Ion Mobility Mass Spectrometry of Peptide Ions: Effects of Drift Gas

Nicole J. Rijs , Thomas Weiske , Maria Schlangen , and Helmut Schwarz ..... Paul S. Blank , Ed Darland , Alex Mordehai , Peter S. Backlund , Alfred L...
1 downloads 0 Views 3MB Size
Article pubs.acs.org/ac

Ion Mobility Mass Spectrometry of Peptide Ions: Effects of Drift Gas and Calibration Strategies Matthew F. Bush,*,† Iain D. G. Campuzano,‡ and Carol V. Robinson§ †

University of Washington, Department of Chemistry, Box 351700, Seattle, Washington 98195-1700, United States Department of Molecular Structure, Amgen, Thousand Oaks, California 91320, United States § Department of Chemistry, Physical and Theoretical Chemistry Laboratory, University of Oxford, South Parks Road, Oxford OX1 3QZ, United Kingdom ‡

S Supporting Information *

ABSTRACT: One difficulty in using ion mobility (IM) mass spectrometry (MS) to improve the specificity of peptide ion assignments is that IM separations are performed using a range of pressures, gas compositions, temperatures, and modes of separation, which makes it challenging to rapidly extract accurate shape parameters. We report collision cross section values (Ω) in both He and N2 gases for 113 peptide ions determined directly from drift times measured in a low-pressure, ambient temperature drift cell with radiofrequency (rf) ion confinement. These peptide ions have masses ranging from 231 to 2969 Da, ΩHe of 89−616 Å2, and ΩN2 of 151−801 Å2; thus, they are ideal for calibrating results from proteomics experiments. These results were used to quantify the errors associated with traveling-wave Ω measurements of peptide ions and the errors concomitant with using drift times measured in N2 gas to estimate ΩHe. More broadly, these results enable the rapid and accurate determination of calibrated Ω for peptide ions, which could be used as an additional parameter to increase the specificity of assignments in proteomics experiments.

I

peptide ion assignments. For example, many different gases are used for IM separations, including He,2,3 N2,11,32−34 and air.1,35 Ions are less mobile in N2 gas than He gas due to the greater mass and polarizability of an N2 molecule, thus ion drift times in N2 gas are significantly longer than those in He gas under similar field strengths. Longer drift times can enable easier coupling with TOF and other multiplexed mass analyzers, e.g., most traveling-wave IM36 separations use N2 gas37 or predominantly N2 gas38,39 prior to TOF mass analysis. Figures of merits for IM separations can also depend on gas selection. For example, IM-MS analysis of a large set of tryptic peptides using N2 drift gas resulted in an enhanced average peak capacity (1700) relative to that when He drift gas was used (1260),32 illustrating the benefits of considering gas composition during IM method development. Even though IM separations are performed using a range of gas compositions, Ω with He gas (ΩHe) are desirable for many applications to enable comparisons with additional experimental and theoretically calculated values.26,40−43 For this reason most Ω reported from traveling-wave IM experiments44−46 were obtained using calibration with calibrant ions whose ΩHe had been determined previously using complementary methods. The resulting calibrated Ω for the analyte ions are treated as ΩHe during analysis. This approach has been justified

on mobility (IM) rapidly separates ions (microseconds to milliseconds) based on their mobilities in a gas under the influence of an electric field, which depend primarily on ion shape (collision cross section value, Ω) and charge (z).1−4 Applications of IM coupled with mass spectrometry (MS) are growing in terms of both diversity and numbers and include analysis of chemical warfare agents,5 explosives,6 polymers,7,8 pharmaceutical compounds,9−11 metabolites,12,13 carbohydrates,14,15 phospholipids,16 and intact biological assembilies.17−20 The integration of IM into MS-based proteomics and systems biology workflows21 has accelerated rapidly in recent years and offers the potential to enhance the dynamic range, peak capacity, and specificity of those experiments.22−26 For example, liquid chromatography (LC)−IM-time-of-flight (TOF) MS analysis of peptides spiked into a tryptic digest of mouse plasma detected additional peptides over a wider dynamic range (19 of 20 spiked peptides identified at concentrations of 100−4 ng/mL) than LC−Fourier-transform MS analysis (13 of 20 spiked peptides identified at concentrations of 102−4 ng/mL).25 IM also enhances the assignment of biomolecule classes,27,28 charge states,29 sequence identities,26 and precursor ions during dissociation experiments.26,30,31 IM-MS experiments are performed using a range of pressures, gas compositions, temperatures, and modes of separation. These differences can make it challenging to use drift times measured using a given instrumental platform and set of conditions, coupled with previous IM experiments or Ω calculated for candidate structures, to improve the specificity of © 2012 American Chemical Society

Received: May 26, 2012 Accepted: July 30, 2012 Published: July 30, 2012 7124

dx.doi.org/10.1021/ac3014498 | Anal. Chem. 2012, 84, 7124−7130

Analytical Chemistry

Article

because calibration should correct for many systematic mobility differences in He and in N2 gases.47 There are, however, relatively few ions for which absolute Ω have been measured under similar experimental conditions using both He and N2 gases,11,47 which has made it challenging to assess the errors associated with using drift times measured in N2 gas to estimate ΩHe. Here, we report Ω measured in both He and N2 gases for 113 peptide ions. Ω in these experiments are determined directly from drift times measured in a low-pressure (∼2 Torr), ambient-temperature drift cell with radio-frequency (rf) ion confinement, without the need for mobility calibration.47 These peptide ions have a wide range of masses (231−2969 Da), ΩHe (89−616 Å2), and ΩN2 (151−801 Å2); thus, they are ideal for calibrating results from IM-MS proteomics experiments. These data are used to quantify the errors associated with travelingwave Ω measurements and with using drift times measured in N2 to estimate ΩHe for peptide ions. These results enable the rapid determination of accurate Ω for peptide ions and the increased specificity of assignments in IM-MS based proteomics experiments.

Table 1. Parameters for Traveling-Wave IM Experiments wave height/V

wave velocities/m·second‑1

40 35 30 25 20

1200, 1000, 800, 600, 500, 400 1000, 800, 600, 500, 400, 350 800, 600, 500, 400, 350, 300 600, 500, 400, 350, 300, 250 350, 300, 250, 210

developed in the lab. The drift time for each arrival-time distribution was determined from the centroid of the best-fit Gaussian distribution, which was first estimated from the moments of the distribution and then optimized using the Levenberg−Marquardt, least-squares minimization algorithm49 implemented in SciPy.50



RESULTS AND DISCUSSION Direct Ω Measurements. Absolute ΩHe and ΩN2 were measured using an rf-confining drift cell.47 Drift times in these experiments depend predominantly on interactions between ions and drift gas molecules in the presence of a weak, uniform electric field, conditions that are well-described in the literature.2,3,51 ΩHe for 16 denatured peptide and protein ions measured using this drift cell47 were indistinguishable from those measured using analogous dc-only, drift-tube experiments,52 suggesting that any contributions of the rf fields to drift times in these experiments are very small. Similarly, ΩHe for a series of tetraalkylammonium ions measured using this drift cell11 are indistinguishable from those measured using a segmented quadrupole.53 Briefly, the measured ion velocities are proportional to the electric field (E) and the mobility of the ion (K) under the conditions in the drift cell. Drift voltages in these experiments ranged from 60 to 200 V. These values correspond to field energies ranging from 1.7 to 5.5 V cm−1 Torr−1, which are significantly less than the low-field limit (20−45 V cm−1 Torr−1) estimated54 for a large set of peptide ions.55 Representative arrival-time distributions for doubly protonated polyalanine ions are shown in Figure 1A. Those arrival-time distributions were measured using 2.5 Torr He and a drift voltage of 60 V, which resulted in resolutions (centroid drift time (tD)/fwhm) of ∼12. That resolution, which is less than that for most doubly protonated tryptic peptides from bovine serum albumin under these conditions (∼16), suggests that the polyalanine ions in these experiments adopt a diverse ensemble of structures.52,56 Furthermore, the resolution of the polyalanine ions did not improve at higher field energies, whereas those for the tryptic peptide ions often improved to ∼20. The arrival-time distributions for all of the observed polyalanine ions and most of tryptic peptide ions were fit well using single Gaussian distributions. Several of the tryptic peptide ions exhibited arrival-time distributions with multiple features in some experiments; these were also assigned using the centroid of the best-fit single Gaussian. Experimental tD for doubly protonated polyalanine ions are plotted as a function of reciprocal drift voltage in Figure 1B. tD depends on both ion velocities in the drift cell and the transport time of ions from the exit of that cell to the TOF mass analyzer (t0):



METHODS Samples. DL-polyalanine (P9003) and polyglycine (P8791) were purchased from Sigma-Aldrich. Tryptic digests of bovine serum albumin (186002329), bovine hemoglobin (186002327), and rabbit phosphorylase B (186002326) were provided by Waters Co. (Manchester, U.K.). All samples were dissolved in 49.5/49.5/1 H2O/acetonitrile/acetic acid. Drift-Cell IM Experiments. Direct Ω were measured using a modified Waters Synapt G1 HDMS instrument in which the traveling-wave IM cell was replaced with an rf-confining drift cell. This 18 cm drift cell has a constant direct-current (dc) electric field along the radial axis to direct ions to the TOF mass analyzer, has a radial rf confinement, and is discussed in detail elsewhere.47 Ions were generated using nanoelectrospray ionization from pulled, gold-coated borosilicate capillaries prepared and used as described previously.48 Drift-cell Ω were determined directly from the slopes of two or more drift times versus reciprocal drift voltage plots measured on separate days, each of which used 10 drift voltages ranging from 60 to 200 V in either 2.6 Torr of He gas or 2.0 Torr of N2 gas. Driftgas pressures were measured using an absolute pressure transducer (MKS Baratron model 626A, Wilmington, MA), and temperatures were measured using three T-type thermocouples attached directly to different positions on the cell. Some additional instrumental conditions are reported in the Supporting Information. Traveling-Wave IM Experiments. All traveling-wave IM experiments were performed using a Waters Synapt G2 HDMS instrument.38,39 Ions were generated using nanoelectrospray ionization at 3.5 kV from 50 μm i.d. TaperTips (New Objective, Woburn, MA) and solutions supplied from a syringe pump (flow rate 0.9998 and average 0.99997.

n

ΩHe

ΩN2

3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33

89 100 114 128 141 157 170 181 194 206 217 228

151 166 181 195 211 228 243 256 271 282 294 306

z=3

ΩHe

Ω N2

197 208 220 232 243 255 265 276 287 297 308 317 327 337 348 358

296 309 320 333 344 357 369 380 393 404 416 428 437 448 458 470

ΩHe

ΩN2

338 348 361 373 386 399 412 425 438 452 465 479 490 502 516

482 491 501 518 532 545 561 576 592 606 621 634 649 666 674

be smaller because many sources of error are common to all measurements. Some of the ions investigated here have also been investigated using dc-only, drift-tube experiments. For example, ΩHe for AlanH+, n = 3−14, reported here and previously57 differ by a standard deviation of 1.2%. ΩHe have been reported previously for seven of the doubly charge tryptic peptides from bovine serum albumin,55 and our values are ∼3% larger. These results provide further evidence that Ω determined using rf-confining, drift cell experiments are indistinguishable from those obtained using dc-only, drifttube experiments, given the uncertainties involved in both experiments. Comparison between ΩHe and ΩN2. These ΩHe and ΩN2 are based on measurements performed on the same instrument, using similar conditions, and from ions generated using the same source. This enables a meaningful direct comparison of ΩHe and ΩN2 for all 113 peptide ions. These values in the two drift gases are correlated strongly. Figure 2A shows that for specific charge states ΩHe and ΩN2 are correlated linearly. ΩN2 are larger than ΩHe, an effect that is strongest for the ions that have the smallest Ω. For example, ΩN2 (296 Å2) is 50% larger than ΩHe (197 Å2) for (Ala12 + 2H)2+, whereas, ΩN2 (470 Å2) is only 31% larger than ΩHe (358 Å2) for (Ala26 + 2H)2+ (Table 2). Those trends are consistent with the increased polarization of gas molecules by ions that have greater charge densities,

experiments, t0 depends strongly on m/z due to a time-of-flight separation in the ion optics prior to the mass analyzer. In contrast, t0 in Figure 1B (0.47−0.53 ms, values increase slightly with m/z) depend less on m/z because the residence times in the transfer cell after the drift cell depend most strongly on the traveling-wave velocity used in the transfer cell (247 m s−1 for these experiments).36,47 The small differences in observed t0 are consistent with a small time-of-flight separation in the region between the transfer cell and the entrance to the TOF mass analyzer. Mobilities in He and N2 gases were measured for ions of polyalanine, polyglycine, and tryptic peptides from bovine serum albumin, bovine hemoglobin, and phosphorylase B. The collision cross section value (Ω) of an ion is related to its mobility through the Mason−Schamp equation:51 1/2 3e ⎛ 2π ⎞ 1 Ω= ⎜ ⎟ 16N ⎝ μkBT ⎠ K

z=2

(2)

where N is the drift-gas number density, μ is the reduced mass of the ion and the drift gas, kB is the Boltzmann constant, and T is the drift-gas temperature. ΩHe and ΩN2 for these ions are reported in Table 2 and Table S1 in the Supporting Information. The absolute uncertainty of these Ω are estimated to be 3%,47 but the relative uncertainty of these values should 7126

dx.doi.org/10.1021/ac3014498 | Anal. Chem. 2012, 84, 7124−7130

Analytical Chemistry

Article

relationship is small (±1.4%) and may represent the lower limit in using mobilities measured in N2 gas to estimate ΩHe. Calibrated Ω Measurements Using Traveling-Wave IM. Because of the wide availability of traveling-wave IM instrumentation, there is growing interest in using those methods to determine Ω. The transport of ions during traveling-wave IM is not fully understood, and unlike ions in drift-tube experiments,51 drift times in traveling-wave IM experiments depend nonlinearly on mobility.37,58 Because of these challenges, several protocols have been reported for determining calibrated Ω based on traveling-wave drift times for calibrant ions whose Ω have been determined from drifttube IM experiments.44−46 Studies of phosphopeptides16 and native-like protein complexes47,59 indicate that using calibrant ions that have similar properties results in significantly smaller absolute errors, which can be challenging to identify due to the limited number of Ω in the literature. Despite substantial progress in this area, no clear consensus has emerged regarding mobility calibration or calibrant ion selection. Polyalanine has many properties that are promising for use in calibration. Polyalanine is inexpensive, is stable, and yields a large number of singly, doubly, and triply protonated peptide ions using electrospray ionization. In order to evaluate the potential of polyalanine in this application, a series of travelingwave IM experiments were performed using a Waters Synapt G2 HDMS instrument38,39 for ions of polyalanine and tryptic peptides from bovine serum albumin (BSA) and bovine hemoglobin (HG). Experiments were performed using five different wave heights. At each wave height, four to six wave velocities were selected (Table 1, 28 total combinations) that yielded drift times between 1.5 and 13 milliseconds for the ions of interest. ΩHe of the doubly and triply protonated polyalanine ions, normalized to charge state, are plotted in Figure 3 as a function of traveling-wave arrival time for each of the 28 sets of traveling-wave conditions. Data for each traveling-wave condition were fit using a binomial regression (Ω/z = atD2 + btD + c); the corresponding fits using linear regressions (Ω/z =

Figure 2. (A) Ω in He and N2 gas are well correlated for individual charge states. (B) Reciprocal mobilities in He and N2 gas are well correlated for singly and multiply charged ions, respectively. (C) Reciprocal mobilities in N2 were estimated based on their reciprocal reduced mobilities in He using the best-fit line for each z in part B, and the ratio of the estimated and experimental mobilities in N2 gas is plotted as a function of the reciprocal mobility in He gas. (D) The histogram of the ratios shows that the estimated and experimental reciprocal mobilities in N2 are very similar.

which will be much more significant for N2 than He due to enhanced long-range, charge-induced dipole effects. Some of the differences between ΩHe and ΩN2 noted above can be minimized by normalizing those values by the charge state and reduced mass of each ion. Figure 2B shows that the reciprocal reduced mobilities (K0, mobility normalized to standard temperature and pressure) of these peptide ions are also correlated linearly, but interestingly, the data can be clearly grouped into results for singly versus multiply charged peptide ions. There are some subtle differences between the relationship for the 2+, 3+, and 4+ ions (Figure S1 in the Supporting Information), but those differences are very small relative to those between the singly and multiply charged ions. The best-fit lines in Figure 2B provide a facile means to relate mobilities in He and N2 drift gases. To demonstrate this, effective mobilities in N2 gas were estimated based on their experimental mobilities in He gas using the best-fit lines determined in Figure 2B. The ratios between the effective (determined from mobilities in He gas) and experimental mobilities in N2 gas are plotted as a function of their experimental mobilities in He in Figure 2C. Unity corresponds to complete correlation between the mobilities in N2 and He gas; the deviation from unity corresponds to the extent that the differences between the mobilities of the two gases are noncorrelated. The standard deviation of the mobility ratios is 1.4%, and the histogram of those ratios is plotted in Figure 2D. These results show that peptide ion mobilities, hence Ω, measured in He or N2 gas can be used to estimate accurately those quantities in the other drift gas. The uncertainty in this

Figure 3. Traveling-wave drift times for doubly and triply protonated polyalanine ions are strongly correlated with the corresponding driftcell ΩHe (normalized for charge state), at all 28 combinations of traveling-wave conditions used (Table 1). Data are fit using regressions with binomial functions, which yield R2 correlation coefficients ranging from 0.998 to 0.9995. In total, 4% of the data were excluded using a Grubbs test with loose tolerances; spot analysis of several of those datum revealed traveling-wave arrival-time distributions with very poor S/N. 7127

dx.doi.org/10.1021/ac3014498 | Anal. Chem. 2012, 84, 7124−7130

Analytical Chemistry

Article

Table 3. rf-Confining Drift-Cell and Traveling-Wave ΩHe for Selected Tryptic Peptide Ions of Bovine Serum Albumin (BSA) and Bovine Hemoglobin (HG)

a All cysteine residues were alkylated with iodoacetamide. bStandard deviation of 28 independent traveling-wave experiments. cPercent deviation between the traveling-wave and drift-cell Ω. dPeptide-level results with N2 are reported in Table S2 in the Supporting Information.

atD + b) yield residuals with clear parabolic character. These binomial regressions were used to estimate ΩHe for the observed 2+, 3+, and 4+ tryptic peptide ions (42 total peptide ions, Table 3). Calibrated Ω obtained using this approach are remarkably insensitive to the wave velocity and height used. For individual peptide ions, the relative standard deviations of the 28 calibrated measurements were all less than 2% and on average were 0.5%. The ΩHe determined indirectly via calibration of the traveling-wave drift times and directly via drift-cell measurements differ by an average of 1.8%. The magnitudes and signs of the deviations between the travelingwave and drift-cell ΩHe appear to be statistically distributed. A few of the peptides may have exceptionally large deviations, e.g., two of the +4 charged peptides exhibit deviations of 6.5 and

9.2%, which may reflect slight differences in the ion structures formed using the two instruments. However, it is difficult to determine the significance of these differences with this sample size. Three additional calibration schemes were evaluated. The data points in Figure 3 were also fit using linear46 and trinomial models (Ω/z = atD3 + btD2 + ctD + d). The former results in average standard deviations and absolute errors of 0.7 and 2.8%, whereas the latter results in values of 0.5 and 1.7%, respectively. The former results are clearly poorer than the binomial model and consistent with the parabolic residuals observed for linear fits to the polyalanine data. The latter results are indistinguishable from those obtained using the simpler binomial model; thus, the use of the higher-ordered function is not justified. The 7128

dx.doi.org/10.1021/ac3014498 | Anal. Chem. 2012, 84, 7124−7130

Analytical Chemistry

Article

use of log/log calibration plots,44 which are analogous to fitting the data in Figure 3 using power functions (Ω/z = atDX),60 results in an average standard deviation of 0.7% and an absolute error of 1.9%. On the basis of these results, binomial regressions of the calibrant ion data appears to be the best choice for peptide calibration protocols. Repeating the process outlined in Figure 3 and using the drift-cell derived ΩN2 in Table 2 yields average standard deviations of 0.4% and absolute errors of 1.2%, respectively. Note the same traveling-wave IM drift times were used to generate both the calibrated effective ΩHe (Table 3) and calibrated ΩN2 (Table S2 in the Supporting Information), the primary difference being whether the ΩHe or ΩN2 drift-cell values were used for the polyalanine ion calibration plots. The use of ΩN2 would be expected to yield more accurate results because the Synapt G2 HDMS was operated with predominantly N2 gas in the traveling-wave IM cell. Slightly poorer figures of merit for the effective ΩHe (0.5 and 1.8%) are consistent with the inherent limitations in interconverting between mobilities in He and N2 gases (Figure 2D).

those measurements prior to validation, as done here for ions of tryptic peptides. Even though effective ΩHe have been used most commonly for traveling-wave studies to date, we expect that ΩN2, which are inherently more accurate, will be used increasingly in the future with the availability of ΩN2 for additional calibrant ions and improved tools for calculating ΩN2 for candidate structures. For example, there has been recent progress in calculating accurate ΩN2 for drug-like molecular ions11 that may be extensible to peptide ions. Determining Ω via calibration offers many advantages. Measuring drift times for calibrant and analyte ions under the same experimental conditions removes the need for knowledge of the drift-gas temperature, composition, pressure, and even the mechanism of separation. Calibration using the results from this study enables different laboratories using different instrumentation to produce consistent Ω and develop universal tools for interpreting those results. Ultimately, such values could be used as additional parameters for comparison with databases of experimental Ω28,47,55 or calculated Ω26,43 to increase the specificity of assignments in proteomics and systems biology workflows. For example, previous studies have highlighted the differences in mobilities of phosphorylated and nonphosphorylated peptides.46,61,62 Similarly, cis and trans proline isomers, which exhibit unique arrival time distributions, were observed in over 60% of doubly charged peptides within a specific study.63 It is therefore highly desirable to move toward routine use of Ω in peptide identification.



CONCLUSIONS Although N2 molecules are heavier and more polarizable than He atoms, these results indicate that IM of peptide ions in these two gases depends on very similar factors and that most differences are systematic to all peptide ions. ΩN2 are larger than ΩHe, an effect that is strongest for the ions that have the smallest Ω, but ΩHe and ΩN2 are correlated linearly for ions of a given charge state. Normalizing the Ω for their charge state and reduced mass minimizes these differences. The reciprocal reduced mobilities in the two gases are also correlated linearly for ions of a given charge state, and the results for all of the multiply charge ions are essentially indistinguishable. To mimic the effects of using drift times in one gas to estimate Ω in another, as done during the calibration of most traveling-wave IM experiments,44−46 the effective mobilities in N2 gas for these peptide ions were estimated based on their experimental mobilities in He gas. Using this approach revealed that the noncorrelated differences between the mobilities in these two gases is very small (1.4%), which may represent the lower limit in using mobilities measured in N2 gas to estimate ΩHe. Ions of polyalanine, which is an inexpensive and stable homopolymer that yields a broad distribution of singly, doubly, and triply protonated peptide ions from electrospray ionization, are particularly effective calibrant ions for IM-MS based proteomics. Traveling-wave IM drift times in N2 gas can be used to determine both calibrated ΩN2 and calibrated effective ΩHe for peptide ions of interest, because both ΩHe or ΩN2 for polyalanine ions are now available for use during analysis. These results show that the average deviation between drift-cell and traveling-wave Ω is small (