Measurements of the self-assembly kinetics of individual viral capsids around their RNA genome

Abstract

Self-assembly is widely used by biological systems to build functional nanostructures, such as the protein capsids of RNA viruses. But because assembly is a collective phenomenon involving many weakly interacting subunits and a broad range of timescales, measurements of the assembly pathways have been elusive. We use interferometric scattering microscopy to measure the assembly kinetics of individual MS2 bacteriophage capsids around MS2 RNA. By recording how many coat proteins bind to each of many individual RNA strands, we find that assembly proceeds by nucleation followed by monotonic growth. Our measurements reveal the assembly pathways in quantitative detail and also show their failure modes. We use these results to critically examine models of the assembly process.

Keywords: RNA virus; kinetics; nucleation and growth; self-assembly; single particle.

Fig. 1.

Overview of the measurement. (A) A structural model of the MS2 capsid (PDB ID: 2ms2) shows its small size and $T=3$ structure. The 2 coat-protein dimer configurations are shown in gray and purple. (B) We inject a solution of unassembled dimers over a coverslip on which MS2 RNA strands are tethered by DNA linkages. As dimers bind to the RNA, the resulting particles scatter light. The particles appear as dark, diffraction-limited spots because of destructive interference between the scattered light and a reference beam. (C) We monitor many such spots in parallel. Shown is a typical image of the field of view, taken 126 s after adding 2-$\mathrm{\mu }$M dimers and representing an average of 1,000 frames taken at 1,000 frames per second. (D) The intensity of each spot is proportional to the number of bound proteins within each particle and changes in the intensity as a function of time reveal the assembly kinetics of each particle. The darker the spot is, the larger its intensity. (D, Top) Time series of images for the boxed spot in C. (D, Bottom) Intensity trace for the same spot using a 1,000-frame average. We discuss the relationship between intensity and number of bound proteins, as well as how we calculate the spot intensity, in Materials and Methods and SI Appendix.

Fig. 2.

Assembly of 2-$\mathrm{\mu }$M coat-protein dimers around surface-tethered RNA strands. (A) Intensity traces for 12 randomly chosen particles from one experiment. x-axis ticks show the start times and y-axis ticks the final intensities. Gray bar indicates the intensity range corresponding to wild-type capsids. Arrows show 2 traces corresponding to overgrown particles. (B) Negatively stained TEM image of particles assembled around RNA strands tethered to a gold nanoparticle (dark region at center). We use a nanoparticle as the substrate because TEM cannot image through a coverslip. (C) The cumulative distribution of start times of all of the traces in the experiment is well fitted by an exponential with delay time $t_0$ of 92 s and a characteristic time $\tau$ of 84 s (see SI Appendix for fit results from repeated experiments). Uncertainties in the start times are smaller than the diameter of the circles.

Fig. 3.

Assembly kinetics at different protein concentrations. (A) Intensity traces for 10 randomly chosen particles at 1.5-$\mathrm{\mu }$M and 4-$\mathrm{\mu }$M coat-protein dimers. (B) Cumulative distributions of the start times show that the rate of nucleation increases with protein concentration. The data are fitted by an exponential with a characteristic nucleation time $\tau$, as described above. The length of each horizontal bar represents the uncertainty in each time measurement. (C) Cumulative distributions of the final intensities show that the fraction of overgrown particles increases with protein concentration. The length of each horizontal bar is the SD calculated from the last 50 s of each trace. (D) TEM images of particles assembled around untethered RNA. At 1.5 $\mathrm{\mu }$M protein (Left), most particles appear to be capsids. At 4 $\mathrm{\mu }$M protein (Right), many particles are clusters of partial capsids.

Fig. 4.

Relative timescales of nucleation and growth and the inferred assembly pathways. (A) Measured nucleation times (${\tau }_{\text{nuc}}$) and median growth times (${\tau }_{\text{grow}}$) at different protein concentrations. Error bars represent the SD from 3 replicate experiments. (B) Cartoon of the inferred assembly pathways. At low protein concentration, ${\tau }_{\text{nuc}}$ is large compared to ${\tau }_{\text{grow}}$, and a nucleus of coat proteins forms on the RNA and then grows into a proper capsid. At moderate protein concentration, ${\tau }_{\text{nuc}}$ is comparable to ${\tau }_{\text{grow}}$, and a second nucleus can form on the RNA before the first one has finished growing, leading to an overgrown structure consisting of a nearly complete capsid attached to a partial capsid. At high protein concentration, ${\tau }_{\text{nuc}}$ is smaller than ${\tau }_{\text{grow}}$, and multiple nuclei can form and grow, leading to an overgrown structure consisting of many partial capsids. Example TEM images of the endpoints of each pathway are shown at Right.