Uncovering translation roadblocks during the development of a synthetic tRNA

Abstract Ribosomes are remarkable in their malleability to accept diverse aminoacyl-tRNA substrates from both the same organism and other organisms or domains of life. This is a critical feature of the ribosome that allows the use of orthogonal translation systems for genetic code expansion. Optimization of these orthogonal translation systems generally involves focusing on the compatibility of the tRNA, aminoacyl-tRNA synthetase, and a non-canonical amino acid with each other. As we expand the diversity of tRNAs used to include non-canonical structures, the question arises as to the tRNA suitability on the ribosome. Specifically, we investigated the ribosomal translation of allo-tRNAUTu1, a uniquely shaped (9/3) tRNA exploited for site-specific selenocysteine insertion, using single-molecule fluorescence. With this technique we identified ribosomal disassembly occurring from translocation of allo-tRNAUTu1 from the A to the P site. Using cryo-EM to capture the tRNA on the ribosome, we pinpointed a distinct tertiary interaction preventing fluid translocation. Through a single nucleotide mutation, we disrupted this tertiary interaction and relieved the translation roadblock. With the continued diversification of genetic code expansion, our work highlights a targeted approach to optimize translation by distinct tRNAs as they move through the ribosome.

The in vitro transcribed tRNA UTu1 , tRNA supD , tRNA Ser , and tRNA UTu1A were aminoacylated with Ser using purified E. coli SerRS (described above). Each aminoacylation reaction contained 4 μM SerRS and 8 μM tRNA in buffer containing 50 mM HEPES (pH 7.3), 4 mM ATP, 10 mM MgCl2, 0.1 mg/mL BSA, 1 mM DTT, and 0.5 mM L-Ser. After incubation of the aminoacylation reaction at 37°C for 45 mins the reaction was quenched with 300 mM sodium acetate (pH 4.5). Standard acid phenol-chloroform extraction and ethanol were used to isolate the serylated tRNAs. Ser-tRNAs were run through a Sephadex G25 spin column (GE Healthcare) to eliminate any ATP. The tRNA concentration was determined using a NanoDrop (Thermo Scientific).
All single-molecule experiments were conducted in a Tris-based polymix buffer consisting of 50 mM Trisacetate (pH 7.5), 100 mM KCl, 5 mM ammonium acetate, 0.5 mM calcium acetate, 5 mM magnesium acetate, 0.5 mM EDTA, 5 mM putrescine-HCl, and 1 mM spermidine. Prior to the single-molecule experiments, the purified 30S and 50S ribosomal subunits (final concentration 1 μM) were mixed in a 1:1 ratio with the fluorescent dye-labeled DNA oligonucleotides complementary to the mutant ribosome hairpin extensions (3,4) at 37 °C for 10 mins and then at 30 °C for 20 mins in the Tris-based polymix buffer system. The 30S subunit was labeled with 5'-Cy3B-labeled DNA and 50S subunit was labeled with 3'-BHQ-2labeled DNA.

RSII instrumentation and data analysis
Single-molecule intersubunit FRET and tRNA occupancy experiments were conducted using a commercial PacBio RSII sequencer. The sequencer was modified to allow the collection of single-molecule fluorescence intensities from individual ZMW wells about 130 nm in diameter in four different dye channels corresponding to Cy3, Cy3.5, Cy5, and Cy5.5 fluorescence. The RSII sequencer has two lasers for dye excitation at 532 nm and 632 nm. In all experiments, data was collected at 10 frames per second (100 ms exposure time) for 6 mins using energy flux settings of the green laser at 0.60 mW/mm 2 and red laser at 0.10 mW/mm 2 .
Data analyses for all experiments were conducted with MATLAB (MathWorks) scripts written in-house (9).
Briefly, fluorescence traces from each ZMW were automatically selected based on fluorescence intensity, fluorescence lifetime, and the changes in intensity. Filtered traces exhibiting intersubunit FRET (Cy3B-BHQ-2) and tRNA Phe binding at the first codon (Cy5) were then manually curated for further data analysis. The FRET states were assigned as previously described (10) based on a hidden Markov model-based approach and visually corrected. Lifetimes measured of the different FRET states were collected across a large sample of ZMW traces (n>100) to provide a representative distribution from which a mean and standard error was calculated on MATLAB.

Cryo-EM specimen preparation
Specimens were composed of vitrified samples occupying UltrAuFoil R 2/2, 300-mesh holey Au/Au grids (Quantifoil Micro Tools). The surfaces of the grids were rendered hydrophilic by glow-discharging using H2 and O2 for 25 s at 10 W with a Solarus 950 plasma cleaner system (Gatan). For vitrification, 3 μL of the sample was applied to each grid, blotted for 3 s at a blot force of 3 inside a Vitrobot Mark IV (Thermo Fisher Scientific), and plunge-froze in liquid propane:ethane mixture (63:37, v/v) cooled with liquid nitrogen. The temperature of the specimen was kept 20°C.

Cryo-EM data processing
The beam-induced motion of the sample and the instability of the stage due to thermal drift was corrected using MotionCor2 (11). The contrast transfer function (CTF) of each micrograph was estimated using CTFFIND4 (12). Imaged particles were picked using the Autopicker algorithm included in the RELION software (13).
For tRNA UTu1 in complex with the ribosome (Supplementary Figure S4), 2,668,059 particles were picked from 11,492 micrographs. 2D classification of the four-times binned particles were used to separate ribosome-like particles from ice-like and/or debris-like particles picked by the Autopicker algorithm. 1,605,175 particles were saved for 3D classification. Of those, 1,207,471 particles were selected from high-resolution 3D classes for focused 3D classification. Focused 3D classification was performed on re-extracted particles without binning. 631,624 particles with tRNA UTu1 density were selected for auto-refinement.
For tRNA UTu1A in complex with the ribosome (Supplementary Figure S7), 10,850,093 particles were picked from 23,738 micrographs. 8,024,280 good particles were saved for 3D classification after 2D classification. Two major classes, non-rotated 70S (682,020 particles) and rotated 70S (179,559 particles) were saved for forced 3D classification based on the tRNA densities. For non-rotated classes, a major class (424,828 particles) containing A site tRNA and a minor class (23,609) containing P site tRNA were obtained from masked 3D classification. For rotated classes, two classes were found. One class (88,852 particles) contained E site tRNA and another class (67,556 particles) contained both A/P and P/E tRNAs.
Un-binned particles from these classes were subjected to auto-refinement. The final density map was sharpened by applying a negative B-factor estimated by automated procedures. Local resolution variations were estimated using ResMap (14) and visualized with UCSF Chimera (15).

Model building and refinement
Models of the E. coli 70S ribosome (5WE4) were docked into the maps using UCSF Chimera (15). All models were manually adjusted, and then de-novo built for the missing residues in Coot (16) followed by Phenix (17) and Refmac (18) refinement. All figures showing electron densities and atomic models were created using USCF Chimera (15) and PyMol Molecular Graphics Systems.

Molecular dynamic simulations
The starting P site cryo-EM tRNA structures of both tRNA UTu1 (PDB:7UR5) and tRNA UTu1A (PDB:7URM) were taken out of the ribosome for simulation in solution. Nine Mg ions were added to the tRNA structures by superimposing the tRNA backbone with tRNA Phe (PDB:1EHZ) and extracting the Mg ion's position from that structure. Each tRNA was solvated in a dodecahedron-shaped box (edge length approximately nm) with TIP3P water molecules (19) and neutralized with 0.15M NaCl. The nucleic acid structure and ions were described using the CHARMM36 force field (20). All simulations were performed using the GROMACS (version 2021.4) package (21,22). To maintain constant temperature and pressure, a V-rescale and Crescale coupling were applied, respectively. The tRNAs were equilibrated to a temperature of 300 K with a coupling time of 0.1 ps and pressure of 1 bar with a coupling time of 1.0 ps. The isothermal compressibility was 4.5 x 10 -5 bar -1 . A leap-frog integrator with an integration time step of 0.002 fs was used. The H-bonds were constrained using P-LINCS algorithm (23) with a short-range electrostatic and van der Waals cutoff of 1.2 nm. Three independent replicas were performed for each molecular system. The three trajectories were combined. From this an average structure was generated and root mean squared fluctuations (RMSF) were calculated for each base.  Figure S7. Engineering an in vivo assay for a low processivity state. (A) Strategy for sfGFP readthrough assay which determines successful suppression of a UAG codon. (B) Schematic to demonstrate that sfGFP with the amino acids UKGE for positions 2-5 results in a high processivity state as observed with the addition of 200 nM Phe while amino acids UGTT results in a low processivity state as observed with the addition of 50 nM Phe. (C) Readthrough assay measured by fluorescence intensity over cell density of tRNA UTu1 compared to tRNA supD in a high processivity (UKGE) and low processivity (UGTT) state. There is a significant decrease (p<0.0001) for tRNA UTu1 compared to tRNA supD in the low processivity (UGTT) state. Error bars shown are the standard deviation from a minimum of four biological replicates. (D) Relative activity of sfGFP readthrough with tRNA UTu1 compared to tRNA supD in a high processivity (UKGE) and low processivity (UGTT) state. There is a significant decrease (p<0.01) for tRNA UTu1 in the low processivity (UGTT) state. Error bars are the standard deviation from a minimum of four biological replicates.