Correction

BIOPHYSICS. For the article ‘‘H/ACA small nucleolar RNApseudouridylation pockets bind substrate RNA to form three-way junctions that position the target U for modification,’’ byHaihong Wu and Juli Feigon, which appeared in issue 16, April17, 2007, of Proc Natl Acad Sci USA (104:6655–6660; firstpublished April 5, 2007; 10.1073�pnas.0701534104), the authorsnote that in Fig. 1A, the sequence of the 5� side of the 5�pseudouridylation pocket of the U65 H/ACA snoRNA, as wellas a portion of the lower stem of the 3� hairpin, were drawnincorrectly. Additionally, in the Fig. 1 legend and elsewhere inthe text, the numbering of the substrate RNA, which was basedon Bortolin et al. [Bortolin M-L, Ganot P, Kiss T (1999) EMBOJ 18:457–469], differs by one nucleotide from that found in thegenomic sequence (refer to the snoRNA database, snoRNA-LMBE-db, at www-snorna.biotoul.fr). The correct numbers forthe pseudouridylation sites are 4427 and 4373 for the 5� and 3�pseudouridylation pockets, respectively. The corrected figureand legend appear below. These errors do not affect the con-clusions of the article.

Fig. 1. Human U65 H/ACA RNA and substrate. (A) Sequence and secondarystructure of U65 H/ACA snoRNA (9). (B) Sequences and secondary structures ofthe model U65 � pocket (U65hp), the substrate rRNA (S14), and the U65hp/S14complex. S14 contains the sequence of human 28S rRNA (4422–4434), includ-ing U4427 (underlined), which is targeted for pseudouridylation, and an extraU (lowercase) at the 5� end.

www.pnas.org�cgi�doi�10.1073�pnas.0709912104

19160 � PNAS � November 27, 2007 � vol. 104 � no. 48 www.pnas.org

Dow

nloa

ded

by g

uest

on

Dec

embe

r 9,

202

0 D

ownl

oade

d by

gue

st o

n D

ecem

ber

9, 2

020

Dow

nloa

ded

by g

uest

on

Dec

embe

r 9,

202

0 D

ownl

oade

d by

gue

st o

n D

ecem

ber

9, 2

020

Dow

nloa

ded

by g

uest

on

Dec

embe

r 9,

202

0 D

ownl

oade

d by

gue

st o

n D

ecem

ber

9, 2

020

Dow

nloa

ded

by g

uest

on

Dec

embe

r 9,

202

0 D

ownl

oade

d by

gue

st o

n D

ecem

ber

9, 2

020

H/ACA small nucleolar RNA pseudouridylation pocketsbind substrate RNA to form three-way junctions thatposition the target U for modificationHaihong Wu and Juli Feigon*

Department of Chemistry and Biochemistry, and Molecular Biology Institute, University of California, Los Angeles, CA 90095-1569

Communicated by Richard E. Dickerson, University of California, Los Angeles, CA, February 20, 2007 (received for review December 6, 2006)

During the biogenesis of eukaryotic ribosomal RNA (rRNA) andspliceosomal small nuclear RNA (snRNA), uridines at specific sitesare converted to pseudouridines by H/ACA ribonucleoprotein par-ticles (RNPs). Each H/ACA RNP contains a substrate-specific H/ACARNA and four common proteins, the pseudouridine synthase Cbf5,Nop10, Gar1, and Nhp2. The H/ACA RNA contains at least onepseudouridylation (�) pocket, which is complementary to thesequences flanking the target uridine. In this article, we showstructural evidence that the � pocket can form the predicted basepairs with substrate RNA in the absence of protein components.We report the solution structure of the complex between an RNAhairpin derived from the 3� � pocket of human U65 H/ACA smallnucleolar RNA (snoRNA) and the substrate rRNA. The snoRNA–rRNA substrate complex has a unique structure with two offsetparallel pairs of stacked helices and two unusual intermolecularthree-way junctions, which together organize the substrate fordocking into the active site of Cbf5. The substrate RNA interacts onone face of the snoRNA in the complex, forming a structure thateasily could be accommodated in the H/ACA RNP, and explains howsuccessive substrate RNAs could be loaded onto and unloaded fromthe H/ACA RNA in the RNP.

NMR � structure � ribonucleoprotein particle � pseudouridine

Pseudouridine (�) is the most frequently occurring posttran-scriptionally modified ribonucleotide and is found in almost all

tRNAs, rRNAs, and small nuclear RNAs (snRNAs) (1, 2). Inhuman rRNA, there are �95 �s, clustered in functionally impor-tant regions of 5.8S, 18S, and 28S rRNAs (3–5), such as the peptidyltransferase region of the 23S rRNA (6, 7). Conversion of uridine(U) to � results in transfer of the glycosidic bond from the N1position of U to the C5 position and is catalyzed by a variety ofspecific pseudouridine synthases (8). Although conversion of U to� in tRNA and rRNA in prokaryotes generally is catalyzed bysingle-polypeptide enzymes, in eukaryotes and archaea the modi-fication of rRNA and most spliceosomal RNAs requires H/ACAsmall nucleolar (sno) ribonucleoprotein particles (RNPs) and smallCajal body-specific (sca) RNPs, respectively (9–12).

The H/ACA snoRNPs contain a common set of four coreproteins and a snoRNA that usually is specific to one or twosubstrate RNAs. The canonical H/ACA snoRNA has a bipartitestructure containing conserved secondary structural elements in a‘‘hairpin–hinge–hairpin–tail’’ arrangement (9, 10). The hinge con-necting the two hairpins includes the consensus (H box) sequence5�-ANANNA-3� (where N is any nucleotide), whereas the 5�-ACA-3� sequence in the tail always is located 3 nt from the 3� endof the mature H/ACA snoRNAs. The two hairpins contain a largeinternal loop, called the pseudouridylation (�) pocket (Fig. 1). This� pocket is complementary in sequence to �3–10 nt on either sideof a UN in the substrate RNA, where U is the nucleotide to bemodified and N is any nucleotide. The nucleotides on the 5� and 3�sides of the � pocket therefore are proposed to be ‘‘guides,’’ whichform base pairs with substrate RNA, leaving the UN unpaired at thetop of the pocket. Either one or both of the � pockets can be guidesfor pseudouridylation of specific Us. In budding yeast, the forma-

tion of all 44 �s in rRNA is guided by 28 H/ACA snoRNAs (13).In archaea, the H/ACA RNAs can have from one to three hairpins.The H/ACA RNAs also are found in the more recently identifiedscaRNPs (12), which are targeted to Cajal bodies, via a conservedCajal body localization signal found in the hairpin terminal loop(14), and are involved primarily in pseudouridylation of spliceoso-mal snRNAs. There also are a few H/ACA RNPs that do not appearto be involved in pseudouridylation, including, e.g., vertebratetelomerase, which has a scaRNA domain at its 3� end (15), and U17(snR30 in yeast), which is required for ribosomal RNA process-ing (16).

The H/ACA RNP proteins in vertebrates and yeast are Cbf5 (17)(Dyskerin in humans) and the basic Nop10, Nhp2 (18), and Gar1(19–21). Cbf5 is highly homologous to the pseudouridine synthaseTruB (8), which is responsible for modification of the conserved�55 found in the T�C stem–loop (TSL) in almost all tRNAs.Biochemical studies have indicated that the pseudouridine synthaseCbf5 can assemble with Nop10 and either Gar1 or Nhp2 to formtwo different ternary complexes (22–26). In vitro, Cbf5 and Nop10are the minimal proteins required with the H/ACA RNA toassemble a pseudouridylation-competent RNP, but this activity isfacilitated by the addition of Gar1 (22).

Cocrystal structures of archaeal Cbf5–Nop10 (27, 28) and Cbf5–Nop10–Gar1 (29) complexes, and most recently an archaealH/ACA RNP (30), have provided insight into the assembly, role ofthe proteins, and structure of the H/ACA RNP. In the proteincomplexes as well as the RNP, Nop10, which largely is unstructuredin solution (27, 31), as well as Gar1 interact extensively with Cbf5but not with each other. The archaeal H/ACA RNP for which thestructure was determined contains a single-hairpin H/ACA RNA incomplex with Pyrococcus furiosus Cbf5, Nop10, Gar1, and L7Ae,which in archaea replaces Nhp2 (30). Nop10 is sandwiched betweenCbf5 and L7Ae, and the three proteins provide a linear platform forbinding the H/ACA RNA, while Gar1 binds to one side andinteracts only with Cbf5. Both Nop10 and Gar1 appear to play a rolein forming or stabilizing the active site of Cbf5 (27–30). In theH/ACA RNP, there are extensive interactions between the bottomof the lower (P1) stem of the H/ACA RNA and the PUA domainof Cbf5 and between L7Ae and the top of the upper (P2) stem ofthe H/ACA RNA, but only a few, from Nop10 and Cbf5, to thebottom of P2 and the � pocket. The � pocket itself, where thesubstrate is proposed to bind, largely is dynamic and not well

Author contributions: H.W. and J.F. designed research; H.W. performed research; H.W. andJ.F. analyzed data; and H.W. and J.F. wrote the paper.

The authors declare no conflict of interest.

Abbreviations: sno, small nucleolar; RNP, ribonucleoprotein particle; sca, small Cajal body-specific; TSL, T�C stem–loop; HSQC, heteronuclear single quantum correlation.

Data deposition: Coordinates and restraints for the 20 lowest-energy structures have beendeposited in the Protein Data Bank, www.pdb.org (PDB ID code 2P89).

*To whom correspondence should be addressed. E-mail: feigon@mbi.ucla.edu.

This article contains supporting information online at www.pnas.org/cgi/content/full/0701534104/DC1.

© 2007 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0701534104 PNAS � April 17, 2007 � vol. 104 � no. 16 � 6655–6660

BIO

PHYS

ICS

defined in the crystal structure. Although substrate binding hasbeen modeled (30), an outstanding question is how the substrateRNA interacts with the RNA in the snoRNP.

U65 is a human H/ACA snoRNA that guides the modification ofhuman 28S rRNA at positions U4374 and U4428 (10, 32, 33) (Fig.1). We previously determined the solution structure of the 3� �pocket of U65 snoRNA (31). In the absence of rRNA substrate andproteins, the � pocket is partially (mismatch) base-paired to forman irregular helix flanked by A-form P1 and P2 stems. Here, wereport the solution structure of the human U65 H/ACA snoRNA3� � pocket in complex with an oligonucleotide corresponding to its28S rRNA substrate RNA sequence. The complex forms in theabsence of proteins, and the substrate RNA binds on only one faceof the pocket. The snoRNA/rRNA substrate complex has a uniquestructure with two offset parallel pairs of stacked helices and twounusual three-way junctions. Analyzed together with the structuresof the archaeal H/ACA RNP (30) and TruB RNA (34, 35)complexes, the structure of the complex reveals that the H/ACARNA functions to not only recruit its specific RNA substratethrough complementary base pairing but also to organize thesubstrate for docking into the active site.

ResultsThe 3�-Pseudouridylation Pocket of U65 H/ACA snoRNA Binds the rRNATarget in the Absence of Proteins. To study the interactions betweenH/ACA snoRNA and rRNA substrate, we used an RNA hairpin(U65hp) derived from the 3� � pocket of human U65 H/ACAsnoRNA and a 14-nt oligoribonucleotide (S14) derived from therRNA substrate sequence as our model system (Fig. 1). U65hpcomprises the 3� � pocket and flanking stems P1 and P2. Additionalnon-wild-type sequence base pairs were added to the end of P1 forin vitro transcription efficiency by T7 RNA polymerase, and aUUCG tetraloop was added to P2 for stability. S14 corresponds tothe sequence flanking U4428 of human 28S rRNA that is predictedto interact with the 3� � pocket (Fig. 1B). An additional U wasadded to the 5� end of S14 in place of G4422 in wild-type rRNA.

To determine whether the rRNA substrate can bind U65hp in theabsence of proteins, we monitored the titration of rRNA substrate

(S14) with U65hp by using NMR spectroscopy (Fig. 2). The freeS14 is in a single-stranded conformation with no observable iminoresonances. Imino proton spectra of free U65hp show resonancescorresponding to the upper stem and UUCG tetraloop, the lowerstem, and a G9–C26 and mismatch A11–G24 bp in the � pocket(Fig. 2) (31). Free U65hp forms an irregular helix with A-formlower and upper stems and partial pairing and stacking in theinternal loop. Thus, in the absence of substrate, this � pocket(internal loop) is in a ‘‘closed’’ conformation. Upon addition of S14to U65hp, new imino resonances emerged, whereas imino reso-nances corresponding to nucleotides in the � pocket decreased inintensity, indicating formation of the complex in slow exchange withthe free U65hp. Substrate binding can be monitored easily by theincreasing intensity of a low-field shifted resonance correspondingto U5 in the complex [Fig. 2 and supporting information (SI) Fig.6]. Imino resonances of nucleotides in P1 and P2 showed littlechemical shift change, except for U5 and U23, which are located atthe top and bottom of P1 and P2, respectively (Fig. 1). The newpeaks were assigned to imino protons of nucleotides in two newlyformed helices (Figs. 1 and 2) by using 2D NOESY and 15N-heteronuclear single quantum correlation (HSQC) experiments.These results indicate that the substrate RNA can bind the � pocketin the absence of protein cofactors. Furthermore, the substratemust displace the mismatch base pairs in the � pocket to bind.

Although the complex is in slow exchange with free RNA on theNMR time scale, binding is weak (KD, �200–300 �M), and complexformation was observed by NMR only at millimolar concentrationsof RNA. Addition of Mg2� and/or a higher NaCl concentrationslightly stabilized the complex but did not result in significantspectral changes (data not shown).

Overview of the Solution Structure of U65hp/S14 Complex. TheU65hp/S14 complex is in equilibrium with free components even atstoichiometric ratios. Therefore, most of the assignments anddistance restraints were obtained from NMR spectra of samplesprepared with one 13C,15N base-selectively labeled RNA and 1.5-fold excess of the other RNA unlabeled. Resonance assignmentsand structure calculations are described in Materials and Methods.A total of 768 nonredundant NOE restraints (for an average of 16per nucleotide) and 288 experimental dihedral angle restraints wereused in the structure calculation. The ensemble of the 20 lowest-

Fig. 1. Human U65 H/ACA RNA and substrate. (A) Sequence and secondarystructure of U65 H/ACA snoRNA. (B) Sequences and secondary structures of themodel U65 � pocket (U65hp), the substrate rRNA (S14), and the U65hp/S14complex. S14 contains the sequence of human 28S rRNA (4423–4435), includ-ing U4428 (underlined), which is targeted for pseudouridylation and an extraU (lowercase) at the 5� end.

Fig. 2. Binding of substrate RNA to the U65 � pocket. Shown are the600-MHz NMR imino proton spectra of free U65hp (0.8 mM) (A), U65hp:S14(1:1.5; 0.8 mM:1.2 mM) (B), and free S14 (0.8 mM) (C) at 15°C. Imino protonresonances in the free RNA and complex are labeled. Note that U17 imino inthe UUCG tetraloop, which is not hydrogen-bonded and is known to besensitive to sample conditions, also showed a large chemical shift change thatis not related to complex formation.

6656 � www.pnas.org�cgi�doi�10.1073�pnas.0701534104 Wu and Feigon

energy structures is well determined, with an rms deviation to themean of 1.1 Å for all heavy atoms (Fig. 3A and SI Table 1).

The solution structure of the complex shows that the 3� � pocketof U65 snoRNA formed two intermolecular short helices with thesubstrate RNA, S14, via formation of 12 new Watson–Crick basepairs. U6 and A29, which form a Watson–Crick base pair to closethe P1 helix in the free U65hp structure, are opened up to form twonew base pairs with A48 and U35 of the substrate, respectively. Toform helices with both the 3� and 5� guide sequences of the � pocket,the substrate RNA is in a U shape and makes a sharp 180° turn atthe two unpaired nucleotides, the target U, and the 3�C. Interest-ingly, S14 binds to the U65hp completely on one face of the �pocket.

The two newly formed helices, P1S and P2S, coaxially stack onP1 and P2 of U65hp, respectively, to form two pseudocontinuoushelices (Fig. 3 B, D, and E and SI Fig. 7). Consistent with thisfinding, ‘‘sequential’’ NOE patterns in the A-form RNA heliceswere observed between U5 at the top of P1 and U35 at the bottomof P1S and between U23 at the bottom of P2 and U43 at the top ofP2S, i.e., NOE cross-peaks were observed between H1� and H2� ofU5 to H5 and H6 of U35 (Fig. 4 A and B) and between H1�, H2�,and H3� to U23 to H5 and H6 of U43 (Fig. 4 A and C). The helicesare unwound slightly at both the P2/P2S (twist �43°) and P1/P1S(twist �38°) helical junctions, consistent with weaker sequentialH6–H2� NOE cross-peaks (Fig. 4A).

The sugar-phosphate backbone of the U65hp has two kinks,between U5 and U6 and between U23 and G24 (Fig. 3C). Thesekinks result from an almost 90° rotation of U6 and G24 so that theycan form base pairs with substrate nucleotides A48 and C40,respectively, at the bottom of P2S and top of P1S. The P1S and P2Sand helices are aligned parallel to each other, but there are nohelical packing interactions between them; rather, there is a narrow

gap (Fig. 3D). The closest phosphate-to-phosphate distance be-tween the adjacent helices is �7 Å.

Two Three-Way Junctions Are Formed upon Substrate Binding. Bind-ing of substrate RNA to the � pocket results in the formation ofunique intermolecular three-way and pseudo-three-way junctions(3H) at the upper (J2) and lower (J1) stems of the � pocket,respectively. The 3H at the top of the � pocket consists of threehelices P2, P2S, and P1S, and a 2-nt loop (U41C42) connecting P1Sand P2S [3HS2, according to International Union of Pure andApplied Chemistry (IUPAC) nomenclature]. The P1S helix isparallel to P2S, which is stacked coaxially with P2. At the lower partof the pocket, P1S, P2S, and P1 form the other intermolecular 3H(J1), where P1 and P1S are stacked coaxially on each other, and P2Sis parallel to P1S. We call J1 a pseudo-3H because there is nocovalent linkage between P1S and P2S at the junction.

Like all 3H, both J2 and J1 in the U65 snoRNA/rRNA complexhave two helices that are stacked on each other to form a contin-uous helix, and there are no (or few, in the general case) nucleotideslinking the two stacked helices. RNA 3Hs have been classified byLescout and Westhof into three families, A, B, and C, based on thenumber of nucleotides in the loops connecting the three helices(36). The J2 (as well as J1) 3H is most similar to family C in thearrangement of helices. However, the junction is unusual amongRNA 3H structures reported to date in the short lengths of the loopsconnecting P2 and P2S (0 nt) and P1S and P2S (2 nt), the absenceof interactions of the longer loop with the minor groove of helix P2,and the absence of interhelical interactions (36). In family C, thelonger loop is 3–9 nt, is structured like a hairpin, and interactsextensively in the minor groove of the adjacent helix. Of the twoloop nucleotides, C42 is stacked partially on the top G24–C40 basepair of the P1S helix, whereas the base of the target U41 (U4428 in

Fig. 3. Solution structure of U65hp/S14 complex. (A) Superposition of the 20 lowest-energy structures. U65hp is green except for the UUCG tetraloop (gray),and S14 is gold except for the unpaired C (light blue) and substrate U (red). (B) Stereoview of the lowest-energy structure. (C) Ribbon and stick depiction of thestructure showing kinks in the two junctions. (D) Surface view of the lowest-energy structure. (E) Schematic showing stacking of helices and location of junctions.In B, D, and E, the helices in the complex are labeled and color-coded as follows: P1 (green), P1S (blue), P2 (magenta), and P2S (gold). U41 is colored red, C42is cyan, and the UUCG tetraloop is gray.

Wu and Feigon PNAS � April 17, 2007 � vol. 104 � no. 16 � 6657

BIO

PHYS

ICS

human 28S rRNA) is exposed to solvent (Fig. 4). The backboneturns between U41 and C42. The unique arrangement of the J2 3Hlargely is because of the geometric restraints imposed by theexistence of the P1 helix of the H/ACA RNA. Family C 3Hs oftenhave apical loops that cap the two parallel helices, which interactwith each other or with the neighboring helix. Here, the P1 helixfunctions in a similar manner to restrain the relative positions of theP2S and P1S helices. Likely because of the physical constraintsimposed by formation of a 3H within an internal loop, J2 (as wellas J1) also is unusual in that it folds in the absence of divalent cationsor protein (37, 38).

DiscussionThe Pseudouridylation Pockets of the H/ACA snoRNAs Are the LoadingDocks for rRNA Target Sites. Our structural results show that theH/ACA snoRNA � pocket can form the predicted base pairs withthe target rRNA sequence flanking the U to be modified. Thisrecognition does not require the presence of any protein compo-nents of the H/ACA snoRNP, although the weak binding suggeststhat H/ACA RNP protein interactions may be important forstabilizing the complex. Significantly, the structure shows that thesubstrate RNA interacts with the snoRNA on one face of thecomplex only (Figs. 3 and 5D). This geometry of binding means thata helicase would not necessarily be required for substrate RNAloading. Because the substrate RNA binds on only one face of thecomplex, and does not thread through, there would be no topo-logical problem with binding substrates to both the 5� and the 3� �pockets simultaneously.

In the crystal structure of the archaeal H/ACA RNP (30), theH/ACA RNA is tied down at each end by extensive interactionsbetween the PUA domain of Cbf5 and the ACA tail and bottom 4bp and between the archaeal H/ACA RNP-specific L7Ae and a

kink–turn (39, 40) in the RNA at the top of the helix and 3� side ofthe terminal loop (Fig. 5D). There are only a few protein interac-tions with the rest of the H/ACA RNA. The � pocket is largelydisordered, with 2 nt being too mobile to be modeled and 4 ntstabilized by crystal-packing forces. Thus, the � pockets in theH/ACA RNP likely do not differ significantly from the free RNA(Fig. 5D) and are not prearranged for binding of the substrate RNA.The H/ACA RNA is oriented on the protein with the 5� side awayand the 3� side proximal to Cbf5, such that the face that thesubstrate RNA binds to in our complex is exposed and ready forloading (and unloading) of the substrate RNA. Together, thestructures of the U65hp/S14 complex and the archaeal H/ACARNP illustrate how this could occur. The H/ACA RNA is believedto remain stably associated with the snoRNP for multiple rounds ofpseudouridylation (24), although a very recent study indicates thatonly Cbf5 is irreversibly associated with the H/ACA RNA (41).

Coaxial Stacking of P1S on P1 Provides a Constant Distance from theACA Tail to the Substrate Binding Pocket. The number of nucleotidesfrom the top of the � pocket to the ACA (or H box) always is 14–16nt, but the loop size on the 3� side varies from 3 to 10 nt. However,because P1 and P1S are stacked coaxially on each other to form a‘‘continuous’’ helix, the distance between the ACA tail (or H box)and the substrate U will be approximately constant for all H/ACARNA–substrate RNA complexes. Thus, binding of the substrateRNA to the H/ACA RNA defines the distance from the ACA tailto the target U to be �14–16 bp for all H/ACA RNPs, as previouslyproposed based on modeling the substrate RNA on the crystalstructure of an archaeal H/ACA RNP (30).

The Three-Way Junctions Position the Target U in the Active Site. Cbf5is highly homologous to TruB, the pseudouridine synthase thatmodifies the conserved � found in the TSL in almost all tRNAs.Because there is no bound substrate in the archaeal H/ACA RNP,we first compared the U65hp/S14 structure with the TSL incomplex with TruB (34). Superposition of P1S and the UC loop (J2)of U65hp/S14 on the upper stem–loop of TSL showed that bindingof the substrate rRNA to the � pocket results in an RNA backboneconformation that mimics that of the TSL loop in the active site(Fig. 5A). The target U sticks out of the loop in the solutionstructure, but its position is not well defined. In the modeledcomplex based on the superposition, both the C and U nucleotidesjust need to be rotated to move into the active site positionsequivalent to those in the TruB complex (Fig. 5 A and B). A similarrotation, facilitated by a conserved histidine, is observed in thestructures of free versus bound TSL (34, 35) (Fig. 5B). All H/ACARNA–substrate RNA complexes are predicted to have two un-paired nucleotides in the substrate, including the target U, at the topof the pocket. This 2-nt loop, which is unique among RNA class C3H structures reported to date, results in an absence of loop–helixinteractions (although they might occur for some sequences) andtherefore assures that the only sequence requirements are forcomplementarity between the 5� and 3� guides of the � pocket and3–10 nt on either side of the UN loop. Thus, all H/ACA RNA–substrate RNA complexes will form the same structure at J2, wherethe target U is located.

In the TruB–TSL complexes, the RNA loop and upper stem arebound in a deep cleft between the body of the protein and aprotruding ‘‘thumb’’ (34). In our model of U65hp/S14 docked ontoTruB, the thumb would sterically clash with major groove of P2S(Fig. 5A and SI Fig. 8). This thumb, which binds in the major grooveof the TSL loop and helps clamp the RNA onto the protein, largelyis deleted in Cbf5 (and Dyskerin). To model U65hp/S14 on thearchaeal H/ACA RNP, we next replaced TruB–TSL with thearchaeal H/ACA RNP by superimposing homologous regions oftheir respective pseudouridine synthases. The shorter thumb loop inCbf5 largely is disordered in the H/ACA RNP (30) (Fig. 5C), as isthe case for apoTruB (35, 42) as well as Cbf5–Nop10 (27–29). It has

Fig. 4. Coaxial stacking of the P1/P1S and P2/P2S helices. (A) Regions from an13C F1-filtered, F2-edited 2D NOESY spectrum of a 13C,15N U-labeled U65hp/unlabeled S14 complex showing some critical NOEs near the J1 and J2 junc-tions that define the stacking interactions. (B) Close-up view of J1 showingobserved NOEs between U65hp U5 and S14 U35. (C) Close-up view of J2showing observed NOEs between U65hp U23 and S14 U43. Nucleotides arecolor-coded by location in helices as in Fig. 3B.

6658 � www.pnas.org�cgi�doi�10.1073�pnas.0701534104 Wu and Feigon

been proposed that this loop would become ordered when substratebinds, interacting with the major groove of PS1 (30), as illustratedin Fig. 5C. The interaction between the shorter thumb of Cbf5 andthe H/ACA RNA–substrate RNA complex would be expected to beweaker than the interaction of TruB with the TSL. We propose thatthe loss of the thumb clamp is compensated for by formation of P2S,which would help fix the RNA complex in place on Cbf5–Nop10.Thus, H/ACA RNP requires the formation of both the intermo-lecular RNA helices 5� and 3� of the target U for the recognitionand positioning of the target U for modification (Fig. 5D).

In the archaeal H/ACA RNP, the binding of the H/ACA RNAcreates a 50° bend toward the protein in the helical axis of P1 andP2 (Fig. 5D). If P1 interacts with the H/ACA RNP proteins in thesame way once the substrate is bound, there would need to be somerearrangement of the P1 helix relative to its position in the freecomplex for the ACA tail to interact with the PUA domain (Fig. 5).This interaction could be accommodated easily by some unwindingor bending at J1. Specific structures of three- and four-way junctionsoften are required for recognition by their target proteins (38) eventhough they rearrange on binding (37). Alternatively, the C-terminal PUA domain could move. A large domain movement ofthe TruB PUA domain, which is connected to the rest of the proteinby a flexible hinge, is observed in the apoTruB versus RNA-boundTruB (35). It also is worth noting that deletion of the ACA tail doesnot abolish catalysis of an archaeal snoRNP (22). At the other endof the � pocket, the P2 helix would need to be reoriented onlyslightly to form the same interactions with Nop10 observed in thesnoRNP. Pseudouridylation-competent snoRNPs can be reconsti-tuted with aCbf5, aNop10, and H/ACA snoRNA alone (22). Thus,the additional interactions between L7Ae and the kink–turn (39,40) in archaeal H/ACA RNA are not required to tie down the RNA.Eukaryotic H/ACA snoRNAs do not have a conserved kink–turn

sequence in the P2 stem–loop, and binding to Nhp2 is likely to differsomewhat (22).

In conclusion, the structure of the U65hp/S14 complex revealshow the rRNA substrate could bind directly to the H/ACA snoRNAon the H/ACA snoRNP, adopting a structure that has a conserveddistance from the ACA tail bound to the PUA domain of Cbf5 tothe target U and an RNA backbone conformation at the target Uthat can be easily modeled into the active site of Cbf5. Thus,formation of both of the unusual 3H is required for properinteraction of the substrate RNA with the H/ACA snoRNP. Basedon the structure of the complex and comparison to TruB–TSLinteractions and the archaeal H/ACA RNP, we propose thatsubstrate RNA binding to the H/ACA RNA induces a conforma-tion of the H/ACA RNA–substrate RNA complex that is recog-nized by initial rigid docking followed by induced fit. The H/ACARNP proteins are not required for positioning of the substrate RNAon the H/ACA RNA � pocket. The absence of topological con-straints means that substrates for one or both � pockets in anH/ACA RNA can be loaded and unloaded successively forpseudouridylation without dissociation of the RNA from the RNP.

Materials and MethodsRNA Synthesis and Purification. Unlabeled, uniformly 13C,15N-labeled, and A, U, G, and C specifically 13C,15N-labeled RNA wereprepared by in vitro transcription with appropriate unlabeled (Am-ersham, Uppsala, Sweden) and/or 13C,15N-labeled (Silantes,Munchen, Germany) nucleoside triphosphates (NTPs) by usingpurified His6-tagged T7 RNA polymerase with synthetic double-stranded DNA templates and purified as described (43, 44). TheRNA construct of human H/ACA snoRNA U65 3� � pocket,U65hp, was designed as described in ref. 31 (Fig. 1B). The substrateRNA, S14 (5�-UUCGGCUCUUCCUA-3�), which contains thetarget site sequence of rRNA (4423–4435) except for the G4422U

Fig. 5. Modeling of the substrate RNA–snoRNP complex. (A) Docking of U65hp/S14 onto the structure of the TruB–TSL complex (1K8W) (34). The P1S and UCloop of U65hp/S14 was superimposed on the TSL. RNA is shown as stick (nucleotides) and backbone ribbon (backbone), with TSL in gray and U65hp/S14 coloredas in Fig. 3B. TruB is shown as ribbon representation in yellow, except for the thumb region in orange. In Cbf5 and Dyskerin, �9, �4, and the following linkerare absent. (B) Expanded view of A showing the modeled position of U65hp/S14 in the catalytic site. Arrows indicate required movement of U41 (red) and C42(cyan) to occupy the positions of these nucleotides (gray) in the TruB–TSL complex. For clarity, the rest of the TSL and nucleotides are omitted. (C) U65hp/S14modeled onto the archaeal snRNP (2HVY) (30) in the same position as in TruB. The shorter disordered thumb region is shown in orange, with disordered residues(dots) ‘‘moved’’ to contact the P1S major groove. (D) Comparison of the structures of free U65hp (2EUY) (31) (Left), the Afu46 snoRNA in the snoRNP (2HVY) (30)(Center), and the U65hp/S14 complex (Right). The UUCG tetraloop on U65hp has been deleted in the figure, and the substrate strand is shown in yellow.

Wu and Feigon PNAS � April 17, 2007 � vol. 104 � no. 16 � 6659

BIO

PHYS

ICS

substitution, initially was synthesized enzymatically as a 22-nt RNA(5�GGCCUUAAUUCGGCUCUUCCUA-3�) with the additional8 nt at its 5� end. The RNA subsequently was cleaved at the A8pU9step by a specifically engineered 10–23 DNAzyme that producesRNA fragments with 5�-OH and 3� phosphates as described in ref.45. The cleavage reaction (0.5 mM RNA/50 �M DNAzyme/50 mMTris�Cl, pH 8.0/100 mM NaCl/60 mM MgCl2) was allowed toproceed at 37°C for 4 h and then quenched by the addition of EDTAto a final concentration of 100 mM. All purified RNAs weredesalted and exchanged extensively into 5 mM sodium phosphatebuffer (pH 6.0) and concentrated by using an Amicon filtrationsystem. The U65hp/S14 complexes were prepared by titrating 0.5–1mM S14 RNA into 0.5–1 mM U65hp (or vice versa) and concen-trating as above to 1–1.5 mM. A total of nine samples was used forNMR structure studies: two samples at 1:1 molar ratio (unlabeledU65hp/unlabeled S14 and uniformly 13C,15N-labeled U65hp/S14)and seven samples at 1:1.5 molar ratio (A, U, G, or C selectively13C,15N-labeled U65hp/unlabeled S14 and A � U, G, or C selec-tively 13C,15N-labeled S14/unlabeled U65hp).

NMR Spectroscopy and Structure Calculations. All NMR spectra wererecorded at 288 K on Bruker (Billerica, MA) DRX 500-, 600-, and800-MHz spectrometers. Spectra were processed and analyzed byusing Bruker XWINNMR and Sparky 3.113 (T. D. Goddard andD. G. Kneller, University of California, San Francisco, CA).Resonances assignments were obtained from analysis of 2DNOESY, 1H-13C and 1H-15N HSQC, 3D HCCH-TOCSY, 2DHCCH-COSY (46), and a suite of 2D-filtered/edited NOESY (F2f,F1fF2e, F1fF2f, and F1eF2e) experiments (47) on the individual13C,15N-base-type-specific labeled RNA samples, as previouslydescribed (48). 3JH3�P and 3JCP were measured by using 31P spin-echo difference constant-time HSQCs to determine � and � torsionangles (49). JNN-HNN-COSY experiments were used to detecthydrogen bonding in Watson–Crick base pairs (50). Residualdipolar couplings (RDCs) were measured for 1JHC in 5% C8E5/

octanol (r � 0.87) (51) by using constant-time, coupling-enhancedHSQC.

Interproton distances were generated from 2D NOESY spectraacquired on unlabeled samples in 2H2O and in H2O as well as2D-filtered/edited NOESY and 3D NOESY-HMQC spectra (47,49) acquired on base-selectively 13C,15N-labeled samples. NOEdistance restraints were classified as very strong (2.5 Å), strong (3.5Å), medium (4.5 Å), weak (5.5 Å), and very weak (6.5 Å), and thelower boundary was defined as the van der Waals radius (1.8 Å). Atotal of 288 experimental dihedral angle restraints for �, �, �, and� angles were incorporated in the structure calculation as describedin ref. 49. Additional 116 A-form dihedral angle restraints for �, �,and angles were included for nucleotides in the helices (46). Finalstructure calculations included hydrogen bonding distance re-straints for the 21 Watson–Crick base pairs, consistent with exper-imental data.

The structures of the U65hp/S14 complex were calculated byusing X-PLOR-NIH (52) starting from extended, unfolded RNAsseparated by 70 Å in space by using NOE distances and dihedralangle restraints, following standard X-PLOR protocols. Structureswith no experimental restraint violations (distance �0.2 Å anddihedral angles �5°) from the initial 200 calculated structures thenwere subjected to refinement against 32 residual dipolar couplingsas described (53, 54). The grid search for the optimal values of themagnitude and asymmetry of the alignment tensor produced anoptimal value of Da � �45.0 Hz and R � 0.55, respectively. Theforce constants for the back-calculated residual dipolar couplingsgradually were increased from 0.001 to 0.2 kcal�mol�1�Hz�2. Ex-perimental restraints and structural statistics for the 20 lowest-energy structures are included in SI Table 1. All structures wereviewed and analyzed with MOLMOL (55), and figures weregenerated by using MOLMOL and Corel Draw.

We thank May Khanna for contributions in the early stages of this work.This work was supported by National Institutes of Health GrantsGM37254 and GM48123 (to J.F.).

1. Charette M, Gray MW (2000) IUBMB Life 49:341–351.2. Ofengand J, Malhotra A, Remme J, Gutgsell NS, Del Campo M, Jean-Charles S, Peil

L, Kaya Y (2001) Cold Spring Harb Symp Quant Biol 66:147–159.3. Decatur WA, Fournier MJ (2002) Trends Biochem Sci 27:344–351.4. Maden BE, Corbett ME, Heeney PA, Pugh K, Ajuh PM (1995) Biochimie 77:22–29.5. Grosjean H, Sprinzl M, Steinberg S (1995) Biochimie 77:139–141.6. Bakin A, Ofengand J (1993) Biochemistry 32:9754–9762.7. Bakin A, Lane BG, Ofengand J (1994) Biochemistry 33:13475–13483.8. Koonin EV (1996) Nucleic Acids Res 24:2411–2415.9. Ganot P, Caizergues-Ferrer M, Kiss T (1997) Genes Dev 11:941–956.

10. Ni J, Tien AL, Fournier MJ (1997) Cell 89:565–573.11. Tollervey D, Kiss T (1997) Curr Opin Cell Biol 9:337–342.12. Darzacq X, Jady BE, Verheggen C, Kiss AM, Bertrand E, Kiss T (2002) EMBO J

21:2746–2756.13. Torchet C, Badis G, Devaux F, Costanzo G, Werner M, Jacquier A (2005) RNA

11:928–938.14. Richard P, Darzacq X, Bertrand E, Jady BE, Verheggen C, Kiss T (2003) EMBO J

22:4283–4293.15. Mitchell JR, Cheng J, Collins K (1999) Mol Cell Biol 19:567–576.16. Enright CA, Maxwell ES, Sollner-Webb B (1996) RNA 2:1094–1099.17. Lafontaine DLJ, Bousquet-Antonelli C, Henry Y, Caizergues-Ferrer M, Tollervey D

(1998) Genes Dev 12:527–537.18. Henras A, Henry Y, Bousquet-Antonelli C, Noaillac-Depeyre J, Gelugne JP, Caizer-

gues-Ferrer M (1998) EMBO J 17:7078–7090.19. Girard JP, Lehtonen H, Caizerguesferrer M, Amalric F, Tollervey D, Lapeyre B (1992)

EMBO J 11:673–682.20. Bousquet-Antonelli C, Henry Y, Gelugne JP, Caizergues-Ferrer M, Kiss T (1997) EMBO

J 16:4770–4776.21. Watkins NJ, Gottschalk A, Neubauer G, Kastner B, Fabrizio P, Mann M, Luhrmann R

(1998) RNA 4:1549–1568.22. Charpentier B, Muller S, Branlant C (2005) Nucleic Acids Res 33:3133–3144.23. Henras AK, Capeyrou R, Henry Y, Caizergues-Ferrer M (2004) RNA 10:1704–1712.24. Wang C, Meier UT (2004) EMBO J 23:1857–1867.25. Baker DL, Youssef OA, Chastkofsky MI, Dy DA, Terns RM, Terns MP (2005) Genes

Dev 19:1238–1248.26. Darzacq X, Kittur N, Roy S, Shav-Tal Y, Singer RH, Meier UT (2006) J Cell Biol

173:207–218.

27. Hamma T, Reichow SL, Varani G, Ferre-D’Amare AR (2005) Nat Struct Mol Biol12:1101–1107.

28. Manival X, Charron C, Fourmann JB, Godard F, Charpentier B, Branlant C (2006)Nucleic Acids Res 34:826–839.

29. Rashid R, Liang B, Baker DL, Youssef OA, He Y, Phipps K, Terns RM, Terns MP, LiH (2006) Mol Cell 21:249–260.

30. Li L, Ye K (2006) Nature 443:302–307.31. Khanna M, Wu H, Johansson C, Caizergues-Ferrer M, Feigon J (2006) RNA 12:40–52.32. Bortolin M-L, Ganot P, Kiss T (1999) EMBO J 18:457–469.33. Ganot P, Bortolin ML, Kiss T (1997) Cell 89:799–809.34. Hoang C, Ferre-D’Amare AR (2001) Cell 107:929–939.35. Pan H, Agarwalla S, Moustakas DT, Finer-Moore J, Stroud RM (2003) Proc Natl Acad

Sci USA 100:12648–12653.36. Lescoute A, Westhof E (2006) RNA 12:83–93.37. Lilley DM (2000) Q Rev Biophys 33:109–159.38. Batey RT, Williamson JR (1998) RNA 4:984–997.39. Rozhdestvensky TS, Tang TH, Tchirkova IV, Brosius J, Bachellerie JP, Huttenhofer A

(2003) Nucleic Acids Res 31:869–877.40. Klein DJ, Schmeing TM, Moore PB, Steitz TA (2001) EMBO J 20:4214–4221.41. Kittur N, Darzacq X, Roy S, Singer RH, Meier UT (2006) RNA 12:2057–2062.42. Chaudhuri BN, Chan S, Perry LJ, Yeates TO (2004) J Biol Chem 279:24585–24591.43. Dieckmann T, Feigon J (1997) J Biomol NMR 9:259–272.44. Batey RT, Battiste JL, Williamson JR (1995) Methods Enzymol 261:300–322.45. Santoro SW, Joyce GF (1997) Proc Natl Acad Sci USA 94:4262–4266.46. Cromsigt J, van Buuren B, Schleucher J, Wijmenga SS (2001) Methods Enzymol

338:371–399.47. Peterson RD, Theimer CA, Wu H, Feigon J (2004) J Biomol NMR 28:59–67.48. Theimer CA, Blois CA, Feigon J (2005) Mol Cell 17:671–682.49. Wu H, Yang PK, Butcher SE, Kang S, Chanfreau G, Feigon J (2001) EMBO J

20:7240–7249.50. Luy B, Marino JP (2000) J Am Chem Soc 122:8095–8096.51. Ruckert M, Otting G (2000) J Am Chem Soc 122:7793–7797.52. Schwieters CD, Kuszewski JJ, Tjandra N, Clore GM (2003) J Magn Reson 160:65–73.53. de Alba E, Tjandra N (2002) Prog Nucl Magn Reson Spectros 40:175–197.54. Theimer CA, Finger LD, Trantirek L, Feigon J (2003) Proc Natl Acad Sci USA

100:449–454.55. Koradi R, Billeter M, Wuthrich K (1996) J Mol Graphics 14:51–55.

6660 � www.pnas.org�cgi�doi�10.1073�pnas.0701534104 Wu and Feigon