Mass photometry enables label-free tracking and mass measurement of single proteins on lipid bilayers

Abstract

The quantification of membrane-associated biomolecular interactions is crucial to our understanding of various cellular processes. State-of-the-art single-molecule approaches rely largely on the addition of fluorescent labels, which complicates the quantification of the involved stoichiometries and dynamics because of low temporal resolution and the inherent limitations associated with labeling efficiency, photoblinking and photobleaching. Here, we demonstrate dynamic mass photometry, a method for label-free imaging, tracking and mass measurement of individual membrane-associated proteins diffusing on supported lipid bilayers. Application of this method to the membrane remodeling GTPase, dynamin-1, reveals heterogeneous mixtures of dimer-based oligomers, oligomer-dependent mobilities, membrane affinities and (dis)association of individual complexes. These capabilities, together with assay-based advances for studying integral membrane proteins, will enable the elucidation of biomolecular mechanisms in and on lipid bilayers.

Fig. 1. Principle and performance of dynamic mass photometry.

a, Schematic diagram of dynamic mass photometry of protein complexes diffusing on an SLB. The images were acquired at 331 Hz and processed with a sliding median filter, which showed individual protein complexes on the bilayer as diffraction-limited spots. This procedure consistently gave similar results (n > 30). b, Histogram of mean trajectory contrasts detected in a dynamic mass photometry movie (n = 1 movie, 4 min) of WT diffusing on an SLB (considering only trajectories of at least 151 ms in length; n = 425 trajectories). c, Contrast–mass calibration curve of the dynamic mass photometry measurement shown in b (n = 1 dynamic mass photometry movie, 4 min) yielding a contrast to mass ratio of 4.40 % MDa⁻¹. Error bars represent the mean contrast ± s.e.m. of each oligomeric species (n_dimer = 34, n_tetramer = 85, n_hexamer = 184, n_octamer = 23 trajectories). d, 2D localization error of our PSF-fitting procedure of WT dimers, tetramers, hexamers and octamers plotted as a function of effective exposure time. Data are given as the mean localization errors in 2D ± the combined s.d. of the mean errors in x and y of particle trajectories detected during the dynamic mass photometry movie in b (n = 1 movie, 4 min), processed with different amounts of frame averaging (n_dimer = 34, 51, 60, 52, 73; n_tetramer = 82, 102, 98, 97, 94; n_hexamer = 177, 229, 224, 208, 173; n_octamer = 22, 29, 37, 38, 33 trajectories for total exposure times of 3.0, 6.0, 9.1, 12.1 and 15.1 ms, respectively). e, Mass trace and histogram of a WT decamer trajectory (n = 6,061 frames). f, Corresponding particle trajectory. g, Corresponding cumulative probability of particle displacements during 1 frame (t = 3 ms) and the fits to a two-component model (equation 4). Scale bars, 500 nm. Source data

Fig. 2. Oligomeric properties and dynamics of dynamin diffusing on an SLB.

a, Major diffusion components versus mean trajectory mass for a dynamic mass photometry movie (n = 1 movie, 4 min) of WT (20 nM); n = 333 trajectories. b, Major diffusion components of each oligomeric species of ΔPRD (10–20 nM) determined from n = 4 replicate dynamic mass photometry measurements (4–5 min each with a total of n_tetramer = 213, n_hexamer = 937, n_octamer = 330 and n_decamer = 83 trajectories) versus the inverse of the number of oligomeric subunits, and a corresponding weighted linear fit (blue dashed line). Error bars represent the mean ± s.d. c, Histogram of SLB residence times of ΔPRD hexamer trajectories from one of the dynamic mass photometry movies with a fit to a 1-component exponential distribution (appropriately scaled here for display) yielding a dissociation rate constant of 19.4 s⁻¹. n_hexamer = 1,123 trajectories. d, Dissociation rate constants of each oligomeric species of ΔPRD versus the inverse of the number of oligomeric subunits from n = 4 independent replicate dynamic mass photometry measurements (10–20 nM, 4–5 min each with a total of n_tetramer = 574, n_hexamer = 5,291, n_octamer = 925 and n_decamer = 69 trajectories) and a corresponding weighted linear fit (blue dashed line). Error bars represent the mean ± s.d. e,f, Examples of a dissociation event (e) and an association event (f). These events were extremely rare (<1 in 1,000 trajectories). g, Effect of GTP addition on the oligomeric distribution of 10–20 nM WT (n = 5 independent replicates of 1 min dynamic mass photometry movies before and after GTP addition). Data are given as mean ± s.d. ^†WT dimer partially overlapped with background noise and could not always be reliably identified. Scale bars, 500 nm. Source data

Extended Data Fig. 1. Background subtraction in standard mass photometry (MP) vs dynamic MP.

Zoom on three consecutive raw images from a dynamic MP movie of WT dynamin in contact with an SLB containing a particle diffusing upwards through the image (top row), which is masked by the large signal from the glass surface roughness. In standard MP (left), images are divided by preceding images to remove the large static signal due to surface roughness. This background subtraction relies on stationary binding of small particles to the surface to visualise them. When particles bind to and diffuse on the surface, the background subtraction used in standard MP results in a signal that is a convolution of the particle’s position at t_i-1 and t_i, (blue and red, respectively), which is challenging to reliably detect and quantify. A sliding median filter (right), that is subtracting each image’s temporal median background obtained from a defined window of images around the image of interest, reveals signals of only the particles in the image of interest (t_i, red). For further explanation see ‘Data processing’ in the Methods section. Scale bar = 500 nm. The effect illustrated in this figure was reproducible in all measurements shown in this study (n > 30).

Extended Data Fig. 2. Comparison between WT in solution and on the SLB.

(a) Mean contrast vs mass calibrations obtained from n = 3 standard MP measurements of WT (40 nM, n_monomer = 832, n_dimer = 2637, n_tetramer = 2404 and n_hexamer = 263 particles) and (b) from n = 6 independent 2 min dynamic MP movies (using only trajectories that lasted at least 20 frames) of 20 nM WT on an SLB (n_dimer = 227, n_tetramer = 1079, n_hexamer = 2482 and n_octamer = 311 trajectories) acquired on the same day. The error bars in (a) and (b) represent the mean ± s.d. of the contrast of each oligomeric species from the repeat measurements. In some cases the standard deviation was less than 1% causing the contrast error bars to overlap. (c) Oligomeric distribution of WT (100 nM) in HKS-100 buffer measured by standard MP (n = 4 combined measurements with a total of n = 16794 particles). The peaks represent WT monomer (0.5%), dimer (1.0%), tetramer (2.0%) and hexamer (3.0%). (d) Oligomeric distribution of 10 nM WT in HKS-100 buffer diffusing on an SLB obtained from n = 2 combined sets of 3 min dynamic MP movies considering only trajectories that lasted at least 50 frames (n = 1187 trajectories). The contrast measured in dynamic MP movies was consistently ~8% lower than that measured in standard MP. This effect is likely a result of particle motion during image acquisition, which results in motion blurring of the PSF (Supplementary Fig. 3 and Extended Data Fig. 3). This effect increased as we increased frame averaging in dynamic MP movies (Extended Data Fig. 4). The standard MP measurements were acquired at 331 Hz and then processed at a final integration time ~24 ms (effective frame rate ~41 Hz), which enabled the detection of WT monomer. *Peak due to background noise. Source data

Extended Data Fig. 3. Contrast decrease in dynamic MP vs standard MP.

Drop in contrast of different WT oligomers (dimer, 0.2 MDa; tetramer, 0.4 MDa; hexamer, 0.6 MDa; octamer, 0.8 MDa) when comparing dynamic MP measurements to standard MP measurements (blue circles) and drop in contrast observed in simulated dynamic MP movies (Supplementary Fig. 3). For experimental data the contrast drop represents the mean reduction in the average contrast of each oligomeric species measured in n = 6 dynamic MP measurements (2 min each, considering only trajectories that lasted at least 20 frames resulting in a total of n_dimer = 227, n_tetramer = 1079, n_hexamer = 2482 and n_octamer = 311 trajectories) compared to the average contrast measured in n = 3 standard MP measurements (Supplementary Table 1, total of n_dimer = 2637, n_tetramer = 2404 and n_hexamer = 263 particles). For simulated data, the contrast drop represents the decrease in contrast of each species detected in n = 3 processed simulated movies compared to the contrast value that was used to simulate the point spread functions onto the raw images (1.00-4.00% for dimer-octamer, Supplementary Table 2, n_dimer = 1780, n_tetramer = 2302, n_hexamer = 2618 and n_octamer = 2745 trajectories). Data is presented as mean values ± s.d.. *In dynamic MP movies the contrast of dimer particles partially overlaps with that of background signal, which most likely causes underestimation of the dimer contrast and an exaggerated decrease in contrast compared to standard MP measurements. ^†WT octamer was not detected in standard MP measurement and the contrast was extrapolated using the contrast vs mass calibration. Source data

Extended Data Fig. 4. Effect of frame averaging on particle contrast.

(a) Mean contrast of WT dimer (red circles), tetramer (orange squares), hexamer (purple crosses) and octamer (blue diamonds) trajectories, (b) mean contrast vs mass calibration slope obtained from the dynamic MP movie in (a) vs single frame length after averaging and (c) contrast precision of our PSF-fitting procedure for each oligomer (same symbols as in (a)) all plotted vs total exposure time of 1 frame after averaging. These trends are most likely a result of particle motion during image acquisition, which becomes more pronounced as more raw images are averaged together and the frame length increases. The plots were obtained from the same movie of WT used in Figs. 1b-d and 2a (n = 1 movie (4 min) of 20 nM WT) with additional frame averaging of 1, 2, 3, 4 and 5 frames, which corresponds to frame lengths of 3.02, 6.04, 9.05, 12.07 and 15.09 ms or frame rates of 331, 166, 110, 83 and 66 Hz, respectively (see Supplementary Information). The data in (a) and (c) are presented as mean values ± s.e.m. for each oligomeric species. The data in (b) is presented as mean values ± s.d.. For these plots n_dimer = 34, 51, 60, 52, 73, n_tetramer = 82, 102, 98, 97, 94; n_hexamer = 177, 229, 224, 208, 173; n_octamer = 22, 29, 37, 38, 33 trajectories for total exposure times of 3.0, 6.0, 9.1, 12.1, 15.1 ms, respectively. Source data

Extended Data Fig. 5. Diffusion coefficient vs inverse of number of subunits of WT oligomers.

Mean diffusion coefficients of each oligomeric species of WT from n = 7 independent repeat measurements (n_tetramer = 498, n_hexamer = 1326, n_octamer = 156; data shown in Supplementary Fig. 6-7) vs the inverse of the number of subunits of each oligomeric species and a corresponding weighted linear fit (blue dashed line). Error bars are presented as mean values ± s.d.. Source data

Extended Data Fig. 6. Effect of lag time on calculated diffusion coefficients.

(a) Average median diffusion coefficient and (b) corresponding interquartile range vs chosen lag time (t) for each oligomeric species from n = 4 independent ΔPRD measurements (n_tetramer= 213, n_tetramer = 937, n_hexamer = 330, n_decamer =83; trajectories data shown in Supplementary Fig. 8-9). As the distribution of diffusion coefficients broadened significantly as the lag time increased, the diffusion coefficient of each oligomer was determined by taking the median of the distribution instead of Gaussian fitting. Each data point represents the mean diffusion coefficient from the median values determined from four repeats of ΔPRD measurements and the error bars are presented as mean values ± s.d.. Source data

Extended Data Fig. 7. Dissociation constant vs inverse of number of subunits of WT oligomers.

Mean dissociation constant from the SLB of each oligomeric species of WT determined from n = 7 independent repeat measurements (n_tetramer = 2263, n_hexamer = 3264, n_octamer = 203 trajectories; data shown in Supplementary Fig. 11) vs the inverse of the number of subunits of each oligomeric species and a corresponding weighted linear fit (blue dashed line). Error bars are presented as mean values ± s.d.. Source data

Extended Data Fig. 8. Effect of GTP and GMPPNP on the oligomeric distribution of WT.

Average oligomeric distribution of 20 nM WT in contact with an SLB in its apo-state, with 1 mM GTP, with 1 mM of GMPNP (non-hydrolysable GTP analogue) and with 1 mM MgCl₂ instead of 2 mM MgCl₂. The 1 mM MgCl₂ measurement was included as a control in case the presence of GTP/GMPNP results in a reduction in free Mg^2 +, which could potentially affect the membrane affinity of WT dynamin oligomers. Each bar plot was generated by taking the mean oligomeric counts from n = 3 repeat measurements (1 min dynamic MP movie each). Error bars are presented as mean values ± s.d.. Source data

Extended Data Fig. 9. Analysis of background noise in dynamic MP movies.

(a) Mean baseline noise of the movie of HKS-100 buffer in contact with an SLB (black circles) shown in Supplementary Fig. 1 vs window size of the sliding median filter used to process the movie and the corresponding theoretical shot noise limit (grey squares). The error bars represent the mean baseline noise ± standard deviation across n = 4599 recorded frames. The inverse trend in baseline noise of the buffer movie compared to the theoretical shot noise is a result of particle-like background features in dynamic MP movies, which are subtracted out by the sliding median filter at short window sizes (for example 20-100 ms) but not at longer window sizes (> 200 ms). (b-c) Mass histogram of n = 182 and n = 30 trajectories, respectively, detected in a single SLB buffer movie processed with the same settings described in the Methods section, filtered for trajectories with a minimum length of 10 and 50 frames, respectively. The detected background features have contrasts corresponding to ~150 kDa, which prevented the detection of ΔPRD dimer and made reliable detection of WT dimer challenging. Source data

Extended Data Fig. 10. Effect of the window size of the sliding median filter on particle contrast.

(a) Mean contrast of WT dimer (red circles), tetramer (orange squares), hexamer (purple crosses) and octamer (blue diamonds) trajectories, (b) mean contrast vs mass calibration slope obtained from the dynamic MP movie in (a) vs single frame length after averaging and (c) contrast precision of our PSF-fitting procedure for each oligomer (same symbols as in (a)) all plotted vs total exposure time of 1 frame after averaging. The plots were obtained from the same movie of WT used in Figs. 1b-d and 2a (n = 1 movie (4 min) of 20 nM WT) with additional frame averaging of 1, 2, 3, 4 and 5 frames, which corresponds to frame lengths of 3.02, 6.04, 9.05, 12.07 and 15.09 ms or frame rates of 331, 166, 110, 83 and 66 Hz, respectively (see Supplementary Information). The data in (a) and (c) are presented as mean values ± s.e.m. for each oligomeric species. The data in (b) is presented as mean values ± s.d.. For these plots n_dimer = 34, 51, 60, 52, 73, N_tetramer = 82, 102, 98, 97, 94; n_hexamer = 177, 229, 224, 208, 173; n_octamer = 22, 29, 37, 38, 33 trajectories for total exposure times of 3.0, 6.0, 9.1, 12.1, 15.1 ms, respectively. Source data