RNA misfolding and the action of chaperones

Abstract

RNA folds to a myriad of three-dimensional structures and performs an equally diverse set of functions. The ability of RNA to fold and function in vivo is all the more remarkable because, in vitro, RNA has been shown to have a strong propensity to adopt misfolded, non-functional conformations. A principal factor underlying the dominance of RNA misfolding is that local RNA structure can be quite stable even in the absence of enforcing global tertiary structure. This property allows non-native structure to persist, and it also allows native structure to form and stabilize non-native contacts or non-native topology. In recent years it has become clear that one of the central reasons for the apparent disconnect between the capabilities of RNA in vivo and its in vitro folding properties is the presence of RNA chaperones, which facilitate conformational transitions of RNA and therefore mitigate the deleterious effects of RNA misfolding. Over the past two decades, it has been demonstrated that several classes of non-specific RNA binding proteins possess profound RNA chaperone activity in vitro and when overexpressed in vivo, and at least some of these proteins appear to function as chaperones in vivo. More recently, it has been shown that certain DExD/H-box proteins function as general chaperones to facilitate folding of group I and group II introns. These proteins are RNA-dependent ATPases and have RNA helicase activity, and are proposed to function by using energy from ATP binding and hydrolysis to disrupt RNA structure and/or to displace proteins from RNA-protein complexes. This review outlines experimental studies that have led to our current understanding of the range of misfolded RNA structures, the physical origins of RNA misfolding, and the functions and mechanisms of putative RNA chaperone proteins.

2. INTRODUCTION

Structured RNAs play essential roles in metabolic processes such as the processing of tRNAs and mRNAs, translation of mRNAs, translocation of proteins into the endoplasmic reticulum, and maintenance of chromosome ends, all of which are carried out by complex enzymes composed of proteins and at least one structured RNA (1–5). RNA also forms catalytic structures like self-splicing introns and a diverse set of smaller ribozymes. Even mRNA, once regarded as little more than ticker tape encoding the sequences of protein, is replete with structured elements in the 5′ -and 3′ -untranslated regions including riboswitches, internal ribosome entry sites, frameshift-inducing structures, and various protein-recognition elements (6–8). Although some protein recognition is based principally on the sequence of an RNA, in many other cases the RNA elements must adopt specific three-dimensional structures for specific protein binding (8). Further, in vitro selection and rational design schemes have expanded the structural and catalytic repertoire of RNA well beyond what has been found in nature (9).

Although there is great diversity in the shapes and sizes of RNA structures, there are clear themes to the types of structure that RNA forms. Structured RNAs typically include a set of short A-form helices (typically <10 base pairs), which together include the majority of nucleotides within the structure. These helices, referred to as elements of secondary structure, are arranged into a higher-order structure by tertiary contacts, many of which are formed by modular contact-forming motifs that can be identified by sequence (10–12). RNA structure is described as being hierarchical because the secondary structure is largely stable in the absence of the tertiary structure, and is largely left unchanged by the formation of tertiary structure (13), although exceptions to this paradigm exist (14).

The requirement that RNAs fold into their functional three-dimensional structures introduces what is termed the ‘RNA folding problem’, analogous to the ‘protein folding problem’, which continues to provide challenges for theoreticians and experimentalists after more than 40 years of research. As previously described by Herschlag, the RNA folding problem is usefully divided into two distinct problems, a thermodynamic one and a kinetic one (15). The thermodynamic problem is that in order to populate a single, functional structure at equilibrium, an RNA chain must specify a native structure that is more stable than all other possible structures. It was pointed out shortly after the first tRNA crystal structures were determined that the independent stability of RNA secondary structure may allow structured RNAs to adopt multiple conformations (16), and a corollary of this property is that it may be difficult for RNAs to specify a single structure. The kinetic problem is that the RNA chain must be able to fold to this single structure on a time scale that allows it to function. Numerous experimental studies on a range of RNAs have shown that a key component of this problem is that RNA has a strong propensity to adopt long-lived misfolded conformations, and it is this problem that is the subject of this review.

The notion that proteins may function as RNA chaperones by facilitating folding transitions was first put forth in the 1970s and early 1980s (17–19), well before many of the structured RNAs listed above were even discovered. Early work focused on the ability of a non-specific RNA-binding protein to facilitate conversion of an inactive conformer of a transfer RNA to its active form (19). Since then, a host of proteins that preferentially bind to single-stranded RNA have been studied and shown to facilitate native folding of a range of RNAs. While some of these effects are probably not of physiological significance, it has become increasingly clear that there is a necessity for proteins to facilitate RNA conformational changes in vivo. The clearest example of this is the recent demonstration that inactivation of a mitochondrial DExD/H-box protein in the fungus Neurospora crassa gives a cold-sensitive phenotype, with several group I introns failing to splice properly (20). Here, I review experimental work that has led to the conclusion that RNAs are prone to forming long-lived misfolded conformations, and the work that has led to our current understanding of the functions and mechanisms of proteins that act as RNA chaperones to reduce the deleterious effects of RNA misfolding.

3. PROPERTIES OF RNA THAT CONTRIBUTE TO MISFOLDING

3.1. Stability of RNA secondary structure

Any consideration of RNA folding and misfolding begins naturally with the secondary structure of RNA. It was noted as early as 1975 from an analysis of tRNA structures that structured RNAs differ fundamentally from proteins in that a large fraction of the nucleotides are present in regular base-paired helices, and these helices are stable enough that they form even when individual helices are generated as isolated molecules (16). Since then, this paradigm of extensive and stable secondary structure has proven to be true for a large number of families of structured RNAs. By 1981, the secondary structures of the 16S and 23S rRNAs had been deduced from comparisons of the sequences of multiple organisms (21–23). Approximately half of the nucleotides in these large RNAs form standard Watson-Crick base pairs, and many of the helices are stable enough that they remain formed under conditions that do not support full ribosome structure and activity. Indeed, chemical footprinting experiments under ‘inactivating’ conditions were used to confirm the accuracy of the secondary structure models (21, 22). More recently, the crystal structures of the ribosomal subunits confirmed that the vast majority of these predicted base pairs are indeed present (24, 25).

The high stability RNA secondary structure is largely independent of all but local interactions, with the stability of each base pair being dominated by the interactions of the two bases with each other and by stacking interactions with the base pairs that are immediately adjacent (26, 27). Indeed, this approximation is accurate enough that the stability of an RNA helix can be estimated from its sequence by what has become known as the ‘nearest neighbor’ method (28–31). With a diversity of only four standard nucleotides, an RNA of any significant length has a good chance of having the potential to form at least a few consecutive non-native base pairs. Further, because of the independence and high stability of RNA secondary structure, non-native helices that do form are often sufficiently long-lived that they interfere with proper folding on a time scale that would present problems for the function of the RNA. Thus, one major origin of misfolding – perhaps the most common one – is the formation of incorrect secondary structure. However, as described below, higher-order components of structure are also likely to play roles, both as origins of misfolding and by providing further stabilization of misfolded species.

3.2. Independent stability of RNA tertiary contacts

The formation of higher-order RNA structure involves the ordering of the secondary elements into defined relative orientations. Much of this ordering is accomplished by formation of contacts between pre-formed elements of secondary structure, and these interactions also have a high degree of energetic independence and stability. Some tertiary contacts are, in fact, composed of standard Watson-Crick base pairs. Known as pseudoknots, these contacts are formed between the terminal loops of two helices or between other single-stranded segments (Figure 1). These interactions, of course, share largely the same thermodynamic parameters as local secondary structure, and are therefore quite stable. Other tertiary contacts are formed by modular structural elements or motifs, which are found in multiple places and in multiple RNAs (12). One of the most common tertiary contact-forming motif is the GNRA tetraloop, which commonly interacts with a ‘receptor’ motif in a neighboring helix to form a tertiary contact (Figure 2). These interactions involve docking of purine bases, most commonly adenine, into the minor groove of the helix, forming an interaction termed an A-minor interaction (32). It has recently been shown that A-minor tertiary contacts can also be formed by motifs within internal loops (33, 34) and are widespread in structured RNAs (33).

Figure 1.

RNA pseudoknots. A, A pseudoknot is formed when loop nucleotides form base pairs with a region outside the loop. Shown is a simple pseudoknot structure from beet western yellow virus (207). B, A double-pseudoknotted RNA structure, the hepatitis delta virus (HDV) ribozyme. Pseudoknots are formed between nucleotides within a hairpin loop and nucleotides outside the hairpin. In the HDV ribozyme, helices P1.1 and P2 are pseudoknots. Outlined nucleotides are non-natural and were introduced to aid in crystallization (reproduced from ref. 208 by permission from Macmillan Publishers Ltd: Nature, copyright 1998).

Figure 2.

A-minor tertiary interactions. A, Crystal structure of the P4-P6 domain of the Tetrahymena group I ribozyme, which includes a tetraloop-receptor interaction (boxed). The GAAA tetraloop sequence (gold) forms tertiary contacts with its receptor sequence (green). The domain also includes an internal loop (orange) that forms A-minor interactions with the adjacent helix. A third helix in the domain (black) forms a tertiary contact composed of base-pairing interactions (the partner was not part of the molecule crystallized and not shown). This panel is reproduced with permission from ref. 209. Copyright 1996, AAAS. B, A-minor tertiary interactions. Adenosine nucleotides form contacts in the minor groove of a helix, most commonly with G-C pairs as shown. The specific contacts for type I and type II interactions are shown. This panel is reprinted from ref. 32 (Copyright 2001 National Academy of Sciences, U.S.A.), and the contacts shown are within the 23S rRNA.

For many RNAs, a large fraction of the tertiary contacts form between elements of structure that are found on the surface of the structure and are therefore termed ‘peripheral elements.’ For these RNAs, the tertiary contacts formed from one peripheral element to another and from peripheral elements to the interior of the RNA stabilize the core structure of the RNA. This stabilization from the periphery is quite different from a major strategy of proteins, which are stabilized from the inside by formation of a hydrophobic core.

Analogous to secondary structure, RNA tertiary contacts can form and be stable in the absence of the overall fold of a structured RNA. Independent formation of tertiary contacts is apparent in recent studies that have probed the structures and properties of tetraloop-receptor complexes that are completely removed from any functional RNA (35–37). The independent stability of local tertiary structure is also highlighted by structural studies in which localized regions of the ribosome have been shown to form the same tertiary contacts, and even adopt essentially the same conformations, as they do in the intact ribosomal subunits (24, 25, 38–40), although it should be noted that there are also examples of RNA elements that differ substantially in conformation when removed from their natural context (41, 42).

A central reason that isolated tertiary contacts can be stable is probably that the secondary elements are stable and are therefore largely pre-organized. When a tertiary contact is formed between pre-organized secondary elements, there is not a large entropic penalty for structural organization of the secondary elements, as there would be if these elements were not independently stable. This does not imply that the tertiary contacts form without cooperativity, as formation of one tertiary contact might be expected to strengthen others by fixing interacting helices or loops in the correct positions and/or orientations for interaction. Nevertheless, even without cooperative formation of an overall structure, isolated tertiary contacts can be favorable enough to form and accumulate, and thus tertiary contacts can also contribute to the problem of RNA misfolding.

In principle, the ability of tertiary contacts to form independently could allow the formation of nonnative tertiary contacts, representing a source of misfolding. However, this type of misfolding has not yet been demonstrated for any RNA. One possibility is that, because the number of tertiary contact-forming regions is relatively small for most structured RNAs and there is significant specificity to the interactions of tertiary partners, non-native tertiary contacts do not represent a major source of misfolding. However, the formation of native tertiary contacts has been demonstrated to stabilize misfolded structures (43), presumably because upon formation of the tertiary contact, resolution of the intermediate requires disruption of both the non-native contact that defines the misfolded structure and the stabilizing native contact. Here, the ability of tertiary contacts to form in isolation is clearly a detriment to proper folding because it enables native tertiary contacts to form in the context of existing nonnative structure.

3.3. Difficulty in switching between topologies

It has been suggested recently that an additional source of misfolding for RNA may stem from difficulty in obtaining the native topology, where ‘topology’ refers to the spatial arrangement of single-and double-stranded structural elements that cross each other in the context of a folded RNA [(44); see section 4.3 and Figure 4C and 4D). If an incorrect topology is adopted during folding, its resolution can require extensive unfolding. The formation of native tertiary contacts would also be expected to exacerbate this type of barrier by placing further constraints on the positions of portions of the RNA that must move to allow such a rearrangement.

Figure 4.

Misfolding of the Tetrahymena group I ribozyme. A, Secondary structure of the ribozyme. Long-range tertiary contacts are shown with thick arrows. The P3/altP3 region is highlighted in gray. B, Exchange of the native P3 and non-native alt P3 secondary structures. The structure at the left shows the native P3, as also shown in panel A. The structure at the right shows formation of the non-native alt P3 structure and disruption of P3. C, A topological difference between misfolded and native structures. The top panel shows the native topology within the core of the ribozyme and the bottom panel shows one of several possible incorrect topologies that may be present in the long-lived misfolded structure. D, A model for how a non-native secondary structure can bias an RNA to misfold by giving an incorrect topology, without the non-native secondary structure remaining present in the most stable misfolded form. Formation of the local and non-native structure alt P3 (purple circle) biases the RNA to adopt an incorrect topology early in folding, and this incorrect topology is maintained throughout folding to the misfolded structure even though the secondary structure is exchanged for the native P3 at a later point in folding. Reprinted from ref. 44 with permission from Elsevier.

4. EXPERIMENTAL STUDIES ON FOLDING AND MISFOLDING OF RNA

The history of experimental studies in RNA folding is filled with examples of RNA misfolding, so much so that misfolding has largely dominated the field. Although recent examples have been discovered of RNAs that do not appear to misfold, or at least do not accumulate misfolded species under the conditions they have been studied (45–48), nearly every RNA whose folding has been studied has been found to adopt misfolded conformations at least transiently (49). In this section I review experimental studies that have provided evidence for RNA misfolding, highlighting the types of misfolded structures that have been identified and their lifetimes. The section begins with tRNA, the first RNA that was shown to misfold.

4.1. Early RNA folding studies: tRNA

Studies of tRNA were prominent in development of the early fields of RNA structure, RNA folding and misfolding, and even the influence of chaperones on folding. At about the same time that the first tRNA sequence was determined and the cloverleaf secondary structure was proposed (50), the groups of Sueoka and Fresco independently reported that tRNAs can adopt two distinct conformations, only one of which is active for aminoacylation (51, 52). The inactive structure, originally referred to as denatured, could be generated by removing or chelating Mg²⁺ ions, and then was long-lived enough that it persisted on the experimental time scale upon subsequent addition of Mg²⁺. An incubation at elevated temperature was shown to restore full activity (> 40 °C), indicating that the RNA had not been irreversibly damaged but rather had folded to an alternative, inactive conformation (52).

A series of subsequent studies of the physical properties of these two conformations showed that the inactive conformation included a somewhat smaller total number of base pairs, was moderately less compact, and was more accessible to binding by certain complementary oligonucleotides (53–55). An extensive change of secondary structure was suggested by the very slow exchange of the inactive and active conformations, and it was suggested that the inactive conformation represented a transition from the cloverleaf secondary structure to an extended secondary structure (56, 57). Although definitive structural studies of the inactive conformation still have not been reported, the conclusion of an extended structure was supported by much more recent work in which chemical and enzymatic footprinting were applied to the inactive conformation (58), and a specific base-pairing pattern was proposed in which nucleotides of the D loop and the T loop form a non-native helix (Figure 3). Although these studies used different tRNAs, the common properties of the inactive conformation suggest a common origin, most likely a significant change in secondary structure.

Figure 3.

Model for misfolding of tRNA. A, Native secondary structure of E. coli tRNA^Glu. B, Proposed misfolded secondary structure. Non-native base pairs are formed between nucleotides in the D loop and the TψC loops, disrupting the corresponding stems. The model is based on data and interpretations in ref. 58.

4.2. Small RNAs: 5S RNA, small regulatory RNAs, and ribozymes

In the years during and immediately after the bulk of the work on tRNAs, interest in RNA structure and folding continued to be centered principally on small RNAs. Crothers and co-workers build on earlier work by Monier by showing that the 5S RNA, a small RNA component of the 50S ribosomal subunit, folds to two distinct conformations that are in slow exchange at physiological temperatures but can be readily exchanged by incubation at elevated temperature (59, 60). Although a specific secondary structure rearrangement was proposed for the exchange of these two structures (61), the nature of the rearrangement was not proven. Other work in this area was on regulatory RNAs in bacteria, principally cis elements, that were shown to respond to the nutritional status of the cell by adopting alternative secondary structures, leading to regulated expression of the mRNAs that contain them (62–64). Pioneered by Yanofsky, work in this area showed that exchange between secondary structures can be slow enough that folding is kinetically controlled. More recently, it has been shown that regulation is also achieved by binding of trans regulatory RNAs to mRNAs (65), and this area has exploded recently with the discovery of a widespread class of RNA regulators collectively known as microRNAs (66).

Several small viral ribozymes have also been the focus of folding studies. The hairpin ribozyme, which adopts a relatively simple structure in which two stem loop structures dock together to form an active site, was observed to give biphasic cleavage kinetics because a fraction of the ribozyme instead adopts an extended conformation (67, 68). More recent single molecule studies have shown that a more complex mixture of conformational states exist, and that these states fluctuate between docked and undocked conformations with distinct kinetics (69). Additional studies have shown that, in its natural context of a four-helix junction, the importance of this misfolded conformation is diminished and the ribozyme is considerably more active (70–72). Another small ribozyme, the hepatitis delta virus ribozyme, has also been shown to adopt misfolded conformations, which have been shown through structure mapping and mutagenesis studies to arise from incorrect formation of secondary structure (73, 74).

The recurring theme in misfolding of these small RNAs is that the misfolded and native conformations differ in their secondary structure. Although this could in part reflect the fact that this type of conformational difference is the one most easily probed experimentally (e.g. by site-directed mutagenesis), the importance of secondary structure in misfolding of these RNAs also undoubtedly results in part from their small size. These RNAs have extensive secondary structure but a relatively small number of tertiary contacts, and therefore when they misfold it is most often by forming non-native secondary structures. The larger RNAs, described below, form correspondingly more elaborate tertiary structures and appear to have a more diverse set of misfolded structures and origins of misfolding.

4.3. Group I intron RNAs

Interest in RNA structure was heightened tremendously in 1982, when it was discovered by Cech and co-workers that an RNA intron of a class referred to as ‘group I’ possessed catalytic activity, splicing itself out of a larger pre-rRNA in the absence of any protein (75, 76). In the nearly 25 years since, this RNA, from the ciliate Tetrahymena thermophila, has proven to be a workhorse for studies of RNA structure, function and folding. More recently, studies on folding of other group I introns have appeared, which have broadened our understanding of the complexities of the folding of large RNAs. Indeed, much of what we know about RNA folding processes and misfolded intermediates has come from studies of group I RNAs.

As several groups began to study catalysis by the Tetrahymena group I intron and its ribozyme derivatives in detail in the late 1980s and early 1990s, it became apparent that the intron was capable of adopting one or more misfolded structures, which could be resolved by incubating the RNA with Mg²⁺ at elevated temperature (77–79). Physical insight into the nature of one misfolded conformation came from studies of varying the length of the exons; it was shown that a phylogenetically conserved helix can form within the 5′ -exon, which competes with the interaction between the 5′ -splice site and the ‘internal guide sequence’ of the intron (80). Subsequent work from Woodson’s lab showed that this inhibitory helix is itself in competition with a helix even farther upstream from the intron, and formation of the second upstream helix prevents formation of the inhibitory helix and therefore restores splicing activity (81, 82).

The next few years saw an explosion in the field of RNA folding, and much of the work was on the Tetrahymena ribozyme. In pioneering work, Williamson and co-workers developed a time-resolved oligonucleotide hybridization method, in which sequences throughout the ribozyme were monitored for accessibility to a complementary oligonucleotide at various times following initiation of tertiary folding by addition of Mg²⁺ ions (83). The principal conclusion from initial work was that folding proceeds through at least one stable intermediate. Knowledge of intermediates and their structures was increased tremendously by the development and application of a time-resolved hydroxyl-radical footprinting method to the folding process (84–86). Thus, it became clear that there are several accumulated intermediates, as tertiary structure forms first in one independently folding domain, termed P4–P6, and subsequently spreads throughout the molecule. More recent work from Brenowitz and colleagues has shown that the conclusion of multiple intermediates hold under a range of solution conditions, and even for folding in the presence of monovalent ions alone (87–89).

In addition to discovering folding intermediates, work in the mid-to late-1990s was directed toward determining the roles of these intermediates in folding. Woodson and co-workers used modification-interference, time-resolved footprinting, and an electrophoretic mobility shift assay (EMSA) to determine that the intron adopted a misfolded conformation and that an alternative secondary structure in the core of the intron, termed alt P3, was at least partially responsible for formation of this misfolded intermediate (90, 91)(Figure 4A and 4B). Considerations of the RNA folding process would suggest that formation of the native P3 would be difficult because alt P3, being local in primary sequence, would be imagined to form first, and would then have to be disrupted to form the native P3 base pairs. On the other hand, the alt P3 helix is only four base pairs and would be expected to exchange on the time scale of milliseconds (92), whereas this inactive conformer was resolved to the native state much more slowly, suggesting that resolution of the misfolded intermediate required more extensive rearrangements than just exchange of this alternative secondary structure. Further work from the Williamson lab showed that mutations that weaken native tertiary contacts on the periphery of the molecule accelerated the formation of a protected core (43, 93, 94), as did addition of the denaturant urea, suggesting that the non-native structure present in the misfolded structure was stabilized by native tertiary contacts.

Further insight into structural features of misfolded intermediates and their folding transitions became possible from application by Herschlag and co-workers of a quantitative ribozyme activity assay to follow formation of the native ribozyme. It was shown that what had been viewed as a single misfolded intermediate was actually two distinct intermediates that form in succession during folding. The longer-lived of the two could be populated by up to 90% of the ribozyme for a period of hours (95, 96). Subsequent studies then characterized this misfolded species and its re-folding transition. The misfolded species was shown by small angle x-ray scattering to be nearly as compact as the native ribozyme (97). It was further shown to have each of five long-range peripheral contacts formed and to require disruption of all of them during its re-folding to the native state (44). Intriguingly, this study also indicated that the alt P3 base pairs, despite biasing the RNA toward the long-lived misfolded conformation, are exchanged to the native P3 base pairs in the misfolded conformer, and a similar proposal was made earlier from oligonucleotide hybridization studies (98). It was suggested that all of the native base pairs are formed in the misfolded conformer, and that the barrier between the native and misfolded species is topological, requiring global unfolding of the RNA to allow one strand or segment to pass to the other side of another within the core (Figure 4C and 4D). While this proposal is intriguing, it will require additional tests to establish whether it is correct.

In recent years, studies on folding of a few other group I introns have broadened the conclusion that misfolded intermediates are common, while also offering some insight into the features that make misfolding more likely. Catalytic activity assays of a ribozyme derived from the phage T4 td intron showed that a non-native extension of the P2 helix can form during folding and inhibit substrate cleavage (99), once again demonstrating a non-native secondary structure. Studies of a Candida albicans group IE intron, a subgroup that is closely related to the Tetrahymena intron, demonstrate that the RNA is prone to misfolding, and in further analogy with the Tetrahymena ribozyme, misfolding can be avoided by pre-incubation with monovalent ions (100–102). Recent studies of a very small group I ribozyme from the bacterium Azoarcus suggest that the ribozyme largely avoids stable misfolding, and acquires all of its tertiary structure in milliseconds (103, 104). As this ribozyme lacks peripheral structure almost entirely, this result reinforces the notion that peripheral structure can exacerbate the consequences of misfolding by stabilizing misfolded species, and it introduces the idea that with fewer structural elements in a smaller RNA there is less that can go wrong.

4.4. RNase P RNA

At about the same time that RNA catalysis was discovered in the group I intron, it was shown by Altman, Pace, and co-workers that the catalytic component of the bacterial RNase P enzyme is RNA (105). RNase P functions in tRNA processing by cleaving pre-tRNAs to generate their mature 5′ -ends (106). In bacteria, the enzyme is composed of one RNA and one protein, whereas in the archaeal and eukaryotic domains the nuclear-encoded enzyme includes RNA and multiple proteins (107, 108). The bacterial RNA consists of two separable domains; the C- (catalytic) domain contains all of the nucleotides that form the active site and the S- (specificity) domain forms some of the contacts with the tRNA substrate (109, 110). Most of the folding studies to date have focused on the bacterial RNase P RNA, as it is able to fold in vitro to a native structure and to function in the absence of accessory proteins.

The first hint that RNase P RNA is prone to misfolding came shortly after the delineation of the RNA as the catalytic component, with the finding that catalysis by the RNA displayed a kinetic lag that could be rescued by pre-incubation of the RNA (111). Detailed characterization of the Mg²⁺-dependent folding of the RNA began a decade later (112). Much of the research on folding of RNase P RNA has been carried out in a long-standing collaboration between the labs of Pan and Sosnick. In 1997, these groups pioneered the use of circular dichroism and absorbance spectroscopy to study RNA folding, reporting that RNase P RNA from Bacillus subtilus requires hundreds of seconds to complete its folding, despite rapid formation of at least one intermediate (113). Further, the folding was found to be strongly accelerated by urea, indicating the presence of kinetically trapped species.

In further work, these groups found that an RNA that was engineered such that one of the connections was broken between the C-and S- domains folded much faster than the wild-type RNA, suggesting that the principal barrier for the wild-type RNA is linked to the interdomain connections (114). When expressed as an independent domain, the C-domain folds much more rapidly and is not accelerated by denaturant, suggesting that its folding is not limited by the resolution of a misfolded structure and reinforcing the notion that the slowest step for folding of the full-length RNA involves resolution of a misfolded structure that involves or is dependent on the connections between the two domain (45, 46). Nevertheless, folding of the isolated C-domain remains complex, as subsequent single molecule fluorescence studies revealed the presence of additional intermediates that are populated at equilibrium (115).

In the last few years, a burst of structural information on RNase P RNA has been obtained from determination of crystal structures, first of the isolated S-domain (116), and then of the entire RNase P RNAs from Bacillus stearothermophilus (117) and Thermotoga maritima (118) [reviewed in (119)]. The structural information is certain to impact research in folding in the coming years and has already been used to infer structural features of a folding intermediate of the S-domain (120).

4.5. The 30S ribosomal subunit

The ribosome is one of the largest and most complex pieces of machinery that the cell must assemble; the three RNAs that make up the bacterial 30S and 50S must fold and assemble with the more than 50 proteins that make up the functional ribosome. With this complexity in mind, it is not at all surprising that assembly involves multiple steps and appears to include misfolded structures. Research into folding and assembly of the 30S ribosomal subunit is more advanced than for the 50S subunit, and much of the research over the last 35 years has been built on the foundation provided by the seminal work of Nomura and co-workers. A series of studies by this group culminated in delineation of an ‘assembly map’ for 30S formation, which outlined a crude order of stable protein association with the 16S rRNA (121). Even at this early stage, it was clear that the folding and assembly process was complicated kinetically by the formation of an intermediate containing only a subset of the proteins that was blocked at low temperature from further assembly but could be pushed along the assembly route by incubating at elevated temperature (122, 123). It was also shown that active 30S subunits could be inactivated by a transient decrease in Mg²⁺ concentration or removal of monovalent ions (124, 125), and that the inactive form was altered in its accessibility to chemical footprinting reagents (126), underscoring the ability of the 30S subunit to adopt alternative conformations.

More recently, detailed physical studies have begun to give physical insights into the nature of the rate-limiting steps for folding and assembly of the 30S subunit, and these studies have been aided tremendously by the advent of high-resolution structural information (25). Holmes and Culver used a chemical footprinting approach to characterize the accessibility of the rRNA bases before and after the principal structural rearrangement, as well as before and after binding of the early and late sets of ribosomal proteins (127, 128). These studies suggested that significant rearrangements of the RNA are involved in the high-temperature conformational change, and these rearrangements may include formation of some of the long-range base pairs present in the functional structure. Development of a powerful method for monitoring the kinetics of protein binding based on mass spectrometry has recently provided a higher-resolution view of the assembly process (129), and has suggested that what has been viewed as a single rate-limiting conformational rearrangement may instead be a series of rearrangements in which the binding of various proteins is coupled to RNA structural changes, some of which may represent disruption of non-native interactions, while others may represent formation of contacts. Further, this work suggested that there is not a single order for completion of these folding steps, but rather a landscape that contains multiple possible routes to the functional structure.

5. CHAPERONE ACTIVITY OF NON-SPECIFIC RNA-BINDING PROTEINS

In several cases, outlined below, it has been shown that proteins that bind single-stranded RNA non-specifically display RNA chaperone activity, facilitating conformational rearrangements of RNA from misfolded to active structures or between functional structures. These proteins are typically basic and are often rich in arginine residues. Presumably, their ability to interact preferentially with single-stranded RNA stabilizes folding intermediates in which RNA-RNA contacts are disrupted, and this stabilization accelerates re-folding transitions. A large excess of these proteins is typically required for activity, consistent with models in which the proteins coat the unstructured RNA intermediates with ill-defined stoichiometry. Although chaperone activity has been clearly demonstrated, for many of these proteins it remains to be determined whether they actually function as RNA chaperones in vivo. Such demonstration may be difficult because cells may express multiple chaperones with overlapping substrate specificity, such that deletion of any one chaperone would not give an observable phenotype. Nonetheless, it remains uncertain whether the chaperone activities reflect the functions of these proteins or whether they are simply a consequence of their physical properties. Below, the experiments are reviewed that led to identification of the major classes of proteins with chaperone activity, while some more recently-discovered examples are omitted for brevity.

5.1. hnRNP proteins

The earliest studies of RNA chaperone activity were performed not long after the first RNA misfolded species were identified. These early studies focused on the activity of a protein referred to as UP1, a proteolytic fragment of hnRNPA1 protein isolated from calf thymus (17, 19). The hnRNP proteins are present in mammalian cell nuclei and are thought to associate with RNAs during their biogenesis. The groups of Karpel and Fresco found that the UP1 protein fragment accelerated reactivation of yeast tRNA^Leu and E. coli 5S RNA from their inactive conformations. The activity of UP1 was inhibited by single-stranded RNA, suggesting that the single-stranded binding activity of the protein was involved in its chaperone activity. The A1 hnRNP protein was subsequently characterized as a protein that facilitates renaturation of DNA as well as RNA, presumably by binding single strands and perhaps bringing them together via multiple sites on the protein or by protein multimerization (130–132).

5.2. Viral nucleocapsid protein

As catalytic RNAs were discovered and characterized in the late 1980s and early 1990s, it became possible to explore whether proteins could increase the activities of these catalytic RNAs by acting as chaperones. In the first such study, Herschlag and co-workers showed that the hammerhead ribozyme was accelerated in its oligonucleotide cleavage reaction by the presence of the HIV p7 nucleocapsid protein (NC) (133). Although the hammerhead ribozyme is not one that is known to adopt long-lived misfolded species during folding (see section 4.2 above), its reaction as a multiple-turnover enzyme requires the formation and dissociation of an RNA helix between the oligonucleotide substrate and the ribozyme, and subsequent work showed that the NC protein accelerated the reaction by facilitating both the formation and dissociation of the helix (134).

At the time this work was performed, the physiological roles of the NC protein were only beginning to emerge, but in the years following it became clear that the NC protein also has substantial chaperone activity for several nucleic acid rearrangements that are vital components of the viral life cycle. These rearrangements include dimerization of the viral RNA (135, 136), annealing of the tRNA primer to the genomic RNA to initiate reverse transcription (135, 137), disruption of template structure during reverse transcription (138, 139), and ‘strand transfer’ of the newly-generated (−) strand DNA to continue reverse transcription (140, 141) [reviewed in (142)]. Thus, it appears highly likely that RNA chaperoning is a central function of the NC protein.

Physical insight into the mechanism of chaperone action has come from an NMR structure of NC in complex with a viral stem-loop structure (143), which shows an extensive set of electrostatic interactions between arginine residues and the backbone of the RNA. Presumably, similar contacts are made and even strengthened in complexes with single-stranded RNA, and recent work from Karpel’s group has provided further evidence for preferential binding to single-stranded RNA (144). Recently, a single molecule fluorescence approach has been applied to the NC-mediated RNA rearrangements, with the finding that the strand-transfer reaction is catalyzed via multiple pathways by the NC protein (145, 146).

5.3. Ribosomal proteins

As many of the ribosomal proteins are intertwined with segments of the rRNAs in the crystal structures of the ribosomal subunits (24, 25, 147), it is perhaps not surprising that many of the ribosomal proteins possess non-specific RNA binding and RNA chaperone activities. RNA chaperone activity was first characterized for the S12 protein, with lesser activity attributed to several other ribosomal proteins, from experiments in which E. coli lysates were fractionated based on their ability to facilitate in vitro splicing of group I introns from the phage T4 (148). More recently, a systematic survey of the ability of large subunit ribosomal proteins to chaperone folding of one of these introns revealed that approximately one-third of the 34 large subunit proteins have RNA chaperone activity (149). It remains unclear whether any of these proteins function as RNA chaperones in vivo.

As a means of testing proteins for RNA chaperone activity in vivo, Schroeder and co-workers have implemented a system in which overexpression in E. coli of a protein with RNA chaperone activity can rescue self-splicing of a group I intron in the phage T4 td (thymidylate synthase) gene, which is otherwise prevented from splicing by formation of a long-lived misfolded structure that includes a non-native secondary structure [(150); reviewed in (151, 152)]. Significant chaperone activity was detected for the S12 protein, as well as for the NC protein (Section 5.2 above) and the StpA protein (Section 5.4 below), and more recently, chaperone activity has been demonstrated for protein components of Ro ribonucleoprotein complexes (153) and has been confirmed by other methods (154). Although this in vivo assay does not establish a chaperone function for these proteins, it establishes that they are capable of RNA chaperone activity inside bacterial cells.

5.4. StpA: a bacterial histone-like protein

Like some of the ribosomal proteins, StpA was shown to function as a chaperone for self-splicing of the group I intron in the phage T4 td gene. It was isolated as a multi-copy suppressor of defective splicing by Belfort and co-workers (155). StpA is a small (15 kDa) and highly basic protein that functions as a transcriptional regulator and is quite similar to the highly expressed H-NS nucleoid protein (156). Interestingly, despite the similarity of these two proteins, they regulate transcription of distinct sets of genes, and StpA is much more efficient as an RNA chaperone (156). Further work on the mechanism of chaperone activity by StpA has shown that it increases the exposure of the group I intron core to chemical footprinting reagents (157). Further, the efficiency of this structure disruption by StpA is related to the stability of the RNA structure, such that StpA can be inhibitory for splicing of variant group I RNAs with decreased stability (158), presumably because it remains associated with the partially unfolded RNA or because it unfolds the RNA faster than it can fully re-fold.

5.5. Cold-shock proteins

Upon temperature downshift, global protein expression in bacteria is decreased, but expression of a relatively small number of proteins (17 in E. coli) is substantially increased (159). Several of these ‘cold-shock’ proteins bind RNA non-specifically and are therefore candidates to function as RNA chaperones. As the stability of RNA secondary structure is strongly temperature-dependent, it would be expected that kinetic traps for RNA would be particularly severe at low temperature, and thus there would be a stringent requirement for RNA chaperones under these conditions (although some of these proteins are also expressed at higher temperature and presumably function similarly). One of the most highly expressed proteins, CspA (70 amino acids), was shown to increase the susceptibility of its own mRNA to cleavage by the single-strand specific RNases A and T1, indicating that CspA reduces the secondary structure content of the mRNA (160). Subsequently, CspA and the related proteins CspE and CspC were shown to function as transcription anti-terminators by preventing the formation of a secondary structure element in nascent RNAs that would otherwise lead to termination (161).

Crystal and solution structures of CspA revealed a five-stranded □eta-barrel structure with an OB-fold (162, 163), and a more recent co-crystal structure has shown that the structurally related Bacillus subtilus CspB protein binds single-stranded nucleic acid in an extended conformation along one surface of the protein (164)(Figure 5). This binding mode, with the nucleic acid extended and single-stranded, is consistent with a role as an RNA chaperone. It remains to be determined whether these proteins function principally as regulators of transcription terminators or whether they have more general roles as RNA chaperones.

Figure 5.

Nucleic acid binding by the cold-shock protein CspB. Two molecular of an oligonucletide of six thymidine nucleotides (dT6, shown in red) were bound to CspB in the crystal, as shown. The protein is shown as a semi-transparent surface, and carbons of residues on the protein that contact the oligonucleotide are colored green. Reprinted from ref. 164 with permission from Elsevier.

5.6. Hfq: a bacterial Sm-like protein

The Hfq protein was originally discovered as a factor required for replication of the phage Q-beta (165), and it was subsequently shown to give wide-ranging phenotypes when inactivated (166) and to be involved in the post-transcriptional regulation of a number of genes [reviewed in (167)]. Hfq shares sequence and structure similarity with the eukaryotic Lsm proteins, forming a hexameric ring structure similar to the hetero-heptameric structures of the Lsm proteins, and binds single-stranded RNA around a central basic cleft (168).

Key insight into the mechanism of regulation by Hfq came from Storz and colleagues, who found in 1998 that Hfq interacts with the small regulatory RNA OxyS (169). At the time, OxyS was one of only a handful of well-characterized, bacterial non-coding RNAs, whereas since then the number has grown substantially [reviewed in (170)]. OxyS was known to activate and repress expression of multiple genes, but the mechanisms of its action were unclear. Subsequent work showed that OxyS binds directly. to the mRNAs of at least two genes that it regulates, and that the complex formation of these RNAs depends on the Hfq protein (171). Thus, Hfq was suggested to function as a chaperone to facilitate annealing of the regulatory RNA with its target mRNAs, and evidence for this mode of action has since been accumulated for several additional regulatory RNAs (170, 172). Although it remains likely that Hfq has multiple modes of action, including some that may be unlinked to RNA chaperoning (173), it is clear that facilitating association of RNAs is a central function of the protein.

Although the mechanism by which Hfq facilitates RNA-RNA complex formation remains uncertain, ideas are beginning to emerge from structural and biochemical studies. Focus was initially on the interaction of Hfq with the small regulatory RNAs, but it was subsequently shown that Hfq interacts with the target mRNAs also, raising the possibility that it disrupts structure within the mRNA, allowing access to the regulatory RNA (174). It has recently been suggested that Hfq has distinct interacting surfaces for multiple RNAs and that it can form ternary complexes with two RNAs (175), raising the possibility that, rather than simply facilitating rearrangements by stabilizing single-stranded intermediates, Hfq actively mediates formation of RNA-RNA complexes by facilitating strand displacement reactions (175, 176).

6. RNA CHAPERONE FUNCTION AND MECHANISM OF DExD/H-BOX PROTEINS

DExD/H-box proteins are ubiquitous throughout nature and have been implicated in nearly all processes that involve structured RNAs [recently reviewed in (177, 178)]. They are ATPases with significant sequence similarity to DNA helicases, and are characterized by the presence of several highly conserved sequence motifs, including one of sequence DExD/H from which they derive their name. At the most general level, DExD/H-box proteins are thought to function by coupling the energy derived from cycles of ATP binding and hydrolysis to acceleration of RNA structural transitions that would otherwise proceed slowly. A very large number of these proteins appear to function with a high degree of specificity for a particular RNA or RNP substrate, and this specificity is achieved because the DExD/H-box protein recognizes an RNA or protein that is part of its target complex (179). On the other hand, recent evidence indicates that a subset of DExD/H-box proteins have more relaxed specificity and apparently function as general RNA chaperones. Unlike the proteins described in the preceding section, which presumably facilitate structural rearrangements of RNA by a relatively passive mechanism in which they bind and stabilize single-stranded intermediates, DExD/H-box proteins have the potential to actively disrupt contacts in an energy-consuming process. Below, I first review research in which model systems have been used to explore the repertoire of structures that DExD/H-box proteins are able to disrupt. Then I review recent evidence for the function of some of these proteins as general RNA chaperones and current proposals and evidence for the mechanisms by which they may achieve this general chaperone function.

6.1. Repertoire of structure-disruption activities by DExD/H-box proteins

Soon after the DEAD box family (a major subset of the DExD/H-box proteins) was identified by sequence comparisons of several newly-sequenced genes (180), it was shown that one of these proteins, the human p68 protein, was capable of unwinding double-stranded RNA (181). RNA unwinding had already been suggested for one other DEAD-box protein, the translation initiation factor eIF4A (182, 183), and over the next several years was characterized further for this protein and for several other DExD/H-box proteins (184–189).

However, despite the apparent similarity of the activity with that of DNA helicases, two features of the RNA unwinding activity of the typical DExD/H-box protein are substantially different. First, the processivity of DExD/H-box proteins tends to be very low, such that significant unwinding usually requires a considerable excess of proteins and is only observed for short helices (typically <15 base pairs). Second, whereas most DNA helicases require a single-stranded extension of a defined polarity for efficient loading, some DExD/H-box proteins have been shown not to require a particular directionality for the single-stranded extension (189, 190).

It might be imagined that the low processivity could reflect the genuine function of many DExD/H-box proteins. After all, the structured RNAs that are implicated in the functions of these proteins – precursors of the ribosomal RNAs, the spliceosomal RNAs, and other structured RNAs – have secondary structures that are composed of short helices, not long continuous ones. A protein whose function it is to facilitate rearrangements of these RNAs, or to help them resolve misfolded species, does not need to unwind long, continuous helices. Instead, it may need to disrupt short, non-native helices, and it also may need to disrupt tertiary contacts or even displace proteins. It is interesting in this regard that one DExD/Hbox protein that is clearly not ‘typical’, as defined above, but instead has properties closer to those of DNA helicases, is the viral protein NPH-II, which is implicated in viral transcription and may function by unwinding or translocating along long stretches of DNA or RNA (191, 192).

With the idea of defining the catalytic repertoire of DExD/H-box proteins, Jankowsky, Pyle and co-workers tested the ability of these proteins to perform activities other than unwinding simple helices. They first found that the NPH-II protein is capable of displacing the tightly-bound U1A protein from an RNA helix in an ATP-dependent fashion (193). While this displacement presumably occurs during unwinding of the duplex by NPH-II and may result from the unwinding reaction rather than a direct interaction of NPH-II with the protein to be displaced, in subsequent work Jankowsky and co-workers showed that NPH-II can also displace a protein, or even a complex of several proteins, from binding sites on single-stranded RNA, ruling out any involvement of RNA helix unwinding (194). Analogous activities were also observed for the DED1 protein from yeast, which is implicated in translation initiation and behaves in unwinding assays as a more ‘typical’ DExD/H-box protein than NPH-II (188, 195). This group and others have also found that DExD/H-box proteins, including DED1, can facilitate formation of RNA duplexes in addition to unwinding (196, 197), an activity reminiscent of the bacterial Hfq protein. For DED1, the annealing activity was shown to be ATP-independent, such that the ratio of ATP to ADP defines a steady-state balance between unwinding and annealing activities (196).

6.2. Evidence for general RNA chaperone activity

The first demonstration that a DExD/H-box protein functions as an RNA chaperone came in 2002, when Lambowitz and co-workers showed that mutations in the Neurospora CYT-19 DEAD-box protein result in defects in splicing of several mitochondrial group I introns in vivo [(20); see also (198)]. Further in vitro experiments indicated that CYT-19 binds structured RNAs with little or no specificity for individual group I introns and that it facilitates a structural rearrangement of the related Tetrahymena group I intron, indicating that the splicing defect in vivo arises from a failure of the introns to fold properly, and that CYT-19 functions as a chaperone in this process.

Subsequent work from the groups of Lambowitz and Perlman suggested further that CYT-19 functions as a broad-specificity or general RNA chaperone, and that a mitochondrial yeast DExD/H-box protein, Mss116p, functions analogously (199). Whereas prior work had detected defects in splicing of only a subset of the group I and group II introns upon mutation of MSS116 (200), the more recent work gave detectable decreases in splicing for all nine group I introns and all four group II introns (199). Although indirect effects of Mss116p inactivation cannot be excluded, particularly as most of the introns require specific-binding proteins whose expression may be compromised by loss of Mss116p activity, this result and the earlier characterization of the activities of Mss116p (201) suggested that Mss116p functions as a general chaperone for both group I and group II introns. Heterologous expression of CYT-19 in the MSS-116-disrupted strain gave partial or complete restoration of splicing for each intron, indicating that CYT-19 is also capable of facilitating folding of a range of structured RNAs that extends beyond group I introns. Further, in vitro studies from Lambowitz, Perlman, and co-workers (202) and from my group (203), described below, have directly demonstrated that CYT-19 acts as a chaperone in folding of group I and group II introns and have begun to probe the mechanisms of this chaperone activity.

6.3. Mechanism of DExD/H-box proteins as general chaperones

To explore the mechanism of RNA chaperone activity by CYT-19, my group designed an in vitro system to follow re-folding of the long-lived misfolded conformation of the group I RNA from Tetrahymena to its native state (203). We found that CYT-19 indeed accelerated re-folding of this misfolded RNA to the native state and could subsequently be removed before the fraction of native ribozyme was determined by activity, indicating that it acts solely to increase the rate of a conformational transition.

Although much about the mechanism by which CYT-19 gives re-folding of this RNA remains to be determined, we made two additional observations that gave insight into the general mechanisms of CYT-19 action. First, we found that CYT-19 could also unwind the six-base-pair helix between the RNA and its oligonucleotide substrate, and it performed this reaction much more efficiently than it unwound the same helix free in solution. This increased efficiency from adjacent secondary and tertiary structure most simply suggested that CYT-19 forms additional interactions with the structured RNA via a site on the protein that is distinct from the one responsible for unwinding, and further work has localized this second site on the protein to an ancillary C-terminal domain (204). The second observation is that this unwinding activity was eliminated by formation of tertiary contacts between the helix and the body of the structured RNA.

These results led to a model for the mechanism of action of CYT-19, and presumably other DExD/H-box proteins that act as general RNA chaperones, that is distinctly different from the mechanism of processive DNA and RNA helicases (Figure 6). In this model, binding via an ancillary domain tethers the unwinding active site of CYT-19 in proximity to its structured RNA substrates. The protein preferentially disrupts structural elements that are not tightly packed to the rest of the RNA structure while remaining bound to the RNA through its ancillary domain. This mechanism accounts for the lack of a polarity requirement for single-stranded extensions in model studies of DExD/H-box proteins (189, 190); presumably the ancillary domain is binding to the extension and remaining bound during unwinding. The mechanism also accounts for the low processivity of unwinding, because continued association by the ancillary domain will most likely prevent significant translocation by the unwinding domain.

Figure 6.

Model for chaperone activity of CYT-19. CYT-19 binds to structured RNA via a site on the protein that is distinct from the site responsible for unwinding helices or performing other structure-disruption activities (labeled RBD). This site remains bound at its attachment point while the ATP-dependent structure-disruption occurs. The figure shows unwinding of the helix of the Tetrahymena ribozyme formed between the ribozyme and its oligonucleotide substrate (colored green and red, respectively). The secondary elements of the ribozyme are depicted as black and blue cylinders for core and peripheral elements, respectively. Reprinted from ref. 203 (Copyright 2006 National Academy of Sciences, U.S.A.).

One feature of this mechanism, the preference for disruption of loosely-associated structure, may provide a means of targeting CYT-19 to act on misfolded rather than native RNAs, because misfolded RNAs are more likely to have structural elements that are unable to pack correctly. Further, the non-native structural elements are the ones that are less likely to pack, so this mechanism may bias chaperones to disrupt the non-native portions of misfolded structures. Similar mechanisms may apply to DExD/H-box proteins that interact with specific RNAs or RNPs, except that in these cases the ancillary domain binding directs the protein to act on particular substrates.

7. SUMMARY AND PERSPECTIVE

Over the course of nearly forty years, in vitro studies of RNA folding have shown time and time again that RNAs are prone to forming misfolded conformations, and considerations of the structures and properties of RNA suggest that misfolded species are likely to be encountered during folding in vivo as well. Numerous proteins that bind RNA non-specifically have been shown to possess RNA chaperone activity in vitro, and some of these proteins have been shown to function by facilitating specific rearrangements or complex formation of RNAs. Whether any of these proteins function as general chaperones, interacting with and facilitating folding of an assortment of RNAs, remains unclear. In contrast, evidence is currently mounting that a subset of the myriad DExD/H-box proteins function more generally, although the range of RNAs that have been shown to be assisted in folding by these RNA chaperones is currently limited to group I and group II introns. Proper folding of some RNAs may be so difficult that cells have devoted chaperones specifically to that function; for example, several DExD/H-box proteins appear to function specifically in ribosome biogenesis (205). On the other hand, the large number and diverse structures of RNAs would suggest that there must be some chaperones that are available to function non-specifically, unfolding whatever misfolded RNA structures happen to arise. The need for general chaperones is underscored by recent revelations that the number of non-coding RNAs is much larger than was previously known (206).

Although our knowledge of the involvement and actions of chaperones in RNA folding is increasing, many questions remain. Most centrally, which RNAs are dependent on chaperones for proper folding? Although a dependency is clear for RNAs and RNPs that require a particular chaperone, it is quite possible that many other RNAs are dependent on chaperones that have broad specificities and are functionally redundant with multiple chaperones, such that folding defects do not arise upon inactivation of any single chaperone. Are there different types of chaperone duties for different types of chaperone proteins? It might be imagined that the unfolding required by some misfolded RNAs to allow them to re-fold to their native states would be sufficiently unfavorable as to require energy in the form of ATP, whereas re-folding of others would proceed quite efficiently as long as chaperone proteins were present to stabilize the less structured intermediates and corresponding transitions states. Presumably the first class of misfolded structures would use DExD/H-box proteins, whereas the second would use other chaperones, but whether such a division of labor occurs among chaperone proteins is completely unexplored. Further, even the mechanistic assumptions underlying this expectation – that DExD/H-box proteins actively disengage contacts, whereas other chaperone proteins stabilize incrementally larger segments of RNA as they unfold on their own – are not established. And finally, if there are general chaperones running around the cell unfolding RNAs and RNPs, how does the cell protect its native RNAs from unwanted attention by chaperones?