Report

Evidence for Macromolecular Protein Rings in the Absence of Bulk Water

See allHide authors and affiliations

Science  09 Dec 2005:
Vol. 310, Issue 5754, pp. 1658-1661
DOI: 10.1126/science.1120177

Abstract

We have examined the architecture of a protein complex in the absence of bulk water. By determining collision cross sections of assemblies of the trp RNA binding protein, TRAP, we established that the 11-membered ring topology of the complex can be maintained within a mass spectrometer. We also found that the binding of tryptophan enhances the stability of the ring structure and that addition of a specific RNA molecule increases the size of the complex and prevents structural collapse. These results provide definitive evidence that protein quaternary structure can be maintained in the absence of bulk water and highlight the potential of ion mobility separation for defining shapes of heterogeneous macromolecular assemblies.

A substantial amount of evidence suggests that native protein structure remains largely intact upon the removal of bulk solvent (1). Specifically, the observation that dehydrated crystal structures of small globular proteins retain a high degree of structural homology to their fully hydrated counterparts highlights the intrinsic stability of the native fold in the absence of solvent (2, 3). Multiprotein complexes have the added complexity of side chains that interdigitate across subunit interfaces to form complicated interaction networks. To date, relatively little is known of the effect of water depletion on these interactions. The noncovalent contact area between protein subunits is thought to be a major stabilizing influence (4), although solvent-related forces may also contribute to the stability of protein interfaces (5). The advent of electrospray ionization (6) has enabled generation of protein ions that are essentially depleted of water. Subsequent analyses using either hydrogen-deuterium exchange (7), optical spectroscopy (8), or ion mobility spectrometry (912), although generally limited to peptides and single proteins, have provided evidence that highly charged gas-phase ions populate a greater number of conformations compared with their solution-phase analogs (7). In contrast, lower charged ions retain compact conformations that resemble more closely the native fold (10, 11). The conformations of multiprotein complexes in the gas phase, however, have not been addressed, primarily because the conditions and instrumentation developed to preserve such interactions are not readily coupled with established approaches that probe higher-order structure. Consequently, although it has been widely recognized that protein complexes can survive within the solvent-depleted environment of the mass spectrometer (1318), their overall topology is unknown.

In order to investigate the architecture of water-depleted protein complexes, we selected the trp RNA binding attenuation protein (TRAP). This small protein (circa 8 kD) forms oligomers with 11 subunits, and x-ray analysis has demonstrated a ring topology for the complex (19). Previous mass spectrometry experiments showed that we can examine the stoichiometry of the complex under a variety of conditions and also enabled us to suggest a tryptophan binding mechanism (20). Given the large number of subunits and their ring arrangement, we reasoned that such a system would allow us to distinguish changes in collision cross section that would occur if the native ring structure of the complex were to collapse as solvent is depleted. By using high-transmission ion mobility measurements of the noncovalent protein assembly, we investigated the collision cross sections of various TRAP complexes generated by nanoflow electrospray from an aqueous solution buffered at pH = 7.0. Ion mobility separation was carried out by using a traveling wave ion guide cell within a modified quadrupole time-of-flight mass spectrometer (Materials and Methods) (21). This instrumental arrangement (fig. S1) allows separation of ions on the basis of their ability to traverse an environment of neutral gas molecules under the influence of a weak electric field. The time taken for an ion to travel a defined distance is recorded, and this transit time is then converted to a collision cross section, a parameter directly related to the size and overall topology of an ion (22). Various computational approaches are used to generate possible model structures on the basis of this information (9, 23).

A mass spectrum recorded for an aqueous solution of apo TRAP shows four broad peaks corresponding to the 19+ to of 11-mer 22+ charge states the protein complex (Fig. 1A). The broadness of peaks arises from the encapsulation of solvent or buffer molecules within the assembly and is consistent with the measured mass being higher than that anticipated for naked protein subunits. For each of the charge states, the time taken to transit the ion mobility device (on the order of 15 to 30 ms) is normalized for the influence of charge and converted to a collision cross section (Fig. 1A, a to d) (Materials and Methods). The 19+ charge state has the largest collision cross section (6600 Å2), and this measurement is in agreement with cross-section estimates derived from the ring structure determined by x-ray crystallography (19). The 20+ charge state exhibits a bimodal distribution, indicative of structural heterogeneity, whereas the peaks assigned to the 21+ and 22+ are the most compact, consistent with closely packed configurations of protein subunits. Overall, the collision cross sections that we have measured for apo TRAP are in accord with the native architecture and close-packed collapsed states but also encompass a range of conformers with intermediate and lower collision cross sections than those anticipated by models constructed from simple, noncompressible spheres.

Fig. 1.

Ion mobility–mass spectrometry data for the TRAP complex in ligand-bound and unbound forms. For all data, blue and green dashed lines correspond to collision cross sections for ring and collapsed structures based on the native state (6500 Å2) and close-packed icosahedral (5600 Å2) arrangements of spheres, respectively. A representation of the crystal structure of TRAP [(19), PDB identification number 1C9S, created using the program SIB Swiss PDB Viewer software] illustrates the positions within the complex of RNA, tryptophan (space-filling representation), and protein subunits (blue). (A) Mass spectrum of the intact apo TRAP complex. (a to d) Ion mobility data for charge states of the intact 11-mer TRAP complex superimposed on a collision cross-section axis. (B) Mass spectrum of the 11-member TRAP complex bound to 11 tryptophan molecules. (a to d) Ion mobility data for the charge states of the tryptophan-bound TRAP complex. (C) Mass spectrum of the intact apo TRAP complex as well as the 11-member complex bound to both 11 tryptophan molecules and a 53-base segment of RNA. Ion mobility data for 22+/21+ charge states of apo TRAP (a and b) and 20+/21+ charge states of RNA/Trp/TRAP complex (c and d).

To generate model structures that more accurately describe both the intermediate and highly compact structures described above, we used molecular dynamics methods. We generated a model that uses compressible spheres of the approximate diameter of TRAP subunits and a simple potential energy surface to drive the ring structure to collapse over the course of the simulation. The conformational space around this pathway is then explored with use of a random vector (R) that increases in magnitude in successive simulations of the same pathway (Materials and Methods). One set of simulations, including 5000 separate quaternary arrangements of TRAP, shows that the range of collision cross sections measured for the 19+ charge state encompasses the native ring architecture and includes partial ring structures as well as buckled rings (Fig. 2). In addition, the simulation was able to generate a small number of highly compact structures that agree closely with the collision cross sections measured for the most compact charge state (22+) of the apo TRAP protein complex. These compact quaternary structures are characterized primarily by protein subunits that overlap with respect to the original boundaries of the subunits defined at the outset of the simulation.

Fig. 2.

Plot of collision cross section versus simulation step for 10 separate molecular dynamics simulations of the collapse of an 11-member protein complex, differing with respect to the amount of random energy given to the subunits that comprise the complex at each step in the simulation (represented by the value of R, the magnitude of which is indicated by the color scale shown in the lower left-hand corner of the plot). Structures A to D were selected from the simulation that correspond to the experimentally determined collision cross sections (A, ∼6500 Å2; B, ∼6000 Å2; and C, ∼5600 Å2). Dashed circles indicate the position of the model structures on the simulated reaction coordinate. The structure labeled D (∼5300 Å2) corresponds to the most compact structure generated by the simulations.

The stability of the native quaternary structure of TRAP is known to increase upon binding tryptophan (19). Mass spectrum and ion mobility data recorded for ions electrosprayed from a solution of TRAP with a two-fold excess of Trp are similar to those observed for apo TRAP (Fig. 1B). The maximum collision cross section measured for the tryptophan-bound holo form is closely similar to the apo form, consistent with tryptophan binding between protein subunits, as established crystallographically (19). However, significant differences can be observed by comparing the ion mobility data for the 20+ charge state of the apo and holo complexes (Fig. 1, Ad and Bc); the former displays a bimodal distribution of transit times assigned to a range of conformers, whereas for the latter a single distribution is observed, consistent with the native-like ring configuration. We therefore conclude that the putative ring-type structure of the holo form of TRAP is more stable than the apo form in the absence of bulk solvent.

It has also been established that, in the presence of tryptophan, a 53-base segment of the trp leader mRNA binds around the perimeter of the TRAP complex and stabilizes the resulting assembly in solution (20). The mass spectrum recorded for the TRAP-Trp RNA solution shows low intensity peaks (22+ and 21+) for TRAP 11-mer, and at higher m/z (mass/charge) values the spectrum shows more intense peaks of the same charge states for the RNA-bound complex (Fig. 1C). For the apo TRAP ions, collision cross sections are similar to those measured for TRAP in the absence of RNA, confirming the reproducibility of our measurements (Fig. 1C, a and b). The RNA-bound species, however, have the broadest distribution of all collision cross sections (by a factor of 2), most likely arising from inherent chemical heterogeneity (Fig. 1C, c and d) (24). Moreover, the longest ion mobility transit times (∼28 ms) of all the structures examined here are recorded for the RNA-bound species, consistent with the largest collision cross section (7400 Å2). This represents a 12% increase in collision cross section with respect to the ring structure of the apo form and is consistent with calculated values for the protein complex upon RNA binding. Our data are therefore in accord with established models that show RNA binding to the periphery of the protein complex. We also find that the contribution from collapsed structures to the distribution of conformers is negligible, in agreement with the enhanced rigidity of the ring when RNA is bound to the perimeter.

The observation that the structure of the apo and holo TRAP protein complexes change as a function of charge state is an unexpected one. We anticipated that the collision cross sections of the complex would be relatively insensitive to changes in charge state because the charge per subunit for all the ions is well below the threshold for Coulombic unfolding observed for monomeric proteins (25, 26), although previous ion mobility experiments have indicated that the gas-phase structures of small proteins can vary dramatically within a narrow distribution of charge states (10, 11, 26, 27). One explanation for the apparent collapse of protein quaternary structure at higher charge states is that minor fluctuations, caused by Coulombic repulsion, could lead to small, local aberrations in structure that subsequently collapse the native-like ring. A second possibility is that, rather than strictly a charge-state dependence, the increased number of solvent and buffer molecules (∼10 to 50) present, particularly in lower charge states, within large protein-protein complexes could contribute to their stability (27). A third possibility is that complexes with higher charge states (21+ or 22+) will be subject to more energetic collisions with neutral gas molecules and will consequently accumulate larger amounts of internal energy, leading to changes in structure.

To investigate the sensitivity of the various conformers to changes in internal energy, we examined collision cross sections of the apo TRAP complex as a function of activation energy by manipulating their acceleration in the atmospheric pressure interface of the instrument (Fig. 3). For the 22+ ions at the lowest activation energies (50 and 125 V), collision cross sections are indicative of close-packed collapsed structures. At 175 V the collision cross section increases, implying that the higher internal energy leads to a larger structure, consistent with the activation of ions initially in very compact configurations. The distribution of collision cross sections for 21+ ions of apo-TRAP is essentially constant throughout the change in acceleration voltage, consistent with a collapsed structure that is less compact than the 22+ ions. For the 19+ charge state as acceleration is increased to 125 V, the collision cross section assigned to the ring-like structure persists, but at 175 V decreases significantly and is consistent with a partially collapsed or buckled ring-type structure. This experiment demonstrates that imparting internal energy to 22+ ions leads to expansion of the collapsed state whereas for 19+ ions we can drive, at least partially, the structural transitions observed for the ring structure as a function of protein charge state, lending further support to our structural interpretation above.

Fig. 3.

Ion mobility data for selected charge states of apo TRAP (19+, 21+, and 22+) as a function of activation energy (175, 125, and 50 V) applied in the high-pressure, sampling-cone region of the instrument. The dashed lines (green and blue) represent the collision cross sections for collapsed and ring structures described in the legend to Fig. 1.

Taken together, the data show that a native-like ring structure of TRAP subunits can be preserved in the absence of bulk water, allowing an opportunity to investigate their stability under a variety of different conditions. We found that increasing internal energy can lead to structural collapse whereas addition of tryptophan or binding of RNA enhances the overall stability of the ring. It follows, therefore, that the influence of bulk solvent on the quaternary structure of TRAP and its complexes is small in comparison with intrinsic forces within binding interfaces that stabilize the rings. Data available from x-ray analysis of the TRAP crystal structure (28) indicate that the inter-subunit interface corresponds to a surface area of about 1100 Å2, dominated by a central hydrogen bonding region and flanked by two smaller areas of hydrophobic interaction. Similar interaction surfaces exist in numerous protein-protein complexes (29) and indicate that many assemblies may maintain their overall topology in the absence of bulk solvent, paving the way for ion mobility mass spectrometry to make significant contributions to quaternary structure determinations of large protein complexes. We anticipate that the area of structural biology most likely to benefit from the use of ion mobility separation is the analysis of protein complexes that exist in multiple functional forms in solution. As shown here, the RNA-bound complex coexists in solution with the apo form, but because they are readily separated by their m/z ratios, it is possible to characterize all components according to their overall topology. This is an important attribute given that heterogeneity is a major problem in structural biology, the potential for which increases dramatically with the size and complexity of the macromolecular complex (30). Current low-resolution structural biology approaches such as small angle x-ray scattering (31), used for determination of the shape and size of monodisperse complexes in solution, are unable to effectively characterize the chemical and structural heterogeneity that often accompanies macromolecular assemblies. Consequently, rather than determining the shape of each of the components present, most structural probes will report an average over all structures. Therefore, the ability to determine the overall topology of individual components within heterogeneous mixtures by using ion mobility coupled to mass spectrometry represents a significant advance. Moreover, knowledge of the shape of macromolecular assemblies obtained from such experiments has the potential to contribute not only to computational approaches to structure determination (32) but also to characterizing the transient and reversible associations that elude many existing structural biology approaches.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1120177/DC1

Materials and Methods

SOM Text

Figs. S1 to S6

Table S1

References and Notes

References and Notes

View Abstract

Navigate This Article