Structural compensation for the deficit of rRNA with proteins in the mammalian mitochondrial ribosome. Systematic analysis of protein components of the large ribosomal subunit from mammalian mitochondria.

The mammalian mitochondrial ribosome (mitoribosome) is a highly protein-rich particle in which almost half of the rRNA contained in the bacterial ribosome is replaced with proteins. It is known that mitochondrial translation factors can function on both mitochondrial and Escherichia coli ribosomes, indicating that protein components in the mitoribosome compensate the reduced rRNA chain to make a bacteria-type ribosome. To elucidate the molecular basis of this compensation, we analyzed bovine mitoribosomal large subunit proteins; 31 proteins were identified including 15 newly identified proteins with their cDNA sequences from human and mouse. The results showed that the proteins with binding sites on rRNA shortened or lost in the mitoribosome were enlarged when compared with the E. coli counterparts; this suggests the structural compensation of the rRNA deficit by the enlarged proteins in the mitoribosome.

Recent advances in structural biology has enabled the observation of the detailed structure of bacterial ribosome at atomic resolution. Electron microscopy has produced high resolution images of Escherichia coli ribosome (1,2). Topological orientation of RNA helices and protein components has been determined with the help of much biochemical data (3). In past few years, crystallographers have reported primary images of crystal structures for 30 S, 50 S, and 70 S ribosome subunits (4 -7), where some proteins and rRNA helices were located. Most recently, the crystallographic resolution was improved greatly, allowing the identification of the complete structures of 50 S (8,9) and 30 S (10, 11) at 2.4 Å and 3.0 -3.3 Å resolutions, respectively. Because almost all nucleotide residues in rRNA and many protein components were unambiguously assigned in the electron density maps, we have a better understanding of how rRNA and proteins assemble to make a functional ribosome. The most notable finding was that the functional regions for peptide bond formation in the large subunit (9,12) and the decoding center in the small subunit (13) consist entirely of rRNA. The ribosome is thus a ribozyme (9,12).
As compared with the bacterial ribosome, the mammalian mitochondrial ribosome (mitoribosome) provides another intriguing model for elucidating the molecular basis of ribosome function. The mitoribosome has a smaller sedimentation coefficient of 55 S (14) consisting of large 39 S subunit and small 28 S subunits; the former contains 16 S, and the latter contains 12 S rRNAs as RNA components without bacterial 5 S rRNA counterpart. Total length of the mitochondrial rRNA is about half that of the bacterial ribosome. Recently, the physiochemical properties of rat mitoribosome were precisely determined (15). Rat mitoribosome has a large molecular mass (3.57 MDa) compared with that of E. coli ribosome (2.49 MDa), which is explained by the fact that the protein to RNA ratio is completely reversed (16). This suggests that large parts of the RNA domains are replaced by protein components during the process of mitochondrial evolution from an eubacteria-like endosymbiont in a progenitor of eukaryotic cells.
Functional equivalency of the mammalian mitoribosome to E. coli ribosome has been suggested by the fact that mitochondrial translation factors are able to function efficiently on E. coli ribosomes. Mammalian mitochondrial elongation factors (EF-Tu mt 1 and EF-G mt) can be replaced by bacterial factors and maintain their efficient activities on the E. coli ribosome (17,18), and even bacterial and chloroplast EF-Tus can cooperate with the bovine mitoribosome and E. coli tRNAs to some extent (17). Furthermore, initiation factor 2 from bovine mitochondria also functions on the bacterial ribosome (19,20). The observations indicate that the mitochondrial factors are exchangeable with bacterial factors on bacterial ribosome.
Systematic analysis of protein components in the mitoribosome is necessary to elucidate the molecular basis of exchangeability of these factors. O'Brien and co-workers (21) were the first to isolate proteins from bovine mitoribosome by two-dimensional PAGE. They identified 52 protein spots in the large subunit and 33 protein spots in the small subunit (21). The increased number of proteins in the mitoribosome as compared with the E. coli ribosome suggested a structural and functional compensation of the shortened rRNA in the mitoribosome by the protein moiety. Peptide analysis and cDNA cloning of each protein component is required to characterize the mitoribosome in more detail. Until 1998, only a few protein sequences had been reported (22,23). Based on the human genome data base and EST sequences, more than 20 mitoribosomal proteins have been identified from rat and bovine mitoribosomes using Nterminal sequencing and data base screening (24 -29).
In this study, we systematically analyzed the large subunit proteins of bovine mitoribosome by peptide analysis using mass spectrometry and N-terminal sequencing. We identified 31 mitoribosomal proteins including 15 new members, with their complete human and mouse cDNA sequences. A striking result was that the mitoribosomal proteins whose binding sites on rRNA are shortened or lost carry an N-or C-terminal extension, whereas the proteins with conserved rRNA binding sites have similar molecular weights when compared with those of the E. coli counterparts. Three-dimensional orientations of the enlarged proteins and the concomitantly lost or shortened rRNA domains were mapped on the crystal structure of the 50 S subunit to investigate how proteins cover the shortened RNA domains to construct the functional mitoribosome. The enlarged proteins together with several extra proteins that are specifically found in the mitoribosome obviously provide both structural and functional compensation for the deficit of rRNA in mitoribosome, so that the whole molecular architecture of the ribosome is conserved in bacteria and mitochondria. The mitoribosome thus serves as a good model system for defining the functional domains of rRNA in more detail and also for an experimental verification of the concept of transition from the "RNA world" to the "RNP world" in the early process of evolution of life.

EXPERIMENTAL PROCEDURES
Purification of Mitoribosome from Bovine Liver-Mitochondria were prepared from a fresh bovine liver according to the literature (46). Crude mitoribosomes were prepared as reported (20) and stored in the 55 S buffer: 20 mM Tris-HCl (pH 7.6), 20 mM MgCl 2 , 80 mM KCl, and 6 mM 2-mercaptoethanol. The 55 S mitoribosome was purified by the sucrose density gradient centrifugation in the 55 S buffer containing sucrose. The concentrations of sucrose were 6% for the top layer and 38% for the bottom layer. Before centrifugation, a linear sucrose gradient was formed in the centrifuge tubes (Seton) by Gradient Mate model 117 (BioComp) according to the user's manual. Eighty A 260 units of mitoribosome, dissolved in 0.6 ml of the 55 S buffer without sucrose, were layered onto the gradient. Centrifugation was run at 20,000 rpm for 18 h using the SW28 rotor (Beckman). The gradient was fractionated from the top to the bottom using the Piston Gradient Fractionater (BioComp). Fractions containing 55 S mitoribosomes were pooled and collected by the ultracentrifugation at 40,000 rpm for 24 h using the 70 Ti rotor (Beckman). The purified 55 S mitoribosome was dissociated into 39 S and 28 S subunits in a buffer containing 20 mM Tris-HCl (pH 7.6), 2 mM MgCl 2 , 200 mM KCl, and 6 mM 2-mercaptoethanol. Each subunit was purified by sucrose density gradient centrifugation at 20,000 rpm for 16 h using the SW28 rotor. The purified 55 S, 39 S, and 28 S ribosomal subunits were dissolved in the 55 S buffer and stored at Ϫ70°C.
Separation and Isolation of Mitoribosomal Proteins by Radical Free High Reducing (RFHR) Two-dimensional PAGE-Total ribosomal proteins were extracted from the purified 55 S mitoribosome by stirring for 1 h at 4°C in 600 l of the solution containing 50 mM magnesium acetate, 50%(v/v) acetic acid; 13.3 A 260 units (425 pmol) of mitoribosome were used. After removing the insoluble fraction by centrifugation, total soluble proteins were precipitated in 5 times the volume of acetone at Ϫ80°C for 1 h. The total proteins were centrifuged, and the protein pellet was dried. The proteins were then denatured and reduced at 40°C for 30 min in 60 l of the reaction mixture consisting of 8 M urea, 0.2 M 2-mercaptoethanol, 0.074% acetic acid, 0.012 N KOH, 0.015% acridine orange, and 0.015% pylonine G. The resultant sample was subjected to RFHR two-dimensional PAGE analysis according to the literature (30,47).
In-gel Digestion of Mitoribosomal Proteins with Trypsin-The protein spots on RFHR two-dimensional gel were visualized by Coomassie Brilliant Blue staining, excised, and completely dried in vacuo. The dried gel pieces were then rehydrated with 5-10 l of the trypsin digest solution (0.2 M NH 4 HCO 3 , 15 ng/ml L-1-tosylamido-2-phenylethyl chloromethyl ketone-trypsin (Pierce)). After the gel pieces absorbed the liquid, 10 -20 l of 0.2 M NH 4 HCO 3 was added to the gel. Complete digestion was carried out overnight at 30°C. The digested peptides were extracted from the gel by shaking in 200 l of 60% acetonitrile and 0.1% TFA solution for 20 min. This step was repeated two more times. The collected fractions containing peptides were batched together and dried in a speedvac, and the obtained sample was dissolved in 80 l of 0.1% formic acid.
For the proteins separated by SDS-PAGE, the gel piece was soaked in a buffer containing 0.2 M NH 4 HCO 3 with 50% acetonitrile and incubated at 30°C for 30 min to remove SDS from the proteins. To assure the complete removal of SDS, the step was repeated twice. Subsequent treatment of the sample was same as described above.
Mass Spectrometry and Protein Identification-A Finnigan LCQ ion trap mass spectrometer (ThermoQuest) equipped with an electrospray ionization source was used for the peptide analysis of mitoribosomal proteins. The LC/MS analysis was performed using an ODS reversephase column (Monitor C18, 0.1 ϫ 15 cm; Michrom BioResource) connected on-line to the electrospray interface. A solvent system consisting of 0.1% formic acid in H 2 O (A) and in acetonitrile (B) was developed from 0% to 70% B in 35 min at a flow rate of 50 l/min by using the Magic 2002 high performance liquid chromatography system (Michrom BioResource). The flow rate of sheath gas and the capillary temperature were kept at 55 arbitrary units and 235°C, respectively. The zoom scan analysis and MS/MS experiment by collision-induced dissociation using the data dependent scan (triple play) were performed in the range of 300 -2000 m/z throughout the separation. An uninterpreted data set of the peptide product ion for each protein generated from the triple play analysis was searched against the nonredundant protein data base and human and mouse EST data bases using the SEQUEST search program (32,48).

Determination of N-terminal Sequence-
The excised gel pieces were dried in vacuo and soaked in a buffer consisting of 10 mM CAPS-NaOH (pH 11) and 10% methanol for 1 h. The each protein in a gel was blotted onto polyvinylidene difluoride membrane (Fluorotrans, Paul) by electroelution using the Electroelutor (Nihon Eido, Japan). Amino acid sequencing was performed using a gas-phase protein sequencer (PPSQ-21, Shimadzu).

Separation of Bovine Mitoribosomal Proteins from 55 S Mitoribosome and 39 S Large Subunit-
To understand the compensation mechanism of the RNA deficit by proteins to maintain ribosomal function, we first analyzed the protein content in the large subunit of the mitoribosome. Since the crude ribosome fraction obtained by ultracentrifugation of the mitoplast extract contained various proteins for metabolisms and amino acid biosynthesis (data not shown), 55 S mitoribosomes were purified by sucrose density gradient centrifugation. Furthermore, the large 39 S subunit was isolated from the purified 55 S mitoribosome by sucrose density gradient centrifugation under a low Mg 2ϩ concentration. The proteins of the purified 55 S mitoribosome were isolated by RFHR two-dimensional PAGE (30, 31), which has a wide separation range in the basic region and, thus, applies well for ribosomal proteins. RFHR two-dimensional PAGE requires no SDS, thereby simplifying the peptide analysis of gel-purified protein by LC/MS spectrometry. More than 70 protein spots were identified in the gel (Fig.  1A). Each protein component of the 39 S subunit was analyzed by one-dimensional SDS-PAGE, as shown in Fig. 1B.
Peptide Analysis of the Large Subunit Proteins and Identification of Their cDNA Sequences-Each protein spot from the 55 S mitoribosome and 39 S large subunit was treated with trypsin in the gel. The tryptic peptides were subjected to mass spectrometric analysis by LC/MS/MS as described under "Experimental Procedures." A total of 27 proteins were identified as ribosomal proteins from the large 39 S subunit (Fig. 1B).
An example (L11mt) for the investigation of the LC/MS/MS analysis is shown in Fig. 2A. The triple play analysis by a data-dependent scan was performed for each bovine protein to obtain the information on the molecular mass (upper panel), charge state (middle panel), and fragment patterns for each peptide (lower panel) within one analysis (see "Experimental Procedures"). The data set of the LC/MS/MS analysis was matched against the nonredundant protein data base of human (and mouse) EST data base by the algorithm SEQUEST (32), and the partial cDNA sequences encoding the analyzed proteins were obtained. High homology in the amino acid sequences of mammalian mitoribosomal proteins led to many primary hits in the human (and mouse) EST data base. A series of related EST sequences were retrieved by BLASTN search (33), with the hit sequence as a query and assembled in silico to make the longest possible cDNA sequences. Sequencing errors were identified and corrected by comparing overlapping EST clones. Since ribosomal proteins are relatively small in general, highly reliable cDNA sequences containing complete open reading frames were obtained. Taking the analysis of L11 homologue (bMRP-32) as an example, two EST sequences from human (AI188527) and mouse (AA272356) were obtained from an initial SEQUEST search. The possible longest cDNA sequence was obtained by assembling numerous related EST sequences. A complete open reading frames of L11 homologue was encoded in the sequence as shown in Fig. 2B. Many peptide ions derived from this protein sequence were identified in the mass spectra and assigned in the mass chromatogram ( Fig. 2A,  left upper panel). Furthermore, a partial or complete sequence of each peptide was assigned in the collision-induced dissociation spectrum, as shown in Fig. 2A (right lower panel). For four small proteins, namely bMRP-59b, -66, -68, and -69, N-terminal sequences were directly determined by the peptide sequencing and the EST data base analysis using the TBLASTN program (33). The human EST clones with significant homology were successfully retrieved for each protein, as shown in Table I.
The protein sequences for L3, L23, and L33 homologues were previously reported not as mitochondrial ribosomal proteins, although their sequence similarity with prokaryotic counterparts had been implied (34 -36). Three protein spots were identified as the above three proteins. The specific EST clones (listed in Table I) from human or mouse were initially retrieved for each protein by SEQUEST or TBLASTN search. With 16  Haemophilus influenzae (Hinfl). Multiple alignment of each sequence has been carried out by CLUSTAL W (50) and displayed by Genedoc multiple sequence alignment editor (51). The homology values between human mitoribosomal protein sequence and others shown at the side of the alignment were calculated by using Genedoc. Protein sequence of bovine B8 subunit of complex I (CI-B8) was also aligned with bMRP-36a homologues. The colored boxes indicated a degree of the sequence similarity. The N-terminal sequence of each bovine mitoribosomal protein was obtained by peptide sequencing. Red letters represent N-terminal sequences with the indicated cleavage site for the importation signal peptide. egans, and yeast mitochondria (Table I). Sequence alignment of each subunit with respective homologues from other species is shown in Fig. 3.
In addition to 27 species that we identified from the large 39 S subunit, four protein homologues of L3, L33, L34, and L36 were found in the 55 S mitoribosome in our two-dimensional PAGE analysis. The small components, such as L33mt, L34mt and L36mt, could not be detected in the SDS-PAGE analysis of the 39 S subunit for technical reasons. But even the relatively large L3mt could not be detected on either the 39 S nor the 28 S subunits for unknown reasons. We suggest that these additional four protein homologues, which have significant sequence similarity with the bacterial counterparts, belong to the large subunit.
A total of 31 proteins was identified in the large 39 S subunit, including 15 new members (Table I, Fig. 3). Fourteen of 15 newly identified protein sequences were recognized as prokary-otic homologues of ribosomal proteins given their sequence similarity. The remaining one sequence, bMRP-36a, had no homology to any of ribosomal proteins but had significant similarity to B8 subunit of NADH-ubiquinone oxidoreductase (complex I) (Fig. 3).
To determine the cleavage site of the signal sequence for the mitochondrial import, N-terminal sequencing was performed. The N-terminal sequences for nine proteins were determined, whereas six proteins may have blocked termini, since we were unable to determine the sequence (Fig. 3). The molecular weight of each protein was calculated by its amino acid sequence of the matured form (listed in Table I (Table I). The amino acid sequence homology between the human mitoribosomal protein and its counterpart from other species was shown in Fig. 3. The significant homology of mitoribosomal proteins throughout the sequence indicates similar topological orientations of the proteins in the mitoribosome to those of the prokaryotic ribosome. The partial conservation of the specific interactions between the rRNA and protein components in the mitoribosome is suggested. High homology was seen among animal and insect mitochondrial proteins, whereas yeast mitoribosomal proteins have lower similarity to human mitochondrial counterparts, which is explained by the fact that no reduction in length of rRNA was observed in the yeast mitoribosome. As shown in the sequence alignment in Fig. 3 Although all proteins in the large subunit are yet to be identified, the average molecular mass of ribosomal proteins, both from E. coli and human mitochondria, was calculated using the identified 24 proteins that belong to the family of prokaryotic homologs. The calculated average molecular mass for these ribosomal proteins were 21.1 kDa for mitochondrial proteins and 13.8 kDa for E. coli proteins. Mitochondrial ribosomal proteins thus show a 1.5-fold increase in the average of their molecular masses when compared with those of the E. coli counterparts. In addition to these enlarged protein homologues, new proteins specific for the mitoribosome participate in the compensation of the decrease of the rRNA moiety in the mitoribosome. Ribosomal proteins L1 and L9 form the L1 ridge together with helices 76 -79 in domain V of 23 S rRNA (3). The extended C terminus of human mitochondrial L1mt is possibly replacing helices 77 and 78 that are lacking in the mitoribosome. The molecular mass of L1mt extends for about 1.5-fold the E. coli counterpart (Table I). Interestingly, in the C. elegans mitoribosome, which lacks helix 76 in addition to helices 77 and 78, the C terminus of the C. elegans L1 homologue is even longer than the C terminus of human mitochondrial L1mt (Fig. 3). In addition, both extended termini of L9mt seem to compensate the collar region of L1 ridge. Several ribosomal protein homologues that are indispensable for peptidyltransferase activity (37)(38)(39) have been identified in this study. L3mt shows a significant homology to its E. coli counterpart over its entire sequence but has elongated N and C termini that seem to compensate for the missing binding sites in domain VI (see below). In the case of the C. elegans mitoribosome having short domain VI rRNA, the L3 homologue contained two characteristic insertions and the longest C terminus (Fig. 3). L4mt shows one of the most prom-inent enlargement. L16mt also has an extended C terminus with a relatively low homology to the E. coli counterpart. L11 is possibly involved in GTP hydrolysis by translation factors in collaboration with L8 complex (L10⅐ (L7/12) 4 ) and the GTPassociated region (helices [42][43][44]. The high homology of L11mt to the counterparts from other E. coli organisms can be explained by the fact that the binding site for L11 on helices 43 and 44 is well conserved in all mitochondrial rRNA. Since the C. elegans mitoribosome contains shorter rRNA (952 bases) in the large subunit than the corresponding rRNA (1558 bases) of human mitoribosome, the progressive compensation of lost rRNA with enlarged protein in C. elegans is a tempting speculation. We detected four protein homologues in C. elegans, L1, L3, L13 and L20, having the longest C terminus among the aligned species (Fig. 3). The findings indicate a correlation between the shortening of rRNA and the enlargement of proteins in the animal mitoribosome.
Structural Compensation for Deficit of Mitoribosomal RNA with Enlarged Protein Components-To visualize the global image of the mammalian mitoribosome composed of shortened rRNAs and enlarged proteins, mitoribosomal proteins were mapped on the secondary structure of rRNA (Fig. 4). The secondary structure of human mitochondrial rRNA (red line) of the large subunit was superimposed on the corresponding E. coli 23 S rRNA (black line) (40). Mitoribosomal proteins with well conserved binding sites on mt rRNA have similar molecular masses as those of the prokaryotic counterparts. In contrast, the proteins having small (reduced) or no distinct binding sites on mt rRNA showed increased molecular masses due to Nor C-terminal extensions. According to the crystal structure of the large subunit, the helices 66, 61, 53, 43/44, and 95 in rRNA are the main binding sites for L2, L22, L23, L11, and L14, respectively. These helices are well conserved in mt rRNA, and the mitochondrial protein homologues have similar molecular  (40), which is described according to the format of Ban et al. (8). The 5Ј region of mitochondrial 16 S rRNA (about 160 bases) could not be aligned with domain I of bacterial 23 S rRNA. Ovals that represent large ribosomal proteins are mapped on the secondary structure of rRNA, with interactions indicated by gray arrows. Solid arrows show interaction maps that were identified by the crystal structure of the 50 S subunit (8). Broken arrows indicate the interaction maps obtained from biochemical studies (52,53). The oval size for each protein represents the relative molecular weight. The degree of protein size enlargement as compared with the E. coli counterpart (Table I) was indicated with colors: red, more than 15 kDa; orange, 10 -15 kDa; yellow, 5-10 kDa; green, less than 5 kDa. masses with the E. coli counterparts (Fig. 4, Table I). L24, L4, and L15 have many contacts with domains I and II in Haloarcula marismortui 50 S subunits; their binding sites are almost lost in the mt rRNA. The molecular masses of L24mt, L4mt, and L15mt, however, are increased by more than 15 kDa when compared with the bacterial counterparts. Similar compensation was also observed in the L19mt homologue, where most of the binding sites were lost in mt rRNA. The lack of the helices 77 and 78 in mt rRNA may be substituted for by the extended termini of L1mt and L9mt. The L3 binding sites in domain VI (helices 94 and 100) are also missing in mt rRNA; L3mt shows long N-and C-terminal extensions. Taken together, these observations strongly suggest that enlarged mitoribosomal proteins may have compensated for the sequence losses of the rRNAs in the mitoribosome. As already mentioned, we also identified a new protein component, bMRP-36a, which has a high homology with B8 subunit of NADH-ubiquinone oxidoreductase (complex I). The rRNA binding site as well as the structure and function of this new ribosomal protein is unknown. DISCUSSION In 1970, Margulis proposed the symbiosis theory stating that mitochondria have evolved from eubacteria-like endosymbionts (41); Rickettsia-like alpha proteobacteria (42) are considered to be the closest relative of the endosymbionts today. The most primitive mitochondrial DNA so far was found in a protozoan Reclinomonas americana (43), in which the largest number of 97 genes coding for proteins and RNAs were identified in the total 64 kilobase pair of mtDNA. Three ribosomal RNA genes including 5 S rRNA have similar lengths with those of eubacterial genes. In addition to the genes related to the respiration, as many as 27 ribosomal protein genes were encoded in the mtDNA. Furthermore, the gene organization and the way of gene expression are also eubacterial-like. Thus that R. americana mtDNA seems to closely resemble the ancestral mitochondrial genome. In the mammalian mitochondrial translation system, the mitoribosome has been categorized as a bacteriatype ribosome with regard to antibiotic susceptibility and sequence similarity of ribosomal proteins, translation factors, and aminoacyl-tRNA synthetases. The bacterial ribosome, therefore, might well be considered as a direct ancestor for mammalian mitoribosome.
Mammalian mitochondria contain small mtDNA (16 kilobase pairs) only capable of encoding 13 proteins, 22 tRNAs, and 2 shortened rRNAs, suggesting that mitochondria would have managed to transfer a large portion of mtDNA to the nuclear genome during the process of mitochondrial evolution (42). Although the molecular mechanism for this gene transfer is unclear, the repairing process of double-strand breaks in yeast chromosomes by mtDNA may be crucial evidence for the gene transfer (44). The mammalian mitoribosome is composed of 2 rRNAs derived from mtDNA and more than 70 different mitochondrial proteins that are all encoded in the nuclear genome. Furthermore, the 5 S rRNA gene and about half-lengths of both large and small rRNA genes were removed from the mtDNA.
The result of our analysis shows that the bacteria-type ribosome was build by covering the gaunt backbone of the shortened rRNAs in the mitoribosome by enlarged ribosomal proteins of prokaryotic homologues and/or new protein families encoded in the nuclear genes. Thus, the structural and functional compensation of the lost rRNA with protein would have resulted by adaptiogenesis.
Because we have identified 14 new proteins in the large subunit of mitoribosome that belong to the family of prokaryotic ribosomal protein in this study, at least 24 homologues of prokaryotic ribosomal proteins of the large subunit are so far reported at this moment. Therefore, ribosomal proteins might be important to maintain ribosomal functions. A molecular mechanism for rRNA-catalyzed peptidyl transfer has been derived (9,12), and it is also known that several protein components are indispensable for the catalytic activity (37,45). Mitoribosomal proteins may serve as a good model to investigate the function of ribosomal proteins in peptide synthesis.
We constructed three-dimensional models for mitochondrial rRNAs from human, C. elegans, and C. fasciculata (Fig. 5) based on the 50 S crystal structure of H. marismortui (8). The shortened mt rRNA forms a spherical cluster at the center of the peptidyltransferase, whereas several lost rRNA portions, which are localized discontinuously in the secondary structure (Fig. 4), form large missing domains at the bottom, the back, the central protuberance, and the left side of the crown view. Recently, Brimacombe and co-workers (3) report a similar model of human mitochondrial rRNA based on their own 23 S rRNA structure. The topological orientation of rRNA helices in their model is in good agreement with the present model based on the 50 S crystal structure. In the case of mammalian mitoribosome (Fig. 4) (8)). The model structures for three mitoribosomes were constructed by removal of nonconserved rRNA portions from the atom coordinates of 1FFK, based on the secondary structure of large mitoribosomal RNA from mammalian (B), C. elegans (C), and Crithidia fasciculata (D) (40). The three-dimensional structures were displayed by Rasmol Version 2.6 (54). The outline shows an edge line of the crystal structure of the 50 S subunit from the crown view. Some functional rRNA domains were colored: red, P loop; blue, A loop; green, S/R loop; light blue, and L2 binding helix (H66). Topological orientation of the ribosomal protein is shown on the model for the mammalian mitoribosome (B). mains are composed mostly of domain I, H25/28 -31/41/42/ 45/46 (domain II), H52 (domain III), H66/68 (domain IV), H77-79/88 (domain V), and H97 (domain VI). A number of ribosomal proteins are localized in rRNA portions that are not present in the mitoribosome. rRNA portions of ribosomal protein binding sites that seemed to be shortened are replaced by certain lengthened ribosomal proteins; ribosomal proteins fill up the empty space by extending their N or C terminus as observed in this study. The binding sites for L2, L22, L23, L11, and L14, whose mitochondrial homologues have similar molecular masses (Fig. 4, Table I), are also conserved in the three-dimensional model (Fig. 5). Typical enlarged protein homologues for L4, L15, L24, and L19 are localized in the region where rRNA is missing (Fig. 5). The elongated peptide sequences of these proteins may contribute in substituting the functional deficit of the lost rRNA. Further study is necessary to clarify in detail how proteins replace the missing rRNA both structurally and functionally.