Putative Domain-Domain Interactions in the Vesicular Stomatitis Virus L Polymerase Protein Appendage Region

ABSTRACT

The multidomain polymerase protein (L) of nonsegmented negative-strand (NNS) RNA viruses catalyzes transcription and replication of the virus genome. The N-terminal half of the protein forms a ring-like polymerase structure, while the C-terminal half encoding viral mRNA transcript modifications consists of a flexible appendage with three distinct globular domains. To gain insight into putative transient interactions between L domains during viral RNA synthesis, we exchanged each of the four distinct regions encompassing the appendage region of vesicular stomatitis virus (VSV) Indiana serotype L protein with their counterparts from VSV New Jersey and analyzed effects on virus polymerase activity in a minigenome system. The methyltransferase domain exchange yielded a fully active polymerase protein, which functioned as well as wild-type L in the context of a recombinant virus. Exchange of the downstream C-terminal nonconserved region abolished activity, but coexchanging it with the methyltransferase domain generated a polymerase favoring replicase over transcriptase activity, providing strong evidence of interaction between these two regions. Exchange of the capping enzyme domain or the adjacent nonconserved region thought to function as an “unstructured” linker also abrogated polymerase activity even when either domain was coexchanged with other appendage domains. Further probing of the putative linker segment using in-frame enhanced green fluorescent protein (EGFP) insertions similarly abrogated activity. We discuss the implications of these findings with regard to L protein appendage domain structure and putative domain-domain interactions required for polymerase function.

IMPORTANCE NNS viruses include many well-known human pathogens (e.g., rabies, measles, and Ebola viruses), as well as emerging viral threats (e.g., Nipah and Hendra viruses). These viruses all encode a large L polymerase protein similarly organized into multiple domains that work in concert to enable virus genome transcription and replication. But how the unique L protein carries out the multiplicity of individual steps in these two distinct processes is poorly understood. Using two different approaches, i.e., exchanging individual domains in the C-terminal appendage region of the protein between two closely related VSV serotypes and inserting unrelated protein domains, we shed light on requirements for domain-domain interactions and domain contiguity in polymerase function. These findings further our understanding of the conformational dynamics of NNS L polymerase proteins, which play an essential role in the pathogenic properties of these viruses and represent attractive targets for the development of antiviral measures.

INTRODUCTION

Much of what is known of nonsegmented negative-strand (NNS) virus RNA synthesis comes from studies with vesicular stomatitis virus (VSV), an important livestock pathogen that has long served as a model system for this diverse group of agents (1). The VSV negative-sense RNA genome encodes five genes in the order 3′-N-P-M-G-L-5′. The core of the virus particle consists of the N protein-encapsidated viral RNA (N-RNA) associated with the L-P polymerase complex, while the viral envelope incorporates M and G proteins. The L-P complex utilizes the N-RNA as a template and initiates transcription near its 3′ end to produce monocistronic transcripts for each gene (2, 3). L protein-catalyzed modifications of these transcripts take place during synthesis and include a unique GDP:polyribonucleotidyltransferase reaction that adds a 5′ cap (4), two biochemically distinct methyltransferase reactions that produce a cap 1 (m⁷GpppNm) structure (5–7), and 3′-end polyadenylation via reiterative copying of a U₇ stretch at gene ends (8). The L-P complex relies on conserved gene junction sequences as signals to carry out successive termination and reinitiation events as it progresses along the template (9). When sufficient amounts of N protein accumulate in infected cells, the virion complex switches to a replication mode and reads through all gene junctions to produce a positive-sense antigenome. This switch is coupled to encapsidation of nascent strands by N protein, and a similar coupling occurs during synthesis of progeny minus-sense genomes copied from encapsidated antigenomes. The underlying factors leading to the shift in the polymerase mode of synthesis are not well understood (10) although a tripartite L-P-N complex has been proposed to serve as a replication mode polymerase (11).

Extensive efforts to obtain high-resolution crystal structures of NNS L proteins, or parts there of, have so far proved unsuccessful (12, 13). Comparison of evolutionarily divergent NNS L proteins (14) has revealed partitioning of conserved sequences into six distinct blocks, conserved region I (CRI) to CRVI, postulated to be domains with specific functions (Fig. 1). CRIII contains a signature polymerase motif and has been confirmed to function as a polymerase domain (15–18). The novel capping enzyme unique to NNS viruses maps to CRV (termed the C domain) (4, 19), while the dual-specificity methyltransferase is encoded by CRVI (M domain) (6, 7, 20, 21). Functions of CRI, CRII, and CRIV have not been clearly established, but these segments appear to be part of the polymerase core ring structure (see below). Three large blocks of nonconserved sequence (∼200 to 300 residues long) are also found in NNS L proteins, positioned at the N- and C-terminal ends of the protein and between CRV and CRVI. The role of these nonconserved regions remains unclear.

FIG 1.

Schematic map of VSV IND L protein showing location of conserved sequence blocks (CRI through CRVI) and associated functions or characteristics. Amino acid coordinates of conserved blocks CRI, CRII, CRIII, and CRIV are as defined by Poch et al. (14), while those of CRV and CRVI are derived from Li et al. (19) and Bujnicki and Rychlewski (20), respectively. VSV NJ L protein segments swapped in place of their VSV IND L counterparts (denoted C, U, M, and T) and their positions are also indicated. The entire VSV NJ L protein sequence is 68% identical to VSV IND L, and positions of conserved and nonconserved regions are the same (39). Junction sites for the serotype swaps are predicted to have no disruptive effect on local secondary structure (Jpred, version 3, server).

Consistent with the notion of a multidomain protein, low-resolution electron microscopy (EM) of the purified VSV L protein has shown that the N-terminal half (∼1,100 residues) forms a ring-like structure similar in dimensions to other known viral RNA polymerases, while the remaining C-terminal region forms an appendage comprised of three adjoining globules of similar dimensions (22). EM imaging of deletion constructs has implicated the capping enzyme (CRV), the methyltransferase (CRVI), and the nonconserved C-terminal region (T domain) as globule-forming regions. The large nonconserved segment between CRV and CRVI was deemed not visible by EM and proposed to act as an “unstructured” linker region (U domain). Notably, the three adjoining globules of the appendage showed no fixed orientation relative to each other, implying freedom of movement and the presence of intervening hinges. The flexible nature of the appendage was all the more apparent in the presence of the P protein, which favored a major rearrangement of the L appendage into a flap structure around the core polymerase, with no clear demarcation between globules (22, 23). The location of the P protein multimer associated with L, however, was not evident from EM analysis.

The presence of at least one hinge region in the L protein was first documented for morbilliviruses on the basis of retention of polymerase activity and virus infectivity following an in-frame enhanced green fluorescent protein (EGFP) insertion just upstream of the methyltransferase domain (24–26). We observed the same phenomenon for the VSV L protein (27), while other groups of investigators extended these findings to additional paramyxoviruses (28, 29) as well Ebola virus (30). Although it seems likely that other regions of the L protein appendage also function as hinges to allow appendage domain movement, these may not be tolerant of large insertions and have yet to be identified.

All of the above supports the view that L operates as a dynamic protein machine with domains interacting in an orchestrated fashion. Notably, a recent study showed that paramyxovirus L protein constructs encoding the region upstream and downstream of the hinge region could not reconstitute polymerase activity when they were mixed together unless first fused to dimerization tags (29). No reconstitution of activity was observed when the two L fragments originated from different paramyxoviruses. These findings provide clear evidence of specific and low-affinity interactions between these two regions of L.

We set out here to explore putative domain-domain interactions in the VSV L protein by swapping segments of the VSV Indiana serotype (VSV IND) L protein appendage with their counterparts from the closely related VSV New Jersey serotype (VSV NJ). Our exchanges aimed to preserve the integrity of the independently folding globules so that disruptive effects on polymerase activity would more likely reflect loss of essential interactions between appendage domains and/or other components of the polymerase machinery.

MATERIALS AND METHODS

Cell culture and viruses.

Baby hamster kidney cells (BHK-21) and HeLa cells were grown as monolayers in minimal essential medium (MEM) containing 7% newborn calf serum. A recombinant IND VSV with an L protein carrying the M domain of NJ virus [VSV-L(M_NJ)] was derived from the recombinant VSV IND wild-type virus (VSV wt) described by Lawson et al. (31) and constructed as outlined below.

Plasmid constructions.

Plasmids encoding VSV NJ (Ogden subtype) L protein (pBluescript II KS/NJ-OL), VSV NJ (Hazelhurst subtype) P protein (pBluescript II KS/NJ-HP), and VSV NJ (Hazelhurst) N protein (pBluescript II KS/NJ-HN) were kindly provided to us by C. Yong Kang. Plasmids encoding the VSV IND L protein harboring exchanged segments from the appendage region of the VSV NJ (Ogden) were engineered by modifying the recipient pGEM-L expression plasmid encoding the VSV IND (Mudd-Summers strain) L protein with a His-hemagglutinin (HA) tag at its N-terminal end, as described in our previous studies (27). VSV NJ L gene donor fragments were obtained by PCR amplification or restriction enzyme digests of the pBluescript II KS/NJ-OL plasmid. Our previously constructed pGEM-L_EGFP plasmid was used as a source of EGFP for in-frame insertion in the nonconserved region between CRV and CRVI. For some constructions, we made use of the pGEM-L_D1595VT construct (containing a unique BstEII site in the hinge region of the L protein, resulting in the replacement of the D1595 with valine and threonine) also engineered in our previous study (27), as well as a pGEM-LΔNsiI construct lacking the NsiI fragment that encodes most of the L protein N-terminal half.

Alignment between the IND and NJ L protein sequences (GenBank accession numbers Q98776 and P16379, respectively) was based on Clustal W and is available from the authors upon request. The capping enzyme and methyltransferase domain borders (CRV and CRVI) were derived from the sequence alignments in Li et al. (19) and Bujnicki and Rychlewski (20), respectively. Junction site positions for the various domain exchanges are noted in Fig. 1.

A variety of strategies were employed to generate single- and multiple-domain exchanges as well as EGFP insertions (L-EGFP variants), including the circular polymerase extension cloning (CPEC) method described by Quan and Tian (32) using Phusion High Fidelity polymerase (New England BioLabs), PCR fragment amplification of selected fragments using Pfu Ultra or Pfu Turbo (both from Stratagene) with addition of appropriate restriction enzyme sites on primers, introduction of silent restriction sites in plasmid templates using QuickChange site-directed mutagenesis (Stratagene), and standard cloning or exchange of fragments into recipient plasmids using restriction sites. Primer sequences and details of each construction are available from the authors upon request. All sequences generated by PCR were verified in the context of the final plasmid construct for the single-region domain exchanges and EGFP insertions. The construct containing an EGFP insertion at position 1472 anomalously contained two juxtaposed in-frame duplications of an 18-amino-acid (aa) stretch of the original L protein sequence (residues 1466 to 1484) 12 residues downstream of the EGFP insertion site. Multidomain exchanges, as well as introduction of the L(M_NJ) exchange into the full-size VSV genome construct, used standard cloning of fragments from the sequence-verified plasmids using restriction enzyme digestion.

Minigenome assays.

Our previously described minigenome assay (27) was modified as follows. Briefly, BHK cell monolayers were grown to ∼75% confluence in 6-cm dishes and cotransfected with a mixture containing VSV IND wt or mutant L plasmids and VSV IND P and N plasmids 18 h prior to vaccinia-T7 virus infection. Plasmid amounts were identical to those we used previously unless otherwise noted. Transfection was carried out using 8 μl of a 7.5 mM solution of polyethyleneimine ([PEI] PEI Max, catalog no. 24765; Polysciences, Inc.) per μg of DNA mixed with 0.4 ml of MEM without serum for 20 min before addition to cell monolayers in 4 ml of fresh MEM plus serum. A constant PEI/DNA ratio was maintained when plasmid amounts were varied by the addition of salmon sperm DNA. For measuring wt NJ L protein activity, cotransfection was carried out using pBluescript II KS/NJ-OL, pBluescript II KS/NJ-HP, and pBluescript II KS/NJ-HN as described above. Transfection supernatants were aspirated, and cell monolayers were washed twice with saline immediately prior to vaccinia-T7 virus infection as described previously (27). Monolayers were then incubated at 37°C for 5 h before infection with minigenome particles, at which point the incubation temperature was lowered to 33°C until harvest of cytoplasmic extracts at 20 h after vaccinia-T7 infection. ³²P-radiolabeled M probes for Northern blots were prepared as described before (27), and bands were quantified using a Typhoon phosphorimager (GE Healthcare). Contrast and brightness adjustments were applied equally to PowerPoint images of samples originating from the same blot.

Western blotting.

Transfer from standard SDS–10% PAGE gels to polyvinylidene difluoride (PVDF) membranes (GE Healthcare) was carried out as described before (27). Membranes were blocked with 1% bovine serum albumin. Anti-HA antibody (Sigma) and horseradish peroxidase-conjugated anti-mouse antibody (Santa Cruz Biotechnology) were used at dilutions of 1:10,000. Antibody detection was carried out using ECL+ chemiluminescent substrate (Thermo Scientific Pierce) and film (GE Healthcare).

Recovery of recombinant IND VSV-L(M_NJ) virus.

Recovery of infectious virus was carried out as previously described (27), except that PEI was used as described above. In this case, support plasmids pBS-N (0.6 μg), pBS-P (1.4 μg), pBS-L (0.9 μg), and the full-genome template plasmid (1.1 μg) were transfected immediately following infection with vaccinia-T7 virus. Successful virus recovery was judged by cytopathic effect and confirmed by plaque assay on BHK cells. Reverse transcription-PCR (RT-PCR) using AccuScript reverse transcriptase (Agilent Technologies) was carried out on infected cell extracts to confirm the presence of the NJ methyltransferase coding sequence in the viral genome. A high-titer working stock of the recovered VSV-L(M_NJ) virus was amplified from a single plaque on BHK cells.

Virus growth assays.

Growth kinetics of IND VSV-L(M_NJ) and VSV wt virus in BHK cells at 37°C were determined in parallel. Monolayers were grown to 70% confluence in 6-cm plates and infected in duplicates at a multiplicity of infection (MOI) of 10. Following virus adsorption in 0.4 ml of MEM for 1 h, 4 ml of fresh MEM was added, and aliquots were collected at the times indicated in Fig. 3 to determine titers (PFU/ml) by plaque assay on BHK cells. Yield determinations in HeLa cells (obtained from Ralph Feuer and representing the same source as HeLa/M cells utilized by Simpson and colleagues [33]) were carried out similarly except that single monolayers grown in 25-cm² flasks were used.

FIG 3.

Growth properties of VSV-L(M_NJ) virus compared to that of isogenic VSV wt in BHK cells and HeLa cells. Average PFU yields from duplicate monolayer infections (MOI of 10) are plotted in panel A (titer differences between duplicates of <20% at each time point). Yields from single monolayer infections are shown in panel B, and similar results were obtained in a separate independent determination. p.i. postinfection.

RESULTS

Validation of modified minigenome assay for quantifying polymerase activity of VSV L protein constructs.

Essential features of the minigenome assay have been described in our previous work (27). In brief, T7 promoter-driven plasmids encoding VSV L, P, and N proteins are transfected into vaccinia-T7 virus-infected BHK cells, followed by infection with a fully assembled VSV particle containing a minigenome template. The latter contains all the RNA synthesis control elements found in the standard virus genome but encodes only the virus M and G genes (34). Replication and transcription of the minigenome template rely entirely on plasmid-expressed viral proteins, and the two processes are then quantified by Northern blotting using plus-sense and minus-sense M gene probes, respectively. Note that transcription in this assay is detectable only if replication occurs first as the latter is necessary to provide sufficient amounts of transcription templates.

We previously showed that insertion of the EGFP coding sequence in the VSV IND L protein upstream of the methyltransferase domain results in temperature-sensitive polymerase activity (27). To maximize detection of polymerase activity in the present study and ensure valid comparisons between samples potentially containing different amounts of L proteins, we modified a number of minigenome assay parameters, including lowering the temperature to 33°C (see Materials and Methods). The sensitivity and L plasmid dose response of the modified assay showed that both transcription and replication activities were detectable at the smallest amount of input L plasmid (250 ng), but the dose response for replication was sharper, reaching a maximum with 3 μg of L plasmid in contrast to 7 μg for transcription (data not shown). Unless otherwise noted, all experiments reported here utilized 3 μg of L plasmid (peak replication) and optimum amounts of P (0.25 μg) and N (6 μg) plasmids as determined previously (27).

Effect of swapping individual VSV IND appendage domains with their VSV NJ counterparts on virus polymerase activity.

The C-terminal half of the VSV L protein comprises four well-defined regions in the following order: capping enzyme (CRV), nonconserved unstructured region, methyltransferase enzyme (CRVI), and nonconserved C-terminal region (Fig. 1). For simplicity, we will refer to each of these segments as domains, abbreviated C, U, M, and T, respectively. When IND and NJ L proteins are compared, these domains display 71%, 67%, 66%, and 42% amino acid identity, respectively.

We constructed a series of T7 polymerase-driven L protein expression plasmids, with each containing one of the above NJ-specific domains within the backbone of the IND L protein (Fig. 1). The boundaries we chose for the exchanges aimed to preserve intradomain structures, and, in most cases, the junctions coincided closely with the already well-established L protein sequence alignments delineating conserved and nonconserved regions (14, 19, 20). The C domain segment initiated with the first residue of CRV (1070) and terminated six residues downstream of its C-terminal end (position 1311). The M domain segment started at the hinge site just upstream of CRVI (position 1595) and ended 20 residues downstream (1863). The C and M domain swaps thus included all known motifs involved in capping and methyltransferase activities, respectively. The 1863 junction site was also utilized for the T domain segment exchange. The U domain segment swap initiated within the second-from-last motif of the capping enzyme domain, corresponding to a perfectly conserved 8-amino-acid stretch between IND and NJ L proteins (position 1293), and ended within a perfectly conserved stretch of 12 amino acids overlapping the start of CRVI (position 1646). As discussed below, our findings strongly support the view that the exchange junctions described above did not disrupt L protein intradomain innate structures, except in the case of the C domain swap.

Figure 2 documents the results of several independent minigenome assays comparing transcription and replication activities of wt L and chimeric L proteins harboring single domain swaps. All exchange constructs were tested in parallel with wt L in at least two independent experiments to ensure consistency and reproducibility. The C, U, and T segment exchanges abolished all polymerase activity, but the M domain swap consistently yielded transcription and replication activities ranging from slightly lower than wt IND L protein to slightly higher, as in the minigenome results shown in Fig. 2A. Lack of polymerase activity for the C, U, and T domain exchanges was clearly not due to defective NJ donor segments or to incompatibility between these segments and the IND minigenome template since the parent NJ L protein was fully active with this template in the presence of NJ P and NJ N proteins (Fig. 2A, top right panel, third lane).

FIG 2.

Effect of swapping individual L protein appendage domains on polymerase activity in the minigenome system. C, U, M, and T domains from VSV NJ were engineered in place of their VSV IND homologues (Fig. 1). (A) Each Northern or Western blot panel displays the results of a representative experiment comparing single-domain swaps to wt L and no-L controls. Transcription blots were first stained with methylene blue before being probed to serve as a loading control and allow normalization of transcription values. The plasmid encoding the wt VSV NJ L protein (top right, _NJ L) was assayed in the presence of plasmids encoding NJ N and P proteins. (B) U domain swap assayed with a range of IND P or NJ P plasmid inputs (top panel) and a range of L plasmid inputs with 250 ng of IND P plasmid (bottom panel). (C) T subdomain swaps (see the text). All L exchange constructs, except for L(M_NJ), yielded transcription and replication values of <1% of wt L.

Significantly, all chimeric L constructs, including the active M domain swap, generated about 2-fold less L protein than the wt (Fig. 2A, lower panels), which not only rules out insufficient L protein expression for lack of activity observed for the C, U, and T domain swaps but also suggests that all L constructs also bound P protein since the latter is required to stabilize L against degradation in transfected cells (35). Under the minigenome assay conditions used here, one would expect very little, if any, L protein accumulation in the absence of P protein binding. This was confirmed here for the U-swapped construct where no significant amount of the chimeric L protein accumulated in the presence of NJ P in place of IND P, and only trace amounts were observed in the presence of both P proteins (Fig. 2B, upper panel). Moreover, increased levels of IND P plasmid had no significant effect on chimeric L protein accumulation (Fig. 2B, upper panels). We infer from these results that NJ P protein, in contrast to IND P, could not stabilize the chimeric L from degradation and that its presence inhibited binding of IND P to L. These findings indicate that lack of IND P protein binding per se did not account for the absence of polymerase activity observed for the C, U, and T domain swaps.

Given that the T domain sequence is the least conserved segment among NNS virus L proteins (42% identity for VSV IND versus VSV NJ), we were somewhat surprised that its exchange abolished polymerase activity. The same junction (residue 1863) was utilized for both M and T domain swaps, arguing against disruption of T intradomain structure as the cause for inactivity. To determine if the loss was due to a specific portion of the exchanged T domain, we swapped its first 97 residues (T1 subregion) and the remaining 150 residues (T2 subregion) separately. Neither the T1 nor the T2 exchange yielded an active polymerase (Fig. 2C), suggesting that the defect involves the T domain as a whole. As in the case of the U domain swap, adding NJ P or substituting NJ P for IND P for the T domain swap did not rescue polymerase activity, and addition of NJ N in the presence of either NJ P or IND P likewise did not restore activity (not shown).

Growth properties of a recombinant virus encoding the NJ methyltransferase domain.

Although the M domain swap generated a fully competent polymerase in the minigenome assay, methylation of viral transcripts is not required for VSV transcription in vitro, and, moreover, the vaccinia-T7 virus employed in this assay encodes trans-acting cap methyltransferase activity. To address whether the exchanged M domain retained catalytic activity, we recovered a recombinant virus, VSV-L(M_NJ), encoding the chimeric L protein in place of IND wt L. As shown in Fig. 3A, the kinetics and yield of VSV-L(M_NJ) growth in BHK cells at 37°C was nearly identical to those of the isogenic wt virus. These results provide strong evidence for retention of viral methyltransferase activity since previous studies have shown that recombinant VSVs deficient in this activity in vitro, even partially so, are growth impaired in BHK cells at 37°C (7). Although earlier studies have shown that the methyltransferase-null VSV host range (hr) mutant hr1 grows to high titers in BHK cells at 35°C (presumably compensated by a BHK host cell activity), it does so with a very significant delay (36), and the recombinant VSV carrying the mutation responsible for the hr1 phenotype (D1671V) displays nearly a 3-log decrease in virus titer at 37°C in BHK cells (7). To provide additional evidence that the exchanged M domain retained catalytic activity, we tested VSV-L(M_NJ) growth in HeLa cells, which are nonpermissive for VSV hr1 growth even at 35°C (33). VSV-L(M_NJ) generated virus titers very close to the wt virus titer in HeLa cells at 37°C (Fig. 3B). These results clearly show that the NJ M domain, in the context of the chimeric L protein, retained sufficient methyltransferase activity for normal or nearly normal virus growth.

Effect of swapping the VSV IND methyltransferase domain with that of Ebola virus on polymerase activity.

We next explored whether swapping the IND M domain with the more evolutionarily distant M domain from Ebola virus L (22% identity) would yield an active polymerase. The resulting chimeric protein accumulated to the same extent as wt IND L, again implying unaltered P protein binding, but displayed no polymerase activity in the minigenome assay (data not shown). We presume that the exchanged M domain in this case did not retain catalytic activity since this function depends on recognizing specific mRNA start sequences that differ substantially between VSV and Ebola. Since this activity is presumably not required for polymerase function in the minigenome system, the loss in this case may be due to disruption of L protein structure and/or loss of critical interaction between the M domain and some other component of the polymerase complex.

Effect of coswapping multiple domains of the L protein appendage.

To test whether inactivity of the C, U, and T domain swaps might be due to loss of domain-domain interactions specifically within the appendage, we constructed chimeric L proteins harboring various combinations of exchanged domains. These compound swaps included the entire appendage (CUMT), two triple-domain exchanges (CMT and UMT), and five double-domain exchanges (CU, CM, CT, UT, and MT). All coexchange constructs generated readily detectable amounts of L proteins, ranging from wt levels to somewhat lower, as seen for the single-domain exchanges and again indicating no major impairment in IND P protein binding (Fig. 4). Importantly, all compound swaps failed to show any polymerase activity except for one telling exception, the M and T domain double exchange [Fig. 4, (MT)_NJ]. Intriguingly, the chimeric MT L protein rescued replication to a greater extent than transcription (67% ± 8.7% and 19% ± 5.3% of wt L, respectively, with the latter value normalized to template amounts). Furthermore, no rescue was observed when only the T2 half of the NJ terminal domain was coexchanged with the NJ M domain [Fig. 4, (MT2)_NJ], suggesting that rescue of activity requires the T region as a whole. Since exchange of the T domain on its own abolished polymerase activity but coexchange of M and T domains rescued activity, our findings imply that the M and T domains of the L appendage must functionally interact with each other to enable polymerase activity (see Discussion).

FIG 4.

Effect of swapping L protein appendage domain combinations on polymerase activity in the minigenome system. Each Northern or Western blot panel shows the results of a representative experiment. All domain swap combinations yielded no significant activity compared to wt L (<1%), except for the MT swap. The latter displayed 67% ± 8.7% (mean ± standard deviation; n = 4) of wt L replication activity and 13% ± 5.3% (n = 6) of wt L transcription activity (19% ± 5.3% when normalized for template amounts). L protein accumulation for the MT2 combination swap (data not shown) was verified in an independent experiment.

The effects on polymerase activity for all single-domain or codomain exchanges are summarized in Fig. 5. Notably, all swaps involving the C domain, including one containing all four segments of the appendage, generated inactive polymerases. Likewise, all exchanges involving the U domain, including the UMT combination, also abolished polymerase activity. Implications of these results are discussed below.

FIG 5.

Graphic summary showing all single and combination domain exchanges tested in this study and their effects on polymerase activity. The appendage domain schematic in the upper left is a rendering of the low-resolution EM structure reported by Rahmeh et al. (22). Polymerase activity is indicated as follows: −, activity abolished; +, transcription and replication activities comparable to those of the wt L; ±, partial activity favoring replication (see the text).

Effect of in-frame EGFP insertions in the unstructured domain.

The U domain was deemed invisible by EM analysis and proposed to function as a large unstructured linker region (22). Assuming that this role would not depend on strict sequence conservation, the loss of polymerase activity we observed when the U domain was swapped by itself was unexpected. Moreover, our previous findings had shown that the C-terminal end of the U domain does behave like a linker region since polymerase activity was retained following in-frame EGFP insertion at residue 1595 (27). To test whether the bulk of the U region might nonetheless function as a linker, we engineered five additional L protein constructs, each containing an in-frame EGFP insertion at different positions spanning the U segment (positions 1318, 1374, 1472, 1522, and 1577) and assessed effects on polymerase activity along with the wt and the previous construct with the EGFP insertion at position 1595.

Surprisingly, none of the newly constructed L-EGFP variants displayed any polymerase activity (Fig. 6). As noted previously for the 1595 construct, all L-EGFP proteins accumulated to levels somewhat lower than the wt, but the reduction was particularly dramatic for the 1522 mutant, which reproducibly accumulated about 10-fold less L protein (Fig. 6), an amount nonetheless sufficient for concluding lack of significant polymerase activity. Increasing input L plasmid amounts from 3 μg to 7 μg for the 1522 insertion mutant resulted in no significant increase in L protein accumulation (data not shown). The decrease in L protein expression for the insertion mutants also correlated with the EGFP fluorescence intensity displayed by transfected cells in the minigenome assays (data not shown). These findings indicate that the integrity of the U domain is important for polymerase function and that it may play some role in P protein binding.

FIG 6.

Effects of in-frame EGFP insertions within the unstructured domain of the VSV IND L protein on polymerase activity in the minigenome assay. Insertion positions are indicated on the top left. Our previous work showed that the L protein with the 1595 insertion retains full polymerase activity in a minigenome assay similar to that employed here (27). Solid lines at the top of adjacent lanes indicate samples from duplicate minigenome transfections. The same extracts were utilized for L protein accumulation and transcription analysis. None of the constructs with inserts upstream of position 1595 yielded significant polymerase activity.

DISCUSSION

Recent EM analysis has made it clear that domains within the L protein appendage region fold independently into well-demarcated globules separated by hinges that allow flexibility of movement (22). The hinges are likely necessary to allow transient interactions between appendage domains encoding mRNA capping reactions and the core polymerase as both cap addition and cap methylation take place on short nascent chains (37). The work presented here was based on the premise that the native globular structure and function of these domains would be retained when they were swapped between the two closely related VSV NJ and VSV IND serotypes while transient domain-domain interactions might be affected by sequence differences between the two.

We exchanged each of the four distinct regions (each ∼300 residues long) that comprise the C-terminal half of the VSV IND L proteins, either singly or in combination, for their counterparts from VSV NJ and determined effects on polymerase activity in a minigenome system. Interestingly, none of the domain exchange constructs appeared to be defective in their ability to form a complex with IND P protein as this binding is required to protect L protein against degradation in the minigenome system (35) (Fig. 2B).

The L protein with the M domain substitution retained replicase and transcriptase activity essentially equivalent to that of the wt IND L protein (Fig. 2). We infer that while the IND and NJ M domains are only 66% identical in sequence, the differences had no disruptive effect on IND L polymerase function, including any putative L domain-domain interactions or interactions with IND P and IND N proteins required for this activity. To address whether the exchanged M domain retained methyltransferase activity, which is not required for polymerase activity in the minigenome system, we recovered a recombinant virus encoding the chimeric L protein. The growth properties of the recombinant VSV-L(M_NJ) virus in BHK and HeLa cells at 37°C cells were indistinguishable from those of the wt virus (Fig. 3), in contrast to all previously characterized methyltransferase-deficient VSV mutants, including the VSV hr1 mutant, which show severe growth restriction under these conditions (7, 33, 36). We therefore conclude that the swapped NJ M domain retained a level of methyltransferase activity close, if not identical, to that of the wt L although future in vitro studies could conceivably uncover more subtle differences. Perhaps not surprisingly, exchanging the IND M domain for the more distant Ebola virus M domain (22% identity) abrogated polymerase activity.

Although the T domain has no known function and is the least conserved appendage region among NNS L proteins (42% identity for VSV IND versus NJ), its exchange abolished polymerase activity (Fig. 2). Prior EM analysis leaves little doubt that the T domain corresponds to the last globule of the appendage (22). In line with the concept of a single functional unit, swapping of either the first ∼40% or last ∼60% of the T segment also abrogated polymerase activity. Interestingly, the analogous T domain of paramyxovirus L proteins bears a motif associated with cellular capping enzymes (38).

Loss of polymerase activity for any of the constructs tested here, including the T domain swap, could in theory result from one or more of the following possibilities: (i) disruption of individual domain folding due to inappropriate segment exchange junctions, (ii) a requirement for interaction of the exchanged NJ domain with homologous NJ P and/or N protein, and (iii) disruption of a required interaction between individual L protein domains. For the T domain swap, the first two possibilities seem highly unlikely. The same junction (20 residues downstream of CRVI) was utilized to generate the active M domain swap and the inactive T domain swap, while coexchange of both domains produced an active polymerase. Moreover, coexchange of M and the T2 domain subregion still abrogated polymerase activity. The T swap construct retained its ability to bind IND P, and addition of NJ P and/or NJ N protein had no discernible rescuing effect on polymerase activity. Taken together, these findings provide very strong evidence that L polymerase activity requires a functional interaction between M and T domains, presumably reflecting a physical interaction. We postulate that this domain-domain interaction is disrupted in the presence of IND M and NJ T domains (but not the converse) as a result of sequence differences between the T domains of the two serotypes. This of course does not preclude possible interactions of M and/or T domains with other components of the polymerase complex that remain unaffected by sequence differences between IND and NJ serotypes.

Interestingly, coexchange of M and T domains favored replication rather than transcription (∼67% and ∼19% of wt L activity, respectively). This bias could conceivably reflect differences in interactions between NJ M and T domains with IND P and N proteins compared to homotypic wt L interactions. Alternatively, M and/or T domain interactions with other domains of L protein might play a role in distinguishing replicase from transcriptase forms of the protein although many other factors are undoubtedly involved in this distinction. While swapping the T domain on its own clearly blocked all replicase activity in the minigenome system, it remains unclear whether the swap also affected transcriptase activity since the latter depends on amplification of templates through replication.

Notably, swapping the C domain by itself or coswapping it with other appendage domains abrogated all polymerase activity (Fig. 5). In this case, we have reason to believe that the swap may well have disrupted proper folding of the C domain or the adjacent polymerase ring structure or both. Based on prior EM analysis, the L protein construct consisting of residues 1 to 1593, which includes the entire C and U regions, displayed the core ring structure and the proximal globule (Fig. 1 shows domain coordinates), whereas the construct consisting of residues 1 to 1114, which includes the first 44 residues of the C domain, showed no globules but a less compact conformation of the polymerase ring structure (22). This suggests that the C domain itself could be part of the ring structure, in which case the U domain would correspond to the core-proximal globule. This alternative hypothesis does away with the need to propose an unstructured and invisible U region and is consistent with our findings. Nonetheless, it is conceivable that our choice of junction between the C region and the core polymerase (first residue of CRV) had no disruptive effect on folding of either domain but, instead, prevented one or more specific interactions between the C domain and some other component of the polymerase complex.

Intriguingly, swapping of the U domain by itself or in combination with any other appendage regions also inactivated polymerase activity while preserving the ability to bind IND P protein (Fig. 5). In this instance, the downstream exchange junction site (position 1646) was localized within a completely conserved region overlapping the start of the M domain. This exchange junction clearly had no disruptive effect on either M domain folding or putative U domain folding since utilization of the upstream hinge site (position 1595) for the M-only swap and coexchange of M and T domains yielded active polymerase proteins. Similarly, it seems likely that the upstream exchange junction for the U domain swap also had no disruptive effect on C and/or U domain folding, given its location within a perfectly conserved 8-amino-acid stretch harboring the second-to-last motif of the capping enzyme. We therefore suggest that the lack of polymerase activity observed for the U domain swap reflects its inability to properly interact with other L protein domains and/or other component(s) of the polymerase complex. Significantly, coswapping of the U domain along with M and T domains also abolished polymerase activity, suggesting that the postulated disrupted U domain interaction involves a component other than the M and T domains. As mentioned above in the context of the T domain swap, our minigenome assay does not address whether any of the U-swapped constructs retained transcriptase activity.

We tested here whether the entire U region behaves as an unstructured and flexible linker, as previously proposed (22), by constructing in-frame EGFP insertions at different positions spanning the region. All EGFP insertions upstream of the previously identified hinge site abrogated polymerase activity (Fig. 6). The integrity of the U domain as a contiguous region thus appears essential for polymerase activity, in agreement with a recent study showing that 10-residue linker insertions within the analogous U region of the measles virus L protein also abrogated polymerase activity in a minigenome system (29). Taken together, these findings strongly suggest that the bulk of the U domain does not simply represent an unstructured region between C and M domains but more likely serves some other role in polymerase function. The linker function appears to be confined to a relatively small region (∼50 residues or less) at the C-terminal end of the U segment.

Curiously, an EGFP insertion at position 1522 of the U region uniquely showed a drastic reduction in mutant L protein accumulation (about 10-fold) (Fig. 6), suggesting a significant impairment in P protein binding. This phenomenon, however, was much less pronounced for the other EGFP inserts. Interestingly, Rahmeh and colleagues (22) showed that native P protein binding to the VSV L protein is severely impaired when a mixture of two L fragments corresponding to the regions upstream and downstream of the hinge site (residues 1 to 1593 and 1594 to 2109) is used. In addition, a recent study employing the equivalent fragments from the measles virus L protein demonstrated that polymerase activity depends on a low-affinity but specific interaction between these two regions of L (29). Since a collapsed form of the VSV L protein appendage is favored in the presence of P protein (22), it is tempting to speculate that the U region, in concert with P protein binding, facilitates hinge-dependent interactions of the M and T domains with upstream regions of L. This global rearrangement of the appendage region, which is likely crucial for polymerase function, could also involve specific interactions between the U domain and other domains of L.

In summary, our work here sheds light on VSV L protein appendage domain function and specific domain-domain interactions required for virus polymerase function. Similar approaches using other NNS viruses may offer additional insights into the structure and function of these complex proteins.