|Home | About | Journals | Submit | Contact Us | Français|
5S rRNAs are ubiquitous components of prokaryotic, chloroplast, and eukaryotic cytosolic ribosomes but are apparently absent from mitochondrial ribosomes (mitoribosomes) of many eukaryotic groups including animals and fungi. Nevertheless, a clearly identifiable, mitochondrion-encoded 5S rRNA is present in Acanthamoeba castellanii, a member of Amoebozoa. During a search for additional mitochondrial 5S rRNAs, we detected small abundant RNAs in other members of Amoebozoa, namely, in the lobose amoeba Hartmannella vermiformis and in the myxomycete slime mold Physarum polycephalum. These RNAs are encoded by mitochondrial DNA (mtDNA), cosediment with mitoribosomes in glycerol gradients, and can be folded into a secondary structure similar to that of bona fide 5S rRNAs. Further, in the mtDNA of another slime mold, Didymium nigripes, we identified a region that in sequence, potential secondary structure, and genomic location is similar to the corresponding region encoding the Physarum small RNA. A mtDNA-encoded small RNA previously identified in Dictyostelium discoideum is here shown to share several characteristics with known 5S rRNAs. Again, we detected genes encoding potential homologs of this RNA in the mtDNA of three other species of the genus Dictyostelium as well as in a related genus, Polysphondylium. Taken together, our results indicate a widespread occurrence of small, abundant, mtDNA-encoded RNAs with 5S rRNA-like structures that are associated with the mitoribosome in various amoebozoan taxa. Our working hypothesis is that these novel small abundant RNAs represent radically divergent mitochondrial 5S rRNA homologs. We posit that currently unrecognized 5S-like RNAs may exist in other mitochondrial systems in which a conventional 5S rRNA cannot be identified.
5S rRNA is a ubiquitous component of prokaryotic (bacterial and archaeal) ribosomes as well as ribosomes of the eukaryotic cytosol and chloroplasts. In all of these cases, 5S rRNA is strikingly conservative in size and higher-order structure, containing diagnostic, strongly conserved sequence motifs (21). This high degree of conservation argues that 5S rRNA plays an important role in ribosome function. What this role might be has been difficult to discern (30) although the biological activity of reconstituted Bacillus stearothermophilus ribosomes has been shown to be critically dependent on the presence of 5S rRNA (5), indicating an essential function for 5S rRNA in translation.
In marked contrast, the mitochondrial ribosomes (mitoribosomes) of many eukaryotic taxa appear to lack 5S rRNAs (reviewed in reference 3). A clearly recognizable 5S rRNA gene is found in the mitochondrial DNA (mtDNA) of land plants, many (but not all) green and red algae, and jakobid flagellates but is absent from the mitochondrial genome of many other protist groups, as well as all animals and fungi (8, 9, 18).
One possible resolution to this enigma is that the functional role of 5S rRNA has been dispensed with entirely in some mitochondrial systems or, perhaps, has been assumed by proteins. Indeed, mitoribosomes are notable for their highly atypical structural characteristics (28, 29), and in certain cases small and large subunit (SSU and LSU, respectively) rRNAs have been drastically reduced in size, with additional novel proteins apparently substituting for some of the deleted RNA segments (20, 22, 28, 35).
Alternatively, a nucleus-encoded 5S rRNA species might be imported into mitochondria and assembled along with the mtDNA-specified small and large subunit rRNAs and mitochondrial ribosomal proteins (some or all of which are encoded in the nucleus and imported into mitochondria). Although import of nucleus-encoded 5S rRNA into mitochondria has been reported (reviewed in reference 3) and although abundant evidence does support the import of nucleus-encoded tRNA species into mitochondria (1), compelling support for the presence and function of nucleus-encoded 5S rRNA in the mitoribosome per se is lacking.
A third possibility is that a mtDNA-specified RNA lacking some or all of the hallmarks of a conventional 5S rRNA is nevertheless able to substitute for the latter in some mitoribosomes. Indeed, a divergent 5S rRNA species, the sequence of which was not immediately recognized during initial genome annotation, was subsequently identified experimentally in the lobose amoeba Acanthamoeba castellanii (3). This small, abundant RNA species was shown to be encoded in the mtDNA, mapping to a previously unassigned region in the A. castellanii mitochondrial genome sequence. On close inspection, this RNA species was found to display primary and secondary structural features (albeit limited) identifying it as a 5S rRNA homolog; moreover, in ultracentrifugation experiments it cosedimented with mitoribosomes.
A. castellanii is a member of the superkingdom Amoebozoa, a group specifically related to opisthokonts (animals plus fungi) and encompassing the lobose amoebae, entamoebae, pelobionts, and mycetozoons (slime molds) (6). Fig. 1 depicts our current understanding of the evolutionary relationships of the amoebozoan species discussed in this report. Here, we describe abundant, 5S-sized RNA species encoded by the mtDNA of several other amoebozoans and provide evidence of their association with mitoribosomes. We discuss whether these novel small abundant RNA species represent radically divergent homologs of mitochondrial 5S rRNA or, alternatively, analogs that have assumed the function normally carried out by a conventional 5S rRNA species.
Hartmannella vermiformis (ATCC 50236; kindly provided by A. J. Lohan) was cultured at room temperature without shaking in a liquid medium (adjusted to pH 6.0 with HCl) consisting of 0.1% yeast extract, 2% proteose peptone, 4 mM MgSO4, 0.4 mM CaCl2, 3.4 mM NaCl, 50 μM Fe(NH4)2(SO4)2, 2.5 mM Na2HPO4, 2.5 mM KH2PO4, and 0.1 M glucose. Cells were pelleted by centrifugation and washed twice with phosphate-buffered saline and resuspended in 10 ml of homogenization buffer (0.35 M sucrose, 50 mM Tris-HCl [pH 8.0], 3 mM EDTA, 1 mM dithiothreitol [DTT], 0.1% bovine serum albumin [BSA]) with a Teflon homogenizer.
H. vermiformis cells were disrupted by passage through a French pressure cell at 1,500 lb/in2. Mitochondria were obtained by differential centrifugation. Purified mitochondria were resuspended in 9 ml of mitochondrial lysis buffer (10 mM Tris-HCl [pH 8.5], 50 mM KCl, and 10 mM MgCl2) with a Teflon homogenizer. Mitochondria were lysed by the addition of 1 ml of 20% Triton X-100, followed by five cycles of 30 s of mixing using a vortex, followed by a 30-s incubation on ice. The resulting mixture was centrifuged at 9,000 × g for 10 min, and RNAs were prepared from the resulting supernatant fraction by the addition of 0.1 volume of 10× detergent mix (10% Sarkosyl, 0.5 M NaCl), followed by extraction with phenol-cresol and precipitation from the aqueous phase with ethanol. Cytosolic RNAs were prepared by the same procedure from the postmitochondrial fraction obtained by differential centrifugation. Methods for cell lysis, organelle purification, and the preparation and visualization of nucleic acids are based on Spencer et al. (31).
RNAs were separated on 10% denaturing polyacrylamide gels (19:1 acrylamide to bis-acrylamide with 7 M urea) and visualized by either staining with ethidium bromide or UV shadowing (11). Selected RNAs were eluted from homogenized polyacrylamide gel slices by shaking overnight at 4°C in a 1:1 mixture of phenol-cresol–buffer [0.5 M NH4OAc, 10 mM Mg(OAc)2, 1.0 mM EDTA], precipitated with ethanol, and analyzed by RNA sequencing (see below).
Microplasmodia were washed, suspended in buffered sucrose, and disrupted in a Waring blender, and mitochondria were isolated by differential centrifugation as described previously (34). After lysis in 20 mM Tris-HCl (pH 7.5), 20 mM EDTA, and 0.5% SDS, mitochondrial nucleic acids were deproteinized via multiple extractions with phenol-chloroform-isoamyl alcohol (25:24:1, vol/vol/vol) and chloroform-isoamyl alcohol (24:1, vol/vol), followed by ethanol precipitation. After treatment with DNase I (Roche), mitochondrial RNAs were deproteinized and precipitated as above.
RNAs were 5′ end labeled (following the protocol of reference 27) with T4 polynucleotide kinase and [γ-32P]ATP. RNAs were 3′ end labeled with [5′-32P]pCp and RNA ligase (23), according to the manufacturer's instructions. Labeled products were precipitated twice with ethanol, separated in polyacrylamide gels, excised, and eluted from gel slices. To determine RNA sequences, 3′- and 5′-end-labeled RNAs were subjected to partial alkaline hydrolysis and partial digestion with RNase T1 (which cleaves 3′ to G residues) and sometimes also with RNase U2 (which cleaves preferentially 3′ to A residues but also after G) and RNase PhyM (which cleaves preferentially 3′ to U residues but also after A). The products of these digestions were separated in adjacent wells of 6% and 20% polyacrylamide gels.
End analysis was performed by digestion of 5′-end-labeled RNAs with nuclease P1 and 3′-end-labeled RNAs with RNase T2 or a combination of RNases T1, T2, and A. The products of nuclease P1 digestion were separated by one-dimensional thin-layer chromatography (TLC) on cellulose plates [predipped in a 10% dilution of a saturated solution of (NH4)2SO4 and allowed to dry] using a 4:1 mixture of 95% ethanol-water as the solvent (17) or by two-dimensional TLC using Polygram cellulose plates with solvents A (isobutryic acid-NH4OH-H2O [66:1:33, vol/vol/vol]) and B [0.1 M sodium phosphate (pH 6.8)–(NH4)2SO4–n-propanol (100:60:2, vol/wt/vol)] (10). Ammonium formate (0.5 M, pH 9.2; 0.2 volume) was added to nuclease P1 digests prior to one-dimensional TLC, as well as to RNase T2 digests, the products of which were separated as described above on cellulose plates using 95% ethanol-water (4:1) solvent (17). Products of digestion with the combination of RNases T1, T2, and A were separated in two dimensions as above. Radioactive products of P1 and T2 digests were visualized by autoradiography. The digestion products (mononucleotides) of the unlabeled Saccharomyces cerevisiae tRNAs added to nuclease P1 and RNase T2 digestions were visualized by UV light and served as migration markers for radiolabeled products.
Five μg of P. polycephalum mitochondrial RNA was heated to 90°C for 5 min, quickly cooled on ice, and then incubated overnight at 37°C in 50 mM HEPES (pH 7.5), 15 mM MgCl2, 3.3 mM DTT, 10% dimethyl sulfoxide (DMSO), 0.01 μg/μl BSA, and 80 μM ATP with 10 units of T4 RNA ligase (Promega). Ligated RNAs were deproteinized, precipitated with ethanol, and resuspended in 10 μl of water. For cDNA synthesis from circularized RNAs, 1 μg of ligated RNA was mixed with 1 pmol of primer 2IG (CACGTCAATTTTGTATATTTTAC), heated to 90°C for 2 min, cooled to room temperature over ~20 min, and incubated on ice for 15 min. Annealed primer-templates were then incubated in 50 mM Tris, pH 8.3 (at 42°C), 50 mM KCl, 10 mM MgCl2, 10 mM DTT, 0.5 mM spermidine, and 60 μM each deoxynucleoside triphosphate (dNTP) with 10 units of avian myeloblastosis virus (AMV) reverse transcriptase (Life Sciences) for 45 min at 42°C. One-tenth of the cDNA was used as the template in subsequent PCRs using Taq polymerase (Roche) under conditions suggested by the supplier. Primers were phosphorylated by incubation with polynucleotide kinase prior to PCR with primers 2IG and 14IG (CGAAATCGGAAAAGCACTAAT). The resulting reverse transcription-PCR (RT-PCR) product was gel purified and ligated to SmaI-digested pBSM13+ (Stratagene) that had been treated with calf intestinal alkaline phosphatase (Roche).
Isolated mitochondria were lysed with 2% Triton X-100 (H. vermiformis) or 1% NP-40 (P. polycephalum) and centrifuged at 9,000 × g (H. vermiformis) or 15,000 × g (P. polycephalum). Cleared lysates were fractionated by ultracentrifugation in 10 to 30% continuous glycerol gradients in mitochondrial lysis buffer at 21,000 rpm for 18.5 h in an SW27 swinging bucket rotor (Beckman). Fractions were collected from the gradient, extracted with phenol-cresol, and precipitated with ethanol. Samples used for Northern blotting were treated with DNase I (Roche) prior to electrophoresis. RNAs were separated on 10% denaturing polyacrylamide gels and visualized by ethidium bromide staining. Selected H. vermiformis RNAs were eluted from homogenized polyacrylamide gel slices as described above and analyzed by RNA sequencing (see below).
RNAs from glycerol gradient fractions were electrophoresed as above and electroblotted onto a Nytran membrane. Hybridizations were carried out for ~20 h at 42°C in 0.5 M sodium phosphate (pH 7.2), 7% SDS, 1% BSA, and 1 mM EDTA with 5′-end-labeled oligonucleotide probes and then washed sequentially with 5× SSC–0.1% SDS (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), 2× SSC–0.1% SDS, and 1× SSC-0.1% SDS at 45°C. Blots were first probed with an oligonucleotide (CAAACATTAGTGCTTTTCCG) specific for the Physarum small mitochondrial RNA characterized here (P. polycephalum RNA [ppoRNA]) (see below) and then stripped and probed with an oligonucleotide complementary to the mitochondrial LSU rRNA (ATGTTCGCTCACCACTAC).
The nine amoebozoan sequences described here were analyzed with the program RNAshapes (32), installed locally. The RNAshapes package allows one to search consensus shapes common to a set of input sequences, scoring these shapes by the sum of the lowest minimum free energy (MFE), calculating shape probabilities, and drawing the shape-representative structure (MFE structure of the given shape class). We used the shape representative folding mode (command line option −a), consensus shape mode (command line option −C), and a relative energy range of 10% (command line option −e 0.1; default); the shape abstraction level value was generally set to 5 (command line option −t 5). When values that were <5 were tested, the value of the -C parameter was increased in order to obtain results for all sequences.
To align the structures and identify the most highly conserved substructure elements, we used RNAforester (14) in the multiple, local alignment mode. Note that both RNAforester and RNAshapes use secondary structure information exclusively and ignore RNA sequence.
In light of the discovery of a divergent but nonetheless bona fide 5S rRNA in A. castellanii mitochondria (3), we expected that the mtDNA of other members of Amoebozoa would also encode such a species. In silico searches of the recently completed mtDNA sequence of H. vermiformis (GenBank accession number GU828005 [G. Burger et al., unpublished data]), a distant amoebozoan relative of A. castellanii (6), failed to identify a potential 5S rRNA sequence or stretches similar to A. castellanii 5S rRNA. Therefore, we prepared mitochondria from H. vermiformis and separated mitochondrial RNAs by size. The resulting profiles revealed a highly abundant small RNA that comigrated with cytoplasmic 5S rRNA during polyacrylamide gel electrophoresis (Fig. 2A). The size, abundance, and subcellular localization of this novel RNA, referred to here as hveRNA, for H. vermiformis RNA, suggested that it might represent a mitochondrial 5S rRNA counterpart. (Table 1 gives a compilation of the small abundant mitochondrial RNAs described here).
5′ End labeling of isolated hveRNA resulted in two labeled RNA species, differing in length by one nucleotide (Fig. 2B) and with the longer RNA several times more abundant than the shorter version. We determined that these two RNAs had identical sequences (Fig. 2C), except that the longer variant (hveRNA-1) contained an additional U at its 5′ terminus (nuclease P1 digestion and TLC confirmed U as the 5′-terminal residue for both length variants) (data not shown). 3′ End labeling of isolated hveRNA from a separate culture of H. vermiformis resulted in three labeled RNAs (Fig. 2D), two of which by enzymatic sequencing (data not shown) proved to be length variants of hveRNA, with one of these two variants (hveRNA-3) containing two more nucleotides at its 3′ end than the other species (hveRNA-4). The third species likely represents contaminating cytosolic 5S rRNA, based on its size and high G content, as determined by enzymatic sequencing (data not shown). The shorter of the two hveRNA 3′ length variants (hveRNA-4) was several times more abundant than the longer variant (compare intensity of pCp-labeled bands in Fig. 2D). The combined sequence data revealed an RNA species that is heterogeneous at both its 5′ and 3′ ends, encompassing 124 nucleotides of sequence in total.
Glycerol gradient ultracentrifugation of a H. vermiformis mitochondrial extract resulted in all detectable hveRNA being found at or near the bottom of the gradient, cosedimenting with mitoribosomes (Fig. 3A, box 2, and B, lane 2), supporting the inference that hveRNA is associated with a large complex, likely the mitoribosome. Cosedimentation of hveRNA with mitoribosomes suggests that despite the absence of expected primary sequence motifs, hveRNA might nevertheless represent a highly derived 5S rRNA counterpart.
In contrast, tRNAs and a separate 5S-sized RNA (Fig. 3A, box 1, and B, lane 1) were found exclusively at or near the top of the gradient, confirming that the high-speed centrifugation did not sediment small RNA species that are not associated with large complexes. The 5S-sized RNA that remained at the top of the gradient (Fig. 3A, box 1) is contaminating cytoplasmic 5S rRNA; the sequence of this RNA (data not shown) was found to be the same as that of the shortest pCp-labeled RNA shown in Fig. 2D.
When the hveRNA sequence was used as a query in BLAST searches of public domain databases, no 5S rRNA sequences (or sequences of any other known RNAs) were retrieved. A search of the complete mtDNA sequence of H. vermiformis located the gene for hveRNA 7 nucleotides (nt) downstream of the tRNAArg(UCG) gene and on the same strand as all other genes in this genome (GenBank accession number GU828005 [G. Burger et al., unpublished]) (Fig. 4A). Processing of the 5′ end of hveRNA and the 3′ end of tRNAArg(UCG) could be linked: e.g., an endonucleolytic cleavage to create the hveRNA 5′ end would also liberate the 3′ end of tRNAArg(UCG) for further processing.
The hveRNA gene falls in a tandem repeat region, whose repeat units end 5 nt upstream of the 3′ end of the longest hveRNA 3′ variant. Each repeat unit consists of 147 nt, and no sequence heterogeneity is observed between repeat units. The region is at least 1.35 kbp long and comprises at least nine repeats; however, the precise length of this region is not known.
Although the sequence of hveRNA does not suggest homology to 5S rRNA or to any other known RNA, it can be modeled to adopt a secondary structure that is consistent with that of a 5S rRNA (Fig. 4B).
In the course of studies of RNA editing in the mitochondria of the myxogastrid slime mold P. polycephalum (phylum Mycetozoa), another abundant small RNA was encountered. Total RNA isolated from crude mitochondrial preparations was 5′ or 3′ end labeled and separated by size (Fig. 5A). The majority of bands resulting from 5′ end labeling appear to be degradation products containing 5′ OH groups because they were labeled to roughly the same extent with or without prior treatment with alkaline phosphatase (Fig. 5A, lanes 5 and 6). In contrast, two prominent species had 5′ P as the 5′ end could be labeled with polynucleotide kinase and [γ-32P]ATP only after phosphatase treatment (Fig. 5A, lane 5 versus lane 6, asterisk and arrowhead). Additionally, these two RNAs had 3′ OH termini as they were labeled by RNA ligase and [32P]pCp without prior phosphatase treatment (Fig. 5A, lane 7). Enzymatic sequencing (data not shown) of the larger of these two species, a molecule of 120 nt (Fig. 5A, asterisk), identified it as the nuclear DNA-encoded cytoplasmic 5S rRNA, a common contaminant of mitochondrial RNA preparations.
The smaller of these two species (Fig. 5A, arrowhead) carries most likely a monophosphorylated 5′ end because it could not be labeled by guanylyltransferase (Fig. 5A, lane 4). Enzymatic sequencing (data not shown) of this RNA following 5′ and 3′ end labeling revealed a 97- to 98-nt RNA (referred to here as ppoRNA) encoded by a previously unassigned region of the P. polycephalum mitochondrial genome, between the SSU rRNA and tRNAMet2 genes. The only discrepancy between the sequence of the RNA and that of the corresponding gene is the presence of two additional C residues in the RNA. These C insertions had been observed previously in in vivo and in vitro studies examining primary transcripts containing the SSU rRNA (4) and in transcripts that include the 3′ end of the SSU rRNA, ppoRNA, tRNAMet2, and tRNALys (7). These two C insertions were previously thought to be the only identified editing events in intergenic regions of characterized P. polycephalum mitochondrial transcripts, but the identification of the stable ppoRNA species negates this interpretation.
The 3′ end of the ppoRNA gene and the 5′ end of the tRNAMet2 gene potentially overlap by a single nucleotide, a C (7). We determined the 5′ and 3′ termini of ppoRNA by hydrolyzing purified end-labeled RNA with specific nucleases and analyzing the single nucleotides by TLC (Fig. 5B and C, respectively). The 5′ end was found to be predominantly pU, with much smaller amounts of pA, pC, and pG. The 3′ end was predominantly Cp (73%), although a significant amount of Up (27%) was also observed. To confirm these results, ppoRNA was circularized by incubation with RNA ligase, and the region encompassing the joined ends was amplified by RT-PCR (procedure based on reference 25), cloned, and sequenced. Consistent with the results from ppoRNA end labeling and our characterization of tRNAMet2 (7), all 10 of the randomly selected clones contained a T at the 5′ end, with eight clones having C and two clones having T at the 3′ end (see Fig. S1 in the supplemental material). This experiment confirms that the 3′ end of the ppoRNA gene overlaps the tRNAMet2 gene by 1 nt and further shows that the RNA has 5′ P and 3′ OH termini.
Support for the idea that ppoRNA is structured and/or complexed with proteins in vivo comes from the observation that in mitochondrial lysates this RNA is largely protected from RNase H cleavage in the presence of oligonucleotides encompassing its entire length, unless the lysate is treated with proteinase and heated prior to the cleavage assay (A. Rhee and J. Gott, unpublished data). In addition, a significant portion of ppoRNA cosedimented with mitochondrial rRNA upon ultracentrifugation (Fig. 6A and B). These observations indicate that, like hveRNA, ppoRNA is associated with mitoribosomes.
As in the case of the other small RNAs discussed in this work, ppoRNA lacks primary sequence features expected of a canonical 5S rRNA although the predicted secondary structure (Fig. 5D) resembles that of a 5S rRNA.
Using PCR primers specific for the 3′ ends of the SSU rRNA and tRNAPro genes in Physarum mitochondria, Antes and colleagues were able to clone the corresponding region of the mitochondrial genome of the closely related myxomycete, Didymium nigripes (2). The gene organization within this region is similar in the two species (Fig. 7A) and includes the genes for tRNAMet2, tRNALys, and tRNAPro (2). Although the Didymium sequence that is positionally homologous to ppoRNA shares only ~50% identity with ppoRNA (Fig. 7B), the hypothetical RNA derived from this region (referred to as D. nigripes RNA, or dniRNA) can be folded into a 5S-like secondary structure (Fig. 7C). However, considering that Didymium mitochondrial rRNAs (16), tRNAs (2), and mRNAs (13, 15, 33) are subject to editing by nucleotide insertion, any mature dniRNA likely differs from the corresponding DNA sequence.
Extensive sequence similarity searches of the complete mtDNA of two other slime molds, Dictyostelium discoideum and Polysphondylium pallidum, failed to uncover a sequence that has any of the diagnostic features expected of a 5S rRNA homolog, including the mitochondrial 5S rRNA of A. castellanii. However, the mtDNA of D. discoideum does encode an abundant, 5S-sized RNA species, originally termed msRNA, for mitochondrial small RNA (24). Based on its size, abundance, and enrichment in mitochondrial preparations, the authors suspected that this 129-nt RNA (D. discoideum RNA, or ddiRNA) might be a 5S rRNA. Although we show here that ddiRNA can be folded into a 5S rRNA-like secondary structure having the expected helical and single-stranded regions (Fig. 8A), it lacks evident sequence motifs characteristic of other 5S rRNAs. Moreover, ddiRNA was found almost exclusively in the supernatant after centrifugation of a mitochondrial lysate at 250,000 × g rather than cosedimenting with mitochondrial ribosomes (24), as would be expected of a bona fide 5S rRNA. Thus, the identity and function of this RNA species remained unclear.
In a sequence similarity search of the mtDNA from a related slime mold, P. pallidum CK8, we found a stretch of identity to the D. discoideum ddiRNA sequence within an unassigned region immediately downstream of the P. pallidum LSU rRNA gene (Fig. 9E). In D. discoideum mtDNA, ddiRNA is also encoded immediately downstream of the LSU rRNA gene (Fig. 9A). Comparative structure modeling of the P. pallidum region flanking the stretch of sequence identity (Fig. 8B) revealed similarities to the proposed ddiRNA secondary structure model, suggesting that this region also encodes a small RNA (P. pallidum CK8 RNA, which we refer to here as ppaRNA-CK8) and that ddiRNA and the proposed ppaRNA-CK8 are homologous (compare colored regions in Fig. 8A and B). Additional sequence conservation (see Fig. S2 in the supplemental material) further supports their common evolutionary origin.
We examined mtDNA sequence from additional Dictyostelium and Polysphondylium species and adduced further support for the existence of mtDNA-encoded homologs of ddiRNA in these amoebae. Sequences corresponding to putative Dictyostelium citrinum RNA (dciRNA), Dictyostelium mucoroides RNA (dmuRNA), Dictyostelium fasciculatum RNA (dfaRNA), and P. pallidum PN500 RNA (ppaRNA-PN500) are located immediately downstream of the LSU rRNA gene in their respective mitochondrial genomes (Fig. 9), as is the case for ddiRNA and ppaRNA. All of these sequences can potentially adopt a 5S rRNA-like secondary structure (Fig. 10). The most highly conserved stretch of sequence corresponds to helix IV and loop D of the canonical 5S rRNA secondary structure (21) (Fig. 8, inset).
The secondary structure diagrams of A. castellanii mitochondrial 5S rRNA (see Fig. S3 in the supplemental material) and the small abundant amoebozoan RNAs described here (Fig. 4B, ,5D,5D, and and8A)8A) and of the positionally equivalent sequences in other amoebozoan mitochondrial genomes (Fig. 7C, ,8B,8B, and and10)10) were modeled manually, guided by the common sequence and folding features of authentic 5S rRNAs and supported by compensating base changes in helical regions. As the diagrams illustrate, all of these sequences have the propensity to fold into a 5S-like secondary structure. To assess whether this particular folding pattern is more likely than other possibilities, we used a bioinformatics approach to assess possible shapes, folding stabilities, and statistical probabilities.
First, we asked whether the 10 amoebozoan sequences are able to adopt structures that share similar features, i.e., a common abstract shape characterized by a particular arrangement of helices, but disregarding the length of helices and unpaired regions. Analysis with the RNAshapes suite (32) detected a consensus shape common to all 10 sequences, consisting of a stem bifurcating into two hairpins. Figure 11 shows, for each of the 10 amoebozoan sequences, the secondary structure having the minimum free energy (the MFE structure) within the consensus shape class. This shape, referred to here as a rabbit shape, is also the abstract shape of conventional 5S rRNAs.
The most highly conserved portion of these MFE structures, as determined by the tool RNAforester (14), is the 5′-proximal “ear” of the rabbit shape (Fig. 11), which always includes two helices separated by an internal loop and closed by a terminal loop. This region corresponds precisely to the most highly conserved structural elements in conventional 5S rRNAs, i.e., helix II, internal loop B, helix III, and terminal loop C (Fig. 8). Less well conserved is the 3′-proximal ear (Fig. 11), which consists of various numbers of internal loops (0 to 2) across the 10 MFE structures.
Next, we inspected the folding stability of the proposed secondary structures. Notably, for each of the 10 sequences, the MFE structure within the rabbit class coincides with the MFE structure of the entire folding space. Thus, the folding energy calculations support the view that the amoebozoan sequences assume a 5S rRNA-like shape.
Even accepting that the in silico and the manually modeled structures have a congruent common abstract shape, the nucleotides engaged in helices or loops might differ among these structures. Visual comparison of the computationally derived MFE structures with the manual ones modeled independently (Fig. 4, ,5,5, ,7,7, ,8,8, and and10)10) shows an impressive degree of conformity in the 5′-proximal ear (Fig. 11). The only exception is hveRNA, where loop B, helix III, and loop C deviate. In the case of the 3′-proximal ears, the in silico and the manually built structures are congruent for A. castellanii mitochondrial 5S rRNA, hveRNA, dniRNA, and ppoRNA but differ substantially in the other instances. The poor fit in the region comprising helices IV and V is not surprising, considering that manual modeling uses sequence conservation wherever this is evident (Fig. 8 and and10,10, colored nucleotides), whereas modeling by RNAshapes is solely based on pairing interactions.
Finally, we asked what the probability is that a given sequence assumes a secondary structure belonging to the rabbit shape class, considering the entire folding space of a molecule. A high cumulative shape probability indicates that a molecule has a single preferred shape. For all 10 amoebozoan sequences, the rabbit shape is the highest-probability class, corroborating that their folding into a 5S-like secondary structure is not only possible but also reasonably probable.
Here, we describe, in several amoebozoan genera (Dictyostelium, Hartmannella, and Physarum), experimentally confirmed mtDNA-encoded RNA species that have several of the generic characteristics of 5S rRNA, namely, high abundance, small size (97 to ~130 nt), ability to adopt a 5S rRNA-like secondary structure, and (in the two cases investigated here) association with mitoribosomes in ultracentrifugation experiments. Moreover, we have identified in silico what appear to be homologous sequences in the mitochondrial genome of several additional Dictyostelium species and in a closely related slime mold genus, Polysphondylium, as well as in a more distantly related myxomycete slime mold, D. nigripes. We predict that these sequences are also expressed as small abundant RNAs in the mitochondria of these other amoebozoan taxa. No conserved open reading frames (ORFs) and no AUG-initiated ORFs of >20 codons in length are contained within these small RNAs or their putative homologs.
Within the two relatively closely related genera Dictyostelium and Polysphondylium, sequence and positional conservation of the genes in question is evident although sequence identity is restricted to relatively short stretches. Thus, even where there is reasonable evidence of homology, the sequences per se are highly divergent. The same situation applies to Physarum and Didymium. In fact, pairwise comparisons demonstrate that the putative mitochondrial 5S rRNA sequences are evolving at a considerably more rapid rate than the corresponding LSU rRNA sequences in the same organisms, with the former consistently exhibiting identity values 10 to 30% lower than the latter. For example, for D. discoideum versus D. citrinum, ddiRNA and dciRNA sequences are only 83.7% identical, whereas the corresponding LSU rRNA sequences (disregarding introns) share 94.0% identity. Very similar values (82.7% for ppaRNA; 94.0% for LSU rRNA) are observed with P. pallidum CK8 versus P. pallidum PN500, whereas a wider spread is seen in a D. fasciculatum-P. pallidum CK8 comparison (86.0% identity for LSU rRNA and 57.8% identity for ddiRNA versus ppaRNA-CK8). In a broader comparison involving Dictyostelium/Polysphondylium, Hartmannella, and Physarum/Didymium, there is no compelling sequence-based evidence of homology among all of the small RNAs described here. Nevertheless, considering that in three instances these sequences are known to be expressed as abundant small RNAs—and in the case of Physarum to undergo the type of insertional editing that characterizes maturation of mitochondrial mRNA, rRNA, and tRNA transcripts in this organism—it is highly likely that these small abundant RNAs are functional in the mitochondria of the organisms in which they are present.
In the absence of both a functional test for 5S rRNA and in vitro mitochondrial translation systems for the amoebozoan taxa investigated here, we cannot definitively assert that these small RNAs are functional counterparts of the 5S rRNA found in other mitoribosomes; however, several of their properties support this hypothesis. Cofractionation of the small RNAs identified here with mitoribosomes (e.g., in ultracentrifugation experiments) is perhaps the strongest indication that these RNAs may play a role in ribosome function. In this respect, it is noteworthy that Pi et al. (24) found that in their ultracentrifugation experiments, the ddiRNA (msRNA in their study) was localized almost exclusively in the high-speed supernatant rather than in the mitoribosome-containing pellet. However, it is likely that dissociation of ribosome-bound ddiRNA was promoted by the relatively high ionic conditions the authors used (buffer containing 250 mM KCl).
In two cases (Hartmannella and Physarum), isolated mitochondrial RNA contained a species identified as cytoplasmic 5S rRNA in addition to the small mtDNA-encoded RNAs characterized here. However, it is unlikely that a nuclear DNA-encoded 5S rRNA is imported into mitochondria to replace the function of a mtDNA-encoded 5S rRNA in these taxa because in ultracentrifugation experiments, Hartmannella cytoplasmic 5S rRNA did not cofractionate with mitoribosomes, whereas hveRNA did.
Computational modeling corroborates the notion that the amoebozoan small RNAs structurally resemble 5S rRNAs, but whether they are derived homologs or analogs originating from convergent evolution cannot be distinguished by the approach we employed.
In view of the extraordinary structural plasticity displayed by mitoribosomes (20, 22, 28, 29, 35) and in marked contrast to the structural conservatism of their bacterial counterparts, it is not unreasonable to suppose that a conventional 5S rRNA might be replaced by some other molecule (RNA or protein) in the mitoribosome of certain organisms. In this regard, we note that in the large subunit of the mammalian mitoribosome, which lacks a 5S rRNA component, the region normally occupied by 5S rRNA (the so-called central protuberance) is, in fact, replaced by protein. Atomic resolution three-dimensional structures suggest that a protein element termed the LSU handle may assume some of the roles of 5S rRNA in the mitoribosome (28). Thus, molecular mimicry, in which a nonhomologous 5S rRNA-like analog is able to structurally and functionally substitute for a conventional 5S rRNA, is a possibility that we cannot entirely discount. However, while there is precedent for substitution of a protein for RNA in mitoribosomes, there is none for substitution of RNA for RNA. Hence, if the mitochondrial RNA species described here do function as 5S rRNA counterparts, in all probability they are 5S rRNA homologs (albeit highly divergent) rather than analogs. The view that a 5S rRNA was encoded in the mtDNA of the common ancestor of Amoebozoa and subsequently diverged through speciation is further supported by the presence of a recognizable mitochondrial 5S rRNA in A. castellanii. By extrapolation, highly derived (but currently unrecognized) 5S rRNAs may also exist in other mitochondrial systems that appear to lack a conventional 5S rRNA.
Definitive proof of the hypothesis advanced here will require a demonstration that a 5S rRNA-like counterpart is not only a component of the large mitoribosomal subunit in one or more of the organisms studied here but also that it occupies the analogous position within the ribosome's three-dimensional structure. In turn, this will require the isolation and crystallization of a sufficient quantity of pure amoebozoan mitoribosomes to allow determination of the tertiary structure.
We gratefully acknowledge Edward Niles (SUNY Buffalo School of Medicine) for his generous gift of vaccina virus guanylyltransferase and Ben Somerlot (Seattle Children's Hospital Research Institute) for cloning the RT-PCR products derived from circularized ppoRNA. Further, we thank Robert Giegerich (Universität Bielefeld, Germany) and Tuana Mesquita (Université de Montreal, Canada) for their tutoring in use of the RNAshapes software.
Funding for this work was provided by CIHR grant MOP-4124 to M.W.G. and NIH grant GM54663 to J.M.G.
†Supplemental material for this article may be found at http://ec.asm.org/.
Published ahead of print on 19 March 2010.