gutell 096.jmb.2006.358.0193

doi:10.1016/j.jmb.2006.01.094 J. Mol. Biol. (2006) 358, 193–212

A Structural Model for the Large Subunit ofthe Mammalian Mitochondrial Ribosome

Jason A. Mears1,2†, Manjuli R. Sharma3†, Robin R. Gutell4

Amanda S. McCook1, Paul E. Richardson5, Thomas R. Caulfield6

Rajendra K. Agrawal3,7 and Stephen C. Harvey1*

1Department of Biology, GeorgiaInstitute of Technology, AtlantaGA 30332, USA

2Department of Biochemistryand Molecular GeneticsUniversity of Alabama atBirmingham, BirminghamAL 35295, USA

3Division of MolecularMedicine, Wadsworth CenterNew York State Department ofHealth, Albany, NY 12201USA

4Institute for Cellular andMolecular Biology and Sectionof Integrative BiologyUniversity of Texas at Austin2500 Speedway, Austin, TX78712, USA

5Department of ChemistryCoastal Carolina UniversityConway, SC 29528, USA

6Department of Chemistry &Biochemistry, Georgia Instituteof Technology, Atlanta, GA30332, USA

7Department of BiomedicalSciences, State University ofNew York at Albany, AlbanyNY 12201, USA

0022-2836/$ - see front matter q 2006 E

† J. A. M. and M.R.S. contributedAbbreviations used: mitoribosom

mitochondrial ribosomal proteins; Lmicroscopy; SRL, sarcin-ricin loop;L1-binding domain.E-mail address of the correspond

Protein translation is essential for all forms of life and is conducted by amacromolecular complex, the ribosome. Evolutionary changes in proteinand RNA sequences can affect the 3D organization of structural features inribosomes in different species. The most dramatic changes occur in animalmitochondria, whose genomes have been reduced and altered significantly.The RNA component of the mitochondrial ribosome (mitoribosome) isreduced in size, with a compensatory increase in protein content. Untilrecently, it was unclear how these changes affect the 3D structure of themitoribosome. Here, we present a structural model of the large subunit ofthe mammalian mitoribosome developed by combining molecularmodeling techniques with cryo-electron microscopic data at 12.1 Aresolution. The model contains 93% of the mitochondrial rRNA sequenceand 16 mitochondrial ribosomal proteins in the large subunit of themitoribosome. Despite the smaller mitochondrial rRNA, the spatialpositions of RNA domains known to be involved directly in proteinsynthesis are essentially the same as in bacterial and archaeal ribosomes.However, the dramatic reduction in rRNA content necessitates evolution ofunique structural features to maintain connectivity between RNAdomains.The smaller rRNA sequence also limits the likelihood of tRNA bindingat the E-site of the mitoribosome, and correlates with the reduced size ofD-loops and T-loops in some animal mitochondrial tRNAs, suggestingco-evolution of mitochondrial rRNA and tRNA structures.

q 2006 Elsevier Ltd. All rights reserved.

Keywords: ribosome; mitochondrial evolution; rRNA; comparativesequence analysis; molecular fitting
*Corresponding author
lsevier Ltd. All rights reserved.

equally to this work.e, mitochondrial ribosome; mito-rRNA, mitochondrial ribosomal RNA; MRPs,SU, large subunit; PTC, peptidyl transferase center; cryo-EM, cryo-electronASF, the A-site finger motif; 5S-BD, the 5 S rRNA-binding domain; L1-BD, the

ing author: [email protected]

194 Structural Evolution in Mitochondrial Ribosomes

Introduction

The mammalian mitochondrial ribosome (mitor-ibosome) is responsible for synthesis of 13 mito-chondrial gene products that are essentialcomponents of the complexes involved in oxidativephosphorylation.1,2 The importance of these pro-teins in generating ATP places a significant burdenof accuracy on the mitoribosome. The mitochon-drion also plays a crucial role in the initiation ofapoptosis,3 and mitochondrial defects have beenimplicated in a wide variety of degenerativediseases, aging, and cancer.4 However, evolutionarypressures to maintain nuclear control of cellularmetabolism following endosymbiosis5 may beresponsible for the significant reduction in animalmitochondrial ribosomal RNA (mito-rRNA)sequence when compared to bacteria, archaea andeukaryotes. This reduction is compensated by anincrease in the size and number of mitochondrialribosomal proteins (MRPs),6–8 whose genes areunder nuclear control.9

Even though mitochondria are believed to havedescended from an endosymbiotic eubacter-ium,10,11 the structural components of theirribosomes are noticeably different. The ratio ofprotein to rRNA mass in animal mitochondria (2:1)is inverted from the ratio found in bacterial andarchaeal ribosomes (1:2). A decrease in particledensity is therefore observed in sedimentationexperiments, yielding a 55 S value for the intactbovine mitoribosome compared to 70 S in bacteriaand archaea. The decreased sedimentation coeffi-cient can also be attributed to a more porousstructure in mitochondrial ribosomes.12,13

The bovine 55 S mitoribosome is comprised oftwo asymmetric subunits, a small (28 S) subunitand large (39 S) subunit. The small subunit containsa 12 S rRNA (955 nucleotides) with 29 pro-teins,6,14,15 and the large subunit contains a 16 SrRNA (1571 nucleotides) with 48 proteins.6–8 All ofthe MRPs are encoded by rapidly evolving nucleargenes,16 while the mitochondrial rRNAs, which arealso evolving at a rapid rate,5 are encoded by themitochondrial genome and are transcribed withinthe mitochondrion. For comparison, the bacterialand archaeal ribosomes are also comprised of twoasymmetric subunits, but the small (30 S) subunitcontains a 16 S rRNA (1500 nucleotides, on average)with roughly 20 proteins and the large (50 S)subunit contains two ribosomal RNA (rRNA)components, 5 S rRNA (120 nucleotides, on aver-age) and 23 S rRNA (3000 nucleotides, on average),with more than 30 proteins. tRNAs migrate throughthree relatively stable binding sites in cytoplasmicribosomes during translation: the A-site (ami-noacyl), the P-site (peptidyl), and the E-site (exit).

Here, we focus our attention on the structure ofthe large subunit of the mammalian mitoribosome.The large subunit (LSU) rRNA is responsible forcatalysis of peptide bond formation. Recent studiessuggest that the mechanism of peptide bondformation can be attributed to positioning transfer

RNA (tRNA) substrates charged with amino acidresidues in a specified proximity during thereaction.17,18 The 2 0-OH of residue A76 of theP-site tRNA has been proposed as the catalyticcomponent.19 The LSU rRNA domain V, whichcontains the peptidyl transferase center (PTC), islargely conserved through all organisms.20 Many ofthe rRNA regions of domain IV that are involved intRNA and inter-subunit interactions21 are alsopreserved.20

The ribosomal protein L11-binding domain(L11-BD) within the LSU rRNA domain II and thesarcin-ricin loop within domain VI, which togetherconstitute the GTPase-associated center of theribosome that interacts with translation cofactors(EF-Tu, EF-G, RF3, etc.),22,23 are also conserved.However, the sequences and structures that connectthe central core and these peripheral structuralelements are truncated in several mitochondrialLSU rRNAs.24 On the other side of the LSU,dynamic motions of the L1 protein and associatedrRNA have been proposed to affect E-site occu-pancy on the ribosome directly.25–27 Together,these functional domains work in concert duringprokaryotic translation, and structural studies arebeginning to elucidate the mechanisms required forprotein synthesis. Even so, it is unclear how theevolving mitoribosome compensates for the largereduction in rRNA sequence while maintaining theprecision that protein synthesis demands.

A recent cryo-electron microscopy (cryo-EM)study has provided the first detailed structure ofthe mammalian mitoribosome.13 Previously deter-mined X-ray structures21,25,28 of large ribosomalsubunits were compared with the mitoribosomestructure, but no effort was made to optimize the fitof the structures to the EM density. Here, a higherresolution (12.1 A) provides an improved rRNA–protein separation in the EM density, allowing adetailed fitting of modeled structures. As docu-mented previously,13 the structure has severalunique features when compared with structures ofcytoplasmic ribosomes from prokaryotic andeukaryotic organisms. Also, the additional proteincontent does not compensate for the missing RNAsequence, as has been proposed.7,16 Instead, manyof the additional proteins assume unique positions inthe mitoribosome structure, thereby leaving regionsof structure vacant where rRNA helices present inbacteria have been deleted in themitoribosome.Also,in contrast to the characteristics of cytoplasmicribosomes, no tRNA was found at the putative exitsite (E-site) of the 55 S mitoribosome, suggesting thatthe E-site is either very weak or non-existent in themitoribosome, consistent with suggestions from acomparative analysis of mitochondrial and nuclear-encoded ribosomes.24

In this study, we analyze the structural pertur-bations in the regions of the mammalian mitoribo-some that have significant changes in size andsequence of rRNA and proteins. We utilize a varietyof methods, including bioinformatics, to study theevolution of RNA and protein sequences, structural

Structural Evolution in Mitochondrial Ribosomes 195

biology (cryo-EM), and novel computational toolsto unify the results in a detailed 3D model of themitoribosome. In most cases, the reduction in mito-rRNA sequence does not alter the 3D spatiallocation of conserved rRNA helical elements.Thus, many of the interactions between themitoribosome, tRNAs and protein cofactors duringthe translation cycle are preserved. However, someof the reductions in rRNA sequence result in theloss or reorientation of functional domains, indicat-ing structures in the bacterial and archaeal ribo-somes that are highly variable and may not beessential for translation. We have modeled thestructures and placement for 16 MRPs that havesufficient levels of sequence identity with homolo-gous prokaryotic proteins whose structures havebeen determined by X-ray crystallography. Thiswork provides a first step towards assigningstructure to the protein mass of the mitochondrialribonucleoprotein structure. Additional structuraland biochemical studies are required to predict theplacement and conformations of the remaining,unmodeled proteins, especially those that areunique to the mitoribosome.

Results

Homology modeling of mitochondrial rRNA(mito-rRNA)

Based on the assumption that the secondary andtertiary structures for rRNA molecules are mostlyconserved for all organisms, we have usedcomparative sequence analysis to identifysequence conservation and variation, and deter-mine similar structural elements that are present insets of distantly related and closely related rRNAsequences. Previous success in determining thesecondary structures for the small and largesubunit rRNAs29 lends confidence to the secondarystructure model for the mitochondrial LSU rRNAderived from comparative sequence analysis(Figure 1). Approximately 86% of the Bos taurusmito-rRNA can be assigned to structural featurespresent in the archaeal Haloarcula marismortuisecondary structure, available from the Compara-tive RNA Website.20 The B. taurus mito-rRNA(Figure 1(a)) has many features in common withall nuclear-encoded rRNAs, including the peptidyltransfer center (PTC), the sarcin-ricin loop (SRL),and L11-BD. However, many differences areevident, because the size of the LSU mito-rRNAis reduced by almost half compared to bacteria.Many of the helical structures in domains I and IIIare lost, as are the A-site finger motif (ASF) indomain II, and the L1- and the 5 S rRNA-bindingdomains (L1-BD and 5S-BD, respectively) indomain V.

Starting with the secondary structure for the largesubunit rRNA of the bovine mitoribosome(Figure 1(a)), we have used homology modeling30

to generate a structural model for the mito-rRNA of

the LSU of the mitoribosome. The X-ray crystallo-graphic structures of large ribosomal subunits fromH. marismortui,28 an archaeon, and Deinococcusradiodurans,25 a bacterium, provide homologousrRNA structures to guide the modeling. For ourstudies, we primarily used the H. marismortuistructure (PDB accession code 1JJ2) as the templatefor modeling the mitochondrial rRNA structure.Helices and loop structures that are conserved inmitochondria are modeled on the basis of geometrydetermined by X-ray crystallography for thearcheon (see Methods). Three significant differenceswere explored: (1) identical secondary structureelements composed of the same base-pairings andunpaired nucleotides are modeled one for one,when the sequences are identical or vary incomposition. The nucleotides from H. marismortuiare replaced by corresponding nucleotides in themito-rRNA. These changes do not affect thebackbone or sugar orientation of the bases. (2) Forcanonical base-pairs that are replaced by non-canonical pairs (i.e. G$U or G$A), the latter pair issuperposed on the canonical pair. The samemethodis used for non-canonical pairs that are replaced bycanonical pairs. None of these changes in base-pairtypes distort the helical geometry severely, anddifferences in the backbone interactions withneighboring nucleotides are easily resolved with around of energy minimization. (3) For moredramatic differences between the two secondarystructures, including changes in the size of loopsand bulges, the mitochondrial sequence is modeledfrom previously determined X-ray crystallographystructures of RNAwith similar sequence (availablefrom the RCSB Protein Data Bank31).

Structure refinement and validation

The homology model of the mito-rRNA providesa starting structure with helical positions based onsimilarities to prokaryotic ribosomes as determinedby comparative sequence analysis. Informationfrom a cryo-EM study13 provides additional,experimental restraints that can be incorporatedusing YAMMP,32 our in-house molecular modelingpackage. The software includes a rigid-body MonteCarlo module with simulated annealing that we useto refine our model. The vector lattice (VLAT)component of YAMMP generates a force–field termthat defines the electron density from experimentalstudies as a 3D potential, providing a score for thefit of the model to the density.Initially, the complete model for the mitochon-

drial rRNA is treated as a single rigid unit using areduced representation.33 The starting model con-tains pseudoatoms representing the phosphatepositions of each nucleotide in the RNA homologymodel (P-atoms). To get an initial placement of themodel in the EM density, the rigid unit wassubjected to Monte Carlo refinement with simu-lated annealing (see Methods). A good fit isobtained, suggesting that the structural organiza-tion is largely conserved in the mitoribosome, with

Figure 1. rRNA secondary structure comparison. (a) Bos taurus mitochondrial rRNA secondary structure based oncomparative sequence analysis. (b) Secondary structure of H. marismortui 23 S rRNA. Regions that align with themitochondrial rRNA are highlighted in red, and those that are absent from the mitoribosome are shown in black.Some relevant functional regions are labeled (green), and six domains of the 23 S rRNA are identified with romannumerals.



small changes due to differences in sequence andsize of some helices and connecting regions.

The mito-rRNA model was then divided into 52rigid units, based on conserved helical structures(Figure 2), and the structure was refined to thecomplete large subunit density with multiple rigid-body Monte Carlo with simulated annealing. Eachrigid unit is topologically connected to neighboringunits using one of two bond-types. (1) Strong bondsare used to connect adjacent conserved structuresthat are separated by distances corresponding to thenormal phosphate–phosphate distances betweenneighboring bases (w6–7 A). (2) Weak bondsconnect nucleotides separated by a gap in themodel due to a lack of secondary structureinformation (Figure 2, grey regions). The weakerbonds allow greater freedom for the connectedstructures, while maintaining reasonable connec-tivity during the Monte Carlo refinement. Non-bond interactions with 7.5 A exclusion diameterswere used for every P-atom in the structure toprevent structural overlap.

All-atom models for the 52 units were thensuperposed on the final reduced representationmodels fit to the EM density, given that the reducedrRNA units were treated as rigid bodies andmaintained their geometries during refinement.The all-atom models for each unit were covalentlylinked, except for the regions where gaps occur(Figure 2, broken grey lines); and a round of energyminimization was used to resolve structural dis-crepancies caused by rearrangements during rigid-body refinement.

The fit of this model was evaluated using theseparated rRNA density. With few exceptions (seebelow), the modeled structures fit the density well.This is due to the large conservation of structuralelements found in both mitochondrial and prokar-yotic rRNAs (seen in Figure 1). H27 in domain II(Figure 2) does have a slightly different positionrelative to the rest of the rRNA structure because ofthe increased size in the adjacent loop whencompared to archaea. This does not affect inter-actions with other conserved rRNA regions. Also,H75 (Figure 2) has a change in sequence that placesan unpaired nucleotide on the opposing side of thehelix when compared with archaea. This changeforces the helix into a position that places the L1protein away from the E-site (discussed below). Themost dramatic changes in rRNA structure are dueto sequences that are missing or unique regionswhose structures could not be determined bycomparative sequence analysis.

Additional and alternative RNA secondarystructure predictions

The only helices in the LSU rRNA comparativemodel that do not fit the EM density afterrefinement are in domains I and III (in Figure 2,helix 13, the end of helix 50 and helix 59.1 didnot find a reasonable fit to the cryo-EM density).We therefore explored the possibility that these two

regions of the rRNA have alternative folds usingAlifold.34 In contrast to comparative analysis of therRNAs that identifies base-pairs from patterns ofvariation and covariation in a set of alignedsequences,35 the Alifold34 program combines ther-modynamic and comparative analyses to predictRNA secondary structure. A multiple sequencealignment was created for a set of mammalianmitochondrial LSU rRNAs, that includes B. taurusand related organisms (see Methods). The Alifoldprogram identified helices that are thermodynami-cally stable and present in the set of alignedsequences. This suggests a secondary structure inthe region of helices H50 and H59.1 (Figure 3, greennucleotides) that differs somewhat from thatproposed by comparative analysis (Figure 2). Themodeled structure matches closely the RNA separ-ated EM density (Figure 4(c)). This suggests one oftwo possibilities: (1) that after endosymbiosis thesequence in domain III of the bovine mito-rRNAhas evolved to generate a somewhat different 3Dstructure; or (2) the structure determined by cryo-EM represents a conformational change in thisregion from the structure predicted by comparativesequence analysis. We cannot rule out eitherpossibility, but our model is based on the alternativesecondary structure (shown in Figure 3) that bestfits the cryo-EM density. We examined alternativesecondary structures for the unstructured regions ofdomain I (including helix 13), but wewere unable tofind secondary structure predictions that could beplaced in the EM density with confidence.Part of the region in domain II that connects the

L11-BD to the highly conserved core of the LSU hasfewer nucleotides in the mammalian mitochondriathan the corresponding region in all nuclear-encoded LSU rRNAs (Figure 1). No base-pairingor helices that are shared in the mammalianmitochondria LSU rRNA were identified withcomparative and covariation analysis for thistruncated region. To examine the possibility ofadditional secondary structure in this region, weused the mfold36 thermodynamic folding programto predict helices to expand the model fromcomparative analysis. Several thermodynamicallystable secondary structure helices were identified,and we tested each of these by examining their fitsto the RNA separated cryo-EM density. From this,we were able to generate a 3D model that matchesthe cryo-EM density very well in this region(Figure 4(a)). This model reveals how, in spite ofthe drastic reduction in rRNA size, the L11-BD ofthe mito-rRNA can remain on the periphery of themitoribosome. A previous modeling study24 pro-posed a movement of L11-BD closer to the core ofthe LSU during the course of evolution in theCaenorhabditis elegans mitoribosome, because nopreviously characterized RNA structure of similarsequence and length was known to span such alarge distance. No such movement is necessary inthe mammalian mitochondria for two reasons: (1)the sequence is not as reduced in mammalianmitochondria as in C. elegans; and (2) the unique

Figure 2. Rigid-body refinement of the rRNA model. A total of 52 independent rigid units were defined for refinement of the RNA structure using Monte Carlo withsimulated annealing (see Methods). Regions are colored to represent the distinct units that were used for the refinement. Unmodeled regions are represented as broken greylines. Most modeled helices maintain a similar structure and relative location when compared with their bacterial homologs. Helices 27 and 75 are highlighted because ofaltered conformations due to changes in sequence. Helices 13, 50 and 59.1, proposed by comparative sequence analysis (http://rna.icmb.utexas.edu), did not fit the EM density.We propose an alternative structure in the region with helices 50 and 59.1 in domain III (see Figures 3 and 4).

http://rna.icmb.utexas.edu

Figure 3. Proposed secondary structure for the mammalian mitochondrial LSU rRNA. The structure is presented as modeled with regions predicted by comparativesequence analysis (blue) comprising roughly 86% of the secondary structure. Additional secondary structure was predicted using mfold34 (orange) to extend the structure forregions where comparative analysis does not predict secondary structure. Alifold34 was used to predict an alternative secondary structure in domain III (green) that differsslightly from that predicted by comparative analysis, but matches the cryo-EM density more closely.13

Figure 4. Additional and alternative secondary struc-tures predicted using thermodynamic and comparativemethods. The left side of each panel shows the predictedsecondary structure, colored as in Figure 3, and the rightside of each panel shows the fit to the correspondingregion in the cryo-EM density map.13 (a) The L11-BD,where additional structure is predicted for the adjacentsequence (orange) using thermodynamic predictionsfrom mfold.36 (b) The L1-BD, where additional pairingis predicted using mfold36 for the rRNA sequence thatinteracts with MRP-L1 (red). A large globular mass ofunidentified protein(s) (marked by * on the semitran-sparent green density) may restrict the movement of theL1 region in the mitoribosome. (c) The domain III regionwhere an alternative secondary was predicted usingAlifold.34 This structure differs from that predicted bycomparative sequence analysis (Figure 1(a)) and fits thecorresponding cryo-EM density much better (not shown).(d) The cryo-EM density for the large mitoribosomesubunit is shown in blue (interface view). The threeregions that have been modeled using additional and


structure found in the mammalian mitoribosomeextends from the core of the subunit to theperiphery (w80 A) by restricting base-pairing andallowing the RNA strand to stretch without theconstraints of helical geometry. In fact, a largeportion of the sequence near the L11-BD does notform helical base-pairs, because the sequenceconsists almost entirely of adenine and cytosinebases (one uracil and no guanine, Figure 3(a)).Cryo-EM density connecting the L11-BD with therest of the rRNA is relatively thin, which suggestsconformational variability in the region, consistentwith non-canonical interactions.

A similar method was used to predict thesecondary structure of an rRNA helix in domainV, which extends to the L1-binding domain (L1-BD).This region of the LSU rRNA is truncated in themammalian mitochondrion, and no base-pairing ispredicted at its apex in the mammalian mitochon-drial comparative structure model (Figure 1(a)).This truncation implies that L1 does not bind to thesame RNA site, despite the conservation of L1 inmammalian mitoribosomes.37 Thermodynamic pre-dictions indicate that a single stem-loop structure isfeasible (Figure 3(b)), and modeling places the RNAin close proximity with the L1 protein (Figure 4(b)).For this reason, an RNA:protein interaction isprobably maintained, with the large bulge in thehairpin easily fitting the cleft of the L1 protein.38

We attempted to determine energetically stablehelices in the remaining regions of the B. taurus LSUrRNA that did not have helices in the comparativestructure model (Figure 3, grey lettering). Unfortu-nately, no common structures were found. In fact,no base-pairings were predicted by mfold36 for thelarge loops in domains II and III, because thesesequences lack the nucleic acid base diversityrequired for canonical pairing (no guanine andfew uracil bases). The abundance of adenine andcytosine bases in these unstructured regionssuggests that these regions do not contain regularbase-pairing and helices. The same is true for thesingle-stranded region in domain VI (Figure 3, oneuracil and no guanine). The unmodeled regions indomain I are not completely devoid of guanine, butthey are also very G-poor. Neither Alifold normfoldwas able to predict common secondary structuresfor this sequence. We have therefore elected not tomodel these regions. Their structures are probablyunique to mitoribosomes.

The final model of mito-rRNA

In total, we have modeled 93% of the mito-rRNAsequence in three dimensions (Figure 5). Much ofthe 16 S mito-rRNA is conserved in domain IV(Figure 5, green), which lies at the interface between

alternative folding predictions are shown in orange andlabeled a, b and c to correspond with the precedingpanels.

Figure 5. Three-dimensional model for the mitochondrial 16 S rRNA. (a) The 16 S rRNA is represented from theinterface and solvent-accessible sides of the structure and colored by domain (I, purple; II, dark blue; III, orange; IV,green; V, red; and VI, yellow). (b) The model RNA fit to EM density that is attributable to RNA,13 except for the tip of adomain V helix that contacts the L1 protein. However, the model of the extended rRNA segment fits into the completeLSU map (also see Figure 4(b)). Coloring is the same as in (a).


the mitoribosome subunits. Interactions with thepenultimate stem of the small ribosomal subunit(SSU helix 44) are maintained near the core of thesubunit. The top of helix 44 contains the decodingcenter, where interactions between the mRNA andthe A-site tRNA are examined for fidelity,39 andsignals for fidelity may be transmitted throughinteractions with domain IVof the LSU rRNAduringtranslation. It is not surprising that these interactionsare largely conserved, but more peripheral inter-actions at the subunit interface are replaced by newinteractions between MRPs in each subunit.13

The central cavity of the 39 S subunit is highlyconserved, because several essential structures arepreserved in domain V (Figure 5, red), including thePTC. The 5S-BD is missing, in agreement with theloss of 5 S rRNA in the mitochondrial genome. Thehelices that radiate from the core of the structure tothe L1 arm are maintained, but are shorter than inbacterial ribsosomes, making interactions with theL1 protein in the mitoribosome unique. Thepositions of functional structures near the peripheryof the structure (L11-BD and SRL) are conserved,thereby preserving critical interactions with elonga-tion factors during the translation cycle.

Protein homology modeling

Of the 48 proteins found in the large subunit ofthe mitoribosome, 28 are homologous to prokar-yotic ribosomal proteins,8 while the remaining 20

are unique to the bovine mitoribosome. All of thebovine MRPs are encoded in the nuclear genome,9

and must be translated in the cytoplasm andtransported into the mitochondrion.40,41 MRPs aremuch larger than prokaryotic ribosomal proteins,and they often have insertions at the N terminus, atthe C terminus or both. It was originally thoughtthat the increase in protein size compensates forreduction in the mito-rRNA,33 but the cryo-EMstructure does not support this theory,13 as onlyw20% of deleted rRNA components are position-ally replaced by mitoribosome-specific proteins orextensions of homologous proteins.We compared all 48 MRP sequences6–8 with

sequences of ribosomal proteins whose structureshave been determined by X-ray crystallography.When significant levels of identity and similaritywere found, we were able to generate proteinhomology models based on templates from X-raycrystal structures of ribosomal large subunitcomplexes.21,25,28,42 We created partial models for16 proteins (Figure 6), and a summary of thosemodels is given in Table 1. (Note that L7 and L12 areidentical.)None of the ribosomal proteins could be modeled

completely. Homology between the MRPs and theprokaryotic ribosomal proteins did not extendacross the entire length of the protein sequences.Commonly the N terminus, the C terminus, or bothtermini were unique. More detailed sequenceanalysis revealed that these regions are insertions,

Figure 6. Structural homology models for MRPs. (a) Models of the 16 MRPs with significant sequence similarity tobacterial ribosomal proteins that have been structurally characterized by X-ray crystallography. (b) Structuralorganization of proteins in the large mitochondrial subunit fit to the cryo-EM density (stereo view).


not mutational differences. This trend has beendescribed,6–8 and the results from our comparativestudy confirm those findings.

It had been proposed that these insertionsrepresent the addition of functional domains tocompensate for the decreased size of rRNA in thelarge subunit.6–8 We analyzed these insertions todetermine if functional roles could be identifiedfrom sequence homology to proteins of knownfunction. Unfortunately, the homology searchagainst the non-redundant (nr) database did notproduce any homology matches, suggesting thatthe sequence insertions in mammalian MRPs areunique.

MRP structure refinement

The homology models for the proteins were fit tothe cryo-EM density in much the same way as the

RNA helices. During protein fitting, the mito-rRNAmodel (modeled again with the P-atom reducedrepresentation) was treated as a rigid scaffold thatwas held fixed. Each protein (modeled withpseudoatoms centered on the Ca coordinates, calledC-atoms) was placed manually into the EM density,and rigid-bodyMonte Carlo with simulated anneal-ing was used to refine protein positions. Refinementincluded VLAT terms for scoring the fits of theproteins to the density plus a set of restraints basedon conserved RNA–protein interactions observedin bacterial crystal structures. These interactionswere determined by measuring distances betweenevery nucleotide in the RNA and every amino acidfor each protein in the crystal structure of thebacterial ribosome. An interaction threshold of 3 Awas used for protein models that had homologousproteins in the H. marismortui structure (where all-atom detail is present),28 whereas a threshold of 6 A

Table 1. Proteins modeled and their homologs

Proteina Homologsb Identity (%) Similarity (%) Model size (residues) MRP size (residues)

MRP-L1 1GIY 28.6 52.4 100–289 303MRP-L2 1NKW 38.2 56.6 126–261 305

1PNU 36.8 54.41GIY 40.4 57.41JJ2 31.6 51.5

MRP-L3 1NKW 28.4 50.7 96–306 3481PNU 28.4 51.1

MRP-L4 1PNU 33.1 56.0 97–271 311MRP-L7 1GIY 31.2 47.1 61–198 198MRP-L12 1GIY 31.2 47.1 61–198 198MRP-L11 1GIY 42.8 61.4 13–157 178

1NKW 35.2 57.21PNU 35.2 57.2

MRP-L13 1NKW 29.1 57.4 1–148 1781PNU 29.1 57.4

MRP-L16 1NKW 27.1 54.2 71–188 2511PNU 27.1 54.2

MRP-L17 1NKW 33.6 59.5 11–126 1751PNU 33.6 59.5

MRP-L19 1NKW 26.5 48.0 96–193 2801PNU 26.5 48.0

MRP-L20 1NKW 36.4 63.6 8–125 1491PNU 36.4 63.6

MRP-L22 1GIY 29.2 47.5 69–178 2061PNU 21.7 48.3

MRP-L24 1PNU 35.4 56.3 59–154 216MRP-L27 1NKW 43.5 65.2 31–99 148

1PNU 43.5 65.2MRP-L33 1NKW 51.9 71.2 9–60 65

1PNU 51.9 71.2MRP-L34 1NKW 44.7 65.8 53–90 92

1PNU 44.7 65.8

a Mitochondrial ribosomal proteins (MRPs) with significant homology to ribosomal proteins that were characterized previously arelisted.

b The homologs that were identified are listed by the Protein Data Bank accession code (1GIY is from Thermus thermophilus, 1NKW isfrom Deinococcus radiodurans, 1PNU is from Escherichia coli, and 1JJ2 is from Haloarcula marismortui).


was set for protein models with homologousstructures available from Thermus thermophilus21 orDeinococcus radiodurans25 (where all-atom detail isnot available). If the nucleotide and amino acidinvolved in the interaction were both conservedin the mitochondrion, harmonic restraints ofequivalent length were included between appro-priate P-atoms and C-atoms during refinement.

Of the 16 protein models, 13 had conservedinteractions with the RNA. The resulting RNA–protein restraints played a critical role in tetheringthe proteins to their conserved positions in themitoribosome model during refinement, becausemuch of the protein density remains unfilled, and itwould otherwise be difficult to guarantee that therefinement would not move these proteins intoinappropriate regions of the density. The positionsof the three proteins that did not have conservedinteractions were refined, although a lower startingtemperature was required during the refinementprotocol to prevent large movements. Also, forproteins with long extended loops, minimalrearrangements of some loops were required toprevent steric overlapbetween theproteinand rRNA.

All 16 proteins are in positions very similar tothose found in the prokaryotic structures(Figure 6(b)). In most cases where restraints areavailable, the nucleotide-binding sites for the

modeled proteins are conserved. For example, thebinding site for protein L2 is conserved in domainIV, which contains helices responsible for position-ing the protein in the same relative orientation as inthe bacterial ribosome. L11 and L7/12 are also inpositions very similar to those in bacteria, so thegeometries of their interactions with protein cofac-tors and the incoming A-site tRNA in mitochondriashould be similar to those in bacteria. However, therRNA-binding site for L1 (L1-BD) is very differentin the mito-rRNA than it is in bacterial rRNA. L1 isrepositioned further from the putative E-site, whichmay be absent from the mitoribosome.13,24 Some ofthis difference is presumably due to the drastictruncation of the L1-BD in mito-rRNA, but it mayreflect the greater flexibility of the L1 arm of thelarge subunit25–27 when compared with nuclear-encoded ribosomes (see Discussion).

The final model of the LSU mitoribosome

The final model for the LSU of the mammalianmitoribosome fits the cryo-EM density nicely(Figure 7). We have placed 93% of the mito-rRNAsequence as well as 16 MRP models with significantsequence similarity to prokaryotic ribosomalproteins. The L1-BD of the RNA that extends outof the density attributed to RNA does fit the density

Figure 7. Stereo-view representation of the 3D model of the 39 S subunit of the mitochondrial ribosome. (a) Theinterface view of the subunit shows that the conserved interface of the mitochondrial ribosome is still dominated byrRNA structure (colored as in Figure 5). (b) The homologous MRPs (grey) are located predominantly towards thesolvent-accessible side of the particle. Upper and lower panels in both sections show the modeled structure (rRNA andproteins), and its fitting into the cryo-EM map,13 respectively.



of the complete subunit, and the placement of theL1 protein suggests that an interaction between theRNA and protein is possible. Our placement ofthe L7/12 dimer is based on an earlier X-raycrystallographic study,21 and the protein doesprotrude slightly from the EM density. Since thisregion is known to be flexible in bacterial ribosomalparticles,43,44 similar flexibility in mitochondriacould explain the weaker density for the proteinin this region. Moreover, in a recent study45 it hasbeen suggested that the traditionally assigned stalkof the LSU actually represents the protein L10 andmultiple copies of L7/12. If this scenario is true inthe mitoribosome also, our placement of L7/12would need further refinement.

Discussion

Our model of the large subunit of the mammalianmitoribosome suggests that the reduction in rRNAsize compared to that of bacteria does little to alterthe spatial organization of structural and functionaldomains common to the LSU of mammalianmitoribosomes and prokaryotic ribosomes. How-ever, some conserved interactions, including partsof the tRNA-binding sites, are absent in themitoribosome. The interactions between the largesubunit and tRNAs at the A-site, P-site and E-site ofthe bacterial ribosome have been characterized byX-ray crystallography studies.21 Figure 8 comparesthe interactions predicted by our model with thoseobserved for bacterial ribosomes.

rRNA segments interacting with the P-site arehighly conserved between mitochondrial and cyto-plasmic ribosomes, and all tRNA contacts with theLSU are preserved in the mitoribosome (Figure 8).The importance of positioning the acceptor stemcorrectly, which holds the nascent polypeptidechain, is evident from the rRNA sequence con-servation for regions interacting with the tRNA.Furthermore, additional contacts are made betweenthe tRNA and the mitoribosome at this positionduring mitochondrial translation by protein(s) inthe central protuberance (the so-called P-sitefinger). The conserved protein models that we fitto the EM density are not in positions to participatein these new interactions, so the P-site fingerinteraction can be attributed to extension(s) of one

of the 12 unmodeled bacterial homologs or, morelikely, to an MRP unique to mitochondrial trans-lation.Interactions at the A-site are also highly con-

served (Figure 8), showing the importance ofpositioning the A-site and P-site tRNAs duringcatalysis of the peptidyltransferase reaction.18

Contacts are maintained also between the highlyconserved rRNA helix 69 and the minor groove ofthe tRNA D-stem. But interactions between the ASFand the D-loop and T-loop of tRNAs in cytoplasmicribosomes21 are lost because of sequence reductionin domain II of the LSU rRNA (Figure 1(b)). ThesetRNA loops exhibit a large degree of variation insize and content in mitochondria,46,47 suggestingthat the rRNA and tRNA sequences that normallyinteract have co-evolved to accommodate structuralchanges. D-loop and T-loop structures are notsensed by the ribosome at the P-site, and newinteractions with the P-site finger are located at theT-stem. It has been suggested that reductions in thesize of D-stems, T-stems and loops in somemitochondrial tRNAs may cause a dramatic changein the preferred angle between the two arms ofthese tRNAs,46,48 and transient electric birefrin-gence studies have supported this suggestion.49

Our model of the LSU mitoribosome would permitthis kind of variability in tRNA conformation at theA-site, but it is not clear what, if any, affect thisdifference may play in translational fidelity.The ribosomal E-site is markedly different in

mitoribosomes, as most of the rRNA regionsresponsible for interacting with the E-site tRNA inthe bacterial ribosome are absent from the mitor-ibosome (Figure 8).24 The one interaction that mightbe structurally conserved occurs near the base of theL1 arm in the rRNA model. The sequence in thisregion is markedly different in the mito-rRNAcompared with prokaryotes, so any interactionwill be different, if not completely lost. Further-more, the observation that the tRNA binds stronglyin the P-site, but not in the E-site, suggests that theE-site is either very weak or non-existent in themitoribosome.13

Many interactions between the prokaryotic 23 SrRNA and E-site tRNA involve contacts near theL1-BD.21 These interactions and the flexibility ofthe L1 arm (Figure 9(b), L1) suggest that E-siteoccupancy on the ribosome may be coupled to

Figure 8. Interactions with thetRNA-binding sites of the mito-ribosome. The A-site, P-site andE-site tRNAs are represented withthe structure from yeast tRNA-Phe.76 Interactions with rRNAsequence that were found in theX-ray crystal structure of theT. thermophilus 70 S particle21 arerepresented as either beingconserved in (red) or absentfrom (yellow) the mitochondrialribosome.

Figure 9. Three-dimensional models of mitochondrial and archaeal large ribosomal subunit rRNAs. Models are shownfrom the subunit interface side (left) and from the solvent side (right). (a) Mitochondrial rRNA, showing a dramaticreduction compared to archaea. (b) H. marismortui rRNA from X-ray crystallography.28 The L1-arm (L1) was modeledusing structural data from other crystallographic structures.21,38 The A-site finger (blue RNA helix adjacent to *) wasmodeled using sequence data and cryo-EM density from E. coli.64 Six domains in both models (a) and (b) are identifiedby different colors: I, purple; II, dark blue; III, orange; IV, green; V, red; and VI, light blue. The 5 S rRNA in the archaeon(yellow) is absent from the mitoribosome. L1 proteins for both models are shown with space-filling representations(grey).


dynamic motions in this structure.25–27 The mito-chondrial L1 protein is readily identifiable inthe EM density, interacting with the shortenedmitochondrial rRNA stem-loop in domain V(Figure 4(b)), but the protein is positioned farfrom the putative E-site (Figure 9(a)). While thisorientation may be due, in part, to the inherentflexibility of the L1-BD, the truncation of the rRNAmay limit the range of motion. The EM densitycorresponding to L1 makes contacts with neighbor-ing, unassigned protein density (Figure 4(b),marked with *), which may further restrict itsmobility. Since the L1 protein is conserved in themitoribosome, it is probably important for somefunction, but it may not have a direct role in tRNAbinding.

On the opposite side of the mitoribosomal LSU,the L11-BD is positioned by a unique structure(when compared to Archaea; Figure 9) because ofsequence reduction in the region that connects theconserved cofactor-binding domain and the core ofthe particle (Figure 3, orange region in domain II).Therefore, the spatial orientation of this conservedstructural domain near the periphery is conserved.The crystal structure of mitochondrial EF-Tu incomplex with GDP50 is similar to homologous

bacterial structures from Escherichia coli51 andThermus aquaticus,52 which would suggest that theoverall binding of cofactor with the ribosome inmitochondria is comparable to that in bacteria.

Unique mito-rRNA structures (i.e. the sequenceleading to the L11-BD) may also be stabilized byadditional protein interactions found in the mitor-ibosome. The placement of 16 MRP models beginsthe assignment of structure to the increased proteinmass in the large subunit of the mitoribosome. Theunique MRP sequences suggest that their role intranslation is likely specific to mitochondria. Theunidentified protein densities are mostly localizedto the peripheral regions of the large subunit, andall of the proteins synthesized bymitoribosomes areinserted into the inner mitochondrial membrane.Therefore, unique MRPs appear to be associatedwith positioning the mitoribosome during cotran-slational insertion of nascent polypeptides into themembrane53 and/or stabilizing extended rRNAstructures created by the large reductions insequence.

The conservation of rRNA structure in bovinemitochondria does not correlate directly withsequence conservation. The overall base content isvery G-poor and A-rich when compared to archaeal

Figure 10. Sequence compositionof LSU rRNAs from: Bos taurusmitochondrion, red; H. marismortui,blue; and E. coli, green. (a) Themitochondrial rRNA exhibits asignificant reduction in guanine(G) content and increase in adenine(A) content relative to H. maris-mortui (an archaeon) and E. coli (aeubacterium). (b) The base-pairedfractions for each species. Thefrequency of base-pairing is sub-stantially lower in the mitochon-drial ribosome, especially forcytosine.


(H. marismortui) and bacterial (E. coli) species(Figure 10(a)). While guanine is the most frequentlyoccurring base in these prokaryotic rRNAs (30%), itis sharply reduced in bovine mitochondria (18%). Incontrast, the adenine content (25% in prokaryotes)jumps to 38% in the mitochondrion. The fraction ofnucleotides found in standard secondary structurebase-pairing drops from w60% in prokaryotes to45% in mitochondria (Figure 10(b)). This is partlydue to the reduction in guanine content without acorresponding reduction in cytosine content, sincethe latter have fewer prospective base-pairingpartners. The increase in adenine content alsocontributes to the decreased helical base-pairing inthe mitochondrial rRNA secondary structure(Figure 10(b)), because adenine bases pair lessfrequently than other bases. For example, 62% ofadenine bases are unpaired in the E. coli 16 S rRNA,while only 30% of G, C and U bases areunpaired.35,54 An even larger fraction of adeninebases are unpaired in the E. coli 23 S rRNA(Figure 10(b)). The increased adenine content inthe mitoribosome results in reduced secondarystructure and facilitates formation of two uniquestructural features: (1) “stretched” regions thatreach across large distances to connect functionalregions whose positions are not changed (e.g. theL11-BD, whose position is maintained at theperiphery to interact with translation cofactors);

and (2) large, unpaired loops at helix ends that mayassume globular structures and serve as recognitionelements for some of the MRP binding, e.g. theA-rich loops in domains II and III (grey in Figure 3).It is well known that adenine bases are involved in anumber of unusual tertiary structures,55–60 and theextended structures reported here are additionalexamples of such structures.It is not clear why evolutionary pressures lead to

such a remarkable decrease in the size of the LSUmito-rRNA and a corresponding increase in proteincontent, but two structural principles have emergedfrom the present study. First, the key functionalsites occupy essentially the same positions as inprokaryotic ribosomes. Second, the significantdecrease in G content and increase in A contentleads to a marked reduction in the fraction of base-paired nucleotides, yielding unique structures thatcan span large distances to maintain the 3Dorganization of structures essential for translation.

Methods

Cryo-electron microscopy and 3D imagereconstruction

Cryo-EM grids were prepared according to standardprocedures.61 Data were collected on a Philips FEI


(Eindhoven, The Netherlands) Tecnai F20 field emissiongun electron microscope, equipped with low-dose kit andan Oxford cryo-transfer holder, at a magnification of50,760!. A total of 194 micrographs were scanned on aZeiss flatbed scanner (Z/I Imaging Corporation, Hunts-ville, AL), with a step size of 14 mm, corresponding to2.76 A on the object scale. The projection-matchingprocedure62 within the SPIDER software63 was used toobtain the 3Dmap. The previous 13.5 A resolution map ofthe 55 S mitoribosome13 was used as the reference for thealignment of the data set. Initially, 217,908 images, sortedinto 54 groups according to defocus value (ranging from1.2 mm to 4.6 mm), were picked. Because of the problem ofconformational heterogeneity, we had to eliminate a largeportion of the data set in order to improve resolution.Only 79,516 images were retained, after manual screeningand removal of images from over-represented groupswithin 83 equi-spaced views of the ribosome and wereincluded in the final 3D reconstruction. The resolution ofthe final CTF-corrected 3D map, estimated using theFourier shell correlation with a cutoff value of 0.5,64 was12.1 A (or 8.5 A by the 3s criterion65). The falloff of theFourier amplitudes toward higher spatial frequencies wascorrected as described.64 RNA and protein components ofthe 55 S mitoribosome map were separated computation-ally using a method66 based on differences in the densitydistribution of the two moieties, taking into account themolecular masses and contiguity constraints.

Comparative sequence analysis

The rRNA sequences were aligned manually with thealignment editor AE2 (developed by T. Macke67). Thisprogram runs on SUN Microsystems computers on theSolaris operating system. Ribosomal RNA sequences arealigned by juxtapositioning the nucleotides that map tothe same elements in the secondary and tertiary structuremodels to the same columns in the alignment. The rRNAstructure models were predicted with covariation anal-ysis,35 a method that identifies a conserved set of base-pairings and helices in a group of aligned sequences. Thesecondary structure diagrams for the B. taurus mitochon-drial LSU rRNA were templated from previouslypredicted mammalian mitochondrial LSU rRNA struc-ture models,20 and modified for unique features in theB. taurus structure. The structure diagrams were drawnwith the interactive secondary structure program XRNA(B. Weiser and H. Noller, University of California, SantaCruz).

RNA homology modeling

The model for the mito-rRNA is based largely on thecrystal structure of the large subunit from H. marismortui(PDB accession code 1JJ2).28 To begin, conserved RNAhelices were identified on the basis of the comparativesequence analysis. Having identified homologous regionsin the structure, simple changes could be made in thecases where a base-pair or an individual base (in a loop orbulge) could be changed to the corresponding nucleo-tide(s) found in the mitochondrial sequence (i.e. A-U pairchanged to C-G) using the Biopolymer module of theInsight-II software package (Molecular Simulations, Inc.,San Diego, CA). Other changes involve mutations thatresult in non-canonical base-pairing where normal base-pairs are found in the H. marismortui structure. In stillother cases, a canonical base-pair may be found inthe mitochondrial structure where a non-canonical pair

exists in the archaeal ribosome. For these cases, the pairfound in the mitochondrial structure is superposed on thepair found in the H. marismortui structure. A round ofenergy minimization was used to satisfy the backbonegeometry while preserving the hydrogen bond inter-actions between the base-pairs.For regions where a greater difference is found between

the sequences, the mitochondrial rRNAwas modeled onthe basis of previously characterized structures withsimilar or identical sequences.30 These include commonmotifs found in RNA tertiary structures;68 such astetraloops,55 U-turns,69 and so on. For structures that arenot as common, RNA structures previously determinedby X-ray crystallography are used as a library, ordatabase, for generating a 3D model. Sometimes,sequence comparisons suggested more than one possiblestructure. In such cases, fits to the cryo-EM density wereused to determine which candidate was more likely.

Additional/alternate secondary structure predictions

The B. taurus mitochondrial rRNA sequence was takenfrom the genomic sequence NC_001567 using the Entrezgenome database at NCBI.70 Potential secondary struc-tures for regions of interest were predicted by mfold36

and the Alifold34 program in the Vienna RNA package.Default settings were used to predict secondary structurewith mfold. Elements of secondary structure that wereconsistent across several thermodynamic predictions,levels of lineage, or both, were selected as potentialsecondary structures. These were used to predict 3Dstructures that were validated or rejected by fitting to thecryo-EM density.Multiple sequence alignments were created with the

mitochondrial LSU rRNAs from B. taurus and relatedorganisms (Bovinae, Bovidae, Pecora, Ruminantia,and Cetarteriodactyla), as defined the NCBI TaxonomyBrowser.70 The sequences were obtained using Entrez(nucleotide or genome) and aligned using CLUSTALW.71

Regions of interest were cut from each alignment andsubmitted to Alifold twice, once with a default covarianceweight of 1, and again with a covariance weight of 10. Inboth cases isolated base-pairs were allowed. Alternativesecondary structures were modeled in three dimensionsand then examined in the cryo-EM density map to selectthe one that fit the best.

Protein homology modeling

The sequences of all 48 proteins of the mitoribosome6–8

were identified and searched against the non-redundantsequence database (nr) and the Protein Data Base (PDB)using the program BLASTP.72 MRPs homologous toribosomal proteins in the same family were modeled forthose cases where the crystal structure has beendetermined. CLUSTALW71 was used to create sequencealignments. We used the aligned sequences in themodeling program MODELLER673 to create structuralmodels. Most of the templates selected for the modelingcontained only the Ca atom of each amino acid residue,due to limited resolution of the crystal structures.Therefore, the remaining atomic positions were extrapo-lated using probability density functions.73 This processcan sometimes lead to positioning that is unfavorable, sosome manual adjustments were made with the proteinusing Insight-II (Molecular Simulations, Inc., San Diego,CA). The models were optimized by steepest decentenergy minimization to remove any unfavorable bonds,


angles and steric conflicts. The structural characteristicsof the models were then examined with PROCHECK.38

Regions that could not be modeled were further checkedagainst the nr and PDB databases to determine ifhomologous sequences or structures could be determinedusing just those short, unmodeled regions.

Determining protein interactions

A simple Python script was written to determine theinteractions between RNA and proteins using variouscutoff distances, based on the resolution of thecrystal structure. This procedure works well for theH. marismortui structure,28 because the all-atom detailallows the use of a 3 A cutoff to determine specifichydrogen bonding pairs between RNA and proteinatoms. Not all mammalian MRPs have a high level ofhomology to proteins in the archaeal structure, so specificinteractions cannot be determined for all proteins, buthomologous proteins are also found in the crystalstructures from Thermus thermophilus21 (PDB accessioncode: 1GIY), and D. radiodurans25 (PDB accession code1NKW). These structures contain only Ca coordinates forthe proteins, and 1GIY provides only the phosphatepositions for each nucleotide. Therefore, P–Ca distanceswere measured using a 6 A cutoff for homologous proteinstructures in these eubacterial complexes, providing a listof neighboring residues between the RNA and protein.MRP-L11 is homologous to E. coli L11, so the all-atomcrystal structure of the L11-RNA fragment74 (PDBaccession code 1KC8) was used to determine specificinteractions (3 A cutoff) between the RNA and protein.Protein–protein interactions were analyzed using thesame approach.Once a list of interactions was determined for each

protein model, restraints in the rigid-body Monte Carlorefinement were applied to pairs of pseudoatomsdefining interacting residues in both the proteins andnucleic acid. Of the 16 proteins, 13 have interactions withthe rRNA that could be expressed in such restraints, butL17, L19 and L24 did not have such interactions, due todeletions in themitochondrial rRNA andMRP sequences.

Structure refinement

Cryo-EM density is incorporated as a structuralrestraint for refinement of the model in YAMMP, our in-house molecular modeling package.32 The vector lattice(VLAT) force–field term defines the cryo-EM density as a3D potential, providing a score for the fit of the model tothe density†. Refinement is done using the rigid bodyMonte Carlo module in YAMMP.The mitochondrial rRNAwas modeled using a reduced

representation with one pseudoatom per nucleotide.6 Theinitial refinement was performed by treating the entireRNA homology model as a rigid unit fit to the completeribosome density, and it was subjected to 2!106 steps ofMonte Carlo refinement with simulated annealing as arigid body, starting at 1000 K with cooling to 10 K.A second round of refinement was used to optimize the

local fit of each helix to the corresponding density. Eachhelix was treated as a separate rigid unit (52 units total,Figure 2), and refinement was used to improve the totalVLAT score. To maintain connectivity between helicesalong the RNA chain, harmonic bonds were included in

†Documentation available at http://rumour.biology.gatech.edu

the energy calculation as tethers between connectedhelices:

Ebi Z kbiðbiKbioÞ2

where Ebi denotes the energy of the ith bond; kbi is theforce constant for the ith bond; bi is the ith bond length;and bio is the corresponding equilibrium or ideal value.The ideal bond length is determined for each bond as thecrystallographic distance between consecutive nucleo-tides in two rigid units. The force constant was set at100 kcal/mol A2 for “normal” distances between con-secutive homologous nucleotides (usually 6–7 A). Insome cases gaps were present due to sequences thatcould not be modeled (grey regions, Figure 2). A smallerforce constant (1 kcal/mol A2) was used for these cases toallow more conformational freedom. A non-bond termwas also added to the energy calculation to preventinterpenetration of the helices:

Eij Z kijðrijKrijoÞ2 if rij%rijo

Eij Z 0 if rijOrijo

where Eij is the non-bond interaction energy betweenatoms i and j; kij is the non-bond force constant (100 kcal/mol A2) for the atom pair ij; rij is the distance betweenatoms i and j; and rijo is the minimum distance allowedbetween the two atoms. The goal of the non-bond term isto provide volume exclusion so that double helices do notinterpenetrate, and a value of rijoZ7.5 A was used toachieve this.This second round of refinement was performed using

multiple rigid-body Monte Carlo with simulated anneal-ing. The starting temperature was set at 100 K, with a finaltemperature of 10 K reached after 2!106 steps. The finalposition of each rigid unit was accepted or rejected basedon the energy score and by visual inspection of the fit withthe RNA-protein separated density using O.75 The rRNAstructures were fit to the complete subunit density, andthe regions corresponding to rRNA structures contained abetter score so the refined helices fit the RNA density.Refined movements were rejected only when a helixshifted (a few angstrom units) allowing one strand to sitin the middle of the helical density (because the best fitwould rarely place phosphate groups at the boundary ofthe density). The structure was still associated with thecorrect density, but the small shift during refinement wasnot accepted because the previous fit placed thephosphate groups at the edge of the helical density,which fit the density more accurately. Finally, the originalall-atom structures were superposed onto the refinedphosphate positions and covalently linked, except for theregions where gaps occur. A final round of energyminimization resolved small structural discrepanciescaused by structural rearrangements during the refine-ment protocol.Having placed the modeled rRNA structure in the

density from EM, the proteins were then placed using asimilar protocol, with the RNA structure held in a fixedposition. Each protein was treated as an independentrigid body (17 rigid bodies total, one RNA molecule and16 protein molecules). The overall placement of the L7/12dimer was based on the results of an earlier X-ray study,21

and fits were done both as a dimer and as two monomers,to determine the structural organization that would bestagree with the density and the conserved interactionsbetween the two proteins. A consensus structure wasdetermined from both methods. Bonds were included inthe calculations using the previously determined P–Ca

http://rumour.biology.gatech.edu

http://rumour.biology.gatech.edu


distances for conserved residues between the RNA andprotein as the ideal bond length (bio) and a bond forceconstant (kb) of 10 kcal/mol A2. The proteins were placedmanually in positions consistent with previous ribosomalcomplexes. Then 1!106 steps of rigid body Monte Carlowith simulated annealing were performed over a range ofstarting temperatures (1000 K, 100 K, and 10 K) becauseof the varied restraints associated with each protein.Those with more restraints could sample conformationsat higher temperatures, while those with fewer restraintsrequired lower temperatures to prevent extensivemotions. All simulations were annealed to a finaltemperature of 1 K. The final position for each proteinstructure was evaluated both visually using O,75 andquantitatively from the final energy score.

Protein Data Bank accession code

Themodel has been deposited to the RCSB Protein DataBank with accession code 2FTC.

Acknowledgements

We thank Linda Spremulli for providing thesamples of mammalian mitoribosomes and JamieCannone for help with generating the secondarystructure diagrams. This work was funded bygrants from the National Institutes of Health toS.C.H. (GM53827), R.R.G. (GM67317) and R.K.A.(GM61576), and from the Human Frontier ScienceProgram to R.K.A. (RGY232003).

References

1. Attardi, G. (1985). Animal mitochondrial DNA: anextreme example of genetic economy. Int. Rev. Cytol.93, 93–145.

2. Chomyn, A., Cleeter, M. W., Ragan, C. I., Riley, M.,Doolittle, R. F. & Attardi, G. (1986). URF6, lastunidentified reading frame of human mtDNA,codes for an NADH dehydrogenase subunit. Science,234, 614–618.

3. Brenner, C. & Kroemer, G. (2000). Apoptosis.Mitochondria—the death signal integrators. Science,289, 1150–1151.

4. Wallace, D. C. (1999). Mitochondrial diseases in manand mouse. Science, 283, 1482–1488.

5. Kurland, C. G. (1992). Evolution of mitochondrialgenomes and the genetic code. Bioessays, 14, 709–714.

6. O’Brien, T. W., Liu, J., Sylvester, J. E., Mougey, E. B.,Fischel-Ghodsian, N., Thiede, B. et al. (2000).Mammalian mitochondrial ribosomal proteins (4).Amino acid sequencing, characterization, andidentification of corresponding gene sequences.J. Biol. Chem. 275, 18153–18159.

7. Suzuki, T., Terasaki,M., Takemoto-Hori, C., Hanada, T.,Ueda, T., Wada, A. & Watanabe, K. (2001). Structuralcompensation for the deficit of rRNAwith proteins inthe mammalian mitochondrial ribosome. Systematicanalysis of protein components of the large ribosomalsubunit from mammalian mitochondria. J. Biol. Chem.276, 21724–21736.

8. Koc, E. C., Burkhart, W., Blackburn, K., Moyer, M. B.,Schlatzer, D. M., Moseley, A. & Spremulli, L. L. (2001).

The large subunit of the mammalian mitochondrialribosome. Analysis of the complement of ribosomalproteins present. J. Biol. Chem. 276, 43958–43969.

9. Kenmochi, N., Suzuki, T., Uechi, T., Magoori, M.,Kuniba, M., Higa, S. et al. (2001). The humanmitochondrial ribosomal protein genes: mapping of54 genes to the chromosomes and implications forhuman disorders. Genomics, 77, 65–70.

10. Yang, D., Oyaizu, Y., Oyaizu, H., Olsen, G. J. &Woese, C. R. (1985). Mitochondrial origins. Proc. NatlAcad. Sci. USA, 82, 4443–4447.

11. Gray, M. W., Burger, G. & Lang, B. F. (2001). The originand early evolution of mitochondria. Genome Biol. 2.REVIEWS1018.

12. Patel, V. B., Cunningham, C. C. & Hantgan, R. R.(2001). Physiochemical properties of rat liver mito-chondrial ribosomes. J. Biol. Chem. 276, 6739–6746.

13. Sharma, M. R., Koc, E. C., Datta, P. P., Booth, T. M.,Spremulli, L. L. & Agrawal, R. K. (2003). Structure ofthe mammalian mitochondrial ribosome reveals anexpanded functional role for its component proteins.Cell, 115, 97–108.

14. Koc, E. C., Burkhart, W., Blackburn, K., Moseley, A.,Koc, H. & Spremulli, L. L. (2000). A proteomicsapproach to the identification of mammalian mito-chondrial small subunit ribosomal proteins. J. Biol.Chem. 275, 32585–32591.

15. Suzuki, T., Terasaki,M., Takemoto-Hori, C., Hanada, T.,Ueda, T., Wada, A. & Watanabe, K. (2001). Proteomicanalysis of the mammalian mitochondrial ribosome.Identification of protein components in the 28 S smallsubunit. J. Biol. Chem. 276, 33181–33195.

16. O’Brien, T. W. (2002). Evolution of a protein-richmitochondrial ribosome: implications for humangenetic disease. Gene, 286, 73–79.

17. Youngman, E. M., Brunelle, J. L., Kochaniak, A. B. &Green, R. (2004). The active site of the ribosome iscomposed of two layers of conserved nucleotideswith distinct roles in peptide bond formation andpeptide release. Cell, 117, 589–599.

18. Sievers, A., Beringer, M., Rodnina, M. V. &Wolfenden,R. (2004). The ribosome as an entropy trap. Proc. NatlAcad. Sci. USA, 101, 7897–7901.

19. Weinger, J. S., Parnell, K. M., Dorner, S., Green, R. &Strobel, S. A. (2004). Substrate-assisted catalysis ofpeptide bond formation by the ribosome. NatureStruct. Mol. Biol. 11, 1101–1106.

20. Cannone, J. J., Subramanian, S., Schnare, M. N.,Collett, J. R., D’Souza, L. M., Du, Y. et al. (2002). TheComparative RNA Web (CRW) Site: an onlinedatabase of comparative sequence and structureinformation for ribosomal, intron, and other RNAs:correction. BMC Bioinformatics, 3, 15.

21. Yusupov, M. M., Yusupova, G. Z., Baucom, A.,Lieberman, K., Earnest, T. N., Cate, J. H. & Noller,H. F. (2001). Crystal structure of the ribosome at 5.5 Aresolution. Science, 292, 883–896.

22. Agrawal, R. K., Penczek, P., Grassucci, R. A. &Frank, J. (1998). Visualization of elongation factor Gon the Escherichia coli 70S ribosome: the mechanismof translocation. Proc. Natl Acad. Sci. USA, 95,6134–6138.

23. Stark, H., Rodnina, M. V., Rinke-Appel, J.,Brimacombe, R., Wintermeyer, W. & van Heel, M.(1997). Visualization of elongation factor Tu on theEscherichia coli ribosome. Nature, 389, 403–406.


24. Mears, J. A., Cannone, J. J., Stagg, S. M., Gutell, R. R.,Agrawal, R. K. & Harvey, S. C. (2002). Modeling aminimal ribosome based on comparative sequenceanalysis. J. Mol. Biol. 321, 215–234.

25. Harms, J., Schluenzen, F., Zarivach, R., Bashan, A.,Gat, S., Agmon, I. et al. (2001). High resolutionstructure of the large ribosomal subunit from amesophilic eubacterium. Cell, 107, 679–688.

26. Agrawal, R. K., Spahn, C. M., Penczek, P., Grassucci,R. A., Nierhaus, K. H. & Frank, J. (2000). Visualizationof tRNA movements on the Escherichia coli 70Sribosome during the elongation cycle. J. Cell Biol.150, 447–460.

27. Tama, F., Valle, M., Frank, J. & Brooks, C. L., III (2003).Dynamic reorganization of the functionally activeribosome explored by normal mode analysis andcryo-electron microscopy. Proc. Natl Acad. Sci. USA,100, 9319–9323.

28. Ban,N.,Nissen, P.,Hansen, J.,Moore, P. B.&Steitz, T.A.(2000). The complete atomic structure of the largeribosomal subunit at 2.4 A resolution. Science, 289,905–920.

29. Gutell, R. R., Lee, J. C. & Cannone, J. J. (2002). Theaccuracy of ribosomal RNA comparative structuremodels. Curr. Opin. Struct. Biol. 12, 301–310.

30. Tung, C. S., Joseph, S. & Sanbonmatsu, K. Y. (2002).All-atom homology model of the Escherichia coli 30Sribosomal subunit. Nature Struct. Biol. 9, 750–755.

31. Bourne, P. E., Addess, K. J., Bluhm, W. F., Chen, L.,Deshpande, N., Feng, Z. et al. (2004). The distributionand query systems of the RCSB Protein Data Bank.Nucl. Acids Res. 32, D223–D225.

32. Tan, R. K.-Z. & Harvey, S. C. (1993). Yammp:development of a molecular mechanics programusing the modular programming method. J. Comput.Chem. 14, 455–470.

33. Malhotra, A., Tan, R. K. & Harvey, S. C. (1994).Modeling large RNAs and ribonucleoprotein particlesusing molecular mechanics techniques. Biophys. J. 66,1777–1795.

34. Hofacker, I. L. & Vienna, R. N. A. (2003). Secondarystructure server. Nucl. Acids Res. 31, 3429–3431.

35. Gutell, R. R., Weiser, B., Woese, C. R. & Noller, H. F.(1985). Comparative anatomy of 16-S-like ribosomalRNA. Prog. Nucl. Acid Res. Mol. Biol. 32, 155–216.

36. Zuker, M. (2003). Mfold web server for nucleic acidfolding and hybridization prediction. Nucl. Acids Res.31, 3406–3415.

37. Suzuki, T., Terasaki,M., Takemoto-Hori, C., Hanada, T.,Ueda, T., Wada, A. & Watanabe, K. (2001). Structuralcompensation for the deficit of rRNAwith proteins inthe mammalian mitochondrial ribosome. J. Biol. Chem.276, 21724–21736.

38. Nikulin, A., Eliseikina, I., Tishchenko, S., Nevskaya, N.,Davydova, N., Platonova, O. et al. (2003). Structure ofthe L1 protuberance in the ribosome.Nature Struct. Biol.10, 104–108.

39. Ogle, J. M., Carter, A. P. & Ramakrishnan, V. (2003).Insights into the decoding mechanism from recentribosome structures. Trends Biochem. Sci. 28, 259–266.

40. Matthews, D. E., Hessler, R. A., Denslow, N. D.,Edwards, J. S. & O’Brien, T. W. (1982). Proteincomposition of the bovine mitochondrial ribosome.J. Biol. Chem. 257, 8788–8794.

41. Graack, H. R. & Wittmann-Liebold, B. (1998).Mitochondrial ribosomal proteins (MRPs) of yeast.Biochem. J. 329, 433–448.

42. Vila-Sanjurjo, A., Ridgeway, W. K., Seymaner, V.,Zhang, W., Santoso, S., Yu, K. & Cate, J. H. X-r. (2003).

ray crystal structures of the WT and a hyper-accurateribosome from Escherichia coli. Proc. Natl Acad. Sci.USA, 100, 8682–8687.

43. Agrawal, R. K., Heagle, A. B., Penczek, P., Grassucci,R. A. & Frank, J. (1999). EF-G-dependent GTPhydrolysis induces translocation accompanied bylarge conformational changes in the 70S ribosome.Natue Struct. Biol. 6, 643–647.

44. Agrawal, R. K., Lata, R. K. & Frank, J. (1999).Conformational variability in Escherichia coli 70Sribosome as revealed by 3D cryo-electron microscopy.Int. J. Biochem. Cell Biol. 31, 243–254.

45. Diaconu, M., Kothe, U., Schlunzen, F., Fischer, N.,Harms, J. M., Tonevitsky, A. G. et al. (2005). Structuralbasis for the function of the ribosomal L7/12 stalk infactor binding and GTPase activation. Cell, 121,991–1004.

46. Steinberg, S. & Cedergren, R. (1994). Structuralcompensation in atypical mitochondrial tRNAs.Nature Struct. Biol. 1, 507–510.

47. Hanada, T., Suzuki, T., Yokogawa, T., Takemoto-Hori,C., Sprinzl, M. & Watanabe, K. (2001). Translationability of mitochondrial tRNAsSer with unusualsecondary structures in an in vitro translation systemof bovine mitochondria. Genes Cells, 6, 1019–1030.

48. Steinberg, S., Leclerc, F. & Cedergren, R. (1997).Structural rules and conformational compensationsin the tRNA L-form. J. Mol. Biol. 266, 269–282.

49. Vacano, E. & Hagerman, P. J. (1997). Analysis ofbirefringence decay profiles for nucleic acid helicespossessing bends: the tau-ratio approach. Biophys. J.73, 306–317.

50. Andersen, G. R., Thirup, S., Spremulli, L. L. &Nyborg, J. (2000). High-resolution crystal structureof bovine mitochondrial EF-Tu in complex with GDP.J. Mol. Biol. 297, 421–436.

51. Song, H., Parsons, M. R., Rowsell, S., Leonard, G. &Phillips, S. E. (1999). Crystal structure of intactelongation factor EF-Tu from Escherichia coli in GDPconformation at 2.05 A resolution. J. Mol. Biol. 285,1245–1256.

52. Kjeldgaard, M., Nissen, P., Thirup, S. & Nyborg, J.(1993). The crystal structure of elongation factorEF-Tu from Thermus aquaticus in the GTP confor-mation. Structure, 1, 35–50.

53. Stuart, R. (2002). Insertion of proteins into the innermembrane of mitochondria: the role of the Oxa1complex. Biochim. Biophys. Acta, 1592, 79–87.

54. Gutell, R. R., Cannone, J. J., Shang, Z., Du, Y. &Serra, M. J. (2000). A story: unpaired adenosine basesin ribosomal RNAs. J. Mol. Biol. 304, 335–354.

55. Woese, C. R., Winker, S. & Gutell, R. R. (1990).Architecture of ribosomal RNA: constraints on thesequence of “tetra- loops”. Proc. Natl Acad. Sci. USA,87, 8467–8471.

56. Nissen, P., Ippolito, J. A., Ban, N., Moore, P. B. &Steitz, T. A. (2001). RNA tertiary interactions in thelarge ribosomal subunit: the A-minor motif. Proc. NatlAcad. Sci. USA, 98, 4899–4903.

57. Lee, J. C., Cannone, J. J. & Gutell, R. R. (2003). Thelonepair triloop: a newmotif in RNA structure. J. Mol.Biol. 325, 65–83.

58. Lee, J. C. & Gutell, R. R. (2004). Diversity of base-pairconformations and their occurrence in rRNA struc-ture and RNA structural motifs. J. Mol. Biol. 344,1225–1249.

59. Cate, J. H., Gooding, A. R., Podell, E., Zhou, K.,Golden, B. L., Szewczak, A. A. et al. (1996). RNA


tertiary structure mediation by adenosine platforms.Science, 273, 1696–1699.

60. Battle, D. J. & Doudna, J. A. (2002). From the cover:specificity of RNA-RNA helix recognition. Proc. NatlAcad. Sci. USA, 99, 11676–11681.

61. Wagenknecht, T., Grassucci, R. & Frank, J. (1988).Electron microscopy and computer image averagingof ice-embedded large ribosomal subunits fromEscherichia coli. J. Mol. Biol. 199, 137–147.

62. Penczek, P. A., Grassucci, R. A. & Frank, J. (1994). Theribosome at improved resolution: new techniques formerging and orientation refinement in 3D cryo-electron microscopy of biological particles. Ultrami-croscopy, 53, 251–270.

63. Frank, J., Penczek, P., Agrawal, R. K., Grassucci, R. A.& Heagle, A. B. (2000). Three-dimensional cryoelec-tron microscopy of ribosomes. Methods Enzymol. 317,276–291.

64. Gabashvili, I. S., Agrawal, R. K., Spahn, C. M.,Grassucci, R. A., Svergun, D. I., Frank, J. & Penczek,P. (2000). Solution structure of the E. coli 70S ribosomeat 11.5 A resolution. Cell, 100, 537–549.

65. Orlova, E. V., Dube, P., Harris, J. R., Beckman, E.,Zemlin, F., Markl, J. & van Heel, M. (1997). Structureof keyhole limpet hemocyanin type 1 (KLH1) at 15 Aresolution by electron cryomicroscopy and angularreconstitution. J. Mol. Biol. 271, 417–437.

66. Spahn, C. M., Penczek, P. A., Leith, A. & Frank, J.(2000). A method for differentiating proteins fromnucleic acids in intermediate-resolution densitymaps: cryo-electron microscopy defines the quatern-ary structure of the Escherichia coli 70S ribosome.Structure, 8, 937–948.

67. Larsen,N.,Olsen,G. J.,Maidak, B. L.,McCaughey,M. J.,Overbeek, R., Macke, T. J. et al. (1993). The ribosomaldatabase project. Nucl. Acids Res. 21, 3021–3023.

68. Hermann, T. & Patel, D. J. (1999). Stitching togetherRNA tertiary architectures. J. Mol. Biol. 294, 829–849.

69. Gutell, R. R., Cannone, J. J., Konings, D. & Gautheret,D. (2000). Predicting U-turns in ribosomal RNAwithcomparative sequence analysis. J. Mol. Biol. 300,791–803.

70. Wheeler, D. L., Church, D. M., Edgar, R., Federhen, S.,Helmberg, W., Madden, T. L. et al. (2004). Databaseresources of the National Center for BiotechnologyInformation: update. Nucl. Acids Res. 32, D35–D40.

71. Thompson, J. D., Higgins, D. G. & Gibson, T. J. (1994).CLUSTALW: improving the sensitivity of progressivemultiple sequence alignment through sequenceweighting, position-specific gap penalties and weightmatrix choice. Nucl. Acids Res. 22, 4673–4680.

72. Altschul, S. F., Gish, W., Miller, W., Myers, E. W. &Lipman, D. J. (1990). Basic local alignment search tool.J. Mol. Biol. 215, 403–410.

73. Sali, A. & Blundell, T. L. (1993). Comparative proteinmodelling by satisfaction of spatial restraints. J. Mol.Biol. 234, 779–815.

74. Conn, G. L., Draper, D. E., Lattman, E. E. & Gittis, A. G.(1999). Crystal structure of a conserved ribosomalprotein-RNA complex. Science, 284, 1171–1174.

75. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard, M.(1991). Improved methods for building protein modelsin electron density maps and the location of errors inthese models. Acta Crystallog. sect. A, 47, 110–119.

76. Kim, S.H., Quigley, G. J., Suddath, F. L.,McPherson, A.,Sneden, D., Kim, J. J. et al. (1973). Three-dimensionalstructure of yeast phenylalanine transfer RNA: foldingof the polynucleotide chain. Science, 179, 285–288.

Edited by D. E. Draper

(Received 19 October 2005; received in revised form 25 January 2006; accepted 27 January 2006)Available online 10 February 2006

Note added in proof: Thearrangement of theproteins in theL7/L12 stalk inFigure 6 is unlikely tobe correct, sinceit has recently been shown that the stalk in bacterial ribosomes is composed of L10, decorated with multiplecopies of L7/L12 (M. Diaconu et al. (2005) Cell 121, 991–2004; P.P. Datta et al. (2005)Mol. Cell 20, 723–731). Ourmodel was based on the earlier crystal structure of the T. thermophilus ribosome by Yusupov et al.

gutell 096.jmb.2006.358.0193

Technology

smaller mitochondrial

mitochondrial ribosomes

mitochondrial rrna sequenceand

mitochondrial defects

mitochondrial rrnas

s mitoribosome

rrna content

animal mitochondrial