Structure and Sequence Requirements for RNA Capping at the Venezuelan Equine Encephalitis Virus RNA 5′ End

ABSTRACT

Venezuelan equine encephalitis virus (VEEV) is a reemerging arthropod-borne virus causing encephalitis in humans and domesticated animals. VEEV possesses a positive single-stranded RNA genome capped at its 5′ end. The capping process is performed by the nonstructural protein nsP1, which bears methyl and guanylyltransferase activities. The capping reaction starts with the methylation of GTP. The generated m⁷GTP is complexed to the enzyme to form an m⁷GMP-nsP1 covalent intermediate. The m⁷GMP is then transferred onto the 5′-diphosphate end of the viral RNA. Here, we explore the specificities of the acceptor substrate in terms of length, RNA secondary structure, and/or sequence. Any diphosphate nucleosides but GDP can serve as acceptors of the m⁷GMP to yield m⁷GpppA, m⁷GpppC, or m⁷GpppU. We show that capping is more efficient on small RNA molecules, whereas RNAs longer than 130 nucleotides are barely capped by the enzyme. The structure and sequence of the short, conserved stem-loop, downstream to the cap, is an essential regulatory element for the capping process.

IMPORTANCE The emergence, reemergence, and expansion of alphaviruses (genus of the family Togaviridae) are a serious public health and epizootic threat. Venezuelan equine encephalitis virus (VEEV) causes encephalitis in human and domesticated animals, with a mortality rate reaching 80% in horses. To date, no efficient vaccine or safe antivirals are available for human use. VEEV nonstructural protein 1 (nsP1) is the viral capping enzyme characteristic of the Alphavirus genus. nsP1 catalyzes methyltransferase and guanylyltransferase reactions, representing a good therapeutic target. In the present report, we provide insights into the molecular features and specificities of the cap acceptor substrate for the guanylylation reaction.

INTRODUCTION

Venezuelan equine encephalitis virus (VEEV) is an encephalitic pathogen transmitted by hematophagous arthropods. The virus circulates throughout the Americas and causes regular outbreaks affecting humans and equids. In horses, the mortality rate is high and can reach up to 80%. In humans, even though the fatality rate is low (<1%), up to 14% of infected individuals can develop long-lasting and even permanent neurological sequelae (1). Over the last few decades, much effort has been devoted to vaccine development. The strategies adopted include the use of live-attenuated, inactivated, chimeric, and various subunit vaccine candidates. While the live-attenuated vaccine, TC-83, is commonly used in Mexico and Colombia, there is still no approved vaccine for human use, urging the development of antivirals (2).

VEEV is a member of the Alphavirus genus, which contains 31 species and belongs to the Togaviridae family. VEEV possesses a single-stranded positive-sense RNA of approximately 11.5 kb that is polyadenylated and contains a 7-methylguanosine (m⁷G) cap structure (cap 0, m⁷GpppN, where N represents the first transcribed nucleotide of the viral RNA). The genome contains two open reading frames (ORFs) and acts as an mRNA after cell entry. The first ORF leads to the synthesis of nonstructural polyprotein precursors, named P123 and P1234, which are ultimately auto-processed to nsP1 to nsP4, forming the membrane-associated replication/transcription complex in virus-driven lipid structures called spherules. The second ORF encodes the structural polyprotein precursor and is translated from the subgenomic (SG) RNA. This polyprotein is ultimately cleaved to yield E1, E2, E3, C, 6K, and TF structural proteins that drive the formation of capsid shells and glycoprotein spikes, directing the packaging of new viral genomes (3, 4). The capping of both genomic and subgenomic RNAs is mediated by the nsP1, harboring both methyltransferase and guanylyltransferase activities. Jones et al. have recently shown that nsP1 can form membranous dodecamer pore complexes, ensuring the transit between the cytoplasm and the spherule contents (5). The reaction catalyzed by nsP1 involves the following three steps: (i) in the presence of S-adenosylmethionine (SAM), GTP is methylated by the methyltransferase activity (MTase) of nsP1 to form m⁷GTP and S-adenosylhomocysteine (SAH); (ii) nsP1 guanylyltransferase activity (GTase1) allows the formation a covalent m⁷GMP-nsP1 intermediate, releasing inorganic pyrophosphate (PPi); and (iii) the methylated guanosine is transferred to the ppRNA, whose γ-phosphate is previously removed by the nsP2 protein (6–12). This results in the formation of a type 0 cap structure, methylated at the N7 position only. While the catalytic amino acids in nsP1 have been relatively well characterized, specifics of the RNA substrate required for efficient capping remain poorly described.

It has previously been demonstrated that specific secondary structures present at the 5′ region of the viral RNA are essential to the virus. These structures, known as conserved sequence elements (CSE), can greatly vary from one alphavirus species to another, suggesting that they play a key role not only in the regulation of the viral replication cycle but also in host diversity, tissue tropism, and probably disease outcome (13–17). Cis-acting elements in the alphavirus genome have been identified in both 5′ and 3′ untranslated regions (UTRs), but also sprawl through coding regions of viral RNA sequences. Experimental evidences confirm the importance of CSE within the 5′ UTR and near the nsP1 start codon for the initiation of viral RNA replication. A 3′ CSE of 19 nucleotides upstream of the poly(A) tail was shown to act in concert with the 5′ CSE for the regulation of minus- and plus-strand RNA synthesis (15, 17, 18). Similarly, the SG 5′ UTR, along with the first capsid protein-coding nucleotides, serves as the SG promoter (19, 20). The role of RNA elements goes beyond RNA synthesis per se. Thus, both genomic and SG 5′ RNA elements contain translational enhancer sequences (21, 22). RNA structural elements spanning nsP1 and nsP2 sequences were shown to act as genome packaging signals (23–25). Moreover, viral RNA structures can serve as docking sites not only for the viral replication machinery but also for the interaction with host factors (26).

In the VEEV-attenuated strain TC-83, which is the basis of the veterinary vaccine, two key single nucleotide changes were shown to contribute to the attenuated phenotype. One mutation is located in the 5′ UTR and corresponds to G3A substitution in the extreme 5′ terminus (27, 28). This replacement strongly destabilizes the secondary structure present at the 5′ terminus. Interestingly, the mutation results in an increase in the synthesis of genomic RNA, reducing SG RNA synthesis and consequently creating an imbalance in the ratio of subgenomic to genomic RNA (29). Moreover, this secondary structure has additionally been shown to play a crucial role in immune escape by resisting IFIT1 restriction, a product of an interferon-stimulated gene that contributes to the regulation of protein synthesis. While many RNA viruses evade IFIT1 restriction through 2′-O methylation of their 5′ cap (^mGpppN_m), alphaviruses lack this second methylation. Instead, the secondary structure located in the 5′ UTR compensates by altering IFIT1 binding, thereby evading translational inhibition (26). Although the role of secondary structure elements in immune escape and RNA synthesis has thus been established, the possible impact of these structural motifs on VEEV RNA capping has never been addressed. The objective of the present study is therefore to investigate the impact of RNA length, sequence, and secondary structure on nsP1-RNA binding and efficiency of RNA capping in order to set the minimal acceptor substrate requirement for m⁷GMP transfer.

RESULTS

In a previous study, we have shown that VEEV-nsP1 is able to cap a synthetic 15-nucleotide-long oligomer mimicking the 5′ end of the VEEV genomic RNA (11). In order to deepen our understanding of the molecular motifs potentially modulating capping efficiency, we investigated the effect of 5′ RNA length and structure on the cap formation. The computer-predicted secondary structure of the first 230 VEEV RNA nucleotides is presented in Fig. 1. The 5′ end of VEEV folds into several stem-loops (SL) similar to those reported for Sindbis virus (SINV) and Semliki Forest virus (SFV) (14, 18, 30). The structure starts with a short stem-loop (SL1) formed by the first 30 nucleotides of the VEEV genome. The second hairpin (SL2) is longer and contains 3 loops. It is comprised of nucleotides 35 to 130 and thus includes part of the nsP1 coding sequence. The latter structure is followed by three short SLs, SL3 to SL5. SL3 and SL4 are believed to form the 51-nucleotide CSE acting as a transcriptional enhancer (16).

FIG 1.

Mfold-predicted secondary structure of VEEV 5′ end genomic RNA. Fold of the first 230 nucleotides showing the 5 stem-loops SL1 to SL5.

Capping of VEEV wild-type RNA SL1.

We first characterized the capping parameters of an in vitro-transcribed VEEV RNA corresponding to SL1 (Fig. 1). A 5′-triphosphate RNA substrate was incubated with recombinant VEEV nsP1enzyme (12) in the presence of [α-³²P]GTP under various experimental conditions. The sequential capping reaction described in the introduction includes the formation of m⁷GTP, which is then covalently complexed to the enzyme (m⁷GMP-nsP1) before getting transferred onto a γ-dephosphorylated RNA (ppRNA). In cells, this 5′-triphosphatase activity is presumably provided by the helicase domain of nsP2. Here, RNA 5′ dephosphorylation was carried out by dengue virus (DENV) nonstructural protein 3 (NS3) (31). Hence, in order to underline the importance of the γ-phosphate dephosphorylation, we used both NS3-treated and untreated RNA. For the characterization of the reaction product, the capped RNA was subsequently treated with P1 endonuclease prior to analysis on thin-layer chromatography (TLC). TLC plates were developed in either (NH₄)₂SO₄ (Fig. 2A) or LiCl (Fig. 2B), the latter allowing a better resolution of high migration spots. The results are presented in Fig. 2. The first 4 lanes in Fig. 2A, as well as lane 1 in Fig. 2B show the capping reaction on a nondephosphorylated RNA (pppRNA). In the presence of SAM (Fig. 2A and B, lane 1), GTP is methylated to form m⁷GTP. The reaction by-products include unused GTP, but also GDP and GMP. However, no m⁷GpppA is detected. It was previously demonstrated that SAH, the by-product of GTP methylation reaction, is required for m⁷GMP-nsP1 complex formation when m⁷GTP is used as the substrate for the capping reaction (7, 11). In the present study, we show that the addition of an exogenous excess of SAH reduces m⁷GTP production (Fig. 2A, lane 2). Moreover, the addition of SAH-hydrolase to the reaction mixture (Fig. 2A, lane 3) promotes the MTase reaction. When the RNA substrate is γ-dephosphorylated, m⁷GpppA is detected (Fig. 2A, lane 5, and Fig. 2B, lane 2). Cap formation is reduced in the presence of an excess of SAH (Fig. 2A, lane 6, and Fig. 2B, lane 3) and slightly increased when SAH-hydrolase is added (Fig. 2A, lane 7, and Fig. 2B, lane 4), in accordance with the production of m⁷GTP observed in lanes 1 to 3 (Fig. 2A). Controls run without enzyme (Fig. 2A, lanes 11 to 12, and Fig. 2B, lane 5) show a main signal corresponding to GTP, which is slightly hydrolyzed to GDP and GMP during the course of the reaction. We note that the addition of SAH-hydrolase alone (Fig. 2B, lane 6) increases GTP hydrolysis to GDP. The addition of a GDPase to the capping reaction did not increase m⁷GTP synthesis (data not shown). In lane 9 (Fig. 2B), the reaction mixture was treated with calf intestinal phosphatase (CIP) to confirm the position of Pi.

FIG 2.

Capping of VEEV WT RNA SL1. RNA was pretreated with DENV NS3 to remove the γ-phosphate. Then, the capping reaction was carried out in the presence of 2 μM nsP1, 100 μM SAM, and 0.33 μM [α-P³²]GTP. m⁷GpppA caps were digested from RNA with nuclease P1 and resolved on polyethylenimine cellulose thin-layer chromatography (TLC) using (NH₄)₂SO₄ (A) or LiCl (B) as a mobile phase. The migration of standards are indicated on the left side of each TLC chromatogram. The plates were revealed by autoradiography. The conditions of each capping reaction are summarized below the TLC plate. Data are representative of three independent experiments.

Another by-product resulting from the formation of the m⁷GMP-nsP1 complex is PPi, which has previously been shown to inhibit the formation of the m⁷GMP-enzyme covalent link (32, 33). In accordance with this, the addition of inorganic pyrophosphatase (PPase) (Fig. 2A, lanes 4 and 8 to 10) promoted formation of m⁷GTP and/or m⁷GpppA (Fig. 2A, lanes 4 and 8, respectively). In the absence of SAM or nsP1 enzyme (lanes 9 and 10), only GMP can be detected when the PPi hydrolase is present. In order to confirm the identity of the cap, a sample of the reaction, lanes 2 and 5 (Fig. 2B), was treated with Cap-Clip acid pyrophosphatase, an enzyme which hydrolyzes pyrophosphate bonds in m⁷GpppN structures. In the absence of nsP1, only GDP and GMP are detected (Fig. 2B, lane 8). In lane 7, we note that the cap structure is degraded into m⁷GMP, confirming that the product is m⁷GpppA.

Effect of RNA length on capping efficiency.

Next, we wondered if the other SL structures present at the 5′ end could affect the capping reaction and if RNA length might play a role in regulating the capping process. We produced VEEV RNAs from different lengths encompassing the different SLs by in vitro transcription, based on the predicted RNA structures (Fig. 1). Five RNAs corresponding to the first 30, 130, 164, 194, and 230 nucleotides were generated. When the five RNAs are used as the substrate for the RNA capping reaction (Fig. 3A), the intensity of m⁷GpppA formation is inversely related to RNA length. A drastic decrease in GTase activity is observed with substrates starting from 130 nucleotides long, with 40% of remained capping compared to the 30-mer RNA. The relative activity drops to 16, 13, and 11% for 164-, 194-, and 230-nucleotide RNA, respectively. These results support that the longer the RNA is, the less efficient the capping synthesis is.

FIG 3.

Effect of RNA length on nsP1 binding and capping efficiency. (A) Capping reaction was carried out for VEEV 5′ RNAs. m⁷GpppA caps were digested from capped RNA using nuclease P1 and subjected to TLC. The data are representative of three independent experiments. (B) Fluorescence polarization experiment assessing nsP1 binding to 3′ Cy5-labeled 30 (•)-, 130 (Δ)-, 164 (♦)-, 194 (□)-, and 230 (♦)-nucleotide VEEV 5′ RNA. Dissociation constants (K_d) were determined using Hill slope curve fitting (GraphPad Prism 7 program) and compiled in Table 1. All data points are the means of three wells, and all error bars represent the standard deviation.

We have previously shown that for short RNAs (15 nucleotides long) mimicking the VEEV 5′ end, the phosphorylation state is crucial for the RNA-enzyme interaction (11). In order to evaluate the role of substrate length in RNA recruitment, the affinity of nsP1 for the five RNA substrates was assessed by fluorescence polarization (FP). To that end, we labeled the 3′ end of each RNA with Cy5 and performed FP experiments with increasing amounts of enzyme (Fig. 3B). Apparent equilibrium dissociation constant (K_d) values are compiled in Table 1. FP results show that the binding constants for 30-, 130-, 164-, 194-, and 230-nucleotide-long RNAs oscillate between 0.36 and 1.3 μM. With the exception of the 194-nucleotide-long RNA (SL1 to SL4), which is the least preferred RNA substrate for nsP1, no major differences in K_d are noted within the set of tested RNAs. Comparison of affinity and catalytic data indicate that while nsP1 is still able to bind RNAs longer than 130 nucleotides, the capping process is largely impeded for longer substrates. Hence, weak variations of RNA affinity toward nsP1 are not likely to play a crucial role in the capping efficacy. The presence of SL3 to SL5 is deleterious not only for the formation of the cap structure itself, i.e., m⁷GpppA, but also for the MTase reaction since the signal corresponding to m⁷GTP is very weak. All the tested RNAs share a denominator, SL1, which therefore requires further evaluation in terms of sequence and structure determinant toward capping efficacy.

TABLE 1.

Effect of 5′ VEEV RNA length on RNA-nsP1 interaction^a

RNA length (no. of nucleotides)	Apparent Kd (μM)
30	0.42 ± 0.03
130	0.64 ± 0.08
164	0.60 ± 0.07
194	1.3 ± 0.18
230	0.36 ± 0.05

^aKinetic parameters were calculated using the SigmaPlot program. Values are expressed as mean ± standard deviation with an R² value of ≥0.98.

Effect of RNA 5′ secondary structure on capping efficiency.

The 5′-end SL1 structure has been shown to be important for immune evasion but also the regulation of genomic and subgenomic RNA synthesis (33). To investigate its role in the capping reaction, we linearized SL1 by mutating positions 3, 5, 7, 9, and 11 to adenosines (linear mutant). In order to evaluate if the effect on the capping efficacy can be influenced by changes in either sequence or structure of SL1 (Fig. 4C), we included in the study (i) two computer-predicted compensatory mutants, CM1 and CM2 (Fig. 4D and E, respectively), that allow the recovery of the SL1 hairpin structure but with different nucleotide composition; (ii) a 30-nucleotide oligomer, the G3A mutation observed in the vaccinal strain TC-83 (26) (Fig. 4F); and (iii) a structural compensatory mutant, TC-83C, allowing the refolding of the TC-83 into a hairpin structure similar to SL1 (Fig. 4G). The set of SL1 RNAs was first tested in the capping reaction. Interestingly, all of them can be capped by nsP1 (Fig. 4A) but with variable efficacy. Relative capping activities were thus calculated and normalized to the total radioactivity in each lane (Fig. 4B). All the mutations leading to a change of the structure or the sequence of the 5′ RNA have a negative effect on RNA capping. The TC-83 RNA, mimicking the RNA of the VEEV veterinary vaccine, has more than 20% reduction in cap synthesis. The linear RNA shows more than 30% loss in capping efficiency. Compensatory mutations fail to restore completely RNA capping, although CM1 reaches nearly 85% of the wild-type (WT) activity. We next tested the effect of SL1 sequence/structure modifications on enzyme binding. FP experiments show that WT SL1 has higher polarization values, with a computed apparent K_d of 0.42 μM (Table 2). The lowest affinity is seen for the linear RNA, with an apparent estimated K_d of 7.5 ± 2.3 μM (Table 2). The TC-83 RNA is also significantly impacted in terms of binding compared to the WT substrate, with a K_d of 4.1 μM. All the compensatory mutants, with an SL similar in fold to the WT, partially restore binding affinity, with estimated K_d values of 2.69, 3.03, and 2.88 μM, for CM1, CM2, and TC-83CM, respectively (Table 2).

FIG 4.

Effect of RNA 5′ secondary structure on capping efficiency. (A) We subjected 30-nucleotide oligomers corresponding to the WT, linear, TC-83, and various compensatory mutant sequences (as shown in Fig. 1) to DENV NS3 treatment and then assessed them under capping conditions. The data shown are representative of at least three independent experiments. (B) Relative activity corresponding to means ± standard deviation of VEEV nsP1 toward various RNA 5′ secondary structures. The relative activity of WT 5′ VEEV RNA was considered 100% for comparison with the other reaction conditions. *, P < 0.05 by t test. (C and F) Computer-predicted fold of the 5′ 30-nucleotide oligomers in WT and TC-83 VEEV RNA, respectively. (D, E, and G) Mfold of compensatory mutants from linear 30-nucleotide oligomers (C and D) and TC-83 (F). The mutations are highlighted with black circles. The computed ΔG values for each RNA are mentioned.

TABLE 2.

Effect of SL1 nucleotide mutations on RNA-nsP1 binding kinetics^a

Type of RNA	Apparent Kd (μM)
WT	0.42 ± 0.03
Linear	7.53 ± 2.3
TC-83	4.14 ± 1.6
CM1	2.69 ± 0.7
CM2	3.03 ± 0.7
TC-83CM	2.88 ± 0.8

^aKinetic parameters were calculated using the SigmaPlot program. Values are expressed as mean ± standard deviation with an R² value of ≥0.95.

Nucleotides as the substrates for N7 cap reaction.

The results obtained so far support the fact that the capping process occurs early, since long RNAs tend to interfere with m⁷GMP transfer on RNA. Moreover, we have previously shown that capping can occur efficiently on 15 mer of 5′ VEEV RNA (11) and non-VEEV RNA sequences of comparable size (data not shown). Studies on Bamboo mosaic virus (BaMV), a member of the alpha-like superfamily, have shown that GDP and ADP can also serve as acceptors for m⁷GMP to form m⁷GpppG or m⁷GpppA (33). We therefore conducted a capping experiment with nucleotides as acceptors for m⁷GMP. In the capping reaction mixture containing VEEV-nsP1, [α-³²P]GTP, and SAM, RNA was replaced by either GDP, ADP, or UDP (Fig. 5). The cap can be formed efficiently with ADP, UDP, and CDP as observed by the presence of m⁷GpppN (where N represents A, U, or C) (Fig. 5A). Caps corresponding to m⁷GpppU and m⁷GpppC comigrate with the m⁷GpppA signal under our conditions, as described previously (34). When GDP is used as the substrate for m⁷GMP transfer, the signal is very weak (Fig. 5A). A sample of each reaction condition was treated with Cap-Clip acid pyrophosphatase to confirm the cap nature of the observed signal. Following this treatment, the m⁷GpppN signal disappeared. A signal corresponding to m⁷GMP is detected (Fig. 5B), confirming the identity of the cap.

FIG 5.

Nucleotides can act as acceptors for N7 capping reaction. (A) ADP, UDP, CDP, and GDP nucleotides were used in replacement of RNA in the capping reaction described in Materials and Methods. (B) Digestion of m⁷GpppN cap with Cap-Clip acid pyrophosphatase enzyme before TLC migration. The experiment was repeated at least three times. One representative experiment is shown.

DISCUSSION

The unconventional capping mechanism of alphavirus nsP1 makes it an attractive drug target. However, exploiting nsP1 drug target potential could rationally be achieved through an extensive and comprehensive characterization of its enzymatic activity. In order to dissect the role of RNA element on VEEV RNA capping, we tested different RNA substrates that were various in length and secondary structure on the capping reaction. The analysis of the extreme 5′-secondary structure SL1 (Fig. 1 and 4C) revealed that this RNA hairpin can be capped efficiently by nsP1, and the capping reaction occurs in concordance with what was previously described in the literature. RNA capping is reduced by an excess of SAH and prompted by the addition of either SAH-hydrolase or PPase. Both SAH and PPi are by-products of the MTase reaction. PPi was shown to play a role in the regulation of cap transfer onto RNA and the regeneration of GTP (35). For eukaryotic guanylyltransferases, the formation of the GMP-protein complex before transfer to RNA was shown to be reversible. When PPi is added in excess to the reaction, GTP cannot be regenerated, and this phenomenon is reversed by the addition of PPase (35). One can speculate that a similar process is occurring in our case, leading to the methylation of all the GTP present in the reaction (Fig. 2A, lanes 4 and 8), thereby enhancing m⁷GpppA formation. It was shown previously that SAH is necessary for m⁷GMP-nsP1 complex formation when m⁷GTP is provided to the enzyme (11, 36). However, when added in excess to the reaction mixture containing GTP as substrate, SAH inhibits GTP methylation (Fig. 2A, lane 2) and, ultimately, m⁷GpppA synthesis (Fig. 2A, lane 6). These data support the proposed Alphavirus capping model according to which SAM and SAH share the same binding site (36). Moreover, the results obtained in the presence of SAH-hydrolase (Fig. 2A, lanes 3 and 7) indicate that SAH is not released from the active complex until the end of the capping process; otherwise, the production of m⁷GTP and m⁷GpppA would have been prevented.

We have shown previously that VEEV nsP1 is able to bind di-, tri-, and, to a certain extent, monophosphorylated 15-nucleotide-long VEEV 5′ RNAs. However, RNA capping was observed almost exclusively with ppRNA. The faint capping observed on pppRNA led us to postulate, at that time, that either a small proportion of pppRNA was hydrolyzed at the γ-phosphate or the enzyme was able to form m⁷GppppA cap, as reported for D1 subunit of vaccinia virus (VV) capping enzyme mutants or vesicular stomatitis virus (VSV) (11, 37–39). In the present conditions, we failed to detect any significant capping reaction with pppRNA. A contrast between RNA binding capacity and capping efficiency appeared when RNA of different lengths was tested. While nsP1 is able to efficiently bind RNA ranging from 30 to 230 nucleotides, it is barely able to cap RNAs longer than 130 nucleotides (Fig. 3). The rationale behind this experiment relied on several assumptions. First, it was shown in SFV that RNA capping occurs concomitantly with RNA biosynthesis (40). Second, many studies highlighted that a minimal RNA chain extension is a prerequisite for capping to occur. For example, Tat protein of human immunodeficiency virus (HIV) stimulates capping when the nascent RNA is 19 to 22 nucleotides long (41). In the case of VSV, RNA capping occurs after reaching 30 nucleotides (42). The reported data for BaMV are puzzling. Huang et al. (33) found that efficient capping occurs when the RNA is longer than 50 nucleotides. However, in the same study, they also showed that the BaMV capping enzyme is able to cap efficiently nucleotides, supporting a possible pre- to cotranscriptional capping event, which might relate to RNA synthesis priming, discussed below. Third, in SINV, we know that the ratio of capped versus uncapped RNAs varies between cell types. In the first 9 h of infection, uncapped RNAs are getting encapsidated to form new virions (43). This might suggest that, indeed, the entire genome can be synthesized without being capped. Our data do not support a posttranscriptional RNA capping event, at least for VEEV nsP1. The other option allowing the generation of uncapped genomes would be a decapping event. A recent paper demonstrated that during viral infection, cellular decapping enzymes DCP1 and DCP2 are directed to viral replication sites for the restriction of viral infection (44).

RNA alignment of different Alphavirus species suggests that, although the presence of CSEs is conserved, they can vary in sequence. Despite these sequence variations, some regions could fold into similar CSEs, suggesting a conserved role for these elements. Similarly, divergence in RNA fold or sequence might contribute to functional diversity (14). Earlier studies reported that modifications of the sequence or the structure of the SL1 have a negative impact on viral replication and alter the ratio of genomic versus subgenomic RNAs (29, 45). Furthermore, it was described that CSEs are hardly transposable between different Alphavirus species. The switch of either sequence or structure had a negative effect on replication (13, 45). Intriguingly, it was confirmed by biophysical techniques that TC-83 stem-loop is less stable than the WT, but this decreased stability did not affect viral RNA translation (29). Concerning the SL1 secondary fold in the capping reaction, even if none of the tested mutants abrogate m⁷GpppA synthesis, our data suggest that both the structure and the sequence of the 5′ hairpin are important for an efficient capping process (Fig. 4). Indeed, SL1 sequence and fold disruption might reduce RNA binding up to 17 times (Table 2), leading to a 30% reduction in capping activity. This suggests that SL1 integrity is important for substrate accessibility and enzyme turnover. Moreover, RNA binding is not rate limiting under these conditions. In our experiments, TC-83 RNA has a more than 20% reduction in cap synthesis. This indicates that, in addition to the negative effect on genome replication and viral escape from IFIT 1 restriction (26), the TC-83 mutant could additionally impact viral RNA capping. This difference might create a disbalance in the ratio of capped versus uncapped VEEV RNAs, shown to be crucial not only for viral replication, translation, and immune escape but also for viral infectivity (26, 29, 46). Similarly, the importance of 5′ SL structure in capping was highlighted in flaviviruses. A 74-nucleotide double-hairpin structure in the flavivirus genome is crucial for N7 cap methylation, whereas 2′-O methylation can proceed on very short RNAs (47).

In VEEV, the accumulated data and the present study tend to suggest that RNA capping is a very early event (11). Using nucleotides as the substrate for capping showed that, in contrast to BaMV, VEEV nsP1 efficiently caps ADP but not GDP (Fig. 5). GDP was shown to inhibit m⁷GTP synthesis in BaMV (48), which might explain the absence of cap signal in our case. It is worth mentioning that both genomic and SG RNAs of VEEV start with an A. The same seems to apply for other alphaviruses when looking at the reported or putative genomic and SG RNA sequences available in the NCBI database (49, 50). Interestingly, a non-VEEV RNA starting with a G can be capped by nsP1 (data not shown), suggesting that even if GDP is not used as the substrate for m⁷GpppG cap formation, the enzyme is able to cap RNAs starting with G. Another difference with the BaMV study resides in UDP capping. We observe an m⁷GpppU and an m⁷GpppC signals comigrating with m⁷GpppA, as described in reference 34. Recent epitranscriptomic studies revealed that a small proportion of m⁷GpppU- and m⁷GpppU_m-capped RNAs do exist in eukaryotic cells (51). In the case of Zika, dengue, polio, and hepatitis C viruses, it was shown that virus infection triggers a modification of epitranscriptomic profiles of the host cell (52). Moreover, m⁷GpppU- and m⁷GpppU_m-capped RNAs can be translated, and, in general, the nature of the cap analogue (m⁷GpppN), present on the RNA, as well as its methylation status on the N nucleotide, regulate the translation process (53). Although the UDP capping described here is obtained in an in vitro setting, one can speculate that it might have a biological significance, potentially for the modification of host cellular mRNA caps in view of innate immunity escape. Finally, different strategies for cap priming have been previously described. Influenza viruses, for example, snatch 10 to 15 cellular mRNA oligomers and use them to prime viral mRNA synthesis (54). In picornaviruses, priming is initiated by a protein-primed mechanism, whereby a small viral protein (VPg) is covalently linked to a UMP via the hydroxyl group of a conserved tyrosine residue. This UMP becomes the first nucleotide of the nascent RNA (55). In L-A yeast virus, it was shown that the cap analogue m⁷GpppG could initiate transcription (56). The latter supports a possible cap analogue synthesis used to prime RNA synthesis, in accordance with our data, which would constitute a novel pathway to achieve cap-protected synthesis of RNA transcripts.

In conclusion, the present study provides insights into VEEV nsP1 substrate preferences for cap structure synthesis. We show that VEEV nsP1 is able to cap RNAs from 1 to at least 130 nucleotides in length and that both the sequence and the structure of the 5′ SL1 are important for efficient binding to nsP1 and, therefore, capping efficiency.

MATERIALS AND METHODS

Expression and purification of recombinant VEEV nsP1 protein.

The cDNA encoding the VEEV nsP1 (strain P676, amino acids 1 to 535) was codon optimized and cloned into the pET28b (Novagen) vector in fusion with a hexa-histidine coding sequence at its 3′ end. The resulting recombinant protein was produced in T7 Express I^q Escherichia coli cells (New England Biolabs) after induction of the expression with 0.5 mM isopropyl-β-d-thiogalactopyranoside (Sigma-Aldrich) for 3 h at 17°C as described in references 11 and 12. Briefly, bacteria pellets were lysed by repeated sonication cycles in lysis buffer (20 mM Tris-HCl, pH 7.5, 300 mM NaCl, 5% glycerol, and 5 mM β-mercaptoethanol) supplemented with 20 μg/ml DNase I, 0.25 mg/ml lysozyme, and a tablet of EDTA-free protease inhibitor cocktail (Sigma) per 50 ml of lysate. The soluble material was recovered by centrifugation at 30,000 × g for 30 min at 4°C and then subjected to immobilized metal affinity chromatography (IMAC) on a 5-ml HisTrap column (GE Healthcare). Following extensive washings with the lysis buffer supplemented with 40 mM imidazole and 1 M NaCl, the protein was eluted with lysis buffer supplemented with 250 mM imidazole and dialyzed against storage buffer (20 mM Tris-HCl, pH 7.5, 100 mM NaCl, and 50% glycerol) for storage at −20°C. For RNA binding assays, VEEV nsP1 was dialyzed in the same storage buffer with 10% of glycerol and stored at −80°C until use.

In vitro RNA transcription.

DNA oligonucleotides were purchased from Eurofins Scientific. Short RNAs 30 nucleotides long were generated from templates of annealed single-stranded DNA (ssDNA) oligonucleotides containing the sequence of T7ϕ2.5 promoter followed by the 5′ sequence of VEEV RNA. Long RNAs >30 nucleotides long were generated using cDNA templates produced by PCR from VEEV strain P676. The forward primers were extended with the T7ϕ2.5 promoter sequence at the 5′ end. The transcription reaction was performed in 40 mM Tris-HCl (pH 7.0), 40 mM MgCl₂, 2 mM spermidine, 0.01% Triton X-100, and 4% polyethylene glycol (PEG) 8000. RNA synthesis was carried out at 37°C for 4 h in the presence of 8 mM nucleoside triphosphates (NTPs; GE Healthcare), T7 RNA polymerase (0.1 μM), and RNase inhibitor (Ambion). RNA was purified by phenol-chloroform extraction technique and then precipitated with ethanol supplemented with 2.5 M ammonium acetate overnight at 4°C. Long RNAs of >30 nucleotides were further purified on agarose gel after electrophoresis using Nucleospin gel extraction kit-NTC buffer (Macherey-Nagel), as described by manufacturer’s protocol. RNA purity was assessed by denaturing acrylamide gel (urea PAGE).

RNA GTase assay (formation of m⁷GpppRNA).

We pretreated 5′-triphosphate RNAs, generated by in vitro transcription, with 1 μM dengue virus (DENV) nonstructural protein 3 (NS3), produced and purified as described elsewhere (31). The γ-phosphate removal was done in 50 mM HEPES (pH 7.5) and 2 mM dithiothreitol (DTT) for 30 min. The enzyme was then heat-inactivated at 65°C for 5 min. GTase reaction was performed with 2 μM nsP1 VEEV in the presence of 10 μCi of [α-P³²]GTP (3,000 Ci/mmol), 50 mM HEPES (pH 7.5), 10 mM KCl, 2 mM MgCl₂, 2 mM DTT, and 100 μM SAM for 2 h at 30°C and then stored at −20°C. Capped RNAs were subjected to nuclease P1 (Sigma) digestion in 30 mM sodium acetate (pH 5.3), 5 mM ZnCl₂, and 50 mM NaCl (2 h, 37°C), followed by proteinase K (NEB) hydrolysis (30 min, 37°C). Digested products were resolved by polyethylenimine cellulose thin-layer chromatography (TLC) (Macherey-Nagel) using 0.45 M (NH₄)₂SO₄ or 1 M LiCl as mobile phase. TLC plates were visualized using Amersham Typhoon phosphor imager (12). Inorganic pyrophosphatase and adenosylhomocysteine-hydrolase (SAH-hydrolase) were purchased from Sigma-Aldrich. Cap-Clip acid pyrophosphatase enzyme was purchased from Cellscript. All commercial enzymes were used as indicated in manufacturers’ protocols. AtAPY1-ΔTM GDPase was produced and used as described in Massalski et al. (57). Quantification of m⁷GpppA signal was carried out using the ImageJ program. Statistical analyses were carried out using two-tailed unpaired Student’s t tests (*, P < 0.05).

RNA binding assay.

(i) RNA labeling. We incorporated 12.5 μM cyanine 5-cytidine-5-phosphate-3-(6-aminohexyl) phosphate (pCpCy5) dye (Jena Bioscience) at the 5′ end of in vitro-transcribed VEEV RNAs by ligation in the presence of 1 mM ATP and 1 U of T4 ligase 1 (NEB) for 1 h at 37°C. The labeled RNA was purified by ethanol precipitation in the presence of 0.3 M sodium acetate, 1 μg/μl glycogen, and 2 volumes of cold ethanol (100%) at −80°C for 1 h. After the incubation time, the Cy5-labeled RNA was centrifuged for 15 min at 10,000 rpm and washed with ethanol (70%) to eliminate the excess dye.

(ii) Fluorescence polarization assay.

The binding between the VEEV nsP1 and each RNA was monitored by FP. Each Cy5-labeled RNA strand was mixed with increased concentrations of nsP1 (0.09 to 20μM) and protein storage buffer in a final volume of 20 μl. Assays were performed in 384-well opaque microplates (Greiner Bio-One). Fluorescence polarization was measured using PHERAstar FS microplate reader (BMG Labtech) with excitation and emission wavelengths of 590 and 675 nm, respectively. All assays were repeated three times and carried out in triplicate, and a blank control without protein was included. The dissociation constants (K_d) were calculated using GraphPad Prism 7 program.