Exploring the Mechanism of Selective Noncovalent Adduct Protein

Apr 12, 2008 - Jae Yoon Ko , Sung Woo Heo , Joon Ho Lee , Han Bin Oh ... Sung Woo Heo , Tae Su Choi , Kyung Man Park , Young Ho Ko , Seung Bin Kim ...
0 downloads 0 Views 608KB Size
Download PDF
Anal. Chem. 2008, 80, 3846–3852

Exploring the Mechanism of Selective Noncovalent Adduct Protein Probing Mass Spectrometry Utilizing Site-Directed Mutagenesis To Examine Ubiquitin Zhenjiu Liu,† Shijun Cheng,‡ Daniel R. Gallie,*,‡ and Ryan R. Julian*,† Departments of Chemistry and Biochemistry, University of California, Riverside, California 92521 Mass spectrometry (MS) is emerging as an additional tool for examining protein structure by way of experiments where structurally related mass changes induced in solution are subsequently detected in the gas phase. Selective noncovalent adduct protein probing (SNAPP) is a recent addition to this type of experiment. SNAPP utilizes noncovalent recognition of lysine residues with 18crown-6 (18C6) to monitor changes in protein structure. It has been observed that the number of 18C6 adducts that attach to a protein is a function of the structure of the protein. The present work seeks to examine the underlying chemistry which controls the differential attachment of 18C6 to lysine by using ubiquitin as a model system. Ubiquitin is a small protein with a structure that has been well characterized by multiple techniques. Sitedirected mutagenesis was used to create a series of ubiquitin mutants where the lysine residues were exchanged for asparagine one at a time. These mutants were then evaluated by SNAPP-MS to determine the relative contribution of each lysine as a binding site for 18C6. It was found that attachment of 18C6 is largely controlled by the strength of intramolecular interactions involving lysine residues. Salt bridges provide the greatest interference, followed by hydrogen bonds. In addition to determining the mechanism for SNAPP, insights are provided about the structure of ubiquitin including confirmation of the existence of two dynamic states for the native structure. These results are discussed in relation to the biological functions of ubiquitin. Protein structure determination remains an area of active interest. Consequently, there are a variety of experimental techniques which can be implemented. For example, X-ray crystallography1 yields unrivaled atomistic structural detail. Nuclear magnetic resonance (NMR) is another popular technique that can also yield subresidue structural information and provide additional data about dynamic processes such as protein folding.2 Circular dichroism (CD) can be used to interrogate secondary structure

and the organization of tertiary structure in simple and rapid experiments.3 In addition, mass spectrometry (MS) based techniques are available. Simple examination of the charge state distributions reveals information about tertiary structure.4,5 Hydrogen/deuterium exchange (HDX) MS can be used as an elegant probe for reporting on the exchangeability of backbone amide hydrogens.6 Structural information can be isolated to small subregions of a protein using HDX-MS, and dynamics and kinetics can be explored.7,8 Variations of HDX can be utilized to translate changes in structure into analytically relevant measurements such as the identification of protein ligands or the quantitative measurement of binding constants.9,10 Also relevant are techniques which attempt to measure binding constants or identify ligands through direct observation in the mass spectrum.11–13 The key to success in these MS based experiments is that the properties observed in the mass spectrometer must reflect solution phase properties, as proteins can undergo significant structural rearrangement in the foreign environment of the gas phase.14 Recently, a complimentary MS based technique using noncovalent interactions to examine protein structure in solution was reported.15 This method, which we will call selective noncovalent adduct protein probing (SNAPP), relies on a specific interaction between lysine residues and 18-crown-6 (18C6).16 This recognition is achieved by three hydrogen bonds between alternating oxygens in 18C6 and the three ammonium hydrogens in the side chain of lysine. The interaction is weak in solution, making it a sensitive probe of solution phase chemistry, but strong once transferred to the gas phase (essentially locking in solution phase information (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

* To whom correspondence should be addressed. E-mail: [email protected], [email protected]. † Department of Chemistry. ‡ Department of Biochemistry. (1) Matthews, B. W. Annu. Rev. Phys. Chem. 1976, 27, 493–523. (2) Dyson, H. J.; Wright, P. E. Chem. Rev. 2004, 104, 3607–3622.

3846

Analytical Chemistry, Vol. 80, No. 10, May 15, 2008

(13) (14) (15) (16)

Sreerama, N.; Woody, R. W. Methods Enzymol. 2004, 383, 18–351. Grandori, R. J. Mass. Spectrom. 2003, 38, 11–15. Dobo, A.; Kaltashov, I. A. Anal. Chem. 2001, 73, 4763–4773. Kaltashov, I. A.; Eyles, S. J. Mass Spectrom. Rev. 2002, 21, 37–71. Wales, T. E.; Engen, J. R. Mass Spectrom. Rev. 2006, 25, 158–170. Hossain, B. M.; Konermann, L. Anal. Chem. 2006, 78, 1613–1619. Tang, L.; Hopper, E. D.; Tong, Y.; Sadowsky, J. D.; Peterson, K. J.; Gellman, S. H.; Fitzgerald, M. C. Anal. Chem. 2007, 79, 5869–5877. Zhu, M. M.; Chitta, R.; Gross, M. L. Int. J. Mass Spectrom. 2005, 240, 213–220. Sun, J. X.; Kitova, E. N.; Wang, W. J.; Klassen, J. S. Anal. Chem. 2006, 78, 3010–3018. Sun, J.; Kitova, E. N.; Sun, N.; Klassen, J. S. Anal. Chem. 2007, 79, 8301– 8311. Nazabal, A.; Wenzel, R. J.; Zenobi, R. Anal. Chem. 2006, 78, 3562–3570. Shelimov, K. B.; Clemmer, D. E.; Hudgins, R. R.; Jarrold, M. F. J. Am. Chem. Soc. 1997, 119, 2240–2248. Ly, T.; Julian, R. R. J. Am. Soc. Mass Spectrom. 2006, 17, 1209–1215. Julian, R. R.; Beauchamp, J. L. Int. J. Mass Spectrom. 2001, 210, 613–623. 10.1021/ac800176u CCC: $40.75  2008 American Chemical Society Published on Web 04/12/2008

in a manner similar to HDX).17 Initial experiments demonstrated that the number of 18C6 adducts that attach to a protein is a function of the structure of the protein,15 making SNAPP a potential tool for identifying structure changing events such as point mutations or ligand binding. It was hypothesized originally that lysine availability, determined by the propensity of a particular lysine side chain to interact intra- versus intermolecularly, controls the number of adducts that attach to a given protein structure; however, the underlying mechanism(s) which might control this lysine availability have not been explored. Furthermore, residue specific information was not obtainable in the initial experiments. Site-directed mutagenesis is a powerful technique for controlling protein sequence.18 Single amino acid substitutions of specific residues can be implemented with this method. For the present work, lysine residues are of interest, which are usually found on the surface of proteins,19 presumably to enhance solubility. Lysine is typically protonated at biological pH, and can therefore participate in salt bridges.20,21 There are also three potential sites for hydrogen bonds, which leads to the favorable interaction between lysine and 18C6. Mutation of lysine to another residue is therefore predicted to reduce attachment of 18C6 in SNAPP experiments. Of the options available, mutation to arginine would enable retention of a charged site, minimizing any changes in electrostatics. However, 18C6 can weakly attach to arginine, particularly in the absence of lysine.22 Therefore, asparagine was chosen as the replacement amino acid for the mutagenesis of ubiquitin in the present work. Asparagine is charge neutral, hydrophilic, similar in size, yet not isobaric with lysine (allowing for easy identification of the mutation by MS). Furthermore, 18C6 will not attach to the side chain of asparagine. Lysine residues are also of particular interest in the ubiquitin system. Ubiquitin is a compact, 76 residue protein which is found in all eukaryotes with a highly conserved sequence.23 Ubiquitin is involved in protein degradation via the post-translational modification of proteins. Interestingly, ubiquitin also forms chains of polyubiquitin where the number of ubiquitins attached and the linkage site between them are important.24 For example, tetraubiquitin linked through isopeptide bonds between the Cterminus and Lys48 signals for rapid protein degradation. However, studies in yeast revealed that polyubiquitin linked through Lys63 affects other pathways related to DNA damage, protein transport, and protein synthesis.25 Thus the chemical microenvironment surrounding each lysine residue (which will be rigorously probed in this work) may play a role in defining the biological activity of the protein. The present work explores the mechanism of SNAPP by examination of a series of ubiquitin mutants where the lysine residues have been exchanged for asparagine one at a time using site-directed mutagenesis. Comparison of these mutants to the (17) Julian, R. R.; May, J. A.; Stoltz, B. M.; Beauchamp, J. L. Angew. Chem., Int. Ed. 2003, 42, 1012–1015. (18) Barik, S. In Site-directed mutagenesis by double polymerase chain reaction; White, B. A., Ed.; PCR Protocols: Totowa, NJ, 1993; pp 277–286. (19) Lins, L.; Thomas, A.; Brasseur, R. Protein Sci. 2003, 12, 1406–1417. (20) Kumar, S.; Nussinov, R. J. Mol. Biol. 1999, 293, 1241–1255. (21) Hendsch, Z. S.; Tidor, B. Protein Sci. 1994, 3, 211–226. (22) Julian, R. R.; Akin, M.; May, J. A.; Stoltz, B. M.; Beauchamp, J. L. Int. J. Mass Spectrom. 2002, 220, 87–96. (23) Hershko, A.; Ciechanover, A. Annu. Rev. Biochem. 1998, 67, 425–479. (24) Pickart, C. M.; Fushman, D. Curr. Opin. Chem. Biol. 2004, 8, 610–616. (25) Pickart, C. M. Trends Biochem. Sci. 2000, 25, 544–548.

wild-type protein enables determination of the relative contribution for each lysine as a binding site for 18C6. Reference to the known structure of ubiquitin can then be used to evaluate the underlying causes controlling 18C6 attachment. It is found that the nature and strength of intramolecular interactions which involve lysine dictate attachment of 18C6 to the protein. Salt bridges provide the greatest interference preventing attachment of 18C6, followed by hydrogen bonds. Solvent accessibility is a less accurate predictor of lysine availability. Most of the lysine residues in ubiquitin contribute to the observed SNAPP distributions, suggesting that SNAPP functions as a sensitive probe for protein structural change. The results also provide interesting insights into biological functions of ubiquitin which are briefly addressed. MATERIALS AND METHODS Oligonucleotides and ubiquitin from bovine red blood cells were purchased from Sigma Chemical Co. (St. Louis, MO) and used without further purification. Escherichia coli BL21 (DE3) cells were purchased from Novagen Inc. (Madison, WI). T4 DNA ligase, pGEM T-vector, BamHI and NdeI were purchased from New England Biolabs (Beverly, MA); Wizard SV Gel and PCR Cleanup System and Plasmid Miniprep Kit were purchased from Promega (Madison, WI). Gene Jet Plasmid Miniprep Kit was purchased from Fermentas Life Science (Glen Bernie, MA). All other chemicals and reagents were purchased from Sigma-Aldrich (St. Louis, MO) or EMD (Gibbston, NJ) unless noted otherwise and used without further purification. Plasmids Constructs. The coding sequence of Arabidopsis ubiquitin (AT4g05320) was amplified by reverse transcription-PCR (RT-PCR). Because AT4g0532 encodes a polyubiquitin, primers were designed to amplify the N-terminal monomer (AtUBI) and a stop codon was introduced in the reverse primer. A NdeI site and a BamHI site were added in the forward and reverse primers, respectively, to facilitate cloning of the amplified product. The PCR product was initially cloned into pGEM T-vector and its correct amplification confirmed by sequencing. The NdeI/BamHI fragment was then introduced into the NdeI/BamHI sites of pET11b. PCR-based mutagenesis was used to introduce specific lysineto-asparagine mutations (i.e., AAG to AAC) into the AtUBI sequence. The PCR products were cloned into pET11b as described for the wild-type sequence, and the sequence of the entire gene was obtained to confirm the mutation. PCR reactions contained 0.5 pmol of template DNA, 15 pmol of each primer and 2 units of Taq1 DNA polymerase. PCR conditions used as follows: 15 min at 94 °C, 30 cycles of 94 °C for 20 s, 60 °C for 30 s, 72 °C for 30 s; and the reactions were incubated at 72 °C for an additional 10 min. Overexpression and Purification of Ubiquitin Variants. Wild-type and mutant AtUBI constructs were transformed into E. coli BL21-(DE3) cells. Protein expression was induced by the addition of 0.25 mM IPTG to midlog cells, followed by incubation at 30 °C for 5 h. Cells were harvested by centrifugation, and the cell pellet was washed twice with water and frozen at –20 °C. The frozen cell pellets were thawed, suspended in water containing 0.5 mM proteinase inhibitor cocktail, and sonicated 30 times on ice for 10 s with 30 s intervals in between using an ultrasonic homogenizer. Following removal of cell debris by centrifugation, the supernatant was fractionated using a Beckman Coulter System Gold HPLC system (Fullerton, CA) and a 10 × 100 mm preparative Analytical Chemistry, Vol. 80, No. 10, May 15, 2008

3847

Figure 1. Representative SNAPP distributions for bovine (white) and Arabidopsis (gray) ubiquitin. The numbers on the x-axis refer to charge state followed by the number of 18C6 adducts. The results are similar for both proteins. The +7 charge state yields a distribution representing a compact structure. The +9 charge state yields distribution corresponding to a more open structure.

C18 column (Waters, Milford, MA). A 55 min gradient of 5–50% acetonitrile in 0.1% aqueous solution of trifluoroacetic acid (TFA) was used with a flow rate of 1.5 mL/min. Chromatograms were recorded by absorbance detection at 280 nm. The purity of the fractions was assessed by SDS-PAGE and ESI-MS. Pooled fractions containing the wild-type or mutant ubiquitin were lyophilized and stored at -20 °C. Ubiquitin concentration was measured using absorbance at 280 nm (extinction coefficient of 0.149 au mL mg-1 cm-1).26 CD Spectral Analysis. Conformational analysis of the wildtype and mutant ubiquitin proteins was determined using circular dichroism (CD) spectroscopy. Spectra were collected in a JASCO J815 spectropolarimeter at 25 °C with a 1 cm path length quartz cuvette. The scan range was from 195 to 450 nm. The average of five scans was used for the analysis. Mass Spectrometry. Mass spectra were obtained using a Finnigan LTQ linear ion trap mass spectrometer (Thermo Electron, San Jose, CA) equipped with a standard electrospray ionization source without modification. Voltages in the source region were optimized for the observation of noncovalent complexes. This was achieved by electrospraying a solution of ubiquitin and 18C6 (Alfa Aesar, Pelham, NH) and optimizing adduct ion intensities by tuning individual voltages. Typical lens settings for observing protein-adduct formation are as follows: capillary voltage 9.0 V, capillary temperature 215 °C, and tube lens offset 145 V. The tube lens voltage was further optimized for maximum total ion count and SNAPP distribution reproducibility. Small variations in the SNAPP distributions can be source dependent; therefore, all experiments were conducted on the same day when possible. Otherwise, the instrument source voltages were recalibrated against a standard solution to reproduce a known SNAPP distribution. The electrospray voltage was set at 4.5 kV. Protein solutions were made in pure water purified by Millipore Direct-Q uv (Millipore, Billerica, MA). A concentration of 5 µM was used for the wild-type and mutant ubiquitin proteins. For all solutions where 18C6 was added, the 18C6 concentration was twice the concentration of lysine residues to ensure stoichiometric excess. RESULTS AND DISCUSSION The SNAPP distributions obtained for wild-type Arabidopsis thaliana and bovine ubiquitin are shown in Figure 1. In this figure, each bar represents the intensity of a corresponding ion observed in the mass spectrum, where the numbers designate charge state (26) Wintrode, P. L.; Makhatadze, G. I.; Privalov, P. L. Proteins 1994, 18, 246– 253.

3848

Analytical Chemistry, Vol. 80, No. 10, May 15, 2008

followed by the number of 18C6 adducts. The SNAPP data are grouped according to charge state, with the shape and size of each distribution revealing the degree of 18C6 attachment. The distributions are each normalized to the zero adduct peak to allow for easy comparison of the relative number of adducts between different charge states (which frequently represent different protein structures).15 The primary sequences for Arabidopsis and bovine ubiquitin differ by three amino acids (S19P, D24E, and A57S). Although the structure of Arabidopsis ubiquitin has not been solved previously by X-ray or NMR, the substitutions in Arabidopsis can easily be accommodated without perturbing any significant backbone or side chain interactions (as verified by in silico mutations). Even the proline mutation, which is the most dramatic change, can be accommodated. Pro19 in bovine ubiquitin occupies a turn region between R1 and β2.27 Replacement of this residue in Arabidopsis does not disrupt the turn and allows a hydrogen bond between Ser19 and the backbone carbonyl of Ile61. The similarity of the SNAPP distributions also suggests that the bovine and Arabidopsis structures are very similar. This is not surprising given the structural stability of the ubiquitin backbone motif.28 In fact, 29 amino acid substitutions in the ubiquitin related protein RUB1 yield a structure with an essentially identical backbone.29 Close inspection of the data in Figure 1 reveals the presence of two distinct SNAPP distributions for both proteins, one dominant in the +7 charge state and the other most prevalent in the +9 charge state. Inspection of all the data suggests that there may be some overlap between these distributions; however, the existence of two SNAPP distributions indicates the presence of at least two protein structures, due to the measurable change in the sum of the lysine-18C6 interactions. The observation that the two structures appear in different preferred charge states suggests that one structure is more open (and thus able to accept more charges) than the other. The distinction between the bovine and Arabidopsis structures is most notable in the open form, where it appears that the Arabidopsis structure can accommodate more 18C6 adducts. All structural interpretations discussed below will be relative to the bovine structure (primarily in the folded +7 charge state), with the underlying assumption that both structures are similar. Comparisons to the SNAPP experimental results will be made with known structures of ubiquitin as determined by X-ray crystallography27 and NMR.30–32 We have evaluated all of the lysine residues in these structures to determine potential intramolecular interactions which might interfere with attachment of 18C6 to lysine. In Table 1, the average heteroatom separation distance between each lysine side chain nitrogen and all other potential hydrogen bonding atoms is shown. This evaluation will also reveal any close contacts that result from the formation of a salt bridge. For structures determined by NMR, the average results from 10 (27) Vijaykumar, S.; Bugg, C. E.; Cook, W. J. J. Mol. Biol. 1987, 194, 531–544. (28) Kony, D. B.; Hunenberger, P. H.; van Gunsteren, W. F. Protein Sci. 2007, 16, 1101–1118. (29) Rao-Naik, C.; delaCruz, W.; Laplaza, J. M.; Tan, S.; Callis, J.; Fisher, A. J. J. Biol. Chem. 1998, 273, 34976–34982. (30) Cornilescu, G.; Marquardt, J. L.; Ottiger, M.; Bax, A. J. Am. Chem. Soc. 1998, 120, 6836–6837. (31) Babu, C. R.; Flynn, P. F.; Wand, A. J. J. Am. Chem. Soc. 2001, 123, 2691– 2692. (32) Kitahara, R.; Yokoyama, S.; Akasaka, K. J. Mol. Biol. 2005, 347, 277–285.

Table 1. Average Nearest Neighbor Hydrogen Bond Distance (in Å) PDB IDa b

Lys6 Lys11 Lys27 Lys29 Lys33 Lys48 Lys63

1d3z 4.19 (0.59)g 3.67 (1.06) 3.10 (0.61) 3.05 (0.47) 4.11 (0.81) 5.23 (0.64) 6.97 (0.52)

c

1g6j 4.46 (0.76) 5.92 (0.83) 3.33 (0.27) 3.87 (0.89) 4.92 (0.87) 4.99 (0.81) 6.08 (1.21)

1v81d 5.26 (0.46) 5.05 (0.87) 4.48 (0.66) 3.96 (0.77) 4.85 (0.95) 4.75 (0.74) 5.25 (0.34)

1ubqe 4.97 (0.32) 5.67 (0.72) 3.91 (0.35) 3.58 (0.59) 5.48 (0.37) 4.30 (0.67) 4.12 (1.11)

1v80f 5.37 3.35 2.90 2.68 3.41 3.38 4.85

a Structures are listed according to PDB ID. b NMR solution phase structure. c NMR solution phase structure in micelles. d NMR low pressure solution phase structure. e NMR high pressure solution phase structure. f X-ray crystal structure. g Standard deviation in parentheses.

structures are reported. In addition, standard deviations for these distances are reported for the NMR structures to serve as a qualitative guide for structural dynamics (i.e., side chains with a large variation in the average distance are potentially in dynamic portions of the protein structure). Of particular interest are the NMR results obtained at high and low pressure. These experiments were designed to explore dynamic states of ubiquitin. The primary changes observed in the high pressure structure coincide with changes observed in the SNAPP distribution between the open and closed structures, as will be discussed in further detail below. In Figure 2a the SNAPP distributions for the +7 and +9 charge states of wild-type ubiquitin and the K6N mutant are shown. As noted in the introduction, 18C6 will not attach to asparagine. Therefore, this mutation should result in a shift in the SNAPP distribution equivalent to the amount of 18C6 which attaches to Lys6 in the wild-type protein. It can be observed in Figure 2a that the amount of 18C6 adducts has decreased for the K6N mutant (white bars) relative to the wild type for both charge states. This is consistent with Lys6 functioning as a binding site for 18C6. Inspection of Lys6 in the crystal structure reveals that the side chain is not in close proximity to any acidic side chains, and the nearest potential hydrogen bond partner is ∼5.4 Å away. Similar results are obtained by inspection of all NMR structures, with average distances of at least 4 Å and small standard deviations. The results for the K11N mutant are shown in Figure 2b. Again, the number of 18C6 adducts is decreased relative to the wild-type protein. In addition, the K11N mutant does not produce a dominant single adduct peak in the +9 charge state, which differentiates K11N from any other mutation. This distribution corresponds to the open structure, suggesting that Lys11 may be particularly important for observing the open state with the characteristically intense monoadduct. Mutation of Lys11 is also unique because it causes the greatest observed shift in the SNAPP distribution. The simplest explanation for this observation would be that Lys11 is the most chemically available lysine; however, inspection of the crystal structure reveals a salt bridge interaction between Lys11 and Glu34. A salt bridge would not be consistent with predicted high lysine availability, an apparent contradiction which is resolved by further inspection. To begin with, the NMR results suggest that the Lys11-Glu34 salt bridge is not always present even in the folded protein. Furthermore, the results from the high pressure experiments (1v81), which were designed to isolate different dynamic states

Figure 2. SNAPP distributions for the wild-type (gray) and mutant (white) proteins. Significant shifts in the SNAPP distributions are observed for both Lys6 and Lys11, indicating that they are binding sites for 18C6. Mutation of Lys27 elicits almost no change, while elimination of Lys29 has an intermediate effect. The results are rationalized in terms of intramolecular interactions as described in the text.

of ubiquitin, are of particular importance. The high pressure structure is a more open conformation of ubiquitin which cannot form a salt bridge between Lys11 and Glu34. This causes the local chemical environment surrounding Lys11 to change dramatically when compared to the fully folded state. Comparison with the remaining lysine residues reveals that Lys11 undergoes the largest structural shift when the crystal structure is compared to the high pressure NMR structure. If the high pressure structure represents a dynamic state which exists in equilibrium with the folded state in solution, then this structure may be the open structure which is observed in the +9 charge state by SNAPP. The switch from weak salt bridge in the fully folded protein to freely available lysine in the open state explains the dramatic difference in the SNAPP distributions between the +7 and +9 charge states for the wild-type ubiquitin and all mutants except K11N. Mutation of Lys11 obscures observation of the open state because it is the key residue with very different 18C6 binding properties in the two states. Therefore, the reason that K11N Analytical Chemistry, Vol. 80, No. 10, May 15, 2008

3849

exhibits the greatest change in the SNAPP distribution is due to dynamic tertiary structural interactions which modulate the local chemical environment for Lys11. The only other likely explanation for the results obtained with K11N would be that this mutation causes the protein structure to be altered substantially. However, previous experiments have demonstrated that perturbation of the native structure of a protein typically leads to increased attachment of 18C6 due to disruption of native intramolecular interactions.15 In Figure 2 the amount of 18C6 attachment decreases substantially, which is not indicative of unfolding. Therefore, the SNAPP results are not consistent with a significant structural rearrangement, an interpretation that is supported by CD experiments presented below. Results for the K27N mutation are shown in Figure 2c. This substitution leads to almost no detectable change in the +7 (or natively folded structure) SNAPP distribution. In fact, this mutation leads to less perturbation in total than any other substitution examined in this work. Inspection of the crystal structure reveals that Lys27 forms a stable salt bridge with Asp52. Further inspection of the solution phase NMR structures reveals that this salt bridge is consistently present, even in the alternate state observed at high pressure. In addition, Lys27 is significantly buried. The solvent accessibility (discussed in greater detail below) for Lys27 is lower than that for all other lysine residues in ubiquitin. The combination of these two factors (salt bridge formation and inaccessibility) prevents Lys27 from interacting with 18C6 to any significant extent, and thus the SNAPP distributions do not change upon mutation of this residue. Importantly, these results also confirm that specific interactions originating from the solution phase structures dictate the experimental observations that are ultimately recorded in the gas phase for SNAPP experiments. If the protein were subjected to denaturation during the ionization process, then all lysine residues would be expected to contribute to the SNAPP distribution. The results for the K29N mutant are shown in Figure 2d. Mutation of Lys29 reduces the number of 18C6 adducts in the SNAPP distributions to a lesser degree than that observed for Lys6 and Lys11. Inspection of the crystal structure reveals that Lys29 forms a single hydrogen bond with the backbone carbonyl of Glu16. NMR reveals that the strength of this hydrogen bond varies, with the heteroatom/heteroatom bond distance ranging from 2.4 Å to 3.5 Å. These results suggest that the presence of a single hydrogen bond will interfere with 18C6 attachment, but not prevent it. As a consequence, structural changes disrupting even a single hydrogen bond involving a lysine residue should be detectable by SNAPP. This is a very high level of chemical sensitivity, which may not be easily achieved by other methods and may help explain why the bovine and Arabidopsis structures are distinguishable by SNAPP. The results for the K33N mutant are shown in Figure 3a. This mutation again leads to a reduction in attachment by 18C6. The magnitude of the change is comparable to that observed for the K29N mutant. Inspection of the crystal structure again reveals a single hydrogen bond (this time to the backbone of Thr14). However, inspection of the NMR structures suggests that Lys33 may be in a dynamic part of the protein because the nearest heteroatom/heteroatom length varies substantially (from 3.2 Å to 5.3 Å), with different heteroatom pairs. These results again are 3850

Analytical Chemistry, Vol. 80, No. 10, May 15, 2008

Figure 3. SNAPP distributions for the wild-type (gray) and mutant (white) proteins. Mutation of Lys33 yields a slight reduction in 18C6 binding. The results for Lys48 suggest less binding than would be predicted by inspection of known structures. Lys63 is also not a good binding site for 18C6, which would not be predicted by comparison with known structures. The results may be related to the biological functionality of these sites.

consistent with a mechanistic picture in which the chemical availability of each lysine side chain determines the amount of 18C6 attachment that is observed. In terms of interfering interactions, salt bridges appear to interfere much more than hydrogen bonds. Figure 3b contains the results for the K48N mutant. Inspection of the results reveals that the degree of 18C6 attachment is reduced to a small extent. The SNAPP results therefore suggest that Lys48 participates in intramolecular hydrogen bonding. The change in the SNAPP distributions suggests interactions that are more interfering than the level observed for Lys33 and Lys29. In the crystal structure, Lys48 is hydrogen bonded to the backbone of Ala46. However, Lys48 is typically free in the NMR solution phase structure with an average heteroatom separation of over 5 Å. The degree of hydrogen bonding observed in the NMR structures is not consistent with the degree of change in 18C6 attachment shown in Figure 3b, which would be expected to be much greater for a noninteracting residue. One possible explanation is that these interactions may be highly dynamic in nature, preventing observation by NMR. The results for the K63N mutant are shown in Figure 3c. The SNAPP distribution does not change dramatically with the mutation of Lys63; furthermore, the number of adducts appears to increase for both the open and closed structures. This unexpected result was verified on different days with different samples. Since the K63N mutant contains fewer binding sites than the wild-type protein, an increase in 18C6 adducts can only occur if the remaining lysine residues become more available with elimination of Lys63. This would most likely occur due to a structural shift; however, the C-terminal tail of ubiquitin has an ill-defined structure which may make structural rearrangements

Figure 4. (a) The solvent accessibility (white) is shown in comparison with the total change in SNAPP distribution for each mutation (gray). This number represents the sum of the differences inside the standard deviations between the mutant and wild type for both charge states. There is no obvious correlation. (b) The total change in the SNAPP distribution for the native structure (gray) and the A-state (white) are shown for comparison. Substantially different results are obtained, indicating significant structural rearrangement.

difficult to detect. In any case, Lys63 is not a good binding site for 18C6, which would not be predicted by comparison with known structures. Circular Dichroism. CD spectra were acquired for the wildtype protein and all mutants to determine if the mutations caused any major structural changes. As seen in the Supporting Information, all spectra are identical within the error of the measurement. These results confirm that the major features of the protein structure, i.e., R-helices and β-sheets, have not been significantly perturbed by the mutation of single lysine residues. These results are in agreement with previous experiments where all charged residues were modified simultaneously without significant impact on the protein structure.29 These results also agree with the observation that the structure of ubiquitin is known to be particularly stable. CD is not likely to detect minor structural changes; therefore, we cannot exclude the possibility that some slight changes may occur with mutation of individual lysine residues. This may be the case for Lys63. However, the CD results do support the conclusion that the dramatic changes in the SNAPP distributions shown in Figure 2b are not the result of a significant structural rearrangement and are therefore best explained in terms of changes related to the availability of Lys11. Surface Availability. It is tempting to assume that one of the primary factors controlling the 18C6 interaction with lysine is the surface availability of the lysine residue. A plot of surface availability versus change in SNAPP distribution is shown in Figure 4a. There is very little correlation between surface availability and attachment by 18C6. Exposed residues can still form salt bridges or other interfering intramolecular interactions, preventing attachment by 18C6. Most lysine residues are not substantially buried in ubiquitin, or other proteins. If a residue is inaccessible, such as Lys27, it is expected that there will be a correlation with reduced attachment of 18C6 due to steric effects, i.e., there will be insufficient room for the adduct to interact with a buried lysine residue. In addition, most buried lysine residues will be bonded intramolecularly via salt bridges and hydrogen

bonds that will further preclude attachment of 18C6. However, exposed residues are not necessarily good binding sites for 18C6, making solvent accessibility a poor overall predictor for attachment by 18C6. Implications for Biological Function. Monoubiquitination occurs by attachment of the C-terminus of ubiquitin to a lysine residue of the target protein via an isopeptide bond. Polyubiquitin most frequently forms by attachment of sequential additional ubiquitins to Lys48, although results also indicate that other lysines may be potential polyubiquitin linkage sites that signal for different processes. A recent comprehensive examination found that ubiquitination occurs to some extent at all lysine residues in ubiquitin; however, Lys48, Lys63, and Lys11 were found to be ubiquitinated much more frequently than the remaining lysine residues.33 Interestingly, these are the three residues where structural agreement between X-ray, NMR, and SNAPP is not achieved. This may suggest that these residues are capable of responding to environmental conditions, enabling them to fulfill different biological functions. Furthermore, the SNAPP results indicate that Lys11 participates in a dynamic breathing motion of the protein. This motion may be required for polyubiquitination to occur at this site. In fact, since the C-terminal portion of the structure is known to be dynamic in nature, which is the side of the protein containing Lys63 and Lys48, all of the important sites for polyubiquitination may be located in dynamic portions of the protein. Ubiquitin A-State. In addition to the natively folded state, ubiquitin exhibits a partially denatured structure in the presence of acid and organic cosolvent termed the “A-state”. The A-state has been studied by MS34,35 and NMR.36 Both techniques reveal that the N-terminal domain of ubiquitin remains nativelike in the A-state, while the remainder of the protein is subject to structural rearrangement. NMR studies suggest that the C-terminal domain becomes R-helical, yet there is not evidence for this in HDX-MS experiments. We examined all of our mutants under conditions that favor the A-state. The results are summarized in Figure 4b, where the magnitude of the change in SNAPP distribution relative to the wild-type protein is shown as a function of lysine mutation for both the native and A-state structures. Both the order and magnitude of the SNAPP shifts are different for the A-state relative to the natively folded protein. For example, Lys6 has shifted from being the second most favorable attachment site to the fifth. Likewise, Lys27 can attach 18C6 in the A-state, which was not observed in the native structure. In general, the results in Figure 4b are in agreement with previous experiments. However, the reduced availability of Lys6 suggests that the N-terminal domain is perturbed, even if the backbone remains primarily nativelike. Similarly, the lack of 18C6 attachment to Lys63 suggests that the C-terminal domain is sufficiently structured to shield this residue. If the C-terminus were R-helical, as suggested by NMR, then some 18C6 attachment to Lys63 would be expected. However, if Lys63 (33) Peng, J. M.; Schwartz, D.; Elias, J. E.; Thoreen, C. C.; Cheng, D. M.; Marsischky, G.; Roelofs, J.; Finley, D.; Gygi, S. P. Nat. Biotechnol. 2003, 21, 921–926. (34) Hoerner, J. K.; Xiao, H.; Kaltashov, I. A. Biochemistry 2005, 44, 11286– 11294. (35) Mohimen, A.; Dobo, A.; Hoerner, J. K.; Kaltashov, I. A. Anal. Chem. 2003, 75, 4139–4147. (36) Brutscher, B.; Bruschweiler, R.; Ernst, R. R. Biochemistry 1997, 36, 13043– 13053.

Analytical Chemistry, Vol. 80, No. 10, May 15, 2008

3851

were buried in a largely disordered, localized molten globule, 18C6 attachment may be prevented without providing significant shielding from HDX (which is observed to occur freely by MS). The fact that shifts in the SNAPP distributions for the A-state are different from those for the folded structure also supports the conclusion that SNAPP extracts information about the solution phase structures of proteins. CONCLUSION Examination of the results reveals a hierarchy of interactions that dictate attachment of 18C6. The order for decreasing 18C6 attachment goes as follows: noninteracting lysines < lysines with hydrogen bonds < lysines in salt bridges. There is some flexibility within each category, for example, a strong hydrogen bond will interfere more than a weak hydrogen bond. Additionally, the degree to which a residue is buried can influence its chemical availability. Although this situation does not frequently arise for lysine residues, the implications must still be considered. Surface availability (which can be measured as solvent accessibility) does not ensure that a strong interaction with 18C6 will be possible; however, lack of surface availability will necessarily interfere with 18C6 attachment. Simultaneously, most buried lysines will be occupied by salt bridges or hydrogen bonds, providing further interference as in the case of Lys27 for natively folded ubiquitin. It is also clear from the present results that SNAPP distributions represent statistically averaged information from a variety of reporter binding sites. For example, the maximum number of adducts for these experiments is five 18C6s, yet contributions to the SNAPP distributions come from at least six lysine residues. This suggests that 18C6 is not attached to the same lysine residues

3852

Analytical Chemistry, Vol. 80, No. 10, May 15, 2008

for each individual ion, and also may indicate that some residues other than lysine may be small contributors to the SNAPP distribution. The statistical averaging and probing of information from multiple sites is likely responsible for the high sensitivity of the method for detecting structural rearrangements. Disruption of even single hydrogen bonds can be observed and used to monitor different dynamic states of proteins or the influence of external factors such as ligand binding. Close inspection of the structure of ubiquitin reveals a paradoxical truth: the overall structure is highly stable, yet dynamic fluctuations that may provide the key to its functionality exist in several sections of the molecule. The combination of sitedirected mutagenesis and SNAPP-MS reveals a dynamic breathing motion that significantly influences the N-terminal portion of ubiquitin in the vicinity of Lys11, as previously suggested by NMR experiments. Similarly, the results indicate that Lys48 and Lys63 exhibit unpredicted behavior, which may be related to their role as polyubiquitin binding sites. ACKNOWLEDGMENT The authors wish to thank J. L. Beauchamp for fruitful discussions and the University of California, Riverside and the American Society for Mass Spectrometry for funding. SUPPORTING INFORMATION AVAILABLE Additional information as noted in text. This material is available free of charge via the Internet at http://pubs.acs.org. Received for review January 24, 2008. Accepted March 11, 2008. AC800176U