Ion mobility–mass spectrometry reveals conformational flexibility in the deubiquitinating enzyme USP5

Many proteins exhibit conformation flexibility as part of their biological function, whether through the presence of a series of well‐defined states or by the existence of intrinsic disorder. Ion mobility spectrometry, in combination with MS (IM–MS), offers a rapid and sensitive means of probing ensembles of protein structures through measurement of gas‐phase collisional cross sections. We have applied IM–MS analysis to the multidomain deubiquitinating enzyme ubiquitin specific protease 5 (USP5), which is believed to exhibit significant conformational flexibility. Native ESI–MS measurement of the 94‐kDa USP5 revealed two distinct charge‐state distributions: [M + 17H]+ to [M + 21H]+ and [M + 24H]+ to [M + 29H]+. The collisional cross sections of these ions revealed clear groupings of 52 ± 4 nm2 for the lower charges and 66 ± 6 nm2 for the higher charges. Molecular dynamics simulation of a compact form of USP5, based on a crystal structure, produced structures of 53–54 nm2 following 2 ns in the gas phase, while simulation of an extended form (based on small‐angle X‐ray scattering data) led to structures of 64 nm2. These data demonstrate that IM–MS is a valuable tool in studying proteins with different discrete conformational states.


Introduction
Ion mobility spectrometry (IMS) linked to MS (IM-MS) is a powerful analytical and structural tool. The orthogonality of the individual techniques affords 2D separation of analytes on the size-to-charge and mass-to-charge axes, respectively [1]. From an analytical perspective, when applied to complex mixtures, this gives far greater separation than is possible by either method alone [2]. IM-MS that utilizes either a static drift-tube IMS cell or a travelling-wave IMS (TWIMS) cell has the added advantage that it is able to measure the collisional cross section (CCS) of ions as absolute values in angstrom square or nanometer square from the ion drift time (t D ) data recorded [3]. When applied in combination with native ESI conditions this ability can be used to provide valuable protein structural information, albeit at relatively low resolution [4,5]. The limited resolution afforded by IM-MS is offset by the speed and sensitivity of the technique, which is orders of magnitude greater than X-ray crystallography and NMR spectroscopy, and by its ability to provide information on large, dynamic or heterogeneous protein complexes [6,7]. Given these considerations, the greatest potential impact of IM-MS in structural biology resides in areas where other, more established, techniques are of limited value for reasons of size, complexity, or time.
A significant consideration when using IM-MS for structural work is that measurements are made in the gas phase. The removal of solvent from a protein molecule during the electrospray process represents a major perturbation, and raises the question whether the method is suitable for providing information appropriate for biological interpretation. Work summarized by Breuker and McLafferty in 2008 [8], and numerous studies performed since, indicate that although structural change (usually compaction) occurs upon desolvation, for many proteins gross reorganization of the fold does not appear to take place on the time frame (< 50 ms) of IMS measurements [9][10][11][12]. The degree of structural collapse is, as one might expect, protein dependent. Flexible proteins with large voids and cavities appear prone to considerable collapse post desolvation, notably in the case of the chaperone protein GroEL [13], but other examples have been reported [14]. With the exception of these extreme examples, and providing care is taken to electrospray the protein from native conditions, using appropriate gas pressures and minimal collisional activation, valuable (biologically relevant) structural information can be gleaned from IMS-MS measurements [15][16][17][18][19][20][21]. This is especially so when experimental data are set alongside gasphase molecular dynamics (MD) simulations to provide more realistic model structures for comparison.
Recently, we reported the detection of two distinct conformers of cytochrome P450 (CYP) reductase using IM-MS [22]. This enzyme, responsible for two separate singleelectron reductions of the CYP heme iron, exists in interconverting compact and extended conformations linked to enzyme function. IM-MS was able to detect two clear gas-phase conformations of CYP reductase, which appeared to reflect the solution structures. Encouraged by these results, we have investigated the deubiquitinating enzyme (DUB) ubiquitin specific protease 5 (USP5), which is believed to exhibit conformational flexibility to enable it to disassemble a range of polyubiquitin chain topologies.
Ubiquitination is a regulatory PTM that involves covalent linkage of ubiquitin to target proteins by formation of an isopeptide bond between the C-terminal Gly76 residue of ubiquitin and lysine residues present in target proteins. Polyubiquitin chains are then often formed as a result of isopeptide, or peptide, linkages between multiple ubiquitin monomers, where any of ubiquitin's seven lysine residues or N-terminal Met can act as the site of attachment [23]. The modifications are associated with a range of functional outcomes, brought about by structural diversity in polyubiquitin topologies, which promotes recruitment of different effectors proteins containing ubiquitin-binding domains. Ubiquitination regulates a diversity of biological processes including protein degradation, intracellular signaling, and trafficking pathways. Like other PTMs, ubiquitination is reversible and ubiquitin modifications can be removed by a family of DUBs [24]. USP5, also known as isopeptidase T, selectively disassembles unanchored (i.e., substrate-free) chains of polyubiquitin [25]. These unanchored polyubiquitin chains have only very recently been realized to be physiologically relevant, acting in different pathways such as in the regulation of 26S proteasome activity [26], as second messengers in NF-B signaling pathways [27], and as regulators of innate immune signaling [28]. USP5 has also been directly implicated in other biological contexts, including control of repair of DNA double-strand breaks [29] and neuropathic and inflammatory pain [30], with aberrant splicing linked to glioblastoma tumorigenesis [9]. At the level of primary structure, USP5 contains a number of functional domains that underlie its binding and hence catalytic activity against unanchored polyubiquitin chains [31]. These include a ZnF-UBP ubiquitin-binding domain that binds with relatively high affinity to the C-terminal di-Gly motif in the proximal ubiquitin of an unanchored chain [32], thereby providing a mechanism for the selective degradation of unanchored polyubiquitin, as well as two ubiquitin-associated domains (UBAs, also ubiquitin binding) that recognize surface patches on other ubiquitin moieties within the chain. Avidity effects underlie high-affinity binding to unanchored polyubiquitin, with catalytic activity provided by an additional UBP domain within the primary sequence. An X-ray crystal structure of USP5 has been solved recently [33]. This identified the positions of the three ubiquitin-binding domains mentioned above, as well as revealing the presence of a second, cryptic ZnF-UBP domain. The structure raised a number of important questions. In particular, the orientation of the original ZnF-UBP with respect to the enzyme active site seemed inconsistent with its proposed function in recognizing unanchored polyubiquitin chains, and may be a result of a crystal artifact. In addition, over 150 residues were missing from the structure. This was believed to be due to inherent flexibility. Indeed, small-angle X-ray scattering (SAXS) analysis, performed in the same study, showed the existence of a structural form(s) significantly more extended than that seen by crystallography. These observations all point toward considerable flexibility in the structure of USP5. Here we report an investigation of USP5 using ESI-IM-MS, which provides further evidence of conformational flexibility in this important regulatory enzyme.

Protein expression
The protein coding region of short variant full-length human USP5 (residues 1-835) was cloned from PCR-amplified human U20S cDNA. This PCR product was ligated into the BamHI/XhoI sites of pGEX-4T-1 (GE Healthcare, Buckinghamshire, UK) and subsequently mutagenized by sitedirected mutagenesis (QuikChange kit; Stratagene, Agilent, Stockport, UK), to introduce an active site (C335A) mutation, preventing deubiquitination activity. The integrity of the construct was verified by DNA sequencing.
Plasmid DNA encoding GST-C335A USP5 was transformed into competent XL-10 gold cells, and grown in a 2 L flask containing 1 L of Luria broth (LB) at 37ЊC. When OD 600 ß0.6 was reached, the temperature was reduced to 15ЊC and overexpression induced with 0.2 mM IPTG. Cell cultures were then grown for 40 h. The cell pellet was collected by centrifugation and snap frozen at −80ЊC.

Sample preparation
USP5 samples were desalted from the gel-filtration buffer into aqueous ammonium acetate (200 mM) by ten washing cycles using Viva-Spin ultrafilters (10-kDa MWCO, 0.5 mL, Sartorius Stedim Biotech), and quantified by nanodrop. Retentate solutions were diluted into ammonium acetate (200 mM) to a final protein concentration of 10 M.

IM-MS
ESI-TWIMS-MS was performed on a Waters (Altrincham, UK) Synapt G1 High Definition Mass Spectrometer (HDMS)-a hybrid quadrupole/ion mobility/orthogonal acceleration TOF (oa-TOF) instrument. Static nanospray ionization was performed using the standard Waters nanospray source and homemade nanospray tips. Capillaries were pulled using a Flaming/Brown P-97 micropipette puller (Sutter Instruments, Novato, CA, USA). Once pulled, the tips were coated in silver using a home-built vacuum evaporator. The nanospray tips were operated at a capillary voltage of 1.3 kV, with the instrument in positive-ion mode. The sample cone was maintained at 30 V, required to avoid gas-phase unfolding. Collisional cooling was applied by partially closing a Speedi-valve attached to the source pumping line until a backing pressure read-back of 4.5 mbar was obtained. The "trap" and "transfer" T-wave collision cells, containing argon gas held at a pressure of 2.5 × 10 −2 mBar, were operated at a collision energy of 5 V. The TWIMS ion mobility cell, which contained nitrogen gas at 0.45 mBar and ambient temperature, was operated with a wave height of 10 V traveling at 300 m/s. Travelling-wave parameters for the "trap" and "transfer" were as follows: trap 300 m/s, 0.5 V and transfer 248 m/s, 4 V. The oa-TOF-MS was operated over the scanning range of m/z 500-8000 at a pressure of 1.8 × 10 −6 mBar. Identical conditions were used to analyze USP5 and the calibrations standards. The instrument was controlled and data were viewed using MassLynx 4.1 software (Waters).
TWIMS measurements on USP5 were calibrated against a set of standard CCS values for beta-lactoglobulin, BSA, and alcohol dehydrogenase (Sigma-Aldrich) taken from [34] using the method of Ruotolo et al. [6] (see Supporting Information). Ion mobility drift traces for each ion of USP5 were exported into Excel (Microsoft) and the arrival-time data points converted onto the CCS scale using the calibration parameters obtained. This allowed the drift traces for each ion to be plotted on a CCS scale, with charge state correction.

Modeling and MD
PDB entry 3IHP was used as a starting point for modeling USP5. The 151 missing residues were built using iTASSER [35] and hydrogens added with MolProbity [36]. Protein molecular modeling was performed using the Amber99SB force field [37] on the European eNMR grid [38]. The protein was modeled using its net charge state under physiological conditions. Prior to MD simulations, the structure was relaxed by 20 000 steps of energy minimization in the gas phase at 300 K (Amber 11 Sander parameters: imin = 1 (energy-minimization mode), maxcyc = 20 000 (number of steps), ncyc = 500 (number of steepest descent algorithm steps), ntb = 0 (period boundary off), cut = 12.0 (nonbonded cutoff inÅ), igb = 0 (no generalized Born solvent, i.e., gas phase); all other parameters Sander defaults). This removed the vast majority of clashes from the original iTASSER structure (from an unacceptable clash score of 131/1000 atoms to an acceptable 0.7/1000 atoms, as assessed by MolProbity). A 2-ns gas-phase MD simulation was then run on the energy-minimized structure at a constant temperature of 300 K (Amber 11 Sander parameters: imin = 0 (minimization off, i.e., MD run), ntb = 0 (period boundary off), cut = 12.0 (nonbonded cutoff inÅ), igb = 0 (no generalized Born solvent, i.e., gas phase), ntt = 3 (Langevin thermostat), gamma ln = 1 (collision frequency), temp0 = 300 (temperature in K), tempi = 300 (initial temperature in K), nstlim = 2 000 000 (number of 1 fs steps), ntpr = 100 (rate of information written to the output file in steps), ntwx = 2000 (rate of coordinates written to the output file in steps, i.e., 1000 sets of coordinates over the 2-ns simulation), ig = 73 277 (random seed); all other parameters Sander defaults). The 1000 output trajectories from this MD simulation were visualized in Visual Molecular Dynamics (VMD), and every fifth structure exported to a PDB file (to keep the file size manageable). The projection approximation CCS of each structure was calculated using the DriftScope 2.1 CCSCalc program (Waters). To convert projection approximation CCS values to "Trajectory Method-like" values they were multiplied by the empirical scaling factor 1.14 [39]. A plot of CCS versus time was generated to assess the change in structure over the time of the MD run. To estimate the theoretical CCS of the more extended conformation of USP5, corresponding to the higher charge states, the energy-minimized structure (above) was adjusted manually to mimic that seen in the SAXS analysis [33]. The resulting structure was then relaxed and submitted to gas-phase MD (2 ns) as described above.

Results
Recombinant human USP5 (short variant form) was expressed in Escherichia coli and purified as described in Section 2. Following desalting into aqueous ammonium acetate (200 mM), USP5 (10 M) was examined by native ESI-MS using a Waters Synapt HDMS spectrometer, and homemade static nanospray capillaries. The resulting spectrum, shown in Fig. 1, revealed a measured   this study was produced as a GST fusion, and the tag cleaved before purification and subsequent MS analysis, so that this could not be the explanation for the observed bimodel distribution of CSDs. Care was taken to optimize MS conditions, and minimize the nanoESI capillary voltage (<1.3 kV), but the resulting spectrum of USP5 was unaltered, and the two CSDs reproducible. It is established that the average charge state exhibited by a protein population, electrosprayed under native conditions, correlates with the surface area of the protein [40,41]. This indicated that two significantly different forms of USP5 were present in solution.
To probe this phenomenon further, IM-MS was employed. The Waters Synapt MS used in this study is equipped with a TWIMS cell. Figure 2 shows the 2D trace of m/z versus t D recorded for USP5 under native conditions. The two CSD populations are clearly visible. Due to the complex nature of the electric field used in TWIMS devices, the CCS of ions cannot be determined directly from mobility t D values, and calibration was required. Using the method of Ruotolo et al. [6], and standard CCSs values provided by Bush et al. [34], the proteins beta-lactoglobulin, BSA, and alcohol dehydrogenase were used to calibrate the TWIMS cell (see Section 2 and Supplorting Information for details). These proteins were chosen to bracket the CCSs of USP5 ions, and electrosprayed under conditions identical to those used for USP5 itself. Figure 3 shows the TWIMS drift traces for the major charge states of USP5 (see Fig. 1). Data were plotted on a CCS axis to allow direct comparison between each species. The two CSDs seen in Figs. 1 and 2 also formed two clearly resolved groups on the CCS axis. Together the lower CSD (17-21 + ) gave a mean CCS value of 52 ± 4 nm 2 , while the higher CSD (24-29 + ) possessed a mean CCS value of 66 ± 6 nm 2 . The CCS values for each individual charge state are listed in the Supporting Information.
In order to provide insights into the structure of USP5 in the gas phase, an MD simulation was performed using the  Amber99B force field (see Section 2). Starting with the crystal structure (PDB code 3IHP), and following modeling-in of 151 missing residues, the energy-minimized structure of USP5 was submitted to 2 ns of MD in the gas phase, at 300 K, using the Amber Sander parameters described in Section 2. Given the nature of USP5's structure, significant collapse was expected following desolvation. Figure 4 shows a plot of the CCS for each structure as a function of time at a sampling rate of 1 structure every 10 ps. A rapid contraction of the structure from its initial CCS of 63.4 nm 2 to approximately 55 nm 2 , representing a reduction of 13%, was apparent over the first 200 ps of the simulation. For the remaining 1.8 ns of the MD run a steady decrease to a CCS of approximately 53-54 nm 2 . The first marker on Fig. 3 shows the position of 53.5 nm 2 on the CCS axis, and is close to the range of CCSs exhibited by the low CSD of USP5 (within ca. 5%). It appears that low charge states of USP5 undergo only a small degree of additional collapse over that seen theoretically on the 2-ns timescale. To provide a guide for the structure of the higher CSD ions a conformation of USP5, which approximated to the extended SAXS structure in [33], was generated. Following relaxation, a 2-ns gas-phase MD simulation was performed, as for the compact form. The structure was seen to collapse from 74 to 64 nm 2 . A marker corresponding to 64 nm 2 is shown on Fig. 3.
Finally, to provide preliminary insights into the relevance of the two USP5 conformations for substrate interactions, we measured a spectrum of the enzyme in the presence of Lys48-linked di-ubiquitin. The results (Supporting Information Fig. 7) show that the complex exhibits low charge states only, meaning that it possesses a compact conformation, perhaps with USP5 domains folding around the di-ubiquitin substrate.

Discussion
Using a combination of CSD analysis and IM-MS measurement, we have shown that USP5 adopts two clear sets of conformational states: a major compact form and a minor extended form. Comparison of these results with SAXS analysis [33] reveals striking similarities. First, each technique shows a similar ratio of compact to extended conformers: approximately 3.5:1 by IM-MS (based on the relative integrals of the two distributions in Fig. 3), and 3:1 by SAXS (based on the three conformers per ensemble representation in [33]). Second, the relative sizes (in units of area) of the two conformations are also similar when measured by IM-MS (CCS: 66:52 nm 2 = 1.3) and SAXS (radius of gyration, R g 2 : 20.:14 nm 2 = 1.4, again using the data for three conformers per ensemble representation in [33]). These findings, together with the good agreement between measured CCSs and those obtained by gas-phase MD simulation, indicate that IM-MS can provide important structural information on conformationally flexible proteins such as USP5 even in the presence of gas-phase structural collapse.
The conformational flexibility of USP5 is believed to be essential for its function in disassembling unanchored polyubiquitin chains possessing a variety of isopeptide linkages, and therefore topological forms. The position of USP5's Cterminal ZnF-UBP domain in the crystal structure (it points away from the active site domain) appears to be at odds with its proposed role in recognizing unanchored polyubiquitin chains and guiding them to the enzyme's active site. Indeed, the authors of the X-ray study discuss the possibility of a crystallization artifact brought about by a nonnative disulfide bond in their structure [33]. The SAXS study [33], performed under reducing conditions, provides further evidence for extended as well as compact forms of USP5 in solution, and is in agreement with the IM-MS data as discussed above.
In summary we present ESI-IM-MS data for the DUB USP5, which support the proposal that it exhibits considerable conformational flexibility in order to recognize and process a variety of polyubiquitin chain topologies. These results further demonstrate that IM-MS is a valuable tool in studying proteins with a number of discrete conformational states, in addition to those possessing intrinsic disorder or misfolding.
We are grateful to the University of Nottingham for funding.
The authors have declared no conflict of interest.