CID fragmentation, H/D exchange and supermetallization of Barnase-Barstar complex

The barnase-barstar complex is one of the most stable protein-protein complexes and has a very wide range of possible applications. Here we report the use of top-down mass spectrometry for the investigation of the structure of this complex, its ionization via ESI, isolation and fragmentation. It was found that the asymmetry of the resulting charge state distributions of the protein monomer product ions increased as the charge state of the precursor ions increased. For the investigation of the 3D structure of the complex, the gas phase H/D exchange reaction was used. In addition, supermetallized ions of the complex with Zn were produced and investigated. It was observed that an increase in the number of metals bound to the complex results in a change in complex stability and the charge distribution between protein fragment. Analysis of the fragmentation pattern of the supermetallized complex [bn-b* + 5Zn]10+ indicated that this ion is present in different conformations with different charges and Zn distributions. Since Zn cannot migrate, such structures must be formed during ionization.

the guanine base of the substrate. Barstar consists of a single polypeptide chain of 89 amino acids 15 . The 3D structure of the barnase-barstar complex was investigated using X-ray 16 , NMR 15 , computer simulation 17 and mass spectrometry 18 .
Mass spectrometry combined with ESI or MALDI ionization is a very powerful method for the investigation of non-covalent protein-ligand complexes 19 , which was used to characterize the barnase-barstar complex 18 . In order to observe gas phase ions of the protein-protein complex, "native" conditions of the experiment should be maintained [20][21][22] . Such requirements impose certain limitations on the use of the conventional H/D exchange approach for mass spectrometric investigations of the 3D structures of complexes 23,24 . The conventional H/D exchange is a bottom-up technique, which consists of protein exposition in deuterated solvent, reaction quenching by lowering pH to 2.6, enzymatic digestion, and LC-MS analysis. This approach requires many steps and is sensitive to experimental conditions. Since special solution composition or pH must be maintained, it may be difficult or even impossible to use this method in an MS experiment. Top-down H/D exchange 25 , on the other hand, especially when it is performed after ionization in the gas phase 26,27 , lacks these disadvantages. Once the complex is ionized and the 3D structure of the ion is likely to be as close as possible to obtain to the structure in the solution, then the H/D exchange reaction in the gas phase labels the outer surface of the complex and the following MS n experiments reveal the exchange-protected sites. Recently it was suggested to perform the H/D exchange reaction directly in the ESI source 28,29 . This approach was applied to study proteins [30][31][32] , oligosaccharides 33 , natural complex mixtures [34][35][36][37] and small organic molecules 38 .
Here we attempt to apply the gas phase H/D exchange approach combined with collision induced fragmentation (CID) in order to get insights into the high order structure of the bn-b* complex ion in the gas phase. One of the advantageous features of the gas phase H/D exchange reaction is the use of elevated temperatures in the region of droplet evaporation, when several unusual gas phase reactions can take place. One of such reactions is the supermetallization 30 , which is also described in the paper. The investigation of the pathways of fragmentation of supermetallized complexes may provide structural information whic cannot be obtained in the H/D exchange experiments because the metal atoms are less labile than protons, so the effect of scrambling would be minimized [39][40][41][42] .

Methods
Samples and Instruments. Barstar A (Cys-40,82-Ala barstar) and barnase were purified from Escherichia coli strain BL21(DE3) harboring the corresponding expression plasmids. Harvested cells with barstar A were sonicated and nucleic acids were precipitated by gradual addition of polyethyleneimine to a final concentration of 0.03%. Cleared lysate was fractionated with ammonium sulfate. The protein fraction precipitated in the saturation interval 40-80% was purified on a XK 16/70 column packed with Sephadex G-50 SF (GE Healthcare). Barstar A containing fractions were pooled, diluted 6-fold with 50 mM Tris-Cl, рН 8.0, and applied on MonoQ 10/100 GL (GE Healthcare). The protein was eluted by NaCl gradient 0-1 M. Peak fractions were exhaustively dialyzed against 10 mM EDTA, pH 7.5.
Barnase was secreted mainly into the growth medium and then was extracted onto phosphocellulose P11 (Whatman). It was further purified by hydrophobic (Phenyl Sepharose FF) and cation-exchange (MonoS 10/100 GL) chromatography, then exhaustively dialyzed against 10 mM EDTA, pH 7.5. The resulting proteins were essentially pure in SDS-PAGE (Fig. 1). Protein concentration was determined with an Ultrospec 7000 spectrophotometer (GE Healthcare), assuming the following molar extinction coefficients at 280 nm (in M −1 cm −1 ): 26 030 for barnase and 20 910 for barstar А. To prove complex formation, a native electrophoresis experiment was performed as described by Laemmli 43 and Ornstein& Davis 44 (see Fig. S1). The gel percentage was 12%.
For mass spectrometric experiments, a solution containing 0.5 M CH 3 COONH 4 , 2% glycerol, and 0.5% m-NBA was prepared. The solutions of bn and b* in water with concentrations of 250 μM were mixed, incubated for 5 minutes at 80 °C, and diluted in the buffer solution in order to obtain a concentration of 10 μM. This final solution was directly introduced into the ESI source. For deuterium labeling, the buffer was prepared in D 2 O. D 2 O was purchased from the "Neogaz" company. Enrichment of D 2 O was 99.9%. Other chemicals were purchased from "Sigma" and were of analytical grade or higher.

MS analysis.
All experiments were performed on a LTQ FT Ultra (Thermo Electron Corp., Bremen, Germany) mass-spectrometer equipped with a 7T superconducting magnet. Ions were generated by an IonMax Electrospray ion source (Thermo Electron Corp., Bremen, Germany) in positive ESI mode. The temperature of the desolvating capillary was varied from 50 °C to 450 °C. The length of the capillary was 105 mm and its inner diameter was 0.5 mm. The infusion rate of the sample was 1 μL/min and the needle voltage was 4000 V.
H/D exchange. For in-ESI source H/D exchange, the atmosphere was saturated with D 2 O vapors in the region between the ESI tip and the inlet of the desolvating capillary by placing 400 μL of D 2 O on a copper plate positioned approximately 7 mm below the ESI needle. More details about the used H/D exchange approach can be found elsewhere 29 . For exchange in the liquid phase, aqueous solutions of bn and b* with concentrations of 250 μM were mixed, incubated at 80 °C for 5 minutes and diluted with a deuterated buffer. Important information about the sequences of proteins and number of labile hydrogens is given in Table S1.

Results and Discussion
Ionization and fragmentation of the bn-b* complex. The broadband mass spectrum of the prepared solution is presented in Fig. 1. It consists of peaks corresponding to the ions of barnase with charges varying from 7+ to 14+, ions of barstar with charges varying from 9+ to 5+, and ions of bn-b* with charge states varying from 9+ to 12+. It was observed that the shape and relative intensities of the different ion peaks in the mass spectra significantly depend on the ion optics settings of the mass spectrometer. In order to observe the ions of bn-b*, the solvent composition should be as described in the "METHODS" section. With other solvents, such as H 2 O:MeOH mixtures, and different additives like formic acid, or NH 4 (HCO 3 ), instead of CH 3 COONH 4 , the intensities of the complex ion peaks were very weak, if they were observed at all. We should mention that for the H 2 O:MeOH mixtures the [b*-b*] 11+ ion was observed 45 . Multiply ionized complex ions [bn-b*] 10+ , [bn-b*] 11+ , and [bn-b*] 12+ were isolated in the ion trap and fragmented using CID (see also Figs S2 and S3). It was found that when electron capture dissociation (ECD) fragmentation was used, the ions' charge decreased without fragmentation. The fragmentation spectra are presented in Fig. 1. It can be seen that when the ion [bn-b*] 10+ is fragmented, it yields almost equal amounts of bn 5+ and b* 5+ monomeric ions and, with lower intensities, b* 6+ and bn 4+ ions, so the charge is split almost symmetrically. When higher charge states of the bn-b* complex are fragmented, a charge distribution asymmetry is observed: barnase monomers retain most of the charge. This can be explained by the assumption that the barnase monomer has more basic sites on its surface available for protonation. The fragmentation of [b*-b*] 11+ produced ions b* 5+ and b* 6+ with almost equal intensities and also ions b* 7+ and b* 4+ , but their intensities were considerably lower. Previously, asymmetric dissociation of the protein homodimers as a function of the charge state and the conformation of the complex was observed and investigated 46,47 . The effect was explained by partial unfolding of one of the monomers, followed by proton migration to the unfolded one. Overall, the problem of conformational changes and correlation between the solution phase structures and gas phase structures is not yet resolved 48 ; recently evidence was provided for the gas phase folding of peptides 49 .
The presence of the highly charged bn and b* ions can be explained by in-source fragmentation of the bn-b* complex. The LTQ FT mass spectrometer has a relatively hard skimmer-nozzle interface, in which non-covalent complexes may undergo fragmentation. As described many times earlier, CID fragmentation of protein complexes results in a charge rearrangement and subsequent ejection of a single highly charged subunit 50-52 . H/D exchange of the bn-b* complex. In order to obtain more structural information about the bn-b* complex, gas phase H/D exchange was applied. If the complex is subjected to H/D exchange in the gas-phase, then the CID fragmentation splits the complex into individual proteins with non-deuterated regions corresponding to the binding sites. Further fragmentation of these protein ions may reveal the exact localization of the protected regions. Unfortunately, neither barnase nor barstar could be fragmented with high efficiency by both CID and ECD approaches in our experimental conditions. Only the highest charge states (bn 15+ Figure S2.
The usage of high temperature capillary requires certain precautions, because in addition to the desirable H/D exchange reaction, ions continuously accumulate vibration energy what eventually may lead to their thermal dissociation, as described previously [53][54][55][56][57][58] . It should be specifically noted that the protein ions are not equilibrated to the temperature of the ion source.
In the case of bn, the number of exchanges decreases with the increasing charge state and for highest temperature it varies from 83 (or 40% of all exchangeable hydrogens or 78% of the fast ones) for bn 13+ to 117 (55% of all labile hydrogens) for bn 7+  The H/D exchange reaction was also performed in the solution phase, and after preparation the sample was immediately sprayed into the mass spectrometer. The total number of exchanges practically did not change with time but decreased with temperature because of the back exchange reaction during ESI (Fig. 2B 3 ). This happens because the traces of atmospheric moisture cannot be fully removed, and as a consequence, an increase in temperature facilitates the back exchange.
To prevent this process, the atmosphere in the ESI region was saturated by D 2 O vapors. The results of the combination of the solution phase and gas phase H/D exchange methods are shown in Fig. 2B 4 . It can be seen that the infusion of D 2 O vapors prevents the back exchange reaction as anticipated. For a wide range of temperatures, the number of exchanges remains almost the same for all charge states: ~155 (73% of all) for bn, ~127 (76% of all) for b* and ~275 (72% of all) for bn-b*. However, when the temperature was increased to 450 °C, ~173 (79%) exchanges for bn, ~142 (84%) for b*, and ~310 (81%) for bn-b* were observed. This indicates that though the infusion of D 2 O initially prevents the back exchange, with an increase in the temperature additional H/D exchange occurs.
Further, ion complexes were isolated after the H/D exchange reaction and fragmented using CID. The results are presented in Fig. 3. When b*-b* complex is fragmented, b* ions in all charge states carried almost equal numbers of deuterium atoms, what means that this complex has a rather symmetrical structure. When the bn-b* complex is fragmented, the number of exchanges in the bn fragments exceeds the number of exchanges in the b* fragments, what is expected due to the asymmetrical nature of this complex. For the gas phase reaction at the maximum achievable temperature, 202 (53%) exchanges in the [bn-b*] 10+ complex was observed, of which the bn 5+ fragment showed 118 (55% of all labile hydrogens) exchanges, the b* 5+ fragment had 86 (52% of all labile protons) exchanges and b* 6+ demonstrated 75 (80% of fast exchangeable and 45% of all labile hydrogens) exchanges. The obtained results lead to the assumption that the [bn-b*] 10+ ion exists in two different states: both subunits carry 5 protons each or bn carries 4 and b* carries 6. This is reasonable, since previously many researchers demonstrated, using the Ion Mobility Spectrometry (IMS) 59, 60 or gas-phase H/D exchange 61 or a combination of these approaches 62,63 , that several conformers can be present within a single charge state of a protein. If two such states are present, then the deuterium distribution between the subunits should differ. Indeed, the efficiency of the gas phase H/D exchange reaction of a protein depends on its charge, so in case of a tightly folded complex, the number of exchanges in the surface region should depend on the local charge distribution.
CID of the protein complex is complicated by proton rearrangement and accompanied by H/D scrambling [39][40][41][42] . After collisional activation, protons on the surface of the complex can migrate, this results in the rearrangement of not only charge but also of deuterium. If the rate of deuterium scrambling is of the same order as the rate of the charge transfer 42 , then all conformers will redistribute the deuterium to the most favorable configuration for the given charge distribution.
In Table 1 the deuterium distribution between bn and b* subunits after the fragmentation of the [bn-b*] 12+ complex is presented. The average number of H/D exchanges in an isolated [bn-b*] 12+ complex is 96. The temperature was 300 °C. The fragmentation of the complex leads to the formation of subunits in different charge states.   Table 1. Deuterium distribution between bn and b* subunits after the fragmentation of the [bn-b*] 12+ complex. Each row corresponds to an individual fragmentation channel so that the sum of charges is equal to 12. The average number of H/D exchanges in the isolated [bn-b*] 12+ is 96. T = 300 °C. Symbol "-" indicates that the data is not available.
Each row in the table corresponds to an individual fragmentation channel so that the sum of charges is equal to the charge of the complex (z = 12). As can be seen, deuterium distribution on the subunits correlates with the charge distribution. These results suggest that either there is no H/D scrambling and no charge migration (otherwise the subunits with the same charge but different numbers of incorporated deuteriums would have been observed) or that the rates of these two processes are the same.
Supermetallization. Supermetallization is a recently observed phenomenon of the formation of complex peptide-metal ions in the gas phase when a peptide accommodates an unexpectedly large number of metal atoms 39,64 . It was found that supermetallization takes place during electrospray ionization when charged droplets are evaporated under relatively high temperatures (~400 °C). The masses of supermetallized ions obey the equation:  34,65 . However, the solution used in this study was different, so it was difficult to observe high intensity peaks of supermetallized ions. In order to perform the supermetallization experiments, different concentrations of Zn(CH 3 COO) 2 were added to the buffer solution. The results are presented in Fig. 4 (see also Fig. S6 for a wider m/z range). x:y Zn means that ions carry from x to y Zn atoms. As can be seen, barnase can attach up to 9 Zn atoms and barstar up to 8. The complex [bn-b*] 10+ can carry up to 8 Zn atoms. Fragmentation of complexes carrying different numbers of Zn atoms demonstrates the asymmetry of Zn distribution between the fragments -the majority of bound Zn atoms are carried by barnase. In addition, an increase in the intensity of the bn 7+ complexes with Zn can be seen with an increase in the number of Zn in the parent complex ion.
Analysis of the charge and Zn distribution between protein monomers (Fig. 5) reveals that the number of Zn atoms attached to the protein monomer tends to increase with the decrease of its charge. Since Zn does not migrate along the protein surface 64 , since it is coordinated by several bonds, these results serve as additional evidence of the charge distribution preservation during the transition from the charged droplet to the gas phase. Based on the fact that monomeric proteins produced via CID fragmentation of complexes of higher order could not be further fragmented using ECD or CID one can conclude that the bn and b* monomers retain their folded 66,67 (though maybe non-native) conformation when transferred to the gas phase and metalized. This is consistent with the observation that cationization of a protein suppresses the effect of scrambling 68 .

Conclusion
In this work the ionization and fragmentation of gas phase barnase-barstar complexes in different charge states were studied. In the in-ESI source H/D exchange reaction, bn ions demonstrated an increase in the number of possible exchanges with the decrease of the charge state. Similar effects were observed during fragmentation of the complex ions. The ion [bn-b*] 10+ fragmented mainly to bn 5+ and b *5+ giving peaks of almost equal intensity. With an increase in the total charge of the complex ion, the charge distribution among the dissociation products changes, due to an increase in the charge carried by barnase. When [bn-b*] 10+ was fragmented, fragment b* 5+ demonstrated fewer possible exchanges than the fragment b* 6+ . Similar results were observed for the [bn-b*] 12+ complex. The fragmentation of free monomers revealed that only the highest charge states (bn 15+ and b* 8+ ) can produce abundant CID or ECD spectra. We also failed to fragment subunits produced after the CID dissociation of the complex [bn-b*] ion. This may be due to the internal hydrogen bonds that keep the 3D structure so that higher energy is required to induce fragmentation. Basing on the H/D exchange results we conclude that either there is no H/D scrambling and no charge migration or that the rate of of these processes is the same. In addition, an ion complex supermetallized with Zn was produced and investigated. Analysis of the fragmentation pattern of the supermetallized complexes indicates that the [bn-b*+5Zn] 10+ ion is present in different conformations with different charge and Zn distributions. Since Zn cannot migrate, such structures must be formed during the ionization process.