Ion mobility in gas and liquid phases: How much orthogonality is obtained in capillary electrophoresis–ion mobility–mass spectrometry?

Ion mobility–mass spectrometry (IM–MS) is an ever‐evolving tool to separate ions in the gas phase according to electrophoretic mobility with subsequent mass determination. CE is rarely coupled to IM–MS, possibly due to similar separation mechanisms based on electrophoretic mobility. Here, we investigate the orthogonality of CE and ion mobility (IM) by analyzing a complex peptide mixture (tryptic digest of HeLa proteins) with trapped ion mobility mass spectrometry (TIMS–MS). Using the nanoCEasy interface, excellent sensitivity was achieved by identifying thousands of peptides and achieving a peak capacity of 7500 (CE: 203–323 in a 150 cm long capillary, IM: 27–31). Plotting CE versus mass and CE versus (inverse) mobility, a clear grouping in curved striped patterns is observed according to the charge‐to‐size and mass‐to‐charge ratios. The peptide charge in the acidic background electrolyte can be estimated from the number of basic amino acids, with a few exceptions where neighboring effects reduce the positive charge. A surprisingly high orthogonality of CE and IM is observed, which is obviously caused by solvation effects leading to different charges and sizes in the liquid phase compared to the gas phase. A high orthogonality of CE and ion mobility is expected to be observed for other peptide samples as well as other substance classes, making CE–IM–MS a promising tool for various applications.


INTRODUCTION
Ion mobility-mass spectrometry (IM-MS) has gained much attention over the last two decades due to the commercialization of various ion mobility (IM) techniques, including drift tube IM (DTIM, Agilent Technologies), traveling wave IM (TWIM, Waters), field asymmetric IM spectrometry (FAIMS, Thermo Fisher Scientific), differential mobility spectrometry (DMS, Sciex) and trapped IM spectrometry (TIMS, Bruker).These devices are already used for the in-depth characterization of complex proteins and samples [1][2][3][4] and isomeric or isobaric structures [5,6].TIMS is a technique where ions are pushed forward by a gas flow proportional to their collisional cross-section (CCS) and trapped by an electrical field based on the molecule's charge.The ions are released by manipulating the electrical field.Parallel accumulation serial fragmentation (PASEF) allows the fragmentation of molecules based on their mobility in the gas phase, enabling very fast and sensitive MS/MS acquisition in TIMS-QTOF MS instruments [7].
CE is a powerful technique to efficiently separate ions due to minor differences in their electrophoretic mobility (charge-to-size ratio) in the liquid phase.Thus, CE-MS is used in many fields including the analysis of complex mixtures of proteins or metabolites [8][9][10][11].New developments in interfacing technology [12][13][14][15][16][17] compensate for the limiting sample introduction volumes regarding achievable concentration sensitivities, enabling the analysis of metabolites and peptides down to sub-ppb levels [18].CE-MS is applied for the analysis of peptides due to its different selectivity compared to LC-MS [8,19,20].HeLa digest is often used as a model for proteomics samples to evaluate separation and MS/MS performance both in LC-MS [4] and CE-MS [21][22][23].
In contrast to CE-MS, only a few studies have been published on the coupling of CE with IM-MS.In one of the first studies, Jooß et al. analyzed protein-derived glycans, showing the benefit of combining both technologies [24], similar to a work on labeled glycans [25].The sensitivity of glycan analysis was increased by developing an in-line SPE for the CE-DTIM-MS set-up by Hooijschuur et al. [26].CE-TIMS-MS has been applied to analyze targeted peptides from low volume samples: Mast et al. showed the separation of diastereomeric neuropeptides in single cells [27], while Delaney et al. analyzed nine targeted angiotensin peptides in mouse tissue samples [28].CE-FAIMS-MS has been used to perform top-down proteomics experiments [29] as well as for the characterization of complex bacterial lipopolysaccharides [30].CE-TWIM-MS has been used for the detection of protein conformational isomers, including affinity experiments, to describe their enzymatic activity and their inhibition [31].Drouin et al. investigated CE-IM-MS for metabolite profiling, showing the benefit for the identification of isobaric analytes, primarily in the neutral (nonseparated) zone [32].In most of these applications, IM assists in the separation of a few analytes not separated by CE or by MS.Gou et al., investigated the application of CE-DTIM-MS for proteome samples in comparison to LC-MS [33].Beside a substantial orthogonality of CE and LC, they also found orthogonality for CE and DTIM.There is hardly any study discussing the orthogonality of CE and IM in detail, which seems to be of great interest, especially since both techniques rely on electrophoretic mobility, leading to potentially small differences in selectivity.
Here, we present CE-TIMS-MS measurements of tryptic HeLa-digest for the discussion of orthogonality in CE-IM-MS.Various capillary lengths were evaluated regarding peak capacity.To obtain a high number of MS/MS experiments, a 1.5-m long capillary was finally applied.Resulting data are discussed regarding peptide charge in liquid and gas phases, respectively, as well as the resulting orthogonality of CE and IM.

Sample preparation and CE separation
Lyophilized Pierce™ HeLa Protein Digest Standard (Thermo Fischer Scientific, Dreieich, Germany) was diluted to a final concentration of 1 µg/µL protein using a mixture of isopropanol (IPA, LC-MS grade, Carl Roth GmbH+Co.KG, Karlsruhe, Germany) and 0.1 M formic acid (FA, ≥98%; Carl Roth GmbH+Co.KG) in water (ultrapure [UPW], 18 MΩ*cm at 25 • C, SG Ultra Clear UV from Siemens Water Technologies, USA) in the ratio of 1:99 (v/v).CE was performed on an Agilent 7100 CE instrument (Agilent Technologies GmbH, Waldbronn, Germany).Fused silica capillaries (separation capillary: 50 µm id, 365 µm od, and SL capillary: 100 µm id, 240 µm od) were purchased from Polymicro Technologies (Phoenix, AZ, USA).PEO-coated capillaries of various lengths were applied, with a length of 1.5 m for the final experiments.To penetrate the glass emitter (30 µm tip opening), the separation capillary was etched at one end to a wall thickness of less than 150 µm using hydrofluoric acid (40%(v/v), Merck, Darmstadt, Germany) prior to the coating procedure [12].The PEO stock solution was prepared by dissolving 100 mg PEO (Mw: 1.000.000;Alfa Aesar, Kandel, Germany) in 45 mL UPW.The solution was heated to 95 • C. A total of 450 µL of that stock solution was acidified using 50 µL 0.1 M hydrochloric acid (conc.HCl: 37%, Sigma, Steinheim Germany).The coating was applied to the capillary using the CE and an external pressure of 4 bar in the following steps: 1 M sodium hydroxide (NaOH, Merck), UPW, and 1 M HCl for 5 min each, followed by the PEO coating solution for 10 min, UPW, and BGE for 5 min.The BGE consisted of 10% IPA in water containing 1 M FA.A large sample plug (200 ng of protein; 6.81% capillary volume) was injected in a dynamic pH-junction/transient ITP mode.For that, a plug of 1 M ammonia (conc.NH 3 : 30%; Carl Roth GmbH+Co.KG) was injected hydrodynamically (100 mbar, 20 s, equimolar to acid amount) before the sample plug.Sample injection was done using the CE flush option (950 mbar) for 21 s.A total of 30 kV was applied for the separation and was executed for 200 min without additional pressure.An additional pressure of 100 mbar was applied after that time point.The capillary was flushed for 5 min with BGE between the runs.

CE-MS
The CE was coupled to the timsTOF Pro2 using the nanoCEasy interface [12,34].The timsTOF was operated in positive ionization mode using captive spray source settings of 1700 V electrospray voltage, 150 • C dry gas temperature, and 3 L/min dry gas flow.The mobility of the ions was measured between 0.6 and 1.6 V*s/cm 2 with a ramp time of 100 ms.One MS acquisition was combined with 10 TIMS cycles for MS/MS experiments using the PASEF algorithm for fragmentation experiments, resulting in an overall duty cycle of 1.1 s.Using the polygonic function in the IM-MS-heatmap singly charged signals were excluded from MS/MS experiments.The mgf file for the Mascot database search was generated based on a threshold intensity (TIC AllMSn) of 25,000 counts, and 10,000 compounds were set as the maximum number.Fragment spectra of the same precursor were combined in a retention time window of 0.5 min with a precursor m/z window of 0.1.Data files were processed using a Mascot search using the SwissProt database and a Homo sapiens (human) taxonomy filter.Only peptides with a charge between +2 and +4 were evaluated.The results were exported to CSV files without showing duplicate peptides.Data were further assessed using Data Analysis 5.3 and Excel 2019.The charge of the peptides at a specific pH was gained using ProtpI.

RESULTS AND DISCUSSION
CE-timsTOF coupling was straightforward using the nanoCEasy interface.The schematic setup is shown in Figure 1.A total of 200 ng of tryptic HeLa digest was injected using stacking experiments.The measurement could be done without further sample preparation in an acidic BGE (1 M FAin 10% IPA).A positive voltage was applied to the CE inlet so that positively charged proteins migrated toward the MS.MS/MS experiments were done using the PASEF algorithm.The overall separation time strongly depended on the capillary length used.
Different capillary lengths were tested to increase the peak capacity.To calculate the peak capacity of a CE separation, the start and end times of the separation were estimated.Nine peaks were chosen (three at the beginning, middle, and end of the separation window) to determine the average full width at half maximum.Calculating the mean width at 13.4% peak height (4σ), a peak capacity for the CE separation was determined.The same approach was done for the gas phase IM for the same set of peptides.The total peak capacity is the product of both individual peak capacities.For a 0.6-m and a 1-m long capillary, an average peak capacities of 1700 (n = 2) and 2700 (n = 2) were achieved, respectively.By increasing the capillary length to 1.5 m, the peak capacity increased to a mean value of 7500 (6223−9497, n = 3).As expected, the peak capacity strongly depends on the capillary length.which is obviously caused by the separation window increase.The results for the individual measurements are shown in Supporting Information S1.
Due to the strongly increased peak capacity in CE, a 1.5-m long capillary was used for the following measurements.Applying 30 kV, the peptides migrated in a time window of > 100 min, as illustrated in Figure 2A, showing the base peak electropherogram of one representative measurement.
The heatmap of migration time and m/z values (Figure 2B) showed a separation of the peptides leading to distinct, slightly curved distributions due to the distinct charges in the acidic liquid phase (see discussion below).CE separates based on the size and charge of the molecule.A large molecule with many charges can, therefore, have the same mobility in the liquid phase as a small molecule with few charges.At the same time, peptides with similar mass and charge can have different mobilities in the liquid phase due to different hydrodynamic radii.The CE has a higher separation efficiency than the IM separation in the gas phase (Figure 2B,C).Plotting the inverse mobility against the CE migration time (Figure 2D), a distinct, slightly curved distribution is observed again.Furthermore, a large area of the heatmap is covered with data points of peptides, demonstrating already the different selectivity of CE and IM, leading to a high orthogonality.
To understand the separation profiles in more detail, the peptides of CE-MS/MS experiments were identified based on the fragment spectra using a Mascot database search.Only the most intense MS/MS spectra were used to obtain unequivocal peptide sequence information.Based on the obtained peptide sequence, the amount of basic and acidic amino acids in the peptide, their charge in the liquid phase, and pI were derived.By calculating the peptide charge in solution at the pH of the BGE (pH = 1.9 for the aqueous part), it was revealed that their charge almost solely depends on the number of basic amino acids.It was almost independent of the amount of acidic amino acids present in the molecule.A peptide with one basic and one acidic amino acid had almost the same charge as a peptide with one basic and eight acidic amino acids (differences in charge < 0.07; see discussion below).Thus, the charge of the peptide in the acidic solution equals to the number of basic amino acids + 1 (N-terminus).Figure 3A shows the migration time-m/z-heatmap of the identified peptides colored by the number of basic amino acids in the peptide.The distinct, slightly curved distributions are related to the different amounts of basic amino acids and therefore different charge states in the liquid phase.However, multiple stripes with the same number of basic amino acids exist and are partially overlapping with stripes containing other numbers of basic amino acids.For example, peptides with two or three basic amino acids overlap between 56 and 76 min.The reason for the existence of several stripes is that the peptides have a distinct charge state in the liquid phase, but exist in several charge states in the gas phase.Thus, a charge deconvolution of the m/z data was performed.The resulting CE-mass-heatmap reveals single stripes for each charge state in solution (Figure 3B).
Peptides containing one basic amino acid (all tryptic peptides should have at least one basic amino acid at the C-terminus) are doubly charged in solution and have a smaller mobility in the liquid phase (higher migration time) than those carrying two or even three basic amino acids.Each curved distribution contains peptides with the same charge in solution, and larger masses migrate slower than smaller ones.Such a picture was already described by Faserl et al., where LC fractions were analyzed by CE using sequential sample injection, and the peptides overlapping could be traced back to their LC fractions depending on their net charge and mass [35].
A few peptides seem not to fit into this scheme.A closer examination of the stray peptides revealed that those peptides that are migrating in CE at a lower charge state than expected either contain a basic amino acid on the Nterminus or two basic amino acids in close proximity (see b, c, and e-l in Supporting Informaion S2).This is obviously due to an overestimation of the charge using the simple approximation without taking neighboring effects into account.Furthermore, the few peptides that are migrating in CE at a higher charge state than expected, are most likely wrong sequence attributions since these peptides appear only in single measurements with a rather low Mascot score (see a and d) in Supporting Information S2.No correlation between peptide size, general amino acid sequence, or character of amino acids was identified as the reason for the misbehaving peptides.
Similar to the migration time-m/z-heatmap, a striped pattern is observed when the inverse mobility is plotted against the migration time (Figure 3C).Peptides with the same mobility in the liquid phase show different mobilities in the gas phase and vice versa.To better understand stripes where peptides with different basic amino acids are mixed, the gas phase charge state was evaluated.In Figure 3D, it is visible that peptides containing one, two, or even three basic amino acids carry two positive charges in the gas phase (MS).Peptides with three charges in the MS have either two, three, or four basic amino acids (Figure 3E), and peptides with four charges in the MS have three or four basic amino acids (Figure 3F).Within an area where peptides with different basic amino acids are mixed, their charge in the gas phase differs.
Since charge deconvolution of the mass led to a better understanding of the behavior of the peptides in the liquid phase, the same was executed for the gas phase mobility as well.The 1/k 0 value was first converted into the K 0 value, divided by the gas phase charge, and then reconverted into a charge-normalized 1/k 0 value.These values were plotted against the migration time to get a "charge normalized" 2D mobility plot.As shown in Figure 4A, the peptides group again in stripes according to their number of basic amino acids.
However, compared to the charge-deconvoluted mass result in Figure 2B, where all peptides of the same charge form distinct sections, the charge-normalized inverse mobility still shows small gaps separating the peptides by their gas phase charge state.This indicates that a charge normalization for the mobility can be done, but a small offset remains.This is most likely because higher gas phase charge states lead to larger CCS values (larger 1/k 0 value) due to the repulsing forces of the charges within one molecule in the gas phase.The high area coverage in the charge normalized inverse mobility-migration time-heatmap (Figure 4A) is retained to a large degree, indicating a high orthogonality of CE and IM even when the gas phase charge state is subtracted.This area is much larger than a similar mass-charge normalized 1/k 0 -plot (Figure 4B).
To obtain a full picture of the orthogonality of CE and IM, the original heatmap of inverse mobility versus migration time (Figure 2D) is considered since the data of Figures 3 and 4 do not contain all peptides (e.g., singly charged peptides were excluded from MS/MS experiments).
The orthogonality of the 2D separation was estimated based on the area covered by signals in the range used for peak capacity calculation (Figure 5).Assuming that the quadratic area represents 100% orthogonality we achieved an orthogonality of around 80%.This is only an estimation (see Gilar et al. [36] for more detailed discussion), however, it clearly shows that CE-MS can significantly benefit from the additional IM dimension.The orthogonality is higher than the one from IM-MS data and rather in the order of a well performed 2D-separation by LCxLC [37].
A more detailed calculation would reduce the orthogonality value to some extent since the empty room between the distinct, slightly curved distributions are considered.However, the information (about the charge state in solution) obtained from these stripes (see discussion above) is valuable.A spreading of the signal could be obtained by increasing the pH.This can be seen in the data of Gou et al. [33]., where a BGE of 2.5% acetic acid (pH of 2.6) was used, making it difficult to observe such a pattern.Indeed, when calculating the expected charge in solution for pH = 2.6, deviations of 0.34 from the nominal charge states derived from the basic amino acids alone are calculated when acidic amino acids are present in the respective peptide.For the BGE at pH = 1.9 (aqueous part of the here applied BGE), this deviation is less than 0.07 (see Supporting Information S3).Although the striped patterns merge into each other at higher BGE pH values, we still expect in this case a high orthogonality of CE and IM due to solvation effects in CE.

CONCLUDING REMARKS
Coupling the CE using the nanoCEasy interface to the timsTOF enabled the sensitive characterization of complex tryptic peptide mixtures derived from a complete proteome.The tryptic HeLa digest could be injected in relatively large volumes (6.81% of capillary) without losing separation performance.A total peak capacity of up to 9500 demonstrates the analytical performance gain for complex samples by adding a factor of around 30 in peak capacity by the TIMS even when fast IM-gradient is used-in conjunction with the ability to generate MS/MS spectra at almost 100 Hz without compromising sensitivity.The peak capacity of the CE reaches a value of 323 for these complex peptide samples.The acidic BGE leads to well-distinct curved patterns in the migration time-m/z profile.Similarly, curved patterns of identical charge in solution and gas phases are obtained in the migration time-inverse mobility heatmap.A charge normalization based on the gas phase charge obtained from the MS reduces the number of stripes to one for each charge state in solution with a slight gap resulting from the initial gas phase charge.This agrees with the expectation that a higher charge in the gas phase leads to larger molecules due to repulsion forces.Even after charge normalization, a high orthogonality for CE and gas-phase IM is observed.Although the underlying separation principle is the same, the solvation leads to differences not only in charge but also in size, resulting in a significantly different selectivity of CE and IM.The high orthogonality obtained here for peptides is expected to be observed also for other analytes, making CE-IM-MS a promising tool for various applications, especially when applying the ultrafast and sensitive MS/MS performance of the TIMS system.

A C K N O W L E D G M E N T S
Thanks to Bruker Daltonics GmbH & Co.KG, especially Christian Albers, Jörg Sauer, and Stefan Slamnoiu for providing the opportunity to couple the CE to the timsTOF.
We thank Lukas Naumann for the support during initial measurements.
Open access funding enabled and organized by Projekt DEAL.

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors have declared no conflict of interest.

D ATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

F I G U R E 2
Separating tryptic HeLa digest using 1.5-m long PEO coated capillary.(A) Base peak electropherogram (BPE; 350−2200 m/z).(B) Heatmap of migration time and m/z value.(C) Heatmap of gas phase mobility and m/z value.(D) Combined heatmap of liquid and gas phase mobility.

F I G U R E 3
Evaluation of the identified peptides.(A) Migration time-m/z-heatmap (analog to Figure 1B).(B) Heatmap after deconvolution, migration time-mass plot.(C) Migration time-mobility heatmap (analog to Figure 1D.(D) Migration time-mobility heatmap with gas phase charge state +2 (cross).(E) Migration time-mobility heatmap with gas phase charge state +3 (cross).(F) Migration time-mobility heatmap with gas phase charge state +4 (cross).In all diagrams, peptides are colored according to the amount of basic amino acids in the peptide.Yellow circle: One, light blue triangle: Two, dark blue square: Three, grey diamond: Four basic amino acids.

F I G U R E 4
Evaluation of the mascot search identified peptides.(A) Migration time-charge normalized inverse mobility plot.(B) Mass-charge normalized inverse mobility plot.In each diagram, peptides are colored according to the number of basic amino acids in the peptide.Yellow circle: One, light blue triangle: Two, dark blue square: Three, grey diamond: Four basic amino acids.Numbers indicate the charge state in the gas phase.F I G U R E 5 Orthogonality of CE-TIMS analysis.Data are identical to Figure 2D with the relevant migration time range and lines indicating the edge of the region where peptides are found.