Comparative sequence analysis of tmRNA

Abstract

Minimal secondary structures of the bacterial and plastid tmRNAs were derived by comparative analyses of 50 aligned tmRNA sequences. The structures include 12 helices and four pseudoknots and are refinements of earlier versions, but include only those base pairs for which there is comparative evidence. Described are the conserved and variable features of the tmRNAs from a wide phylogenetic spectrum, the structural properties specific to the bacterial subgroups and preliminary 3-dimensional models from the pseudoknotted regions.

Prediction of consensus RNA secondary structures including pseudoknots

“…tmRNA: The structure of tmRNA contains four H-type pseudoknots and is roughly globular [47]. The consensus structure predicted by our program is based on the alignment of 8 bacterial tmRNA sequences.…”

“…Comparative sequence analysis revealed conserved pseudoknots e.g. in rRNAs [6], RNase P RNAs [5,18], and tmRNA [47].…”

Evaluation and refinement of tmRNA structure using gene sequences from natural microbial communities

“…The extensive structural variation in diverse tmRNA sequences, and the sporadic occurrence of some helical regions, has been noted previously (Williams & Bartel 1998;Zwieb et al+, 1999)+ Extensive structural variability occurs even within the gamma-group proteobacterial sequences reported here, confirming the notion that much of the tmRNA structure can vary with little effect on its basic function+ Indeed, experimental modifications of the E. coli tmRNA have shown that three of the pseudoknots (pk2, pk3, and pk4; Fig+ 4) are completely interchangeable and can even be replaced by unstructured regions with no significant loss of function (Nameki et al+, 2000)+ We suspect that these phylogenetically volatile regions of the structure occur on the surface of the functional unit because of spacefilling constraints (Burgin et al+, 1990), whereas conserved structural elements and sequences are expected to form the functional core of tmRNA+ A recent crosslinking study also may shed some light on the variability of tmRNA across different lineages+ In this study, the investigators discovered substantial crosslinking between tmRNA and ribosomal protein S1 (Wower et al+, 2000)+ Ribosomal protein S1, however, is not found in all bacterial lineages, specifically the Low GϩC group of Gram-positive bacteria+ Thus, at least in this group of Bacteria, tmRNA must find an alternative binding site+ If the binding site of tmRNA can change radically, as this study suggests, then such extensive variability in tmRNA secondary structure may be expected+…”

“…The new tmRNA sequences were first aligned to the known gamma-proteobacterial sequences using Clustal W (Aiyar, 2000) leaving the original alignment unchanged+ We then added the natural community sequences to 87 other tmRNA sequences collected from the internet (Knudsen et al+, 2001) and manually refined the alignment by keying on elements of structure using the ARB Program (Strunk & Ludwig, 1999)+ The basic proposed tmRNA secondary structure model was used as a template to align homologous regions (Williams & Bartel, 1996;Zwieb et al+, 1999)+ All of the comparative analyses were based on multiple alignments that included only the positions that were homologous to the E. coli nucleotides+ Other sections of the alignments contained numerous gaps and were not considered reliable+ To predict tmRNA structure, we calculated the phylogenetically based R ij and H ij statistics for the data set using the, appropriately titled, Rij and Hij programs compiled on a PC running Slackware Linux version 3+5 + These statistics incorporate information from the phylogenetic relationships among the sequences and have been shown to be more accurate than standard mutual information methods (Akmaev et al+, 1999+ The phylogenetic tree for the 107-sequence tmRNA data set used in the statistical analysis was calculated using the neighbor-joining algorithm in the PHYLIP phylogeny package (Felsenstein, 1993)+ For given pairs of positions, the R ij and H ij statistics compare the rates of evolution of each position independently (calculated as the independent likelihood) to the joint rates of evolution (joint likelihood) of the two positions+ If the independent rate of evolution for two positions is very high (low independent likelihoods) but the positions always change together (high joint likelihood) then the values for these statistics will be high+ Both statistics calculate the independent likelihoods using the phylogenetic tree, but only H ij utilizes the phylogenetic tree to calculate the joint likelihoods, making this statistic much slower to calculate+ However, the H ij statistic approximates a x 2 distribution with nine degrees of freedom, allowing for an assessment of confidence in particular proposed pairings+ In the course of the analysis, we utilized the computationally faster R ij method to identify the set of initial pairs with high correlation values and then used the computationally intensive H ij statistic to determine which of these pairs were statistically significant+ We also performed mutual information (MI) analyses using the in-house ALEX program (version 1+0) designed by Dan Frank, which calculated MI values based on previous work (Gutell et al+, 1992)+…”

“…Phylogenetic comparative methods have proven effective for inferring the secondary and tertiary structures of RNA molecules (Woese et al+, 1980;Williams & Bartel, 1996;Frank & Pace, 1998;Wassarman et al+, 1999)+ Comparative methods use sequences of homologous RNAs from diverse organisms as natural sources of variation to detect correlated change (covariation) between nucleotide positions (Gutell et al+, 1992)+ Such correlated changes are evidence of direct interactions in homologous RNAs+ Comparative analysis has been used to establish the structures of several large and small RNA molecules, for instance the rRNAs (Fox & Woese, 1975;Woese et al+, 1980;Gutell et al+, 1990), tRNA (Gutell et al+, 1992), and RNase P RNA (Brown et al+, 1996)+ Accurate structure predictions using methods of statistical comparative analysis typically require a large number of sequences (.30) from diverse organisms (Gutell et al+, 1992;Akmaev et al+, 1999)+ This extent of data is available for some RNAs, but other stable RNAs such as 6S RNA, OxyS RNA, and CsrB RNA have too few sequences available to analyze with statistical comparative methods (Wassarman et al+, 1999)+ On the other hand, some small RNAs, such as tmRNA, are well represented by 50 or more highly variable sequences that have been used for structure prediction (Williams, 2000;Zwieb & Samuelsson, 2000)+ tmRNA, named for its dual tRNA and mRNA roles, is a small RNA so far found exclusively in the Bacteria, and is conserved throughout that phylogenetic domain (Keiler et al+, 2000)+ tmRNA frees ribosomes stalled on mRNAs that lack stop codons (Williams & Bartel, 1996;Williams et al+, 1999)+ tmRNA contains a structural element that mimics a tRNA, for interaction with the ribosome, and a translational open reading frame (mRNA element)+ Translation of the mRNA element adds a short polypeptide to the incomplete protein that targets it for degradation (Nameki et al+, 1999)+ Recent studies indicate that tmRNA may require an RNA-binding protein (smpB) to function properly as well (Karzai et al+, 1999(Karzai et al+, , 2000+ Although the core regions of the proposed tmRNA structural model seem sound, some elements of the structure vary substantially in diverse organisms and are difficult or impossible to align (Williams & Bartel 1998;Zwieb et al+, 1999)+ Considering this rapidly evolving nature of tmRNA, the best approach to a statistical comparative analysi...…”