How does the polymerase of non-segmented negative strand RNA viruses commit to transcription or genome replication?

ABSTRACT

The Mononegavirales, or non-segmented negative-sense RNA viruses (nsNSVs), includes significant human pathogens, such as respiratory syncytial virus, parainfluenza virus, measles virus, Ebola virus, and rabies virus. Although these viruses differ widely in their pathogenic properties, they are united by each having a genome consisting of a single strand of negative-sense RNA. Consistent with their shared genome structure, the nsNSVs have evolved similar ways to transcribe their genome into mRNAs and replicate it to produce new genomes. Importantly, both mRNA transcription and genome replication are performed by a single virus-encoded polymerase. A fundamental and intriguing question is: how does the nsNSV polymerase commit to being either an mRNA transcriptase or a replicase? The polymerase must become committed to one process or the other either before it interacts with the genome template or in its initial interactions with the promoter sequence at the 3´ end of the genomic RNA. This review examines the biochemical, molecular biology, and structural biology data regarding the first steps of transcription and RNA replication that have been gathered over several decades for different families of nsNSVs. These findings are discussed in relation to possible models that could explain how an nsNSV polymerase initiates and commits to either transcription or genome replication.

INTRODUCTION

The non-segmented negative-strand RNA viruses (nsNSVs) are an order of RNA viruses, known as the Mononegavirales, that encompasses several viral families (1). These families include, among others, the Filoviridae, which includes viruses such as Ebola virus (EBOV) and Marburg virus (MARV); the Paramyxoviridae, which includes childhood pathogens, such as measles virus, mumps virus, the parainfluenza viruses (PIV-3, PIV-5), and the highly pathogenic Nipah virus and Hendra virus; the Pneumoviridae, which includes the respiratory pathogens, respiratory syncytial virus (RSV) and human metapneumovirus (HMPV), and the Rhabdoviridae, which includes rabies virus (RABV) and the well-studied animal pathogen, vesicular stomatitis virus (VSV). An intriguing facet of the nsNSVs is that they encode a single polymerase that is capable of transcribing and replicating their genomic RNA. Moreover, both transcription and replication are initiated using the same promoter region at the 3´ end of the viral genome. These features lead to the key question addressed in this review, namely, how does the nsNSV polymerase become committed to being either a transcriptase or replicase?

OVERVIEW OF THE NSNSV TRANSCRIPTION AND REPLICATION PROCESSES

The key players in nsNSV transcription and replication are the encapsidated viral genome, and the cis-acting elements contained within the polymerase comprised of a complex of the large polymerase subunit (L), phosphoprotein (P; or viral protein 35, VP35 in the filoviruses), and nucleoprotein (N). Each nsNSV genome is a single strand of negative-sense RNA, with the viral genes positioned sequentially (Fig. 1). Each gene is flanked by conserved elements: a gene start (gs) signal at the beginning of each gene and a gene end (ge) signal at the end (2–9). In most cases, the genes are separated by short intergenic regions. In some nsNSVs, such as VSV and Sendai virus (SeV), these are short (2 or 3 nt) and identical at each junction, whereas in other nsNSVs, such as RSV and EBOV, they can be variable in length and sequence, and genes can overlap in some cases. Preceding the gs signal at the 3´ end of the genome is a leader (le) region of approximately 50 nt, and after the last ge signal at the 5´ end of the genome is a trailer (tr) region, which is variable in length, depending on the virus (8) (Fig. 1). Transcription is initiated at the 3´ end of the genome and involves sequential synthesis of each of the viral genes. During this process, the polymerase is governed by the gs and ge sequences (10–20). The gs at the beginning of the first gene directs the polymerase to initiate pre-mRNA synthesis, the polymerase then caps the mRNA, methylates the cap, and elongates the pre-mRNA until it reaches the ge signal (21) (Fig. 2). At the ge signal, the polymerase stutters on a U-tract to polyadenylate the RNA, which is then released as a mature mRNA (11) (Fig. 2). Following mRNA release, the polymerase can remain attached to the template, scan to locate the next gs signal, and then reinitiate RNA synthesis (22). It continues in this vein to transcribe the remainder of the genome. The polymerase can only initiate transcription at or near the 3´ end of the genome and cannot initiate directly at gs signals of downstream genes. Thus, transcription initiation from the 3´ end of the genome is essential for the synthesis of all viral mRNAs (23–25). The polymerase does not reinitiate at gs signals with 100% efficiency, resulting in a gradient of transcripts, with those transcribed from the genes at the 3´ end of the genome being more abundant than those transcribed from genes at the 5´ end (24, 26–29). It is unknown if the polymerase that fails to reinitiate dissociates from the template RNA or continues moving along the template in a non-synthesis mode. Replication also begins at the 3´ end of the genome. In this case, the polymerase disregards the gs and ge signals and synthesizes a full-length, uncapped complement of the genome, known as the antigenome. The antigenome is encapsidated with N protein as it is synthesized, and this is thought to allow the polymerase to be super-processive and able to read through the ge signals (Fig. 2). The antigenome in turn acts as a template for the synthesis of uncapped genomic RNA.

Fig 1.

Genome organization and arrangement of nsNSVs. Schematic diagrams of the genomes of the rhabdovirus, VSV, the pneumovirus, RSV, the paramyxovirus, SeV, and the filovirus, EBOV. The genomes are drawn based on Genbank numbers EF197793.1, U39661.1, DQ219803.1, and AF499101.1, respectively. They are drawn to scale, except for the cis-acting elements, and the gene overlaps. The white boxes represent a gs signal, blue boxes represent the labeled genes, and black boxes represent a ge signal. The 3´ le and 5´ tr are on the left and right of each genome, respectively. A representative intergenic region and/or gene overlap is shown for each virus. In the case of the RSV genome, the M2 and L genes overlap by 68 nt. Note that throughout the review, viral gene names and sequences containing cis-acting elements are italicized.

Fig 2.

Transcription and replication by nsNSV polymerases. Schematic diagrams of transcription and replication by the nsNSV polymerases. The polymerase (L) is shown as an orange oval at different positions within a gene; the white boxes represent a gs signal, blue boxes represent the genes, and black boxes represent a ge signal. A U₄ tract within the ge signal is labeled. The positive-sense RNA product is shown as a black line coming out of the polymerase; mRNA with a 5´ cap and 3´ poly(A) tail for transcription, and antigenome that is encapsidated with N protein (blue circles) for replication.

OVERVIEW OF THE NSNSV TRANSCRIPTION-REPLICATION MACHINERY

The template for the nsNSV polymerase is a nucleocapsid, which is comprised of the RNA coated by a linear N oligomer (30, 31). As noted above, the polymerase is a complex of L-P or L-VP35. The nsNSV L proteins contain the enzymatic domains necessary for all the steps of mRNA and replicative RNA synthesis, including RNA synthesis, cap addition, and cap methylation (Fig. 3A). The P or VP35 protein serves as an adaptor that forms a bridge between L and the N oligomer coating the RNA; in addition, P and VP35 proteins recruit soluble, unassembled N protein (N⁰) to encapsidate nascent replicative RNA (32) (Fig. 3B). High-resolution polymerase structures are now available for the representatives of four nsNSV families: the rhabdoviruses VSV and RABV (33–35); the pneumoviruses RSV and HMPV (36–38); the paramyxoviruses PIV-3, PIV-5, Newcastle disease virus (NDV), and mumps virus (39–42); and the filovirus EBOV (43, 44) (Fig. 3C). These structures reveal that L consists of five domains, known as the RNA-dependent RNA polymerase (RdRp) domain, which synthesizes RNA; the capping domain, which adds a guanosine cap to the 5´ end of each mRNA; connector domain (CD); methyltransferase (MT) domain, which methylates the cap; and C-terminal domain (CTD) (35, 45, 46) (Fig. 3A and C). The RdRp and capping domains make up the core of the L protein structure and contain the active sites for RNA synthesis (which includes a characteristic GDN motif) and a polyribonucleotidyltransferase activity (PRNTase; a key step of the distinctive nsNSV capping reaction), respectively (47–51). The connector, methyltransferase, and C-terminal domains are globular appendages. The methyltransferase and C-terminal domains function cooperatively to methylate the mRNA cap (52, 53) and might also participate in the GTPase step of the capping reaction (50, 52). These globular appendages are flexible relative to the core and have been observed in different arrangements (54) (Fig. 3C). Accumulating evidence indicates that the polymerase undergoes multiple conformational changes to juxtapose the different globular domains appropriately for the different steps involved in transcription and RNA replication (45, 46). In addition, the L-P complex might adopt a dimeric form: negative stain electron microscopy images of rhabdovirus polymerases have revealed L dimers (33, 34, 54), and in recent work, it was shown that the PIV-3 L-P complex forms a functional dimer through a capping-connector domain interface (42).

Fig 3.

The nsNSV polymerases (L) and phosphoproteins (P/VP35). (A) Schematic diagram of an nsNSV polymerase, based on the VSV polymerase structure. The domains are colored as follows: RdRp in cyan, PRNTase/capping domain in green, CD in yellow, MT in orange, and CTD in red. (B) Schematic diagram of the NDV phosphoprotein tetramer, showing the flexible N-terminal domain (P_NTD) and C-terminal domain (P_CTD) arms, and the ordered oligomerization domain (P_OD). (C) Representative structures of nsNSV polymerases from each family available: rhabdovirus VSV L-P (PDB: 6U1X), paramyxovirus PIV-5 L-P (PDB: 6V85), pneumovirus RSV L-P (PDB: 6PZK), and filovirus EBOV L-VP35 (PDB: 7YES). The L domain colors are the same as in A and the P tetramer is the same color as in B. Note that most of the flexible N- and C-terminal arms of PIV-5 and RSV P and EBOV VP35 are not visible in the cryo-EM structure, and only three small segments of the VSV P are visible. Comparison of the VSV and PIV-5 polymerase structures shows how the CD, MT, and CTD can be rearranged relative to the RdRp-PRNTase domain core.

The P protein of the pneumoviruses, paramyxoviruses, and filoviruses is a tetramer (although other oligomeric forms have also been detected), whereas the rhabdovirus P protein is a dimer; in each case, the P/VP35 protein has a modular organization with a central oligomerization domain and N- and C-terminal extensions (39, 55–63) (Fig. 3B). In most nsNSVs, the C-terminal extensions of P (P_CTD) and VP35 proteins bind to the N oligomer that coats genomic and antigenomic RNA, allowing the polymerase complex to associate with the template (64–71). The N-terminal extensions of P (P_NTD) and VP35 chaperone the monomeric N⁰ protein by preventing N-N binding and probably deliver the N⁰ protein to the nascent RNA being extruded from the polymerase during RNA replication (72–84).

In some viruses, other proteins also come into play during RNA synthesis. For example, RSV encodes a transcription elongation factor, M2-1, which is a component of the transcribing polymerase complex (85, 86). Although it is tempting to speculate that M2-1 might switch the polymerase between a transcriptase and replicase function, M2-1 is not required for replication and has no effect on the transcription-replication balance (87). The filoviruses encode viral protein 30 (VP30), which is required for transcription of genes that contain RNA secondary structures at their start (88–94) and has another less-defined transcription role (94–96). In contrast to M2-1, VP30 does appear to have a regulatory function, as the phosphorylation status of VP30 can regulate the polymerase between transcription and replication (97, 98). However, VP30 is not essential for either transcription or replication to occur, indicating that although it has a regulatory function, there is a distinct mechanism for committing the filovirus polymerase to one process or the other.

ROLES OF RNA SIGNALS IN DIRECTING CAPPING AND ENCAPSIDATION

Although there is only a single L polymerase protein, the events that occur during transcription involve the enzymatic activities of cap addition and cap methylation that do not occur during replication. These enzymatic activities, coupled with the fact that transcription involves recognition of gene junction sequences, whereas replication does not, highlight the differences between the transcription and replication processes. Because the L protein can adopt different conformations, it is possible that the polymerase can distribute into different pools, with one pool having a conformation suitable for transcription and another pool having a conformation appropriate for RNA replication. However, this is not necessarily the case, and instead, the polymerase might be regulated between different activities depending on where it has initiated the nascent RNA that it is synthesizing. This is because transcription and replication depend on a “dialog” between the polymerase and signals in the RNA. As noted above, the template contains sequences that allow polymerase binding, RNA synthesis initiation at the 3´ end of the template and at internal gs signals, and mRNA polyadenylation. In addition to these template signals, signals at the 5´ end of the nascent RNA products also coordinate polymerase activity. The complement of the gs sequence at the 5´ end of each pre-mRNA contains signals that direct the polymerase to cap the mRNA and methylate the cap (52, 99–104). Checkpoints have evolved such that the polymerase must successfully complete each step involved in transcription to be able to proceed to subsequent steps. For example, studies with VSV and RSV have shown that perturbation of cap addition causes the polymerase to abort pre-mRNA synthesis prematurely (47, 48, 50, 102, 105, 106). In rare cases in which the polymerase fails to cap the pre-mRNA and does reach the ge signal, it is able to release the RNA, but the RNA is not appropriately polyadenylated, and the polymerase is unable to re-initiate at the next gs signal (106, 107). Together, these findings indicate that the sequence at the 5´ end of the nascent, pre-mRNA communicates with the polymerase to activate its capping and methyltransferase enzymatic activities, which in turn signal the polymerase to proceed to perform sequential transcription. This can explain why the le(+) and replicative RNAs are not capped, namely, they do not have the necessary cis-acting capping signal at their 5´ ends. Regarding RNA replication, there is evidence to suggest that the 5´ ends of newly synthesized antigenomic and genomic RNAs contain a “seed” signal that allows for the initiation of RNA encapsidation (108–112). Once the encapsidation process has been initiated, elongation of the N chain could involve the delivery of N⁰ onto the existing assembled N protomers by the viral polymerase (73, 113, 114). Encapsidation facilitates highly processive polymerase elongation and prevents gene junction recognition (115–118). Thus, if the polymerase initiates at the 3´ end of the le region and N⁰ has reached a critical threshold, the le(+) RNA product could become encapsidated, beginning the replication process. If there is insufficient N⁰ protein for encapsidation, the le(+) RNA would be neither encapsidated nor capped, causing the polymerase to be non-processive and release the RNA. Thus, while it is possible that there are distinct pools of transcriptase and replicase, each with the appropriate properties for capping and cap methylation versus encapsidation, respectively, this is not necessarily the case. Instead, it could be the same pool of polymerase that performs mRNA synthesis and replicative RNA synthesis, with the difference in encapsidation and cap addition being determined by the RNA initiation site and the resulting sequence at the 5´ end of the nascent RNA. In this case, the different polymerase conformations that have been observed could represent polymerase at different stages of RNA synthesis, rather than distinct transcriptase or replicase pools.

THE SINGLE PROMOTER MODEL: THE POLYMERASE INITIATES OPPOSITE NUCLEOTIDE 1 OF THE LE PROMOTER TO BEGIN BOTH RNA REPLICATION AND TRANSCRIPTION

Key to understanding how nsNSV polymerases perform transcription or replication is determining the initial steps in these processes. As noted above, a key difference between the two processes is that the replication products are encapsidated with N protein, whereas transcription products are not (Fig. 2). This finding, coupled with the early observation that the VSV polymerase synthesizes abundant amounts of short RNAs complementary to the le region, referred to here as le(+) transcripts, led to a simple model for nsNSV transcription-replication control (119), which can be described as follows. The le region contains a promoter that directs the initiation of both transcription and replication, and the two processes are initiated in an identical fashion opposite the first nucleotide of the le (Fig. 4A). According to one version of this model, there is a single pool of polymerase, and by default, the polymerase is relatively non-processive. Having transcribed the le region, the polymerase releases the le(+) transcript but stays attached to the template and can scan to locate the first gs signal. Here, it can reinitiate RNA synthesis and begin synthesizing mRNA. The polymerase can then continue in a transcription mode, recognizing and responding to the downstream gs and ge signals, which signal it to cap and polyadenylate the mRNAs. According to this model, the default polymerase activity is altered when unassembled N⁰ protein reaches a critical concentration threshold that is high enough for the le(+) RNA to become encapsidated. Concurrent encapsidation prevents the polymerase from releasing the le(+) transcript and causes the polymerase to be super-processive, allowing it to synthesize antigenomic RNA to the end of the genome template (Fig. 4A). A modification of this model is that there are conformationally distinct pools of transcriptase and replicase that both initiate from the 3´ end of the le promoter region (Fig. 4B). According to this model, the replicase could be primed to concurrently encapsidate the nascent RNA (perhaps by being bound to N⁰ protein), whereas the transcriptase could be a distinct pool of polymerase that has a conformation that allows it to recognize and respond to gene junction signals and cap and polyadenylate the mRNAs.

Fig 4.

Four nsNSV transcription initiation models. Schematic diagrams of four different transcription-replication initiation models. The negative-sense genome is shown with the 3´ leader (le), the gene start (gs) signal in white, and part of the first gene in blue. A putative nucleation encapsidation signal (not shown) is present at the 5´ end of the le(+) RNA and antigenome. Only when nucleoprotein (N, blue circles) is present at a sufficiently high concentration will the RNA become encapsidated. (A) Single promoter - single polymerase model: in this model, the polymerase (pol) begins transcription and replication at the 3´ end of the le region and becomes either a transcriptase or replicase depending on the levels of N⁰. If N⁰ is limiting, the polymerase produces a short le(+) transcript and then reinitiates RNA synthesis at the first gs signal. If N⁰ has reached a critical threshold, the RNA becomes encapsidated, and the polymerase can perform RNA replication. (B) Single promoter - replicase/transcriptase model: in this model, there are two pools of polymerase, replicase (rep) and transcriptase (trans), which have different functional capabilities but compete for binding to a single promoter and initiation site. (C) Dual promoter – single polymerase model: in this model, there is a single pool of polymerase that can either begin replication by initiating at the 3´ end of the le region or transcription by initiating directly at the first gs signal. It should be noted that according to this model, short le(+) transcripts might be synthesized as abortive replication products (at low N protein concentrations). (D) Dual promoter - replicase/transcriptase model: in this model, there are distinct pools of replicase and transcriptase (as described for B), which initiate directly at the 3´ end of the le region or the gs signal, respectively. In B and D, replicase and transcriptase and their respective RNA products are shown in purple or pink, respectively.

THE DUAL PROMOTER MODEL: THE POLYMERASE INITIATES OPPOSITE NUCLEOTIDE 1 OF LE REGION TO BEGIN RNA REPLICATION AND DIRECTLY AT THE FIRST GS SIGNAL TO BEGIN TRANSCRIPTION

An alternative model is that the polymerase initiates transcription and replication by two distinct mechanisms, initiating at the 3´ end of the le region to begin RNA replication or at the first gs signal to begin transcription (Fig. 4C and D). In this model, polymerase that binds to the 3´ end of the le region would engage in replication, whereas polymerase that binds at the first gs signal would engage in transcription. This model could involve a single pool of polymerase that can bind a promoter at the 3´ end of the le region or at the gs signal and distribute between these two initiation sites according to the relative affinities for the two signals (Fig. 4C). In this case, the model for replication could be similar to that described above; specifically, there could be an encapsidation nucleation site within the le(+) RNA that allows the polymerase that has initiated at the 3´ end of the le promoter to become a replicase, whereas the polymerase that has initiated at the gs signal would have the appropriate cis-acting signal at the 5´ end of the nascent RNA to allow it to cap the pre-mRNA and become a transcriptase. Alternatively, there could be two conformationally distinct pools of polymerase, a transcriptase and replicase with different properties, as described above, and different propensities for binding to the two different initiation signals (Fig. 4D).

Currently, there are insufficient data to know if nsNSV polymerases form conformationally distinct pools of transcriptase and replicase, and hence, it is not clear if the models presented in Fig. 4B and D are valid. However, there is a body of evidence available regarding the initiation of nsNSV transcription and replication, allowing us to distinguish between the single promoter model shown in Fig. 4A and B and the dual promoter model shown in Fig. 4C and D. These findings are described in the following sections.

RHABDOVIRUS TRANSCRIPTION AND REPLICATION INITIATION

The rhabdovirus VSV is well studied because it grows to high titer, permitting experiments that are not possible with other nsNSVs. Importantly, transcription and replication can be studied using an in vitro assay in which RNA synthesis is either reconstituted from nucleocapsids released from detergent-disrupted virions or by incubating a purified N-RNA template with L-P complexes. A key study using these approaches revealed that L-P complexes that had been released from the N-RNA template must synthesize the le(+) RNA before being able to initiate mRNA synthesis at the N gs signal (120). This finding is consistent with a model in which the process of transcription is initiated at the 3´ end of the le region as depicted in Fig. 4A and B. This conclusion is bolstered by experiments measuring the kinetics of RNA synthesis, which revealed a lag time prior to N mRNA synthesis, consistent with N mRNA synthesis occurring after synthesis of the le(+) RNA (24). Results from promoter mapping studies are also consistent with this model. The 3´ 17 nt of the VSV le region contains a promoter sequence that can signal initiation of RNA synthesis (121, 122), and studies using a cell-based minigenome showed that sequence within le nucleotides 1–17 is required for mRNA transcription in addition to replication (123). It was shown that the site for nucleocapsid assembly by N protein is located within the 5´ end of the le(+) RNA (108, 111) and that the presence of N protein enables the synthesis of encapsidated antigenomic RNAs otherwise unencapsidated le(+) RNA is synthesized (119, 124–126). Finally, it was shown that there is competition between transcription and replication for a common pool of polymerase (127). Together, these studies provide compelling evidence that the VSV polymerase begins transcription at the 3´ end of the le promoter region and synthesizes le(+) RNA prior to reinitiating RNA synthesis at the N gs signal, with replication occurring when the le(+) RNA becomes encapsidated prior to being released. These data are consistent with the models shown in Fig. 4A and B.

PNEUMOVIRUS TRANSCRIPTION AND REPLICATION INITIATION

Most work on pneumovirus transcription and replication has been performed with RSV. The RSV cis-acting elements involved in transcription and replication initiation have been finely mapped (116, 117, 128–130). By using a cell-based minigenome assay, it was shown that RSV le nucleotides 3C, 5C, 8U, 9U, 10U, 11U, and 12A comprise a core promoter element that is required for both transcription and replication (128, 130, 131). Although additional elements are required in addition to this core promoter for transcription and replication to occur, the core promoter element alone is sufficient for initiation of RNA synthesis (128, 132).

Although RSV has a common core promoter element for transcription and replication, the viral polymerase efficiently initiates from two sites within the promoter, at positions 1U and 3C, with position 3C being the dominant initiation site (116, 117, 131–133). In vitro RNA synthesis assays utilizing purified L-P complexes revealed that the L-P complex alone, without other viral proteins, initiates at positions 1U and 3C and is regulated between the two different initiation sites by the relative concentrations of the initiating NTPs, ATP, and GTP (131, 132, 134, 135). Polymerases that initiate at position 3C produce short heterogeneous le(+) RNAs that are predominantly ~20–25 nt in length and released (131). Several lines of evidence indicate that le(+) synthesis is the first step in RSV transcription. For example, a small molecule inhibitor, AZ-27, does not inhibit gs initiation by polymerase molecules that are already associated with the nucleocapsid template but does block initiation from position 3C and inhibits new rounds of mRNA transcription (136). A series of experiments with a small molecule inhibitor of the RSV polymerase, termed BI-D, showed that this compound inhibits le(+) transcript release, with an inverse correlation between the failure to release the le(+) RNA and initiation at the first gs signal (47). These findings, coupled with the promoter analysis, suggest that to begin transcription, the polymerase initiates RNA synthesis at position 3C, synthesizes and releases the le(+) RNA, and then reinitiates RNA synthesis at the first gs signal to begin the transcription process. RSV genome replication is performed by polymerase that initiates at 1U (131). In this case, the 5´ end of the nascent RNA contains a pppApC motif that is lacking in the nascent RNA initiated at 3C. The presence of 5´ pppApC correlates with the ability of the nascent RNA to be encapsidated (117), and this motif presumably functions in conjunction with other sequences within the le template or le(+) RNA to facilitate encapsidation (116). The RNA initiated at 3C lacks this 5´ pppApC motif, which could explain why this RNA is not encapsidated and is released. Thus, RSV follows a slight modification of the single promoter model shown in Fig. 4A and B, in which transcription and replication are both initiated from a promoter at the 3´ end of the le region but at two closely positioned sites within this promoter.

PARAMYXOVIRUS TRANSCRIPTION AND REPLICATION INITIATION

The paramyxoviruses differ significantly from the rhabdoviruses and pneumoviruses in two ways. First, their genome (and antigenome) nucleotide lengths must be an integer of 6 to be efficiently replicated (137–139). This correlates with each molecule of N protein binding six nucleotides, allowing the replicative products to be entirely and precisely coated with N protein (140–145). Second, the paramyxoviruses have two promoter elements (146). One element lies at the 3´ end of the le region, and the other lies after the first gs signal, within the untranslated region of the first gene. In the context of the nucleocapsid, these promoter elements, and gs signals, are recognized in conjunction with the N protein that encases them in the nucleocapsid template (139, 147, 148). Fine mapping of the PIV-3 promoter showed that nucleotides 1U, 2G, 3G, 6U, 7G, 8U, 9U, 11U, and 12C represent the 3´ promoter element (149). The second promoter element consists of a hexamer repeat in which most nucleotides can consist of any base, but one or two nucleotides in the hexamer repeat must be of a specific sequence. For example, promoter element 2 of SeV is three hexamer repeats of GNNNNN, positioned from nt 79–96 (150), and in PIV-5, it is (CGNNNN) x 3 (151). The positioning of the two promoter elements relative to each other is key. Interestingly, they are positioned on the same face of the helical nucleocapsid, suggesting that the polymerase, perhaps in association with a co-factor, recognizes them simultaneously (146).

As with RSV, paramyxovirus promoter mapping studies strongly support the single promoter initiation model for transcription-replication (Fig. 4A). Studies with SeV have shown that transcription and replication signals compete with each other, consistent with the idea that both processes are performed by polymerase binding to a single site for initiation (152, 153). Minigenome studies in which the first gs signal was separated from promoter elements 1 and 2 and moved further downstream showed that the first gs signal could still direct mRNA synthesis, even if it was moved from position 56 to position 864, provided that the 3´ end of the template maintained promoter elements 1 and 2 (153–155). This finding strongly suggests that recognition of the gs signal occurs as a downstream event after the polymerase has recognized and initiated at the 3´ promoter, rather than concurrently with recognition of promoter elements 1 and 2. Encapsidation occurs concurrently with RNA replication (115). If the le(+) transcript is not encapsidated, the RNA is released, whereas if it is encapsidated, it can be elongated to form antigenomic RNA (115, 118). In the case of measles virus and Nipah virus, increasing the intracellular level of N protein causes a switch from transcription to replication, indicating that either there is a common pool of polymerase that switches from being a transcriptase to a replicase or that transcriptases and N-associated replicases compete for the same promoter and initiation site (156, 157). Thus, although the paramyxoviruses have a very different template requirement and promoter structure compared with the rhabdoviruses and pneumoviruses, they also appear to follow a model in which transcription and replication are both initiated from the 3´ end of the le region, as shown in Fig. 4A and B.

FILOVIRUS TRANSCRIPTION AND REPLICATION INITIATION

The filoviruses are like the paramyxoviruses in that they have bipartite promoters, and both promoter elements 1 and 2 and the gs signal must follow hexamer phasing to function efficiently (158, 159). The promoter or promoters for filovirus transcription and replication lie within the 3´ 70 nt of the MARV genome and within the 3´ 128 nt of the EBOV genome (94, 160). An in vitro assay using purified MARV polymerase and a naked RNA oligonucleotide template showed that the first 17 nt of the tr promoter region, which is almost identical to the MARV and EBOV le regions, are sufficient to signal RNA synthesis initiation, indicating that promoter element 1 lies within this sequence (161). The second promoter element lies after the gs signal of the first (N) gene and consists of three consecutive hexameric repeats of UN₅ (94, 160). In the case of EBOV, there are additional hexamers in the leader and non-translated region of the N gene, which also contribute to replication efficiency (94, 158, 160). Intriguingly, ebolavirus genomes have heterogeneity at their 3´ termini, having either the sequence 3´ CCUGU, 3´ ACCUGU, or 3´ GCCUGU (162). This heterogeneity arises through the addition of either an A or G residue to the 3´ end of the genomic RNA, possibly through a back-priming mechanism (162). However, despite this heterogeneity, the EBOV polymerase consistently initiates opposite the first C residue of the template (162, 163).

A distinctive feature of the filoviruses is that their genomes have predicted RNA stem-loop structures that encompass the gs signal at the beginning of each gene. It is not known if these RNA structures can form in the template RNA (perhaps in a situation in which N protein does not completely cover the entire genome or antigenome), but their existence raised the possibility that the N gs signal could serve as an internal polymerase entry site for transcription, akin to the model shown in Fig. 4C and D. However, studies performed in the minigenome system showed that although the N gs stem-loop can influence transcription efficiency, this secondary structure is not required for transcription (94, 95). In addition, nucleotides at the 3´ end of the le promoter region are required for transcription, consistent with promoter element 1 being necessary for transcription initiation (163). The EBOV polymerase generates le(+) transcripts, which are longer than those of other nsNSVs being ~60–80 nt in length (163). Thus, these le(+) transcripts extend beyond the 55 nt le region. At first glance, this could suggest that the generation of le(+) transcripts is not the first step in transcription, as the polymerase would be situated downstream of the first gs signal following le(+) RNA release. However, given that the EBOV polymerase is able to scan backward to access a gs signal following RNA release at the overlapping gene junctions (12), polymerase that has released a 60–80 nt le(+) transcripts would be well positioned to scan backward to reinitiate RNA synthesis at the N gs signal. Together, the filovirus data are also consistent with a model in which transcription and replication are initiated from a core promoter at the 3´ end of the genome, as illustrated in Fig. 4A and B, although the possibility of internal initiation at the first gs signal has not been completely excluded.

SUPPORT FOR A MODEL OF TRANSCRIPTION INITIATION FROM THE FIRST GS SIGNAL

Together, the studies described above provide strong support for the single promoter model of transcription and replication initiation shown in Fig. 4A and B. However, there are some VSV studies that suggest an alternative model in which the polymerase can initiate transcription directly at the N gs signal (Fig. 4C and D). Work with a variant of VSV called polR1 showed that the number of initiations at the N gs signal was in excess of the number of le(+) transcripts that were generated, indicating that the polymerase could initiate at the N gs signal without synthesizing le(+) RNA (164). A second piece of evidence that the VSV polymerase can initiate internally at the N gs signal comes from a study that employed a UV inactivation mapping technique to determine if the N gene could only be transcribed after le(+) synthesis or independent of le(+) synthesis (165). When nucleocapsids from detergent and salt-disrupted virions were analyzed, the results showed that the polymerase needed to transcribe le(+) RNA prior to initiation at the N gs signal, consistent with previously described results and the model in Fig. 4A and B (120). However, when cells were infected with UV-treated virus and primary viral transcription was assessed, it was found that the polymerase could initiate at the N gs signal without prior synthesis of le(+) RNA, suggesting that in a cellular environment, transcription can be initiated directly at the first gs signal, consistent with Fig. 4C and D. There are also promoter mapping studies that support this model: while the VSV le region contains a promoter and transcription signals at its 3´ end, minigenome studies have revealed transcription signals throughout the le region, which could potentially guide the polymerase to initiate transcription at the gs signal (123, 166). These studies raise the possibility that internal initiation at the first gs signal can occur and might be favored in some conditions (Fig. 4C and D).

RELATIONSHIP BETWEEN POLYMERASE STRUCTURES AND INITIATION OF RNA SYNTHESIS

As noted above, structures of several different polymerases have been solved, allowing us to begin to unite the functional studies described above with an understanding of polymerase structure. Probably the best-understood step in nsNSV polymerase activity from a structure-function perspective is initiation at the 3´ end of the le promoter. Initiation of RNA synthesis at the 3´ end of a template presents a challenge for the polymerase, requiring it to have a stable interaction with the promoter to position the 3´ end of the template RNA precisely relative to the active site (including the GDN motif) and recruit the two initiating NTPs and position those appropriately for phosphodiester bond formation (167). There are three nsNSV polymerase-promoter RNA structures: RSV L-P-le RNA, RSV L-P-tr RNA, and EBOV L-VP35-le RNA (Fig. 5) (43, 168). These structures reveal an extensive network of protein-RNA and RNA-RNA interactions that function to position the promoter appropriately within the polymerase (43, 168). Although the core promoter sequences of different virus families are distinctive, it is likely that other nsNSVs have also evolved a network of polymerase-promoter interactions to stabilize the template in the initiation complex.

Fig 5.

Structures of RSV and EBOV L proteins in complex with their cognate promoter RNAs. Structures of the (A) RSV L + nucleotides 1–10 of the le promoter (PDB: 8SNX), (B) RSV L + nucleotides 1–10 of the tr promoter (PDB: 8SNY), and (C) EBOV L + nucleotides 1–10 of the le promoter (PDB: 8JSM) complexes. For a better visual of the RNA, only the RdRp domains of L are shown in cyan. The GDN motif within the RdRp is shown as red spheres. Promoter RNA is colored in orange, except for the initiation sites colored in black (position 3C for RSV le 1–10, 1U for RSV tr 1–10, and 2C for EBOV le 1–10; note that the EBOV le RNA used for this analysis has the sequence 3´ GCCUGUGUGU). The distance between the initiation site of the RNA and the aspartic acid in the GDN motif is shown as a yellow dotted line.

It is anticipated that in addition to a network of interactions to position the promoter RNA precisely, there are features that help stabilize the initiating NTPs (167). In other viruses, a key polymerase feature to stabilize the initiating NTPs is a priming residue, which is typically within a priming loop (167, 169–171). The function of a priming residue is to form base-stacking interactions with the initiating NTPs (169). As noted above, there are conformational differences between nsNSV polymerases. In addition to the large-scale differences in the arrangement of the globular domains that are described above, there are differences in the arrangement of features within the RdRp-capping domain core, including the putative priming loop. In the rhabdoviruses, RABV and VSV, the putative priming loop is extended into the active site cavity of the RdRp domain (Fig. 6A) (33–35). Consistent with this being a priming loop, a tryptophan at the tip is essential for initiation of RNA synthesis at the 3´ end of the template, but not from an internal site (51). The equivalent loop of the RSV L protein also contains amino acid residues required for RNA synthesis initiation (172), but in RSV polymerase structures, the loop is folded into the capping domain (Fig. 6B) (36, 37, 168, 173, 174). Direct evidence that this loop can adopt different conformations comes from comparison of the different EBOV L-VP35 structures. For example, in the EBOV L-VP35 apo structure and L-VP35-promoter structure, the priming loop is disordered, whereas in the EBOV L-VP35-template-primer structure, the priming loop is folded into the capping domain (43, 44). Together, these structures, combined with functional studies of the RSV and VSV L proteins (45, 47, 105, 173, 174), paint a picture in which during initiation of RNA synthesis, the priming loop extends into the RdRp active site to help buttress the 3´ end of the template and stabilize the initiating NTPs; then, as RNA synthesis proceeds, the priming loop retracts out of the active site and folds into the capping domain.

Fig 6.

The priming loops and channels of L. Structures of two domains are shown from the (A) VSV L (PDB: 6U1X) and (B and C) RSV L (PDB: 6PZK) proteins. The RdRp domains are colored in cyan, the PRNTase domains are colored in green, and in (C), the RSV phosphoprotein (P) is colored in purple. The GDN motif within the RdRp is shown as red spheres. The priming loop is colored in magenta, and the putative priming residues are shown as pink spheres (Trp1167 for VSV; Pro1261 and Trp1262 for RSV). (A) The Trp1167 residue on the VSV priming loop is located near the active site, suggesting an initiation conformation. (B) The RSV priming residues and priming loop are tucked away into the PRNTase domain, suggesting a non-initiation conformation. (C) A surface representation of the RSV L-P complex with a sliced cross-section (gray regions) is shown to indicate the putative channels for RNA templates, transcripts, and NTPs. The GDN motif in the RdRp active site and priming loop are overlaid onto the cross-section to show the rotation with respect to panel B. The nascent RNA transcript will exit toward the methyltransferase domain.

DO POLYMERASE STRUCTURES SHED LIGHT ON TRANSCRIPTION-REPLICATION INITIATION MODELS?

The experiments described above help us understand how the polymerase can initiate RNA synthesis at (or near) the 3´ end of the template, but does it help us distinguish between the different models for transcription initiation? Mutation of the VSV priming residue in L, Trp1167 (Fig. 6A), not only inhibited le(+) RNA synthesis but also inhibited downstream transcription (105). Likewise, mutation of the putative priming residues of the RSV L protein inhibited initiation at position 3C and downstream transcription, as did an inhibitor, MRK-1, that inhibited early elongation (172, 173). These data are consistent with the model that in VSV and RSV, transcription is initiated from the 3´ end of the le region, from position 1U in the case of VSV, and from position 3C in the case of RSV, although it should be noted that the mutations and MRK-1 inhibitor could have had a direct effect on initiation from the first gs signal, independently of an effect on initiation from the 3´ end of the genome.

Additional support for the model that transcription is initiated from the 3´ end of the le region comes from consideration of the channels within the polymerase. The polymerase has a template entrance and template exit channel that threads the template through the active site, and a transcript exit channel (Fig. 6C). The template channel that has been identified in the nsNSV polymerase structures runs through the interior of the polymerase (43, 45, 168). Although the possibility exists that the polymerase can “open up” to expose the template channel, allowing it to attach directly at an internal site on an RNA template, such a significant structural change would appear to be energetically unfavorable, suggesting that the polymerase is more likely to thread onto the 3´ end of the template like a bead being threaded onto a string. A mechanism by which the polymerase could thread onto the 3´ end of the template is suggested by structural analysis of measles virus N-RNA complexes. Specifically, it was shown that the three nucleotides at the 3´ end of the N-RNA template are exposed, as is a P binding site on the 3´ terminal N protomer (109). These findings suggest a model in which P (in association with L) could bind to the 3´ terminal N protomer (at a binding site that would be occluded at an internal N protomer) positioning L at the 3´ end of the template RNA (109).

If we accept that the polymerase threads on at the 3´ end of the template, a second question arises, namely, is the polymerase obliged to initiate le(+) RNA synthesis to begin transcription or could it thread onto the 3´ end and then scan along the template without synthesizing RNA and initiate transcription directly at the first gs signal. As noted above, the EBOV and RSV polymerases make multiple base-specific contacts with the promoter, resulting in a high-affinity interaction. Movement away from the promoter would necessitate breaking these bonds. In the case of DNA-dependent RNA polymerases, the energy required to break polymerase-promoter interactions is derived from the initiation of RNA synthesis and initial extension of the nascent RNA (175–178), and it seems likely that this is also the case for the nsNSV polymerases. Together, these observations suggest that the polymerase threads onto the 3´ end of the RNA template, forms specific interactions with the bases of the promoter sequence, and then needs to initiate RNA synthesis to break those interactions and move along the RNA template. This sequence of events is most consistent with a model in which the polymerase initiates both transcription and replication at the 3´ end of the le region (as in Fig. 4A and B), rather than initiating replication at the 3´ end of le and transcription directly at the first gs signal (as in Fig. 4C and D).

IS THERE AN EXPLANATION FOR VSV TRANSCRIPTION INITIATION AT THE FIRST GS SIGNAL?

With these biochemical and structural data providing strong support for transcription and replication being initiated from a common promoter element at the 3´ end of the le region, how can we explain the findings with the VSV polR1 mutant and the cell-based UV-inactivation experiment that strongly suggested that the polymerase can initiate transcription directly at the first gs signal? First, it should be recognized that there are technical explanations for these unusual findings. For example, it was shown that the polR1 mutant virus generates large quantities of le(+)-N readthrough transcripts that were aborted within the N gene, in addition to generating large amounts of N mRNA (164, 179). Given that nsNSV polymerases can scan the template in both directions after having released an RNA transcript, it seems plausible that the polR1 polymerase always initiates at the 3´ end of the le promoter but often continues RNA synthesis into the N gene; having released a le(+)-N abortive transcript, the polymerase could scan backward and reinitiate at the N gs signal. Indeed, the amounts of le(+)-N readthrough transcripts that were synthesized are consistent with this model (164). In the case of the experiments using UV-inactivated VSV, the experiments performed in vitro used detergent and salt-disrupted virions and likely assessed the initiation properties of polymerase that had been released from the nucleocapsid template under those buffer conditions; this free polymerase would likely have needed to thread onto the template and initiate at the 3´ end of the le (165). In contrast, experiments in which cells were infected with UV-treated virus and primary transcription was analyzed could have assessed nucleocapsid-associated polymerase (165). Earlier biochemical experiments indicated that virion-associated VSV nucleocapsids have polymerase associated with the N-RNA template able to initiate at internal gs signals (120). Thus, it is possible that during primary transcription (when the nucleocapsid is first released into the cytoplasm), the template-associated polymerase can start transcription from any of the gs signals but that newly synthesized, free polymerase generated during the course of infection needs to start transcription from the 3´ end of the le region.

Although the 3´ transcription-replication initiation model seems the most plausible based on polymerase structures and most biochemical and molecular biology studies, it is important not to ignore the possibility that the transcriptase can sometimes initiate RNA synthesis directly at the first gs signal. As noted above, there is strong evidence that nsNSV polymerases can scan intergenic regions. Paramyxovirus studies have indicated that scanning is not limited to the gene junction regions: in the case of SeV, the polymerase can scan 200 nt of non-specific sequence at the 3´ end of the template to find the promoter (180), suggesting that the polymerase can thread onto the template RNA in a conformation that allows it to scan. As noted above, several nsNSV polymerase structures show the priming loop in the retracted position, with the template exit channel open. It is possible that if a polymerase is in this conformation when it threads onto the 3´ end of the template, it could fail to form a specific interaction with the promoter, allowing it to scan to the first gs signal instead. If this were to happen, it could allow transcription to initiate directly at the first gs signal. It is possible that in the cell-based study with UV-inactivated VSV, the intracellular environment promoted the adoption of a scanning polymerase conformation, allowing efficient transcription initiation from the first gs signal.

CONCLUDING REMARKS AND FUTURE DIRECTIONS

The accumulated data regarding nsNSV transcription and replication initiation show that transcription can be initiated at the 3´ end of the le region, and this might be the dominant or even exclusive mechanism of transcription initiation for most nsNSVs. This model is consistent with the known polymerase structures, which suggest that the polymerase has to thread onto the 3´ end of the template RNA. However, there is tantalizing evidence suggesting that the polymerase can sometimes initiate directly at the first gs signal, without prior transcription of the le region, and it is possible that the polymerase can thread onto the template in a scanning conformation that will allow this to happen, at least occasionally.

Now that these findings are available as a foundation, the next step is to determine if there is a single pool of undifferentiated polymerase that is regulated between transcription and replication simply by the level of N⁰ and/or its initiation site or if template-free polymerase can adopt two distinct conformations as either a transcriptase or replicase. There are some recent findings that hint at a mechanism for how a polymerase could be differentiated into either a transcriptase or replicase. As noted above, several nsNSV polymerases have been shown to adopt a dimeric form (33, 34, 42, 54). Mutation analysis of the PIV-3 L protein dimer interface, coupled with complementation studies, revealed that the PIV-3 polymerase dimer is required for RNA replication (42). In this study, the role of the dimer in mRNA transcription was not determined. However, previous mutation analysis of the closely related SeV L protein showed that mutations that would be expected to disrupt the L-L dimer interface inhibited replication, without impeding le(+) synthesis or transcription (181). Together, these findings suggest a model in which monomeric polymerase serves as a transcriptase, whereas dimeric polymerase serves as a replicase. Additional studies are required to test this model. It is anticipated that future structure-function analyses of nsNSV polymerases will determine if nsNSV polymerases exist as a single undifferentiated polymerase or as two distinct pools of transcriptase and replicase in their template-free state.