Article pubs.acs.org/ac
Mass-Selective Soft-Landing of Protein Assemblies with Controlled Landing Energies Victor A. Mikhailov, Todd H. Mize, Justin L. P. Benesch,* and Carol V. Robinson* Department of Chemistry, Physical and Theoretical Chemistry Laboratory, University of Oxford, Oxford, OX1 3QZ, United Kingdom S Supporting Information *
ABSTRACT: Selection and soft-landing of bionanoparticles in vacuum is potentially a preparative approach to separate heterogeneous mixtures for high-resolution structural study or to deposit homogeneous materials for nanotechnological applications. Soft-landing of intact protein assemblies however remains challenging, due to the difficulties of manipulating these heavy species in mass-selective devices and retaining their structure during the experiment. We have developed a tandem mass spectrometer with the capability for controlled ion soft-landing and ex situ visualization of the soft-landed particles by means of transmission electron microscopy. The deposition conditions can be controlled by adjusting the kinetic energies of the ions by applying accelerating or decelerating voltages to a set of ion-steering optics. To validate this approach, we have examined two cage-like protein complexes, GroEL and ferritin, and studied the effect of soft-landing conditions on the method’s throughput and the preservation of protein structure. Separation, based on mass-to-charge ratio, of holo- and apo-ferritin complexes after electrospray ionization enabled us to soft-land independently the separated complexes on a grid suitable for downstream transmission electron microscopy analysis. Following negative staining, images of the soft-landed complexes reveal that their structural integrity is largely conserved, with the characteristic central cavity of apoferritin, and iron core of holoferritin, surviving the phase transition from liquid to gas, soft-landing, and dehydration in vacuum.
P
properties. As such, there is a clear need for means to prepare nanocages that are assembled from certain numbers of subunits, to carry a known quantity of cargo, or to accommodate specific modifications in the protein shell, even when their synthesis is itself heterogeneous. Mass spectrometry (MS) is an inherently selective approach that has in the last 20 years become an important means for characterizing the self-assembly of proteins.12,13 In a nondenaturing MS approach, protein complexes are introduced into the mass spectrometer intact by means of electrospray ionization (ESI) from aqueous solutions in which native noncovalent interactions are retained and manipulated and measured in the vacuum of the instrument.14−16 Notably, on the time scale of the MS experiment, the overall architecture of protein assemblies is typically retained.17 This has not only allowed the direct structural interrogation of proteins within the mass spectrometer, but has also opened the door to the exciting possibility of using the instrument as a high-resolution means of purifying intact protein assemblies.18 The potential benefits of the use of MS as a preparative technique for ex situ investigation of protein assemblies, or for their use in biotechnological applications, are significant, as MS
roteins assemble into a myriad of structures, providing inspiration to exploiting both natural and engineered forms for nanotechnological applications that span the biological, physical, and engineering sciences.1,2 The structural and dynamical diversity of protein assemblies renders their molecular study and manipulation challenging, both theoretically and experimentally. Many examples, including viral capsids, components of the protein folding and degradation machinery, and metal-storage proteins, have distinctive cagelike structures.2 In nature, these protein nanocages act to sequester important and sensitive cargo from the surrounding environments. The potential applications of mimicking such activity are considerable and include serving as contrast agents in magnetic resonance imaging,3 as drug delivery vehicles,4 catalytic nanoreactors,5,6 templates for nanoparticle synthesis,6,7 and nanoelectronic devices.8 Transmission electron microscopy (TEM) is often used to characterize such assemblies, having the advantage of providing sub-2-nm structural resolution relatively rapidly and without requiring much sample.9 TEM has been used to validate predicted nanocage structures10 and to characterize encapsulated nanocrystals to help optimize crystallization protocols.5,6,11 Three-dimensional particle reconstruction from TEM data however requires homogeneous particles, which frequently necessitates purification before imaging, or in silico afterward. Similarly, technological applications of nanocages often rely on their homogeneity, so that they have precisely defined physical and chemical © 2014 American Chemical Society
Received: May 16, 2014 Accepted: July 15, 2014 Published: July 15, 2014 8321
dx.doi.org/10.1021/ac5018327 | Anal. Chem. 2014, 86, 8321−8328
Analytical Chemistry
Article
holo-ferritin freshly buffer-exchanged into ammonium acetate using BioSpin 6 microcolumns (Bio-Rad Laboratories, Hercules, CA). Control TEM experiments revealed that after 24 h in ammonium acetate the holo-protein complexes lose their mineral cores. Purchased GroEL was further purified prior to buffer-exchange.27,43 Mass Spectrometry and Ion Soft-Landing. The modified quadrupole time-of-flight mass spectrometer (QToF 2, Waters, Milford, MA) has been described elsewhere.18,29 Briefly, it was modified for high m/z transmission29 with additional ion optics introduced after the ToF pusher to steer and focus the ion beam toward a target surface.18 Additional modifications include the probe now being prepumped using an oil-free scroll pump (Edwards, Sanborn, NY) and an additional disk electrode with a 2.7 mm diameter central hole, placed 1 mm in front of the TEM grid. The scroll pump has improved the vacuum conditions in the soft-landing stage while concomitantly eliminating condensation of oil vapor onto the TEM grid. The extra electrode allows us to decouple the decelerating and focusing fields in the ion optics. The ion flux is optimized using a miniature channeltron (Burle Industries, Lancaster, PA) placed behind the soft-landing stage. Additionally, the ion current on the TEM grid during the soft-landing process can be measured directly using a picoammeter (6487, Keithley Instruments, Cleveland, OH). Ion soft-landing was carried out with the quadrupole in RFonly mode, with all ions passing through the quadrupole filter and reaching the TEM grid. Ions (GroEL, 803 kDa; apoferritin, 443 kDa) were nanoelectrosprayed using a capillary voltage of 1.55 kV, a sample cone voltage 200 V, an extractor cone voltage 100 V, and the hexapole ion guide pressure in the range of 5 × 10−3 to 1.4 × 10−2 mbar. First, ToF MS spectra were obtained and optimized for the ions of interest; then the reflectron and detector voltages were turned off to avoid unwanted ion deflection during soft-landing, and lastly, the probe was moved to position the TEM grid in the ion beam behind the ToF pusher. Ion trajectories for GroEL were simulated using the SIMION 7 suite of programs (Scientific Instrument Services, Ringoes, NJ), with m = 810 kDa; z = 65+; an initially square beam of 1 × 1 mm; angular distributions randomized ±25° from the elevation and azimuth of the axis; and an Ek0/z of 1, 5, or 10 eV randomized ±25%. A total of 200 ion trajectories were run for each energy value. Electron Microscopy. Carbon-coated copper TEM grids, 200−300 mesh, were purchased from Electron Microscopy Sciences (Hatfield, PA). The grids were cleaned by washing in chloroform and glow-discharged in low-pressure oxygen for 10−20 s immediately prior to the soft-landing experiment. TEM grids with soft-landed complexes were stained using uranyl acetate (2% w/w) using standard procedures. Uranyl acetate (Sigma-Aldrich, St. Louis, MO) is a mildly radioactive and chemically toxic compound, requiring appropriate handling, according to safety regulations. TEM images were obtained on a Tecnai 12 electron microscope (FEI, Hillsboro, OR, USA) using a FEI Eagle 4k × 4k CCD camera. Magnification was in the range 45−90 k, defocusing −0.8 to −1.2 μm. Images from soft-landing experiments were compared with control samples pipetted directly from solution and acquired on the same microscope with the same operational settings.
has unsurpassed selectivity based on high-resolution separation of different ionized species according to their mass-to-charge (m/z) ratios. This advantage is critical in structural studies when dealing with polydisperse or heterogeneous proteins that frustrate characterization by other traditional methods.19 MS has been used for selective preparation of monodisperse chemical samples, such as metal clusters,20−22 organic molecules,23 or polymers,24 followed by downstream analysis of the sample by various methods. The progress of MS for protein complexes is hindered by the fact that the ion transfer rate of ESI MS is low, with initial ESI ion currents typically in the nA range,25,26 and dropping to the pA level before the detector.26 This problem is exacerbated for protein assemblies, because of the need to modify the instrument for elevated pressures in the transmission regions to cool and focus these typically 10 kDa to 10 MDa ions.27−29 The greatest challenge is, however, the successful deposition of protein assemblies onto a surface in the vacuum of the mass spectrometer without significant distortion of their structure. This concept is known as ion “soft-landing”30−32 and involves transportation of mass-selected ions with hyperthermal kinetic energies ( +30 eV, the absolute value of the potential at electrodes 1 and 4 decreases proportionally and the focusing conditions are compromised. Both effects lead to a dramatic reduction in the ion flux in either direction from ΔEland/z = +30 eV. Significantly, it is not possible to achieve ΔEland/z < 0, i.e. decelerate apo-ferritin and GroEL ions, by varying Vbias. In order to broaden the range of landing energies available for soft-landing and achieve the decelerating conditions for ferritin and GroEL ions, we fixed Vbias and V1 to V5 at the values corresponding to the maximum of the channeltron signal (Figure 3, blue circles) and varied V6 with respect to Vbias (Figure 3, red squares). In accordance with our SIMION simulations, this manipulation of landing energies just before the TEM grid (Figure 2) does not affect focusing conditions in the preceding steering optics. Therefore, this arrangement renders it possible to decelerate the ions just before their softlanding, while maintaining sufficient ion flux across a wider range of Eland. Importantly, in all cases when the landing energy was varied, the ion flux decreased with decreasing ΔEland/z (Figure 3 and Figure S2). This is because, in order to reduce the velocity of the ions, a potential barrier has to be raised either at the entrance to the soft-landing optics or at the TEM grid (Figure 3). In either case, all ions with Ek0/z below the barrier will not reach the grid (detailed diagrams given in Figure S3). It is possible to estimate the spread of ion kinetic energies by extrapolation to negative values of ΔEland/z where the ion flux
(2)
Voltages here are reported in volts, e = 1 elementary charge, and ΔEland/z is reported in electronvolts (eV) per elementary charge. From the equations above, it follows that ions are accelerated before soft-landing when Ventr > Vgrid, Eland > Ek0, and ΔEland/z > 0. Ions are decelerated when Ventr < Vgrid, Eland < Ek0, and ΔEland/z < 0. It also follows that a distribution of kinetic energies at the ToF entrance, Ek0, results in a distribution of landing energies offset by ΔEland. Knowledge of the spread of Ek0 is important for evaluation of the spread of Eland. Furthermore, changing steering voltages will affect focusing and, thereby, both ΔEland and ion flux onto the TEM grid. It was therefore crucial to investigate this effect, with the ultimate goal of optimizing landing voltages to maximize ion flux onto the TEM grid while minimizing the ions’ kinetic energies. First, we modeled the ion optics to estimate steering and focusing potentials necessary for deposition on the grid (Figure 2c). The simulations of GroEL65+ show that focusing conditions are readily achieved, and that the landing energies and number of ions successfully reaching the TEM grid depend on the initial ion energy and the voltage applied to the grid, Vgrid (regulated by either Vbias or V6). Vbias affects the focusing conditions in all steering optics, while V6 only affects the field near the grid. When Vgrid is increased, the landing energies are decreased (eq 1). However, the convergence efficiency of ions onto the grid decreases as Vgrid approaches Ventr, i.e. as ions are decelerated for soft-landing. For example, for Ek0/z = 10 eV, 70.0% of the ions reach the grid in the case of Vgrid −Ventr = −10 V while only 50.5% reach the grid for Vgrid − Ventr = +10 V. This demonstrates that achieving lower landing energies by deceleration reduces the ion flux reaching the grid. To test this prediction, we measured the ion flux through an uncoated TEM grid for different landing energies. Specifically, we manipulated ΔEland/z by varying Vgrid and monitored the signal on the channeltron (Figure 3, Figure S2). In the case of the tetrameric protein concanavalin A ions (102 kDa, z = 17− 21+), a large range of landing energies could be achieved by 8324
dx.doi.org/10.1021/ac5018327 | Anal. Chem. 2014, 86, 8321−8328
Analytical Chemistry
Article
unloaded nanocages in the solution. Additionally, to explore whether the prolonged exposure to vacuum during particle collection in the mass spectrometer affects the appearance of the complexes after staining, we prepared control samples in which the complexes were pipetted directly onto the grids, the excess solution wicked away, and the grids air-dried for 10−15 min. The grids were then exposed to the vacuum inside the mass spectrometer for time periods comparable to those experienced in typical soft-landing experiments (1−3 h). After removal from the vacuum, the grids were stained with uranyl acetate and imaged by means of TEM. For both GroEL and apoferritin, particles are clearly observed with dimensions comparable to those for the proteins that had not been exposed to a vacuum (Figures 4c,d, S4, and S5 respectively). Notably the particles in this case are positively stained and therefore reveal fewer details than the negatively stained images obtained from traditional preparations (Figure 4a,b). The ability of uranyl acetate and other metal salts to cause this phenomenon in some cases is well-known.45,46 We speculate that two effects take place in a vacuum that may be responsible for the observed change: complexes are dehydrated, and the hydrophilicity of the carbon film decreases during extended exposure within the mass spectrometer, thus reducing the adhesion of the stain to the carbon film and creating conditions more favorable for positive staining. We also tried polylysine coated carbon films to maintain the film hydrophilicity but did not observe an increased propensity for negative stain. Attempts to rehydrate the samples by pipetting water onto grids, after they were removed from the vacuum, resulted in either positive stain or loss of the sample. A similar effect of vacuum exposure on TEM images of soft-landed complexes might also be anticipated. Ion Soft-Landing at High Landing Energies. The voltage settings that provide ΔEland/z > 0 (accelerated ions) result in relatively high ion flux on the TEM grid (Figure 3) and therefore were the first soft-landing conditions we explored. Accordingly, we obtained TEM micrographs of soft-landed GroEL and apoferritin (Figures 4e,f and S5) at Eland/z ≥ 30 eV, with ≈1−3 h deposition time. In both cases, we observe positively stained particles and, albeit at coarse resolution, it is apparent that the integrity of the soft-landed complexes is retained. Furthermore, the images of apoferritin can be distinguished from those of GroEL: the former have a large central cavity, which is remarkably well preserved after softlanding. The integrity of the deposited GroEL complexes is better than those from vacuum-exposed complexes (Figures 4c, S4), where some complexes have dissociated. Nonetheless, changes in the structure upon soft-landing are likely, as evidenced by the dimensions of the apoferritin complexes being larger than in the controls (Figure 4b,f). This could be due to excess landing energy causing the hollow nanocage to deform upon impact. Notably, however, the background noise in the EM micrographs is significantly reduced in comparison with our previous results.18 We attribute this to an improved cleaning protocol for TEM grids and better vacuum conditions. Ion Soft-Landing at Low Landing Energies. To minimize deformation induced by the impact on the surface, we reduced landing energies by deceleration, i.e. ΔEland/z < 0. As demonstrated above, the ion flux is significantly attenuated under these conditions (Figure 3), hindering identification of GroEL and apoferritin complexes that were soft-landed in much smaller numbers than at ΔEland/z > 0. Therefore, we elected to soft-land holoferritin, as its protein shell with an iron
becomes zero. For both concanavalin and apo-ferritin, the extrapolated values are 0) and decelerated (ΔEland/z < 0) ions, respectively (Figure 3). In order to compare the soft-landed complexes with those deposited directly from solution, we first obtained traditional negative stain TEM images for GroEL and a mixture of apoand holoferritin (Figure 4). In both cases, the cage-like architecture of the protein complexes is clearly visible. Dark spots in the middle of some ferritin complexes correspond to the iron mineral cores (Figure 4b), while other protein shells being empty indicates the heterogeneous mixture of loaded and
Figure 4. Control TEM images and images from soft-landing at high landing energy. Box size 200 × 200 nm. (a) Standard staining of GroEL with uranyl acetate. (b) Standard staining of apo- and holoferritin. (c, d) GroEL and apoferritin, respectively, stained after vacuum exposure for 2 h. (e) Stained soft-landed GroEL complexes: ΔEland/z = 30 eV, 49 min deposition (first two in column); ΔEland/z = 40−62 eV, 147 min deposition (last in column). (f) Stained softlanded apoferritin complexes ΔEland/z = +50 eV, 145 min deposition. 8325
dx.doi.org/10.1021/ac5018327 | Anal. Chem. 2014, 86, 8321−8328
Analytical Chemistry
Article
“Washing” TEM grids with a droplet of water after soft-landing and prior to staining led either to the reduction of the total number of detected complexes or their aggregation (Figure 5e) or apparent distortion of the protein shells and loss of subunits from the complex (Figure 5f, small bright spots). It is possible that changes are induced by washing, with adhesion between the protein subunits weakened by dehydration in a vacuum, and subsequent dissolution upon addition of water and acidic uranyl acetate solution. Regardless, even though the ion flux and the number density of the complexes on the TEM grid are greatly reduced for ΔEland/z < 0, it is clear that soft-landing of holoferritin, and its identification, is possible at low landing energies.
core gives readily identifiable TEM images featuring dark spots surrounded by brighter protein shells.47 We found that positive staining affected the images of vacuum-exposed holoferritin less than those of GroEL and apoferritin since the mineral cores still display higher contrast than stained protein shells (Figure 5),
■
DISCUSSION We have demonstrated that it is possible to soft-land protein assemblies with controlled kinetic energies in a mass-selective manner. We have validated this preparative approach by means of TEM and applied it to two large protein nanocages, GroEL and ferritin. We effected separation of apo- from holoferritin by manipulating the ion transfer conditions in the front part of the mass spectrometer. Our TEM images show that the integrity of the soft-landed complexes is largely preserved, validating our strategy as an approach for preparing purified protein assemblies. Recombination of positive charges on the softlanded ions with electrons may take place on the surface of the carbon film,38 but, apparently, liberation of the recombination energy does not lead to dissociation of protein assemblies. Therefore, the complexes survive the phase transition from liquid to gas, soft-landing, and vacuum exposure. The prototype instrument that we used is based on a standard Q-ToF modified for transmission of high-m/z ions and equipped with a relatively simple set of additional ion optics to control soft-landing. This design is readily translatable to other regions of the Q-ToF instrument, or different mass spectrometer geometries altogether. Future iterations of the approach will be concerned with improving ion flux, and consequently the number density of the soft-landed complexes, preservation of the structure of the complexes during and after the soft-landing, and optimizing the signal-to-noise ratio of the TEM analysis. Particle Number Density of the Soft-Landed Complexes. A high initial flux of complexes is crucial for highresolution m/z selection, for example in isolating and softlanding complexes in a particular oligomeric form or charge state, or for nanocages carrying specific cargo. Lower deposition times will result in higher throughput and a concomitant reduction in vacuum-induced structural changes. Ideally, particle counts > 100 μm−2 of soft-landed complexes are desirable for single-particle TEM analysis, where typically thousands of particles are averaged to reconstruct a low resolution structure. Lower signal-to-noise ratios typical in electron cryo-microscopy (cryo-EM), which is favored for high resolution and reducing deleterious effects attributable to staining,9 will require greater numbers of particles. At present, the number density we have achieved is approximately 10−20 μm−2 for ΔEland/z > 0 with deposition times of 50−150 min (particle flow ca. 106 min−1 for 3 mm diameter grid), falling by an order of magnitude for ion decelerating conditions, ΔEland/z < 0. Assuming an efficiency of soft-landing of unity, in other words that every ion that reaches the TEM grid is soft-landed and retained until visualization, we can put a lower bound on the ion current of 0.14 pA. This estimate is in agreement with
Figure 5. Control holoferritin TEM images and holoferritin images from soft-landing at low landing energy. Box size 100 × 100 nm. Top row: control staining after vacuum exposure (a, b) for 2 h and (c) for 30 min. (d) Staining after soft-landing at ΔEland/z = −10 eV, 78 min deposition. (e) Staining after washing the grid in water following the soft-landing at ΔEland/z varied from 0 to −30 eV, 89 min deposition. (f) Staining after washing the grid in water following the soft-landing at ΔEland/z varied from −10 to −30 eV, 77 min deposition.
helping us to identify these complexes unambiguously. Furthermore, aggregation of holoferritin resulting in groups of homogeneous particles was often observed, thus aiding identification of holoferritin (Figure 5b,e). A similar aggregation of soft-landed BSA molecules has been observed in AFM images by others42 and was attributed to the mobility of proteins on the carbon surface. TEM images of holoferritin complexes (>650 kDa) selected in the mass spectrometer as described above and soft-landed at ΔEland/z < 0 are similar to those from vacuum-exposed control samples (Figure 5d−f), and demonstrate that the integrity of holoferritins is therefore clearly preserved upon soft-landing. 8326
dx.doi.org/10.1021/ac5018327 | Anal. Chem. 2014, 86, 8321−8328
Analytical Chemistry
Article
S7). While the washing causes significant particle loss, excellent images are obtained, demonstrating the feasibility of this approach for future soft-landing implementations that allow for higher ion current. In principle, our method is not limited by its application to TEM grids; other methods of visualization, including label-free optical methods,53 could be employed for detection of single proteins.
our direct current measurements of