Neutral mass spectrometry of virus capsids above 100 megadaltons with nanomechanical resonators

See allHide authors and affiliations

Science  23 Nov 2018:
Vol. 362, Issue 6417, pp. 918-922
DOI: 10.1126/science.aat6457

Bridging the mass gap

Viruses and many large biomolecule complexes are in a mass range that is challenging to measure with conventional mass spectrometry methods. Nanomechanical resonators can determine masses of impacting molecules, but separation methods often lose too much of the sample to be efficient. Dominguez-Medina et al. used an aerodynamic lens that improved separation and focusing of nebulized molecules with increasing mass. The mass of both filled and empty viral capsids was determined with an array of 20 nanoresonators.

Science, this issue p. 918


Measurement of the mass of particles in the mega- to gigadalton range is challenging with conventional mass spectrometry. Although this mass range appears optimal for nanomechanical resonators, nanomechanical mass spectrometers often suffer from prohibitive sample loss, extended analysis time, or inadequate resolution. We report on a system architecture combining nebulization of the analytes from solution, their efficient transfer and focusing without relying on electromagnetic fields, and the mass measurements of individual particles using nanomechanical resonator arrays. This system determined the mass distribution of ~30-megadalton polystyrene nanoparticles with high detection efficiency and effectively performed molecular mass measurements of empty or DNA-filled bacteriophage T5 capsids with masses up to 105 megadaltons using less than 1 picomole of sample and with an instrument resolution above 100.

The current lower limit to weighing individual objects is in the picogram range and is obtained with piezoelectric resonators (1). Twelve orders of magnitude lower on the mass scale, mass spectrometry (MS) uses ionization, electromagnetic fields to manipulate ions, and ensemble averaging of mass-to-charge ratios to identify species on the basis of their specific molecular mass. Since the advent of soft ionization techniques (2, 3), commercial spectrometers have performed routine proteomics analysis of peptides and proteins in the range of 1 to 100 kDa (1 kDa = 1.66 × 10−24 kg = 1.66 × 10−9 pg). Recent MS research has expanded the mass range accessible with conventional techniques, up to the 10-MDa range (4). Alternative MS techniques, such as charge-detection MS (CDMS) (5, 6), can analyze single ions with masses up to 1 GDa. Nevertheless, unmodified commercially available MS instruments measure masses only up to 1 MDa.

Nanomechanical resonators measure the mass of individual particles accreting on their surface. Because the frequency-to-mass relation scales with the resonator’s characteristic dimension d−4, researchers in the field of nanomechanics have engaged in a race to measure the smallest detectable mass with ever smaller resonators (711). These efforts were driven by the hope of competing with commercial MS down to the Dalton mass range, whereas the MS community has worked in the opposite direction to measure larger masses. Nanomechanical resonators fabricated by scalable silicon processes are actually ideally suited for analysis of masses in the mega- to gigadalton range, which includes most viruses, many disease biomarkers, and species with defined molecular mass.

The analysis of viruses by nanomechanical MS has been pursued for several years (12). Unfortunately, early attempts to perform nanomechanical MS were hampered by a combination of losses associated with ionization yield and ion transfer, and the small cross-section (typically a few square micrometers) of the capture area on nanoresonators. These issues led to prohibitively long analysis times and excessive sample consumption (13, 14). However, nanomechanical MS of metallic nanoparticles of any charge state (ionized or neutral) (15) was demonstrated, and recently, the possibility of measuring particle stiffness simultaneously with mass and position was demonstrated with a system that operates without ion guides in the tens to hundreds of gigadaltons range (16).

To achieve the full potential of nanomechanical MS in the MDa to GDa range, we report here on a system architecture specifically designed for this mass range (Fig. 1) that features high efficiency and excellent resolution by circumventing the requirement for ionization. It relies on nebulization of analytes from solution at atmospheric pressure, exploits the particles’ inertia to efficiently transfer and focus them without any need for electromagnetic fields, and determines the mass of individual particles by using arrays of nanomechanical resonators in high vacuum.

Fig. 1 High-transmission system architecture for nanomechanical resonator–based charge-independent single-particle mass sensing.

The setup consists of three chambers with decreasing pressures. Analytes in solution are nebulized by SAWN or nano-ESI and aspirated through a heated metal capillary inlet at atmospheric pressure. An aerodynamic lens focuses the particle stream (shaded blue area), which is then transferred onto an array of frequency-addressed nanomechanical resonators. Left inset: Simplified schematic of the aerodynamic lens (actual design shown in fig. S8). Right inset: SEM image of a portion of an array showing 12 out of 20 resonators and magnified SEM image of a single resonator with its metallic layer and silicon material, falsely colored in yellow and gray, respectively. Resonator dimensions: 160 nm (thickness), 300 nm (width), and 7 to 10 μm (length).

Nebulization of particles was performed by surface acoustic wave nebulization (SAWN) (17) or by electrospray ionization (ESI) (4). With SAWN, a liquid sample deposited on a piezoelectric surface is nebulized as a thin mist upon propagation of an acoustic wave across the surface. Compared to ESI, which produces a rapidly expanding jet through Coulomb repulsion, SAWN produces droplets with lower kinetic energy, resulting in very efficient uptake into the system’s inlet capillary. Nevertheless, ESI was also used as spraying conditions have been extensively optimized for numerous species of interest, including viruses (4).

The small capture area of nanoresonators necessitates a low particle-beam–to–detector-size ratio for efficient particle detection. We used an aerodynamic lens composed of a pressure-limiting orifice followed by a series of apertures between relaxation volumes (Fig. 1). We designed and optimized this lens for the mass range of interest (MDa to GDa) following previously reported guidelines (18). Compared to conventional ion guides, which counteract inertia to maintain heavy particles on a stable trajectory, an aerodynamic lens exploits the particles’ inertia for focusing, so its performance improves with increasing mass (19).

The detector consisted of 20 nanomechanical resonators (arranged in a four-by-five array covering a 50 μm by 230 μm area). We fabricated doubly clamped beams with electrostatic actuation and differential piezoresistive readout using large-scale silicon processes (20). Particles landing on the vibrating part of a resonator add to its total mass M and cause its resonance frequency f to down-shift Embedded Image. As these frequency shifts also depend on the landing position on the resonator’s surface, the frequencies of two resonance modes were monitored simultaneously to resolve the two unknowns (i.e., mass and position) (14, 15, 21). With an array of 20 resonators multiplexed in time, an increase of more than an order of magnitude in capture cross-section was obtained without degrading mass resolution (2224).

We initially analyzed ~45-nm-diameter polystyrene nanoparticles [PS NPs, National Institute of Standards and Technology (NIST) size standards]. The transmission and focusing performances of our system were characterized by exposing silicon targets to the NP beam produced by SAWN (Fig. 2A). The transmission efficiency, defined as the ratio of the number of NPs on the target to the number of NPs in the nebulized solution, ranged from 3.5 to 7.6%. This transmission was four to five orders of magnitude greater than that achieved with previously reported architectures combining ion guides with nanomechanical resonators (14), and a factor 30 greater than that reported by (16) where, unlike in the present study, efficiency was measured from sprayed NPs, in a two- to three-orders-of-magnitude higher mass range. The measured full width at half maximum (FWHM) of the beam profile at the detector position (1.5 mm) was smaller than the aerodynamic lens outlet and the skimmer’s orifice (Fig. 2B). Moreover, the beam’s solid angle was less than 1°, indicating near-perfect aerodynamic collimation of the NP beam.

Fig. 2 Characterization of SAWN-based transmission and focusing efficiency, and high-throughput mass spectrum of PS NPs.

(A) Diagram showing how the particle beam was characterized by placing silicon targets at the skimmer (~4 cm from aerolens outlet) and nanomechanical detector (~8 cm) positions. (B) Optical images of the targets observed after deposition of NIST PS NPs, with respective Gaussian fit to horizontal sections of the beam profile in particles per μm2 (normalized scale) as measured on a series of SEM images. (C) Normalized raw frequency traces as a function of time for the fundamental mode of 15 nanoresonators exposed to the particle beam produced by SAWN. Each color represents a different nanomechanical device, and each step corresponds to a particle landing event. Some resonators were discarded because of poor signal (fig. S24). (D) Accumulated histogram of mass measurements of the same PS NPs for a nanoresonator array exposed to the particle beam, fit to a normal distribution.

A 1.7 × 107 NP·μl−1 (28.2 pM) solution was nebulized, and NPs were focused onto the resonator array. Frequency time traces obtained for one resonance mode of 15 resonators within the array over 10 min of acquisition (Fig. 2C) showed downward frequency jumps corresponding to individual NP landing “events.” An event rate ranging from 0.3 to 1.8 NP·min−1·resonator−1 was recorded over 128 min of nebulization at an average flow rate of 2 μl·min−1. The NP average size and standard size deviation σ determined beforehand by scanning electron microscopy (SEM) (45 ± 3 nm and σ = 13 nm) were consistent with the NIST specifications (46 ± 2 nm and σ = 7 nm). Converting these size measurements to mass, we could expect a broad mass distribution, with a central mass ranging from 28 to 36 MDa (σ = 15 to 25 MDa) (24). The mass histogram constructed from the nanomechanical measurements of 173 individual NPs (Fig. 2D) showed that the central mass and standard deviation of the normal fit to this histogram (29.5 MDa and σ = 17 MDa) were in good agreement with the expected mass distribution. The absolute mass error in our measurement was mainly caused by errors in the determination of the resonators’ effective mass. We estimated this absolute error to be between 1 and 2% (0.3 to 0.6 MDa) (24), which was well below the uncertainty in central mass expected from size measurements (4 to 6 MDa). Moreover, the standard deviation of the mass distribution expected from size measurement ranged from 15 (NIST) to 25 MDa (SEM), whereas we measured a 17-MDa deviation. This difference was not caused by our measurement noise level: The mass resolution for each measured NP could be inferred from the frequency noise and landing position on the resonator (24), and this estimation yielded an average mass resolution of 0.6 MDa for the 173 events.

The total amount of NPs consumed during this measurement was only 7.3 fmol (4.4 × 109 particles), and the detection efficiency was 1 NP per 2.2 × 107 in solution. These numbers are six orders of magnitude better than previous nanomechanical resonator–based systems using ion guides (14), but still two orders of magnitude lower than the system in (16) that used resonators with two orders of magnitude greater capture area. Indeed, this system operates in a mass range three orders of magnitude higher where efficient detection can be obtained without a focusing device, by accelerating particles through a nozzle with a viscous flow (low vacuum) and placing the resonator at close proximity, at the cost of a mass resolution four orders of magnitude higher than in our work.

We then performed measurements of biological particles: capsids of bacteriophage T5. Bacteriophage T5 is a Siphoviridae family member that infects Escherichia coli bacteria. Its ~90-nm icosahedral capsid is connected to a 250-nm tail, which plays a role in host cell recognition and genome delivery. The capsid itself is composed of 775 copies of the major capsid protein pb8 and 12 copies of the portal protein pb7. Well-controlled capsid assembly and expansion without [“empty” capsids (25)] and with [“filled” capsids (26)] genome packaging inside the capsid have been demonstrated. Once loaded with its 121.75-kbp double-stranded DNA viral genome, the capsid is completed by the head completion protein p144, which constitutes the docking site for tail attachment. Phage T5 is one of the only viruses with such large genome content amenable to in vitro studies. Unlike PS NPs, empty and filled capsids have well-defined molecular masses, calculated to be 26.0 and 105.4 MDa, respectively (see Fig. 3).

Fig. 3 Molecular mass of the bacteriophage T5 capsid.

Top: Negatively stained electron microscopy image of the native bacteriophage T5. The capsid is falsely colored in blue. Bottom: Three-dimensional reconstruction of the assembled capsid measured here [EMDB-8423 (27)], also falsely colored in blue. Dynamic light-scattering size measurements of empty and filled capsids: Both capsids display very similar sizes, whereas their mass differs widely. The table shows the components of the capsid, with theoretical molecular mass calculations for both types of capsid.

Empty capsids were produced from infection of E. coli F cells by a T5stAmN5 mutant defective in genome packaging and purified by anion exchange chromatography (24). Their expansion was triggered by decreasing the salt concentration to 0 mM NaCl through dialysis against 25 mM Hepes buffer at pH 7.2. Expanded empty capsids were stored at 4°C at a concentration of 0.46 mg/ml (~1013 capsids/ml). T5 DNA–filled capsids were produced as described in (27) by infection of E. coli F strain with the mutant T5D18am-dec, defective in tail assembly. The viral particles were purified by precipitation in NaCl–polyethylene glycol followed by centrifugation on a CsCl gradient. The capsids were then dialyzed in phage buffer for storage at a concentration of 2×1013 capsids·ml−1.

Ionization and transfer of large objects such as virus capsids with SAWN has not been reported in conventional MS experiments. Although capsids could be nebulized with SAWN and transferred to the detector chamber, relevant spectra could not be acquired, which was attributed to previously reported aggregation issues (28). We thus chose to use nano-ESI to nebulize the capsids, given the success of this approach for native MS of high-mass biological analytes (4). The capsid solutions were dialyzed against 25 mM ammonium acetate solution for at least 24 hours at 4°C, then slowly diluted to a final buffer concentration of 12.5 mM in 10% (v/v) methanol. The dialysis against ammonium acetate, addition of methanol, and an adequate capsid concentration were necessary for stable spray conditions. Empty and filled capsids were thus further diluted and sprayed at a concentration of ~5.9 × 1011 and ~8.8 × 1010 capsids·ml-1, respectively. The dialysis procedure did not affect the stability of the capsids in solution, as almost identical dynamic light-scattering measurements were obtained for the filled capsids in their original buffer and after dialysis in ammonium acetate (fig. S21).

Empty and filled capsids were electrosprayed into the system and detected by the array of nanoresonators. Figure 4, A and B, show representative frequency traces of eight resonators for empty and filled T5 capsids, respectively. Spraying the buffer solution alone produced almost no detectable mass events nor measurable drift, suggesting efficient buffer desolvation given its volatility, the heated inlet capillary (150°C), and the high-vacuum environment. Figure 4C shows the mass distribution obtained for empty capsids. The spectrum displayed a bimodal distribution, which was fitted by two Gaussian peaks. The main peak fit had a central mass of 27.2 MDa (σ = 2.3 MDa) with the most abundant bin at 26 MDa, near the calculated 26.02 MDa; the second peak had a mass of 33.4 MDa (σ = 2.8 MDa). From the mass resolutions obtained for each single particle in the main peak, it was possible to deduce a 1.0-MDa FWHM resulting from the instrument alone (i.e., considering negligible sample heterogeneity) (fig. S22). This value corresponds to a resolution of 27 (MMFWHM), which is comparable to charge-detection MS measurements in this mass range (6). Moreover, it suggests that the two peaks could be ascribed to two distinct capsid populations present in the sample. This result was further substantiated by a differential scanning calorimetry experiment (fig. S20) showing two denaturation profiles, and could be attributed to the presence of a remaining short DNA fragment associated to some T5 capsids during purification. Finally, 363 events within the mass range of Fig. 4C were obtained with 645 fmol of capsids, over the course of 302 min (one particle detected per 1 × 109 in solution), with an event rate ranging from 1.15 to 1.35 capsid·min−1. This result confirms the advantages of using our architecture over previous NEMS-MS systems using ion guides.

Fig. 4 Molecular mass measurement of empty and filled bacteriophage T5 capsids.

(A and B) Representative raw frequency traces while spraying empty (A) and filled (B) capsids of bacteriophage T5, shown in fractional frequency change (Δf/f0) with respect to their initial frequency f0. For the sake of readability, the two modes of only eight nanoresonators are shown for 1600 s out of the whole experiment. (Other devices and time samples are shown in figs. S24 and S25.) (C and D) Accumulated mass histograms of 363 empty (C) and 648 filled (D) capsids nebulized using nano-ESI (with 1- and 2-MDa bin size, respectively), fitted to two and single normal distributions, respectively. Mass resolution per particle for empty (E) and filled capsids (F), with their respective particle landing position histograms (insets) in normalized units (beam extremities are at position 0 and center is at position 0.5). Capsids landing below 0.25 were discarded because of their poor mass resolution.

We then nebulized filled capsids and measured their mass distribution (Fig. 4D). In the filled capsid spectrum, a clear peak emerged with a fit at a mass of 108.4 MDa (σ = 6.0 MDa), and the most abundant bin at 107.5 MDa. The central mass of this distribution was within 2.5% of its calculated molecular mass (105.4 MDa). The slight remaining discrepancy and the peak asymmetry could be the result of salt adduction during the electrospray process, which reportedly yields masses slightly greater than theoretical estimates (29). This discrepancy is higher in absolute value for the filled than empty capsids (3.1 versus 1.2 MDa), presumably because the filled capsids’ storage buffer contained a much higher salt concentration than that of the empty capsids. In addition, it is likely that some of the salt adducts bind to the viral DNA inside the filled capsids, making them more difficult to remove by dialysis. Individual polyhedral particles of the expected size (~93 nm) were clearly discernible on the nanomechanical resonators observed under SEM (fig. S26).

Notably, most capsids not only maintained their integrity during nebulization and transfer across the vacuum system, but they also survived landing on the nanoresonator surface at nearly supersonic speeds. The SEM pictures also show a finite contact area (i.e., not point contact) between capsids and silicon. Capsid stiffness could then induce a downward shift in the measured central mass and a broadening of the peak (16). Finite-element simulations have shown that not only is this effect negligible compared to both these quantities, but it is also below our average mass resolution (fig. S23). We attribute this to the use of an in-plane mode, to the moderate aspect ratio of our resonators and to the small displacement and strain induced by our high-frequency devices. Also, the measured capsid mass was very close to (even slightly greater than) the known molecular one. We detected 648 capsids within the mass range of Fig. 4D with a total sample of 286 fmol (one capsid detected per 2.6 × 108 in solution) over the course of 655 min with an event rate ranging from 0.8 to 1.15 capsid·min−1. This fourfold improvement in detection efficiency compared to that of empty capsids was likely the result of improved focusing of these higher-mass particles. Notably, Fig. 4, E and F, show that the average mass resolution per detected filled capsid was the same as that obtained with empty capsids (~0.5 MDa), with capsids evenly spread over the beam area (Fig. 4, E and F, insets), and confirmed that nanomechanical MS had a higher resolution (MMFWHM) at higher masses. From the mass resolutions obtained for each single capsid in the main peak, we determined a 0.95-MDa FWHM from the instrument alone (i.e., in the absence of sample heterogeneity), which corresponds to an instrument resolution of 114 at 106 MDa.

The MS architecture can analyze ionized or neutral species and overcomes many of the limitations associated with earlier nanomechanical resonator–based systems, including inefficient detection. Simple nebulization techniques like SAWN can be used, leading to efficient uptake as ionization yield or Taylor cone expansion do not come into play. Moreover, these techniques may be less prone to dissociating noncovalently bound supramolecular analytes. Rather than being counteracted by electromagnetic fields, the inertia of massive particles is exploited for efficient guiding and focusing by using an aerodynamic lens. Nanomechanical resonators directly measure the inertial mass of individual analytes, avoiding separate measurements of m/z (mass-to-charge ratio) and charge. Experiments were performed with concentrations and sample quantities that were close to those typically used in MS experiments, and the total duration of experiments was only a few hours.

Supplementary Materials

Materials and Methods

Figs. S1 to S26

Tables S1 to S8

References (3036)

References and Notes

  1. See supplementary materials and methods.
Acknowledgments: We thank V. Brun, E. Boeri Erba, and C. Breyton for support and fruitful discussions. We also thank M. Aumont-Nicaise for assistance with differential scanning calorimetry experiments and V. Zhong for help with electrospray ionization. This work has benefited from the facilities and expertise of the Macromolecular Interaction Platform of I2BC (UMR 9198). We also thank Malvern for the use of their facilities in Vénissieux, France. Funding: We acknowledge support from the European Union through the ERC Enlightened project (616251), from the LETI Carnot Institute NEMS-MS project, from the DGA Astrid NEMS-MS project, and from incoming CEA fellowships (M.S., S.D.-M.) from the CEA-Enhanced Eurotalents program, cofunded by FP7 Marie-Skłodowska-Curie COFUND program (Grant Agreement 600382). Author contributions: S.D.-M., S.F., M.D., A.-K.S., M.A.H., and T.A. performed the MS experiments. E.V. and P.B. prepared the T5 capsid samples and performed their characterization experiments. M.S., M.D., and G.J. worked on the operation set point and control of the nanomechanical arrays. M.G. fabricated the nanomechanical arrays. C.M. and S.H. supervised the research. All authors participated in the writing of the manuscript. Competing interests: C.M. and S.H. are co-inventors of the patents US9506852B2, EP2779209A1, and JP6352004B2. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper or the supplementary materials.

Correction (28 November 2018): A new reference 8 has been added, and subsequent references in the main text and supplementary materials have been renumbered accordingly.

View Abstract

Stay Connected to Science

Navigate This Article