Cryo-EM structure and rRNA model of a translating eukaryotic 80S ribosome at 5.5-Å resolution

Abstract

Protein biosynthesis, the translation of the genetic code into polypeptides, occurs on ribonucleoprotein particles called ribosomes. Although X-ray structures of bacterial ribosomes are available, high-resolution structures of eukaryotic 80S ribosomes are lacking. Using cryoelectron microscopy and single-particle reconstruction, we have determined the structure of a translating plant (Triticum aestivum) 80S ribosome at 5.5-Å resolution. This map, together with a 6.1-Å map of a Saccharomyces cerevisiae 80S ribosome, has enabled us to model ∼98% of the rRNA. Accurate assignment of the rRNA expansion segments (ES) and variable regions has revealed unique ES–ES and r-protein–ES interactions, providing insight into the structure and evolution of the eukaryotic ribosome.

In all living cells, the translation of mRNA into polypeptide occurs on ribosomes. Ribosomes provide a platform upon which aminoacyl-tRNAs interact with the mRNA as well as position the aminoacyl-tRNAs for peptide-bond formation (1). Ribosomes are composed of two subunits, a small subunit that monitors the mRNA–tRNA codon-anticodon duplex to ensure fidelity of decoding (2, 3) and a large subunit that contains the active site where peptide-bond formation occurs (4). Both the small and large subunits are composed of RNA and protein: In eubacteria such as Escherichia coli, the small subunit contains one 16S rRNA and 21 ribosomal proteins (r proteins), whereas the large subunit contains 5S and 23S rRNAs and 33 r proteins. Crystal structures of the complete bacterial 70S ribosome were initially reported at 5.5 Å (5), with an interpretation based on atomic models of the individual subunit structures (6–8), and are now available at atomic resolution (9). These structures have provided unparalleled insight into the mechanism of different steps of translation (1) as well as inhibition by antibiotics (10).

Compared to the bacterial ribosome, the eukaryotic counterpart is more complicated, containing expansion segments (ES) and variable regions in the rRNA as well as many additional r proteins and r-protein extensions. Plant and fungal 80S ribosomes contain ∼5,500 nucleotides (nts) of rRNA and ∼80 r proteins, whereas bacterial 70S ribosomes comprise ∼4,500 nts and 54 r proteins. The additional elements present in eukaryotic ribosomes may reflect the increased complexity of translation regulation in eukaryotic cells, as evident for assembly, translation initiation, and development, as well as the phenomenon of localized translation (11–15).

Early models for eukaryotic ribosomes were derived from electron micrographs of negative-stain or freeze-dried ribosomal particles (16) and localization of r proteins was attempted using immuno-EM and cross-linking approaches; see, for example, refs. 17–20. The first cryo-EM reconstruction of a eukaryotic 80S ribosome was reported for wheat germ (Triticum aestivum) at 38 Å (21). Initial core models for the yeast 80S ribosome were built at 15-Å resolution (22) by docking the rRNA structures of the bacterial small 30S subunit (6) and archaeal large 50S subunit (8), as well as docking of corresponding homology models of the r proteins. Recently, reconstructions at about 9-Å resolution of fungal and dog 80S ribosomes were used to extend the molecular models to include rRNA expansion segments (23, 24). However, due to the modest resolution, the completeness and accuracy of these models are also limited.

Here we have determined a cryo-EM structure of a wheat germ (T. aestivum) translating 80S ribosome at 5.5-Å resolution, enabling us to systematically model ∼98% of the rRNA. This effort encompasses the de novo modeling of 1,885 nucleotides comprising structurally variable regions and rRNA expansion segments. The model reveals direct interaction between ES3 and ES6 as predicted previously by Alkemar and Nygård (25), as well as r-protein–ES interactions, such as L6e and L28e with ES7^L and L34e and L38e with ES27^L. The accurate modeling of the rRNA has enabled the localization of 74 (92.5%) of the 80 r proteins of the 80S ribosome (see ref. 26).

Results and Discussion

Cryo-EM Reconstructions of T. aestivum and Yeast 80S Ribosomes.

Cryo-EM and single-particle analysis were used to reconstruct the T. aestivum translating 80S ribosome (Fig. 1A) at 5.5-Å resolution (Fig. S1). Similarly, we have previously reported a cryo-EM structure of a translating Saccharomyces cerevisiae 80S ribosome at 6.1-Å resolution (Fig. 1B) (27). For both reconstructions, ribosome-nascent chain complexes in the posttranslocational state were utilized (27, 28), after in silico sorting (see Experimental Procedures) to increase conformational homogeneity. The final reconstruction of the T. aestivum 80S ribosome was derived from 1,362,920 particles sorted for the presence of peptidyl-tRNA in the P site (Fig. 1A). The resulting cryo-EM map displays characteristics similar to X-ray crystallographic maps of ribosomes at similar resolution, namely, the Haloarcula marismortui 50S subunit at 5 Å (29) and the Thermus thermophilus 30S (30) and 70S structures (5) at 5.5 Å (Fig. 1 C–F). At this resolution, well-resolved density for double-helical RNA is observed, with defined minor and major grooves as well as distinctive bumps indicative of phosphate groups located along the backbone ridges (Fig. 1 C and D). In many regions, single-stranded rRNA sections are traceable and assignment of bulged nucleotides is possible, as reported previously for the 5.8-Å cryo-EM map of TnaC-stalled bacterial 70S ribosome (31). For r proteins, α-helices are observed as rod-like densities (Fig. 1E) and β-sheets are represented by smooth surfaces (Fig. 1F). The α-helix pitch and strand separation for β-sheets are indiscernible, as expected at this resolution.

Fig. 1.

Cryo-EM reconstruction of eukaryotic 80S ribosomes. (A) T. aestivum and (B) S. cerevisiae 80S ribosomes, with small (40S) and large (60S) subunits colored yellow and gray, respectively and the P-tRNA, green. (C–F) Selected views of the T. aestivum 80S density map (blue mesh) and corresponding molecular model, with r protein in yellow and rRNA in white (backbone) and red (bases).

Near-Complete Models for the T. aestivum and Yeast 80S Ribosomes.

The majority of the conserved core of the T. aestivum and yeast 80S ribosomes was modeled based on homology of the eukaryotic rRNA with the available bacterial and archaeal ribosome structures using Assemble (32). On this basis, it was possible to generate a template-based model for the T. aestivum 80S with a total of 3,466 (1,051/40S and 2,415/60S) nts of the 5485 rRNA, incorporating isosteric base substitutions (33) (Fig. 2). Nucleotides that were not available in the sequences for T. aestivum (120 nts, 2.2%) were substituted with those from the closely related Oryza sativa. One-hundred sixteen (67 and 49) nts, mostly single-stranded linker regions, could not be modeled (orange in Fig. 2 A and B; enlargement in Figs. S2–S7) due to unreliable secondary structure predictions and/or ambiguity in the electron density. The remaining 1,903 nts comprising structurally variable regions and rRNA expansion segments were modeled de novo (green in Fig. 2 A–D) using Assemble (32), RNAfold (34), and RNAshapes (35). Similarly, models for 44 of the 80 r proteins of the T. aestivum 80S ribosome (gray in Fig. 2 E and F) were built using the templates present in the bacterial and archaeal ribosome structures (29, 30), as well as 44 of 79 r proteins of the yeast 80S ribosome (see ref. 26). Models were fitted into the density using Molecular Dynamics Flexible Fitting (MDFF) (36). The T. aestivum and yeast 80S models contain all five expansion segments (ES3^S, ES6^S, ES7^S, ES9^S, and ES12^S following the ES nomenclature of ref. 37) and five variable regions (h6, h16, h17, h33, and h41) of the small subunit, as well as the 16 expansion segments (ES3^L, ES4^L, ES5^L, ES7^L, ES9^L, ES10^L, ES12^L, ES15^L, ES19^L, ES20^L, ES24^L, ES26^L, ES27^L, ES31^L, ES39^L, and ES41^L) and two variable regions (H16–18 and H38) of the large subunit (Fig. 3). On the small subunit, the majority of the additional rRNA is clustered at the spur or foot region, except for ES9^S which is positioned at the head (Fig. 3 C and E). On the large subunit, most ES are located on the back and sides of the particle, leaving the subunit interface and exit tunnel regions essentially unaffected (Fig. 3 D and F).

Fig. 2.

An atomic model for the T. aestivum 80S ribosome. (A and B) Secondary structures for the (A) small (18S) and (B) large subunit (5S, 5.8S, and 28S) ribosomal RNAs, with the newly modeled regions colored green. Expansion segments and variable regions are indicated in gray and unmodeled regions are orange. (C and D) Newly modeled regions of rRNA (green) are highlighted on the (C) small and (D) large subunit density map (Left) and as molecular models (Right). (E and F) Newly modeled proteins are highlighted on the (E) small and (D) large subunit density map (Left) and as molecular models (Right). Newly identified proteins are colored red, whereas de novo modeled extensions are colored light green, and modeled but unassigned proteins are yellow.

Fig. 3.

Ribosomal RNA expansion segments and variable regions. (A and B) Secondary structures for the T. aestivum (A) small (18S) and (B) large subunit (5S, 5.8S, and 28S) ribosomal RNAs, with the ES and variable regions (VR) colored distinctly. (C and D) Cryo-EM maps of the (C) small and (D) large subunits with assigned ES and VR colored as in A and B. (E and F) Molecular models of the ES and VR of rRNA colored as in C and D.

Comparison of Expansion Segments Between Yeast and Wheat Germ.

Interestingly, the density maps of the T. aestivum and yeast 80S reconstructions support a direct interaction of the loop of ES6^Sd with the internal loop between ES3^Sb and ES3^Sc (Fig. 4A), as predicted previously by Alkemar and Nygård (25). Although this interaction was not modeled in the fungal or canine 80S ribosomes (23, 24), covariation analysis suggests that the ES3^S–ES6^S base-pairing interaction is also conserved in mammalian 80S ribosomes (25).

Fig. 4.

Molecular models for expansion segments ES3^S/ES6^S, ES7^L, and ES27^L. (A) Isolated density for ES6^Sd (blue) and ES3^Sa,c (gold) on the 40S subunit (Left) and transparent with a molecular model (Center). rRNA secondary structure prediction highlighting interaction between the loop of ES6^Sd and the bulge in ES3^Sc (Right), as proposed by ref. 59. (B) Isolated density for ES7^L from T. aestivum (T. a., blue) and S. cerevisiae (S. c., gold) on the 80S ribosome (Left) and transparent with a molecular model (Center). Ribosomal proteins L28e (red) stabilizes ES7^La in the T. aestivum 80S ribosome, whereas the extension of r-protein L6e appears to pass through the three-way junction formed by helices ES7^Lc–e (Right). Molecular models for the (gold) and (blue) positions (Left), as observed in S. cerevisiae 80S ribosomes (Thumbnail Insets) (39) and an intermediate position (, gray) observed in the T. aestivum 80S ribosome. In yeast, r-protein L34e (green) and L38e (red) interact with the and positions, respectively. The tunnel exit (TE) and L1 stalk (L1) are indicated for reference. (C) Schematic (Top Right) and molecular model (Middle Right) indicating that the interchange between the (gold) and (blue) positions involves a rotation of ∼110° of ES27^La–c relative to H63. Secondary structure for the junctions of S. cerevisiae ES27^La–c and H63.

The largest ES in the T. aestivum and yeast ribosomes is ES7^L, which is located at the back of the 60S subunit (Fig. 4B). Overall, ES7^L is similar between T. aestivum and yeast, however, at least two clear differences are evident: Firstly, density for ES7^La in yeast is only seen at low thresholds (Fig. S8), suggesting it is more flexible than in T. aestivum (Fig. 4B). The reason for this flexibility appears to be that ES7^La in T. aestivum is stabilized through an interaction with r-protein L28e (Fig. 4B), which is not present in the S. cerevisiae genome (38). Secondly, T. aestivum contains a three-way junction formed by ES7^Lc–e, whereas this architecture is not present in yeast due to the absence of ES7^Ld,e (Fig. 4B and Figs. S2–S7). Surprisingly, the N terminus of T. aestivum r-protein L6e, which is shorter in yeast, appears to insert through the three-way junction of ES7^L (Fig. 4B), an RNA-protein interaction that has to our knowledge not been reported previously.

ES27^L is unique for its highly dynamic behavior, being found in two distinct positions in yeast 80S ribosomes (39); one oriented toward the L1 stalk, termed and one away from the L1 stalk but toward the tunnel exit, termed (Fig. 4C). Modeling of both conformations reveals that interchange between the (gold) and (blue) positions involves a rotation of ∼110° of ES27^La–c relative to H63 (Fig. 4C). Weak density for ES27^L in the reconstruction of the T. aestivum ribosome suggests that ES27^L exhibits a continuum of different conformational states. Nevertheless, at low thresholds, one preferential state is observed, intermediate in position () to the yeast and positions (Fig. 4C and Fig. S8). All three positions appear to be stabilized through interaction with newly identified eukaryotic-specific r proteins: The yeast and the T. aestivum conformations directly contact r-protein L38e (Fig. 4C), whereas r-protein L34e stabilizes the yeast position. In Tetrahymena, deletion of ES27^L is lethal (40), suggesting a functionally important role for this RNA insertion. Despite the high variability in length of ES27^L, ranging from ∼150 nucleotides in T. aestivum and yeast to ∼700 nucleotides in mammals (41), the ES27^L deletion can be complemented with a corresponding ES27^L from other species (40). ES27^L has been suggested to play a role in coordinating the access of nonribosomal factors at the tunnel exit (39).

Evaluation of RNA Models for the Eukaryotic Ribosome.

A reconstruction at 8.7 Å of a canine ribosome was used for a model including models of ∼50% of the ES by fitting of A-form helices into the ES density (23). Recently, a more comprehensive model of the yeast S. cerevisiae ribosome was built on the basis of an 8.9-Å cryo-EM reconstruction of a 80S ribosome from a related thermophilic fungus, Thermomyces lanuginosus (24), which, however, shares only ∼85% sequence identity with S. cerevisiae rRNA. With the exception of ES10^L, ES27^L, and the tip of ES15^L, molecular models were built for all the remaining expansion segments and variable regions (24). Yet, a number of significant differences between the yeast model presented by Taylor et al. (24) and the yeast model presented here are evident (Fig. S9). Taken together, using the correct sequences for modeling into corresponding maps of improved resolution allowed for a significant improvement in completeness and accuracy of both the RNA and protein models.

Evolution of RNA Expansion Segments.

A comparison of genomic sequences from diverse organisms, ranging from bacteria to mammals, indicates additional mass with increasing organism complexity (Fig. 5). However, the composition of mammalian ribosomes, e.g., from human, is surprisingly similar to those of other eukaryotes, such as yeast and plants described here. Human ribosomes have the same 80 r proteins that are found in T. aestivum ribosomes and, in terms of rRNA, differ significantly only in the length of four ES on the large subunit (ES7^L, ES15^L, ES27^L, and ES39^L). These are longer in human (∼850, ∼180, ∼700, and ∼220 nts) than in T. aestivum/yeast (∼200, ∼20, ∼150, and ∼120 nts, respectively), and cryo-EM reconstructions of mammalian ribosomes (23, 42–44) show that the longer ES in mammalian ribosomes are generally highly mobile elements for which little to no density is visible (Fig. 5). Evolution has thus favored the development of two apparently distinct layers of mass gain for the ribosome: A first layer of tightly intertwined additional proteins and rRNA expansions rigidly positioned on the subunit surfaces (with the only exception of the mobile ES27^L), which was followed by a second layer comprising a few drastically extended highly mobile rRNA elements with hitherto unknown function.

Fig. 5.

Cryo-EM reconstructions of ribosomes from (A) the eubacterium Escherichia coli (31), (B) the yeast S. cerevisiae (27), (C) wheat germ T. aestivum (this work), and (D) Homo sapiens (44). The small and large subunits are shown in yellow and gray, respectively and the P-tRNA (green) is indicated for reference. The dashed lines and numbers indicate the number of nucleotides of the rRNA expansion segments that are not visualized.

Experimental Procedures

Sample Preparation and Cryo-EM.

The cryo-EM map used for modeling of the yeast 80S ribosome was previously deposited in the Electron Microscopy Data Bank (EMDB ID 1669; ref. 27). T. aestivum ribosome nascent chain complexes (RNCs) were generated using a homemade wheat germ in vitro translation system (based on ref. 45) and were purified according to ref. 46. As described previously (47), T. aestivum RNC samples were applied to carbon-coated holey grids, and micrographs were recorded under low-dose conditions on a Polara field emission gun electron microscope at 300 kV (FEI) in a defocus range of 1.0–4.5 μm. Micrographs were scanned on a Heidelberg Primescan D8200 drum scanner, resulting in a pixel size of 1.24 Å on the object scale. The data were analyzed by determination of the contrast transfer function using CTFFIND (48). The data were further processed with the SPIDER software package (49). After automated particle picking followed by visual inspection, 2,108,230 particles of the T. aestivum RNC dataset were sorted in a supervised manner (50) into programmed (with P-tRNA) and unprogrammed/empty (without P-tRNA) ribosome subdatasets, using reconstructions of programmed and unprogrammed ribosomes as initial references. Removal of unprogrammed ribosome particles resulted in 1,362,920 particles that were used for reconstruction of the wheat germ 80S ribosome. The final contrast transfer function corrected reconstruction has a resolution of 5.5 Å, based on the Fourier Shell Correlation with a cutoff value of 0.5 (Fig. S1). Densities for the 40S subunit, the 60S subunit, and the P-site tRNA were isolated using binary masks.

Ribosomal RNA Sequences.

The rRNA sequences of the S. cerevisiae 5S, 5.8S, 18S, and 25S were taken from GenBank Accession number (Acc.) U53879. The rRNA sequence for the T. aestivum 5S (Acc. X06094), 5.8S (Acc. FM998894), 18S (Acc. AY049040), and 28S (Acc. AY049041) rRNAs were available, with the exception of five and four nucleotides at the 5′ end of the 18S and 28S, respectively, 46 nts from the 3′ end of the 28S, and 65 nts (487–551) in the 28S corresponding to ES7, which were filled with the corresponding sequences of O. sativa (Acc. M11585). Sequence alignments between the available T. aestivum and O. sativa rRNAs show a 98% sequence identity, indicating the suitability of using the O. sativa sequence for filling the missing 120 (2.2%) nucleotides in the T. aestivum model.

Modeling of the Ribosomal RNA Core.

The structure-based sequence alignment of both the 18S of the small subunit and the 5S, 5.8S, and 28S rRNA of the large subunit was done using the X-ray structure of the large ribosomal of H. marismortui [Protein Data Bank (PDB) 1FFK] (8) and the small ribosomal subunit of T. thermophilus (PDB 1J5E) (6). For regions like H5–H7, the stalk base (H42–H43), and the L1 stalk (H76–H78), the X-ray structure of E. coli (PDB 3FIK) (51, 52) was used as template. The alignment was constructed semiautomatically using S2S (53). The multiple sequence alignment for the 5S, 5.8S, and 28S was constructed between H. marismortui, T. aestivum/O. sativa, and S. cerevisiae and for the 18S between T. thermophilus, T. aestivum/O. sativa, and S. cerevisiae, respectively. The resulting core models for S. cerevisiae and T. aestivum were deduced from the alignments using Assemble (32) and core models consist only of isosteric base substitutions (33, 54).

Modeling of the Ribosomal RNA Expansion Segments.

Primary sequences were used as an input for RNA secondary structure prediction tools RNAfold (34) and RNAshapes (35). The core model was used as an anchor point for modeling the ES. According to the secondary structure predictions and the electron density, the ES were constructed semiautomatically using Assemble (32). The applied structural motifs for loops and inner helical non-Watson-Crick base-pairing motifs were extracted from known structures of the PDB and Structural Classification of RNA database (55).

Refinement and Fitting of the rRNA and r-Proteins into the EM Densities.

The de novo modeled RNA parts were initially refined using the internal refinement tool of Assemble. A preliminary rigid body fitting of the models was done without proteins using Chimera (56) with low-pass filtered electron densities. Subsequently, all RNA segments were merged using visual molecular dynamics (VMD) (57), and MDFF (36) was applied to fit the rRNA to the density map while preserving canonical and noncanonical base-pair interactions identified by RNAview. Subsequently, proteins were introduced using VMD, and an extended version of interactive molecular dynamics (58), namely, interactive MDFF, was used to refine the proteins into the density while fixing protein–RNA and protein–protein clashes, followed by an MDFF refinement of the entire 80S model.

Visualization and Figure Preparation.

Cryo-EM maps and models were visualized and all figures were generated using VMD (57), Chimera (56), and/or PyMol (http://www.pymol.org).