Introduction

The ability to identify functional proteins with post-translational modifications (PTMs) is a crucial element in understanding cellular function. For example, acetylation of acidic residues can impact protein function and acetylation of protein N-terminal amines can protect proteins from degradation [1]. Furthermore, the acetylation of primary amines on histones often regulates protein-DNA interactions [2,3,4,5]. Protein PTMs have emerged in the post-genomic era as critical features in regulating and diversifying the biological activity of proteins. However, identifying specific proteoforms with PTMs and understanding their function on individual proteins is currently limited by the lack of effective and accessible analytical methods. Therefore, there is a need to develop a high-throughput approach to characterize intact proteins and their modified proteoforms at the systems level.

Proteomics is a high-throughput bio-analytical approach to study proteins in a complex sample using mass spectrometry (MS)-based methods. Currently, there are two fundamental approaches to MS-based protein identification and characterization: bottom-up and top-down proteomics. Recent developments in bottom-up proteomics significantly expanded the depth of proteomic analysis, both qualitatively and quantitatively [6, 7]. Two independent bottom-up-based human proteomics studies confirmed ~ 20,000 protein-coding genes from human samples, which provided direct evidence for the actual translation of over 90% of putative protein-coding genes [8, 9]. The throughput of top-down proteomics has also improved significantly during the last decade due to the advancements in chromatographic separation and high-resolution MS instrumentation [10,11,12,13,14,15,16,17].

Reversed-phase liquid chromatography (RPLC) is the most prevalent separation technique in top-down proteomics. Long capillary column (i.e., 1 m length, 75 μm i.d.) ultra-high-pressure liquid chromatography has been coupled to high-resolution top-down mass spectrometry (UPLC-HRMS) for the analysis of intact proteins in complex mixtures [18, 19]. However, in cases of high complexity and large dynamic range, such as in human cell lysate, 1D-RPLC may not provide sufficient proteome coverage due to a limited sample loading capacity. To overcome this challenge, different LC separation formats are often combined to increase the separation power based on the orthogonality. LC techniques often used in proteomics include reversed-phase chromatography (RPLC), ion exchange chromatography (IEX), hydrophilic interaction chromatography (HILIC), hydrophobic interaction chromatography (HIC), size exclusion chromatography (SEC), etc. [20,21,22,23,24,25]. Several electrophoresis-based separation methods such as gel-eluted liquid fraction entrapment electrophoresis (GELFrEE) [26] and solution isoelectric focusing (sIEF) [27] can be applied as pre-fractionation methods for RPLC; however, these methods cannot be directly coupled to MS for detection. Capillary electrophoresis (CE) is a very promising electrophoresis-based separation technique that can be coupled after RPLC separation due to the small sample loading amount. CE can be directly coupled with MS either by the sheathless interface [28] or by a sheath-flow-based interface [29]. McCool et al. reported a three-dimensional platform, SEC-RPLC-CZE-MS/MS, that was applied to study intact proteins in E. coli lysate. A total of 5705 proteoforms from 850 proteins were identified using this method [14].

Recently, our lab developed a two-dimensional separation platform using high-pH and low-pH RPLC for top-down proteomics [23]. This method was applied to analyze intact proteins and proteoforms in E. coli cell lysate, and our results demonstrated that the platform provided high-resolution protein separation due to the comparatively good separation resolution and orthogonality between the two RPLC formats. Another advantage of this approach is the simple sample handling procedure which utilizes MS-compatible buffers. In this study, we further applied 2D pH RP/RPLC coupled with top-down MS to characterize intact proteins and proteoforms in human cell lysate (e.g., HeLa cell lysate). A fraction-to-fraction orthogonal selectivity analysis between high-pH and low-pH separations was performed. In addition, we compared different cell lysis sample preparation methods (i.e., with and without the removal of small proteins with a 10 kDa MW cutoff filter). In total, 2778 proteoforms from 628 proteins were identified and manually confirmed. A total of 20 different types of PTMs were characterized, including intact proteoforms with phosphorylation, lipoylation, and glutathionylation. Overall, our results suggested that the two-dimensional high-pH and low-pH RPLC-MS can be easily adapted to complex protein samples for deep proteoform characterization.

Materials and Methods

Chemicals and Reagents

All chemicals including LC-MS grade water and organic solvents were purchased from Sigma-Aldrich (Milwaukee, WI) unless noted otherwise. Jupiter particles used as column packing material were purchased from Phenomenex (Torrance, CA).

HeLa Cell Culture and Cell Lysate Preparation

The HeLa cells were incubated at 37 °C in a humidified atmosphere containing 5% CO2 in Dulbecco’s modified Eagle medium (DMEM) with 10% fetal bovine serum (FBS) and 2% penicillin-streptomycin. The growth status of the cells was monitored using a microscope. When the cells were 80–90% confluent, they were washed with ice-cold PBS buffer and scratched off the petri dish. The cells were centrifuged at 2500×g, 4 °C for 30 min. Ten plates (10 cm diameter) of HeLa cells were resuspended to 10 mL in lysis buffer (1 μM PMSF and 20 mM NaF in PBS, pH = 7.4). The cells were broken using a sonication homogenizer on ice for 30 min, and insoluble matter was removed by centrifugation at 15,000×g, 4 °C for 30 min. The lysate (unfiltered sample) was aliquoted and stored at − 80 °C. For a portion of the lysate sample, small proteins or degradation products were removed using a 10 kDa MWCO spin filter at 4 °C (two times with 25 mM ammonium bicarbonate), referred to as “filtered sample.”

High-pH RPLC Fractionation

The intact proteins from HeLa cell lysate were separated and analyzed using the 2D pH RP/RPLC-MS/MS platform as previously described [23]. Briefly, 1 mg of proteins was fractionated by an offline high-pH RPLC separation (pH = 10) using an XBridge Protein BEH C4 column (300 Å, 3.5 μm, 2.1 mm × 250 mm) from Waters, Inc. (Milford, MA). The mobile phase A (MPA) was 20 mM ammonium formate in water (pH = 10), and mobile phase B (MPB) was 20 mM ammonium formate in acetonitrile (pH = 10), the pH was adjusted to 10 using ammonium hydroxide. The unfiltered sample was separated using a gradient from 3 to 90% MPB over 60 min. The eluted proteins were fractionated into 24 2.5-min fractions starting from 16 min after sample injection (i.e., the first detected UV peak). Based on the results from the unfiltered sample, we fractionated the filtered sample into nine fractions (Supplementary Figure 1A). All the fractions were vacuum dried and resuspended to a final volume of 50 μL using 25 mM ammonium bicarbonate before injecting onto the second-dimension column.

Low-pH RPLC-MS Analysis

Ten micrograms of protein (filtered, unfiltered, or fractions from high-pH RPLC separation) from the high-pH fractions was loaded onto a trapping column (150 μm i.d., 10 cm length, Jupiter particles, 5 μm diameter, 300 Å pore size) and separated using a home-packed C5 RPLC capillary column (75 μm i.d., 70 cm length, Jupiter particles, 5 μm diameter, 300 Å pore size) on a modified Thermo Scientific (Waltham, MA, USA) Accela LC system [19]. Specifically, MPA was 0.01% TFA, 0.585% acetic acid, 2.5% isopropanol, and 5% acetonitrile in water. MPB was 0.01% TFA, 0.585% acetic acid, 45% isopropanol, and 45% acetonitrile in water [18, 30, 31]. A 200-min gradient from 10 to 65% of MPB was applied with a flow rate of 400 nL/min. The LC elution was sprayed from an etched capillary tip through a customized nano-ESI interface to an LTQ Orbitrap Velos Pro mass spectrometer (Thermo Fisher Scientific, Hanover Park, IL, Bremen, Germany, USA) [32]. The temperature of the inlet capillary was set to 275 °C and the spray voltage was 2.6 kV. Full MS scans used the resolving power setting of 100,000 (at m/z 400) with two micro scans. Collision-induced dissociation (CID) with a normalized collision energy of 35 eV was applied. The top 6 most abundant precursor ions in the full MS scans were selected for MS/MS fragmentation. The MS/MS data were obtained at a resolving power setting of 60,000 (at m/z 400) with two micro scans and an isolation window of 3.0 m/z. The AGC target was set as 5 × 105 for full mass scans and 3 × 105 for MS/MS scans. All the data was collected using Xcalibur 3.0 software (Thermo Fisher Scientific, Bremen, Germany).

Data Analysis

Raw data were deconvoluted and searched against the Swiss-Prot reviewed Homo sapiens protein database (published on May 2018, 20,355 proteins) using TopPIC Suite [33]. All parameters were set as follows: the error tolerance was 15 ppm; the maximum value for the unexpected mass shift was 500 Da; the maximum number of the unexpected mass shifts was 1. The decoy database was utilized to calculate PSM-level FDRs and proteoform-level FDRs in TopPIC. The E-value cutoff value was 8.81E−8 for PSM-level FDRs < 1% and 5.31E−9 for proteoform-level FDRs < 1%. For each dataset, we only considered proteoforms with E-values less than the proteoform-level FDR cutoff values. Proteoform identifications in different fractions were aggregated based on their retention time (~ 3 min normalized retention time window) and detected masses (15 ppm). No protein level FDR was calculated. All proteins with proteoforms that passed the E-value cutoff were reported. All identified proteoforms were also manually evaluated. MASH Suite [34] and ProSight Lite [35] were used for manual interpretation and spectrum presentation.

Results and Discussion

2D pH RP/RPLC Top-Down MS on HeLa Cell Lysate

In this study, we applied the previously developed 2D pH RP/RPLC separation and top-down MS to characterize intact proteins and proteoforms in human HeLa cell lysate. A total of 24 first-dimension fractions were analyzed with long-column high-pressure RPLC with top-down MS [23] and by SDS-PAGE (Figure 1a, b). In total, 2778 unique proteoforms from 628 proteins were identified and manually confirmed (Supplementary Table 1). Seventy-seven percent of the proteoforms were detected in one fraction, 14% of the proteoforms were detected in two fractions, and 9% of the proteoforms were detected in three or more fractions. Our results suggest that high-pH RPLC provides good separation power as the first dimensional separation.

Figure 1
figure 1

Protein and proteoform identification count of HeLa cell lysate. (a) First-dimension fractions performed by SDS-PAGE. (b) Protein and proteoform identification count in each fraction under high pH conditions. (c) Venn diagram of proteins identified using the 1D and 2D platforms. (d) Venn diagram of proteoforms identified using the 1D and 2D platforms

Full size image

The same sample was also analyzed using 1D low-pH RPLC top-down MS without pre-fractionation; 124 intact proteins and 225 intact proteoforms were identified. (Supplementary Table 2). We compared the results of 1D and 2D separation of the same sample (Figure 1c, d). We observed an increase in the number of proteoform identifications when 2D separation was used, and there are two possible reasons for this increase: (1) there is a higher chance that MS/MS analysis will be performed on low-abundance proteoforms in 2D analysis due to extended MS analysis time and (2) improved separation performance in 2D analysis reduces sample complexity and allows for the detection of proteoforms that cannot be detected in 1D analysis because many co-eluted species split the ion current. To evaluate the effects of improved separation, we performed two 1D RPLC experiments using E. coli cell lysate: one with a 70-min gradient (lower peak capacity) and one with a 200-min gradient (higher peak capacity). All the detected mass features were deconvoluted with a 1-min moving average window (i.e., all spectra in a 1-min window were summed and peak deconvolution was performed). We compared the detected mass features in different MW ranges (Supplementary Figure 1). In total, we detected 137 mass features > 5 kDa from the 70-min run, and 367 mass features > 5 kDa from the 200-min run. Our results suggested better separation facilitated the detection of more species and provided higher chances to identify them using MS/MS analysis. Previously, we demonstrated that 2D pH RP/RPLC separation of E. coli cell lysate has high peak capacity due to good separation power on both high-pH RPLC and low-pH RPLC. Here, we applied 2D pH RP/RPLC to human cell lysate resulting in significantly improved detection of intact mass features (2637 mass features > 5 kDa) compared to the 1D 200-min RPLC run.

We further evaluated two cell lysate preparation approaches for 2D pH RP/RPLC top-down MS analysis (i.e., with and without the removal of small proteins using a 10-kDa MW cutoff filter). Based on the 24-fraction identification results from the unfiltered sample, the filtered samples were fractionated into nine fractions (Supplementary Figure 2A) to reduce the MS operation time. A total of 1710 unique mass features > 5 kDa were detected in nine fractions vs. 2637 unique mass features > 5 kDa in 24 fractions. Using the same MS conditions, fewer mass features were detected when nine fractions were collected, but much less MS analysis time was required. We also compared the masses of detected proteoforms in different MW ranges (Supplementary Figure 2B). The MW distribution is similar between these two fractionation schemes, and both approaches have similar percentages of detected mass features less than 10 kDa. One possible reason for the discrepancy in percentage of proteoforms is that some proteins may degrade during the concentration process (i.e., speedVac) before MS analysis. If protein degradation during concentration is an issue, filter-based concentrators or an online 2D system can be adapted in the future to reduce sample degradation.

In general, current commercially available MS instruments are limited in their ability to detect large proteins (MW > 30 kDa) with low abundance due to the low S/N ratios inherent to mass measurement. In our study, only a very small portion of proteins > 20 kDa were detected even though many large proteins were observed in the same fractions by SDS-PAGE. Gel-eluted liquid fraction entrapment electrophoresis (GELFrEE) has been coupled with RPLC for deep intact human proteoform characterization [27, 36], and different fragmentation approaches such as UPVD [37] and Ai-ETD [13] have been coupled with GELFrEE for better sequence coverage of PTM localization in human samples. In addition, many proteoforms > 30 kDa were successfully detected in GELFrEE fractions that contained enriched large proteoforms using a 21T FTICR-MS [15]. By removing co-eluting small proteoforms, GELFrEE allows for the detection of large proteoforms with low S/N ratios in complex samples. However, GELFrEE often requires MS-incompatible detergents such as SDS as well as specific instrumentation. Therefore, GELFrEE is challenging to use online with MS analysis. SEC is another size-based separation that may enhance MS detection of large proteoforms and has been coupled online with MS analysis [38, 39]. However, SEC in general does not provide good separation resolution in the full MW range because the separation selectivity of SEC depends on the resin pore size. Our results suggested that both high-pH RPLC and low-pH RPLC provide good separation power, which allows for deep proteoform detection in complex samples. With an additional dilution step, our approach can be easily adapted into an online platform to improve the robustness and sensitivity of the 2D pH RP/RPLC. Additionally, larger proteins can be enriched prior to 2D pH RP/RPLC separation using size-based approaches such as GELFrEE or SEC for better detection of larger proteins.

Orthogonality Between High-pH RPLC and Low-pH RPLC for Intact Protein Separation

In bottom-up proteomics, it has been reported that high-pH RPLC only has semi-orthogonality with low-pH RPLC for peptide analysis [40, 41]. For example, more hydrophobic peptides tend to elute late regardless of high or low pH. Therefore, the separation window of the secondary dimension cannot be fully utilized, which limits the usage of online 2D pH RP/RPLC analysis in bottom-up proteomics. Our previous study in top-down proteomics suggested that high-pH RPLC had good orthogonality against low-pH RPLC [23]. To further assess if these two separation forms have semi-orthogonality or full orthogonality for intact proteoform analysis, we evaluated proteoform elution patterns in high-pH fractions (Figure 2). For each fraction (each column on Figure 2), a heatmap was generated using the relative number of uniquely identified proteoforms in each bin (10-min windows). It has been demonstrated that in an ideally orthogonal system, area coverage > 63% represents full orthogonality [40]. Our results suggested that proteoforms were distributed across the entire efficient elution window of the secondary dimension in most of the fractions (81% bins containing 5 or more mass features > 2.5 kDa). The good orthogonality between RPLC performed under different pH conditions can be explained by the change in charge distribution caused by the change in pH of the mobile phase [40, 42, 43]. Compared to peptide chains, intact proteins are more complicated with more dynamic and versatile changes in charge distribution under different pH conditions. These changes in charge distribution can significantly affect protein retention and elution orders [23, 44].

Figure 2
figure 2

The proteoform elution patterns in high-pH fractions. Each bin in the heatmap represents the number of detected mass features > 2.5 kDa in each 10-min elution window in the low-pH RPLC separation

Full size image

To further evaluate the effects of pI values on intact protein retention behavior in high-pH RPLC separation, we filtered all identified intact proteoforms from the nine-fraction samples. The pI values were calculated for each filtered proteoform, and the average pI values and their standard deviations were plotted for each fraction (Supplementary Figure 2C). The average pI values were between 4 and 7 for most fractions, which suggests that the pI value alone does not alter the retention behavior. Still, the average pI value (4.48) in fraction 1 was significantly lower than other fractions indicating that pI may have some effect on separation. At pH 10, most proteins with pI values around 4 carry multiple negative charges, which reduces their retention times. In general, we did not detect many proteins with pI values larger than 8. One possible reason is that proteins with high pI values do not carry many charges and may not be easily eluted from the high-pH RPLC column.

Identification and Characterization of Intact Proteins

Overall, a total of 20 different types of PTMs were characterized in the proteoforms detected using the 2D analysis (Supplementary Table 3). Only 6 types of PTMs were detected using 1D analysis. Several commonly observed PTMs were detected including phosphorylation, acetylation, and glutathionylation. We were able to identify 96 intact proteoforms with phosphorylation and 771 intact proteoforms with acetylation. Several less common functional PTMs were also detected using the 2D platform including lipoylation and octanoylation. To increase the number of proteoforms detected using 1D RPLC separation, a combined top-down and intact mass analysis, such as that developed by Smith’s group, may be utilized [45]. The Smith group proposed a strategy to combine both top-down MS and intact mass determination, which addresses the issue that some proteoforms may be detected, but not identified, using MS/MS. The mass shift between experimentally detected species in the same run was calculated, and experimental delta mass histograms were plotted (Supplementary Figure 3). Mass shifts (16 Da and 80 Da) were often detected in pairs in both nine-fraction analysis and 24-fraction analysis, which may correspond to oxidation and phosphorylation. Interestingly, although we identified 771 intact proteoforms with acetylation, we did not observe many proteoform pairs with 42 Da shifts. This may be due to the fact that many acetylated species we detected have static N-term acetylation.

We also conducted a comparison of the proteoforms detected using both the 1D and 2D platforms (Figure 3). We observed significant improvements of the S/N ratios with higher signal and lower background noise when 2DLC separation methods were used, which is consistent with previous literature [46]. For example, calmodulin-1 (Swiss-Prot, P0DP23) was detected in both 1D and 2D results, but there was a much higher S/N ratio in 2D (S/N = 647) than in 1D (S/N = 37) (Figure 3). Additionally, a proteoform of calmodulin-1 which lost three amino acids (i.e., TyrAlaLys) at the C-terminus was detected using 2D (S/N = 23) but not 1D separation. The C-terminus deletion mutants of calmodulin-1 were reported to reduce the Ca2+ binding ability because it had minor conformers [47]. This suggest that the C-terminus of calmodulin-1 is active in inter-domain communication while Ca2+ induces structural transition so the truncated proteoform detected is functionally relevant. To summarize, our results suggest that the 2D pH RP/RPLC separation and top-down MS can be applied to complex protein samples for deep proteoform characterization, especially to detect and characterize low-abundance proteoforms and PTMs. Additionally, the application of 2D LC separation improves S/N ratio, which increases the limit of detection of the analytes [46].

Figure 3
figure 3

MS detection of calmodulin-1 proteoforms. (a) Full MS spectrum and extracted ion chromatograms (EICs) of proteoforms after 2D separation. (b) Full MS spectrum and extracted ion chromatograms (EICs) of proteoforms after 1D separation

Full size image

Characterization of Intact Proteoforms with Phosphorylation

The 2D LC technique applied in our study increases the detection of phosphorylated proteoforms. Many low-abundance PTM-modified proteoforms were detected using the 2D platform with good S/N ratios, but were not observed in the 1D analysis. Two proteoforms of parathymosin (Swiss-Prot, P20962), with and without phosphorylation, were characterized in fraction 5 during the 2D analysis (Figure 4a). The MS/MS spectra elucidated mass differences between b51 and b53 ions indicating the possible locations of the phosphorylation (Figure 4b). We observed a decrease in the elution time of the phosphorylated proteoforms in both 1D and 2D results because the phosphorylation added an additional charge to the protein that may have decreased the hydrophobicity. Decreased hydrophobicity of the phosphorylated proteins weakened the binding of these proteins to the stationary phase in RPLC, so they often eluted earlier compared to non-phosphorylated proteins [48]. Additionally, some Fe[III] adducts were detected from phosphorylated parathymosin proteins (* in Figure 4a). The Fe[III] adducts may come from the stainless-steel parts of the high-pressure system, such as the frit and the union. To avoid the production of Fe[III] adducts and increase the sensitivity for phosphoprotein detection, a “metal-free” RPLC-ESI-MS platform could be adapted to our 2D pH RP/RPLC platform [18].

Figure 4
figure 4

MS detection and MS/MS identification of parathymosin proteoforms. (a) Full MS spectrum of proteoforms. The asterisk (*) shows the Fe[III] adduct. The EIC of each proteoform is shown. (b) MS/MS identification of proteoforms. The orange serine residue on the N-terminus is acetylated. The tyrosine residue highlighted in blue is phosphorylated

Full size image

Ribosomal proteins are involved in protein synthesis by providing control mechanisms for transcription and translation [49]. The 60S acidic ribosomal protein P2 (Swiss-prot, P05387, RPLP2) is coded by the gene RPLP2 and participates in the elongation step in protein synthesis and GTPase activation [50, 51]. Phosphorylation has been proven to stimulate interaction with eukaryotic translation elongation factor 2 (eEF-2) [51]. Phosphorylation of serine residues located near the C-terminus increases the affinity of RPLP2 for eEF-2. The phosphorylation of these residues also causes changes in eEF-2 activity, which suggests the C-terminus is involved in this functionality. The phosphorylation on the C-terminus of RPLP2 is also related to autoimmune disease. Autoantibodies against the C-terminus peptide of 60S acidic ribosomal proteins exist in 15% of systemic lupus erythematosus patients [52].

We observed a total of ten proteoforms of RPLP2 using 2D methods with modifications including phosphorylations, oxidations, and a sequence variation, Tyr- > Gln (Supplementary Figure S4). Figure 5 a shows the MS spectrum of 60S acidic ribosomal protein P2 proteoforms in fraction 14 from the 2D separation. The 60S acidic ribosomal protein P2 variations include the mature proteoform RPLP2_P0, and the proteoforms RPLP2_P1 and RPLP2_P2, which have 1 and 2 phosphorylations, respectively. Figure 5 b illustrates the identification and PTM characterization of RPLP2_P0, P1, and P2. The MS/MS spectra elucidated the mass differences between b96 and b107 ions of the three proteoforms indicating the possible locations of the phosphorylations. However, there was no fragmentation between Ser 102 and Ser 105, so the location of the phosphorylation of RPLP2_P1 could not be determined. The NET of proteoforms shifted from 119 to 126 min and the extracted ion chromatogram (EIC) of RPLP2_P1 was split into two peaks (i.e., 120.42 and 120.87). This splitting might suggest the existence of two proteoforms with one phosphorylation on Ser102 or Ser105. To determine the location of the phosphorylation, the fragmentation method should be optimized to enhance the sequence coverage using electron transfer dissociation (ETD) and/or UVPD in the future [53]. We also found that phosphorylated proteoforms were eluted later than non-phosphorylated proteoforms, which disagrees with previous observations [48]. This may be due to the fact that both phosphorylated sites were located on the highly flexible and acidic C-terminus of the protein.

Figure 5
figure 5

MS detection and MS/MS identification of RPLP2 proteoforms. (a) Full MS spectrum of proteoforms: RPLP_P0, P1, P2. The asterisk (*) indicates the Fe[III] adduct. The EIC of each proteoform is shown. (b) MS/MS identification of RPLP_P0, P1, P2. The serine residues highlighted in blue are phosphorylated

Full size image

Identification of Intact Proteoforms with Functional Modifications

The 2DLC technique studied here allowed for the identification of functional PTMs that have not been commonly studied. For example, octanoylation and lipoylation were characterized on the glycine cleavage system H protein (Swiss-Prot, P23434, GcvH). The GcvH protein was reported as a subunit of the glycine cleavage system (GCV). GCV is active in glycine and serine catabolism [54]. The GcvH protein shuttles the methylamine group of glycine from the P protein to the T protein [55]. We confidently identified two GcvH proteoforms with octanoylation and lipoylation (Figure 6), which have never been reported in previous top-down studies. The transit peptides of both proteoforms were cleaved suggesting that these proteoforms were mature and transferred to the mitochondrion. The difference between the y66 and y67 ions, observed via MS/MS fragmentation, suggests two different PTMs (octanoylation and lipoylation) are located at the same amino acid, Lys50, of the proteoforms. Octanoylation and lipoylation are rare modifications characterized by the addition of a heptane chain and a heptane chain ending in a 5-membered ring containing two sulfur atoms, respectively. Lipoylation plays a critical role in the function of GcvH. The oxidized form of the lipoylation serves as the acceptor of the intermediary methylamine group derived from glycine as the two sulfur atoms bind to the glycine residue [55, 56]. Lipoylation was also reported to bind with the substrates of other enzymes (e.g., pyruvate dehydrogenase and alpha-ketoglutarate dehydrogenase) [57]. Octanoylation is an intermediate of lipoylation. Jordan et al. published a mechanism proposing that octanoyl-ACP is converted to lipoyl-ACP in E. coli. In this mechanism, LipA lipoate synthase inserts two sulfur atoms at C6 and C8 on the octanyl chain [58, 59]. Our results demonstrated the first detection of octanoylation on GcvH on human proteins, which suggests that the lipoyl-GcvH was synthesized through a similar pathway in human biological systems.

Figure 6
figure 6

MS detection and MS/MS identification of GcvH proteoforms. (a) Full MS spectrum of GcvH proteoforms with octanoylation and lipoylation. (b) MS/MS identification of the GcvH with octanoylation and lipoylation. The lysine residue highlighted in orange is lipoylated. The yellow cysteine residues do not participate in disulfide bonds

Full size image

Another example of a rare PTM identified by our method is hypusine on the eukaryotic translation initiation factor 5A-1 (Swiss-Prot, P63241, eIF5A-1). eIF5A-1 is an mRNA-binding protein and is involved in translation elongation. It plays such an essential role in eukaryotic cell growth and animal development that its inactivation causes death in yeast [60, 61]. We characterized two proteoforms of eIF5A-1 containing a hypusine residue and acetylation at Lys47 (Figure 7). The hypusine was found on both proteoforms using a y110 and y98 ion with + 87.08 Da. Hypusine, a unique PTM only reported on eIF5A, is derived from polyamine spermidine [60]. Interestingly, the hypusine residue is highly conserved from yeast to mammals [62]. Hypusine has a longer side chain than lysine and promotes the activity of eIF5A by enhancing translation termination through stimulating peptide release from peptide bond formation at polyproline stretches and ribosome pausing sites [60]. It was reported that the basic charge from hypusine at Lys50 is important for acetylation at Lys47 [61]. The acetylation on Lys 47 was characterized by the mass difference of a y110 and y98 ion (+ 42.01 Da) from two proteoforms and has been previously reported in the literature [62]. The acetylation at Lys47 also regulates the activity of eIF5A. Previous research suggests a basic charge at Lys47 is critical for eIF5A activity [62]. Overall, the deep proteoform characterization made possible by 2DLC separation allows for the detection and quantification of functional PTMs and proteoforms, which are not commonly detected using other methods.

Figure 7
figure 7

MS detection and MS/MS identification of eIF5A-1 proteoforms. (a) Full MS spectrum of eIF5A-1 proteoforms with hypusine and 1 or 2 acetylations. (b) MS/MS identification of the eIF5A-1 proteoform with 1 acetylation on N-terminus. The cysteine residues highlighted in gray participate in a disulfide bond. The methionine residue highlighted in gray is removed. The yellow cysteine residues do not participate in disulfide bonds. The orange alanine residue on the N-terminus is acetylated. The Lys50 highlighted in blue is modified with a hypusine

Full size image

Conclusion

In our study, the 2D pH RP/RPLC-MS/MS platform was applied to study the human proteins and PTMs in HeLa cell lysate. The protein, proteoform, and PTM identifications were enhanced using the 2D platform compared to 1D. The 2D separation implemented here made possible the characterization of proteoforms with complicated PTM distribution, such as multiple phosphorylations. In addition, we presented the first detection of intact human GcvH proteoforms with rare modifications such as octanoylation and lipoylation. The 2D platform is “salt-free” with good orthogonality, making the platform easily optimized to online 2D methods. Size-based separation approaches such as GELFrEE or SEC can be used before the 2D pH RP/RPLC for large protein fractionation and characterization. Different buffer conditions and column materials can be evaluated to improve the elution of hydrophobic proteins with high pI values. To characterize specific proteins, the 2D platform may be coupled with other sample preparation methods, such as immobilized metal affinity chromatography for phosphoprotein study [63]. In summary, our application provides a comprehensive aerial view of proteins in human cancer cells. This technique may also be used with isobaric-labeling-based quantification (TMT or iTRAQ) and applied to other human proteomics studies.