RNA-guided isomerization of uridine to pseudouridine—pseudouridylation

Abstract

Box H/ACA ribonucleoproteins (RNPs), each consisting of one unique guide RNA and 4 common core proteins, constitute a family of complex enzymes that catalyze, in an RNA-guided manner, the isomerization of uridines to pseudouridines (Ψs) in RNAs, a reaction known as pseudouridylation. Over the years, box H/ACA RNPs have been extensively studied revealing many important aspects of these RNA modifying machines. In this review, we focus on the composition, structure, and biogenesis of H/ACA RNPs. We explain the mechanism of how this enzyme family recognizes and specifies its target uridine in a substrate RNA. We discuss the substrates of box H/ACA RNPs, focusing on rRNA (rRNA) and spliceosomal small nuclear RNA (snRNA). We describe the modification product Ψ and its contribution to RNA function. Finally, we consider possible mechanisms of the bone marrow failure syndrome dyskeratosis congenita and of prostate and other cancers linked to mutations in H/ACA RNPs.

Abbreviations

DC

dyskeratosis congenita

HH

hoyeraal-hreidarsson syndrome

PIKK

phosphatidylinositol 3-kinase-related kinase

ψ

pseudouridine, 5-ribosyluracil

PUA

pseudouridylase and archaeosine transglycosylase

RNP

ribonucleoprotein

rRNA

ribosomal RNA

sca

small Cajal body

SMN

survival of motor neuron protein

sno

small nucleolar

snoRNA

small nucleolar RNA

snRNA

small nuclear RNA

SSD

SHQ1 specific domain

tRNA

transfer RNA

U

uridine

X-DC

X-linked dyskeratosis congenita

Introduction

Posttranscriptional modification of RNA is a common and diverse event with over 140 distinct RNA modifications.¹ Of all these, pseudouridine (Ψ, 5-ribosyluracil) is the most common and at the same time perhaps the most unusual. Unlike other RNA modifications that simply add functional groups while leaving the nucleoside untouched, pseudouridylation involves the detachment of the base from the ribose, its rotation by 180° and reattachment to form a nearly unique carbon-carbon glycosidic bond, i.e. the isomerization of uridine to Ψ (Fig. 1A). Ψ is so abundant that it was originally recognized as a “fifth” nucleotide.² This is not surprising given that it so far has been identified as significant component of only the most abundant stable RNAs in the cell, transfer RNA (tRNA), ribosomal RNA (rRNA) and spliceosomal small nuclear RNA (snRNA). Pseudouridylation is performed by phylogenetically related enzymes, known as pseudouridine synthases, with or without the help of additional proteins and guide RNAs. Guide RNAs identify target uridines by site-directed base pairing. Pseudouridylation guide RNAs are known as H/ACA RNAs, one of the 2 major classes of small nucleolar RNAs (snoRNAs), the other being C/D RNAs responsible for guiding 2′-O-methylation. The ribonucleoproteins (RNPs), formed by H/ACA RNAs, the pseudouridine synthase, and 3 other core proteins, are called H/ACA RNPs. In this review, we will concentrate on RNA-guided pseudouridylation, which limits us to the following modification targets, archaeal and eukaryotic rRNA, archaeal tRNA and yeast and mammalian snRNA. For the abundant pseudouridylation of tRNAs and its dedicated stand-alone enzymes, we refer to the following excellent sources.^3-5

Figure 1.

Pseudouridylation, guide RNA, and guide mechanism. (A) Chemical structure of uridine (U) and pseudouridine (Ψ). The 180° rotation of the base frees up the N-glycosidic nitrogen in U as an additional hydrogen donor in Ψ (red). The numbering of base and ribose positions for potential nucleophilic attacks by a conserved aspartate of the Ψ-synthase is indicated. Note the aspartate is highlighted in yellow in the dark green catalytic domain of the Ψ-synthase (Fig. 2, inset). (B) Schematic of an H/ACA guide RNA hybridized to a substrate RNA (blue) in its 3′ pseudouridylation pocket. The hinge region and ACA conserved sequence elements responsible for the H/ACA name are highlighted (red; N = any nucleotide). (C) Example of the pseudouridylation pocket of the yeast H/ACA RNA snR81 hybridized to its substrate RNA (blue). Perfect Watson-Crick base-pairs between the guide sequence and the substrate RNA result in constitutive pseudouridylation at the target site (left panel, constitutive pseudouridylation). In contrast, if the guide sequence forms imperfect base-pairs (there are 2 mismatches) with the substrate RNA, pseudouridylation occurs only under stress (nutrient-deprivation) conditions (right panel, inducible pseudouridylation).

Pseudouridine

Ψ, 5-ribosyluracil, is the most abundant modification of RNA (Fig. 1A). The isomerization of uridine to Ψ changes the chemical properties of the nucleoside in 2 ways, the additional imino group in the base provides an extra hydrogen bond donor and the carbon-carbon glycosidic bond is more stable than the nearly universal nitrogen-carbon bond.⁴ Compared to uridine, these differences improve local base stacking, rigidify the phosphate-sugar backbone (through bound water molecules) and enhance the stability of pseudouridylated duplexes.^6-12 Although measurable, the differences between uridine and Ψ containing RNAs are subtle and Ψs are thought to fine-tune the function of their RNAs (see below).

Guide RNA Machinery

H/ACA RNP overview

All H/ACA RNPs consist of one name and function specifying H/ACA RNA and the same 4 core proteins (Fig. 2). The core proteins are essential for the stability of all H/ACA RNAs and of each other, and for catalysis of pseudouridylation. Surprisingly, it appears that for the majority of H/ACA RNPs, at least in mammalian cells, the enzymatic activity of the pseudouridine synthase is irrelevant but its function in maintaining the RNAs is paramount. These structural classes of H/ACA RNPs include the telomerase RNP, intron-encoded Alu (AluACA) RNPs and small nucleolar-long noncoding (sno-lnc) RNPs (Fig. 2)^13-15. Although it is not clear on which side the orphan H/ACA RNPs will come down, structural H/ACA RNPs outnumber catalytic ones in diversity by about 4 to 1, whereas the catalytic RNPs are far more abundant in terms of copy number. While the structural aspects of the RNPs discussed below apply to all H/ACA particles, we will focus here specifically on those guiding pseudouridylation, i.e., H/ACA small nucleolar (sno) and small Cajal body-specific (sca) RNPs. For information on all H/ACA RNPs, we refer to the following reviews.^16-22

Figure 2.

H/ACA RNPs. Schematic of the different classes of H/ACA RNPs defined by the H/ACA RNAs. One set of the 4 core proteins assembles with each hairpin of every H/ACA RNA. Guide sequences of the RNAs are indicated (thickening of lines) and the snoRNAs and scaRNAs that are the focus of this review are highlighted (bold). The functions, targets, and importance, where known, are listed underneath. The approximate number of species within each class of H/ACA RNA (which in most cases is still growing) is specified in parentheses. Inset: model of the structure of a single hairpin of a human H/ACA RNP. The color code of the core proteins is the same as in the schematic. The RNA is in gray and the ACA triplet at the bottom of the molecule is in purple. The catalytic aspartate of the green Ψ-synthase is highlighted (yellow). Archaeal and bacterial structures served as basis for the model.¹⁴⁰

Guide RNAs

In RNA-guided pseudouridylation, site-directed base pairing of small H/ACA RNAs with the target RNA positions the uridine at the active site of the modification machinery. Identified in 1996/7,^23,24 H/ACA RNAs have a median length of 133 nucleotides (137.6 +/-18.8; average +/-standard deviation)²⁵ forming an evolutionarily conserved hairpin-hinge-hairpin-tail secondary structure that carries a consensus ANANNA sequence in the hinge region (H box) and an ACA triplet exactly 3 positions from the 3′-end (ACA box) (Fig. 1B). Identification of target uridines is achieved by 2 3-10 nucleotide-long antisense elements in the bulge of one or both hairpins (pseudouridylation pocket) directly 5′ and 3′ of the upper stem (Fig. 1B). These 2 elements hybridize to sequences immediately 5′ and one nucleotide 3′ of the target uridine thereby framing it.^26,27 This positions the target uridine 14-15 nucleotides from the H or ACA box.

Although this is the basic structure and guide mechanism for most H/ACA RNAs, there are exceptions. For example in archaea, H/ACA RNAs are simply termed small (s) RNAs and consist mostly of single hairpins but may have up to 3 hairpins.^28,29 In mammalian cells, H/ACA RNAs that guide pseudouridylation are snoRNAs, which guide the modification of rRNA, and scaRNAs, which guide the modification of spliceosomal snRNAs. A fraction of some of the sno and scaRNAs is further processed to pre-microRNAs (Fig. 2, gray boxes). In yeast, all H/ACA RNAs are snoRNAs, irrespective of target RNA. A few of the H/ACA scaRNAs are hybrids that also contain C/D boxes and that guide both pseudouridylation and 2′-O-methylation.^30,31 This is particularly of practical importance because isolation of box H/ACA RNPs will therefore also yield box C/D components and vice versa.³²

H/ACA core proteins

The central core protein is the enzyme of H/ACA RNPs, the pseudouridine synthase. It was originally identified in yeast and mammalian cells before its function was known and subsequently revealed its true identity through sequence homology to the bacterial pseudouridine synthase specific for U55 in tRNA and through genetic analysis in yeast.^33-37 The protein goes under a variety of names, NAP57 and dyskerin in mammalian cells, Cbf5 in yeast, protists and archaea, minifly and Nop60B in fly, and as the stand-alone enzyme TruB in bacteria.^33-35,38-42 The 514-amino acid human NAP57 (Fig. 2, green) features highly charged, lysine-rich N- and C-termini, a central catalytic domain followed by an RNA-binding pseudouridylase and archaeosine transglycosylase (PUA) domain. The X-ray structure of the bacterial, archaeal and yeast enzymes show a high degree of structural conservation, particularly in the catalytic and PUA domains.^43-48 The charged and low complexity terminal extremes are apparently unstructured and were omitted. The PUA domain and part of the N-terminus combine to form a domain that is separate from the catalytic module. The catalytic domain is conserved in all pseudouridine synthases including all stand-alone enzymes.^22,49 The 2 domains form an L-shaped structure with one arm representing the catalytic (Fig. 2, dark green) and the other the N- and C-terminal domains (light green).

The other 3 core proteins are NOP10, NHP2 and GAR1. Like NAP57, they are all essential proteins and depletion of their yeast orthologs, with the exception of GAR1,^50,51 causes the loss of all H/ACA RNAs. NOP10 (Fig. 2, red) is a 64-amino acid polypeptide that stabilizes and lines the catalytic domain of NAP57 serving as a dock for NHP2 binding.^43-45,47 NHP2 (Fig. 2, yellow) is a 153-amino acid protein with a phylogenetically conserved central RNA binding domain, which in the case of NHP2 binds RNA non-specifically.^52,53 Unlike the other H/ACA core proteins, NHP2 remains stable when the other core proteins are depleted in yeast, but depletion of NHP2 leads to loss of the other 3 proteins.^54,55 The 217-amino acid long GAR1 (Fig. 2, light blue) also binds to the catalytic domain of NAP57 but independently of NOP10 and NHP2. As suggested by its name, the hallmarks of GAR1 are its 2 terminal glycine–arginine-rich domains that together account for about half of the protein.^50,51 GAR1 can contact the active site of the particle and is required for substrate turnover during the pseudouridylation reaction^48,53 (see below).

Structure

The crystal structures of many pseudouridine synthases have been solved. In case of NAP57 homologs, this starts with the structure of the bacterial TruB in complex with part of a tRNA, then encompasses the structures of several archaeal partial and complete H/ACA protein complexes with guide RNA and substrate RNA and finally, the structure of a yeast H/ACA RNP lacking Nhp2.^{43-46,48,56-58} Although the structure of the human complex is still elusive, copious information can be gained from these structures because of the high degree of phylogenetic conservation of the enzymes and RNPs. All structures include only a single hairpin of an H/ACA RNA leaving the organization of the intact particle open to speculation. However, based on low resolution structures derived from negative stained EM images of purified yeast H/ACA RNPs, the holoparticle forms a v-shaped structure with each arm apparently representing one hairpin and 4 core proteins.^41,59 Therefore, as documented for the telomerase RNP, one H/ACA RNA accommodates 2 sets of the 4 core proteins to yield an intact eukaryotic H/ACA RNP.⁶⁰ The structure of a human H/ACA RNP modeled after that of the complete structure of an archaeal RNP shows the organization of the proteins and the single hairpin RNA (Fig. 2, inset). NAP57 forms the core of the particle, with NOP10 and GAR1 binding independently to its catalytic domain and NHP2 docking on NOP10. The RNA is draped over the core trimer of NAP57, NOP10, and NHP2 without seemingly engaging GAR1. The conserved ACA triplet is anchored in the PUA domain of NAP57, whereas the loop of the hairpin binds to NHP2 thereby placing the pseudouridylation pocket over the central catalytic domain (Fig. 2, inset). Although this structure is firmly established through crystallization of several homologs and based on biochemical evidence, it is somewhat misleading as significant parts of the individual proteins were truncated or are not visible in the structure. For example, the charged termini of eukaryotic NAP57 and the glycine–arginine-rich domains of GAR1 are missing from the archaeal orthologs and were removed from the yeast proteins for crystallization purposes. Nevertheless, these “partial” structures are functional readily pseudouridylating substrate RNAs.^48,61,62 In general, H/ACA RNPs are very stable. Although all core proteins, except NAP57 can exchange, the H/ACA RNA cannot be replaced without disassembly or degradation of the particle, i.e. requiring de novo synthesis.^53,63

Mechanism

Isomerization of uridine to Ψ requires significant molecular gymnastics, the breakage of the N-glycosidic bond and the 180° rotation of the base. Based on the universal conservation of an aspartic acid residue in the active site of all pseudouridine synthases identified to date, all proposed mechanisms involve a nucleophilic attack by the aspartate. However, the site of this attack has been questioned, be it at the base (Fig. 1A, C6) or at the ribose (C1′).⁶⁴ This confusion has been mainly brought about by the use of 5-fluorouridine (5-FU) in substrate RNAs. Such substrates lead to covalent attachment of 5-FU to some pseudouridine synthases concomitant with their inhibition, whereas others efficiently convert 5-FU to 5-fluoro-6-hydroxy-Ψ (5FhΨ).^46,65,66 The latter is true for the TruB class of enzymes to which NAP57 belongs. Indeed, in vitro pseudouridylation experiments with 5-FU substrates indicated that also mammalian H/ACA RNPs isomerize substrates with 5-FU at their target sites⁶⁷ (Wang and Meier, unpublished results). The more recent identification of arabinose (instead of ribose) in a minor product of 5-FU isomerization by TruB, suggests a reaction mechanism wherein the nucleophilic attack by the aspartate occurs at C1′ of the ribose ring and proceeds via a glycal or acylal intermediate.^68,69 However, analysis of structures of uridine substrate analogs bound to archaeal H/ACA RNPs reveals a closer proximity of the catalytic aspartate to the C6 of the base favoring the originally proposed mechanism.^65,70 So, the discussion about the reaction mechanism continues. How the enzymes accomplish the separation and rotation of the base will have to await more sensitive technology that allows following detailed conformational changes, live and in structures.

Regardless, it has become abundantly clear that the target uridine within a substrate RNA is being flipped out for isomerization into the active pocket of the enzymes, be that in the stand-alone enzymes or in H/ACA RNPs.^46,56,57 An important role in substrate RNA binding and turnover is played by a thumb-loop element (b7_10 loop) in the catalytic domain of both kinds of pseudouridine synthases, guide-RNA dependent and independent.^46,56,57,71 Thus, upon substrate binding, the thump-loop swings over to lock the substrate RNA in place. In H/ACA RNPs this movement is supported by GAR1, which is required for multi-turnover reactions.^57,72 The central conserved domain of GAR1 is sufficient for the reaction, with its 11 C-terminal residues playing a crucial role by directly interacting with the thumb loop.⁴⁸ This interaction is essential for the multi-turnover reaction of the enzyme complex.⁴⁸ Unlike the rest of the central GAR1 domain, these 11 amino acids are not conserved in the assembly factor NAF1, which, while binding to the same domain of NAP57, is unable to support pseudouridylation.^48,73,74 The function of assembly factors is explored in the next section.

Biogenesis

Biogenesis of the only 5-component pseudouridylation guide machinery is surprisingly complex involving at least 6 assembly factors in the cytoplasm and nucleus (Fig. 3). NAF1 and SHQ1 are H/ACA-specific chaperones and were identified in yeast through their interaction with the yeast NAP57 ortholog, Cbf5.^74-76 They are required for stable expression of all H/ACA RNAs without being part of the mature particles in nucleolus and Cajal bodies. SHQ1 functions first by tightly binding to NAP57, apparently right after translation.⁷⁷ Doing so, SHQ1 achieves 2 purposes, protecting the degradation and aggregation prone NAP57 and physically preventing non-cognate RNAs from binding to NAP57 before its cotranscriptional association with its guide RNAs. Specifically, SHQ1 is an RNA mimic taking up the same position on NAP57 as an H/ACA RNA will in the mature RNP and forming many of the same molecular interactions (compare insets in Figs. 2 and 3).^78,79 Thus, the C-terminal SHQ1-specific domain (SSD, 487 amino acids) of SHQ1 binds predominantly to the domain of NAP57 formed by its C-terminal PUA and N-terminal parts (Figs. 2 and 3, light green). Additionally, the HSP20-like N-terminal CS domain of SHQ1 (90 amino acids) binds to the same domain of NAP57 but on the opposite side thereby closing a tight clamp around NAP57.^77,80-82 In fact, the NAP57-SHQ1 interaction is so tight that its resolution depends on additional chaperones, the R2TP complex.⁸²

Figure 3.

Schematic of the surprisingly complex biogenesis of H/ACA RNPs. Details of the at least 4-step process are described in the text. The approximate size and sites of interactions of the proteins is drawn to scale. Inset: Surface rendering of the structures of yeast NAP57 alone and in complex with the SHQ1 specific domain (SSD; pdb: 3uai). Amino acid positions equivalent to those mutated in human NAP57 in patients with dyskeratosis congenita are highlighted (orange). Note many of these positions cluster in the C-terminal segment (CTS, knob on the left) that serves as handle for the SSD, although they are generally poorly visible in this surface rendering.

The R2TP complex consists of the 2 closely related AAA+ ATPases (ATPases associated with diverse cellular activities) pontin and reptin (Rvb1/2, TIP48/49, RuvBL1/2, etc.), PIH1D1 (Nop17, Pih1) and RPAP3 (hSpagh, Tah1), arranged in (hetero)hexameric rings, and is named after the yeast homologs Rvb1, Rvb2, Tah1 and Pih1 (Fig. 3).⁸³ In addition to snoRNP biogenesis of both, box H/ACA and C/D RNPs, the R2TP complex has been implicated in the assembly of RNA polymerase II and of phosphatidylinositol 3-kinase-related kinases (PIKKs, i.e., mTOR and SMG1).^82,84-90 Pontin and reptin alone or together are associated with many other macromolecular assemblies involved in a wide variety of cellular activities, e.g. chromatin remodeling, histone acetyl transfer, transcription, and mitotic spindle formation.^83,91,92 Studies on the separation of the NAP57•SHQ1 complex resulted in mechanistic insight into the function of the ATPases (Fig. 3). They bind to both proteins and require the charged C- (but not N-) terminus of NAP57 for SHQ1 release.⁸² Presumably, freeing of NAP57 from SHQ1 occurs immediately before its association with a guide RNA at the site of transcription to prevent aggregation and binding to non-cognate RNAs.^77-79

At the site of H/ACA RNA transcription, the core proteins NOP10 and NHP2 as well as the assembly factor NAF1 are present in and likely join the nascent RNP complex (Fig. 3).^93-95 As stated above, NAF1 lacks the amino acids for interaction with the thumb loop domain of NAP57 from its central GAR1-homology domain.⁷³ In this manner, NAF1 keeps the pseudouridylation machinery inactive until replacement of NAF1 with GAR1 thus preventing modification of random RNA targets until formation of the mature particles.⁹⁵ The latter, unlike the assembly factors, concentrate in nucleoli and Cajal bodies, their major sites of action. Exactly where in the cell these exchanges and maturation steps occur remains to be elucidated; however, they are essential for viability, as cells depleted of the individual assembly factors lose all H/ACA RNAs and consequently fail to properly process and modify rRNA.^74-77,82,96

Substrates

Overview

Pseudouridylation guide RNAs are known to target mainly 2 kinds of RNAs, rRNA and spliceosomal snRNA.^97,98 In vertebrate cells, these 2 types of RNAs are pseudouridylated mostly (if not exclusively) by an RNA-guided mechanism. Additionally, vertebrate snoRNAs U3 and U8 can also be pseudouridylated by this pathway.^99,100 In contrast, yeast snRNA is pseudouridylated by both RNA-guided and stand-alone enzymes, the latter recognize and modify the RNA without the help of guide RNAs.¹⁰¹ This mixed mechanism of pseudouridylation of the same RNA may mark an evolutionary transition from single-protein to RNA-guided pseudouridine synthases.¹⁰¹ This transition is further evident with the switch to RNA-guided pseudouridylation concomitant with increasing number of modification sites in target RNAs, e.g., from 11 sites in bacterial rRNA (stand-alone enzymes) to ∼50 in yeast and ∼100 in mammalian rRNA (both RNA-guided).^102,103 Despite the extensive pseudouridylation of all tRNAs in all species, archaea offer the only known example of RNA-guided modification of tRNA.^104,105 In summary, only stable noncoding RNAs are targets of RNA-guided pseudouridylation, which shall be the focus of this review. Nevertheless, given the difficulty of detecting partial pseudouridylation in low-abundant RNAs (e.g. mRNAs) and the existence of many H/ACA RNAs without identified target (orphan H/ACA RNAs) or with imprecise complementarity, it is reasonable to predict pseudouridylation extending beyond these types of noncoding RNAs.

Eukaryotic rRNAs

Eukaryotic rRNAs contain a large number of pseudouridines. For instance, there are 46 pseudouridines in S. cerevisiae rRNAs.¹⁰⁶ In mammalian rRNAs, the amount of pseudouridine is even higher; currently, 97 pseudouridines have been identified.¹⁰³ Since the discovery of the box H/ACA RNA-guided pseudouridylation mechanism approximately 17 years ago,^26,27 a great effort has been made to search for new box H/ACA RNAs. Using a combination of computational and experimental approaches, researchers have identified a large number of new box H/ACA RNAs in a variety of organisms. Subsequent computational analysis and visual inspection (comparing the guide sequences of box H/ACA RNAs with the target sequences of rRNAs) have generated nearly complete sets of box H/ACA RNAs that are potentially capable of directing rRNA pseudouridylation, not only in S. cerevisiae, but also in humans.^103,106,107 For some, the predicted guide RNA-substrate relationship has been experimentally verified. For instance, it has been shown that deleting several box H/ACA RNAs individually or in combination results in loss of pseudouridine(s) at expected sites in yeast rRNAs. Importantly, introduction of a plasmid(s) containing the box H/ACA RNA gene(s) into the deletion strain leads to restoration of rRNA pseudouridylation.^{26,27,108-113} Using various reconstitution systems, several labs have recapitulated box H/ACA RNA-guided rRNA pseudouridylation in vitro.^{31,48,61,62,67,113} Although quite successful, the experiments performed to date have tested only a small fraction of predicted box H/ACA RNA-guided rRNA pseudouridylation; for most sites, box H/ACA RNA guided rRNA pseudouridylation has yet to be experimentally verified. However, given the clear base-pairing interactions between the guide sequences of box H/ACA RNAs and their predicted substrate sequences in rRNAs, it is widely believed that eukaryotic rRNA pseudouridylation is catalyzed mostly (if not exclusively) by the box H/ACA RNA-guided mechanism.

Eukaryotic spliceosomal snRNAs

Pseudouridine is the most abundant modified nucleotide in eukaryotic spliceosomal snRNAs.¹⁰⁰ There are a total of 24 pseudouridines in the 5 vertebrate snRNAs (U1, U2, U4, U5 and U6), and 13 of them are concentrated in U2 snRNA accounting for approximately 7% of all nucleotides.^100,114 A total of 6 pseudouridines have been identified in S. cerevisiae snRNAs (3 in U2, 2 in U1 and one in U5).¹¹⁵ Over the years, an intense search for enzymes responsible for snRNA pseudouridylation has been carried out, resulting in the identification of a large number of snRNA-specific box H/ACA RNAs in various eukaryotic organisms, including, among others, S. cerevisiae, X. laevis, mouse, human, and fruit fly.^{101,107,116,117} Some of the identified putative box H/ACA RNAs have been experimentally proven to guide spliceosomal snRNA pseudouridylation. For instance, it has been shown that depletion of pugU2-34/44, a box H/ACA RNA that was predicted to guide U2 pseudouridylation at positions 34 and 44, results in the loss of Ψ34 and Ψ44 in U2 snRNA in Xenopus oocytes; injection of in vitro-transcribed pugU2-34/44 into the depleted oocytes restores the formation of Ψ34 and Ψ44 in U2 snRNA.¹¹⁷ Likewise, it has also been experimentally demonstrated that mammalian U85 is capable of directing U5 pseudouridylation at position 46 both in vitro and in vivo.³¹ Recently, the Gall group has identified several box H/ACA RNAs in Drosophila, and shown experimentally that these RNAs are capable of guiding spliceosomal snRNA pseudouridylation in fruit fly.¹¹⁶ Although a large number of snRNA-specific H/ACA RNAs have been identified, there are still many known pseudouridylation sites in spliceosomal snRNAs that are unaccounted for. Thus, it remains a challenge to identify the missing H/ACA RNAs for these sites. However, it is still possible that Ψ formation at some of these sites is not guided by H/ACA RNAs. In this regard, of the 3 characterized pseudouridylation enzymes responsible for S. cerevisiae U2 pseudouridylation, only one is a box H/ACA RNP (snR81), catalyzing the formation of Ψ42¹⁰¹; the other 2 are stand-alone protein enzymes (responsible for the formation of Ψ35 and Ψ44, respectively).¹¹⁸ Similarly, pseudouridylation of S. cerevisiae U6 snRNA (Ψ28) is catalyzed by the stand-alone Ψ synthase Pus1.¹¹⁹ Further research is necessary to address this issue.

Potential substrates

Both computational and experimental approaches have been used in the search for novel box H/ACA RNAs. Because computational searches often involve sequence alignments with known pseudouridine sites/sequences in rRNAs and spliceosomal snRNAs, the thus identified H/ACA RNAs are specific for those already known. In contrast, the experimental approach uses biochemical techniques to detect and identify any H/ACA RNAs present in the cell, and is thus unbiased. As a result, although many biochemically identified H/ACA RNAs match known pseudouridine sites, many others have no known target sites and are thus called orphan H/ACA RNAs¹⁰⁷ (Fig. 2). It is quite possible that these orphan H/ACA RNAs are able to target other types of RNA (other than rRNA and snRNA) for pseudouridylation. In this sense, it is expected that a simple BLAST search against RNA databases (e.g. mRNA database) will generate a number of novel target sites. However, whether these predicted sites are in fact pseudouridylated will have to be experimentally tested. Finally, an entire new set of pseudouridylation sites are potentially out there due to imprecise base pairing of known H/ACA guide RNAs, exemplified by the case of regulated pseudouridylation below (Fig. 1C).

Artificial substrates

In an effort to study possible mRNA/pre-mRNA pseudouridylation, Karijolich and Yu designed an artificial H/ACA RNA, derived from snR81 (a verified endogenous H/ACA RNA), to target mRNA in S. cerevisiae.¹²⁰ All sequence elements remain unchanged from the original gene, with the exception of the single-stranded guides, which are targeted to specific mRNA sequences. Experiments using this system have demonstrated that the artificial H/ACA RNA is capable of directing mRNA pseudouridylation at the target site. Although the modification efficiency is not as robust (∼7–10%) as in rRNAs and snRNAs, the fact that the artificial H/ACA RNA can target mRNA pseudouridylation indicates that different types of RNA can indeed be modified, and that these additional RNAs may even be naturally modified, perhaps at low enough levels that have eluded detection by current techniques. Likewise, using a similar strategy, Chen et al. show that such artificial H/ACA RNAs can also site-specifically pseudouridylate pre-mRNA in Xenopus oocytes.¹²¹ BLAST searches using the guide sequences of some known S. cerevisiae H/ACA RNAs have revealed a handful of complementary sequences in naturally occurring S. cerevisiae mRNAs, suggesting that these sites might indeed be pseudouridylated.¹²⁰ Unfortunately, the pseudouridylation assays currently available are generally not sensitive enough to check the modification status of low-abundance mRNA species, and thus, development of more sensitive pseudouridylation assays is necessary to clarify this issue.

Function of Pseudouridines

Overview

Given the fact that pseudouridylation endows the modified uridine with chemical properties that are distinct from those of all other known nucleotides, introducing Ψ(s) into an RNA will likely alter its function. Close inspection of the sequences and structures of both rRNAs and spliceosomal snRNAs has revealed that Ψs are not only abundantly present in these RNAs, but are also clustered in regions/structures of functional importance,^100,122 thus further underlining the notion that pseudouridylation may affect RNA function. Indeed, over the years, it has been reported that pseudouridines contribute significantly to the functions of rRNAs and spliceosomal snRNAs in protein translation and pre-mRNA splicing, respectively.

In rRNA

The consequences of Ψ deletion from rRNA have been studied in yeast ribosomes, where Ψs can be individually and globally ablated. For instance, the Fournier group has shown that the deletion of 5 H/ACA RNAs responsible for the formation of pseudouridines in the ribosome peptidyl transferase center results in defects in protein translation and cell growth.¹⁰⁹ Likewise, eliminating pseudouridylation in either the ribosome decoding center of 18S rRNA and the A-site finger region of 25S rRNA leads to functional defects.^111,123 Surprisingly, ablation of Ψs from helix 69 of 25S rRNA demonstrated variable effects, such that deletion of 5 Ψs showed less effect on cell growth/translation than that of only 3, even if they were contained within the 5 Ψs, highlighting the subtleties and intricacies of Ψ function.^110,112 Furthermore, the Dinman group has shown that hypo-pseudouridylated rRNAs—resulting from catalytically defunct H/ACA RNPs—form impaired ribosomes that exhibit low affinity for tRNA and poor fidelity in protein translation.¹²⁴

In spliceosomal snRNA

The functional importance of snRNA-specific H/ACA RNPs has also been extensively studied over the past 17 years. For instance, depletion of pugU2-34/44 from Xenopus oocytes results in the loss of 2 pseudouridines (Ψ34 and Ψ44) in the U2 branch site recognition region, which, in turn, impacts pre-mRNA splicing.^117,125 Deletion of snR81 from yeast cells eliminates the formation of Ψ42 in the U2 branch site recognition region, creating a mutant strain, which, when combined with a U2 point mutation, exhibits a synthetic growth (and splicing) defect phenotype.¹²⁶ In an exciting recent observation, a single Ψ residue in U6 snRNA was demonstrated to initiate the yeast filamentous growth program highlighting potential effects of Ψs in development.¹¹⁹

Artificially introduced

When artificially introduced into stop codons, Ψs cause nonsense suppression.¹²⁰ X-ray crystallography revealed that pseudouridine is capable of promoting unusual base-pairing interactions between the stop codon and the tRNA anticodon in the ribosome decoding center, allowing effective amino acid incorporation at the pseudouridylated stop codon.¹²⁷ Experimental results, although restricted to artificial H/ACA RNAs targeting reporter mRNAs, have demonstrated that mRNA pseudouridylation can play an important role in protein translation.¹²⁰ For the reasons discussed above, pseudouridylation probably occurs naturally to mRNA, and thus, pseudouridine-mediated translation recoding is likely to be biologically relevant. One consequence of pseudouridine-mediated codon-specificity change is proteome diversity: by pseudouridylating specific codons, multiple proteins could be produced from a single mRNA (gene), representing gene regulation at the RNA modification level.

Regulation

For decades, RNA pseudouridylation, including RNA-guided RNA pseudouridylation, had been considered a constitutive process. However, this notion was challenged 3 years ago, with the discovery of inducible pseudouridylation.^128,129 Specifically, S. cerevisiae cells, when stressed, can catalyze RNA pseudouridylation at novel sites. For instance, when cells are deprived of nutrients or grown to stationary phase, novel pseudouridylation activities are activated, resulting in spliceosomal snRNA pseudouridylation at several new sites that are otherwise unmodified when cells are grown in nutrient-rich media (or under log-phase conditions). One of these induced pseudouridylation activities involves snR81 H/ACA RNP, which catalyzes U2 pseudouridylation at position 93 under stress conditions. Interestingly, in this guide-substrate base-pairing scheme, there are 2 mismatches (Fig. 1C). It appears that the 2 mismatches are required for induced pseudouridylation, as changes from 2 mismatches to 3 (or more) mismatches, or to complete complementarity, result in no pseudouridylation or constitutive pseudouridylation at position 93, respectively. Importantly, induced spliceosomal snRNA pseudouridylation seems to have a negative impact on pre-mRNA splicing.

Thus, besides the known constitutive pseudouridines, at least 2 new inducible pseudouridines have been identified in spliceosomal snRNAs.¹²⁸ The fact that Ψ93 of U2 snRNA affects pre-mRNA splicing suggests that, like the regulatory modifications of DNA and protein, induced pseudouridylation of RNA may play an important regulatory role in gene expression. However, the question of whether induced pseudouridylation is reversible remains to be addressed. Given that induction requires imperfect or less stringent base-pairing (2 mismatches) between the guide sequence and the substrate sequence, it is possible that induced pseudouridylation is widespread (perhaps also occurring to mRNA) and that current computer algorithms and visual inspection are inadequate to identify these sites. Further investigation is necessary.

Disease

The pseudouridine synthase NAP57 is the target of the predominant and severe X-linked form of dyskeratosis congenita (X-DC). Hence, NAP57 is also known as dyskerin and its gene is termed DKC1.⁴⁰ DC is an inherited bone marrow failure syndrome caused mainly by the depletion of haematopoietic but also other stem cells. The classic phenotypical triad of abnormal skin pigmentation, mucosal leukoplakia, and nail dystrophy manifests itself in the first 2 decades of life, but is often followed by severe anemia in the third decade with dire consequences.^130-132 In the autosomal forms of DC, mutations have been identified in 5 other H/ACA RNP components, the core proteins NOP10 and NHP2, telomerase RNA (which ends in an H/ACA domain), telomerase reverse transcriptase, and WDR79 (WRAP53, TCAB1; responsible for binding to the CAB box motif of scaRNAs and for their Cajal body localization). Additionally, the telomere proteins TIN2, CTC1 and RTEL1 are affected.^133,134 Given the abundance of telomeric targets in patients with DC and Hoyeraal-Hreidarsson syndrome (HH, a severe variant of DC), it is not surprising that telomeres are affected in these individuals.^135,136 However, it is still debated to what extent translation and pre-mRNA splicing are impaired through insufficiently modified rRNA and snRNA, respectively.^137,138 In this context, it was recently demonstrated that even minor fluctuations in protein synthesis rate can impair the function of haematopoietic stem cells.¹³⁹

Regardless, mutations in the NAP57 gene DKC1 account for some 40% of all known DC mutations and include cases of HH. Most of the 55 missense mutations identified so far cluster in the N- and PUA containing C-terminal domain, which is enveloped by SHQ1 in the NAP57•SHQ1 complex (Fig. 3, inset).¹³⁴ In fact, about a quarter of all X-DC amino acid substitutions are present in the short C-terminal segment (CTS, 35 amino acids) that is right adjacent to the PUA domain and serves as a handle for binding of the SSD of SHQ1 (Fig. 3 inset, left knob).^78,79 This clearly implicates the interface between NAP57 and SHQ1 as a target of X-DC. Indeed, also binding of the N-terminal CS domain of SHQ1 is affected by X-DC mutations.^82,140 A consequence of impaired or enhanced binding of NAP57 to SHQ1 is reduced cellular availability of NAP57 through degradation or sequestration. This predicts insufficient NAP57 for the stabilization of all cellular H/ACA RNAs, such that the least abundant RNAs, i.e. telomerase RNA, may lose out. Such a mechanism is analogous to that proposed for spinal muscular atrophy where reduced levels of the survival of motor neuron protein (SMN) affects levels of snRNAs varyingly and in a cell type-specific manner.¹⁴¹ Moreover, when focusing on the molecular consequences of single mutations, it is important to keep in mind that bone marrow failure is additionally strongly influenced by genetic background. This is perhaps best illustrated by the alanine 353 to valine mutation in NAP57 that accounts for about half of X-DC cases. Thus, the penetrance of this amino acid substitution ranges from causing growth retardation and immunodeficiency in utero (in HH) to going unnoticed until over 30 years of age (in DC).¹⁴² Additionally, the equivalent amino acid position in the yeast ortholog Cbf5p is already a valine. Obviously, there is an enormous complexity underlying the pathogenesis, which we have yet to understand. Perhaps, whole genome sequencing will make a significant contribution. Another confounding issue is the fact that with the many NAP57 mutations adorning the interface of the 2 proteins, no mutations in SHQ1 have been observed in DC.¹⁴³ Does SHQ1 have other functions that elude this simple concept?

In the case of somatic mutations observed in several tumors, the NAP57-SHQ1 interaction is affected by mutations in both proteins. Thus, in an integrative genomic profiling study of prostate cancer, the chromosomal 3p region that contains SHQ1 was identified as a tumor suppressor in conjunction with the androgen-driven serine protease-transcription factor fusion TMPRSS2-ERG, which is observed in some 50% of prostate cancers.¹⁴⁴ Importantly, SHQ1 was the only gene in the 3p region that also harbored tumor associated mutations.^144-146 Indeed, the growth-suppressive properties of SHQ1 and its loss in primary prostate tumors has been observed.^147,148 A more general role of the 3p region in tumorigenesis is supported by its tumor suppressing activity in the pathogenesis of lung and other cancers, as well as by the acquisition of invasiveness in cervical cancer after its loss.^149,150 Importantly, tumor associated mutations have now also been identified in NAP57 (and more in SHQ1) in prostate, colon, lung, uterine and glioblastoma tumors.^145,151-153 Many of these amino acid substitutions lie in the NAP57-SHQ1 interaction domains. However, even some mutations outside affect the interaction of the recombinant proteins likely through allosteric effects (Machado-Pinilla and Meier, in preparation). Finally, it is interesting to note that DC also predisposes to malignant tumor formation, possibly providing an intersection between the somatic and inherited mutations observed in NAP57. Regardless, nature teaches us that fine-tuning of the NAP57-SHQ1 interaction and with it the biogenesis of H/ACA RNPs is crucial for the wellbeing of the cell and organism.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Funding

The work in the authors’ laboratories is supported by grants from the National Institutes of Health (GM104077 and AG039559 to Y.T.Y. and GM097752 to U.T.M.)