Gas-phase protein conformation/multimer ion formation by electrospray ion mobility-mass spectrometry: bovine insulin and ubiquitin

Abstract

Ion mobility-mass spectrometry (IMMS) is a very attractive method for studies in structural biology because of the ability of rapid isolation by nearly simultaneous m/z characterization and size separation, leading to an emergence of IMMS as a complimentary biochemical tool. Earlier, we developed a method based on varying the protein concentration in solution prior to electrospray ionization (ESI) with subsequent m/z selection and dissociation of protein multimers by IMMS of cytochrome c. The focus of this work will be to correctly distinguish truly different ion conformations formed by ESI versus homomultimeric complexes with the same m/z for well-studied proteins bovine ubiquitin and insulin. These proteins were chosen due to their large difference in solution phase structures: insulin tightly bound by disulfide linkages, and ubiquitin—a protein that may adopt a range of states from compact to extended. Our preliminary results, as with cytochrome c reveal false negatives for protein oligomer formation and false positives for protein conformational states. In addition, these results will be couched in terms of the need for quantification of IMMS analysis of proteins given the total area under IMMS peaks can also distinguish conformation versus aggregation as higher order oligomers have more mass per ion.

This article is part of the themed issue ‘Quantitative mass spectrometry’.

1. Introduction

Ion mobility-mass spectrometry (IMMS) is a very attractive method for studies in structural biology because m/z characterization and size separation are nearly simultaneous, leading to an emergence of IMMS as a complimentary biochemical tool. Determining the structure/conformation of a protein is key to understanding a proteins function and activity. Conventional analytical methods such as nuclear magnetic resonance or circular dichroism are fundamentally limited while IMMS can characterize the size of biomolecules even in trace amounts present in complex mixtures. Yet, as with mass spectrometry alone, the controversial issue of whether IMMS being a measurement in the absence of solvent can be related to or is even relevant to the native form of protein oligomers in solution remains [1,2]. There is still no experiment capable of monitoring a protein in the gas phase at an atomic level and interpretations using available gas-phase techniques vary considerably; in addition, there is usually a tendency for researchers to cast their results in a positive light with data that is ‘in contrast to’, ‘unexpected’, ‘surprising’, ‘discrepant’ and even ‘an exception to prove the rule’. For example, charge state distributions (CSDs) are routinely used to determine a protein's conformational properties even though we unambiguously showed that cytochrome c CSDs changed as a result of concentration/aggregation followed by multiply charged multimer (MCM) dissociation not conformation [3]. More recently, the CSDs of the intrinsically disordered protein (IDP) apolipoprotein C-II were found to be narrow and completely unaffected by solution conditions or temperature while the IDP α-synuclein shows the expected bimodal, wide CSD albeit with the presence of MCMs [4]. The discrepancy in CSDs is rationalized as being due to different ionization processes but clearly the CSD for apolipoprotein C-II is not related to its solution properties. Hydrogen deuterium exchange (HDX) is another common MS-based tool for probing solution and gas-phase protein structure and attempts have been made to relate ion collision cross sections measured by ion mobility to HDX measurements. For ubiquitin, there is no correlation of HDX distributions to charge state conformation measured by ion mobility and even though a correlation is expected the authors call the two methods complementary, able to resolve more conformations by combining methods [5]. If HDX is performed during ionization at atmospheric pressure every charge state of ubiquitin exists in a single conformation, yet ion mobility results suggest a wide range of conformations for charge states +7 through +13 up to at least 12 conformations identified with a high resolving power (more than 250) device [6–8]. A recent unexpected result for ubiquitin involved ‘supermetallization’, where the electrospray ionization (ESI) desolvation capillary is heated to temperatures up to 450°C in the presence of Zn atoms [9]. The higher charge states (+8 to +11) incorporate a few Zn atoms at high temperatures while the lower charge states (+5 to +7) incorporate up to 15 Zn atoms. This is opposite to the expected result because the higher charge states are supposed to be unfolded and the lower charge states compact according to previous ion mobility results. The unexpected results are rationalized because they agree with previous unexpected results using HDX [6]. There have been other attempts to shed light on gas phase versus solution phase structures of proteins, e.g. by combining various dissociation methods or gas-phase spectroscopy with IMMS, but they all rely on the assumed hypothesis that the presence of multiple ion mobility peaks is a direct measurement of conformation, e.g. the gas-phase electron photodetachment spectra for ubiquitin are independent of solution conditions [10–13]. As noted above, cytochrome c has been previously investigated by concentration-dependent ion mobility and CID showing that multiple ion mobility peaks for a given charge state are the result of dissociation of MCMs. The question remains as to whether the results for cytochrome c hold for other proteins. Rather than argue that ‘limited data in the literature suggest they do’ [14] we have chosen two well-studied proteins: (i) ubiquitin, which adopts a compact native (N) state in aqueous solution, a partially extended A-state in solution containing greater than 40% methanol, and as stated above a number of ion mobility peaks are present in the gas phase, and (ii) bovine insulin, a rigid molecule constrained by two disulfide bonds linking chains A and B restricting conformational changes in the gas phase that should provide single ion mobility peaks for each charge state (see electronic supplementary material). As with the previous study on cytochrome c, the goals of this research are to correctly distinguish true ion conformations versus homomultimeric complexes of the same m/z. Our method consisting of varying the protein concentration while probing the resulting IM peaks by m/z selection/dissociation prior to IM analysis reveals the true origin of multiple IM peaks for protein ions of a single m/z as MCMs not ion conformations as previously reported. An alternative method is also discussed involving either absolute or relative quantitation of IMMS peaks to determine their true origin as the total area under all IMMS peaks must account for the total mass not m/z. Given the increasing demand for characterizing the higher order structure of proteins and the rapid advances in MS-based methods in protein therapeutic applications, this research is of particular importance for the biopharmaceutical industry as reliability and robustness must be demonstrated by quantitative measurements for broad acceptance.

2. Experimental set-up

IMMS experiments were performed on an IM-TOF-MS (Synapt HDMS G1, Waters Corp.) operating under recommended conditions for proteins. The capillary of the ESI source was held at a voltage of 3 kV and a temperature of 80°C, with the source operating in positive ion mode and a nitrogen drying gas region held at 250°C. The sample cone was operated at 10–90 V, and the injection voltage into the TW drift cell was varied between 10 and 25 V to observe the effect of ion temperature at different times from ion formation. Samples were infused to the standard electrospray (z-spray) source at a rate of 5 µl min⁻¹. The concentrations of protein solutions were varied from 1 µM to 100 µM in various solvent compositions. Bovine insulin and ubiquitin were purchased and used as supplied from Sigma-Aldrich.

3. Results and discussion

(a). Ubiquitin

Ubiquitin is a conserved regulatory protein of 76 amino acids and the focus of a large number of studies aimed at probing the extent of solution protein structure maintained in the gas phase [5–22]. Figure 1 shows ESI-IM mass spectra acquired for acidic 50% MeOH solutions as a function of concentration. Note that when low concentrations of ubiquitin are used only high charge state ions are produced by ESI (figure 1a). Using a typical ESI concentration shifts the CSD to higher m/z (figure 1b). A high concentration of ubiquitin results in low charge states dominating the mass spectrum (figure 1c). Note that the abundance of odd charge state dimers of ubiquitin and even charge state dimers of the same m/z as their corresponding monomers also increase with analyte concentration. As with cytochrome c [6], the data shown in figure 1 suggest that as no low CSD is observed at low concentration and the CSD shifts with the production of MCMs, the typically observed bimodal CSD of electrosprayed ubiquitin is the result of dissociation of higher order highly charged clusters of the protein and is unrelated to the protein conformation.

Figure 1.

Positive ion mass spectra of ubiquitin as a function of ESI concentration: (a) 1 µM, (b) 10 µM and (c) 100 µM; 10 V injection voltage, 50% methanol, 1% acetic acid.

Figure 2 shows IMMS distributions for ubiquitin using a 10 µM solution at a pH of 2.8, 6.9 and 11.3 in parts a, b and c, respectively. The differences in peak shape for individual charge states are negligible, with the main feature being the significant abundances of odd charge state MCMs. These results are not surprising considering that the CSDs for ubiquitin ions are concentration dependent, not solution dependent.

Figure 2.

IMMS distributions of 10 µM ubiquitin as a function of pH: (a) pH 2.8, (b) pH 6.9 and (c) pH 11.3. (Online version in colour.)

In the IMMS literature, most charge states of electrosprayed ubiquitin ions produce multiple IM peaks ranging from 1 to 15, depending on the laboratory, experimental conditions and differing interpretations [15–19]. The reasons for these differences are for the most part unknown, but rationalized in terms of kinetic trapping of solution phase structures, kinetically trapped in the gas phase. Broad IM peaks are even fit with a number of ‘Gaussian conformer’ [7,20] peaks correlated to gradual changes in solution compositions, yet other experimental evidence including our results show that IM peak profiles are not dependent on solution conditions [14]. Therefore, we have chosen a number of charge states of ubiquitin to investigate in more detail by m/z selection followed by energy-dependent injection/MCM dissociation prior to mobility analysis. Figure 3a,b shows IMMS distributions for ubiquitin ions formed at high concentrations and m/z selected prior to 25 V injection into the IM cell for the assumed +4 and +5 charge states, respectively. The +11 trimer (T + 11), +7 dimer (D + 7) and the +9 dimer (D + 9) can be identified in the distributions. These ions must have a mobility proportional to charge state and fragmentation of each charge state indicates that the multiple ion mobility distributions are due to a combination of MCMs. The labile MCMs dissociate upon injection into the drift cell as evidenced by the additional peaks present at higher and lower charge states than the parent ions charge state. Note also that the D + 8 and D + 10 are very stable and at 25 V injection no monomers are present in the + 4 or +5 IMMS distributions. Figure 3c,d shows IMMS distributions for the m/z selected ions of the assumed +6 charge state for 10 V and 25 V injection voltages, respectively. With 10 V injection, two peaks are present in the IMMS distribution consisting of MCMs as confirmed by increasing the injection voltage to 25 V revealing the presence of higher charge states of monomer ion and both the D + 8 and D + 10 ions. To observe the intact MCM ions for the m/z selected +7 charge state, the injection voltage must be lowered to 5 V as shown in figure 4a. Figure 4b shows the resulting dissociation products for the MCMs at 10 V injection voltage, as for the lower charge states peaks at m/z both lower and higher charge states are present in the spectrum. The results shown for the assumed +4 through +8 charge states can be compared to the stability of MCMs using the relationship between coulomb energy and collision cross section as calculated by Counterman et al. for spherical ions [21]. For ubiquitin, we have shown that the IMMS peaks typically reported as ‘native’ or compact conformations are actually MCMs, so collision cross sections for peaks typically reported as ‘denatured’ or elongated will be used for each charge state. All of the charge states for +4 to +8 have monomer collision cross sections that predict that the D + 8 (M + 4), D + 10 (M + 5) and D + 12 (M + 6) have coulomb energies where the gas phase MCM would be stable. The D + 14 (M + 7) has a coulomb energy located between regions of stability and instability, while the D + 16 (M + 8) would not be stable. Our results are in agreement with the coulomb energy calculations: D + 8 and D + 10 are very stable with evidence they exist as higher order multimers; D + 12 will dissociate with 25 V injection; D + 14 is metastable, dissociating with low injection voltages and within the drift cell as evidenced by the filling in of the drift time profile between dimer and monomer; and D + 16 is the least stable and would require lower ion temperatures to increase in abundance relative to M + 8. It is interesting to compare our results to recent cryogenic IMMS results for the +7 charge state [22]. They observe that the D + 14 ion undergoes near complete desolvation before subsequently dissociating to the M + 7 ion, an indication that water stabilizes dimer formation. They suggest that the M + 7 ion exists in the ‘native’ compact state, but here we have shown unambiguously that the D + 14 ion is actually the misinterpreted ‘native’ state and the ‘streaking’/filling in of peaks between D + 14 and M + 7 ions are dissociated D + 14 ions typically misinterpreted as intermediate conformations. The authors also argue that the ‘streaking’/filling in of the IM distribution between dimer and monomer determines the conformational preferences of ubiquitin, but this is unassociated with conformation, the result of dissociation of the dimer ion at different drift times/positions in the IM drift cell. Finally, if the injection voltage is increased to 30 V and ions are m/z selected for the assumed +4 and +5 charge states, an abundance of charge states resulting from ubiquitin dimer dissociation are present in their corresponding mass spectra as shown in figure 5a,b, respectively, even for a typical 10 µm solution.

Figure 3.

The m/z selected IMMS distributions of 100 µM ubiquitin. (a) The m/z 2141 selected; injection voltage, 25 V; (b) m/z 1713 selected; injection voltage, 25 V; (c) m/z 1428 selected; injection voltage, 10 V and (d) m/z 1428 selected; injection voltage, 25 V. (Online version in colour.)

Figure 4.

The m/z selected IMMS distributions of 100 µM ubiquitin. (a) The m/z 1224.4 selected; injection voltage, 5 V; (b) m/z 1224.4 selected; injection voltage, 10 V; (c) m/z 1071 selected; injection voltage, 10 V and (d) m/z 1071 selected; injection voltage, 25 V. (Online version in colour.)

Figure 5.

(a) The m/z 1713 selected post mobility ion mass spectrum of 10 µM ubiquitin with 25 V injection. (b) The m/z 2141 selected post mobility ion mass spectrum of 10 µM ubiquitin with 25 V injection.

(b). Quantification

Our method of concentration dependence and m/z selection prior to dissociation and IM analysis is a pseudo-quantitative method for distinguishing between MCM formation and structure/conformation but no measurement is made of the total mass. Alternatively, a traditional quantitative analysis could be performed, i.e. absolute quantitation with isotope labelling or relative quantitation by measurement of total IM peak areas correlated to concentration. For example, if the total area under IM peaks (not m/z peaks) does not correlate to concentration the presence of MCMs must be taken into account because the mass of the monomer and MCMs differ:

3.1

where c_X is the molar concentration of the multimer x, c_M is the molar concentration of the monomer, A_x is the peak area of the nth multimer, n is the order of the nth multimer, and i is the highest order multimer observed. A quantitative measurement of IM peaks would also be very useful for IMMS of a large variety of molecules not just proteins as aggregation is a common phenomenon. In future studies, we will combine quantitative measurements with our current procedure to correctly identify IMMS peaks due to MCMs or conformations and extend this work to other proteins and peptides especially those of interest in protein and peptide therapeutics.

4. Conclusion

In conclusion, here we report unambiguous evidence from IMMS experiments showing that multiple IM peaks of bovine ubiquitin and insulin ions produced by ESI are due to non-specific MCMs and their product ions, not protein conformational states as previously reported. Our results revealed false negatives for protein oligomer formation and false positives for protein conformational states. The stabilities of MCM ions as a function of coulomb energy for spherical monomer ions agree with the experimental results, i.e. low charge state MCMs are very stable and no monomer ion is present under low energy conditions. Quantitation of IM peaks has been suggested to determine their identity in a reliable and robust manner. It is important to note that if IMMS is used as a measurement of structure/function of proteins or peptides by correlating arrival time distributions to conformation the results will be in error without correcting for MCMs by quantitative IMMS.

Competing interests

I declare I have no competing interests.

Funding

The author acknowledges grants from MOST (MOST 102-213-M-001-002-MY5) and NHRI (NHRI-EX104-10301EI) in Taiwan and the Genomics Research Center, Academia Sinica.