Sub-Nanometer Structural Analysis of Respiratory Syncytial Virus Matrix Mutants by Cryo-Electron Tomography and Sub-Tomogram Averaging

Respiratory syncytial virus (RSV) is an enveloped, negative-sense RNA pneumovirus that infects the human respiratory tract [1]. Although vaccines are available for pregnant women and older adults, similar protective measures do not exist for infants, who are at risk of severe disease from RSV infection [2]. Antivirals targeting RSV replication remain under development and would be valuable therapeutics for mitigating severe outcomes [3]. Advancing our understanding of how RSV replicates and assembles into infectious particles would support therapeutic design and vaccine development. The RSV matrix (M) protein directs particle assembly into pleomorphic filaments [4-6]. Using cryo-electron tomography (cryo-ET) and sub-tomogram averaging (STA), our group and others have demonstrated that RSV M forms a helical-like lattice of dimers under the inner leaflet of the viral membrane [5-8]. Prior research suggests that coordination of M oligomerization is necessary for assembly of infectious RSV particles [9,10]. Determining the molecular interactions that stimulate and stabilize M oligomerization could further our understanding of RSV assembly and particle structure. To investigate RSV M lattice formation, we used a robust STA workflow to generate density maps for the RSV M lattice to sub-nanometer resolution. RSV wild-type M and M-mutant strains (T205A, T205D, E231A) were created with site-directed mutagenesis and rescued as previously described [9-11]. BEAS-2B cells were seeded on gold EM grids (Quantifoil, 200 nm mesh, Au R2/1), infected with virus, then vitrified at 24 hours post-infection using a Leica EM GP [8]. Using SerialEM, tilt-series were collected on a Thermo Fisher Titan Krios G3i 300 kV electron microscope equipped with a Gatan Bioquantum energy filter and a K3 direct electron detector [12]. A dose symmetric scheme was used to collect tilt-series from -60° to 60° with 3° increments, a total dose of ~100 e/Å2, an unbinned pixel size of 1.698 Å, and a defocus range of -2 to -6 µm with 0.25 µm steps [8,13]. Pre-processing steps, including frame alignment and CTF correction, were conducted using RELION 5.0 [14]. Tilt-series fiducial-based alignment was performed in IMOD using the RELION 5.0 wrapper [15]. Tomogram reconstruction was conducted in RELION 5.0, and tomograms binned by six were denoised with IsoNet [14,16]. Segmentation of the M layer in virus particles was performed with Dynamo on IsoNet-corrected tomograms binned by twelve [17]. The surface models were scaled to bin six, and duplicate points (i.e., points within 30 Å of one another) were removed. The central axis through each cylindrical model was defined in IMOD, and the PEET spikeInit program was used to generate initial particle orientations [18]. The angles of each cylindrical model were adjusted to align the model along the y-axis. An initial reference point on the matrix lattice near the top of the virus particle was selected from each model. For initial alignment and template matching in PEET, each tomogram was processed separately. Initial sub-tomogram averaging was performed in PEET using bin 6 tomograms generated in IMOD without IsoNet correction. Classification with Principal Component Analysis (PCA) was performed to remove erroneous particles. Particles from all tomograms were pooled, and an average was reconstructed from bin 4 tomograms. An additional round of PCA classification was performed before particles were imported back into RELION 5.0 for bin 2 reconstruction. Geometric cleaning was performed to remove erroneous particles based on z-plane position and increased distance from neighboring particles. The final, unbinned reconstruction was performed with the cleaned particles and refined with CTF refinement, 3D auto-refinement, and Bayesian Polishing [14]. To investigate the molecular interactions involved in M lattice organization, we examined RSV particles with mutations in M (T205A, T205D, E231A) known to influence particle assembly [9,10]. Using our workflow, we generated density maps of the M lattice for each RSV M-mutant strain at sub-nanometer resolution. Among the average volumes, we note the E231A mutant M lattice appears similar in organization to wild-type, whereas the helical placement of M dimers in the T205A and T205D lattices appears shifted, suggesting an alteration in the dimer-dimer interface propagated across the large-scale lattice architecture in virus particles [19]. Sub-tomogram averaging workflow for RSV M lattice. M mutant sub-tomogram averages compared to wild-type M [8]. Isosurfaces of sub-tomogram averages of M lattices for (left to right) WT, E231A, T205A, T205D. Yellow circles are placed at the center of M dimers and connected by a yellow line to demonstrate the difference in M dimer placement in the M lattice. The dashed yellow line represents the central axis used to define the angular rotation of the M lattice relative to WT. Angular rotation of the M lattice is denoted above the dashed yellow line. Scale bar=5 nm. Density maps are averages from n=38,771 (WT), 15,891 (E231A), 8,705 (T205A), 13,755 (T205D) sub-tomograms.