• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of rnaThe RNA SocietyeTOC AlertsSubscriptionsJournal HomeCSHL PressRNA
RNA. Nov 2009; 15(11): 2046–2056.
PMCID: PMC2764483

A variant riboswitch aptamer class for S-adenosylmethionine common in marine bacteria

Abstract

Riboswitches that sense S-adenosylmethionine (SAM) are widely distributed throughout a variety of bacterial lineages. Four classes of SAM-binding riboswitches have been reported to date, constituting the most diverse collection of riboswitch classes that sense the same compound. Three of these classes, termed SAM-I, SAM-II, and SAM-III represent unique structures that form distinct binding pockets for the ligand. SAM-IV riboswitches carry different conserved sequence and structural features compared to other SAM riboswitches, but nucleotides and substructures corresponding to the ligand binding pocket are identical to SAM-I aptamers. In this article, we describe a fifth class of SAM binding aptamer, which we have termed SAM-V. SAM-V was discovered by analyzing GC-rich intergenic regions preceding metabolic genes in the marine α-proteobacterium “Candidatus Pelagibacter ubique.” Although the motif is nearly unrepresented in cultured bacteria whose genomes have been completely sequenced, SAM-V is prevalent in marine metagenomic sequences. The consensus sequence and structure of SAM-V show some similarities to that of the SAM-II riboswitch, and it is likely that the two aptamers form similar ligand binding pockets. In addition, we identified numerous examples of a tandem SAM-II/SAM-V aptamer architecture. In this arrangement, the SAM-II aptamer is always positioned 5′ of the SAM-V aptamer and the SAM-II aptamer is followed by a predicted intrinsic transcription terminator stem. The SAM-V aptamer, however, appears to use a ribosome binding site occlusion mechanism for genetic regulation. This tandem riboswitch arrangement exhibits an architecture that can potentially control both the transcriptional and translational stages of gene expression.

Keywords: pseudoknot, metagenome, noncoding RNA, SAM, sulfur metabolism, tandem riboswitch

INTRODUCTION

Riboswitches are noncoding RNA genetic regulatory elements found in intergenic regions (IGRs) of organisms from all three domains of life (Montange and Batey 2008; Roth and Breaker 2009). They fold into complex three-dimensional structures and bind cellular metabolites with high specificity to regulate gene expression via a range of mechanisms (Barrick and Breaker 2007). Most riboswitches are formed from a single ligand-binding aptamer domain and an expression platform that controls gene expression. In eubacteria, riboswitches are almost always found in 5′ untranslated regions (5′ UTRs) of mRNAs and typically use one of two expression platform mechanisms. The first mechanism involves transcription termination, where transcription is halted by an intrinsic terminator stem that causes RNA polymerase to separate from the DNA (Mandal and Breaker 2004; Tomsic et al. 2008). Riboswitch-controlled transcription termination occurs upstream of the open reading frame for the associated mRNA, and thus ligand binding regulates gene expression by regulating mRNA production. The second mechanism involves ribosome binding site occlusion, where translation initiation is prevented by the RNA structure after the ribosome binding site and start codon have been produced by transcription (Fuchs et al. 2006). Thus, these two mechanisms function at two different stages of gene expression. In addition, ribozyme-mediated self-cleavage has also been observed as a riboswitch expression platform (Winkler et al. 2004). In eukaryotes, riboswitches usually function by controlling alternative splicing (Cheah et al. 2007; Wachter et al. 2007).

Sulfur metabolism in many bacteria is extensively regulated by riboswitches that selectively recognize S-adenosylmethionine (SAM) or S-adenosylhomocysteine (SAH) (Fig. 1A; Wang and Breaker 2008). SAM is synthesized from methionine and ATP and is a key methyl group donor for the methylation of DNA, RNA, proteins, phospholipids, hormones, and neurotransmitters (Grillo and Colombatto 2008). SAH is the by-product of SAM methyl group donation (Takusagawa et al. 1998), and the similar chemical structures of SAM and SAH create a molecular recognition challenge for RNA aptamers. Previous publications describe four SAM-binding riboswitch classes (Epshtein et al. 2003; McDaniel et al. 2003; Winkler et al. 2003; Corbino et al. 2005; Fuchs et al. 2006; Weinberg et al. 2008) and one SAH-binding riboswitch class (Wang et al. 2008) that form binding pockets able to discriminate strongly against natural chemical derivatives.

FIGURE 1.
Comparison of SAM-II and SAM-V RNA motifs. (A) Chemical structures of SAM and SAH. (B) Consensus sequences and secondary structure models for SAM-V (left) and SAM-II (right) aptamers. Nucleotides forming the binding pocket of SAM-II and the putative binding ...

Representatives of the SAM-I riboswitch class are the most widely distributed SAM-sensing RNAs (Barrick and Breaker 2007). The consensus sequence and structure were initially named the S box (Grundy and Henkin 1998), and were found residing in the 5′ UTRs of gram-positive mRNAs coding for proteins involved in sulfur metabolism. In Bacillus subtilis alone, SAM-I controls 11 transcriptional units comprising 26 genes involved in sulfur metabolism, methionine biosynthesis, cysteine biosynthesis, and SAM biosynthesis (Winkler et al. 2003; Rodionov et al. 2004; Fuchs et al. 2006). Each SAM-I riboswitch typically functions by facilitating the formation of an intrinsic terminator stem when SAM is bound, thus terminating transcription (Winkler et al. 2003; Tomsic et al. 2008). A variation of the SAM-I class, called SAM-IV, was later discovered (Weinberg at al. 2007) using the CMfinder comparative genomics pipeline (Yao et al. 2007). SAM-IV riboswitches are more narrowly distributed and are present primarily in the order Actinomycetales. Despite significant differences in conserved nucleotide sequence and overall architecture, the two classes share conserved nucleotides within their ligand binding cores, suggesting that the two aptamers use similar ligand binding pockets. The binding pocket similarities and phylogenetic distributions of SAM-I and SAM-IV riboswitches support the hypothesis that SAM-IV riboswitches most likely evolved from a parental SAM-I representative (Weinberg et al. 2008).

SAM-II riboswitches are present predominantly in α-proteobacteria, but are also found in other proteobacteria and bacteroidetes. The first examples of SAM-II aptamers were identified during a comparative sequence analysis focusing on intergenic regions of α-proteobacteria (Corbino et al. 2005), and the sequence and structural features of the RNAs are entirely distinct from those of SAM-I and SAM-IV. Moreover, the phylogenetic distribution of SAM-II riboswitches does not overlap with that of SAM-I and SAM-IV. Structural studies have shown that SAM-I and SAM-II riboswitch aptamers also bind SAM in different conformations (Montange and Batey 2006; Gilbert et al. 2008).

SAM-III or SMK box riboswitches are narrowly distributed mainly within the order Lactobacillales (Fuchs et al. 2006). The first representatives were identified by examining the 5′ UTRs of metK genes, which encode SAM synthetase. This gene is known to be regulated by SAM riboswitches in many other organisms. SAM-III riboswitches commonly regulate gene expression through occlusion of the ribosome binding site or Shine–Dalgarno (SD) sequence. When SAM is bound, the SD base pairs with an anti-SD sequence to form the aptamer, thus preventing ribosome access (Fuchs et al. 2007). The distinctive sequence and secondary structures of SAM-III aptamers yield a tertiary structure and binding pocket that are distinct compared to the other SAM riboswitches previously described (Lu et al. 2008).

As with SAM-IV, the SAH riboswitch class was discovered using the CMfinder comparative genomics pipeline (Weinberg et al. 2007; Wang et al. 2008). SAH riboswitches are always located in the 5′ UTRs of mRNAs coding for proteins involved in SAH degradation and SAM regeneration pathways. Representatives are present in β- and γ-proteobacteria with a few examples residing in species of δ-proteobacteria and actinobacteria. SAH riboswitch aptamers also appear to form a structure that is distinct from all known SAM riboswitch classes (Wang and Breaker 2008).

In the current article, we describe the experimental validation of a fifth class of SAM-binding aptamers, which we have termed SAM-V (Fig. 1B). SAM-V aptamer representatives exclusively have been found in marine α-proteobacteria and bacteroidetes (Meyer et al. 2009). SAM-V exhibits a number of distinct features from the known SAM-binding riboswitch classes, although the consensus sequence and structure has similarities with the binding site of SAM-II riboswitch aptamers. This suggests that the cores of these two aptamer classes may form near identical structures and ligand contacts. Despite this similarity, bioinformatics search algorithms cluster representatives of the two classes separately. We also examined a SAM-II/SAM-V tandem arrangement, which appears to allow both transcriptional and translation control of gene expression.

RESULTS AND DISCUSSION

SAM-V is a SAM binding aptamer

The SAM-V motif was discovered by conducting a comparative sequence analysis of long and GC-rich IGRs in the marine α-proteobacterium “Candidatus Pelagibacter ubique” (Meyer et al. 2009). “Cand. P. ubique” has the smallest genome of any free-living organism, an exceptionally low GC content of 29.7%, and a mean intergenic length of only 3 nucleotides (nt) (Giovannoni et al. 2005). Of the four likely cis-regulatory motifs identified in this study, we were especially interested in the SAM-V motif identified in the IGRs preceding metY, mmuM, and bhmT, which encode O-acetylhomoserine (thiol)-lyase, homocysteine S-methyltransferase, and betaine-homocysteine methyltransferase, respectively. The first enzyme catalyzes the reaction that produces L-homocysteine from O-acetyl-L-homoserine and the other two enzymes catalyze the subsequent reaction producing L-methionine from L-homocysteine. These three genes are also up-regulated during stationary phase suggesting a regulatory role for this motif (Sowell et al. 2008). The genomic context for SAM-V RNAs suggests that each example may function as the aptamer component of a SAM-sensing riboswitch. While riboswitch expression platforms are not always immediately obvious, all three of these motifs in "Cand. P. ubique" are immediately followed by a purine rich region that might act as a ribosome binding site, allowing the riboswitches to act through an SD occlusion mechanism.

To assess the ability of SAM-V RNAs to bind SAM, we tested several RNA constructs using in-line probing (Soukup and Breaker 1999), including the full-length IGRs preceding the genes metY, bhmT, and mmuM, and a 116-nt region of the metY 5′ UTR (116 metY RNA) that is 76% identical in sequence to the corresponding region of the mmuM 5′ UTR. In-line probing relies on the inherent instability of internucleotide linkages in RNA molecules to reveal information about their structural flexibility. RNAs can undergo intramolecular phosphoester transfer reactions when a ribose 2′-oxygen nucleophile is collinear with a downstream 5′-oxygen leaving group and with the electrophilic phosphorus center located between the two (Soukup and Breaker 1999). Single-stranded RNA is more susceptible to spontaneous cleavage because the phosphodiester linkages are better able to explore conformational space and are more likely to be arranged in an in-line conformation compared to double-stranded RNA or RNA that is otherwise tightly structured. Changes in RNA aptamers brought about by ligand binding will yield patterns of spontaneous cleavage products that differ in the absence versus the presence of saturating concentrations of ligand (Nahvi et al. 2002; Soukup and Breaker 1999; Soukup et al. 2001). Moreover, estimated values for ligand dissociation constant (KD) can be made by determining the concentration of ligand needed to bring about half-maximal change in these spontaneous cleavage patterns. All of the constructs noted above exhibit structure modulation when subjected to in-line probing with SAM (data not shown), confirming that SAM-V RNAs are SAM-binding aptamers.

SAM-V aptamers are far more narrowly distributed than SAM-I and SAM-II riboswitch aptamers, and therefore are more similar to SAM-III and SAM-IV riboswitch aptamers in distribution. SAM-V representatives are only present in the fully sequenced HTCC 1062 strain of "Cand. P. ubique," as well as the incomplete "Cand. P. ubique" strain HTCC 1002 and Psychroflexus torquis, a psychrophilic bacterium isolated from Antarctic sea ice belonging to the phylum Bacteroidetes (Bowman et al. 1998). However, homology searches of environmental sequence databases revealed a total of 429 unique SAM-V sequences belonging mostly to oceanic α-proteobacteria and bacteroidetes (Meyer et al. 2009). Due to the high proportion of marine metagenomic DNA originating from species similar to Cand. P. ubique (Rusch et al. 2007), it is likely that many of these SAM-V representatives belong to various strains of this organism.

SAM-V RNAs shares sequence and structural characteristics with SAM-II RNAs

Upon careful examination of an atomic-resolution model of a SAM-II aptamer from the Sargasso sea metagenome bound to SAM (Fig. 1C; Gilbert et al. 2008), we noted several sequence and structure similarities between SAM-V and SAM-II RNAs (Fig. 1B). SAM-II aptamers form a classic H-type pseudoknot architecture, with three Watson–Crick base-paired stems: P1, P2a, and P2b, and three loop regions: L1, L2, and L3 (Gilbert et al. 2008). Stems P1 and P2b form a contiguous near-colinear helical structure, with two loop regions, L1 and L3, folding over the major and minor grooves of the helix, respectively. The SAM binding pocket (Fig. 1C) consists of an extended triplex between L1 and the major groove of P2b. SAM binds in an extended conformation along the major groove face of the P2b-L1 triplex, forming direct contacts with five successive base pairs and base triples. The adenosyl moiety of SAM participates in the formation of a base triple between aptamer residues U10 in L1 and U44 from stem P2b. The positively charged sulfur is recognized by the carbonyl oxygen groups of U11 in L1 and U21 in P2b (Gilbert et al. 2008). On ligand binding, P1 and L3 do not undergo substantial structural changes, whereas P2b and L1 become protected from cleavage by spontaneous phosphoester transfer (Gilbert et al. 2008). This suggests that the interactions between SAM and the aptamer are required to create the stable major groove triplex.

The consensus sequence and structure for SAM-V RNAs share several key characteristics with SAM-II RNAs. SAM-V RNAs are predicted to form a pseudoknot structure with two stems (P1 and P2) and two loops (L1 and L2). Moreover, nucleotides (U10, U11, U12, U20, U21, G22, U44, A45, and A47) that form key interactions with the ligand in the crystal structure of a SAM-II aptamer are highly conserved and present at sites corresponding with those in SAM-V (Fig. 1B). Therefore, we speculate that SAM-II and SAM-V aptamers form similar ligand binding pockets using distinct architectural features to precisely arrange the ligand-binding residues in space. This is analogous to the likely correspondence between SAM-I and SAM-IV riboswitch aptamers that appear to use distinct secondary and tertiary structure features to form similar ligand-binding pockets (Weinberg et al. 2008). However, it is not clear whether SAM-II and SAM-V aptamers are the result of convergent or divergent evolution. The aptamers exist in overlapping lineages, such as α-proteobacteria, and their occurrence in tandem suggests a potential duplication event and therefore divergent evolution. However, the motifs are quite small, and are examples of a common RNA fold—the H-type pseudoknot (Gilbert et al. 2008), thus increasing the plausibility of convergent evolution.

62 metY is a minimized SAM-V aptamer that selectively binds SAM

We generated a 62-nt truncation of the 116 metY RNA, termed 62 metY (Fig. 2A), that carries all sequence and structural features characteristic of SAM-V RNAs. This RNA was examined by in-line probing (see Materials and Methods) to establish the KD values for SAM and various analogs of this coenzyme. A series of in-line probing reactions were conducted using a range of ligand concentrations, and the apparent dissociation constant (KD) was calculated by plotting the extent of structural modulation versus the concentration of ligand. 62 metY RNA binds SAM with a KD of ~150 μM, whereas the 116 metY RNA construct exhibits a KD of 15 μM (Fig. 2B,C). Although the 54 nt at the 5′ end of 116 metY RNA are not conserved and do not show modulation upon SAM binding (data not shown), it is possible that they aid SAM binding indirectly by facilitating the folding of the RNA into its functional tertiary structure or by inhibiting competing folding pathways. Effects on ligand-binding affinity by flanking nucleotides have been observed with other riboswitch aptamers (e.g., Winkler et al. 2002; Wickiser et al. 2005; Roth et al. 2007). Although the apparent KD for SAM-V is higher than those of other riboswitches, except that measured for a glmS ribozyme with GlcN6P (200 μM) (Winkler et al. 2004), intracellular concentrations of SAM grown in media supplemented with methionine have been measured at 300 μM (Tomsic et al. 2008). Furthermore, KD values may be irrelevant for riboswitches that do not have time to reach thermodynamic equilibrium with their ligands, and therefore are kinetically driven (Wickiser et al. 2005; Rieder et al. 2007; Breaker 2008; Wang and Breaker 2008).

FIGURE 2.
The 62 metY RNA construct binds SAM. (A) Sequence and secondary structure model for the 62 metY RNA. The two guanosyl residues at the 5′ end represented by lowercase g are nonnative and were added to facilitate in vitro transcription by T7 RNA ...

To assess the ability of the aptamer to discriminate against compounds related to its ligand, we performed a series of in-line probing experiments in which 62 metY RNA was incubated with SAM analogs previously used to study the specificity of other SAM riboswitch classes. Of these, AzaAdoMet, MeAzaAdoMet, SAH, and SAH sulfone carry modifications at the sulfur atom (Fig. 3A). None of these compounds has a measurable interaction with SAM-V (data not shown). These results are expected given the similarities between SAM-II and SAM-V aptamers and based on the results reported previously (Lim et al. 2006), wherein a SAM-II aptamer did not tolerate modifications at the sulfur position.

FIGURE 3.
The SAM-V aptamer discriminates strongly against analogs of SAM. (A) Both SAM-II and SAM-V reject compounds that carry modifications (X) at the sulfur position. (B) Equilibrium dialysis was conducted with a two-chamber system wherein chambers A and B ...

Equilibrium dialysis confirms SAM binding

To confirm SAM binding using an alternative methodology, we performed equilibrium dialysis experiments using 3H-SAM with SAM-V constructs 62 metY RNA and 116 metY RNA. We also used this technique to examine the SAM binding activity of two constructs termed 1-65 bhmT RNA and 120-174 bhmT. Nucleotides 1–65 from the bhmT 5′ UTR of "Cand. P. ubique" are predicted to form a SAM-II aptamer, whereas nucleotides 120–174 are predicted to form a SAM-V aptamer. The nature of this tandem aptamer arrangement will be discussed in greater detail below.

We used equilibrium dialysis devices consisting of two chambers separated by a 5000-Da molecular weight cutoff dialysis membrane, which allows diffusion of SAM but not the RNAs. A buffered solution containing RNA was added to one chamber of the device and a 3H-SAM solution was added to the other chamber (for details, see Materials and Methods). The chambers were allowed to equilibrate and aliquots were removed from each chamber and measured by liquid scintillation counting (Fig. 3B). With the 116 metY, 1–65 bhmT, and 120–174 bhmT RNAs, we observed shifts of 84%, 84%, and 88%, respectively, of the 3H-SAM to the chamber containing the RNA (Fig. 3B, left panel). The 62 metY RNA caused a more modest shift of 57%. This result was anticipated given the high apparent KD value of the RNA measured by in-line probing (Fig. 2C) and the relatively low concentration of RNA in the equilibrium dialysis chamber. The 46 metY construct, a truncated version of 62 metY wherein key aptamer features are deleted, fails to cause a shift of 3H-SAM. Likewise, the absence of RNA also produces no shift.

The addition of excess unlabeled SAM to equilibrium dialysis systems previously equilibrated with 3H-SAM and 116 metY samples causes a redistribution of radiolabeled compounds (Fig. 3B, right panel), indicating that unlabeled SAM competes with 3H-SAM for RNA binding sites. In contrast, the addition of excess SAH does not cause a redistribution of the 3H-SAM, indicating that the 116 metY does not bind SAH. These results are consistent with our conclusions based on in-line probing assays with SAM and SAM analogs.

Characterization of a tandem SAM-II/SAM-V construct

As demonstrated by the equilibrium dialysis experiments above using the 1–65 bhmT and 120–174 bhmT RNAs, the bhmT 5′ UTR in Cand. P. ubique carries two SAM-binding aptamers that represent SAM-II and SAM-V classes. Similar tandem aptamer arrangements were identified in 120 instances of the 429 unique SAM-V environmental sequences (Meyer et al. 2009). The SAM-II/SAM-V tandem aptamer arrangement is the first known example where two different aptamer classes sensing the same metabolite occur in tandem. Several tandem riboswitch arrangements have been described to date, and different architectures appear to yield distinctive biochemical properties (Stoddard and Batey 2006; Sudarsan et al. 2006; Breaker 2008). For example, tandem glycine aptamers are commonly found wherein two glycine molecules are bound cooperatively and gene expression is regulated via a single expression platform (Mandal et al. 2004). This arrangement provides a more digital genetic response allowing small changes in glycine concentration to bring about more substantial changes in gene expression. Tandem arrangement of two complete and independently functioning riboswitches can yield a similar digital response. This is likely the purpose of tandem TPP riboswitches, such as that present in Bacillus anthracis (Welz and Breaker 2007). Finally, a SAM-I/AdoCbl tandem riboswitch consists of two independently functioning riboswitches that bind and respond to two different ligands to function as a Boolean NOR logic gate (Sudarsan et al. 2006).

To assess ligand binding to the tandem SAM-II/SAM-V system, we performed in-line probing experiments with a 174 nt section of the 5′ UTR of the bhmT gene (175 bhmT RNA) (Fig. 4A). SAM-mediated structural modulation of the RNA is observed both in the region corresponding to the SAM-II aptamer (nucleotides 1–65) and the region corresponding to the SAM-V aptamer (nucleotides 120–175). All spontaneous cleavage product band changes are consistent with a single KD value of ~120 μM. The Hill coefficient was determined to be ~1 (Fig. 4B), indicating that cooperative ligand binding between aptamers is not occurring. Moreover, although the region corresponding the SAM-II aptamer does undergo structural modulation upon SAM addition, the pattern of spontaneous cleavage products does not reflect the pattern expected for SAM-II riboswitch aptamers as has been reported previously (Corbino et al. 2005).

FIGURE 4.
Tandem SAM-II and SAM-V riboswitches (174 bhmT RNA) bind SAM independently. (A) Sequence and secondary structure model of the SAM-II (nucleotides 1–65) and SAM-V (nucleotides 120–174) aptamers in a tandem arrangement preceding the bhmT ...

To further evaluate whether the two aptamers might act cooperatively, we tested truncations of the bhmT 5′ UTR corresponding to the SAM-II aptamer (1–65 bhmT RNA) and the SAM-V aptamer (120–175 bhmT RNA). Both the 1–65 bhmT (KD ~ 1.2 μM) and the 120–174 bhmT (KD ~ 0.8 μM) RNAs exhibit patterns indicative of structural modulation that are characteristic of their aptamer classes (Fig. 4A,B). This finding indicates that the structure modulation of the SAM-II aptamer region in the context of the tandem arrangement may not be due to binding of SAM by this aptamer, but may reflect distal structural changes brought about by ligand binding by the SAM-V aptamer. Furthermore, if the aptamers acted cooperatively, we would expect the affinity of individual aptamers to decrease compared to that of the tandem arrangement (Mandal et al. 2004). In contrast, it appears that the presence of the SAM-V aptamer or its flanking sequences precludes proper structure formation of the SAM-II aptamer.

To further assess possible cooperative function, we created mutations that selectively prevent ligand binding in each aptamer within the context of the 175 bhmT construct. The U57C mutation prevents SAM binding to the SAM-II aptamer because this residue forms a critical interaction with the adenosyl moiety of SAM (corresponds to U44 in Fig 1C). Similarly, the corresponding U165C mutation should prevent SAM binding to the SAM-V aptamer. Interestingly, the U57C mutation has essentially no effect on the pattern of spontaneous cleavage products and does not substantively alter the measured SAM KD value exhibited by the 175 bhmT RNA (Fig. 4B,C). In contrast, this mutation eliminates SAM-II aptamer function when tested in the context of the 1–65 bhmT construct (data not shown). The U165C mutation, however, abolishes SAM-dependent structure modulation throughout the entire 175 bhmT RNA (Fig. 4C). Based on these findings, we conclude that the SAM-II aptamer is not capable of binding SAM in context of the 175 bhmT RNA construct that carries both aptamers. However, the SAM-II RNA is able to bind SAM before the region carrying the SAM-V aptamer is transcribed.

Of the 120 tandem RNAs identified, the SAM-II aptamer always occurs first and is followed by an apparent intrinsic transcription terminator stem (Yarnell and Roberts 1999), suggesting the first aptamer in the tandem arrangement will function as an independent riboswitch to control transcription of the downstream ORF. The SAM-V aptamer follows the terminator stem after a linker of ~20 nt (including the run of uridines that are characteristic of intrinsic terminator stems) and is in turn closely followed by the start codon of the associated ORF. The second loop of SAM-V contains a purine-rich sequence AGGAG that could serve as an SD sequence, which resides 7 nt upstream of the start codon. This arrangement is consistent with a riboswitch that functions by occluding ribosome access to its binding site upon ligand binding. Other riboswitch examples in this organism also appear to act through an SD-occlusion mechanism (Tripp et al. 2008; Meyer et al. 2009; Worden et al. 2009). This common arrangement of SAM-II and SAM-V aptamers, in conjunction with our experimental data, leads us to speculate that the two aptamers independently function as riboswitches at two different stages of the gene expression pathway (Fig. 5).

FIGURE 5.
Proposed gene control mechanisms for tandem SAM-II/SAM-V riboswitches. The SAM-II aptamer may control transcription termination (top), whereas the SAM-V aptamer may control translation initiation (bottom).

If true, then the SAM-II/SAM-V architecture may provide a newly recognized capability for tandem riboswitches. Upon transcription initiation, if the intracellular SAM concentration is high, the SAM-II aptamer will bind its ligand and permit formation of the intrinsic terminator stem to halt transcription before the second aptamer and downstream ORF are synthesized (Fig. 5, top). Transcription termination at this stage would prevent protein expression in the same manner as many other riboswitches that carry a single aptamer. However, if the intracellular SAM concentration is low, transcription continues beyond the intrinsic transcription terminator and both the SAM-V aptamer and the adjoining ORF are synthesized (Fig. 5, bottom). If the concentration of SAM should increase, while this full-length RNA is present in the cell, the SAM-V aptamer could bind SAM and regulate gene expression by sequestering the SD sequence within the ligand-bound aptamer structure. Since the SAM-II aptamer is predicted to have no functional utility after transcription has passed the intrinsic terminator stem, there is no evolutionary advantage gained by retaining ligand binding function by this aptamer in longer RNAs such as the 175 bhmT RNA construct.

In summary, the tandem SAM-II/SAM-V riboswitch architecture permits an organism to control both transcription and translation of an ORF at different times (Fig. 5). This particular feature may be largely unnecessary in organisms with rapid mRNA turnover because genetic control in response to changing concentrations of SAM may be satisfied by a riboswitch that controls only transcription or translation. However, "Cand. P. ubique" exists in a severely nutrient limited habitat. The cells are very small (volume ~0.01 μm3) (Morris et al. 2002), and the genome occupies nearly 30% of the cellular volume. Since all cellular processes are metabolically costly (Giovannoni et al. 2005), it is possible that there is no rapid turnover of mRNAs. In such a situation, a genetic control mechanism allowing both transcription termination and translation control may be very advantageous. If transcription is initiated, the SAM-II riboswitch terminates unnecessary transcription of the downstream ORF if SAM is plentiful. If full-length mRNA has been produced, and if this mRNA is long lived, then SAM-mediated inhibition of translation initiation via the SAM-V riboswitch without unnecessary destruction of the mRNA would be energetically favorable.

CONCLUSION

SAM-V represents a new member of the family of SAM-binding riboswitch aptamers that are positioned to regulate the expression of sulfur metabolism of bacteria. Among all riboswitch aptamers reported to date, SAM-V in particular illustrates how small structured RNAs may be difficult to identify by searching the genomic data of cultured bacterial species. This aptamer is found in only one organism whose genome has been completely sequenced, although numerous examples were identified in metagenomic data. The growing number of SAM-sensing RNA molecules illustrates the diversity of structures that RNA can form to selectively sense and respond to a single metabolite. In addition, the existence of tandem SAM-II/SAM-V architectures may also represent another type of more complex biochemical function that can be derived by combining the functions of simple riboswitches.

MATERIALS AND METHODS

Chemicals and DNA oligonucleotides

SAM and SAH were purchased from Sigma-Aldrich. 3H-SAM was purchased from American Radiolabeled Chemicals. Preparations of SAM analogs were described previously (Lim et al. 2006). All DNA oligonucleotides were synthesized by Sigma-Genosys.

RNA preparation

The 5′ UTRs of the metY, mmuM, and bhmT genes were amplified by PCR from "Cand. P. ubique" genomic DNA using the following primers, respectively:

  • 5′-AAGTAATACGACTCACTATAGGGTGTGCGAGGATTAATACTATTATTAG-3′ and 5′-CTTAATACTCCTTTTTTAGTACTTTTGC-3′;
  • 5′-CCAAGTAATACGACTCACTATAGGGTTTACTAATCAGTCTAATAGATATA-3′ and 5′-TTTTATTCCTTACTTTTTTAGTGC-3′; or
  • 5′-CCAAGTAATACGACTCACTATAGGGGATTATATTTAGTTGCGCTGATTTA-3′ and 5′-CTCCTTTTTAGTGCTTAAGCAAATG-3′.

These fragments were cloned into pCR2.1 using a TOPO TA cloning kit (Invitrogen) and their integrity confirmed by DNA sequencing (The W. M. Keck Foundation Biotechnology Resource Center, Yale University). We used these plasmids as templates for subsequent PCR reactions. The primers used to generate DNA templates for various RNA constructs are listed below.

116 metY RNA:

  • 5′-CCAAGTAATACGACTCACTATAGGGCCGATTTGTTCCTATTGCAG-3′, and
  • 5′-CTTTTTTAGTACTTTTGCTTTATGCAAGGC-3′;

62 metY RNA:

  • 5′-CCAAGTAATACGACTCACTATAGGGTTATCGGTAGGCGCATTTGAAC-3′, and
  • 5′-CTTTTTTAGTACTTTTGCTTTATGCAAGGC-3′;

46 metY RNA:

  • 5′-CCAAGTAATACGACTCACTATAGGGAGGCGCATTTGAACTGTATTGTACG-3′, and
  • 5′-GTACTTTTGCTTTATGCAAGG-3′;

174 bhmT RNA:

  • 5′-CCAAGTAATACGACTCACTATAGGGGATTATATTTAGTTGCGCTGATTTA-3′, and
  • 5′-CTCCTTTTTAGTGCTTAAGCAAATG-3′;

1–65 bhmT RNA:

  • 5′-CCAAGTAATACGACTCACTATAGGGGATTATATTTAGTTGCGCTGATTTA-3′, and
  • 5′-GCGTCGCTTTAGCAGG-3′; and

120–174 bhmT RNA:

  • 5′-CCAAGTAATACGACTCACTATAGGGAAATTAAGCCGGGCAGTTGAACC-3′, and 5′-CTCCTTTTTAGTGCTTAAGCAAATG-3′.

The T7 promoter sequence and two additional guanosyl residues were added to the 5′ end of each forward primer to enable in vitro translation using T7 RNA polymerase.

The DNA template for the 174 bhmT RNA construct carrying the U57C mutation was prepared by PCR combining DNA constructs made with oligonucleotides 5′-ACAAATCCTGCcAAAGCGACGC-3′ and 5′-GCGTCGCTTTgGCAGGATTTGT-3′ and the wild-type primers listed above for the 5′ and 3′ ends. The DNA template for the 174 bhmT RNA carrying a U165C mutation was generated using the wild-type forward primer and 5′-CTCCTTTTTAgTGCTTAAGCAAATG-3′ reverse primer. The DNA template for the 1–65 bhmT RNA carrying a U57C mutation was generated using primers 5′-CCAAGTAATACGACTCACTATAGGGGATTATATTTAGTTGCGCTGATTTA-3′ and 5′-GCGTCGCTTTgGCAGGATTTGTAACGCGC-3′. Lowercase nucleotides designate the mutations introduced into the templates.

RNAs were generated by in vitro transcription of double-stranded DNA templates (~0.5 μg) in a 30 μL reaction volume containing 80 mM HEPES-KOH (pH 7.5) at 23°C, 24 mM MgCl2, 2 mM spermidine, 40 mM DTT, 2.5 mM each of the four ribonucleoside 5′ triphosphates (NTPs), and 5 U/μL T7 RNA polymerase (T7 RNAP). Reactions were incubated for 2.5 h at 37°C and RNA products were separated by denaturing (8 M urea) 6% PAGE. Bands corresponding to full-length transcripts were visualized using UV shadowing and excised from the gel. RNA was eluted from the excised gel fragments using 400 mL of a solution containing 10 mM Tris-HCl (pH 7.5) at 23°C, 200 mM NaCl, and 1 mM EDTA. The solution was incubated at 4°C with agitation for 12 h. RNA was precipitated from the supernatant using 1 mL (2.5 vol) of 100% ethanol, the sample was centrifuged at 13,000 rpm for 20 min, and the resulting RNA pellet was air dried for 30–45 min and resuspended in 30 μL of water.

RNAs whose sequences are not provided in the figures are listed below.

116 metY RNA:

  • 5′-CCGAUUUGUUCCUAUUGCAGGCUAGUUAUUGCUAAAAAGGAUUUUUUAACCACAUUAUCGGUAGGCGCAUUUGAACUGUAUUGUACGCCUUGCAUAAAGCAAAAGUACUAAAAAAG-3′;

46 metY RNA:

  • 5′-AGGCGCAUUUGAACUGUAUUGUACGCCUUGCAUAAAGCAAAAGUAC-3′;

113 mmuM RNA:

  • 5′-CCGAUUUGAUCCAUAUUGCAGGUGCUAACUGCUAAAAAGGAAAUCAUUAUAUAAUUGGUAAAGGUAGUUGAACUGUAUUGUGCACCUUGCAUAGAGCUAAAGCACUAAAAAAG-3′;

174 bhmT U57C RNA:

  • 5′-GAUUAUAUUUAGUUGCGCUGAUUUAAGUUAAAUUGCAGAGCGCGUUACAAAUCCUGCCAAAGCGACGCUUCAACCACCGCUUUAGACAGGUGGUUUUUUUUUAUCUUUCAUAAAAUUAAGCCGGGCAGUUGAACCAUAUUGUGCGCCCUGCAUUUGCUUAAGCACUAAAAAGGAG-3′;

174 bhmT U165C RNA:

  • 5′-GAUUAUAUUUAGUUGCGCUGAUUUAAGUUAAAUUGCAGAGCGCGUUACAAAUCCUGCUAAAGCGACGCUUCAACCACCGCUUUAGACAGGUGGUUUUUUUUUAUCUUUCAUAAAAUUAAGCCGGGCAGUUGAACCAUAUUGUGCGCCCUGCAUUUGCUUAAGCACCAAAAAGGAG-3′; and

1–65 bhmT U57C:

  • 5′-GAUUAUAUUUAGUUGCGCUGAUUUAAGUUAAAUUGCAGAGCGCGUUACAAAUCCUGCCAAAGCGACGC-3′.

In-line probing assays

RNA molecules prepared as described above were dephosphorylated using alkaline phosphatase (Roche Diagnostics) and radiolabeled with [γ-32P] ATP and T4 polynucleotide kinase (New England Biolabs) according to the manufacturers instructions. The resulting 5′ 32P-labeled RNAs were purified on 6% PAGE. 32P-labeled RNA was incubated for 40 h in the presence of varying concentrations of SAM (50 nM–4 mM) in a solution containing 50 mM Tris-HCl (pH 8.3) at 23°C, 20 mM MgCl2, and 100 mM KCl. The cleavage products were resolved using denaturing (8 M urea) 10% PAGE. Gels were dried and imaged using a Molecular Dynamics PhosphorImager and band intensities were quantified using ImageQuaNT software. Product band intensities at sites of interest were quantified and normalized to a range of 0 (lowest concentration of SAM tested) to 1 (highest concentration of SAM tested). The average of these values at each site was plotted against the logarithm of the molar concentration of SAM. Apparent dissociation constant (KD) values were estimated by plotting the relative extents of structural modulation at several sites and fitting a standard binding curve to the points.

Equilibrium dialysis

Equilibrium dialysis experiments were conducted using DispoEquilibrium Dialyzer (Harvard Biosciences), in which two chambers (A and B) are separated by a 5000 MWCO membrane. Chamber A was loaded with a 20 μL mixture containing 2 μM 3H-SAM and 50 mM MOPS (pH 7.2) at 20°C, 20 mM MgCl2, and 500 mM KCl; whereas chamber B was loaded with 20 μL mixture containing 10 μM RNA (116 metY, 1–65 bhmT and 46 metY RNAs), 19.2 μM RNA (62 metY RNA), or 6.95 μM RNA (120–174 bhmT) in the same buffer mixture. The chambers were equilibrated at 25°C for 10 h, and 3 μL aliquots were removed from each chamber to establish the distribution of 3H label by using scintillation counting.

For competitive binding experiments, 30 μL mixtures similar to those described above were prepared, incubated and analyzed as described above. Subsequently 3 μL of the buffer mixture containing 4 mM of either SAH or SAM (unlabeled) was added to chamber A, and an equivalent volume of the buffer mixture alone was added to chamber B. The chambers were allowed to equilibrate for an additional 10 h before the redistribution of 3H label was measured as described above.

ACKNOWLEDGMENTS

We thank members of the Breaker laboratory for helpful discussions. We also thank Michael S. Schwalbach and Stephen J. Giovannoni (Oregon State University) for the "Cand. P. ubique" genomic DNA. M.M.M. was supported by a NRSA postdoctoral fellowship (F32GM079974). Research in the Breaker laboratory is also supported by the Howard Hughes Medical Institute.

Footnotes

Article published online ahead of print. Article and publication date are at http://www.rnajournal.org/cgi/doi/10.1261/rna.1824209.

REFERENCES

  • Barrick JE, Breaker RR. The distributions, mechanisms, and structures of metabolite-binding riboswitches. Genome Biol. 2007;8:R239. doi: 10.1186/gb-2007-8-11-r239. [PMC free article] [PubMed] [Cross Ref]
  • Bowman JP, McCammon SA, Lewis T, Skerratt JH, Brown JL, Nichols DS, McMeekin TA. Psychroflexus torquis gen. nov., sp. nov., a psychrophilic species from Antarctic sea ice, and reclassification of Flavobacterium gondwanense (Dobson et al. 1993) as Psychroflexus gondwanense gen. nov., comb. nov. Microbiology. 1998;144:1601–1609. [PubMed]
  • Breaker RR. Complex riboswitches. Science. 2008;319:1795–1797. [PubMed]
  • Cheah MT, Wachter A, Sudarsan N, Breaker RR. Control of alternative RNA splicing and gene expression by eukaryotic riboswitches. Nature. 2007;447:497–500. [PubMed]
  • Corbino KA, Barrick JE, Lim J, Welz R, Tucker BJ, Puskarz I, Mandal M, Rudnick ND, Breaker RR. Evidence for a second class of S-adenosylmethionine riboswitches and other regulatory RNA motifs in alpha-proteobacteria. Genome Biol. 2005;6:R70. doi: 10.1186/gb-2005-6-8-r70. [PMC free article] [PubMed] [Cross Ref]
  • Epshtein V, Mironov AS, Nudler E. The riboswitch-mediated control of sulfur metabolism in bacteria. Proc Natl Acad Sci. 2003;100:5052–5056. [PMC free article] [PubMed]
  • Fuchs R, Grundy F, Henkin TM. The SMK box is a new SAM-binding RNA for translational regulation of SAM synthetase. Nat Struct Mol Biol. 2006;13:226–233. [PubMed]
  • Fuchs R, Grundy F, Henkin TM. S-adenosylmethionine directly inhibits binding of 30S ribosomal subunits to the SMK box translational riboswitch RNA. Proc Natl Acad Sci. 2007;104:4876–4880. [PMC free article] [PubMed]
  • Gilbert S, Rambo R, Van Tyne D, Batey R. Structure of the SAM-II riboswitch bound to S-adenosylmethionine. Nat Struct Mol Biol. 2008;15:177–182. [PubMed]
  • Giovannoni SJ, Tripp HJ, Givan S, Podar M, Vergin KL, Baptista D, Bibbs L, Eads J, Richardson TH, Noordewier M, et al. Genome streamlining in a cosmopolitan oceanic bacterium. Science. 2005;309:1242–1245. [PubMed]
  • Grillo M, Colombatto S. S-adenosylmethionine and its products. Amino Acids. 2008;34:187–193. [PubMed]
  • Grundy FJ, Henkin TM. The S box regulon: A new global transcription termination control system for methionine and cysteine biosynthesis in gram-positive bacteria. Mol. Micro. 1998;30:737–749. [PubMed]
  • Lim J, Winkler WC, Nakamura S, Scott V, Breaker RR. Molecular-recognition characteristics of SAM-binding riboswitches. Angew Chem Int Ed Engl. 2006;45:964–968. [PubMed]
  • Lu C, Smith AM, Fuchs RT, Ding F, Rajashankar K, Henkin TM, Ke A. Crystal structures of the SAM-III/SMK riboswitch reveal the SAM-dependent translation inhibition mechanism. Nat Struct Mol Biol. 2008;15:1076–1083. [PMC free article] [PubMed]
  • Mandal M, Breaker RR. Adenine riboswitches and gene activation by disruption of a transcription terminator. Nat Struct Mol Biol. 2004;11:29–35. [PubMed]
  • Mandal M, Lee M, Barrick J, Weinberg Z, Emilsson G, Ruzzo W, Breaker RR. A glycine-dependent riboswitch that uses cooperative binding to control gene expression. Science. 2004;306:275–279. [PubMed]
  • McDaniel BA, Grundy FJ, Henkin TM. Transcription termination control of the S box system: direct measurement of S-adenosylmethionine by the leader RNA. Proc Natl Acad Sci. 2003;100:3083–3088. [PMC free article] [PubMed]
  • Meyer MM, Ames TD, Smith DP, Weinberg Z, Schwalbach MS, Giovannoni SJ, Breaker RR. Identification of candidate structured RNAs in the marine organism ‘Candidatus Pelagibacter ubique.’ BMC Genomics. 2009;10:268. doi: 10.1186/1471-2164-10-268. [PMC free article] [PubMed] [Cross Ref]
  • Montange RK, Batey RT. Structure of the S-adenosylmethionine riboswitch regulatory mRNA element. Nature. 2006;441:1172–1175. [PubMed]
  • Montange RK, Batey RT. Riboswitches: Emerging themes in RNA structure and function. Annu. Rev. Biophys. 2008;37:117–133. [PubMed]
  • Nahvi A, Sudarsan N, Ebert M, Zou X, Brown K, Breaker RR. Genetic control by a metabolite binding mRNA. Chem Biol. 2002;9:1043. [PubMed]
  • Morris SR, Connon SA, Vergin KL, Giovannoni SJ. Cultivation of the ubiquitous SAR11 marine bacterioplankton clade. Nature. 2002;418:630–633. [PubMed]
  • Rieder R, Lang K, Graber D, Micura R. Ligand-induced folding of the adenosine deaminase A-riboswitch and implications on riboswitch translational control. ChemBioChem. 2007;8:896–902. [PubMed]
  • Rodionov DA, Vitreschak AG, Mironov AA, Gelfand MS. Comparative genomics of the methionine metabolism in gram-positive bacteria: A variety of regulatory systems. Nucleic Acids Res. 2004;32:3340–3353. [PMC free article] [PubMed]
  • Roth A, Breaker RR. The structural and functional diversity of metabolite-binding riboswitches. Annu Rev Biochem. 2009;78:305–334. [PubMed]
  • Roth A, Winkler WC, Regulski EE, Lee BW, Lim J, Jona I, Barrick JE, Ritwik A, Kim JN, Welz R, et al. A riboswitch for the queuosine precursor preQ1 contains an unusually small aptamer domain. Nat Struct Mol Biol. 2007;14:308–317. [PubMed]
  • Rusch DB, Halpern AL, Sutton G, Heidelberg KB, Williamson S, Yooseph S, Wu D, Eisen JA, Hoffman JM, Remington K, et al. The Sorcerer II Global Ocean Sampling Expedition: Northwest Atlantic through eastern tropical Pacific. PLoS Biol. 2007;5:e77. doi: 10.1371/journal.pbio.0050077. [PMC free article] [PubMed] [Cross Ref]
  • Soukup GA, Breaker RR. Relationship between internucleotide linkage geometry and the stability of RNA. RNA. 1999;5:1308–1325. [PMC free article] [PubMed]
  • Soukup GA, DeRose EC, Koizumi M, Breaker RR. Generating new ligand-binding RNAs by affinity maturation and disintegration of allosteric ribozymes. RNA. 2001;7:524–536. [PMC free article] [PubMed]
  • Sowell SM, Norbeck AD, Lipton MS, Nicora CD, Callister SJ, Smith RD, Barofsky DF, Giovannoni SJ. Proteomic analysis of stationary phase in the marine bacterium Candidatus Pelagibacter ubique. Appl Environ Microbiol. 2008;74:4091–4100. [PMC free article] [PubMed]
  • Stoddard CD, Batey RT. Mix-and-match riboswitches. ACS Chem Biol. 2006;1:751–754. [PubMed]
  • Sudarsan N, Hammond MC, Block KF, Welz R, Barrick JE, Roth A, Breaker RR. Tandem riboswitch architectures exhibit complex gene control functions. Science. 2006;314:300–304. [PubMed]
  • Takusagawa F, Fujioka M, Spies A, Schowen RL. S-adenosylmethionine (AdoMet)-dependent methyltransferases. In: Sinnott M, editor. Comprehensive biological catalysis. Academic; San Diego, CA: 1998. pp. 1–30.
  • Tomsic J, McDaniel B, Grundy F, Henkin T. Natural variability in S-adenosylmethionine (SAM)-dependent riboswitches: S-box elements in Bacillus subtilis exhibit differential sensitivity to SAM in vivo and in vitro. J Bacteriol. 2008;190:823–833. [PMC free article] [PubMed]
  • Tripp HJ, Schwalbach MS, Meyer MM, Kitner JB, Breaker RR, Giovannoni SJ. Unique glycine-activated riboswitch linked to glycine–serine auxotrophy in SAR11. Environ Microbiol. 2009;11:230–238. [PMC free article] [PubMed]
  • Wachter A, Tunc-Ozdemir M, Grove BC, Green PJ, Shintani DK, Breaker RR. Riboswitch control of gene expression in plants by splicing and alternative 3′ end processing of mRNAs. Plant Cell. 2007;19:3437–3450. [PMC free article] [PubMed]
  • Wang JX, Breaker RR. Riboswitches that sense S-adenosylmethionine and S-adenosylhomocysteine. Biochem Cell Biol. 2008;86:157–168. [PubMed]
  • Wang JX, Lee ER, Morales DR, Lim J, Breaker RR. Riboswitches that sense S-adenosylhomocysteine and activate genes involved in coenzyme recycling. Mol Cell. 2008;29:691–702. [PMC free article] [PubMed]
  • Weinberg Z, Barrick JE, Yao Z, Roth A, Kim JN, Gore J, Wang JX, Lee ER, Block KF, Sudarsan N, et al. Identification of 22 candidate structured RNAs in bacteria using the Cmfinder comparative genomics pipeline. Nucleic Acids Res. 2007;35:4809–4819. [PMC free article] [PubMed]
  • Weinberg Z, Regulski EE, Hammond MC, Barrick JE, Yao Z, Ruzzo W, Breaker RR. The aptamer core of SAM-IV riboswitches mimics the ligand-binding site of SAM-I riboswitches. RNA. 2008;14:1–7. [PMC free article] [PubMed]
  • Welz R, Breaker RR. Ligand binding and gene control characteristics of tandem riboswitches in Bacillus anthracis. RNA. 2007;13:573–582. [PMC free article] [PubMed]
  • Wickiser JK, Winkler WC, Breaker RR, Crothers DM. The speed of RNA transcription and metabolite binding kinetics operate an FMN riboswitch. Mol Cell. 2005;18:49–60. [PubMed]
  • Winkler W, Nahvi A, Breaker RR. Thiamine derivatives bind messenger RNAs directly to regulate bacterial gene expression. Nature. 2002;419:952–956. [PubMed]
  • Winkler W, Nahvi A, Sudarsan N, Barrick J, Breaker RR. An mRNA structure that controls gene expression by binding S-adenosylmethionine. Nat Struct Biol. 2003;10:701–707. [PubMed]
  • Winkler WC, Nahvi A, Roth A, Collins JA, Breaker RR. Control of bacterial gene expression by a natural metabolite-responsive ribozyme. Nature. 2004;428:281–286. [PubMed]
  • Worden AZ, Lee JH, Mock TM, Rouzé P, Simmons MP, Aerts AL, Allen AE, Cuvelier ME, Derelle E, Everett MW, et al. Green evolution and dynamic adaptations revealed by genomes of the marine picoeukaryotes Micomonas. Science. 2009;324:268–272. [PubMed]
  • Yao Z, Barrick J, Weinberg Z, Neph S, Breaker RR, Tompa M, Ruzzo WL. A computational pipeline for high-throughput discovery of cis-regulatory noncoding RNA in prokaryotes. PLoS Comput Biol. 2007;3:e126. doi: 10.1371/journal.pcbi.0030126. [PMC free article] [PubMed] [Cross Ref]
  • Yarnell WS, Roberts JW. Mechanism of intrinsic transcription termination and antitermination. Science. 1999;284:611–615. [PubMed]

Articles from RNA are provided here courtesy of The RNA Society
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links