Ion Mobility Separation of Variant Histone Tails Extending to the “Middle-down” Range

Abstract

Differential ion mobility spectrometry (FAIMS) can baseline-resolve multiple variants of post-translationally modified peptides extending to the 3 - 4 kDa range, which differ in the localization of a PTM as small as acetylation. Essentially orthogonal separations for different charge states expand the total peak capacity in proportion to the number of observed states that increases for longer polypeptides. This might enable resolving localization variants for yet larger peptides and even intact proteins.

Introduction

Proteins are replete with various post-translational modifications (PTM) that often govern their biological function.^1-4 As proteomic technologies mature, research shifts to the determination, detection, and quantification of 3-D protein structures (conformations) and PTMs. Complete characterization of PTMs includes their precise localization on the protein, as multiple variants (some with a PTM transposed by just one residue) generally coexist and have differing activity.^5-9 Proteolytic digestion in “bottom-up” or “middle-down” proteomic methods translates such variants into isomeric peptides. Variants are normally identified by tandem mass spectrometry (MS/MS) via collision-induced dissociation (CID) and/or electron capture/transfer dissociation (EC/TD). Constraints of this approach are: (1) major sensitivity losses due to common preference for facile PTM elimination (over informative backbone severance) in CID and low efficiency and largely indiscriminate fragmentation in EC/TD that partitions the useful ion signal into multiple channels,⁴ (2) zero or very low EC/TD yield for smaller peptides with 1+ and 2+ charge states or peptides with “electron predator” PTMs such as nitrate,¹⁰ (3) difficulty of distinguishing the PTMs on a basic and adjacent residue (including the two N-terminal residues), as the bond between them is hard to cleave,¹¹ (4) possibility of missing or mistaken assignments caused by shifts of labile PTMs or other isomerizations during the CID process,¹² and (5) fundamental inability to distinguish some variants when the number of options n exceeds two because (n – 2) peptides yield no unique-mass fragments in either CID or EC/TD. For example, a mixture of X₁mZX₂ZX₃ZX₄, X₁ZX₂mZX₃ZX₄, and X₁ZX₂ZX₃mZX₄ (where X_1-4 and Z are arbitrary amino acids and m is a modification of Z) produces unique fragments for the 1^st and 3^rd, but not the 2^nd peptide.¹³ Variant mixtures are more challenging and usually must be separated prior to MS analysis, as the competitive fragmentation of same species frequently masks multiple variants in CID data. However, chromatographic and electrophoretic separations of localization variants require lengthy gradients and often fail, especially for cases of alternative PTMs on proximate residues where MS/MS is most problematic.¹⁴

Localization variants of modified peptides were recently resolved employing ion mobility spectrometry (IMS), including conventional (drift-tube) IMS¹⁵ based on absolute mobility (K) at low electric field intensity (E) and differential or field asymmetric waveform IMS (FAIMS)^13,16-18 that exploits the difference between K values at high and low E. Unlike dispersive drift-tube IMS, FAIMS is a filtering technique: species with a given derivative of K(E) over a range of E are selected while pulled through a gap between two electrodes carrying the waveform.^19,20 Scanning a dc voltage (the compensation voltage) superposed on the waveform, commonly expressed²⁰ as the compensation field (E_C), generates the FAIMS spectrum. The initial efforts^13,16,17 have focused on phosphorylation, perhaps the most common and important PTM. Drift-tube IMS (with ion funnel interfaces to the preceding ion source and subsequent MS stage) is exceptionally sensitive, but, at the resolving power (R) of ~80, separated the phosphopeptide variants only in part.¹⁵ The commercial FAIMS stage with cylindrical gap geometry and thus inhomogeneous electric field provides a much lower²¹ R ~ 10 (for peptides) and, despite FAIMS being more orthogonal to MS than conventional IMS is,^22,23 the separation is even worse.¹⁵ Planar-geometry FAIMS systems utilize homogeneous fields that permit much greater R, up to ~300 for multiply-charged peptides using the He/N₂ or H₂/N₂ gas mixtures.^24-26 This performance has allowed routine baseline (and frequently better) separation of variant phosphopeptides,^16,17 including those from a real protein.¹⁷ The degree of separation appears uncorrelated with the attachment site (S or T), number of PTMs, or the PTM shift distance: mono- and bis-phosphorylated variants and those with alternative modifications on adjacent and remote sites are resolved equally well.¹⁷ This renders FAIMS especially attractive for disentangling variants with multiple PTMs and/or small PTM shifts that challenge MS/MS.

One concern is the transferability of approach to other PTMs. With low-resolution FAIMS, variant glycopeptides (with O-linked N-Acetylgalactosamine, GalNAc) were partly separated¹⁸ about as well as phosphopeptides. However, GalNAc (221 Da) is larger than phosphate (80 Da) while the underlying peptide was smaller (1501 vs. 1649 Da). One would expect more difficult separations with smaller PTMs such as acetylation (42 Da) and/or for larger peptides. To probe both issues in a biological context, we have investigated histone tails where ubiquitous variant acetylations and other PTMs are thought to contribute to the “histone code”.^27,28 In particular, H4 permits ~3×10⁶ combinations for just the known histone PTMs and attachment sites, of which <80 were observed so far.^29,30 Such tremendous diversity of histones makes their analysis a topical application for IMS-MS.

Experimental methods

We have utilized the custom planar FAIMS device with gap width of 1.88 mm and length of ~50 mm, coupled to an ion trap MS analyzer (Thermo LTQ) via electrodynamic funnel interface.^21,24-26 Operational parameters were close to those in earlier studies using He/N₂ and H₂/N₂ gases^16,17,24,26 with the waveform amplitude (dispersion voltage, DV) of 5.4 kV and E_C scan rate of ~2.5 V/(cm × min). The gas flow to FAIMS unit was 2 L/min, translating into the separation time of t ~ 0.2 s. The mono-emitter ESI source was biased at ~1.6 kV above the FAIMS inlet, with the ~2 - 4 uM solutions in 50/49/1 methanol/water/acetic acid (pH = 2.9) infused at ~0.5 uL/min. Acetylation occurs on lysines, and the human H4 histone tails (SGRGK⁵GGK⁸GLGK¹²GGAK¹⁶RHRK²⁰VLRDN²⁵ …) were found acetylated in vivo on K5, K8, K12, and K16 (but not K20).³⁰ These four monoacetylated variants of above 1 - 25 segment with the biotinylated terminal GSGSK linker were purchased from Anaspec (Fremont, CA). With 30 residues total and (monoisotopic) mass of 3273 Da, these variants weigh about twice the largest ones separated by FAIMS to date^13,16,17 and are well into the “middle-down” size range. Under present conditions, these species create protonated ions with charge states of z = 3 - 7, the most abundant being 5 and 6. Following our established protocol,^16,17 we first processed each sample individually with the 2+ and 3+ ions of common peptide Syntide 2 or S2 (PLARTLSVAGLPGKK) as internal calibrants (selected because S2 yields abundant ions with multiple well-defined reproducible features for z = 2 - 4 spaced over a wide E_C range). While this calibration appears most accurate when the charges of analyte and benchmark peptides match, it has greatly helped in other cases such as 1+ nitropeptides.³¹

Results

We have tested He/N₂ mixtures first. For z = 3, the signal at any He fraction is low while E_C spectra are broad, featureless, and near-identical for all variants. Hence we have focused on the data for z = 4 - 7 (Fig. 1). Increasing the He fraction augments the collisional heating of ions in FAIMS rf field, because the heating magnitude (ΔT) depends on the gas and is proportional to Ω⁻² (where Ω is the ion-molecule collision cross section)²⁴ and Ω values are smaller with He than with N₂. This commonly brings on conformational transitions that affect spectral profiles.^{16,17,24,31,32} Elongated peptides and proteins tend to have lower absolute E_C values than compact conformers.^32-34 Thus, the diminution and disappearance of high-E_C features (c and then b) relatively to lower-E_C peaks a for 6+ ions of K8, K12, and K16 with increasing He content indicates unfolding driven by stronger rf field heating, as was encountered for unmodified and phosphorylated peptides.^16,17,24 In contrast, for 6+ ions of K5, the feature a drops relative to b at higher E_C above 20% He, hinting at the conversion to more compact conformers. Of course, the shape of global minimum may depend on the variant, with some folding and others unfolding upon annealing. The spectra for z = 5 change little. With 4+ ions, K12 seems to unfold as c peak vanishes. For other variants, both a and c greatly diminish or disappear to leave b as the dominant peaks (Fig. 1). This suggests the annealing of most compact and elongated conformers to the intermediate geometries that perhaps lie lower in energy for this charge state, in line with reports for some peptides and proteins in conventional IMS.^35,36

Fig. 1.

Normalized FAIMS spectra for the localization variants of H4 histone tail (color-coded on the top) with z = 4 - 7, measured using He/N₂ with up to 46% He (v/v), as labeled. Peaks for different conformers are marked by letters. The widths (w, V/cm) are shown for the well-shaped major peaks at 46% He, with R values added for z = 6 and 7.

Adding He also raises E_C values for all charge states and narrows the peaks. Thus the resolving power increases, while signal decreases because of accelerating diffusion and wider ion oscillation in the gap.²⁰ At the highest He fraction tried (46%), the peak widths for z = 6 and 7 [w = 0.9 - 1.2 V/cm with mean w (<w>) of 1.0 V/cm, Fig. 1] match those for shorter peptides with z = 3 or 4 (e.g., 1.0 - 1.1 V/cm for S2 ions),²⁴ although the R metrics are slightly less (~140 - 190 vs. ~180 - 200 for S2) because of lower E_C. This evidences no or small peak broadening for larger peptides due to the greater number and diversity of unresolved conformers. Such non-instrumental broadening is a major limitation in either conventional or differential IMS of proteins, even smaller ones such as ubiqiutin (8.6 kDa) or cytochrome c (12.3 kDa).^32,34,37-39 Lack of this phenomenon here means that peptides in the ~3 - 4 kDa range are too small to form the sufficient quantity and variety of conformers surviving for t ~ 0.2 s despite the field heating. The peaks for lower z are mostly broader (<w> of 1.6 V/cm for 5+ and 1.8 V/cm for 4+ ions, Fig. 1), likely reflecting multiple albeit similar conformers. As Coulomb repulsion within an ion scales as z², higher charge state facilitates unfolding of a peptide to a single elongated geometry. Therefore IMS and FAIMS spectra of protein ions typically exhibit narrower peaks for higher z, especially at elevated ion injection energy or gas temperature that also promote unfolding.^34,36-38

The higher resolving power and fewer peaks upon annealing at greater He content broadly improve the variant resolution: only K5 and K12 can be filtered from others (as minor features for z = 7 and 6, respectively) using N₂, while at 46% He each variant is separable from others at the major peak apex (Fig. 1). Indeed, the only substantial peaks overlapping for z = 6 are the major a features of K8 and K16 that are baseline-resolved as slightly smaller peaks b or as the dominant peaks for z = 5 and 7. Of most analytical benefit are 6+ ions, providing the highest overall resolution and sensitivity, while overlaps can often be disentangled using other z (as exemplified above). Indeed, the separations in all charge states are orthogonal: all pairwise linear correlations for z = 4 - 6 have χ² < 0.36 (Fig. 2). The relative E_C ranges spanned in Fig. 1 for z = 4, 5, and 6 at 46% He are respectively 1.13, 1.10, and 1.08: much smaller than the values for four monophosphorylated variants (1603 Da) of the τ protein 226 - 240 sequence¹⁷ (1.2 for z = 2 and 1.4 for z = 3). This supports the idea that localization variants become less separated for larger peptides and smaller PTMs, although does not isolate the two effects. Seeking to quantify this, we note that the PTM makes 1.3% of the peptide mass here versus 5.0 % for phosphorylation of the above τ-peptide, and the maximum E_C spans (~0.13 and 0.4, respectively) are roughly proportional to those percentages. How broadly this relationship holds remains to be determined.

Fig. 2.

Correlations of separations in Fig. 1 (major peaks at 46% He) across charge states for z = 4, 5 and 6, showing the linear regressions and χ² correlations.

Ions have comparable mobilities in H₂ and He, and He/N₂ and H₂/N₂ buffers with equal N₂ fractions provide close FAIMS resolving powers.²⁶ However, much greater electrical breakdown resistance of H₂ permits lower N₂ fractions (down to ~10% versus ~50% with He/N₂ at the presently maximum DV = 5.4 kV) and thus higher R values.^20,26 The gain for multiply-charged peptides was less than that for large singly-charged ions, but still significant.²⁶ In agreement with those findings, here the E_C values and peak widths at 46% H₂ (<w> of 1.0 V/cm for 6+ and 7+ ions, 1.5 V/cm for 5+, and 1.6 V/cm for 4+ from interpolation of the data at 40% and 50% H₂) about match those at 46% He, leading to the same range of R ~ 140 - 190 for z = 6 and 7 (Fig. 3). As H₂ content grows to 70%, the peaks narrow (to <w> of 0.9 V/cm for z = 6 and 1.3 V/cm for z = 4) while E_C continue rising, and the resolving power (for z = 6) increases to ~230 - 310. This does not enhance the variant resolution as all four variants can be separated at lower R, but may be useful for more complex mixtures, e.g., comprising further acetylations and/or methylations.

Fig. 3.

Same as Fig. 1, but using H₂/N₂ mixtures with 40 - 70% H₂ (v/v).

Overall spectral profiles (the conformer distributions for each variant and their relative positions across variants) at 46% He differ from those at 40 - 50% H₂, but are close to those at 60% H₂ (Fig. 3), while those at 30% He best match those at 40% H₂. As H₂ molecules (polarizability a = 0.8 ų) are larger and more polarizable than He atoms (a = 0.2 ų) but smaller and less polarizable than N₂ molecules (a = 1.7 ų), the Ω values for ions in H₂ lie²⁶ between those in He and N_2. By the Blanc’s law stipulating that the inverse mobility for an ion in a gas mixture is the interpolation of values in its components, this carries over to Ω and ΔT values with H₂/N₂ and He/N₂ compositions.²⁶ Hence adding H₂ causes less conformational changes than an equal He fraction. The heating at >60% H₂ exceeds that at the maximum He fraction, inducing further transitions such as the vanishing a feature for 6+ ions of K5 (Fig. 3).

The folding distinctions between variants elicited by altering the gas influence the separation outcomes. For example, K5 can be fully resolved from K12 as 6+ ion in 46% He or 50 - 60% H₂ because the K12 b feature that overlaps with the dominant K5 b peak at lower He or H₂ fractions is destroyed by field heating. However, the major peaks of K5 and K12 move closer at higher He or H₂ concentrations and coincide at 70% H₂ (Fig. 3), rendering them inseparable (for z = 6) despite the greater resolving power. Other species such as K5 (z = 7) and K12 (z = 4 and 7) are eliminated at higher He or H₂ fractions, presumably by FAIMS “self-cleaning” upon extensive unfolding.^20,32 This process may limit the room for higher H₂ fractions to ranges below the electrical breakdown threshold, depending on the analytical targets. One path around this obstacle may be thermal or collisional heating of ions before FAIMS analysis.

As previously,^16,17,31 separations using all He/N₂ and H₂/N₂ compositions tried were confirmed by analyses of 1:1 binary mixtures of K8 with K5, K12, or K16 (Fig. 4 and Figs. S1 - S4 in the Supporting Information). All significant spectral features were attributed employing the data for individual variants, which validates the calibration procedure and verifies the utility of FAIMS for separation of histone variants.

Fig. 4.

FAIMS spectra for the K5Ac/K8Ac, K8Ac/K12Ac, and K8Ac/K16Ac mixtures with z = 4 - 7, measured using 46:54 He/N₂. The vertically scaled spectra for individual components are overlaid.

Conclusions

Planar FAIMS analyzers using He/N₂ or H₂/N₂ mixtures can separate large peptides in the “middle-down” size range (~3 - 4 kDa) and charge states of 4 - 7 with resolving power similar to that reached for smaller peptides in the “bottom-up” range. Here, we baseline-resolved all four biological monoacetylated histone tail variants with differing modification sites. The exact separation mechanism is unclear, but both the initial 3-D peptide geometry and its evolution under the field heating likely matter. Moving a PTM may strongly affect both factors by modulating the charge distribution along the backbone and/or steric hindrances that control the relative energies of competing folds and barriers to their unfolding. The PTM position should intuitively make a lesser difference for longer peptides, and indeed the separation range in any charge state is narrower for these histones than for shorter phosphopeptide variants. However, separations in different charge states are virtually orthogonal as for sequence inversions,³¹ hence the total peak capacity approximately scales with the number of available states. With standard ESI sources, that number grows roughly in proportion to (peptide mass)^1/2 or faster:⁴⁰ e.g., from sole z = 1 for leucine enkephalin⁴¹ (0.56 kDa) to 5 states for present histone tails to 24 states (z = 15 - 38) for carbonic anhydrase (29 kDa).³⁹ These ranges can be expanded to higher z via supercharging⁴² and lower z by proton stripping.³⁶ This dramatically raises the odds for resolving larger peptide variants in at least one charge state. These findings raise a remarkable possibility that such localization variants, even involving smaller PTMs like acetylation, could be separated for intact proteins.