Pseudouridine: Still mysterious, but never a fake (uridine)!

Abstract

Pseudouridine (Ψ) is the most abundant of >150 nucleoside modifications in RNA. Although Ψ was discovered as the first modified nucleoside more than half a century ago, neither the enzymatic mechanism of its formation, nor the function of this modification are fully elucidated. We present the consistent picture of Ψ synthases, their substrates and their substrate positions in model organisms of all domains of life as it has emerged to date and point out the challenges that remain concerning higher eukaryotes and the elucidation of the enzymatic mechanism.

Keywords:

• enzymatic mechanism
• modified nucleoside
• pseudouridine
• regulation
• rRNA
• RNA Modification
• tRNA
• snRNA

Abbreviations

Pus	=	pseudouridine synthase
Ψ	=	Psi, pseudouridine
rRNA	=	ribosomal RNA
snRNA	=	small nuclear RNA
E. coli	=	Escherichia coli
S. typhimurium	=	Salmonella typhimurium
S. cerevisiae	=	Saccharomyces cerevisiae
H. volcanii	=	Haloferax volcanii and/or Halobacterium volcanii.

Introduction

More than 150 nucleoside modifications fine-tune conformation, structure and function of RNA.^1,2 In 1951 the first modified nucleoside was discovered in RNA hydrolysate³, shortly after termed the fifth nucleoside,⁴ identified as 5-ribosyl uracil⁵ and named pseudouridine (Ψ).⁶ Eventually the development of a tritium release assay for Ψ formation led to the identification of the first pseudouridine synthase gene,⁷ HisT, later renamed to TruA, which modifies tRNA in S. thyphimurium⁸ and E. coli.⁹ Although Ψ is the most abundant nucleoside modification,¹ the actual advantage of pseudouridylation that warrants this abundance, remains hard to grasp and is usually described as stabilization by ‘additional hydrogen bonds’ and ‘improved base stacking’.¹⁰

The importance of Ψ is reflected and documented in the variety of existing reviews, be it general,¹¹ centered on structural biology of either stand-alone protein Ψ synthases^12,13 or H/ACA box ribonucleic particles (RNPs),¹⁴ or, even more recently, focused on H/ACA box RNPs and Ψ formation and function in snRNA and rRNA.¹⁵

In the last 2 decades a more consistent picture of Ψ synthesis and Ψ distribution in model organisms of all domains of life has emerged, of which the outlines will be presented here. Despite significant progress however, a clear catalytic role assignment to amino acids is still lacking, and hence the catalytic mechanism of Ψ formation remains elusive even now, almost 15 years after publication of the first cocrystal structure.¹⁶ We will outline why the elucidation of this mechanism remains a challenge, while research on Ψ is about to move to complex organisms and transcriptome wide analyses.

Physicochemical Properties of Ψ

Pseudouridine is a C-C glycosidic isomer of uridine (U), and the isomerization reaction, which incorporates the C5 into the glycosidic bond, is shown in Figure 1. Both nucleosides share a similar UV spectrum⁵ and identical molecular mass⁵, but differ in mass spectrometric dissociation.^17,18 Early methods for the detection of Ψ were based on random alkaline hydrolysis followed by TLC detection of ³²P-labeled nucleotides¹⁹, rendered semi-quantitative by biased, non-quantitative hydrolysis and incomplete labeling of Ψ.²⁰ In a known sequence context, random hydrolysis can be substituted by site-specifically cleaving DNAzymes²⁰ or RNase H²¹ or by making use of the decreased ligation efficiency of a complementary probing strand, thereby circumventing cleavage.²² The arguable most popular, albeit technically demanding technique for sequence specific Ψ detection includes specific derivatization of Ψ with CMCT followed by primer extension.²³ This technique found recent application in genome-wide pseudouridine profiling by deep sequencing in yeast and human.^24,25 Specific derivatization of Ψ with CMCT, acetonitrile, or methylvinylsulfone is also applied in sequence specific detection via LC-MS approaches, (reviewed in ref. ¹⁷). Incomplete reactions, side products, and unstable response factors prevent quantitative analysis by this approach. In consequence there is an increasing interest in derivatization-free MS/MS-based approaches that allow quantitative analysis,^26,27 and which have recently included isotope labeling^28,29

Figure 1. Isomerization of uridine into pseudouridine (Ψ). Post-isomerization several derivatives discovered to date¹ can be formed by further modification at either position 1 (R₁), 3 (R₂) or 2’-O (R₃), while several modifications at once are possible.

So called hypermodified Ψ derivatives are formed via further modification of Ψ. At present, they comprise Ψm, m¹Ψ, (found in all domains of life, predominantly in tRNA¹), m³Ψ (in rRNA of Eubacteria¹), as well as 3-(3-amino-3-carboxypropyl)-Ψ (αχαπ3Ψ)^30,31 and m¹acap³Ψ¹ in rRNA of eukaryotes (Fig. 1).

General Function/Structural Aspects of Ψ

Although the identical Watson-Crick faces of Ψ (Fig. 1) and U enable both to engage in classical Watson-Crick base pairing with adenosine (A), Ψ base pairs with any of the 4 major bases^32,33 are more stable than their U equivalents. For Ψ-A base pairs, NMR revealed that NH1, which is situated in the major groove, was being protected from proton exchange with solvent water.^10,34,35 This protective effect is probably caused by hydrogen bonding of ΨNH1 to the 5’-phosphate oxygen atoms via water, for which several lines of evidence lend support.^11,33,36-40 Thus, conferred increased backbone rigidity may be the cause for a presumably secondary effect of Ψ formation: improved base stacking, which was concluded from a preference of the 3’-endo-conformation.¹⁰

The Presence of Ψ in Various RNAs

Ψ was first identified in rRNA,¹ (recently reviewed by Ge and Yu¹⁵) and tRNA.¹ Further occurrences of Ψ are known in small nuclear RNAs of various eukaryotes, as is reviewed in, e.g., refs ^15,41. As an example, the Ψ in spliceosomal branch site of U2 snRNA will be discussed below in some detail. Further RNAs containing Ψ include snoRNAs U3 of rat and U8 of mouse¹, tRNA-like domains of plant viruses,^42,43 SRA RNA,⁴⁴ and human telomerase RNA,^25,45 long non-coding RNAs and mRNA.^24,25 The following section will illustrate the distribution of Ψ in tRNA and rRNA, along with their respective enzymes, in all domains of life based on model organisms. In the subsequent sections, the functional and structural aspects of Ψ will be discussed in more detail.

Enzymatic Formation of Ψ residues

Enzyme families

Six families of pseudouridine synthases (Pus enzymes) have been identified, each named for a prominent representative: TruA, TruB, TruD, RsuA, RIuA, (reviewed in ref. ¹²) and Pus10p.⁴⁶ They share the same overall fold and require an active site aspartate for catalysis,^12,47 implying a common mechanism, to which we will turn our attention later. Different N- or C-terminal domains govern substrate specificity, as reviewed in ref.¹² In contrast, few Ψ-hypermodification enzymes are known: E. coli m³Ψ methyltransferase RImH^48,49 and 3 m¹Ψ methyltransferases in Archaea^50,51 and yeast.⁵²

The most versatile enzyme family may be the ribonucleic particles (RNPs) depicted in Figure 2. These particles contain a subgroup of small nucleolar RNAs (snoRNAs), called H/ACA RNAs, and were proven to catalyze Ψ formation, at first in eukaryotes,^53,54 later in Archaea.⁵⁵ The snoRNA (called sRNA in Archaea) acts as guide for the protein components with Nop10 and the Ψ synthase NAP57 (higher eukaryotes) or Cbf5 (yeast, Archaea) as minimal requirements.^56,57 Non-essential components Gar1 and L7Ae (or Nhp2 in Eukarya) are involved in catalysis and product release⁵⁸ or in substrate binding by interaction with Nop10^59-61, respectively. Cbf5 is also capable of guide RNA free catalysis, the activity of which is increased by Gar1 and Nop10.⁶² Investigation of guide RNA specificity⁶³ enabled artificial guide RNAs to target specific uridines for Ψ formation.^64-66

Figure 2. Structure of the archaeal ACA RNP¹⁹⁸ (left) and the eukaryotic H/ACA RNP¹⁹⁹ (right). Guide RNA in black, substrate RNA turquoise. Catalytically active component is light blue Cbf5 (NAP57)

From biochemical data⁶⁷ and crystal structures⁶⁸ a specific degradation pathway for Ψ in Eubacteria is evident: Ψ is first phosphorylated by a dedicated kinase and subsequently converted to uracil and ribose-5’-phosphate. The remarkable cleavage of a C-C glycosidic bond was reported to be reversible⁶⁷ and to proceeds via a ribose ring opening mechanism.⁶⁸ Mammals do not degrade Ψ, but urinary excrete the intact nucleoside.^69,70 Recently, a pseudouridine-5’-phosphatase that dephosphorylates Ψ in human was described.⁷¹ As assays performed in cell extracts indicated conversion of pseudouridine-5’-phosphate into triphosphate,⁷² dephosphorylation might prevent accidental incorporation of pseudouridine into RNA transcripts. ⁷¹

Occurrence and formation of Ψ in model organisms of all domains of life

As indicated above, pseudouridine formation in cellular RNAs is ensured either by stand-alone protein enzymes or by H/ACA sno(s)RNA-dependent RNP particles or by both. In Eubacteria, E. coli is taken as a model (Fig. 3), Ψ synthases acting on RNA belong to 5 distinct families, the Pus10-related family was not detected. Altogether 11 enzymes ensure complete modification of tRNAs and rRNAs in Eubacteria. Pseudouridine modification of other eubacterial RNAs have never been reported in the literature. Although no knockout of a single Ψ synthase has proven to be lethal in Eubacteria, certain single-knockouts suffer from disadvantages compared to their unaffected counterparts.^73-75

Figure 3. Distribution of Ψ and Ψ synthases in E. coli: Enzymes and their substrates positions color-coded: TruA in purple, RIuA and family members green, RsuA family members orange, TruB blue, all reviewed in,²⁰⁰ and TruD¹²⁴ yellow. Substrate residues of RIuD are shown in a dashed box to indicate model helix H69.

In this light it may seem surprising that the total number of pseudouridine modification sites is much higher, and thus many enzymes demonstrate so-called region-specificity (like TruA or RluD, see Fig. 3) or even multisite-specificity (RluA and RluC).^73,74 This balance between substrate specificity and promiscuity typical for Ψ synthases is evident in, e.g., E. coli TruA, the only dimeric Ψ synthase, which uses the intrinsic flexibility of its substrate tRNA to access either positions 38, 39 or 40.⁷⁶ In contrast the specificity of RIuF and RIuB for adjacent sites in the ribosome, is achieved by substrate binding in different conformations.^77,78 This specificity is compromised by a weak activity of RIuF for the substrate position of RIuB.⁷⁹ TruB recognizes the shape of the T-stem loop and therewith its substrate position in its single substrate tRNA.⁸⁰ Strikingly, the preference for structured 50S subunits over free 23S rRNA of RIuD⁸¹ coincides with few sequence requirements, in contrast to the associated m³Ψ methyltransferase RImH.⁸²

The modification pattern of archaeal RNAs (including pseudouridine residues) was only studied for a limited number of species, among which the halophilic Archaea H. volcanii is the best studied organism (see Fig. 4 for modification positions and responsible enzymes). Direct RNA sequencing of isolated tRNA species^83,84 pointed out a modification profile similar to the one observed in bacteria, but Ψ32 was absent and some additional sites were detected in D-and TΨ-loops. Genomic studies and direct analysis of Ψ synthase activities^85,86 confirmed the absence of RIuA-related activities in Archaea, while instead, an additional family of Pus10-related proteins was found. One of the best studied members of this family, Pus10p from H. volcanii,⁸⁷ fills out the role of the TruB enzymes by acting as Ψ synthase on positions 54 and 55 in archaeal tRNA,⁸⁸ using a different recognition mechanism for each position.⁸⁹ Recognition by Pus10 proteins probably involves the characteristic N-terminal THUMP domain,⁴⁷ that binds to the tRNA acceptor stem in a docking model of the human Pus10 homolog.⁹⁰ This binding mode is supported by a recent cocrystal structure of a THUMP domain-containing enzyme, 4-thiouridine synthetase.⁹¹ Generation of Ψ55 is undisturbed by deletion of Cbf5,⁸⁷ which can also generate Ψ55 in vitro.⁹² Whether Cbf5 can substitute Pus10p in generating Ψ55 in vivo cannot be tested since a Pus10p knockout is lethal.⁸⁷ In contrast to Eubacteria, the pseudouridine formation in archaeal rRNA is insured by H/ACA sRNA RNPs,^86,87 and thus the RsuA related family is also missing. Several rare sites of pseudouridine modification in archaeal tRNAs still have not been assigned to a particular enzyme,^86,87 but highly promiscuous enzymes like TruD (Pus7)⁹³ may be responsible for Ψ formation at these locations.

Figure 4. Distribution of Ψ and Ψ synthases in H. volcanii: TruA purple⁸⁷, TruD⁸⁶ yellow, Pus10p⁸⁷ brown, Cbf5^86,87 blue, positions with yet unknown enzyme⁸⁶ in gray. Note that position 52 is only partially modified⁸³ and that ribosomal Ψs are only available for 16S and not for 23S and 5S rRNA. ⁸⁶

The best studied lower eukaryote, S. cerevisiae, displays 4 common Ψ synthase families (see Fig. 5, RsuA- and Pus10-related families are missing). In Eukarya, not only tRNAs and rRNA are modified to pseudouridine, but also snRNAs. Recent genome-wide pseudouridine profiling even revealed hundreds of Ψs in mRNA and provided further evidence on Ψs in snoRNAs.^24,25 Additional complexity comes from distinct cellular compartments (and their respective specific RNA species) coexisting in eukaryotic cells. Thus, nuclear (cytoplasmic) tRNAs and rRNA are not necessarily modified by the same machinery as their mitochondrial counterparts. This duality clearly exists for Pus1/Pus2^94,95 and Pus8/Pus9⁹⁶ pairs for tRNA modification and for Cbf5/Pus5^15,97 for rRNA pseudouridine formation. However, some enzymes like Pus3⁹⁸, Pus4⁹⁹ and Pus6¹⁰⁰ have dual functions and are partially imported to mitochondrial compartment. As for bacterial Ψ synthases, many yeast enzymes demonstrate both region-specificity and multisite-specificity to account for the large number of modification in all types of cellular RNAs. Formation of some pseudouridine residues in yeast RNAs, notably U2 snRNA and U6 snRNA (dashed circles in Fig. 5, see also below) and mRNA is stress-regulated.^24,66,101 Upon heat shock the localization of Pus7p changes from nuclear to in part cytosolic.²⁵ TruA family member Pus1p, as well as yeast TruD homolog Pus7p,¹⁰² in contrast to their bacterial counterparts, modify various positions in a large variety of substrates, including U2 snRNA^66,103,104, various positions in various tRNAs^95,105,106, 5S rRNA¹⁰⁷ and mRNA.^24,25 The loose specificity of eukaryotic Pus1p may be related to its additional C-terminal domain, which, in contrast to E. coli TruA, causes it to act as monomer.¹⁰⁸ This difference in structure results in substrate specificity for a minimal substrate defined solely by shape and not by sequence.¹⁰⁹ In contrast, Pus7p acts on a specific recognition sequence.¹⁰² This striking difference in substrate recognition could be confirmed in pseudouridine profiling of mRNA.^24,25

Figure 5. Distribution of pseudouridine and pseudouridine synthases in yeast: Cellular location of enzyme and substrates as well as substrate position are given for TruA family members Pus1p^95,103,106, Pus2p⁹⁴ and Pus3p⁹⁸ (purple), RIuA family members Pus5p,⁹⁷ Pus6p,¹⁰⁰ Pus8p⁹⁶ and Pus9p⁹⁶ (green), TruD homolog Pus7p^104,105,107 (yellow) and stand-alone TruB homolog Pus4p⁹⁹ (blue), as well as the RNA-guided TruB homolog Cbf5 (blue) for U2 RNA²⁰¹ und U5 snRNA²⁵. Modification sites without attributed enzymatic activity are indicated in gray. Mitochondrial LSU rRNA contains only one Ψ residue at position 2819 generated by Pus5.⁹⁷ Note that for clarity the at least 44 ribosomal Ψs formed by Cbf5¹⁵ are only suggested and that U2 snRNA positions 56 and 93 and U6 snRNA at position 28 have a dashed outline due to their inducibility.^66,100 Pus7p is shown in the cytoplasm with dashed outline, since the enzymes changes its localization from nuclear to cytoplasmic upon heat shock.²⁵

Occurrence and formation of Ψ in human

The precise pseudouridylation pattern of human RNAs remains only partially uncovered (see Fig. 6). Despite hard efforts in direct RNA sequencing of cytoplasmic and mitochondrial tRNAs, only some species have been analyzed in detail,¹ and some existing pseudouridine sites still escape identification. However, the overall profile of human tRNA modification is similar to the one from S. cerevisiae, even if some minor sites have not (yet) been detected in human. For instance, Ψ32, very common in Eubacteria and in yeast, has been mapped in only one cytoplasmic human tRNA so far, tRNA^His.¹ Known human Ψ synthases belong to 5 families, only the RsuA-related family is missing (like it is also the case for S. cerevisiae). One can also notice duplication of some Ψ synthase genes, as Pus1/Pus1L, Pus7/Pus7L, TruB1/TruB2.¹¹⁰ All stand-alone human Ψ synthases are supposed to modify mostly tRNAs, since the great majority of known sites in rRNA and snRNA are attributed to specific H/ACA-snoRNA-guided machinery. However, the implication of stand-alone enzymes (like highly promiscuous Pus1 or/and Pus7) in modification of these species cannot be formally excluded. Only a few predicted human Ψ synthases have been studied up to date, only the specificity of hPus1 was experimentally confirmed,^109,111 assignment of the other proteins is mostly based on the sequence homology and the properties of the human and archaeal counterparts and thus remains only tentative. Some of human Ψ synthases are predicted to have preferential mitochondrial localization and are thus supposed to modify tRNAs in this compartment. Like in S. cerevisiae, Ψ13 and Ψ35 are missing in mitochondrial tRNA, while other sites are quite well conserved.

Figure 6. Distribution of Ψ and Ψ synthases in Homo sapiens: Cellular location of substrates and substrate position are given for TruA family members Pus1,¹¹¹ Pus1L, Pus3 (UniProt Acc. number Q9BZE2) (purple), TruB family members TruB1¹¹⁰, TruB2¹¹⁰ and Cbf5²⁰² (blue), TruD family members Pus7 and Pus7L (UniProt Acc. number Q9H0K6), RIuA family members PusD1 (UniProt Acc. number Q9UJJ7.1), PusD3 (UniProt Acc. number Q6P087.3) and PusD4 (UniProt Acc. number Q96CM3.1) (green) and Pus10⁹⁰ (brown). In addition to tRNA and rRNA snRNA and snoRNA are modified. Note that, to current knowledge, Ψ-positions in snRNAs U2, U4 and U6, exclusively formed by H/ACA Box RNPs.²⁰³ Positions with known or putative guide RNAs are depicted in blue, while gray positions await guide RNA identification.²⁰³ Not shown are Ψ-containing SRA RNA²⁰⁴ and human telomerase RNA.⁴⁵

Regulation of and via Ψ

Levels of Ψ differ from tissue to tissue¹¹² and may be cell cycle dependent.¹¹³ This implies that Ψ levels are regulated and, in turn, that there is a biological benefit to this regulation. Consistent with this picture, additional Ψs can be induced in yeast U2 snRNA and U6 snRNA in site-specific and stimulus specific manner^66,101 and the mTOR pathway induces a higher Ψ content in 28S rRNA of CHO cell cultures.¹¹⁴ In mouse, Ψ is directly involved in activation of nuclear receptors via pseudouridylation of steroid receptor RNA activator (SRA).¹¹⁵ Such regulatory function in transcription is related to the concept of a regulatory role of Ψ in translation. Interestingly, Ψ can suppress non-sense codons in vitro and in vivo, if it is artificially and site-specifically introduced into mRNA.⁶⁵ This led to a detailed study on possible effects pseudouridine modified nonsense and sense codons.¹¹⁶ Nonsense suppression may be caused via a Ψ-A base pair, which is thought to stabilize the 2 non-canonical base pairs completing the codon-anticodon interaction.¹¹⁷ Indeed 2 recent studies reported various inducible Ψs in yeast mRNA.^24,25 Further investigation identified the enzyme Pus7p to be mainly responsible for heat shock induced pseudouridylation in yeast: A change in localization of the enzyme from mainly nuclear to also cytosolic seems to allow mRNA pseudouridylation that presumably contributes to mRNA stability.²⁵

The mechanism of Ψ formation

Kinetics

Kinetic studies on Ψ synthases depict them as slow in catalysis under multiple turnover conditions (see Table 1) with changes in RNA conformation¹¹⁸, catalysis^119,120 and catalysis or product release¹¹⁹ as rate limiting steps.

Table 1. Overview on K_m and k_cat of Ψ synthases

Enzyme	Organism	Family	KM / nM	kcat / s^−1
RIuD^205	E. coli	RIuD	980 ± 180	∼0.033
TruB^116,120, 123,135	E. coli	TruB	146-780	0.12-0.7
TruA^119,131	E. coli	TruA	940	0.18-0.7
RIuA^138, 206,207	E. coli	RIuA	108-308	0.1
TruD^208	E. coli	TruD	380	0.001
Pus1p^111	H. sapiens	TruA	32	—
Pus1p^118	S. cerevisiae	TruA	420-740	∼0.006
Pus10p^47	P. furiosus	Pus10p	400	0.9

Judging from the apo-enzyme¹²¹, cocrystal structure¹⁶ and kinetic studies^119,120 E. coli TruB, serving as a general role model for Ψ synthases, acts via an induced fit mechanism that consists of at least 4 steps: (i) initial RNA binding (ii) induced fit (iii) catalysis (iv) product release. The process of base-flipping involves a non-essential^12,122 histidine⁵ for TruB family members or an arginine in other Ψ synthases.^12,47,77,78 The most obvious explanation for Ψ formation being slow is that the chemistry of the reaction is rate limiting and may not allow faster catalysis.¹¹⁹ Several Ψ synthases were found to act more efficiently on weakly structured RNAs and avoid modification of stable RNAs.^{76,80,102,123,124} Stabilizing RNA modifications are of cooperative and/or pleiotropic nature^73,125,126 and single modifications were often found to be non-quantitative.^83,127

Mechanistic studies on Ψ synthases using 5-fluorouridine

Inhibition of Ψ synthases by the anti-cancer drug 5-fluorouracil¹²⁸ (5FU) was investigated in several organisms.^129-132 While the original target of 5FU is thymidylate synthase, it may also inhibit formation of ribothymidine if incorporated into RNA, where it is also commonly regarded as inhibitor of Ψ formation. SDS-PAGE stable, but heat disruptable 5FU-RNA-Ψ synthase complexes,^77,131-134 requiring the catalytic aspartate, were reported for several Pus enzymes, leading to the proposal of a Michael addition like mechanism of Ψ formation.^131,133 In this mechanism, the catalytic aspartate would attack the Michael acceptor C6 of the base (see Fig. 7), while the alternative, so called “acylal mechanism” would involve an aspartate attack on the C1’ of the ribose (see non-gray reaction path in Fig. 8a).¹³³

Figure 7. The “Michael” addition-like mechanism of Ψ formation modified from Czudnochowski and coworkers.⁷⁸ The substrate is either 5-fluorouridine (R = F) or uridine (R = H ). To account for the “generally accepted covalent adduct” of the substrate base’ C6 to the catalytic aspartate of the enzyme (if the substrate is 5FU), the aspartate would have to attack in an Michael addition-like manner. The protonation- and deprotonation steps proposed by Czudnochowski et al. would be carried out by yet unidentified bases (¹B, ²B, ³B). Please note that turnover of U and 5FU both result in compound 5. This final intermediate is either deprotonated to eventually result in pseudouridine or hydrated in case of 5FU (gray shaded reaction step) to generate 5S-6R-6-hydroxy-5-fluoro-pseudouridine.

Figure 8 (See previous page). The acylal mechanism and the glycal mechanism for Ψ formation in a version modified from ref. ¹³⁹. (A) In case of 5FU the acylal intermediate can result in compound 5 to eventually yield the 5S-6R-6-hydroxy-5-fluoro-pseudouridine found in the crystal structures. However, an equilibrium of the 5FU-acylal intermediate with an oxocarbonium intermediate (compound 5b) might open an additional gray shaded reaction manifold exclusively to 5FU. This would account for the arabino-isomer as minor product of E. coli TruB action on 5FU RNA that was discovered by Miracco and Mueller.¹³⁹ Pseuoduridine could be formed by the not-shaded acylal mechanism, the only difference would be the last step: The ‘F’ would be a proton that is abstracted to generate the product. (B) Miracco and Mueller proposed that pseudouridine could also be formed by a third glycal mechanism. This mechanism resembles the gray reaction manifold in a) but yield only one product in ribo conformation.

Cocrystal structures of active, e.g. refs.^{16,77,78,121,134-136}, but not of inactive¹³⁷ Ψ synthases with 5FU RNA contain a hydrated and rearranged 5FU, 5S-6R-6-hydroxy-5-fluoro-pseudouridine. Evidence that the hydration is caused by attack of water^132,138,139 does not favor one mechanism over the other, but is strengthened by a fortuitous adduct of RNA with Ψ synthase RIuB, where a conserved, but not catalytically essential¹³⁵ Tyrosine⁷⁸ substitutes water. In one case the SDS PAGE stable adduct proved sensitive to X-ray exposure¹³⁶, implying that the covalent adduct cannot be visualized in crystals because it was destroyed during measurement.

One Ψ synthase, E. coli TruB, failed to form a SDS PAGE stable complex with 5FU-containing RNA and failed to be inhibited in kinetic studies,¹³² which is consistent with turnover of 5FU to the same rearranged, hydrated product by several E. coli enzymes.¹³⁹ In depth NMR analysis of E. coli TruB-5FU-RNA products, revealed a second, minor product in the arabino conformation, specifically resulting from turnover of 5FU-RNA.¹⁴⁰ To account for the lack of arabino product in U turnover compared to 5FU turnover, Miracco and Mueller suggested that U and 5FU might be turned over by different mechanism. Pseudouridine could either be formed by the acylal mechanism, which is shown in the non-gray reaction path of Figure 8a, or by a third ‘glycal mechanism’ shown in Figure 8b. Miracco and Mueller hypothesize that 5FU turnover by the acylal mechanism (Fig. 8a) might open an additional, reaction manifold, shaded gray in Figure 8A, which is unavailable to uridine.¹⁴⁰ They suggested that step ‘iv b’ and the following reaction path leading to the arabino product are restricted to 5FU due to lower reactivity: The electron-withdrawing fluorine substituent might stabilize the free anion of the fluorinated pyrimidine, thereby decreasing its nucleophilicity.¹⁴⁰ Please note that the glycal intermediate in Figure 8a can, in contrast to its counterpart in Figure 8b, be converted to either the ribo product (“H to top face”) or to the arabino product (“H to bottom face”), again due to the assumed long lifetime of the intermediate. E. coli RIuA might also form an arabino product, as 2 products detectable in preliminary NMR data imply.¹³⁹ Undoubtedly, this analysis is the most sophisticated and most reliable analysis of 5FU-turnover by a Ψ synthase reported until now. In this respect it is particularly surprising that a minor arabino product was not reported in any of the available cocrystal structures of 5FU-RNA and Ψ synthases. We checked the B-factors of the respective O2’ in cocrystal structures of 3 different enzyme families for irregularities: Indeed we found them to be mostly unremarkable.^{16,76,77,120,133-135} This indicates a confidence of the ribo conformation compared to an arabino conformation that is similar to the accuracy of the whole structure. Seemingly the arabino product is either not contained in the crystals or not detectable for yet unknown reasons.

Of note, related modification enzymes use both, the attack on C1’^68,141 and the C6 Michael addition mechanism, respectively.^142-147 The most instructive hint in this case might be, that related transglycosylases actually proceed by a C1’ attack as reviewed in ref. ¹⁴⁸.

Functions of Ψ residues in RNAs

Structural effects - tRNA

The most conserved Ψ modifications stabilize the tertiary structure of tRNA, be it at position 32,¹⁴⁹ 39^150,151 or 55.¹⁵² Conformational effects caused by Ψ39 influence anticodon recognition^153,154, missreading and frame shifting in yeast (together with Ψ38)¹⁵⁵, and interaction with HIV RNA.¹⁵⁶

Several eukaryotic cytoplasmic tRNAs carry Ψ at the anticodon positions 34, 35 and 36, where the modification is introduced in intron-dependent manner, as reviewed in ref. ¹⁵⁷. Ψ35, the only modification tolerated at that position¹⁵⁸, is especially conserved in tRNA^Tyr of a large variety of eukaryotes¹, including, e.g., the amobea Tetrahymena thermophila¹⁵⁹ and Xenopus.¹⁶⁰ Presumably, Ψ35 confers superior stabilization to the anticodon by replacing a (U33)O2’-HC5(U35) hydrogen bond by the stronger (U33)O2’-H-N1(Ψ35).¹⁵⁸ Until now there is no mechanistic basis for other anticodon Ψs, namely at positions 34 and 36, that can occur single¹ or as pair.¹⁶¹

The function of Ψ in mitochondrial tRNAs is less characterized. In case of human Ψ occurs at positions 27, 28, 41, 42, 49, 40, 50 and 67, and occasionally at 55.¹⁶² A well understood, but special case demonstrating a possible role of nucleoside modifications is human mitochondrial tRNA^Lys. The conformational equilibrium of this tRNA is influenced by nucleoside modifications, including 2 Ψs. These Ψs, located at positions 27 and 28, have, in contrast to the usual role of Ψ, a slight destabilizing effect on the canonical cloverleaf structure.¹⁶³

Role of Ψ in the helix 69 of the ribosome

The role of Ψ in ribosomes was reviewed recently,¹⁵ a deeply investigated motif conserved over all domains is helix 69 (H69). The three Ψs in the isolated H69 of E. coli (indicated by the dashed box in Fig. 3) show complicated pleiotropic effects,¹⁶⁴ potentially involving increased base stacking and N1H hydrogen bonding^165,166 and influence of a m³Ψ modification¹⁶⁷ and pH.¹⁶⁸ These effects are equally present in human H69^169,170 and in whole ribosomes¹⁷¹, and influence ribosomal subunit association.¹⁷² These conformational effects still await full clarification.

Ψ in spliceosomal branch-site architecture

Ψ in small nuclear RNAs was thoroughly reviewed recently, e.g. in refs. ^15,41 A prominent example is a Ψ residue in eukaryotic U2 RNA that stabilizes and fine-tunes spliceosomal branch-site interaction^39,173, involving a water-ΨNH1 hydrogen bond.^174,175

Functional importance for RNA

Ψ in artificial mRNAs

Synthetic replacement of all uridines by Ψ renders mRNAs non-immunogenic¹⁷⁶, increases biological stability^176-178 and enhances translation in vivo^176,179,180, while reducing PKR activation.¹⁸¹ In contrast, studies with in vitro assays suggested that mRNAs where all Us were changed to Ψs inhibit translation at the initiation and elongation levels.¹⁸²

Ψ in eukaryotic mRNAs

Recently at least 260 Ψs in 238 mRNAs of Saccharomyces cerevisiae could be identified with most frequent occurrences in the GUA valine codon and an initial screen of highly expressed genes identified 96 Ψs in 89 human mRNAs.²⁴ A second study could link 41 Ψs in 41 mRNAs to specific Ψ synthases in yeast and 136 mRNA sites in human to specific Ψ synthases.²⁵ Although the majority of modifications could be induced by starvation²⁴ or heat shock²⁵, their actual functional relevance remains to be proven. In case of yeast most pseudouridines are introduced not by H/ACA box RNPs but by 4 out of 9 stand-alone protein Ψ synthases: Mainly by Pus1p and Pus7p, but also by Pus2p and Pus4p.^24,25 Occurrences of Ψ in mRNA are widely distributed over coding, as well as non-coding 5’ and 3’ sequences. It is therefore possible that a portion of modification sites mimic Pus substrates rather by coincidence than due to an actual advantage gained from pseudouridylation.²⁴

RluD/ribosomal assembly

Knockout of RIuD, the enzyme generating the 3 Ψs H69 of the E. coli ribosome (see dashed box in Fig. 3), interferes with ribosome assembly,¹⁸³ implying requirement for normal growth in E. coli K12,¹⁸⁴ in contrast to wild type E. coli.¹⁸⁵ A mutated release factor 2 rescues ΔRIuD E. coli K12,¹⁸⁶ which is consistent with increased affinity of native release factor 2 to pseudouridylated H69.¹⁸⁷ In yeast the loss of Ψs in Helix69 impairs growth and influences ribosome synthesis¹⁸⁸ and function synergistically,¹⁸⁹ but also with pleiotropic effects.^188,190

Implications in human pathologies

Pseudouridine related enzymes have been implicated in various human diseases, e. g. in Crohn's disease and Celiac disease¹⁹¹ and X-linked ichthyosis.⁷¹ The involvement of NEP1, a N1-Ψ specific methyltransferase, in the Bowen-Conradi syndrome¹⁹² and dyskerin in X-linked dyskeratosis congenita¹⁹³ may not be directly related to Ψ, but rather caused by involvement of the proteins in ribosomal assembly¹⁹⁴ and telomere maintenance, respectively.^195,196 A recent study detected a slightly lower pseudouridylation level in dyskeratosis congentia patients compared to healthy individuals and verified Ψs in the telomerase RNA component that may be involved in the disease.²⁵

A mutation in the human PUS1 gene leads to hypomodification in mitochondrial tRNAs by preventing hPus1p activity, resulting in mitochondrial myopathy and sideroblastic anemia.¹⁹⁷ Due to the wide substrate specificity of Pus1p discussed above, hypomodification of RNAs other than tRNA might contribute to the disease. ¹⁶²

Conclusions and outlook

To date enzymes and substrate positions for Ψ formation are quite well understood in the major model organisms E. coli, H. volcanii and S. cerevisiae. In contrast the chemical mechanism of Ψ formation is as elusive as ever. Possible are either an acylal mechanism¹³⁹, where the catalytic Asp acts as general base as inferred from the pH dependency of the TruB reaction¹²² or a Michael addition mechanism that would not account for a (still not directly characterized) covalent adduct of the enzyme to C6 of the target base in RNA.⁷⁸ These mechanistic studies suffer from ambiguous mutagenesis approaches, which were unable to identify the major basic and acidic residues required for either mechanisms (abbreviated as ‘B’ in Fig. 7 and 8).

The next task on hand is undoubtedly the functional characterization of Ψ in mRNA and elucidating the modifications regulatory properties. Such properties should intensify the interest in human Ψ synthases, of which only hPus1p is characterized^109,111 and all others lack evidence on protein level (Fig. 6).

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.