Binding of Glycyl-tRNA synthetase to Mengovirus RNA stimulates translation

Abstract

Picornaviruses are small viruses with a plus-strand RNA genome in which RNA secondary structures bind cellular proteins to support viral translation and replication. Here, we characterize tRNA anticodon stem-loop-like structures in the 5′- and 3′ untranslated regions (UTRs) of the RNA of Mengovirus, a member of the Cardiovirus group in the Picornaviridae family. These RNA elements specifically bind cellular Glycyl-tRNA synthetase (GARS). Mutation of the conserved CCA motifs in the loops of these GARS binding elements (GBEs) impairs binding, as does deletion of the anticodon binding domain of GARS. Mutation of the 3′-UTR GBE reduces Mengovirus translation early after transfection, independent of viral polymerase activity. The 3′UTR GBE is a stronger GARS binding site, and in reporter RNAs with the Mengovirus 5′- and 3′-UTRs, the 3′UTR GBE strongly contributes to recruitment of translation factors and ribosomes, thereby stimulating translation. In contrast, the 5′UTR GBE is a weaker GARS binding site, but its mutation has a stronger effect on translation. Therefore, we hypothesize that a GARS dimer binds strongly to an “anchor” site in the 3′UTR with one monomer, while the other monomer interacts with the 5′UTR to stimulate recruitment of translation factors and ribosomes.

Graphical Abstract

Graphical Abstract.

Introduction

The large group of positive-strand RNA viruses in the family of Picornaviridae comprises many pathogens, like Enteroviruses including Poliovirus (PV) and Rhinovirus, Coxsackie Virus, Hepatitis A Virus (HAV) and foot-and-mouth-disease virus (FMDV) [1]. These viruses have RNA genome lengths of about 7–10 kb (Fig. 1) and come with little own protein equipment in the virus particle, but their genomes encode functions that allow rapid and efficient capture of the cellular machineries for virus replication [2, 3].

Figure 1.

GARS binding elements (GBEs) in Picornavirus RNA genomes. (A) The Poliovirus RNA genome with the 5′UTR structure (stem-loop structures numbered with roman numerals) and blowups showing the stem-loop domain (dom) II and the apical part of dom V. The GBE in dom V [22] is in dark blue, and the putative second GBE in dom II [41, 68] is in light blue. The color code shows base pair probability (bpp) values for bases predicted to pair, but single-strand probabilities for bases predicted to be unpaired according to Vienna RNAalifold output style, as detailed in the methods section. (B) The Mengovirus genome with the GBE in the 3′UTR (dark blue) and the GBE in the 5′UTR (light blue), with blowups of the apical part of dom 3 in the 5′UTR and the 3′UTR, both of Cardiovirus clade 1 (which includes Mengovirus).

After entering the cell, the positive-strand RNA genome of these viruses is directly used – like an mRNA – for the first round of translation in which viral proteins are synthesized [3]. Viral translation is initiated at the Internal Ribosome Entry Site (IRES) element in the 5′ untranslated region (5′UTR, see Fig. 1) [4, 5]. This arrangement comes with two major advantages. On the one hand, viral translation can be regulated independently of a cap nucleotide (nt) at the 5′end, which allows these viruses to shut down cellular cap-dependent translation and efficiently capture the cellular translation machinery. On the other hand, recruitment of the ribosomes to an internal location on the viral RNA allows the viral genome ends to be fully available for RNA cis-signals involved in RNA genome replication [6, 7] without any functional constraints due to the cap at the 5′end. Moreover, interaction of the 5′- and 3′UTRs of picornaviral RNA genomes also stimulates translation in cis [8–10], likely to make sure that the viral RNA genome is not degraded by nucleases and by that is “worth” to be efficiently translated, a phenomenon reminiscent of the stimulation of cap-dependent mRNA translation by the poly(A)-tail [11, 12].

Picornaviruses hijack many cellular components (usually proteins) that are recruited to the viral RNA in order to support its translation. The picornaviral IRES elements are conserved and highly structured RNA regions (Fig. 1) [6, 13] that use a variety of cellular proteins (largely RNA binding proteins) for the – usually positive – regulation of IRES activity [10, 14–17]. These proteins are therefore called IRES trans-acting factors (ITAFs). Such cellular RNA-binding proteins often have multiple RNA-binding domains. One well-known example of cellular RNA-binding proteins that have no direct role in normal cellular translation regulation but are recruited to viral IRES elements is PTB (Polypyrimidine tract-binding protein). PTB binds to two different regions in the EMCV IRES [18, 19] and the related FMDV IRES [20] and likely acts as an RNA chaperone that stabilizes a certain IRES RNA structure to support efficient translation initiation [18, 20]. Some ITAFs likely support recruitment of canonical translation initiation factors (eIFs) to the IRES. For example, the large adaptor protein eIF4G binds to the mid and lower portion of the domain V of the Poliovirus IRES, or to the domain IV of the FMDV IRES or the corresponding J-K-L domain of Encephalomyocarditis Virus (EMCV) [10, 20, 21]. Several other host proteins are degraded to allow viral replication [10].

A cellular protein that actually has a direct role in canonical cellular translation but is also recruited by the Poliovirus IRES [22] is glycyl-tRNA synthetase (or Glycine tRNA ligase, abbreviated GARS, GlyRS or GRS). In addition to charging its cognate tRNA^Gly with glycine, GARS was also reported to have non-canonical or “moonlighting” functions, including 3′end processing of mRNAs [23, 24], the production of the cellular stress molecule (“alarmone”) Ap4A [25, 26] and a role in the defense against ERK-activated tumors [27]. Also, some other aminoacyl-tRNA synthetases have such non-translational functions [24, 28]. Moreover, various mutations in GARS are associated with a neurological disorder called Charcot-Marie-Tooth Disease, CMT [28, 29].

GARS is a dimeric class II aminoacyl-tRNA synthetase which links glycine directly to the 3′-OH end of glycyl-tRNAs [30–34]. In the anticodon stem-loop of glycyl-tRNAs, loop positions 3 - 5 (i.e. pos. 34 – 36 of the tRNA^Gly) represent the anticodon sequence 5′-NCC-3′ [35–37], which is the reverse complement of glycine codons (5′-GGN-3′, with N for any nucleotide) in the mRNA. Directly downstream of the conserved CC in the anticodon loop of glycyl-tRNA, an A residue at tRNA position 37 is extremely conserved both in bacteria and eukaryotes [35, 36, 38]. This A residue has close contact to the anticodon binding domain (ABD) of glycyl-tRNA synthetase [34, 39, 40]. Therefore, the tRNA^Gly anticodon loop contains a conserved CCA motif at anticodon loop positions 4 – 6, i.e. at positions 35 – 37 in tRNA^Gly.

We have previously shown that GARS binds to the apical stem-loop of the Poliovirus IRES RNA domain V (see Fig. 1B) and stimulates Poliovirus translation [22]. This GARS-binding element (GBE) with a small stem and the CCA sequence at positions 4–6 of a 6-nucleotide loop is conserved among all enteroviruses [13, 41]. Binding of GARS to the Poliovirus IRES stimulates not only placement of the ribosome to the classical Poliovirus polyprotein start codon at position 743 but also to the AUG 586 in the IRES domain VI [22]. This AUG 586 is now known to start a uORF in Poliovirus [42]. In agreement with our findings, Aviner et al. found that GARS selectively accumulates in virus polysomes during poliovirus but not flavivirus infections [43].

Based on the above knowledge about GARS binding elements, we performed bioinformatic searches for putative GARS binding elements in Picornavirus RNA genomes. The results revealed that the Cardiovirus group of Picornaviruses also contains potential GBEs. Surprisingly, in the Cardiovirus group, these GBEs are located both in the 5′UTR and in the 3′UTR of the viral RNA genome. We experimentally validated the function of these GBEs in Mengovirus (MV), an isolate of Encephalomyocarditis Virus (EMCV) [44]. GARS binds to the 5′- and 3′UTRs of the Mengovirus RNA genome and stimulates Mengovirus translation. Thus, GARS plays an important role in the regulation of the Mengovirus replication cycle.

Materials and methods

Reagents: Enzymes: T7 RNA polymerase (NEB M0251L), Maxima Reverse Transcriptase (ThermoFisher No. EP0743); Antibodies: GARS (proteintech No. 67893–1-Ig), GAPDH (Merck No. CB1001-500UG). Kits: NEB Monarch RNA isolation kit (NEB #T2040L).

Non-standard chemicals: Biotin-X Hydrazide (Lumiprobe 2730), also called Biotinamidocaproyl hydrazide (Merck B3770); Lipofectamine 2000 (Invitrogen 11 668 027); Acidic phenol/chloroform/isoamylalcohol (125/24/1, pH 4.5, ThermoFisher AM9722); SsoAdvanced Universal SYBR Green Supermix (Bio-Rad #1 725 271), SNAP-capture magnetic beads (NEB # S9145S), Nano-Glo HiBiT Lytic Reagent (Promega N3030).

Biological Resources: Attenuated Mengovirus (MV) genome from plasmid pMC₀ [45]; Expression vectors for human GARS (Gene ID: 2617) and its derivative [22]; Cells: HEK293T, HeLa, SK-N-AS (human neuroblastoma), Neuro-2A (murine neuroblastoma).

Statistical Analyses: Data analysis was performed using Microsoft Excel and GraphPad Prism. Raw data processing, calculations and generation of initial diagrams were usually done in Excel. GraphPad was used to display data and calculate Mann–Whitney-U or one-sample Wilcoxon significance tests in some cases. n: number of biological replicates; significances: * (P < 0.05), ** (P < 0.01), *** (P < 0.001) and **** (P < 0.0001); n.s. = not significant (P > 0.05).

Novel Programs, Software, Algorithms: none.

Web Sites/Database Referencing: tRNA database [35]; Database NCBI Virus (download: 30.5.2022); Infernal [46]; RNABOB (version 2.2.1; Eddy, S.R. RNABOB: a program to search for RNA secondary structure motifs in sequence databases. Unpublished); Vienna RNAalifold WebServer (http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAalifold.cgi) with the new RNAalifold with RIBOSUM scoring [47]. For visualization of predicted RNA structures, RNA sequences and Vienna dot-bracket outputs from RNAalifold [47] were loaded into VARNAv3-93 (https://varna.lisn.upsaclay.fr) [48]. MaxQuant version 2.5.1.0 was used to map peptide sequences to all reviewed human proteins of the Uniprot database ([49] UniProt release 2024_04). Perseus software (version 1.6.15) was used for further analyses of protein intensity values [50]. ImageJ (NIH) was used for densitometric quantification of WB results.

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [51] partner repository with the dataset identifier PXD057754. Access at https://www.ebi.ac.uk/pride: Project accession: PXD057754, Token: xbtnn4EFqeTJ.

Bioinformatic analyses

From the conserved structure of the experimentally validated strong GARS binding element in the Poliovirus IRES domain V [22], the cocrystal structure of GARS with Glycyl-tRNA [39] and the conserved structure of the anticodon stem-loop of Glycyl-tRNAs in the tRNA database tRNAdb [35], the following constraints were defined for a bioinformatic search for RNA structures that might function as GARS binding sites. The GARS binding site of Glycyl-tRNA is mainly characterized by the two conserved C residues of the glycine anticodon (5′-NCC-3′) plus a conserved A residue directly downstream of the anticodon [35] that has close contact with the synthetase [34, 39, 40], resulting in a conserved “CCA” sequence in the loop. The loop has either six or seven nucleotides (NNN*CCA*), with “N” for any single nucleotide and the asterisks for either no or one additional single nucleotide upstream or downstream of the “CCA”. The sequence and the exact base pairs of the stem appear not to be conserved; therefore, we have defined an uninterrupted stem of five base pairs and, in addition, 8 bp as context on both sides with a minimal fold energy (MFE) of -10 kcal/mol or less as the minimal requirements for the stem.

Using the above constraints, we have performed a global search in all RNA plus-strand viruses with a complete genome in the database NCBI Virus (download: 30.5.2022) to identify new GARS binding elements (GBEs) in the RNAs. However, the use of an infernal [46] model based on the anticodon loop of 608 tRNA(Gly) sequences from tRNAdb [35] turned out as too unspecific. In contrast, a model based on 2089 SL V sequences from Picornaviruses returned just a few candidates (basically in Picornaviruses) and turned out as too specific.

We therefore decided to choose a very basic approach based on RNABOB, which is a fast pattern searching tool for RNA secondary structures. We used the following input for RNABOB:

s1 h1 s2 h1' s3

s1 0 NNNNNNNN

h1 0:0 NNNNN:NNNNN

s2 0 NNN*CCA*

s3 0 NNNNNNNN

which corresponds to these three RNA structure models:

NNNNNNNN|NNNNN|NNNCCA|NNNNN|NNNNNNNN

…..|((|……|)|……..

and

NNNNNNNN|NNNNN|NNNCCAN|NNNNN|NNNNNNNN

…..|((|…….|)|……..

and

NNNNNNNN|NNNNN|NNNNCCA|NNNNN|NNNNNNNN

…..|((|…….|)|……..

This corresponds to a hairpin with a stem of 5 bp and a loop with 6 or 7 nts. This loop has a sequence of either three or four arbitrary nucleotides (of which the last is the first and non-conserved nucleotide of the Gly anticodon), followed by the two conserved nucleotides of the Gly anticodon sequence (CC) and the conserved A residue as described above, plus one optional nucleotide. This hairpin is surrounded by 8 flanking nucleotides on each side that contextualize the structure; these nucleotides can extend the stem. A limitation of RNABOB is that only the first fitting pattern is considered. This is irrelevant to our analysis since the output of RNABOB is only a preselection of the sequences, and the folding is still validated with RNAfold. For all resulting sequences, we calculated the RNA secondary structure and MFE value with RNAfold [52] version 2.5.0. To get a more specific candidate list, we only considered stable stem-loop structures with an MFE smaller than -10 kcal/mol.

From the resulting list of virus isolates, we selected those in the Cardiovirus group of Picornaviruses that contain a predicted putative GARS binding signal described above. Sequences from 5′ and 3′UTRs from these selected isolates were used to predict conserved RNA secondary structures using the Vienna RNAalifold WebServer (http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAalifold.cgi) with the new RNAalifold with RIBOSUM scoring [47]. Basic options were set to “avoid isolated base pairs”, advanced folding options to “dangling energies on both sides of a helix in any case” with the Turner model, 2004 [53]. Predicted RNA structures were visualized according to the RNAalifold output, which returns a colour code for base pair probability (bpp) values for bases predicted to pair, but single-strand probabilities for bases predicted to be unpaired.

Transcription templates and RNA synthesis – full-length mengovirus

We used the Mengovirus (MV) genome from plasmid pMC₀ [45], which is attenuated by deletion of the poly(C) tract in the 5′UTR but grows in mice [54, 55]. For mutating the GARS binding element (GBE) in the 3′UTR, the sequence CGGTAAGCCAACCG (pMC₀ pos. 7607–7624) was mutated to CGGTTAGGGTACCG.

For mutation of the active center of the 3D polymerase of Mengovirus, the sequence GGTGATGAT in the 3Dpol gene segment was mutated to GCTAGTAGT, thereby changing the conserved GDD sequence (EMCV 3Dpol amino acid pos. 332 - 334) [56] to ASS.

The sequence for the HiBiT tag ([57, 58] and Promega) (GTGAGCGGCTGGCGGCTGTTCAAGAAGATTAGC) followed by a Gly-Ser linker (GGTAGCGGCAGCGGTAGC) was inserted in frame downstream of amino acid no. 6 of the 2B gene segment, between the sequences CACGTTTAAACCA-3847 and 3848-AGACAACGGCCGG. This insertion site is upstream of the 2A binding and frameshift site in the 2B sequence [59, 60].

For in vitro transcriptions, pMC₀ or derivative full length Mengovirus DNA template was linearized with BamHI downstream of the viral poly(A) tract sequence and used at 20 mg/ml in 1 x T7 RNA polymerase reaction buffer (NEB), additional 5 mM MgCl₂ and 10 mM DTT, 1 µl SUPERase RNase inhibitor, each 3.75 mM of ATP, CTP, GTP and UTP, and 1 U/µl final concentration of T7 RNA polymerase. After 2 h incubation at 37°C, another aliquot of T7 RNA polymerase was added and the reaction incubated for another 3 h. Then, the transcription reactions were digested with 1 µl (2 units) of DNase I (NEB) for 1 h. RNAs were phenol-chloroform and chloroform extracted, ethanol precipitated, 70% ethanol-washed 2–3 times, briefly air dried and dissolved in water. RNA amounts were quantified by Qubit measurement (Invitrogen), and RNA integrity was checked on 1% agarose gels.

Transcription templates and RNA synthesis - translation reporter RNAs

For construction of the short Mengovirus HiBiT translation reporter RNAs, the Mengovirus 5′UTR/IRES sequences directly downstream of the deleted poly(C) tract (pMC₀ pos. 148–697) were cloned downstream of a T7 promoter sequence including one additional G. Then follows an 11 amino acid linker sequence (ATGGGAAGCAGATCTGGAGCGGCCGCCAGCGGC) and the 11 amino acid HiBiT tag sequence (GTGAGCGGCTGGCGGCTGTTCAAGAAGATCAGC), followed by the Mengovirus 3′UTR including 23 A residues (pMC₀ pos. 7580–7719). The mutation of the GBE was as above.

The Mengovirus secNLuc reporter constructs were cloned using the secretable Nanoluc luciferase sequence from plasmid pNL3.3[secNluc/minP] (Promega, GenBank No. JQ513371.1) between the Mengovirus 5′UTR/IRES and 3′UTR as above. Mutations of the GBEs were as above. In vitro transcriptions were performed as above.

Short RNAs with the 5′UTR GBE were as shown in Fig. 1B and Supplementary Fig. S1 left panel, with the first A replaced by G, and synthesized by in vitro transcription as above. The conserved loop sequence AACCCCA was replaced by UACCGGU in the GBE mutant. Short RNAs with the 3′UTR GBE were as shown in Fig. 1B and Supplementary Fig. S1 right panel, with the first A replaced by G, and synthesized by in vitro transcription as above. In the GBE mutant, the conserved loop sequence AAGCCA was replaced by UAGGGU. The 3′UTR sequence of Theiler´s Murine Encephalitis Virus (TMEV) is essentially as shown in the right panel of Supplementary Fig. S1B. Here, the loop sequence AAGCCA of the GBE was replaced by UAGGGU in the mutant.

GARS binding assays

To prepare DNA templates for in vitro transcription of short RNAs from Poliovirus 5′UTR and Mengovirus 3′UTR sequences, corresponding DNA oligonucleotides were annealed and extended with Phusion HS Flex polymerase. Specifically, 100 µl reaction mix which contains 0.5 µM of each forward and reverse primer, 0.2 mM dNTPs, 1 x Phusion HF buffer and 1 µM of Phusion HS Flex polymerase (NEB) was combined and PCR amplified under the following conditions: 98°C 1 min; 60°C 0.5 min (with T increment 0.2°C/s); 72°C 1 min. dsDNA templates were isolated with GeneJet PCR Purification Kit (Thermo Scientific) according to manufacturer´s instructions.

The following combinations of oligos were used to prepare short RNAs of the Poliovirus 5′UTR domain V - PV wt (Pr1 and Pr2, Poliovirus Type I positions 474 - 531), Mengovirus 3′UTR wt (Pr3 and Pr5, attenuated Mengovirus vMC₀ positions 7580 - 7700) and Mengovirus 3′UTR mut (Pr4 and Pr5).

Pr1 (PV_domV for):

GTGAATTGTAATACGACTCACTATAGGCCTCGGAGCAGGTGGTCACAAACCAGTGATTGGCCTGTCGTAACGCGCAAGTCCGTGG;

Pr2 (PV_domV rev): CCACGGACTTGCGCGTTACGACAGG;

Pr3 (Mengo_WT_for):

GTGAATTGTAATACGACTCACTATAGGTAGCGCGGTCACTGGCACAACGCGTTACCCGGTAAGCCAACCGGGTGTACACGGTCGTCATACCGCAGACAG;

Pr4 (Mengo_mut_for):

GTGAATTGTAATACGACTCACTATAGGTAGCGCGGTCACTGGCACAACGCGTTACCCGGTTAGGGTACCGGGTGTACACGGTCGTCATACCGCAGACAG;

Pr5 (Mengo_rev):

AAAACTATTTATTTTACTACTCTAGTTTATCTTGCAAAGTAGAAGAACCCTGTCTGCGGTATGACGACCGTGTA.

For preparation of radiolabeled RNAs, corresponding DNA were used as templates for T7 transcription with T7 RNA polymerase (NEB) according to the manufacturer´s instructions. The reaction mixes contained 0.5 µM UTP, 0.5 µM CTP, 0.5 µM GTP, 0.08 µM ATP and 20 µCi of α-³²P-labeled ATP (Perkin Elmer, # NEG003 × 250UC) and were incubated for 2.5 h at 37°C, followed by RNA purification with GeneJET RNA Purification Kit (ThermoFisher, K0731).

For expression of SNAP-GARS fusion proteins, the constructs pcDNA3.4-SNAP-GARS wt and pcDNA3.4-SNAP-GARS ΔABD were created from corresponding constructs pGEX-6P1 GARS wt and pGEX-6P1 GARS ΔABD described in [22] and the plasmid pcDNA3.4-SNAP-POLGARF described in [61]. pcDNA3.4-SNAP-POLGARF was treated with BamHI and NotI, and the POLGARF coding sequence was replaced with corresponding fragments containing the GARS sequence from pGEX-6P1 GARS wt and pGEX-6P1 GARS ΔABD. The resulting constructs code for fusion proteins, which contain the SNAP-tag at their N-terminus and the corresponding variants of GARS at their C-terminus.

To provide SNAP-GARS fusion proteins for the following binding assay, lysates from Hek293T cells, which express SNAP-GARS fusion proteins, were prepared. 10 cm dishes of HEK293T cells at 70% confluency were transfected with 8 µg of pcDNA3.4-SNAP-GARS wt or pcDNA-3.4-SNAP-GARS ΔABD with Lipofectamine 2000 (Invitrogen 11 668 027) according to manufacturers instructions. After 6 h, the media was replaced with fresh DMEM and the cells were incubated for an additional 30 h. Then the cells were washed with 10 ml of PBS and lysed in lysis buffer (100 mM NaCl, 20 mM Tris HCl pH 7.5, 1.5 mM MgCl₂, 0.5% Triton X100, Protease inhibitor (Cell Signalling # 5871S) and 1 mM DTT). The lysate was split into aliquots and stored at -80°C.

The GARS binding assay was performed as follows: 5 µl of radiolabeled RNA was mixed with 25 µl of the corresponding HEK293T lysate (with either GARS wt or negative control GARS ΔABD), 0.2 µl of RNAse inhibitor SUPERASEin (Invitrogen # AM2694) and 30 µl SNAP-capture magnetic beads (NEB # S9145S) prewashed 4 times with lysis buffer. The reaction mix was incubated at 30°C in a thermoshaker at 800 rpm for 1 h. During the incubation time, the SNAP moiety of the SNAP-GARS fusion protein is covalently attached to the SNAP-tag substrate benzylguanine on the beads, while the RNA interacts with the GARS moiety. After incubation, the beads were washed 2 times with 0.5 ml of lysis buffer at 37°C for 15 min and separated on a magnetic rack. After the second wash step, the beads were resuspended in 100 µl of lysis buffer, and the radioactive RNA which remain attached to the beads was measured in a Packard counter.

Electroporations

Cells were washed with PBS, detached with 0.5% Trypsin/EDTA, resuspended in DMEM including 10% FBS, counted and transferred to a 14 ml Falcon tube. After centrifugation at 700 rpm for 5 min, the cell pellet was resuspended in PBS, centrifuged again and resuspended in cytomix (120 mM KCl, 0.15 mM CaCl₂, 10 mM K₂HPO₄/KH₂PO₄ pH 7.6, 25 mM HEPES pH 7.6, 2 mM EGTA, 5 mM MgCl₂). Around 400 µl of this cell suspension was adjusted to 10⁷/ml, mixed with 1 pmol RNA (full-length Mengovirus RNAs or short Mengovirus HiBiT translation reporter RNAs) and transferred to a 4 mm electroporation cuvette. Electroporation was performed with a square wave pulse for 20 ms at 240 V. The resulting cell suspension was directly transferred to a Falcon tube with 12 ml DMEM, mixed, and each 1 ml transferred to the wells of a 6-well plate. Cells were incubated at 37°C and 5% CO₂. After adherence (at about 6 h after transfection), dead cell debris was removed and fresh DMEM/10% FBS added for further incubation, if appropriate.

Short Mengovirus HiBiT translation reporter RNAs and Mengovirus secNLuc reporter were also electroporated following an improved protocol using Opti-MEM medium (Gibco). Around 0.45 pmol RNA was pipetted to 4 mm electroporation cuvettes. 2.5 × 10⁶ cells per electroporation were trypsinized, collected using 1.5 volumes of DMEM with 10% FBS, centrifuged at 900 rpm for 5 min and resuspended in 10 ml Opti-MEM. Cells were collected again by centrifugation at 900 rpm for 5 min, and this washing step was repeated one more time. The resulting cell pellet was resuspended in Opti-MEM to a cell density of 2.5 × 10⁶ per ml. Around 400 µl of this cell suspension (1 × 10⁶ cells) was pipetted per electroporation cuvette and mixed with the RNA by repeatedly pipetting up and down. Electroporation was performed with a square wave pulse for 20 ms at 240 V. Around 2 ml of DMEM including 10% FBS and penicillin/streptomycin were added to the cuvette, the cells were carefully resuspended, transferred to a well of a 6-well plate and incubated at 37°C for 2 h.

Lipofectamine transfections

For transfections of full-length Mengovirus RNA with Lipofectamine 2000, 0.5 µg full-length Mengovirus RNA was used per 6-well plate [62]. Per well, either 3 × 10⁵ HeLa cells, 5 × 10⁵ SK-N-AS cells or 9 × 10⁵ Neuro-2A cells were seeded one day before transfection, resulting in about 90% confluent cells on the day of transfection. Transfection solution 1 contains 900 µl DMEM/0% FBS and 1.5 µl Lipofectamine, solution 2 contains 900 µl DMEM/0% FBS and 0.5 µg of Mengovirus full-length RNA. Solutions 1 and 2 were incubated at room temperature (RT) for 5 min. Then, both solutions were combined and incubated for another 30 min. Cells were washed with PBS, and the lipofection mixture was added dropwise to the cells. Three hours post-transfection, the lipofection medium was replaced by DMEM with 10% FBS. The cells were then further incubated at 37°C and 5% CO₂. Cells were then lysed for measuring HiBiT expression, or virus-containing supernatants were used for infection of cells and plaque assays.

Plaque assays

For plaque assays, 1% agarose was heat-dissolved and incubated at 55°C. An equal volume of 2 x DMEM with 3% FBS prewarmed to 37°C was added to the 1% agarose and well mixed. The cells of a 6-well plate were washed with PBS 2 h after MV infection, the PBS removed, and 4 ml of the 0.5% agarose/DMEM/FBS solution was carefully pipetted to the cells. The cells were then incubated at 37°C and 5% CO₂.

3% paraformaldehyde (PFA) solution was prepared by solubilizing 3 g PFA in 80 ml H₂O. After the addition of 0.1 ml 5 M NaOH, the solution was heated to 65°C and allowed to cool down. 10 ml of 10 x PBS was added, and the pH was adjusted to 7.4.

For fixing the cells, the agarose medium was carefully removed from the cells with a spatula. Usually, 48 h after transfection, the cells were washed with PBS, and then the cells were fixed with 3% PFA solution for 25 min. Then, the PFA solution was removed, the cells were washed with PBS and stained with 1% crystal violet in 20% ethanol/80% PBS for 10 min. Residual crystal violet was removed by repeated washings with PBS, and the plates were allowed to dry.

HiBiT assay

The HiBiT assay ([57, 58] and Promega) was performed in 12-well plates for HeLa cells and neuronal cells. 24 h before transfection, 1.2 × 10⁵ HeLa cells, 2.1 × 10⁵ SK-N-AS cells or 3.8 × 10⁵ Neuro-2A cells were seeded per well of a 12-well plate. The next day, 0.3 µg of RNA was transfected per well using Lipofectamine. Cells in the wells were then harvested and HiBiT expression measured at the times indicated. Medium supernatants were removed, and the cell layers were washed repeatedly with PBS. Cells were lysed by adding 50 µl PBS and 50 µl Nano-Glo HiBiT Lytic Reagent (Promega N3030). Nano-Glo HiBiT Lytic Reagent is pipetted together from LargeBit Protein (1:100), Nano-Glo HiBiT Lytic Substrate (1:50) and Nano-Glo HiBiT Lytic Puffer immediately before use. Lysates were transferred to a 5 ml tube and measured for 2 s in a Berthold Lumat 9501 single tube luminometer. Alternatively, plates with lysed cells were read in a Promega GloMax Discover reader.

Secretable nano-luciferase assay

For measuring secretable Nanoluciferase activity in the medium, transfected cells were left in their medium, scraped, and everything was transferred to a tube. After centrifugation for 5 min at 2000 rpm, the supernatant was transferred to a new tube and used for measuring secNLuc activity, while the cell pellet could be used for RNA isolation. Around 100 µl of the supernatant was transferred to a new tube, and 100 µl of Nano-Glo Luciferase Assay Buffer (Promega N3030) was added. The reaction was started by adding 1 µl Nano-Glo Luciferase-Assay-Substrate. After mixing and incubation for 3 min at room temperature, fluorescence was measured in a Berthold Lumat LB 9501 luminometer.

Data analysis was performed using Microsoft Excel and GraphPad Prism. Raw data processing, calculations and generation of initial diagrams were usually done in Excel. GraphPad was used to display data in some cases (Figs. 4F, 5B–E, 7) and calculate Mann-Whitney-U or one-sample Wilcoxon significance tests. n: number of biological replicates; significances: * (P < 0.05), ** (P < 0.01), *** (P < 0.001) and **** (P < 0.0001); n.s. = not significant (P > 0.05).

Figure 4.

The GARS binding element affects translation efficiency. Two different translation reporter RNA systems were used. In (A), the translation reporter RNA contains the MV IRES driving expression of a very small ORF consisting of a short 11 AS linker and the 11 AS HiBiT tag, followed by the MV 3′UTR with the GBE in wild-type (wt) or mutated (mt) form. (B) HiBiT expression in HeLa cells transfected with the small HiBiT ORF reporter RNA constructs shown in (A). In (C), the ORF from (A) was replaced by the ORF for the secretable Nanoluciferase (secNLuc, grey box), and the GBEs in the 5′UTR and the 3′UTR were mutated individually or in combination. (D) HiBiT expression in the medium of HeLa cells transfected with the secNLuc reporter RNA constructs shown in (C). (E) Reporter RNA stability assays in the HeLa cells of the experiments shown in (D). Reporter RNA abundance was analyzed by RT-qPCR. (F) HiBiT expression normalized to reporter RNA stability. Data points with the same colour correspond to results obtained within the same experiment.

Figure 5.

Physical binding of the GARS protein to the GARS binding elements. (A) The GARS protein with a SNAP fusion tag was used for the binding assays. The protein and its derivatives were expressed in HeLa cells after transfection of corresponding expression plasmids, and the cell lysates were then used as a source for the GARS fusion proteins. In the SNAP-GARS fusion protein, the anticodon binding domain (ABD) was either present (wt) or deleted (ΔABD). (B) The RNA of the apical part of the Poliovirus (PV) IRES domain V with the GBE in wild-type (wt) or mutated (mt) form. (C) The RNA of the Mengovirus (MV) 3′UTR with the GBE in wild-type (wt) or mutated (mt) form. (D) Binding assays. ³²P-labeled radioactive RNA of the PV IRES domain V or the MV 3′UTR with the GBE in wt or mt form were added to HeLa cell lysates made after transfection of plasmids expressing SNAP-GARS fusion proteins. ³²P-labeled RNA binding to the SNAP-GARS protein was pulled down using SNAP beads, and radioactivity in RNA was measured. (E) 3′-end-biotinylated MV reporter RNAs containing the HiBiT tag and the GBE in the Mengovirus 5′UTR or the 3′UTR in wt or mt form or in combination. (F) NeutrAvidin bead affinity purification and Western Blot of GARS using the 3′-biotinylated RNA with the GBE in the 3′UTR in wt or mutated form, shown in E (upper part); GAPDH detection was used as a control (lower part). 97.5% of the total lysate was used for the affinity purification (pull-down, lanes 1 and 2), and 2.5% of the total lysate was used as an input control (lanes 3 and 4). Lanes 3′ and 4´ (right panel) show a shorter exposure of lanes 3 and 4 for comparison of protein amounts in the total lysate. Additionally, a WB from a similar experiment is shown in Supplementary Fig. S2. (G) Western Blot from pull-down experiments like in (F) with 5′UTR and 3′UTR GBEs mutated individually or in combination. In (H), densitometry scan quantification from the pull-down WB GARS bands in (G) is shown as blue dots, and the GARS band scan results from the right panel in Supplementary Fig. S3A as green dots and from the right panel in Supplementary Fig. S3B as red dots.

Figure 7.

Mengovirus 3′UTR GBE mutation impairs the interaction with protein networks involved in RNA metabolism and translation. (A) Visualization of protein-protein-interaction (PPI) networks of the 166 factors whose association differs between wild type and mutant RNA (Fig. 6B). Grey lines visualize physical PPIs based on experimental evidence according to the STRING data base, version 12.0. The network components were further categorized for their biological function by overrepresentation analysis. Blue and red colors of the network nodes mark components of two highly enriched pathway terms involved in RNA metabolism and translation based on the databases WikiPathways, STRING Clusters and KEGG (as shown in Supplementary Fig. S4). Translation factors that are included in the displayed network and that have known physical interactions within the network are marked in orange for initiation factors and in green for elongation factors. (B) The five top pathways, of which No. 1 and 2 are shown in the above network, plus translation factors. Underlying data are shown in Supplementary Fig. S4.

RNA isolation and RNA stability assay

Cells were scraped from the dishes, transferred to a new tube and centrifuged for 5 min at 2000 rpm. After carefully removing the supernatant, the cells were carefully dissolved in 1 ml TRIzol (ThermoFisher #15 596 018) and incubated at room temperature for 20 min. Around 200 µl of chloroform was added, the suspension was gently inverted 40 times and incubated for 3–5 min at room temperature. After centrifugation at 16 000 rpm at 4°C, 400 µl of the supernatant was transferred to a new tube, very carefully avoiding the interphase. Around 400 µl of acidic phenol/chloroform/isoamylalcohol (125/24/1, pH 4.5, ThermoFisher #AM9722) was added, the tubes were gently inverted 40 times and centrifuged for 10 min at 16 000 rpm and 4°C. The supernatant was transferred to a new tube, extracted with chloroform, the RNA ethanol-precipitated, washed, air-dried and dissolved in water. The RNA was then further purified by NEB Monarch RNA isolation kits (NEB #T2040L) following the “protocol after TRIzol extraction”.

Reverse Transcription (RT) was performed with 600 ng RNA per reaction using Maxima Reverse Transcriptase (ThermoFisher No. EP0743) for 30 min at 50°C, followed by heat-inactivation at 85°C for 5 min. RT-Primers were 1 mM of CATAAGACAAGACCTTCACGTCATC for full-length Mengovirus RNA, GTGACCGCGCTATTATTACGC for secNLuc mRNA and GATCTCGCTCAAGATG for GAPDH mRNA. qPCR was performed with 3 µl cDNA obtained in the previous step (diluted 1:40) using the SsoAdvanced Universal SYBR Green Supermix (Bio-Rad #1 725 271), with a denaturing step at 95°C for 30 s and 40 cycles of amplification (each 3 s 95°C, 1 min 62°C). qPCR primers were CCATTCAGTAGCTGTATTTAGGTTATTG (Mengovirus for), CACAACCAGAAGGAAGACCACCGG (Mengovirus rev), GGACCAAGTCCTTGAACAGGGAG (secNLuc for), GTCGCCGCTCAGACCTTCATAC (secNLuc rev), GAGTCAACGGGGTCGT (GAPDH for) and GATCTCGCTCAAGATG (GAPDH rev).

3′-end biotinylation, coprecipitation and Western blot

RNA was in vitro-transcribed, treated with DNase and purified as above. For 3′-labelling RNA with Biotin [63], 20 µl of RNA was incubated with 180 µl of 50 mM KIO₄ (adjusted to pH 7.0 with NaOH) for 1 h in the dark. The reaction was stopped by adding 200 µl of 50% ethylene glycol. The RNA was precipitated by adding 40 µl 3 M NaAc and 1 ml ethanol and incubation at -20°C for 20–30 min, followed by centrifugation at 18 000 x g for 20 - 30 min at 4°C. The RNA pellet was washed three times with 70% ethanol, briefly air dried and carefully redissolved in 100 µl of 10 mM biotinamidocaproyl hydrazide. After incubation at 37°C for 2 h, 100 µl of 0.2 M NaBH₄ and 200 µl of 1 M Tris-Cl pH 8.2 were added, and the solution was incubated for 30 min on ice in the dark. Then the RNA was purified by standard phenol/chloroform extraction and ethanol precipitation and checked for integrity by agarose gel electrophoresis. 3′-end biotinylation efficiency was routinely checked by a “Streptavidin Shift” by mixing 150 ng of 3′-biotinylated RNA and 0.5–2 µg of soluble Streptavidin in StrepShift buffer (10 mM Tris-Cl pH 7.4, 2.5 mM MgCl₂, 100 mM NaCl), followed by agarose gel electrophoresis beside RNA alone in StrepShift buffer as a control. In this test, the retardation of RNA migration in the gel by Streptavidin binding revealed that virtually all of the RNA was successfully 3′-biotinylated.

For RNA-protein coprecipitation assays, for each sample 20 µl packed volume of Pierce HighCapacity NeutrAvidin Agarose Beads (ThermoFisher 29 204) were used along with 80 pmol of 3′-biotinylated RNA recovered after transfection of cells. The beads were prepared by washing 3 x with Physiological Wash Buffer (PWB, 150 mM KCl, 20 mM HEPES pH 7.5, 1 mM DTT, 1 mM MgCl₂, and 0.01% Nonidet P 40 (octylphenoxypolyethoxyethanol), or alternatively containing 0.001% of the stronger NP-40 (nonylphenoxypolyethoxyethanol, Dow Chemicals)). Then the beads were blocked for at least 1 h at 4°C in blocking solution (0.2 mg/ml tRNA, 0.2 mg/ml BSA, 0.2 mg/ml glycogen and 0.1 mM DTT in PWB). After blocking, the beads were again briefly washed 3 x with PWB before use.

For transfection, 7 × 10⁷ HeLa cells were grown in twelve 175 cm² flasks to about 80% confluence. Cells were trypsinized, and each 10⁷ cells were used per electroporation transfection sample. Cells were collected by centrifugation at 100–200 x g for 5 min, the supernatant discarded, the cells washed 3 x with 10 ml Opti-MEM and resuspended in 600 µl Opti-MEM per sample. Around 80 pmol of RNA was added to each dry 4 mm cuvette, 600 µl of cell solution (10⁷ cells) was added and electroporated with a square wave pulse for 20 ms at 240 V. Cells were resuspended in DMEM as above, plated to a 10 cm plate per sample and incubated at 37°C for 1 h. Cells were harvested by scraping, washed with PBS and lysed in IP lysis buffer (20 mM Tris-Cl, pH 7.5, 150 mM NaCl, 0.5% Nonidet P-40, 2 mM EDTA, 0.5 mM DTT, 1 mM NaF, 1 mM Pefablock protease inhibitor (Carl Roth)) at 4°C for 20 min on a rotating wheel. After centrifugation at 10 000 x g at 4°C for 10 min, the supernatant was transferred to a new tube, and 2.5% were saved in another tube as input control.

Coprecipitation (pull-down) of cellular proteins bound to the Mengovirus reporter RNA was performed by combining the blocked and washed beads with the cleared cell lysate plus Complete EDTA-free Protease Inhibitor (Merck) and incubation for 30–60 min at 4°C on a rotating wheel. After centrifugation at 10.000 x g at 4°C for 2 min, the supernatant was removed, and the beads were washed four times for 2 min with PWB (with protease inhibitor), each followed by centrifugation. Proteins were released from the RNA bound to the beads using 60 µl of 1% SDS, 100 mM NaCl for 1 min. Beads were centrifuged at 10.000 x g at 4°C for 1 min, and the supernatant was removed immediately. Samples for mass spec analyses were temporarily stored at -20°C.

To samples for Western Blot (WB), 2 x Laemmli sample buffer was added, samples heated to 95°C for 5 min and loaded on a Laemmli protein gel with a 6–20% acrylamide separating gel. Color Protein Standard Prestained Marker (NEB #P7719) was used as size marker. Proteins were transferred to nitrocellulose (0.2 µm, Bio-Rad) in a wet-blot tank at 280 mA for 90 min at 4°C. Membranes were blocked with 5% skim milk in PBST for at least 2 h at RT or overnight at 4°C. The membrane was cut horizontally, and primary antibodies were used separately for the membrane segments against GARS (proteintech No. 67893–1-Ig, 1:20 000) and GAPDH (Merck No. CB1001-500UG, 1:5 000) in blocking solution for either 1–2 h at RT or over night at 4°C. After washing 3 times in PBST with 0.1% Tween20, the secondary antibody (Anti-Mouse HRP, Sigma-Aldrich A9044) was incubated in blocking solution for 1 h at RT. After washing 3 times in PBST with 0.1% Tween20, the membranes were developed using the Cytiva Amersham ECL Prime Western Blot detection reagent (FisherScientific 12 994 780). After analysis on a Bio-Rad ChemiDoc, equally sized chemiluminescence (for WB) and visible light images (colorimetric, for marker bands) were taken and automatically merged in the ChemiDoc.

Mass spectrometry

Protein samples solubilized in 1% SDS, 100 mM NaCl from NeutrAvidin affinity purification were reduced by the addition of Tris-(2-chlorethyl)-phosphate (TCEP) (using a 40 x stock-solution in 100 mM ammonium bicarbonate) to a final concentration of 5 mM and incubation for 15 min at 90°C. Samples were cooled down, spun down by a brief centrifugation step, and iodoacetamide was added to a final concentration of 10 mM (using a 40 x stock-solution in water). Samples were incubated for 30 min at room temperature in the dark.

Two types of SP3 beads (GE45152105050250 and GE65152105050250) were mixed 1:1 (20 µl each) and washed by the addition of 200 µl water, vortexed and separated on a magnetic rack [64]. These washes were repeated 3 times. Finally, 100 µl of water was added, and equilibrated beads were directly used or stored at 4°C. Around 4 µl of this slurry was added to 25 µl of each sample. After the addition of 29 µl acetonitrile, samples were briefly vortexed and incubated for 15 min at room temperature. Then, the beads were separated using a magnetic rack. The supernatant was discarded, and 500 µl 70% ethanol was added to each sample. After separation of the beads using a magnetic rack, the supernatant was discarded again, and the ethanol wash was repeated once. Subsequently, beads were washed in the same way by the addition of 200 µl of acetonitrile and finally dried.

Samples were reconstituted in 100 µl ammonium bicarbonate buffer (50 mM, pH 8.0). Trypsin (0.1 µg in 50 µl ammonium-bicarbonate buffer) was added to the beads, and samples were incubated in a thermomixer at 1200 rpm and 30°C overnight. The next morning, beads were separated on a magnetic rack, and the supernatant was transferred into fresh collection tubes. To the remaining beads, 30 µl of 2% DMSO were added, and samples were incubated for 5 min in an ultrasonic bath. Subsequently, beads were magnetically separated again, and the supernatant was combined with the previous supernatant in the corresponding collection tubes. Finally, 30 µl of water (HPLC grade) was added to the beads, and the beads were vortexed and subsequently separated again by a magnet. Supernatant was transferred to the corresponding collection tubes. Then, 10 µl of 5% Trifluoroacetic acid (TFA) was added to each collection tube containing the tryptic peptides, samples were vortexed and finally shortly spun down.

Tryptic peptides in the supernatants were desalted and concentrated using Chromabond C18WP spin columns (Macherey-Nagel, # 730 522) according to the manufacturer´s protocols. Finally, peptides were dissolved in 10 µl of water with 5% acetonitrile and 0.1% formic acid.

The mass spectrometric analysis of the samples was performed using a timsTOF Pro mass spectrometer (Bruker Daltonic). A nanoElute 2 HPLC system (Bruker Daltonics), equipped with an Aurora Ultra C18 RP column (25 cm x 75 µm ID) filled with 1.7 µm beads (IonOpticks, Australia), was connected online to the mass spectrometer. A portion of 2 µl of the peptide solution was injected directly into the separation column. Sample loading was performed at a constant pressure of 800 bar.

Separation of the tryptic peptides was achieved at 60°C column temperature with the following gradient of water/0.1% formic acid (solvent A) and acetonitrile/0.1% formic acid (solvent B) at a flow rate of 400 nl/min: Linear increase from 2% B to 17% B within 18 min, followed by a linear gradient to 25% B within 9 min and linear increase to 37% solvent B in additional 3 min. Finally, B was increased to 95% within 10 min and held at 95% for an additional 10 min. The built-in “DDA PASEF-standard_1.1sec_cycletime” method developed by Bruker Daltonics was used for mass spectrometric measurement.

Raw data analysis of mass spectra was performed using MaxQuant version 2.5.1.0, and peptide sequences were mapped to all reviewed human proteins of the Uniprot database ([49] UniProt release 2024_04). Perseus software (version 1.6.15) was used for further analyses of protein intensity values [50]. One of five replicates of samples from cells transfected with the GBE mutant RNA was omitted from the data sets due to low input protein concentration and low protein coverage. The protein intensity values were log2-transformed, and their distribution was visualized as violin plots using Graphpad Prism (version 9.5.1, GraphPad Software). No normalization was performed in order to preserve the true differences between the samples. In order to enable calculation of ratios between samples, missing values were imputed a log2 intensity value of 9, which was below the lowest intensity value measured across all samples. Proteins that were enriched or depleted compared to the wild type condition were identified based on significant differences of mean intensities by Student’s t-tests using significance levels of -log10 P-value > 1.3 (P < 0.05%) or -log10 p-values > 1 (P < 0.1). Filtering and heatmap visualizations were performed in Excel 2016 according to the criteria described in the figure legends. Protein network analysis was inferred from filtered gene ID lists using information from the STRING database (https://string-db.org, version 12.0). Networks were visualized and annotated with enriched pathway terms using Cytoscape version 3.10.3 and the integrated StringApp version 2.1.1 [65, 66].

Results

GARS binding elements (GBEs) are conserved in the 5′ and 3′UTR of cardiovirus RNA

In Poliovirus, a strong GARS binding element is located in the domain V (dom V) of the 5′UTR [22] (Fig. 1A, right part). This GBE is characterized by a stem with five uninterrupted base pairs that include two C-G pairs. The loop has 6 nucleotides (nts) with the sequence NNNCCA, with a conserved CCA motif in loop positions 4 - 6 [13]. In this loop, the sequence “NCC” (loop positions 3–5) constitutes the glycine anticodon sequence. Also, the A residue directly downstream in position 6 of the anticodon loop, the A in the CCA motif, is highly conserved not only in tRNAs [35, 36, 38] but also in the Polioviruses IRES [13] which strongly binds to GARS [22]. Although this downstream A residue is not part of the actual anticodon sequence, it makes close contact to the anticodon binding domain of the glycyl-tRNA synthetase protein [34, 39, 40, 67]. Therefore, it must be considered to essentially contribute to the specific interaction of the synthetase with a GARS binding element. In addition to the GBE in Poliovirus IRES dom V, we also bioinformatically confirmed that the apical stem-loop of the Poliovirus 5′UTR dom II in which the CCA motif is also conserved [13] (Fig. 1D, left part) can be considered to be a second GARS binding element [68].

The activity of the Poliovirus IRES in an in vitro-translation assay was competed strongly by the Poliovirus domain V [22]. Also, binding of a putative GARS binding element found in the Cadicivirus (CDV) IRES with a 6 nt AAACCA loop sequence to the complete GARS protein was about twofold more efficient compared with an AAACCC sequence in the loop [41]. Moreover, the CDV IRES AAACCA loop GARS binding element and the GARS anticodon binding domain underwent induced fit movements in Molecular Dynamics Simulations [69].

Based on these observations, we considered a strong putative GARS binding element to have an RNA stem-loop with a 6 nt loop with the sequence NNNCCA and a considerably stable stem. Therefore, we bioinformatically searched in the family of Picornaviridae [1, 2] for a stem-loop structure with an uninterrupted stem of at least five base pairs with a minimal fold energy (MFE) of -10 kcal/mol or lower and a loop of six nucleotides with the sequence “NNNCCA”. This search confirmed the presence of putative GARS binding elements in the IRES domain V of virtually all Enteroviruses, consistent with their conservation [13, 41].

Interestingly, we found that members of the genus Cardiovirus contain a putative GARS binding element in the 3′UTR (Fig. 1B). According to the conservation of the RNA secondary structures in their 3′UTRs, these Cardioviruses essentially separate in two clades. Clade 1 (Fig. 1B; Supplementary Fig. S1B, left part; Supplementary Fig. 1E) contains members of the Cardiovirus A group, namely Mengovirus and other isolates of EMCV, while clade 2 (Supplementary Fig. S1B, right part, Supplementary Fig. 1F) includes members of Cardiovirus B and D, like Theiler´s Murine Encephalitis Virus (TMEV) as well as Saffold Virus (SAFV). A second bioinformatic search also including a 7 nts loop (NNNNCCA or NNNCCAN) revealed that the 5′UTR of Cardioviruses contains a putative GBE with a 7 nts loop in the apical part of the large central IRES domain (named domain 3, or I) (Fig. 1B, left part, and Supplementary Fig. S1A, C, D). Looking back to the sequence comparisons of Richard Jackson and colleagues [13], we realized that they already found the sequence corresponding to the 5′UTR GBE to be conserved among Type II IRES elements. We have therefore also included analyses of the putative 5′UTR GBE in the reporter RNA translation and pull-down assays.

The secondary structures of the 3′UTR GBEs appear quite similar among different Cardiovirus isolates (Supplementary Fig. S1B, E, F). In both clades, an apical stem of not only five but eight uninterrupted base pairs forms, including four C-G base pairs, forming a rather stable stem that exposes the NNNCCA loop. In contrast, the primary sequence as well as the secondary structure of the lower parts of the GBE containing stems appear less conserved and/or stable. Interestingly, in most Cardiovirus isolates, the GBE is located about 25 to 28 nts downstream of the MV polyprotein ORF stop codon. Thus, GARS binding can be speculated not to physically interfere with binding of a ribosome to the translation termination site, since a ribosome bound to a codon covers about 28 nts, with the A site being approximately in the center of the ribosome [70]. Accordingly, a ribosome with the Mengovirus termination codon in its decoding center would unfold and cover about 14 nts of the Mengovirus 3′UTR, but it would not unfold the apical stem-loop of the 3′UTR GBE. Therefore, GARS binding to the Mengovirus 3′UTR can be considered to be possible during ongoing translation and provide a kind of “anchor” site for GARS.

The apical part of the Cardiovirus 5′UTR central domain (3 or I) appears quite conserved [13], with the conserved 5′UTR GBE with a 7 nts loop (Fig. 1B, left part, and Supplementary Fig. 1A, C, D). This conserved central domain can be speculated to provide some kind of “IRES-organizing” function, while the downstream Y-shaped stem-loop domain 4 or J-K-L binds canonical translation initiation factors [10, 14, 21].

Efficient mengovirus replication depends on the 3′UTR GBE in HeLa and neuronal cells

To investigate whether the GBE in the Mengovirus RNA 3′UTR affects the viral life cycle, we mutated the GBE in the context of the full-length virus genome. To monitor viral translation and replication, full-length MV genomes were tagged with the HiBiT tag [57, 58] in the N-terminal 2B region (Fig. 2A). We used MV attenuated by removal of the poly(C) tract in the 5′UTR (vMC₀), which grows well in HeLa cells and in mice [62]. The 11 amino acid (AA) HiBiT sequence was inserted together with a downstream 6 AA Gly-Ser linker between codons 6 and 7 of the 2B sequence, shortly upstream of the 2A binding and frameshift site in the 2B sequence [59, 60].

Figure 2.

Replication of the Mengovirus genome depends on the 3′UTR GARS binding element. (A) The Mengovirus (MV) genome with the HiBiT tag inserted near the N-terminus of 2B. (B) Plaque assays after infection of HeLa cells with Mengovirus contained in the medium diluted supernatant of HeLa cells transfected with the wild-type (wt) or mutated (mt) 3′UTR GBE. (C) Quantification of plaque numbers after infection of wt or mt MV in HeLa cells in a time course from 24 to 96 h post-infection (hpi). (D) Quantification of plaque sizes from the experiment in (C). (E) HiBiT expression from MV genomes in HeLa cells as well as in SK-N-AS and Neuro-2A neuroblastoma cells at 24 h post-transfection (hpt). n: number of biological replicates; significances with two-sided P-values: * (P < 0.05), ** (P < 0.01), *** (P < 0.001) and **** (P < 0.0001); n.s. = not significant (P > 0.05).

These MV genomes with the wild-type or mutated 3′GBE were transfected into HeLa cells for virus production. The supernatants were used in different dilutions for the infection of HeLa cells and plaque assays. The results show that the virus with the HiBiT tag replicates well in the cells (Fig. 2B). The number as well as the size of the plaques were reduced with the 3′GBE mutant MV genome. In time courses, plaque numbers (Fig. 2C) and sizes (Fig. 2D) were significantly reduced with the GBE mutant MV genome.

We also analyzed MV replication in neuronal cells, since MV is known to cause paralysis in man [71] and mouse [55, 62]. SK-N-AS (human neuroblastoma) cells and Neuro-2A (murine neuroblastoma) cells were used since it is known that MV replicates in these cells [72]. Since plaque assays in particular with the Neuro-2A cells revealed difficult, since the cells did not grow in monolayers but showed patchy growth on the cell culture dish surface, we analyzed MV replication by measuring HiBiT expression after transfection. HeLa cells were used as a control. The results (Fig. 2E) show that mutation of the 3′GBE impairs replication of MV not only in HeLa cells but also in the neuronal cells, irrespective of whether human or mouse cell lines were used.

GARS binding stimulates Mengovirus gene expression independent of the viral polymerase

HiBiT-tagged MV genomes with the wild-type or mutated 3′UTR GBE (Fig. 3A) were then transfected with Lipofectamine into HeLa cells, and HiBiT expression was analyzed in a time course (Fig. 3B). In addition, the amino acid sequence GDD in the active center of the 3D polymerase (3Dpol) was mutated to ASS, disabling replicase activity [56]. After transfection of the MV genome with the wt 3′UTR GBE, HiBiT expression increased from 3 to 24 h post-transfection (hpt) by more than two orders of magnitude (Fig. 3B, GBE + P+). In contrast, mutation of the 3′UTR GBE resulted in significantly reduced HiBiT expression (GBE- P+). However, HiBiT expression still increased over time, indicating that MV replication is reduced but not disabled by the GBE mutation.

Figure 3.

Expression of full-length Mengovirus genomes at early time points. (A) The Mengovirus genome with the HiBiT tag and the 3D polymerase (P) with the 3′UTR GBE in either wild-type (+) or mutant (-) version. (B) HiBiT expression from full-length MV RNA genomes after lipofectamine transfection in HeLa cells at the indicated hours post-transfection (hpt). MV genomes contained the 3′UTR GARS binding element in the wild-type (GBE+) or mutated (GBE-) form and the 3D polymerase (P) in its active wild-type form (P+) or inactivated (P-). Expression with the GBE + P + construct at 3 h was set to 1. Below the graph, n(P+) indicates the number of experiments for the P + constructs for each time point, and p(P+) indicates the significance with two-sided p-values comparing GBE + P + and GBE-P + constructs. Significances were calculated with GraphPad´s nonparametric Wilcoxon signed-rank test, with * (P < 0.05), ** (P < 0.01), *** (P < 0.001) and **** (P < 0.0001); n.s. = not significant (P > 0.05). (C) blowup from (B) at early time points. (D) HiBiT expression from full-length MV RNA genomes as in (B) but using electroporation transfection. Expression with the GBE + P + construct at 0.5 h was set to 1. Here, the n and P values comparing GBE+ (blue) and GBE- (grey) status are indicated for both the P + and the P- constructs below the graph. (E) Blowup from (D) at very early time points. (F) RNA stability assays measuring the abundance of Mengovirus RNA genomes at 2 h after electroporation into HeLa cells. MV genomes had the GBE in wild-type form (GBE+) or mutated (GBE-). Both forms had intact active 3D polymerase (P+). Reporter RNA abundance analyzed by RT-qPCR, normalized to GBE + P + values. “n” indicates the number of biological replicates; “n.s.” indicates that there is no significant difference.

Inactivation of the 3D polymerase (P-) resulted in an overall decline of HiBiT expression over the time course, indicating that the virus genomic RNA could not replicate any more but was degraded over time. Both with the replication competent (P+) and the replication deficient (P-) MV genome, HiBiT expression was significantly higher already from early time points (starting at 3 hpt) with the GBE + genomes in comparison to the GBE- genomes (Fig. 3C). In conclusion, GARS acts on the MV genome expression independent of the activity of the viral 3D polymerase.

The GBE affects full-length mengovirus genome expression at early time points

Based on the above observations, we used electroporation instead of lipofection to achieve an immediate uptake of the electroporated RNA into the cells. HiBiT expression from the MV genomes was then analyzed as early as 0.5 h after electroporation (Fig. 3D, E). The results show that the difference in HiBiT expression very early after electroporation is significantly higher in GBE + constructs than in GBE- constructs, independent of 3Dpol activity. In conclusion, the MV genomes are translated with higher efficiency when GARS can bind to the MV 3′UTR. Moreover, there is no evidence for a more rapid decline of HiBiT expression at early time points when the GBE is mutated (Fig. 3E), indicating that differential degradation (i.e. stability) of the electroporated MV RNA does not essentially contribute to the difference in HiBiT expression. This conclusion was confirmed by analysis of RNA stability by RT-qPCR (Fig. 3F), which shows that - despite variability in RNA stability from experiment to experiment - on average the RNA stability of the full length Mengovirus RNA genomes is not affected by mutation of the GBE. This indicates that GARS regulates MV RNA genome translation rather than RNA stability.

The GBE affects mengovirus translation efficiency

Following the above conclusion that GARS likely stimulates MV RNA genome translation, we cloned the MV 3′UTR along with the Mengovirus 5′UTR into “translation only” reporter systems (Fig. 4) to directly analyze the effect of GARS on MV translation. A minimal translation reporter system containing only the 11 AA HiBiT tag with an N-terminal 11 AA Gly-Ser linker peptide was inserted between the MV 5′UTR and 3′UTR sequences, with a poly(A) tract of 23 A residues at the 3′end (Fig. 4A). The 3′UTR GBE was either wild-type or mutated as described above. These translation reporter RNAs were transfected into HeLa cells, and HiBiT expression was analyzed after 4 and 6 h (Fig. 4B). The results show that mutation of the GARS binding element in the MV 3′UTR significantly impairs translation of the reporter RNA.

The MV 5′UTR and 3′UTR sequences were also cloned into an alternative reporter construct using secretable Nanoluciferase, secNLuc (Fig. 4C). This system comes with the advantage that expression of the secreted translation reporter can be easily and very efficiently measured in the medium without lysing the cells, and by that it allows to measure intracellular reporter RNA stability in the same experiment. According to our second bioinformatic search revealing also GARS binding element candidates with a 7 nts loop (see introduction), we also included mutations of the putative 5′UTR GBE into this secNLuc reporter RNA system (Fig. 4C), generating Mengovirus reporter RNA constructs with the GBEs mutated either in the 5′UTR (5′mt) or the 3′UTR (3′mt) or in both UTRs (5′3′mt).

Reporter protein expression (Fig. 4D) shows that the translation efficiency is higher in the presence of the wt GARS binding elements compared with their mutated form (mt) in either the 5′UTR or the 3′UTR. Interestingly, mutation of the GBE in the IRES has a stronger effect on translation efficiency. Alongside, the stability of the reporter RNAs was not significantly affected by mutation of the GBEs (Fig. 4E), and the ratios of translation efficiency normalized to RNA amounts (Fig. 4F) yield the same results as the translation assays only (Fig. 4D). Taken together, these results confirm that both GARS binding elements in the 5′UTR and in the 3′UTR stimulate translation controlled by the MV untranslated regions.

GARS physically binds to the GBEs in the mengovirus 5′- and 3′UTR

In order to test if the GARS protein binds to the putative GARS binding elements in the Mengovirus RNA, we first performed binding assays with a SNAP-GARS fusion protein (Fig. 5A) and a small radioactively labeled MV 3′UTR RNA containing the putative 3′UTR GBE (Fig. 5B). The SNAP-GARS fusion protein has an N-terminal SNAP-tag fusion moiety [73], and the C-terminal GARS moiety was used in wild-type (wt) form or with the anticodon binding domain deleted (ΔABD) to disable GARS binding to the GBE [22]. The SNAP-GARS fusion proteins were expressed from plasmids transfected into HEK293T cells. Lysates made from the transfected HEK293T cells were then incubated with ³²P-labeled RNAs containing GBEs. On the one hand, small MV 3′UTR RNAs (Fig. 5B) with the putative GBE (wt or mutated) were used. The ³²P-labeled RNA of the MV 3′UTR including the GBE started directly downstream of the polyprotein ORF stop codon and was 121 nts long (positions 7580–7700 of the attenuated Mengovirus vMC₀ [45], renumbered as nts 1 - 121 in the sequence alignment outputs in Fig. 1B and in Supplementary Fig. S1B, E, F). On the other hand, the apical part of Poliovirus dom V RNA with the GBE [22] (Figs. 1A and 5C) was used as a positive control. This ³²P-labeled RNA was 58 nts long and included the upper half of the Poliovirus 5′UTR dom V (Poliovirus Type I positions 474–531; the apical part of this dom V RNA is shown in the right part of Fig. 1A).

After incubation of these RNAs with the cell lysates, the SNAP-GARS fusion proteins were pulled down using SNAP-capture magnetic beads. The results (Fig. 5D) show that the MV 3′UTR binds to the SNAP-GARS fusion protein. Mutation of the GARS binding site (“GBE mt”) strongly reduces binding, as does deletion of the anticodon binding domain of GARS (“ΔABD”) or the combination of both. The Poliovirus dom V GBE [22] was used as a positive control with wt SNAP-GARS or as a negative control with ΔABD SNAP-GARS.

The above physical binding of GARS to the GBE was confirmed by analyzing the proteins associated with the translation reporter RNA construct containing the Mengovirus 5′UTR, a short open reading frame with a linker and the HiBiT tag, and the Mengovirus 3′UTR (Fig. 5E). This RNA was 3′-biotinylated and electroporated into HeLa cells. After cell lysis and NeutrAvidin affinity purification (pull-down) of the 3′-biotin-RNA, a Western Blot (WB) for GARS was performed, with GAPDH detection as a control (Fig. 5F). The results show that GARS binds less efficient to the 3′UTR GBE mutant RNA (Fig. 5F, lane 2) compared with wt GBE RNA (lane 1). In contrast, GAPDH binding was virtually lost. The very faint GAPDH band in lane 2 in this particular experiment may be due to the ability of GAPDH to weakly bind RNA with its NAD⁺ binding site [74, 75]; in other pull-down experiments GAPDH binding was completely lost (Supplementary Fig. S2, and see below for Fig. 5G, Supplementary Fig. S3 and also Fig. 10D). Taken together, these results demonstrate that GARS protein binds specifically to the MV 3′UTR, dependent on the MV 3′UTR GBE and the GARS anticodon binding domain.

Figure 10.

Binding of GARS and translation machinery components to single UTR GBEs. (A) Short 3′-end-biotinylated MV reporter RNAs containing the 5′UTR alone with the GBE in wt or mt form. (B) Short RNA of the Mengovirus 3′UTR alone with the GBE wt or mt. (C) Short RNA of the TMEV 3′UTR alone with the GBE wt or mt. (D) NeutrAvidin bead affinitiy purification and Western Blot of GARS using the above 3′-biotinylated RNAs (upper part); GAPDH detection was used as control (lower part). 97.5% of the total lysate were used for the affinity purification (pull-down, left panel), and 2.5% of the total lysate were used as input lysate control (right panel). (E–H) Relative abundance of proteins binding to 3′-end-biotinylated short RNAs of the apical loop of IRES domain I (with the GBE wt or mt) as shown in (A) and on the left side of Fig. 1B, of the Mengovirus 3′UTR (with the GBE wt or mt) as shown in (B) or the right side of Fig. 1B. Thereby, (E) shows proteins of the small ribosomal 40S subunit, (F) of the 60S subunit, (G) translation initiation factors and (H) translation elongation factors. As in Fig. 9, “n” indicates the number of different protein species used for calculation of mt/wt ratios, significances were calculated over the ratios (mt/wt) for all proteins, and asterisks indicate significance levels as in previous figures.

Extending these pull-down and WB assays also to the 5′UTR GBE mutations, Fig. 5G shows that mutation solely of the 5′UTR GBE only had a very slight effect on GARS binding (Fig. 5G right panel, lane 2), whereas mutation of solely the 3′UTR has a strong effect on GARS binding (lane 3), as does the double mutation of both GBEs (lane 4). A quantitative evaluation of this and two additional Pull-down Western Blots (shown in Supplementary Fig. S3A and B) shows that mutation of the 5′UTR has a mild (non-significant) effect on GARS binding, whereas mutation of the 3′UTR has a strong effect (Fig. 5H).

The GBE enhances recruitment of components of the translation machinery

To further support the conclusion that GARS is involved in translation stimulation, we analyzed the recruitment of translation machinery components to the Mengovirus reporter RNAs by a proteomics approach. 3′-biotinylated RNAs of the MV HiBiT reporter RNA (Fig. 5E) were transfected into HeLa cells. After cell lysis, we performed NeutrAvidin affinity purification, followed by identification and quantification of the purified protein fraction components by liquid chromatography tandem mass spectrometry (LC-MS/MS). Both wild-type and mutant RNA constructs were associated with several hundred proteins with similar abundancy distributions across the different biological replicates, demonstrating that the efficiency and reproducibility of the pull-down experiments were largely comparable (Fig. 6A). These protein sets were analyzed for statistically significant differential binding events between wild-type and mutant RNA based on mean intensity values. In total, the interaction of 166 or 92 proteins was reduced or abolished with mutant RNA, based on p-values of 10% or 5%, respectively (Fig. 6B and C).

Figure 6.

Mutation of the GARS binding elements in the Mengovirus 5′UTR and 3′UTR impairs the interaction with proteins. Cells were transiently transfected with 3′-end-biotinylated short reporter RNAs with 5′UTR and 3′UTR GBE either wild-type (wt) or mutated (mt) individually or in combination (see Fig. 4A). After cell lysis, RNA constructs were affinity purified on NeutrAvidin beads, and the associated proteins were identified by LC-MS/MS. (A–C) Violin plots show total numbers (below violins) and log₂ protein intensities (y-axis) of all identified proteins across five or four biological replicates (A), while the other panels show the distribution of values from protein sets with significantly altered RNA binding compared to the wild RNA based on P-value threshold of < 0.1 (B) or < 0.05 (C) (Student’s t-test).

Mutation of the 3′UTR GBE resulted in a stronger loss of overall protein binding (Fig. 6B and C). According to the information for this data set on physical and experimentally validated protein-protein interactions (PPI) deposited in the STRING database [76], the 166 proteins form a dense and highly enriched PPI network (Fig. 7). This network was segregated into three subnetworks based on ontology annotations of proteins and overrepresentation analyses. With this approach, we identified two distinct subnetworks comprising 21 proteins involved in mRNA recognition and mRNA processing (red colors in Fig. 7A and B) and 18 ribosomal proteins (blue colors in Fig. 7A and B). A protein-by-protein inspection of the experimental raw data demonstrates that the interaction of all factors of these two pathways with the 3′UTR GBE mutant RNA is reduced compared to the wild-type RNA (Supplementary Fig. S4). However, the measured protein intensities were too variable between experiments to fulfil the filter criteria for follow-up (network) analyses.

Moreover, analysis of the above mass spec data for the binding of functional groups of proteins (Fig. 8) shows that mutation of the 3′UTR GBE (B) strongly reduces GARS binding itself to about 25%, which is consistent with the results shown in Fig. 5F, G, and H). In particular also the binding of ribosomal proteins (blue), hnRNP proteins (red, translation initiation factors and translation elongation factors is reduced. In contrast, mutation of the 5′UTR GBE (A) has a weaker effect on the binding of GARS (with a reduction to about 70%) and on binding of the above-mentioned groups of proteins. These results are separately displayed for selected functional protein groups in Fig. 9. Several proteins of both the large ribosomal 60S subunit and the small ribosomal 40S subunit were detected in significantly reduced amounts in the RNA-protein complexes with the Mengovirus reporter RNAs with the mutant version of the 3′UTR GBEs, including proteins of the small (A) and the large (B) ribosomal subunit, hnRNP proteins (C), translation initiation factors (D) and elongation factors (E). These analyses confirm the above conclusions that GARS binding to the Mengovirus 3′UTR stimulates recruitment of components of the translation machinery. Thereby, in these reporter RNA constructs containing both GBEs, mutation of the 3′UTR GBE caused a more dramatic loss of these factors, whereas mutation of the 5′UTR GBE has virtually no influence on binding of these factors. This points to the idea that binding of GARS to the 3′UTR GBE is required as a stronger “anchor” and important for recruitment of translation factors, while GARS interaction with the 5′UTR is functionally involved in translation initiation, and both binding sites may act together to stimulate translation at the 5′UTR (compare Fig. 4).

Figure 8.

Mutation of the 3′UTR GBE reduces the binding of many proteins. Volcano plots of changes in protein intensities from the data shown in Fig. 6, obtained with pull-downs with the Mengovirus reporter RNA shown in Fig. 4A, and variations in the protein intensities obtained with the GBE mutant RNAs in relation to the wild-type reporter RNA. Negative Log₂ difference values mean less protein binding to the mutant RNA compared with the WT RNA; positive values mean more protein binding. (A) Only the 5′UTR GBE mutated, (B) only the 3′UTR GBE mutated, and (C) both 5′ and 3′UTR GBEs mutated. GARS, as well as selected protein groups, are marked in color.

Figure 9.

Reduced binding of selected functional protein groups to Mengovirus RNA with a mutated GARS binding element. Relative abundance of proteins bound to 3′-end-biotinylated short Mengovirus reporter RNA with the wild-type (wt) or mutated (mt) GBE in the 5′UTR and in the 3′UTR (see Fig. 4A) after NeutrAvidin bead affinity purification and mass spectrometry analysis. The data shown in Fig. 6 were selected for functional protein groups like in Fig. 8. Each data point shows the relative abundance of a specific protein binding to the 5′UTR or 3′UTR mutant GBE RNA or the 5′-3′-UTR double mutation, while the mean value includes all proteins in the respective group. Binding to the wild-type RNA was normalized to 1.0. Significances were calculated with GraphPad´s nonparametric Wilcoxon signed-rank test, with * (P < 0.05), ** (P < 0.01), *** (P < 0.001) and **** (P < 0.0001); n.s. = not significant (P > 0.05). (A) Proteins of the small ribosomal 40S subunit. (B) Proteins of the large ribosomal 60S subunit. (C) Heterogeneous nuclear ribonucleoproteins (hnRNP proteins). (D) Translation initiation factors. (E) Translation elongation factors. “n” indicates the number of different proteins used for calculation of mt/wt ratios; significances were calculated over the ratios (mt/wt) for all proteins, asterisks indicate significance levels as in previous Figures. Underlying data are shown in Supplementary Fig. S5.

In conclusion, the results of the above unbiased systematic proteomic and bioinformatics analysis support a major role of GARS in the recruitment of specific components of RNA metabolism and the translation machinery.

The 5′UTR and 3′UTR GBEs act together to recruit translation machinery components

According to the possible cooperative role of the GARS dimer binding to both GARS binding sites in the 5′UTR and 3′UTR in recruiting translation machinery components mentioned above, we also tested GARS binding as well as recruitment of proteins involved in translation, using the isolated GARS binding elements in small RNAs. One of these RNAs constitutes only the apical part of the large central IRES domain (3 or I), which contains the 5′UTR GBE (Fig. 10A). The other small RNA constitutes the Mengovirus 3′UTR with its GBE (Fig. 10B). In addition, we also tested if the putative GBE that we also found conserved in the 3′UTR of a Cardiovirus clade 2 member (see Supplementary Fig. S1B and F), Theiler´s Murine Encephalitis Virus (TMEV), binds GARS.

Pull-down and WB experiments (performed as in Fig. 5F, G and Figs. 6–9) show that the apical part of the central IRES domain does not show strong GARS binding (Fig. 10D, lanes 1–2). The limited corresponding mass spec data obtained with the 5′UTR GBE mutation in this small RNA (not shown) are consistent with a slight reduction in GARS binding to this GBE and by that consistent with the data obtained with the longer reporter RNA context with two GBEs (Fig. 5G and H, Fig. 8A) in which the 5′UTR GBE was shown to contribute only slightly to GARS binding. In contrast, in accordance with those above results, the small 3′UTR RNA strongly binds GARS (Fig. 10D, lanes 3–4). Also, the 3′UTR of the related TMEV specifically binds GARS with similar efficiency (lanes 5–6).

Even though the sole Mengovirus 3′UTR strongly binds GARS, this Mengovirus 3′UTR alone does not specifically recruit proteins of the translation machinery (Fig. 10E – H), including ribosomal proteins and elongation factors, while it does so when it is combined with the 5′UTR in the larger reporter RNA context (Fig. 9). In contrast, the role of the 5′UTR GBE appears similar in the isolated context (Fig. 10) and in the larger reporter RNA context (Fig. 9). Recruitment of translation initiation factors is slightly impaired with the Mengovirus 3′UTR mutant alone (G), but the extent of reduction of binding of these factors is not as strong as with the 3′UTR GBE in the context of the longer RNA containing both 5′UTR and 3′UTR GBEs (compare Fig. 10G with Fig. 9D). Thus, it appears that the two GBEs at the 5′- and the 3′-end of the viral genome need to act together to recruit proteins involved in translation, as evident by the strong reduction of ribosome and translation factor recruitment by the 3′UTR mutation in the dual UTR reporter construct (Fig. 9).

Discussion

Our comprehensive search for potential GARS binding elements in the family of Picornaviridae not only confirmed that Poliovirus and other Enteroviruses contain GBEs in the IRES domain V of their 5′UTRs [22, 41], but also confirmed the presence of a second predicted GBE in the Poliovirus IRES domain II [41, 68], consistent with the apical stem and 6 nts loop of PV domain II with the conserved NNNNCCA [13]. Thus, these results can be regarded as a positive control for the bioinformatic search routines we used. In the Cardiovirus group of Picornaviruses, our bioinformatic search revealed two predicted conserved GARS binding elements, one in the Mengovirus 3′UTR, and another one in the large central domain of the 5′UTR. These GBEs are conserved irrespective of sequence variations in surrounding regions.

The GBE in the Mengovirus 3′UTR is conserved among Cardioviruses to be located within a certain distance from the polyprotein translation stop codon (with rare exceptions). A minimal GBE can form starting at nucleotide 26 of the 3′UTR, a more stable extended GBE can form starting from nucleotide 21 of the 3′UTR when a ribosome is positioned at the polyprotein stop codon (see Fig. 1B and Supplementary Fig. S1B, E, F); numbers are for Cardiovirus clade 1 members, but apply roughly also to clade 2 for which we show here that the 3′UTR of TMEV also binds GARS. Similar RNA stem-loops of about 32 – 35 nts in length with a GBE in the apical stem-loop structure (from Poliovirus, Rhinovirus and Cadicivirus) were shown to bind to GARS with K_D values of 10 to 37 nM, while the tRNA^Gly has a K_D of about 2100 nM [41, 77]. Thus, this tight binding of GARS to the viral GBE makes sure that the interaction is not outcompeted by cellular tRNA^Gly. When no ribosome is terminating polyprotein translation at the stop codon, i.e. during early steps of first round translation, an even more extended and by that more stable hairpin with an apical GBE can form, starting immediately downstream the stop codon. This means that neither translation initiation, nor ongoing translation elongation or termination would interfere with binding of GARS to the GBE in the 3′UTR. Consistent with our finding that mutation of this GBE reduces Mengovirus translation efficiency by about 50%, a previous analysis of the Mengovirus 3′UTR secondary structures revealed roughly similar results. Deletion of the Stem-Loop containing the GBE (SL I in [62]) showed a similar mild reduction in overall replication efficiency, largely observed as reduction of plaque size rather than plaque number, whereas deletion of 3′UTR sequences downstream of the GBE was essentially abolishing Mengovirus replication [62].

The GBE in the Mengovirus 5′UTR is located close to the cruciform-like apical part of the large central IRES domain that harbors a GNRA loop and a RAAA loop (with R = purine and N = any nucleotide) [10, 13, 14]. The conserved small stem-loop that contains the GBE downstream and below the cruciform was called “loop B” in [13] and “C-rich loop” in [10]. While the lower stem of the large central IRES domain is less conserved, the apical region with the cruciform and the “C-rich loop” with the GBE is more conserved not only in structure but also in primary sequence [13] (please also see Fig. 1B and Supplementary Fig. S1A, C, D). Mutations in this region impair initiation of translation [10, 14, 78], and this structure appears to be involved in the “self-organization” of IRES structure [79]. Thus, we can assume that this region essentially interacts with components of the translation machinery, even though the canonical initiation factors bind to the downstream Y-shaped IRES domain 4 or J-K-L [80, 81, 82]. Moreover, the activity of the Cardiovirus EMCV IRES is not dependent on Poly(rC) binding protein 2 (PCBP2), whereas the activity of the IRES elements of enteroviruses like Poliovirus depends on PCBP2 [83].

In analogy to the stimulation of translation initiation at the 5′end of capped mRNAs by PABPC [11, 12], it appears reasonable that also in plus-strand viruses the 3′UTR stimulates translation regulated by the 5′UTR in order to ensure efficient translation only of intact, undegraded viral genomes. However, the poly(A) tail of EMCV is only about 30 nts short [84, 85], and the generation of the poly(A) tail of Picornaviruses is a function of the viral 3D polymerase [84], suggesting a role of the picornaviral poly(A) tail in replication. In accordance with these observations, PABPC causes only very little translation stimulation from an EMCV IRES translation reporter RNA [86]. Moreover, during ongoing EMCV replication, PABPC is cleaved by the EMCV protease 3C [87], perhaps because PABPC binding would interfere with VPg-pUpU priming of minus-strand synthesis [7]. Taken together, the above findings argue against a classical role of PABPC in stimulation of EMCV/Mengovirus translation. Therefore, it appears reasonable that a surrogate function in the 3′UTR needs to stimulate translation at the 5′UTR as a “check” for RNA integrity.

Recruitment of proteins of the translation machinery by the 5′UTR and 3′UTR GARS binding sites is more dependent on the 3′UTR GBE (Figs. 5, 9, 10), while mutation of the 5′UTR GBE has a stronger effect on functional translation efficiency (Fig. 4). Since GARS is a dimeric class II aminoacyl-tRNA synthetase, these findings support an RNA loop model with a strong GARS anchor site in the 3′UTR that is more important for overall binding of the GARS dimer to the RNA undisturbed by translation initiation events, while its interaction with the 5′UTR appears to functionally modulate the efficiency of the translation machinery at the IRES.

Thereby, the Mengovirus 3′UTR does not function as a “translation enhancer” in a classical sense like by directly binding to translation initiation factors [11, 12], but it does so in a more general sense by providing a binding site in the 3′UTR for another protein (GARS) that is involved in recruitment of translation machinery components at the 5′UTR [88–90].

A very intriguing question is why exactly this specific tRNA synthetase is used for this purpose. GARS was shown [22, 68] or suspected [41] to be involved in the regulation of translation of various picornaviruses, which evolved corresponding tRNA mimicry elements in their untranslated regions.

Firstly, apart from the fact that expression of many RNA-binding proteins is restricted to particular cell types, GARS, as well as any other canonical tRNA synthetases must be sufficiently abundant in virtually any cell of the organism to allow protein synthesis. Therefore, every virus will find a sufficient number of GARS molecules in the cell, which will ensure efficient virus replication, irrespective of the cell or tissue tropism of the virus.

Secondly, at least one of the viral GARS binding elements present in the RNA of Poliovirus (domain V) [22, 68] as well as Mengovirus (3′UTR) appears to have a high affinity for GARS. Therefore, it is likely that the GARS binding element of these viruses can efficiently recruit GARS to the viral RNA irrespective of tRNA concentrations. At the same time, due to the sufficient abundance of GARS and other tRNA synthetases, a limited number of viral GARS binding sites will not interfere with the overall GARS aminoacylation capacity in the infected cell, which is also important for viral polyprotein translation.

Thirdly, GARS and many other tRNA synthetases work as a dimer [31–33, 34], and in the case of the Poliovirus IRES, GARS actually binds viral mRNA as a dimer to IRES domains II and V ([68], and unpublished observations). Thus, GARS dimers not only can mediate 3′- to 5′-end communication like in Mengovirus but also can locally serve as an IRES structure organizer like in Poliovirus. Dimer binding even opens the possibility that GARS can simultaneously recruit two RNA motifs between two RNAs to promote genome dimerization.

Finally, tRNA synthetases are components of the translation apparatus and therefore may have a capacity to bind to the translation machinery in order to efficiently deliver charged tRNAs to translating ribosomes. Consistently, GARS has been shown to weakly interact with the delta subunit of translation elongation factor eEF1 [91, 92]. In this context, it is interesting to note that GARS is not part of the Multi-Aminoacyl-tRNA Synthetase (MARS or MSC) complex [92]; perhaps, interaction of the MSC could cause steric hindrances during specific interactions with viral RNAs just due to the huge size of the complex of 1.5 MDa.

Regarding a possible resistance to infectious diseases conferred by genetic disorders [93], we can only speculate about a possible interplay of Charcot-Marie-Tooth disease-associated GARS mutations and the susceptibility to picornaviral infections. If certain pathogenic GARS mutations can restrict certain picornaviral infections, this advantage may restrict the clearance of pathogenic mutations from the human population.

Notes

Present address: Genevention GmbH, Göttingen, Germany

current address: Institute for Virology, Faculty of Medicine, Philipps-University Marburg, 35043 Marburg, Germany

Supplementary data

Supplementary data is available at NAR online.

Conflicts of interest

None declared.

Funding

This work was supported by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – [grant number 197785619] – SFB 1021 (Project A03 to D.E.A. and M.N., Project Z03 to U.L., A.W., and M.K.); DFG Research Training Group RTG 2355 (to F.D., D.V.N., and M.N.); Russian Science Foundation (RSF) [grant number 24-14-00213] (to D.E.A. and I.N.S.); and a “Promotionsabschlussförderung” of Justus-Liebig-University Giessen to D.V.N. Funding to pay the Open Access publication charges for this article was provided by: Justus-Liebig-University Giessen.

Data availability

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (52) partner repository with the dataset identifier PXD057754.