Neutralizing Antibodies Inhibit Chikungunya Virus Budding at the Plasma Membrane.

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Graham Simmons ( gsimmons@bloodsystems.org ).

EXPERIMENTAL MODEL AND SUBJECT DETAILS Cell lines

Human muscle-cell line RD (Cat. #CCL-136), bone-cell line U2OS cells (Cat. #HTB-96) and adenovirus 5 DNA transformed human kidney epithelial cell line 293 c18 (Cat. #CRL-10852) are all female cell lines purchased from American Type Culture Collection (ATCC). Hamster fibroblast cell line BHK21 cells (Cat. #CCL-10) were purchased from ATCC. Cells were maintained at 37 °C with 5% humidified CO 2 in DMEM (Invitrogen) supplemented with penicillin and streptomycin, 10 mM HEPES, non-essential amino acids, and 10% FBS (Hyclone).

Virus strains

CHIKV strain 37997 and vaccine strain 181/clone 25 (CHIKV 181) were amplified in BHK21 cells.

METHOD DETAILS Transfection

RD or 293 c18 cells were transfected using TransIT 2020 transfection reagent (Mirus) at a ratio of 3 μl of TransIT 2020 to 1 μg DNA diluted in Opeti-MEM (Thermo Fisher Scientific). Cells were at ~80-90% confluency. Antibody production Sequences for all antibody light chain (LC) and heavy chain (HC) were sub-cloned in pTT5 expression vector. Paired LC and HC expressing vectors were transfected into 293 c18 cells and antibodies were purified from the supernatant of the transfected cells. N297Q mutation was introduced to HC of C9 and IM-CKV063 by PCR-based site-direct mutagenesis to generate aglycosylated C9 and IM-CKV063.

Antibody labeling Alexa 488-conjugated

C9 was generated with the Alexa Fluor™ 488 protein labeling kit, per the manufacturer’s instructions (Invitrogen).

Fluorescence microscopy of CHIKV-infected cells

Antibody concentration and treatment time and multiplicity of infection

(MOI) were optimized to reach the most efficient inhibition of virus release by NAbs with a majority of cells infected to facilitate microscopy observation. U2OS cells were infected with CHIKV 181 at a multiplicity of infection (MOI) of 10 for 3 h. After extensive washing, the cells were cultured for 7 h in the presence of 10 μg/ml NAb C9 or IM-CKV063, or the non-NAb IM-CKV066 as a control. At 10 h post infection, the cells were fixed, permeabilized, and then probed with rabbit anti-CHIKV 181 polyclonal antibody (IBT Bioservices, Gaithersburg, MD, USA) followed by staining with Alexa 594-conjugated anti-rabbit and Alexa 647-conjugated anti-human antibodies (Invitrogen). Confocal and STED super-resolution microscopy images were taken with a Leica TCS SP8 STED 3X Nanoscope.

TEM of CHIKV-infected cells

RD cells were infected with CHIKV 37997 at an MOI of 1 for 3 h. After extensive washing, the cells were cultured in the presence or absence of 2 μg/ml human NAb C9, IM-CKV063, or mouse NAb CHK9, CHK152, or the human non-NAb IM-CKV066, IM-CKV062. Then, 20 mM NH 4 Cl were added to prevent further rounds of infection. At 18 h post infection, the cells were fixed with 2.5% glutaraldehyde. After dehydration, the cells were embedded in an epoxy resin, sectioned ultra-thin (60-80 nm), and examined using a transmission electron microscope. IEM RD cells were infected with CHIKV 37997 at an MOI of 1 for 3 h. After extensive washing, the cells were cultured in the presence or absence of 2 μg/ml NAb C9, IM-CKV063, or the non-NAb IM-CKV066. Then, 20 mM NH 4 Cl were added to prevent further rounds of infection. At 18 h post infection, the cells were fixed with 2% paraformaldehyde and 0.025% glutaraldehyde. RD cells were transfected with pCAGGS CHIKV E3/E2/E1 plasmid ( Salvador et al., 2009 ). At 6 h post transfection, the cells were cultured in the presence or absence of 2 μg/ml NAb C9. At 18 h post transfection, the cells were fixed with 2% paraformaldehyde and 0.025% glutaraldehyde. Cells were probed with goat anti-human antibodies conjugated to 6-nm colloidal gold before post-fixation with osmium tetroxide. After dehydration, the cells were embedded in an epoxy resin, sectioned ultra-thin (60-80 nm), and examined using a transmission electron microscope.

IEM tomography

Electron tomography was performed using the IEM samples. In those experiments, semi-thick sections (~150 nm thick) were cut, and dual-axis tiltseries were collected using a Tecnai TF30 microscope operated at 300 kV. Tomographic reconstruction was carried out using the IMOD software package as previously described ( Kremer et al., 1996 ).

CryoET of CHIKV-infected cells

U2OS cells grown on fibronectin-coated gold R2/2 Quantifoil finder EM grids were infected with CHIKV 181 at an MOI of 10 for 3 h. After extensive washing, the cells were cultured for 7 h in the presence or absence of 10 μg/ml Alexa 488-conjugated C9. At 10 h post infection, the grids were imaged with a fluorescence microscope to assess antibody binding to infected cells. Upon confirming binding of antibodies, the grids were one-side blotted with filter paper and plunged into liquid ethane for rapid vitrification. After optimizing vitrification having screened multiple grids in different days, we collected 20 cryoET tiltseries at the regions of interest in the presence or absence of NAbs from different grids during multiple imaging sessions on different days using a JEM2200FS (JEOL Ltd., 3-1-2 Musashino, Akishima, Tokyo 196-8558, JAPAN) electron microscope operated at 200 kV equipped with an in-column energy filter (omega type), using a condenser lens aperture of 100 mm, no objective aperture, spot size 3. The energy filter slit was set to a width of 20 eV. Micrographs were recorded using a DE-20 direct electron detector (Direct Electron, LP, San Diego, CA, USA) at a recording rate of 25 raw frames per second and a total exposure time of 0.6 s, yielding 15 frames per stack at each tilt angle. Tiltseries were acquired semi-automatically using SerialEM software ( Mastronarde, 2005 ) in low-dose mode with a target defocus at the tilt axis of −6 μm. For each tiltseries, the images were acquired from a target range of −60° to +60° in increments of 2°. Frames were motion corrected using DE software (Direct Electron LP, San Diego, CA, USA). The magnification was set to 10,000x for six tiltseries to acquire a larger field of view, yielding images with a sampling size of 5.74 Å/pixel, and to 15,000x for the other 14 tiltseries, yielding images with a sampling size of 3.58 Å/pixel (total cumulative dose for the latter ~ 60 e/Å 2 ). ADCC assay An ADCC reporter bioassay core kit (Promega) was used to measure the induction of ADCC by NAbs that crosslinked viral spikes at the surface of CHIKV-infected cells. U2OS cells grown on 96-well plates (1.25×10 4 cells/well) were infected with CHIKV 181 at an MOI of 5 for 3 h. After extensive washing, the cells were cultured in the presence or absence of serial dilutions of anti-CHIKV or control anti-DENV antibodies. At 7 h post infection, 7.5×10 4 effector cells that stably expressed FcγRIIIa and an NFAT response element driving the expression of firefly luciferase were added to each well. Luciferase activity was measured 10 h later according to the manufacturer’s instructions (Promega). The fold of induction was calculated as relative luciferase unit (RLU) (induced-background)/RLU (no antibody-background).

QUANTIFICATION AND STATISTICAL ANALYSIS

We analyzed following features in our datasets: 418 fluorescent puncta in CHIKV-infected cells treated with non-NAb and 107 fluorescent patches in CHIKV-infected cells treated with NAb in STED dataset; 98 NCLPs in CHIKV-infected cells treated with NAb in IEM tomogram dataset; 47 egressed virions and 82 membrane patches in CHIKV-infected cells, and 1,163 NCLPs, 243 membrane patches and 99 spherules in CHIKV-infected cells treated with NAb in cryoET dataset. Detailed classification and parameters used for STA are described below. NCLP size measurement from IEM tomograms of fixed CHIKV-infected RD cells Nucleocapsid-like particle (NCLP) subtomograms (n=98) were manually picked and extracted from the IEM tomograms using e2spt_boxer.py , publicly available through EMAN2. The NCLPs appeared pleomorphic, varying widely in size. To quantify the variation in size, we computed radial density profile plots from spherically averaged subtomograms and measured the hydrodynamic radius ( i.e. , the radius of gyration, or spherically averaged radius) of individual particles by identifying the radius at which the steepest slope occurred after the maximum peak in their corresponding radial density profiles. Lacking adequate models for heterogeneous NCLPs, we fine-tuned the centering of the manually picked particles by aligning them translationally to a spherical blob with a soft Gaussian falloff. The particles were low-pass filtered to 40 Å and spherically masked with a soft Gaussian mask to optimize their translational alignment prior to computing their radial density profiles.

Reconstruction of cryoET tiltseries into tomograms

CryoET tiltseries were subjected to fiducial-based alignment in IMOD and CTF corrected using EMAN2 ( Galaz-Montoya et al., 2016 ) with defocus values derived through CTF fitting for the images and tiltseries for which the power spectrum showed clear CTF oscillations. Otherwise, the average defocus value between flanking images in the tiltseries was used based on the assumption of defocus trends. For tiltseries for which the CTF could be fitted only in a minority of images, or could not be fit in any of them, the target defocus at the tilt axis was assumed to be approximately correct. The tiltseries were reconstructed into tomograms using a compressed sensing algorithm implemented in the ICON-GPU package ( Chen et al., 2017b ). The tiltseries were pre-filtered with a band-pass filter (a low pass at 100 Å, and high-pass zeroing out the first four pixels in Fourier space) to improve gold fiducial visibility and tracking during tiltseries alignment; however, the alignment parameters were applied to the unfiltered tiltseries, and gold fiducials were erased with IMOD prior to tomographic reconstruction with ICON-GPU. The tiltseries were shrunk by a factor of two using Fourier cropping to increase the reconstruction speed with ICON-GPU. Tomographic annotation was performed semi-automatically using the neural network algorithm in EMAN2 ( Chen et al., 2017a ). We used Chimera UCSF ( Pettersen et al., 2004 ) for all isosurface visualizations. STA All particle picking was done manually using e2spt_boxer.py available in EMAN2 on 4x down-sampled and heavily filtered tomograms (low-pass to 100 Å, high-pass erasing the first four Fourier pixels). Subtomogram alignment and averaging, and all related analyses and controls, were performed using the tools for single-particle tomography in EMAN2 ( Galaz-Montoya et al., 2015 ) as follows. Released virions were extracted from control tomograms in the absence of NAbs (n=47). Reference-free initial models were built using self-symmetry alignment ( Dai et al., 2013 ) and used to seed iterative, gold-standard refinement ( Galaz-Montoya et al., 2015 ) until convergence was achieved. For “membrane particles” (patches of membrane) from the control tomograms in the absence of NAbs (n=82) and from tomograms in the presence of NAbs (n=243), an initial model was built using hierarchical ascendant classification for each case, after which the particles in each set (with or without NAbs) were iteratively refined against each respective initial model and averaged. We also extracted arrested nucleocapsid-like particle (NCLP) subtomograms (n=1,163) from tomograms of CHIKV-infected cells in the presence of NAbs. Given previous reports of icosahedral subpopulations existing among isolated nucleocapsids (NCs) in vitro ( Lamb et al., 2010 ; Paredes et al., 2003 ), we hypothesized that assembled NCs inside cells might also acquire an icosahedrally symmetric conformation. To test this hypothesis, we carried out extensive STA analyses of the NCLPs from tomograms at the highest magnification and with the best contrast for STA (n=763). We first subjected the arrested NCLPs to self-symmetry alignment but the algorithm failed to produce any consistent icosahedrally symmetric initial models using the same method that worked on released virions, even when the algorithm was run for many more iterations (n=110) than those needed to align typical icosahedral viruses (n=10-50). Given our successful alignment of released virions using the same published methodology ( Dai et al., 2013 ) and software ( Galaz-Montoya et al., 2015 ; Galaz-Montoya et al., 2016 ) as well as tomograms reconstructed using the same pipeline from tiltseries of similar quality, the results for the arrested NCLPs suggested that the vast majority were indeed not icosahedrally symmetric but rather pleomorphic and structurally heterogeneous. The absence of readily detectable symmetry in NCLPs persisted even for trials in which we computationally deleted the central region of NCLPs (at a radius of ~20 pixels, corresponding to ~14 nm) with a soft mask to abrogate any influence from the genome on alignment, since the genome has been hypothesized to lack an icosahedral arrangement, even in mature virions. Given that the averages after self-symmetry alignment did not resemble the known structure in mature virions, we performed visual examination of randomly selected individual particles (n=~100). We did not find any NCLPs resembling the known structure of mature NCs in budded virions. While some individual subtomograms exhibited angular features and some of the symmetry imposed structures might correspond to legitimate conformations for a few individual NCLPs (not shown) even if they differ from the known structure inside mature virions, further analyses or studies with larger datasets at higher resolution are needed to validate them. We then iteratively refined the entire set of NCLPs without enforcing symmetry. The average of the NCLPs converged to a low-resolution globular structure, ~20 nm in radius, that did not resemble the known structure inside mature virions. Quintile classification based on correlation after alignment against the global average, followed by iterative refinement and averaging of the particles within each quintile, yielded five structures that were all globular yet structurally pleomorphic in different ways. This strongly suggested that presence of extensive structural heterogeneity among the NCLPs in our specimens. To determine whether a “good” initial model could help to computationally isolate a consistent, icosahedrally symmetric subpopulation, we aligned all of the NCLP subtomograms to the known icosahedral structure of the nucleocapsid in mature CHIKV. The initial average resembled the known structure due to model bias. However, when we iteratively refined the data to prevent model bias, the average structure across iterations drifted quickly from the initial biased average, reverting to a low-resolution, globular density. This strongly suggested that the bulk of NCLPs were not in a conformation like that of NCs in mature virions. We then took the 10% of particles that had the best correlation with the known structure after a single round of alignment and aligned them de novo to each other, in an all-versus-all fashion. The average structure did not resemble that of the nucleocapsid inside mature virions. These analyses suggested that there was no significant subpopulation among our in situ budding-arrested NCLPs dataset with the same icosahedral arrangements as those seen in NCs isolated in vitro and in NCs inside mature virions. Indeed, in our cryoET tomograms, the NCLPs appeared pleomorphic upon visual inspection, in agreement with the observations from our 2D and 3D TEM/IEM experiments with fixed cells. Next, we again used radial density profiles, as described above for the analysis of NCLPs in IEM tomograms, to measure the size distribution of the NCLPs and classify them by size. Of the 17 size classes by radius, those at the tails at either end of the distribution (representing particles with radii between ~8-15 nm, and ~23-28 nm) were sparsely populated (n