The Crystal Structure of the Ubiquitin-like Domain of Ribosome Assembly Factor Ytm1 and Characterization of Its Interaction with the AAA-ATPase Midasin*

The synthesis of eukaryotic ribosomes is a complex, energetically demanding process requiring the aid of numerous non-ribosomal factors, such as the PeBoW complex. The mammalian PeBoW complex, composed of Pes1, Bop1, and WDR12, is essential for the processing of the 32S preribosomal RNA. Previous work in Saccharomyces cerevisiae has shown that release of the homologous proteins in this complex (Nop7, Erb1, and Ytm1, respectively) from preribosomal particles requires Rea1 (midasin or MDN1 in humans), a large dynein-like protein. Midasin contains a C-terminal metal ion-dependent adhesion site (MIDAS) domain that interacts with the N-terminal ubiquitin-like (UBL) domain of Ytm1/WDR12 as well as the UBL domain of Rsa4/Nle1 in a later step in the ribosome maturation pathway. Here we present the crystal structure of the UBL domain of the WDR12 homologue from S. cerevisiae at 1.7 Å resolution and demonstrate that human midasin binds to WDR12 as well as Nle1 through their respective UBL domains. Midasin contains a well conserved extension region upstream of the MIDAS domain required for binding WDR12 and Nle1, and the interaction is dependent upon metal ion coordination because removal of the metal or mutation of residues that coordinate the metal ion diminishes the interaction. Mammalian WDR12 displays prominent nucleolar localization that is dependent upon active ribosomal RNA transcription. Based upon these results, we propose that release of the PeBoW complex and subsequent release of Nle1 by midasin is a well conserved step in the ribosome maturation pathway in both yeast and mammalian cells.

Ribosomes are large macromolecular machines composed of four pieces of ribosomal RNA (rRNA) 2 and 80 (79 in yeast) associated ribosomal proteins (1,2). Ribosomes are responsible for carrying out the synthesis of all proteins within a cell, and the eukaryotic ribosome is composed of two subunits known as the small subunit (40S) and the large subunit (60S). The assembly of ribosomes begins in the nucleolus with the transcription of the rRNAs. Three of the rRNAs are transcribed as a single polycistronic precursor (18S, 5.8S, and 25S), which must then be modified, folded, processed, and exported to the cytoplasm in a carefully orchestrated manner (3,4). Ribosome biogenesis in eukaryotic cells is an incredibly complex and energetically demanding process that requires more than 200 essential nonribosomal assembly factors (3)(4)(5)(6).
Defects in the mammalian ribosome biogenesis pathway are linked to a group of human diseases that are collectively called ribosomopathies. These are all congenital, inherited disorders with a broad clinical spectrum that have perplexed researchers for years because they cause tissue-specific effects although ribosomes are essential in all cell types (7). Ribosome biogenesis has also been emerging as a new target for cancer therapy. Recent studies have shown that several important oncogenes, including cMYC, RAS, and PI3K, play key roles in promoting hyperactive ribosome biogenesis, and deregulated ribosomal DNA transcription is a requirement for the transformed phenotype (8). Moreover, studies have also shown that the nucleolus is a key element in regulating the cellular stress response and is directly involved in regulating the activity of the tumor-suppressor p53 in response to stress (8 -12). Furthering our understanding of the ribosome biogenesis pathway is essential for the development of new cancer therapeutics and treatments for ribosomopathies.
The majority of what is known about eukaryotic ribosome biogenesis is based on extensive studies in the budding yeast Saccharomyces cerevisiae (recently reviewed in Ref. 4). Although the overall process is thought to be well conserved among eukaryotes, much less is known about ribosome biogenesis in higher organisms (14). One complex that has been well characterized in both yeast and mammalian cells is the PeBoW complex (Nop7 complex in S. cerevisiae). The PeBoW complex was named for the mammalian assembly proteins Pes1 (Pescadillo), Bop1 (block of proliferation), and WDR12 (WD repeat domain 12), whereas the yeast homologues are Nop7, Erb1, and Ytm1, respectively. Knockdown of any of the proteins within the mammalian PeBoW complex blocks processing of the large subunit 32S pre-RNA and triggers p53-dependent cell cycle arrest (15,16). Similarly in yeast, Nop7, Erb1, and Ytm1 are all essential proteins required for the processing of the 27SA 3 pre-rRNA (17,18). All components of the PeBoW/Nop7 complex are thought to be multifunctional proteins and have been linked with roles in various cellular processes, including DNA replication, cell cycle regulation, and cardiac function in addition to ribosome biogenesis (19,20). The PeBoW/Nop7 complex is thought to primarily localize to the nucleolus. The complex is not found on ribosomal particles within the nucleoplasm or cytoplasm, indicating that the PeBoW complex must be released from preribosomal particles before their transport out of the nucleolus (16,18,21).
One way in which ribosome maturation factors are released from preribosomal particles is through the aid of three different AAA-ATPases, Rea1/midasin, Rix7, and Drg1/Afg2, which all utilize both nucleotide binding and hydrolysis to drive release of distinct ribosome maturation factors (reviewed in Ref. 22). Rea1 is responsible for driving release of both the Nop7 complex and the maturation factor Rsa4 from pre-60S particles in S. cerevisiae. Rea1 is the largest protein in S. cerevisiae and contains an N-terminal domain, six concatenated ATPase domains, a 260-kDa linker domain, and a C-terminal MIDAS domain that is well conserved across all eukaryotes (23). Electron microscopy studies reveal that Rea1 has a large AAA motor domain connected to a flexible tail-like structure, which contains the MIDAS domain at the end (24). Electron microscopy studies also indicate that Rea1 contacts preribosomal particles adjacent to the Rix1-Ipi1-Ipi3 subcomplex with the tail being able to reach the preribosomal factor Rsa4 (24). Further studies in S. cerevisiae demonstrated that Rea1 and Rsa4 interact with one another in vivo and in vitro through the MIDAS domain of Rea1 and the conserved N-terminal UBL domain of Rsa4, which was previously referred to as the MIDO (MIDASinteracting) domain (24). Interaction between Rea1 and Rsa4 coupled with ATP hydrolysis triggers release of Rsa4 from preribosomal particles, probably through a large scale mechanical conformational change (22,24). The mammalian homologue of Rsa4 is called Notchless (Nle1) and was so named because it is a direct regulator of the Notch signaling pathway in Drosophila melanogaster in addition to its role in ribosome maturation (25).
In addition to driving release of Rsa4, Rea1 has also been shown to drive release of the Nop7 complex from nucleolar preribosomal particles in S. cerevisiae. Bioinformatic analysis revealed that the N-terminal UBL domain of Rsa4 is homologous to the N-terminal UBL domain of Ytm1, suggesting that Rea1 could interact with Ytm1 through its MIDAS domain (21). Earlier work in mice also suggested that the N-terminal domain of WDR12 showed similarity to the N-terminal domain of Nle1, and it was originally referred to in the literature as the Nle1 domain (26). Subsequent experiments in S. cerevisiae demonstrated that Ytm1 interacts with Rea1 through its UBL domain. This interaction is essential for an earlier step in the 60S maturation pathway, and coupled with ATP hydrolysis, it drives release of the Nop7 complex from preribosomal particles (21).
Rea1 is therefore able to act on two different substrates at two different stages of assembly of the large ribosomal subunit, including the nucleolar Nop7 complex and the nucleoplasmic Rsa4 (22).
Here we present the crystal structure of the UBL domain of the S. cerevisiae homologue of WDR12 (Ytm1), which revealed that it is structurally homologous with the UBL domain of Rsa4/Nle1. We demonstrate that human midasin and WDR12 as well as midasin and Nle1 interact through their MIDAS and UBL domains, respectively. The interaction between the MIDAS and UBL domains is reminiscent of integrin receptor ligand binding and is dependent upon coordination of a metal ion. However, the MIDAS domain of midasin also contains a well conserved extension region that we show is required for binding WDR12 and Nle1. We also demonstrate that nucleolar localization of WDR12 in U2OS cells is dependent upon active rRNA transcription and that midasin primarily localizes to the nucleoplasm. Taken together, our data reveal the first piece of structural information for WDR12 and suggest that interactions between midasin with the UBL domains of WDR12 and Nle1 are important evolutionarily conserved interactions in the ribosome assembly pathway.

Experimental Procedures
Reagents-Normal rabbit IgG antibodies against midasin and WDR12 for use in immunohistochemistry were obtained from Sigma. Normal mouse antibodies conjugated with horseradish peroxidase against GST (B-14) and His (H3) for use in Western blots were obtained from Santa Cruz Biotechnology. Monoclonal mouse antibody against the Strep-Tag peptide and normal rabbit antibody against actin for use as primary antibodies in Western blots were obtained from Novagen and Sigma-Aldrich, respectively. Goat anti-rabbit IgG and anti-mouse IgG antibodies conjugated with horseradish peroxidase for use as secondary antibodies in Western blots were obtained from Sigma-Aldrich (anti-rabbit) and Millipore (anti-mouse). Goat anti-rabbit IgG antibodies with Alexa Fluor 633 conjugate for fluorescence was obtained from Molecular Probes, Inc. Normal goat serum was obtained from Santa Cruz Biotechnology, Inc. Actinomycin D, paraformaldehyde, and bovine serum albumin were obtained from Sigma-Aldrich. 4Ј,6-Diamidino-2-phenylindole (DAPI) was purchased from Life Technologies, Inc.
Cloning, Protein Expression, and Protein Purification of the Ytm1 UBL Domain-Residues 1-94 of S. cerevisiae Ytm1 were PCR-amplified from genomic DNA and inserted into the pGST2 parallel vector between the BamHI and NotI restriction sites (27). The Ytm1-UBL plasmid was transformed into BL21(DE3) star cells (Life Technologies). Cells were grown in the presence of 100 g/ml ampicillin in LB broth at 37°C to an A 600 of 0.6, and then protein expression was induced with 0.9 mM isopropyl-␤-D-thiogalactopyranoside for 3 h at 37°C. Cells were harvested and then frozen at Ϫ20°C until further use. Cells were resuspended in lysis buffer (50 mM Tris, pH 8.0, 500 mM NaCl, 10 mM MgCl 2 , 0.01% Triton X-100) containing one EDTA-free protease inhibitor tablet (Roche Applied Science) and lysed by sonication at 4°C. Clarified lysate was applied to a gravity flow column containing 5 ml of GST resin (GE Healthcare) and incubated for 1 h at 4°C. Unbound protein was removed by successive washes with lysis buffer, and the protein was eluted with 20 ml of elution buffer (50 mM glutathione, 50 mM Tris, pH 8, 200 mM NaCl). The sample was then concentrated and run over a Superdex 75 16/600 size exclusion column (GE Healthcare) pre-equilibrated with 20 mM Tris, pH 8, 200 mM NaCl, 0.1 mM tris(2-carboxyethyl)phosphine. Fractions containing the GST-tagged UBL domain were pooled, and then the GST tag was cleaved overnight at 4°C with tobacco etch virus protease. The cleaved sample was then run back over the Superdex 75 16/600 size exclusion column to remove GST and any uncleaved protein. Fractions containing the UBL domain were concentrated to 14.4 mg/ml and used immediately for crystallization. Selenomethionyl (SeMet) incorporation was achieved by expression of the UBL domain in BL21 (DE3) star cells grown in M9 minimal medium that was supplemented with SeMet. The SeMet UBL domain was purified as described above, and incorporation of SeMet was verified by mass spectrometry.
Crystallization of the UBL Domain-The Ytm1 UBL domain was crystallized by hanging drop vapor diffusion using equal volumes of protein and well solution containing 0.1 M sodium cacodylate, pH 6.2-6.8, and 0.9 -1.2 M sodium citrate. Crystals grew as plate clusters in 2-3 days. Crystal quality was improved by several rounds of microseeding with seed beads (Hampton Research) to yield single plate crystals with dimensions up to 200 ϫ 200 ϫ 50 m. Crystals were cryoprotected by the stepwise addition of ethylene glycol to a final concentration of 35% (v/v) and then flash-frozen in liquid nitrogen.
Data Collection and Structure Determination-Native x-ray diffraction data were collected at the Advanced Photon Source on the SER-CAT 22-ID beamline (Chicago, IL), and SeMet diffraction data were collected at the Advanced Photon Source on the SER-CAT 22-BM beamline (Chicago, IL). The crystals are in the P1 space group with two copies in the asymmetric unit. The structure was initially solved by molecular replacement using a homology model derived from the structure of the UBL domain of Rsa4 (Protein Data Bank entry 4WJS) in the program MRage (28). Bucanneer (29) was used to build the model, followed by several rounds of manual model building with COOT and refinement with Phenix (28). To confirm that the assignment of the backbone was correct, we used the molecular replacement phases to calculate an anomalous difference map from the SeMet data set, which confirmed the correct placement of the two Met residues. The structure was refined to a final R work /R free of 17.5%/23.2%.
Yeast Growth Assays-The S. cerevisiae strains used in the study are listed in Table 2. The tetO 7 parental strains were obtained from Open Biosystems. Ytm1 together with its endogenous promoter were PCR-amplified from S. cerevisiae genomic DNA and cloned into the centromeric plasmid YCplac111 (30). Mutations to Ytm1 were generated using QuikChange mutagenesis, and all mutations were verified by sequencing. To carry out the dilution plating assays, cells were grown in YPD and then diluted to an A 600 of 0.05. 1:10 serial dilutions were then plated on YPD plates with and without 10 g/ml doxycycline and incubated at 20, 30, and 37°C for 3 days.

Mammalian Transfections and GST Pull-down Assays-
Full-length human Bop1 and Pes1 were amplified from cDNAs (Open Biosystems) and cloned into the pLexM vector (31) containing either an N-terminal His 12 tag or N-terminal GST tag. The DNA encoding for residues 5287-5596, which includes the MIDAS domain of human midasin, was codon-optimized for expression in HEK293 cells and cloned into the pLexM vector encoding an N-terminal GST tag and the pCAG-OSF vector (32) encoding an N-terminal OSF (One-Strep-FLAG) tag. Fulllength human WDR12 and Nle1 were cloned into the pCAG-OSF vector encoding an N-terminal OSF tag. To create WDR12⌬UBL, residues 84 -423 of WDR12 were PCR-amplified from cDNA encoding WDR12 and cloned into the pCAG-OSF vector. To create Nle1⌬UBL, residues 100 -485 of Nle1 were PCR-amplified from cDNA encoding Nle1 and cloned into the pCAG-OSF vector. Mutations were introduced using QuikChange mutagenesis (Stratagene). 40-ml cultures of HEK293 cells adapted to grow in suspension were transfected with 1 g/ml purified plasmid DNA and 2 g/ml PEI. Cells were harvested 72 h after transfection, washed with PBS, and frozen at Ϫ80°C until further use. Cells were lysed by resuspension in lysis buffer (50 mM Tris, pH 7.4, 500 mM NaCl, 10 mM MgCl 2 , 10% (v/v) glycerol, 1% (v/v) Triton X-100, EDTA-free protease inhibitors (Roche Applied Science), and 10 units of Benzonase (Sigma)) with gentle rocking at 4°C for 1 h. Lysates were clarified by centrifugation and then incubated with 100 l of glutathione resin (GE Healthcare) for 1 h at 4°C. Unbound protein was removed by three 200-l washes with lysis buffer, followed by two 200-l washes with lysis buffer, including 5 mM ATP, and followed by a final 200-l wash with lysis buffer. Protein that was retained on the resin was analyzed by both SDS-PAGE and Western blotting.
Cell Cultures-U2OS cells were cultured in Dulbecco's modified Eagle's medium high glucose supplemented with 10% (v/v) fetal bovine serum, antibiotics (100 units/ml penicillin and 100 g/ml streptomycin), and 2 mM glutamine. Cells were incubated at 37°C in a CO 2 incubator.
Immunohistochemistry/Fluorescence-U2OS cells were cultured on 35-mm glass bottom culture dishes (MatTek Corp.) until 80 -100% confluence was reached. Medium was discarded; cells were rinsed with 1ϫ PBS, pH 7.4; and cells were fixed in 3-4% paraformaldehyde solubilized in 1ϫ PBS for 1 h. Paraformaldehyde solution was removed, the cells were rinsed with 1ϫ PBS, and cells were permeablized with 0.5% Triton X-100, 1ϫ PBS (Sigma-Aldrich) for 1.5 h. Cells were then washed three times (10 min each wash) with 1ϫ PBS. Cells were then incubated for a minimum of 1.5 h with blocking buffer (1% bovine serum albumin, 4% normal goat serum, and 0.4% Triton X-100 in 1ϫ PBS). Blocking buffer was then removed, and cells were incubated overnight at 4°C with a primary antibody from Sigma at the manufacturer's recommended dilutions. Cells were then washed three times (10 min each wash) with 1ϫ PBS and then protected from light and incubated with the Alexa Fluor secondary antibodies for 1 h at the manufacturer's recommended dilutions. Cells were then washed again with 1ϫ PBS, as described previously. Cells were finally incubated with 300 nM DAPI, washed once with 1ϫ PBS, and then covered with 1ϫ PBS and visualized with a Zeiss Axio Observer Z1 Epifluorescence Microscope with a Zeiss EC Plan-Neofluar 40ϫ/0.9 objective and aperture lens. The Alexa Fluor 633 conjugate fluorochrome was imaged at room temperature on the Zeiss AxioCam MRm camera with Zen Pro 2012 (Blue Edition) software and processed with ImageJ.

Results
Structure of the UBL Domain of S. cerevisiae Ytm1-To further our structural knowledge of Ytm1/WDR12 (Fig. 1A), we determined a crystal structure of the UBL domain (density was visible for residues 5-91) of Ytm1 from S. cerevisiae at 1.7 Å resolution (Fig. 1B). The structure was solved by molecular replacement with the recently published UBL domain of Rsa4 (Protein Data Bank entry 4WJS (33)) as a search model, and the backbone assignment was confirmed by calculation of an anomalous difference map from a selenium methionine derivative. The structure was refined to an R work /R free of 17.5%/ 23.2% (see Fig. 1C for a representative section of unbiased electron density; for further data collection and crystallography statistics, see Table 1). There are two nearly identical copies of the domain in the asymmetric unit, but there is no evidence to suggest that the domain forms a dimer in solution or in vivo. The structure revealed that the UBL domain of Ytm1 has a classic, compact ubiquitin ␤-grasp fold, composed of four ␤-strands that form a central ␤-sheet and two ␣-helices located on one side of the ␤-sheet (Fig. 1B).
Despite little sequence homology between the UBL domains of Rsa4 and Ytm1 (Fig. 2B), the structures of the two domains superimpose well with one another with a root mean square deviation of 4.5 Å over 80 C␣ residues ( Fig. 2A). Both Rsa4 and Ytm1 contain well conserved glutamate residues that are thought to coordinate the metal ion with the Rea1 MIDAS domain in S. cerevisiae (21,24). The conserved glutamate residue (residue Glu 80 in S. cerevisiae) in Ytm1, important for binding the Rea1 MIDAS domain lies in an extended loop before the final ␤-strand in the same structural location as the conserved glutamate in Rsa4 (Figs. 1B and 2A). A search of the Dali server revealed that the Ytm1 UBL domain is most similar to the autophagy-associated ubiquitin-like protein, Atg12, and the E3 ubiquitin-protein ligase, RING2, highlighting the range of diversity of biological activity of UBL domains. The Ytm1 UBL domain also superimposes reasonably well with ubiquitin (Protein Data Bank entry 1UBQ) with a root mean square deviation of 5.5 Å over 71 C␣ residues (Fig. 2C). Overall, given the structural similarity between Rsa4 and Ytm1 along with the overlapping glutamate residues, Rea1 undoubtedly interacts with both proteins in an analogous fashion.
Mutations in the UBL Domain of Ytm1 Cause Growth Defects in Yeast-The UBL domain of Ytm1 is essential for viability in S. cerevisiae and contains many well conserved residues (Fig.

Interactions between Mammalian WDR12 and Midasin
2B). It has been previously shown in S. cerevisiae that truncation of the UBL domain or mutation of residue E80A in yeast is lethal (18,21). However, there are many other well conserved residues within the UBL domain. To determine which of these residues were essential for cell viability, we made mutations to the UBL domain and tested the effects on growth. We transformed plasmids encoding wild type or mutant Ytm1 into cells harboring endogenous Ytm1 under a Tet promoter (Hughes Tet-Promoter Collection, Open Biosystems) and assayed for growth in the presence and absence of 10 g of doxycycline at 20, 30, and 37°C (see Table 2 for a list of yeast strains used in this study). The addition of doxycycline turns off expression of endogenous Ytm1. The E80A mutation was used as a positive control, and as expected, both depletion of endogenous Ytm1 and expression of the E80A mutant conferred a severe growth defect relative to growth of the plasmid expressing wild type Ytm1 (Fig. 2D). Surprisingly, mutation of the well conserved residues Arg 65 and Glu 86 to alanine had no effect on growth of yeast, indicating that they are not essential for function (Fig. 2D). Given the high structural similarity between the UBL domain of Ytm1 and ubiquitin, we hypothesized that Ytm1 may contain an essential hydrophobic patch required for function. It has been well established in ubiquitin that residue Ile 44 forms part of an essential hydrophobic patch along with residues Leu 8 and Val 70 that is important for maintaining interactions with ␣-helical ubiquitin-associated domains, such as S5a of the 26S proteasome (34). There are two well conserved hydrophobic residues in Ytm1, Phe 57 and Leu 58 , that superimpose near Ile 44 in ubiquitin (Fig. 2C). Individual mutation of Phe 57 or Leu 58 to
Midasin Binds the PeBoW Complex through the UBL Domain of WDR12-After demonstrating that WDR12 binds to midasin through its UBL domain, we wanted to determine how midasin interacts with the PeBoW complex. First we transfected HEK293 cells with plasmids harboring full-length GST-tagged Bop1 as bait, full-length Pes1, and full-length WDR12 or WDR12⌬UBL, to determine what effect the UBL domain of WDR12 has on formation of the PeBoW complex. We assayed for binding by GST pull-down followed by Western blot analy- sis, which revealed that the UBL domain of WDR12 is not required for formation of the PeBoW complex (Fig. 3C). We do observe differences in the amounts of WDR12 that is pulled down. However, if we normalize everything to the amount of Bop1, which was used as bait, the levels of Pes1 and WDR12 are fairly uniform. All of our pull-downs were carried out in cells that were transiently transfected with multiple plasmids, so we attribute these differences to levels of transfection efficiency.
After we confirmed that the UBL domain of WDR12 is not required for formation of the PeBoW complex, we looked at binding to midasin. HEK293 cells were transfected with plasmids harboring full-length GST-tagged Bop1 as bait, full-length Pes1, full-length WDR12 or WDR12⌬UBL, and residues 5287-5596 of midasin. GST-tagged Bop1 was able to pull down the midasin MIDAS domain only in the presence of full-length WDR12 (Fig. 3C). Therefore, midasin can interact with WDR12 alone and within the PeBoW complex; however, both interactions are dependent upon the UBL domain of WDR12.
Human Midasin Binds Nle1 through Its UBL Domains-We further demonstrated that human Nle1 also binds to midasin through its UBL domain. Because we showed that the UBL domains of the yeast homologues of Nle1 (Rsa4) and WDR12 (Ytm1) are structurally homologous ( Fig. 2A), we assumed that Nle1 would also bind midasin though its UBL domain in mammalian cells. Nle1 has previously been shown to bind to midasin/Rea1 in several organisms, including S. cerevisiae and Solanum chacoense (24,36). To determine whether mammalian midasin and Nle1 are binding partners, we transfected HEK293 cells with plasmids encoding residues 5287-5596 of midasin and Nle1 with or without the UBL domain and assayed for binding by GST pull-downs. GST-tagged residues 5287-5596 of midasin were able to pull down full-length Nle1 but not Nle1⌬UBL (Fig. 3D). Expression of both Nle1 and Nle1⌬UBL was confirmed by Western blot analysis, and full-length Nle1 does not interact with GST-bound glutathione resin in the absence of midasin (Fig. 3B). Our results confirm that, as in yeast, mammalian midasin is able to bind both WDR12 and Nle1 through their respective UBL domains.
The Extension Region of the Midasin MIDAS Domain Is Required to Bind WDR12 and Nle1-After confirming that midasin binds to both WDR12 and Nle1, we mapped the binding interface between the proteins and found that midasin contains a well conserved region required for binding WDR12 and Nle1. Just upstream of the MIDAS consensus motif, there is a region that is highly conserved (46% invariant) in midasin (Fig.  4A) and is predicted to be ␣-helical but shares no significant homology with other known proteins (23). This will be referred to as the MIDAS extension region for the remainder of this paper. We hypothesized that, given its high sequence conservation, the MIDAS extension region must play an important role in midasin function and may be important for binding WDR12 and Nle1. To test this hypothesis, a series of N-terminal truncations to the MIDAS extension region were created and then assayed for binding to WDR12 and Nle1 by GST pull-down. Starting with the midasin truncation 5310 -5596, we see a diminished capacity for WDR12 binding to midasin (Fig. 4B). N-terminal truncations past residue 5320 were unable to pull down WDR12, indicating that the entire MIDAS extension region is required for binding WDR12 (Fig. 4B). Given that the extension region is not part of the canonical MIDAS consensus sequence, the interaction between midasin and WDR12 must require more than the typical MIDAS/ligand metal ion interface. We repeated the same experiment with Nle1 as well and obtained similar results (Fig. 4C), indicating the importance of the midasin MIDAS extension region for binding both Nle1 and WDR12.
Glu 78 of WDR12 Is Required for Binding Midasin-We mapped the interface between midasin and WDR12 and confirmed that the conserved glutamate, residue Glu 78 of WDR12 (Glu 80 in S. cerevisiae Ytm1), is required for binding midasin. Mutation of either the conserved aspartic acid residues in the MIDAS consensus motif or the conserved glutamate in the UBL domains of Ytm1 and Rsa4 abolishes the interactions between midasin and Ytm1/Rsa4 in yeast (21,24). To determine whether this also holds true in mammalian cells, we mutated the corresponding glutamate in WDR12 (Glu 78 ) to alanine and found that the E78A mutation was unable to bind the midasin MIDAS domain (Fig. 5A). Mutation of S78L in yeast Ytm1 has been shown to be synthetic lethal when combined with a mutant of Rea1 (21). This serine residue is conserved in mammalian WDR12 (Fig. 2B), and mutation of Ser 76 to leucine significantly reduced binding to the midasin MIDAS domain (Fig. 5A). Mutation of Ser 76 to a glutamate residue, however, did not affect binding to midasin (Fig. 5A). Sequence alignments of WDR12 across numerous species (Fig. 2B) revealed that higher organisms have a small insertion in the loop following the first ␣-helix, and deletion of residues 47-50 of mammalian WDR12 also significantly reduced binding to the midasin MIDAS domain (Fig. 5A). We also mutated the conserved aspartic acid residues within the MIDAS consensus motif in midasin but were unable to determine whether they abolish WDR12 binding because we could not detect expression of the midasin-MIDAS double aspartic acid mutant when transfected in HEK293 cells.
Metal Ion Is Required for Midasin-MIDAS and WDR12 Binding-After verifying that the residues thought to be important for coordinating the metal ion between midasin and WDR12 are important for binding, we demonstrated that complex formation is dependent upon the presence of a metal ion. MIDAS domains have been extensively studied in integrins, where they play important roles in mediating ligand binding through coordination of a metal ion that is essential for ligand binding (22,37,38). We generated a homology model for the WDR12 UBL domain and midasin MIDAS domain interaction using our Ytm1 UBL structure for WDR12 and the ␣L␤2 integrin (Protein Data Bank entry 1T0P) MIDAS domain structure for midasin (Fig. 5B).
In the crystal structure of the ␣L␤2 integrin bound to its ICAM ligand, the metal bound between the two proteins is a magnesium ion. The midasin MIDAS domain contains the full canonical MIDAS consensus sequence (hhhhDXSXS, ϳ70 residues T, and an hhhh(S/T)DG in ϳ30 residues, where h is any hydrophobic residue and X is any residue (Fig. 4A)), as is found in the ␣L␤2 integrin (39). WDR12 and Nle1 also contain the conserved glutamic acid residue like the ICAM ligand. Thus, all six residues that coordinate the metal ion are conserved, so we believe that the metal ion coordination will be the same as the ␣L␤2 integrin bound to its ICAM ligand. Missing from our homology model is the MIDAS extension region because we have no structural information for this region. To determine the importance of the metal ion for WDR12 binding to the midasin MIDAS domain, we transfected HEK293 cells with plasmids containing human midasin MIDAS and human WDR12. 72 h after transfection, the cells were harvested, split in half, and then lysed in the presence or absence of 10 mM EDTA to ablate metal coordination. EDTA greatly diminishes the amount of WDR12 that can be pulled down by GST-tagged midasin-MIDAS (Fig. 5C). This result confirms that a metal ion is required to promote complex formation between the MIDAS domain of midasin and WDR12.
WDR12 but Not Midasin Localizes to the Nucleolus in U2OS Cells-To further characterize mammalian WDR12 and midasin, we carried out immunofluorescence experiments in U2OS osteoscarcoma cells with specific antibodies for WDR12 and midasin. Endogenous WDR12 showed prominent nucleolar localization, as expected (15), whereas endogenous midasin localized to both the nucleoplasm and the cytoplasm (Fig. 6).
Interestingly, midasin also appears to have prominent localization in a distinct location in the cytoplasm, suggesting that it may have additional non-ribosomal roles in the cell (Fig. 6). The localization of endogenous WDR12 to the nucleolus is consistent with it playing a role in the early stages of ribosome biogenesis that take place in the nucleolus, whereas the localization of midasin in the nucleoplasm is consistent with its role in the later stages of ribosome biogenesis (40).
Next, we wanted to determine whether the localization of WDR12 and midasin was affected by cellular stress and rRNA transcription. Disruption of rRNA transcription by various cellular stressors can cause a relocalization of proteins involved in ribosome biogenesis (41). For example, Pes1, one of the other proteins associated with WDR12 in the PeBoW complex, relocalizes from the nucleolus to the nucleoplasm upon induction of cellular stress (42). To mimic cellular stress, we treated U2OS cells with a low concentration (10 M) of actinomycin D to inhibit rRNA transcription by RNA polymerase I, and we repeated our immunofluorescence experiments using DMSO as a control. WDR12 loses its nucleolar localization upon treatment with actinomycin D but not DMSO alone. It is instead . The predicted secondary structure of the MIDAS domain based on secondary structure alignments in HHPRED is shown below the sequence, and the numbers at the bottom of the first row correspond to the H. sapiens midasin sequence. The metal ion consensus sequence in MIDAS domains is composed of the following: hhhhDXSXS, followed by a conserved threonine in ϳ70 residues and an hhhh(S/T)DG in ϳ30 residues, where h is any hydrophobic and X is any residue (23,37). MIDAS consensus sequence residues are indicated above the sequence. B, Western blot analysis of GST pull-downs of midasin MIDAS N-terminal truncations with WDR12. The constructs used contained the following residues of midasin fused to an N-terminal GST tag: 5287 (residues 5287-5596), 5300 (residues 5300 -5596), 5310 (residues 5310 -5596), 5320 (residues 5320 -5596), and 5530 (residues 5530 -5596). C, Western blot analysis of GST pull-downs of midasin MIDAS truncations with Nle1. WCL, whole cell lysate.
found in the nucleoplasm, confirming that cellular stress affects the localization of WDR12 (Fig. 6). In contrast, the nucleoplasmic localization of midasin did not change upon treatment of cells with actinomycin D or DMSO (Fig. 6).

Discussion
The assembly of eukaryotic ribosomes requires two essential proteins, Ytm1/WDR12 and Rsa4/Nle1, that both harbor UBL domains followed by C-terminal WD40 ␤-propeller domains. Using x-ray crystallography, we determined the crystal structure of the UBL domain from the yeast homologue of Ytm1/ WDR12. UBL domains mimic the properties of ubiquitin through incorporation of the ubiquitin ␤-grasp fold within their protein coding regions. They are used in a wide variety of cellular processes, such as protein degradation, DNA repair, cell division, and autophagy, and are often part of protein-protein interaction motifs (34,43,44). WDR12 and Nle1 are the only proteins known to contain a UBL domain in the ribosome assembly pathway. The function of this domain is to recruit the tail of the large motor protein midasin and drive release of both the PeBoW complex and Nle1 from preribosomal particles, presumably through factor-relay mechanisms that cause conformation changes within the pre-rRNA (33). The recent crystal structure of the yeast homologue of Rsa4/Nle1 contained FIGURE 5. Metal ion is required for WDR12 and midasin binding. A, Western blot analysis of GST pull-downs with WDR12 UBL mutations. B, model of the interaction between midasin (blue) and the UBL domain of WDR12 (purple). The homology model of midasin was generated from a crystal structure of the ␣L␤2 integrin (Protein Data Bank entry 1T0P). Residues important for coordinating the metal ion (red) are shown as stick representations. C, GST pull-downs of midasin and WDR12 with and without EDTA. Cells were lysed in the presence or absence of EDTA, and the binding was assayed by GST pull-downs followed by SDS-PAGE analysis (Simply Blue Stain (Thermo)). WCL, whole cell lysate. two copies of the protein in the asymmetric unit, and interestingly, the UBL domain in each had a different orientation to the ␤-propeller, suggesting that the rotational flexibility of the UBL domain could be important for function (33). During the revision of this manuscript, another group published the crystal structure of Chaetomium thermophilium Ytm1 bound to the C-terminal end of Erb1. As was observed for Rsa4, they also found that the UBL domain of Ytm1 was found in different orientations, emphasizing the flexible orientation of this domain with respect to the ␤-propeller domain (45). Given the structural similarity of the UBL domain of Ytm1/WDR12 presented here and Rsa4/Nle1 ( Fig. 2A), including the overlapping glutamate residues, the interactions between the midasin MIDAS domain and the UBL domains of WDR12 and Nle1 should be very similar. This is reminiscent of integrins, which can interact with a range of different ligands through their MIDAS domains (46).
Whereas WDR12 and Nle1 are well conserved, little is known about their roles in mammalian ribosome biogenesis. In this study, we present the first evidence that interactions with midasin and WDR12 as well as midasin and Nle1 are conserved in higher organisms. As in yeast, the UBL domain of WDR12 is required for binding midasin but not for formation of the PeBoW complex along with Pes1 and Bop1 (Fig. 3C). The UBL domain of Nle1 is also required for binding midasin (Fig. 3D). These results suggest that release of the nucleolar PeBoW complex and the nuclecoplasmic Nle1 by the large motor protein, midasin, is a well conserved step in the eukaryotic ribosome maturation pathway.
Roles of WDR12 and midasin in mammalian ribosome biogenesis are also further supported by their endogenous localization. WDR12 displays prominent nucleolar localization; however, cellular stress induced by treatment of cells with actinomycin D halts rRNA transcription and leads to a relocalization of WDR12 in the nucleoplasm (Fig. 6). Human midasin did not display nucleolar localization but rather was found in the nucleoplasm and cytoplasm under both normal and stress conditions (Fig. 6). The localization of human midasin is similar to that of its yeast homologue, which is not found in the nucleolus except in the presence of the E80A glutamate mutant of Ytm1 that halts ribosome maturation (21). How midasin gets recruited to preribosomal particles in the nucleolus is still unknown. Previous work has suggested that WDR12 and the Rix complex are the two most likely candidates for recruitment, although other trans-acting factors could be involved (21). Collectively, these results suggest that midasin is only transiently associated with the nucleolus to drive release of the PeBoW complex.
We also identified that the MIDAS extension region in mammalian midasin is required for binding WDR12 and Nle1 (Fig.  4). The extension region is not part of the canonical MIDAS domain but is highly conserved and essential for binding WDR12 and Nle1, suggesting that it may also play an important role in metal ion coordination. Integrins are dependent upon metal ions for their function, and their dependence upon metal ions has been well studied (38). Some integrin domains contain metal ion-binding sites in addition to the MIDAS domain. For example, the ␤I integrin domain is made up of three interlink-ing metal ion-binding sites, including the MIDAS site, adjacent to MIDAS (AMIDAS) site, and synergistic metal ion-binding site (SyMBS). The exact in vivo roles of all three sites are unclear, but in vitro all three sites appear to play important roles in mediating ligand binding (38). There is no sequence similarity between the MIDAS extension region and the SyMBS or AMIDAS metal ion-binding sites, but the MIDAS extension region could represent a new type of metal ion coordination site. Bioinformatic analysis suggested that there is a weak similarity between the MIDAS extension region and the D subunit of magnesium chelatases based upon the high density of basic and hydrophobic residues just upstream of the MIDAS domain (23). Magnesium chelatases, which play an important role in chlorophyll biosynthesis by inserting Mg 2ϩ into protoporphyrin, also contain C-terminal MIDAS domains and two N-terminal AAA domains (47). In addition to magnesium chelatases and midasin, there is only one other known protein, VWA8 (only found in higher organisms; function is unknown), that contains the combination of AAA and MIDAS domains (37). Further work will be needed to determine the exact role of the MIDAS extension region in midasin, but it will be interesting to see whether there is a functional relationship between the AAA domains and the MIDAS domain because this combination also occurs in two other proteins.
In this paper, we present the first high resolution structure of the UBL domain of the yeast homologue of Ytm1/WDR12 and demonstrate that mammalian midasin binds to both WDR12 and Nle1, two proteins that have interesting implications for human health. Recent work has demonstrated that overexpression of WDR12 is induced by cardiac overload, and genome association studies have associated the WDR12 gene with early onset myocardial infarction and coronary artery disease (20,48). Up-regulation of WDR12 has also been shown to lead to activation of the p38 MAPK pathway, an increase in levels of Bop1, and ultimately myocardial dysfunction, suggesting that WDR12 could be a novel therapeutic target for patients with failing hearts (20). Moreover, elevated levels of the proteins associated with the PeBoW complex have been found in many different types of human cancers, including neuroblastoma, hepatocellular carcinoma, and breast, colon, and ovarian cancers (42, 49 -55). Nle1 was also recently shown to be a crucial factor for intestinal homeostasis and a requirement for organogenesis, especially in axial skeleton formation in mice (13,56). Further investigations into the roles of WDR12 and Nle1 are needed to determine the mechanisms by which they are involved in cardiac function and intestinal homeostasis, but our work identifies that the interactions between the UBL domains of WDR12 and Nle1 with midasin are both evolutionarily conserved interactions important for the synthesis of eukaryotic ribosomes.