Solution Structure and Rpn1 Interaction of the UBL Domain of Human RNA Polymerase II C-Terminal Domain Phosphatase

The ubiquitin-like modifier (UBL) domain of ubiquitin-like domain proteins (UDPs) interacts specifically with subunits of the 26 S proteasome. A novel UDP, ubiquitin-like domain-containing C-terminal domain phosphatase (UBLCP1), has been identified as an interacting partner of the 26 S proteasome. We determined the high-resolution solution structure of the UBL domain of human UBLCP1 by nuclear magnetic resonance spectroscopy. The UBL domain of hUBLCP1 has a unique β-strand (β3) and β3-α2 loop, instead of the canonical β4 observed in other UBL domains. The molecular topology and secondary structures are different from those of known UBL domains including that of fly UBLCP1. Data from backbone dynamics shows that the β3-α2 loop is relatively rigid although it might have intrinsic dynamic profile. The positively charged residues of the β3-α2 loop are involved in interacting with the C-terminal leucine-rich repeat-like domain of Rpn1.


Introduction
An enzymatic cascade composed of the enzymes E1, E2, and E3 conjugates ubiquitin to target proteins, followed by translocation to the 26 S proteasome, where ubiquitinated protein is removed by proteolysis [1]. Despite low sequence homology, ubiquitin-like proteins are classified as ubiquitin-like modifiers (UBLs) and ubiquitin-like domain proteins (UDPs) [2]. UBLs, such as NEDD8, SUMO, and FAT10, modify target proteins in a manner similar to ubiquitinylation [3]. UDPs, including Rad23, the human homolog of Rad23 (HHR23), Dsk2, ubiquilins 1-4 (human homologs of Dsk2), and parkin, bind to the 26 S proteasome in a UBL domain-dependent manner [4,5]. The UBL domain of UDPs interacts specifically with subunits of the 26 S proteasome. Rad23 and Dsk2 preferentially interact with the Rpn1/S2 subunit. UDPs have been implicated in neurodegenerative diseases caused by dysfunction of the ubiquitin proteasome system [6]. Recently, a novel UDP protein, ubiquitin-like domaincontaining C-terminal domain phosphatase (UBLCP1) has been identified [7]. UBLCP1 consists of two independent domains, UBL and a phosphatase domain. It has been proposed that UBLCP1 directly interacts with Rpn1 of the 26 S proteasome through the UBL domain and that it serves as a transcription regulator via the phosphatase domain [8,9,10]. UBLCP1 also inhibits proteasome activity, dephosporylating the 26 S proteasome [10]. The crystal structure of Drosophila UBLCP1 shows that the UBL domain and the phosphatase domain are connected by a flexible linker and that the UBL domain has a b-grasp fold with four b-strands and two a-helices [10]. Because UBLCP1 has diverse functions related to regulation of the phosphorylation state of the 26 S proteasome, it is of the essence to understand the detailed interactions between UBLCP1 and Rpn1 of the proteasome component. We determined the solution structure of the UBL domain of human UBLCP1 (hUBLCP1). Structural information indicated detailed intermolecular interactions between the UBL domain and Rpn1 on the atomic scale. Nuclear magnetic resonance (NMR) data revealed that the secondary structure of the UBL domain of hUBLCP1 differs from that of known UBL domains, especially b3 and b4. In addition, the structure of the UBL domain of hUBLCP1 is dramatically different from that of the fly UBLCP1 [10]. Interestingly, the positively charged clamp formed by the unique b3-a2 loop of the UBL domain of hUBLCP1 serves as a key motif in the interaction with Rpn1.

Materials and Methods
Cloning, Expression, and Purification of Human UBLCP1 and Rpn1 Polymerase chain reaction (PCR) products of both hUBLCP1 and Rpn1 were amplified from a Homo sapiens cDNA library. All sense primers encoded the recognition site of tobacco etch virus protease (ENLYFQG) for the clearance of affinity tags. For both the UBLCP1 and UBL domain (UBLCP1 1-81 ), sense primers and antisense primers were designed as previously described [8], and amplified PCR products were ligated into the pGEX 4T-1 vector (Amersham Pharmacia Biotech, Uppsala, Sweden). The sense and antisense primers for three Rpn1 constructs, Rpn1 1-908 , Rpn1 394-568 , and Rpn1 640-772 incorporated BamHI and XhoI restriction enzyme sites, and PCR products were ligated into the pET32a vector. The plasmids were transformed into E. coli BL21 (DE3) cells for overexpression. The transformed cells were induced by 0.1 mM isopropyl b-D-thiogalactopyranoside at an OD 600 of 0.6. All purified proteins were applied to a size exclusion chromatography column (Amersham Pharmacia Biotech) to improve protein purity and exchange the buffer solution.

NMR Spectroscopy and Structure Calculations
To obtain the 13 C/ 15 N-and 15 N-labeled UBL domains of hUBLCP1, cells were cultured in M9 media containing 99% 15 NH 4 Cl or 99% 15 NH 4 Cl and 99% 13 C-D-glucose (Cambridge Isotope Inc., Andover, MA, USA). Purified UBL domain was concentrated to 1.5 mM with an Amicon Ultra-15 concentration device (Millipore, Bedford, MA, USA). All NMR experiments were performed on Bruker DRX 500 MHz and Bruker Avance 900 MHz spectrometers equipped with a cryoprobe at 25uC. The 15 N-edited two-dimensional (2D) HSQC, HNCACB, CBCA(-CO)NH, HNCA, HNCO, HBHA(CO)NH, and HCCH-TOCSY experiments were performed for resonance assignments of backbone and side-chain atoms [11]. To obtain nuclear Overhauser effect (NOE) constraints for structure calculations, 15 Nedited and 13 C-edited three-dimensional (3D) NOESY experiments with two mixing times (t = 100 and 150 ms) were performed. All NMR data were processed using NMRPipe [12] and analyzed using the SPARKY program. The in-phase-antiphase (IPAP) experiment for residual dipolar coupling (RDC) measurements was performed using polyacrylamide gels prepared as described by Sass et al. [13]. The 6% polyacrylamide gels were made in a 5-mm inner-diameter tube, and dried after 3 days of dialysis. The 15 N-labeled UBLCP1 1-81 (300 ml) was incorporated in the dried gel and placed in a Shigemi NMR tube, and the gel was compressed by the plunger. The 1 D NH dipolar coupling constants in both isotropic and anisotropic media were measured using the 2D IPAP experiments [14] on the Bruker DRX 500 spectrometer. The RDC constants were calculated from the difference of the 1 J NH splitting in isotropic and anisotropic media. The alignment tensor was calculated using REDCAT software [15].
Structure calculations were performed using the CYANA 2.1 program [18] installed on a 16-node Linux cluster computer. The alignment tensor was used during the structure refinement procedure. From the RDC, Da and R values were calculated as 1.44860.015 and 0.62260.016, respectively. The value of the Qfactor after back calculations was determined as 0.34. A total of 2080 NOE constraints, consisting of 1009 short-range NOEs (|i-j| #1), 365 medium-range NOEs (1, |i-j| ,5), 706 long-range NOEs (|i-j| $5), and 100 angle constraints were used for structure calculations [19]. The PROCHECK program was used for structure evaluation [20] and the MOLMOL, NACCESS, and Pymol programs were used for structural analyses [21].
Isothermal Titration Calorimetry VP-ITC system (MicroCal) was used for ITC experiments of UBL domain and Rpn1 regulatory subunit 1 at 25uC in NMR buffer. In each titration, 20 mM Rpn1 regulatory subunit 1 in the cell was titrated with 25 injections of 400 mM UBL domain. Each injection was 6 mL. The resulting data were fitted to a one-site binding isotherm using Microcal Origin program for ITC data analysis.

NMR Structures of the UBL Domain of hUBLCP1
All backbone resonance assignments were completed with data from HNCA, CBCACONH, and HNCACB experiments (Fig. 1A). Most of the side chain assignments were made based on 3D HCCH-TOCSY and 15 N-edited TOCSY-HSQC experiments. Secondary structures were determined from the chemical shift indices (CSIs), NOEs, and 3 J HNa coupling constant values.
The UBL domain of hUBLCP1 consisted of four short bstrands (b1, b2, b3, and b4) and two a-helices (a1 and a2) (Fig. 1B  and Fig. S1). A total of 50 distance geometry structures were used as starting conformers for dynamical simulated annealing calculations. The structures were further refined using the calculated alignment tensor from the IPAP experiment. The 20 lowest-energy structures were selected for the final analysis ( Table 1). The final structures were well converged, with a root-mean-square deviation (RMSD) of 0.4460.13 Å and 0.7460.11 Å for all backbone and heavy atoms, respectively (Fig. 1C). Especially, a unique b3-a2 loop was well defined by a number of intra-loop NOEs (Fig. 1D and Fig. S2) observed in that region. The average structure was calculated from the geometrical average of the 20 final structures and subjected to restraint energy minimization to correct covalent bonds and angle distortions (Fig. 1D).

Structural Comparison with other UBL Domains
The molecular topology of the UBL domain of hUBLCP1 was quite different from that of the canonical UBL domain (Fig. 1B), especially in the organization of the secondary structures (b3 and b4; Fig. 2A and Fig. 2B). The pairwise RMSDs of C a of ubiquitin and the UBL domains of hHR23a, hPLIC-2, and parkin with respect to that of hUBLCP1 were 2.405 Å , 1.391 Å , 3.457 Å , and 3.785 Å , respectively. The fourth strand (b4) in the other UBLs was not observed in hUBLCP1; however, a unique b3-a2 loop was observed in that region ( Fig. 1B and Fig. 2A). In addition, unlike other UBL domains, the UBL domain of hUBLCP1 has a short b3, comprising residues Q43, K44, and L45. Most of the unique b3-a2 loop in UBLCP1 1-81 is exposed to the solvent (Fig. 2B). Very recently, the crystal structure of dmUBLCP1 derived from Drosophila melanogaster was reported [10]. Although the UBL domain of dmUBLCP1 has a high percentage of sequence identity (54%) with that of human UBLCP1, dramatic differences between the two structures were observed (Fig. 2C, 2D, and 2E). Surprisingly, two a-helices (a1 and a2) are connected by a short linker in the UBL domain of the dmUBLCP1. This is very unusual because a2 is located far from a1 (next to b3/b4) in most of the UBL domains, including that of hUBLCP1 ( Fig. 2A, 2B, and Fig.  S1). However, the structural folds in the UBL domains of

Molecular Interaction between hUBLCP1 and Rpn1
Because the structure of the UBL domain of hUBLCP1 is unique, detailed analysis of Rpn1 binding is of the essence in understanding the mechanism underlying the interaction between the two molecules. Using an immunoprecipitation assay, we showed that hUBLCP1 directly interacts with the Rpn1 of the regulatory particle of the 26 S proteasome (Fig. 3A). A recent study suggests that leucine-rich repeat-like domains of Rpn1 recognize the ubiquitin motif [25]. Homology modeling and secondary structure predictions show that Rpn1 has two structural subunits, regulatory subunits 1 (Rpn1 394-568 ) and 2 (Rpn1 640-772 ) with leucine-rich repeat-like-rich sequences (data not shown). Data from the GST pull-down assay showed that regulatory subunit 1 (Rpn1 394-568 ) directly interacts with UBLCP1 1-81 , whereas regulatory subunit 2 (Rpn1 640-772 ) of Rpn1 does not (Fig. 3B).
We performed NMR titration experiments to identify binding residues of the UBL domain of hUBLCP1 upon Rpn1 binding in solution. A number of chemical shift perturbations were observed upon Rpn1 394-568 titration ( Fig. 3C and 3D). Most of the resonance perturbations in the UBL domain of hUBLCP1 were found in the residues of the loop regions (i.e., W9, G11, T17, K49, K51, A55, and K65; Fig. 3E

Backbone Dynamics
Although the UBL domain of hUBLCP1 contains the canonical fold of UBL domains, the secondary structures and b3-a2 loop are very distinct. To correlate structure and dynamic properties of the UBL domain of hUBLCP1, we performed NMR backbone relaxation ( Fig. 4A and 4B) and 15 N-1 H heteronuclear NOE experiments (Fig. 4C). The overall average R1, R2, and NOE values were determined as 2.4460.13 s 21 , 6.7360.61 s 21 , and 0.7160.013, respectively. Interestingly, the average R1, R2, and XNOE values of the residues in the b3-a2 loops were determined as 2.3660.139 s 21 , 6.82560.88 s 21 , and 0.62, indicating that the b3-a2 loop is relatively rigid although it might have intrinsic dynamic profile. The order parameters and conformational exchange supported all dynamic data, with the average value of order parameters calculated as 0.86160.037 (Fig. 4D). The conformational exchange parameters suggest an evidence of rapid exchange characteristics for some residues in the b3-a2 loop (Fig. 4E).

Functional Implications
Although UBLCP1 has been classified as a member of the UDP family, it is important to note that the UBL domain of hUBLCP1 has a unique structural feature, which could be related to its specific function. It is well known that UBLs recognize the ubiquitin-associated domain or ubiquitin-interacting motif through their conserved hydrophobic residues [26]. The hydrophobic residues in the b3-b4 loop and b5 of hPLIC-2, hHR23a, and parkin are mainly involved in molecular interactions with their partner proteins [27][28][29]. Our findings suggest that the UBL domain of hUBLCP1 interacts with regulatory subunit 1 of Rpn1 via hydrophilic residues of the unique b3-a2 loop. Therefore, we hypothesize that the unique structure of the UBL domain of hUBLCP1 could be responsible for specific recognition of partner molecules, such as Rpn1. Data from proteomics studies have suggested that the peptidase activity of individual subunits of the murine cardiac 20 S proteasome is enhanced by phosphorylation and that the activity is tightly regulated by protein phosphatase 2A and protein kinase A [30,31]. The UBL domain of hUBLCP1 interacts with Rpn1 of the regulatory unit of the 26 S proteasome, and hUBLCP1 has been identified as one of the phosphatases functionally integrated with dephosphorylation of regulatory particles in the 26 S proteasome. Therefore, our findings will be directly applicable to future investigation of molecular interactions  Immunoblotting was conducted using HA-and FLAG-antibodies to distinguish the two proteins. IP was performed for the a-FLAG-RPN1 and coprecipitation was confirmed by IB. (B) In vitro GST pull-down assay using the GST-fused UBL domain of hUBLCP1 (UBLCP1 1-81 ), the TRX-His 6 -fused regulatory subunit 1 (Rpn1 394-568 ), and subunit 2 (Rpn1 640-772 with the proteasome complex during regulation of the 26 S proteasome through posttranslational modification of regulatory particles.