Capillary Electrophoresis Coupled to Electrospray Ionization Tandem Mass Spectrometry for Ultra-Sensitive Proteomic Analysis of Limited Samples

Abstract

In this work, we developed an ultra-sensitive CE-MS/MS method for bottom-up proteomics analysis of limited samples, down to sub-nanogram levels of total protein. Analysis of 880 and 88 pg of the HeLa protein digest standard by CE-MS/MS yielded ~1100 ± 46 and ~160 ± 59 proteins, respectively, demonstrating higher protein and peptide identifications than the current state-of-the-art CE-MS/MS-based proteomic analyses with similar amounts of sample. To demonstrate potential applications of our ultra-sensitive CE-MS/MS method for the analysis of limited biological samples, we digested 500 and 1000 HeLa cells using a miniaturized in-solution digestion workflow. From 1-, 5-, and 10-cell equivalents injected from the resulted digests, we identified 744 ± 127, 1139 ± 24, and 1271 ± 6 proteins and 3353 ± 719, 5709 ± 513, and 8527 ± 114 peptide groups, respectively. Furthermore, we performed a comparative assessment of CE-MS/MS and two reversed-phased nano-liquid chromatography (RP-nLC-MS/MS) methods (monolithic and packed columns) for the analysis of a ~10 ng HeLa protein digest standard. Our results demonstrate complementarity in the protein- and especially peptide-level identifications of the evaluated CE-MS- and RP-nLC-MS-based methods. The techniques were further assessed to detect post-translational modifications and highlight the strengths of the CE-MS/MS approach in identifying potentially important and biologically relevant modified peptides. With a migration window of ~60 min, CE-MS/MS identified ~2000 ± 53 proteins on average from a single injection of ~8.8 ng of the HeLa protein digest standard. Additionally, an average of 232 ± 10 phosphopeptides and 377 ± 14 N-terminal acetylated peptides were identified in CE-MS/MS analyses at this sample amount, corresponding to 2- and 1.5-fold more identifications for each respective modification found by nLC-MS/MS methods.

Graphical Abstract

Bottom-up proteomics is a widely used mass spectrometry (MS)-based method for in-depth characterization of complex biological samples.¹ Reversed-phase chromatography (RP-LC) is one of the most common separation methods for bottom-up proteomics applications. Conventional nanoflow (~100–400 nL/min) LC (nLC)^2,3 and currently less common ultra-low flow (<25 nL/min) LC methods (ULF LC)^4,5 are used to increase the sensitivity and depth of proteomic analyses. Capillary-based columns of varying dimensions and stationary phases may be used for RP-nLC separations, including more conventionally used silica particle-packed columns ranging 50–100 μm ID² and monolithic^4,6 and open tubular columns⁵ which are typically ultra-narrow bore (i.e., ≤20 μm ID) and operate under ULF conditions.

Capillary electrophoresis (CE) is an alternative nano-scale flow separation technique, orthogonal to RP-LC, that provides certain advantages over LC methods. Generally, CE provides a faster and higher efficiency separation without the need for pumps and high-pressure connections and consumes lower volumes of buffers and samples. Moreover, the bare fused silica capillary surface can be easily restored and regenerated with sodium hydroxide or other rinses to eliminate the carryover issues, extend the capillary lifetime, and improve the reproducibility of separation. As an open tubular separation with no retention medium and a plug-like flow velocity profile, CE is not as prone to peak broadening that can be seen with separations that use a pressure-driven laminar flow and columns packed with a stationary phase.⁷ Additionally, CE can offer key benefits in the analysis of biologically relevant classes of peptides.^8–10 Since CE separates based on charge and hydrodynamic volume of the analyte, unlike in RP-LC, post-translational modifications (PTMs) that affect the charge of a peptide, such as phosphorylation, may cause a dramatic shift in electrophoretic mobility.^9,11 This shift can separate low-abundance, modified peptides from the bulk of the eluting peptides.

CE can be coupled directly to MS through a sheathless electrospray ionization (ESI) interface, which is able to reportedly achieve higher sensitivity over other CE interfaces that rely on sheath flow-driven electrospray that dilutes the sample.^12,13 The ULF generated in CE has been shown to reduce ion suppression and increase ionization efficiency,^14,15 proving beneficial to high-sensitivity analysis, which is increasingly important for rare or limited samples containing only low nanogram or sub-nanogram amounts of protein. Reducing or eliminating electroosmotic flow by controlling the surface charge of the capillary has been shown to increase the migration window and improve the separation resolution in analysis of complex mixtures.^16,17 Additionally, preconcentration methods such as isotachophoresis and pH stacking can help improve sensitivity, particularly in the analysis of less concentrated samples.^18–20

PTMs are an extremely important part of biological functioning. Phosphorylation, for example, plays a critical role in signaling transduction pathways that regulate gene expression.²¹ N-terminal acetylation and glycosylation are other common modifications that are crucial components of protein regulation and function.^22,23 Identifying and characterizing these modifications in biological samples can provide key information and understanding about cellular functioning, which warrants that modified peptides are highly relevant target analytes in proteomic profiling. These modifications are often challenging to detect from complex biological samples due to their low abundance.²⁴ Ion suppression and co-isolation of multiple precursors may also interfere with the identification and characterization of modified peptides. To tackle the challenges in the analysis of PTMs, enrichment techniques are often used to concentrate specific modifications and enhance coverage of peptides of interest.²⁵ These approaches have been very successful and robust with high starting protein amounts, but protein losses may be a significant concern for limited samples, when enrichment techniques may become impractical.^26,27

In this work, we compared the performance of this CE-MS/MS method with two RP-nLC-MS/MS-based high sensitivity methods at the level of ~10 ng of the sample to demonstrate the complementarity of CE-MS/MS and RP-nLC-MS/MS proteomic profiling for HeLa cell tryptic digests. We performed searches for modified peptides and mined the dataset for other differences between types of peptides identified by each method. Demonstrating a ~2–4 increase in phosphopeptide identifications over nLC-MS/MS methods, these results highlighted the strengths CE-MS/MS can offer in the analysis of generally challenging to detect PTMs (e.g., O-phosphorylation) in low-nanogram samples when enrichment strategies may not be possible or beneficial.

We present a CE-MS/MS method for high-sensitivity analysis of low-nanogram and sub-nanogram complex peptide mixtures. On average, we achieved ~160 and ~1100 protein identifications from 88 and 880 pg of the HeLa protein digest standard, respectively, which compares favorably with previously reported results achieved by state-of-the-art CE-MS/MS-based methods.^10,28–30 Furthermore, we lysed and digested low numbers of HeLa cells in solution and analyzed down to single-cell equivalent injections to demonstrate the applicability of our high-sensitivity CE-MS/MS method to a limited sample proteomics workflow. From 1-, 5-, and 10-cell equivalents injected from the resulted digests, we identified approximately 750, 1140, and 1270 proteins, respectively.

MATERIALS AND METHODS

In-House Miniaturized HeLa Cell Digestion.

HeLa cells were washed three times with ice cold 1× phosphate-buffered saline (PBS), counted, and diluted to 1000 and 500 cells/μL aliquots in ice cold 1× PBS. 1 μL of each cell suspension was transferred to a plastic CE insert for 1000 cell or 500 cell processing. Lysis buffer containing 10 M urea, 2.5 M thiourea, 6 mM TCEP, and 25 mM ammonium bicarbonate, pH 8, was added to each replicate and incubated for 10 min at room temperature with shaking at 300 rpm. The samples were alkylated for 30 min at room temperature in the dark with 5 mM iodoacetamide followed by trypsin and Lys-C digestion overnight at 37 °C at 300 rpm (E/S ratios of 1:50 each assuming 0.4 ng/cell). 3% formic acid was added to quench digestion and the samples were dried. The samples were resuspended in 200 mM ammonium acetate pH 9 and sonicated for 20 s prior to injection.

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). For in-house HeLa digested samples, a 2 min 0.1 mol/L sodium hydroxide rinse followed by a 3 min water rinse (both at 100 psi) was used in between injections. Hydrodynamic injection of ~44 or ~50 nL of the sample was performed by applying 5 psi to the sample vial for 60 or 68 s, followed by an injection of the background electrolyte (BGE) at 0.5 psi for 25 s. The separation voltage applied was 20 kV in normal polarity. The BGE used was 30 or 40% acetic acid, and the conductive line was filled with 10% acetic acid. The capillary was interfaced with a Nanospray Flex ion source mounted at the front end of an Orbitrap Fusion Lumos Tribrid mass spectrometer (both from Thermo Fisher Scientific). The same injection volume and same sample buffer were used for each analysis of 10-fold serial dilutions of the protein digest to not alter the isotachophoretic and pH stacking preconcentration effects from the sample buffer.

LC–MS/MS Proteomic Profiling.

LC–MS/MS analyses using monolithic and bead-packed columns were conducted applying LC and MS conditions as we described in our recent studies.^2,4

Data Analysis.

All raw files were analyzed using Proteome Discoverer (v. 2.3 & 2.5) software (Thermo Fisher Scientific). Processing workflow parameters are described in the Supporting Information.

The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE³¹ partner repository with the dataset identifier PXD027226. Additional experimental details about materials, reagents, and methods, including HeLa cell culture, miniaturized sample processing, preparation of monolithic column, LC conditions, MS parameters, and data analysis, are provided in the Supporting Information.

RESULTS AND DISCUSSION

Overview of the CE-MS Method and Workflow.

CE-based separations for complex proteomic analyses typically result in higher throughput and lower carryover and sample losses, and consume smaller volumes of the buffer and sample than LC-based methods. However, CE-MS is often overlooked due to concerns with reproducibility of migration times, mostly caused by changes in BGE composition over time and the injection variability. In this work, we developed a CE-MS/MS method for ultra-sensitive proteomic analysis of limited samples from low numbers of cells containing low-nanogram or even sub-nanogram amounts of the HeLa protein digest standard. We hypothesized that the absence of a packed stationary phase in CE and the minimal surface area of the open tube will lead to decreased sample losses and negligible carryover between runs. We compared the performance of our CE-MS/MS-based method with RP-nLC-MS/MS methods in proteomic analysis of the HeLa protein digest standard (Figure 1A). Then, we explored the capabilities of CE-MS/MS for the analysis of sub-nanogram and low cell protein digests. Additionally, we assessed the capabilities of each of the examined methods in the detection and identification of post-translationally modified peptides, including O-phosphopeptides and N-terminal acetylated peptides, as well as other non-enzymatic peptide modifications and characteristics to investigate the advantages of CE-MS over LC–MS in deep characterization of limited samples without enrichment or additional sample processing steps.

Figure 1.

Experimental workflow and comparison of CE-MS and RP nLC-MS proteomic profiling modes. (A) Overview of the ultra-sensitive CE-MS/MS proteomics and low-nanogram protein digest comparison experimental workflow. Net charge at pH 2.2 calculated for peptides identified in (B) CE-MS/MS shows a gradient of decreasing net charge with increasing migration time in CE. GRAVY scores calculated for peptides identified in (C) bead-packed nLC-MS/MS (0.2% data not shown), and (D) monolith nLC-MS/MS (1% data not shown) shows a gradient of increasing hydrophobicity with retention time in nLC methods.

The developed CE-MS/MS method relies on CE separation conducted in a bare fused silica capillary coupled to MS using the commercially available sheathless interface to achieve the highly sensitive proteomic analysis of low numbers of cells or sub-nanogram amounts of the whole-cell digest. Base peak chromatograms show the reproducibility of migration times in technical replicates (Figure S1). In our hands, CE-MS conducted on a bare fused silica capillary demonstrated better separation reproducibility, capillary lifetime, and nESI stability than the evaluated coated capillaries (e.g., linear acrylamide- and polyetheleneimine-coated capillaries; data not shown).

Few recent studies reported results of CE-MS/MS proteomic analyses of low sample loads similar to the levels shown in this work.^28,29 Typically, the depth of proteomic characterization and the number of protein identifications in CE-MS are limited due to the speed of analysis and the short migration window that leads to co-migration of multiple analytes and ion suppression effects, which make it challenging to detect lower abundance peptides. In the CE method reported in this work, the highly acidic BGE (40% acetic acid in water) decreases the electroosmotic flow and increases the viscosity, thereby extending the effective separation window to ~60 min to match the length of the elution gradients and elution windows used in the nLC methods. In other studies, similar separation windows have been achieved, typically through suppression of electroosmosis with neutral coatings such as linear polyacrylamide;^10,11,17 however, in this work, we developed a separation method with a bare fused silica capillary to avoid the challenges of working with coated capillaries, as mentioned above. With this extended peptide migration window, as described below, we achieved higher numbers of protein and peptide identifications than have been previously reported with sub-nanogram protein digest sample loads in CE-MS/MS without the need for capillary surface coatings (Table S1).

Orthogonality of RP-nLC and CE Separation Methods for Bottom-Up Proteomics.

We investigated CE-MS/MS and two different RP-nLC-MS/MS approaches (monolithic and bead-packed column) as orthogonal approaches for proteomic analysis of the HeLa protein digest. High performance CE separation of digests is based on different electrophoretic mobilities of the peptides, which are directly related to the hydrodynamic volume (size) and charge of the peptide, while RP-LC separates peptides based on hydrophobicity and interactions with stationary and mobile phases. CE separation yielded the narrowest peak widths and highest efficiency separation of the three methods tested (Table S2). The high efficiency associated with CE and minimized sample losses make it an attractive option for ultra-sensitive analysis of nanogram and sub-nanogram amounts of protein.

To demonstrate the effect of peptide charge on separation by CE and LC, we calculated the net charge of each identified peptide with the pH of the CE BGE solution and LC mobile phase A as pH 2.2 and pH 2.7, respectively. As expected, in CE, peptides with a higher net charge migrated faster with some time delay for larger peptides, and no trend was observed for net charge in either reversed-phase method (Figures 1B and S2B,C). To demonstrate the effect of hydrophobicity on separation, we calculated the GRAVY index of each peptide. An expected trend of greater retention for more hydrophobic peptides was observed for the RP-nLC methods, and no migration trend based on hydrophobicity was observed in CE (Figures 1C,D and S2A).

In both LC and CE, the injection volume and the sample concentration dictate the amount of sample loaded on the column. We aimed to subject similar sample amounts to analyses by orthogonal techniques. In LC, the injection amount is typically more straightforward due to fixed volume injections by the autosampler. As previously described, the estimated hydrodynamic injection volume in CE was ~44 nL, resulting in ~8.8 ng loaded amount from 0.2 mg/mL HeLa protein digest standard. While the sample amounts used in LC–MS and CE-MS analyses were slightly different, we found the generated results sufficiently representative. The CE-MS/MS method presented here was able to identify on average nearly 2000 ± 53 protein groups (N = 3, RSD = 2.7%) and 7300 ± 256 peptide groups (N = 3, RSD = 3.5%) from only 8.8 ng of the sample (Figure 2A). However, both nLC-MS/MS methods outperformed the CE-MS/MS method at the level of ~10 ng of the same sample (~1 μL of 10 ng/μL injected), which corresponds to ~40–50 HeLa cells, in terms of overall protein group and peptide group identifications, with the analysis using the ULF monolithic column achieving 62 and 72% more protein and peptide identifications than CE and 31 and 30% more protein and peptide identifications than the bead-packed column for the injected protein amount, only ~10% different between the methods (Figure 2A).

Figure 2.

CE and RP-nLC separation comparison of protein group and peptide group identifications from ~10 ng of the HeLa protein digest standard. (A) Average protein groups and peptide groups identified from N = 3 replicates analyzed by each method. Error bars represent the standard deviation of each calculated average. Assessment of similarity in (B) protein groups and (C) peptide groups identified by each method demonstrated by Venn diagrams. Colored circles represent the total pool of protein groups and peptide groups for CE-MS/MS (green, ~8.8 ng injected), bead-packed nLC-MS/MS (blue ~10 ng injected), and monolith nLC-MS/MS (red, ~10 ng injected). Overlapping regions indicate protein groups, or peptide groups, identified by both (or all) methods. Calculated percentages indicate the percentage of protein or peptide groups represented by each region.

Protein group identifications show a 52% overlap between the three methods (Figure 2B). The high overlap in protein identification indicates that the identification rate at the protein level is similar between the methods, but the protein coverage was the highest for proteins identified in all three methods likely due to the higher abundance of the proteins, better compatibility with the bottom-up proteomic workflow, and the complementary nature of the above methods for peptide identification that has been described in the literature.^32,33 Peptide group identifications indicated even more complementarity and orthogonality between the CE-MS/MS method and nLC-MS/MS methods, with ~21% of all identified peptides being found uniquely by CE-based methods and only ~12% identified by all three methods (Figure 2C). While the depth of proteomic profiling provided by the CE-MS technique was somewhat lower than in the nLC-MS analyses at the used sample amounts, the fraction of peptides uniquely identified by CE-MS was significant (~26% of the IDs resulted from both LC-MS-based analyses). The two nLC-based methods showed a relatively low (~46%) overlap in peptide identifications mainly due to the differences in the stationary phases and the flow rates (Figure S3).

Investigation of the Coverage of Post-translationally and Non-enzymatically Modified Peptides.

We assessed the capability of the CE-MS for profiling common, biologically significant PTMs, including O-phosphorylation and protein N-terminal acetylation, as well as other modifications that may be related to sample preparation and its common artifacts such as methionine oxidation and carbamidomethylation of cysteine residues. The CE-MS/MS method identified substantially more phosphopeptides and N-terminal acetylated peptides than either RP-nLC method, with 53% of all identified phosphopeptides and 42% of all identified N-acetylated peptides found only in the CE-MS/MS data (Figure 3A,B). In total, CE-MS/MS was able to identify 308 phosphopeptides and 455 N-terminal acetylated peptides across all three analyses (Figure 3A,B) without any prior phosphoenrichment of the sample. Phosphoenrichment may boost the number of phosphopeptides identified, as has been reported by several other groups.^9,10,26 However, any enrichment techniques, where the sample is exposed to additional surface areas and chemistries and transfer steps, may result in sample losses and decreased profiling sensitivity for low-ng and sub-ng samples. A comparison between carbamidomethylated peptides, however, demonstrates similar complementarity between CE-MS/MS and nLC-MS/MS with only 11% of carbamidomethylated peptides identified by all three methods and 29 and 23% uniquely identified by CE-MS/MS or monolith nLC-MS/MS, respectively (Figure 3C). Only 4% overlap is observed for identified peptides with methionine oxidation (Figure 3D). To rule out concerns of higher incidence of oxidation in CE-MS compared to nLC-MS because of the differences in nESI interfaces, we investigated the percentage of peptides with oxidation modifications identified in each method. The percentage of oxidized peptides found in the CE and bead-packed nLC data sets was comparable at 2.2 and 2.1%, respectively. A slightly higher percentage of oxidized peptides was identified in the monolith nLC-MS/MS experiments (Figure S4). Similarly, the percentage of carbamidomethylated peptides is comparable in the CE and bead-packed nLC data sets and higher in the monolith nLC data. However, the fractions of phosphorylated and N-terminal acetylated peptides are highest in the CE-MS/MS data (Figure S4).

Figure 3.

Assessment of post-translationally and non-enzymatically modified peptide identifications for the HeLa protein digest standard for sample amounts indicated in Figure 2. Comparison of (A) phosphorylated, (B) N-terminal acetylated, (C) carbamidomethylated, and (D) oxidized peptides identified from ~10 ng of the HeLa protein digest standard by each separation method. Total numbers of modified peptides identified across three replicates of each method are indicated next to the colored circles. Overlapping regions indicate peptides identified by both (or all) methods. Calculated percentages indicate the percentage of the total represented by each region. CE-MS/MS migration times of (E) N-terminal acetylated indicated by pink dots and (F) phosphorylated peptides indicated by red dots.

Trends in Migration times for Modified and Miscleaved Peptides in CE-MS/MS.

To further investigate why some peptide modifications have better coverage in CE-MS/MS, we examined trends in migration times for modified peptides. Our results show that the majority of phosphopeptides and N-acetylated peptides found in CE-MS analyses expectedly migrate later than most other peptides (Figure 3E,F), while the same trend is not observed for these types of modified peptides in nLC, as also expected (Figure S5). This migration pattern in CE is expected since the addition of a phosphate group to a peptide adds a negative charge to the peptide and acetylation at the N-terminus removes a positive charge, either of which leads to a decrease in the overall net charge of the peptide and consequently the effective mobility of the peptide in CE, a phenomenon that has been described in previous studies.^9,34 In data-dependent MS acquisition, this shift in modified peptide migration times helps increase the identification rate by separating phosphopeptides and N-terminal acetylated peptides from the bulk of other peptides. Both phosphorylated and acetylated peptides are known to show lower ionization efficiency than unmodified peptides (Figure S6). Therefore, the later elution of these modified peptides after the bulk of other peptides improves the detection sensitivity of these peptides by minimizing ion suppression and co-isolation effects.

A noteworthy trend in CE migration time is also observed for peptides containing one or more missed cleavage sites. We observe from the CE-MS data that most peptides containing one or more missed cleavage sites migrate faster than most other peptides (Figure 4A). Since trypsin cleaves the protein at the C-terminus of lysine and arginine residues, missed cleavages in tryptic digest result in longer peptide sequences that contain an extra basic residue, therefore raising the overall net charge of the peptide. Expectedly, no trend in retention time was observed for miscleaved peptides in nLC-MS (Figure S7). Overall, the developed CE-MS method identified a higher number of peptides with missed cleavage sites than either of the nLC methods because of the same reason pointed out above regarding phosphorylated and acetylated peptides, that is, the miscleaved peptides were mostly separated from the bulk of other peptides due to the differences in the net charge. This finding raises a concern that LC–MS methods may introduce biases against detecting miscleaved peptides and misrepresent proteomic profiling results, especially in digestion efficiency assessment studies. The overlap of miscleaved peptides identified between methods was only 16%, again indicating a complementarity between the methods (Figure 4B). The foundation for this complementarity can be observed in the differences in net charge distributions between miscleaved peptides and non-miscleaved peptides identified in the CE- and nLC-based analyses (Figure 4C). For the examined LC- and CE-based methods, the distribution shifts towards a higher predicted net charge for peptides with one or two missed cleavage sites (Figure 4C), indicating that miscleaved peptides are more basic. The distribution also demonstrates a shift to higher MS-detected ion charge states for peptides with more missed cleavage sites expectedly due to the presence of more basic amino acid residues (Figure S8).

Figure 4.

Assessment of peptides with missed cleavage sites in CE-MS/MS and nLC-MS/MS from the HeLa protein digest standard. (A) Peptides with one or two missed cleavage sites in CE-MS/MS data indicated with blue for one missed cleavage site and orange for two missed cleavage sites in the scatter plot. (B) Comparison of peptides identified from ~10 ng (same amounts specified in Figure 2) of the HeLa protein digest standard with one or more missed cleavage sites between each method demonstrated by the Venn diagram. Overlapping regions indicate peptides identified by both (or all) methods. Calculated percentages indicate the percentage of all miscleaved peptides represented by each region. (C) Histograms show distributions of calculated net charges for peptides with 0, 1, and 2 missed cleavage sites. The y-axis refers to the density of peptides in each bin across the distribution for each respective histogram.

Since CE-MS/MS appeared to show better coverage of acidic and basic peptides, we also investigated the MS-detected ion charge states of peptides identified in CE-based methods compared to those found in nLC-based methods. In general, CE-MS/MS shows a higher percentage of identified peptides with ion charge states above 2⁺ than either nLC-MS/MS method with 38% of all peptides identified as 3⁺ ions in CE and 14 and 15% of peptides with ion charge state 3⁺ in bead-packed and monolith nLC methods, respectively (Figure 5A–C). A shift towards a higher ion charge state that is experimentally detected by MS is observed in peptides with higher calculated net charge in all evaluated LC- and CE-based methods (Figure 5E), which is expected since more basic peptides are prone to higher ionization efficiency in the positive ESI mode. In addition, our data show that peptides identified by CE-MS/MS have a wider distribution of net charge (Figure 5E), indicating that CE-MS/MS has advantages in identifying more acidic and more basic peptides over the tested RP-nLC-MS/MS methods. Additionally, higher molecular weight peptides appear to produce ions with higher charge states in all methods (Figure 5F). However, in CE-MS, there is a trend of higher ion charge states being observed for many of the early migrating peptides regardless of molecular weight (Figure 5D). As reported before,¹⁶ CE-MS/MS identified peptides with higher molecular mass than either of the nLC-MS/MS methods. Consequently, the distribution of sequence lengths for peptides uniquely identified by the CE-MS/MS method in this work is shifted toward longer sequence peptides with a maximum of around 19–20 amino acid residues, while each nLC-MS/MS method uniquely identified peptide distribution showed a maximum of approximately 10–11 amino acid residues (Figure 5F).

Figure 5.

Investigating peptide-detected ion charge state distributions from ~10 ng of the HeLa protein digest standard. Pie charts indicate overall charge state distributions for all identified peptides in (A) CE-MS/MS, (B) bead-packed nLC-MS/MS, and (C) monolith nLC-MS/MS. (D) Scatter plot of identified peptides molecular weight and migration times in CE-MS/MS indicates the charge state of the peptide by color. (E) Distribution of calculated net charge states for uniquely identified peptides in each method shown with stacked bars to demonstrate counts of each charge state in each bin. (F) Distribution of peptide sequence lengths in the number of amino acids is plotted for each method with a bin width of 1. Stacked bars indicate the detected ion charge states of the peptides at each sequence length.

Analysis of (a) Sub-Nanogram Amounts of the HeLa Protein Digest Standard and (b) 1–10 HeLa Cell Equivalents from 500 to 1000 Cultured and Processed HeLa Cells.

Next, we tested the performance of our CE method with low-nanogram and sub-nanogram amounts of the HeLa protein digest standard. While there are several methods for injection in CE,^35–37 we used a hydrodynamic injection performed by applying a specified pressure for a specified time. Yang et al. demonstrated the benefits of dynamic pH-junction online sample stacking strategy to increase sample loading,³⁸ and here, we have utilized a similar strategy to increase the injection volume to ~7% of the total volume of the capillary. With the injection conditions applied, 8800, 880, and 88 pg were calculated as loading amounts from 200, 20, and 2 ng/μL HeLa protein digest standard, respectively. Replicates (N = 4) of each loading amount were analyzed and showed an average of ~2000, ~1100, and ~160 protein group identifications, respectively (Figure 6A). Blank sample buffer injections were also analyzed between the digest runs to assess carryover. The acquired CE-MS/MS data resulted in identification of averages of 253 ± 13 and 33 ± 15 phosphopeptides and 396 ± 16 and 158 ± 44 N-acetylated peptides identified from 8.8 ng and 880 pg, respectively (Figure 6C).

Figure 6.

Proteomic profiling results from sub-nanogram HeLa protein digest standard and low numbers of digested cultured HeLa cells. (A) Average numbers (±STDEV) of protein groups, peptide groups, and PSMs identified from N = 4 replicates of HeLa protein digest standards. (B) Average numbers (±STDEV) of protein groups, peptide groups, and PSMs identified from N = 3 technical replicates of HeLa cell equivalents from the low-cell number digest. ~1 cell and ~10 cell equivalents injected from ~1000 cell digest and ~5 cell equivalents injected from ~500 cell digest as in panels (C–F). Gray bars represent the average of the first injection of 1 cell equivalent for N = 3 biological replicates as in panel (D). (C) Average phosphopeptide and N-terminal acetylated peptides identifications from four technical replicates of the ~8800, 880, and 88 pg HeLa protein digest standard. N-terminal acetylated peptides include peptides with N-terminal methionine truncation followed by acetylation as in panel (D). (D) Average phosphopeptide and N-terminal acetylated peptides identifications from three technical replicates of ~1 cell, ~5 cell, and ~10 cell equivalents. (D) Bar charts show average values (±STDEV) for protein identifications using different data processing workflows from approximately 1, 5, and 10 cell equivalents. All data are filtered for ≤1% FDR, but the Byonic data is also shown with proteins filtered for at least medium confidence (≤1% FDR) and high confidence (<0.1% FDR). The same Sequest HT results are shown for panels (C–F). (F) Average identifications of phosphorylated and acetylated peptides using Sequest HT or Byonic algorithm are shown for each cell equivalent level. (G) Bar charts show the percentage of proteins identified in each cellular component using GO term annotated lists generated from the UniProt human database. The data shown was filtered for p-value ≤ 0.001 for all cell amounts.

Proteomic analysis of low numbers of cells introduces additional challenges in achieving high sensitivity due to protein losses incurred during sample preparation.³⁹ To evaluate the capability of the developed CE-MS/MS platform for deep proteomic profiling of quantity-limited biological samples, we assessed the performance of our ultra-sensitive CE-MS/MS method on low numbers of cultured HeLa cells processed in our laboratory using a miniaturized in-solution digestion workflow. To minimize sample losses, all processing steps after washing and counting cells were completed in the CE injection vial in low microliter volumes (i.e., ≤20 μL), and desalting steps were omitted. Aliquots of ~500 cells and ~1000 HeLa cells were processed in this way, and dried protein digests were resuspended in sample buffer and directly analyzed by CE-MS/MS. Replicates of the estimated 1-cell, 5-cell, and 10-cell equivalents (corresponding to 0.1 and 1% of the total material resulting from digesting small cell populations, ≤1000 cells) were injected, keeping the injection volume the same for each CE-MS/MS analysis. Blank sample buffer was also injected and analyzed between the digest analyses to assess the carryover effects. Due to the absence of desalting steps in the sample processing method, abundant signals, presumably corresponding to singly charged ions of salts and adducts, were observed migrating from ~25 to ~34 min in the limited sample analyses (Figure S9). To mitigate the negative effects of the ionic build-up on the capillary walls and the nESI emitter, rinsing steps with sodium hydroxide and water were performed between analyses. Our laboratory has recently shown the benefits of coupling high-field asymmetric waveform ion mobility (FAIMS) with ULF LC-based high sensitivity proteomic analysis.⁴ Coupling FAIMS with our CE-MS/MS approach may limit the interference of salts and other common background ions that we observed in the presented here results. Even without filtering out interfering ions with the FAIMS interface or by other means, we demonstrate an average of 400 ± 72 proteins from a single cell equivalent, and ~800 ± 25 and ~1050 ± 21 proteins from 5-cell and 10-cell equivalents, respectively (Figure 6B), in three technical replicates of each cell equivalent. In each set of replicate CE-MS/MS analyses of single-cell equivalents (i.e., technical replicates) for all three biological replicates (i.e., three independently processed samples), a higher identification rate and better depth of proteomic coverage were achieved in the analysis that was conducted first (Figure 6B, Table S3). Thus, the numbers of identifications in analyses following the first one dropped by approximately 63, 54, 28, and 56% for proteins, peptides groups, phosphopeptides, and acetylated peptides, respectively. These results suggest that protein is lost over time, likely due to adsorption to the plastic walls of the sample vial.^40,41 Alternative materials used for sample vials are needed to avoid such losses in the future. Vials made from glass or with lower surface area contact with the sample, for example, result in less non-specific adsorption to vessel walls.^2,5 Modified peptides were also monitored in these experiments, showing averages of 4 ± 3, 33 ± 6, and 41 ± 3 phosphopeptides and 103 ± 32, 391 ± 23, and 439 ± 12 N-terminal acetylated peptides from 1, 5, and 10 HeLa cells, respectively (Figure 6D).

Next, to evaluate alternative data processing workflows and more fully characterize the potential of this limited sample analysis method, we analyzed these data using the novel INFERYS rescoring algorithm with Sequest HT, as well as Byonic software. All protein search results were filtered to ≤1% FDR for peptide identifications, and Byonic protein identifications were further filtered based on confidence levels (medium ≤1% FDR and high <0.1% FDR). The INFERYS rescoring algorithm calculates intensity-based similarity scores for PSMs identified by Sequest HT to improve the confidence in identification results, and it increased overall protein identifications for all cell amounts by ~20–30%, while the Byonic search algorithm identified ~60–80% more proteins from single-cell equivalents than Sequest HT alone and ~20–30% from the 5 and 10 cell equivalents (Figure 6E). Byonic identified an average of 744 ± 127 proteins from a single-cell equivalent and 1139 ± 24 and 1271 ± 6 proteins from 5 cell and 10 cell equivalents with at least medium confidence and averages of 702 ± 114 proteins from 979 ± 16 and 1038 ± 2 proteins from 1 cell, 5 cell, and 10 cell equivalents with high confidence (Figure 6E). Furthermore, Byonic was able to identify ~2.5–5 times more phosphopeptides and ~1.5–2 times more acetylated peptides than Sequest HT, demonstrating an average of 22 ± 8 phosphopeptides and 181 ± 47 acetylated peptides from single-cell equivalents, 84 ± 14 phosphopeptides and 549 ± 40 acetylated peptides from 5-cell equivalents, and 137 ± 15 phosphopeptides and 621 ± 22 acetylated peptides from 10-cell equivalents (Figure 6F).

Gene ontology (GO) enrichment analysis of these data demonstrated consistent representation of each cellular component, scaling expectedly with the cell amount. Notably, the percentage of proteins identified in each cellular component from the first injection of single-cell equivalent was similar to the coverage from the 5-cell and 10-cell equivalent data (Figure 6G). The results showed nearly 80% coverage of the cytosolic large ribosomal subunit (~44 of 57 total proteins annotated in this term) and cytosolic small ribosomal subunit (~33 of 42 total proteins annotated in this term) cellular components, even from the single-cell equivalent, and ~40% coverage of the ribonucleoprotein complex (~63 of 169 total proteins annotated in this term) cellular component (Figures 6G and S10). These findings suggest the potential application for this method in limited sample and single-cell profiling of ribosomal and ribosome-associated proteins and possibly phenotypic differentiation of cells based on their ribosomal specialization reflected in compositional and structural alterations of ribosomal proteomes. In recent years, the regulatory role of ribosomes and “specialized ribosomes” in translation has been explored to reveal high heterogeneity of ribosomal protein makeup and certain PTMs of ribosomal proteins that have been linked to human disease.^42,43 Profiling small sections of tissue or single-cell isolates from heterogeneous populations using the CE-MS/MS method described in this work could help drive further understanding of ribosome heterogeneity and its effects on translation and potentially identify markers of ribosomal makeup that are associated with disease.

CONCLUSIONS

Here, the complementarity between CE-MS/MS and RP-nLC-MS/MS methods at the peptide level has been thoroughly evaluated. The superior coverage of phosphorylated and acetylated peptides indicates the potential for CE-MS/MS to be used in phosphoproteomics and analysis of acetylated and potentially other post-translationally modified peptides in limited samples without applying any enrichment techniques, particularly for analyses where high sensitivity is required. Improved identification of larger peptides and peptides with missed cleavage sites in CE-MS/MS indicate strengths in CE-MS/MS for specific targeted work or as a supplement to LC-based methods to increase protein coverage. CE-MS/MS may also have advantages for profiling of other post-translational and non-enzymatic chemical modifications not investigated in this work at high sensitivity in scarce samples.

Furthermore, we demonstrate a robust, reproducible, ultra-sensitive CE-MS/MS method for low-nanogram and sub-nanogram proteomic profiling using a commercially available CESI-MS system. This method can routinely identify 1091 ± 46 proteins and 4426 ± 182 peptide groups from ~880 pg of the HeLa protein digest standard without fractionation. From a ~1 cell equivalent injection of a miniaturized in-solution digest of ~1000 in-house cultured HeLa cells with no cleanup steps included or booster/carrier samples added, 744 ± 127 proteins and 3353 ± 719 peptide groups were reliably identified. To the best of our knowledge, these identification numbers are the highest reported by a CE-MS/MS method with such limited protein amount. Moreover, we demonstrated 41 ± 3 phosphopeptides and 439 ± 12 acetylated peptides identified, corresponding to 45 phosphorylated and 441 acetylated proteins, respectively, from 10-cell equivalents without any prior enrichment or pre-fractionation. Through GO term enrichment analysis, we showed a high coverage of ribosomal proteins even down to the level of single cell injections. These results suggest that the developed method is potentially suitable for the analysis of important biological, clinically relevant proteins from individual single cells processed either alone or using multiplexed approaches when multiple cells are individually processed, lysed, barcoded with tandem mass tags (TMT or similar), and combined and mixed with a carrier sample derived from typically over 100 cells of a relevant origin.⁴⁴

We believe that the results shown in this work demonstrate the potential for CE-MS/MS-based proteomics to make an impact in the cutting-edge limited sample and single-cell analysis. Also, further improvements in the method sensitivity may be achieved using other materials and improved techniques for reducing losses during sample processing and coupling CE to the same or more advanced MS instruments via a FAIMS interface to reduce interference from singly charged background ions. Additionally, it is important to note that the small injection volumes typically used in CE-MS/MS (~10–50 nL) allow the analysis of only a small fraction of the total sample in the vial in most applications. In this study, we were able to analyze up to 1% of HeLa cell digests from either 500 or 1000 cultured cells per injection (50 nL). Future improvements in sample processing and injection vessel formats that can enable processing of small sample amounts in nanoliter-scale volumes and reduce the minimum volume requirements for injection would greatly improve the capabilities of CE-MS/MS-based molecular profiling in the analysis of quantity-limited samples and single cells.