Pattern analysis of RNA secondary structure similarity and consensus of minimal-energy folding.

We describe an automated procedure to search for consensus structures or substructures in a set of homologous or related RNA molecules. The procedure is based on the calculation of optimal and sub-optimal secondary structures using thermodynamic rules for base-pairing by energy-minimization. A linear representation of the secondary structures of the related RNAs is used so that they can be compared and classified using standard alignment and clusterings programs. We illustrate the method by means of two sets of homologous small RNAs, U2 and U3, and a set of alpha-globin mRNAs and show that biologically interesting consensus structures are obtained.

Deconvolution of combinatorial libraries for drug discovery: a model system.

Iterative synthesis and screening strategies have recently been used to identify unique active molecules from complex synthetic combinatorial libraries. These techniques have many advantages over traditional screening methods, including the potential to screen large numbers of compounds to identify an active molecule while avoiding analytical separations and structural determination of unknown compounds. It is not clear, however, whether these techniques identify the most active molecular species in the mixtures and, if so, how often. Two key factors which may affect success of the selection process are the presence of many active compounds in the library with a range of activities and the chosen order of unrandomization. The importance of these factors has not been previously studied. Moreover, the impact of experimental errors in determination of subset activities or in randomization during library synthesis is not known. We describe here a model system based on oligonucleotide hybridization that addresses these questions using computer simulations. The results suggested that, within achievable experimental and library synthesis error, iterative deconvolution methods generally find either the best molecule or one with activity very close to the best. The presence of many active compounds in a library influenced the profile of subset activities, but did not preclude selection of a molecule with near optimal activity.

Deconvolution of combinatorial libraries for Drug discovery: theoretical comparison of pooling strategies.

Synthesis and testing of mixtures of compounds in a combinatorial library allow much greater throughput than synthesis and testing of individual compounds. When mixtures of compounds are screened, however, the possibility exists that the most active compound will not be identified. The specific strategies employed for pooling and deconvolution will affect the likelihood of success. We have used a nucleic acid hybridization example to develop a theoretical model of library deconvolution for a library of more than 250,000 compounds. This model was used to compare various strategies for pooling and deconvolution. Simulations were performed in the absence and presence of experimental error. We found iterative deconvolution to be most reliable when active molecules were assigned to the same subset in early rounds. Reliability was reduced only slightly when active molecules were assigned randomly to all subsets. Iterative deconvolution with as many as 65,536 compounds per subset did not drastically reduce the reliability compared to one-at-a-time testing. Pooling strategies compared using this theoretical model are compared experimentally in an accompanying paper.

Strategies for rapid deconvolution of combinational libraries: comparative evaluation using a model system.

Synthesis and testing of complex mixtures maximize the number of compounds that can be prepared and tested in a combinatorial library. When mixtures of compounds are screened, however, the identity of the compound(s) selected may depend on the deconvolution procedure employed. Previously, we developed a model system for evaluation of deconvolution procedures and used it to compare pooling strategies for iterative and noniterative deconvolution [Freier et al. J. Med. Chem. 1995, 38, 344-352]. We have now extended the model studies to include simulations of procedures with overlapping subsets such as subtractive pooling [Carell et al. Angew, Chem., Int. Ed. Engl. 1994, 33, 2061-2064], bogus coin pooling [Blake and Litzi-Davis. Bioconjugate Chem. 1992, 3, 510-513], and orthogonal pooling [D'Prez et al. J. Am. Chem. Soc. 1995, 117, 5405-5406]. These strategies required synthesis and testing of fewer subsets than did the more traditional nonoverlapping iterative strategies. The compounds identified using simulations of these strategies, however, were not the most active compounds in the library and were substantially less active than those identified by simulations of more traditional strategies.

Structural analysis of a group II intron by chemical modifications and minimal energy calculations.

Folding of the yeast mitochondrial group II intron aI5c has been analysed by chemical modification of the in vitro synthesised RNA with dimethylsulfate and diethylpyrocarbonate. Computer calculations of the intron secondary structure through minimization of free energy were also performed in order to study thermodynamic properties of the intron and to relate these to data obtained from chemical modification. Comparison of the two sets of data with the current phylogenetic model structure of the intron aI5 reveals close agreement, thus lending strong support for the existence of a typical group II intron core structure comprising six neighbouring stem-loop domains. Local discrepancies between the experimental data and the model structures have been analyzed by reference to thermodynamic properties of the structure. This shows that use of the latest refined set of free energy values improves the structure calculation significantly.

Mutant U2 snRNAs of Xenopus which can form an altered higher order RNA structure are unable to enter the nucleus.

We studied the nuclear targeting of U snRNAs by microinjection of wild-type and mutant U2 small nuclear RNA transcripts into the cytoplasm of Xenopus oocytes. It has previously been shown that a mutant U2 RNA (delta C) which does not bind certain common U snRNP proteins, some of which carry epitopes recognized by anti-Sm antisera, does not enter the nucleus. We show here that several mutant U2 RNAs which bind to Sm antigens do not enter the nucleus, demonstrating that this RNA-protein interaction is insufficient to produce a nuclear targeting signal. Computer predictions of the secondary structures of the RNAs, derived from minimal energy calculations, show that those which are unable to enter the nucleus have the potential to form an additional secondary structure interaction due to base complementarity between sequences near to their 5' and 3' ends. The data suggest that this structural feature inhibits nuclear targeting.

U1 snRNA: the evolution of its primary and secondary structure.

In this paper we first show that the primary structure of U1 snRNA is homologous to that of tandem repeated pre-tRNA. Two sets of polymerase III promoter sites (the a and b boxes) are clearly recognisable at the appropriate positions in U1, although neither is functional; these sites occur in a degenerate form and their transcription is initiated by polymerase II. Moreover, several of the conserved subsequences of tRNAs that are not associated with transcription initiation (and supposedly are conserved because of their role in translation) are conserved in U1 as well, one of them being the pattern Py-Py-anticodon-Pu-Pu (for both anticodons of tandem tRNA). Second, we show that the secondary structure of U1 is apparently formed after fixation of the "B-hairpin loop' by one of the associated proteins. If and only if this hairpin loop is fixed, a consensus secondary structure is produced by the minimisation-of-free-energy technique. Moreover, we show that this B-hairpin loop has been destabilised relatively recently in evolutionary time by deletions (e.g., in the polymerase III box). If we reinsert the deleted bases, the so constructed hypothetical "ancestral" molecule folds into the consensus secondary structure by unconstrained energy minimisation (i.e., without fixation of the B-loop). Some features of the secondary structure of tandem repeated pre-tRNA are conserved in U1, but the overall structure has changed dramatically. Like tRNA, U1 has a cloverleaf-like structure, but its overall size has doubled. By comparing their secondary structures and by alignment of the sequences, we trace the local events associated with the global change in secondary structure (and apparently in the function of the molecule). Finally, we discuss our results from the perspective of informatic prerequisites for heterarchical multilevel evolution.

A comparison of thermodynamic foldings with comparatively derived structures of 16S and 16S-like rRNAs.

To increase our understanding of the dynamics and complexities of the RNA folding process, and therewith to improve our ability to predict RNA secondary structure by computational means, we have examined the foldings of a large number of phylogenetically and structurally diverse 16S and 16S-like rRNAs and compared these results with their comparatively derived secondary structures. Our initial goals are to establish the range of prediction success for this class of rRNAs, and to begin comparing and contrasting the foldings of these RNAs. We focus here on structural features that are predicted with confidence as well as those that are poorly predicted. Whereas the large set of Archaeal and (eu)Bacterial 16S rRNAs all fold well (69% and 55% respectively), some as high as 80%, many Eucarya and mitochondrial 16S rRNAs are poorly predicted (approximately 30%), with a few of these predicted as low as 10-20%. In general, base pairs interacting over a short distance and, in particular, those closing hairpin loops, are predicted significantly better than long-range base pairs and those closing multistem loops and bulges. The prediction success of hairpin loops varies, however, with their size and context. Analysis of some of the RNAs that do not fold well suggests that the composition of some hairpin loops (e.g., tetraloops) and the higher frequency of noncanonical pairs in their comparatively derived structures might contribute to these lower success rates. Eucarya and mitochondrial rRNAs reveal further novel tetraloop motifs, URRG/A and CRRG, that interchange with known stable tetraloop in the procaryotes.

Biochemical and genetic evidence for a pseudoknot structure at the 3' terminus of the poliovirus RNA genome and its role in viral RNA amplification.

The sequences in the plus-stranded poliovirus RNA genome that dictate the specific amplification of viral RNA in infected cells remain unknown. We have analyzed the structure of the 3' noncoding region of the viral genome by thermodynamic-based structure calculation and by chemical and enzymatic probing of in vitro-synthesized RNAs and provide evidence for the existence of an RNA pseudoknot structure in this region. To explore the functional significance of this structure, revertants of a mutant bearing a lesion in the proposed pseudoknot and exhibiting a temperature-sensitive defect in viral RNA synthesis were isolated and mapped. The results of this genetic analysis established a correlation between the structure of the 3' terminus of the viral RNA and its function in vivo in RNA amplification. Furthermore, phylogenetic analysis indicated that a similar structure could be formed in coxsackievirus B1, a related enterovirus, which further supports a role for the pseudoknot structure in viral RNA amplification in infected cells.

Equal G and C contents in histone genes indicate selection pressures on mRNA secondary structure.

Protein-specific versus taxon-specific patterns of nucleotide frequencies were studied in histone genes. The third positions of codons have a (well-known) taxon-specific G+C level and a histone type-specific G/C ratio. This ratio counterbalances the G/C ratio in the first and second positions so that the overall G and C levels in the coding region become approximately equal. The compensation of the G/C ratio indicates a selection pressure at the mRNA level rather than a selection pressure or mutation bias at the DNA level or a selection pressure on codon usage. The structure of histone mRNAs is compatible with the hypothesis that the G/C compensation is due to selection pressures on mRNA secondary structure. Nevertheless, no specific motifs seem to have been selected, and the free energy of the secondary structures is only slightly lower than that expected on the basis of nucleotide frequencies.

Multiple coding and the evolutionary properties of RNA secondary structure.

This article evaluates evolutionary properties of the transition from RNA primary sequence to RNA secondary structure. It focuses on the restrictions that the conservation of a protein code in an RNA sequence puts on its potential to evolve towards a specific secondary structure. Restricting the mutations to those that do not affect the coding for a protein restricts both the accessibility and the connectivity of the sequence space. The accessibility is restricted because only certain point mutations are allowed. The connectivity is restricted because no insertions and deletions are allowed. Simulating an evolutionary search process for a specific secondary structure shows that (i) the reduction of allowable point mutations allows for adaptation to some large-scale topology, but strongly reduces the possibility of small-scale adaptations, (ii) the abolition of insertions and deletions has very little effect on the results of the search process. During the evolutionary search process for a secondary structure with a specific topology and a high frequency of base-pairing the quasispecies moves into a subspace in which the similarity between secondary structures of neighboring sequences is relatively high. Increased similarity between second structures of neighboring sequences is also found in the Rev responsive element (RRE) in the lentiviruses Caprine arthritis-encephalitis virus and Visna virus. In these viruses a biased nucleotide frequency in the RRE region suggests that selection for the RRE RNA secondary structure affects the amino acid sequence of the env gene. Our results show a variation in the ruggedness of fitness landscapes which are based on a high degree of epistatic interactions. Fitness landscapes play an essential role, not only in biotic evolution, but also in all kinds of optimization processes (Hill Climbing, Simulated Annealing, Genetic Algorithms, etc). Variation in their ruggedness should therefore be taken into account in the analysis of these processes.

Evolution of the primary and secondary structures of the E1a mRNAs of the adenovirus.

In this paper we investigate and compare (evolutionary) patterns in the primary and secondary structure of four homologous E1a mRNAs of the adenovirus. Our main results are as follows: (1) The similarity of the coding regions of the mRNA sequences reflects both similarity in function (i.e., oncogenicity) and evolutionary divergence. (2) The similarity of the leader and the trailer regions reflects host specificity (i.e., human or simian) and must therefore arise from convergence. (3) Minimal energy foldings of the mRNAs show similar secondary structures (in particular around the splice sites). The conservation of pre-mRNA secondary structure shows that mRNAs are subject to selection constraints in addition to those associated with proteins. (4) The conserved secondary (helical) structures consist of nonhomologous subsequences, i.e., shifts have occurred. The observed shifts near the splice sites seem to be the simplest way of dealing with the dual constraints.

The yeast homologue of U3 snRNA.

snR17, one of the most abundant capped small nuclear RNAs of Saccharomyces cerevisiae, is equivalent to U3 snRNA of other eukaryotes. It is 328 nucleotides in length, 1.5 times as long as other U3 RNAs, but shares significant homology both in nucleotide sequence and in predicted secondary structure. Human scleroderma antiserum specific to nucleolar U3 RNP can enrich snR17 from sonicated yeast nuclear extracts. Unlike other yeast snRNAs which are encoded by single copy genes, snR17 is encoded by two genetically unlinked genes: SNR17A and SNR17B. The RNA snR17A is more abundant than snR17B. Deleting one or other of the genes has no obvious phenotypic effect, except that the steady-state level of snR17B is increased in snr17a- strains. Haploid strains with both genes deleted are inviable, therefore yeast U3 is essential.

Structural analyses of plastid-derived 16S rRNAs in holoparasitic angiosperms.

Higher-order structures have been constructed for plastid-encoded small-subunit (SSU, 16S), rRNAs from representatives of seven nonphotosynthetic holoparasitic angiosperm families: Apodanthaceae, Cynomoriaceae, Cytinaceae, Balanophoraceae, Hydnoraceae, Mitrastemonaceae, and Rafflesiaceae. Whereas most pairwise comparisons among angiosperms differ by 2-3% in substitutions, the 16S rRNAs of the holoparasites show an increasingly greater number of mutations: Cynomorium (7.3%), Cytinus (8.0%), Bdallophyton (12.7%), Mitrastema (14.9%), Hydnora (19.4%), Pilostyles (30.4%) and Corynaea (35.9%). Despite this high level of sequence variation, SSU structures constructed for all species except Pilostyles possess the typical complement of 50 helices (that contain numerous compensatory mutations) thereby providing indirect evidence supporting their functionality. Pilostyles, likely with the most unusual plastid 16S rRNA yet documented, lacks four major helices and contains lengthy insertions for four others. Sequences of products generated via RT-PCR show that these structural modifications are present on a mature (transcribed) rRNA. The trend toward increasing numbers of base substitutions in the holoparasites is accompanied by a marked increase in A+U content of the rRNA. This 'A/T drift' phenomenon of rDNA is especially apparent in Corynaea whose SSU rDNA sequence is 72% A+T. A comparison of Cytinus to tobacco showed that substitution rates appear to be dependent upon the composition of neighboring bases. Transversions represented 26% of the mutations when flanking bases were G or C whereas transversions increased to 36% when the flanking bases were A to T. The underlying molecular mechanism associated with these high substitution rates is presently unknown, however, relaxation of selection pressure on ribosome function resulting in altered DNA replication and/or repair systems may be involved.

Statistics of RNA secondary structures.

A statistical reference for RNA secondary structures with minimum free energies is computed by folding large ensembles of random RNA sequences. Four nucleotide alphabets are used: two binary alphabets, AU and GC, the biophysical AUGC and the synthetic GCXK alphabet. RNA secondary structures are made of structural elements, such as stacks, loops, joints, and free ends. Statistical properties of these elements are computed for small RNA molecules of chain lengths up to 100. The results of RNA structure statistics depend strongly on the particular alphabet chosen. The statistical reference is compared with the data derived from natural RNA molecules with similar base frequencies. Secondary structures are represented as trees. Tree editing provides a quantitative measure for the distance dt, between two structures. We compute a structure density surface as the conditional probability of two structures having distance t given that their sequences have distance h. This surface indicates that the vast majority of possible minimum free energy secondary structures occur within a fairly small neighborhood of any typical (random) sequence. Correlation lengths for secondary structures in their tree representations are computed from probability densities. They are appropriate measures for the complexity of the sequence-structure relation. The correlation length also provides a quantitative estimate for the mean sensitivity of structures to point mutations.

Minimal energy foldings of eukaryotic mRNAs form a separate leader domain.

We have investigated the minimal energy foldings of 38 mature mRNAs, including the globin family, the insulins, the growth hormones and interleukin-2, and have compared these foldings with those of fully and partly randomised sequences. The mRNAs differ from the random sequences in that they form a separate leader hairpin of 40-60 nucleotides, with the initiation codon typically located downstream of this hairpin, followed by a main fold in which a region flanking the initiation codon is basepaired with the trailer: resulting in a close proximity of the 5' and 3' end of the mRNA. The formation of this conformation depends not only--or primarily--on the structure of the leader, but on both the leader and trailer sequence and their interaction with the coding sequence. Thus if, as the frequent occurrence of this pattern suggests, the secondary structure of the leader regions plays a role in the initiation of translation, possibly accounting for the specificity of initiation and the different translational efficiencies of various mRNAs, we expect that these features may be influenced both by leader and trailer mutants.

Novel GACG-hairpin pair motif in the 5' untranslated region of type C retroviruses related to murine leukemia virus.

We searched for the presence of common RNA structural motifs in mammalian type C retroviruses related to murine leukemia viruses and the closely related avian spleen necrosis virus. A novel motif consisting of a pair of hairpins, called hairpin pair motif, was detected in the 5' untranslated regions of the genomes of these retroviruses. A combination of computational analyses that included the assessment of phylogenetic sequence conservation by multiple alignment, the search for regions with unusual RNA folding properties, and the analysis of RNA secondary structure by suboptimal free-energy calculations highlighted the significance of this hairpin pair motif. The hairpin pair motif encompasses 70 to 80 nucleotides between the splice donor site and the gag translational initiation codon of these viruses. The motif is composed of two adjacent hairpins both with a perfectly conserved GACG tetraloop. We propose that the novel GACG-hairpin pair motif described here constitutes an essential component of the regulatory machinery in these type C retroviruses.

Differential premature termination of transcription as a proposed mechanism for the regulation of coronavirus gene expression.

We propose that the different subgenomic mRNA levels of coronaviruses are controlled through differential premature termination of transcription, and are modulated by the relative strength of transcriptional initiation/blockage events. We present the complete set of sequences covering the leader encoding and intergenic regions of the MHV-A59 strain. A computer-assisted analysis of the two now complete sets of these sequences of strain IBV-M42 and MHV-A59 shows that, in contrast to the previous theory, differences amongst stabilities of intermolecular base-pairings between the leader and the intergenic regions are not sufficient to determine the mRNA gradients in both MHV and IBV infected cells. Neither can the accessibility of the interacting regions on the leader and the negative stranded genome, as revealed by secondary structure analysis, explain the mRNA levels. The nested gene organisation itself, on the other hand, could be responsible for observed mRNA levels gradually increasing with gene order. Relatively slow new initiation events at intergenic regions are proposed to block elongation of passing transcripts which, via temporary pausing, can cause premature termination of transcription. This effects longer transcripts more than shorter ones.

'Mutational SURF': a strategy for improving lead compounds identified from combinatorial libraries.

Synthesis and testing of mixtures of compounds in a combinatorial library offers the potential of much greater throughput than the 'one compound, one well' approach. When mixtures of compounds are screened, however, pooling and deconvolution strategies must be employed to identify the most active compound in the library. The possibility exists that the most active compound will not be identified. We have developed a theoretical model of library deconvolution using the well characterized properties of nucleic acid hybridization to calculate activities of individual molecules in libraries of more than 250,000 compounds. Calculations using this model have been employed to evaluate strategies for pooling and deconvolution. In the presence of errors in synthesis and testing, iterative deconvolution or position scanning sometimes identified a compound with sub-optimal activity. We describe a procedure called 'mutational SURF' in which 'mutants' of the selected compound are individually synthesized and tested. Simulations of mutational SURF using our model libraries suggest that mutational SURF provides an efficient method for improving the activity of lead compounds identified from combinatorial libraries.

Nucleotide sequence and transcriptional analysis of molecular clones of CAEV which generate infectious virus.

The lentivirus caprine arthritis-encephalitis virus (CAEV) is closely related by nucleotide sequence homology to visna virus and other sheep lentiviruses and shows less similarity to the other animal and human lentiviruses. The genomic organization of CAEV is very similar to that of visna virus and the South African ovine maedi visna virus (SA-OMVV) as well as to those of other primate lentiviruses. The CAEV genome includes the small open reading frames (ORF) between pol and env which are the hallmarks of the lentivirus genomes. The most striking difference in the organization of CAEV is in the env gene. The Env polyproteins of visna virus and the related SA-OMVV contain 20 amino acids between the translational start and the signal peptide not present in CAEV. In addition to nucleotide sequence analysis, the transcriptional products of CAEV were determined by Northern analysis. The viral mRNA present in cells transfected with the infectious clone reveal a pattern characteristic of the mRNAs observed in other lentivirus infections. The putative tat ORF of CAEV could be identified by genomic location and amino acid homology to the visna virus tat gene. However, the CAEV rev gene could not be identified in a similar fashion. Thus, to determine the location of the rev ORF cDNA clones were obtained by PCR amplification of the mRNA from infected cells. To determine if a Rev response element was contained in the CAEV genome, secondary structural analysis of the viral RNA was performed. A stable stem loop structure which is similar in location, stability, and configuration to that determined for the Rev response element of HIV was found.