Initiation of protein-primed picornavirus RNA synthesis

Abstract

Plus strand RNA viruses use different mechanisms to initiate the synthesis of their RNA chains. The Picornaviridae family constitutes a large group of plus strand RNA viruses that possess a small terminal protein (VPg) covalently linked to the 5’-end of their genomes. The RNA polymerases of these viruses use VPg as primer for both minus and plus strand RNA synthesis. In the first step of the initiation reaction the RNA polymerase links a UMP to the hydroxyl group of a tyrosine in VPg using as template a cis-replicating element (cre) positioned in different regions of the viral genome. In this review we will summarize what is known about the intiation reaction of protein-primed RNA synthesis by the RNA polymerases of the Picornaviridae. As an example we will use the RNA polymerase of poliovirus, the prototype of Picornaviridae. We will also discuss models of how these nucleotidylylated protein primers might be used, together with viral and cellular replication proteins and other cis-replicating RNA elements, during minus and plus strand RNA synthesis.

1. Introduction

RNA templated RNA synthesis is a central step in the life cycle of plus strand RNA viruses, which replicate and transcribe their genomes in the cytoplasm of the infected host cell. RNA replication is carried out on membranous structures by the viral RNA-dependent RNA polymerase, in conjunction with other viral and cellular proteins and cis-replicating RNA elements (cres). The genomes of plus strand RNA viruses are linear molecules that have to utilize special steps to retain the integrity of the 5’-end of their genomes. The very first stage in RNA replication involves initiation of minus and plus strand RNA synthesis, processes important to preserve the integrity of the ends of the RNA genome. Although different plus strand viruses use a variety of replication strategies there are only two primary mechanisms by which RNA synthesis can be initiated (Fig. 1) (162). The first is a de novo, or primer-independent, initiation in which the 3’-hydroxyl group of the starting nucleotide, usually a purine triphosphate, serves as the acceptor of the next nucleotide. This is followed by the addition of subsequent nucleotides yielding a full-length RNA chain (17, 69). The second is the so-called primer-dependent mechanism that employs either a free oligonucleotide (99) or a protein primer that provide the hydroxyl group for the formation of a phosphodiester bond with the first nucleotide (115).

Fig. 1.

Initiation mechanisms of RNA synthesis by plus strand RNA viruses. Two basic types of mechanisms for RNA synthesis by plus strand RNA virus RNA polymerases are shown (162). (A) Primer-independent (de novo) initiation. De novo initiation involves starting an RNA chain, usually with a purine nucleoside triphosphate, templated by a pyrimidine at the 3’-end of the template strand or at an internal site. (B) Primer-dependent initiation. The primer is either an oligonucleotide or a protein. A 3’-hydroxyl group of a nucleotide or the hydroxyl group of a tyrosine or serine residue of a peptide/protein provides the hydroxyl group for the formation of a phosphodiester bond with the first nucleotide.

Most plus strand RNA viruses that use de novo initiation for both replication and transcription, initiate synthesis at the 3’- or 5’-ends of the genomes. However, some viruses employ an internal site for the process of transcription (93, 105). The very first step in protein-primed initiation, the nucleotidylylation of the terminal protein, is usually templated, as in the case of picornaviruses (82, 111, 144). The template is contained within the loop of a small RNA hairpin that is located at different positions of the RNA genome, in different genera of Picornaviridae. The nucleotidylylation reaction can, however, also occur in a template-independent manner, for example with caliciviruses (54).

The most thoroughly studied family within the plus strand RNA viruses is the Picornaviridae. This family includes a large and diverse group of medically important human and animal pathogens such as poliovirus, rhinovirus, foot-and-mouth disease virus and hepatitis A virus. The family currently contains 26 genera of which the best known is the Enterovirus genus (PV: poliovirus) that also includes a large number of rhinoviruses, e.g. HRV2 and HRV14 (www.picornaviridae.com). Other important genera and member viruses include the Aphthovirus (FMDV: foot-and-mouth disease virus), Hepatovirus (HAV: hepatitis A virus) and Cardiovirus (EMCV: encephalomyocarditis virus). Although there are many different picornaviruses they all share several common properties. They possess a plus strand RNA genome (7.5-8 kb) with a small peptide, VPg linked to the 5’-end (Fig. 2). Their genome organization is similar with a long highly structured 5’ nontranslated region (NTR), a single large open reading frame (ORF) and a short 3’NTR, terminated with a poly(A) tail (20 nt to 150 nt long). The ORF is translated in the cytoplasm of the host cell into a polyprotein that contains one structural (P1) and two nonstructural domains (P2 and P3). The polyprotein is processed into precursor and mature proteins primarily by two viral proteinases, 2A^pro and 3CD^pro (Fig. 2). The “maturation cleavage” between VP4 and VP2 occurs by an autocatalytic process during maturation of the provirion (68). The role of the P2 proteins is primarily to induce the biochemical and biophysical changes that occur in the infected cell. The P3 proteins are more directly involved in the process of RNA synthesis.

Fig. 2.

Genome structure of poliovirus and processing of the polyprotein. The RNA genome of poliovirus contains a long 5’ NTR, a single large open reading frame, and a short 3’NTR, which is terminated by a poly(A) tail. At its 5’-end the RNA is covalently linked to a small peptide called VPg. The polyprotein made during translation of the RNA contains one structural (P1) and two nonstructural precursors (P2, P3). The polyprotein is processed into precursor and mature proteins by proteinases 2A^pro and 3C^pro/3CD^pro. The maturation cleavage of VP0 into VP2 and VP4 occurs by an autocatalytic mechanism.

It is generally accepted that picornavirus RNA replication proceeds by the following steps:

$$\begin{array}{cc}\hfill & \text{Plus}\text{strand}\text{RNA}----\to \text{minus}\text{strand}\text{RNA}----\to \text{replicative}\text{form}\left(\text{RF}\right)\hfill \\ \hfill & \text{RF}----\to \text{replicative}\text{intermediate}\left(\text{RI}\right)----\to \text{plus}\text{strand}\text{RNA}\hfill \end{array}$$

The RF consists of a double stranded structure, which has a specific infectivity 30 fold higher than ssRNA (20). It is not yet known how RF molecules initiate an infectious cycle but it is presumed that the 2 strands must be separated (74). Interestingly, the infectivity of the RF is dependent on a nuclear factor since enucleated HeLa cells cannot be transfected with PV RF (36, 74). Although the RF is normally considered to be a true intermediate during RNA replication, the possibility cannot be ruled out that it is a byproduct of rapid hybridization between plus and minus strands occurring during isolation. The RI, formed in small amounts during plus strand RNA synthesis, consists of a full-length minus-strand template strand with 6-8 nascent positive strands (9). It contains both double and single stranded regions as shown by their partial resistance to ribonuclease digestion (74, 171). The opposite polarity of RI, with a full-length positive strand RNA and incomplete nascent minus strands, has never been observed (21). It is important to note that since the picornavirus RNA genome contains two different types of termini the RNA polymerase has to be able to initiate RNA synthesis at different ends. The 5’-end consists of VPg linked to a heteropolymeric sequence that starts with two uridylates (4, 77, 130). At the 3’-end the RNA is terminated with a genetically encoded poly(A) tail (38).

In this review an attempt will be made to summarize what is known about the initiation of protein-primed RNA synthesis by the RNA polymerases of picornaviruses, which involves the uridylylation of terminal protein VPg. In addition, a model will be proposed to explain how the nucleotidylylated protein primer might be utilized for minus and plus strand RNA synthesis. It should be emphasized that these models are highly speculative and may change in the future. The review is organized such that the reader can easily find any particular topic of interest but unfortunately this has lead to some redundancy in the discussion of a few subjects, for which we apologize. It will concentrate on poliovirus, the prototype of Picornaviridae, simply because of the greater abundance of information with this virus. However, wherever important differences exist among members of the family, these will be discussed separately. Specifically, we will discuss the following topics: (i) viral and cellular factors involved in the initiation reaction; (ii) what is known about the initiation reaction and the “slide-back” model of VPgpUpU synthesis; (iii) models of protein-primed minus and plus strand RNA synthesis; (iv) unanswered questions about the initiation and elongation reactions of RNA synthesis. We regret that because of space limitations it will not be possible to discuss every publication that deals with this topic or to give credit to all of those investigators who have studied this intricate step in the life cycle of picornaviruses. However, we hope that this article will provide the reader with a summary of the accomplishments and unsolved questions of this topic.

2. Factors involved in the initiation of protein-primed RNA synthesis

2.1. RNA structures

A requirement for cis-acting RNA elements (cres) for RNA synthesis by picornavirus RNA polymerases was already suspected from early in vitro studies, which showed that purified poliovirus RNA polymerase was able to copy many different RNAs with a complementary oligonucleotide primer (46). However, in vivo the enzyme is highly specific for the copying only of its own RNA template suggesting that specific RNA structures must be present within the genome that provide specificity to this process. After years of extensive analyses it has become clear that at least 3 different cis-acting RNA elements are required for the initiation of picornavirus RNA replication. Here we will summarize what is known about these elements in enteroviruses, with an emphasis on poliovirus RNA.

2.1.1. The 5’-terminal cloverleaf

All picornaviruses contain complex RNA structures at the 5’-end of their RNA genomes. In enteroviruses and rhinoviruses, the 5’-terminal structure is similar to that of a cloverleaf (Fig. 3) (5, 6, 124) while with cardioviruses, aphthoviruses and hepatoviruses these structures are less well-defined (172). With all picornaviruses the 5’-terminal UMP of the plus strand cloverleaf is covalently linked to a hydroxyl group of a tyrosine in a small peptide called VPg (Fig. 3) (4, 77, 130).

Fig. 3.

RNA structures involved in poliovirus RNA replication. There are three RNA structures in poliovirus RNA that are required or important for RNA replication (section 2.1) (131). A cloverleaf-like structure is located at the very 5’-end of poliovirus RNA. It contains the terminal UMP that is linked to the hydroxyl group of tyrosine in VPg. The amino acid sequence of VPg is shown below and the residues important for RNA replication are shown in red. The cloverleaf binds viral protein 3CD^pro to stem-loop d and cellular protein PCBP2 to stem-loop b and a C-rich spacer. Viral protein 3AB also binds to stem loop b. The cre(2C) is a hairpin located in the coding sequence of 2C^ATPase. It is the template for the linkage of 2 UMPs to VPg during the initiation of RNA synthesis. The third RNA structure used during PV RNA replication is the 3’NTR, which contains two stem-loops, X and Y. These stem loops are involved in a “kissing” interaction (not shown). The binding of cellular poly(A) binding protein to the poly(A) tail is believed to be important for the circularization of the RNA genome.

The role of the plus strand PV cloverleaf in RNA replication was discovered by Andino and coworkers (5, 6). Subsequently, the binding of 3CD^pro to stem-loop d and of cellular proteins PCBP1/PCBP2 and of 3AB to stem-loop b was demonstrated and proposed to have a function in the initiation of minus strand RNA synthesis (22, 50, 51, 63). Poly(A) binding protein (PABP), which binds to the poly(A) tail, also interacts with 3CD^pro and is believed to facilitate the circularization of the RNA genome prior to minus strand RNA synthesis (15, 64). The observation that minus strand RNA synthesis requires the cloverleaf suggests that circularization of the genome is a necessary part of the replication process.

The sequence complementary to the plus strand cloverleaf forms a similar structure, the minus strand cloverleaf, which serves as the template for the initiation of plus strand RNA synthesis. Viral proteins 2C^ATPase and 2BC^ATPase were shown to bind specifically to the minus strand cloverleaf but the biological significance of this interaction has not yet been determined (10, 11). Similarly, specific binding of cellular protein hnRNPC1/C2 to the 3’-end of minus strand cloverleaf RNA was demonstrated (25). These proteins interact with 3CD^pro and the P2 and P3 precursor polypeptides and this interaction has been found to stimulate plus strand RNA synthesis (24, 25). The importance of the two As at the 3’end of the minus strand cloverleaf for efficient VPgpUpU-primed plus strand RNA synthesis in preinitiation replication complexes (section 3.5) was suggested by Sharma and coworkers (138).

It has been known for some time that the cloverleaf structure is a critical RNA element for both minus and plus strand RNA synthesis but determining the exact role of each cloverleaf separately has been difficult. To dissect the function of the cloverleaf, Vogt and Andino has used a novel construct of PV with dual cloverleaves in the 5’NTR, one of which regulated minus strand synthesis while the other mediated plus strand RNA synthesis (167). Using this construct they demonstrated that the entire 5’-plus strand cloverleaf, including intact stem-loops b and d for PCBP and 3CD^pro binding, were required for the initiation of plus strand RNA synthesis. Moreover, it was shown that specific sequences in stem loop a were required for this process but not for minus strand RNA synthesis.

2.1.2. The 3’NTR-poly(A)

The 3’NTRs of picornaviruses are highly diverse. Those of enteroviruses form a highly ordered structure with 2 stem-loops (X, Y) (Fig. 3), which are involved in a “kissing” interaction (91, 121). Other picornaviruses such a rhinoviruses contain only a single stem-loop in their 3’NTRs while the 3’NTRs of coxsackieviruses and echoviruses consist of three stem-loops (X, Y, Z) (120). The importance of the 3’NTR in RNA replication was first suggested by mutational analyses (91, 120, 129). However, subsequent studies have shown that this domain is interchangeable between PV, CVB4 and HRV14 (129). In addition, it was shown that the 3’NTRs of poliovirus, HRV14 and CVB3 are not required for infectivity (23, 156, 163). In contrast those of FMDV and EMCV are essential for viral viability (40, 132). Deletion of the Z domain of CVB3 had no significant effect on growth in tissue culture but in mice this mutant exhibited reduced viral pathogenesis (92).

Numerous studies have attempted to find specific binding of viral or cellular proteins to the 3’NTR of picornavirus RNAs. Both PV 3AB and 3CD^pro were shown to specifically interact with the PV 3’NTR (63). The RNA polymerase of EMCV binds to its cognate 3’NTR while 3AB and 3ABC of HAV interact with the 3’NTR (33, 76).

The poly(A) tail of picornaviruses, first identified by Yogo and Wimmer, is variable in length (180). That of poliovirus is reported to be on average between 50 and 90 nucleotides long (70, 180). It is genetically encoded, that is, it is transcribed from poly(U) in the minus strand (38) and that seems to be true for all picornaviruses. It was recently reported for poliovirus that 3D^pol polyadenylylates viral RNA during RNA replication using a reiterative transcription mechanism (70, 147). Previous evidence indicated that shortening the poly(A) tail reduces the infectivity of genomic RNA (143). The poly(A) tail was also proposed to play a role in the formation of a circularized RNP complex, which is stabilized by protein/protein interactions (64). In the in vitro translation/RNA replication system a poly(A)_13-20 RNA was required to achieve wt-like minus strand RNA synthesis (141). This is also about the same length necessary for human PABP to interact with poly(A) in vitro (64).

2.1.3. The cre elements of picornaviruses

For many years only the 5’ and 3’NTRs of picornaviral genomes were considered as ample sources of cis-replicating RNA elements for RNA replication. The first indication of a cis-replicating RNA element in the coding sequences of a picornavirus came from the finding that deletion of the P1 coding region of HRV14 resulted in the replicons’ inability to replicate (90). A hairpin structure in VP1 was identified that was absolutely required for genome replication. This genetic element was named as a cis-acting replication element, cre. Subsequently, Goodfellow and coworkers reported the presence of an essential RNA hairpin, cre(2C), in the coding sequence of poliovirus 2C^ATPase ( Fig. 3) (55). The biological function of cre(2C) was demonstrated both by biochemical and genetic experiments to be the template for the uridylylation of VPg, generating the primer (VPgpUpU) for RNA synthesis (113, 114, 123). Since then cre elements with the same function have been discovered or predicted to exist in the coding sequences of various picornaviruses. These structures are similar in the different genera of Picornaviridae although the location of these elements varies and overall their nucleotide sequences are highly divergent (Figs. 4A, 4B): cres of PV and CVB3 (Enterovirus) in 2C^ATPase (55, 164); cres of HRV14, HRV3 and HRV76 (Enterovirus) in VP1 (89, 90); cres of HRV2, HRV1a, HRV16 (Enterovirus) (52, 89); cre of species C Rhinovirus (Enterovirus) in VP2 (30); cre of TMEV and Mengovirus (Cardiovirus) in VP2 (83); cre of HAV (Hepatovirus) in 3D^pol (178); cre of HPeV (Parechovirus) in VP0 (2); cre of Porcine Sapelovirus (Sapelovirus) in 2C^ATPase (142). The only known exception is FMDV (Aphthovirus) whose cre is located in the 5’NTR rather than the coding sequences (88).

Fig. 4.

Secondary structures of the picornavirus cre elements and their locations in the genome. (A) A simplified version of the enterovirus and rhinovirus RNA genome is shown on top. Open loops of the PV1, HRV2 and HRV14 cre elements (section 2.1.3) are shown in detail with G₁ and A₁₄ shown in blue and A₅ and A₆ that template VPgpUpU synthesis are shown in red. Only the upper part of the stem is shown. (B) A simplified version of the cardiovirus and aphthovirus RNA genomes is shown on top. The hepatovirus genome is the same except it lacks the L protein. The loops of the EMCV, FMDV, and HAV cre elements (section 2.1.3) are shown in detail and A₅ and A₆ that template VPgpUpU synthesis are shown in red.

The diversity of the location of cre elements in different picornavirus genomes is astounding since, as we will show below, their basic function is identical. Since cre elements function strictly by their RNA structure that, as sequence, has also to accommodate the coding activity of the viral mRNA, the location may be determined by the best “fit”into messenger function of the viral open reading frame. Equally surprising, the location of the functioning cre elements by genetic engineering can be changed within a specific genome. It has been shown for several viruses (PV, FMDV and HRV14) that the location of the cre elements can be changed within the genomes (55, 88, 178, 179). In this case, the endogenous cre function can be completely inactivated by mutation without loss of viral proliferation. In poliovirus, a synthetic PV-specific cre element can even be moved to the 5’NTR (into spacer I between the cloverleaf and IRES) (179). In this construct the inactivation of the cre(2C) by mutation has very little influence on the replication in HeLa cells although, remarkably, its neurovirulence in CD155 tg mice was lowered five orders of magnitude (34, 160). It should be stressed, however, that the location of the cre is conserved within each particular picornavirus species (30).

The cre elements of human enteroviruses and rhinoviruses are characterized by a loop that is 14 nt long and contains a conserved motif: R₁NNNA₅A₆RNNNNNNR₁₄ (Fig. 4) (176, 179). The first base, the 7^th base and the 14^th base are all purines (R) and the As at positions 5 and 6 in the loop are involved in templating the two Us during the synthesis of VPgpU/VPgpUpU (52, 100, 123, 164, 176, 178). The other genera of Picornaviridae where cre elements have been identified possess cre structures in which the length of the loop is more variable (15-23 nt) except for the core sequence of AAA/GC. The double stranded stem of the cres with internal bulges and loops are variable in length and in sequence. The minimal functional HRV14 cre(VP1) consists of the 14 nt loop and a 9 bp stem (154, 176). With PV cre(2C) the minimal structure that retained activity in vitro was about 26-29 nt long, including the 14 nt loop (108, 179).

2.2. Viral and cellular proteins

In the simplest in vitro reaction on a poly(A) template only two purified poliovirus proteins (3D^poland VPg) are required for the synthesis of the nucleotidylylated protein primer, VPgpU/VPgpUpU, while on the cre(2C) template the optimal reaction also requires 3CD^pro (section 3.2). However, in preinitiation replication complexes (PIRCs) (section 3.5), the initiation reaction appears to be much more complicated because VPg-uridylylation requires the cloverleaf (85), the ATPase activity of protein 2C^ATPase (13, 14) and membranes (47). This section of the review will discuss briefly what is known about the role of the viral and cellular proteins that are involved in the initiation reaction.

2.2.1. RNA polymerase 3D^pol

The central player in RNA synthesis is 3D^pol, the RNA-dependent RNA polymerase, which catalyzes both the VPg-uridylylation and RNA chain elongation reactions (46, 115). Three dimensional structures have been reported for several members of the virus family such as PV, HRV14, CVB3, FMDV, EV-71 (7, 27, 43, 44, 59, 61). The 3D^pol of Picornaviridae are 460-470 amino acids long and possess high primary sequence similarity and structural homology (26). They adopt a fold resembling a cupped right hand with fingers, thumb and palm domains and a strong interaction between the fingers and thumb domains accounts for the closed protein structure. The palm domain is the most conserved where two aspartic acid coordinate two metal ions essential for catalysis. The roles of the fingers and thumb domains are in primer, NTP and template binding. In vitro the protein oligomerizes through an interaction between the palm and thumb domains, which enhances the RNA binding and elongation acitivities of the polypeptide (65, 107). Although in infected cells most of the protein is in a soluble form, a fraction of it is anchored to the membranous replication complexes, possibly via an interaction with the 3B domain of membrane protein 3AB (66, 173). Recently, it was proposed that PI4P (phosphatidylinositol 4-phosphate) lipids, produced by cellular protein PI4KIIIβ (phosphatidylinositol-4 kinase III beta), recruit 3D^pol to the membranes for coxsackie virus B3 (CVB3) RNA replication (67). However, CVB3 mutants that, interestingly, can bypass the requirement for PI4KIIIβ have been described (161), which suggests that the current model for enterovirus genome replication in infected tissue culture cells may have to be modified (67).

2.2.2. Terminal protein VPg and its precursors

A 5’-terminal viral protein VPg that is covalently bound to the genome of poliovirus was first identified by Wimmer and his colleagues (77). They are small, basic peptides 19-26 amino acids in length (Table 1) (148). They contain a fully conserved tyrosine at position 3 of the VPg amino acid sequence, the attachment site to the RNA. With the exception of EMCV, picornavirus VPgs also contain a conserved glycine at position 5 (Table 1). It should be noted that the conserved spacing of Y3 and G5, downstream of the 3A/3B cleavage site, can be potentially used in the identification of VPg sequences in newly identified picornaviral polyproteins. Other residues required in PV VPg for viral growth are the glycine at position 5, two lysines at positions 9 and 10 and an arginine at position 17 (Fig. 3) (53, 75, 113, 122). All of these residues are important for an interaction of VPg with 3D^pol of PV when assayed in a yeast two-hybrid system (113). Interestingly, enterovirus VPgs contain 3 prolines at positions, 7, 11 and 14. The structure in solution of PV VPg in the presence of an organic solvent, TMAO, was determined by NMR (136). The structure consists of a long loop (residues 1-14) and a short C-terminal helix. An NMR structure of chemically synthesized VPgpU was more stable than that of free VPg and exhibited a defined globular structure (135). FMDV is unique in that it encodes 3 distinct VPg sequences all of which function as substrates in vitro for uridylylation and all of them can be observed attached to viral RNAs (71, 100). Interestingly, the VPg of PV could be replaced by that of HRV14 both in the in vitro uridylylation assay and in vivo, yielding a viable virus (113) that acquired a mutation (L11P) in HRV14 VPg after passages. Similarly, the VPg of HRV16 was also compatible with the growth of PV but HRV16 replication did not tolerate the VPg of PV (28).

Table 1.

VPg sequences of Picornaviridae

Genus	Virus	VPg amino acid sequence*
Enterovirus
	PV1(M)	GAYTGLPNKKPNVPTIRTAKVQ
	PV2s	GAYTGLPNKRPNVPTIRTAKVQ
	PV3s	GAYTGLPNKRPNVPTIRAAKVQ
	CAV21	GAYTGLPNKKPNVPTIRIAKVQ
	CVB3	GAYTGVPNQKPRVPTLRQAKVQ
	EV71	GAYSGAPNQVLKKPVLRTATVQ
	Echo6	GAYTGMPNQKPKVPTLRQAKVQ
Rhinovirus
	HRV14	GPYSGNPPHNKLKAPTLRPVVVQ
	HRV2	GPYSGEPKPKTKIPERRVVTQ
Aphthovirus
	FMDV1	GPYAGPLERQKPLKVRAKLPQQE
	FMDV2	GPYAGPMERQKPLKVKAKAPVVKE
	FMDV3	GPYEGPVKKPVALKVKAKNLIVTE
Cardiovirus
	EMCV	GPYNETARVKPKTLQLLDIQ
Hepatovirus
	HAV	GVYHGVTKPKQVIKLDADPVESQ
Sapelovirus
	SPV	GAYTGAPKPETRKPVLRKAVVQ

^*The conserved Tyrosine at position 3 is shown in red and bold.

There are 5 precursors of VPg (3AB, 3ABC, 3BC, 3BCD and P3) and it is not yet known whether in vivo VPg itself is uridylylated by 3D^pol or one of its precursors. 3AB, a small basic protein (12 kDa) has multiple functions in vitro. It is a nonspecific RNA binding protein but when complexed with 3CD^pro it specifically interacts with the 5’ cloverleaf of PV RNA and with the 3’NTR (63, 174, 175). In addition it possesses in vitro chaperone and helix destabilizing activities (35). Experiments to uridylylate 3AB in vitro, however, have failed (49). Interestingly, mutations in 3AB that cause a severe RNA replication defect can be complemented in trans only by P3 but not by the mature 3AB polypeptide (157).

Polypeptide 3BC is normally not observed in PV-infected cells but it functions efficiently as substrate for 3D^pol in in vitro VPg-uridylylation reactions on a cre(2C) template and it can also replace the function of 3CD^pro in the reaction (110). The larger 3B-containing precursors, 3BCD and P3, can be observed in small amounts as uncleaved polypeptides in PV-infected cells. 3BCD can also be uridylylated by 3D^pol in vitro though less efficiently than either VPg or 3BC (110). A 3BC cleavage site mutant replicated in vivo and produced 3BC-linked RNA (104). Infectious virus, however, was not produced. A pseudorevertant of this virus, which partially restored 3BC cleavage, also restored virus infectivity.

2.2.3. Proteases 3C^pro and 3CD^pro

Protein 3C^pro is a Cys-reactive proteinase with a chymotrypsin like structure (97), which is also an important RNA-binding protein during RNA replication (108, 177, 179). 3C^pro and its precursor 3CD^pro are responsible for the processing of most of the polyprotein at Q/G cleavage sites and they also affect several cellular functions (39, 87). The crystal structures of the 3C^pro polypeptide have been determined for HRV14, PV, HRV2, HAV and FMDV (87). 3CD^pro is the precursor of both proteinase 3C^pro and of RNA polymerase 3D^pol. Although, it possesses proteinase activity it totally lacks RNA polymerase activity. As noted above (section 2.1.1), as an RNA-binding protein 3CD^pro interacts with the 5’ cloverleaf in a complex with PCBP2 or 3AB, with the cre(2C) RNA element in a complex with VPg and 3D^pol, and with the 3’NTR in a complex with 3AB (50, 63, 108, 123). The RNA-binding activity of 3CD^pro resides in the 3C^pro domain but the activity is modulated by the 3D^pol domain of the polypeptide (31). The crystal structure of 3CD^pro revealed that the 3C/3D cleavage site in the polypeptide is on the opposite side from the catalytic site of the 3C^pro protein (86). The RNA binding activity of 3CD^pro appears to be also important for virus maturation (48).

2.2.4. 2C^ATPase

The 329 amino acid long 2C^ATPase is a complex multifunctional polypeptide, which contains an important ATP binding domain and possesses ATPase activity in vitro that is inhibited by guanidine hydrochloride (94, 119, 126). It was classified as a member of superfamily III helicases with 3 conserved motifs two of which are related to ATP binding (57). However, so far no helicase activity has been detected in in vitro assays with the purified protein (119). The N-terminal domain of 2C^ATPase, which anchors the protein to membranes, contains an amphipathic helix, an oligomerization domain and an RNA-binding domain (1, 41, 112, 125). Near the C-terminus there is a Zn⁺⁺ binding domain and another amphipathic helix that is likely involved in membrane binding (118, 152). Genetic studies have implicated 2C^ATPase in numerous functions in the viral life cycle, such as uncoating (78), host cell membrane rearrangements (3, 29, 152), RNA binding (125), RNA replication (14, 79, 112, 153), and encapsidation (81, 165, 169).

2.2.5. Poly(rC) binding protein 2: PCBP2

PCBP2 is a cellular RNA binding protein belonging to a family of proteins that contain K-homologous domains. It recognizes and binds poly(rC) residues and interacts specifically (i) with the IRES of PV to enhance translation and (ii) with the cloverleaf and a C-rich sequence (see below) to facilitate minus strand RNA synthesis (16, 22, 50, 51, 106, 168). During PV infection PCBP2 is cleaved by proteinase 3C^pro/3CD^pro (117). Mammalian cell extracts depleted of PCBP2 can rescue PV translation and RNA replication. PCBP1, another member of the PCBP group, also binds to the 5’NTR of PV but with less affinity than PCBP2 (131). PCBP1 can rescue only RNA replication in extracts depleted of this protein.

Recent studies have shed new light to the binding of PCBP2 to the PV cloverleaf (158, 181). It was shown that PCBP2 needs two cytosine-rich clusters just downstream of the cloverleaf (mapping to spacer between the cloverleaf and the IRES) in order to form a PCBP2 /cloverleaf complex for function in RNA replication (158).

2.2.6. Heterogeneous nuclear ribonucleoprotein C: hnRNP C

Cellular protein hnRNPC is a primarily nuclear polypeptide that exists in human cells in two spliced isoforms C1 and C2 (25). It forms heterotetramers consisting of 3 C1 molecules and one C2 molecule, which bind RNA (58). Previous studies have identified RNP complexes between hnRNP C and the 3’-end of PV minus strand RNA (127). The protein also interacts with 3CD^pro and with the P2 and P3 precursor polypeptides (24). More recently Ertel and coworkers have demonstrated that hnRNP C also binds to the 5’-end of PV minus strand RNA (42). In addition, it was shown that depletion of hnRNP C in cell culture results in reduced plus strand RNA synthesis and virus yields. These authors proposed that the role of hnRNP C in PV replication is to bind to the 5’- and 3’-ends of the genome and to stabilize the interactions between them. This would facilitate the initiation of plus strand RNA synthesis at the 3’-end of the minus strand.

2.2.7. Poly(A) binding protein: PABP

A role for PABP1 in poliovirus RNA replication was first proposed by Herold and Andino (64). These authors have observed that PABP interacts in vitro with both the poly(A) tail of PV RNA and also with proteins PCBP and 3CD^pro. It was suggested that these interactions are required for the circularization of the genome and initiation of poliovirus minus strand RNA synthesis. Using preinitiation replication complexes Silvestri and coworkers have determined the minimum length of the poly(A) tail on PV RNA as 12 nt in length for optimal binding of PABP, for VPg uridylylation, and for negative strand RNA synthesis (141).

3. Initiation of protein-primed picornavirus RNA synthesis

Prior to our studies on the intitation of poliovirus RNA synthesis, DNA synthesis by a protein-primed mechanism was already demonstrated for a number of viruses. These viruses all possess linear double-stranded DNA genomes such as adenovirus, phages Φ29, PRD1 and GA-1 (133, 134). The DNA polymerases of these viruses use as template the 3’-terminus of their DNA strands for the nucleotidylylation of their terminal or preterminal proteins. In addition, it was known that the reverse transcriptase of hepatitis B virus uses a nucleotidylylated protein-primer for cDNA synthesis (137). With RNA viruses protein-primed RNA synthesis was first demonstrated in vitro with a double stranded RNA virus (infectious necrosis virus) whose RNA polymerase itself becomes guanylylated and serves as primer for RNA synthesis (37).

3-1. Discovery of the in vitro VPg uridylylation reaction

The role of VPg as a primer for poliovirus RNA synthesis was already proposed more than thirty years ago (170) shortly after the discovery of VPg covalently linked to both minus and plus strands of poliovirus RNA (4, 77, 130). This model was consistent with the absolute dependence of RNA synthesis by the RNA polymerase on a primer (46). Subsequently, the presence of free VPg and of VPgpUpU in poliovirus-infected cells was demonstrated (32). In addition, the synthesis of VPgpU/VPgpUpU in crude replication complexes was observed and the elongation of these precursors into plus strand RNAs could also be demonstrated (149, 150, 159). Although these results strongly supported the possibility that VPg is the primer for RNA synthesis direct evidence for this model remained elusive. Moreover, alternative models of primed RNA synthesis in poliovirus RNA replication were proposed: amongst them initiation by elongation of a snap-back oligo(U) tail that the 3D^pol polymerized onto the 3’ poly(A) of plus strands (155).

Our studies with the in vitro synthesis of VPgpU/VPgpUpU started when the late Jacques van Boom of Leiden University sent us a sample of chemically synthetized VPgpU. We first tried to elongate it to VPgpUpU on a poly(A) template with purified PV 3D^pol and ³²P-UTP, under conditions normally used for the elongation of oligonucleotides on RNA templates. These experiments, however, were unsuccessful. After testing various reaction conditions to achieve an elongation reaction we accidentally discovered that the only change needed was an increase in 3D^pol concentration from nanomolar to micromolar concentrations. As a control, we also tested synthetic VPg under the same reaction conditions and observed ³²P-labeling of VPg to yield products migrating at the expected positions of VPgpU, VPgpUpU and VPg-poly(U) on polyacrylamide gels (115). This critical experiment proved that the viral RNA polymerase was able to uridylylate VPg and then elongate it by transcription of the poly(A). In fact, uridylylation itself was critically dependent on a poly(A) template (115).

Although the in vitro reaction on a poly(A) template could in principle explain the initiation of minus strand RNA synthesis at the 3’-end of genomic RNA the yield was poor even with full length PV transcript RNAs terminated by poly(A). More importantly, considering that in the infected cell mRNA molecules carrying poly(A) tails are abundant, the uridylylation of VPg did not account for the observed specificity in vivo. Surprisingly, transcripts from which the poly(A) tail was removed worked equally well suggesting that the real template for the uridylylation reaction was somewhere else in the genome. The most likely location of the template sequence was in the 5’ or 3’NTR but there was no detectable VPg-linked product made in vitro with transcripts of these RNA sequences/structures. At about the same time there was a report of an RNA hairpin in the 2C^ATPase coding sequence, called cre(2C), that is required for RNA replication (55). To our surprise transcripts of this hairpin yielded the expected VPgpU/VPgpUpU products in the in vitro reaction suggesting the possibility that the unknown template sequence resided in this RNA structure (55, 114). Since the yield of the product was still relatively low we tried to stimulate the synthesis of the VPg-linked products with various purified viral and cellular proteins that were available in our lab. These experiments led to the discovery that 3CD^pro stimulates the in vitro reaction about 100 fold (114). Subsequent mutational analyses of the cre(2C) led us to suggest that 2 consecutive A residues, A₅ and A₆ in a conserved motif (A₅A₆A₇C₈A₉) located in the loop of cre(2C), are of critical importance to cre function (Figs. 4A and 5) (55, 113, 114, 123).

Fig. 5.

“Slide-back” model for VPgpUpU synthesis during initiation of RNA synthesis. The upper stem of cre(2C) interacts with 2 molecules of 3CD^pro (or 3C^pro) (section 3.7). 3CD^pro binds VPg with the back side of its 3D^pol domain, where another molecule of 3D^pol links UMP to the hydroxyl group of tyrosine in VPg. A₅ in the loop of cre(2C) is the template for the linkage of the first UMP (U₁) to VPg yielding VPgpU₁ . VPgpU₁ “slides-back” to hydrogen bond with A₆ and the second UMP (U₂) is templated again by the A₅ nucleotide during the elongation step yielding VPgpU₁U₂. The nucleotides involved in the “slide back” are boxed. Nucleotides A₅ and A₆ of the conserved motif are shown in bold.

3.2. Characterization of the in vitro VPg uridylylation reaction catalyzed by PV _3Dpol

When purified PV RNA polymerase 3D^pol is incubated with a PV RNA template or with other 3’-adenylylated RNA templates, NTPs, Mg⁺⁺ and an oligo(U) primer, full length minus strand RNA copies can be synthesized (12). Poly(A) also serves as a template for the uridylylation of VPg in vitro by 3D^pol to yield VPgpU and VPgpUpU (115). The final product is VPg-poly(U), the 5’-end of minus strand RNA:

$$\text{Poly}\left(\mathrm{A}\right)-\text{RNA}+\text{VPg}+{\text{Mg}}^{++}+\text{UTP}+3{\text{D}}^{\text{pol}}---\to \text{VPgpU}+\text{VPgpUpU}---\to \text{VPgpoly}\left(\mathrm{U}\right)$$

The poly(A) templated in vitro reaction can be enhanced 100 fold by replacing Mg⁺⁺ by Mn⁺⁺, which, however, is known to decrease the specificity of RNA polymerases (8, 113).

When full length PV RNA is used in the in vitro reaction, instead of poly(A)-RNA, the template for the synthesis of VPgpU and VPgpUpU is the cre(2C) hairpin (114), located in the coding sequence of protein 2C^ATPase (55). Protein 3CD^pro strongly stimulates the cre(2C) templated reaction and this reaction is equally active with Mg⁺⁺ and Mn⁺⁺:

$$cre\left(2\mathrm{C}\right)\text{RNA}+{\text{Mg}}^{++}+\text{UTP}+3{\text{CD}}^{\text{pro}}+3{\text{D}}^{\text{pol}}---\to \text{VPgPU}+\text{VPgpUpU}$$

Protein 3CD^pro can be replaced by 3C^pro, 3BC or by 3BCD (108, 110, 179). It was proposed that the 3C^pro domain of 3CD^pro determines the specificity of the reaction while the 3D^pol domain enhances affinity (108). The cre RNAs of HRV14 and to a small extent that of HRV2 can also serve as template in the uridylylation reaction of PV VPg by PV 3D^pol and 3CD^pro (52, 114). The template nucleotide in vitro is A₅ while A₆ provides specificity to the reaction, and A₇ only serves to enhance the uridylylation reaction (113, 116, 123). The elongation of VPgpU on cre(2C) RNA is aborted after the addition of the second U most likely because of structural restrictions imposed by the uridylylation site of the enzyme. Interestingly, an exact spacing of A₅ and A₆ within the loop of cre(2C) is required for optimal uridylylation of VPg in vitro (Fig. 4A). The optimal size of the cre(2C) loop is 14 nt (Fig. 4A) in accordance to previous findings with HRV14 cre(VP1) (176, 179). Mutations that lead to base pairing with A₅ and A₆ within the loop abolished cre function in the reaction (179). In addition to A₅ and A₆ in the loop A₁ and G₁₄ at the bottom of the loop are also required for efficient production of the VPg-linked precursors (179). Transversions such as G₁C or G₁U or A₁₄C are highly detrimental to uridylylation activity while transitions (G₁A or A₁₄G) are better tolerated at these nucleotide positions. Studies analyzing the optimal length of the cre(2C) element revealed that the upper stem and the loop with a total length of about 26 nt carry the minimal structural requirements for a uridylylation complex formation (108, 179). The upper stem provides the specificity for protein binding and the lower stem plays a structural role (110).

A mutational analysis of PV VPg identified several amino acids that are critical or important for VPg uridylylation by PV 3D^pol in vitro either on a poly(A) or a cre(2C) template (Fig. 3) (113). The most important residues in VPg are Y3, the attachment site of the peptide to genomic RNA, G5, K9K10 and R17. The same residues were found to be important for uridylylation of HRV2 VPg by its cognate 3D^pol (53). PV 3D^pol is able to use as substrate, in addition to its own VPg, the VPg of HRV14 but the yield of product is reduced. In contrast the VPgs of HRV2 and HRV89 had no detectable substrate activities in the in vitro assay. The 3D^pol of HRV2 on the other hand was able to use all four viral VPgs (PV, HRV14, HRV2 and HRV89) as substrates with nearly equal efficiency (53).

To demonstrate that the VPg-uridylylation is a common mechanism used by picornaviral RNA polymerases during the initiation of RNA synthesis we also examined the reactions catalyzed by the RNA polymerase of HRV2. We have demonstrated that the purified 3D^pol of HRV2 catalyzes the same type of synthetic reactions with either poly(A) or a cre RNA template, located in the coding sequence of 2A^pro template [cre(2A)] (52, 53). Subsequent in vitro studies with purified RNA polymerases of other picornaviruses have confirned these results. The RNA polymerase of HRV14 uridylylates VPg on a cre located in the VP1 capsid protein coding sequence (176), which was first identified by genetic experiments by McKnight and Lemon (90). In contrast the FMDV RNA polymerase uses a cre element derived from the 5’NTR for the in vitro uridylylation of all of its three distinct VPgs, 3B₁, 3B₂, 3B₃ (88, 100). The reaction is enhanced not only by FMDV 3CD^pro but also to a lesser extent by 3C^pro and precursors 3B₃3C and 3B₁₂₃3C. The residues within 3C^pro involved in the interaction with the cre have been identified and are located on the face of the polypeptide opposite of the catalytic site (101). In contrast to the polymerases of PV and HRV2 the FMDV enzyme does not use poly(A) as template in vitro either in the presence of Mg⁺⁺ or Mn⁺⁺ (100).

3.3. Correlation of VPg uridylylation in vitro with viral growth phenotypes in vivo

Complete mutational analyses of the A₅ and A₆ positions of the PV and CVB3 cre(2C) loops, and the HRV2 cre(2A) loop, yielding either lethal or quasi-infectious progeny, fully correlated with the inability of such mutants to uridylylate VPg in vitro (53, 123, 164). Substitution of A₇ with pyrimidines resulted in reduced synthesis of VPg-linked products in vitro and to a moderately defective growth in vivo. We have confirmed these results with a dual cre virus in which the endogenous cre was fully inactivated by an A₅C mutation and an extra wt copy of the cre was inserted between P1 and P2 (179). The results with this virus indicated that the positively charged lysine encoded by the A₅A₆A₇ sequence in the 2C^ATPase protein could be functionally fully replaced by either Gln or Glu and to a smaller extent by Asn (116). Therefore the lethal growth phenotype of an A₅C mutant is not related to 2C^ATPase protein function but rather to cre function.

A subsequent mutational analysis of the HRV14 cre(VP1) was also in full agreement with our results (176). In addition, it pointed out the functional importance of purines at positions 1 and 14 of the enteroviral and rhinoviral cre elements, forming a non-Watson-Crick closing pair at the base of the loop, which is possibly contributing to the structure of the cre loop.

FMDV is so far the only exception among picornaviruses in that its cre is not located in the ORF but rather in the 5’NTR, contiguous with domain I of the IRES (88). Mutations introduced into this hairpin resulted in reduced RNA replication but did not influence viral mRNA translation. The position of the cre in the 5’NTR was not important since functional replicons and viruses could be obtained if a wt cre was added to the genome following the 3D^pol coding sequences. Genetic experiments with the FMDV cre have shown that the cre can function in “trans” and for this reason it has been suggested that the name cre be replaced by “bus” (3B-uridylylation site) (100).

Mutations in PV VPg or the HRV2 VPg, which reduced uridylylation in vitro were found to abolish or reduce viral growth (53, 113). A chimeric PV with the VPg of HRV14 was viable but substitutions with the VPgs of HRV2 or HRV89 were lethal (113).

3.5. VPg-uridylylation in vivo and in preinitiation replication complexes

As noted above (section 3.2), the first indication that VPg might serve as a protein primer for the RNA polymerase was that VPgpUpU was identified in PV-infected HeLa cells (32). In addition, it was observed that in crude replication complexes isolated from PV-infected cells, VPg was uridylylated in the presence of UTP to yield VPgpU and VPgpUpU (149, 150, 159). These precursors could be chased into VPg-linked UUAAAACAG, the first nine nucleotides of plus strand RNA and into full length plus strand RNAs. The treatment of such crude replication complexes with detergent abolished the synthesis of VPgpUpU suggesting that the initiation reaction is dependent on the integrity of membranes (47, 150, 159). This is in contrast to the in vitro reaction where 3D^pol is able to uridylylate VPg in the absence of membranes (53, 114, 115). It has been suggested that in vivo membranes might enhance the concentration of ingredients that is required for the uridylylation reaction (107, 111).

In addition to isolating replication complexes from infected cells they can be also made in vitro in a translation and RNA replication system in a HeLa cell free extract, which produces viable poliovirus (95). Preinitiation replication complexes (PIRCs), formed in the presence of GnHCl, an inhibitor of RNA replication, are isolated and used for studies of poliovirus RNA replication (13, 14, 85). Upon removal of GnHCl the PIRCs immediately initiate both cre-dependent VPg uridylylation and minus strand RNA synthesis. Both VPg uridylylation and minus strand synthesis were found to require the GnHCl-inhibited activity of protein 2C^ATPase (13, 14). It was suggested that PV RNA functions coordinately as a template for both RNA replication and cre-dependent VPgpUpU synthesis (85, 145). Whether or not 2C^ATPase has a function directly in the initiation reaction itself or indirectly in the production of membranous replication complexes required for the reaction is not yet known. In PIRCs negative strand RNA is made before positive strand RNA and RNA replication is asymmetric, with an excess of plus strand RNA made for each negative strand RNA template. During the initiation reaction hundreds of excess molecules of VPgpUpU per minus strand are made (85) and each minus strand is copied into 40-70 copies of plus strand progeny RNA (102). Barton and coworkers have demonstrated that PIRCs produced RF RNA and VPg-linked genome length plus strand RNA in a large excess over minus strands (13). Unlike in the simple in vitro reaction that uses purified components, the synthesis of VPgpUpU in PIRCs is sensitive to detergents suggesting a requirement for intact membranes in this process (47).

3.6. Protein/protein and protein/RNA interactions involved in the VPg-uridylylation reaction

In vitro three proteins are required for the uridylylation of PV VPg on cre(2C) RNA: RNA polymerase 3D^pol, the substrate VPg, and an RNA binding protein 3C^pro/3CD^pro (section 3.2) (114). While 3C^pro alone can stimulate the reaction it is not as effective as 3CD^pro suggesting that the 3D^pol domain also contributes to its function (101, 109, 140, 177, 179). The enhancing effect can be abolished by mutations within the RNA–binding domain of PV 3C^pro (R84S, I86A) (114). Direct binding between 3C^pro and cre(2C) transcript RNA was demonstrated in vitro by gel-shift and filter binding assays (179). Both structural and biochemical studies have shown that PV 3C^pro can form dimers in vitro and furthermore genetic studies indicate that PV 3C^pro dimer functions in VPg-uridylylation (108). Recent studies have suggested that 3C^pro binding to cre(2C) occurs in two steps. First a dimer of 3C^pro binds to the upper stem of cre(2C) and unwinds it such that each 3C-domain interacts with a strand of the stem in a sequence specific way (108). Subsequently, 3D^pol is recruited to this complex by an interaction between the back of the thumb subdomain of 3D^pol and the top of the 3C^pro dimer (26, 110).

Since in vitro VPg functions well as substrate for uridylylation by 3D^pol, one would expect that it binds into the active site of the enzyme. However, previous genetic studies have indicated the presence of a second VPg-binding site (in the context of 3AB) on PV 3D^pol on the palm subdomain near motif E, far from the active site (84). This finding was supported by structural studies of CVB3 in which 3D^pol was cocrystalized with VPg and the peptide was found to be bound to the back of the thumb, again distant from the active site (59). One possible explanation of how VPg could be uridylylated at this site is that the nucleotidylylation reaction is carried out by a second molecule of 3D^pol (59, 103, 151). Such a “trans” uridylylation mechanism was supported also from structural studies of the EV71 polymerase complexed with VPg, in which VPg bound at the bottom of the palm domain could not access the catalytic site (27). It should be noted, however, that a front-loading model of VPg-binding and a “cis” uridylylation mechanism was proposed for FMDV 3D^pol (43) and for the RNA polymerase of rhinovirus (7). These conflicting results suggest that there exists some variation in the uridylylation reaction among members of the Picornaviridae. A recent study with PIRCs have provided strong evidence that PV VPg binds at the back site in the 3D^pol domain of PV 3CD^pro, and that a second molecule of 3D^pol carries out the linkage of UMP to the tyrosine of VPg (45).

3.7. A “slide-back” mechanism of VPg uridylylation

In our initial studies on VPg-uridylylation we demonstrated that A₅ and A₆ in the conserved A₅A₆ACA sequence in the cre(2C) loop are essential not only for the in vitro synthesis of the VPgpU/VPgpUpU precursors but also for RNA replication in vivo and for viral growth (113, 114, 123). Since RNA polymerases synthesize new chains in the 5’----✧ 3’ direction we expected that A₆ would serve as template for the addition of the first UMP to VPg followed by the addition of the second UMP on the A₅ template nucleotide. In addition, we assumed that the RNA polymerase possesses a strict specificity toward UTP in the nucleotidylylation reaction since in both minus and plus strand RNAs VPg is linked to UMP, the first nucleotide of the RNA chains. Surprisingly, neither of these predictions turned out to be correct. We observed that beside U, 3D^pol could also link C, G, or A to VPg on mutant templates that contained G, C, or U at the A₅ position of the loop, respectively (Fig. 5) (116). The nucleotidylylated peptides, VPgpN (where N is C, G or A) could not be efficiently elongated into a dinucleotide such as VPgpNpN or VPgpNpU. In contrast, nucleotide changes at the A₆ position of the conserved motif did not lead to any linking of C, G, or A to VPg but resulted in a much reduced yield of products, particularly of VPg-linked dinucleotides.

These results are consistent with a “slide-back” mechanism for the synthesis of VPgpUpU, similar to what was proposed for the initiation of protein-primed DNA synthesis by the DNA polymerases of adenovirus or phages Φ29, PRD1 or G-1 (Fig. 5) (111, 133, 134). According to this model the first UMP (U₁) is linked to the hydroxyl group of Y3 in VPg on the A₅ template nucleotide of the conserved cre(2C) sequence in the loop, yielding VPgpU (116). In the next step VPgpU₁ “slides-back” so that U₁ can hydrogen bond to A₆. This is followed by the addition of a second UMP (U₂), again templated by A₅, producing the final product VPgpU₁pU₂. In this model A₅ is the template for the addition of both U₁ and U₂ to VPg and A₆ provides the specificity to the reaction. Although A₆ itself is not used as a template the role of this nt is very important because it ensures that the only dinucleotide produced will be VPgpUpU. The polymerase 3D^pol has a relaxed nucleotide specificity in the nucleotidylylation reaction but it does not possess a proof-reading activity. Therefore the enzyme would be unable to remove any incorrect nucleotide from VPgpN (N is C, G or A), if it were produced by mistake. However, during the “slide-back” step VPgpN cannot hydrogen bond with A₆ therefore it will not be converted into a dinucleotide and only VPgpU will be elongated into VPgpUpU by the enzyme.

The elongation of VPgpU is aborted after the addition of the second U on the cre(2C) template, to produce VPgpUpU (116). For example, the polymerase is not able to synthesize VPgpUpUpU on a mutated template in which C₄ is substituted with an A (116). In addition, further copying of the loop sequence in cre(2C) would result in VPgpUpU being linked to a G and then a C, products that would not be complementary to the 3’ sequence of the minus strands during plus strand RNA synthesis. Therefore the termination of VPgpUpU elongation is likely due to some structural characteristics of 3D^pol associated with the nucleotidylylation reaction. Even in the poly(A)-templated reaction the primary products are VPgpU/VPgpUpU and only a small fraction of the precursors are elongated into VPg-poly(U) (115). Such abortive synthesis of VPgpUpU is also evident in crude replication complexes isolated from PV-infected cells (149, 150, 159) and in PV-infected HeLa cells in vivo (32).

Both biochemical and genetic experiments suggest that VPg is delivered to the replication complex in the form of one of its precursors (80, 110, 157). However, the identity of the precursor that is used as substrate for uridylylation remains to be elucidated. Evidence for 3AB (or a 3AB precursor) as the VPg-precursor in vivo was provided by Liu and coworkers who showed that a replication defect caused by an Y3F mutation in VPg could only be rescued by wt P3 (80). Other in vivo studies showed that a mutation in the 3A domain of 3AB that caused a replication defect that could only be rescued by the P3 polypeptide and not by 3AB (157). Surprisingly, 3AB can not be uridylylated by 3D^pol in vitro on the cre(2C) template with Mg⁺⁺ as the cofactor while 3BC is an excellent substrate for this reaction under the same conditions (110). Although 3BC is not normally observed in infected cells, 3BC-linked RNA was found with a PV replicon that contains a mutation in the 3B/3C cleavage site (104). This suggests the possibility that the VPg precursor in vivo is 3BC or 3BCD or P3.

3.8. Is VPgpUpU synthesized on the cre required for both plus and minus strand RNA synthesis?

The fact that the picornaviral cre templates the synthesis of VPgpU and VPgpUpU for priming RNA synthesis has been well established for several viruses by both biochemical and genetic experiments (52, 100, 114, 123, 176, 178). However, it was not clear whether the VPgpUpU made on the cre was used in vivo for both minus and plus strand RNA synthesis. As far as minus strand RNA synthesis is concerned, beside cre(2C), the 5’-end of the poly(A) tail could also be considered a potential template for the uridylylation of VPg (98, 115). Similarly, instead of cre(2C), the two As at the 3’end of minus strands might serve as templates for VPgpUpU synthesis during plus strand RNA synthesis (138). However, early studies in vivo with full-length transcript RNAs of PV or CVB3, lacking the first two Us of the genome, indicated that they were infectious but the two Us were regained during replication in transfected cells (62, 73). These observations suggested that the first two As at the 3’-end of minus strand RNA are not essential as templates for VPgpUpU synthesis in vivo during plus strand RNA synthesis. Subsequently, PIRCs have provided useful, though sometimes conflicting information, on this subject (14, 56, 164). Initial experiments with this system suggested that the VPgpUpU made on the cre(2C) was used only for plus strand RNA synthesis (56, 96, 98). Subsequently, Steil and Barton have shown that cre-dependent VPgpUpU is made both before and during minus strand RNA synthesis, with 100-400 molecules of VPgpUpU made per negative-strand (145). This observation is consistent with the possibility that VPgpUpU also primes negative strand RNA synthesis. The hypothesis was further supported by several lines of evidence: (i) guanidine, a potent inhibitor of RNA replication and 2C^ATPase activity, reversibly inhibits both cre-dependent VPg uridylylation and minus strand RNA synthesis (98, 119); (ii) the A₅C mutation in the CVB3 cre blocked both VPg-uridylylation in vitro and minus strand RNA synthesis in PIRCs (164); (iii) under conditions of low UTP concentration VPgpUpU made on the cre is the preferred primer for negative strand RNA synthesis in PIRCs (146). The consensus opinion emerging on this topic is that cre(2C) templates VPgpUpU synthesis for the production of both minus and plus strands of PV RNA.

4. Models of poliovirus minus and plus strand RNA synthesis

The following models aim to integrate all the data currently available from both genetic and biochemical experiments, including preinitiation replication complexes (72, 111, 131, 144). It is clear that picornavirus RNA replication is an extremely complex process, which takes place on membranous structures (19) where several cis-acting RNA elements and viral and cellular proteins have to function in a coordinated fashion. The membranous structures involved in this process are not yet well understood and will not be discussed in the models presented below. Therefore the models, as presented, are highly simplified versions of what might happen in vivo.

Minus strand RNA synthesis begins with the circularization of the PV plus strand RNA via interaction between 3CD^pro and PCBP2 (and/or 3AB) bound to the 5’cloverleaf and PABP, which interacts with the poly(A) tail (Fig. 6A). Whether the interaction between the RNP complexes at the 5’- and 3’-ends of the RNA occurs only in cis or can also take place in trans is not yet known. VPg or one of its precursors is uridylylated on the cre(2C) template to yield VPgpUpU. This is then transferred to the 3’-end of the poly(A) tail but the mechanism of transfer and what provides specificity to this process are not yet known. Since the two uridylates attached to VPg are insufficient to provide sequence specificity to the transfer reaction it is likely that this is accomplished through binding of VPgpUpU to 3D^pol alone or complexed with 3CD^pro, which could recognize a sequence or structure near the 5’-end of the genome. In the next step 3D^pol elongates the VPgpUpU primer on the plus strand RNA template into a VPg-linked minus strand yielding a double stranded intermediate RNA (RF). In order to begin plus strand RNA synthesis the end of the RF structure has to be unwound or at least destabilized so that the plus strand cloverleaf can refold and interact with a 3CD^pro/PABP/PCBP2 complex (Fig. 6B). This might be facilitated by the binding of 2C/2BC^ATPase to the 3’-end of the minus strand cloverleaf. A cellular protein hnRNP C then might bind to both the 5’- and 3’-ends of the minus strand thereby stabilizing the interactions between the termini of the RF. Alternatively, 3AB, which possesses RNA chaperone and helix destabilizing activity, and binds in a complex with 3CD^pro to the plus strand cloverleaf, assists in the unwinding process. VPgpUpU, made on cre(2C) prior to the formation of RF, is then transferred to the 3’-end of the minus strand cloverleaf. As noted above, it is likely that VPgpUpU is in a complex with 3D^pol and/or 3CD^pro during the translocation step. The two terminal adenylate residues align with the two Us of VPgpUpU and 3D^pol elongates the plus strand RNAs on the minus strand template. Multiple plus strands are made off the same minus strand template yielding a replicative intermediate (RI) and finally the full length plus strands are completed.

Fig. 6.

Models of minus and plus strand RNA synthesis. (A) The poliovirus RNA circularizes through the interaction of the cloverleaf/PCBP2/3CD^pro complex with PABP that is bound to the poly(A) at the 3’-end of the genome (section 4). VPgpUpU is synthesized on the cre(2C) and is probably transferred in a complex by 3D^pol and/or 3CD^pro to the 3’-end of the poly(A) tail. There it serves as the primer for minus strand RNA synthesis yielding a double-stranded replicative form (RF). (B) The 5’-end of the RF is destabilized by the binding of viral and cellular proteins. 3CD^pro in a complex with PCBP2 or 3AB binds to the plus strand cloverleaf after destabilizing the 5’-end of the RF. 2C^ATPase and hRNP C bind to and stabilize the minus strand cloverleaf (section 4). VPgpUpU that was made prior to or during minus strand synthesis is transferred in a complex with 3D^pol and or 3CD^pro to the 3’-terminal two As of the minus strand. There VPgpUpU is elongated by 3D^pol yielding a replicative intermediate (RI) that contains multiple plus strands at various stages of elongation. The final products are the full length VPg-linked plus strands.

5. Unanswered questions about the initiation of picornavirus RNA replication

Although great progress has been made in identifying the viral and cellular factors involved in the initiation of picornavirus minus and plus strand RNA synthesis, many important questions remain unanswered:

1. what is the true protein substrate in vivo for uridylylation?

2. why are the cre elements of picornaviruses located in different regions of the RNA genomes?

3. how is VPgpUpU translocated from the cre to the 3’-end of plus strand RNA and 3’-end of minus strand RNA?

4. what provides specificity to the transfer of VPgpUpU from the cre to the ends of plus and minus strand RNAs?

5. what factors (viral and/or cellular) are required for the transfer reaction?

6. what is the role of membranes in VPg uridylylation in vivo?

7. why is there a large excess of VPgpUpU molecules synthesized relative to the number of progeny RNAs made?

8. how is the end of the RF unwound prior to plus strand RNA synthesis?

9. what is the role of 2C^ATPase in VPgpUpU synthesis?

6. Conclusions

As we discussed above (section 3), it is well established that picornavirus RNA polymerases utilize a protein-primed reaction to synthesize both their minus and plus RNA strands. Although initially the poly(A) tail was proposed to template VPgpUpU synthesis for minus strand RNA synthesis, it is now generally accepted from both in vivo studies and those with PIRCs that VPgpUpU made on the cre is used to prime both minus and plus strand RNA synthesis. Assuming that the mechanism of protein-primed RNA synthesis is similar with all viruses that possess terminal proteins this model is consistent with the existence of some plant viruses (Sobemovirus, Luteovirus, Enamovirus and Barnavirus) with VPg-linked genomes that do not possess poly(A) tails (139). It is also clear that the template for the linkage of two UMPs to VPg (or VPg precursor) is a conserved sequence in the loop of a cre element, located usually in the ORF, and the reaction is strongly enhanced by the presence of viral protein 3CD^pro. The PV VPg-uridylylation reaction is strictly dependent on the presence of the cloverleaf, of 2C^ATPase function, and of membranes in vivo and in PIRCs but not in vitro. The “slide back” mechanism of VPg uridylylation ensures the integrity of the ends of the viral RNA genomes.

Studies with other plus strand viruses are limited but so far they suggest that protein-primed RNA synthesis is a general mechanism used by viruses that contain protein-linked genomes. Recently, the RNA polymerase and/or RNApolymerase precursor of norovirus was shown to nucleotidylylate its VPg protein (18, 60, 128). There are a large number of RNA virus families, in addition to Picornaviridae, that possess genome-linked proteins (82): Caliciviridae, Astroviridae, Potyviridae, Comoviridae, Dicistroviridae and Sequiviridae (133). The terminal proteins of these viruses vary in length from 2 kDa to ~24 kDa and also vary in amino acid sequence.

Cre elements have been either discovered or their existence has been predicted for several but not all genera of Picornaviridae (144). It should be noted that the conserved AAA/GC sequence within the loop of the typical picornavirus cre hairpin, together with conserved RNA structure predictions, might be useful in the future in predicting the location of unknown cre elements in newly identified genera of Picornaviridae. Outside of picornaviruses there is very little known about cre elements in members of other plus strand RNA virus families that might be templates for nucleotidylylation of their VPgs. However, stimulation of the VPg uridylylation reaction by RNA sequences in the genome of norovirus, a member of Caliciviridae, was reported (18, 60). In addition, the presence of a cre-like sequence was observed in the genome of norovirus (166).

The insights we have gained from studies of protein/protein and protein/RNA interactions involved in the process of the initiation step of picornavirus RNA replication will hopefully facilitate studies of RNA replication of other plus strand RNA viruses with genome-linked proteins. Since protein-primed RNA replication normally does not occur in eukaryotic cells this unique reaction is particularly well suited for studies of drug development to fight picornaviral infections.

Highlights.

• Picornaviruses use a protein-primed mechanism of RNA synthesis

• The viral RNA polymerase links two UMPs to a tyrosine in VPg, the terminal protein

• Sequences in an RNA hairpin are the templates for the linkage of the UMPs to VPg

• VPgpUpU is the primer for the elongation of RNA chains by the RNA polymerase