Coupling High-Field Asymmetric Ion Mobility Spectrometry with Capillary Electrophoresis-Electrospray Ionization-Tandem Mass Spectrometry Improves Protein Identifications in Bottom-Up Proteomic Analysis of Low Nanogram Samples

Abstract

In this work, we pioneered the assessment of coupling high-field asymmetric waveform ion mobility spectrometry (FAIMS) with ultrasensitive capillary electrophoresis hyphenated with tandem mass spectrometry (CE-MS/MS) to achieve deeper proteome coverage of low nanogram amounts of digested cell lysates. An internal stepping strategy using three or four compensation voltages per analytical run with varied cycle times was tested to determine optimal FAIMS settings and MS parameters for the CE-FAIMS-MS/MS method. The optimized method applied to bottom-up proteomic analysis of 1 ng of HeLa protein digest standard identified 1314 ± 30 proteins, 4829 ± 200 peptide groups, and 7577 ± 163 peptide spectrum matches (PSMs) corresponding to a 16, 25, and 22% increase, respectively, over CE-MS/MS alone, without FAIMS. Furthermore, the percentage of acquired MS/MS spectra that resulted in PSMs increased nearly 2-fold with CE-FAIMS-MS/MS. Label-free quantitation of proteins and peptides was also assessed to determine the precision of replicate analyses from FAIMS methods with increased cycle times. Our results also identified from 1 ng of HeLa protein digest without any prior enrichment 76 ± 9 phosphopeptides, 18% of which were multiphosphorylated. These results represent a 46% increase in phosphopeptide identifications over the control experiments without FAIMS yielding 2.5-fold more multiphosphorylated peptides.

Graphical Abstract

1. INTRODUCTION

In recent years, high-field asymmetric waveform ion mobility spectrometry (FAIMS), a variant of ion mobility spectrometry (IMS), has generated a lot of interest for its utility in enhancing depth of coverage in single-shot proteomics.^1-3 The mechanism of FAIMS separation and its potential applications in protein analysis have been thoroughly described in other works.^4,5 In brief, FAIMS utilizes an assembly of two coaxial cylindrical electrodes mounted at the front end of the mass spectrometer in which ions are displaced by an applied oscillating asymmetric waveform, where the amplitude of the waveform is the dispersion voltage (DV). Alternating between high and low electric fields between the electrodes, the difference in ion mobility creates a lateral offset between the electrodes, and a direct current compensation voltage (CV) applied to one of the electrodes dictates which ions are filtered through to the MS. The FAIMS Pro interface by Thermo Fisher Scientific is a recently developed, commercially available cylindrical electrode assembly FAIMS instrument. Although the resolving power of cylindrical electrode-based FAIMS devices is relatively low compared to their planar electrode FAIMS counterparts,^6-8 transmission of multiply charged ions occurs at different CVs than singly charged ions, therefore enabling FAIMS to effectively reduce the background noise levels and increase the signal-to-noise of the analytes of interest.^9,10 Despite the lower resolving power in cylindrical FAIMS, the ion focusing effect due to the curved geometry of the electrode assembly gives the cylindrical FAIMS interface a sensitivity advantage over planar FAIMS interfaces.¹¹ For complex samples, the fractionation of analytes of interest prior to MS reduces duty cycle strain, which enables a broader dynamic range for precursor ions targeted for fragmentation.¹² Recent studies have demonstrated the advantages of coupling the FAIMS Pro interface with LC-MS-based proteomic applications,^2,13 including for ultrasensitive analyses.^10,14 However, to the best of our knowledge, coupling the FAIMS Pro interface with CE-MS-based proteomic approaches has not been previously described.

The capabilities of CE-MS for proteomic profiling at ultrasensitive levels have been previously demonstrated by our lab and others.^15-17 Profiling post-translationally modified peptides has been highlighted as an advantage in CE-MS analysis^17,18 due to the shift in electrophoretic mobility imparted by chemical modifications that alter peptide charge.¹⁹ FAIMS has been reported to improve the identification of phosphorylated peptides by enhancing the signal-to-noise ratio.²⁰ A previous report demonstrated the coupling of a FAIMS device with sheath-flow CE-MS for analysis of lipopolysaccharides.²¹ This work demonstrated a reduction of background noise and increased detection limits for glycolipid species and the resolution of isomers. Another report demonstrated improvements in separation and sensitivity of nitrosamines from drinking water using CE-FAIMS-MS with a sheath-flow interface.²² Here, we utilize a sheathless electrospray interface for CE-MS/MS analysis,²³ which has been reported to gain higher sensitivity than sheath-flow interfaces.²⁴ To our knowledge, the compatibility of a sheathless CE electrospray interface with FAIMS has not been assessed.

In this work, we evaluated the coupling of ultrasensitive sheathless CE-MS/MS with the FAIMS Pro interface (CE-FAIMS-MS/MS) to determine the compatibility of these two technologies and to understand the benefits that FAIMS can bring to ultrasensitive CE-MS/MS proteomic profiling. We demonstrated that FAIMS decreases background noise levels and increases signal-to-noise for CE-MS proteomic analyses. With the FAIMS device, a combination of three CVs at 20 and 25 V intervals alternated throughout a single CE-MS run can be used to achieve a 16% increase in protein identifications, a 25% increase in peptide group identifications, and a 46% increase in phosphopeptide identifications. Furthermore, the FAIMS results show improved protein sequence coverage and peptide spectral matching over control experiments without FAIMS.

2. MATERIALS AND METHODS

2.1. Sheathless CE-ESI-MS/MS

Separation was performed with a CESI 8000 (SCIEX, Brea, CA) on a bare fused silica capillary (90 cm, 30 μm ID, 150 μm OD) with a sheathless electrospray interface (SCIEX OptiMS cartridge). Hydrodynamic injection of ~50 nL of the sample was performed by applying 5 psi for 68 s followed by an injection of background electrolyte (BGE) at 0.5 psi for 25 s. The separation voltage applied was 20 kV in normal polarity. The BGE used was 40% acetic acid in water, and the conductive line was filled with 10% aqueous acetic acid. For CE-FAIMS-MS/MS methods, 0.5 psi supplemental pressure was applied at 30 min for the remainder of the separation to support stable spray; otherwise, no supplemental pressure was used. The capillary was interfaced with a Nanospray Flex ion source mounted at the front end of an Orbitrap Fusion Lumos Tribrid mass spectrometer (both Thermo Fisher Scientific). MS acquisition was started after the initial 1 min CE voltage ramp up and ended at the start of the 5 min CE voltage ramp down. The BGE was refreshed between each analysis with a 4 min rinse at 100 psi. Similarly, the conductive liquid was refreshed with a 3 min rinse at 100 psi between each analysis. The total runtime per analysis is ~109.5 min.

2.2. Mass Spectrometry and FAIMS Settings

The mass spectrometer was operated in positive ionization mode with an ion transfer tube temperature of 110 °C and a spray voltage of +1.3 kV during data collection. The total data collection time was 96.5 min. For MS1, full scan data was collected at 60,000 resolution (at 200 m/z) from 375 to 1600 m/z with a maximum ion injection time of 100 ms and an AGC target setting of 3 × 10⁶. The RF lens was set to 40%. Data-dependent MS2 scans were analyzed in the linear ion trap (rapid scan mode) with a 35 ms ion inject time and an AGC target setting of 8 × 10⁴. Precursor ions were isolated with a 0.7 m/z window and fragmented with an HCD collision energy of 28%. The total MS cycle time was varied from 3 to 7 s, and parallelization mode was on. The dynamic exclusion window was set to 10 s, and a minimum precursor intensity threshold of 5 × 10³ was used. In CE-FAIMS-MS/MS experiments, a FAIMS Pro interface was used (Thermo Fisher Scientific), the temperatures of the FAIMS inner electrode, outer electrode 1, and outer electrode 2 were set to 100 °C (Standard Resolution Mode), and a FAIMS user set gas flow of 0 L/min was used. The gas flow setting was modified from the factory setting of 5–4 L/min. Three or four compensation voltages (CVs) were evaluated within each single-stage MS data acquisition method (“internal CV stepping” type of experiment, see Table S1) followed by tandem MS data acquisition as described above.

2.3. Data Analysis

All raw files were analyzed using Proteome Discoverer software (v.2.5, Thermo Fisher Scientific), searched against the UniProtKB/Swiss-Prot human database (downloaded on 3/11/2020, containing 20,302 sequences). For protein and peptide group identifications, the SequestHT algorithm with INFERYS rescoring²⁵ was used with carbamidomethylation of cysteine residues set as a static modification. Precursor feature detection was performed with the Minora Feature Detector node for label-free quantitation (LFQ) of all files with SequestHT and INFERYS rescoring. Carbamidomethylation of cysteine residues was set as a static modification. Precursor abundances were reported based on intensity. Additional searches for modified peptides used SequestHT alone with carbamidomethylation of cysteine residues set as a static modification and oxidation of methionine residues, phosphorylation of serine, threonine, and tyrosine residues and N-terminal acetylation and methionine loss with N-terminal acetylation set as dynamic modifications for the protein N-terminus. At least one high confidence peptide sequence was required for protein identifications, and a target false discovery rate (≤1% FDR) was used for protein identification. A multi-consensus report was generated for triplicate analyses of all FAIMS and control methods tested. The open-source data parsing and a quality control analysis program, RawTools,²⁶ was used to determine average peak widths for selected data files. The IMP-apQuant node²⁷ available for Proteome Discoverer was used to extract peak shape characteristics, including full width half max (FWHM) and the number of data points per peak.

Further data analysis was performed in the R environment.²⁸ Graphical outputs were generated using the package ggplot2.²⁹

The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE³⁰ partner repository with the dataset identifier PXD034255.Additional experimental details about materials, reagents, methods, mass spectrometry parameters, and data analysis are provided in the Supporting Information file.

3. RESULTS AND DISCUSSION

3.1. Coupling CE-MS/MS with FAIMS and Optimization of CE-FAIMS-MS/MS Method Cycle Time

In this work, we evaluated coupling the FAIMS Pro interface with ultrasensitive sheathless CE-MS/MS¹⁷ for proteomic profiling of low nanogram samples. The ultrasensitive CE-MS/MS method used here was optimized to produce a minimal electroosmotic flow (EOF), thereby resulting in a wider separation window than typically observed with bare fused silica (BFS) capillaries (~60 min). Even with the increased migration window, ~77% of PSMs are observed in a 15 min window (Figure S1), suggesting a need for another dimension of separation such as FAIMS to improve the depth of proteome coverage. In part due to the low flow rate, coupling sheathless CE-MS presented challenges in maintaining spray stability and efficient ion transmission. Additionally, due to the nitrogen carrier gas flow utilized with FAIMS, difficulties in maintaining stable spray were observed. To alleviate spray stability issues, the carrier gas flow factory setting on the FAIMS was dropped from 5 to 4 L/min. All FAIMS experiments were conducted with 4 L/min carrier gas flow. Additionally, an added supplemental pressure of 0.5 psi was applied at the start of the peptide migration window (30 min) to support stable spray with minimal interference to the separation. Internal stepping FAIMS methods³¹ apply a number of discrete compensation voltages (CVs) alternated through a single analytical run (Figure 1A-C). The peak capacity of the CE method was estimated to be ~174 by multiplying the average width at 13.4% (4σ) of five peaks selected across the migration window by the length of the migration window (60 min). While mass spectrometry provides an additional dimension in ion separation, the peak capacity in the m/z domain is identical for both FAIMS and no FAIMS control CE-MS methods since the same MS instrument was used in both types of experiments. With FAIMS internal stepping methods, where ion separation by the FAIMS interface serves as an additional separation dimension, the peak capacity can be theoretically increased by multiplying the number of CVs used, assuming 0% overlap in peptide identifications. This difference in theoretical maximum peak capacity may explain why our preliminary studies using two CV methods resulted in lower identification rates compared to three and four CV methods. Based on these findings, in this work, we focused our efforts on exploring three and four CV internal stepping methods. Here, we have tested stepping through three or four CVs spaced apart by at least 15 V to cover the range of voltages where peptides have been demonstrated to transmit through the FAIMS interface and to minimize the amount of overlap in peptide identifications.¹ In a data-dependent MS experiment, either a cycle time or topN setting is applied to determine either how long the instrument should spend acquiring MS2 scans or how many precursor scans should be isolated for fragmentation based on intensity. Here, we use cycle time with precursor scan events occurring in the Orbitrap and MS2 scanning in the ion trap of the Fusion Lumos Tribrid instrument. In an internal stepping FAIMS method, there is a distinction between the cycle time for each CV (i.e., the time between each precursor scan) and the overall cycle time (i.e., the time between precursor scans of the same CV). Initial experiments maintaining the same overall cycle time of 3 s used in the control (i.e., no FAIMS) experiments consisted of a 1 s cycle time for each CV in which a precursor (MS1) scan was acquired in the Orbitrap for a given CV with the concurrent acquisition of data-dependent fragment (MS2) scans in the ion trap for the previous CV (Figure 1B). The length of the cycle time dictates the time between precursor scans and how many MS2 scans can be acquired per precursor. In this work, we investigated the effect of increasing the cycle time for each compensation voltage and, therefore, increasing the overall cycle time (Figure 1C). With increased CV cycle time, an increased number of MS2 scans per precursor scan was also observed (Figure 1D). For the FAIMS method with 1 s cycle time per CV for a total of 3 s cycle time, the number of MS2 scans per cycle does not exceed 16. The range of MS2 scans per precursor increased as the cycle times per CV were extended to 1.5 and 2 s. The range of MS2 scans per precursor scans doubled with 2 s cycle time per each of three CVs compared to 1 s cycle time, reaching a similar range of values as the no FAIMS method with 3 s cycle time.

Figure 1.

Overview of CE-FAIMS-MS/MS internal stepping mode experimental workflow and FAIMS-MS cycle time parameters. (A) CE-FAIMS-MS/MS interface demonstrating an internal CV stepping method with three CVs alternated through a single analytical run. Colored dots represent ions transmitted through FAIMS to MS at different CVs. As shown in this image, green ions are transmitted when −65 V is applied to the inner FAIMS electrode. Schematics of the FAIMS 3 CV internal stepping method for (B) 3 s (1 s/CV) and (C) 6 s (2 s/CV) total cycle times. (D) Distribution of the number of MS2 scans acquired per precursor scan for three CV methods with increasing cycle times compared to no FAIMS with 3 s cycle time.

3.2. Reduction of Background Ions with FAIMS

One of the main advantages of FAIMS that has been well described in other recent studies^9,10 is the reduction of background ions. Many singly charged ions that are common contaminants in LC-MS-based proteomics, such as polysiloxanes,³² do not transmit favorably at the same CVs that are optimal for peptides, therefore nearly eliminating the background chemical noise originating from these ions. We observe the same phenomenon of significantly reduced background in this CE-MS-based work. With FAIMS, the background signal is 3 orders of magnitude lower than in the control experiments without FAIMS (Figure 2A,B). Polysiloxanes and other singly charged ions that are typically present are notably absent from ion density plots of FAIMS data (Figure 2C,D). Also, as we previously described,¹⁰ this reduction in the background is associated with an increase in signal-to-noise for data acquired with FAIMS (Figure S2). Base peak electropherograms from all FAIMS methods tested show similar reduction in background noise compared to no FAIMS (Figure S3). Minimizing the transmission of contaminating ion species to the mass spectrometer is particularly important for the analysis of low sample amounts, similar to the level analyzed in this work. However, the impact of filtering out singly charged species may be even greater for the analysis of limited samples with higher matrix complexity, such as those obtained from microbiopsy, microscale liquid biopsy, or laser-capture micro-dissection.

Figure 2.

Background noise reduction and signal-to-noise increase with FAIMS. Base peak electropherograms of representative analyses (A) CE-MS/MS alone and (B) CE-FAIMS-MS/MS normalized to 3E7. Zoomed-in insets show background intensity from a segment of the analysis after the peptide migration window. Ion density maps of (C) CE-MS/MS and (D) CE-FAIMS-MS/MS normalized to 1E6.

3.3. Comparison of CE-FAIMS-MS/MS and CE-MS/MS Identifications from 1 ng of HeLa Protein Digest

In this study, we aimed to show that FAIMS can provide an improvement to the level of proteome coverage currently attainable with ultrasensitive CE-MS/MS and to establish the optimal parameters of the FAIMS method. First, we observed that maintaining an overall cycle time of 3 s and allowing each CV only 1 s cycle time did not show improvement over CE-MS/MS alone. The use of three CVs (−85, −65, and −40 V) with 1 s cycle time for each CV resulted in 1061 ± 43 proteins and 3982 ± 170 peptide groups identified, which is comparable to the 1131 ± 20 proteins and 3865 ± 104 peptide groups identified in control experiments without FAIMS (Figure 3A). We demonstrated that extending the cycle time for each CV resulted in an increased number of MS2 scans between each precursor scan (Figure 1D). Extending the cycle time to 2 s per CV (overall 6 s cycle time) resulted in 1314 ± 30 protein and 4829 ± 200 peptide group identifications representing a 16 and 25% improvement over CE-MS/MS without FAIMS, respectively (Figure 3A). Blank injections of sample buffer between experiments were processed without percolator to determine a minimal level carryover with ≤3 PSMs identified. Control experiments were also conducted, increasing cycle times to 4 and 6 s without FAIMS; however, the 3 s cycle time provided the best results, and the increased cycle times appeared to have a negative impact on the number of identifications (Figure 3B). Although cycle times up to 7 s were tested, the additive effects of extending the cycle time for the FAIMS method appeared to plateau at 6 s. The evaluated FAIMS methods also demonstrated improvement in MS2 spectral matching. The optimal FAIMS method showed a nearly 2-fold increase in the percentage of acquired MS2 spectra that resulted in PSMs with 28% of MS2 spectra matched compared to 15% in no FAIMS experiments (Table S2). The improvement in spectral matching is likely due in part to lower co-isolation interference that was observed in all FAIMS experiments compared to no FAIMS analyses (Figure 3C).

Figure 3.

Protein, peptide group, and PSM identification results from 1 ng of HeLa digest with and without FAIMS. (A) Average identifications (+/−STDEV) from 1 ng of HeLa digest with the FAIMS methods using three or four CVs with varied cycle times (N = 3) and (B) without FAIMS at varied cycle times (N = 3). Statistically significant increases (p-value ≤0.05) compared to no FAIMS control are indicated with an asterisk (*) symbol. Average MSMS values were divided by 10 to be displayed on the same scale as other metrics. (C) Boxplots show precursor isolation interference statistics by the method. (D) All PSM and non-redundant PSM totals from triplicate analyses for each method. PSM overlap between compensation voltages from a representative run with (E) three CVs (7507 total PSMs) and (F) four CVs (8493 total PSMs) applied.

FAIMS methods with four CVs (−85, −70, −55, and −40 V) were also tested with cycle times of 1, 1.5, and 1.75 s/CV. The results demonstrated a similar trend with increased identifications at longer cycle times (Figure 3A). The total number of PSMs from the four CV methods is ~23% higher than the three CV methods; however, the unique peptide group identifications (with one or more high confidence PSMs) and protein group identifications did not reflect the same increase due to the overlap between CVs (Figure 3A,F). When redundant PSMs are removed for each method, the numbers of non-redundant PSMs from the four CV methods are comparable to the numbers observed in the three CV methods (Figure 3D). Compared to the ~14% overlap in PSMs from −40 and −55 V from a FAIMS 4 CV method (Figure 3F), −40 and −65 V used in a FAIMS 3 CV method show only 7% overlap (Figure 3E). FAIMS 4 CV methods also demonstrated a higher ratio of PSMs to peptide groups than what was observed for FAIMS 3 CV and no FAIMS methods, which is likely due to the overlap between CVs (Table S2). Although the total number of non-redundant PSMs identified for FAIMS methods was higher than for no FAIMS analyses (Figure 3D), the number of unique PSMs identified for any single CV value did not surpass the number of PSMs identified in the no FAIMS experiments (Figure S4), indicating the advantages of combining multiple CVs during the analysis.

3.4. Protein and Peptide LFQ with FAIMS Compared to No FAIMS Control

The data were further analyzed using a label-free quantitation workflow to compare the performance of CE-FAIMS-MS methods to CE-FAIMS alone. The numbers of quantitative peptides and proteins were significantly increased (p ≤ 0.05) in all FAIMS methods compared to the no FAIMS control, except the three CV 3 s cycle time/CV method where the number of quantified proteins was lower (Figure S5). The numbers of quantitative peptides and proteins in the three CV 2 s cycle time/CV method were 5393 ± 135 and 1306 ± 19, respectively. The distribution of peak FWHM for each method showed similar median values (Figure S6). A decrease in the average number of data points per peak from ~9 to ~6 was observed when the overall cycle time was extended from 3 to 6 s (Figure S7); however, this did not appear to negatively impact the quantitation. Peptide abundances were reproducible between replicates for the FAIMS method using three CVs with 2 s cycle time/CV and replicates of the no FAIMS control method demonstrated by strong positive correlation of peptide abundances between replicates (Figure S8) and similar median coefficient of variations in percent (%CVs) for peptide abundances for each method (Figure S9). The distribution of all peptide abundances determined from the LFQ analysis is similar for all FAIMS methods using three CVs and for all FAIMS methods using four CVs, although the median peptide abundance observed in three CV FAIMS methods was ~1.5× lower than those observed in no FAIMS and four CV methods (Figure S10).

3.5. Evaluation of CV Selection on Characteristics of Recovered Peptides and PSMs

One of the primary benefits of FAIMS in the analysis of complex mixtures is the additional fractionation of ions prior to MS by applying different CVs, which reduces the duty cycle strain on the mass spectrometer and enables a greater dynamic range for the analysis.¹² Alternating through multiple CVs during one analysis, also known as internal CV stepping, increases peptide coverage compared to using a single CV.¹ To demonstrate the importance of combining CVs and selecting appropriate CV intervals, we assessed the differences in peptides and PSMs for each CV from the optimal FAIMS method using three CVs with 2 s/CV cycle time. Different distributions of peptide transmission during the migration window are observed from base peak electropherograms for each CV (Figure 4A). The differences in peptide groups identified at each CV were also demonstrated by minimal overlap (1%) in peptide groups identified by all CVs and the majority (83%) of peptide groups identified across three replicates uniquely observed at a single applied CV (Figure 4B). Additionally, in comparing the PSM charge states observed at each CV, the majority of PSMs with a charge state of +3 were identified with a CV of −65 V, while the majority of PSMs with an ionic charge state of +2 were identified at −40 V (Figure 4C and Figure S11). The ratios of +2 to +3 PSMs identified at −85 and −65 V were skewed toward the +3 charge state, while the majority of PSMs observed for −40 V had a charge state of +2 (Figure S4). The PSM results without FAIMS likewise showed a greater number of +2 ions. Comparing the overall number of PSMs identified by each method, the most significant difference is observed in the number of +3 ions, where FAIMS shows an ~1.5× increase over CE-MS/MS without FAIMS (Figure 4C). The additive benefits of peptide profiling with FAIMS ultimately resulted in higher protein sequence coverage and protein identification with FAIMS (Figure 4D). The median sequence coverage for proteins identified by both methods increased from 11% with no FAIMS to 14% with the optimal FAIMS method (three CVs and 2 s cycle time/CV). Although, protein identifications were higher in the FAIMS method, and a 62% overlap between FAIMS and no FAIMS protein identifications was observed (Figure 4E).

Figure 4.

Comparative assessment of peptide characteristics and protein coverage from optimal FAIMS and no FAIMS methods. (A) Base peak electropherograms for each CV in a representative analysis with three CVs. (B) Overlap of peptide groups identified for each CV. (C) Charge states of non-redundant PSMs identified with and without FAIMS (N = 3). Stacked bars of PSMs identified in the FAIMS method indicate the contribution of each CV to the total. (D) Protein sequence coverage of proteins identified in both FAIMS and no FAIMS methods (N = 3) ranked from highest to lowest based on the FAIMS data. Smoothed conditional means with 95% confidence interval are shown for each data set. Zoomed-in inset shows a lollipop plot connecting corresponding values from FAIMS and no FAIMS for the top 100 proteins based on the FAIMS data ranking. (E) Venn diagrams demonstrate overlap in protein identifications FAIMS and no FAIMS (N = 3).

3.6. Effect of CV in FAIMS Experiments on Identification of Post-Translationally Modified Peptides and PSM Distribution

Next, we assessed the number of unique phosphorylated and N-terminal acetylated peptide identifications profiled with each FAIMS method and without FAIMS. Post-translational modifications (PTMs) such as O-phosphorylation and N-terminal acetylation have a crucial role in biological function,^33,34 and abnormal PTMs have been associated with numerous diseases.³⁵ Due to the biological significance of PTMs, it is necessary to have highly sensitive proteomic methods that can detect these low abundance modified peptides. We have demonstrated an advantage previously in CE-MS/MS for profiling these modified peptides without the need for any prior enrichment,¹⁷ and FAIMS has been reported to improve identification of phosphorylated peptides due primarily to the increased signal-to-noise ratio.²⁰ For the optimal CE-FAIMS-MS/MS method, 76 ± 9 unique phosphopeptides were identified from 1 ng of the HeLa protein digest standard, representing an ~46% increase over CE-MS/MS alone (Figure 5A). Multiphosphorylated peptides with two or three phosphorylation sites represented ~18% of the phosphopeptides identified by this FAIMS method, while only 8% of the phosphopeptides identified without FAIMS contained multiple phosphorylation sites (Figure 5A). Further investigation into the migration times for phosphorylated peptides compared to all other peptides showed a migration time shift for phosphorylated peptides, which would be expected due to the introduction of negative charges from the phosphate group (Figure S12). Specifically, for ~50 pairs of peptides where both non-phosphorylated and phosphorylated forms of the peptides with the same sequence were detected, the observed shift in migration time for the phosphorylated form helped to increase confidence in the phospho-identification (Figure S13). Migration time shifts between peptide sequences with 1 or 2 and 2 or 3 phosphosites could also be observed. In some cases, additional modifications such as oxidation or N-terminal acetylation were also identified in phosphopeptides. One observation of phosphorylated peptide, ESEDKPEIEDVGpSDEEEEK, with a missed cleavage site resulting in an additional lysine at the C-terminus for tryptic peptide (ESEDKPEIEDVGpSDEEEEKK), showed that the miscleaved phosphopeptide migrated faster than the unphosphorylated peptide (ESEDKPEIEDVGSDEEEEK), likely due to the increased charge from the additional lysine residue. Interestingly, FAIMS did not show an improvement in unique acetylated peptide identifications within the evaluated CV range (Figure 5A). To understand this further, we investigated the trends for modified peptide spectrum matches at each CV. The number of non-redundant PSMs found for each CV shows a positive correlation between the CV magnitude and the number of PSMs corresponding to phosphorylated peptides (Figure 5B) and to acetylated peptides (Figure 5C). From −55 to −40 V, there is an ~2.5-fold increase in acetylated PSMs, suggesting the possibility that the optimal CV for acetylated peptides may be higher than the values used in this study. The CV ranges selected intentionally did not include values above −40 V because it was determined with a CV sweep of background ions that −30 V is the peak for transmission of contaminant polysiloxane ions (data not shown). At −40 V, the background is higher relative to the other CVs, and the distribution of PSM m/z values is more similar to the distribution observed with no FAIMS (Figure 5D). Interestingly, a trend was observed for m/z values transmitted at each CV. The median value of PSM m/z increased as the CV increased, from ~520 m/z at −85 V to ~655 m/z at −40 V, and the distribution of m/z values was wider at higher CVs (Figure 5D). These differences in peptides transmitted at different CVs highlight the importance of utilizing multiple CVs across a wide range with FAIMS experiments to ensure that peptides are not filtered out of the analysis.

Figure 5.

Investigating post-translationally modified peptides identified in FAIMS. (A) Stacked bars demonstrate average singly and multi- (doubly and triply) phosphorylated peptide identifications with at least one high confidence PSM for FAIMS methods and CE-MS/MS alone (left panel). Standard deviations are shown for singly and multiphosphorylated peptides. Statistically significant increases (p-value ≤0.05) compared to no FAIMS control are indicated with an asterisk (*) symbol. Average N-terminal acetylated peptide identifications with at least one high confidence PSM (+/−STDEV) for FAIMS methods and CE-MS/MS alone (N = 3, right panel). (B) Non-redundant phosphorylated PSMs and (C) acetylated PSMs for each CV. Values plotted for each method are represented by different shapes. (D) Density plots show the distribution of PSM m/z values identified by each CV and for no FAIMS.

4. CONCLUSIONS

In this work, for the first time, we have demonstrated the compatibility of sheathless CE-MS/MS and the FAIMS Pro interface and the benefits of coupling these technologies for MS-based proteomics of low nanogram samples. With the FAIMS Pro, we achieved significantly greater depth of proteome coverage of 1 ng of the HeLa digest sample by CE-MS/MS, higher protein sequence coverage, and improved profiling of phosphorylated peptides. Proteomic analysis by CE-FAIMS-MS/MS has not been reported previously, according to our knowledge.

In this study, we demonstrated a decrease in background signal by 3 orders of magnitude with FAIMS, thereby enabling a signal-to-noise increase for analytes of interest and a greater dynamic range for precursor intensities. The additional fractionation of peptide ions in the gas phase prior to the mass analyzer with FAIMS internal CV stepping methods increased the peak capacity of the analysis and resulted in reduced co-isolation interference and a 2-fold increase in the percentage of matched spectra. The optimized FAIMS method identified 1314 ± 30 proteins, 4829 ± 200 peptide groups, and 7577 ± 163 peptide spectrum matches (PSMs) from 1 ng of HeLa protein digest, corresponding to a 16, 25, and 22% increase over CE-MS/MS alone, respectively. The numbers of quantified proteins and peptides are also increased with FAIMS methods. The effects of extending the cycle time on quantitation were investigated to confirm that the precision and accuracy of peptide quantitation were not compromised at an optimal cycle time of 6 s. At this low level of the sample, profiling of modified peptides is very challenging. At this sample level, equivalent to approximately five cells, CE-MS/MS analysis alone identified an average of 52 ± 2 phosphopeptides from three replicates without applying any enrichment or targeted data acquisition techniques. Phosphorylated peptides were observed migrating at later migration times compared to non-phosphorylated peptides, which allows profiling of these low abundance species in CE-MS/MS. The FAIMS Pro facilitated a 46% increase in phosphopeptide identifications over CE-MS/MS alone, including an increase in multiphosphorylated peptides. We believe that this CE-FAIMS-MS/MS method for ultrasensitive proteomic analysis holds tremendous potential for applications in limited sample analysis and provides an excellent alternative to LC-MS/MS approaches. The picoliter-nanoliter regime of injection volumes typical to CE-MS-based approaches can be particularly beneficial for analysis of single cells or subcellular regions where sample processing volumes can be scaled down to minimize losses. We anticipate that the FAIMS interface may enable even more substantial gains in CE-MS/MS profiling sensitivity for scarce, “real-world”, biological and clinical samples, especially those with higher matrix complexity and a more predominant presence of small molecules resulting in prevalent singly charged ion species.