea virus
Hepaciviruses	HCV

^aGroupings are based on picornavirus IRES sequence homologies (42, 192, 483). 

Picornaviruses, pestiviruses, and hepaciviruses carry one copy of an uncapped, single-stranded RNA of positive polarity that functions directly as the viral mRNA and substrate for translation (117). These virus families translate their genomic RNA as a single large polyprotein that is posttranslationally processed into distinct structural and nonstructural polypeptide products through a series of proteolytic cleavage events (for reviews of the replication of picornaviruses, pestiviruses, and hepaciviruses, see references 23, 33, 117, 390, and 398). Picornavirus infections generally exhibit an intense host shutoff (with the exception of hepatitis A virus [HAV]) that is characterized by IRES-mediated selective translation of the viral mRNA (Fig. 4, lanes 1 and 2) and rapid lysis of the host cell within 6 to 12 h after infection (398). In contrast, the prototypic hepacivirus, HCV, mediates persistent infection in which the effects, if any, on host shutoff are less well understood. What, then, is the role of the IRES element in the contrasting life cycles of these two virus families?

The functional role of the IRES is perhaps best understood by considering IRES structure, the context of the IRES within the 5′ UTR, and the advantages conferred by IRES-mediated translation. IRES secondary structure has been concisely modeled from several different viruses, and the prototypic IRES structures in poliovirus, EMCV, HAV, and HCV are depicted in Fig. 6. The extensive secondary structure of the IRES makes the long viral 5′ UTR (which ranges from 341 nt in HCV to over 1,400 nt in various picornaviruses) incompatible with standard 5′ ribosome entry, scanning, and AUG site selection. Among picornaviruses, the IRES itself spans approximately 450 nt and begins a variable distance from 5′ terminus of the RNA (101). The enteroviruses and rhinoviruses comprise a structurally conserved IRES group that is distinct from a second IRES group represented by the cardioviruses and aphthoviruses (193, 223). A third and structurally distinct picornavirus IRES group is represented by HAV (156).

FIG. 6.

IRES structure. Structural representation of the 5′ UTR from EMCV (top left), poliovirus (top right), HAV (bottom left), and HCV (bottom right). The major stem-loops are labeled according to previous designations (42, 193, 206, 483). The region encompassing the IRES is underlined. Pyrimidine-rich sequence elements are shown as solid rectangles. AUG denotes the position of the translation initiation codon. The box on the HCV IRES denotes the core protein-coding region.

The IRES groups also diverge in relation to the position of the authentic initiator AUG codon. Both the cardiovirus-aphthovirus IRES group and HAV initiate translation from the AUG codon at the immediate 3′ boundary of the IRES (193). Thus, the site of ribosome entry actually corresponds to the initiation codon. In contrast, the enteroviruses and rhinoviruses initiate translation from an AUG codon located approximately 40 and 160 nt downstream, respectively, from the 3′ IRES boundary (23, 223). In this case, the ribosome scans the mRNA from the point of ribosome entry to the authentic AUG codon, in accordance with the conventional scanning model. The prototypic HCV IRES diverges in both length and structure from the three picornavirus IRES groups (Fig. 6). The actual boundaries of the HCV IRES have yet to be precisely defined, but initial studies suggest that the IRES begins approximately 20 nt from the RNA 5′ terminus and extends at least 30 to 40 nt into the actual coding region of the viral core protein (211, 391). Ribosome entry on the HCV RNA takes place within the IRES itself rather than at the 3′ IRES boundary. This coincides with the actual site of translation initiation, which resides approximately 40 nt internal to the 3′ IRES boundary (391).

Structural conservation between picornavirus IRES groups is limited to a 10- to 15-nt pyrimidine-rich sequence element located near the 3′ IRES boundary (101) and, to a lesser extent, a common 3′ structural core related to the group I intron (290). The significance of the latter is not clear, although it may reflect common structural features of RNA required for the assembly of mRNP complexes. Several independent studies have identified the pyrimidine-rich sequence element as a conserved IRES feature, indicative of a role for this element in IRES-mediated translation. Indeed, partial deletion or purine substitution of the pyrimidine-rich sequence abolished the function of the poliovirus IRES and, to a lesser extent, limited the translation efficiency of the EMCV IRES (192, 351). Together, these experiments demonstrated that the number of bases residing between the 3′ end of the pyrimidine-rich sequence and the actual initiator AUG codon was an important variable influencing the site of ribosome entry and AUG codon selection. These results suggested that (i) the length of the pyrimidine-rich sequence element and its proximity to the initiator AUG codon may influence translation start site selection and (ii) this sequence element may function in the process of IRES-mediated recruitment of the translation initiation complex to the site of initiation. The latter notion is supported by the findings that several host proteins, including the cellular poly(rC)-binding protein 2 (PCBP2) and the pyrimidine tract-binding protein (PTB), may interact with the pyrimidine-rich sequence element and affect the translation efficiency of the picornaviral IRES (34, 35, 238, 264, 394).

Interestingly, the HCV genome contains a pyrimidine-rich sequence element located just outside the IRES, at the 3′ end of the core-coding region (206) (Fig. 6). Recent analyses indicate that the HCV core region pyrimidine-rich sequence functions in cis to regulate translation from the HCV IRES. In these studies, Ito and Lai found that the core region pyrimidine-rich sequence had a suppressive effect on IRES-mediated translation, most likely induced through interaction with PTB (217, 218). This translational suppression was relieved by an interaction of PTB with the highly conserved 98 nt of the HCV 3′ UTR. These results are consistent with the idea that sequences outside the HCV IRES impart control over the translation initiation process and that this may involve (i) the recruitment of a host factor(s) to the HCV RNA, and (ii) cross talk between the viral 3′ UTR, internal elements in the HCV RNA, and the IRES itself. Such a model may provide a mechanism allowing the virus to modulate polyprotein synthesis in response to changing conditions within the host cell through specific protein interactions with the viral RNA.

IRES-mediated translation avoids the potential limitations posed by cap dependency and provides important advantages for viral replication. By mediating internal ribosome entry, the IRES allows the viral mRNA to bypass rate limitations on translation imposed by the cap-binding reaction. This can be seen as critical for the picornaviruses, which can reach an astounding rate of 5 × 10⁶ virion particles/cell over a 6-h period (398), and HCV (10¹² virion particles/day/ml of infected blood examined [348]). Cap-independent translation bypasses the requirement for a virion-encoded mRNA-capping enzyme or alleviates the requirement for host capping enzymes and nuclear localization of viral mRNA synthesis and replication. The picornaviruses have taken this a step further by encoding mechanisms to disrupt cap-dependent translation in favor of viral protein synthesis. As described in “Viral modification of cellular factors” (below), these involve cleaving the eIF4G component of the cap-binding complex (174, 468), regulating the activity of eIF4E (155), and cleaving PABP (233, 259). Perhaps most importantly, the IRES allows the 43S preinitiation complex to avoid scanning the long 5′ UTR of the viral mRNA. Thus, IRES entry of ribosomes avoids the pitfalls of scanning through highly structured regions within the viral mRNA and bypasses any initiation interference from upstream AUG codons and uORFs (Fig. 5). This also ensures that the translation machinery avoids interfering with the genome replication signals embedded within the 5′ UTR of the viral RNA (37).

As suggested by the IRES-mediated translation model (223), IRES function is determined, in part, through interactions with host proteins in addition to the translational machinery itself. IRES-host protein interactions may contribute to host range specificity and virulence phenotype. This was suggested by early experiments that measured the in vitro translation efficiencies of the poliovirus and EMCV IRES elements within a rabbit reticulocyte lysate translation system (41, 99). These experiments found that the cardiovirus and aphthovirus RNAs were efficiently translated in vitro whereas, in contrast, the enterovirus and rhinovirus RNAs were translated with low fidelity and inefficiency. Supplementation of the translation mixture with HeLa cell extract significantly increased IRES efficiency and restored the accuracy of translation (371). These studies immediately suggested that viral host range might directly reflect the requirements for specific host factors, in addition to the canonical translation factors, to facilitate IRES function and viral protein synthesis. Indeed, IRES-binding host factors have now been identified that play a functional role in IRES-mediated translation (23, 372, 394) (see below). The identification of such factors, and their cognate binding sites within the IRES, should lead to a better understanding of the impact of translational control on host range restriction and viral pathogenesis.

The concept that the translational efficiency of viral mRNA affects pathogenesis is supported by the results of analyses of poliovirus Sabin vaccine strains (400). A subset of attenuating mutations within the Sabin strains were mapped to the major stem-loop V (Fig. 6) of the poliovirus IRES (reviewed by Ehrenfeld [101]). These mutations resulted in reduced translation efficiency of the viral mRNA, resulting in the attenuated vaccine phenotype. These studies suggested that the stem-loop structures within the IRES were critical determinants of translational efficiency and, in poliovirus, were important elements of neurovirulence. Analyses of the neuropathogenic phenotype of rhinovirus/poliovirus chimeras support this notion. Studies of poliovirus neurovirulence in which the poliovirus IRES was replaced with the structurally related IRES from human rhinovirus type 2, conducted by Wimmer and colleagues (167), demonstrated that the neurovirulent phenotype of poliovirus was dependent on the authentic poliovirus IRES. Interestingly, the chimeric virus retained its ability to replicate in nonneuronal tissues, indicating that the IRES itself is an important determinant of host range specificity. More recent structure-function analyses have identified stem-loops V and VI as the poliovirus genetic elements responsible for the neurovirulent phenotype (168). Together, these results suggest that poliovirus IRES stem-loops V and VI provide the capacity for viral replication within the central nervous system. It is therefore possible that these genetic determinants may function to recruit tissue-specific host factors to the poliovirus IRES. Future experiments to test this idea should aid in our understanding of the relationship between IRES structure and viral pathogenesis.

Substantial progress is now being made toward understanding how elements within the HCV IRES may affect viral replication and pathogenesis. An initial comparison of IRES structure and IRES-dependent translation from representative isolates of the various HCV genotypes has revealed some interesting features. First, by measuring the relative translation rates of chloramphenicol acetyltransferase (CAT) reporter constructs placed under control of HCV IRES sequences, Buratti et al. demonstrated a lower translation efficiency for HCV genotype 3 compared to genotypes 1 and 2 (47). Moreover, these investigators found that conservation of both the secondary structure and the sequence within linear domains of HCV IRES stem-loop III was important for maintaining IRES function. A broader analyses of HCV IRES translational efficiency supported and extended these results, revealing a lower efficiency of translation mediated by the IRES elements of HCV genotypes 3 to 6 (71). It is interesting that the level of polymorphism between the IRES elements used in the latter study was limited to a mere total 17 nucleotide positions. Most of these differences could be mapped to within the regions on the major stem-loop III and, to a lesser extent, stem-loop IV. Thus, it appears that subtle differences in nucleotide sequence can confer sufficient alterations in RNA secondary structure to affect the function of the HCV IRES and the efficiency of viral mRNA translation. In accordance with this idea, Honda et al. have revealed that major alterations in stem-loops III and IV of the HCV IRES resulted in loss of function, presumably due to the induction of gross aberrations in IRES structure (204–206). However, these studies also identified differences in translation efficiencies between HCV genotypes 1A and 1B that were not attributed to major structural differences between the respective IRES sequences (206). In follow-up studies, it was found that the reduced translation efficiency of the HCV 1B RNA was attributed to the presence of an AG dinucleotide sequence (HCV nt 34 and 35) present within a single-stranded region that preceded the IRES (207). Mutation of the AG dinucleotide to GA (present in HCV 1A) increased the translation efficiency of HCV 1B RNA, demonstrating that the wild-type (wt) AG dinucleotide sequence had an inhibitory effect on translation. Remarkably, translation inhibition was attributed to an RNA-RNA interaction mediated between the AG dinucleotide sequence and a region far downstream within the viral core coding sequence (207). These results suggest that the long-range RNA-RNA interaction between the viral 5′ UTR and the core-coding region functions cooperatively to influence the activity of the HCV IRES. Such an interaction may sufficiently alter the structure of the IRES to affect the efficiency of viral RNA translation. An intriguing question is whether the correlation between HCV genotype and IRES translation efficiency in vitro extends to the phenotypic differences exhibited by these viral genotypes during the course of HCV infection.

Molecular epidemiological studies have identified an association between HCV genotype, response to the current anti-HCV interferon (IFN) therapy, and severity of infection. In these studies, patients who were infected with HCV genotype 1 or 2 consistently presented a more severe pathology than did patients infected with HCV genotype 3, 4, 5, or 6 (7, 45). Moreover, analyses of the virus load in serum revealed an association between increased viral titer, resistance to therapy, and a poor prognosis, especially in patient groups infected with HCV genotype 1 (103, 115). It has been proposed that these biological differences between HCV genotypes may be due, in part, to the variations in translation efficiency conferred by the subtle polymorphisms between the respective genotype-specific IRES. Secondary-structure analyses of the HCV IRES isolated from healthy patients with low viral titers who responded to anti-HCV therapy and from those who did not respond and maintained high viral loads revealed that a significant level of IRES structural variation was associated with a low viral titer and complete response to therapy (439). In stark contrast, conservation of IRES structure was consistently associated with maintenance of a high virus load and resistance to therapy. Taken together, these studies suggest that IRES structure and the corresponding translation efficiency of a given HCV strain are critical factors affecting the virulence of HCV and the sensitivity to antiviral therapy. Such results have identified the HCV IRES as a potential target for therapy of HCV infection (31).

Ribosome shunt.

The processes of cap-mediated ribosome entry, 5′ scanning, and internal initiation are combined features of the ribosome shunt mechanism of translation. Ribosome shunting allows the small ribosomal subunit to avoid the problems of scanning a long and complex linear sequence preceding the major ORF. Unlike IRES-mediated translation, shunting is dependent on the 5′ cap and the cap-binding complex for ribosome entry onto the mRNA (410) (Fig. 5). The best evidence for ribosome shunting, and the general ribosome shunt model, comes from studies of the pararetrovirus cauliflower mosaic virus (CaMV) (94, 410), although shunting also takes place in other viral systems (reviewed in reference 203). Upon engaging the mRNA, the ribosome scans until it reaches a cis-acting shunting element that promotes ribosomal translocation to a downstream receiving element(s) (94). By this process, scanning becomes nonlinear and the ribosome can bypass a large portion of the 5′ UTR to initiate translation at the downstream cistron (Fig. 5 and 7).

FIG. 7.

The CaMV ribosome shunt. Specific details are described in the text. The relative positions of sORFA, the major stem-loop, and the gene VII ORF of the 35S RNA are shown. Ribosome shunting (denoted by the lower arrow) occurs around the major stem-loop encompassing nt 70 to 550 (196). After terminating sORFA translation, the 40S ribosomal subunit is shunted around the major stem-loop structure and resumes scanning to initiate translation at the ORF VII AUG codon (upper arrow).

CaMV encodes two major RNAs, a monocistronic 19S RNA that encodes a translational transactivator and the 35S pregenomic RNA (123). The 35S RNA spans the complete CaMV genome, serves as the template for virus-mediated reverse transcription, and functions as the major viral mRNA for the synthesis of at least seven viral proteins. Translation from the 35S RNA is cap dependent and is under the control of the 600-nt multifunctional RNA leader (196). Structural analyses of the CaMV leader revealed that it contains several short uORFs and assumes a complex stem-loop structure (123, 195). Although these features of the CaMV leader are incompatible with the 5′-scanning model, translation from downstream ORFs still occurs with reasonable efficiency in vitro and in vivo (410).

Previous studies of CaMV translation demonstrated that ribosome shunting could occur in trans using separate RNA molecules (124). These experiments involved RNAs that were designed to restore the secondary structure of the 5′ leader by annealing and adopting the native structure of the 35S RNA. The results identified the 35S 5′-UTR stem-loop structure as a requisite element in the ribosome shunting process and implicated the immediate flanking sequences as the shunt donor and shunt acceptor sites. Subsequent structure-function analyses of the CaMV stem-loop demonstrated that translational control of the CaMV 35S RNA was dependent on the presence of an elongated hairpin comprising nt 70 to 550 of the 5′ leader (195, 196). Moreover, experiments examining the influence of the 35S uORFs (sORFA to sORFF) on CaMV translation found that the integrity of sORFA was required for translation of the major downstream cistrons (196). sORFA is the 5′-proximal 35S uORF and is followed closely by the elongated hairpin structure (123). These studies demonstrated that mutation of the sORFA initiation codon impaired translation past the elongated hairpin on the 35S RNA. Together, these results are consistent with the model that CaMV translation initiates at sORFA, allowing assembly of the 80S ribosomal complex onto the 35S RNA (Fig. 7). Perhaps concomitantly while encountering the termination codon of sORFA, the 80S ribosomal complex encounters the shunt donor element at the base of the elongated hairpin structure. sORFA translation then terminates, with the shunt donor accepting the 40S subunit and promoting a shunt around the elongated hairpin to the site of the shunt acceptor. Shunting is then followed by resumption of mRNA scanning by the 40S ribosomal subunit. The 80S ribosomal complex then reassembles at the downstream ORFVII AUG codon, and CaMV translation begins. Alternatively, the shunt donor may facilitate passage of the intact 80S ribosomal complex around the elongated hairpin structure without a requirement for complex disassembly, although this would require the shunt donor to deposit the ribosomal complex directly at the ORFVII initiation codon (Fig. 7). The exact sequence elements responsible for ribosome shunting in CaMV are not yet known. Similar to the IRES mechanism of translation, shunting may involve assembly of host factors at and round the shunt donor and shunt acceptor sites. Recent evidence indicates that shunting, at least in CaMV and related viruses, may be host independent (410). Thus, the ribosome shunt may represent a general mechanism of translational adaptability among viruses. Future studies to determine the nature of the host “shunting factors” will certainly aid in understanding the molecular mechanisms of ribosomal shunting.

What advantages does the ribosome shunt mechanism of translation confer to viral replication? The shunt has clearly evolved as a mechanism to avoid the problems associated with scanning a highly structured 5′ UTR. In this respect, shunting confers translational efficiency to the viral mRNA. More specifically, however, ribosomal shunting may preclude or reduce the requirement for the host eIF4F helicase activity to melt the secondary structures within an mRNA that impede the normal scanning and the AUG site selection process. This would also relieve the competition with cellular mRNAs for the limited amounts of eIF4F present within the host cell. Thus, the ribosome shunt may facilitate the selective translation of viral mRNA, especially under cellular conditions in which the pool of functional eIF4F becomes limiting. Schneider and colleagues have investigated this idea by examining the mechanisms of selective translation of late adenovirus mRNAs within infected cells (reviewed in reference 411). Adenovirus induces the dephosphorylation of eIF4E and concomitant reduction in the functional pool of eIF4F during late-stage infection, resulting in inhibition of host cell protein synthesis (210, 499) (see below). However, the translation of late adenovirus mRNAs remain uncompromised, due to the function of the conserved tripartite leader.

The adenovirus late mRNA tripartite leader sequence is composed of a linear 5′ end followed by regions of stable secondary structure (91). In uninfected cells, the tripartite leader can direct the translation of a downstream cistron by both conventional 5′ scanning and ribosome shunting. However, during late-stage adenovirus infection, the translation of late viral mRNAs is mediated exclusively by the ribosome shunt (494). Structure-function analysis of the tripartite leader has shown that (i) the linear 5′ end is essential for the efficient recruitment of ribosomes when eIF4F is limiting, (ii) the stem-loop structural elements inhibit ribosome scanning during late-stage adenovirus infection, and (iii) the selective translation of late adenovirus mRNAs is influenced by the actions of one or more viral gene products. These results, taken together, suggest that limitations in eIF4F select for expression of the late adenovirus gene products through the function of the late mRNA tripartite leader. Stem-loop elements within the tripartite leader, along with virus-encoded transacting factors, then direct the selective translation of late mRNAs through a process of ribosome shunting, independent of eIF4F helicase activity. Thus, it appears that the ribosome shunt has evolved the dual functions of allowing the ribosome to avoid the impediments on translation imposed by regions of mRNA secondary structure and the limitations in the quantity and quality of the host translation machinery.

Leaky scanning.

Translation initiation site selection is determined, in part, by the context of the nucleotide sequence surrounding the first AUG codon encountered by the scanning ribosomal subunit. Departure from the canonical accAUGg sequence (the initiation codon is shown in bold letters) often results in the scanning ribosome initiating translation from this weak AUG at a low frequency or bypassing it completely in favor of a stronger downstream AUG start site (221). This AUG selectivity is referred to as leaky scanning. Leaky scanning allows the translation of multiple ORFs from a common mRNA substrate (Fig. 5). The versatility of leaky scanning is quite evident when one considers that each ORF need not be in the same reading frame. Thus, the process of leaky scanning allows the virus to maximize its genome coding capacity and encode functionally distinct proteins from a common mRNA.

Leaky scanning is widely used by viruses and is perhaps best defined from studies on retrovirus replication. Human immunodeficiency virus (HIV) encodes a heterogeneous class of mRNAs that include several multicistronic species. Among these are the bicistronic mRNAs encoding the viral Vpu and Env proteins (368). The ORFs for Vpu and Env are tandemly arranged such that the Vpu coding region precedes the Env ORF (Fig. 8A). Synthesis of Env is essential for viral replication and takes place through a mechanism of leaky scanning from the upstream Vpu ORF (368, 412). Env synthesis requires a weak Vpu translation initiation codon. Mutation of the weak Vpu start site to a sequence more closely matching the canonical start site sequence resulted in suppressed Env translation from the bicistronic Vpu-Env mRNA (412). Thus, the weak context of the Vpu initiation codon allows the ribosome to scan pass the Vpu ORF and to initiate translation at the downstream Env AUG codon (Fig. 8A). Analyses of HIV mutants in which the vpu gene was deleted or lacked the Vpu initiation codon revealed a stimulation of Env synthesis (413). Accordingly, vpu mutant viruses exhibited defects in virus release and showed increased syncytium formation in vitro (368). These results support a model for the coordinate expression of Vpu and Env during HIV infection, which is dependent on the presence of the vpu ORF. By this model, the synthesis of Vpu occurs inefficiently via a weak initiation codon, perhaps allowing Vpu to coordinately accumulate during infection to levels sufficient for its function in late-stage HIV replication. In accordance with the leaky-scanning mechanism, the synthesis of Vpu itself may coordinate Env production by impeding ribosome scanning to the downstream Env ORF. This model remains to be directly examined by analyzing the polyribosome distribution of the relevant mutant Vpu-Env mRNAs. However, it suggests that the limitations placed on Env translation by the upstream vpu ORF may allow Env expression to coincide with late-stage replication events, including virion assembly, and release. Thus, while ensuring that both viral proteins will be produced during HIV infection, translational control by leaky scanning provides for the coordinate expression of Vpu and Env during viral replication.

FIG. 8.

Leaky scanning and translational frameshifting during retroviral mRNA translation. (A) Leaky scanning during HIV mRNA translation accounts for synthesis of the Vpu and Env proteins. The HIV gene structure (upper) consists of several overlapping cistrons encoded within a heterogeneous array of mRNAs (438). Synthesis of the Vpu and Env proteins (lower diagram) of HIV proceeds via a leaky-scanning mechanism in which the scanning ribosome bypasses the weak vpu AUG codon (shown) to initiate translation from the env AUG codon (bent arrow). Translation initiation occurs at the vpu AUG codon at a low frequency and may account for the stoichiometric ratios of Vpu and Env protein accumulation during HIV infection (413). (B) Translational frameshifting at the gag-pro junction during MMTV mRNA translation. An RNA pseudoknot near the 3′ end of the gag gene causes the elongating ribosome to pause at the gag-pro slippery sequence element. As a result, the mRNA slips backward by 1 nt and the ribosome-bound tRNAs mediate new anticodon base pairing in the −1 reading frame (12). Frameshifting is favored by weak codon-tRNA anticodon base pairing in the original reading frame and strong base pairing in the new reading frame. Synthesis of the entire Gag-Pro-Pol polyprotein is facilitated by a second frameshift at the downstream pro-pol slippery site. Model adapted from reference 12.

Frameshifting.

As in the example provided by the HIV genome, retroviral genomes exhibit overlapping gene arrangements (Fig. 8A). The mouse mammary tumor virus (MMTV) genome exhibits an overlapping gag, pro, and pol gene arrangement (72). Translation of these gene products involves the process of frameshifting, in which the translating ribosome shifts position by +1 or −1 nt, resulting in a change of reading frame (reviewed by Gesteland and Atkins [151]) (Fig. 5). The process of frameshifting was first described from studies of Rous sarcoma virus replication (220) and has been extensively defined in the MMTV and infectious bronchitis virus (IBV) (a coronavirus)] systems, although it occurs widely throughout other eukaryotic RNA viruses (151). Frameshift sites within the viral mRNA correspond to heptanucleotide sequences in which the mRNA slips 1 base with respect to the tRNAs in the A and P sites on the translating ribosome (40, 108). The frameshift site, also known as the slippery site, allows the tRNA to move along the mRNA template by 1 base (forward or back) and reestablish codon-anticodon pairing, resulting in a stable +1 or −1 reading frame shift.

Frameshifting is stimulated by the presence of an RNA pseudoknot structure located 2 to 4 nt downstream of the slippery site (59, 219, 319). A general model for retroviral frameshifting has been proposed (108), in which the elongating ribosome pauses on the mRNA upon encountering the pseudoknot structure (Fig. 8B). Ribosomal pausing facilitates the realignment of the slippery-sequence-decoding tRNAs in the −1 reading frame. The heptanucleotide sequence that comprises the slippery site typically conforms to the motif XXXYYYN. Frameshifting occurs at this site through the slipping of two ribosome-bound tRNAs that are translocated from the current reading frame of X-XXY-YYN, to the −1 reading frame of XXX-YYY. In MMTV, translation initiation of the gag-pro mRNA begins at the 5′ end of the gag gene (Fig. 8B). The translating ribosomes encounter a slippery site and a pseudoknot structure near the 3′ end of the gag gene. The majority of the ribosomes read through this region, but approximately 25% hesitate at the heptanucleotide site, where the mRNA will slip backward by 1 nt. This event is stabilized by the new pairing with the two tRNAs in the −1 reading frame. Meanwhile, most of the translating ribosomes will terminate to make Gag-Pro but another 10% will slip again at the pro-pol site to make the requisite Gag-Pro-Pol polyprotein (108).

What are the molecular mechanisms by which the tRNAs and pseudoknot contribute to frameshifting during mRNA translation? Evidence has accumulated to indicate that the actual frameshift occurs at the second (underlined) codon of the tandem slippery codon pair, XXXYYYN, corresponding to the ribosome aminoacyl (A) site. Slippery A sites within eukaryotic viruses correspond to the codon sequence of AAC, AAU, UUA, UUC, and UUU (108). Interestingly, these codons are decoded by tRNAs with a highly modified base in the anticodon loop (185). Thus, it has been suggested that hypomodified variants of these tRNAs may function to promote shifting by being less bulky and therefore more easily moved within the slippery site (186). However, this idea remains controversial, since other researchers have proposed that frameshifting is mediated by standard cellular tRNAs and is simply dependent on the strength of the codon-anticodon tRNA interaction (40, 465). In either case, frameshifting requires a pseudoknot structure near the slippery site to stimulate the frameshifting events.

Recent evidence indicates that the actual secondary structure of the pseudoknot is important for stimulating frameshifting. Analysis of the IBV frameshifting signals clearly demonstrated that the pseudoknot causes ribosome pausing. Replacement of the IBV pseudoknot with a simple stem-loop structure of equivalent base pairs did induce ribosome pausing but, remarkably, did not stimulate frameshifting (427). These results suggested that ribosome pausing was necessary but not sufficient for frameshifting to occur and support the hypotheses that (i) conservation of pseudoknot structure is essential for frameshifting and (ii) pseudoknot interactions with specific trans-acting factors may promote the frameshift events. Atomic modeling of the MMTV gag-pro pseudoknot supports the former hypothesis, in that this pseudoknot does not have coaxially stacked helices but, rather, assumes a wedge conformation induced by an A nucleotide between the helices (418). Structure-function analyses of the MMTV pseudoknot revealed that this A nucleotide was essential for stimulating frameshifting activity. Structural analyses of other viral pseudoknots should provide further insight into the contribution of pseudoknot sequence and structure in ribosome frameshifting.

Control of termination and reinitiation.

Translational control by reinitiation involves two or more tandemly arranged ORFs on a common mRNA. In the simplest model of reinitiation, a short uORF controls the translation of the major downstream ORF by impeding ribosome scanning (reviewed by Geballe [146]). In this sense, the uORF commonly asserts a suppressive effect upon translation of the downstream ORF. However, and as in the CaMV 35S RNA translation, exceptions to this rule do apply. As described above, the sORFA uORF of CaMV actually plays a stimulatory role in translation of downstream ORFs within the 35S RNA (195). An example of viral use of reinitiation to negatively control translation from a downstream cistron comes from studies of cytomegalovirus replication. During cytomegalovirus infection, expression of the gp48 product of the polycistronic viral gpUL4 mRNA is coordinately controlled to reach peak levels during late-stage viral replication. gp48 is translated from the third of three cistrons within the gpUL4 mRNA (146, 313). Gelballe and colleagues have determined that coordinate control of gp48 expression is mediated through the actions of the second gpUL4 uORF (uORF2) (147). Remarkably, the uORF2 inhibitory effect on gp48 translation is dependent upon the sequence of uORF2 (48, 85); introduction of uORF2 missense mutations severely diminished the inhibitory signal upon gp48 translation, while introduction of mutations that preserved the coding content of uORF2 led to retention of gp48 translational inhibition. In vitro and in vivo expression studies revealed that the translational control actions of uORF2 (i) function exclusively in cis to repress gp48 synthesis through ribosome stalling at the uORF2 termination codon and (ii) are mediated through interference of uORF2 translation termination by the uORF2 peptide product itself. Analysis of uORF2 translation revealed that the 20-kDa peptide product remained bound to the ribosome complex as a peptidyl-tRNA covalently linked to tRNA_pro, which decodes the uORF2 carboxyl-terminal codon (49). Recent studies have now demonstrated that the uORF2 peptidyl-tRNA_pro blocks its own hydrolysis and ribosome release to remain stably bound to the ribosomal complex (50). These results suggest that inhibition of uORF2 peptidyl-tRNA_pro hydrolysis blocks the translation of gp48 by creating a barrier that obstructs ribosome scanning to the downstream gp48 ORF (Fig. 9).

FIG. 9.

CMV control of gp48 synthesis by inefficient termination of uORF2. During CMV infection the viral gp48 glycoprotein is translated from the third of three cistrons (rectangles) within the gpUL4 mRNA. 1, gp48 synthesis is facilitated by a leaky-scanning mechanism in which the scanning ribosome bypasses the weak upstream AUG codons within the gpUL4 mRNA to initiate translation at the gp48 AUG (indicated by arrow); 2, translation initiation at the uORF2 AUG occurs at a low frequency and results in control of gp48 translation. Synthesis of the uORF2 peptide produces a stable peptide-tRNA_pro-ribosome complex that prevents peptide release and stalls the elongating ribosome at the uORF2 termination codon. As a result, ribosome scanning and reinitiation at the downstream gp48 AUG codon is blocked.

If uORF2 blocks gp48 translation, what facilitates synthesis of the gp48 protein during CMV infection? Sequence analyses has shown that the initiator AUG codon of uORF2 is presented within a “weak” context for optimal translation initiation (147). Accordingly, it was found that the uORF2 AUG codon is recognized by the ribosomal complex only at a low frequency and is actually bypassed by a leaky-scanning mechanism in favor of initiating translation from the gp48 start codon. Alteration of the uORF2 initiation codon to within an optimal context for translation initiation results in nearly a complete block in gp48 synthesis (48). Together, these results support a bipartite model for the control of gp48 translation by uORF2 (Fig. 9). It is not clear how uORF2 peptidyl-tRNA_pro blocks its own hydrolysis and release from the uORF2 termination site. Possible explanations may be found by examining the influence of the uORF2 product on the function of translation elongation and release factors.

Another example of reinitiation control comes from studies of reovirus translation. Reoviruses are double-stranded RNA (dsRNA) viruses with a segmented genome. The S1 mRNA of reoviruses is bicistronic and encodes the ς1 capsid protein and the ς1 nonstructural protein (409, 417). Initial analyses of S1 mRNA translation revealed that it was translated inefficiently compared to other reovirus mRNAs (345). Subsequent in vitro studies showed that ribosomes paused at several positions on the S1 mRNA relative to the S4 mRNA, suggesting that translating ribosomes were less evenly distributed along the coding region of the inefficiently translated S1 mRNA than of the efficiently translated S4 mRNA (97). These results were supported by in vivo studies in which the distribution of translating ribosomes on polyribosome-bound reovirus S1 and S4 mRNAs was examined in reovirus-infected cells (98). The pattern of ribosome pausing in vivo showed that ribosomes were less evenly distributed along the poorly translated bicistronic S1 mRNA. Consistent with a model of S1 mRNA translational control by reinitiation, expression of the downstream S1 ORF was significantly increased by mutation of the upstream AUG codon to a less favorable context for translation initiation (106). Interestingly, however, the synthesis of the upstream S1 mRNA translation product was not decreased by the same mutations. Identification of differential codon usage between the S1 ORFs suggests that the translation efficiency of the S1 uORF may be due to codon usage that confers a low elongation rate through the utilization of low-abundance tRNAs (106). The diminished elongation rate of the S1 uORF may then limit the efficiency of reinitiation and synthesis of the downstream S1 cistron. Studies aimed at understanding the influence of differential codon usage on the elongation rate may uncover additional examples of this type of translation control among eukaryotic viruses.

As described in the examples cited above, control of reinitiation has been attributed largely to processes inherent within the 5′ UTR of the viral mRNA. Analyses of alfalfa mosaic virus (AMV) replication now suggests that the 3′ UTR may likewise play an important role in translation reinitiation and the efficiency of viral protein synthesis. The single-stranded RNA genome of AMV is capped but not polyadenylated. Translation studies have revealed that AMV RNAs are efficiently translated in spite of lacking the traditional poly(A) tail and the advantages to RNA stability and translation afforded to polyadenylated transcripts (157, 200). In contrast to the many viruses that induce host translational shutoff during infection, AMV infection, AMV infection is not associated with a decrease in host protein synthesis (149). How, then, do AMV mRNAs adequately compete for available translation factors? Early evidence suggested that the 3′ UTR of the AMV coat protein played a stimulatory role in mediating coat protein synthesis (399, 496). Examination of coat protein mRNA translation in vitro and in vivo revealed this mRNA to be efficiently translated even in the presence of large quantities of a cellular mRNA competitor. A functional role for the 3′ UTR in coat protein mRNA translation was demonstrated by conducting similar experiments with mutant mRNAs lacking the 3′ UTR; loss of the 3′ UTR consistently reduced the efficiency of coat protein synthesis without altering mRNA stability (177, 399). Interestingly, it was found that the 3′ UTR was required for assembly of the coat protein mRNA into polyribosome complexes, indicating that the 3′ UTR was an important determinant for ribosome binding (177). Mutagenesis studies were used to identity the 3′ UTR nucleotide sequence element GAUG as an important determinant in AMV coat protein synthesis. This tetranucleotide sequence encompasses an initiation codon downstream from the coat protein termination codon and is thought to stimulate coat protein synthesis through a process of reinitiation (177). With this model, reinitiation would facilitate ribosome-mRNA interaction and continued coat protein synthesis.

How might reinitiation within the 3′ UTR actually contribute to increased translational efficiency? One possibility is that reinitiation may retain the mRNA within the pool of active ribosomes, thereby increasing the probability of 5′ UTR-ribosome interactions and promoting further rounds of authentic translation initiation. On the other hand, the viral 3′ UTR may stimulate coat protein translation through a process independent of reinitiation, although this idea remains inconsistent with experimental evidence. In this case, it remains possible that specific 3′ UTR-protein interactions may impart increased translational efficiency.

Functional recoding.

In addition to frameshifting, viruses partake in functional recoding whereby translation proceeds through an in-frame termination codon (Fig. 5). This occurs though a process of redefining the termination codon to encode glutamine at UAG or tryptophan or selenocysteine at UGA (151). Functional recoding has been extensively studied in the Moloney murine leukemia virus (MuLV) system. MuLV redefines the stop codon at its gag-pol junction through the insertion of glutamine at the UAG stop codon. This allows the translating ribosomal complex to read through the gag-pol junction and synthesize the Gag-Pol polyprotein. During viral replication, the ribosome reads through the MuLV gag-pol stop codon 5% of the time, and this exact frequency seems to be essential for replication (491). If the UAG codon is replaced with an in-frame GAG codon, no viral particles are formed (111). In contrast, replacement of the native gag-pol stop codon with either the UAA or UGA stop codon permitted translational readthrough with similar efficiency. These results suggest that MuLV utilizes functional recoding as a mechanism to control the level of Gag-Pol polyprotein synthesis during the course of infection.

Moreover, it appears that redefining the stop codon is not codon specific, suggesting that other elements within the viral mRNA are responsible for translational readthrough. It is now clear that sequences downstream of the stop codon are necessary for functional recoding. In MuLV, this includes a pseudoknot sequence that appears to stimulate the recoding process (112, 482) (Fig. 10). Structure-function analyses of the MuLV pseudoknot revealed several interesting features required for stimulating recoding, including (i) specific nucleotide sequence within the spacer region between the stop codon and the pseudoknot; (ii) nucleotide conservation within stem-loop 2 of the pseudoknot; and (iii) a nonhelical pseudoknot structure (482). These results support the idea that recoding may be dependent upon recruitment of trans-acting factors to the termination site, possibly mediated through sequences within the spacer region between the stop codon and the pseudoknot and/or the pseudoknot structure itself. Identification of such factors may have implications for the development of future antiviral drugs, since disruption of the recoding process results in a block in virion formation within the infected cell (111).

FIG. 10.

Functional recoding at the MuLV gag-pol junction. Synthesis of the MuLV Gag protein terminates at the gag stop codon (top). However, approximately 5% of the time, the elongating ribosome will read through the gag stop codon to produce the Gag-Pol polyprotein (bottom). During this process, the gag stop codon is redefined to encode glutamine and is shown by the presence of tRNA_Gln at the redefined UAG codon. Stop codon redefinition is dependent on specific downstream sequences and a 3′-proximal pseudoknot structure. The elongating ribosome eventually melts out the pseudoknot to complete Gag-Pol synthesis. Low-frequency gag-pol stop codon redefinition is essential for MuLV replication. Figure adapted with modification from references 12 and 108.

Coupling the Virus Life Cycle to Translational Control

Translation control programs in eukaryotic cells play important roles in governing cellular metabolism. In many cases, these programs are implemented in response to specific environmental cues. Viruses, specifically the more complex DNA viruses, including the poxviruses, papillomaviruses, and herpesviruses, have similarly incorporated into their own life cycles translational control programs that play important roles in replication, latency, and virulence. In particular, recent evidence from studies of HSV replication indicates that translational control programs are implemented by the virus to (i) facilitate host shutoff, (ii) control global and specific viral gene expression, (iii) maintain or exit latency, and (iv) overcome the host antiviral response. In this section, we use the herpesvirus life cycle to illustrate how viruses utilize translational control programs to direct replicative decisions and how this translational programming contributes to the control of viral replication.

The herpesviruses.

The human herpesviruses establish latent infection in either neural cells (HSV and varicella-zoster virus) or hematopoietic cells (Epstein-Barr virus and cytomegalovirus) with negligible damage to their respective host. Latency permits viral persistence in the face of an active immune response. In response to certain stimuli, the viruses may periodically reactivate from latency throughout the life of the host to enter the productive phase viral life cycle, which sheds sufficient virus progeny to infect new hosts. This life-style is bound to impose a distinctive set of evolutionary pressures on the control mechanisms regulating herpesviruses gene expression. Thus, herpesviruses present a challenging and attractive model system for studying the ever-complex mechanisms of viral gene expression and regulation at the level of translation. Furthermore, mutant herpesviruses deficient in a particular function can be isolated or genetically engineered with deletions in certain genes to assess the roles of specific virus-encoded proteins in viral translation and replication in the host. In the following sections, we examine how the alphaherpesviruses HSV-1 and HSV-2 have implemented control of viral and host translation into their latent and productive phases and discuss how translational control of HSV mRNA might contribute to viral pathogenesis.

Because HSV gene expression is coordinately modulated during a productive infection, the viral genes can be categorized into three kinetic classes: immediate-early (α), early (β), and late (γ) genes. α genes are transcribed in the absence of de novo protein synthesis; this process peaks at 2 to 4 h postinfection, and the transcripts continue to accumulate until late in infection. Most products of α genes are potent transcriptional trans-activators that cooperate to activate the transcription of β and γ genes. β genes encode proteins required for HSV DNA synthesis, as well as a number of auxiliary replication factors. Viral structural proteins are the products of γ genes, whose expression occurs at the onset of DNA synthesis. While much research on HSV has centered on the transcriptional events responsible for the differential expression of α, β, and γ genes, accumulating evidence indicate that translational mechanisms are also important for HSV gene expression.

HSV and the shutoff of host cell protein synthesis.

Like other cytolytic viruses, HSVs are thought to facilitate their replication by preferentially producing viral proteins at the expense of host cell gene expression. In tissue culture cells infected with HSV-1 or HSV-2, host protein synthesis and mRNA levels decrease by approximately 90% within 3 h postinfection and viral proteins dominate thereafter (424). This remarkable feature of the shutoff of host protein synthesis induced by HSV infection, presumably to alleviate competition for precursors, is a multistep process that involves several mechanisms and can be separated into two stages: primary shutoff and secondary shutoff. Primary shutoff, which is characterized by rapid disintegration of preexisting polyribosomes and degradation of preexisting cellular and viral mRNAs, occurs very early after HSV infection in the absence of de novo protein synthesis. In contrast, secondary shutoff takes place later in the course of infection and requires viral gene expression.

Despite extensive research efforts, the exact mechanisms responsible for the shutoff events during HSV-1 infection are poorly understood, but encouraging progress has been made over the past few years. In the primary shutoff, at least one viral factor, the virion host shutoff (VHS) protein, is necessary for mRNA destabilization, and this may, at least in part, account for the disassociation of polyribosomes (280). Encoded by HSV gene UL41, the VHS protein is a 58-kDa phosphoprotein located between the capsid and envelope regions of the virion (called tegument) and is delivered into the cytoplasm of newly infected cells. How does VHS function to specifically induce mRNA degradation? Apart from having limited homology to a small segment of PABP (389), the primary sequence of the VHS protein provides little clue to its function. Despite the lack of any primary sequence similarity to known RNases, several lines of evidence suggest that VHS is associated with RNase activity: (i) incubation of polyribosomes from uninfected cells with postpolysomal (S130) supernatant from HSV-infected cells, but not S130 from uninfected cells or from cells infected with a VHS-defective virus, resulted in rapid degradation of stable mRNAs; (ii) crude extracts from host shutoff-competent virions or reticulocyte lysate containing wild-type VHS protein displayed enhanced RNase activity, while VHS mutants did not; and (iii) antibodies against VHS protein inhibited the RNase activity of wild-type VHS protein in the cell-free reactions. However, there is as yet no direct evidence, obtained using highly purified proteins, to demonstrate that wild-type, but not mutant, VHS protein itself indeed contains RNase activity in vitro. Thus, the possibility that the VHS protein works in conjunction with another factor(s) with RNase activity has not been entirely excluded.

The mechanism(s) by which the VHS protein specifically targets mRNA for degradation is not known. Polyadenylated RNAs are degraded faster than nonpolyadenylated substrates (225), and deadenylated mRNAs congregate in HSV-infected cells (180). Thus, it is conceivable that the VHS protein recognizes polyadenylated mRNAs, perhaps through the putative poly(A)-binding region of VHS. In this context, it would be interesting to test the effect of mutating the conserved residues in this region on HSV-induced mRNA degradation and translational arrest. Furthermore, the ability of the VHS protein to bind poly(A) mRNA in vivo (for example, through UV cross-linking procedures) or the RNA sequences or structural regions recognized by VHS have not been determined. Alternatively, the VHS protein may interact with the PABP complex for localized poly(A) cleavage of mRNA. This possibility is consistent with the observation that mRNAs in infected cells or in cell-free reaction mixtures are preferentially degraded over protein-free RNAs, suggesting that the VHS protein might achieve specificity through interaction with a factor(s) present in the ribonucleoprotein (RNP) complex. Therefore, the identification of VHS-interacting factors may also provide insights into the mechanistic action of the VHS protein. At any rate, considering the specificity of the VHS protein and the generation of discrete decay mRNA intermediates, VHS probably operates, either on its own or in cooperation with another factor(s), by cleaving mRNA molecules at one or a few critical sites, such as poly(A), that normally function to protect mRNA from destruction by host RNases.

The ICP27 protein of HSV-1 is thought to play an important role in the secondary shutoff of host protein synthesis (180). A nuclear phosphoprotein of 63 kDa, ICP27 was originally known for its capability to both stimulate and repress the expression of different target genes. Subsequent studies demonstrated that ICP27 is also capable of interfering with host cell splicing, causing a reduction in the levels of several cellular spliced transcripts and an accumulation of pre-mRNA in the nucleus during infection (180, 181). Furthermore, ICP27 expression causes a dramatic redistribution of the splicing small nuclear RNPs and other splicing factors during HSV-1 infection (406). Finally, an interaction between ICP27 and the splicing machinery has been demonstrated (181, 406). On the basis of these studies, it has been hypothesized that ICP27-mediated impairment of host cell splicing may contribute to the secondary shutoff, because unspliced cellular transcripts remain in the nucleus, where they become degraded. This leads to fewer cellular transcripts being exported to the cytoplasm for translation, and thus selective synthesis of virus-specified proteins is favored, since most viral transcripts are not spliced.

Selective repression of mRNA translation initiation during HSV infection.

HSV induces shutoff of most host protein synthesis. However, a few remaining cellular mRNAs, whose protein products presumably play an important role in the survival of the virus within the host, are continuously being translated after infection (285). This paradox raises some important questions. (i) What is the function and nature of these cellular mRNAs? (ii) Is the sustained translation of cellular mRNAs induced by HSV or an inherent feature of the mRNAs? (iii) What is the mechanism of the persistence of translation of these mRNAs after infection?

Recent studies demonstrate that at least one set of cellular mRNAs that are persistently translated after HSV-1 infection encode ribosomal proteins (164, 425). An analysis of ribosomal protein mRNA expression across a polyribosome gradient revealed that there is a discernible shift between the untranslated subpolysomal (prepolysomal) fraction to the polyribosomal fraction as infection proceeded (164). It is known that the 5′ leader sequence of vertebrate ribosomal protein mRNAs contains a terminal oligopyrimidine tract (known as the 5′ TOP element) that is sufficient to confer translational regulation and migration between the polyribosomal fractions (13). The specific recruitment of 5′ TOP mRNAs by ribosomes is closely associated with an increase in phosphorylation of ribosomal protein S6 (231, 232, 456). In this regard, S6 proteins in preexisting ribosomes were phosphorylated to greater extent than were those found in newly assembled ribosomes in infected cells (164). This suggests that translation of ribosomal protein mRNAs may occur preferentially on preexisting ribosomes. Interestingly, in parallel with this study, a progressive shift of β-actin and GAPDH mRNAs from polyribosomes to 40S subunits was observed during the course of infection. This phenomenon appeared to be independent of VHS-mediated mRNA degradation, since the protein did not affect mRNA recruitment by polyribosomes. With the caveat that only two cellular mRNAs have been examined, these studies suggest that HSV-1 may employ an additional strategy to selectively suppresses host mRNA translation, namely at the initiation step.

Other mechanisms may also contribute to the persistence of ribosomal protein synthesis. In addition to S6 ribosomal protein, HSV-1 infection induces phosphorylation of at least two other proteins, including the product of the US11 late gene (90). Furthermore, the possibility that an increase in the elongation rate of translation might also account for the sustained translation of ribosomal protein mRNAs has not been eliminated. In this regard, two HSV-1 proteins, the trans-activator ICP0 protein (10) and the protein kinase encoded by the U(L)13 gene (257), interact with and phosphorylate elongation factor-1 delta (EF1-δ), respectively. However, a role for EF1-δ phosphorylation in HSV-1 replication has not been demonstrated.

Selective translation of HSV mRNA.

How does HSV achieve selective viral translation during the shutoff of protein synthesis? In VHS-induced shutoff, in which both cellular and viral mRNAs undergo concomitant degradation, HSV would need to ensure that the viral mRNAs continue to accumulate after cellular mRNAs have been degraded. The potency and indiscriminate nature of the VHS activity suggests that it would have to be negatively controlled in a temporal fashion during the course of infection. Evidence for this hypothesis came from a study demonstrating that the virion trans-activator VP16, which forms a specific complex with VHS in the infected cell, is capable of suppressing VHS activity (282). Specifically, viral protein synthesis and mRNA levels were significantly reduced at intermediate times after infection with a VP16 null mutant virus. Additionally, a stably transfected cell line expressing VP16 was refractory to VHS-induced host shutoff of protein synthesis. Although it remains to be shown, the VHS binding function of VP16 is likely to be important for inactivating VHS, either by masking one or more functional domains, inducing a conformational change, or by targeting VHS into the nucleus and/or the virion assembly pathway. Moreover, the mechanism by which the VP16-VHS interaction is modulated is not known. Finally, it is noteworthy that two other viral factors are involved in HSV-induced host shutoff: the virion-associated protein kinase encoded by the U(L)13 gene (358) and the ICP22 protein encoded by the US1.5 gene (349). The possibility that these gene products may mediate host shutoff by regulating VHS and/or VP16 function will undoubtedly be explored in forthcoming studies.

Another strategy by which HSV may exert selective translation of viral mRNAs over host mRNAs is suggested by recent studies that demonstrate a role of the ICP27 protein in the export of HSV-1 intronless mRNAs (405). Thus, ICP27 appears to mediate preferential viral translation via two mechanisms: (i) it impedes the translation of cellular mRNA by preventing the export of this mRNA to the cytoplasm through the impairment of host splicing, and (ii) it binds viral transcripts and delivers these RNAs to the cytoplasm for translation. However, several key aspects will have to be elucidated in future studies to strengthen the proposed dual role of ICP27. These include assessing (i) the functional significance (an effect on splicing) of the interactions and changes of ICP27 and small nuclear RNPs, (ii) the specificity of ICP27 RNA binding, and (iii) the interaction of ICP27 with the cellular nuclear export complex.

Finally, the observations that ribosomal proteins are persistently synthesized and new ribosomes are assembled after HSV-1 infection might suggest another mechanism for the preferential translation of viral mRNAs. Because new ribosomes contain underphosphorylated S6 ribosomal protein, they are presumably less effective in the translation initiation of mRNAs that possess 5′ TOP sequences (reviewed in reference 232). Thus, viral mRNAs which lack 5′ TOP sequences may be selectively initiated by newly synthesized ribosomes over cellular mRNAs containing 5′ TOP. Although the idea is not unreasonable, especially since reinitiation of translation progressively becomes a limiting factor during shutoff (10, 285), it remains to be supported by any experimental evidence.

Inhibition of eIF2α phosphorylation during HSV infection.

Cells modulate the synthesis of proteins in response to external stimuli, including viral infection, through the modification of translation factors. Phosphorylation of eIF2α is perhaps the best-characterized mechanism by which this occurs, particularly within the context of virus infection. As described in detail in “Viral modification of translation factors” (below), viral replication produces highly structured viral transcripts in the form of dsRNA that can bind to and activate the host PKR, which in turn phosphorylates eIF2α (131). As a result, the cellular translational machinery is incapacitated and viral protein synthesis and replication are restricted within the infected cell. Accumulating evidence now indicate that HSV-1, like many viruses (see Table 4), has evolved ways to circumvent the virally induced translational block by counteracting PKR function (Fig. 11).

TABLE 4.

Viral mechanisms of PKR inhibition

Target	Virus/inhibitor	Mechanism of action	Reference(s)
dsRBMs	Adenovirus/VAI RNA	Competitively inhibits dsRNA-binding reaction	320, 321
	Epstein-Barr virus/EBER RNAs	Competitively inhibits dsRNA-binding reaction	64, 70
	Vaccinia virus/E3L	Sequesters dsRNA	55, 475
	Influenza virus/NS1	Sequesters dsRNA	312
	Reovirus/ς3	Sequesters dsRNA	20, 307, 493
	HIV TAR RNA	Competitively inhibits activator dsRNA binding; may activate PKR; could target TAR RNA binding protein to PKR	169, 314
	TAR RNA binding protein	Binds to PKR, inhibits kinase activity; may sequester dsRNA	24, 73
Dimerization	Influenza virus/P58IPK	Disrupts dimerization	294
	Hepatitis C virus/NS5A	Disrupts dimerization	129, 133, 134
Substrate interaction	Vaccinia virus/K3L	Binds to PKR as a pseudosubstrate; blocks PKR-substrate interaction	21, 255
	HIV Tat		39, 327
Expression and stability	Poliovirus	Induces PKR proteolysis	29, 30
	HIV Tat	Lowers expression of PKR	397
Regulation of eIF2α	HSV/γ134.5 gene product	Complexes with eIF2α and induces an eIF2α phosphatase	188, 338
	Simian virus 40/T antigen	Targets processes downstream of eIF2α	386, 444

FIG. 11.

Translational control during HSV-1 infection. 1, Infection with HSV-1 induces a rapid shutoff of host cell protein synthesis, due in part to the actions of the viral VHS and ICP27 proteins, which indirectly affect mRNA translation. VHS-induced translational shutoff may be directly modulated through the viral VP16 protein. 2, Repression of VHS by VP16 may contribute to the selective translation of viral mRNAs. The viral U(L)13 and ICP0 proteins may similarly modulate viral mRNA translation by phosphorylating and binding, respectively, to host EF1-δ. During infection, HSV-1 ensures that the host cell remains translationally competent by blocking the phosphorylation of eIF2α. Disruption of eIF2α phosphorylation may occur through the actions of the viral Us11 protein, which prevents PKR activation, and by the viral γ₁34.5 gene product, which directs the dephosphorylation of eIF2α by PPIα. Finally, viral modulation of ribosomal protein S6 phosphorylation may contribute to the sustained translation of host 5′ TOP mRNAs, including those that encode ribosomal proteins. Sustained ribosomal protein synthesis and disruption of eIF2α phosphorylation facilitates HSV persistence by supporting viral mRNA translation.

Initial evidence that HSV-1 can antagonize PKR-mediated translational arrest stemmed from studies of mutant viruses that lack the γ₁34.5 gene. These viruses fail to grow on many human malignant neuronal cells, which displayed increased PKR and eIF2α phosphorylation, as well as premature shutoff of protein synthesis late in infection. The phenomenon appeared to be independent of VHS function and mRNA degradation (62, 379). Interestingly, an unknown 90-kDa phosphoprotein (p90) coprecipitated with anti-PKR antibody from lysates of cells infected with γ₁34.5⁻ viruses. While the function of p90 is not known, the correlation between p90 phosphorylation and the premature shutoff of protein synthesis suggests that it may play a positive role in modulating PKR activity in phosphorylation of eIF2α. It was thought that the γ₁34.5 gene product might inhibit protein kinase activity by blocking the interaction between p90 and PKR. However, subsequent studies demonstrated that the γ₁34.5 protein operates through a different mode of action. Using the yeast two-hybrid approach, He et al. (190) found that the γ₁34.5 protein associated with the cellular protein phosphatase 1α (PP1α). Further, the γ₁34.5 protein formed a complex with PP1α in HSV-1-infected cells, and fractions containing the complex were capable of dephosphorylating purified eIF2α. Thus, the γ₁34.5 protein is likely to function as a regulatory or targeting subunit of PP1α to redirect the phosphatase to dephosphorylate eIF2α, therefore neutralizing PKR activity. However, it is not clear how the γ₁34.5 protein guides PP1α to eIF2α; an interaction between γ₁34.5 and eIF2α has not been demonstrated. Interestingly, we have recently obtained evidence that PP1α can also directly inhibit PKR function by binding to and dephosphorylating PKR (S.-L. Tan and M. G. Katze, unpublished data). Whether dephosphorylation of PKR by PP1α is also triggered by HSV-1 infection or other signals is not known. Furthermore, it will be important to identify the regulatory subunit of PP1α that targets the phosphatase to PKR (Fig. 11).

The story becomes more complicated with recent studies describing the isolation of second-site suppressor mutant viruses that lack the γ₁34.5 gene (53, 338). These variant viruses regained the ability to grow on otherwise nonpermissive neuronal cells and contained additional mutations that affect a distinct viral genetic element, the SUP locus. Moreover, deletion of the SUP locus prevented the accumulation of phosphorylated eIF2α. Consequently, extragenic suppressor γ₁34.5 mutants could sustain protein synthesis and multiply on cells that failed to support replication of the parental the γ₁34.5 variants. As it turns out, these dominant suppressor alleles compensated for the loss of the γ₁34.5 function by overproducing a viral RNA-binding, ribosome-associated protein (US11) that reduced PKR activation (343). Taken together, these results suggest that HSV-1 encodes at least two strategies (US11 and γ₁34.5) to modulate cellular translation by targeting both PKR and eIF2α (Fig. 11). Interestingly, the γ₁34.5 protein contains a region of significant homology to the cellular protein GADD34, which is induced in response to agents that promote growth arrest, DNA damage, and differentiation. Indeed, GADD34 could also interact with PP1α and could functionally replace γ₁34.5 in prolonging late-protein synthesis in infected cells (189). These studies suggest that signals that trigger differentiation, growth arrest, and DNA damage may be intimately linked to translational control.

Implications of viral modulation of translation in HSV pathogenesis and disease treatment.

An important question that has remained unanswered in the study of HSV pathogenesis concerns how the viruses undergo latency in their host. It is unanswered because of the lack of a reliable cell culture system to support latent HSV infection. What is known from in situ hybridization studies of latently infected neuronal cells is that the latent state is characterized by the transcription of specific colinear transcripts. These RNAs, known as latency-associated transcripts (LATs), persist despite the absence of virtually all gene expression. Although experiments with LAT-negative mutants showed that the LATs are not required for HSV-1 lytic replication and establishment or maintenance of latency, they appear to be necessary for efficient in vivo reactivation in infected animal models. Thus, there is great interest in determining the stage of latency at which the LATs modulate, their mechanisms of action, and the exact sequences responsible for reactivation. In this regard, Steiner and colleagues have recently reported that the LATs are associated with polyribosomes in vitro and during latent in vivo infection (158, 159). These observations are highly suggestive of translation of a functional LAT protein, although efforts to demonstrate the presence of HSV-1 LAT protein products in latently infected tissues have been futile to date. To begin to gain an understanding of how translational control may play a role in HSV reactivation and thus contribute to viral pathogenesis, it will be critical to identify mechanism(s) of ribosome binding to the LAT transcripts and to characterize the LAT proteins.

Finally, the cytolytic and neurotropic properties of HSV-1 render this virus a potential tumoricidal agent for destroying malignant cells in the central nervous system. In this regard, HSV-1 γ₁34.5 mutant viruses, which are attenuated and nonneurovirulent in animals, are able to discriminate between normal and malignant cells (8). However, these variant viruses grow poorly on neuronal tumors, which imposes a major limitation on their effectiveness in destroying tumor tissue. The suppressor mutants of the γ₁34.5 allele, which have regained the ability to grow on neoplastic cells but retained the attenuated phenotype of the γ₁34.5 parent virus, may represent a prototype virus to destroy malignant cells in the CNS. Furthermore, the use of attenuated HSV as a vector for gene therapy in the study and treatment of neurodegenerative diseases is under investigation.

Recruitment of Host Factors for the Efficient Translation of Viral mRNA

The previous examples of HSV-host interactions and the resulting modification of host factors reflects a common theme among viruses, which target and modify host process to facilitate viral replication. Indeed, the pressures of translational dependence on viral replication have resulted in a wide variety of viral strategies to maximize translational efficiency. Such strategies often involve recruitment of specific host factors that function to improve the efficiency or mediate the selectivity of viral mRNA translation. This section discusses how viruses recruit and utilize specific host factors, in addition to the conserved repertoire of canonical translation factors, to facilitate efficient mRNA translation during infection. As presented below, recruitment of host factors to the viral mRNA is broadly used in both cap-dependent and cap-independent translational strategies. This section will begin by examining the host factors that are recruited to the IRES element and the viral 3′ UTR and discussing the roles that such proteins may play in mediating viral mRNA translation and mitigating virus host range. This is followed by a detailed examination of how influenza virus ensures the efficient and selective translation of its mRNAs though a process of host factor recruitment and modification.

IRES binding proteins and proteins that bind the viral 3′ UTR.

As the functional element for cap-independent translation initiation, the IRES promotes viral mRNA translation through the recruitment of canonical translation factors to the initiation site (223). In addition, it is now known that several noncanonical factors are recruited to the IRES that stimulate, or in some cases repress, IRES-mediated translation. Structure-function analyses have provided insights into the roles of IRES-binding proteins in viral replication and have identified sequences within the IRES that direct interaction specificity (162, 238, 239, 394) (reviewed by Belsham and Sonenberg [23], Jackson and Kaminski (223), Hellen and Wimmer [193], and Ehrenfeld [101]). Meanwhile, protein interactions with the viral 3′ UTR have similarly been implicated in providing translational efficiency and selectivity during viral infection (122, 209, 218, 302, 360, 496). A role for 3′ UTR-binding proteins in viral mRNA translation may reflect the efforts of viruses to take full advantage of the translational efficiency provided by the closed-loop translation model.

The early observations from in vitro studies revealed that efficient translation from enterovirus or rhinovirus IRES in rabbit reticulocyte lysate (RRL) required supplements derived from HeLa cell extracts (41, 99). In contrast, the cardiovirus-aphthovirus IRES conferred efficient translation in native RRL. These results were significant in that they indicated that IRES translation required a specific trans-acting factor(s) supplied by the host cell, pointing the way to the identification of IRES-binding proteins. Moreover, these observations implicated IRES-binding proteins in the determination of virus host range specificity. IRES-binding proteins have since been identified through a combination of gel mobility shift assays and the use of UV light to cross-link RNA probes with resident cytoplasmic factors within cell extracts. Several distinct cellular IRES-binding proteins have been identified and characterized (reviewed in reference 23). Among the best-characterized IRES-binding proteins are the systemic lupus erythematosus autoantigen (La) and PTB (36, 191, 194, 328, 330). Both proteins specifically bind the poliovirus IRES.

La is a 52-kDa RNA-binding protein that is known to interact with RNA polymerase III-transcripts and to play a role in the transcription termination reaction (160). The La protein resides predominantly within the nucleus of uninfected cells but, importantly, is redistributed to the cytoplasm during picornavirus infection (330). Structure-function analyses have identified nucleotides 559 to 624 of the poliovirus RNA as the major binding site for the La protein (328). This region maps to within the polypyrimidine tract of the poliovirus IRES (Fig. 6). The best evidence in support of a functional role for the La protein in poliovirus translation comes from experiments in which RRL was supplemented with purified recombinant La protein. Addition of recombinant La stimulated poliovirus translation to an efficiency similar to that observed when RRL was supplemented with HeLa extract (330). However, the concentration of recombinant La greatly exceeded the level of La within HeLa extracts, suggesting that (i) recombinant La was only partially active or (ii) additional factors may participate with La to promote poliovirus translation. It should be noted that the La protein also binds to the 5′ UTR of influenza virus and HIV RNA, where it may also play roles in facilitating efficient viral protein synthesis (361, 443). Thus, the functional role of La in viral protein synthesis may not be limited to promoting IRES-mediated translation but may reflect a more general function. Although a direct biochemical role for La in IRES-mediated translation remains to be demonstrated, it is postulated that La may function as an RNA chaperone to maintain RNA structure in a conformation that favors translation.

PTB is a 57-kDa cellular protein that plays a role in RNA splicing (304). PTB binds to multiple sites within the poliovirus IRES (191), but a direct role for PTB function in poliovirus translation has yet to be demonstrated. PTB also binds with high affinity to the EMCV IRES (229) and, more recently, to the IRES of HCV (3, 4, 217). Major PTB-binding sites have been identified within stem-loop H and the polypyrimidine tract of EMCV (229) (Fig. 6). Similar to the La protein, PTB is postulated to function as an RNA chaperone, where it may stabilize IRES structure in a translation-competent conformation. Evidence to support this idea comes from analyses of EMCV RNA that possessed point mutations within the stem-loop H PTB-binding site (229). Mutations that disrupted stem-loop base pairing prevented PTB binding and abrogated IRES function. Meanwhile, compensatory mutations that restored the native stem-loop conformation restored both PTB-binding and IRES function. A direct role for PTB in EMCV IRES-mediated translation was suggested from competition experiments in which PTB was functionally depleted from in vitro translation reaction mixtures by the addition of competitor RNA that contained a PTB-binding site (38). Addition of competitor RNA was found to specifically inhibit EMCV translation, and the addition of exogenous PTB relieved this competition to restore EMCV translation. However, recent results from Kaminski and Jackson (239) suggest that PTB binding may not be an absolute requirement for EMCV translation but, rather, is conditional upon the structure of the major PTB binding site within the EMCV IRES.

In addition to La and PTB, several other factors have been identified that may play a role in IRES-mediated translation, although the nature of such factors remain to be determined (23). Recent analyses of HAV IRES function indicate a role for PCBP2 in HAV translation (162). PCBP2 was found to bind specifically to within the polypyrimidine tract of the HAV IRES, corresponding to HAV nt 1 to 157. In these experiments, affinity column depletion of PCBP2 from HeLa extracts resulted in only low levels of HAV translation while translation was restored by adding back recombinant PCBP2 to the depleted extracts. PCBP2 has also been implicated in stimulating poliovirus translation (35), where it has been shown to bind to stem-loop IV of the poliovirus IRES (34). An essential role for PCBP2 in HAV and poliovirus translation awaits further studies. However, like La and PTB, it appears that PCBP2 may play a role in maintaining IRES structure in a translationally competent conformation. HAV RNAs lacking the 5′-terminal 138 nt, which are not part of the HAV IRES but do include a major region of the PCBP2-binding site, retained translational efficiency independent of PCBP2 (162). This suggests that deletion of the first 138 nt of the HAV RNA may mimic the effects of PCBP2 binding and allow the IRES to spontaneously adopt the correct structure needed to promote translation. Considering that the picornavirus genomic RNA must serve as template for both translation and transcription, one may propose that at least one role for IRES-binding proteins like La, PTB, and PCBP2 is to functionally differentiate between translation and transcription by making the RNA accessible to the translational machinery. By this model, these and other proteins may bind to their cognate motif(s) within the IRES to induce and/or stabilize the specific RNA tertiary structures that promote mRNA translation over genome transcription. On the other hand, modification, or masking, of IRES-binding proteins by viral and/or cellular factors may provide the signals promoting the switch from translation to transcription.

Recent studies indicate that the IRES and IRES-binding proteins may not act alone to promote cap-independent viral mRNA translation. In addition to an IRES, the HCV genome contains a highly structured 98-nt 3′ UTR (32, 265). Sequence analyses revealed that the HCV 3′ UTR is highly conserved among viral isolates, suggesting that it may function as a requisite element in HCV replication. This region of the HCV genome contains a high-affinity PTB-binding site (216, 464). Interestingly, Lai and colleagues found that the presence of this PTB-binding site actually relieved translational repression conferred by PTB binding to a second high-affinity site located within the HCV IRES (217, 218). It has been proposed that PTB may control IRES-mediated translation by interacting with both the viral RNA and an unknown factor(s) to (i) bind the viral 3′ UTR and relieve translational repression from within the HCV IRES and (ii) enhance translation through circularization of the HCV RNA into a closed-loop translation complex. The exact role of PTB-mediated translational suppression and stimulation in the context of HCV replication have yet to be determined. One possibility is that PTB interactions with discrete regions of the HCV genome ensure the efficient translation of only genome-length RNA, and provide translational fidelity to HCV polyprotein synthesis.

Another 3′ UTR-directed mechanism of stimulating viral mRNA translation has been described in rotavirus-infected cells. In contrast to cellular mRNAs, rotavirus mRNAs lack a poly(A) tail. Viral mRNA stability is achieved in part through the actions of the viral NSP3A protein, which binds to the 3′ end of virus-encoded mRNAs. During rotavirus infection, NSP3A directly interacts with the eIF4G isoform, eIF4GI (376). Interestingly, NSP3A may represent a virus-specific analog of PABP, in that it appears to compete with PABP for interaction with eIF4G and participation in the translation initiation process. Results from in vitro experiments suggest that NSP3A may actually “evict” PABP from the eIF4F translation initiation complex. These results support a model in which NSP3A mediates selective translation of viral mRNA by (i) disrupting the cellular mRNA translation through “eviction” of PABP from the “closed-loop” translation complex and (ii) facilitating viral mRNA translation through interactions with eIF4G and the viral mRNA 3′ end. Moreover, eviction of PABP from the translation complex is likely to contribute to host protein synthesis shutoff during rotavirus infection by reducing the overall efficiency of host mRNA translation.

Influenza virus.

Similar to the previous examples provided by the picornaviruses, influenza virus depends on the recruitment of specific host factors to mediate selective and efficient translation of viral mRNAs during infection. Influenza virus is responsible for up to 70,000 deaths a year in the United States and 20,000,000 worldwide in the worst epidemic years and remains one of the most dreadful threats for a recurring virus pandemic. Like other cytopathic viruses, influenza virus dramatically perturbs the normal synthesis of host macromolecules. However, unlike most nononcogenic RNA viruses, the replicative cycle of influenza virus includes both a nuclear and cytoplasmic phases. Accordingly, the evolution of the translational strategies of this virus is likely to be dictated by a different set of selective pressures (274). Given its small negative-strand, segmented RNA genome size, it is almost intuitive that such a successful virus must have evolved a plethora of ingenious schemes to ensure selective and efficient translation of viral mRNAs. In the sections that immediately follow, we review our present understanding of how influenza virus imposes selective translation of viral mRNAs over cellular mRNAs, but yet manages to keep the infected cells translationally competent, such that the steps involved in the protein synthetic pathway are not compromised.

Selective translation of influenza virus mRNAs.

The mechanisms by which influenza virus mediates selective translation of viral mRNAs are summarized in Table 3. In influenza virus-infected cells, cellular mRNAs are subjected to a modest decrease in transcription rates (approximately twofold) (250, 251). In addition, degradation of cellular mRNAs is evident late after infection (22, 215). Moreover, newly synthesized cellular mRNAs fail to reach the cytoplasm after infection due the degradation of nuclear RNA early after infection (250). Although the mechanism has not been determined, it is thought that influenza virus-induced cellular mRNA degradation in the nucleus is initiated by the cleavage of the 5′ ends of cellular RNA polymerase II transcripts by the viral cap-dependent endonuclease. The decapped RNAs would probably be more susceptible to degradation by cellular nucleases, since it is well established that the 5′ cap structure stabilizes mRNAs (274).

TABLE 3.

Translational regulatory mechanisms of influenza virus

Strategy	Reference(s)
Host shutoff of cellular protein synthesis	250, 288
Inhibition of cellular mRNA transport and degradation of mRNAs in the nucleus	250
Inhibition of PKR by P58IPK	135, 249, 294–296, 447
Inhibition of PKR by NS1	184, 312, 448
Cap-dependent, selective translation of influenza virus mRNAs	140, 142, 247, 250, 362
Inhibition of cellular mRNA translation at initiation and elongation stages	140, 247
Innate ability of influenza virus mRNAs for preferential translation	140, 142, 247
Viral 5′ UTR-mediated preferential translation of influenza virus mRNAs	361, 362
Recruitment of eIF4G to the viral mRNA	9
Dephosphorylation of eIF4E	110
Temporal regulation of influenza virus protein synthesis	102, 486, 487

However, the above events cannot completely account for the dramatic cessation of cellular protein synthesis, because preexisting cytoplasmic cellular mRNAs are stable and functional when tested in cell-free translation systems (143, 250). Thus, the shutoff of cellular protein synthesis is not primarily due to the degradation or modification of cellular mRNAs. Indeed, polyribosome analysis unveiled that there is a significant diminution in the association of cellular mRNAs with large polyribosomes (247). Furthermore, many cellular mRNAs remained polyribosome associated even though these mRNAs were not translated inside the infected cell. Although there were hints in the literature that actively translated mRNAs were associated with a pool of polyribosomes that were cytoskeleton bound (298), subsequent studies showed that the polyribosomes containing both the viral and nontranslated cellular mRNAs were associated with the cytoskeleton (252). This is in contrast to poliovirus-infected cells, in which cellular mRNAs are dissociated from the polyribosomes and the cellular cytoskeleton (252, 298). These observations suggest that there are other mechanisms that direct preferential influenza viral protein synthesis.

Contribution of influenza virus mRNA structure upon selective translation.

It was conceivable that translational selectivity, at least at the level of initiation, could be due to competition between cellular and viral mRNAs for limiting components of the translational machinery as shown in other systems, including the reoviruses (309, 388, 471). Since viral mRNAs do not have an advantage merely due to sheer mass (143), it was likely that the influenza virus mRNAs are intrinsically better initiators of translation due to certain unknown structural qualities. Such structural features could deceive the cellular protein-synthesizing machinery into making only viral proteins (141, 143). The first persuasive evidence that influenza virus mRNAs are intrinsically efficient initiators of translation was provided by studies in which cells were doubly infected with influenza virus and adenovirus (246, 247, 249). These early experiments also showed that influenza virus also has a strategy to sustain overall high levels of protein synthesis.

In these adenovirus-influenza virus doubly infected cells, influenza virus proteins accumulated essentially to the same levels as in cells infected by influenza virus alone (246). These data suggested that influenza virus is able to overcome the halts on host cell mRNA transport and translation imposed by adenovirus (14), demonstrating that the virus establishes its own translational and transport regulatory mechanisms. Furthermore, influenza virus mRNAs were more efficiently translated than late adenovirus mRNAs, which are believed to be strong mRNAs due to the presence of the tripartite leader sequences (310). Further evidence that the structure of influenza virus mRNAs plays a key role in their selective translation was obtained from a study conducted by Alonso-Caplen et al. (5). The authors found that the translation rate of the influenza virus nucleocapsid protein (NP) mRNA, expressed from a recombinant adenovirus, was equivalent to that of the native NP expressed by influenza virus itself. The recombinant NP mRNA translation rates were remarkably efficient despite the absence of all other influenza virus gene products. Thus, these studies indicated that the sequence and/or structure alone of an influenza virus RNA can confer enhanced translatability.

Because influenza virus is not readily amenable to genetic studies, researchers have resorted to alternative strategies to study the contribution of the viral mRNA structure to selective translation. Definitive evidence that influenza virus mRNAs have an innate ability to be preferentially translated was obtained from transfection and infection studies in which representative viral or cellular cDNAs were transfected into COS-1 cells, which were then infected with influenza virus (140). It was shown that mRNA translation, directed by cellular transfected genes such as interleukin-2 or secreted embryonic alkaline phosphatase (SEAP) was markedly shut off after viral infection. In contrast, an exogenously introduced influenza virus gene encoding NP was not subjected to the translational blocks imposed on the cellular genes. Subsequent studies using chimeric constructs in which the viral 5′ UTR was fused to a cellular mRNA demonstrated that the selective translation of influenza virus mRNAs is mediated almost exclusively by the viral 5′ UTR (142).

Influenza virus recruitment of Grsf-1.

Unlike the IRES of poliovirus or the tripartite leader of adenovirus, influenza virus mRNAs do not possess any extensive secondary structures. Instead, they contain short and relaxed 5′ leaders with no upstream AUGs. Using gel mobility shift and UV cross-linking analysis, several factors have been identified that bind to cis-acting sequences present in the viral 5′ UTR (361). A yeast three-hybrid screen using a HeLa cell cDNA library revealed that one of these 5′ UTR-interacting factors turns out to be the cellular protein GRSF-1 (362). In vitro and in vivo binding analyses demonstrated that GRSF-1 can specifically bind to the NP 5′ UTR but not select NP 5′ UTR mutants or cellular RNA 5′UTRs (362) (Fig. 12A). Importantly, recombinant GRSF-1 was found to specifically stimulate the translation of an NP 5′ UTR-driven template in cell-free translation systems. Furthermore, the translation efficiency of NP 5′ UTR-driven templates was markedly reduced in GRSF-1-depleted HeLa cell extracts but was restored by GRSF-1-in reconstituted extracts. Competition experiments using NP 5′ UTR sequences similarly demonstrated a requirement for GRSF-1 binding in the translation of viral but not cellular mRNA (Fig. 12B). Taken together, these results demonstrate a specific interaction between GRSF-1 and the influenza virus 5′ UTR. More importantly, these results suggest that influenza virus may recruit GRSF-1 to the 5′ UTR to ensure the selective translation of viral mRNAs in infected cells.

FIG. 12.

GRSF-1 binds the 5′ UTR of influenza virus mRNAs to stimulate viral mRNA translation. (A) Specific interaction of GRSF-1 with the 5′ UTR of influenza virus NP mRNA. The 5′ UTR of the viral NP mRNA has been functionally divided into at least three contiguous sequence elements (denoted by A to C) (361). The 12-nt A region (underlined) is conserved in all influenza virus mRNAs (274). Dashed lines indicate deleted regions within the respective 5′ UTR constructs used to assess GRSF-1 binding. A_INV refers to the NP 5′ UTR in which the sequence of the A region has been inverted. The sequence of a control 5′ UTR from the cellular SEAP gene is shown at the bottom. The A and B regions of the NP 5′ UTR are both required for binding to GRSF-1 (362). (B) Competition for GRSF-1 selectively blocks translation from the NP 5′ UTR in a cell-free system. To determine the contribution of GRSF-1 to influenza virus mRNA translation, the expression of a luciferase reporter protein was placed under control of the NP 5′ UTR. Relative luciferase activity from the resulting NP-Luc construct was assessed in a cell-free system in the absence (No comp) or presence of competitor oligonucleotides encoding the NP 5′ UTR (NP), NP-A, SEAP, or NP-C. (C) Structural representation of GRSF-1. The three RRMs are indicated in black. The 41-amino-acid acidic domain is shown in gray.

It is not yet known how GRSF-1 affects the overall translation efficiency of influenza virus mRNAs within the infected cell. GRSF-1 was first cloned by probing protein blots with a labeled G-rich RNA element probe (384), which is predominantly present in the cytoplasm, contains three RNA recognition motifs (RRM), and belongs to the RRM protein superfamily (Fig. 12C). Consistent with its RNA-binding properties, GRSF-1 also possesses a carboxyl-terminal acidic domain. Several models have now been proposed to explain how GRSF-1 mediates selective translation (Fig. 13) (362). Potentially relevant is the observation that eIF4E, the subunit responsible for recognition of the cap structure, is modestly dephosphorylated in influenza virus-infected cells (110). Furthermore, influenza virus mRNAs display exceptional ability to be translated in adenovirus-infected cells where eIF4E is severely dephosphorylated. eIF4E dephosphorylation leads to a decrease in the cap-binding activity (392), which may contribute to the inhibition of the cellular mRNA translation in influenza virus-infected cells. As an RNA-binding protein, it is conceivable that GRSF-1 may promote eIF4E function and interaction with the components of the cap-binding complex to stimulate mRNA translation. Alternatively, GRSF-1 may directly participate in ribosome recruitment to the mRNA, possibly bypassing the limitations imposed by reduced eIF4E function. One possible scenario is that GRSF-1 participates in the formation of the RNA-binding complex with eIF4G and eIF4A, independent of eIF4E (Fig. 13). In either case, eIF4G is likely to be required, since influenza virus mRNA translation is acutely cap dependent (244, 274).

FIG. 13.

Models for the selective translation of influenza virus mRNAs by GRSF-1. GRSF-1 binding may be facilitated by the AGGGU sequence spanning the junction of the A and B regions within the NP 5′ UTR (362). The two models shown reflect the cap dependency and requirement for eIF4G in influenza virus mRNA translation. (I) GRSF-1 may participate in recruitment of the cap-binding complex to the mRNA by physically interacting with one or more components of eIF4F. Active recruitment of the eIF4F complex by GRSF-1 may overcome the reduced affinity of phosphorylated eIF4E for the m⁷G cap that occurs during influenza virus infection (110). As a result, GRSF-1 may enhance ribosome binding to the mRNA by increasing the stability of the eIF4F-mRNA complex (bottom). (II) Alternatively, GRSF-1 may allow viral mRNA translation to proceed independently of eIF4E by directly participating in ribosome recruitment with the other components of the eIF4F complex. In this model, GRSF-1 would function in the absence of eIF4E to promote ribosome binding to the mRNA (bottom), thereby avoiding the limitations on translation due to eIF4E phosphorylation. 4E, eIF4E; 4G, eIF4G, 4A, eIF4A.

Temporal regulation of influenza virus mRNA translation.

In addition to regulating cellular mRNA translation, influenza virus carries out temporal regulation of its own viral gene expression at the level of translation. For example Yamanaka et al. (486, 487) transfected HeLa cells with CAT reporter gene appended to the 5′ UTR of each of the viral mRNAs separately and superinfected these cells with influenza virus. They measured CAT activity at early and late times after infection. When the CAT construct contained the 5′ UTR of an mRNA encoding an early protein such as NS, CAT activity was increased early after infection. Conversely when this was done with a late viral protein such as neuraminidase, CAT activity was higher at late times after infection. This avenue of investigation has not been pursued further, and exact sequences or trans-acting factors responsible for this regulation have not been identified. More recently, however, it was found that the influenza virus NS1 protein could stimulate the synthesis of the viral M1 protein (102). Site-directed mutagenesis studies showed that specific sequences within the M1 mRNA 5′ UTR are required for this stimulation. These data, taken together with results discussed above, suggest a key role for the 5′ UTR of influenza virus mRNAs in dictating temporal and selective translation in the infected cell.

Maintenance of translation in influenza virus-infected cells.

Not only does influenza virus have to exert preferential translation of its viral mRNAs, but also it has to ensure that the cell remains optimally translationally competent during infection (244). Without both these major strategies, viral replication might be compromised, a scenario unacceptable to an actively replicating cytopathic virus. The translational competence of the infected cell is assured because influenza virus has developed an intricate strategy to repress the activity of the PKR protein kinase. That influenza virus represses PKR was first demonstrated by analyzing cells doubly infected with influenza virus and the adenovirus VA RNAI-negative mutant dl331 (246, 247). In cells infected by dl331 alone, there was a dramatic decline in the levels of both viral and cellular protein synthesis (458). This was due to excessive phosphorylation of the eIF2α subunit by an active PKR, which cannot be inhibited due to the absence of the virus-encoded VA RNAI (reviewed by Mathews and Shenk [323]). In contrast, when dl331-infected cells were superinfected with influenza virus, a reduction of the protein kinase activity normally detected during dl331 infection was observed (246, 247). These data provided the first evidence that influenza virus encodes or activates a gene product that, analogous to VA RNAI, inhibits PKR and prevents any resultant inhibition of protein synthesis initiation. It was subsequently shown that the suppression of PKR activity also occurs in cells infected by influenza virus alone (253).

Recruitment of P58^IPK and inhibition of eIF2α phosphorylation during influenza virus infection.

A number of eukaryotic viruses have devised one or more strategies to minimize the deleterious effects on protein synthesis caused by activation of PKR (for a recent review, see reference 131). Unlike the strategies used by other viruses described above, influenza virus utilizes at least two strategies, involving a cellular and a viral protein, to block PKR activity. During viral infection, influenza virus mobilizes a cellular protein, termed P58^IPK, to repress PKR activity by blocking both the autophosphorylation of PKR and the subsequent phosphorylation of eIF2α by an active form of PKR (294, 295). In uninfected cells, P58^IPK appears to form an inactive complex with its own inhibitor, termed I-P58^IPK (296). In response to activating stimuli, such as viral infection or other cellular stresses, P58^IPK is released from its inhibitor. As a result, PKR activity is repressed by a physical interaction between P58^IPK and PKR (130, 135, 378). To further complicate the story, two different inhibitors of P58^IPK, the molecular chaperone Hsp40 (332) and another cellular protein P52^IPK (128), have recently been identified. It has been postulated that these independent P58^IPK complexes are regulated by distinct cellular pathways (128). Once released from its inhibitor, P58^IPK may negatively regulate PKR activity through multiple steps, including the recruitment of molecular chaperone Hsp70 to inhibit kinase activity (333) and disruption of PKR dimerization (447).

Role of the influenza virus NS1 protein in viral mRNA translation.

The NS1 protein of influenza virus is an RNA-binding factor that inhibits both the nuclear export of poly(A)-containing mRNA and the splicing of pre-mRNA (86, 119, 311, 385). In addition to binding to poly(A) mRNA and a stem-bulge region in U6 small nuclear RNA, the NS1 protein binds to dsRNA (183). Evidence has accumulated to indicate that NS1 is involved in the translation of select viral mRNAs, including those encoding the viral matrix and nucleocapsid proteins (361). In this regard, NS1 is thought to interact with the viral 5′ UTR to selectively stimulate the initiation of viral mRNA translation (244). To shed some light on how such translational selectivity may occur, recent evidence now indicates that NS1 can also form a complex with eIF4G in extracts from influenza virus-infected cells (T. Aragon, S. de la Luna, I. Novoa, L. Carrasco, J. Ortin, and A. Nieto, submitted for publication). Thus, we may hypothesize that NS1 can recruit eIF4G to the viral 5′ UTR to facilitate translation initiation. Such an interaction could also contribute to the host shutoff due to competition with cellular mRNAs for eIF4G. Clearly, the exact mechanisms of NS1 specificity in mediating selective viral mRNA translation remain to be determined. It will be interesting to examine whether NS1 may interact with GRSF-1 to synergistically promote selective viral mRNA translation.

NS1 may also function during influenza virus infection to block the actions of PKR. Given the RNA-binding properties of NS1 and its ability to bind dsRNA, it was hypothesized that NS1 could compete with PKR in influenza virus-infected cells for binding to dsRNA. This tenet is supported by the observation that NS1 inhibits the activation of PKR and, as a result, the phosphorylation of eIF2α in vitro (312). Furthermore, the NS1 protein also blocks the inhibition of translation caused by dsRNA-mediated activation of PKR in reticulocyte lysate extracts. The relevant role of NS1 in PKR regulation is further strengthened by the finding that the two proteins can form a specific complex in vitro (448). An inactive mutant of NS1, which lacks a functional RNA-binding domain, was unable to bind to PKR. Moreover, a PKR mutant defective in dsRNA binding did not interact with or inhibit the NS1 protein in vivo. These results suggest that NS1 exerts its effect, at least in part, through heterodimerization with PKR, possibly in an RNA-dependent manner. However, discretion is advised when using dsRNA-binding mutants of PKR to determine the mechanism of these interactions because both the dsRNA-binding and protein interaction properties of PKR are closely embedded in the same regions. At any rate, the fact that NS1 plays an in vivo role in modulating PKR function was recently demonstrated by studies of mutant influenza viruses with a defective NS1 protein (184). These variant viruses could not block the activation of PKR in infected cells, leading to enhanced phosphorylation of eIF2α and suppression of mRNA translation. Furthermore, the level of phosphorylation of PKR and eIF2α was well correlated with the defect in virus protein synthesis. Consistent with its role in translation modulation, the NS1 protein has been shown to stimulate the translation of viral mRNAs (86, 102), although it has not been determined if the PKR pathway is involved. Collectively, these results suggest that NS1 may facilitate viral replication by blocking the critical PKR-dependent arm of the cellular IFN response. This idea is supported by work by Garcia-Sastre et al., who used reverse genetics to engineer a recombinant influenza virus, termed delNS1, which lacks the NS1 gene (139). Similar to wild-type influenza virus, delNS1 replicated to high titer within cells deficient in IFN signaling pathways. However, delNS1 replication was severely limited in cells in which IFN signaling remained intact. Similarly, delNS1 replicated to lethal titers in mice with a target deletion in the STAT1 gene, which renders cells unable to respond to IFN (77, 334). IFN-competent control mice effectively suppressed delNS1 replication. Thus, the NS1 gene of influenza virus may not play a direct role in viral replication, but, rather, it functions to block the antiviral effects of the host IFN system. In this regard, NS1 may confer translational competence to influenza virus by removing the translational blockade imposed by PKR.

Influenza virus appears to encode more than one strategy to repress PKR function (Table 4). Influenza virus is known to generate large amounts of both negative-strand and positive-strand viral RNAs during infection, forming dsRNAs that are capable of activating PKR. It seems logical that the virus uses a cellular factor (P58^IPK) and a viral protein (NS1) to inhibit the activation of PKR in order to ensure efficient protection against the resulting inhibition of translation that would block virus replication. Alternatively, some of these PKR inhibitors may interfere with activities of PKR not directly related to the regulation of protein synthesis. Perhaps such viral multistrategies are designed to fine-tune the activities of the enzyme at different stages in the viral life cycle. Furthermore, since PKR is found in both the nucleus and cytoplasm (68) the use of NS1, a predominantly nuclear protein (166), and P58^IPK, a cytoplasmic protein (268), may allow influenza virus to control PKR functions in both cellular compartments. Several key questions remain to be addressed in future studies. (i) How does influenza virus activate the P58^IPK-PKR regulatory pathway? (ii) Would cells devoid of P58^IPK be more susceptible to influenza virus infection and replication? (iii) Do NS1 and P58^IPK work in a synergistic manner to inhibit PKR function and enhance the translation of viral mRNAs?

VIRAL MODIFICATION OF CELLULAR FACTORS

Viral replication requires a large amount of energy and thereby demands almost total metabolic control of cellular resources. As discussed in the previous sections, the competition for resources imposed by these conditions has created a playing field in which viruses have evolved mechanisms to supersede cellular mRNA translation through the recruitment and/or modification of translation factor function. Many of these processes of translation factor modification may have arisen through the efforts by viruses to block cellular countermeasures aimed at disrupting viral mRNA translation. As described in this section, the effects of translation factor modification range from altering the efficiency of cap-dependent translation and translation elongation to altering the rate of global mRNA translation and the efficacy of the innate antiviral response of the host cell.

Inactivation of eIF4E and modulation of the eIF4E-binding proteins.

eIF4E may be considered the pivotal translation initiation factor. Compared to the other translation factors, it is present in limiting amounts in the cell, where it is required for assembly of the 5′ cap-binding complex. As described above, eF4E binds to the 5′ ends of both cellular and viral mRNAs and interacts with eIF4A and eIF4G. This macromolecular complex constitutes eIF4F and facilitates the binding of the mRNA to the 43S preinitiation complex (335). The affinity of eIF4E for the mRNA 5′ cap is increased by phosphorylation on serine 209 and occurs in response to mitogenic stimulation (237, 431, 432). Analyses of eIF4E activity in the presence or absence of serine 209 phosphorylation has indicated that phospho-eIF4E is stimulatory for mRNA translation (432). The cellular enzyme responsible for serine 209 phosphorylation has been putatively identified as the Mnk1 protein kinase. Mnk1 was shown to phosphorylate eIF4E on serine 209 in vitro (473). More recent studies demonstrated that overexpression of Mnk1 could induce high levels of eIF4E phosphorylation in vivo (474), although the effects of Mnk1 overexpression on mRNA translation were not directly examined. These results are consistent with previous observations demonstrating that mitogen-induced eIF4E phosphorylation was dependent on the activity of the extracellular signal-related kinase and mitogen-activated protein kinase (339, 341, 472). Thus, Mnk1 may link eIF4E phosphorylation and translation stimulation through the mitogen-activated protein kinase-signaling pathway (387). In general, eF4E function is a requisite for cap-dependent translation but is thought to play a more critical role in the translation of mRNAs that have long 5′ UTRs with regions of extensive secondary structure that are nonconducive to ribosomal scanning. These mRNAs are translated inefficiently due to limitations in eIF4E and the associated helicase activity of the assembled eIF4F complex (reviewed by Sonenberg [432]). Serine 209 phosphorylation is thought to increase the translational efficiency of these mRNAs by stimulating eIF4E cap-binding activity and increasing the level of mRNA-bound eIF4F helicase activity sufficiently to melt mRNA secondary structure.

The activity of eIF4E is also regulated through direct interaction with the eIF4E-binding proteins, 4E-BP1 to 4E-BP3 (367, 380). The 4E-BPs are low-molecular-weight proteins that link cap-dependent translation to mitogenic signaling pathways and are themselves directly regulated by phosphorylation (105, 305, 367). A model for eIF4E regulation by the 4E-BPs has been presented (387, 432). This model is based on the observations that interaction of eIF4E with 4E-BP1 disrupted the assembly of an eIF4E-eIF4G complex (172). 4E-BPs may facilitate the suppression of cap-dependent translation by binding to eIF4E and preventing the formation of an active eIF4F complex. Phosphorylation of the 4E-BP results in dissociation of the eIF4E–4E-BP inhibitory complex (367), thereby stimulating mRNA translation (387). 4E-BP phosphorylation and translation stimulation occur in response to mitogenic signaling and other cell growth-modulatory stimuli (154, 469). Elucidation of such cellular signaling cascades is currently an intense area of research. Viral disruption of eIF4E regulation may therefore have far-reaching consequences beyond supporting viral replication, including implications for apoptosis and oncogenic transformation (432).

Viruses have targeted the processes of eIF4E regulation to facilitate translational selectivity for viral mRNAs during infection. In mammalian cells, the level of eIF4E serine 209 phosphorylation is reduced during infection with a number of viruses, including adenovirus and influenza virus (110, 244, 411, 499). In short, the block in eIF4E phosphorylation corresponds to a decrease in the level of host mRNA translation with little or no effects on translation of viral mRNAs. The molecular mechanisms by which eIF4E phosphorylation is reduced by influenza virus are not clear. Examination of mRNA translation in cells infected with influenza virus revealed that (i) virus infection significantly reduced the extent of eIF4E phosphorylation and the pool of active eIF4E and (ii) the efficiency of viral mRNA translation was insensitive to the resulting low levels of functional eIF4E and the eIF4F complex (110). All influenza virus mRNAs contain a short, conserved 5′ UTR predicted to possess limited, if any, secondary structure (274). This short linear 5′ UTR directs viral mRNA translation with a remarkably high efficiency. Such “strong” mRNAs may be less sensitive to limitations in the availability of eIF4F and associated helicase activity imposed by eIF4E dephosphorylation. As a result, dephosphorylation of eIF4E during influenza virus infection may favor the translation of viral mRNAs over the host mRNA and therefore may contribute to the host shutoff phenomenon. At this point, it is not known if influenza virus induces an eIF4E phosphatase activity, inhibits an eIF4E kinase including possibly Mnk1, or simply makes eIF4E unavailable for phosphorylation. These possibilities should provide a fertile area for future research in viral translational control mechanisms.

In cells infected with adenovirus, the block in eIF4E phosphorylation plays a dual role of contributing to the host shutoff during late-stage infection and selecting for the translation (by ribosome shunt) of viral mRNAs that contain the tripartite leader. The actual mechanism by which adenovirus prevents eIF4E phosphorylation is not clear. Evidence from experiments that examined infection kinetics and levels of eIF4E phosphorylation has implicated a late adenovirus gene function in the coordinate reduction in eIF4E phosphorylation (499). An attractive hypothesis is that adenovirus may directly or indirectly inhibit Mnk1 or another cellular protein kinase(s) that phosphorylates eIF4E.

Viral regulation of eIF4E function also occurs at the level of 4E-BP activity. Both poliovirus and EMCV inhibit 4E-BP phosphorylation within infected cells (155). Inhibition of 4E-BP phosphorylation may therefore contribute to the host shutoff of protein synthesis observed during picornavirus infection. This idea is supported by analyses of 2A-pro deficient strains of EMCV. Loss of 2A-pro decreased the efficiency of viral protein synthesis and abolished virus-induced host shutoff in infected cells (441). Interestingly, mutant viruses exhibited enhanced viral replication and increased efficiency of viral mRNA translation during infection in the presence of rapamycin and wortmannin, chemical inhibitors of 4E-BP phosphorylation (25). Thus, inhibition of 4E-BP phosphorylation complements EMCV mutations in 2A-pro to rescue viral mRNA translation (441). Together, these results indicate that inhibition of 4E-BP and repression of eIF4E function contributes to the host shutoff induced by picornavirus infection (26).

Cleavage of eIF4G.

With the exception of HAV, the host shutoff induced during picornavirus infection is extremely intense and results in nearly complete disruption of cellular mRNA translation to favor IRES-mediated translation of the viral mRNA. A main feature of this host shutoff involves disruption of eIF4F function through virus-mediated cleavage of eIF4G (Fig. 5), although it is now clear that other factors, including inhibition of 4E-BP phosphorylation, play a role in the shutoff process (174, 468). Cleavage of eIF4G proceeds in part through the actions of the virus-encoded 2A-pro during EMCV and poliovirus infections or through the actions of protease-L in foot-and-mouth disease virus infection (reviewed in references 197, 258, and 340). Cleavage of eIF4G effectively selects for cap-independent, IRES-mediated viral mRNA translation by removing the competition for the translational machinery imposed by cap-dependent cellular mRNA translation.

The implications of eIF4G cleavage upon host cap-dependent mRNA translation are best understood by examining eIF4G structure and function. eIF4G has been cloned from several different species and is present as structurally distinct isoforms that contain binding sites for interaction with eIF4E, eIF4A, eIF3, PABP, and the Mnk1 protein kinase (153, 214, 258, 340). The current model for eIF4F function in cap-dependent translation proposes that eIF4G serves as a molecular bridge for the assembly of the cap-binding complex upon the 5′ end of the mRNA and facilitates its interaction with the 43S preinitiation complex. Furthermore, eIF4G may potentiate eIF4E phosphorylation and increased eIF4E activity by recruiting Mnk1 to the eIF4F complex (153, 383, 474). Poliovirus 2A-Pro cleaves eIF4G to yield discrete amino- and carboxyl-terminal cleavage products (Fig. 14). This prevents assembly of a functional eIF4F complex, thereby disrupting an essential step in cap-dependent translation initiation. Interestingly, cap-independent viral protein synthesis of select picornaviruses is stimulated by the carboxyl-terminal eIF4G cleavage product, which contains the binding sites for eIF3 and eIF4A (340, 354).

Recent evidence indicates that eIF4G isoforms are differentially targeted and cleaved upon picornavirus infection. The host shutoff of protein synthesis during poliovirus infection has been attributed to cleavage of eIF4GII. Importantly, eIF4GII cleavage was shown to coincide with the kinetics of host shutoff of protein synthesis and occurred during later-stage infection, after the cleavage of eIF4GI (161). Similarly, cleavage of eIF4GII was shown to occur during later-stage human rhinovirus infection, again coinciding with the kinetics of the host protein synthesis shutoff (440). Together, these results demonstrate that the different eIF4G isoforms are differentially targeted during picornavirus infections and suggest that eIF4GII cleavage may be the rate-limiting step in the shutoff of host protein synthesis during poliovirus and rhinovirus infection.

Cleavage of PABP: disruption of the closed-loop translation complex.

In addition to cleavage of eIF4E, picornaviruses may disrupt host mRNA translation through the modulation or modification of PABP. PABP is a 70-kDa RNA-binding protein that has remained highly conserved in evolution and plays a direct role in mRNA stability through its interaction with the poly(A) tails of mRNA (225). Recent evidence indicates that PABP directly participates in mRNA translation by functioning to bring the 3′ UTR of the mRNA into the proximity of the 5′ cap and the cap-binding complex. As described above, this circularization of the translation initiation complex is facilitated through PABP interactions with the mRNA poly(A) tail and eIF4G. The resulting “closed-loop” translation initiation complex is thought to stabilize assembled initiation factors and increase translation efficiency (225, 401). Thus, viral disruption of PABP function can be expected to (i) alter mRNA stability and (ii) reduce the overall rate and efficiency of cap-dependent mRNA translation.

Interestingly, it now appears that PABP is targeted for cleavage by 2A-Pro of the enteroviruses, coxsackievirus, and poliovirus. Analysis of PABP during coxsackievirus infection revealed that it was proteolytically cleaved during infection (259). In these studies, PABP was efficiently cleaved in vitro and in vivo by the virus-encoded 2A-Pro. Cleavage of human PABP occurred in a unique position that resulted in separation of the RNA-binding activity from the homodimerization activity. Importantly, the proteolytic fragments were inefficient at stimulating mRNA translation, suggesting that PABP cleavage may impart host translational suppression. Similar observations were extended to poliovirus infection, in which PABP cleavage was shown to occur through the actions of the viral 2A-Pro and, to a lesser extent, the viral 3C protease (233). In both studies PABP cleavage was accompanied by a dramatic loss of cellular translational activity in vitro and in vivo (233, 259). Taken together, these results suggest that direct cleavage of PABP may contribute to the inhibition of host mRNA translation during enterovirus infection. However, the extent to which PABP cleavage contributes to host protein synthesis shutoff, versus cleavage of eIF4G, has not been directly compared. Moreover, using infected cells, it will now be important to separate the effects of PABP cleavage on mRNA stability from direct translation-regulatory effects due to disruption of the “closed-loop” translation complex. It is tempting to propose that the cleavages of the PABP and eIF4G isoforms occur at distinct stages during viral infection. In this case, PABP cleavage may function to disrupt ongoing host translation while eIF4G cleavage events may be directed toward blocking de novo cap-dependent translation during infection.

Modification of EF-1.

In addition to modifying the components necessary for translation initiation, viruses may target the process of translation elongation to facilitate the efficient translation of viral mRNA. Translation factor EF-1 catalyzes the critical step of delivering the aminoacyl-tRNAs to the elongating ribosome. The viral regulatory protein, ICP0, forms a stable complex with EF-1δ during HSV infection (256). Analysis of the translation efficiency of a reporter protein in RRL revealed that the addition of recombinant ICP0 repressed translation in a dose-dependent fashion. Although the actual rates of translation elongation and the efficiency of translation initiation were not addressed, it was concluded that ICP0 may function during specific stages of HSV infection to modulate the translation of cellular and viral mRNAs. Interestingly, EF-1δ exists as hypo- and hyperphosphorylated isoforms within HSV-infected cells (256). Analysis of HSV mutants suggests that EF-1δ phosphorylation is mediated by a protein kinase encoded by the product of the viral U(L)13 gene (257). Although a physiological role for U(L)13-mediated phosphorylation of eEF-1δ during HSV infections has not been demonstrated, it is interesting to speculate that EF-1δ phosphorylation may promote viral protein synthesis by inducing the coordinate dissociation of EF-1δ from ICP0 at a specific point(s) during viral replication. Finally, EF-1δ has also been shown to interact with the Tat protein of HIV-1 (485). In this case, the Tat–EF-1δ interaction resulted in a dramatic reduction in the efficiency of cellular but not viral mRNA translation. As with HSV, however, confirmation of EF-1 regulation by Tat, and its physiological role during HIV infection remain an open question.

Disruption of eIF2α phosphorylation.

Eukaryotic cells respond to stress conditions, including viral infection, in part by down-modulating the overall rate of protein synthesis. This translational control response to stress occurs largely through the modification of eIF2. eIF2 functions to deliver the Met-tRNA_i to the 40S ribosome, and this constitutes a rate-limiting step to translation initiation when eIF2 is modified through phosphorylation by specific cellular serine-threonine protein kinases. Currently, at least five distinct eukaryotic protein kinases have been identified that play a role in translational control by modulating eIF2 function; these are the HRI, PERK, PEK, and PKR enzymes and the GCN2 protein kinase (423). Known as the eIF2α protein kinase family, these enzymes respond to specific signals to phosphorylate serine 51 of eIF2α. Functional analyses of the eIF2α protein kinases indicate that each enzyme provides the cell with a unique ability to modulate mRNA translation in response to specific cellular stresses (reviewed in references 201 and 477). For example, the HRI protein kinase is expressed in mammalian reticulocytes and mediates eIF2 α phosphorylation in response to heme depletion (58). PERK and PEK reside on the endoplasmic reticulum, where they mediate translational control in response to endoplasmic reticulum stress (178, 419). The yeast GCN2 enzyme presents an example of both global and specific control of mRNA translation by phosphorylating eIF2α in response to amino acid starvation (202). This results in the specific stimulation of GCN4 translation and concomitant repression of global protein synthesis (201). GCN4 stimulates amino acid production by inducing the expression of amino acid-biosynthetic components. PKR is ubiquitously expressed in most mammalian tissues. As described below, PKR is a component of the IFN-induced cellular antiviral response and a pleiotropic mediator of extracellular signals (68). However, PKR is best known for its ability to phosphorylate eIF2α and repress mRNA translation. Phosphorylation of eIF2α on serine 51 by PKR or the other eIF2α protein kinases inhibits the guanine nucleotide exchange reaction on eIF2 (Fig. 15). The resulting eIF2 [S51-phospho]-GDP binary complex has a higher affinity for the eIF2B guanine nucleotide exchanger than does the nonphosphorylated eIF2 isoform. The increased affinity for eIF2B impedes eIF2B function and results in sequestration of eIF2B within an inactive complex with eIF2 [S51-phospho]-GDP. eIF2B sequestering blocks the requisite recycling of GDP for GTP on eIF2 and prevents de novo eIF2–GTP–Met-tRNAi ternary-complex formation. As a result, mRNA translation initiation is blocked.

FIG. 15.

Translational control by eIF2α phosphorylation. eIF2 is composed of several subunits, including the alpha subunit (eIF2α), which is targeted for phosphorylation by the eIF2α protein kinases. eIF2 participates in mRNA translation by delivering the Met-tRNA_i to the incoming 40S ribosomal subunit in the form of an eIF2–GTP–Met-tRNA_i ternary complex (3°) (335). After Met-tRNA_i binding, eIF2 is released from the complex in an inactive state bound to GDP. The bound GDP is recycled for GTP by the eIF2B guanine nucleotide exchange factor (dotted and gray lines), and the initiation process continues. PKR, PERK, PEK, HRI, and GCN2 can block the guanine nucleotide exchange reaction by phosphorylating serine 51 of eIF2α. As a result, the pool of functional eIF2 is depleted and translation arrests at the initiation stage.

(i) PKR structure and function.

PKR and its role in cellular metabolism have been extensively reviewed (58, 66, 201, 335, 404). Similar to many protein kinases, PKR is regulated through a combination of transcriptional and posttranscriptional processes including regulatory interactions with P58^IPK (135, 294, 378). PKR, however, is unique among the protein kinase superfamily in that it is the target for regulation by virus-encoded inhibitory molecules (241–244, 321, 322). Structurally, PKR is composed of an NH₂-terminal regulatory domain and a COOH-terminal protein kinase catalytic domain (Fig. 16A) (68, 165, 336, 459). Ubiquitously expressed at low levels in virtually all mammalian tissues examined (15), PKR is transcriptionally induced by IFNs, which are secreted by host tissues in response to viral infection (69) (see below). PKR is rapidly activated after binding dsRNA or even single-stranded RNA species that possess regions of extensive secondary structure (70, 165, 248, 254, 316, 325, 331). This clearly imposes problems for many eukaryotic viruses, which possess dsRNA within the virion or produce high levels of dsRNA intermediates during replication. PKR binds dsRNA via its two NH₂-terminal regulatory domain-binding motifs (dsRBM) (Fig. 16A) (113, 165, 224, 363, 395). As a result of these events, the kinase may undergo a conformational alteration which triggers the catalytic activities of PKR (see reference 68 for a detailed review of PKR structure and function). Typically, low levels of dsRNA are required to initiate PKR activation while high levels of dsRNA actually result in inhibition of PKR activation and suppression of kinase function (70). Once bound to activator dsRNA, PKR autophosphorylates on several critical serine and threonine residues (453), rendering the kinase active. Recent evidence by Zhu et al. (503) indicates that in addition to mediating dsRNA activation of PKR, the dsRBMs function to target PKR to the ribosome, thereby potentially providing access to the PKR substrate, eIF2α (335). Activation of PKR can also proceed through interaction with Pact, a novel PKR-activating protein that also binds dsRNA (366). However, Pact appears to activate PKR by a process that is independent of dsRNA, which presumably involves alteration of PKR conformation.

FIG. 16.

Structure and activation of PKR. (A) Domain structure of PKR and sites of viral regulation. The PKR regulatory domain spans amino acids 1 to 264 and includes two dsRBMs (indicated in black) which mediate binding to activator dsRNA. The protein kinase catalytic domain comprises the PKR C terminus (amino acids 265 to 551) and contains the 11 subdomains (denoted by Roman numerals) conserved in all eukaryotic protein kinases (176). Bars indicate the regions that participate in dsRNA binding, dimerization, and substrate interaction. Virus-directed inhibitors that target PKR function are listed beneath their specific target sites and are referenced in Table 4. (B) PKR activation. PKR is transcriptionally induced by IFNs and becomes active through a process of dsRNA binding and dimerization. Once active, PKR can phosphorylate eIF2α. Within a virus-infected cell, PKR-mediated phosphorylation of eIF2α results in a block in protein synthesis, cell growth arrest, and inhibition of viral replication (241, 242). Additionally, PKR may participate in the regulation of other IFN-induced genes by signaling the phosphorylation of IkB or the activation of IRF-1 (277).

Comprising the COOH-terminal region of the molecule, the PKR catalytic domain facilitates the recognition of PKR substrates and the phosphorylation of serine and/or threonine substrate residues. Characteristic of all protein kinases, this region of PKR presents a regulatory target for kinase inhibitory molecules, including protein kinase pseudosubstrates (454). Indeed, recent studies indicate that the COOH-terminal region of the PKR catalytic domain mediates the substrate interaction and is targeted for regulation by virally encoded pseudosubstrate inhibitors (73, 74, 135, 255).

Several studies indicate that dimerization between two PKR molecules may be required for the autophosphorylation process and subsequent catalytic activity (18, 51, 284, 357, 364, 395, 460). Structure-function analyses have shown that PKR dimerization can occur through both dsRNA-dependent and -independent processes (364, 365, 395). Our laboratory has identified a region of PKR, amino acids 244 to 296, which can mediate the dimerization process independently of dsRNA (135, 447) (Fig. 16A). Interestingly, this region also mediates a direct interaction with a diverse set of virus-directed PKR inhibitors (133, 294) that disrupt the PKR dimerization process.

PKR clearly has other substrates in addition to eIF2α, consistent with its reported roles in growth factor and calcium-mediated signal transduction (344, 435), the regulation of transcription (266, 277, 317, 488), and the induction of apoptosis (87, 293, 490) (reviewed by Proud [382] and Williams [481]). However, it is the phosphorylation of eIF2α by PKR that is most frequently targeted for regulation by viruses. A diagram of the events leading to PKR activation and eIF2α phosphorylation is shown in Fig. 16B. IFNs, produced in response to viral infection, induce the transcriptional activation of PKR, resulting in a high level of PKR expression. During virus infection, the binding of virus-encoded dsRNA to PKR initiates the PKR activation process, which includes the aforementioned dimerization and autophosphorylation events (84, 301, 373). Once activated, PKR phosphorylates eIF2α to limit mRNA translation. As a result of these PKR-mediated processes, viral replication is blocked at the level of protein synthesis (Fig. 15 and 16). High levels of eIF2α phosphorylation lead to an antiproliferative state (19, 61, 89, 337, 395). Such a situation would be incompatible with viral infection and replication, providing another reason why viruses must inhibit PKR.

(ii) Mechanisms of PKR inhibition by eukaryotic viruses.

To avoid the deleterious effects on viral replication due to PKR-mediated eIF2α phosphorylation, many viruses have developed successful strategies to block PKR function, thus avoiding, at least in part, the antiviral effects of IFN. As summarized in Table 4, viruses have employed a range of mechanisms to inhibit PKR function, from targeting the PKR activation process to regulation of catalytic function and beyond. These include directing inhibitors that (i) interfere with the dsRNA-mediated activation of PKR, usually by binding to the conserved dsRNA-binding domains or sequestering RNA activators; (ii) interfere with kinase dimerization; (iii) block the kinase catalytic site and PKR-substrate interactions; and (iv) alter the physical levels of PKR; and (v) may regulate eIF2α phosphorylation directly, or affect components downstream from eIF2α. Each PKR-inhibitory group includes a diverse set of viruses, largely unrelated except for their ability to inhibit the kinase. In addition, some viruses employ multiple strategies for inhibiting PKR, resulting in pleiotropic effects on PKR function. It is noteworthy that such a diverse range of viruses has utilized often-limited genomic resources to target PKR for regulation. This reflects the pivotal role played by the kinase within the IFN-induced antiviral response of the host cell.

Disruption of the IFN-induced cellular antiviral response through inhibition of PKR.

Upon infecting the cell of a vertebrate host, viruses must overcome the innate antiviral response provided by the cellular IFN system. IFNs are a family of cytokines which are secreted by the cells of vertebrate animals in response to viral infection and other cellular stresses. Once exposed to IFN, responsive cells initiate a signaling cascade which culminates in the specific induction of multiple IFN-inducible genes (88), only a small subset of which have been extensively characterized (for a review of the IFN system, the reader is directed to references 230, 346, 374, 403, 414, 415, and 437). The IFN-induced gene products, which play a role in fighting virus infection, include the 2′-5′ oligoadenylate synthetase (28, 422), RNase L (421, 500), the Mx proteins (275, 369), and PKR (15, 212, 336, 451, 459). As a result, IFNs direct a block in viral gene expression at multiple levels, including the inhibition of viral RNA transcription and translation, and the degradation of viral transcripts.

Paramount to the IFN-induced antiviral response is the function of PKR, which not only imposes a limitation on viral mRNA translation but also participates in dsRNA- and IFN-induced signaling events (Fig. 16). These functional properties of PKR greatly contribute to the ability of IFN to provide the body's first level of defense against viral infection. PKR is required for the induction of IRF-1 activation (261, 277), which in turn regulates the transcription of a variety of IFN-inducible genes (260), including the type 1 IFNs, IFN-α and IFN-β (277, 415, 488). PKR also contributes to the IFN-mediated response through its ability to activate the transcription factor, NF-κB (277, 297). Not surprisingly, inhibition of PKR function results in attenuation of the IFN response by blocking one or more of these PKR-dependent events (277, 488). Thus, PKR presents an attractive target for virus-mediated inhibition of the host IFN response (Fig. 16).

(i) Viral inhibition of PKR: HCV.

The implications stemming from viral inhibition of PKR and disruption of the IFN response are perhaps best demonstrated by very recent work on the clinically relevant HCV. HCV now infects more than 2% of the worldwide population, including over 4 million in the United States (6). HCV infection often assumes a persistent course, which can lead to chronic hepatitis and liver cirrhosis, and is strongly associated with the development of hepatocellular carcinoma and lymphoproliferative disorders (355, 356, 463).

HCV infection is currently treated by parenteral administration of type I IFN alone or in combination with ribiviran, a nucleoside analog (326). Problematically, an increasingly high proportion of HCV-infected individuals (60 to 80%) fail to respond to IFN therapy or relapse after therapy cessation (208, 213). Response to IFN therapy differs among the six HCV genotypes but is observed, at some level, in all HCV genotypes worldwide. In an effort to understand the molecular mechanism(s) of HCV-mediated resistance to IFN therapy, several research groups have focused attention on sequencing clinical isolates of HCV from individuals who did or did not respond to IFN therapy (103, 104, 240, 278). What is clear from these studies is that sequence variation from the prototypic IFN-resistant HCV J strain (240) within the nonstructural 5A (NS5A) protein of the HCV polyprotein cleavage product is associated with sensitivity to IFN in Japanese HCV 1B subtypes (103, 104, 278). Viral isolates with multiple amino acid substitutions within a region of NS5A, termed the IFN sensitivity-determining region (ISDR; amino acids 2209 to 2248), were eliminated from HCV-infected patients during IFN therapy, while those exhibiting the prototypic ISDR sequence were IFN resistant, persisting at therapy cessation. These results suggested that HCV NS5A may mediate viral sensitivity to IFN through specific sequences located within or around the ISDR.

Accordingly, it was demonstrated that NS5A from IFN-resistant strains of HCV 1A and 1B can physically bind PKR by an ISDR-dependent mechanism to inhibit kinase function, implicating NS5A as a mediator of the IFN-resistant HCV phenotype (133). It was subsequently hypothesized that mutations within the ISDR may similarly disrupt NS5A function to render HCV sensitive to the PKR-mediated antiviral effects of IFN. Subsequent analyses revealed that ISDR sequence variants of NS5A corresponding to IFN-resistant and -sensitive clinical isolates of HCV 1B (103) exhibited differential abilities to control PKR function in vivo (129, 132, 134). These studies demonstrated that NS5A from IFN-resistant HCV disrupted a critical step of PKR activation, resulting in repression of PKR function. In contrast, clinically defined mutations within the ISDR abrogated the PKR-regulatory function of NS5A. Further experiments mapped the PKR-interactive domain of HCV NS5A to a 64-amino-acid motif that includes the ISDR and the immediate carboxyl-terminal flanking region (129). Thus, it now appears that HCV may mediate resistance to IFN in part by blocking the PKR-dependent arm of the IFN response, through a direct NS5A-PKR interaction. In support of this, mutations within the PKR-binding domain on NS5A, including those within the ISDR, may confer IFN sensitivity by disrupting the NS5A-PKR interaction. Recent work indicates this to be true; infection of stable cell lines expressing wt or mutant NS5A with VSV revealed striking differences in the sensitivity of VSV to IFN in the presence or absence of functional NS5A (134, 377). In these studies, VSV replicated to significantly high levels within IFN-treated cells expressing wt NS5A from IFN-resistant HCV. In contrast, VSV replication was suppressed in cell lines expressing a nonfunctional NS5A mutant, reflecting the innate sensitivity of VSV to the antiviral effect of IFN. Analyses of PKR function and eIF2α phosphorylation attributed the apparent rescue of VSV replication to the NS5A-mediated inhibition of PKR and a resulting block in eIF2α phosphorylation (134) (Fig. 17). Moreover, very recent evidence indicates that HCV disruption of PKR function is not restricted to effects on eIF2α phosphorylation alone. NS5A expression also renders cells refractory to PKR-dependent signaling (134; M. Gale, Jr., unpublished data). Disruption of host PKR signaling pathways is now considered an important mechanism by which HCV may facilitate its domination of the host IFN response.

FIG. 17.

Translational control by HCV. (A) Structural representation of the HCV polyprotein. The individual positions of the polyprotein cleavage products are shown. NS5A (black region) from IFN-resistant HCV can bind and repress PKR (129, 133). (B) NS5A blocks PKR-dependent eIF2α phosphorylation. eIF2α phosphorylation from control (Neo) NIH 3T3 cell lines and those stably expressing NS5A from IFN-resistant HCV (NS5A-1A) or a nonfunctional NS5A mutant (ΔISDR) was assessed by single-dimension isoelectric focusing of cell extracts and anti-eIF2α immunoblot analysis (133). Cells were mock infected (lanes 1, 3, and 5) or infected with VSV (lanes 2, 4, and 6). Arrows denote the positions hypo- and hyper-phosphorylated isoforms of eIF2α (eIF2α and eIF2αP, respectively). Hyperphosphorylated eIF2α is phosphorylated on serine 51 by PKR and is sufficient to block mRNA translation (61, 89).

VIRAL PERSISTENCE AND TRANSLATIONAL CONTROL

As is clear from the material presented above, a common theme in viral translation programming is the preferential selection for the translation of viral mRNAs at the expense of host mRNA translation. While such programming is often compatible with the life cycle of viruses that mediate rapid, lytic infection, such as influenza virus, poliovirus, and EMCV, it is not always compatible with viruses that mediate persistent infection. Little is known of the nature of viral translational programming as it pertains to persistent infection, although it clearly requires that host mRNA translation remain sufficient to sustain the host cell and support viral persistence. In this section, we will take a brief look at the relationship between translational programming and viral persistence. As described below, persistent infection may involve viral translational programming that impacts the host cell at several levels, including the control of translation, apoptosis, and cell growth-regulatory pathways.

Translational programming and maintenance of viral persistence.

Persistent infection requires the virus to ensure that the host cell remains translationally competent. This is important not only for the synthesis of viral proteins but also for the synthesis of cellular proteins and continued cellular viability. Analyses of the mechanisms by which viruses may mediate persistence and latency suggest that host cell integrity and translational competence are maintained through (i) viral modulation of specific cellular mRNA translation and (ii) viral modification of host signaling and translational regulatory pathways. As described above, the continued synthesis of ribosomal proteins during HSV-1 infection can be considered critical for maintaining viral persistence and latency. Interestingly, the level of ribosomal protein S6 phosphorylation is modulated in response to HSV-1 infection, thereby favoring the translation of host 5′ TOP RNAs, including those that encode the ribosomal proteins (164). Modification of S6 phosphorylation may thus facilitate translational competence, in part by ensuring that the synthesis of ribosomal proteins remain uncompromised during infection. The mechanisms by which HSV-1 induces the modification of S6 have not been determined. It is conceivable that viral infection results in activation of one or more cellular protein kinases that phosphorylate S6 and/or that a virus-encoded protein kinase may effect S6 phosphorylation.

In addition to maintaining host translational competence during persistent infection, viruses must ensure that the synthesis of their own proteins remains uncompromised. As presented above, many eukaryotic viruses encode mechanisms to repress the cellular PKR protein kinase and avoid the deleterious effect upon protein synthesis due to high levels of eIF2α phosphorylation (131). Thus, inhibition of PKR-dependent eIF2α phosphorylation can be seen as a mechanism to ensure overall translational competence during viral infection. Through the subsequent targeting of specific translational processes, including the cap-binding reaction, elongation, and termination, viruses can then manipulate host mRNA translation to the overall benefit of viral replication. However, the implications for viral disruption of PKR-dependent eIF2α phosphorylation are far-reaching and are not limited to the effects on mRNA translation. As presented above, repression of PKR provides the added advantage of avoiding, in large part, the antiviral effects of the host IFN response. Moreover, viral inhibition of eIF2α phosphorylation may disrupt critical host apoptotic and tumor suppressor pathways, which rely specifically on the ability to modify eIF2α activity (67, 68, 450).

Translational control, persistent infection, and regulation of host apoptosis.

What are the implications of blocking PKR function and preventing the phosphorylation of eIF2α during viral infection? Recent evidence suggests that inhibition of PKR function is a key feature in the establishment of persistent viral infection. Lau and colleagues (489) examined EMCV infection in established cell lines that were deficient in PKR expression or in control cells that expressed normal levels of PKR. EMCV normally mediates a cytolytic infection in vitro that is characterized by massive apoptosis of permissive cells. However, EMCV infection of PKR-deficient cells conferred persistent infection to this otherwise cytolytic virus. Although viral RNA translation was not directly examined in these studies, the results suggest that constitutive disruption of PKR-dependent eIF2α phosphorylation may facilitate the establishment of persistent viral infection. Such a relationship between viral control of PKR and establishment of persistent infection may prove important for HCV infection. HCV mediates persistent infection within a majority of infected individuals (6). Persistent HCV infection is strongly associated with the development of hepatocellular carcinoma and lymphoproliferative disorders (355, 356, 463). Recent observations now indicate that viral persistence and disruption of host apoptosis are linked to the block in eIF2α phosphorylation mediated by the HCV NS5A protein (134). Expression of NS5A in mammalian cells induces a block in eIF2α phosphorylation and concomitant stimulation of mRNA translation through NS5A-mediated repression of PKR. NS5A may contribute to HCV persistence by removing the translational blockade imposed by PKR-dependent eIF2α phosphorylation (Fig. 17). Consequently, however, constitutive expression of NS5A rendered cells refractory to PKR-dependent apoptosis (134). These results support previous studies indicating that eIF2α phosphorylation is an important component of cellular apoptotic signaling (436). The mechanisms by which eIF2α phosphorylation promote apoptosis are not well understood. Current thinking proposes that the phosphorylation of eIF2α is required to block the synthesis of antiapoptotic gene products during the apoptotic response (450). More recently, however, high-level PKR expression and PKR-dependent eIF2α phosphorylation have been associated with increased synthesis of proapoptotic effector proteins, including Bax and Fas (16, 95). Thus, eIF2α phosphorylation may result in the selective translation of specific proapoptotic mRNAs. Constitutive disruption of eIF2α phosphorylation that may occur during persistent HCV infection may therefore render the host cell refractory to apoptotic signaling. It is problematic for the host that disruption of eIF2α phosphorylation and the ensuing block in apoptotic signaling is associated with oncogenic transformation (16, 134).

Cell growth control, eIF2α phosphorylation, and oncogenic transformation.

Results from recent studies suggest that eIF2α phosphorylation may regulate cell growth, in part by enforcing a translational control and apoptosis checkpoint on cell proliferation. By this model, oncogenic potential is conferred to persistent viral infections in which the eIF2α checkpoint is targeted and constitutively disrupted. Thus, it is interesting that a wide range of tumorigenic viruses possess mechanisms to repress PKR-dependent eIF2α phosphorylation during infection (reviewed in reference 131). Recent work suggests that PKR activity and eIF2α phosphorylation are strictly regulated during the cell division cycle (495; M. Gale, Jr., C. Zhou, E. J. Firpo, M. G. Katze, A. Rudendsky, and B. R. Franza, Jr., submitted for publication) and that inhibition of PKR function results in perturbation of cell cycle control (16, 95). Our studies suggest that the NS5A protein from IFN-resistant HCV may confer an oncogenic potential to infected cells through the constitutive inhibition of PKR (134). It is fitting to speculate that inhibition of PKR function and disruption of eIF2α phosphorylation may contribute to the development of hepatocellular carcinoma in patients persistently infected with HCV.

MRNA TRANSLATION AS A TARGET FOR ANTIVIRAL THERAPY

As our understanding of viral mRNA translation increases, it is becoming quite clear that viruses often utilize rather unorthodox tactics to ensure their efficient mRNA translation. Such viral strategies that deviate from the processes of conventional cellular mRNA translation may represent potential targets for the therapeutic intervention in viral replication. The utility of targeting translation for the development of antimicrobial therapeutics has been successfully demonstrated through the use of antibiotics such as tetracycline and erythromycin, which target bacterium-specific elements of translation. The challenges to developing successful antiviral therapeutics that target viral translational programming remain in (i) identifying translational targets that are specific to the virus and (ii) ensuring that host translation remains selectively undisturbed by the therapeutic intervention. As is evident from the numerous examples cited in this review, viral translational programming intersects with nearly every aspect of cellular mRNA translation, thereby making these challenges all the more spectacular. However daunting a task, therapeutic intervention in viral mRNA translation remains a promising arena for the development of effective antiviral compounds, which in large part have been limited to viral polymerase or protease inhibitors.

Recent work has used antisense oligonucleotide strategies, dsRNA selection and targeting strategies, and direct mutagenesis of viral mRNA to demonstrate that effective viral translational targeting can be achieved. Here we briefly present some interesting highlights in this developing field. This section features strategies that are under development as potential therapeutics to combat the global HCV pandemic. The development of antiviral drugs for HCV infection is problematic, due to the high mutation rate of HCV and the enormous level of quasispeciation that occurs during infection (46, 92). Thus, one can expect that potential HCV therapeutics that target the viral polymerase, helicase, or protease activities may ultimately be of limited value, since drug-resistant strains will certainly emerge. Indeed, this has already been demonstrated for HIV, which exhibits a similarly high mutation rate during infection (92). The roles of the HCV 5′ and 3′ UTRs in viral mRNA translation and replication and the fact that these regions are highly conserved among the different HCV genotypes make them viable targets for the development of antiviral therapies. Thus, strategies which target the function of the HCV IRES and the translation-stimulatory activity of the viral 3′ UTR may represent efficient means of blocking HCV protein synthesis. Moreover, the development of compounds that may disrupt the ability of the NS5A protein to bind and repress PKR may serve to increase the efficacy of IFN and the existing IFN therapeutic regimens. For further information on therapeutic targeting of viral mRNA translation, the reader is referred to a review by Harford, and references therein (182).

IRES-mediated translation is not a common feature among cellular mRNAs, suggesting that it may represent a viable target for therapeutic intervention in viral mRNA translation. In support of this idea, Dasgupta and colleagues have identified a cellular RNA that possess an IRES-inhibitory function (467). This 60-nt RNA, termed IRNA, was isolated from S. cerevisiae and was first identified by examining the efficiency of poliovirus IRES-mediated translation in S. cerevisiae. These investigators found that poliovirus IRES translation was blocked in yeast and that this was due to a trans-acting factor that could similarly prevent poliovirus IRES translation when added to translationally competent HeLa extracts (79). Purified IRNA was shown to specifically inhibit IRES-mediated translation without having any effect on cap-dependent translation of cellular mRNAs. Interestingly, the IRNA was found to bind the cellular La protein (80), a major IRES-binding protein and effector of IRES-mediated translation (23). Evidence for other IRNA-polypeptide interactions was also demonstrated. These results suggested that the IRNA functions as an IRES-binding protein competitor to block translation from the poliovirus IRES. Subsequent studies revealed that expression of the IRNA in hepatoma cell lines similarly rendered a block to translation from the HCV IRES (81). Moreover, cells expressing the IRNA were refractory to infection with chimeric poliovirus under control of the HCV IRES. These results support the idea that the poliovirus and HCV IRES elements may bind a similar repertoire of cellular proteins. Structure-function analysis of the IRNA indicates that specific secondary structures are required for IRNA to bind cellular factors that promote IRES function (467). Thus, it appears that the IRNA may functionally mimic the IRES, at least in the context of secondary structure, and thereby compete for cellular proteins that mediate IRES translation. However, many questions remain to be addressed regarding the nature and origin of the IRNA itself. What is the cellular function of this RNA species, and are there homologous sequences present in the cells of higher eukaryotes? How may this IRNA control or interfere with cellular gene expression? Until such questions are answered, the potential therapeutic value of IRNA sequences will be significantly limited.

The targeted disruption of HCV genome translation has similarly been achieved using antisense oligonucleotides. Evaluation of the translation-inhibitory properties of a limited library of chemically modified oligonucleotides, directed to various regions of the HCV IRES and core protein-coding region, identified at least two antisense sequences that effectively inhibited HCV gene expression (175). In these studies, the expression of a truncated HCV genome in immortalized human hepatocytes was ablated by hybridization of an antisense oligonucleotide corresponding to a region encompassing the initiator AUG codon of the HCV core protein. Analysis of HCV RNA levels revealed that inhibition of genome expression was achieved without reducing genomic RNA expression. These results indicate that inhibition of HCV genome expression occurred at the level of translation and was not dependent on the activation of endogenous RNase H activity by duplex RNA. The translational block imposed by antisense oligonucleotides has yet to be confirmed by polyribosome analyses of HCV genome expression in the presence or absence of oligonucleotide treatment.

The possible utility of antisense oligonucleotides as an HCV antiviral therapy was demonstrated in a mouse model of HCV infection (498). This system utilized a recombinant vaccinia virus expressing an HCV 5′ UTR-core region construct fused to the firefly luciferase gene. Translation of this HCV-core–luciferase construct was under control of the authentic HCV IRES. Mice infected with this vaccinia virus recombinant exhibited a block in liver-specific luciferase activity when treated with antisense oligonucleotides directed to the core protein initiation codon and flanking sequences. Luciferase expression remained high in infected mice that received oligonucleotide controls. These results are subject to the following criticisms: (i) they used a heterologous virus system, which may not faithfully represent events of HCV infection, and (ii) the actual mechanism(s) contributing to inhibition of HCV (luciferase) expression was not determined. However, considering the enormous number of applications of antisense strategies and the efficiency with which antisense transcripts disrupt gene expression (2, 262), antisense targeting of HCV replication remains a viable means of developing anti-HCV therapies. Taken together, these results indicate that antisense oligonucleotides may provide a potent mechanism by which to target viral mRNA translation as an antiviral therapy.

Another possible target of antisense-oligonucleotide strategies may be found within the HCV 3′ UTR. As detailed above, this region of the HCV genome is highly conserved in all viral isolates, where it is thought to play an important a role in genome replication. It is appropriate to speculate that antisense oligonucleotides directed to within the HCV 3′ UTR may disrupt transcription and block the translation-stimulatory activity induced by PTB binding (218). Finally, it should be noted that the use of antisense-oligonucleotide strategy to target specific RNAs for degradation by the IFN-induced RNase, RNase L, has been achieved (317, 461). Although this strategy is not directly aimed at blocking viral mRNA translation, it presents a viable option for targeting the HCV RNA for specific degradation. In contrast to strategies that depend on RNA degradation by RNase H, a predominant nuclear protein, RNase L is a resident cytoplasmic enzyme and would be available to disrupt HCV replication, which also takes place in the cytoplasm (500). The possibilities for targeting the HCV 3′ UTR and the use of the RNase L pathway remain exciting areas of research into antisense oligonucleotides.

Similar to the use of antisense oligonucleotides, ribozymes, which are enzymatic RNA molecules that catalyze the cleavage of RNA, can be constructed to target specific RNA sequences (145). Ribozymes have been shown to be effective in blocking translation directed from the HCV IRES, although this probably occurs indirectly by RNA cleavage. Ribozymes constructed to target conserved sites within the viral 5′ UTR blocked the translation of a luciferase reporter protein under control of the HCV IRES and 5′ UTR in a tissue culture system, with little or no apparent toxicity to the host cell (402). Ribozymes have the added advantage of being efficiently packaged and expressed by various viral vectors, including vaccinia virus, adeno-associated virus, and various retroviruses. Moreover, ribozymes have been effective in catalyzing the degradation of both positive and negative strands of the HCV RNA (478). Strand-specific targeting of HCV RNA by ribozymes may thus hold promise for blocking the emergence of drug-resistant stains of HCV by eliminating new RNA variants.

Viral resistance to the current IFN-based therapeutic regimes for HCV infection is a major problem (121, 208, 263) and is in part responsible for the high frequency of persistent infections within the HCV-infected population. As described above, HCV resistance to IFN has been attributed in part to viral repression of the IFN-induced protein kinase PKR. The NS5A-PKR interaction may provide a useful therapeutic target for increasing the sensitivity to IFN in individuals who initially fail to respond to IFN therapy. Disruption of the NS5A-PKR interaction may restore eIF2α phosphorylation and the translational blockade imposed by PKR, thereby increasing the antiviral activity of IFN. Moreover, it is conceivable that restoration of PKR-dependent eIF2α phosphorylation may prove useful in reducing the incidence of cellular proliferative disorders associated with persistent HCV infection, since our results indicate that NS5A repression of PKR may provide an oncogenic potential to HCV (134).

CONCLUSIONS AND PERSPECTIVES

Investigations into the mechanisms and controls of viral protein synthesis have led the way to understanding the processes of cellular mRNA translation. Analyses of viral systems have introduced us to a better understanding of cellular antiviral pathways and signaling processes that affect mRNA translation. It is now clear that the translational control is an intimate part of most, if not all, metabolic processes of the cell. With this said, it now becomes important to understand the mechanisms of translational specificity; that is, how do cells and viruses impart translational control of specific mRNAs in a sea of translationally competent transcripts, and how do the extracellular environment and environmental cues effect mRNA translation?

Selective mRNA translation is a hallmark of many viral infections. While it is clear that viruses often encode mechanisms to disrupt cellular mRNA translation, the molecular mechanisms of selective viral mRNA translation under conditions of host protein synthesis shutoff remain poorly understood. In particular, the molecular mechanisms of IRES-mediated translation have yet to be elucidated. What are the basal factors that are required to support IRES-mediated translation? What is the actual role of IRES-binding proteins, and how do they function to stimulate viral mRNA translation? Similar questions extend to the mechanisms of cap-dependent selective translation. Identification of trans-acting factors that directly interact with viral RNA and understanding the implications of such interactions for viral translational programming remain the logical course of action. However, these approaches have been limited in scope by the nature of the traditional biochemical and molecular techniques often used when conducting such studies.

Recent advances in protein and nucleic acid analyses, such as combined mass spectrometry and sequence database searching (170), tandem mass spectrometry (171), and the current genomic technologies (44, 56, 236, 299, 407), make these techniques viable tools for investigating the mechanisms of translation control. The application of such techniques should allow for characterizing novel RNA-protein interaction. Genomic technologies, such as the use of high-density genome arrays, have already offered spectacular and sometimes surprising insights into the differential spectrum of cellular gene expression under various environmental conditions (88, 308, 350, 375, 408, 434). Similar applications to assess the array of both viral and cellular gene expression in virus-infected cells are proving equally interesting. In particular, high-density genome arrays can be used for the parallel identification of transcriptionally and translationally regulated mRNAs (150, 502). Recent studies have demonstrated the utility of this latter approach. Morris and colleagues (504) separated cellular mRNAs from resting or mitogenically activated fibroblast cultures into discrete pools based on the number of mRNA-bound ribosomes. By interrogating cDNA microarray filters with probes generated from the mRNA pools, these investigators found that translational control of cellular mRNA, at least in the context of mitogenic stimulation, was remarkably selective and represented less than 1% of the mRNAs in this cross-sectional analysis. Taking this application a step further, Sarnow and colleagues (234) interrogated cDNA microarrays with probes derived from polyribosome-associated mRNAs prepared from poliovirus-infected cells. This application allowed the investigators to identify cellular mRNAs that could be translated independent of a functional eIF4F complex. Remarkably, it was found that approximately 2 to 3% of the mRNAs analyzed were associated with polyribosomes under these conditions. When examined for their ability to direct the cap-independent translation of a bicistronic reporter protein, it was shown that at least a subset of these mRNAs contained functional IRES sequences within their 5′ UTR. An emerging theme from these analyses is that mRNAs, whose gene products have been implicated in a variety of stress responses, are translated with little or no requirement for eIF4F. Thus, IRES-mediated translation may be prevalent among mRNAs that are involved in acute cellular responses.

Development of an understanding of the acute cellular signaling pathways that modulate mRNA translation has now attracted the attention of those in the signal transduction field (43). It has become very clear that cellular mRNA translation is controlled in response to specific environmental signals that modulate development, cell proliferation, and apoptosis. Recent evidence now indicates that viruses impinge on these pathways during infection to promote viral replication. An important question now is how viral infection affects these pathways and how this may contribute to viral pathogenesis.

Determination of the three-dimensional structure of translation-regulatory proteins and translation factors represents another fruitful area for future research. Structural determination of translation factors is essential for a full understanding of the complexity and control of translation factor interactions and function. This was demonstrated by the recent elucidation of the crystal structure of eIF4E (318). As with eIF4E, translation factors play prominent roles in cell proliferation and malignancy (67, 267, 433). Understanding translation factor structure may pave the way for rational drug design of anticancer and antiviral therapeutic compounds.