Capillary zone electrophoresis-high field asymmetric ion mobility spectrometry-tandem mass spectrometry for top-down characterization of histone proteoforms

Abstract

Characterization of histone proteoforms with various post-translational modifications (PTMs) is critical for a better understanding of functions of histone proteoforms in epigenetic control of gene expression. Mass spectrometry (MS)-based top-down proteomics (TDP) is a valuable approach for delineating histone proteoforms because it can provide us with a bird’s-eye view of histone proteoforms carrying diverse combinations of PTMs. Here, we present the first example of coupling capillary zone electrophoresis (CZE), ion mobility spectrometry (IMS), and MS for online multi-dimensional separations of histone proteoforms. Our CZE-high-field asymmetric waveform IMS (FAIMS)-MS/MS platform identified 366 (ProSight PD) and 602 (Top-PIC) histone proteoforms from a commercial calf histone sample using a low microgram amount of histone sample as the starting material. CZE-FAIMS-MS/MS improved the number of histone proteoform identifications by about 3 folds compared to CZE-MS/MS alone (without FAIMS). The results indicate that CZE-FAIMS-MS/MS could be a useful tool for comprehensive characterization of histone proteoforms with high sensitivity.

1 ∣. INTRODUCTION

Histones, serving as the backbone of the nucleosome where the genomic DNA is packed along, are essential for the epigenetic gene regulation and the structural stability of the chromatin [1-6]. Histones are consisted of two parts: the core histones (H2A, H2B, H3 and H4) as the key components of the nucleosome and the linker histone (H1) as the linkage between nucleosomes. The long N-terminal tails of all four core histones, which are protruded from the nucleosome, have high diversity of post-translational modifications (PTMs) such as methylation, acetylation, phosphorylation and citrullination [2, 7]. “Histone code” was proposed to state that the PTMs of histone directly contribute to the alternation of chromatin structure and thereby influence the DNA transcription [6]. It is crucial to characterize histone PTMs for a better understanding of the epigenetic gene regulation. With the diverse PTMs and high heterogeneity of histones, fully deciphering the “histone code” using proteomics is super challenging [8, 9].

Top-down proteomics (TDP), middle-down proteomics (MDP), and bottom-up proteomics (BUP) are three popular ways applied into the decryption of histone code [10-15]. Unlike partial or full digestion of protein into peptides by MDP or BUP, TDP targets intact proteins for the utmost preservation of the PTM information of proteoforms via mass spectrometry (MS) and tandem mass spectrometry (MS/MS) [13, 15, 16]. To delineate rich PTMs of the histone, TDP is definitely the ideal choice among these three methods. In the meantime, TDP requires sufficient separation techniques to improve proteoform detection and identification, especially for complex proteoform mixtures. Reversed-phase liquid chromatography (RPLC)-MS has been frequently used for TDP of histones [17-21]. To advance the separation and throughput of histone variants for TDP, Tian et al. reported a weak cation exchange-hydrophilic interaction liquid chromatography (WCX-HILIC) platform for two-dimensional (2D) separations prior to MS analysis, leading to over 700 histone proteoform identifications [22]. Besides LC-MS, capillary zone electrophoresis (CZE)-MS has also been developed for TDP of histone proteoforms by our lab [11]. CZE provides efficient separation of proteoforms due to PTMs by their charge-to-size ratios and CZE-MS has been proven as a power analytical technique for large-scale delineation of proteoforms of histones, bacteria, brains, and human cancer cells [11, 23-27]. Usually, offline LC fractionation is coupled to CZE-MS/MS for TDP of a complex mixture (e.g., histones) to achieve an in-depth proteoform characterization [11]. Development of online multi-dimensional separation technique involving CZE-MS/MS will be invaluable for TDP of histone proteoforms and complex proteomes in general because of potentially much higher throughput and much less sample loss during sample transfer.

High-field asymmetric waveform ion mobility spectrometry (FAIMS) is drawing more attention as a highly efficient gas-phase online separation technique, which could significantly reduce the background chemical noise, improve sensitivity, and fractionate ions based on their mobility differences under the asymmetric oscillation between high and low electric fields [28-31]. In FAIMS device, a compensation voltage (CV) is uniquely applied to the inner electrode and amends the trajectory of passing ions based on their mass, charge and shape. FAIMS has been online coupled with RPLC-MS/MS for TDP to efficiently fractionate proteoform ions based on their masses via applying different CVs [32-35]. Also, Pham et al. reported customized tandem nonlinear and linear IMS (FAIMS-TIMS) coupled to MS for bettering the characterization of human histone H2A and H4 proteoforms [36, 37]. Therefore, we expect that coupling FAIMS to CZE-MS/MS to build an online 2D platform will be useful for further advancing TDP of complex samples, for example, histones.

Here, we present the first example of 2D-CZE-FAIMS-MS/MS for TDP of histones. We optimized the CZE separation of histone proteoforms by adjusting the pH of background electrolyte (BGE) and sample buffer. We achieved nearly 400 and 600 histone proteoform identifications by CZE-FAIMS-MS/MS analyses of a commercial calf histone sample with nine different FAIMS CVs by using two different data analysis tools (ProSight PD and TopPIC Suite).

2 ∣. MATERIALS AND METHODS

2.1 ∣. Materials and reagents

Histone extract (from calf thymus, Product No. 10223565001) and all chemicals were obtained from Sigma-Aldrich (St. Louis, MO) unless stated otherwise. Acetic acid, formic acid, methanol, LC/MS grade water, and ammonium hydroxide were ordered from Fisher Chemical (Hampton, New Hampshire). Ammonium acetate (NH₄OAc) was purchased from Invitrogen (Waltham, MA). Acrylamide was purchased from Acros Organics (NJ, USA). Fused silica capillaries (50 $\mu \mathrm{m}$ i.d./360 $\mu \mathrm{m}$ o.d.) were purchased from Polymicro Technologies (Phoenix, AZ).

2.2 ∣. Capillary coating

The linear polyacrylamide (LPA) coated capillary was prepared according to the previous publications [38, 39]. In brief, a one-meter-long fused silica capillary (50 $\mu \mathrm{m}$ i.d. 360 $\mu \mathrm{m}$ o.d.) was successively flushed by 1 $\mathrm{M}$ sodium hydroxide (NaOH), water, 1 $\mathrm{M}$ hydrochloric acid (HCl), water, and methanol, followed by overnight nitrogen flow. Then, the capillary was treated with 50% (v/v) 3-(trimethoxysilyl) propyl methacrylate in methanol at room temperature (RT) for 24 h. Next, the capillary was flushed with methanol and dried under overnight nitrogen flow. For coating inner wall of capillary, 500 $\mu \mathrm{L}$ of 4% (w/v) acrylamide in water was mixed with 3.5 $\mu \mathrm{L}$ of 5% (w/v) ammonium persulfate (APS) in water, followed by a 15 min degassing procedure using nitrogen flow. This mixture was then introduced to the capillary using a vacuum suction. Both ends of capillary were sealed before the incubation in 50°C water bath for 1 h. At last, the capillary was flushed with water to remove any residue reactants and kept at RT before use.

2.3 ∣. Sample preparation

The histone extract was dissolved in 50 mM NH₄OAc (pH 6.5 or pH 9.0) to prepare 2 mg/mL histone samples for CZE-MS/MS analysis.

2.4 ∣. Optimization of CZE conditions for CZE-MS/MS of histones

CZE-MS/MS platform was built up by connecting a CESI 8000 Plus CE system (Sciex) to an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher Scientific) with an in-house constructed electrokinetically pumped sheath-flow nano-electrospray ionization (nanoESI) interface [40, 41]. A glass electrospray emitter with orifice size ranging 30–35 $\mu \mathrm{m}$ was pulled using a Sutter P-1000 flaming/brown micropipette puller, and the sheath liquid contained 0.2% (v/v) formic acid and 10% (v/v) methanol in water. The electrospray voltage was set at 2.2–2.4 $\text{kV}$ to the sheath liquid reservoir for ionization. The inlet of the capillary was installed in the cartridge of the CE system and the outlet was fit into the glass electrospray emitter (around 0.5 mm to the emitter tip), following the previous procedures [41].

For the mass spectrometer settings, the ion transfer tube temperature was set to 320°C, and the RF lens was 60%. The application mode was set to intact protein mode with low C-trap pressure. The isolation window was 0.7 m/z for isolating parent ions for high-energy collision dissociation (HCD).The normalized collision energy (NCE) of HCD was 25%. The MS/MS experiments were performed using data-dependent acquisition (DDA). Full MS scan was performed with the following parameters: orbitrap resolution of 480,000 (at m/z of 200), m/z range of 300–2000, normalized AGC target of 300%, microscans of 1. The top 6 most intense precursors with charge states in the range of 5–60 in full MS spectra were isolated and fragmented, and the threshold of precursors was set at 10,000. Other parameters for MS/MS include the resolution of 120,000 (at m/z 200), m/z range of 100–1500, microscans of 3, normalized AGC target of 100%, auto maximum injection time and dynamic exclusion of 30 s.

To achieve a better histone proteoform separation for more proteoform identifications, we optimized the BGE of CZE. Two BGEs were evaluated, and they were 5% (v/v) acetic acid (pH 2.4) and 20 mM NH₄OAc (pH 5.0). For the BGE 5% (v/v) acetic acid (pH 2.4), we used a sample buffer 50 mM NH₄OAc (pH 6.5) for dynamic pH junction sample stacking [39]. For the 20 mM NH₄OAc BGEs (pH 5.0 achieved by adding acetic acid), we chose the 50 mM NH₄OAc (pH 9) as the sample buffer to maintain the pH difference between sample buffer and BGE for efficient dynamic pH sample stacking. A 1-meter long LPA-coated separation capillary (50 $\mu \mathrm{m}$ i.d./360 $\mu \mathrm{m}$ o.d.) was used for the project. The histone sample was injected under 5 psi for 5 $\mathrm{s}$ to introduce about 25 nL for each run (50 ng loaded). Then, a 30 $\text{kV}$ voltage was applied for separation. After the separation, 30 $\text{kV}$ voltage and 15 psi pressure were applied for 10 min to clean up the capillary.

2.5 ∣. CZE-FAIMS-MS/MS

All CZE separation conditions of histones and mass spectrometer settings were the same as that mentioned above unless stated otherwise. For the CZE separation, the BGE was 20 mM NH₄OAc (pH 5.0). The sample injection volume was about 25 nL and about 50-ng histone was loaded for each run. The separation voltage was 30 kV, and the separation time was 50 min.

For the FAIMS fractionation, FAIMS Pro Duo interface (Thermo Fisher Scientific) was installed prior to the mass spectrometer. After the auto DV tune, FAIMS Pro Duo interface was set to standard resolution and the nitrogen carrier gas was set as default (4.6 L/min). Different CV voltages (−60 V, −50 V, −40 V, −30 V, −20 V, −10 V, +10 V, +20 V, and +30 V) were individually tested for triplicate CZE-FAIMS-MS/MS runs to examine the fractionation performance of the FAIMS.

For the mass spectrometer settings, intact protein mode, low C-trap pressure and isolation window 0.7 m/z for MS/MS were employed. 28% HCD energy was applied for FAIMS CV ranging from −60 V to −40 V, and 30% HCD energy was applied for FAIMS CV ranging from −30 V to −10 V, and 35% HCD energy was applied for FAIMS CV ranging from +10 V to +30 V. The top 6 most intense precursors with charge states in the range of 5–60 in full MS spectra were isolated and fragmented for FAIMS CV ranging from −50 V to +30 V, and precursor charge states in the range of 3–60 were selected for FAIMS CV at −60 V.

2.6 ∣. Data analysis

Proteome Discoverer 2.2 software (Thermo Fisher Scientific) with the ProSightPD 1_1 node for TDP was used for database search [42]. The detailed database searching setup was the same as our previous work [11]. Briefly, the MS1 spectra were firstly averaged using the cRAWler algorithm in Proteome Discoverer. The precursor m/z tolerance was set to 0.2 m/z. For both precursor and fragmentation Xtract parameters, the signal-to-noise ratio threshold, the lowest and the highest m/z were set to 3, 200, and 4000, respectively. Then deconvolution was performed by the Xtract algorithm followed by database searching against a Bos taurus database (downloaded from http://proteinaceous.net/-database-warehouse-legacy/in April 2018). A three-prone database searching was performed: (1) a search was performed with a 2-Da and 10-ppm mass tolerance of absolute mass for MS1 and MS2, respectively; (2) a subsequent biomarker search was performed to find unreported truncated proteoforms with 10 ppm tolerance for both MS1 and MS2; (3) the last search was performed with a 1000-Da mass tolerance for MS1, and a 10-ppm mass tolerance for MS2 for matching unexpected PTMs. The target-decoy strategy was exploited for evaluating the false discovery rates (FDRs) [43, 44]. FDR estimation was performed for each of three search strategies. The identified proteoform-spectrum matches (PrSMs) and proteoforms were filtered using a 1% FDR.

For the CZE-FAIMS-MS/MS data, the raw files for FAIMS CV ranging from −60 V to −10 V were searched by ProSightPD, and the raw files from CV ranging from +10 V to +30 V were analyzed with the TopPIC (Top-down mass spectrometry based proteoform identification and characterization) software (version 1.6.2) [45]. The raw files were firstly converted to mzML files with the MsConvert software [46], and spectral deconvolution was performed with the TopFD (Top-down mass spectrometry feature detection) software, generating msalign files, which were used as the input for database searching using TopPIC. The spectra were searched against a Bos taurus database (downloaded from Swiss-Uniprot, March 2022). FDRs were estimated using the target-decoy approach [43, 44]. A 1% PrSM-level FDR and a 5% proteoform-level FDR were employed to filter the identifications. The mass error tolerance was 15 ppm. The mass error tolerance was 1.2 Da for identifying PrSM clusters. The maximum mass shift was 500 Da. The maximum number of mass shift was set to 2.

All the CZE-MS/MS and CZE-FAIMS-MS/MS data were further analyzed by the TopPIC software (version 1.6.2) [45]. The parameters were the same as previously described except several differences. A 1% PrSM-level FDR and a 1% proteoform-level FDR were employed to filter the identifications. The identified proteoforms were further filtered by the E value lower than 0.001. The maximum variable PTM number was set to 5.

2.7 ∣. Experimental and Predicted Electrophoretic Mobility (${\mu }_{\mathbf{ef}}$)

We followed the procedure in our previous work for calculating the experimental and predicted ${\mu }_{\text{ef}}$ [11]. The experimental ${\mu }_{\text{ef}}$ was calculated by Equation (1),

$$\text{experimental}{\mu }_{\text{ef}}=\mathrm{L}∕((30-2)∕\mathrm{L}\times t_{\mathrm{M}})({\text{unit of cm}}^2{\text{kV}}^{-1}{\mathrm{s}}^{-1})$$ (1)

where $\mathrm{L}$ is the capillary length in cm, 30 and 2 are the separation voltage and electrospray voltage in $\text{kV}$. The Equation (1) is obtained from the literature [47, 48].The predicted ${\mu }_{\text{ef}}$ was calculated by Equation (2),

$$\text{predicted}{\mu }_{\text{ef}}=\text{ln}(1+0.350\times \mathrm{Q})∕{\mathrm{M}}^{0.411}$$ (2)

where $\mathrm{Q}$ is the number of charges of the proteoform in the BGE by counting the number of positively charged amino acid residues in the proteoform sequence (K, R, H, and N-terminus). $\mathrm{M}$ is the molecular mass obtained by MS measurement in Da. The Equation (2) is obtained based on previous publications [47-49].

3 ∣. RESULTS AND DISCUSSIONS

3.1 ∣. Optimization of CZE for better separation and identification of histone proteoforms

CZE separates histone proteoforms according to their electrophoretic mobilities, which relate to their charge to size ratios. Histones are super basic and are highly positively charged under our typical CZE BGE condition (i.e., 5% (v/v) acetic acid (~pH 2.4)) for TDP [50]. The pH of BGE will influence the charge of histone proteoforms and impact the CZE separations. A BGE with a higher pH value decreases the charge of histone proteoforms, resulting in potentially bigger differences in charge-to-size ratios of histone proteoforms, which eventually leads to better separation resolution and more histone proteoform identifications. To test our hypothesis, we studied two different BGEs: 5% (v/v) acetic acid (~pH 2.4) and 20 mM NH₄OAc (pH 5.0). We maintained the same sample injection volume and protein injection amount for the two conditions (25 nL and 50 ng). To maintain a sufficient pH difference between sample buffer and BGE for dynamic pH junction sample stacking, we employed 50 mM NH₄OAc (pH 9.0) as the sample buffer for the BGE of 20 mM NH₄OAc (pH 5.0). For the 5% (v/v) acetic acid (~pH 2.4) BGE, the sample buffer was 50 mM NH₄OAc (pH 6.5).

As shown in Figure 1, CZE-MS/MS using a BGE pH 5.0 produced a substantially wider separation window and better resolution for histone proteoforms compared to a BGE pH 2.4. The separation window of histone proteoforms was about 3 min for the BGE pH 2.4 and was about 9 min for the BGE pH 5.0. Histone H2A and H2B co-migrated under the pH 2.4 BGE condition, agreeing well with our previous data [11]. Interestingly, Histone H2A and H2B were well separated using a BGE pH 5.0. Due to the much better separation and dramatically wider separation window, CZE-MS/MS using a BGE pH 5.0 identified 60% (TopPIC) or 85% (ProSight PD) more proteoforms that are larger than 10 kDa than that using a BGE pH 2.4 in triplicate runs, Figure 1. CZE-MS/MS with a BGE pH 5.0 produced reproducible separations of histone proteoforms in terms of separation profiles and proteoform intensity, Figure S1. Considering the overall number of large intact histone proteoforms (over 10 kDa), the separation profiles, and proteoform intensity, the BGE pH 5.0 was used for all the following experiments. We observed that the histone proteoforms from pH 2.4 and 5.0 were substantially different, and only 16% of identified histone proteoforms were shared, Figure S2. We need to point out that decrease of separation voltage is another potential way to increase separation window for histone proteoforms, but this approach will most likely reduce the separation efficiency.

FIGURE 1.

Electropherograms of CZE-MS/MS analysis of histone proteoforms under different BGE conditions. Two different BGEs with pH 2.4 (5% (v/v) acetic acid) and pH 5.0 (20 mM NH₄OAc by adding acetic acid to achieve the pH) were studied. The peaks of H1, H2A, H2B, H3, and H4 are marked with red arrows. The total number of proteoform identifications from ProSightPD and TopPIC Suite, and the number of proteoforms larger than 10 kDa from triplicate analyses are labeled.

3.2 ∣. CZE-FAIMS-MS/MS as an online two-dimensional technique for characterization of histone proteoforms

Because of the extreme complexity of histone proteoforms, multi-dimensional (MD) separations are crucial for delineating histone proteoforms. Here, we integrated FAIMS to the CZE-MS/MS system to carry out additional proteoform separations in the gas phase based on the ion mobility principle between liquid-phase CZE and gas-phase MS separations. After initial liquid phase CZE separation, histone proteoforms are further online fractionated in the gas phase by FAIMS based on their charges and sizes prior to MS and MS/MS. For FAIMS fractionation, nine different CVs ranging from −60 V to +30 V with 10-V increments were studied (triplicate runs for each CV). The identified proteoforms from each CV by ProSight PD, and the identified proteoforms from combined search of all CVs by TopPIC Suite are listed in the Supporting information II and III, respectively.

As shown in Figure 2A, each CV displays its unique histone separation profile. Main peaks of H2A and H2B were detected from −40 V to −10 V and from −50 V to −20 V CVs, respectively. Interestingly, the main peak of H1 emerged at CV of −10 V and became the only one at CV of +20 V and +30 V. This is consistent with the database searching result using the TopPIC. Almost all the identified proteoforms at CV of +20 V and +30 V were from histone H1. Because the size of H1 (>20 kDa) is larger than that of H2A and H2B, this phenomenon is in agreement with the literature that protein ions are fractionated by FAIMS according to their masses [34, 35].

FIGURE 2.

(A) Electropherograms of CZE-FAIMS-MS/MS analysis of histone proteoforms under different CVs.CV values from −60 V to +30 V are listed from top to bottom. (B) Overlap of identified proteoforms of histones between FAIMS CVs. (C) Violin plots of mass distributions of identified histone proteoforms by different FAIMS CVs.

Totally, we identified 366 (from ProSight PD) and 602 (from TopPIC Suite) histone proteoforms with the combination of 9 CVs by CZE-FAIMS-MS/MS, and the number of histone proteoforms is improved by about 3 folds compared to that from CZE-MS/MS alone (without FAIMS) (366 vs. 113 proteoforms from ProSightPD, 602 vs. 194 proteoforms from TopPIC Suite). We previously coupled size-exclusion chromatography (SEC) to CZE-MS/MS for TDP of histones with the identification of about 400 histone proteoforms from the same calf histone sample [11]. Both offline 2D-SEC-CZE-MS/MS and online 2D-CZE-FAIMS-MS/MS are efficient for histone proteoform characterization. The unique advantage of online 2D-CZE-FAIMS-MS/MS is the much lower requirement for initial histone material compared to offline 2D platforms. The offline 2D-SEC-CZE-MS/MS used hundreds of micrograms of protein material to start the analysis and online 2D-CZE-FAIMS-MS/MS only required less than 10 $\mu \mathrm{g}$ of histone material to initiate and complete the analyses because it avoided any potential sample loss due to, for example, LC fraction collection and sample transfers. We expect the online 2D-CZE-FAIMS-MS/MS will be a powerful tool for TDP of mass-limited biological samples.

To further investigate the histone proteoform fractionation performance of FAIMS, we studied the histone proteoform overlaps between different CVs, Figure 2B. The proteoform overlap coefficients between any two different CVs became smaller when the CV difference increased. For example, the proteoform overlap coefficient was close to 0.4 between −60 V and −50 V CVs; the coefficient was reduced to lower than 0.2 between −60 V and −40 V CVs. The data suggests that FAIMS can perform proteoform fractionation efficiently in the gas phase. Next, the violin plots in Figure 2C show the mass distributions of histone proteoforms identified under each CV condition. It is clear that CZE-FAIMS-MS/MS with a larger CV tends to identify histone proteoforms with higher masses. The median mass of identified histone proteoforms increased from 3.4 kDa to 21.3 kDa when the CV was enlarged from −60 to +30 V. The results indicate that CV value of FAIMS and proteoform mass have a clear correlation in our CZE-FAIMS-MS/MS condition. The histone proteoform overlaps between different CVs and proteoform mass distributions across different CVs from TopPIC Suite (Figure S3) were in consistent with that from ProSightPD (Figure 2).

Besides global analyses of the histone proteoform data from CZE-FAIMS-MS/MS, we also tried to investigate the performance of FAIMS for separations of near isobaric histone proteoforms. For example, we found that the H2B type 1-N (Protein Accession: Q32L48) had two nearly isobaric proteoforms well separated by FAIMS. The one with two acetylation (Theo. MH+: 13868.4992 Da) was only identified by −40-V CV, while the one with one phosphorylation (Theo. MH+: 13864.4444 Da) was only identified by −20-V CV. The data suggests that CZE-FAIMS-MS/MS is promising for the characterization of isobaric or nearly isobaric histone proteoforms, which will be further systematically studied in our future work.

3.3 ∣. Electrophoretic mobility prediction of histone proteoforms

Recently, our lab published the first examples of accurately predicting proteoforms’ ${\mu }_{\text{ef}}$ by optimized semiempirical models using large-scale CZE-MS/MS datasets of E. coli, zebrafish brain, plant leaf, and calf histone samples [11, 47, 51]. This new approach could help validate the confidence of proteoform identifications by examining the correlation between predicted and experimental ${\mu }_{\text{ef}}$ of proteoforms. The previous works employed 5% (v/v) acetic acid (~pH 2.4) as the BGE in CZE–MS/MS analysis for denaturing histone proteoforms. Here, we employed an optimized BGE (20 mM NH₄OAc, pH 5.0) to better separations of histone proteoforms. Under the pH 5.0 condition, the histone proteoforms most likely tend to carry fewer positive charges and unfold less compared to the pH 2.4 condition, which could substantially influence the ${\mu }_{\text{ef}}$ prediction of histone proteoforms by the semiempirical model used in our previous studies [11, 47]. Here we further studied the ${\mu }_{\text{ef}}$ prediction of histone proteoforms under the two BGE conditions, pH 2.4 and pH 5.0. The details of calculating the predicted and experimental ${\mu }_{\text{ef}}$ of the proteoforms were described in the Materials and Methods part. Only histone proteoforms without unknown modifications were used in this study. Both CZE-MS/MS datasets from BGE 2.4 and 5.0 without FAIMS were utilized.

For the CZE-MS/MS dataset from the BGE pH 2.4, we first optimized the prefactor of ‘$\mathrm{Q}$’ and the power factor of ‘$\mathrm{M}$’ in Equation (2) using proteoforms without any PTMs by independent adjustment of the factors. We found that 0.218 as the prefactor of ‘$\mathrm{Q}$’ and 0.411 as the power factor of ‘$\mathrm{M}$’ produced the best linear correlation coefficient (R²) as 0.9824, Figure 3A. Then, the optimized ${\mu }_{\text{ef}}$ prediction equation was used for histone proteoforms with PTMs (i.e., acetylation and phosphorylation). Proteoform acetylation and phosphorylation were reported to reduce the charge ($\mathrm{Q}$) by roughly one unit according to our previous studies [47]. Some of the histone proteoforms with acetylation and/or phosphorylation were clearly off the trendline without charge $\mathrm{Q}$ corrections, Figure 3B. After we applied charge $\mathrm{Q}$ reduction for those acetylated and/or phosphorylated proteoforms, the linear correlation coefficient between predicted and experimental ${\mu }_{\text{ef}}$ increased from 0.9058 to 0.9548, Figure 3C. The nice linear correlations between experimental and theoretical ${\mu }_{\text{ef}}$ of identified histone proteoforms suggests high-confidence proteoform identifications in this study, which agrees with the P-Score distribution of proteoforms, Figure 3D. The histone proteoforms identified by CZE-MS/MS have P-scores centering around 10⁻²⁰. The low P-score value indicates confident proteoform identifications [50].

FIGURE 3.

Linear correlations between theoretical and experimental ${\mu }_{\text{ef}}$ of unmodified histone proteoforms (A) and unmodified plus phosphorylated and acetylated proteoforms (B) identified in a single CZE-MS/MS analysis under the BGE of 5% (v/v) acetic acid (pH 2.4). The black dots represent proteoforms without PTMs and light blue dots represent proteoforms with phosphorylation and/or acetylation. (C) Linear correlations between theoretical and experimental ${\mu }_{\text{ef}}$ of unmodified histone proteoforms plus phosphorylated and acetylated proteoforms after charge $\mathrm{Q}$ corrections due to PTMs. (D) Distribution of the −log(P-score) of the identified histone proteoforms by CZE-MS/MS.

We further studied the CZE-MS/MS dataset from the BGE pH 5.0, Figure S4. After optimizations, the best linear correlation coefficient (R² = 0.9417) for histone proteoforms without PTMs was obtained using a prefactor of ‘$\mathrm{Q}$’ as 0.277, Figure S4, which is substantially different from that for the BGE pH 2.4. The best linear correlation coefficient under BGE pH 5.0 is significantly lower than that under BGE 2.4 (0.9417 vs. 0.9824). We also explored the CZE-MS/MS dataset by Top-PIC Suite from both BGE pH 2.4 and pH 5.0, Figures S5 and S6. After applying charge $\mathrm{Q}$ reduction for acetylated and/or phosphorylated proteoforms, the linear correlation coefficient between predicted and experimental ${\mu }_{\text{ef}}$ increased from 0.8538 to 0.8963 at pH 2.4, while the linear correlation coefficient at pH 5.0 almost had no changes (from 8774 to 0.8792). We suspected that the rise of BGE pH to 5.0 led to a more folded condition of histone proteoforms and made accurate charge and size calculations more difficult, resulting in a lower correlation coefficient.

The results here demonstrate that the pH of BGE can strongly affect the prediction of histone proteoforms’ ${\mu }_{\text{ef}}$. To achieve better ${\mu }_{\text{ef}}$ prediction of histone proteoforms under BGE pH 5.0, more efforts need to be made regarding collection of much larger histone proteoform datasets and more systematic investigations of factors that potentially influence the charge and size of histone proteoforms.

3.4 ∣. Histone PTMs

The individual PTMs and the combination of diverse PTMs located on histone proteoforms are critical for the epigenetic control of gene expression. Some examples of histone proteoforms with PTMs identified in our CZE-FAIMS-MS/MS study by the TopPIC software are shown in Figure 4. All the four examples of histone proteoforms were identified with multi-PTMs under high confidence, less than 10-ppm mass errors and better than 1 × 10⁻¹³ E-values. Figure 4A,B show proteoforms of H2B type 1K and H2B type 1, while Figure 4C,D display two proteoforms from H2A type 1. The proteoform in Figure 4A has four PTMs at S6 (phospho), K15 (methyl), R86 to R92 (citrullination), L101 to A110 (methyl). The phosphorylation of H2B at S6 was reported to occur during the early mitotic phases and may prevent the chromosomal instability and aneuploidy [52]. The proteoform in Figure 4B has one phosphorylation at the S14 and one citrullination at $\mathrm{Q}22$. The phosphorylation of H2B at S14 was associated with the apoptotic chromatin condensation pathway and regulation of monoubiquitination of H2B [53, 54]. The proteoform in Figure 4C has N-terminal acetylation at S1, one citrullination at R3 and another citrullination between N68 and N89. The citrullination of histones by PADs (protein arginine deiminases) was reported to be correlated with both transcriptional activation and repression [55]. The proteoform in Figure 4D carries N-terminal acetylation and one phosphorylation at T101. Phosphorylation of H2A at T101 may play a crucial role in creating distinctive binding sites for DNA double-strand break (DSB) response proteins, and lead to alterations in the local chromatin structure [56]. We noted that ProSight PD also identified similar histone proteoforms to that shown in Figure 4B,C. ProSight PD did not identify the proteoforms shown in Figure 4A,D, which is most likely due to the fact that Top-PIC and ProSight PD employ drastically different database search strategies.

FIGURE 4.

Sequences and fragmentation patterns of four histone proteoforms (A–D) with various PTMs. (A) and (B) belong to H2B. (C) and (D) are from H2A. The data is from the TopPIC Suite.

We further compared some identified histone proteoforms carrying PTMs in this work with that identified in one recent study from the Brodbelt group, which employed ultraviolet photodissociation (UVPD) for TDP of the same calf histone sample as our study [57]. Several histone proteoforms with PTMs were identified with high fragmentation coverage by ProSight PD and highlighted in the Brodbelt group’s study. The first one was histone H4 carrying N-terminal acetylation, R3 dimethylation, and K12 acetylation (acH4R3me2K12ac). The second one was histone H4 with N-terminal acetylation and R3 dimethylation (acH4R3me2). The third one was histone H2A with N-terminal acetylation (acH2A). In our study, we also identified similar histone proteoforms by ProSight PD, Figures S7-S9. We identified three possible H4 proteoforms, acH4R3me2K5ac, acH4R3me2K8ac, and acH4R3me2K12ac, Figure S7. Due to the limited backbone cleavage at the N-terminus, it is impossible for us to distinguish those three H4 proteoforms. Figure S8 shows another possible H4 proteoform, acH4R3me2, identified in this study. We noted that there are other possible explanations regarding the PTMs in Figure S8 due to the limited fragmentation coverage at the N-terminus. Figure S9 shows a high-confidence identification of one H2A proteoform, acH2A, in this work. The data demonstrate that our histone proteoform data and the Brodbelt group’s data agree reasonably well regarding the specific histone H4 and H2A proteoforms discussed above.

4 ∣. CONCLUSIONS

We presented the first example of coupling CZE, IMS, and MS as a multi-dimensional platform for characterization of histone proteoforms with the identification of 366 histone proteoforms (from ProSightPD) and 602 histone proteoforms (from TopPIC Suite) using a low microgram amount of histone sample as the starting material. We revealed that the pH of BGE could affect the CZE separation and ${\mu }_{\text{ef}}$ prediction of histone proteoforms substantially. We documented that FAIMS is an efficient gas-phase separation method for histone proteoforms and can fractionate histone proteoforms according to their masses.

One limitation of our current CZE-FAIMS-MS/MS platform is the low backbone cleavage coverage for histone proteoforms with HCD fragmentation, which impedes accurate localizations of PTMs on histone proteoforms. We expect that the integration of alternative gas-phase fragmentation techniques like UVPD [57], electron capture dissociation [58-61] and electron transfer dissociation [62, 63] will drastically benefit the characterization of histone proteoforms.

Significance Statement.

Histones contain a super complex mixture of proteoforms carrying various combinations of post-translational modifications (PTMs). Accurate delineation of histone proteoforms is crucial for bettering our understanding of epigenetic gene regulation and requires analytical tools with high capacity and high sensitivity for histone proteoform separation and detection. In this work, we developed an online multi-dimensional platform for characterization of histone proteoforms via coupling capillary zone electrophoresis (CZE), ion mobility spectrometry (IMS), and MS for the first time. The platform identified hundreds of histone proteoforms from a commercial calf histone sample using only several micrograms of histone proteins as the starting material. We expect the CZE-IMS-MS and MS/MS platform will be valuable for advancing histone proteoform characterization.

Abbreviations:

CZE

capillary zone electrophoresis

FAIMS

field asymmetric waveform ion mobility spectrometry

MS/MS

tandem mass spectrometry

PTMs

post-translational modifications

TDP

top-down proteomics.

DATA AVAILABILITY STATEMENT

The MS proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [64] partner repository with the dataset identifier PXD041357.