Visualizing a viral genome with contrast variation small angle X-ray scattering

Despite the threat to human health posed by some single-stranded RNA viruses, little is understood about their assembly. The goal of this work is to introduce a new tool for watching an RNA genome direct its own packaging and encapsidation by proteins. Contrast variation small-angle X-ray scattering (CV-SAXS) is a powerful tool with the potential to monitor the changing structure of a viral RNA through this assembly process. The proteins, though present, do not contribute to the measured signal. As a first step in assessing the feasibility of viral genome studies, the structure of encapsidated MS2 RNA was exclusively detected with CV-SAXS and compared with a structure derived from asymmetric cryo-EM reconstructions. Additional comparisons with free RNA highlight the significant structural rearrangements induced by capsid proteins and invite the application of time-resolved CV-SAXS to reveal interactions that result in efficient viral assembly. – exam-ples

Despite the threat to human health posed by some singlestranded RNA viruses, little is understood about their assembly. The goal of this work is to introduce a new tool for watching an RNA genome direct its own packaging and encapsidation by proteins. Contrast variation small-angle X-ray scattering (CV-SAXS) is a powerful tool with the potential to monitor the changing structure of a viral RNA through this assembly process. The proteins, though present, do not contribute to the measured signal. As a first step in assessing the feasibility of viral genome studies, the structure of encapsidated MS2 RNA was exclusively detected with CV-SAXS and compared with a structure derived from asymmetric cryo-EM reconstructions. Additional comparisons with free RNA highlight the significant structural rearrangements induced by capsid proteins and invite the application of time-resolved CV-SAXS to reveal interactions that result in efficient viral assembly.
Viruses exploit many strategies to encapsidate the genome that codes for its container. The exact packaging process depends on the nature and form of the genetic material as well as the container or capsid. For some dsDNA viruses, there is clear structural separation between container and message; the genome is pumped into the preformed protein capsid by ATPactivated motor proteins (1). The packaged genome, under high pressure, uniformly fills the capsid (2). In contrast, viruses with single-stranded RNA genomes appear to exploit RNAprotein interactions to facilitate encapsidation. The few examples of spatially resolved, encapsidated ssRNA genomes suggest that the genetic material is not uniformly distributed, often asymmetric and with higher density at the protein interface (3-8). The distinct packaging strategies are at least partially because of the very different biophysics of DNA and RNA (9,10).
Because ssRNA viruses are prevalent in human disease, there is significant interest in revealing their assembly mechanisms and packaged structures. Fortunately, a fair amount is known about RNA structure from work on other biological RNAs. These polymers contain self-complementary regions and fold on themselves to form series of short, base-paired duplexes joined by a wide range of non-base-paired regions, including junctions, bulges, or loops. Many of these latter motifs serve as protein-binding sites. In viral genomic RNA, recent work suggests that some RNA motifs that bind proteins, so-called packaging signals, are essential for viral assembly (11)(12)(13)(14)(15)(16)(17)(18). Their presence may explain why ssRNA viruses selectively pack their own genome relative to other cellular RNAs. Thus, the prevailing picture of how ssRNA genomes are encapsidated by proteins in viruses may be best viewed as a cooperative, specific protein-RNA folding process, as opposed to a sequential assembly of protein capsid followed by genome insertion (19). The self-assembly of ssRNA viruses can therefore be viewed as a process that exploits the delicate balance between RNA-RNA interactions, RNA-protein interactions and, of course, protein-protein interactions (20,21).
From this vantage point, the packaging of some ssRNA genomes can be recast as an RNA folding problem. In the past 10-20 years, significant progress has been made toward defining the rules for RNA folding, largely in response to the growing appreciation of the role of RNA conformational dynamics in biology (22). Many experiments have provided details about RNA folding as it interacts with counterions (23)(24)(25)(26), small ligands (27)(28)(29)(30), and even proteins (31)(32)(33). Time-resolved experiments have been particularly useful at revealing the dynamics and transiently populated states of RNA. When compared with simulations or models, these studies elucidate the principles that direct folding of small to moderately sized (,1 kb) RNAs (24).
Of the many biophysical tools applied to study the dynamic restructuring of RNAs, solution small-angle X-ray scattering (or SAXS) is particularly useful for revealing large-scale, timedependent conformational changes. SAXS has been recently applied to monitor the changing structures that accompany capsid self-assembly, providing new information about the underlying mechanisms for the simplest systems (34)(35)(36)(37). These works are informative but confounded by SAXS' sensitivity to all components present: both capsid components and nucleic acid. In fact, most of the information extracted was related to the number, and/or conformation of protein subunits. Thus, past kinetic studies of virus assembly by SAXS primarily reported the assembly of the capsid. Fortunately, contrast variation (CV-) SAXS can be used to extract the structures of individual components of complexes, when these components have different electron densities. The nucleic acid genome, for example, can be selectively detected even when encapsidated. The high X-ray flux available at synchrotron sources has already enabled time-resolved CV-SAXS studies of nucleic acid structuring by proteins in other systems (38). Applying this technique to study viral assembly could reveal valuable information about the role of RNA in the assembly process.
The goal of this work is to assess the feasibility of timeresolved CV-SAXS studies of virus assembly. Here we report static CV-SAXS studies of bacteriophage MS2, a model system. Bacteriophage MS2 is an Escherichia coli phage from the leviviridae family with an ssRNA genome of 3569 nucleotides, and a T = 3 icosahedral capsid. This container is composed of 180 copies of the capsid protein and a maturation protein. MS2 is commonly used as model system to study 1ssRNA viral assembly. As such, it has been widely studied both biochemically (39)(40)(41)(42)(43)(44) and structurally, using SAXS (45)(46)(47)(48), as well as electron microscopy (4)(5)(6)(7)(8). Contrast variation studies of bacteriophage MS2 have been accomplished using small-angle neutron scattering (46) which more readily yields information about the protein conformation within the MS2 capsid. In these studies, structural information about the encapsidated RNA was inferred, not directly measured. Furthermore, the ability to perform these measurements with X-rays exploits the higher signals available from synchrotron sources relative to neutron sources. Recent asymmetric cryo-EM reconstructions have produced electron density models of the encapsidated RNA for comparison with our findings (5-7). The MS2 system is an excellent candidate for future time-resolved studies, as recent work suggests that, as a result of specific interactions with capsid proteins, its assembly pathways are orderly (49)(50)(51)(52). Thus, it is an ideal candidate to benchmark the performance of contrast variation SAXS in virus self-assembly experiments.

Results and Discussion
With the goal of directly measuring the encapsidated viral RNA, we performed CV-SAXS on bacteriophage MS2, as well as a recombinantly produced MS2 virus-like particle (VLP) devoid of viral RNA. Contrast variation SAXS, illustrated in Fig. 1 and described in full detail in Ref. 53, exclusively detects the nucleic acid component of a protein-RNA complex, enabling structural studies of the RNA genome contained within the fully assembled virus and potentially throughout the assembly or disassembly process.
Time-resolved CV-SAXS has been successfully implemented in studies of DNA unwrapping from protein (histone) cores in nucleosome core particles (38). However, in nucleosome core particles, the nucleic acid surrounds the proteins, whereas the opposite is true for viruses. We first address the feasibility of CV-SAXS for particles where the nucleic acid is enclosed by protein. Previous work from Zipper et al. (54) suggests that contrast variation works equally well in either geometry. They performed a limited set of contrast variation SAXS on bacteriophages fr and R17. They extrapolated the radial electron density distribution for the phages but did not thoroughly explore the contrast matched condition. The use of high intensity X-ray sources and sophisticated SAXS data analysis tools make it worthwhile to revisit this approach.

Contrast variation on VLPs
To establish the feasibility of CV-SAXS in virus-like systems, we first performed control experiments on the empty capsid, the MS2 VLP. These studies directly address two primary concerns of CV-SAXS in these systems. The first is that the capsid may be impermeable to the contrast agent, resulting in different electron densities on its inside and outside. As described in Ref. 53, we use high concentrations of sucrose to effect the dramatic changes in solvent electron density required to match the protein density. If the capsid is impermeable to sucrose, the solvent inside the capsid would have an excess negative electron density relative to the solvent outside. Because scattering depends only on density differences, which can be positive or negative, a signal would result. This effect is illustrated using a simple model that treats the capsid as a spherical shell. An analytical solution (see "Experimental procedures") is used to obtain scattering profiles where the density inside the sphere is different from the density outside. Spheres are particularly amenable to characterization using SAXS; their dimensions are encoded in the positions of (multiple) extrema in the scattering profiles. In particular, the positions of the minima reflect the size of the particle. Fig. 2 shows the predicted behavior of the two cases, a permeable versus nonpermeable capsid, as the (outer) solvent electron density is raised. Here, the expected intensity of the scattered X-rays (I on the y axis) is plotted against the momentum transfer q, defined to be q = (4p sin(u)/l), where l is the X-ray wavelength and 2u is the scattering angle. Noise was added to simulate realistic experimental conditions. In the former case, the overall signal decreases without significant change in the shape of the scattering profile, assessed through the positions of the extrema. In contrast, for the latter case of a Figure 1. Illustration of the principle of contrast variation SAXS. In CV-SAXS the electron density of the solvent is increased so that it matches that of one of the components of a multicomponent system. In this figure, electron density is represented as color. In a protein-nucleic acid complex the electron density of the solvent can be increased (schematically shown as color changed from white to red to blue) by the addition of sucrose until it matches the electron density of the protein (red). Under this matched condition any scattering signal that contains a contribution from the protein component blends into the background, e.g. disappears. Only the scattering from the denser nucleic acid is detected.
nonpermeable capsid, the positions of the maxima and minima of the scattering profile change, and the signal intensity never drops substantially. A second concern is that sucrose could create osmotic pressures that swell, disrupt, or cause other structural modifications of the capsid. These changes are harder to predict; however, any significant alterations would manifest through a change in the positions of the extrema. The absence of a significant change in the scattering profile of the VLP through the contrast series would suggest that the structure of the capsid is not altered (at the resolution of our measurements) by osmotic effects.
Experimentally determined SAXS scattering profiles of the MS2 VLP, acquired in solutions containing different amounts of sucrose, are shown in Fig. 3. As the protein contrast is reduced by increasing the sucrose concentration, the intensity of the scattering profile decreases, but no change in shape is detected: The positions of the minima remain constant despite the change in solvent electron density. Thus, the capsid appears permeable to sucrose and its structure remains roughly constant. With a further increase in sucrose concentration, the characteristic scattering features of the capsid disappear, and we conclude that it is rendered as transparent as possible. A and B, the predicted scattering profiles of an empty protein capsid (modeled as a spherical shell) at different contrast conditions in the case that it is permeable (A) or nonpermeable (B) to the contrast variation agent. The % blanked refers to the contrast match between solvent and solute. Perfect contrast matching corresponds to 100% blanked (the line would be difficult to display on a logarithmic axis; we show 99.9% as an alternative). In the inset, the color designates density as in Fig. 1. The curve colors correspond to the fractions quoted in the box. These models suggest that the permeability of the capsid can be readily evaluated from the shape changes in a contrast series. For the desired case of a permeable capsid (top plot), the signal drops dramatically as the contrast match is achieved (purple curve). An impermeable capsid (bottom plot) has a distinctly different signature.
Visualizing a viral genome with contrast variation SAXS These measurements suggest that an ;60% (w/w) sucrose solution represents the match condition. Some signal remains, likely because of impurities or the presence of a small amount of mRNA encapsidated during VLP expression. With the knowledge that the capsid can be (largely) blanked we proceed to measure the scattering from the encapsidated RNA inside the WT bacteriophage MS2.

Contrast variation on WT MS2
As a first step toward measuring the scattering from the genomic RNA encapsidated within WT MS2 phage particles, we acquired SAXS profiles of the full particle in solutions containing different sucrose concentrations. Measurements acquired at different solution contrast levels allow us to monitor both protein and nucleic acid components as their scattering strengths are varied. Fig. 4A shows SAXS profiles acquired at sucrose concentrations between 0 and 65% (w/w) in discrete steps. It is interesting to compare the shape of the profiles shown here with those of Fig. 2. In the WT MS2 the signal from the RNA becomes apparent; minima in the scattering profile shift to higher q values as the protein contrast is reduced. This trend is consistent with an increased signal from density within the capsid (smaller radius), relative to the capsid itself. The coincident decrease in the depth of the minima can be explained by a loss of spherical symmetry of the molecule (55). We note that the curves change qualitatively above 60% added sucrose. When coupled with the above measurements of the empty capsid, it appears that the "match point," where the sucrose density equals that of the protein, is around 60%. At higher concentrations, the protein signal reappears, but with lower electron density than the sucroserich solvent.
Pair distance distributions functions (P(r)) can be computed from the scattering curves and used to interpret the effects of the changing contrast on the signal from the various components of the phage. This formalism displays information from SAXS profiles in real space; as opposed to momentum transfer (q), the x axis in these plots shows real space distances in angstroms. The pair distance distribution functions shown in Fig.  4B were computed using GNOM (56) (ATSAS). All curve features remain robust against variations in selection of parameters for the P(r) computation, the q range, and maximum particle dimension, Dmax. With increasing sucrose concentration, the maximum of the P(r) curve shifts toward a smaller radius. When the sucrose concentration exceeds the match point of about 60%, a local minimum appears above 200 Å, which likely indicates that the solvent density exceeds that of the protein.
Under this condition, the signal from the capsid re-emerges, and creates a negative interference term with the RNA. In this case, the second peak above 225 Å could reflect capsid autocorrelation, and the minima above 200 Å could indicate the interference of the signals from RNA (positive contrast) and protein (negative contrast). In any case, all the above measurements are consistent with a match point near 60% sucrose for both the empty and full capsid. Under this condition, the SAXS profile from the native phage most closely represents the scattering from just the encapsidated RNA. Although the contrast match is imperfect, we can nevertheless extract low-resolution structural features about this RNA, for comparison with other measurements on the RNA, and as a proof of principle of the method.

Comparison to EM structure
To assess the power and validity of CV-SAXS applied to viruses and virus-like particles, we compare our contrastmatched data (blue curve from Fig. 4) to RNA density derived from a recent cryo-EM study of the MS2 phage by Koning et al. (7). To make this comparison, we computed the pair distance distribution from the electron density map EMD-3404 for voxels above the recommended contour level. Fig. 5 compares the pair distance distribution functions derived from the SAXS data in 60% sucrose (blue curve) with the one computed from this EM density map. The derived distributions are quite similar. Because the spatial resolution of SAXS is lower than that of EM, some of the features in the blue curve are understandably smeared out. Furthermore, the cryo-EM density map accounts for only 95% of the RNA. This loss is attributed to the flexibility near one of the ends of the RNA molecule. In contrast, SAXS retains sensitivity to all the RNA present. Finally, the slight variations at the largest distances may result from incomplete blanking from the capsid shell, e.g. the exact contrast variation point might be 59%, not 60% as measured.
Although it is simplest to interpret data derived from a single molecular component, such as the RNA as described above, several computational tools are available to interpret SAXS data of multicomponent systems acquired under different contrast conditions. To explore this approach, we performed ab initio multiphase reconstruction of the SAXS data using the MONSA algorithm from ATSAS (57). With MONSA we  Fig. 2, this result suggests that the capsid is permeable and therefore can be blanked. Near ;60% added sucrose there is a significant change in the scattering profile as the signal merges into the noise; this sucrose concentration appears to contrast match, or cancel the protein scattering. The mismatch at low q can be attributed to impurities in the sample; however, we note that even small signals appear amplified when displayed on a logarithmic scale. exploit measurements at many different contrast values to obtain bead models for both the protein and the RNA phases of this complex. Results of these multiphase reconstructions are shown in Fig. 6, adjacent to the EM density map of bacteriophage MS2 (EMD-3403 and 3404) (7). Orthogonal cross-sections are shown to help visualize the structure. All molecules were rendered in UCSF Chimera (58).
As with all SAXS reconstructions, the solution is neither unique nor high resolution, but it accurately captures several characteristic features present in the cryo-EM model. Specifically, both models suggest that the RNA is localized close to the capsid shell and the genome displays an asymmetry reflecting the position of a maturation protein (top of model). This agreement provides additional confidence in the validity of the contrast variation method as applied to viruses.

Comparison with Mg 21 -induced compaction
The contrast variation method offers new opportunities for structural studies of viral genomes. Time-resolved approaches The observed changes are consistent with an increased contribution from the RNA core (relative to the protein contribution) as the solution contrast increases. All curves are normalized to enable comparison and curves in part (A) are offset to aid in visualization. Beyond the match point, near (but likely just below) 60% added sucrose the contribution from the protein shell reappears as the second peak in pair distance distribution. Visualizing a viral genome with contrast variation SAXS can be applied to follow the genome structure as it folds during assembly. Despite the lack of fully detailed structural information, even low-resolution studies can distinguish different models by measuring the extent and global structure(s) of the RNA; very large changes in conformation are required to compress a long RNA into a capsid. Distinct theoretical models, shown as cartoons in Ref. 59, could be readily distinguished as their global structural signatures are quite distinct. Four different models are discussed in that work and each would present a unique signature. The first, a nucleation-elongation model, would be distinguished by a rapid compaction of RNA, which precedes protein binding. The second, micellar condensation, is most likely consistent with a gradual condensation of RNA as it is slowly condensed by protein. The third, an RNA antenna model, would be distinguished by extended RNA structures as protein binds locally. Finally, a packaging signal model would likely display discrete folding steps because of the cooperative nature of the packaging. When coupled with modeling, the key distinct features (and kinetics) of these different models would be easily distinguished.
Finally, the global compaction and folding of viral RNAs differs from typical structural of catalytic RNAs. Many of the latter RNAs fold to compact structures following the addition of divalent Mg, which aids in screening the large negative backbone charge and directs catalytic RNAs toward structures that engage tertiary contacts to secure compact states. Fig. 7 shows SAXS profiles and dummy atom reconstructions of free MS2 RNA in a solution containing 150 mM NaCl (purple, top curve) and following the addition of Mg 21 (orange, middle curve) at concentrations that fold many catalytic RNAs compared with the MS2 RNA in virio (blue, bottom curve). Here, scattering profiles are displayed as Kratky plots of Iq 2 versus q. This representation of the SAXS profiles emphasizes compaction and is useful in studies of RNA folding, where, in the absence of Mg 21 , the molecule assumes more extended states. Dummy atom reconstructions of free MS2 RNA were performed through DAMMIF (60) (ATSAS). Multiple reconstructions are shown for each of the models. Although these reconstructions are not unique, they are useful in exemplifying the changes the RNA undergoes. SAXS studies of other functional RNAs reveal large changes upon the addition of Mg 21 (23,25). For this RNA, the addition of divalent ions does not lead to a large change in structure or Rg. In contrast, large differences in the SAXS signal between encapsidated (lower, blue curve) and free RNA are striking. This change is also accompanied by a collapse in the radius of gyration from 169 to 95 Å within the capsid, consistent with measurements of MS2 in solution by Zipper et al. (61), the spatial distribution of MS2 (46), and the change in genome size measured in other viruses (62). Folding of the viral RNA seems to require the protein. Perhaps this is by design; the architecture of this RNA has evolved to fold with capsid proteins so that it can assemble efficiently.

Conclusion
In conclusion, these studies not only underscore the important role of protein-RNA interactions in compacting/ packaging RNA (as suggested by others) but present a unique strategy for watching this process as it occurs, using CV-SAXS. There is much to learn about the folding of viral RNAs by the proteins that they encode (59). The method, demonstrated here, should be readily transferrable to other nonenveloped viruses and may be useful in unraveling the mechanism of novel antiviral drugs that target assembly, a topic of great current importance.

Sample preparation
Escherichia coli bacteriophage MS2 (ATCC ® 15597-B1 TM ) was propagated in Escherichia coli (Migula) Castellani and Chalmers (ATCC ® 15597 TM ) in the recommended growth media. After overnight growth, E. coli were pelleted by precipitation, and the supernatant was collected. PEG and NaCl were added to final concentrations of 10% w/v and 0.5 M, re- . Protein is shown in red and RNA is shown in blue. The full reconstruction and three orthogonal cross-sections are shown for each case. Although the spatial resolution obtained through SAXS is lower and the reconstruction is not unique, SAXS data are much simpler to acquire than asymmetric cryo-EM reconstructions. Similar structural features are captured by both methods, including a small, protruding piece of RNA that may reflect the position of the maturation protein.
spectively. After incubation for an hour, the bacteriophages were precipitated by centrifugation. The precipitate was resuspended in buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl), filtered, and further purified by gel filtration chromatography though a Superdex 200 column. Contrast variation samples were then dialyzed overnight into the appropriate sucrose concentration. This was done to ensure complete matching between buffer and sample.
MS2 capsid protein (CP) was produced recombinantly in E. coli (BL21). A plasmid containing the MS2 CP sequence was produced de novo by ATUM. Transformed E. coli were grown in LB-Lennox until mid-log phase, after which expression was induced with 1 mg/ml of isopropyl 1-thio-b-D-galactopyrano-side. Protein expression continued for 4 h. Afterward E. coli were pelleted by centrifugation, and the pellet was stored at 220°C. The pellet was thawed, resuspended in buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl), sonicated, and clarified by centrifugation. MS2 CP was then purified as selfassembled VLPs using the same protocol as WT MS2. UV absorbance ratios suggest that the VLPs contain a small amount of mRNA. Because of the exploratory nature of the blanking procedure, sucrose was added to samples with a positive displacement pipette.
MS2 RNA was purchased from Sigma and frozen. After thawing, it was buffer exchanged into 50 mM Tris-HCl, pH 7.5, 150 mM by serial concentration. Lastly it was annealed by Visualizing a viral genome with contrast variation SAXS heating to 95°C for 5 min and then cooled with ice. Magnesium was added immediately before measurement with a pipette.

SAXS data collection
SAXS data were collected at the BioCAT sector of the Advanced Photon Source, in two separate studies. Protocols for acquiring contrast variation SAXS data are detailed in Ref. 53. During the first SAXS data were acquired on MS2 VLP using the standard BioCAT equilibrium setup. Contrast variation data were acquired on MS2 WT in a second beamtime, using a custom built set up that employed coaxial-sheathed continuous flow. In both cases data were collected on a Pilatus 3M detector. Samples were suspended in 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, with added sucrose to enable CV-SAXS. Data reduction was performed using the RAW package (63). Further analysis was performed with the ATSAS suite (64).

Contrast variation shell modeling
The possible effects of permeability on an empty capsid were calculated using the theoretical form factor formula for a coreshell model.
In the above equations, R shell ; R core are respectively the outermost and innermost radii of a shell, and Dr shell ; Dr core are the excess electron density above the solvent for both the shell and core, respectively. In the case of a permeable capsid, Dr core was held at zero as Dr shell was reduced. In the case of the nonpermeable capsid, Dr core was reduced by the same amount as Dr shell . To further simulate experimental conditions, the theoretical SAXS intensity was convoluted with a gaussian function to simulate experimental broadening. Lastly a random noise background was added to simulate the level of signal to noise seen as contrast is reduced.

Data availability
Data will be made available upon request to the corresponding author, Lois Pollack, lp26@cornell.edu.