Mass spectrometry has been used to examine the subunit interactions in the capsular F1 antigen from Yersinia pestis, the causative agent of the plague. Introducing the sample using nanoflow electrospray from solution conditions in which the protein remains in its native state and applying collisional cooling to minimize the internal energy of the ions, multiple subunit interactions have been maintained. This methodology revealed assemblies of the F1 antigen that correspond in mass to both 7-mers and 14-mers, consistent with interaction of two seven-membered units. The difference between the calculated masses and those measured experimentally for these higher-order oligomers was found to increase proportionately with the size of the complex. This is consistent with a solvent-filled central cavity maintained on association of the 7-mer to the 14-mer. The charge states of the ions show that an average of one and four surface accessible basic side-chains are involved in maintaining the interactions between the 7-mer units and neighboring subunits, respectively. Taken together, these findings provide new information about the stoichiometry and packing of the subunits involved in the assembly of the capsular antigen structure. More generally, the data show that the symmetry and packing of macromolecular complexes can be determined solely from mass spectrometry, without any prior knowledge of higher order structure
The application of mass spectrometry to the characterization of antigens has been highly successful for monomeric proteins, typically in the presence of denaturants (Heller et al. 2000). These conditions have, however, limited the application of the technique to probe the oligomeric nature of antigens. The ability to examine the interactions in macromolecular complexes using mass spectrometry has arisen from the coupling of two enabling technologies. The first involves the refinement of electrospray ionization (ES) to a nanoflow process (Wilm and Mann 1994). This technique was developed for improved efficiency and sensitivity. However, the smaller droplet size, requiring lower voltages to initiate the spray, and milder desolvation conditions for evaporation favor the preservation of noncovalent complexes (Chung et al. 1999). In addition to these advantages, the nanoflow method permits the use of relatively high concentrations of buffer salts, providing the necessary ionic strength to maintain electrostatically driven interactions during the transition from solution to the gas phase of the mass spectrometer (Vis et al. 1998).
The second enabling technology is the coupling of ES techniques with time-of-flight (ToF) mass analysis (Verentchikov et al. 1994). This enables the analysis of high mass/charge ratios (m/z) in excess of the restrictions typically imposed by standard quadrupole analyzers (typically, m/z <4000). Thus, much larger macromolecular assemblies with relatively few charges can now be addressed. Before the introduction of ToF analysis, this m/z restriction limited the study of complexes by ES to those with masses <100 kD (Nettleton et al. 1998). ToF mass analyzers, however, have an infinite mass range in principle, and using this type of instrumentation, it has been possible to probe the structure of oligomeric enzymes (Fitzgerald et al. 1996; Van Berkel et al. 2000) and even larger macromolecular complexes (Rostom et al. 2000; Tito et al. 2000). These studies have highlighted the importance of maintaining low internal energy in macromolecular ions. This is achieved by allowing a large number of low-energy collisions with an inert gas such as nitrogen and, in practical terms, involves increasing the pressure in the ion source and intermediate pressure regions of the mass spectrometer. This process is termed collisional cooling and imparts to the neutral gas molecule the higher translational energy that typically exists with highly charged gas phase ions (Krutchinsky et al. 1998). In contrast, fewer high-energy collisions, under high vacuum conditions, result in the dissociation of macromolecular complexes to their component subunits. Collisional cooling has been shown to be effective in the transmission of whole particles through the mass spectrometer, allowing them to reach the detector intact (Rostom et al. 2000; Tito et al. 2000). In combination, nanoflow ES coupled with ToF mass analysis and collisional cooling has the potential to probe the symmetry and assembly of even the largest macromolecular complexes.
Yersinia pestis is the causative agent of plague, a disease associated with a high level of mortality in man (Perry and Fetherstone 1997). Most cases of the disease are of the bubonic form, arising from the bite of a flea that has previously fed on the blood of an infected rodent (Butler 1984). Some cases of bubonic plague progress to the pneumonic form, which is readily transmitted from human to human, leading to epidemics (Cowling and Moss 1994). Although no longer the cause of high levels of mortality in man, plague is still endemic in parts of the world, and a recent outbreak of pneumonic plague in India has heightened the search for effective vaccines. Current plague vaccines are based on whole-cell formulations of heat-treated or formaldehyde-killed Y. pestis (Russell et al. 1995) and induce protection against the bubonic form of the disease. However adverse side-reactions have been reported for patients having repeated immunization with these vaccines (Marshall et al. 1974), and some cases of pneumonic plague have been reported in vaccinated individuals (Meyer 1970). As an alternative to whole-cell vaccines, immunization with purified recombinant Y. pestis proteins, including fraction 1 (F1), has been proposed (Williamson et al. 1997). Although the F1 antigen has been characterized immunologically, solution-based techniques, including circular dichroism, and gel filtration have shown a high proportion of β-sheet structure and a population of mutimeric species of indeterminate length (Miller et al. 1998). The F1 antigen is thought to assemble on the surface of the bacterium to form a capsular-like structure, but its molecular associations in solution remain unknown and are the subject of this mass spectrometry investigation.
Results and Discussion
Figure 1a shows the nanoflow ES mass spectrum of the F1 antigen acquired from solution conditions in which the protein is in its native state. Under the standard pressure conditions used for normal operation of the Q-ToF mass spectrometer, essentially only the monomeric form of the protein is observed with a molecular weight of 15,565±1 Da. This is consistent with the calculated value of the molecular weight from the known amino acid sequence of the protein (15,564 Da; Galyov et al. 1990). At low intensity and high m/z values, some aggregation of the monomeric species is apparent between m/z 4000 and 8000. The origin of these aggregates is most likely due to nonspecific association of the monomer during the evaporative stages of the nanoflow ES process, when the local concentration of protein molecules within droplets can increase (Kebarle 2000).
Decreasing the internal energy of the ions by increasing the pressure in the ion source of the mass spectrometer, inducing collisional cooling, and manipulating the accelerating lenses from 150 V to 120 V leads to the disappearance of the charge states assigned to the monomer and the appearance of a distinct charge state ensemble centered at m/z 5500 (Fig. 1b). The ions observed give rise to a molecular weight of 114,040±5 Da. In comparison with the molecular weight obtained for the monomeric protein, the mass measured for this species is consistent with a 7-mer. The value obtained for the measured mass of 114,040±5 Da compared with 108,955 Da calculated by multiplication of the molecular mass of the monomer gives rise to a mass difference of 5085 Da. It is likely that no small protein ligands are bound within the complex because the same solution was used to record the spectrum of the monomer, and other species were not detected even at low m/z (Fig. 1a). The most likely origin of this extra mass is either water molecules or ammonium acetate ions derived from the buffer solution, presumably trapped within the multimeric assembly. Consistent with this extra mass is the observation of the increased peak width for the 7-mer from ∼30 to ∼1300 for the molecular ions of the monomer and 7-mer, respectively. The peak width observed for a particular charge state in a mass spectrum is the convolution of a number of different effects, including the isotope composition of the ion, the resolving power of the instrument, and the heterogeneity of small molecule binding. For oligomeric complexes, such as those observed in this study, the overriding factor in determining the peak width is that of small molecule binding within both subunit interfaces and channels (Nettleton et al. 1998). Because the most likely arrangement of the seven subunits would be in a seven-membered ring (Wolynes 1996), we propose that many small molecules are trapped within a central cavity formed by this arrangement of subunits.
Reducing the internal energy of the ions still further, using SF6 for collisional cooling, results in the observation of an additional higher mass oligomer centered at m/z 8000 (Fig. 1c). The four well-resolved charge states enable an accurate mass measurement of 226,417±30 Da, consistent with a 14-mer. The theoretical mass calculated from the amino acid sequence of 217,910 Da shows that the 14-mer carries an excess mass of 8507 Da. This is less than twice that of the extra mass observed for the heptamer, suggesting that the volume occupied by small molecules is conserved within each assembly unit. This additional mass may be related to the labile nature of these assemblies and is not always a feature of high molecular weight complexes analyzed by mass spectrometry (Zhang et al. 1999). It is interesting to note that although both assemblies are more massive than would be anticipated from straightforward association of monomeric subunits, the peak widths observed for the spectra in Figure 1c (440 and 725 for the 7-mer and 14-mer molecular ions, respectively) are significantly narrower than that measured for the 7-mer in Figure 1b. The most likely explanation for this difference is that the heterogeneity of small molecule binding to the surface of the assembly is reduced in the presence of SF6. It is interesting to note that the use of SF6 increases the effectiveness of collisional cooling for this assembly, enabling the detection of the 14-mer. The observation of the 14-mer and its excess mass is consistent, therefore, with the stacking of two sevenfold symmetry rings in which the geometry of the central channel is not perturbed by the association.
It is interesting to note that the sevenfold symmetry observed in the high molecular weight assemblies is also seen in the different charge states observed in the mass spectra. It is widely accepted that under the conditions for positive ion ES, the charge state of the protein is dictated by the number of surface accessible residues that will be protonated (Kebarle 2000). The negatively charged carboxylic acid groups will either be protonated or involved in interactions with ammonium ions derived from the buffer (Kebarle 2000). Figure 1 shows that the average overall charge of the 14-mers is +28, giving an average net positive charge of two per subunit. The average overall charge on the 7-mers is +21, yielding a net charge of three per subunit. This suggests that during the dissociation of the 14-mers, an average of one positive charge per subunit is exposed for protonation, presumably through the disruption of ionic interactions between the rings. Similarly, during the dissociation of the 7-mers, each of the monomers released in the dissociation gains an average of four additional sites for protonation. This implies that the interactions between neighboring subunits in the 7-mers involve four basic sites per subunit. It is of course highly probable that other factors are involved in maintaining the subunit interfaces, in particular hydrophobic interactions, but because these will not produce a net change in the overall average charge, they cannot be addressed by mass spectrometry. The changes in charge state observed in these mass spectra, however, are consistent with the exposure of one and four basic sites per subunit when 14-mers and 7-mers dissociate.
A key question arises as to whether or not the dissociation of the oligomers is occurring in the gas phase or whether monomers, 7-mers, and 14-mers are in equilibrium in solution. Given the fact that no monomeric species were observed in the mass spectra recorded for the 7-mers and 14-mers (Fig. 1b,c), this would suggest that the monomeric subunits are not in equilibrium in solution. Because the observation of the 14-mers is highly dependent on the presence of effective collisional cooling, this implies that in solution the F1 antigen exists as a large multimeric assembly. Consequently, the intrinsic stability of the macromolecular ions appears to be the key factor in preventing their gas phase dissociation to monomers. Furthermore, when multiprotein complexes are deliberately subjected to high-energy collisions in the gas phase, the dissociation pathway invariably gives rise to highly charged monomeric ions (Lei et al. 1998). These are not observed in the spectra recorded for the F1 antigen. Thus, the inter-ring interactions involving a single basic site per subunit are only preserved through many low-energy collisions that absorb the excess translational energy of the ions. In the absence of these effects, the multimers observed for the F1 antigen in solution (Miller et al. 1998) rapidly dissociate to monomer presumably during or immediately after the ES process. These observations are therefore consistent with the stacking of seven-membered building blocks to give a fragile assembly, readily dissociated to yield monomeric subunits in the absence of collisional cooling.
These results are interesting in the light of structural information obtained for projectile structures assembled on the surface of bacteria. The molecular machinery involved in the export and assembly of the F1-subunit appears to be similar to the chaperone-usher systems that are involved in the assembly of composite adhesive pili on the surface of bacteria (Sauer et al. 1999). Moreover, the molecular chaperone involved in the export of F1-antigen (caf1M) shows sequence homology with the PapD chaperone involved in biogenesis of Escherichia coli P-pili, and like PapD, Caf1M is thought to adopt an immunoglobulin-like fold (Chapman et al. 1999). The structure of pili from Neisseria gonorrhoeae and E. coli has been found to give helical structures and to contain solvent filled central cavities (Bullitt et al. 1995; Parge et al. 1995). This helical assembly of pilin-subunits into pili seems to be relatively common, and intuitively it seems likely that F1 antigen subunits may assemble in a similar manner. It is possible that the F1 antigen 7-mers observed in this study form from stepped, rather than closed, ring structures assembled into a helical arrangement. On dissociation of the larger assembly in the gas phase, 14-mer and 7-mer units are released (Fig. 2). This would imply that the F1 antigen subunits associate into a helical structure with 7 subunits per turn. In common with other bacterial pili, there is a central solvent-filled cavity in accord with the data presented here for the capsular F1 antigen.
It is interesting to note that from the structures of type IV pili that have been deduced, diameters of 60 Å and 68 Å have been reported (Bullitt and Makowski 1995; Parge et al. 1995), despite the difference in molecular weights between the component proteins, 17.4 kD for N. gonorrhoeae pilin (Parge et al. 1995) and 16.5 kD for PapA (Baga et al. 1984). Given the close similarity in diameters of pili, a relationship between molecular mass and the number of subunits per helical turn might be anticipated. The number of subunits per helical turn has been reported as 3.28 for P-pili and 5 for N. gonorrhoeae pilin (Bullitt and Makowski 1995; Parge et al. 1995). However, for P-pili the definition of subunits states that they may not correspond to the boundaries of single PapA molecules, rather each subunit appears as a pair of approximately globular domains (Bullitt and Makowski 1995). Assuming two molecules of PapA per subunit, therefore, and multiplication of the number of subunits in each helical turn by their molecular mass gives values of 87 kD and 108 kD for the N. gonorrhoeae pilin and P-pili from E. coli, respectively. For the F1 antigen protein, the molecular mass of 15.5 kD and observed sevenfold symmetry gives a value of 108.5 kD per turn, in very close agreement with the value presented for the N. gonorrhoeae pilin.
Before this study, there had been no indication of specific interactions of monomeric F1 protein either within the capsular antigen or in solution, although it has been shown that the antigen forms pores in lipid bilayers (Rodrigues et al. 1992) and extremely large multimers in solution (Miller et al. 1998). These mass spectrometry results define the symmetry and packing of monomeric subunits in solution that presumably also relate to their arrangement in the capsular F1 antigen. We propose that the formation of the macromolecular capsular antigen structure involves seven monomeric subunits in each helical turn and that a solvent filled channel is maintained on association of the 7-mers. The mass spectrometry results reported here show no evidence for any other assemblies and allow us to propose that different numbers of ionic interactions are involved in maintaining subunit interactions both within and between the 7-mer building blocks. More generally this example illustrates that modern mass spectrometry techniques can make a significant contribution to our understanding of intractable macromolecular assemblies such as the capsular F1 antigen.
Materials and methods
The F1 antigen was prepared as described previously (Miller et al. 1998). Before analysis, the protein was buffer exchanged into 250 mM ammonium acetate (analytical grade, Sigma Chemical Co.) using Amicon PD-10 columns. Adjustment of sample concentration was achieved using a Savant Speed Vac system (model No. SC-110). pH was typically between 6.8 and 7.0 and was not adjusted. Buffers for all protein samples were prepared with Milli-Q water.
Mass spectra were recorded on Q-ToF and LC-ToF mass spectrometers (Micromass UK Ltd.). For both mass spectrometers, the cone voltage was typically 150 V, with a needle voltage of 1.4 kV, except where stated otherwise. Mass spectra were acquired without source heating and processed with Mass Lynx software, version 3.1. The spectra represent the raw data to which a mean smoothing algorithm has been applied, corresponding to the peak width at half height.
The Q-ToF mass spectrometer is equipped with a Z-spray nanoflow ES source. Ions were focused by a radio frequency lens before transmission to the quadrupole, which was used in the RF-only mode as a wide-bandpass filter. Ions were transferred through a hexapole collision cell pressurized with dry argon, except where stated. Transmission into the ToF was achieved with an acceleration voltage of 8 kV and with a pulse rate of 4 kHz for detection with a multi channel plate (MCP). A 1-GHz time-to-digital converter was used, and the MCP was set at 2700 V. Collisional cooling was achieved by manually throttling the analyzer roughing pump and applying up to 30 mBa of argon gas in the collision cell.
The LC-ToF is also fitted with a Z-spray nanoflow ES source in an airtight housing isolated from atmosphere and under negative pressure. Acceleration into the ToF was achieved with a 3-kV pulse. Gold-coated borosilicate glass capillaries were used for sample introduction and prepared as described previously (Nettleton et al. 1998). Collisional cooling for the spectrum shown in Figure 1c was achieved by leaking SF6 at an analyzer pressure of 6.0e−4 mBa into the first hexapole of the LCT mass spectrometer. Using a heavier gas such as SF6 was found to be more efficient than argon or nitrogen in reducing the excess translational energy of the ions.
We acknowledge helpful discussions with Don Daley and Paula Tito. This is a contribution from the Oxford Centre for Molecular Sciences, funded by the BBSRC, EPSRC, and MRC. M.A.T. is grateful for support from DERA-CBD (Porton Down). C.V.R. acknowledges support from the Royal Society.
The publication costs of this article were defrayed in part by payment of page charges. This article must therefore be hereby marked “advertisement” in accordance with 18 USC section 1734 solely to indicate this fact.