An RNA-Binding Complex Involved in Ribosome Biogenesis Contains a Protein with Homology to tRNA CCA-Adding Enzyme

The structure of a complex of two ribosome synthesis factors and identification of their ribosomal binding sites provides insights into early stages of ribosome biogenesis.


Introduction
Ribosomes, the protein factory, are large RNA-protein complexes composed of small and large subunits (SSUs and LSUs). The core structure and function of ribosomes are universally conserved, but eukaryotic ribosomes are 40% larger than their bacterial counterparts mainly due to the presence of additional ribosomal proteins (r-proteins) and expansion segments in rRNAs [1,2]. In addition to their greater structural complexity, eukaryotic ribosomes are assembled through a substantially more complex process that involves approximately 200 trans-acting protein factors and 75 small nucleolar RNAs (snoRNAs) in the yeast Saccharomyces cerevisiae. In contrast, only several tens of ribosome synthesis factors have been found in bacteria [3]. These conserved eukaryotic factors are involved in modification and processing of pre-rRNAs, coordination of rRNA folding, and the assembly of ribosomal proteins and exportation of preribosomal particles from the nucleolus to the cytoplasm (see recent reviews [4][5][6][7]).
In yeast, ribosome synthesis begins in the nucleolus with the RNA polymerase I-mediated transcription of a 35S pre-rRNA, which encodes 18S, 5.8S, and 25S rRNAs as well as four external and internal transcribed spacers (59-ETS, 39-ETS, ITS1, and ITS2). The 35S pre-rRNA associates cotranscriptionally with nearly 50 nonribosomal proteins, U3 snoRNA, and a subset of SSU r-proteins into the enormous 90S preribosome or small subunit processome, which can be visualized as a terminal ball on nascent rRNAs by electron microscopy [4,8,9]. Within the 90S preribosome, the 35S pre-rRNA is sequentially cleaved at sites A0, A1, and A2, and these early cleavages can occur during or after transcription [10]. Following these cleavages and a dramatic change in protein composition, a pre-40S particle is released that contains 20S pre-rRNA, which is the 59-product of A2 cleavage, most of SSU r-proteins, and a handful of nonribosomal factors [11,12]. The pre-40S particle is exported to the cytoplasm and associates with LSU to complete its maturation [13,14]. The 39 product of A2 cleavage, 27SA2 pre-rRNA, is assembled into pre-60S particles and maturates into 5.8S/25S rRNA.
The nucleolus harbors numerous box C/D and box H/ACA snoRNAs that mostly function to guide 29-O-methylation and pseudouridylation of rRNA. In addition, four snoRNAs in yeast, the U3 and U14 C/D snoRNAs, and the snR30 (U17 in humans) and snR10 H/ACA snoRNAs are required for early processing of 18S rRNA (see review [15]). U3, U14, and snR30 are essential in yeast and thought to be universally conserved in eukaryotes, whereas snR10 is nonessential and appears to be yeast-specific. The former three are known to function by binding to complementary sites on pre-rRNA, but the mode of action remains unknown for snR10 [16,17].
Most ribosome synthesis factors probably have been identified in yeast through genetic study and biochemical purification of various preribosomal particles. Their association with rRNA processing steps and preribosome types has been generally assigned. The current major challenge is to understand the molecular function of individual factors and the structure and assembly of preribosomal particles. A few factors of the 90S preribosome have been shown to form independent complexes, including UTP-A (t-UTP), UTP-B, UTP-C, U3 snoRNP, the Mpp10-Imp4-Imp3 complex, the Bms1-Rcl1 complex, and the Noc4-Nop14 complex [8,[18][19][20][21][22]. Several complexes and factors were shown to assemble into the 90S preribosome in a hierarchical order [23,24]. Recently, the UV crosslinking and analysis of cDNA (CRAC) method was applied to locate precise binding sites of ribosome synthesis factors on pre-rRNAs [25][26][27][28][29]. Cryoelectron microscopy structures have been determined for late pre-40S and late pre-60S particles [12,30,31]. However, very little is currently known for structure of early-acting SSU synthesis factors and how and where they associate with 90S preribosomes.
Utp22 and Rrp7 are two proteins associated with early 90S preribosomes and form the UTP-C complex together with four subunits of casein kinase 2 [8,18]. The UTP-C complex is also associated with transcription factor lfh1 to form the CURI complex, which is implicated in coordination of r-protein production with ribosome biogenesis [32]. Like most early-acting SSU synthesis factors, Utp22 and Rrp7 are required for early processing of 18S rRNA and 40S ribosome formation [33][34][35].
Utp22 and Rrp7 do not contain any recognizable domain, rendering their function highly mysterious.
In this study, we determined the cocrystal structure of the large complex of Utp22 (1,237 residues) and Rrp7 (297 residues) and the NMR structure of a C-terminal fragment of Rrp7, which was not visible in the crystal structure. We found unexpected structural homology between Utp22 and class I tRNA CCA-adding enzyme and between Rrp7 and the RNA-recognition motif (RRM). We identified functionally important domains in the two proteins with structure-based mutagenesis analysis in yeast. We further studied how and where the Utp22/Rrp7 complex assembles into the 90S preribosome. We found that the flexible C-terminal tail of Rrp7 is the key RNA-binding domain that anchors the complex into preribosomes. We mapped the in vivo RNA-binding target of Rrp7 using UV crosslinking and found that Rrp7 binds to the central domain of 18S rRNA and shares a neighborhood with the processing H/ACA snoRNAs snR30 and snR10. We demonstrated that snR30 is required for the stable incorporation of Rrp7 into preribosome. Our comprehensive structure-function analysis of Utp22 and Rrp7 provides important insight into their evolutionary origin and functional context in preribosomes.

Structural Determination and Overall Structure of the Utp22 and Rrp7 Complex
Both Utp22 and Rrp7 are essential genes in yeast and conserved in eukaryotes ( Figure 1A, Texts S1 and S2). However, analysis of their sequences failed to reveal any recognizable domain. We sought to solve the crystal structure of Utp22 and Rrp7 to gain insight into their molecular function. The full-length proteins of yeast Utp22 and Rrp7 were coexpressed using recombinant baculoviruses in insect cells and then copurified and cocrystallized. The structure of the complex was determined by singlewavelength anomalous dispersion phasing based on a Se-labeled crystal and refined to 1.97 Å resolution with an R work /R free of 0.210/0.239 (Table S1). N-terminal residues 1-80 of Utp22, Cterminal residues 190-297 of Rrp7, and several internal loops of each protein were not visible in the crystal structure, likely due to structural flexibility. SDS-PAGE analysis of dissolved crystals showed that Utp22 was intact and Rrp7 was partially degraded (unpublished data).
Utp22 and Rrp7 form a 1:1 dimer that adopts a saddle-like structure with approximate dimensions of 137 Å666 Å669 Å ( Figure 1B,C). The N-terminal half (N-half) and C-terminal half (C-half) of Utp22 are structurally similar to each other and are arranged in tandem along the longest dimension. Rrp7 binds at the C-half of Utp22, forming a raised end.

Utp22 Is a Structural Homolog of Dimeric Class I tRNA CCA-Adding Enzyme
We searched for structural homolog of Utp22 using the DALI server [36] and surprisingly found that Utp22 shares significantly structural homology with class I tRNA CCA-adding enzymes. CCA-adding enzymes are responsible for the synthesis or repair of the universally conserved CCA sequence at the 39 end of tRNAs [37]. These enzymes catalyze three different polymerization reactions using a single active site and no nucleic acid template. CCA-adding enzymes are classified into two classes: class I is found in archaea, and class II is distributed in eukaryotes and bacteria. Enzymes of both classes are composed of four domainsnamely, the head, neck, body, and tail domains. The head domain is the catalytic domain, which is also conserved in the superfamily of nucleotide polymerases. The neck domain constitutes part of the

Author Summary
Ribosomes are large RNA-protein complexes that manufacture proteins in all living organisms. Synthesis of large and small ribosomal subunits is a fundamental and enormous task that requires activities of approximately 200 assembly factors in eukaryotic cells. These factors transiently associate with the ribosome, forming a series of pre-ribosomal particles. We currently have a poor understanding of the structure and assembly of ribosome precursors. Utp22 and Rrp7 are two interacting proteins present in early precursors of the small ribosomal subunit. In this study, we determined the structure of the Utp22 and Rrp7 complex by X-ray crystallography and NMR and dissected their functional domains by mutagenesis. The structure of Utp22 reveals an unexpected structural similarity to the tRNA CCA-adding enzyme, providing insight into the evolutionary origin of Utp22. Utp22 apparently lacks any enzymatic activity and functions instead as a structural building block. Rrp7 associates extensively with Utp22 and appears to be anchored to preribosomes via a flexible RNA-binding tail. We used RNAprotein crosslinking to identify the binding site and neighboring factor of Rrp7 on pre-ribosomes. Our study provides a detailed insight into the structure of small ribosomal subunit precursors.
nucleotide-binding pocket, and the body and tail domains bind the tRNA acceptor stem. The two classes share a similar head domain but differ significantly in the remainder of their structures.
The structure of Utp22 can be divided into eight domains (D1 though D8) (Figures 1 and 2A). Both the N-half (D1-D4, residues 81-689) and C-half (D5-D8, residues 699-1237) are structurally similar to class I CCA-adding enzyme in all four individual domains (Figures 3A and S1A-D). We compare the structure of the N-half and C-half with that of Archaeoglobus fulgidus CCAadding enzyme (AfCCA) bound to a tRNA acceptor stem [38]. In the N-half, D1 and D2 combined are superimposable on the head and neck domains of AfCCA. D3 can be aligned with the body domain, but the orientation of D3 with regard to D1-D2 is not conserved. D4 is a small insertion in D3 and shares topology with the tail domain. The four domains in the C-half of Utp22 also bear strong structural similarity with the four domains of AfCCA. Nevertheless, Utp22 and class I CCA-adding enzyme display considerable variations in the length and orientation of secondary structural elements, which precludes detection of their homology based on sequence.
Class I CCA-adding enzymes form a symmetric homodimer. Utp22 is likewise an intramolecular dimer composed of two copies of CCA-adding enzyme modules. The N-half and C-half structures of Utp22 are roughly related by a 180 degree rotation along a pseudo-dyad axis at the interface of D2 and D6 ( Figure 1B), and contact each other at D2, D3, and D6 with an extensive interface of 1,621 Å 2 . However, the dimer interface is significantly different between class I CCA-adding enzyme and Utp22 ( Figure  S1E).
Despite its structural similarity with CCA-adding enzymes, Utp22 is unlikely to possess any polymerase activity. The catalytic domain of CCA-adding enzyme contains three carboxylates (Glu59, Asp61, and Asp110 in AfCCA) that are responsible for binding Mg 2+ ions and catalyzing the phosphoryl transfer reaction. These catalytic residues are degenerate and nonconserved in D1 and D5 of Utp22 (Figures 2A and S1A-B). The only exception is the Asp204 residue, which is conserved at the corresponding active site of D1. However, Ala substitution of Asp204 did not affect yeast growth (see below), indicating that Asp204 is functionally dispensable. In addition, several structural elements in D1 and D5, including the loop between b5 and a6, the N-terminal part of a6, a20, and the N-terminal part of a21 would occlude the RNAbinding paths in both halves of Utp22 and prevent access of substrate tRNA to the active site. These structural observations indicate that Utp22 is inactive as a CCA-adding enzyme.

The N-Terminal Domain of Rrp7 Is a Deviant RRM Domain
The structure of Rrp7 is composed of an N-terminal domain (NTD, residues 1-156), a linker region (residues 157-189), and a C-terminal domain (CTD, residues 190-297). The CTD is not visible in the crystal structure. The NTD adopts a two-layered a/b fold, which is similar to the fold of RRM according to DALI search (best z-score = 4.8, Figure 3B,C). The structure of Rrp7 NTD can be aligned with that of U2AF65 RRM1 domain with a root mean standard deviation (RMSD) of 1.777 Å over 49 Ca pairs ( Figure 3C) [39]. The classic RRM fold has a topology of b1-a1-b2-b3-a2-b4 with juxtaposed N-and C-termini. By contrast, the NTD of Rrp7 displays a cyclic permutation of RRM topology: the strand equivalent to RRM b4 is shuffled to the N-terminus of the strand equivalent to RRM b1. Moreover, Rrp7 has an extra strand b1, which, together with other four b-strands, forms an antiparallel five-stranded b-sheet. Other atypical RRM domains generally have a similar fold as the canonical RRM domain, but differ in RNA-binding mode [40].
The RRM domain is known to recognize single-stranded RNA through its b-sheet surface. Two exposed aromatic (sometimes hydrophobic) residues on strands b1 and b3 are key residues that stack on RNA bases ( Figure 3D). The two equivalent residues in Rrp7-that is, Phe54 and Leu125-are conserved ( Figure 2B, Text S2) and appear to be accessible for RNA binding. However, replacement of Phe54 to Ala caused no effect on yeast growth (see below), indicating that the putative RNA-binding residue is dispensable. Utp22 associates with the NTD of Rrp7 near the putative RNA-binding site and may interfere with RNA binding. Hence, the RRM-like NTD of Rrp7 is distinct from classic RRM domains in terms of structure and function.

NMR Structure of a C-Terminal Fragment of Rrp7 Reveals Two Flexible a-Helices
The CTD of Rrp7 is invisible in the crystal structure but is highly conserved and functionally important (see below). We set to determine its solution structure using NMR. A fragment spanning highly conserved residues 256-297 of Rrp7 was expressed in E. coli and labeled with 15 N and 13 C to facilitate resonance assignment.
The structure was determined using 563 NOE-based distance constraints and 68 chemical shift-based backbone dihedral constraints ( Figure S2 and Table S2). The structure shows that  Figure 1A. Dashed lines denote disordered regions. Residues whose surface areas are buried by 30 Å 2 and 10 Å 2 due to the intermolecular association of Utp22 and Rrp7 are marked with solid and empty circles, respectively. Utp22 residues that are equivalent to three catalytic acidic residues in class I CCA-adding enzyme are labeled with solid triangles. doi:10.1371/journal.pbio.1001669.g002 the C-terminal 40-residue fragment is composed of two a-helices linked by a flexible hinge ( Figures 1C and S2B). The two helices are not packed because no long-range NOE was identified between them.

Dimer Interface Between Utp22 and Rrp7
The NTD and linker region of Rrp7 associate with D6, D7, and D8 of Utp22 through an extensive interface, which buries 3,116 Å 2 of solvent accessible surface area per subunit ( Figure 4A,B). At the center of the interface, one end of the bsheet of the Rrp7 NTD, which is composed of strands b1, b2, and b3 and surrounding loops, packs against D7 and D8 of Utp22. In addition, two prominent tentacle-like structures project from the NTD to reach the more distant D6 of Utp22. One tentacle comprises a long loop between strands b4 and b5. This loop is disordered in the C-terminal half and its sequence is highly variable among Rrp7 orthologs ( Figure 2B and Text S2). The other tentacle is the linker (a4 and a5) that connects the NTD with the flexible C-terminal tail. The intermolecular association is stabilized through a large number of hydrophobic, polar, and electrostatic interactions ( Figure 4C,D). Somewhat surprisingly, the dimer interface is only moderately conserved, including the hydrophobic faces on helix a5 and strands b2 and b3 of Rrp7 and a few scattered sites of Utp22 ( Figure 4A,B).

The NTD and CTD of Rrp7 Are Essential for Function
Next, we investigated the contribution of each domain of Rrp7 to Utp22 binding and function. Individual domains of Rrp7 were deleted or mutated and assessed for effect on the interaction with Utp22 using a two-hybrid assay ( Figure 4E) and their effect on yeast growth by complementation with the rrp7D strain ( Figure 4F).
Rrp7 and its CTD deletion mutant (D190-297) strongly bound Utp22 in two-hybrid assays, whereas the CTD alone (D1-190) failed to bind Utp22. This is consistent with the structural observation that the CTD is not involved in the intermolecular interaction. Nevertheless, deletion of the CTD was lethal, indicating that it plays an essential role.
Removal of the b4-b5 loop (D95-105)-that is, the Utp22binding tentacle within the NTD-had no effect on the interaction with Utp22. Deletion of the linker region (D163-188)-that is, the other tentacle-reduced the interaction with Utp22 because the two-hybrid reporter strain grew under intermediate stringent but not highly stringent conditions. However, neither tentacle is required for yeast growth ( Figure 4F).
Phe38 is located at the hydrophobic interface between the NTD of Rrp7 and Utp22 D5. Substitution of Phe38 with Asp decreased the interaction with Utp22 but did not detectably affect yeast growth, suggesting that the weakened intermolecular association was tolerated. Furthermore, the interaction between Utp22 and Rrp7 was unaffected by removal of either half of or the entire NTD (D1-89, D1-156). Apparently, the linker region is sufficient for Utp22 association in these cases. The F38D mutation was more disruptive to the interaction with Utp22 than the domain deletion mutations, likely because the negatively charged Asp residue drives the NTD away from the hydrophobic binding face of Utp22 and affects the conformation of the linker region. In contrast with the viable F38D mutation, the NTD deletion mutants cannot support yeast growth, indicating that the NTD has an additional essential function other than Utp22 binding. These results show that Rrp7 and Utp22 are associated with multiple and somehow redundant interfaces and that disruption of a single interface is tolerated in vivo. The redundancy of interface also provides an explanation for its moderate conservation. (E) Two-hybrid interaction between Utp22 and Rrp7. Utp22 was fused to GAL4 DNA-binding domain (BD) as bait, and Rrp7 and its mutants were fused to the GAL4 activation domain (AD) as prey. Ten-fold serial dilutions of AH109 cells containing bait and prey vectors were spotted onto plates containing SC medium lacking Leu and Trp as a growth control, and onto plates containing SC medium lacking Leu, Trp, and His and containing 5 mM 3-amino-1,2,4-triazole (3AT) and onto plates containing SC medium lacking Leu, Trp, His, and Ade to assess protein interaction with increasing stringency. (F) Yeast growth assay of rrp7 mutants. The rrp7D strain, complemented by a URA3 RRP7 plasmid, was transformed with LEU2 plasmids containing wildtype (WT) RRP7, no RRP7 (Vector), or the indicated rrp7 mutations. Ten-fold serial dilutions of the transformants were spotted onto plates containing SC medium or SC with 5-fluoroorotic acid (5-FOA) to counterselect for the URA3 RRP7 plasmid and grown at 37uC, 30uC, or 20uC. (G) Sedimentation behavior of Rrp7 in sucrose gradients. Extracts of RRP7-HTP cells overexpressing FLAG-tagged Rrp7 or mutants D190-297, D1-89, or D1-156 from a 2 m plasmid were fractionated on 7%-50% sucrose gradients and analyzed by SDS-PAGE and Western blotting with peroxidase anti-peroxidase (PAP) and anti-FLAG antibody. (H) EMSA of snR5. 59-32 P-labeled snR5 was bound with 0, 100, 150, 200, 300, 400, 600, and 1,000 nM of indicated proteins. Rrp7 189-297 contains a His 6 -SMT3 tag and other proteins have a short His 6 -tag or not. doi:10.1371/journal.pbio.1001669.g004 The CTD of Rrp7 Is an RNA-Binding Domain Required for Association with Preribosomes Given the essential role of the Rrp7 NTD and CTD, we asked whether they function in preribosome association ( Figure 4G). We used a yeast strain that expresses His 7 -TEV-ProtA (HTP)-tagged Rrp7 from chromosome as well as FLAG-tagged Rrp7 from plasmid. Sucrose gradient sedimentation analysis shows that wildtype Rrp7 expressed from either chromosome or plasmid was distributed broadly from free protein fractions to large complexes that sediment at positions corresponding to those of 80S to polysomes and should correspond to 90S preribosomes. The CTD deletion mutant of Rrp7 was exclusively present in free protein fractions, indicating that the CTD is necessary for association with preribosomes. Conversely, the deletion of the NTD led to a predominant distribution in 80S-sized and larger particles, suggesting that the NTD is required for the dissociation of Rrp7 from preribosomes.
These mutant Rrp7 proteins were overexpressed under the control of the GAL1 promoter from a multicopy plasmid. We found that overexpression of the CTD deletion mutant caused slow growth and decreased levels of 40S ribosome, indicating a dominant negative effect ( Figure S3). Excessive CTD-lacking Rrp7, which is capable of binding Utp22 but unable to bind preribosomes, would sequester Utp22 in a nonfunctional state. Two NTD truncation mutants (D1-89, D1-156) displayed no dominant negative effect ( Figure S3), likely because they have an incomplete Utp22 binding interface and cannot compete with endogenous Rrp7.
The structural similarities with CCA-adding enzymes and RRM domains suggest that Utp22 and Rrp7 may directly contact RNA in preribosomes. We tested their RNA-binding activity using electrophoretic mobility shift assay (EMSA) with snR5, a yeast H/ ACA snoRNA with abundant secondary structures ( Figure 4H). Although snR5 is unlikely to be a natural target, such an analysis is useful to identify which protein and domain in the complex are involved in RNA binding. The Utp22 and Rrp7 complex and individual proteins all show at least general RNA-binding activities. The isolated NTD of Rrp7 displayed virtually no RNA binding, but the CTD of Rrp7 still strongly bound RNA. The RNA-binding activity of the Rrp7 CTD may account for its essential role in preribosome association.

The D2 and D4 Domains of Utp22 Are Functionally Important
To reveal the functional sites of Utp22, we identified the exposed conserved residues on Utp22 structure and assayed their functional importance using mutagenesis and complementation assays with the utp22D strain. Overall, the exposed surface of Utp22 in the Rrp7 complex is moderately conserved, and there are three conserved patches on D1, D2, and D4 ( Figure 5A). The conserved patch on D1 is composed mainly of basic residues. Substitutions of Lys217, Arg223, and Arg316 in this region with negatively charged glutamate, both singly and in combination, had no detectable effect on yeast growth ( Figure 5B), indicating that these residues do not play a significant role. The conserved patch on D2 around helix a2 consists of amino acids that have different properties. Incorporation of the L104E/L105D double mutation or the E109K single mutation into helix a2 caused no detectable effect on yeast growth, but the corresponding triple mutation slightly inhibited growth ( Figure 5B). This indicates that the conserved patch on D2 is functional.
The small D4 domain protruding from the main body displays the most conserved surface of Utp22. One face of D4 is mixed with highly conserved basic and hydrophobic residues, including Arg656 and Arg657. Although the single mutations of R656E and R657E caused no obvious growth phenotype, the R656E/R657E double mutation inhibited yeast growth at 30uC and inhibited growth more significantly at 37uC and 20uC ( Figure 5B). Deletion of the entire D4 domain resulted in a similar degree of growth defect as the double mutation, indicating that these two arginine residues are major functional residues in D4. These results show that D4 is a key functional domain of Utp22.

Rrp7 Crosslinks to the Central Domain of 18S rRNA and to snR10
The general RNA-binding activity of Utp22 and Rrp7 suggest that they directly bind to the pre-rRNA in preribosomes. We attempted to map their RNA-binding sites using the CRAC crosslinking approach [27]. To this end, the UTP22 or RRP7 chromosomal gene was tagged with a C-terminal HTP tag. Following UV-crosslinking in vivo, the HTP-tagged protein was DD4 is deletion of residues 630-667. Ten-fold dilutions of the transformants were spotted on plates containing SC medium or SC with 5-FOA to counterselect for the UTP22 URA3 plasmid and grown at 37uC, 30uC, or 20uC. doi:10.1371/journal.pbio.1001669.g005 affinity purified via two steps including one conducted under denaturing conditions. The crosslinked RNA was cloned into cDNA and subjected to Solexa sequencing (Table S3). Utp22 crosslinked rather weakly with RNA, and its CRAC result appears to be contaminated by Rrp7-crosslinked RNAs and is therefore not discussed.
Rrp7 crosslinked efficiently with RNA, as evident by the intense radioactive signal of 32 P-labeled crosslinked RNA ( Figure 6A). Alignment of sequence reads to the reference genome sequence of S. cerevisiae revealed that 92.75% of the mapped reads are derived from pre-rRNA (Table S3). A major peak of rRNA reads was mapped to helix E of extension segment 6 (ES6E) of 18S rRNA, and minor peaks were also found in helix h26 ( Figure 6B). ES6E and h26 belong to the central domain of 18S rRNA, which constitutes a major part of the platform of ribosome structure and covers the body with ES6 ( Figure 6E). One crosslinking peak at 39end of 25S rRNA was a frequent contamination [25,26,28]. Nucleotide deletions and substitutions in mapped reads are highly indicative of actual crosslinking sites. Such analyses revealed several cross-linking sites on ES6E (nt 812, 814-816, 822, 829-830 based on deletions) and one cross-linking site (nt 1051-1052) on h26 ( Figure S4A-C).
Interestingly, the crosslinking region of Rrp7 on ES6E was flanked by two motifs previously found to be targeted by snR30, a conserved H/ACA snoRNA essential for 18S rRNA processing ( Figure 6C) [41][42][43]. The two 6-nt motifs, termed rm1 and rm2, are complementary to the bipartite sequences, termed m1 and m2, at the base of an internal loop in the 39 hairpin of snR30 ( Figure 6D) [41].

snR30 Is Required for the Stable Association of Rrp7 to Preribosomes
The spatial proximity between the binding sites of Rrp7 and snR30 on 18S ES6 raises a question as to whether they are dependent on each other to bind preribosomes. To examine whether association of snR30 with preribosomes depends on Rrp7, the HTP-tagged RRP7 chromosomal gene was placed under the control of the GAL1 promoter, which is active in the presence of galactose and repressed in the presence of glucose. The accumulation of Rrp7 in the GAL::RRP7-HTP strain was efficiently depleted 12 h after shifting from galactose-to glucose-containing medium ( Figure 7A). Sucrose gradient sedimentation analysis showed that depletion of Rrp7 led to the disappearance of free 40S peak but did not affect the distribution of U3 in large preribosomes ( Figure 7B,C), consistent with the previous results [24]. In normal cells, only a small fraction of snR30 cosediments with large preribosomes [41,44]. Depletion of Rrp7 caused no detectable change of the sedimentation profile of snR30, suggesting that Rrp7 does not control the association or dissociation of snR30. The distribution of snR10, which Rrp7 crosslinks, was not altered either in the absence of Rrp7.
Next, we examined whether snR30 affects the association of Rrp7 with preribosomes. To this end, the chromosomal SNR30 gene in the RRP7-HTP strain was placed under the control of the GAL1 promoter. The expression of snR30 can be efficiently repressed in glucose medium ( Figure 7D). In wild-type cells, majority of Rrp7 was distributed in large particles. Upon depletion of snR30, Rrp7 was still distributed in large particles but an increase in free protein fractions was observed, suggesting that snR30 may affect the strength or dynamics of Rrp7 binding to preribosomes. In addition, the depletion of snR30 seemed not to change the distribution of snR10 in sucrose gradients.
To directly analyze the association of Rrp7 with preribosomes, we determined RNA species coimmunoprecipitated with Rrp7-HTP. The 90S preribosome could contain 35S or 23S pre-rRNA; 23S pre-rRNA is resulted from cleavage at site A3 of 35S pre-rRNA without prior cleavage at sites A0, A1, and A2 ( Figure 7G). Immunoprecipitation of Rrp7-HTP coprecipitated much less U3 snoRNA and 35S/23S pre-rRNA in the absence of snR30 than in the presence of snR30 ( Figure 7H). This indicates that snR30 is required for the stable incorporation of Rrp7 into preribosomes.

Discussion
We have conducted a comprehensive structure-function analysis of Utp22 and Rrp7 and illustrated the way by which they assemble into the 90S preribosome. The complex structure of Utp22 and Rrp7 shows that they are unlikely to possess any enzymatic activity and that they rather function as an essential building block in the 90S preribosome. The binding interface between Utp22 and Rrp7 is so extensive that disruption of the intermolecular interaction at the NTD or the linker region of Rrp7 is well tolerated in vivo (Figure 4). The two proteins most likely function as a single stable module during association and dissociation with preribosomes.
We demonstrate that Rrp7 is an RNA-binding protein and efficiently crosslinks 18S rRNA. Somewhat surprisingly, the RNAbinding activity of Rrp7 is principally located on the CTD, not the RRM-like NTD. The CTD deletion mutant of Rrp7 failed to associate with preribosomes, even though it can still bind Utp22 in this case. We can infer that Utp22 cannot assemble into preribosomes on its own and should depend on Rrp7 for assembly. The CTD of Rrp7 serves as the primary anchor of the Utp22/Rrp7 complex on preribosomes.
Our NMR analysis reveals that the highly conserved C-terminal 40 residues of Rrp7 form two flexible a-helices. Such a structure is reminiscent of long tails present in many r-proteins, which are flexible in isolation yet contact rRNA over a long distance once assembled into the ribosome. We speculate that the CTD of Rrp7 could assume a similar mode during rRNA binding. While our data suggest that the CTD of Rrp7 anchors the Utp22/Rrp7 complex on 90S preribosomes through binding RNA, we cannot exclude other mechanisms, such as protein interaction, are involved.
Once assembled into the 90S preribosome, the Utp22/Rrp7 complex is expected to make contact with neighboring RNAs and proteins. Our mutagenesis data suggest that the NTD of Rrp7 and the conserved surface patches on D2 and D4 of Utp22 may be involved in interaction with other molecules to maintain a functional conformation of the 90S preribosome. In the absence of the essential NTD, Rrp7 can assemble into preribosomes but appears not to dissociate ( Figure 4G). In this case, maturation of the defective 90S preribosome may be inhibited, subsequently blocking the release of assembly factors. The central domain of 18S rRNA consists of helices h19-h26 and ES6. ES6 is the largest eukaryotic-specific extension segment in 18S rRNA and composed of five helices named A to E. In the 40S structure, helices h19-h26, together with the 39-end region of 18S rRNA, make up the platform, whereas ES6 lies over the solvent side of the body ( Figure 6C,E) [1,2].
Our observation that Rrp7 binds to ES6E and h26 in the central domain of 18S rRNA is correlated with several previous genetic and biochemical results ( Figure 6G). The r-protein S27 was found to be a high copy suppressor of the lethal phenotype of rrp7 deletion [35]. In the 40S structure, S27 binds h26 adjacent to the Rrp7 crosslinking site, corroborating the genetic interaction between S27 and Rrp7. In addition, depletion of two r-proteins S13 and S14, which bind to the platform, reduced the association of Utp22 and Rrp7, among other proteins, with 90S preribosomes [45]. S13 contacts S27 in the 40S structure and is also close to the Rrp7 crosslinking sites on ES6E and h26, whereas S14 binds at one edge of the platform. S13, S14, and S27 are all required for early processing of 18S pre-rRNA [46], and they may assemble together with Utp22/Rrp7 and other ribosome biogenesis factors around the central domain to form a structural module in 90S preribosomes.
shown in green. S27 is shown in blue, S13 and S14 in cyan, and other ribosomal proteins in grey. Major features of the ribosome and the 59   and immunoprecipitations (IP) of IgG Sepharose were analyzed by Western blotting to detect Rrp7-HTP and by Northern blotting to detect copurifed U3 snoRNA and pre-rRNAs. A probe D-A2 that hybridizes to a region between sites D and A2 was used to detect 35S, 23S, and 20S pre-rRNAs. The minor fast-migrating band of Rrp7-HTP marked by asterisk might be degradation or modification products and its identity was not studied. doi:10.1371/journal.pbio.1001669.g007 We find that the major binding region of Rrp7 on ES6E is flanked by two snR30-binding sites: rm1 and rm2. The middle sequence between rm1 and rm2 is predicted to adopt a hairpin when snR30 is bound to 18S rRNA ( Figure 6D) [41]. However, in the mature 40S structure, rm1 is part of helix C of ES6, rm2 forms a long-range base-pairing interaction with ES3 at the left foot, and the middle region is unpaired or comprises one strand of the ES6E helix ( Figure 6C). Apparently, dramatic structural changes should occur when the ES6E region is transformed from the snR30bound state to the mature state. Which state of ES6E is recognized by Rrp7 is unknown. Given that snR30 is required for the stably association of Rrp7 to preribosomes but not vice versa (Figure 7), Rrp7 might be recruited downstream of snR30 and recognize the intermediate snR30-bound hairpin structure of ES6E.
Among four processing snoRNAs present in yeast, snR10 is the only one that still has an unknown binding site in preribosome [16,17]. Our finding of Rrp7 crosslinking snR10 provides the first glimpse into the location of snR10 in preribosomes. Rrp7 crosslinks with the 39-hairpin of snR10 ( Figure S4D,E), however the function of snR10 39-hairpin remains uncertain [16,47]. The interaction between Rrp7 and snR10 is also supported by their genetic interaction with a common factor, Rrp5 ( Figure 6G). Mutations of Rrp5 displayed a synthetic lethal phenotype with snR10 deletion [48], and snR10 is a high-dose suppressor of an Rrp5 mutant [47]. Moreover, incorporation of Rrp7 in preribosome was found dependent on prior association of Rrp5 [24].
Our data suggest that Rrp7 is located near to two processing H/ ACA snoRNAs, snR30 and snR10, in preribosomes. To provide insight into other factors that are potentially associated with them around the central domain of 18S rRNA, we complied from the literature an interaction network map focused on the three molecules ( Figure 6G). In addition to the interactions and factors discussed above, the map also includes Utp23, Kri1, and Rok1. Utp23 and Kri1 are two early-acting SSU synthesis factors that bind the snR30 snoRNP [49][50][51]. Rok1, an essential RNA helicase, was identified in a synthetic lethal screen with snR10 deletion [52]. Rok1 is involved in release of snR30 [53] and is a high copy suppressor of an Rrp5 mutant [54]. In this map, snR30 plays a key role in preribosome assembly since it is required for assembly of Utp23, Kri1, and Rrp7 ( [49], this work) and the formation of a compact 90S particle at the terminus of nascent rRNAs [44].
Eukaryotic rRNAs contain many extension segments that contribute to increased structural complexity of eukaryotic ribosomes. The exact function of extension segments is elusive in most cases. The interaction of ES6E with snR30 [41] and Rrp7 shows that extension segments can play a role in binding ribosome synthesis factors. Another example is provided by recent cryo-EM structures of late pre-60S particles, which show that extension segment 27 of 25S rRNA interacts with the nuclear export factor Arx1 [30,31]. The interaction between rRNA extension segments and ribosome synthesis factors illustrates that the structure of eukaryotic ribosome coevolved with its assembly machinery.

Evolutionary Insight
The structural homology of Utp22 with dimeric class I CCAadding enzyme is intriguing. It appears that the eukaryotic rRNA processing machinery has borrowed a factor that is involved in tRNA processing during evolution. Finding a connection between tRNA and rRNA processing machinery is, however, not unprecedented. The MRP nuclease, which is responsible for pre-rRNA cleavage at site A3 in ITS1, is homologous to RNase P, which processes the 59-end of tRNA [55]. In addition, the catalytic subunit Cbf5 of H/ACA RNP is closely related to TruB, the synthase for tRNA pseudouridine 55 [56].
It is difficult to envision how a CCA-adding enzyme that processes tRNA evolved into an rRNA processing factor. In one scenario, the primordial eukaryotic SSU rRNA might bind a tRNA or contain a tRNA-like structure that recruits a dimeric CCA-adding enzyme. The CCA-adding enzyme might have been initially recruited for its RNA-binding property, thus allowing the unneeded polymerase active site to mutate. During the course of evolution, Utp22 recruited Rrp7 and began to rely on Rrp7 rather than its own RNA-binding ability to assemble into preribosomes. The original tRNA-binding channels of Utp22 were subsequently blocked. The D4 domain in the N-half of Utp22 remains functionally important; however, the two conserved arginine residues in Utp22 D4 do not correspond to the original tRNAbinding residues in the tail domain of CCA-adding enzyme, suggesting that D4 has a different mode in RNA binding or assumes a different function.
The Utp22 gene apparently evolved after duplication and conjugation of a class I CCA-adding enzyme gene. Notably, class I CCA-adding enzymes are specifically distributed in archaea, suggesting that Utp22 has an archaeal origin or shares a common ancestor with archaeal enzymes. In this regard, archaeal homologs have also been found for a subset of eukaryotic ribosome synthesis factors. These include RIO-type kinases, the ATPase Fab7, the RNA-binding protein Dim2/Pno1, the dimethyltransferase Dim1 (which is also present in bacteria), the nuclease Nob1, the RNA methyltransferase Emg1, and Brix domain proteins (Imp4, Ssf1, Rpf1, Rpf2, and Brx1), many of which function at late stages of 40S synthesis. In addition, H/ACA RNPs and C/D RNPs are conserved in archaea. They direct rRNA modification but are not known to mediate rRNA processing in archaea. The archaeal origin of Utp22 supports the notion that the eukaryotic ribosome synthesis machinery evolved from an archaeal-like system.

DNA Cloning and Protein Purification
Utp22 and Rrp7 were coexpressed in insect cells using the Bacto-Bac system (Invitrogen). The Utp22 gene was amplified by PCR from yeast genomic DNA and cloned into pFastBac-1 with no tag. The Rrp7 gene was cloned similarly with an N-terminal His 6 -tag followed by a PreScission cleavage site. The recombinant viruses were generated in SF21 cells according to the manufacturer's instruction. For coexpression of Utp22 and Rrp7, High Five cells were cultured in SF-900 II SFM medium at 27uC to a density of 2610 6 cells/ml and coinfected by viruses expressing each protein for 48-60 h. Cells were harvested from 1 L medium and resuspended in 100 ml of lysis buffer (50 mM Tris, pH 8.0, 500 mM NaCl, 30 mM imidazole, 5% glycerol, and 2 mM bmercaptoethanol). The sample was supplemented with two complete, EDTA-free protease inhibitor cocktail tablets (Roche) and lysed by sonication. After centrifugation at 200,000 g, the supernatant was loaded onto a 5-ml HisTrap column (GE Healthcare). After washing with lysis buffer, the protein was eluted with a linear gradient of imidazole. The combined fractions were diluted 3-fold with buffer A (50 mM Tris, pH 8.0, 5% glycerol) and incubated with PreScission protease overnight at 4uC to cleave the His 6 -tag from Rrp7. The protein was loaded onto a heparin column, washed with 500 mM NaCl, and eluted with 725 mM NaCl in buffer A. The protein was further purified with a HiLoad 16/60 Superdex 200 column using buffer 10 mM Tris (pH 8.0) and 200 mM NaCl, and then concentrated to 6.5 mg/ml for crystallization.
For selenomethionine labeling, the infected cells were spun down 8 h postinfection and resuspended in 1 L of SF-900 II methionine-free, cystine-free SFM media supplemented with 200 mg/L L-cysteine. The cells were cultured for 8 h, supplemented with 250 mg selenomethionine per liter, and harvested after an additional 36 h of growth. The labeled protein was purified in the same way as the unlabeled protein.
For purification of Utp22 alone, Utp22 was fused with an Nterminal noncleavable His 6 -tag. Utp22 was expressed and purified in the same way as the Utp22/Rrp7 complex. The Rrp7 protein and its fragments were expressed in E. coli. Rrp7 and its fragments were cloned into the plasmid pETDuet-1 and fused to an Nterminal His 6 -tag, the SMT3 protein, and a PreScission cleavage site. The protein was induced for expression in the Rosetta (DE3) strain using 0.1 mM IPTG for 16 h at 16uC. The cells were resuspended in buffer containing 50 mM Tris, pH 8.0, 300 mM NaCl, 5% glycerol, and 30 mM imidazole, which was supplemented with 100 mM phenylmethylsulfonyl fluoride and disrupted using a high-pressure cell disruptor (JNBIO). After clarification, the supernatant was applied to a 5-ml HisTrap column and the protein was eluted with imidazole. The N-terminal His 6 -tag and the SMT3 fusion protein were removed by overnight PreScission digestion at 4uC. The protein was further purified through a heparin column and a gel filtration column equilibrated in 10 mM Tris, pH 7.5, 200 mM NaCl. For NMR study, Rrp7 256-297 was labeled with 15 N and 13 C in M9 minimal medium containing 1 g/ L of ( 15 NH 4 ) 2 SO 4 and 2 g/L of 13 C-glucose (Cambridge Isotope Laboratories).

Crystallization and Structure Determination
Crystals of the Utp22 and Rrp7 complex (6.5 mg/ml in 10 mM Tris, pH 8.0, and 200 mM NaCl) were grown from 100 mM sodium cacodylate pH 6.2-6.5, 30% (w/v) PEG 400, and 200 mM lithium sulfate by hanging drop vapor diffusion method at 20uC and were directly frozen in liquid nitrogen without further cryoprotection. The Se-labeled protein was purified and crystallized in the same way as the native protein. A Se-derivative dataset was collected to 3.0 Å resolution at beamline BL17U of the Shanghai Synchrotron Radiation Facility, processed with HKL2000 [57], and used for SAD phasing in SHARP [58]. After density modification, the electron density map was of sufficient quality to allow automatic model building in ARP/wARP [59]. The model was further adjusted in Coot [60] and refined with PHENIX and refmac [61,62]. A native dataset was collected at Japan SPring-8 beamline BL41XU and used for final refinement at 1.97 Å resolution. The current model contains Utp22 residues 81-274, 282-317, 326-445, 453-983, 1010-1116, and 1128-1237; Rrp7 residues 3-27, 32-105, and 120-189; 764 water molecules; 11 sulfate ions; and three PEG molecules. Analysis with RAMPAGE showed that 98.5% of the residues are in favored regions, 1.4% are in allowed regions, and 0.1% are in outlier regions. Structural figures were prepared using PyMOL [63].

NMR Structure Determination of Rrp7 256-297
The NMR sample contained 1.0 mM 15 N/ 13 C-labeled Rrp7 256-297, 50 mM potassium phosphate (pH 6.0), and 10% (v/v) 2 H 2 O. NMR spectra were measured at 298 K on a Bruker DMX600 spectrometer equipped with a triple resonance cryoprobe. Spectra 1 H- 15  NOESY-HSQC (t m 200 ms) and 3D aliphatic 1 H-13 C NOESY-HSQC (t m 200 ms) spectra were recorded to derive NOE distance restraints. Backbone dihedral angle restraints were calculated by analyzing HN, Ha, Ca, Cb, C9, and N chemical shifts in TALOS+ [65]. The structure was calculated in CYANA and further refined in CNS by incorporating additional dihedral angle restraints [66,67]. The 20 lowest energy structures out of 100 calculated structures were analyzed. EMSA snR5 RNA was in vitro transcribed, dephosphorylated, labeled with 32 P at the 59-end, and column-purified using standard methods. Approximately 0.1 nM labeled RNA was incubated with protein in a 10 ml reaction containing 25 mM HEPES-K (pH 7.6), 100 mM NaCl, 2 mM MgCl 2 , 1 mM DTT, 0.01% NP-40, and 10% glycerol at room temperature for 10 min. The reactions were resolved in 5% native polyacrylamide gels running in 16 Trisglycine (pH 8.3) buffer at room temperature. The gels were dried and autoradiographed using a Typhoon PhosphorImager (GE Healthcare).
Gene cloning was mainly preformed using the non-ligationbased In-fusion (TaKaRa) or Transfer-PCR approaches [68]. Mutagenesis was conducted with QuikChange. All plasmids were verified by DNA sequencing. The strains, primers, and plasmids generated are listed in Tables S4, S5, and S6.
Chromosomal tagging was performed using the one-step PCR strategy. The GAL1 promoter cassette was amplified from plasmid pFA6a-His3MX6-PGAL1 [69]. To generate the RRP7 shuffle strain, the heterozygous deletion diploid rrp7D/RRP7 (Euroscarf) was transformed with a URA3 pRS416 plasmid carrying RRP7 under its endogenous promoter. The transformants were sporulated, and isolated spores were germinated to select for the rrp7D haploid complemented with the URA3 RRP7 plasmid in Uradeficient SC medium containing G418. The UTP22 shuffle strain was generated in a similar manner from the utp22D/UTP22 strain.
To construct a HTP cassette for genomic tagging, the ProtA-TEV-His 7 tag in plasmid pYM9 [70] was modified to remove the original His-tag and incorporate a new His 7 -tag before Protein A, yielding plasmid pYM9-HTP. The RRP7-HTP and UTP22-HTP strains were generated by integrating the HTP cassette into strain BY4741.

Spot Assay
Yeast cells were inoculated into 2 ml of YPDA liquid medium and cultured at 30uC until OD 600 reached 0.6-1.0. The culture was adjusted to OD 600 = 0.6 and serially diluted 10-fold with sterile water. The sample was spotted on plates containing SC medium with or without 0.1% 5-FOA and incubated at 37, 30, and 20uC for 4 d.

Sucrose Gradient Sedimentation
To deplete GAL-driven genes, logarithmically growing cells (OD 600 = 0.6-1.0) cultured in YPGA medium were harvested, washed with water, and re-suspended in YPDA medium. The GAL::SNR30 strain was grown in YPDA medium for 14 h and the GAL::RRP7-HTP strain was grown in YPDA medium for 16 h.
Polysome profile analysis was preformed as previously described [71]. Yeast cells (250-300 ml) were grown to OD 600 = 0.8-1.0 and supplied with 0.1 mg/ml of cycloheximide (Sigma) immediately before harvesting. Pelleted cells were resuspended in 500 ml of lysis buffer (10 mM Tris, pH 7.5, 100 mM NaCl, 30 mM MgCl 2 , 0.1 mg/ml cycloheximide, and 0.2 mg/ml heparin) and lysed by vortexing with acid-washed, baked glass beads. After clarification by centrifugation at 15,000 g for 10 min at 4uC, 350 ml of extracts equivalent to 15-20 OD 260 units were layered onto a 10 ml 7-50% sucrose gradient prepared in 50 mM Tris-acetate, pH 7.5, 50 mM NH 4 Cl, 12 mM MgCl 2 , and 1 mM DTT. Samples were centrifuged in a SW41 Ti rotor (Beckman) at 39,000 rpm at 4uC for 165 min. Gradients were manually fractionated in ,0.5 ml volume using a gradient collector (ISCO). Ribosome profiles were recorded by measuring UV absorbance at 254 nm. Proteins from 20 ml of fractions were separated by SDS-PAGE and analyzed with Western blotting. RNA was extracted from 100 ml of gradient fractions and analyzed with Northern blotting.

Immunoprecipitation
Yeast cells were lysed using glass beads in lysis buffer containing 20 mM Tris-HCl (pH 8.0), 5 mM Mg-acetate, 10 mM NaCl, and 0.2% Triton X-100, supplemented with one tablet of EDTA-free protease inhibitor cocktail (Roche), 0.5 U/ml RNasin (Promega), and 1 mM DTT. After clarification by centrifugation, IgG Sepharose beads (100 ml) were incubated with 100 OD 260 units of supernatant for 2 h and washed seven times with 800 ml of lysis buffer containing 200 mM NaCl. Twenty percent of the beads were used for protein analysis, and the remaining beads were used for RNA extraction.

Yeast Two-Hybrid Assay
Two-hybrid assays were performed using the MATCHMAK-ER GAL4 two-hybrid system (Clontech). Utp22 was cloned into the GAL4 DNA-binding domain (BD) vector pGBKT7 as bait. Rrp7 was cloned into the GAL4 DNA activation domain (AD) vector pGADT7 as prey. The two plasmids were co-transformed into strain AH109, which expresses the HIS3 and ADE2 reporter genes under the control of the GAL4 promoter. The Leu + Trp + transformants were grown in 3 ml of SC medium lacking Leu and Trp overnight at 30uC and adjusted to an OD 600 of 0.6. The cells were 10-fold serially diluted with water and spotted on plates with SC medium lacking Leu and Trp; on SC medium lacking Leu, Trp, and His and containing 5 mM of 3-amino-1,2,4-triazole (3-AT); or on SC medium lacking Leu, Trp, His, and Ade. The plates were incubated for 3 d at 30uC.

CRAC
CRAC experiments were performed as previously described with the following changes [27]. Briefly, yeast cells were grown from 1 L YPDA medium to OD 600 ,0.5 (1.5 g), UV crosslinked in Petri dishes with a Stratalinker (Stratagene), and lysed with glass beads. HTP-tagged proteins were bound to IgG Sepharose beads and eluted after TEV cleavage. The samples were incubated with RNase A/T1 mixture for 10 min at 37uC to partially digest crosslinked RNA. Guanidine-HCl was added to 6 M and the samples were then bound to MagnetHis Ni-Particles (Promega). Crosslinked RNAs were dephosphorylated, ligated to the 39 linker and 59-end 32 P-labeled on beads. The 59-linker was not ligated at this step. Proteins were eluted with imidazole, resolved in Bis-Tris NuPAGE gels (Invitrogen), and blotted onto nitrocellulose membranes. After exposure to X-ray film, the radioactive RNP band was excised, sliced, and incubated with proteinase K. The released RNA was purified by phenol extraction and ethanol precipitation and ligated to the 59 linker. The ligation reaction was resolved in a 20% polyacrylamide/8 M urea gel. The gel band containing radioactive RNA was excised and crushed. RNA was soaked out in 0.4 M NaCl overnight at 4uC, filtered through a Costar Spin-X column (Sigma), and ethanol precipitated. cDNA was synthesized by reverse transcription using the primer DP3 and amplified by PCR (25-35 cycles) using the primers DP3 and DP5 (Table S7). PCR products were resolved in a 10% denaturing polyacrylamide gel, and DNA of expected size was purified using a QIAEX II kit (Qiagen). For Sanger sequencing, DNA was cloned into the pCR4-topo vector. For Solexa sequencing, 2 ml of the first PCR product was PCR-amplified (6-14 cycles) using the primers SBS3 and SBS5 (Table S7). The second PCR product was purified in 3% MetaPhor agarose gels (Lonza) and sent for deep sequencing (Illumina).

Accession Numbers
The atomic coordinates and experimental data have been deposited in the Protein Data Bank under accession codes 4M5D for the Utp22 and Rrp7 complex and 2MBY for Rrp7 256-297. The NMR resonance assignments for Rrp7 256-297 have been deposited in BioMagResBank under accession number 19416. The AfCCA tRNA complex structure is shown in grey. In the active site of AfCCA, a bound ATP is shown as red sticks, a Mg 2+ ion as a green sphere, and three catalytic acidic residues as silver sticks. Utp22 residues that are equivalent to three catalytic residues of AfCCA are shown as sticks and labeled.       Text S1 Alignment of 151 Utp22 sequences. The alignment file is in the fasta format.