Experimental model and culturing

HEK293S-derived cells (T501, originally purchased from Thermo Fisher Scientific) were grown in Freestyle 293 expression medium containing 5% tetracycline-free FBS in vented shaking flasks at 37 °C, 5% CO2 and 120 rpm (550g). The cell line tested negative for Mycoplasma contamination. The culture was scaled up sequentially, by inoculating at 1.5 × 106 cells per ml and subsequently splitting at a cell density of 3.0 × 106 cells per ml. Finally, a final volume of 2 l of cell culture at a cell density of 4.5 × 106 cells per ml was used for mitochondrion isolation, as described below61.

Mitoribosome purification

HEK293S-derived cells were collected from the 2-l culture when the cell density was 4.2 × 106 cells per ml by centrifugation at 1,000g for 7 min at 4 °C. The pellet was washed and resuspended in 200 ml PBS. The washed cells were pelleted at 1,000g for 10 min at 4 °C. The resulting pellet was resuspended in 120 ml MIB buffer (50 mM HEPES–KOH, pH 7.5, 10 mM KCl, 1.5 mM MgCl2, 1 mM EDTA, 1 mM EGTA, 1 mM dithiothreitol (DTT) and complete EDTA-free protease inhibitor cocktail (Roche)) and allowed to swell in the buffer for 15 min in the cold room by gentle stirring. About 45 ml of SM4 buffer (840 mM mannitol, 280 mM sucrose, 50 mM HEPES–KOH, pH 7.5, 10 mM KCl, 1.5 mM MgCl2, 1 mM EDTA, 1 mM EGTA, 1 mM DTT and 1× complete EDTA-free protease inhibitor cocktail (Roche)) was added to the cells while stirring before pouring into a nitrogen cavitation device kept on ice. The cells were subjected to a pressure of 500 psi for 20 min before releasing the nitrogen from the chamber and collecting the lysate. The lysate was clarified by centrifugation at 800g and 4 °C for 15 min to separate the cell debris and nuclei. The supernatant was passed through a cheesecloth into a beaker kept on ice. The pellet was resuspended in half the previous volume of MIBSM buffer (three volumes MIB buffer + one volume SM4 buffer) and homogenized with a Teflon–glass Dounce homogenizer. After clarification as described before, the resulting lysate was pooled with the previous batch of the lysate and subjected to centrifugation at 1,000g for 15 min at 4 °C to ensure complete removal of cell debris. The clarified and filtered supernatant was centrifuged at 10,000g for 15 min at 4 °C to pellet crude mitochondria. Crude mitochondria were resuspended in 10 ml MIBSM buffer and treated with 200 U of RNase-free DNase (Sigma-Aldrich) for 20 min in the cold room to remove contaminating genomic DNA. Crude mitochondria were again recovered by centrifugation at 10,000g for 15 min at 4 °C and gently resuspended in 2 ml SEM buffer (250 mM sucrose, 20 mM HEPES–KOH, pH 7.5 and 1 mM EDTA). Resuspended mitochondria were subjected to a sucrose density step gradient (1.5 ml of 60% sucrose, 4 ml of 32% sucrose, 1.5 ml of 23% sucrose and 1.5 ml of 15% sucrose in 20 mM HEPES–KOH, pH 7.5 and 1 mM EDTA) centrifugation in a Beckmann Coulter SW40 rotor at 28,000 rpm (139,000g) for 60 min. Mitochondria seen as a brown band at the interface of the 32% and 60% sucrose layers were collected and snap-frozen using liquid nitrogen and transferred to −80 °C.

Frozen mitochondria were transferred on ice and allowed to thaw slowly. Lysis buffer (25 mM HEPES–KOH, pH 7.5, 50 mM KCl, 10 mM magnesium acetate, 2% polyethylene glycol octylphenyl ether and 2 mM DTT, 1 mg ml−1 EDTA-free protease inhibitors (Sigma-Aldrich)) was added to mitochondria and the tube was inverted several times to ensure mixing. A small Teflon–glass Dounce homogenizer was used to homogenize mitochondria for efficient lysis. After incubation on ice for 5–10 min, the lysate was clarified by centrifugation at 30,000g for 20 min at 4 °C. The clarified lysate was carefully collected. Centrifugation was repeated to ensure complete clarification. A volume of 1 ml of the mitochondrial lysate was applied on top of 0.4 ml 1 M sucrose (v/v ratio of 2.5:1) in thick-walled TLS55 tubes. Centrifugation was carried out at 231,500g for 45 min in a TLA120.2 rotor at 4 °C. The pellets thus obtained were washed and sequentially resuspended in a total volume of 100 µl resuspension buffer (20 mM HEPES–KOH, pH 7.5, 50 mM KCl, 10 mM magnesium acetate, 1% Triton X-100 and 2 mM DTT). The sample was clarified twice by centrifugation at 18,000g for 10 min at 4 °C. The sample was applied onto a linear 15–30% sucrose gradient (20 mM HEPES–KOH, pH 7.5, 50 mM KCl, 10 mM magnesium acetate, 0.05% n-dodecyl-β-d-maltopyranoside and 2 mM DTT) and centrifuged in a TLS55 rotor at 213,600g for 120 min at 4 °C. The gradient was fractionated into 50-μl volume aliquots. The absorption for each aliquot at 260 nm was measured and fractions corresponding to the monosome peak were collected. The pooled fractions were subjected to buffer exchange with the resuspension buffer.

Cryo-EM data acquisition

A volume of 3 μl of ~120 nM mitoribosome was applied onto a glow-discharged (20 mA for 30 s) holey carbon grid (Quantifoil R2/2, copper, mesh 300) coated with continuous carbon (of ~3-nm thickness) and incubated for 30 s in a controlled environment of 100% humidity and 4 °C. The grids were blotted for 3 s, followed by plunge-freezing in liquid ethane, using a Vitrobot MKIV (Thermo Fisher). The data were collected on FEI Titan Krios (Thermo Fisher) transmission electron microscope operated at 300 keV, using a C2 aperture of 70 μm and a slit width of 20 eV on a GIF quantum energy filter (Gatan). A K2 Summit detector (Gatan) was used at a pixel size of 0.83 Å (magnification of ×165,000) with a dose of 29–32 e− per Å2 fractionated over 20 frames.

Cryo-EM data processing

The beam-induced motion correction and per-frame B factor weighting were performed using RELION-3.0.2 (refs. 62,63). Motion-corrected micrographs were used for contrast transfer function (CTF) estimation with gctf64. Unusable micrographs were removed by manual inspection of the micrographs and their respective calculated CTF parameters. Particles were picked in RELION-3.0.2, using reference-free followed by reference-aided particle picking procedures. Reference-free two-dimensional (2D) classification was carried out to sort useful particles from falsely picked objects, which were then subjected to three-dimensional (3D) classification. The 3D classes corresponding to unaligned particles and LSU were discarded and monosome particles were pooled and used for 3D autorefinement yielding a map with an overall resolution of 2.9–3.4 Å for the five datasets. Resolution was estimated using a Fourier shell correlation (FSC) cutoff of 0.143 between the two reconstructed half maps. Finally, the selected particles were subjected to per-particle defocus estimation, beam-tilt correction and per-particle astigmatism correction followed by Bayesian polishing. Bayesian polished particles were subjected to a second round of per-particle defocus correction. A total of 994,919 particles were pooled and separated into 86 optics groups in RELION-3.1 (ref. 65) on the basis of acquisition areas and date of data collection. Beam tilt, magnification anisotropy and higher-order (trefoil and fourth-order) aberrations were corrected in RELION-3.1 (ref. 65). Particles with bound P-site tRNA and mRNA that showed comparatively higher occupancy for the unmodeled density potentially corresponding to the LRPPRC–SLIRP module were pooled and re-extracted in a larger box size of 640 Å. The re-extracted particles were subjected to 3D autorefinement in RELION-3.1 (ref. 65). This was followed by sequential signal subtraction to remove the signal from the LSU, all of the SSU except the region around mS39 and the unmodeled density, in that order. The subtracted data were subjected to masked 3D classification (T = 200) to enrich for particles carrying the unmodeled density. Using a binary mask covering mS39 and all the unmodeled density, we performed local-masked refinement on the resulting 41,812 particles within an extracted subvolume of 240-Å box size leading to a 3.37-Å resolution map.

Model building and refinement

At the mRNA channel entrance, a more accurate and complete model of mS39 could be built with 29 residues added to the structure. Improved local resolution enabled unambiguous assignment of residues to the density, which allowed us to address errors in the previous model. A total of 28 α-helices could be modeled in their correct register and orientation. Furthermore, a 28-residue-long N-terminal loop of mS31 (residues 247–275) along mS39 and a mitochondrion-specific N-terminal extension of uS9m (residues 53–70) approaching mRNA were modeled by fitting the loops into the density maps.

For building the LRPPRC–SLIRP module, the initial model of the full-length LRPPRC was obtained from the AlphaFold2 Protein Structure Database (UniProt P42704). On the basis of the analysis, three stable domains were identified that are connected by flexible linkers (673–983 and 1,035–1,390). We then systematically assessed the domains against the map and the N-terminal region (77–660) could be fitted into the density. The initial model was real-space refined into the 3.37-Å resolution map of the mS39–LRPPRC–SLIRP region obtained after partial signal subtraction using reference restraints in Coot (v.0.9)66. The N-terminal region covering residues 64–76 was identified in the density map and allowed us to model 34 helices of LRPPRC (residues 64–644). Helices α1–α29 could be confidently modeled. An additional five helices, as predicted by AlphaFold2 (ref. 35), could be accommodated into the remaining density. After modeling LRPPRC into the map, there was an unaccounted density that fit SLIRP. The initial model of SLIRP was obtained from the AlphaFold2 Protein Structure Database (UniProt Q9GZT3). The unmodeled density agreed with the secondary structure of SLIRP. The model was real-space refined into the density using reference restraints as for LRPPRC in Coot (v.0.9)66. Five additional RNA residues could be added to the 3′ terminal of mRNA to account for the tubular density extending from it along the mRNA-binding platform. The A/A P/P E/E state model was rigid-body fitted into the corresponding 2.85-Å resolution consensus map. The modeled LRPPRC was merged with the rigid-body fitted monosome model to obtain a single model of the mitoribosome bound to LRPPRC and SLIRP. The model was then refined against the composite map using PHENIX (v.1.18)67 (Table 1).

Phylogenetic analysis

The phylogenetic distribution of proteins was determined by examining phylogeny databases60, followed by sensitive homology detection to detect homologs outside of the Bilateria. Orthologs were required to have identical domain compositions and Dollo parsimony was used to infer the evolutionary origin of a protein from its phylogenetic distribution. When multiple homologs of a protein were detected in a species, a neighbor-joining phylogeny was constructed to assess monophyly of putative orthologs to the human protein. The short length of the SLIRP candidate protein from Trichoplax adhaerens (B3SAC0_TRIAD), which is part of the large RRM family, precludes obtaining a reliable phylogeny to confidently assess its orthology to human SLIRP; therefore, the assessment is tentative.

TLSMD analysis

The TLSMD analysis36,37 was performed with the full-length LRPPRC model obtained from the AlphaFold Database (AF-P42704-F1) and the mitochondrion-targeting sequence (residues 1–59) was removed. The model was divided into TLS segments (N) and single-chain TLSMD was performed on all atoms using the isotropic analysis model. Instead of using atomic B factors, the values for a per-residue confidence score of AlphaFold called the predicted local distance difference test (pLDDT) were used as reference to calculate the least-squared residuals against the corresponding values calculated by TLSMD analysis. This is based on the assumption that local mobility of the model should be inversely correlated with the pLDDT score. AlphaFold pLDDT values and the corresponding calculated values were plotted for every iteration to monitor the improvement in prediction across the length of LRPPRC. The data in Extended Data Fig. 2 are presented for N = 4, where segments 1 and 2 (residues 60–373 and 374–649) correspond to the modeled region, whereas segments 3 and 4 correspond to the remaining domains that could not be modeled.

Helicase sequence analysis

To address the possibility that LRPPRC may serve as a helicase, we inspected the sequence of full-length LRPPRC (UniProt ID P42704). First, we checked the sequence for matches with consensus motifs characteristic of helicases using regular-expression search. The following motifs were searched, GFxxPxxIQ, AxxGxGKT, PTRELA, TPGR, DExD, SAT, FVxT and RgxD (DDX helicases); GxxGxGKT, TQPRRV, TDGML, DExH, SAT, FLTG, TNIAET and QrxGRAGR (DHX helicases); AHTSAGKT, TSPIKALSNQ and MTTEIL (others). Next, we carried out multiple sequence analysis against representative member helicases of the DHX and DDX families to verify the results of the regular-expression sequence search and to find potentially valid weaker matches.

Human cell lines and cell culture conditions

Human HEK293T embryonic kidney cells (CRL-3216, RRID: CVCL-0063) were obtained from the American Type Culture Collection. The HEK293T LRPPRC-KO cell line was engineered in-house and previously reported31. The LRPPRC-KO cell line was reconstituted with either the WT LRPPRC gene31 or a variant causing LSFC. The LSFC variant carries a single-base change (nucleotide 1119C>T transition), predicting a missense A354V change at a conserved protein residue47.

Cells were cultured in high-glucose DMEM (Thermo Fisher Scientific, cat. no. 11965092), supplemented with 10% FBS (Thermo Fisher Scientific, cat. no. A3160402), 100 μg ml−1 uridine (Sigma, cat. no. U3750), 3 mM sodium formate (Sigma cat. no. 247596) and 1 mM sodium pyruvate (Thermo Fisher Scientific, cat. no. 11360070) at 37 °C under 5% CO2. Cell lines were routinely tested for Mycoplasma contamination.

To generate an LRPPRC-KO cell line reconstituted with the LSFC variant of the gene, a Myc-DDK-tagged LRPPRC open reading frame (ORF) plasmid was obtained from OriGene (cat. no. RC216747). This ORF was then subcloned into a hygromycin resistance-containing pCMV6 entry vector (OriGene, cat. no. PS100024) and used to generate an LRPPRC-KO cell line reconstituted with a WT LRPPRC gene as reported31. To generate the LRPPRC LSFC variant carrying the 1119C>T mutation, we used the Q5 site-directed mutagenesis kit from New England Biolabs. Approximately 10 pg of template pCMV6-A-Myc-DDK-Hygro-LRPPRC vector was used, along with the primers LSFC-Q5-F 5′-GGAAGATGTAGTGTTGCAGATTTTAC and LSFC-Q5-R 5′-AATTTTTCAGTGACTAAAAGTAAAATG, designed to include the codon to be mutated. After exponential amplification and treatment with kinase and ligase, 2.5 µl of the reaction was transformed into competent E. coli cells. Several transformants were selected and their plasmid DNA was purified before sequencing to select the correct pCMV6-A-Myc-DDK-Hygro-LRPPRC-LSFC construct.

For transfection of the construct into LRPPRC-KO cells, we used 5 μl EndoFectin mixed with 2 μg vector DNA in OptiMEM-I medium according to the manufacturer’s instructions. The medium was supplemented with 200 μg ml−1 hygromycin after 48 h and drug selection was maintained for at least 1 month.

Whole-cell extracts and mitochondria isolation

For SDS–PAGE, pelleted cells were solubilized in radioimmunoprecipitation assay (RIPA) buffer (25 mM Tris-HCl, pH 7.6, 150 mM NaCl, 1% NP-40, 1% sodium deoxycholate and 0.1% SDS) with 1 mM PMSF and mammalian protease inhibitor cocktail (Sigma). Whole-cell extracts were cleared by centrifugation at 20,000g for 5 min at 4 °C.

Mitochondrion-enriched fractions were isolated from at least ten 80% confluent 15-cm plates as described previously68,69,70. Briefly, the cells were resuspended in ice-cold TKMg buffer (10 mM Tris-HCl, 10 mM KCl and 0.15 mM MgCl2; pH 7.0) and disrupted with ten strokes in a homogenizer (Kimble/Kontes). Using a 1 M sucrose solution, the homogenate was brought to a final concentration of 0.25 M sucrose. A postnuclear supernatant was obtained by centrifugation of the samples twice for 5 min at 1,000g. Mitochondria were pelleted by centrifugation for 10 min at 10,000g and resuspended in a solution of 0.25 M sucrose, 20 mM Tris-HCl, 40 mM KCl and 10 mM MgCl2 (pH 7.4).

Denaturing and native electrophoresis, followed by immunoblotting

Protein concentration was measured by the Lowry method71. First, 40–80 μg of mitochondrial protein extract was separated by denaturing SDS–PAGE in the Laemmli buffer system72. Then, proteins were transferred to nitrocellulose membranes and probed with specific primary antibodies to the following proteins: β-actin (dilution 1:2,000; Proteintech, 60008-1-Ig), ATP5A (1:1,000; Abcam, ab14748), CORE2 (1:1,000; Abcam, ab14745), COX1 (dilution 1:2,000; Abcam, ab14705), LRPPRC (dilution 1:1,000; Proteintech, 21175-1-AP), NADH:ubiquinone oxidoreductase subunit A9 (1:1,000; Proteintech, 20312-1-AP), succinate dehydrogenase complex flavoprotein subunit A (1:1,000; Proteintech, 14865-1-AP) or SLIRP (1:1,000; Abcam, ab51523). Horseradish peroxidase-conjugated anti-mouse or anti-rabbit IgGs were used as secondary antibodies (dilution 1:10,000; Rockland). β-Actin was used as a loading control. Signals were detected by chemiluminescence incubation and exposure to X-ray film.

Blue-native PAGE analysis of mitochondrial OXPHOS complexes in native conditions was performed as described previously73,74. To extract mitochondrial proteins in native conditions, we pelleted and solubilized 400 μg of mitochondria in 100 μl buffer containing 1.5 M aminocaproic acid and 50 mM Bis-Tris (pH 7.0) with 1% n-dodecyl-β-d-maltoside. Solubilized samples were incubated on ice for 10 min in ice and pelleted at 20,000g for 30 min at 4 °C. The supernatant was supplemented with 10 µl of 10× sample buffer (750 mM aminocaproic acid, 50 mM Bis-Tris, 0.5 mM EDTA and 5% Serva Blue G-250). Native PAGENovex 3–12% Bis-Tris protein gels (Thermo Fisher) were loaded with 40 μg of mitochondrial proteins. After electrophoresis, the gel was stained with 0.25% Coomassie brilliant blue R250 or proteins were transferred to PVDF membranes using an eBlot L1 protein transfer system (GenScript) and used for immunoblotting.

Pulse labeling of mitochondrial translation products

To determine mitochondrial protein synthesis, six-well plates were precoated at 5 μg cm−2 with 50 μg ml−1 collagen in 20 mM acetic acid and seeded with WT or LRPPRC cell lines (two wells per sample per timepoint). Then, 70% confluent cell cultures were incubated for 30 min in DMEM without methionine and then supplemented with 100 μl ml−1 emetine for 10 min to inhibit cytoplasmic protein synthesis as previously described68. Next, 100 μCi of [35S]methionine was added and allowed to incorporate into newly synthesized mitochondrial proteins for increasing times from 15–60-min pulses. Subsequently, whole-cell extracts were prepared by solubilization in RIPA buffer and equal amounts of total cellular protein were loaded into each lane and separated by SDS–PAGE on a 17.5% polyacrylamide gel. Gels were transferred to a nitrocellulose membrane and exposed to a Kodak X-OMAT X-ray film. The membranes were then probed with a primary antibody against β-actin as a loading control. Optical densities of the immunoreactive bands were measured using the Histogram function of the Adobe Photoshop software in digitalized images.

Whole-cell transcriptomics

Cells were grown to 80% confluency in a 10-cm plate (two plates per sample) and were collected by trypsinization and washed once with PBS before resuspending in 1 ml of Trizol (Thermo Fisher Scientific). RNA was extracted following the Trizol manufacturer’s specifications. The aqueous phase was transferred to a new tube and an equal volume of 100% isopropanol and 3 μl of glycogen were added to precipitate the RNA. The sample was incubated at −80 oC overnight and centrifuged at 15,000g for 45 min at 4 °C. RNA was resuspended in 50 μl of RNAse-free water and quantified by measuring absorbance at a wavelength of 260 nm. Then, 2 μg of RNA was sent to Novogene for further processing. Novogen services included library preparation, RNAseq on an Illumina HiSeq platform according to the Illumina Tru-Seq protocol and bioinformatics analysis. The raw data were cleaned to remove low-quality reads and adaptors using Novogen in-house Perl scripts in Cutadapt75. The reads were mapped to the reference genome using the HISAT2 software76. The transcripts were assembled and merged to obtain an mRNA expression profile with the StringTie algorithm77. The RNAseq data were then normalized to account for the total reads sequenced for each sample (the read depth) and differentially expressed mRNAs were identified by using the Ballgown suite78 and the DESeq2 R package79. GraphPad Prism v.9.0 software was used to prepare the volcano plots.

MitoRPF

MitoRPF, matched RNAseq and data analysis were performed as previously described31. Briefly, human and mouse cell lysates were prepared and mixed in a 95:5 ratio of human to mouse. For mitoRPF, the combined lysates were subjected to RNaseI treatment and fractionated across a linear sucrose gradient. Sequencing libraries were prepared from the monosome fraction after phenol–chloroform extraction. For RNAseq, RNA was extracted from the undigested combined lysate and fragmented by alkaline hydrolysis and sequencing libraries were prepared. Reads were cleaned of adaptors and filtered of rRNA fragments, and PCR duplicates were removed. Read counts were summed across features (coding sequences) using Rsubread feature Counts80 and then normalized by feature length and mouse spike-in read counts. TE was calculated by dividing spike-in normalized mitoRPF reads per kilobase by spike-in normalized RNAseq reads per kilobase. Values are expressed as the log2 fold change in the LRPPRC-KO cells compared to the LRPPRC rescue cells. MitoRPF and RNAseq data for LRPPRC-KO and LRPPRC reconstituted cell lines were deposited to the Gene Expression Omnibus (GEO) under accession number GSE173283. MitoRPF and RNAseq data for the LSFC reconstituted cell line are deposited to the GEO under accession number GSE221586.

The mitoRPF length distribution was determined from mitochondrial mRNA aligned reads. First, soft-clipped bases were removed using jvarkit81 and then the frequency for each length was output using SAMtools stats82.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.