Although RNA is generally thought to be a passive genetic blueprint, some RNA molecules, called ribozymes, have intrinsic enzyme-like activity — they can catalyse chemical reactions in the complete absence of protein cofactors. In addition to the well-known small ribozymes that cleave phosphodiester bonds, we now know that RNA catalysts probably effect a number of key cellular reactions. This versatility has lent credence to the idea that RNA molecules may have been central to the early stages of life on Earth.
How life began on Earth is one of the great scientific mysteries. Molecular biologists have long suspected that RNA molecules were key to the process, in part because RNA has essential roles in a most fundamental process — protein synthesis — within all cells. The first example of an RNA molecule that forms a catalytic active site for a series of precise biochemical reactions was reported 20 years ago: the self-splicing pre-ribosomal RNA (rRNA) of the ciliate Tetrahymena. Although there was only one example, the word 'ribozyme' was coined for the general concept of an RNA molecule with enzyme-like activity1. The following year catalytic activity was discovered in the RNA component of a ribonucleoprotein enzyme, ribonuclease (RNase) P, providing the first example of a multiple-turnover enzyme using RNA-based catalysis2. These findings lent increased credibility to the hypothesis of an RNA world, where RNA served both as the genetic material and the principal cellular enzyme, probably assisted in the latter role by metal ions, amino acids and other small-molecule cofactors. The RNA world hypothesis posits that as cellular metabolism became more sophisticated, increasing demands on biocatalysts provided the impetus for the transition to protein enzymes. Descendants from this proposed RNA-dominated era inhabit today's world in the form of naturally occurring ribozymes present in organisms ranging from bacteria to humans (Table 1).
Although the known natural cellular and viral ribozymes catalyse only phosphodiester transfer chemistry, ribozymes obtained through in vitro selection techniques can exhibit the sort of biochemical sophistication necessary to support cellular metabolism. Starting with a pool of random RNA sequences, molecules possessing a desired activity are isolated through successive cycles of activity selection, reverse transcription of the 'winners' into DNA and amplification of those sequences by the polymerase chain reaction. This methodology has allowed identification of ribozymes that form a nucleotide from a base plus a sugar3, synthesize amide bonds4,5, form Michael adducts such as those involved in the methylation of uridine monophosphate to give thymidine monophosphate6, and form acyl-coenzyme A, which is found in many protein enzymes7. It is tantalizing to think that these ribozymes are analogues of missing links in a transition from an RNA world to contemporary biology (ref. 8, and see review in this issue by Joyce, pages 214–221). Because the structures and chemical mechanisms of in vitro-selected ribozymes are largely unknown at present, we focus here on the more extensively studied natural ribozymes.
How do RNA catalysts compare to their better-known protein enzyme counterparts? First, ribozyme rate enhancements can be substantial. For example, the rate constant for chemistry of the self-cleaving hepatitis delta virus (HDV) ribozyme is estimated at 102–104 s−1, which is close to the maximal cleavage rate of RNase A (1.4×103 s−1 at 25 °C). The Tetrahymena group I self-splicing intron also has a rate constant for its chemical step that is comparable to those of protein enzymes. Second, ribozymes can use cofactors such as imidazole during catalysis9,10, and they can be switched on and off by the binding of small-molecule allosteric effectors11. Finally, molecular structures have revealed that ribozymes, like protein catalysts, fold into specific three-dimensional shapes that can harbour deep grooves and solvent-inaccessible active sites. These tertiary structures facilitate catalysis in part by orienting substrates adjacent to catalytic groups and metal ions.
To facilitate chemical transformations, catalysts stabilize the transition state between substrate and product. Both protein and RNA catalysts may achieve this by adding or removing protons during a reaction, orienting substrates so that they are optimally positioned to react, and using binding interactions away from the reaction site to 'force' an unfavourable contact that is relieved in the transition state. Here we discuss the structural and chemical basis for RNA catalysis as determined for several of the naturally occurring ribozymes. The lack of diverse functional groups in RNA molecules and the propensity for RNA to bind metal ions led to early hypotheses that all ribozymes might act as metalloenzymes, positioning metal ions for direct roles in catalysis. This seems to be true for group I and group II self-splicing introns and RNase P, and in the Tetrahymena group I intron these metal ions and their specific functions have been identified. Surprisingly, however, those ribozymes that perform site-specific strand scission — the hammerhead, hairpin, Neurospora Varkud satellite (VS) and HDV ribozymes — may use diverse catalytic mechanisms, and none has emerged clearly as a metalloenzyme.
Site-specific RNA self-cleavage
The hammerhead, HDV, hairpin and VS ribozymes are small RNA structures of ∼40–160 nucleotides that catalyse site-specific self-cleavage (Table 1). Found in viral, virusoid or satellite RNAs, they process the multimeric products of rolling-circle replication into genome-length strands. Although the reaction catalysed by these ribozymes is the same as that of many protein RNases (Fig. 1a), they act only at specific phosphodiester bonds by using base-pairing and other interactions to align the cleavage site within the ribozyme active site. The evolutionary maintenance of these sequences may result from the relative simplicity and efficiency of RNA-catalysed RNA strand scission.
The hammerhead, at ∼40 nucleotides, is the smallest of the naturally occurring ribozymes, and mediates rolling-circle replication within circular virus-like RNAs that infect plants. Recent experiments show that the hammerhead motif (Fig. 1b) is the most efficient self-cleaving sequence that can be isolated from randomized pools of RNA, suggesting that it may have arisen multiple times during the evolution of functional RNA molecules12. Consisting of three short helices connected at a conserved sequence junction, the hammerhead catalyses site-specific cleavage of one of its own phosphodiester bonds via nucleophilic attack of the adjacent 2′-oxygen at the scissile phosphate (Fig. 1a).
The simplicity of the hammerhead secondary structure lent itself to the design of two-piece constructs in which the strand containing the cleavage site was separated from the rest of the self-cleaving RNA. By treating one strand as the substrate and the other as the enzyme, multiple-turnover cleavage occurred with a typical rate of 1 molecule per minute at physiological salt concentrations, consistent with a substantial 109-fold rate enhancement over the uncatalysed rate of nonspecific RNA hydrolysis. Initial studies revealed a requirement for a divalent metal ion for catalysis, leading to the idea that site-specific positioning of a metal ion such as magnesium might enable efficient deprotonation of the attacking 2′-hydroxyl nucleophile. Later studies at much higher ionic strength (4-M monovalent salt) showed that the hammerhead as well as the hairpin and VS ribozymes could react nearly as fast in the absence of divalent ions13. This discovery suggested two distinct possibilities: either these ribozymes use a different catalytic mechanism in the presence of high, non-physiological concentrations of monovalent salts, or the divalent metal ion requirement at low salt concentrations serves a structural rather than a purely chemical function.
The unveiling of the crystal structure of the hammerhead ribozyme in 1994, the first ribozyme structure to be determined, revealed a Y-shaped conformation in which nucleotides essential for catalysis were clustered at the junction of the three helical arms14 (Fig. 1c). Since then, additional crystal structures of the hammerhead ribozyme have provided 'snapshots' of the RNA at several steps along the catalytic reaction pathway: the initial pre-cleaved state, two sequential conformational changes that precede cleavage and rotate the scissile phosphate to be in line with the attacking 2′-oxygen nucleophile, and the post-cleavage product state15,16,17,18,19. Together these structures have led to a model of ribozyme cleavage involving precise positioning of the reactive groups by the structure of the ribozyme. However, it has been difficult to ascertain from these structures what the role of bound divalent metal ions might be. Although several divalent ions were identified unambiguously in the crystal structures, they were not situated close enough to the site of catalysis to support a direct role in RNA cleavage.
Site-specific substitution of phosphate oxygens with sulphur atoms, which is readily achieved by solid-phase synthetic methods, enabled direct analysis of the effects of disrupting divalent metal-ion binding sites that were potentially involved in catalysis20. These investigations led to evidence for the direct simultaneous coordination of a single metal ion by the scissile phosphate and a second phosphate oxygen located 20 Å away in the crystal structure21,22. It was proposed that the crystal structures might represent the 'ground state' conformation of the hammerhead ribozyme, and that prior to catalysis the RNA conformation changed significantly but transiently to bring the critical catalytic metal ion proximal to the cleavage site. More recently, molecular modelling and kinetic analysis of the hammerhead cleavage reaction in the presence of monovalent versus divalent salts support the idea that divalent metal ions are not essential to the catalytic step, but instead stabilize the active ribozyme structure19,23,24,25.
Whether hammerhead catalysis requires a global conformational change or merely a local rearrangement is not yet resolved. In either case, orientation of reactants within the ribozyme active site probably contributes significantly to the rate of site-specific strand scission in the hammerhead. This is achieved through the unique structure of this RNA motif, conferred by the secondary structure and the presence of multiple conserved nucleotides in the active site. Whether these nucleotides provide anything else, such as general acid–base catalysis, is unknown.
Hepatitis delta virus and hairpin ribozymes
The HDV and hairpin ribozymes catalyse the same chemical reaction as that of the hammerhead, and they are likewise responsible for cleaving intermediates generated during rolling-circle replication of the HDV and a plant virus satellite RNA, respectively. Crystal structures of these ribozymes showed that in each case the RNA forms an enclosed cleft in which strand scission takes place26,27. Furthermore, neither ribozyme seems to coordinate a divalent metal ion at the site of catalysis, but instead positions functionally essential nucleotides proximal to the substrate in a configuration suggesting the possibility of their direct role in catalysis.
The potential for RNA to use general acid–base chemistry during catalysis was unexpected. Functional groups within proteins that have pKa values near neutrality, and thus can donate or accept a proton readily under physiological conditions, can act as general acids or general bases to shuttle protons during enzyme catalysis. But the lack of functional groups within RNA with pKa values near physiological pH (6–7) means that for RNA to function in this way, one or more of its functional groups must have a pKa significantly shifted towards neutrality. In RNA, adenine and cytosine have the potential for protonation of their ring nitrogens, N1 and N3, respectively, but the pKa values for the free nucleosides are 3.5 and 4.2. However, A and C residues with substantially shifted pKa values have been detected in small functional RNAs, presumably owing to the structural environment of the nucleotide28,29,30,31,32.
In the HDV ribozyme, a cytosine base essential to catalytic function is positioned in a cleft adjacent to the site of cleavage in the RNA (Fig. 2a, b). A network of potential hydrogen bonds to this nucleotide is consistent with stabilization of a protonated form of the cytosine that might allow it to donate or accept a proton at some stage during catalysis26. This feature could be useful for mediating catalysis by pulling a proton off the attacking 2′-oxygen nucleophile, or by providing a proton to the 5′-oxygen leaving group (Fig. 1a). But does this in fact occur? Although detecting the movement of protons within an enzyme, or ribozyme, active site is difficult to achieve directly, several clever experiments have provided indirect evidence that such a mechanism is at work in the HDV ribozyme.
In one set of experiments, catalytic activity of ribozymes with point mutations at the critical cytosine was partially restored in the presence of imidazole, the side chain of histidine that readily accepts or donates a proton in many protein enzyme active sites9. Furthermore, the measured pKa values of a series of restored reactions correlated with those of the imidazole analogues that promoted cleavage in the mutants33. In a second series of experiments, kinetic isotope effects and correlation of reaction pKa with the pKa of different bases placed at the position of the critical C supported a direct role of this nucleotide in proton shuttling during catalysis34,35. However, the pKa of the active-site C is apparently transiently shifted during the reaction, as the shift is not detected within the product or precursor forms of the ribozyme36. Together, the current data support a model in which the C acts as a general acid during the reaction to donate a proton to the 5′-bridging oxygen (Fig. 2c). A hydrated metal ion coordinated near the ribozyme active site may abstract a proton from the 2′-hydroxyl nucleophile37 (Fig. 1a).
In the hairpin ribozyme the situation is less clear. Although the crystal structure of a precursor form of the RNA suggested that an active-site adenosine might adopt the role of a general acid or base27 (Fig. 3a, b), nucleotide analogue interference experiments have not provided evidence for a shifted pKa at this position38. However, the lack of a requirement for divalent metal ions during hairpin ribozyme cleavage implies that a metal ion-independent mechanism is at work39,40,41.
Group I introns
Group I introns have been found to interrupt genes for rRNA, transfer RNA (tRNA) and messenger RNA (mRNA) in far-reaching corners of biology, including the nuclei of protozoa, the mitochondria of fungi, the chloroplasts of algae, and bacteria and their phages. They are defined as a group by their common core secondary structure, consisting of an array of nine base-paired elements (P1–P9), and by their common mechanism of self-splicing. They accomplish splicing by a two-step transesterification mechanism initiated by an exogenous molecule of guanosine or guanosine triphosphate (GTP; Fig. 4a).
Because the excised intron still contains the active site for transesterification, it can be re-engineered to give a true catalyst that can cleave or ligate exogenous substrate molecules intermolecularly ('in trans'). The Tetrahymena version of this RNA enzyme, together with oligonucleotide substrates that can be synthesized with subtle chemical variations, has facilitated the dissection of the reaction pathway into its elemental steps. First, the RNA substrate base-pairs to an internal guide sequence within the ribozyme, forming the P1 helix. Second, specific ribose 2′-hydroxyl groups along the minor groove of the P1 helix promote its docking into the active site. Third, guanosine (or a 5′-phosphorylated analogue such as GTP) binds to the G site within P7. Fourth, the 3′-hydroxyl of the G acts as a nucleophile, cleaving the 5′ splice-site phosphate with inversion of configuration. Finally, products are released; the slow release of the product bound by base pairing plus tertiary interactions is rate limiting for multiple-turnover cleavage under typical conditions42.
Many protein enzymes that promote phosphoryl transfer reactions use metal ions for catalysis, and group I ribozymes use the same trick43. The number and location of the active-site metal ions has been investigated by substituting individual phosphate or ribose oxygen atoms with sulphur or with an amino group, and then testing for changes in metal-ion specificity (a procedure known as 'thiophilic metal-ion rescue'). The most highly supported current model is shown in Fig. 4b. One metal ion (MB) helps deprotonate the 3′-oxygen of the G nucleophile, while another (MA) stabilizes the developing negative charge on the leaving-group oxygen in the transition state. MC could assist in positioning the substrates with respect to one another and, along with MA, could help stabilize the trigonal, bipyramidal transition state44.
The enzyme mechanism of the group I introns — involving a nucleotide-binding site, nucleophilic attack and metal-ion catalysis — is commonplace in the world of protein enzymes. But clearly the ribozyme active site must be constructed differently, given the charged and hydrophilic nature of the nucleic acid building blocks and the limited diversity of their side chains compared to amino acids. So how might a catalytic active site be built out of ribonucleotides?
The first detailed view of active-site construction came with the crystal structure of the 160-nucleotide P4–P6 domain of the Tetrahymena intron, an attractive target because it folds into the same structure as an excised domain as it does in the context of the whole ribozyme. This structure revealed how long-range base triples and divalent cation-mediated structures can fold an RNA molecule into a globular structure with an interior that is relatively inaccessible to solvent45,46. The way in which this domain and the G-site-containing domain combine to create a concave active site was seen at modest resolution in the crystal structure of a 240-nucleotide active ribozyme47, a structure whose general features had been predicted by modelling based on comparative phylogenetic analysis48. Finally, the way the structure embraces the P1 substrate helix was modelled by mapping sites of chemical modification that perturb the reaction49. Future goals are to obtain a high-resolution crystal structure of an entire group I ribozyme with bound substrates, and to locate the proposed three catalytic metals within that structure.
Group II introns
Group II introns, found in bacteria and in organellar genes of eukaryotic cells, catalyse precise self-excision and ligation of the flanking RNA sequences to form a mature transcript. In a mechanism distinct from that of the group I introns, the group II reaction involves nucleophilic attack by the 2′-hydroxyl of a specific adenosine within the intron — the 'branch site' — to form a branched or lariat-type structure (Fig. 4a). Magnesium ions coordinated within the intron are thought to have a direct role in catalysis50,51,52, and several studies have revealed aspects of the intron tertiary structure that are essential to catalytic function53,54,55,56. Models of group II intron architecture have been proposed based on chemical probing, phylogenetic covariation and mutagenesis results57,58. Interestingly, some group II introns encode proteins that assist RNA splicing and can also enable efficient integration of the intron RNA into double-stranded DNA by reverse splicing and reverse transcription59. This reverse splicing activity promotes intron mobility by enabling insertion into targeted genes.
RNase P, found in all cells, catalyses site-specific hydrolysis of precursor RNA substrates including tRNA, 5S rRNA and the signal recognition particle RNA60,61. These substrates probably share structural features that enable efficient recognition by the RNase P substrate-binding site, positioning the reactive phosphate in each case for nucleophilic attack by a coordinated water molecule. The ribozyme is thought to be a metalloenzyme, a hypothesis supported by data from phosphorothioate substitution at the scissile phosphate of a substrate pre-tRNA62,63.
RNase P is in fact an RNA–protein complex whose activity, at least in bacteria, resides within the RNA component. Furthermore, it is a true catalyst in the sense that each ribozyme complex catalyses the cleavage of multiple substrate RNAs. As with group I and II introns, much of the information about the secondary and tertiary structure of RNase P RNA has come from extensive phylogenetic covariation analysis of related sequences. Typically 300–400 nucleotides in length, it comprises two domains containing the substrate-recognition site and the ribozyme active site, respectively. Structural models of a bacterial RNase P RNA have been proposed64,65, but a crystal structure is not yet available. In human cells, the RNase P complex is larger and includes multiple protein components in addition to the RNA66,67,68. The human RNase P RNA is not catalytically active in the absence of protein, which has made it challenging to determine whether its active site is composed of RNA, protein or some combination of the two.
Ribozyme activity, folding and dynamics
Protein enzymes that catalyse nucleophilic attack at a phosphate within RNA or a ribonucleotide apparently utilize different chemical mechanisms depending on the enzyme. For example, mammalian adenylyl cyclases function by a two-metal-ion mechanism69, RNase A uses two histidines for general acid–bases catalysis70, and the anthrax adenylyl cyclase exotoxin uses one histidine and a coordinated metal ion to activate the attacking nucleophile and stabilize the leaving group, respectively71. The fact that ribozymes also catalyse phosphodiester bond cleavage by a variety of mechanisms shows that RNA has a breadth of catalytic potential similar to protein enzymes.
Furthermore, like protein enzymes, ribozymes must fold into specific three-dimensional structures to function catalytically. How do these RNAs reach their active conformations? This problem has been the subject of intense study using in vitro systems, and several themes are emerging. RNA structures generally fold via a cooperative, hierarchical pathway in which tertiary interactions follow the formation of a stable secondary structure. Rates of tertiary structure formation vary from tens of milliseconds to several minutes, and for the large ribozymes, can be dominated by folding intermediates that transiently trap the RNA in non-native or partially folded conformations72,73,74,75. Single-molecule experiments, in which individual ribozyme molecules are analysed by fluorescence microscopy or mechanical tethering, reveal that multiple folding pathways can exist for a particular RNA sequence76,77,78. It is not yet clear how these observations relate to RNA folding processes in vivo, because proteins may assist ribozyme assembly either through direct RNA binding79,80 or through covalent modification of specific nucleotides. In addition, RNA tertiary structures and even secondary structures can undergo rapid conformational changes81,82,83,84,85. Such dynamics may be essential for progress through a catalytic cycle. On the other hand, too much structural flexibility may hamper catalysis, providing the incentive for evolution of ribonucleoprotein enzymes rather than pure ribozymes.
A starring role for ribozymes?
If the RNA world had a lengthy head-start over the protein catalyst world, why are RNA catalysts relatively minor players in modern cells? In fact, they may be much more central to cell biology than was previously believed. The ribosome, which is responsible for information-directed protein synthesis in all of life, is composed of three (or in some cases four) RNA molecules along with several dozen proteins. No protein subunit has ever been identified as a peptidyl transferase enzyme, and for more than 20 years, evidence for a primary role of RNA in this activity has accumulated86. The most direct evidence came with the deduction of the crystal structure of the large subunit87, in which the peptidyl transferase centre was precisely located by binding a small-molecule inhibitor that is an analogue of the anionic tetrahedral intermediate in amide bond formation88. Remarkably, only RNA and no protein lies in the vicinity of the reaction centre, so the catalysis must be ribozymic. The authors suggested one possible mechanism involving a conserved adenine acting as a general base to abstract an amino proton from the amino acid89, but subsequent mutagenesis of the key A has not provided strong support90,91. Identifying the rRNA's catalytic strategy is an important direction for future research.
Given the large size of the rRNA (Table 1), scientists are using in vitro evolution to find smaller peptidyl transferase ribozymes that might model the biological reaction92. Indeed, RNA catalysts have also been identified that accomplish two other steps of the protein synthesis pathway: formation of activated amino acid adenylates93 and transfer of an amino acid to the 3′-oxygen of a tRNA-like acceptor94,95.
Another ribonucleoprotein catalyst found in eukaryotic cells is the spliceosome, which assembles with nuclear pre-mRNAs and splices out the major class of introns (which are not self-splicing). For many years, a popular hypothesis has been that the spliceosome is also a molecular fossil from the RNA world — with several enzymatic RNAs acting intermolecularly via a mechanism analogous to that used intramolecularly by self-splicing group II introns96,97. As with the ribosome, most of the evidence for this hypothesis has been circumstantial until very recently.
Now Valadkhan and Manley98 have shown that two spliceosomal RNAs, the U2 and U6 small nuclear RNAs, can bind an RNA substrate containing the sequence of the intron branch site and promote a splicing-related reaction in the absence of any of the numerous spliceosomal proteins. The reaction product is not the natural 'branch' — consisting of a nucleotide forming both 2′-5′ and 3′-5′ phosphodiester bonds — but instead a new product consistent with a phosphotriester. (This is surprising from a chemical perspective, because it would require hydroxyl as the leaving group from a pentavalent phosphorous intermediate or transition state.) This new RNA-catalysed reaction will undoubtedly stimulate fresh investigations of the mechanism by which spliceosomal RNAs catalyse mRNA splicing, and adds weight to the proposition that remnants of the RNA world are still among us.
As some of the first RNAs to be studied in structural and mechanistic detail, ribozymes have provided many important insights into RNA function at a fundamental chemical level. Although much progress has been made, many interesting questions remain to be addressed. Determining detailed reaction mechanisms for RNA catalysts, including large RNA–protein complexes such as the ribosome and the spliceosome, will be a priority, as well as exploring the chemical mechanisms of ribozymes identified by in vitro selection. In addition to revealing new aspects of RNA biology, these investigations may shed light on aspects of the proposed RNA world and the role of RNA in early evolution.
We acknowledge D. Battle for extensive help with figure preparation, and V. Rath for comments on the manuscript.
About this article
Nature Reviews Chemistry (2017)