key: cord-0033623-7fbzbxa2
authors: Limanskaya, O. Yu.; Murtazaeva, L. A.; Limanskii, A. P.
title: Potential thermosensitive riboswitches in the genome of Salmonella
date: 2013-10-11
journal: Cytol
DOI: 10.3103/s009545271305006x
sha: eb7a3a927d9ff98bbe0daf1815ef98362bc93bca
doc_id: 33623
cord_uid: 7fbzbxa2

Currently, a number of structurally and functionally different thermosensitive elements, such as structurally and functionally different RNA thermometers, for controlling a variety of biological processes in bacteria, including virulence are known. These well-known RNA thermometers are structures, whether matched or mismatched, which are represented by either a single stretched hairpin structure or a few hairpins. Based on computer and thermodynamic analyses of 25 isolates of Salmonella enterica with complete genome, we have developed an algorithm and criteria to search for potential RNA thermometers, which will enable us to undertake a future search for potential riboswitches in the genomes of other socially significant pathogens. In addition to the well-known 4U RNA thermometer, another four hairpin-loop structures have been identified in S. enterica as new potential RNA thermometers and two of them are localized in 5′-UTR of virulence regulators gltB and yaeQ. They are highly conserved noncanonical structures and correspond to the necessary and sufficient conditions for forming RNA thermometers, since they are found in each of the 25 S. enterica genome isolates. We analyzed the thermosensitive motif in the pXO1 plasmid of Bacillus anthracis—an anthrax-causative pathogen—and visualized matched hairpins that form a cruciform structure in pUC8 supercoiled plasmid by atomic force microscopy.

According to the central dogma of molecular biol ogy, the role of RNA molecules is limited to being a passive intermediary between proteins and DNA mo lecules. It has now been established that regulatory RNAs can store not only genetic information but can also perform other functions. Some research laborato ries are especially interested in the regulatory RNAs of prokaryotes, their structure and identification, as well as in their potential use in medicine and biotechnol ogy. Regulatory RNAs have been established to exhibit a catalytic activity in the absence of cofactors, cleave phosphodiester bonds (ribozymes), and bind small molecules (ligands) and ions with a high affinity and specificity via monitoring cell metabolism and con trolling gene expression, translation efficiency, trans cription termination, and mRNA stability; they also regulate the translation of genes that encode heat and cold shock proteins, as well as virulence genes, due to temperature shifts in the surrounding environment [1] [2] [3] .

Free living microorganisms are periodically exposed to shifts in surrounding environmental condi tions, such as temperature, pH, and the presence of nutrients that constantly vary. To counteract the con sequences of temperature fluctuations, bacteria have developed a complex network of defense mechanisms. The potential thermosensitive elements used in nature include components from membranes to DNA, RNA, and protein molecules. Pathogenic microorganisms often respond to temperatures reaching 37°C by inducing virulence gene expression. The genes that control the sensitivity to the environment may be reg ulated at the transcription level through the interac tion of regulatory proteins. However, a few posttran scriptional mechanisms based on RNA molecules were recently discovered [3] . Some tRNA molecules are known to be not only a substrate for ribosomes but also to include controlling elements that modulate their own expression in a condition dependent way. Structural changes in such sensory RNAs are caused by specific surrounding shifts.

There are two principally different classes: cis act ing RNA elements with their regulatory potential in the middle of an mRNA sequence and trans acting, small, and noncoding RNA molecules, which func [4] . Unlike classic attenuators that regulate the struc ture of an RNA leader sequence in correspondence with the translating ribosome position, the cis acting RNAs alter their conformation in response to physical or chemical signals. The so called riboswitches moni tor the cell metabolism through binding metabolites with a high specificity and affinity. They are localized in the 5' untranslated region (UTR) of the genes that encode biosynthesis, intake, and degradation of small metabolites and provide control over the reversible binding for these metabolic pathways.

The binding of a small molecule drives a conforma tional switch that alters gene expression through one of the following three mechanisms: (a) premature trans cription termination, (b) translation initiation, and (c) mRNA processing. The majority of riboswitches switch off expression in a bound state; however, an insignificant number of switches have been found to switch on gene expression.

Unlike highly specific metabolite binding riboswitches, RNA thermometers (a closely related type of sensory mRNAs) act in response to a general physical signal, namely, to the intracellular tempera ture-an important parameter, which affects, in par ticular, the expression of both the genes that encode the heat and cold shock proteins and the virulence genes and are under the constant control of duplexes (formed, for example, by the hairpin stem of RNA thermometers). Melting at an elevated temperature is a well known characteristic of nucleic acids. Temper ature shifts are capable to modulate the conformation of regulatory RNA molecules, that is, the transition of molecule fragments from the intramolecular hairpin conformation to a single stranded state.

There is a group of structurally and functionally different RNA thermometers that control a diversity of cellular processes. All known molecular thermometers (whether cis or trans acting) control translation through isolating ribosome binding fragments, and the majority of them are localized in 5' UTR of bacte rial heat shock or virulence genes. At a low tempera ture, the Shine-Dalgarno sequence (SD sequence, 5' aaggag 3', 5' rraggak 3' which is the consensus sequence for prokaryotes, and 5' uygcu 3' which is the one for Gram negative bacteria) [4] is masked (located in the middle of a hairpin loop structure). An increase in temperature destabilizes the hairpin loop structure, so that the ribosome binding site (the SD sequence) becomes accessible and thus it permits translation ini tiation (AUG is the translation initiation start codon).

The first RNA thermometer that acts through the mechanism of melting was found in the E. coli rpoH gene, which encodes alternative sigma factor σ 32 or RpoH [5] . The alternative RpoS sigma factor plays a central role in the regulation of virulence associated external surface proteins OspC and OspA in Lyme dis ease caused by Borrelia burgdorferi spirochete. Tem perature is one of the RpoS controlled surrounding environment key parameters, while DsrA Bb -a small non coding RNA molecule-regulates the increase in the amount of RpoS, caused by a temperature shift. Lybecker et al. [6] formulated a hypothesis that DsrA Bb is stable as a secondary structure at t = 23°C, at which base pairing with the rpoS transcript does not take place. At increasing the temperature, a small RNA molecule secondary structure becomes degraded due to melting which leads to the binding of the anti SD rpoS mRNA fragment. This may stimulate transla tion under virulent conditions (37°C) through the with drawal of the SD sequence and the translation initiation site from the secondary structure in rpoS mRNA.

It is possible that the most widespread bacterial RNA thermometer is the ROSE element (Repression of Heat Shock Gene Expression), which suppresses heat shock gene expression. This was found in many α and β proteobacteria, including E. coli and Salmo nella enterica [7] . The ROSE element with length from 60 to 100 nucleotides is as a rule localized in 5' UTR of heat shock genes. Its rather complicated structure includes 2-4 hairpin loop structures with one of them containing the SD sequence and, in some cases, also the start AUG codon. Another widespread RNA thermometer is the 4U element initially found in the S. enterica heat shock agsA gene [8] . The supposed structure contains two hairpins (four uridine residues forming complementary pairs with the SD sequence) (Fig. 1 ). The temperature dependent melting has been experimentally confirmed in one of the hairpins, while the binding of the ribosome with the SD sequence takes place only at heat shock temperatures.

The 4U element is often used for controlling the heat shock and virulence genes in bacteria, since this element can bind to the 5' agga 3' fragment of the SD sequence [8] . For example, the hypothesis regarding the control has been proven, using an RNA thermometer, for the Yersinia sp. lcrF (virF) gene encoding the viru lence controlling response regulator [9] . Gene transla tion does not occur at 26°C but is induced at t = 37°C.

An analysis of well known RNA thermometers and their regulatory principles has shown that for their functioning only a few nucleotides should form com plementary bonds with the SD sequence nucleotides or a flanking region to prevent the ribosome binding. It means that there have to exist in nature other still unknown types of RNA thermometers. This hypothe sis has been confirmed using bioinformatic analysis. The synthesis of efficient individual RNA thermome ters provided another proof of this assumption [11, 12] . Nevertheless, molecular details of the bacterial mechanisms of sensitivity to temperature shifts are still obscure in many cases.

Our computer analysis of the chromosomal DNA sequences of a variety of S. enterica isolates from data bases has shown that the 4U thermometer analogous to the supposed thermometer for the S. enterica agsA gene and confirmed by in vivo and in vitro experiments by Waldminghaus et al. [8] (Fig. 1 ) was absent in three out of the 25 analyzed isolates (nos. CP001138, FM200053, and NC_006511). The necessity of adap tation and replication under different conditions, including temperature fluctuations, in particular, between the external environment and the host's organism, leads to the inevitable emergence in patho gens of the corresponding gene expression regulation mechanisms, where RNA thermometers belong as elements. The absence of 4U RNA thermometers in some S. enterica isolates indicates that they may con tain certain thermosensitive elements that are differ ent from both the 4U thermometers and, possibly, from the ROSE elements. In this paper the search of the for new earlier unknown potential RNA thermom eters in the chromosomal DNA sequences of Gram negative Salmonella enterica isolates was performed.

Search for hairpin structures. Chromosomal DNA sequences of 25 S. enterica isolates with complete genome from the GenBank database were chosen for computer analysis.

The search for chromosomal DNA fragments with a potential to form hairpin loop structures in bacteria was performed, using the formula 5' aaggag(n) k atg 3', where aaggag is the SD sequence; n is any possible nucleotide; k = 6-10; and atg is the transcription ini tiation site (alternative transcription initiation sites ttg and gtg were also used). Furthermore, the following formulas were used to search for hypothetical 4U RNA thermometers in sequenced "+" and "-" strands of the S. enterica chromosomal DNA 5' cat(n) 7 ctcct(n) 12 aaaa 3' and 5' tttt(n) 12 agggag(n) 7 atg 3'.

The found sequences were analyzed using the established criteria.

We used the Mfold software (version 3.2) [13] for predicting the secondary structure of the found linear DNA fragments and the corresponding RNA tran scripts, as well as for determining the melting temper atures of potential hairpins at physiological ionic strength (I = 0.2 M Na + , [Mg] 2+ = 0.0 mM or I = 0.15 M Na + , and [Mg] 2+ = 0.2 mM). In addition, the RNA2 program of the Gene Bee software was used for building hairpin loop structures [14] .

Atomic force microscopy (AFM). The supercoiled pUC8 DNA molecule (with length of 2665 bp) was visualized by Nanoscope III Multimode AFM with a D scanner (Veeco Instruments Inc., USA). AFM images of DNA were recorded in tapping mode, in the air under the "height" regime, using standard unsharpened probes (NT MDT, Russia) with a reso nance frequency of 300-360 kHz. To obtain amino mica, we used a freshly cleaved mica modified with amino groups in vapours of distilled (3 aminopro pyl)triethoxysilane (Aldrich Chemical Co., United States). The modification technology has been described earlier [15] .

To date, the functioning of ROSE thermometersone of the well known types of RNA thermometershas been experimentally confirmed only for several of the forty ones proposed for proteobacteria [16] . The possibility of existence of the 4U element-another potential RNA thermometer-has been established for the Brucella melitensis dnaJ gene and for the upstream region of the Staphylococcus aureus groES gene. However, it is still obscure whether they would function as RNA thermometers [8] .

Analysis of specificities in the structural organiza tion of experimentally confirmed RNA thermometers have allowed us to develop an algorithm for the search of the corresponding fragments in the chromosomal DNA sequences in prokaryotes and RNA transcripts, one element of which is the distance between the SD sequence that is included in hairpin loop structure and the translation initiation site, accounting for 6 (for ROSE elements) to 12 nucleotides (for 4U thermom eters). We used the following main criteria for the potential RNA thermometers: (a) the melting temper ature for hairpins is within the range from 32 (at ionic strength of 0.1 M Na + ) to 43°C (at physiological ionic strength); (b) the Shine-Dalgarno sequence is com pletely or partly localized in the stem of the hairpin structure; (c) the start codons are the canonical AUG codon (ATG for DNA) which ensures that more than 90% of all encoded sequences start in prokaryotes, as well as the alternative GUG and UUG codons (GTG and TTG for DNA) [17] . The well known RNA thermometers are structures represented by either a single stretched hairpin or sev eral hairpin loop structures, which may be either per fect or imperfect. As the NMR research results have shown, a hairpin stem that has a ROSE element includes several noncanonical base pairs [18] ; there fore, we have analyzed both perfect and imperfect hairpin loop structures in our search of hypothetical RNA thermometers.

As has been earlier established for supercoiled DNAs that contain palindrome, it is possible under physiological conditions to form hairpins as a frag ment of a cruciform structure with a stem at least 7 bp long and a loop with a size of no more than 4-5 nucle otides [19] [20] [21] . The stem of hairpins formed in 16S rRNA is on average 3-4 bp, reaching 10 bp [22] , whereas a loop with 6-7 nucleotides is the most ener getically preferable for RNA hairpins [23] . It has been shown by the thermal denaturation method that RNA hairpin structures are more stable, compared to the corresponding structures formed in DNA [24] . Based on literature data on both in vivo and in vitro experi ments [25] [26] [27] [28] and considering that hairpin loop struc tures containing a SD sequence effectively affect trans lation only when their free energy is about -6 kcal/mol [29] , we chose for the further analysis hairpins with loop length not exceeding eight nucleotides and a stem not exceeding 7 bp with free energy ΔG reaching approxi mately -2 to -6 kcal/mol. In addition, we looked into the localization of the potential hairpins relative to the hsp genes encoding heat shock proteins.

A combination of methods for immobilizing DNA on amino mica with the AFM advantages enabled us to visualize the hairpin formed cruciform structure of supercoiled pUC8 DNA. An analysis of AFM images of pUC8 DNA on amino mica (Fig. 2) has shown that hairpins look like sharply expressed protrusions on DNA threads, the length of which can be directly eval uated from AFM images. An analysis of our experi mental results has shown that 11-12 bp participated in the formation of a hairpin, and according to the con ducted thermodynamic analysis of inverted repeats a hairpin was shown to be formed by 26 nucleotides, while its free energy ΔG was -17.8 kcal/mol. We note for comparison that the sizes of the cruciform struc tures that were registered, using two dimensional electrophoresis and nuclease treatment, in DNA of ϕX 174 phage and pBR322, ColE1, and pAO3 plas mids are equal to 9-13 bp in the helical fragments of each of the cruciform hairpins, while their loops contain 3-5 nucleo tides [19, 21] .

Plasmid pUC8 is known to contain several palin dromes that can form cruciform structures in aqueous solutions. Using the Oligo program [30] , we have found that the cruci with the structure given in Fig. 3 is thermodynamically more preferable. (The concen tration of plasmid pUC8 was selected so that 4-5 sin gle supercoiled DNAs with a cruciform conformation were visualized per image of 2 μm × 2 μm in size.) One of the hairpins forming a cruci had free energy ΔG reaching -17.8 kcal/mol. The search for self comple mentary fragments in the pUC8 DNA sequence, using the GeneBee software (RNA 2 program) [14] , has also confirmed that the hairpin consists of 11 nucleotide model of pairs, while the loop length is four nucle otides. The model of supercoiled pUC8 DNA possess ing a cruciform structure is presented in Fig. 3 . There are several variants of the hairpin forming for cruci. For example, the hairpin may be formed by both complementary inverted repeats of one strand and repeats belonging to complementary DNA strands. We believe that further improvements in the AFM tech nology will enable researchers to determine the cruci form structure in detail, that is, what fragments of the DNA strand are complementary out of several possible formation variants of this noncanonical structure.

It is significant that RNA molecules look like con densed structures under AFM visualization. We earlier visualized RNA trascripts, immobilized on mica, which formed rod like condensed structures 122 ± 10 nm in length with a length-width ratio of 4.5-5 [31] . We think that, for visualizing stretched uncondensed RNA molecules that contain hairpin loop structures, it is necessary to change the surface properties of the substrate (mica). The noted morphological specifici ties of AFM visualized RNA molecules can be explained by a significant effect of the surface proper ties of the mica where RNA transcripts are immobi lized. The surface properties of the substrate are in turn determined by the hydrophobicity and density of the cations localized on the mica surface. The point is that the same mica was used for visualizing both RNA and DNA molecules, i.e., the mica possessing the hydrophobicity and surface density of cations at which linear and supercoiled double stranded DNA mole cules do not form condensed structures when immo bilized on the mica surface but are characterized by an uniform distribution of DNA fragments. We have established earlier that even insignificant changes in the hydrophobicity and density of cations on the mica surface lead to significant modifications in the mor phology of immobilized DNA molecules [32] [33] .

Since thermometers may be localized at any locus in an RNA molecule, to find the hairpins, the main components of those new potential RNA thermome ters differing from the well known 4U RNA ther mometers, we at first analyzed the chromosomal DNA sequences of three S. enterica isolates, which did not contain any of the 4U RNA thermometers.

Our computer and thermodynamic analysis of the fully sequenced S. enterica genome (no. FM 200053 for GenBank) allowed us to reveal four hairpin struc tures (Fig. 4) that satisfy the required conditions for a potential RNA thermometer: availability of the Shine-Dalgarno sequence; the translation initiation site at a distance of no more than 15 nucleotides; the corresponding melting temperature (nearly 40-42°C) for a hairpin; and localization in the 5' UTR. The presence of 13 known heat shock genes (with one of which the 4U RNA thermometer is bound) in isolate FM 200053 enables us to assume that there are several types of RNA thermometers in the genome of salmo nella apart from the one currently known as the 4U RNA thermometer.

It is important that hairpin loop structures (Fig. 4 ) do not only meet the necessary and sufficient condi tions for forming RNA thermometers but also have highly conserved noncanonical structures. They are present in the genomes of all studied 25 isolates and are characterized by 100% similarity in nucleotide sequences.

The expression of virulence factors, such as hemol ysin and lipopolysaccharides, is regulated in proteo bacteria by transcription elongation factor RfaH. The S. enterica yaeQ gene was identified as a high copy suppressor of the hemolytic defect in the deleted seg ment of the rfaH gene. Proceeding from this, Vicari et al. [34] assumed that the direct role of the yaeQ gene that encodes glutamate synthetase is in the transcrip tional control of bacterial virulence.

It is known that different RNA thermometers have different melting temperatures.

For example, the ROSE thermometer melts at 42°C, whereas the pfrH and lcrF thermometers melt at 37°C [35] . Therefore, for consideration we selected as the potential RNA thermometers only hairpin loop structures with a melting temperature within the range 37-42°C.

It is sufficiently difficult to predict for many 5' UTR sequences the possibility of their functioning as RNA thermometers. To confirm the functioning of a hairpin loop structure as a thermosensor, it is neces sary to undertake some experiments, for example, in vivo testing of the efficiency of the reporter gene expression at various temperatures or melting the hair pin loop structure (which contains the SD conse quence). The real melting temperature of the deter mined RNA thermometers is measured, for example, Vol. 47 No. 5 2013 POTENTIAL THERMOSENSITIVE RIBOSWITCHES 273 using circular dichroism spectroscopy or UV spectros copy. Since the melting temperature of nucleic acids depends on the ionic strength, it is considered in assessing the thermodynamic parameters of thermom eters that the concentration of Mg 2+ ions within the 1-2 mM interval corresponds to the physiological value inside bacterial cells.

The probability of the functioning of the found hairpin loop structures as thermosensors in the S. enterica genome is confirmed by the fact that Neupert et al. [11] have established for the functioning of a hairpin loop structure (fully analogous to one of the found potential RNA thermometers) as a thermosen sor (Fig. 4c) . A comparison of the functioning effi CYTOLOGY AND GENETICS Vol. 47 No. 5 2013 ciencies of 12 synthetic RNA thermometers, con ducted by the authors of paper [11] under physiological conditions, has shown that the indicated RNA thermom eter (Fig. 4c) is one of the two RNA thermometers that most efficiently control the temperature dependent expression of the lacZ gene encoding β galactosidase in E. coli [11] . Our potential RNA thermometers (Fig. 4 ) significantly differ in their secondary structure. Although RNA thermometers are characterized by a small loop (4-5 nucleotides), the experimentally con firmed RNA thermometer (Fig. 4c) has a rather large loop, although all found hairpin structures have a per fect stem. Since it has been established in [11] that sig nificantly differing hairpins equally efficiently switch on/off gene expression, we can expect that other hairpin structures (Fig. 4) will also be able to play the role of RNA thermometers.

To check the reliability of the algorithm for deter mining secondary structures, we modeled the second ary structures in a Saccharomycetes cerevisiae mRNA fragment. Lu et al. [36] proposed three secondary structure variations for a 74 bp fragment, each being characterized by 19 canonical base pairs. Simulation of the secondary structure, using the GeneBee soft ware, has also confirmed the possibility of forming 19 canonical base pairs. Thus, the conducted testing has shown that the results obtained in determining the secondary structures are in good agreement with liter ature data; therefore, the methods for predicting potential hairpin loop structures can be regarded as reliable.

Another evidence for the reliability of the method ology used for searching for hypothetical riboswitches in this work can be their subsequent testing. In general, any theoretical studies can be considered successful only after their experimental testing. Therefore, within a pilot project, we conducted preliminary experimental studies in the Salmonella enterica genome for testing a series of RNA thermometers, which we had theoretically determined before. We cloned synthesized nucleotide sequences of potential RNA thermometers into E. coli cells close to the 5' end of the gene that encodes GFP and conducted several preliminary experiments with temperature shifts, observing the changes in the fluorescence of the GFP protein in E. coli. At temperature shifts from 22 to 37°C, a weak GFP induction was recorded but not so strong as observed for synthetic RNA thermometers during temperature controlled gene expression in bacteria.

Thus, based on a computer and thermodynamical analysis of 25 Salmonella enterica isolates with com plete genome, we have established an algorithm and criteria for the search for new potential RNA ther mometers, which will enable us to undertake a search for potential RNA thermometers in the genomes of other socially significant pathogens. In addition to the well known 4U RNA thermometer, hairpin loop structures that can be new RNA thermometers have been determined for S. enterica. They meet the neces sary and sufficient conditions for the formation of RNA thermometers and are highly conserved nonca nonical structures, since they are present in the genomes of all studied S. enterica isolates.

The chemical repertoire of natural ribozymes

The glmS riboswitch inte grates signals from activating and inhibitory metabolites in vivo

Regulatory RNAs in prokaryotes: here, there and everywhere

Microbial thermosen sors

Translational induction of heat shock transcription factor r32: evidence for a built in RNA thermosensor

Temperature induced regulation of RpoS by a small RNA in Borrelia burgdorferi

RNA thermometers are common in alpha and gamma proteobacteria

FourU: a novel type of RNA thermometer in Salmonella

Temperature sensing in Yersinia pestis: translation of the LcrF activator protein is thermally regulated

Genome wide bioinformatic prediction and experi mental evaluation of potential RNA thermometers

Design of sim ple synthetic RNA thermometers for temperature con trolled gene expression in Escherichia coli

RNA quadruplex based modulation of gene expression

UNAFold: software for nucleic acid folding and hybridization

The software package for biopolymer sequence analysis: GeneBee

Imaging compac tion of single supercoiled DNA molecules by atomic force microscopy

RNA thermometers are common in α and γ proteobacteria

Struktura ribo somy i biosintez belka (Molecular Biology. The Struc ture of Ribosomes and Protein Synthesis)

Molecular basis for temperature sensing by an RNA thermometer

Hairpin loop formation by inverted repeats in supercoiled DNA molecules

Cruciform transitions in DNA

The absence of cruciform structures from pAO3 plasmid DNA in vivo

Structures, kinetics, thermodynamics and biological functions of RNA hair pins

The Behavior of Biological Macromolecules

Ghermodynamic parame ters for loop formation in RNA and DNA hairpin tetraloops

A sharp structural transition in pAO3 plasmid DNA caused by increased superhelix density

Kinetics of cruciform formation and stability of cruciform struc ture in superhelical DNA

Con served structural motifs in the 3' untranslated region of genomic RNA of SARS CoV virus

Visualization of cruciform structure in supercoiled DNA by atomic force microscopy

Secondary structure of the ribosome binding site determines translational efficiency: a quantitative analysis

Optimi zation of the annealing temperature for DNA amplifi cation in vitro

Imaging of T7 RNA polymerase elongation complexes by atomic force microscopy

over supercoiled DNA with a 1.94 to 2.19 Å rise per base pair

Compaction of single supercoiled DNA molecules adsorbed onto amino mica

Virulence regulators RfaH and YaeQ do not operate in the same pathway

An RNA thermosensor controls expression of virulence genes in Listeria monocytogenes

A set of nearest neighbor parameters for predicting the enthalpy change of RNA secondary structure formation

The work was partially supported by the National Academy of Medical Sciences of Ukraine (grant no. AMN 95/2010). The authors thank the reviewer for the critical remarks, ideas, and propositions aimed at improving the paper.