Main

Previously, we determined the full composition as well as the overall architecture of the Arabidopsis thaliana mitoribosome6. However, owing to the difficulty of purifying large amounts of A. thaliana mitoribosomes—mainly because of the low quantities of plant material usable for mitochondrial extraction—only a low-resolution cryogenic electron microscopy (cryo-EM) reconstruction was derived. To obtain a high-resolution structure of the plant mitochondrial ribosome, we purified mitoribosome from a closely related species, Brassica oleracea var. botrytis, or cauliflower (both Arabidopsis and cauliflower belong to the group of Brassicaceae plants), as previously described6 (see Methods). We have recorded cryo-EM images for ribosomal complexes purified from two different sucrose-gradient peaks (see Methods), corresponding to the small ribosomal subunit (SSU) and the full 78S mitoribosome. After extensive particle sorting (see Methods) we have obtained cryo-EM reconstructions for both types of complexes. The SSU reconstruction displayed an average resolution of 4.36 Å (Extended Data Fig. 1). After multi-body refinement (3 bodies) and particle polishing using RELION 3.0 (ref. 12; see Methods), we derived reconstructions of the body of the SSU at a resolution of 3.77 Å, the head at a resolution of 3.9 Å and the head extension at a resolution of 10.5 Å. The combined structure revealed the full plant mitoribosome SSU (Fig. 1a–d). For the full mitoribosome, focused refinement of the LSU, SSU head and SSU body generated reconstructions at resolutions of 3.50 Å, 3.74 Å and 3.66 Å, respectively (Fig. 1e–g, Extended Data Fig. 1).

Fig. 1: The overall structure of the plant mitochondrial ribosome.
figure 1

a–g, Composite cryo-EM maps of the SSU alone (beak side (a), intersubunit side (b), platform side (c) and solvent side (d)) and the complete mitoribosome (beak side (e), LSU side (f) and platform side (g)). rRNAs are coloured in light blue (LSU) and yellow (SSU), and ribosomal proteins are indicated in blue (LSU) or orange (SSU). rPPR proteins are shown in red. h–k, The Arabidopsis mitoribosome rRNAs compared to the E. coli ribosome; the SSU is compared in h and j, and the LSU is compared in i and k. Extensions in Arabidopsis rRNAs are shown in green (h and i); deletions are shown in blue (j and k). CP, central protuberance.

Full size image

Density segmentations of our various cryo-EM reconstructions revealed the fine architecture of the plant mitoribosome, showing that a large rRNA core interacts with numerous ribosomal proteins. Among those ribosomal proteins, 8 densities located at the surface of both subunits (3 on the LSU and 5 on the SSU) unambiguously display α-helical motifs that are characteristic of pentatricopeptide repeat proteins (Fig. 1a–g). Most of these ribosomal PPRs (rPPRs) are in direct interaction with large rRNA expansion segments (ESs) such as the head extension of the SSU (Fig. 1a–d). However, some of these SSU rPPRs appear to show a higher level of flexibility in the context of the full 78S, as they appear more scant compared with the SSU-only reconstruction, along with the ESs that they interact with. As a consequence, we focused our structural analysis of these SSU rPPRs and their associated ESs on the SSU-only reconstruction.

Our cryo-EM reconstructions at near-atomic resolutions, along with our previous extensive tandem mass spectroscopy (MS/MS) analysis6, enabled us to build near-complete atomic models of the 78S mitoribosome (see Methods). In contrast to its mammalian10,13 and trypanosoma8 counterparts, the plant mitoribosome is characterized by its largely expanded rRNAs, completely reshaping the overall structure of this mitoribosome. The 26S, 18S and 5S rRNAs are 3,169, 1,935 and 118 nucleotides in length, respectively, therefore making the plant mitochondria SSU and LSU rRNAs 20% and 9% larger than their prokaryote counterparts, respectively6 (Fig. 1h,i). Nevertheless, although plant mitoribosomes contain more rRNA in general compared to bacteria, they have lost a few rRNA helices that are present in bacteria (Fig. 1j,k).

Most of the ribosomal proteins (45 in the LSU and 37 in the SSU) are either universally conserved or mitochondria specific. They comprise the common protein core of almost all of the known mitoribosomes—such as mS23, mS26 and mS29 on the SSU or mL46 and mL59/64 on the central protuberance of the LSU—therefore confirming that these proteins were acquired early during eukaryote evolution (Extended Data Fig. 9).

The structure of the SSU revealed the exact nature of its several specific features, such as its large and elongated additional head domain, the body protuberance and its elongated foot (Fig. 1). The body protuberance is mainly formed by the mitoribosome-specific r-protein mS47, which is shared with yeast7 and trypanosoma8. Interestingly, mS47 is absent only in mammals, suggesting that this protein was lost during animal evolution. The body protuberance also contains one additional protein (mS45) and extensions of uS4m (Extended Data Fig. 7), as well as additional protein densities that could not be identified. The foot of the SSU is mainly reshaped by rRNA ESs and deletions, stabilized by plant-specific r-proteins, namely ribosomal PPR (rPPR) proteins. The SSU-characteristic helix 44 is 47 nucleotides longer, therefore slightly extending the SSU and forming part of the foot extension (Figs. 2 and 3e). This extension is stabilized by a rPPR protein that is connected to ES-h6 forming a three-way junction that is also stabilized by an additional rPPR protein. We could not identify these rPPRs with certainty on the basis of their sequence owing to a lack of resolution at these regions; however, based on their number of repeats, they can only correspond to rPPR1, 3a or 3b, identified in our previous study6 and we refer to them here as rPPR*. By contrast, h8, h10 and h17 are reduced compared with their bacterial counterparts (Fig. 1j), leaving space for additional plant-specific proteins.

Fig. 2: Atomic model of the plant mitoribosome.
figure 2

a, The overall model of the plant mitoribosome; individual proteins are annotated. rRNAs are shown in blue (LSU) and light brown (SSU), and ribosomal proteins are shown in blue and green (LSU) or yellow and red (SSU) on the basis of their conservation. b, Intersubunit interfaces with conserved and novel observed intersubunit bridges are highlighted. Bacterial bridges are coloured in purple, mitoribosomes bridges in pink and plant bridges in green.

Full size image
Fig. 3: The plant mitoribosome PPR proteins.
figure 3

a–e, Overall view of the mitoribosome; PPR proteins are highlighted in red, with a magnified view on each of the rPPR proteins (rPPR4 (a), rPPR9 (b), rPPR6 (c), rPPR5 (d) and rPPR* (e)) and their rRNA target and mode of binding. In a and d, rPPR4 and rPPR5 stabilize single-stranded segments, whereas in b, c and e, rPPR proteins mainly interact with the backbone of the rRNAs.

Full size image

Our analysis of the large SSU head extension revealed that it is indeed primarily shaped by a 370-nucleotide rRNA novel domain inserted in h39. Owing to its high flexibility, it was refined with an average resolution of about 10 Å. Indeed, owing not only to its movement relative to the head, but also the head movement relative to the body, the overall movement of the head extension is of large amplitude (~30°; Extended Data Fig. 2), impairing the local resolution of this area. Nevertheless, the composition of the head extension can be determined unambiguously, as our data identified secondary-structure elements for both rRNA and rPPRs (Fig. 3c). The head extension is mainly composed of rRNA, rooting from h39. From there, the extension forms a four-way junction that is stabilized by a long PPR protein (rPPR6 or mS80), locking the whole additional domain into a position that is perpendicular to the intersubunit side. Past the four-way junction, two of the rRNA helices organize into two parallel segments, forming the core of the extension—one of the two helices ends in a three-way junction that shapes the tip of the head-extension. The two parallel RNA helices are themselves contacted by a rPPR (Fig. 3c). However, the local resolution is too low to clearly determine its exact identity (rPPR*). Interestingly, the protein bTHXm, which was previously identified using MS6, was found buried deep inside the SSU head. This protein is found only in the plant mitoribosome as well as in chlororibosomes4,5 and ribosomes from the Thermus genus14.

On the back of the SSU, a large cleft extends the exit of the mRNA channel. This cleft is delimited by mS26, uS8m and h26 on one side and by mS47 and the rPPR mS83 on the other side (Fig. 4a). Similar to all known PPRs, mS83 is predicted to be an RNA binder. In plants, the processes that underlie the recruitment and correct positioning of mRNAs during translation initiation are unknown. Similar to other known mitochondrial translation systems, the Shine–Dalgarno (SD) and the anti-SD sequences are absent from both plant mRNAs and SSU rRNAs. Moreover, mRNAs have long 5′ untranslated regions (UTRs), similar to yeast mitochondrial mRNAs7. Interestingly, half of the plant mitochondrial mRNA 5′ UTRs harbour an A/purine-rich sequence, AxAAA, located about 19 nucleotides upstream of the AUG (Extended Data Fig. 8). This distance correlates with the size of the extended mRNA exit channel and would put the purine-rich sequence in close vicinity to the rPPR protein mS83. We therefore hypothesized that this plant-specific cleft may act as a recruitment platform for incoming mRNAs and/or additional factors. mS83 might recognize the AxAAA motif, therefore recapitulating a SD–antiSD-like recognition system, using an RNA–protein interaction instead of an RNA–RNA interaction (Fig. 4a). An example of such a possible rPPR-mediated initiation system may be found in mammalian mitochondria in which the rPPR mS39 located on the SSU was proposed to accommodate the 5′ UTR-less mRNAs from their 3′ end through a U-rich motif15. It is important to note that this A-rich cis-element is not found in all plant mitochondrial mRNAs, suggesting that this proposed mechanism for translation initiation would therefore not be universal in plant mitochondria. Similarly, in chloroplasts, a third of the mRNAs do not possess a SD-sequence, suggesting that at least two different mechanisms co-exist for translation initiation16.

Fig. 4: Specific features of the plant mitoribosome.
figure 4

a, View of the SSU back channel and its protein constituents. The possible mRNA path is indicated by the dashed blue line; the blue star indicates the position of the AxAAA motif present in the 5′ UTR of half of plant mitochondrial mRNAs. b, The plant mitoribosome peptide channel and a possible mode of membrane attachment of the plant mitoribosome. Similar to yeast with Mba1 (ref. 30), the plant mitoribosome would require an accessory protein factor (green linker) to tether the mitoribosome to the Oxa1 insertase. The green line represents the possible surface of interaction with the inner membrane (IM) of the mitochondria, revealing an extensive surface mainly formed by the remodelled domain III of the LSU. OM, outer membrane. c, Comparison of the different mitoribosomes described to date. A schematic of the ancestral alphaproteobacterial ribosome is compared with the actual mitoribosomes. Core bacterial rRNAs are indicated in light blue and core bacterial r-proteins are indicated in yellow. Acquisition of mitochondria-specific rRNAs and r-proteins is represented by blue and orange colouring, respectively. Two main different evolutive paths have been taken by mitoribosomes, a mainly constructive one (i) represented by the yeast and plant mitoribosomes, in which few proteins were lost and novel r-proteins were acquired along with extended rRNAs, whereas in human and Trypanosoma (ii), rRNAs were highly reduced and more novel r-proteins were acquired.

Full size image

Similar to the SSU, the overall shape of the LSU is also strongly remodelled, even though a large portion of the core components is conserved (Fig. 2). Indeed, core functional components, such as the peptidyl-transferase centre, the L7/L12 stalk and the central protuberance (CP) are similar to those found in bacteria. This conservation is in line with the globally bacterial-like intersubunit bridges (Fig. 2b, Extended Data Fig. 10). However, the LSU is reshaped by the conserved mitochondria-specific proteins, for example mL41 or mL59/mL64 that connect the body of the LSU to the CP. In contrast to all of the other mitoribosomes described to date, the plant mitoribosome CP includes a 5S rRNA. The core structure of the plant CP is quite conserved compared to the CP of prokaryotes. However, mitochondria-specific ribosomal proteins, namely mL40, mL59/64 and mL46, complement the classical bacterial-like CP and bind to the top of the r-proteins bL27m, uL5m and uL18m. Interestingly, the same mitochondria-specific ribosomal proteins that form part of the plant CP are also present in other mitoribosomes even though they have lost their 5S rRNA, indicating that the acquisition of these mitoribosome-specific proteins occurred before the loss of the 5S in other mitoribosomes (Extended Data Fig. 5).

Interestingly, the universally conserved uL2 is split into two parts in the plant mitoribosome, its N-terminal part is encoded by a mitochondria-encoded gene whereas its C-terminal part is encoded by a nucleus-encoded gene, which could constitute a mechanism of mitochondria–nuclear crosstalk (Extended Data Fig. 7).

In yeast7, mammals10,13 and trypanosoma8 mitoribosomes, the peptide exit channel is highly remodelled by species-specific proteins (such as mL45 or mL71). However, in plants, the major part of the peptide channel and its exit are rather similar to that of bacteria, with only minimal rearrangement of the surrounding rRNA helices and a small extension of uL29m (Fig. 4b). In contrast to humans and yeast17, a substantial portion of the mitochondria-encoded proteins are soluble proteins in plants (for example, 7 soluble r-proteins are mitochondria encoded in Arabidopsis); it is therefore conceivable that the plant mitoribosome does not systematically require an association with the inner mitochondrial membrane, as it is the case in humans and yeast18,19. However, it is probable that, at least in some cases, a non-ribosomal protein could link the mitoribosome to the mitochondrial insertase Oxa1. This hypothesis is supported by the observation that Oxa1 co-purifies with immunoprecipitated plant mitoribosomes6(Fig. 4b).

The main plant-specific features of the LSU are the plant-specific proteins and rRNA ESs that completely reshape the back of the LSU below the L1-stalk region. Indeed, running along the back of the LSU, from uL15m and H31 and contacting the largely extended and remodelled domain I rRNAs (ES-H21–22 and ES-H16–18), the 19-repeat rPPR protein mL102 (rPPR5) stabilizes these additional rRNAs extensions (Fig. 3d). Moreover, the domain III is extensively remodelled and holds several expansion segments; helices of this domain were therefore renamed pH53–59 (Fig. 1h, Extended Data Fig. 4). Indeed, this remodelled domain III has two main helices that largely extend into the solvent and are stabilized by two PPR proteins (rPPR4 and 9 or mL101 and mL104; Fig. 3a,b). These two rPPRs appear to be rigid and present numerous interactions with the rRNA. Thus, rPPR9 encapsulates the tip of helix H10 and stabilizes the end of pH59, on the newly formed helices of domain III. rPPR9 also directly contacts rPPR4, which wraps around a single-stranded rRNA extension (nucleotides 1635–1644) and contacts the two major helices of domain III pH55 and pH57 (Fig. 3b). Interestingly, the rPPRs described here, along with those of the SSU, seem to have a different mode of RNA binding compared with the RNA-recognition process of canonical PPR proteins. Indeed, PPR proteins usually bind to single-stranded RNA through a combinatorial recognition mechanism that mainly involves two specific residues in each repeat (5 and 35) enabling each repeat to bind to a specific nucleotide20, similar to other helical repeat modular proteins21. However, the structure obtained here revealed that several rPPR proteins bind to the convex surface of double-stranded rRNA, mainly through positively charged residues contacting the phosphate backbone of the rRNAs. This novel mode of RNA binding evidenced for rPPR proteins (Fig. 3) extends our understanding of the diversity of functions and modes of action held by PPR proteins. The position of this remodelled domain III on the full 78S suggests that rPPRs 4 and 9 might be involved in the attachment to the inner mitochondrial membrane, as they appear to strongly stabilize the structure of the whole domain (Fig. 4b).

In conclusion, the plant mitoribosome with its large rRNA ESs shows yet another route taken during mitoribosome evolution. It represents an augmented prokaryote-type ribosome that is based on a bacterial scaffold, and it has both expanded rRNAs and an expanded set of proteins that were specifically recruited during eukaryote evolution in the plant clade. The structure of the plant mitoribosome could reflect the so-called ‘constructive phase’ of mitoribosome evolution in which both rRNAs and protein sets were augmented, in strong contrast with mammals and Trypanosoma in which rRNAs were considerably reduced3 (Fig. 4c). The structure presented here therefore provides further insights into the evolution of mitoribosomes and the elaboration of independent new strategies to perform and regulate translation.

Methods

Mitochondrial ribosome purification

Cauliflower (Brassica oleracea var. botrytis) was used here as it is best suited for large-scale biochemical, structural analyses compared to Arabidopsis. Cauliflower belongs to the same family of plants as Arabidopsis—that is, Brassicaceae—therefore making it an optimal model for this study, enabling the preparation of large quantities of highly pure mitochondria. However, the genome of cauliflower is not sequenced, and the closest fully sequenced member of the family (Brassica oleracea subsp. oleracea) is poorly annotated. However, protein sequence identities between members of the Brassicaceae family are higher than 90% and, therefore, facilitate proteomics identification of cauliflower proteins. Combining the proteomics results from Waltz et al.6 and those obtained in this study, it is evident that no difference in terms of protein composition could be observed between Cauliflower and Arabidopsis. Thus, to facilitate comprehension and analysis, we aligned Arabidopsis proteins to the cauliflower map.

For the mitochondria purification, fresh cauliflower inflorescence tissue was blended in extraction buffer containing 0.3M mannitol, 30 mM sodium pyrophosphate (10·H2O), 0.5% BSA, 0.8% (w/v) polyvinylpyrrolidone-25, 2 mM beta-mercaptoethanol, 1 mM EDTA, 20 mM ascorbate and 5 mM cysteine, pH 7.5. Lysate was filtered and clarified by centrifugation at 1,500g for 10 min at 4 °C. The supernatant was retained and centrifuged at 18,000g for 15 min at 4 °C. The organelle pellet was resuspended in wash buffer (0.3 M mannitol, 10 mM phosphate buffer and 1 mM EDTA, pH 7.5) and the precedent centrifugations were repeated once. The resulting organelle pellet was resuspended in wash buffer and loaded onto a single-step 30% Percoll gradient (in wash buffer without EDTA) and run for 1 h 30 min at 40,000g. Mitochondria were retrieved, washed twice before being flash-frozen in liquid nitrogen.

For mitoribosome purification, mitochondria were resuspended in lysis buffer (20 mM HEPES-KOH pH 7.6, 100 mM KCl, 30 mM MgCl2, 1 mM dithiothreitol, 1.6% Triton X-100, 100 µg ml−1 chloramphenicol, supplemented with proteases inhibitors (cOmplete EDTA-free)) to a concentration of 1 mg ml−1 and incubated for 15 min at 4 °C. Lysate was clarified by centrifugation at 30,000g for 20 min at 4 °C. The supernatant was loaded onto a 40% sucrose cushion in monosome buffer (same as the lysis buffer, but without Triton X-100 and 50 µg ml−1 chloramphenicol) and centrifuged at 235,000g for 3 h at 4 °C. The crude ribosome pellet was resuspended in monosome buffer and loaded onto a 10–30% sucrose gradient in the same buffer and run for 16 h at 65,000g. Fractions corresponding to mitoribosomes were collected, pelleted and resuspended in monosome buffer.

Grid preparation

We applied 4 µl of the samples at a concentration of 2 µg µl−1 proteins onto a Quantifoil R2/2 300-mesh holey carbon grid, which had been coated with thin home-made continuous carbon film and glow-discharged. The sample was incubated on the grid for 30 s and then blotted with filter paper for 2.5 s in a temperature- and humidity-controlled Vitrobot Mark IV system (T = 4 °C, humidity 100%, blot force 5) followed by vitrification in liquid ethane that was precooled with liquid nitrogen.

Collection of single-particle cryo-EM data

For the two datasets (full and dissociated complexes), data collection was performed using a Talos Arctica instrument (Thermo Fisher Scientific) at 200 kV using EPU (Thermo Fisher Scientific) for automated data acquisition. Data were collected at a nominal underfocus of −0.5 to −2.7 µm at a magnification of ×120,000, yielding a pixel size of 1.21 Å. Micrographs were recorded as a video stack using a Falcon III direct electron detector (Thermo Fisher Scientific), each video stack was fractionated into 20 frames for a total exposure of 1 s corresponding to an electron dose of 60 e− Å−2.

EM image processing

Drift and gain correction and dose weighting were performed using MotionCor2 (ref. 22). A dose-weighted average image of the whole stack was used to determine the contrast transfer function using Gctf23. The following process was achieved using RELION 3.0 (ref. 24). Particles were picked using a Laplacian of Gaussian function (minimum diameter, 260 Å; maximum diameter, 460 Å). For the full mitoribosome, after 2D classification, 153,608 particles were extracted with a box size of 400 px and binned fourfold for three-dimensional (3D) classification into ten classes. Four classes depicting high-resolution features were selected for refinement. The complex was focus-refined using a mask on the LSU, the body and the head of the SSU, yielding resolutions of 3.50 Å, 3.66 Å and 3.74 Å, respectively. Determination of the local resolution of the final density map was performed using ResMap25.

For the dissociated subunits, following a 2D classification, 132,130 particles for the LSU and 120,350 particles for the SSU were extracted with a box size of 400 px and binned fourfold for 3D classification into eight classes for each subunit. Five subclasses depicting high-resolution features were selected for the SSU refinement with 73,670 particles. After Bayesian polishing, a multi-body refinement was performed using a mask on the body, the head and the RNA expansion on the head, yielding resolutions of 3.77 Å, 3.9 Å and 10.5 Å, respectively. Three classes were selected for LSU refinement yielding a resolution of 3.96 Å. Determination of the local resolution of the final density map was performed using ResMap25.

Structure building and model refinement

The atomic model of the plant mitoribosome was built into the high-resolution maps using Coot, Phoenix and Chimera. Initial protein composition was derived from Waltz et al.6. Atomic models from E. coli ribosome (Protein Data Bank (PDB), 5KCR)26, yeast mitoribosome (PDB, 5MRC)7, human mitoribosome (PDB, 6GAW)15 and Trypanosoma mitoribosome (PDB, 6HIV)8 were used as starting points for protein identification and modelling. As a result, almost all of the proteins initially identified by Waltz et al.6 could be placed in the final model. Additional proteins, such as mS38 or bL33, that were not originally identified in the initial study were directly identified from the density. The online SWISS-MODEL service was used to generate initial models for bacterial and mitochondria conserved r-proteins. Models were then rigid-body fitted to the density in Chimera27 and all of the subsequent modelling was performed using Coot28.

For the LSU and SSU rRNA, 16S and 23S from E. coli were docked into the maps and used as templates from positioning and reconstruction. A multiple-sequence alignment of several plant mitochondrial ribosomes and E. coli ribosome was performed to determine the additional or depleted domains of A. thaliana mitochondrial ribosome. Punctual differences were performed in Chimera using the swapna command line.

To build the additional rRNA domains of A. thaliana, the co-variation algorithm LocARNA webservice (http://rna.informatik.uni-freiburg.de) was used to determine the secondary structure of these domains. At that point, the secondary-structure prediction was used to build the 3D model in Chimera using the ‘build structure’ tools followed by manual adjustments.

Proteins with clear homologues in either mammalian or yeast mitoribosomes that were as long as E. coli ribosome were built using Phyre2 and SWISS-Model.

The global atomic model was refined with VMD using the molecular dynamic flexible fitting then with PHENIX using a combination of real and reciprocal space refinement for proteins and ERRASER for RNA.

Proteomic and statistical analyses of mitochondrial ribosome composition

MS analyses of the ribosome fractions were performed at the Strasbourg Esplanade proteomic platform as previously described6. In brief, proteins were trypsin digested, MS analyses and quantitative proteomics were performed using nano liquid chromatography with electrospray ionization MS/MS analysis on AB Sciex TripleTOF mass spectrometers, and quantitative label-free analysis was performed through in-house bioinformatics pipelines.

Figure preparation

Figures featuring cryo-EM densities as well as atomic models were visualized using UCSF ChimeraX29.

Reporting Summary

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