Single-particle studies of the effects of RNA–protein interactions on the self-assembly of RNA virus particles

Understanding the pathways by which simple RNA viruses self-assemble from their coat proteins and RNA is of practical and fundamental interest. Although RNA–protein interactions are thought to play a critical role in the assembly, our understanding of their effects is limited because the assembly process is difficult to observe directly. We address this problem by using interferometric scattering microscopy, a sensitive optical technique with high dynamic range, to follow the in vitro assembly kinetics of more than 500 individual particles of brome mosaic virus (BMV)—for which RNA–protein interactions can be controlled by varying the ionic strength of the buffer. We find that when RNA–protein interactions are weak, BMV assembles by a nucleation-and-growth pathway in which a small cluster of RNA-bound proteins must exceed a critical size before additional proteins can bind. As the strength of RNA–protein interactions increases, the nucleation time becomes shorter and more narrowly distributed, but the time to grow a capsid after nucleation is largely unaffected. These results suggest that the nucleation rate is controlled by RNA–protein interactions, while the growth process is driven less by RNA–protein interactions and more by protein–protein interactions and intraprotein forces. The nucleated pathway observed with the plant virus BMV is strikingly similar to that previously observed with bacteriophage MS2, a phylogenetically distinct virus with a different host kingdom. These results raise the possibility that nucleated assembly pathways might be common to other RNA viruses.

Since the 1950s, the question of how RNA viruses self-assemble has inspired theoretical and experimental work in many fields of basic and applied science (1–5). Simple RNA viruses, which consist of a single-stranded RNA genome inside an ordered capsid made up of multiple copies of a single protein (Fig. 1A), have served as model systems for studying the physical principles of structural virology involving virus particles of all shapes and sizes (1, 2, 6, 7). However, the mechanisms and pathways by which these viruses assemble into the correct structure, while avoiding the many possible malformed structures, are not yet understood.Open in a separate windowFig. 1.Overview of the system and the measurement. (A) A 3-dimensional model of BMV reconstructed from cryoelectron microscopy data (51) shows the protein capsid (gray) surrounding the RNA (gold). The model reveals most of the icosahedral capsid but only a small portion of the RNA, the rest of which adopts a disordered arrangement within the capsid. (B) A cartoon of the experiment shows viral coat proteins assembling around RNA strands that are tethered by DNA linkages to the surface of a functionalized glass coverslip. (C) The assembling proteins are imaged at 1,000 Hz for 600 s using iSCAT microscopy. Each dark spot that appears in the images corresponds to proteins bound to an individual RNA strand. The darkness, or intensity, of each spot is proportional to the number of proteins bound to that RNA. The displayed images are the average of 1,000 consecutive frames. (D) Traces of the intensity as a function of time (1,000-frame moving average) reveal the assembly kinetics for each particle. Experimental conditions are 0.135 μmol/L protein and 250 mmol/L NaCl. The initial spike in intensity present in many of the traces is associated with vibrations introduced into the system as the coat protein is injected. The thick, black trace corresponds to the boxed particle in (C). We compare the final intensities of the traces to the estimated intensity range of full capsids, which is shown as a vertical bar to the right of the traces.Although many different RNA viruses self-assemble (8–10), our interest is in comparing the assembly of virus-like particles from two well-studied virus families: Bromoviridae, a family of plant-infecting viruses that includes brome mosaic virus (BMV) and cowpea chlorotic mottle virus (CCMV), and Fiersviridae (previously Leviviridae), a family of bacteria-infecting viruses that includes MS2 and Qβ. These families are as distinct phylogenetically as any two RNA virus families can be, having a last common ancestor that is thought to predate the emergence of eukaryotic cells (11). Accordingly, there are many well-established physical and biological differences among viruses in these families and virus-like particles derived from them. Yet the four most studied members—BMV, CCMV, MS2, and Qβ—do have some structural commonalities: They have icosahedral capsids with a triangulation number (T) of 3 (2), they have no lipid envelope, and each capsid surrounds approximately 3,000 to 4,000 nucleotides of single-stranded RNA.The assembly of such structures is a nontrivial process. Identical coat proteins must adopt nonequivalent positions to make a T = 3 capsid, with some arranging in pentagonal configurations and others in hexagonal configurations (2, 7, 12). Furthermore, these configurations must form in the correct proportions and positions for the capsid to close. Despite these challenges, assembly of virus-like particles of CCMV (13–15), BMV (14, 15), and MS2 (16) occurs in high yield even in vitro and in the absence of host-cell factors. The ability of viruses to avoid the many possible metastable states en route to complete assembly has been likened to the Levinthal paradox of protein folding (17, 18).But unlike proteins, RNA viruses have a template for assembly: their own RNA. Current theoretical models of RNA virus self-assembly posit markedly different roles for the RNA, depending on the relative strengths of RNA–protein and protein–protein interactions, sequence-dependent RNA–protein interactions, RNA-mediated protein–protein interactions, and several other factors (19). Although specific interactions between RNA substructures and coat proteins have been hypothesized to help the virus avoid malformed configurations (18), viruses from different families differ greatly in their RNA structures and RNA–protein interactions. It is therefore unclear whether there are common features of the assembly process for different T = 3 viruses or if there are distinct assembly pathways that depend on RNA–protein interactions.Recent measurements of assembly kinetics suggest the latter: that the assembly of viruses from different families follows different pathways. Fluorescence correlation spectroscopy experiments (20, 21) of the kinetics of binding of MS2 coat protein and RNA indicate that assembly starts with a small cluster of RNA-bound proteins that trigger a change in the hydrodynamic radius of the RNA. In contrast, cryoelectron microscopy (22) and small-angle X-ray scattering (23) experiments of the assembly of the CCMV coat protein and RNA show that disordered RNA–protein complexes formed at neutral pH anneal over several thousand seconds into well-formed capsids when the pH drops below 6.But because these experiments involve different assembly conditions and different measurement techniques, their outcomes might not reflect fundamental differences in the assembly pathways of these viruses but rather technical differences in the methods and protocols used to study them. Furthermore, most of the techniques that have been used do not measure the assembly process directly at the scale of individual particles because—one way or the other—they involve averaging over many particles. Such averaging can obscure the mechanisms and pathways that underpin stochastic assembly processes like viral assembly, in which each individual particle can follow its own unique sequence of intermediate states. Thus, it remains an open question whether a common assembly pathway might exist between these viruses.We recently demonstrated that interferometric scattering (iSCAT) microscopy (24) can resolve the assembly kinetics of individual virus-like particles (25), providing a method to directly measure and compare the assembly pathways of different viruses. To perform the iSCAT experiment, we first tether viral RNA molecules to the surface of a functionalized glass coverslip under the desired buffer conditions (26) (Fig. 1B). Next, we begin collecting iSCAT images of the RNA-decorated coverslip as we inject viral coat proteins at the desired concentration and in the appropriate buffer. As the proteins bind to the surface-tethered RNA, dark spots appear in the iSCAT images (Fig. 1C). Subtracting the intensity associated with the RNA then yields images in which the intensity of each dark spot is proportional to the number of proteins that have accrued onto each individual RNA. Previous measurements by Young and coworkers (27) show that the iSCAT intensities of protein assemblies are, in general, linearly proportional to the total mass of the assemblies. Accordingly, in our experiments, plotting the trace of the intensity of a spot as a function of time reveals the assembly kinetics for that particle, and plotting the collection of traces reveals the assembly kinetics for the ensemble of particles (Fig. 1D).In our previous work (25) we examined the assembly of bacteriophage MS2. We found that well-formed capsids could assemble around surface-tethered RNA strands and that the assembly kinetics were consistent with a nucleation-and-growth pathway in which a small cluster of RNA-bound proteins must exceed a critical size before the binding of additional proteins becomes favorable. Despite an apparently small critical nucleus size of only a few coat–protein dimers, we found that MS2 capsids grow monotonically to full or nearly full size with high yield.Although this previous study highlighted the importance of the RNA in the assembly process, the strong and specific RNA–protein interactions in MS2 (28–30), which are thought to occur at a dozen or so positions on the RNA molecule (31, 32), make it difficult to systematically address the central question of how the RNA affects the pathway. By contrast, the RNA in BMV interacts with the coat proteins primarily through nonspecific electrostatic interactions (33), with the possible exception of a single, specific RNA–protein interaction occurring at the 3′-end of the RNA (34). As a result, the strength of RNA–protein interactions in BMV can be largely tuned by changing the ionic strength of the buffer solution (22, 35, 36). BMV therefore offers not only an interesting comparison to MS2—it is phylogenetically distinct but structurally similar—but also the means to understand the role of RNA–protein interactions.In the current study, we infer the assembly pathways of BMV from iSCAT measurements under different RNA–protein interaction strengths, allowing us to critically assess of competing models of the assembly process. We follow the assembly trajectories of more than 500 individual virus particles under different assembly conditions, and we correlate the results with the absence and presence of ordered capsids as detected with negative-stain transmission electron microscopy (TEM). We find that BMV can assemble by a nucleation-and-growth process that is qualitatively similar to that of MS2. We show that the strength of RNA–protein interactions strongly affects the nucleation time but only weakly affects the growth time, suggesting that RNA plays a central role in nucleating the viral capsid but a relatively minor role in its growth kinetics. We discuss these observations in the context of recent models and hypotheses of RNA virus self-assembly.