Sordarin bound eEF2 unlocks spontaneous forward and reverse translocation on CrPV IRES

Abstract The Intergenic Region Internal Ribosome Entry Sites (IGR IRESs) of Discistroviridae promote protein synthesis without initiation factors, with IRES translocation by elongation factor 2 (eEF2) being the first factor-catalysed reaction. Here, we developed a system that allows for the observation of intersubunit conformation of eukaryotic ribosomes at the single-molecule level by labeling rRNA. We used it to follow translation initiation and subsequent translocation of the cricket paralysis virus IRES (CrPV IRES). We observed that pre-translocation 80S–IRES ribosomes spontaneously exchanged between non-rotated and semi-rotated conformations, but predominantly occupied a semi-rotated conformation. In the presence of eEF2, ribosomes underwent forward and reverse translocation. Both reactions were eEF2 concentration dependent, indicating that eEF2 promoted both forward and reverse translocation. The antifungal, sordarin, stabilizes eEF2 on the ribosome after GTP hydrolysis in an extended conformation. 80S–CrPV IRES–eEF2-sordarin complexes underwent multiple rounds of forward and reverse translocations per eEF2 binding event. In the presence of sordarin, neither GTP hydrolysis nor a phosphate release were required for IRES translocation. Together, these results suggest that in the presence of sordarin, eEF2 promotes the mid and late stages of CrPV IRES translocation by unlocking ribosomal movements, with mid and late stages of translocation being thermally driven.


INTRODUCTION
Viruses employ alternati v e translation initiation mechanisms to translate their mRNAs. Internal ribosome entry sites (IRESs) are highly structured, untranslated regions of the viral mRNA that promote efficient cap-independent transla tion initia tion ( 1 , 2 ). The intergenic region (IGR) IRESs of Dicistroviruses are the simplest IRESs that start translation in the absence of all initiation factors (3)(4)(5). The most studied IGR IRESs are the Cricket paralysis virus (CrPV), Taura syndrome virus (TSV) and Plautia stali intestine virus (PSIV) IRESs, which all initiate through a similar mechanism. It begins with the IRES directly binding to the 40S ribosomal subunit, forming the 40S-IRES complex ( 4 ). This complex then recruits the large ribosomal subunit. In the resulting 80S-IRES ribosomes, the IRES occupies the intersubunit space spanning from the A-to E-site, where it interacts with the phylo geneticall y-conserved 80S core (6)(7)(8). Consequently, CrPV IRES is acti v e in a broad range of organisms and can promote translation in insect cells ( 9 , 10 ), mammalian cells and extracts ( 4 , 11 ), and yeast ( 12 , 13 ). CrPV IRES is composed of three structural domains ( 14 ). Domains I and II recruit the ribosome ( 11 , 15 ). Domain III, composed of RNA pseudoknot I (PKI) with the GCU translation start codon, is located immediately downstr eam. In 80S-IRES complex es, PKI occupies the Asite of the ribosome and spans into the decoding center, where it mimics the anticodon stem of tRNA base paired to a mRNA ( 16 , 17 ). Thus, subunit joining results in pretranslocation ribosomes, and the 80S-IRES complexes undergo a translocation in order to move PKI from the A-site into the P-site to, in turn, allow for the first codon decoding (17)(18)(19)(20). IRES translocation r equir es eEF2 and places the first GCU codon into the A-site, where it is decoded by eEF1A-tRNA Ala .
Mechanisms of elongation are largely conserved between all kingdoms of life. A majority of our understanding of elongation comes from studies of prokaryotic translation. Peptidyl transfer unlocks the conformational dynamics of the ribosome. After peptidyl transfer, the acceptor ends of the A-and P-site tRNAs move into the P-and E-sites. This results in tRNAs occupying hybrid A / P and P / E states (where the letters before and after the slash designate the position of the tRNA anticodon and acceptor stems, respecti v ely) ( 21 , 22 ). Transition into the hybrid state is concurrent with an ∼9 degr ees counter clockwise (forward) rotation of the small subunit relati v e to the large subunit ( 22 , 23 ). Sim ultaneousl y, the L1 stalk moves inward and contacts the elbow of the P-site tRNA (24)(25)(26). The pre-translocation ribosomes spontaneously exchange between the classical (A / A and P / P states) and hybrid states ( 27 , 28 ), with the hybrid state being favored at physiological magnesium and polyamine concentrations ( 29 , 30 ). Translocation proceeds via a number of chimeric states, where tRNAs occupy intermediate positions between canonical tRNA binding sites (31)(32)(33)(34). A translational GTPase, EF-G, a prokaryotic homolog of eEF2, triggers conformational rearrangements of the ribosome ( 35 ), with the ribosomes transiently occupying the rotated state prior to GTP hydrolysis ( 25 , 27 , 36 , 37 ). In rotated ribosomes, domain IV of EF-G extends into the A-site and head of the 30S subunit, swi v elled by 18 degrees. These r earrangements ar e accompanied by the movement of the anticodon stem-loop, placing the tRNA into chimeric ap / P and pe / E sites. In the chimeric ap / P state, the anticodon stem-loop of the A-site tRNA moves toward the P-site into the intermediate position between the P-and A-sites, while the acceptor end is located in the P-site. In the pe / E state, the anticodon stem-loop of the P-site tRNA moves closer to the intermediate pe position, while the acceptor end of the tRNA is located in the E-site ( 31 , 32 , 38 ). Post-GTP hydrolysis, the body of the small subunit undergoes re v erse (clockwise) rotation and the head of the small subunit swi v els back. The combination of these two motions leads to translocation of the tRNAs into the P-and Esites ( 38 , 39 ). While translocation can occur in a GTP-and factor-independent manner ( 40 , 41 ), EF-G greatly accelera tes transloca tion, and GTP hydrolysis further increases the rate 10-100 fold (42)(43)(44). During translocation, the L1 stalk maintains contact with the tRNA ( 25 , 45 ). In bacteria and higher eukaryotes, after translocation is complete, the L1 stalk moves outward and allows the E-site tRNA to dissociate ( 25 , 46 , 47 ). Howe v er, in fungi, after tRNA mov ement is complete, the L1 stalk remains in a closed position and its opening is facilitated by elongation factor 3 (eEF3) ( 48 , 49 ).
The translocation of the 80S-IRES complex bears similarity to regular tRNA translocation. The 80S-CrPV IRES pr e-translocation complex es occup y non-rotated and semirota ted conforma tions tha t dif fer by a 5 degrees counterclockwise rotation of the 40S subunit body ( 17 , 18 ). It was proposed that these conformations are at equilibrium that resembles the rotated -non-rota ted sta te exchange in pretranslocation ribosomes. Reminiscent to tRNA translocation, eEF2 binding to 80S-CrPV IRES complexes induces an additional 3 degr ees counter clockwise rotation, r esulting in fully rotated ribosomes. Simultaneously, tRNA-mRNA mimicking PKI moves into the chimeric ap position on the small subunit ( 16 ). The Cryo-EM structure of 80S ribosomes with related TSV IRES ( 50 , 51 ), eEF2 and the antifungal, sordarin, showed ribosomes in five intermediate stages of IRES translocation ( 52 ). Ther e, the r e v erse body rotation and forward head swivel are accompanied by progressi v e mov ements of PKI towar d the P-site. This mov ement is followed by a re v erse head swi v el that coincides with the final placement of PKI in the P-site, yielding a translocated IRES and a non-rotated ribosome ( 16 , 52 ). The translocated 80S-CrPV IRES complex is not stable and back translocates, as was shown by both toeprinting analysis and Cryo-EM ( 5 , 53 , 54 ). Despite these studies of the 80S-CrPV IRES complex, the molecular mechanisms of IRES initiation and translocation remain unclear.
Eukaryotic translation was previously observed at the single-molecule le v el. A-site codon decoding and tRNA dynamics before and during translocation in mammalian ribosomes were tracked with tRN A-tRN A FRET (55)(56)(57). The FRET between uS19 and uL18 was used to follow subunit joining and dissociation during translation initiation and termination in yeast (58)(59)(60). Howe v er, intersubunit conformation was not directly probed in these experiments. Her e, we r eport the de v elopment and application of a single-molecule FRET system that allows for the observation of intersubunit conformation of S. cerevisiae ribosomes in real-time. We used it to follow translation initiation and translocation of CrPV IRES. We showed that 60S subunit joining places ribosomes in both semi-rotated and nonrota ted conforma tions. Pre-transloca tion 80S-CrPV IRES complex es ar e exchanging between these two conformations. Using ribosomal rotation as a proxy for translocation, w e show ed that both forward and re v erse translocation are dri v en by eEF2. GTP hydrolysis is r equir ed for rapid forwar d translocation. Howe v er, in the presence of the antifungal sordarin, which stabilizes eEF2 on the ribosome, neither GTP hydrolysis nor a phosphate r elease wer e needed, and forwar d and re v erse translocation occurred spontaneously and repea tedly a t ra tes tha t were comparable with normal translocation. Thus, sordarin captures the ribosome in an unlocked state in which the mid and late steps of IRES translocation are thermally dri v en. The e xperiments with GDP and GDPNP indica ted tha t GTP hydrol ysis is likel y needed to achie v e an unlocked state. Together, these results provide support for the Brownian model of translocation.

rRNA mutagenesis, mutant selection and validation
Ribosome mutagenesis was performed, as previously described ( 61 ), using the RDN mutagenesis system ( 62 , 63 ). Extended helix 44 of 18S rRNA was amplified from RDN-Ura plasmid containing extended hairpin ( 61 ) using F h44 SexAI and R h44 MluI primers (Supplementary Table S1). The PCR product was cloned into the RDN operon bearing the pJD180.Trp plasmid ( 64 ) by SexAI and MluI sites. Extension of helix 101 of the 25S rRNA was introduced into the plasmid, pJD180.Trp, by two-step megaprimer PCR. First, primers F h101 b and R h101.3 b were used to add the 5 extension and amplify Nucleic Acids Research, 2023, Vol. 51, No. 13 7001 the region downstream of helix 101. Primers F h101.3 a and R h101 a were used to introduce the 3 extension and amplify the r egion upstr eam of helix 101. In the second step, upstream and downstream PCR products were annealed together to get the full helix 101 extension. The insertion-bearing hairpin (called sp22) was then cloned into pJD180.Trp by BamHI and MluI sites. The resulting plasmids, pAL783 (pJD180.Trp.h44.sp68) and pAL797 (pJD180.Trp.h101.3.sp22), were transformed into an AL14 S. cerevisiae strain ( 61 ), which expressed mutant rRNA from pJD694 (URA3) plasmid. Transformants were initially selected on − Trp medium, followed by two passages of replica plating onto 5-FOA medium. RDN mutagenesis and lack of wild type loci were confirmed by RDN PCR and sequencing. A pair of primers, which are about 150 bp up and downstream of the insertion site, were used to amplify the insertion region. The product is 383 bp for the helix 44 mutant and 336 bp for the helix 101 mutant, which are 19 bp and 21 bp longer than the PCR product from the wildtype RDN oper on. The PCR pr oduct was analyzed by electrophoresis on 4% agarose gel, which migrated as a single band with a size corresponding to the mutant hairpin. Furthermore, sequencing of the PCR product confirmed the presence of the mutant hairpin and absence of wildtype contamination. All strains used and generated in this study are listed in the Supplementary Table S6.

Purification and labeling of ribosomal subunits
Ribosomal subunits were purified and labeled, as previously described ( 61 , 65 ), with the following modifications: A 5 ml liquid YPAD culture was inoculated with a single colony from a fresh YPAD plate and grown at 250 rpm in a 30 • C shaker until OD 600 = 0.8-1.0. 1 ml of the resulting culture was used to inoculate 1 l of YPAD media. Cells were grown until OD 600 = 0.8-1.0, harvested by centrifugation, and washed twice with ice-cold buffer A (30 mM HEPES-KOH pH 7.4, 100 mM KCl, 15 mM MgCl 2 , 2 mM DTT). Cells were lysed by a bead-beater and the lysate was cleared by centrifugating twice in an A27-8 × 50 rotor for 15 min at 15 000 rpm. Fines were removed by passing the lysate through a 0.45 m filter. The lysate was over lay ed on a 3 ml cushion of buffer B (30 mM HEPES-KOH pH 7.4, 100 mM KCl, 12 mM MgCl 2 , 2 mM DTT), supplemented with 1 M sucrose and spun down in a Type 70Ti rotor for 5 h at 55 000 rpm. The ribosome pellet was resuspended in buffer B and loaded on top of 5-47% sucrose gradients in buffer B. Gradients were centrifuged at 22 000 rpm for 12 h in a SW32 rotor. Gradients were fractionated to collect 80S ribosomes. Resulting ribosomes were buffer exchanged into buffer B supplemented with 250 mM sucrose, aliquoted and frozen with liquid nitrogen. Purified 80S ribosomes were labeled with corresponding oligonucleotide dye at a 1:2 molar ratio by sequentially incubating at 42 • C for 1 min, 37 • C for 10 min and 30 • C for 10 min. The reaction was loaded on top of 10-30% sucrose gradients (30 mM HEPES-KOH pH 7.4, 500 mM KCl, 7.5 mM MgCl 2 , 2 mM DTT supplement with sucrose) and centrifugated at 24 000 rpm for 12 h in a SW32 rotor. Gradients were fractionated and 40S and 60S ribosomal subunits were collected. Labeled ribosomal subunits were then concentrated and the buffer was ex-changed to a storage buffer (30 mM HEPES-KOH pH 7.4, 100 mM KCl, 5 mM MgCl 2 , 2 mM DTT and 250 mM sucrose) using Amicon ® Ultra-15 100 kDa centrifugal filters. Ribosomal subunits were aliquoted and frozen with liquid nitrogen. The purity and integrity of the subunits were validated by composite agarose-acrylamide gel electrophoresis. The labeling efficiency was measured spectrophotometrically and was found to be 92% and 88% for 40S and 60S subunits, respecti v ely. The measurement was confirmed by quantifying the fluorescence intensity of the 40S-Cy3B and 60S-Cy5 subunit bands of the composite gel.
eEF1A. Untagged eEF1A was purified from the S. cerevisiae strain, CB010, as previously described ( 66 ) with the following modifications: The lysate was clarified by centrifugation in an A27-8 × 50 rotor at 15 000 rpm for 30 min, followed by centrifugation in a Type 70 Ti rotor at 50 000 rpm for 90 min. The clarified lysate was applied to a gravity-flow 7 ml DEAE column equilibrated with eEF1A buffer A (20 mM Tris-HCl, pH 7.5, 100 mM KCl, 0.1 mM EDTA, 1 mM DTT, 25% glycerol). After loading the lysate, the column was washed with two column volumes (CV) of eEF1A buffer A. The flow-through and column wash were pooled and loaded on a gravity-flow 5 ml SP Sepharose column equilibrated with eEF1A buffer A. The SP Sepharose column was washed with 5 CV of eEF1A buffer A before eluting. eEF1A was eluted with eEF1A buffer 150B (eEF1A buffer A with 150 mM KCl) for 5 CV at first, then eluted with buffers 250B, 350B and 500B for 4 CV. 5 ml fractions were collected. eEF1A-containing fractions were pooled and were concentrated to ∼5 ml. The concentrated protein was applied to a 26 / 600 Super de x 75 size exclusion column equilibrated with a storage buffer (20 mM Tris-HCl, pH 7.5, 100 mM KCl, 0.1 mM EDTA, 1 mM DTT, 25% glycerol). Fractions containing pur e eEF1A wer e identified by SDS-PAGE, pooled, and concentrated using Amicon ® Ultra-15 30 kDa MWCO centrifugal filter units. eEF1B α. The pET28b plasmid, carrying eEF1B ␣-TEV site-His6, was transformed into BL21-RIPL Esc heric hia coli cells. The protein was expressed and purified, as previously described ( 65 ). eEF2. 6 × His tagged eEF2 was purified from the S. cerevisiae strain, TKY 675, as previously described ( 67 ), with the following modifications. Cells were grown in YPAD at 30 • C to OD 600 = 1.5 and harvested by centrifugation. The cells were resuspended with ice-cold lysis buffer (50 mM potassium phosphate, pH 7.6, 300 mM KCl, 1 mM DTT, 10 mM imidazole) and lysed using a bead-beater. After lysis, the pH of the lysate was adjusted to 7.7 with 1 M un-titrated Tris and clarified by centrifugation in an A27-8 × 50 rotor at 12500 rpm for 20 min. The resultant supernatant was centrifuged at 50 000 rpm in a Type 70 Ti rotor for 90 min. The lysate was passed through a 0.22 m filter and loaded onto a 5 ml HisTrap HP column (GE Healthcare). The column w as w ashed with 25 ml of lysis buffer containing 30 mM imidazole. Proteins were eluted with lysis buffer containing 250 mM imidazole. The eluate was buffer exchanged into buffer A (20 mM Tris-HCl pH 7.6, 30 mM KCl, 5 mM MgCl 2 , 1 mM DTT). Protein was loaded onto a 5 ml HiTrap Q column. Column was washed with 6 CV of buffer A. Proteins were eluted with a 15 CV gradient of 30-500 mM KCl. Fractions containing eEF2 were pooled and concentrated using an Amicon ® Ultra-15 30 kDa MWCO centrifugal filter and buffer exchanged into stora ge b uffer (20 mM Tris-HCl, pH 7.5, 100 mM KCl, 0.1mM EDTA, 1 mM DTT, 25% glycerol).

Pr epar ation and biotinylation of CrPV IRES
CrPV IRES was pr epar ed by T7 runoff transcription. The template plasmid, T7-IGR-CrPV-IRES ( 65 ), bears a 6025-6232 nts region of CrPV genome that contains a CrPV IGR IRES loca ted a t the 6030-6219 genomic coordina tes. The plasmid was linearized with NarI and transcribed with NEB HiScribe ™ T7 High Yield RNA Synthesis Kit. The resulted RNA was purified by gel filtration on a SEC650 column, as described before ( 69 ). RNA was concentrated using a 30 kDa MWCO spin concentrator. The 5 -biotinylated DNA oligonucleotide was annealed to the 3 end of the RNA to a region located 39-54 nts downstream of pseudoknot I of CrPV IRES, as described before ( 65 ).

Single-molecule imaging and data analysis
A home-built total internal r efection fluor escence (TIRF) microscope was used for single-molecule imaging. The prism-based system was built on the base of a Nikon Ti2-A inv erted microscope. The vie wing area of ∼3000 m 2 ( ∼ 55 × 55 m) was illuminated with a 532 nm continuous wave fiber laser (MPB Communications). The excitation laser was spectrally cleaned up with 532 / 20 nm bandpass filter. The laser power was 50 mW for 100 ms imaging and 150 mW for 25 ms imaging. The sample was imaged with a CFI Plan Apo 60 ×, N.A. = 1.2, water immersion objecti v e. Cy3B and Cy5 fluorescence was separated using Photometric Quad View with a 640 long pass dichroic mirror and imaged in the 585 / 70 and 700 / 75 channels for Cy3B and Cy5 dyes, respecti v ely. All filters and mirrors were obtained fr om Chr oma Technology. Non-binned images were recorded using an Andor iXon Ultra 897 EMCCD camera with a 600 EM gain and no pre-gain. The solvent delivery system was built using a J-Kem syringe pump equipped with a 100 l syringe. The imaging chambers were constructed out of a quartz slide, double-sided tape, and a coverslip. The chamber volume was a pproximatel y 8 l. Both slides and coverslips were treated with PEG 5000 and PEG-biotin 5000, thus passivating the surface and providing molecular handles for immobilization ( 28 ). PEG-Biotin-treated quartz slides were washed twice with 200 l of TP50 buffer (50 mM Tris-HCl, pH 7.5, 100 mM KCl) and incubated with 1 M neutravidin, 0.67 mg / ml BSA, and 1.3 M of blocking DNA oligonucleotide duplex for 5 min. Slides were washed twice with 200 l of TP50 and one time with a reaction buffer containing 30 mM HEPES-KOH, pH 7.4, 100 mM KCl, 5 mM MgCl 2 and 1 mM Spermidine. 40S-CrPV IRES complex es wer e pr epar ed by incubating 100 nM of 40S-Cy3B fluorescent subunits and 100 nM CrPV IRESbiotin in a reaction buffer at 30 • C for 5 min. 40S-CrPV IRES complex es wer e diluted to 0.8 nM with the reaction buffer and immobilized for 5 min. Non-immobilized ribosomes were removed by washing the chambers with reaction buffer. Finally, the chambers were washed with the reaction buffer supplemented with 1mM GTP, 2 mM TSY, and an oxygen scavenging system composed of 2.5 mM protocatechuic acid and 0.06 U / l protoca techua te dehydrogenase ( 70 ). A 50 l deli v ery mix was flowed to the slide simultaneously with the start of observation. For experiments with sordarin (Cayman Chemical #26255), the drug concentration was 20 M. All experiments were performed at 21 • C.
Fluor escent traces wer e extracted using home-written MATLAB scripts. Time-dependent donor and acceptor fluorescence intensities were converted into FRET efficiency using the formula, FRET = I acceptor / ( I acceptor + I donor ). Individual molecules were identified by single step photobleaching e v ents. FRET trajectories w ere idealized using e bFRET software ( 71 , 72 ). According to ELBO criterion, the threestate model (two FRET states and FRET off) was preferable in most experiments. However, the minority of traces ( ∼1%) clearly showed three distinct FRET states. Unfortunately, signal-to-noise hampered the identification of this third state in the majority of molecules. Ther efor e, to avoid overinterpretation of the results, we selected a simpler two-FRET-state model.
All statistical analysis was performed in MATLAB. Kinetic parameters were extracted by building a cumulati v e probability plot and fitting observed data to linear, single-, or doub le-e xponential functions. The reported 95% confidence interval is a goodness of fit. The rotating molecules were defined as molecules with at least two unambiguous FRET transitions between the fitted FRET states. All rela ted da ta size and statistical parameters wer e r eported in Supplementary Tables S2-S5.

Following 80S-CrPV IRES complex formation in real-time
To follow the conformation of eukaryotic ribosomes in realtime, we site-specifically labeled yeast small and large ribosomal subunits with fluorescent dyes. Phylo geneticall y variable loops of helix 44 of 18S rRNA and helix 101 of 25S rRNA were extended with distinct weak hairpins (Figure 1 A). Isolated strains were confirmed to have only mutant rDNA by PCR and RDN loci sequencing (Supplementary Figure S1). Low thermodynamic stability of the inserted hairpins allows for site-specific annealing of fluorescent DNA oligonucleotides at physiological temperatures ( 61 , 73 ). Small ribosomal subunits were site-specifically labeled with Cy3B dye and large ribosomal subunits were labeled with Cy5 dye by annealing fluorescently labeled DNA oligonucleotides to the h44 and h101 extensions, respecti v ely (Supplementary Figure S1).
The Cryo-EM structures of both 80S ribosomes, in complex with tRNAs and IGR IRESs, show that the apical positions of h44 and h101 are within FRET distance and are ∼60 Å apart in the non-rotated conformation ( 18 , 52 , 74 ).
The distance between dyes increases by ∼10 Å when the ribosome transitions from the non-rotated to rotated conformation (PDB IDs: 3J77 and 3J78) ( 74 ). This distance between dyes and FRET efficiency are anticorrelated. This allows us to follow 60S subunit joining by the appearance of FRET and ribosome conformation by measuring FRET efficiency.
Using this system, we first followed 80S ribosome assembly on CrPV IRES. The 40S-CrPV IRES complex is e xceedingly stab le with a lifetime of more than 400 s ( 65 ). This allowed us to pr epar e and surface-immobilize 40S-Cy3B-CrPV IRES complexes using biotinylated CrPV IRES. Then, Cy5-labeled large ribosomal subunits were deli v ered sim ultaneousl y with the start of observations. 80S ribosome formation was detected by an appearance of FRET (Figure 1 B). 60S ribosomal subunit binding was efficient, with 70% of the 40S-IRES ribosomes forming 80S-IRES complexes. The large subunit arrival rate was fast, with k obs = 0.19 ± 0.012 s −1 at 50 nM 60S subunits (Supplementary Table S2). Consistent with bimolecular reaction, the arrival rate was concentration dependent (Supplementary Figure S2A). The rate constant was determined as a slope of concentration dependence and was found to be 3.5 ± 1.0 M −1 s −1 (Figure 1 C). The measured rate is comparable with 60S subunit recruitment rates during factordri v en initiation measured by ensemble and single-molecule approaches (0.076 and 0.2 s −1 , respecti v ely, at 100 nM 60S ribosomal subunits ( 59 , 75 )). Consistent with previous studies ( 65 ), 80S-CrPV IRES complex es wer e stable, with lifetimes longer than the fiv e minute observation window (Supplementary Figure S2B). Thus, labeled ribosomal subunits are acti v e in IRES-dri v en initiation.

Spontaneous rotations in pr e-tr anslocation 80S-CrPV IRES comple x es
The Cryo-EM structures show 80S-IRES complexes in both semi-rotated and non-rotated conformations. It was proposed that 80S-IRES complexes are spontaneously ex-changing between these conformations ( 17 , 18 ). In the experiments described above, ribosomes predominantly occupied the low, ∼0.22, FRET state and transiently transitioned into the higher, ∼0.34, FRET state (Figure 2 B). Cryo-EM structures of 80S IRES complexes showed that the distance between the end of helix 44 and the end of helix 101 decreases by ∼5 Å as the ribosome changes from the semi-rotated to the non-rota ted conforma tion (Supplementary Figure S3) ( 17 , 18 ). Thus, we assigned the low (0.22) FRET state to the semi-rotated ribosomes and the high (0.34) FRET state to the non-rotated ribosomes. Consistent with two conformations, a total FRET efficiency histogram of 80S-IRES complexes had a right shoulder and was best approximated by a double Gaussian fit. The FRET states of the double Gaussian fit matched individual FRET distributions and had an efficiency of 0.2 ± 0.06 (95% CI) for the low FRET state and 0.33 ± 0.07 (95% CI) for the high FRET state (Supplementary Figure S2C).
The experiments described above were done at a 100 ms exposure time. At this condition, 27% of 80S ribosomes showed spontaneous rotations, with an average of 2.3 rotations per trace (average FRET duration is 169 ± 10.3 s) (Supplementary Table S2). This is substantially less frequent than what was observed in pre-translocation bacterial and human ribosomes ( 27 , 76 ). Importantly, spontaneous transitions of pre-translocation ribosomes in complex with tRNAs, are fast and occur on the sub-second timescale ( 27 , 28 , 77-79 ). We hypothesized that spontaneous transitions (and particularly the transient high state) are faster than 100 ms , and, thus , escape detection due to timeaveraging by the imaging camera ( 80 ). To determine if the low fr ame r ate masked spontaneous tr ansitions, we imaged 80S-CrPV IRES complexes at a 25 ms resolution. The increased time resolution made deli v ery e xperiments challenging, due to a lower signal-to-noise ratio (SNR) caused by lower photon budget and SNR decrease due to mechanical stresses produced by the solv ent deli v ery system. Therefore, 25 ms experiments were done with pre-formed 80S-CrPV IRES complex es. Ther e, 83% of pr e-translocation ribosomes underwent spontaneous rotations with an average of 23. The synchronization point was set as time 0. The FRET distributions for semi-rotated ribosomes and non-rotated ribosomes were best described by a single Gaussian. The estimated mean FRET efficiency was 0.22 ± 0.08 (95% CI) for the semi-rotated state and 0.34 ± 0.08 (95% CI) for the high FRET state. ( C ) Bar chart of mean dwell times for re v erse (b lue) and forwar d (orange) rota tions a t various Mg 2+ concentra tions a t 25 ms exposure time. Table S2). The dwell times were best described by a double exponential function (Supplementary Figure S4), indicating kinetic heterogeneity. Ribosomes highly pr eferr ed the semi-rota ted sta te, with a mean dwell time being 0.97 s, while the non-rotated state was a pproximatel y three times shorter at 0.38 s (Figure 2 C). The median high FRET state dwell time was 0.05 s, which explains why spontaneous rotations wer e rar ely observ ed at 100 ms e xposure time and confirms that they have been masked by low fr amer ate (Supplementary Table S2). Together, this confirms the hypothesis that 80S-IRES complexes spontaneously exchange between semi-rotated and non-rotated conformations.

eEF2 facilitates intersubunit dynamics
Translocation is accompanied by a re v erse subunit rotation and results in non-rotated ribosomes. Translocation of 80S-tRNA complexes is unidirectional and results in stable translocated ribosomes. However, translocated IRES complex es ar e unstable and back-translocate ( 17 , 54 ). To follow translocation of CrPV IRES, we co-deli v ered 60S subunits, eEF2, and GTP to the immobilized 40S-CrPV IRES complexes. Spontaneous rota tions complica te identifica tions of conforma tional transitions associa ted with translocation. Thus, we conducted these experiments at 100 ms time resolution, so that spontaneous transitions in the pr e-translocation complex es ar e masked by camer a aver aging and predominantly eEF2-induced rotations are visible.
In the presence of eEF2-GTP, the 80S ribosomes began exchanging between the semi-rotated (0.23) and nonrotated (0.34) FRET states (Figure 3 A) at rates that are drastically different from spontaneous rotations (Figure 2 C, Supplementary Figure S5A). The semi-rota ted sta te mean dwell time was 8.4 s at 100 nM eEF2 and 0.97 s in the ab-sence of eEF2. The non-rota ted sta te mean dwell time was 5.9 s at 100 nM eEF2 and 0.38 s in the absence of eEF2, indica ting tha t these transitions were eEF2-induced (Supplementary Tables S2 and S3). Translocation places the ribosome into the non-rotated, high FRET state. Thus, a lowto-high FRET transition, in the presence of eEF2, corresponds to the clockwise intersubunit rotation and forward translocation. By extension, a high-to-low FRET transition corresponds to the counterclockwise rotation, and back translocation.
To understand the translocation mechanism, we varied eEF2 concentration in the deli v ery mix from 10 to 1000 nM. The rates of both forward translocation (low-tohigh FRET) and, surprisingly, re v erse (high-to-low FRET) translocation were dependent on eEF2 concentration, indicating a bimolecular mechanism. The kinetics of both reactions wer e r emar kab l y similar. Both were best a pproximated by doub le e xponential fits. The rates of both fast and slow components increased with eEF2 concentrations, while relati v e amplitude increased for the fast component and decreased for the slow component (Supplementary Table S3). Both fast and slow components sa tura ted a t ∼200 nM eEF2 for both the forward and reverse rotations (Figure 3 C). Interestingly, this is comparable with the previously measured K m of GTP hy drolysis b y EF-G (80 nM ( 81 ) to 400 nM ( 82 )). Thus, eEF2 efficiently engages both conformations of the 80S-IRES complexes and promotes both forward and re v erse transloca tion. The na tur e of both phases r emains to be delineated. In these experiments, we are observing multiple cycles of translocation. Presumably, eEF2 dissociates and rebinds after e v ery reaction, which makes this a multistep process. Such behavior might potentially result in double exponential kinetics. Furthermore, both fast and slow phases are sa tura ted a t 1 and 0.2 s −1 , which is substantially slower than tRNA transloca tion ( 83 ), indica ting that IRES imposes additional barriers for translocation that may be surpassed in a multistep fashion. An insufficient imaging fr amer ate might also lead to the kinetic hetero geneity ( 80 , 84 ). Lastl y, ribosomal complex es ar e likely not uniform, as shown by kinetic heterogeneity of spontaneous intersubunit rotations, potentially leading to double exponential kinetics of translocation.
As a control, we replaced GTP with GDP and the nonhydrol yzable GTP analo g, GDPNP. eEF2-GDPNP stabilizes ribosomes in the fully rotated state, characterized by an additional 3 • counterclockwise rotation ( 16 ) that should further decrease the FRET efficiency. In accordance with this, 80S-IRES-eEF2-GDPNP complexes had a lower FRET of 0.17 ± 0.04 (Figure 3 Table S5) that corresponds to the fully rotated ribosomes. Translocation, in the presence of GDP, is slow due to a low affinity of eEF2-GDP to the ribosome and the absence of GTP hydrolysis ( 42 , 43 ). Here, eEF2-GDP had a modest effect on ribosome conformation ( Supplementary Figure S6). It slightly increased the number of rotated to nonrota ted sta te transitions, but did not result in stably translocated ribosomes, potentially due to a limited observation time.

Sordarin bound eEF2 accelerates intersubunit rotation
Sordarins are a class of tetracyclic diterpene glycosides that specifically inhibit fungal protein synthesis, but do not affect bacterial, plant, or mammalian translation (85)(86)(87). Sordarin acts by stabilizing eEF2 on the ribosome ( 86 ). It does not inhibit GTP hydrolysis and does not affect phosphate release ( 52 , 86 ). It binds to the le v er arm of eEF2 in the cleft between domains III, IV and V and restricts compaction of eEF2, thus stabilizing eEF2 in the extended conformation and pre v enting eEF2 dissociation ( 85 , 88 ). We used sordarin to probe the role of eEF2 in IRES translocation. First, w e follow ed 80S ribosome formation in the presence of sordarin by delivering 60S ribosomal subunits and sordarin (no eEF2) to 40S-CrPV IRES complexes. As expected, sordarin alone did not affect 80S complex formation or spontaneous rotations in pre-translocation 80S-IRES complexes (Supplementary Table S5). Then, we repeated these experiments with sordarin and eEF2. Similar to the experiments without the drug, eEF2-soradarin promoted intersubunit rotations (i.e. forward and reverse translocation) (Figure 4 A). Howe v er, reaction rates were greatly increased at low eEF2 concentrations. Furthermore, the satura tion concerta tion of eEF2 was 20 times lower in the presence of sordarin than in the absence of the drug (10 nM vs 200 nM eEF2) (Figures 4 C, Supplementary Table S4). Together, with pr evious r eports showing that sordarin stabilizes eEF2 on the ribosome, this suggests that when sordarin is present, eEF2 promotes multiple rounds of forward and re v erse rotations without dissociating from the ribosome.
The Cryo-EM structure of 80S ribosomes with TSV IRES, eEF2 and sordarin provide us with structural information needed to connect ribosomal rotations to the functional states of the ribosome in the presence of sordarin. As shown in the 80S-TSV IRES-eEF2-sordarin complex structur e, the degr ee of subunit rotation correlates with the position of PKI (Supplementary Figure S3), which shows a gradual movement of PKI between the A-and P-sites as the ribosome rotates. Over all, an un-tr anslocated 80S-IRES complex, characterized by the rotated conformation, transitions into a fully translocated complex, characterized by the non-rota ted conforma tion. This implies tha t the subunit rotations of the 80S-TSV IRES-eEF2-sor darin comple x are correlated with IRES translocation ( 52 ). We used this relationship between IRES translocation and subunit rotations in our data interpretation. In our experiments with sordarin and eEF2, a clockwise rotation always corresponds to forward translocation and a counterclockwise rotation corresponds to re v erse translocation. Thus, in the presence of sor darin, 80S-CrPV IRES-eEF2 comple xes continuously translocate and back translocate.
EF-G, a prokaryotic analog of eEF2, cannot efficiently exchange nucleotides while being ribosome-bound ( 42 ), suggesting that in the presence of sordarin, the energy of GTP hydrolysis is not r equir ed for repeated rounds of forwar d and re v erse translocation. To determine if GTP hydr olysis pr ovides energy to IRES transloca tion, we repea ted experiments with GDPNP and GDP. In the presence of sordarin, more than 80% of 80S-IRES-eEF2-GDPNP complex es wer e characterized by stable FRET (Figur e 5 A, Supplementary Table S5), which is consistent with the previous proposal that sordarin acts post-GTP hydrolysis ( 45 , 60 ). On the other hand, eEF2-sordarin promoted subunit rotations in the presence of GDP (Figures 4 B and 5 A). The first conformational transition in the presence of GDP and sordarin was slightly slower, taking 8.4 s (versus 6 s in the presence of GTP), while the subsequent transitions were indistinguishable from transitions in the presence of GTP (Figure 5 B, Supplementary Figure S7). Thus, in the presence of sordarin, GTP hydrolysis is not r equir ed for r epeated rounds of forward and reverse translocation.
To exclude the possibility that eEF2 rebinding is driving these e v ents, we conducted wash-off e xperiments. In these experiments, immobilized 80S-IRES ribosomes were preincubated with eEF2-GTP, either in the presence or absence of sordarin, and immobilized on the surface of a microscope slide. Ribosomes were then imaged to confirm the expected behavior. After 30 s of imaging, eEF2 and nucleotides were replaced with buffer that did not contain eEF2 and GTP and observation continued. Prewash imaging served as a control and showed that ribosomes underwent rotations identical to ones seen in the experiments described above. After the wash-off, control ribosomes (preincubated without sordarin) showed stable FRET with an average of 0.4 rotations per ribosome (in a 30 s observation window after the wash). Howe v er, ribosomes that were preincubated with sordarin continued to fluctuate between high and low FRET states, with an average of 6.1 FRET transitions per ribosome ( Figure 6 , Supplementary Table S5). Because the wash buffer did not contain eEF2 and GTP, the wash precluded the possibility of eEF2 rebinding to the 80S-IRES complexes or nucleotide exchange. Ther efor e, r epeated rounds of translocation are not caused by eEF2 rebinding, but rather promoted by stably bound eEF2 and are thermally dri v en. This interpretation is also consistent with faster translocation rates in the presence of sordarin, where the reaction essentially becomes The dwell times before the first rotations were fitted by a doub le e xponential function. The first rotation is slower with GDP, consistent with a decreased eEF2-GDP affinity to the ribosome.
unimolecular and, thus, is expected to be faster than bimolecular translocation in the absence of the drug.

eEF3 does not affect translocation of CrPV IRES
To investigate the role of yeast elongation factor 3 (eEF3) in IRES translocation, 500 nM eEF3-ATP, Cy5-60S, and 10 or 200 nM eEF2-GTP were co-deli v ered to immobilized 40S-CrPV complexes. A kinetics analysis re v ealed that the rates of both forward and reverse rotations were unaffected by the presence of eEF3 (Supplementary Figure S8). FRET intensity distributions were similarly unaffected. It is possible that eEF3 functions ra pidl y at timescales faster than the time resolution of our experiments (100 ms). To examine this possibility, we repeated the experiments outlined above in the presence of the non-hydrolyzable ATP analogue, ADPNP. It stabilizes eEF3 on the ribosome and prevents the L1 stalk opening by eEF3 ( 49 , 89 ). Similarly, rota tion ra tes and the FRET intensity remained unchanged (Supplementary Figure S8). Thus, we concluded that eEF3 has no effects on the translocation of the 80S-CrPV IRES complex. states. This directly confirms a hypothesis, based on Cryo-EM structures, that 80S-CrPV IRES complexes sample different rota tional conforma tions ( 17 , 18 ). Importantly, 80S-IRES complex es ar e kineticall y hetero geneous, as spontaneous exchange was best described by double exponential fits. Furthermor e, ther e wer e two populations of ribosomes. About 80% of the pre-translocation complexes were predominantly occupying the semi-rotated state, while 20% of the pr e-translocation complex es mainly occupied the nonrota ted sta te (Supplementary Figure S9). This implies tha t the system is not at a simple two state equilibrium and additional, unobserved intermediates might exist. Spontaneous rotations in pre-translocation 80S-IRES complex es r esemble the behavior of tRNA pretranslocation complexes, thus indicating similarities between the two types of tr anslocation. The r ate of spontaneous rotations in pr e-translocation 80S-IRES complex es is comparable to that of bacterial pre-translocation ribosomes complexed with tRNA, as measured by smFRET studies ( 46 , 90 , 91 ). When eEF2 binds to the 80S-IRES complex, it transiently places the ribosome into the rotated conformation. One point of contention is whether eEF2 can bind to the non-rotated conformation or whether eEF2 binding captures the rotated state of the 80S-IRES complex. The ability of translocase to interact with nonrotated ribosomes is well described. EF-G sampling of non-rotated ribosomes with empty A-sites was detected by single-molecule fluorescence ( 43 ) and visualized by Cryo-EM ( 92 ). L11-tRNA ( 33 , 93 ) and S6-L9 FRET ( 91 ) indica te tha t EF-G engages both rotated and non-rotated ribosomes. The ensemble measurements of translocation kinetics argue that EF-G can engage both intersubunit conformations ( 94 ). Similarly, eEF2 also can recognize non-rotated ribosomes. eEF2-GDPNP causes a single nucleotide toeprint shift in the 5 direction during posttranslocation in non-rotated ribosomes ( 95 ). At a high . The ribosome oscilla tes between the non-rota ted and rota ted sta tes before the w ash. After the w ash, rotations disappear in the experiment without sordarin, but the ribosome continues fluctuating in the presence of sordarin. ( B ) Number of transitions before and after the wash for eEF2 (blue) and eEF2 plus sordarin (red). ( C ) Traces were synchronized with the wash time set to zero. Ribosomes preincubated with eEF2 (left) ra pidl y settled in the low FRET state, while ribosomes preincubated with eEF2 and sordarin (right) experienced multiple rounds of rotations.

DISCUSSION
concentration of E-site tRNA, eEF2 can promote re v erse translocation, a reaction in which a factor is also expected to engage non-rotated ribosomes ( 96 ). Our kinetics analysis showed eEF2 promoted both re v erse and forwar d subunit rotations in a concentration dependent manner, which also suggests eEF2 can efficiently recognize both semir otated and non-r otated ribosomes. Together, these results demonstra te tha t the ability of translocase to interact with non-rotated ribosomes is conserved between prokaryotic and eukaryotic elongation, as well as in tRNA and IRES translocation.

eEF3 in IRES translocation
The lack of eEF3 effects on CrPV translocation are in line with our understanding of eEF3 function. During tRNA translocation, the L1 stalk contacts the elbow of the P-site tRNA. Post-translocation eEF3 opens the L1 stalk, thus allowing E-site tRNA dissociation. In the pre-translocation 80S-IRES complex, the domain I of CrPV IRES displaces the L1 stalk, placing it in a position that is more open than in the tRNA pre-translocation complex ( 16 ). The translocated IRES pushes the L1 stalk open e v en further, beyond to what is found in post-translocation ribosomes complexed with tRNA ( 54 ). This movement of the L1 stalk is a result of outwar d mov ements of domains I and II from the E-site that occur concurrently with re v erse head swi v eling during the late stages of IRES translocation ( 18 ). Thus, opening of the L1 stalk is a part of IRES translocation, which explains why eEF3 had no effect on the dynamics of intersubunit rotations and, congruently, had no effect on the recruitment of the first tRNA ( 65 ).

Role of eEF2 in IRES translocation
In the presence of eEF2 and GTP, 80S-CrPV IRES complex es r epea tedly transloca ted and back transloca ted. Ribosomes became static after eEF2 was washed off, indica ting tha t repea ted transloca tion e v ents were due to multiple rounds of eEF2-ribosome interactions. Both reverse and forward rotations are promoted by eEF2 and require GTP hydrolysis (at least for forward translocation), as eEF2-GDP had only modest effect on ribosome conformation (Supplementary Figure S6). In prokaryotes, translocation in the presence of GDPNP is slower than in the presence of GTP ( 42 , 43 ). Similarly, < 20% of 80S-IRES-eEF2 complexes underwent translocation in the presence of GDPNP, as most ribosomes were stabilized in the fully rota ted sta te (Figures 3 B and 5 A). Thus, eukaryotic IRES translocation is remar kab ly similar to tRNA translocation. The Brownian mechanism of translocation has been long proposed ( 97 , 98 ). There, the movement occurs stochastically due to Brownian forces and directionality is provided by rectifying energy. It is supported by a number of studies that showed translocation and conformational rearrangements associated with translocation can occur spontaneously (re vie wed in ( 99 , 100 )). Structures of translocation intermedia tes show tha t EF-G / eEF2 extends into the Asite of the ribosome. This suggests that translocase acts as a 'pawl', pre v enting re v erse sliding of the tRNA. Our results show that the ribosome spontaneously translocated and back translocated in the presence of eEF2 and sordarin. Neither r eaction r equir ed GTP hydrolysis nor a phosphate release, as shown by wash experiments. The spontaneous and rapid nature of these reactions indicates that sordarin captures the ribosome in the unlocked state, whereas IRES translocation occurs over a flat energy landscape, as pre viously suggested ( 52 ). Remar kab ly, after the first somewha t slower transloca tion e v ent in the presence of GDP and sordarin, subsequent rounds of translocation were as fast as translocation in the presence of GTP (Supplementary Figure S7). This suggests that GTP hydrolysis is used to achie v e the unlocked state. Once the ribosome is unlocked, IRES translocation is thermally dri v en in the presence of the drug. Here, eEF2 acts as a classical enzyme that decreases the energy barrier of translocation making possible for it to occur ra pidl y due to lo w ener gy Bro wnian fluctuations. The release of rectifying energy gi v es translocation directionality. The ribosome being unlocked during intersubunit rotation suggests that release of rectifying energy occurs after intersubunit rotation and is possibly associated with the dissociation of eEF2 and / or final positioning of tRNA with polypeptide chain in the Psite. Together, these results support the Brownian mechanism by directly showing that the mid and late stages of translocation are thermally dri v en in the presence of sordarin.
CrPV IRES mimics tRNA and translation factors. Despite the ability of the IRES to trigger conformational changes in the ribosome associated with initiation and elonga tion, the d ynamics and conforma tion of the 80S-IRES pre-transloca tion complex dif fer from tha t of tRNA transloca tion. Further investiga tion of the d ynamics of eukaryotic tRN A translocation, specificall y high-resolution structures of the 80S-tRNA complex with eEF2 and sordarin, will be needed to see if tRNA translocation is thermally dri v en by the eEF2-sor darin comple x.

DA T A A V AILABILITY
The data underlying this article will be shared on reasonable request to the corresponding author.

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.