Nonuniform Isotope Patterns Produced by Collision-Induced

pH and pD values were measured with an AB15 pH meter (Fisher Scientific, Nepean, ON). Reported ...... Harry J. Sterling and Evan R. Williams. Analytic...
0 downloads 0 Views 605KB Size
Anal. Chem. 2008, 80, 4078–4086

Nonuniform Isotope Patterns Produced by Collision-Induced Dissociation of Homogeneously Labeled Ubiquitin: Implications for Spatially Resolved Hydrogen/Deuterium Exchange ESI-MS Studies Peter L. Ferguson and Lars Konermann* Department of Chemistry, The University of Western Ontario, London, Ontario, N6A 5B7, Canada There is an ongoing debate whether collision-induced dissociation (CID) of electrosprayed proteins after solution-phase hydrogen/deuterium exchange (HDX) is a viable approach for determining spatially resolved deuteration patterns. This work explores the use of two methods, source-CID and hexapole tandem mass spectrometry (MS/MS) on a quadrupole time-of-flight (Q-TOF) mass spectrometer, for measuring the fragment deuteration levels of regioselectively labeled ubiquitin. Both methods reveal that b-ions exhibit HDX levels significantly below that of the intact protein, whereas several y′′ fragments are labeled to a much greater extent. These results are consistent with earlier source-CID data (Akashi, S.; Naito, Y.; Takio, K. Anal. Chem. 1999, 71, 4974–4980). However, the measured b-ion deuteration levels are in disagreement with the known solution-phase behavior of ubiquitin. Partial agreement is observed for y′′-ions. Control experiments on homogeneously labeled ubiquitin (having the same average deuteration level at every exchangeable site) result in highly nonuniform fragment HDX levels. In particular, b-ions exhibit deuteration levels significantly below that of intact ubiquitin, thereby mimicking the behavior seen for the regioselectively labeled protein. This effect is likely caused by isotope fractionation during collisional activation, facilitated by the high mobility of charge carriers (scrambling) in the gas phase. The observation that the b-ion labeling behavior is largely independent of the spatial isotope distribution within solution-phase ubiquitin invalidates these ions as reporters of the protein deuteration pattern. This work questions the common practice of interpreting any nonuniformities in fragment deuteration as being indicative of regioselective solution-phase labeling. Artifactual deuterium enrichment or depletion during collisional activation may have contributed to the current lack of consensus as to whether HDX/CID represents a potentially viable tool for measuring solution-phase deuteration patterns. Hydrogen/deuterium exchange (HDX) techniques with mass spectrometry (MS) detection have become a popular approach * To whom correspondence should be addressed. E-mail: [email protected].

4078

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

for exploring the structure and dynamics of proteins in solution.1–5 Upon incubation of a protein in a D2O-containing solvent, hydrogens along the amide backbone and in several side-chain functional groups can be replaced with deuterium. For natively folded proteins the rates of HDX are drastically lowered at sites that are protected either sterically or by hydrogen bonding. Exchange of these hydrogens is mediated by the conformational dynamics of the protein, i.e., by short-lived local or global unfolding/refolding events that induce the disruption of hydrogen bonds and provide transient solvent access.6 Thus, the HDX behavior of specific residues reflects their structural environment and conformational flexibility. In the most common version of these HDX experiments,4 isotope exchange is terminated by acid quenching and rapid cooling, followed by limited proteolysis with pepsin. Subsequent liquid chromatography and electrospray MS (LC-ESI-MS) allows deuteration patterns to be determined by measuring the mass shifts of individual proteolytic fragments as a function of exchange time. The spatial resolution of this approach is on the order of a few residues.7 Side chains undergo rapid backexchange in the protiated LC mobile phase, such that only deuterons incorporated into backbone amide groups can normally be retained. Amide back-exchange takes place more slowly; however, it also has to be taken into account for the data analysis, and it represents a factor that can lead to complications under some conditions.8 Numerous recent studies9–33 have explored the use of gasphase fragmentation approaches instead of (or in combination with) the classical proteolytic digestion/LC methodology for spatially resolved HDX experiments. A few of these were based on electron capture dissociation,30,33 providing some very promising results.29,32 The overwhelming majority of previous work in this area, however, has been based on collision-induced dissociation (HDX/CID). CID represents by far the most widely available (1) Kaltashov, I. A.; Eyles, S. J. Mass Spectrometry in Biophysics; John Wiley and Sons, Inc.: Hoboken, NJ, 2005. (2) Wales, T. E.; Engen, J. R. Mass Spectrom. Rev. 2006, 25, 158–170. (3) Englander, S. W. J. Am. Soc. Mass Spectrom. 2006, 17, 1481–1489. (4) Smith, D. L.; Deng, Y.; Zhang, Z. J. Mass Spectrom. 1997, 32, 135–146. (5) Busenlehner, L. S.; Armstrong, R. N. Arch. Biochem. Biophys. 2005, 433, 34–46. (6) Krishna, M. M. G.; Hoang, L.; Lin, Y.; Englander, S. W. Methods 2004, 34, 51–64. (7) Engen, J. R.; Smith, D. L. Anal. Chem. 2001, 73, 256A–265A 10.1021/ac8001963 CCC: $40.75  2008 American Chemical Society Published on Web 05/07/2008

ion activation method on commercial instruments. Among these studies, two types of experiments can be distinguished:31 those employing nonspecific fragmentation caused by an elevated declustering potential in the ion sampling interface (HDX/sourceCID) and those using tandem mass spectrometry with specific precursor ion selection (HDX/MS/MS). The considerable interest in the combination of solution-phase HDX with gas-phase fragmentation is based on the fact that this approach could open up new avenues for exploring biomolecular folding, structure, dynamics, and function.34 Of particular interest in this context is the application of HDX-CID to intact proteins.13,24 A conceptual problem with all HDX/CID studies is that backbone cleavage by low-energy collisions involves the rapid intramolecular migration of protons and/or deuterons, as stipulated by the mobile proton model.35–39 The resulting H/D scrambling would be expected to randomize the isotope exchange pattern originally imprinted upon the protein during solution-phase labeling.10 However, it may still be feasible to extract useful sitespecific information if H/D scrambling is restricted to within a few residues. HDX/CID experiments carried out over the past (8) Wu, Y.; Kaveti, S.; Engen, J. R. Anal. Chem. 2006, 78, 1719–1723. (9) Kaltashov, I. A.; Eyles, S. J. J. Mass Spectrom. 2002, 37, 557–565. (10) Johnson, R. S.; Krylov, D.; Walsh, K. A. J. Mass Spectrom. 1995, 30, 386– 387. (11) Jørgensen, T. J. D.; Ga˚rdsvoll, H.; Ploug, M.; Roepstorff, P. J. Am. Chem. Soc. 2005, 127, 2785–2793. (12) McLafferty, F. W.; Guan, Z.; Haupts, U.; Wood, T. D.; Kelleher, N. L. J. Am. Chem. Soc. 1998, 120, 4732–4740. (13) Hoerner, J. K.; Xiao, H.; Dobo, A.; Kaltashov, I. A. J. Am. Chem. Soc. 2004, 126, 7709–7717. (14) Ferguson, P. L.; Pan, J.; Wilson, D. J.; Dempsey, B.; Lajoie, G.; Shilton, B.; Konermann, L. Anal. Chem. 2007, 79, 153–160. (15) Deng, Y.; Pan, H.; Smith, D. L. J. Am. Chem. Soc. 1999, 121, 1966–1967. (16) Kim, M.-Y.; Maier, C. S.; Reed, D. J.; Deinzer, M. L. J. Am. Chem. Soc. 2001, 123, 9860–9866. (17) Demmers, J. A. A.; Haverkamp, J.; Heck, A. J. R.; Koeppe, R. E.; Killian, A. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 3189–3194. (18) Demmers, J. A. A.; Rijkers, D. T. S.; Haverkamp, J.; Killian, J. A.; Heck, A. J. R. J. Am. Chem. Soc. 2002, 124, 11191–11198. (19) Anderegg, R. J.; Wagner, D. S.; Stevenson, C. L.; Borchardt, R. T. J. Am. Soc. Mass Spectrom. 1994, 5, 425–433. (20) Kraus, M.; Bienert, M.; Krause, E. Rapid Commun. Mass Spectrom. 2003, 17, 222–228. (21) Akashi, S.; Takio, K. J. Am. Soc. Mass Spectrom. 2001, 12, 1247–1253. (22) Akashi, S.; Naito, Y.; Takio, K. Anal. Chem. 1999, 71, 4974–4980. (23) Akashi, S.; Takio, K. Protein Sci. 2000, 9, 2497–2505. (24) Eyles, S. J.; Speir, J. P.; Kruppa, G. H.; Gierasch, L. M.; Kaltashov, I. A. J. Am. Chem. Soc. 2000, 122, 495–500. (25) Eyles, S. J.; Dresch, T.; Gierasch, L. M.; Kaltashov, I. A. J. Mass Spectrom. 1999, 34, 1289–1295. (26) Xiao, H.; Kaltashov, I. A. J. Am. Soc. Mass Spectrom. 2005, 16, 869–879. (27) Hoerner, J. K.; Xiao, H.; Kaltashov, I. A. Biochemistry 2005, 44, 11286– 11294. (28) Cai, X.; Dass, C. Rapid Commun. Mass Spectrom. 2005, 19, 1–8. (29) Charlebois, J. P.; Patrie, S. M.; Kelleher, N. L. Anal. Chem. 2003, 75, 3263– 3266. (30) Kweon, H. K.; Hakansson, K. Analyst 2006, 131, 275–280. (31) Hagman, C.; Hakansson, P.; Buijs, J.; Hakansson, K. J. Am. Soc. Mass Spectrom. 2004, 15, 639–646. (32) Rand, K. D.; Adams, C. M.; Zubarev, R. A.; Jorgensen, T. J. D. J. Am. Chem. Soc. 2008, 130, 1341–1349. (33) Hagman, C.; Tsybin, Y. O.; Hakansson, P. Rapid Commun. Mass Spectrom. 2006, 20, 661–665. (34) Konermann, L.; Simmons, D. A. Mass Spectrom. Rev. 2003, 22, 1–26. (35) Dongre´, A. R.; Jones, J. L.; Somogyi, A´.; Wysocki, V. H. J. Am. Chem. Soc. 1996, 118, 8365–8374. (36) Tang, X.-J.; Thibault, P.; Boyd, R. K. Anal. Chem. 1993, 65, 2824–2834. (37) Harrison, A. G.; Yalcin, T. Int. J. Mass Spectrom. 1997, 165/166, 339–347. (38) Paizs, B.; Suhai, S. Mass Spectrom. Rev. 2005, 24, 508–548. (39) Summerfield, S. G.; Gaskell, S. J. Int. J. Mass Spectrom. 1997, 165/166, 509–521.

Figure 1. Amino acid sequence of ubiquitin (PDB code 1ubq ) (ref 42), showing the locations of major b- and y′′-ion cleavage sites and secondary structure elements. Previous NMR-based experiments (ref 43) reveal the presence of 41 slowly exchanging amides that closely match the H-bond pattern seen in the X-ray structure of the protein (ref 42). The corresponding 41 residues are highlighted in bold italics.

few years have provided conflicting results. Complete scrambling was observed in several HDX/MS/MS studies,10–14 whereas others reported the apparent retention of spatial deuteration patterns to various degrees.15–19,28 A considerable number of experiments employing HDX/source-CID reported very little apparent scrambling.13,20–27 It has been suggested that different types of fragment ions may not be equally subjected to scrambling15,16,31 and that intermolecular H/D migration is most pronounced when using very slow heating methods such as SORICID (sustained off-resonance irradiation collision-induced dissociation).9,12,13 It is unfortunate that those previous investigations could not resolve the ongoing debate whether HDX/CID represents a potentially viable approach for measuring spatially resolved deuteration patterns. This lack of consensus is partly due to the fact that those studies involved different proteins or peptides, different types of mass spectrometers (Q-TOFs (quadrupole timeof-flight), ion traps, triple quadrupoles, as well as FTICR (Fourier transform ion cyclotron resonance) instruments), and different fragmentation approaches (source-CID vs MS/MS). Another noteworthy aspect is the lack of stringent control experiments in previous studies that reported the apparently successful application of HDX/CID. In those studies it is often assumed that complete scrambling has to result in identical fragment deuteration levels, matching that of the intact protein. Thus, any deviations from such a uniform behavior are considered to be the hallmark of incomplete scrambling, implying the retention of “real” information.18,40 The validity of this basic premise has been questioned in a recent study from our laboratory, which suggests that HDX/CID of peptides may result in nonuniform deuteration patterns, even under conditions of complete scrambling.14 The aim of the current work is to provide additional insights into the feasibility of HDX/CID as a tool for studying protein structure and dynamics. As a model system we focus on ubiquitin. The native structure of this protein is remarkably resistant to chemical and pH-induced denaturation.41 It is composed of five β-strands, a three-turn R-helix, a single turn of 310 helix, and six short loops (Figure 1). The C-terminal tail (residues 71-76) is highly flexible and provides no protection from amide exchange.42,43 (40) Rand, K. D.; Jorgensen, T. J. D. Anal. Chem. 2007, 79, 8686–8693. (41) Katta, V.; Chait, B. T. J. Am. Chem. Soc. 1993, 115, 6317–6321. (42) Vijay-Kumar, S.; Bugg, C. E.; Cook, W. J. J. Mol. Biol. 1987, 194, 531– 544.

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

4079

The choice of ubiquitin is based on the fact that its structure and dynamics have been exceptionally well characterized by X-ray crystallography,42 molecular dynamics simulations,44 and NMR spectroscopy.43–46 In addition, Akashi et al. have previously conducted detailed ubiquitin HDX/source-CID experiments on an FTICR mass spectrometer.22 In order to determine whether the outcome of HDX/CID studies depends on the instrumentation used, we employ source-CID and hexapole MS/MS on a Q-TOF and compare the results to the earlier FTICR data of Akashi et al.22 It is demonstrated that unexpected nonuniform isotope distributions, previously observed for small peptides after homogeneous labeling,14 also occur for a relatively large system such as ubiquitin. The implications of this and other phenomena (such as scrambling) for the interpretation of HDX/CID data are discussed. EXPERIMENTAL SECTION Chemicals. Bovine ubiquitin (MW 8565 Da) and porcine pepsin were purchased from Sigma Chemical Co. (St. Louis, MO) and used without further purification. Deuterium oxide (99.9% D) was from Cambridge Isotope Laboratories (Andover, MA). Acetic acid was obtained from Caledon Laboratories (Georgetown, ON), and deuterated acetic acid-d was from Isotec Inc. (Miamisburg, OH). pH and pD values were measured with an AB15 pH meter (Fisher Scientific, Nepean, ON). Reported values were corrected for isotope effects by using the relation pD ) pH meter reading + 0.40.47 Electrospray HDX/CID Experiments. Source-CID measurements were carried out by mixing 500 µM ubiquitin in water with D2O in a 1:9 volume ratio for a final pD of 5.5, thereby initiating isotope exchange. Alternatively, lyophilized ubiquitin dry powder was hydrated directly in D2O (pD 5.5), as described by Akashi et al.22 The two procedures resulted in data that were virtually indistinguishable (not shown). Following exposure to the deuterated solvent, the solution was continuously infused into a mixing tee (Upchurch) at 8 µL/min, where it joined the flow from a second syringe containing 100 mM acetic acid-d (90% D atoms) in 90% D2O. The resulting mixture was injected on-line into the ESI source of the mass spectrometer at a total flow rate of 16 µL/min and a final protein concentration of 25 µM. All solutions in these experiments were kept at room temperature (22 ± 1 °C). The flow delay between mixing tee and the outlet of the ESI capillary was 20 s. Mass spectra were recorded on a Q-TOF instrument (Waters Ultima API) fitted with a Z-spray ESI source that was operated at a voltage of 3 kV, using dry nitrogen produced by a generator (Whatman) both as nebulizer (120 °C) and as cone gas (80 °C). Normally the instrument is operated at a declustering voltage (potential difference between sampling cone and rf 1 lens) of ca. 20 V, resulting in negligible levels of fragmentation in the ion sampling interface. Source-CID was induced by increasing this potential difference to 110 V. For early HDX data (up to 30 min after initiation of labeling) total ion chromatogram (TIC) sections (43) Pan, Y.; Briggs, M. S. Biochemistry 1992, 31, 11405–11412. (44) Lindorff-Larsen, K.; Best, R. B.; DePristo, M. A.; Dobson, C. M.; Vendruscolo, M. Nature 2005, 433, 128–132. (45) Zech, S. G.; Wand, A. J.; McDermott, A. E. J. Am. Chem. Soc. 2005, 127, 8618–8626. (46) Ferrage, F.; Pelupessy, P.; Cowburn, D. A.; Bodenhausen, G. J. Am. Chem. Soc. 2006, 128, 11072–11078. (47) Glasoe, P. K.; Long, F. A. J. Am. Chem. Soc. 1960, 64, 188–190.

4080

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

of 2 min centered around the time points of interest were combined to yield spectra. At later times, 5 min sections were used to increase the signal-to-noise ratio. For HDX/MS/MS studies, direct fragmentation of selected ubiquitin ions during on-line labeling was not practical due to the relatively low signal intensities. Instead, an off-line strategy was used. An amount of 300 µL of 500 µM ubiquitin stock was mixed with 2.7 mL of D2O and incubated at room temperature (pD 5.5). At designated time points 200 µL aliquots were removed, mixed with an equal volume of ice-cold 100 mM acetic acid-d (90% D) in 90% D2O, and flash frozen in liquid nitrogen prior to storage at -80 °C. The ubiquitin samples (now at 25 µM) were thawed on ice and directly infused into the mass spectrometer at 5 µL/min, while cooling the solution with ice packs. The declustering voltage used was 10 V to eliminate source activation. Fragmentation of the [M + 9H]9+ precursor in the hexapole collision cell was carried out at a collision voltage of 45 V. Data collection times of 20 min were combined to obtain spectra. All spectra shown represent unprocessed data without any smoothing. Homogeneous Labeling of Ubiquitin. Homogeneously labeled ubiquitin was generated by incubation at 50 °C for 3 days at pD 5.5 in solvent mixtures containing a well-defined percentage of D2O. Control experiments confirmed that labeling under these conditions had indeed gone to completion, i.e., the deuteration level of the protein did not undergo any additional change when the incubation period was extended further. Three different D2O percentages were tested, 60%, 75%, and 90%. The resulting ubiquitin deuteration levels were 0.5-2.5% lower than expected based on the nominal D content of the solvent. One factor contributing to these small discrepancies is isotope back-exchange in the ion source of the mass spectrometer.41,48 Also, a slightly elevated H2O concentration in the commercially obtained D2O (nominally 99.9% D) cannot be ruled out, as well as small degree of H2O contamination during the 3 day water bath heating period, and residual moisture in the lyophilized protein powder. In the subsequent text we will refer to the actually measured deuteration levels for homogeneously labeled ubiquitin, not to the nominal isotope composition of the solvent. Source-CID and MS/MS studies on homogeneously labeled protein were carried out as described above for the kinetic experiments, using a final protein concentration of 25 µM. Pepsin digestion of homogeneously labeled ubiquitin was carried out by incubating the protein at 10 mg/mL (1.2 mM) with 0.2 mg/mL pepsin in 100 mM acetic acid for 1 min at 0 °C, followed by direct infusion into the mass spectrometer using a syringe that was kept chilled by ice packs. Digestion and subsequent infusion were carried out using solvent systems that matched the D2O percentage used for labeling. The high protein concentration in these experiments was necessitated by the considerable stability of ubiquitin toward acid-induced unfolding41 which makes it highly resistant against pepsin digestion, causing poor proteolysis yields. Data Analysis. Peptide deuteration levels were determined based on a comparison of the measured data with modeled isotope distributions as described previously.14 Briefly, the modeled distributions D were calculated as a convolution, D ) NB, where (48) Hossain, B. M.; Simmons, D. A.; Konermann, L. Can. J. Chem. 2005, 83, 1953–1960.

Figure 2. (A) ESI mass spectrum of 25 µM nondeuterated bovine ubiquitin in 50 mM acetic acid. The inset shows an expanded view of the 9+ peak. (B) Source-CID spectrum of ubiquitin recorded after 10 min of HDX at pD 5.5, followed by 20 s of acid quenching in 50 mM acetic acid. Some of the more prominent fragment ions are labeled.

N is the natural isotope distribution of the undeuterated species (ProteinProspector, UCSF) and B is a binomial distribution that contains the deuteration level p (0 e p e 100%) as the only adjustable parameter. The value of p for each ion was determined from the simulated distribution D that resulted in the best fit to the experimental data. Error bars were determined as described in ref 14. RESULTS AND DISCUSSION CID Analysis of Regioselectively Labeled Ubiquitin. The time dependence of ubiquitin HDX in 90% D2O (pD 5.5, 22 °C) was monitored on a Q-TOF instrument. After isotope labeling the sample was quenched by mixing with acetic acid, thereby closely matching the conditions previously employed by Akashi et al.22 The ESI mass spectrum obtained for an unlabeled control shows a charge state distribution that extends from 5+ to 10+ (Figure 2A). In contrast, spectra recorded under source-CID conditions are shifted to lower charge states (Figure 2B). This change is attributed to preferential fragmentation of higher protonation states, caused by the elevated collision energies experienced by the latter.49 Close examination of the spectrum in Figure 2B reveals a host of CID fragments, the most intense of which were selected for analysis of their deuteration behavior. The cleavage (49) Thomson, B. A. J. Am. Soc. Mass Spectrom. 1997, 8, 1053–1058.

positions of these fragments are indicated in Figure 1, and some of the more prominent signals are labeled in Figure 2B. A further enhancement of the fragment ion yield could be achieved by operating the mass spectrometer at even higher declustering voltages (data not shown). However, the considerable spectral complexity and signal overlap under those conditions prompted us not to employ those harsher settings for the experiments discussed below. Figure 3 shows HDX/source-CID data for selected fragment ions, generated after labeling for 5 min (A-D), 120 min (E-H), and 300 min (I-L). Also shown in each panel is the modeled isotope distribution (solid symbols) that best matches the corresponding experimental data. The deuteration levels obtained in this way are plotted as a function of time in Figure 4A. Results for y′′-ions are depicted in blue, and those for b-ions are in red. Black symbols represent the intact protein. Figure 4B shows the results of measurements analogous to those in Figure 4A, except that the fragment ions were generated by selected fragmentation (HDX/MS/MS) of the [ubiquitin + 9H]9+ precursor. Although many of the observed CID products are common to both types of experiment, a few variations in the fragmentation patterns are apparent. For example, weak y′′122+ and y′′142+ signals after HDX/ MS/MS prevented the determination of the corresponding deuteration levels for the data set in Figure 4B. Overall, there is a striking difference between the labeling kinetics for b- and y′′ions in Figure 4, parts A and B. C-Terminal fragments track at deuteration levels either much higher (y′′122+ and y′′142+) or relatively close (y′′182+, y′′243+) to the intact protein. In contrast, N-terminal fragments (b-ions) exhibit a degree of labeling considerably lower than intact ubiquitin, with b3+ being the least deuterated species for all time points studied. The deuteration levels of intact ubiquitin for t ) 5 min from Figure 4, parts A and B, are 66.8% ± 0.5% and 68.0% ± 0.5%, respectively. These values are in quite good agreement with the presence of 41 slowly exchanging amide hydrogens43 (Figure 1), based on which a burst-phase amplitude of 65.8% would be expected. The latter value can be calculated as 0.9(72 side-chain H + 31 unprotected amide H + 9 charge carrier H)/(144 total H + 9 charge carrier H), where the factor of 0.9 represents the D2O mole fraction of the solvent. One important goal of this study is to investigate whether spatially resolved HDX experiments yield different results when using source-CID or MS/MS. As noted above, there are slight differences in sequence coverage. Also, the overall spread in deuteration level among individual fragments is somewhat more pronounced for the source-CID data. For example, y′′182+ is 8.5% more deuterated than b3+ at t ) 240 min in Figure 4A, whereas the corresponding difference in Figure 4B is 6%. These minor discrepancies may be due to the different sample handling (see the Experimental Section). Most importantly, however, the overall deuteration trends are very similar for the data obtained by HDX/ source-CID and HDX/MS/MS. For instance, both approaches reveal that y′′182+ closely follows the deuteration kinetics of intact ubiquitin, whereas the HDX levels of b2+, b3+, and b5+ are consistently much lower. For the system studied here, it is concluded that the choice of fragmentation method (source-CID vs MS/MS on a Q-TOF) is not a major determinant of the resulting fragment deuteration levels. Our results also agree with the HDX/ Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

4081

Figure 3. Source-CID data, depicting the HDX properties of selected ubiquitin fragment ions (b5+, b162+, y122+, and y142+) for different labeling times: (A-D) 5 min; (E-H) 120 min; (I-L) 300 min. The fragment ion labeling patterns of homogeneously 87.5% deuterated ubiquitin are shown in panels M-P. Solid circles represent fitted isotope distributions, corresponding to the percent deuteration level indicated in each panel.

source-CID data obtained by Akashi et al. on an FTICR instrument employing a capillary-skimmer-type interface.22 Taken together, these observations demonstrate that the HDX properties of CID fragments are remarkably robust, showing little dependence on the design of the mass analyzer and ion sampling interface. It is clear that parameters such as excessive activation during declustering32 or the very slow time scale of SORI-CID13 can have an effect on the measured deuteration levels. Nonetheless, the results presented here imply that differences in instrumentation are a secondary factor when deciding on the general feasibility of spatially resolved HDX/CID experiments. Homogeneously Labeled Ubiquitin. Having observed different HDX characteristics for various fragment ions, it is tempting to proceed immediately to an interpretation of these data in a structural context. Prior to taking any such steps, however, it is prudent to carry out control experiments in order to determine whether any unexpected factors have to be considered for HDX/ CID analyses of ubiquitin. We attempted to produce ubiquitin having the same average deuteration level at every exchangeable site by exposing the protein to well-defined D2O/H2O mixtures at elevated temperature (50 °C) for an extended period of time (3 days). We will refer to these samples as ”homogeneously” labeled, in contrast to the ”regioselective” labeling used for Figure 4, parts A and B. As expected, HDX under homogeneous labeling conditions results in deuteration levels for the intact protein that closely match the D2O mole fraction of the solvent (see the Experimental Section). Before proceeding further, we have to address a side aspect related to the number of exchangeable sites in ubiquitin. 4082

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

Normally only hydrogens in N-H, O-H, and S-H bonds are labile, whereas those in C-H bonds are not exchangeable.22,50 The Cε1-H in the imidazole side chain of His represents a possible exception. This hydrogen has been shown to be exchangeable in some cases due to its unusual acidity.51 Peptic digestion of ubiquitin generates the peptide HLVL (residues 68-71, Figure 5A), which contains the only His residue in the protein. On the basis of the number of hydrogens in O-H and N-H bonds, one would expect a mass increase of eight units upon complete deuteration of this peptide. This expectation is consistent with data obtained after incubating the digestion product in 100% D2O at room temperature for 40 min (Figure 5, parts B and C). In contrast, peptic digestion of ubiquitin after homogeneous labeling in 100% D2O (3 days at 50 °C) results in a HLVL peptide that is shifted by nine mass units (Figure 5D). Clearly, the Cε1-H of His68 represents the most likely candidate for the additional exchangeable site.51 On the basis of these findings our analyses assumed 145 exchangeable hydrogens in ubiquitin under homogeneous labeling conditions. These include 72 backbone N-H, 72 hydrogens in side-chain O-H and N-H functional groups,41 and the Cε1-H of His68. Under regioselective labeling conditions, characterized by much shorter incubation times and lower temperature, Cε1-H was not considered to be exchangeable (144 exchangeable hydrogens).41 Several NMR studies have suggested that long-term incubation of proteins in D2O/H2O mixtures may not always result in truly homogeneous amide labeling. Although the interpretation of those (50) Englander, S. W.; Kallenbach, N. R. Q. Rev. Biophys. 1984, 16, 521–655. (51) Niimura, N.; Chatake, T.; Kurihara, K.; Maeda, M. Cell Biochem. Biophys. 2004, 40, 351–369.

Figure 4. (A) Deuteration kinetics of ubiquitin (black symbols, dotted line) and several source-CID fragments (solid lines). Data for Cterminal fragments (y′′-ions) (ref 67) are shown in blue; those for N-terminal fragments (b-ions) are in red. (B) Same as in panel A, except that the fragment ions were produced by MS/MS of the [ubiquitin + 9H]9+ precursor. Note that the time scale for (A) and (B) is logarithmic. (C) Source-CID data for homogeneously labeled ubiquitin at three different protein deuteration levels. (D) MS/MS analysis of [ubiquitin + 9H]9+ after homogeneous labeling at three deuteration levels. The experimental error of the measured HDX levels does not exceed (2% (see Figure 3). As an example, some (2% error bars are depicted in (A). Error bars have been omitted from the other data points to prevent cluttering.

measurements is complicated by NMR-specific experimental factors,52 it appears that proteins can exhibit solution-phase isotope fractionation, where certain sites exhibit long-term deuteration levels either above or below that of the bulk solvent.53–55 Thus, one cannot necessarily assume that the ”homogeneous” labeling conditions employed here do indeed produce a perfectly uniform spatial isotope distribution. To assess the possible occurrence of solution-phase isotope fractionation, ”homogeneously” labeled ubiquitin was subjected to pepsin digestion in solutions containing the same D2O mole fraction as the original labeling buffer. The resulting peptides were analyzed immediately by ESI-MS under quenching conditions (0 °C, pH 2.8, after 1 min of pepsin digestion). In Figure 6 the peptide deuteration levels of segments 1-15, 68-71, and 70-76 are plotted versus the deuteration level of the intact protein. Data virtually indistinguishable from those depicted in Figure 6 were obtained for peptides 1-4, 4-15, 16-45, 46-67, and 46-69 (not shown). Large-scale isotope fractionation (52) (53) (54) (55)

LiWang, A. C.; Bax, A. J. Am. Chem. Soc. 1996, 118, 12864–12865. Bowers, P. M.; Klevit, R. E. Nat. Struct. Biol. 1996, 3, 522–531. Englander, S. W.; Poulsen, F. M. Biopolymers 1969, 7, 379–393. Khare, D.; Alexander, P.; Orban, J. Biochemistry 1999, 38, 3918–3925.

Figure 5. (A) Singly protonated peptic ubiquitin fragment HLVL (residues 68-71). Eight exchangeable hydrogens (those in N-H and O-H bonds) are shown in bold. The His C1-H hydrogen is highlighted by a circle. The exact location of the charge carrier (tentatively shown as being attached to the N-terminus) is irrelevant for the considerations of this work. (B) Mass spectrum of nondeuterated HLVL (singly protonated). (C) Mass spectrum measured after exposing HLVL to 100% D2O at room temperature for 40 min. (D) Mass spectrum measured after exposing ubiquitin to 100% D2O for 3 days at 50 °C, followed by pepsin digestion to generate the peptide HLVL. The fitted isotope distributions (solid circles) in (C) and (D) correspond to models with eight and nine exchangeable sites, respectively. The HDX levels determined in this way are indicated in the corresponding panels.

in solution would result in some peptides showing deuteration levels above, and others below, that of the intact protein. Evidently, this is not the case for the data in Figure 6. Instead, all peptides show deuteration levels matching that of the intact protein to within 1%. These data demonstrate that solution-phase isotope fractionation is negligible under the conditions employed here. CID Analysis of Homogeneously Labeled Ubiquitin. SourceCID data obtained for 87.5% homogeneously labeled ubiquitin are depicted in Figure 3M-P. Surprisingly, the deuteration levels of the resulting fragments exhibit considerable variations, ranging from 81.5% for b5+ to 89.5% for y′′142+. The results of source-CID experiments for three different levels of homogeneous ubiquitin Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

4083

Figure 7. Simulated deuterium distributions for homogeneously labeled ubiquitin, assuming average HDX levels of p ) 60% (A) and 90% (B). The distributions were calculated from the binomial expression B(p, n, k) ) (n!/(k!(n - k)!))pk(1 - p)n-k, with n ) 145 (number of exchangeable hydrogens in uniformly labeled ubiquitin), and k ) 0, ..., 145. Further explanations are given in the text.

Figure 6. Deuteration percentage of peptic fragments plotted vs deuteration level of homogeneously labeled ubiquitin. Panels A, B, and C correspond to residues 1-15, 68-71, and 70-76, respectively. Some representative error bars are shown.

labeling are depicted in Figure 4C. The deuteration of b-ions is 4-6% lower, and that of y′′-ions up to 2% higher, than that of the intact protein. Similar observations were made after fragmentation of the [ubiquitin + 9H]9+ precursor in MS/MS experiments (Figure 4D). The occurrence of nonuniformities in fragment deuteration levels may seem counterintuitive. On the basis of the simplistic framework used for previous HDX/CID studies, a homogeneously labeled protein would be expected to produce fragment ions having exactly the same labeling level as the intact protein.13,18 In Figure 4, parts C and D, such a behavior would result in all points falling on the black dotted line. Evidently, the experimental data do not match this expectation. We previously reported the occurrence of similar nonuniformities for much smaller peptides.14 The data presented here demonstrate that variations in the HDX levels of CID fragments also occur for homogeneously labeled proteins. As discussed in the previous section, we can exclude the possibility that the nonuniformities seen in Figure 4, parts C and D, reflect the spatial H/D distribution of the protein in solution. It is concluded that the observed effects must be caused by gas-phase processes, and we will briefly discuss three possible contributing factors. (A) CID fragments might undergo forward or backward isotope exchange in the gas phase to a different degree.31,33 Closer examination reveals that such a scenario appears unlikely for the current work. The data obtained under source-CID conditions 4084

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

(Figure 4C) are the result of fragmentation in a region where nitrogen (used as cone and nebulizer gas) is present together with residual D2O and H2O vapor. In contrast, the MS/MS data of Figure 4D involve fragmentation in an Ar-filled collision cell that is virtually free of residual D2O or H2O. Any forward or backward exchange taking place under these nonidentical conditions should lead to very different results. The fact that the behavior seen in Figure 4, parts C and D, is similar implies that differential gasphase HDX does not play a major role here. In addition, the results obtained for the various peptic fragments (Figure 6) also do not show any indication of differential gas-phase HDX. (B) The stability of individual amide bonds under CID conditions depends on neighboring amino acid side chains, conformation, and charge state.56,57 We previously hypothesized14 that the CID behavior could additionally be modulated by kinetic isotope effects.58–60 The H/D distribution of homogeneously labeled ubiquitin covers quite a range, as demonstrated in Figure 7 for average HDX levels of 60% and 90%. Fragmentation at bonds that cleave more readily in the -CO-NH- form would preferentially occur within the low mass portion of the distribution (red brackets in Figure 7), resulting in CID fragments with HDX levels below that of the intact protein. Conversely, cleavage at bonds that are more unstable in the -CO-ND- form (blue brackets) would provide fragment HDX levels higher than that of the precursor. The H/D distribution for 60% deuteration is much (56) Kapp, E. A.; Schutz, F.; Reid, G. E.; Eddes, J. S.; Moritz, R. L.; O’Hair, R. A. J.; Speed, T. P.; Simpson, R. J. Anal. Chem. 2003, 75, 6251–6264. (57) Huang, Y.; Triscari, J. M.; Tseng, G. C.; Pasa-Tolic, L.; Lipton, M. S.; Smith, R. D.; Wysocki, V. H. Anal. Chem. 2005, 77, 5800–5813. (58) Mirza, S. P.; Krishna, P.; Prabhakar, S.; Vairamani, M.; Giblin, D.; Gross, M. L. Int. J. Mass Spectrom. 2003, 230, 175–183. (59) Herrmann, K. A.; Kuppannan, K.; Wysocki, V. H. Int. J. Mass Spectrom. 2006, 249-250, 93–105. (60) Derrick, P. J.; Donchi, K. F. In Comprehensive Chemical Kinetics; Bamford, C. H., Tipper, C. F. H., Eds.; : Amsterdam, 1983; Vol. 24.

broader (fwhm ) 14, Figure 7A) than for 90% deuteration (fwhm ) 9, Figure 7B). Accordingly, the overall spread in HDX levels of homogeneously labeled ubiquitin (Figure 4, parts C and D) should be less pronounced at the highest deuteration levels used. This trend is not observed experimentally, arguing against kinetic isotope effects of the type discussed here as a major factor underlying the observed nonuniformities. (C) Although we can rule out the occurrence of H/D fractionation in solution under the conditions used here (Figure 6), it is possible that isotope fractionation occurs in the gas phase during CID. The mobile proton model implies that protons and deuterons can migrate along the polypeptide backbone under CID conditions.35–39 If the proton affinities61,62 of the various sites within a protein differ from the corresponding deuteron affinities, the reshuffling of H and D during collisional activation would be accompanied by deuterium enrichment at some sites and deuterium depletion at others. Significant isotope fractionation has been reported for low molecular weight species in the gas phase, a phenomenon that has been ascribed to vibrational zero-point energy and rotational factors.63–65 Although we are not aware of any studies on isotope fractionation in gas-phase proteins, it is possible that effects of this type are chiefly responsible for the nonuniformities seen in Figure 4, parts C and D. This hypothesis is supported by known changes in H-bond energies upon desolvation,66 a factor that has previously been linked to isotope fractionation.52 Implications for Spatially Resolved HDX Analyses. We will now return to the key question whether the CID data for regioselectively labeled ubiquitin (Figure 4, parts A and B) contain information on the solution-phase deuteration pattern of the protein. Consistent with Akashi et al.22 we observe that b2-b16 are deuterated to a significantly lesser extent than the intact protein. These low labeling levels have previously been attributed to substantial solution-phase protection in the N-terminal region.22 However, the current work reveals that strikingly low b-ion deuteration levels also occur for homogeneously labeled ubiquitin (Figure 4, parts C and D), implying that the HDX behavior of these fragments is largely independent of the deuteration pattern in solution. This disqualifies b2-b16 as reporter ions for the spatial solution-phase isotope distribution. It is not surprising, therefore, that a comparison of the b-ion HDX levels of regioselectively labeled ubiquitin with the corresponding NMR data from the literature43 reveals major discrepancies (Figure 8A). The y′′-ions generated from homogeneously labeled ubiquitin exhibit only relatively small deviations from the HDX level of the intact protein (e2%, Figure 4, parts C and D), suggesting that the factor(s) responsible for artifactual nonuniformities are less prevalent for these fragments. According to Akashi et al.22 the high deuteration levels of short C-terminal fragments obtained after regioselective labeling are indicative of low protection in these protein regions. Indeed, the measured y′′ deuteration levels seem to show some correlation with the expected HDX behavior Harrison, A. G. Mass Spectrom. Rev. 1997, 16, 201–217. Deakyne, C. A. Int. J. Mass Spectrom. 2003, 227, 601–616. Larson, J. W.; McMahon, T. B. J. Am. Chem. Soc. 1986, 108, 1719–1720. Larson, J. W.; McMahon, T. B. J. Am. Chem. Soc. 1988, 110, 1087–1093. Graul, S. T.; Brickhouse, M. D.; Squires, R. R. J. Am. Chem. Soc. 1990, 112, 631–639. (66) Nie, B.; Stutzman, J.; Xie, A. Biophys. J. 2005, 88, 2833–2847. (61) (62) (63) (64) (65)

Figure 8. Measured deuteration levels for b-ions (A) and y′′-ions (B), taken from Figure 4A for t ) 5 min (squares). Expected b-ion (A) and y′′-ion deuteration levels (B) are depicted as pink circles, determined from the NMR protection data (ref 43) that are summarized in Figure 1. The expected deuteration level x for each fragment was calculated as x ) ( 0.9F)/T, where F ) fast exchanging sites (side-chain H + rapidly exchanging amide H (ref 43) + terminal H + charge carrier H) and T ) total exchangeable H + charge carrier H (ref 67); 0.9 is the D2O mole fraction in the labeling buffer. Two of the experimental data points (b8 and y′′24) are taken from Figure 4B, rescaled to account for the slightly different HDX level of the intact protein.

(Figure 8B), certainly more so than the b-ions (Figure 8A). Thus, it appears that y′′-ions retain the original solution-phase deuteration pattern to some degree. CONCLUSIONS This work provides new information pertinent to the debate on the general feasibility of CID methods as a tool for spatially resolved HDX measurements. Differences in instrumentation appear to be a secondary factor, because vastly different designs (FTICR22 vs Q-TOF) and different fragmentation methods (sourceCID vs MS/MS) provide consistent data, at least for the ubiquitin model system studied here. When comparing the CID behavior of uniformly and regioselectively labeled ubiquitin, a number of interesting observations can be made. The fact that the b-ion deuteration behavior is largely independent of the spatial labeling pattern in solution implies that the production of these species involves extensive scrambling. Furthermore, the factors responsible for the below-average deuteration levels of uniformly labeled ubiquitin also appear to be highly prevalent during CID of the selectively labeled protein. This behavior is likely caused by isotope fractionation during collisional activation, facilitated by the high mobility of charge carriers in the gas phase. Regardless of the underlying mechanism, scrambling under these conditions does not result in fragment ions exhibiting the same deuteration level as the intact protein. This conclusion puts into question the common practice of interpreting any nonuniformities in fragment deuteration as being indicative of regioselective solution-phase labeling. Artifactual deuterium enrichment or depletion can easily Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

4085

be misinterpreted as “real” spatially resolved information. This is particularly true if (i) control experiments on homogeneously labeled proteins are lacking and (ii) if validation based on detailed comparisons with reference data is not being carried out. We propose that unrecognized deuterium enrichment or depletion effects may have contributed to the lack of consensus in the literature as to whether HDX/CID represents a potentially viable tool for measuring solution-phase deuteration patterns. For example, the previous ubiquitin work of Akashi et al.22 is generally considered to be a highly successful HDX/CID study, since both b- and y′′-ions apparently reveal detailed information on the protein dynamics in solution. In fact, a cursory comparison of the observed trends in that work (low HDX levels for b-ions, high levels for y′′-ions) provides qualitative agreement with the presence of a highly mobile C-terminus42,44–46 that is devoid of amide protection.43 The fact that the b-ion labeling behavior is highly affected by artifactual nonuniformities is only uncovered after carrying out experiments on uniformly labeled ubiquitin (Figure 4, parts C and D) and a comparison with previous HDX/NMR data43 (Figure

4086

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

8). In contrast, the y′′-ion data observed here and in Akashi’s study22 are not inconsistent with a certain retention of the solutionphase HDX pattern. Interestingly, this behavior is opposite to earlier results obtained for small peptides, which suggested y′′ions to be less reliable indicators of solution-phase deuteration patterns than b-ions.15,16,28 Future studies will be required to gain more detailed insights into the mechanisms underlying the nonuniform labeling behavior observed in this work. ACKNOWLEDGMENT This study was funded by the Natural Sciences and Engineering Research Council of Canada, the Canada Foundation for Innovation, and by the Canada Research Chairs Program.

Received for review January 26, 2008. Accepted April 4, 2008. AC8001963