Electron microscopy studies of the coronavirus ribonucleoprotein complex

Coronaviruses are enveloped viruses that cause different diseases in humans and animals (Su et al., 2016). Murine hepatitis virus (MHV) causes hepatitis, enteritis and central nervous system diseases in rodents and is one of the beststudied coronaviruses. MHV belongs to the genera betacoronavirus. Members from the same genera also include highly pathogenic coronaviruses such as the severe acute respiratory syndrome coronavirus (SARS-CoV) and the Middle-East respiratory syndrome coronavirus (MERS-CoV) (Vijay and Perlman, 2016). The coronavirus has a single strand, positive sense RNA genome of about 30 kb, which encodes 4–5 structural proteins, including the nucleocapsid (N) protein, the matrix (M) protein, the small envelope (E) protein, the spike (S) glycoprotein and for some betacoronaviruses, the hemagglutinin esterase (HE) protein (Su et al., 2016). The N proteins bind the viral RNA genome and play important roles in packaging and stabilizing the virus genome, in viral particle assembly and envelope formation, and in the genomic RNA synthesis (McBride et al., 2014). Moreover, it was reported that coronavirus nucleoprotein can regulate host cell cycle, cell stress response, and influence the immune system and other cellular responses (Lu et al., 2011; McBride et al., 2014; Cui et al., 2015; Chang et al., 2016). The N proteins of different coronaviruses are homologous and can be divided into five parts and domains: the N terminal flexible arm, the N terminal domain (NTD), the middle disordered region (LKR), the C terminal domain (CTD), the C terminal flexible tail. The N terminal arm, C terminal tail and the LKR are flexible (Chang et al., 2014). The NTD structures of MHV, SARS-CoV, infectious bronchitis virus (IBV), human coronavirus strain OC43 (HCoV OC43) and the CTD structures of MHV, SARS-CoV and IBV were determined using either x-ray crystallography or NMR (Chang et al., 2016). The determined NTD or CTD structures are highly similar among different coronaviruses. Both the NTD and CTD are shown to interact with the genome RNA while the CTD is also responsible for the dimerization of the nucleoproteins (Chang et al., 2014). The domain crystal structures have provided useful information on the assembly of the ribonucleoprotein complex (RNP), but a lack of the full-length N protein structure and the RNP structure limits our understanding to the assembly and function of coronavirus RNP. Previous analysis of the RNP extracted from the virus by using negative staining electron microscopy showed that coronavirus RNP might be a long helix with a diameter between 9 nm to 16 nm (Macneughton and Davies, 1978). In this study, we isolated the RNPs from MHV and performed negative staining EM and cryo-EM images analysis of the isolated intact and degraded RNPs. We found that the isolated RNPs are in either relaxed helical sausage-like or supercoiled flower-like structures. Interestingly, we also found that the isolated intact RNPs degraded into small pothook-like subunits. These small subunits could be the building blocks of the long loose helical and the supercoiled flower-like RNP structures. We performed both negative staining EM and cryo-EM analysis of the MHV (strain MHV-A59) particles. Negative staining images of the intact MHV particles showed that most viral particles had a round shape while some distorted particles were also observed (Fig. 1A). Cryo-EM image analysis of the same sample showed almost all round shaped particles (Fig. 1B), indicating that the distortion in the negative staining images might be caused by the staining procedure. The corona-like spikes around the envelope could be identified in both the negative staining images (Fig. 1A) and cryoEM images (Fig. 1B). The cryo-EM MHV particles were picked and subjected for 2D classification analyzes. The results showed that the particles have a diameter of ∼80 nm to 90 nm (Fig. 1C), which is consistent with the previous EM results (Neuman et al., 2006; Barcena et al., 2009). A dense interior core corresponding to the intertwined RNP is encapsulated inside the envelope (Fig. 1C). We then broke the MHV particles by incubating the particles in a lysis buffer containing ∼3% CHAPS. Negative staining analysis showed that most of the virus particles were broken and the RNPs were released after the treatment. The released RNPs are in either a loose filament structure or in a compact flower-like assembly that may be similar to the intact RNP assembly in virus particles (Fig. 1D). There were also some smaller particles, which might be the RNP fragments (Fig. 1D). SDS-PAGE gel analysis of the intact virus showed that the N protein is about 55 kDa (Fig. 1E). To investigate the

centrifugation for 5 min at 21,130 × g. The aged RNPs were prepared by placing the purified RNPs at 4 degree for about one week. The intact MHV particles, fresh purified and aged RNPs were analyzed using 12% SDS-PAGE gels and EM.

Negative staining EM analysis of the intact MHV virus and the purified RNPs
For negative staining specimen preparation, 4 microliters of samples were applied onto a glow discharged continuous carbon grid (Beijing XinxingBraim Technology Co., Ltd.). After waiting for approximate 1 minute to let the grid adsorb enough materials, excess sample was blotted away from the grid with filter paper. The grid was immediately washed twice with 1% uranyl acetate (UA) solution and was then incubated with the UA solution for additional 1 min. The UA solution was completely removed with filter paper and the grids were air-dried at room temperature.
For cryo-EM specimen preparation, 3 microliters of samples were applied onto a glow discharged continuous carbon grid with a layer of continuous ultrathin carbon film (TED PELLA, INC.). Each grid was blotted for 9-11 s, plunged into liquid ethane using a Gatan Cryoplunge 3 System. Cryo-EM specimens and the negative staining specimens of the purified aged RNP were examined under an FEI Tecnai F20 electron microscope equipped with an FEG filament. The microscope was operating at 200-kV acceleration voltage, using a nominal magnification of 62,000 × (pixel size: 0.135 nm). Negative staining specimens of the fresh purified RNPs were examined under an FEI Tecnai Spirit Bio-Twin electron microscope equipped with a LaB 6 filament operating at 120 kV acceleration voltage. All images were recorded using Gatan 895 4 k × 4 k CCD cameras with an exposure dose of 20-30 e -/Å 2 . The defocus range used was -1 μm to -3.5 μm.
Cryo-EM tomographic tilt series were collected with an FEI Titan Krios microscope equipped with a FalconⅡ camera at a pixel size of 0.29 nm. The FEI tomography software was used for the data collection. Each tilt series covered an angular range from -64° to 64° with a tilt step of 2°. The defocus range used was -4 um to -6 um. The total dose was about 110 e -/Å 2 for each tilt series.

Image Processing
For the negative staining images of the purified aged RNP, 24,826 particles from 55 micrographs were boxed using the EMAN2 program e2boxer.py (Tang et al., 2007).
The initial model was generated by the EMAN2 python script e2initialmodel.py and low-pass filtered to 60Å. Reference-free 2D class averages were performed by using EMAN2 and RELION1.4 (Scheres, 2012). Particles of bad class averages were discarded and 22,060 good particles were used for the 3D auto-refinement. To validate the accuracy of the reconstructions, we did the refinement and reconstruction with both EMAN2 and RELION1.4 using the same low-pass filtered initial model and same particle dataset. Both 3D refinements were performed following a ''gold standard'' procedure, wherein the raw data were randomly divided into two subsets that were then refined independently. The resolution was 25 Å at Fourier shell correlation value 0.5 and 19 Å at Fourier shell correlation value 0.143. Docking of the crystal structures into 3D reconstruction map was performed in UCSF Chimera (Pettersen et al., 2004).
Particle boxing and reference-free 2D class average analyzes of the fresh purified RNPs were performed using the EMAN2 software package. 3D density maps were visualized using UCSF Chimera.
The tomographic data were processed with the software Protomo (Noble and Stagg, 2015). The tilt series images were aligned for several iterations and then the aligned images were back-projected to generate the 3D reconstructions. The tomograms were denoised by band-pass filtering and virtualized using the software IMOD (Kremer et al., 1996).

Tilt-pair Validation
For tilt-pair validation experiment, data were collected at 0° and 20° tilt angles for each region of the grid, using an FEI Tecnai Spirit Bio-Twin electron microscope equipped with a LaB 6 filament and operated at 120-kV acceleration voltage at a pixel size of 0.16 nm. A total of 39 particle pairs were used for the validation test. The EMAN2 reconstruction density map was scaled to the same pixel size and the same boxsize as those of the tilt-pair images before it was used as the validation reference.
The tilt-pair validation test was performed using the EMAN2 python script e2tiltvalidate.py. About 17 of the 39 particle pairs were inside the red cycle around the corresponding angle (~44%).