2.1 Riboswitches: gene regulation by small molecule-binding RNAs

Riboswitches are gene-regulatory messenger RNA (mRNA) domains that recognize small molecules without employing proteins, and modulate gene expression in cis. Binding to their specific ligands stabilizes a conformation of the ligand-binding domain (the ‘aptamer domain’) of the riboswitch that in turn affects the conformation of an RNA segment (the ‘expression platform’) that interfaces with the transcription, translation or RNA processing machinery that ultimately leads to the modulation of gene expression. Although the proposal that gene-regulatory RNAs may themselves be directly responsible for sensing the intracellular concentration of small molecules goes at least as far back as the idea of the ‘repressor’ (Jacob & Monod, Reference Jacob and Monod1961), and characterization of the T-box system demonstrated that gene regulatory RNAs can sense the aminoacylation state of tRNA (i.e. a small-molecule modification of an RNA) (Grundy & Henkin, Reference Grundy and Henkin1993), formal demonstration of the existence of riboswitches did not occur until the 21st century (Mironov et al. Reference Mironov, Gusarov, Rafikov, Lopez, Shatalin, Kreneva, Perumov and Nudler2002; Winkler et al. Reference Winkler, Nahvi and Breaker2002). At present, riboswitches have been discovered that respond specifically to more than a dozen different metabolites and one bacterial second messenger. Their structural and biochemical characterization has progressed apace (reviewed in Baird et al. Reference Baird, Kulshina and Ferré-D'amaré2010; Edwards et al. Reference Edwards, Klein and Ferré-D'amaré2007; Henkin, Reference Henkin2008; Serganov, Reference Serganov2009).

2.2 Catalytic activation of the glmS ribozyme by glucosamine-6-phosphate (GlcN6P)

Most riboswitches known to date have been discovered by bioinformatic approaches in which phylogenetically conserved elements are sought by aligning non-coding RNA sequences. This approach has been particularly successful in the discovery of riboswitches in bacterial mRNAs, whose 5′-untranslated regions (5′-UTRs) are short compared to UTRs and introns of typical eukaryotic transcripts. Once a candidate riboswitch is discovered, its cognate ligand has to be established. Often, the biochemical function of the structural genes that are adjacent in the genome, and therefore potentially under control of the candidate riboswitch, offers useful clues. By employing this heuristic, Barrick et al. (Reference Barrick, Corbino, Winkler, Nahvi, Mandal, Collins, Lee, Roth, Sudarsan, Jona, Wickiser and Breaker2004) discovered a conserved element that is widespread in Gram-positive bacteria upstream of the glmS gene. This gene encodes the protein glucosamine-6-phosphate synthetase (GlmS), an enzyme that catalyzes the conversion of fructose-6-phosphate and glutamine to GlcN6P and glutamate (reviewed in Milewski, Reference Milewski2002), the first committed step in the metabolic pathway that leads to synthesis of the bacterial cell wall. As expected from its function, glmS is an essential gene in the model organism Bacillus subtilis (Kobayashi et al. Reference Kobayashi, Ehrlich, Albertini, Amati, Andersen, Arnaud, Asai, Ashikaga, Aymerich, Bessieres, Boland, Brignell, Bron, Bunai, Chapuis, Christiansen, Danchin, Debarbouille, Dervyn, Deuerling, Devine, Devine, Dreesen, Errington, Fillinger, Foster, Fujita, Galizzi, Gardan, Eschevins, Fukushima, Haga, Harwood, Hecker, Hosoya, Hullo, Kakeshita, Karamata, Kasahara, Kawamura, Koga, Koski, Kuwana, Imamura, Ishimaru, Ishikawa, Ishio, Le Coq, Masson, Mauel, Meima, Mellado, Moir, Moriya, Nagakawa, Nanamiya, Nakai, Nygaard, Ogura, Ohanan, O'Reilly, O'Rourke, Pragai, Pooley, Rapoport, Rawlins, Rivas, Rivolta, Sadaie, Sadaie, Sarvas, Sato, Saxild, Scanlan, Schumann, Seegers, Sekiguchi, Sekowska, Seror, Simon, Stragier, Studer, Takamatsu, Tanaka, Takeuchi, Thomaides, Vagner, Van Dijl, Watabe, Wipat, Yamamoto, Yamamoto, Yamamoto, Yamane, Yata, Yoshida, Yoshikawa, Zuber and Ogasawara2003). Chemical probing of the glmS 5′ UTR RNA in the presence of GlcN6P did not reveal protection patterns diagnostic of RNA folding in the presence of the metabolite. Instead, a dramatic increase in the rate of scission of one particular phosphodiester bond of the presumed riboswitch domain was detected. Further characterization established that physiologic concentrations of GlcN6P activate a latent self-cleavage activity of the RNA domain that leads to specific scission of one internucleotide phosphodiester bond, with production of products bearing 2′,3′-cyclic phosphate and 5′-OH termini, analogous to those produced by the previously well-characterized hammerhead, hairpin, HDV and VS ribozymes. GlcN6P was not consumed in the reaction. Winkler et al. (Reference Winkler, Nahvi, Roth, Collins and Breaker2004) showed that, in vitro, it is possible to convert the glmS 5′-UTR into a GlcN6P-dependent multiple-turnover catalyst, that is, a ribozyme.

2.3 Regulation of gene expression by the glmS ribozyme

How does the GlcN6P-induced self-cleavage activity of the glmS ribozyme lead to the modulation of gene expression? In Gram-positive bacteria, intact mRNAs are capped by a 5′ triphosphate. Activation of the glmS ribozyme by high levels of GlcN6P releases a short (~25 nt) leader sequence that bears the triphosphate, and exposes a new 5′-OH group in the ribozyme domain. This hydroxyl group is recognized by RNase J1, an RNase conserved among bacteria, which degrades the mRNA. As the GlmS protein is unstable, the degradation of its mRNA results in down-regulation of the enzymatic activity, thus completing a negative-feedback loop (Fig. 1). The physiologic importance of the glmS ribozyme has been demonstrated by employing ribozyme–riboswitch mutants that lack GlcN6P-induced self-cleavage activity. B. subtilis strains harboring glmS genes downstream of such catalytically defective mutant RNA domains fail to sporulate under conditions that lead to sporulation of wild-type bacteria. Presumably, the differentiation program leading to sporulation requires a precise regulation of GlcN6P levels (Collins et al. Reference Collins, Irnov, Baker and Winkler2007). The ribozyme regulation of the GlmS activity appears to be restricted to Gram-positive bacteria, whose GlcN6P synthetase is not an allosterically regulated enzyme. The paralogous enzymes of Gram-negative bacteria and eukaryotes are allosterically regulated by GlcN6P; thus, the activity of the enzyme is controlled at the protein level (Milewski, Reference Milewski2002). In addition, it was recently found that the model Gram-negative bacterium Escherichia coli regulates expression of its glmS gene employing trans-acting non-coding RNAs that do not bind GlcN6P and are not catalytic (i.e. are neither riboswitches nor ribozymes) (Görke & Vogel, Reference Görke and Vogel2008; Reichenbach et al. Reference Reichenbach, Maes, Kalamorz, Hajnsdorf and Görke2008).

Fig. 1. Schematic representation of the mechanism of gene regulation by the glmS ribozyme. (a) The glmS mRNA comprises a triphosphate cap, a leader sequence (green) preceding the ribozyme domain (black) and a conventional Shine–Delgarno sequence (S–D) followed by the initiation codon for the open reading frame (ORF) encoding the protein GlmS. This enzyme catalyzes conversion of fructose-6-phosphate and glutamine into glutamate and GlcN6P. The latter serves as the starting point for the synthesis of the bacterial cell wall. (b) When GlcN6P accumulates cytoplasmically, it binds to the glmS ribozyme domain, activating a latent self-cleavage activity. This releases the leader sequence, exposing a new 5′-OH terminus in the cleaved mRNA. RNase J1 recognizes the 5′-OH, and degrades the glmS mRNA (Collins et al. Reference Collins, Irnov, Baker and Winkler2007).

3.1 A triply pseudoknotted ribozyme fold

From the biochemical standpoint, the most remarkable feature of the glmS ribozyme–riboswitch is its GlcN6P-induced catalytic activation. Depending on the precise experimental conditions employed, the change in the rate of cleavage between the inactive and fully activated ribozyme can approach 10⁷ (Brooks & Hampel, Reference Brooks and Hampel2009; Klein et al. Reference Klein, Wilkinson, Been and Ferré-D'amaré2007b). Deletion analyses demonstrated that RNAs comprising sequences spanning from one nucleotide upstream of the scissile phosphate, denoted as residue (−1) to approximately 75 nt downstream exhibit GlcN6P-dependent catalytic activity. However, RNAs comprising residues from (−1) to ~145 nt 3′ of the cleavage site exhibited maximal activity. Based on the comparative analysis of glmS ribozyme homologs from a number of bacteria, Barrick et al. (Reference Barrick, Corbino, Winkler, Nahvi, Mandal, Collins, Lee, Roth, Sudarsan, Jona, Wickiser and Breaker2004) and Winkler et al. (Reference Winkler, Nahvi, Roth, Collins and Breaker2004) proposed a secondary structure for the minimal core (residues −1 to 75) comprised of three conserved stem-loops, denoted as paired regions P1 through P3. These authors demonstrated biochemically that the nucleotides that form the loop (L1) that closes helix P1, which are not phylogenetically conserved, are not functionally important. Moreover, the glmS ribozyme can be split at L1, and the resulting bimolecular RNA constructs perform multiple-turnover catalysis in a GlcN6P-dependent manner. The additional sequences between residues ~75 and ~150, required for maximal activity, were proposed to fold into a long helix (P4) interrupted by a central bulge (Barrick et al. Reference Barrick, Corbino, Winkler, Nahvi, Mandal, Collins, Lee, Roth, Sudarsan, Jona, Wickiser and Breaker2004; Winkler et al. Reference Winkler, Nahvi, Roth, Collins and Breaker2004). Subsequent sequence and biochemical analyses led Wilkinson and Been (Wilkinson & Been, Reference Wilkinson and Been2005) to suggest the formation of a canonical H-type pseudoknot between nucleotides near the 3′ terminus of the glmS ribozyme and those in the loop capping helix P3.

In order to elucidate the mechanism of GlcN6P binding and ligand-dependent catalysis by the glmS ribozyme, we set out to determine its three-dimensional structure. As is customary in crystallographic structure determination, homologs from a number of different species were subjected to crystallization experiments. A ribozyme construct spanning nucleotides −1 through 145 of the glmS ribozyme from the thermophilic Gram-positive bacterium Thermoanaerobacter tengcongensis (Xue et al. Reference Xue, Xu, Liu, Ma and Zhou2001) produced highly ordered crystals (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2009). This crystallization construct is comprised of two RNA chains: a synthetic oligonucleotide that corresponds to residues (−1) to the tip of loop L1 and a longer strand (made by in vitro transcription) that includes residues from L1 to the end of the conserved ribozyme domain. The assembly of the crystallization employing a synthetic oligonucleotide that spans the scissile phosphate between residues (−1) and (+1) allowed a facile introduction of chemical modifications nearby to inhibit cleavage, analogous to previous studies on the hammerhead (Scott et al. Reference Scott, Finch and Klug1995) and hairpin (Rupert & Ferré-D'Amaré, Reference Rupert and Ferré-D'amaré2001) ribozymes. The structure of the ribozyme was solved by multiple isomorphous replacement (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006) and refined, ultimately against diffraction data extending to 1·7 Å resolution (Klein et al. Reference Klein, Wilkinson, Been and Ferré-D'amaré2007b) (Fig. 2). Subsequently, the structure of the glmS ribozyme from Bacillus anthracis (identical in sequence that from Bacillus cereus, which has been employed for some biochemical studies) was determined at ~2·5 Å resolution (Cochrane et al. Reference Cochrane, Lipchock and Strobel2007). As expected from the high degree of sequence conservation, the glmS ribozymes from T. tengcongensis and B. anthracis adopt virtually identical structures. As the precision of the atomic coordinates of the T. tengongensis structure is markedly higher, this structure shall be discussed hereafter and its numbering scheme employed throughout.

Fig. 2. Overall structure of the glmS ribozyme. (a) Schematic secondary structure of the glmS ribozyme from Thermoanaerobacter tengcongensis based on its crystal structure. Nucleotides that flank the scissile phosphate are colored red. Helices (or paired regions) are numbered according to convention. In the RNA construct employed for crystallization, there is a covalent break in loop L1. This construct comprises two RNA chains: the substrate (red and gray) and the ribozyme (blue). The core of the ribozyme is colored in cyan; the peripheral P4–P4.1 region in dark blue. Nucleotides that are >90% conserved throughout phylogeny are in upper case letters. Lines with embedded arrowheads denote connectivity. (b) Ribbon representation of the crystal structure, colored as in (a); the bound GlcN6P is yellow (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006). (c) View from the side opposite (b). Structure figures prepared with Ribbons (Carson, Reference Carson1997).

The crystal structure revealed that the glmS ribozyme is comprised not of four stem-loops and a pseudoknot, but of three parallel helical stacks (Fig. 2) comprising a total of three pseudoknots (Fig. 3). Parallel packing of the three helical stacks results in a molecule with approximate dimensions of 100×50×20 ų (the smallest dimension being that of the width of an RNA double helix). The minimal functional core of the ribozyme [red, gray and cyan in Fig. 2, residues (−1) to 73] is comprised of four helices P1, P2, P2.1 and P2.2. Three of these (P1, P2 and P2.2) stack coaxially, and this stack packs side-by-side with the short helix P2.1, thus forming a compact, contiguous structure. The four helices are connected by four strand crossovers that define two nested pseudoknots (Fig. 3a). The scissile phosphate and the bound GlcN6P lie at the center of this core structure, with the metabolite binding site in the major groove of P2.2. Consistent with this, the most highly conserved nucleotides of the glmS ribozyme are in P2.1, P2.2, and the non-helical segments connecting these two helices (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006).

Fig. 3. Three pseudoknots form the core of the glmS ribozyme (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006). (a) Ribbon representation of the core of the ribozyme shown from the same direction as Fig. 2c. Nucleobases that form Watson–Crick pairs are depicted as short cylinders. Nucleotides not forming such pairs are omitted for clarity. Arrows denote the direction of the polynucleotide chain. M and m denote the major and minor grooves, respectively. (b) Canonical H-type pseudoknot formed by helices P3 and P3.1. Dashed arrows indicate that the minor groove crossover is comprised of the entirety of helices P4 and P4.1.

3.2 Stabilization of the structure by a peripheral element

The peripheral domain comprised of the coaxial stack of helices P4 and P4.1 (dark blue in Fig. 2) is connected covalently to the ribozyme core through the predicted H-type pseudoknot formed by helices P3 and P3.1 (Figs 2 and 3b), and two sets of tertiary interactions: packing of the oblique A-minor stack against P2.1 (see below), and a class-I A-minor interaction (Nissen et al. Reference Nissen, Ippolito, Ban, Moore and Steitz2001) of the GNRA tetraloop capping the peripheral domain with the conserved C10-G31 base pair in P1. Association of the peripheral and core domains of the ribozyme buries a total of ~2500 Ų of solvent accessible surface area, and generates the most solvent-inaccessible interface in the RNA. The pseudoknotted helices P3 and P3.1 stack coaxially with P1, P2 and P2.2 to form the longest helical stack (~100 Å) of the ribozyme. An H-type pseudoknot is defined by two helices and three connecting loops (Aalberts & Hodas, Reference Aalberts and Hodas2005). In the glmS ribozyme peripheral domain, the first loop crosses the major groove of P3.1 and is comprised of a single nucleotide (U79). As in the majority of H-type pseudoknots (Klein et al. Reference Klein, Edwards and Ferré-D'amaré2009), the P3/P3.1 pseudoknot has a second loop of length zero. The third loop crosses the minor groove of P3 and is unusually long, being in effect comprised of the entire 48 nt P4/P4.1 coaxial stack (Fig. 3) (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006).

3.3 The oblique A-minor motif

The central portion of the P4/P4.1 stack, which includes five highly conserved adenosine residues, folds into an overwound helix that packs at an oblique angle against P2.1 (Fig. 4a). The conformation of this portion of the peripheral domain results in the purine nucleobases of three residues from each strand stacking on each other and projecting into the minor groove (Fig. 4b). Oblique (~70°) docking of this structure against the minor groove of P2.1 differs from a stack of canonical A-minor interactions (Nissen et al. Reference Nissen, Ippolito, Ban, Moore and Steitz2001) in which each adenosine contacts base and sugar atoms from a single base pair. Instead, A105 and G128 of the glmS ribozyme contact two, and A104 and A127 contact three consecutive base pairs of P2.1 (Fig. 4c). Although stacks of inclined adenosines that pack against the minor groove of a helix have been observed previously [e.g. in the thiamine pyrophosphate (TPP) riboswitch (Serganov et al. Reference Serganov, Polonskaia, Phan, Breaker and Patel2006)], the stack in the glmS ribozyme bridges both strands of the helix against which it packs. This oblique A-minor motif, which was first described in the glmS ribozyme (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006), has also been observed in the crystal structure of the Class-II S-adenosylmethionine riboswitch, in which loop nucleotides of an H-type pseudoknot contact the minor groove of a helix in a similar manner (Gilbert et al. Reference Gilbert, Rambo, Van Tyne and Batey2008). The P4/P4.1 peripheral domain of glmS riboswitches is quite variable across phylogeny (Roth et al. Reference Roth, Nahvi, Lee, Jona and Breaker2006), and it is possible that other structural solutions to the stabilization of the core domain exist in ribozymes from organisms distantly related to T. tengcongensis and B. anthracis.

Fig. 4. The oblique A-minor motif forms the interface between the core and the peripheral domains (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006). (a) Representation of the entire ribozyme, seen from the direction of P4 and P4.1. In this figure, all the nucleotides in the substrate strand are colored gray. Nucleobases of the ascending and descending strands of the oblique A-minor motif are filled in yellow and red, respectively. Note oblique angle of the region boxed in gold (dark blue) with respect to the core of the ribozyme (cyan). (b) Detailed view of the boxed region in panel (a). Black dashed lines denote hydrogen bonds between nucleobases. (c) Ball-and-stick representation of the oblique A-minor interaction formed between the peripheral domain nucleotides 104–106 and 127–129, and the core domain. Red spheres labeled “w” depict water molecules.

4.1 RNase A: a canonical protein nuclease

In discussing the catalytic mechanism of the glmS ribozyme, the mechanism of a protein nuclease and those of the other well-characterized natural self-cleaving ribozymes provide useful frames of reference (reviewed in Fedor, Reference Fedor2009; Ferré-D'Amaré & Scott, Reference Ferré-D'amaré and Scott2010; Lilley & Eckstein, Reference Lilley, Eckstein, Lilley and Eckstein2008; Raines, Reference Raines1998). RNase A is a protein nuclease that catalyzes RNA cleavage through the same overall chemical mechanism as the glmS ribozyme and other small self-cleaving RNAs (unlike the ribozymes, the protein enzyme hydrolyzes the cyclic phosphate in a subsequent step.) Although RNase A exhibits limited sequence specificity, it achieves a rate enhancement of ~10¹¹ over the uncatalyzed reaction. The principal factors responsible are schematized in Fig. 5a (Raines, Reference Raines1998). Like all enzymes, RNase A overcomes the entropic and steric penalties of bringing the reactants into the active site in a productive conformation through the binding energy. The transesterification reaction proceeds through a concerted S _N2 mechanism which requires that the 2′ ribose oxygen, the phosphorus and the 5′ ribose oxygen be aligned. Richards et al. (Reference Richards, Wyckoff, Carlson, Allewell, Lee and Mitsui1971) solved a crystal structure of the enzyme bound to a dinucleotide mimic inhibitor (made non-cleavable by replacing the 5′-oxo leaving group with a methylene) and discovered that RNase A splays apart the nucleotides flanking the scissile phosphate to achieve a nearly linear arrangement of the three reactive atoms. [This angle, denoted τ (Soukup & Breaker, Reference Soukup and Breaker1999), is typically ~150°, not 180°, in inhibitors bound to crystal structures, which represent the ground state – greater τ angles require that atoms be brought closer than their van der Waals radii.] This structure revealed the location of two catalytic histidines and one lysine. His12 functions as a general base catalyst, deprotonating the 2′-OH, and His114 functions as a general acid, protonating the 5′-oxo leaving group. The reaction proceeds through a trigonal bipyramidal oxyphosphorane transition state whose excess negative charge is stabilized by the ammonium group of Lys41. As predicted by this mechanism, mutating these catalytic amino acids greatly impairs RNase A (Fig. 5a).

Fig. 5. Comparison of the active sites and mechanisms of RNase A and the hairpin ribozyme. (a) Active site of RNase A bound to the dinucleotide mimic inhibitor UpcA (Richards et al. Reference Richards, Wyckoff, Carlson, Allewell, Lee and Mitsui1971). In this compound, the atom corresponding to the 5′-oxo leaving group has been replaced with a methylene group. The degree of impairment of the enzyme when the catalytic histidines are replaced with alanines, or the catalytic lysine with cysteine (Raines, Reference Raines1998) is indicated. τ is the angle formed between the nucleophile, electrophile and the leaving group of the transesterification reaction. (b) Active site of the hairpin ribozyme (Rupert et al. Reference Rupert, Massey, Sigurdsson and Ferré-D'amaré2002). The cleavage reaction was inhibited by replacing the hydrogen atom of the nucleophile 2′-OH by a methyl group (magenta). The degree of impairment of the ribozyme when the catalytic nucleotides are replaced by abasic substitutions (Kuzmin et al. Reference Kuzmin, Da Costa, Cottrell and Fedor2005; Lebruska et al. Reference Lebruska, Kuzmine and Fedor2002) is indicated.

4.2 The small, self-cleaving ribozymes

The hammerhead, hairpin, HDV and VS ribozymes all adopt different three-dimensional structures. Crystal structures have been determined for the first three (Ferré-D'Amaré et al. Reference Ferré-D'amaré, Zhou and Doudna1998; Pley et al. Reference Pley, Flaherty and Mckay1994; Rupert & Ferré-D'Amaré, Reference Rupert and Ferré-D'amaré2001; Scott et al. Reference Scott, Finch and Klug1995), while molecular models for the VS ribozyme have been developed based on biochemical characterization, fluorescence resonance energy transfer (FRET) and small-angle X-ray scattering (SAXS) data (Lilley, Reference Lilley2004). The hammerhead, hairpin and VS ribozymes are organized around multi-helical junctions (Lilley, Reference Lilley1999), and differ in this fundamental manner from the glmS ribozyme. The hammerhead ribozyme folds around a three-helix junction, and its overall Y-shaped architecture (de la Peña et al. Reference De La Peña, Dufour and Gallego2009) is stabilized by loop–loop interactions distal to the junction (Martick & Scott, Reference Martick and Scott2006). The active site of the hairpin ribozyme results from a precise juxtaposition of the minor grooves of two irregular helical stems (Rupert & Ferré-D'Amaré, Reference Rupert and Ferré-D'amaré2001). Association of the two helices is facilitated by a four-helix junction from which they radiate (Murchie et al. Reference Murchie, Thomson, Walter and Lilley1998). The VS ribozyme is larger than the hammerhead or hairpin ribozymes (being ~150 nt long) and its structure comprises two separate three-helix junctions (Lipfert et al. Reference Lipfert, Ouellet, Norman, Doniach and Lilley2008). All three ribozymes bind their substrate through canonical Watson–Crick base-pairing to nucleotides both 5′ and 3′ to the cleavage site (although not to the nucleotides immediately adjacent to the scissile phosphate). This base-pairing specifies the cleavage site, and has the consequence that the substrate RNA is presented to the ribozyme in the context of an RNA duplex. Because of this, both products of the RNA cleavage may remain associated with the ribozyme through base-pairing, allowing the reverse ligation reaction to be catalyzed by the hammerhead, hairpin and VS ribozymes. The efficiency of the cleavage and ligation reactions (i.e. the internal equilibrium) depends on the specifics of the RNA constructs and solution conditions. The HDV ribozyme differs from the hammerhead, hairpin and VS ribozymes in that its structure is organized not around a helical junction, but by a double pseudoknot. As in the case of the core of the glmS ribozyme, the four helical crossovers that define its fold allow close parallel packing of two helical stacks, with the active site lying in the interhelical interface. Nonetheless, the glmS and HDV ribozyme folds are completely different: the scissile phosphate faces the major groove of P2.1 in the glmS ribozyme and the minor groove of P3 in the HDV ribozyme. Unlike the ribozymes organized around helical junctions, the HDV and glmS ribozymes only base pair with their substrates 3′ to the site of cleavage. As a result, the 5′ product of the cleavage reaction (which bears a 2′,3′ cyclic phosphate) readily dissociates from the ribozymes after scission, and neither ribozyme catalyzes the ligation reaction (reviewed in Ferré-D'Amaré & Scott, Reference Ferré-D'amaré and Scott2010).

4.3 Nucleobase participation in ribozyme catalysis

In large measure, the catalytic power of RNase A derives from its use of histidine side-chains, with pK _a's close to neutrality, and a lysine side-chain, whose side-chain amino group is positively charged at physiologic pH (Fig. 5a). RNA has no functional groups with analogous ionization properties. The groups with pK _a's closest to neutrality are the N1 nitrogens of purines (3·5 and 9·2 for A and G, respectively) and the N3 nitrogens of pyrimidines (4·2 and 9·2 for C and U, respectively). As a polyanion, however, RNA binds cations, and in principle, these can function either as electrostatic catalysts, as Lewis acids that perturb the pK _a of RNA functional groups bound to them, or as Brønsted–Lowry acids that lower the pK _a of the cation-coordinated waters. Thus, RNA-bound cations might assist in catalysis by providing a localized positive charge, or by providing hydroxide or hydronium ions to function as reactants or specific acid or base catalysts. As the Group I intron has been demonstrated to employ tightly bound Mg²⁺ ions for electrostatic transition-state stabilization and as Lewis acids, it was widely believed that all catalytic RNAs would employ bound metal ions as cofactors (Pyle, Reference Pyle1993). It is now known that this is not the case for the small self-cleaving ribozymes.

The hammerhead, hairpin and VS ribozymes are fully active in the presence of high concentrations of monovalent ions alone, or if Mg²⁺ is replaced with cobalt (III) hexammine. The latter is an isoster of hexahydrated Mg²⁺. However, unlike the waters of hydration of Mg²⁺, the amine ligands of cobalt hexammine do not dissociate, and therefore this complex ion cannot function as a Lewis acid or perturb the pK _a of hydration waters (Cowan, Reference Cowan1993). Thus, the activity of the hammerhead, hairpin and VS ribozymes implies that RNA functional groups are catalytic. The structure of the hairpin ribozyme (Fig. 5b) shows that the bound substrate is distorted to τ~158°, and three conserved nucleobases are positioned in a manner reminiscent of catalytic amino acid residues in the active site of RNase A. Thus, G8 and A38 may function as general base and acid catalysts, respectively, in the cleavage reaction (their roles would be reversed for the ligation reaction), and A9 may provide an electrostatic stabilization of the transition state. As the total rate enhancement achieved by the hairpin ribozyme (~10⁶) is modest compared to that of RNase A, individual deletion of these nucleobases, while detrimental to catalysis, does not have effects of nearly the same magnitude as individual mutation of RNase A active site residues (Fig. 5a). Nonetheless, the effect of deletions is approximately additive as in the case of the protein enzyme (Fig. 5b).

The HDV ribozyme, which can achieve rate enhancements of nearly 10¹⁰, supplements a catalytic nucleobase with a catalytic metal ion (Nakano et al. Reference Nakano, Chadalavada and Bevilacqua2000). The highly negatively charged environment of the active site of the HDV ribozyme perturbs the pK _a of the N3 group of the catalytic C75 residue of the RNA, raising it from 4·2 to ~7. Experiments in which the stability of the leaving group of the cleavage reaction was altered demonstrate that C75 functions as a general acid catalyst (Das & Piccirilli, Reference Das and Piccirilli2005). While this nucleobase provides the bulk of the rate enhancement achieved by the HDV ribozyme, the ribozyme activity decreases in the absence of divalent cations. This is because a hydrated cation in the active site facilitates cleavage by delivering water with a perturbed pK _a that functions as a specific base catalyst.

5.1 Allosteric activator or coenzyme function for GlcN6P?

In principle, activation of the glmS ribozyme by GlcN6P can occur through two different mechanisms. Binding of GlcN6P to the RNA can drive a conformational change of the RNA; that is, GlcN6P might function as an allosteric activator. Alternatively, GlcN6P may bind in the active site of the ribozyme, and provide a catalytic functional group, that is, function as a coenzyme. Although the two mechanisms are not mutually exclusive, the available biochemical and structural evidence indicate that GlcN6P is a coenzyme of the glmS ribozyme (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006; McCarthy et al. Reference Mccarthy, Plog, Floy, Jansen, Soukup and Soukup2005).

5.1.4 Active site structure

What is the organization of the glmS ribozyme active site, and if GlcN6P is a coenzyme, how does it complement it? The nucleotides flanking the scissile phosphate, (A−1) and (G+1) traverse the thin dimension of the core of the ribozyme in a conformation that brings the reactive groups of the substrate into a near-in-line conformation (τ=155° for the 2′-aminoribose structure). The nucleobase of (A−1) makes a hydrogen bond with that of G65. An adenosine at the (−1) position is conserved among glmS ribozymes. While the single hydrogen bond would not specify an A at this position, replacing a G into this position would result in steric clash with G65 (Fig. 6a). The scissile phosphate is held in place by hydrogen bonds from G39 and G65. Nucleotide (G+1) projects onto the other face of the ribozyme, where its nucleobase forms the roof of the GlcN6P-binding pocket (Fig. 6b). Although the nucleobase of (G+1) does not make any hydrogen bonds with the rest of the RNA, its position is defined by its stacking underneath the base of A35, and the sharp bend in the phosphate-ribose backbone following it. This bend results from the formation of canonical Watson–Crick pairs in P2.2 starting with residue C2, and appears to be stabilized by a tightly bound magnesium ion that coordinates the phosphates of C2, C36 and G37 (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006).

Fig. 6. Active site of the glmS ribozyme. (a) View from the direction of Fig. 2b. A partially dehydrated hexacoordinate magnesium ion is depicted as a black sphere, linked to its coordinating atoms with thin black lines. Note the position of the 5′-OH of nucleotide (−1). Nucleotides upstream of the ribozyme domain in the natural transcript would approach the active site from this direction. For this crystal structure, the ribozyme was prevented from cleaving the substrate by replacing the ribose of nucleotide (−1) with a 2′-deoxyribose. (b) View from the direction of Fig. 2c. The GlcN6P-binding pocket includes several well-ordered magnesium ions (black spheres), and is extensively hydrated. Only selected water molecules are shown for clarity.

5.2.1 GlcN6P recognition

The activator, GlcN6P, binds to an open, hydrated binding site that is lined with phylogenetically conserved residues, burying ~80% of its solvent-accessible surface area (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006). Nucleotides that line the ligand-binding pocket and active site are highly conserved phylogenetically. In addition to (A−1) and (G+1), nucleotides conserved in more than 90% of glmS ribozyme sequences include residues 35–40 and 64–65 on the side opposite to the GlcN6P-binding pocket, and residues 2, 35–36, 50–53 and 59–65 in the GlcN6P pocket (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006). Although GlcN6P exists as a mixture of the α and β anomers in solution in a ratio of ~40:60, with less than 1% in the acyclic conformation (Lim et al. Reference Lim, Grove, Roth and Breaker2006), it binds to the RNA exclusively in the α-axial anomeric conformation. The interactions made by GlcN6P are summarized in Fig. 7a. Studies of the ability of non-cognate small molecules to activate the glmS ribozyme point to the importance of the stereochemical arrangement of the anomeric OH and the amine at position 2. Thus, L-serine, whose side-chain OH and amine groups are in the same stereochemical relationship as the anomeric OH and amine groups of GlcN6P (Fig. 7b) is an activator of the glmS ribozyme, but D-serine is not (McCarthy et al. Reference Mccarthy, Plog, Floy, Jansen, Soukup and Soukup2005). Non-chiral activators of the ribozyme all have vicinal amine and hydroxyl groups. Thus, Tris (Fig. 7b) and ethanolamine are activators, but neither methylamine nor ethanol are (McCarthy et al. Reference Mccarthy, Plog, Floy, Jansen, Soukup and Soukup2005).

Fig. 7. Recognition of GlcN6P and analogs by the glmS ribozyme (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006). (a) Schematic depiction of interactions responsible for the recognition of GlcN6P. Hydrogen bonds are shown as two-headed arrows. Curved lines depict stacking of the nucleobase of G1 on the sugar. Numbering scheme for the sugar is indicated. (b) L-serine and Tris function as weak activators of the glmS ribozyme in vitro; D-serine does not (McCarthy et al. Reference Mccarthy, Plog, Floy, Jansen, Soukup and Soukup2005). The three molecules are drawn to emphasize their similarity to pyranose ring positions 1 and 2 of GlcN6P.

The functionally critical anomeric OH and amine groups of GlcN6P each make three hydrogen bonds. The anomeric OH hydrogen bonds to the N1 and N2 groups of G65 and G66, respectively, and to non-bridging oxygen of the scissile phosphate (Figs 6a and 7a). The amine hydrogen bonds to the 5′ oxygen of (G+1) (the leaving group of the cleavage reaction), to O4 of U51, and to a well-ordered water molecule seen in all high-resolution structures (Figs 6b and 7a). The OH group in position 3 of GlcN6P hydrogen bonds to both the 2′-OH of A50 and to non-bridging phosphate oxygen of A50. The OH group at position 4 of GlcN6P only makes water-mediated contacts to the RNA, and the pyranose ring oxygen accepts a hydrogen bond from the exocyclic amine of C3. The N1 imine of residue (G+1) of the glmS ribozyme donates a hydrogen bond to one of the non-bridging oxygen of the phosphate of GlcN6P. In addition, the nucleobase of (G+1) stacks on the sugar (Fig. 7a) (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006; Klein et al. Reference Klein, Wilkinson, Been and Ferré-D'amaré2007b). Most of these interactions between the ribozyme and GlcN6P are also observed in the crystal structure of the B. anthracis RNA (except for the water-mediated interactions, not resolved due to the low resolution of that structure) (Cochrane et al. Reference Cochrane, Lipchock and Strobel2007).

The importance of the interaction between the N1 imine of (G+1) and the phosphate of GlcN6P for ribozyme function was explored by substituting this RNA residue with a series of natural and unnatural purine nucleotides and comparing the rates of the cleavage of the variant RNAs when activated by either GlcN6P or GlcN (Klein et al. Reference Klein, Wilkinson, Been and Ferré-D'amaré2007b). With the wild-type ribozyme bearing a guanine at position (+1), GlcN6P is the stronger activator. This preference for the phosphate-bearing activator is maintained when (G+1) is replaced with inosine, another purine whose N1 position is protonated at neutral pH. However, when (G+1) is replaced with a purine whose N1 position is not protonated at neutral pH, such as A, 2-aminopurine, or 6-dimethyladenine, the selectivity of the ribozyme switches. These ribozymes are more active in the presence of GlcN than in the presence of GlcN6P. These observations were interpreted to mean that if the N1 imine of the purine residue at position (+1) of the glmS ribozyme is not protonated, then steric clash with the phosphate of GlcN6P is sufficiently detrimental to ribozyme activity that GlcN, not having the phosphate, is a better activator. In these studies, the variant ribozymes with mutations or non-natural purines at position (+1) were all less active than the wild-type, regardless of whether GlcN or GlcN6P was employed for activation. This indicates that aspects other than the hydrogen bond between the N1 imine and the phosphate of GlcN6P (or the relief of steric clash with the phosphate) are important for ribozyme function (Klein et al. Reference Klein, Wilkinson, Been and Ferré-D'amaré2007b).

5.3 A role for GlcN6P in proton transfer

The fundamental biochemical phenomenon that any proposed catalytic mechanism for the glmS ribozyme must explain is the complete lack of measurable activity of the ribozyme in the absence of GlcN6P. Given the compelling biochemical and structural evidence for a pre-organized active site in which the substrate is distorted into a near-in-line conformation prior to GlcN6P binding, the most parsimonious model is that the small molecule provides an essential catalytic functional group. The location of GlcN6P in the active site of the glmS ribozyme is reminiscent of those of Lys41 and His119 in the active site of RNAe A (Figs 5a and 8a). In solution, the pK _a of the amine of GlcN6P is 8·2 (the pKa of the phosphate is ~6·2) (McCarthy et al. Reference Mccarthy, Plog, Floy, Jansen, Soukup and Soukup2005); therefore, free GlcN6P will mostly exist in its ammonium form under physiologic conditions. Since the ribozyme is strongly negative, the ammonium form of GlcN6P would appear to be a better ligand for the RNA than the amine form. Binding of the ammonium form of GlcN6P to a pre-organized glmS ribozyme would place a positive charge in van der Waals contact with the scissile phosphate, analogous to the ammonium group of the lysine side-chain of RNase A. Thus, GlcN6P could function as an electrostatic catalyst. Moreover, the ammonium group would also hydrogen bond to the 5′-oxo leaving group of the cleavage reaction, and may be able to stabilize it by donating a proton, that is, by carrying out general acid catalysis, analogous to His 199 of RNase A (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006).

Fig. 8. Hypothetical catalytic mechanisms of the glmS ribozyme (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006). (a) GlcN6P deprotonates the 2′-OH nucleophile through a proton relay employing two tightly bound water molecules, functioning as a general base catalyst. The resulting ammonium form of GlcN6P functions to stabilize the transition state electrostatically and also donates a proton to the 5′-oxo leaving group, functioning as a general acid catalyst. (b) The N1 imine of G40 functions as a general base catalyst. GlcN6P, in its ammonium form, functions as an electrostatic and general acid catalyst. The ribozyme is inactive in the absence of GlcN6P or if a G40A mutation is made. The degree of impairment of the ribozyme indicated in parentheses is therefore an estimate based on the in vitro cleavage rate.

The analogy to RNase A can be extended further if it is postulated that the amine form of GlcN6P binds to the ribozyme. The active site of the ribozyme is heavily hydrated and two well-ordered water molecules are seen in all high-resolution structures (Fig. 6) that might form a proton relay connecting the amine of GlcN6P to the 2′-OH of (A−1). Thus, the amine of GlcN6P might be able to deprotonate the 2′-OH indirectly, functioning as a general base catalyst (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006). Once this has happened, GlcN6P would exist in the ammonium form, and would be able to carry out electrostatic and general acid catalysis as proposed above (Fig. 8a). While it might appear fanciful, this proposed mechanism has two attractive features. First, it is consistent with the observation that the RNA itself, even though pre-organized, is completely inactive in the absence of GlcN6P. Therefore, a mechanism in which GlcN6P is responsible for all the catalytic chemistry is parsimonious. Second, mechanisms involving a proton relay through water molecules have been proposed for a number of protein enzymes, for instance, some DNA polymerases (Wang et al. Reference Wang, Yu, Hu, Broyde and Zhang2007; Wang & Schlick, Reference Wang and Schlick2008), the F₁-ATPase (Dittrich et al. Reference Dittrich, Hayashi and Schulten2008) and some β-lactamases (Damblon et al. Reference Damblon, Raquet, Lian, Lamotte-Brasseur, Fonze, Charlier, Roberts and Frère1996).

Analysis of the activity of the glmS ribozyme as a function of pH also supports a role in proton transfer for GlcN6P. The rate of the ribozyme was found to increase linearly with pH and plateau above pH ~8 (Winkler et al. Reference Winkler, Nahvi, Roth, Collins and Breaker2004), with a pK _a of 7·8 (McCarthy et al. Reference Mccarthy, Plog, Floy, Jansen, Soukup and Soukup2005). If the pK _a of the reaction corresponds to that of the amine of GlcN6P, this indicates that binding to the ribozyme lowers the pK _a of the coenzyme by ~0·5 pH units, potentially making it a better general acid–base catalyst. These authors also compared the apparent pK _a of the cleavage reaction when the ribozyme is activated by either GlcN or serinol, and found them to be 7·9 and 8·7, respectively. Free in solution, GlcN and serinol have pK _a's of 7·8 and 8·8, respectively. That the reaction pK _a tracks closely that of the activator employed is evidence that an ionization event involving the amine group of the activator is part of the rate-limiting step of the reaction. A similar argument has been made in support of the participation of a nucleobase in general acid–base catalysis in the HDV ribozyme, where mutation of the proposed catalytic residue C75 to an A changed the pK _a of the reaction from 6·1 to 5·7. Since the pK _a's of the unperturbed N3 of C and N1 of A are 4·2 and 3·5, respectively, this suggests that C75 is a general acid–base catalyst, and that the local environment of C75 in the ribozyme active site raises its pK _a by ~2 pH units (Nakano et al. Reference Nakano, Chadalavada and Bevilacqua2000; Perrotta et al. Reference Perrotta, Shih and Been1999).

5.4 A role for G40 in catalysis

G40 is highly conserved in all glmS ribozymes, and crystal structures show that its nucleobase is adjacent to the ribose of (A−1), with the N1 imine of the guanine ~3·2 Å from the nucleophilic 2′-OH (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006). G40, therefore, can be postulated to function as a general base, deprotonating the nucleophile, analogous to His12 in RNase A and G8 in the hairpin ribozyme (Fig. 5). Then, GlcN6P can function as an electrostatic catalyst, a general acid, or both (Fig. 8b). A fundamental problem with such a model is that in the absence of GlcN6P, the glmS ribozyme is completely inactive, even though the substrate is in a reactive conformation and G40 is in position. This contrasts greatly from the behavior of other catalysts. Mutation of either of the catalytic histidines of RNA A lowers the activity of the enzyme several hundred to several thousand-fold, but the mutant enzymes are still active (Fig. 5a). Likewise, the deletion of either of the general acid–base catalysts of the hairpin ribozyme lowers the activity of the ribozyme several hundred- to several thousand-fold, but the ribozymes with the abasic substitutions also retain activity (Fig. 5b).

The importance of G40 was evaluated by mutagenesis. A G40A mutant of the glmS ribozyme was found to be completely inactive in the presence of saturating concentrations of GlcN6P. Indeed, the rate of cleavage measured for this mutant was indistinguishable from that of the spontaneous degradation of RNA under the experimental conditions. In principle, this result could simply indicate that the mutant RNA cannot fold correctly, or bind GlcN6P productively. However, the mutant RNA could be crystallized readily. Its structure was solved bound to an all-ribose (cleavable) substrate in the presence of GlcN6P. The crystal structure revealed that the substrate is uncleaved, and that GlcN6P is bound in precisely the same location as in the wild-type ribozyme. Residue 40 is nearly in the same position in the mutant as in wild-type: the distance between the N1 imine of A and the 2′-OH of (A−1) is 4·1 Å instead of 3·2 Å as observed in the wild-type (Klein et al. Reference Klein, Been and Ferré-D'amaré2007a).

As the N1 of G40 is near the nucleophile (the 2′-OH of A−1), it might hydrogen bond to it in the transition state, helping to orient it. Previously, vanadate was employed to mimic the transition state of the hairpin ribozyme (Rupert et al. Reference Rupert, Massey, Sigurdsson and Ferré-D'amaré2002). Crystal structures showed that ribozyme makes more hydrogen bonds to the transition state than to either the precursor or product state. Presumably because the glmS ribozyme does not bind tightly to the 5′ cleavage product (i.e. A−1), it could not be successfully crystallized with vanadate serving as a transition state mimic. Instead, structures were solved of the wild-type and G40A T. tengcongensis glmS ribozymes bound to an RNA strand in which (A−1) and (G+1) are linked through a 2′–5′ phosphodiester linkage. This type of linkage has been proposed to mimic the transition state of the hairpin ribozyme (Torelli et al. Reference Torelli, Krucinska and Wedekind2007). The structures of the glmS ribozyme bound to such 2′–5′ linked transition-state mimics did not show any additional interactions between G40 (or A40) and the substrate RNA (Klein et al. Reference Klein, Been and Ferré-D'amaré2007a). Thus, the available structural information does not support a role for G40 in preferential hydrogen bonding to the transition state.

The inactivity of the G40A mutant is surprising for three reasons. First, if GlcN6P is responsible for all catalysis in the wild-type, its presence in the active site of the mutant structure implies that the mutation should not abrogate all (or any) ribozyme activity. Second, if the N1 nitrogen of the purine at position 40 functions as a general base catalyst, then A, whose unperturbed pK _a is 3·2 (and is therefore unprotonated at physiologic pH) would be better suited than a G, whose unperturbed pK _a is 9·2 (and is protonated at physiologic pH). Third, although the 0·9 Å difference in distance between the N1 imine of G40A and wild-type to the 2′-OH nucleophile can well lead to a decrease in activity, it would be surprising if it led to complete inactivation of the catalyst. Thus, this experiment appears to have uncovered interdependency in the catalytic activity of GlcN6P and G40: neither is functional in the absence of the other. This is even more surprising in the light of the structure of the active site (Fig. 6): the closest approach of GlcN6P and residue 40 is 7 Å, measured from the amine of GlcN6P to the N1 imine of G40. These two functional groups which appear to interact functionally in the transition state are, in fact, on opposite sides of the scissile phosphate. How the active site of the glmS ribozyme modulates the ability of GlcN6P to function as a catalyst (and vice versa) is a question that remains unanswered.

5.5 pH dependence of GlcN6P binding

Another unexplained feature of the glmS ribozyme is that the affinity of the ribozyme for GlcN6P increases with pH. This increase in affinity is not sufficient entirely to explain the increase in rate with pH (Cochrane et al. Reference Cochrane, Lipchock and Strobel2007; McCarthy et al. Reference Mccarthy, Plog, Floy, Jansen, Soukup and Soukup2005), but is nonetheless very apparent, even in crystallization experiments. Thus, if the glmS ribozyme is crystallized at pH 5·5, GlcN6P binding is not observable crystallographically (even though the ribozyme is active in the crystal at that pH) (Klein & Ferré-D'Amaré, Reference Klein and Ferré-D'amaré2006), but at pH 8·5 electron density for the bound coenzyme is readily apparent (Klein et al. Reference Klein, Wilkinson, Been and Ferré-D'amaré2007b). The phenomenon is likely not due to titration of the phosphate of GlcN6P (which could lead to weaker binding to Mg²⁺ ions, and decrease the through-cation binding of the phosphate of GlcN6P to the ribozyme) as suggested by Cochrane et al. (Reference Cochrane, Lipchock, Smith and Strobel2009), because Glc6P binds to the ribozyme comparably at low or high pH. Therefore, it appears that the glmS ribozyme has a ‘selectivity filter’ that allows it preferentially to bind to the amine form of GlcN6P over the ammonium form.