Chemicals, viruses, and cells
H84T-BanLec was prepared as previously described2. MERS-CoV (EMC/2012) was kindly provided by Ron Fouchier (Erasmus Medical Center, the Netherlands)4. SARS-CoV-2 (HKU-001a strain, and the B.1.1.7 and P.3 variants) were isolated from the respiratory tract specimens of COVID-19 patients in Hong Kong. Archived clinical strains of SARS-CoV, HCoV-OC43, and HCoV-229E were obtained from the Department of Microbiology, The University of Hong Kong (HKU). The cell lines used in this study were available in our laboratory as previously described42-44. All experimental protocols involving live MERS-CoV, SARS-CoV-2, and SARS-CoV followed the approved standard operating procedures of HKU Biosafety Level 3 facility.
Cytotoxicity and antiviral assays
The cytotoxicity of H84T-BanLec in various cell lines was determined using the CellTiter-Glo® luminescent cell viability assay (Promega Corporation, Madison, WI) according to manufacturer’s instructions and as previously described24,45. Viral load reduction, CPE inhibition, and plaque reduction assays for coronaviruses were performed as we previously described46,47. To test the vulnerability of H84T-BanLec to induce resistant escape mutants, a drug resistance assay was performed as we described previously48. In brief, MERS-CoV and SARS-CoV-2 (MOI=0.01) was passed in the presence of about 5× EC50 of H84T-BanLec (30nM) for 8 passages and in VeroE6 cells. After the 8 passage, the antiviral activities of HB84T-BanLec against the wild-type and passaged viruses were compared using viral load reduction assay.
Ex vivo human lung tissue organ culture
The anti-MERS-CoV activity of H84T-BanLec was evaluated in an established ex vivo human lung tissue organ culture model as we described previously22,23. The human lung tissues were obtained from patients with lung tumours who underwent surgical operations at Queen Mary Hospital, Hong Kong. Briefly, the freshly isolated human lung tissues were rinsed with the primary tissue culture medium, which contained the advanced Dulbecco’s Modified Eagle Medium (DMEM)/F12 medium supplemented with 2mM HEPES (Gibco, Thermo Fisher Scientific, Waltham, MA, USA), 1×GlutaMAX (Gibco), 100U/ml penicillin, 100μg/ml streptomycin, 20μg/ml vancomycin, 20μg/ml ciprofloxacin, 50μg/ml amikacin, and 50μg/ml nystatin. Human lung tissues were cut into small pieces at comparable sizes and then infected with MERS-CoV at 1×106 plaque-forming units (PFU)/ml in a 6-well culture plate. After 2h, the inoculum was removed and the specimens were washed with the primary tissue culture medium. The infected specimens were then transferred to inserts of 12-well transwells (Corning Life Sciences, Tewksbury, MA, USA) pre-coated with 200µl 60% Matrigel (Corning Life Sciences) diluted with the primary tissue culture medium. An additional 60% Matrigel was then added to the insert to seal the explant. The basolateral compartment was filled with 1ml of the primary tissue culture medium supplemented with 0, 5, or 20µM H84T-BanLec. The samples were harvested at 24hpi for confocal microscopy and qRT-PCR analysis. The virus genome copies in cell lysate samples were normalized with GAPDH copy numbers. Immunofluorescence staining of MERS-CoV nucleocapsid protein in the lung tissues was performed as previously described22,49,50.
Co-localization of H84T-BanLec and SARS-CoV-2 proteins in autopsied lung sections
Autopsied lung sections of a deceased COVID-19 patient were obtained from the Department of Pathology, University of Michigan51 and sequentially stained with biotinylated H84T-BanLec (1:25k, Opal 620 reporter), and antibodies to SARS-CoV-2 nucleocapsid (Genetex, GTX635679, 1:1000, Opal 570 Reporter) and spike proteins (Genetex, GTX632604, 1:50, Opal 690 Reporter). After each round of antigen detection, the slides were exposed to microwave epitope retrieval in 10mM citrate buffer to remove the antibody complex, leaving behind the deposited fluorochrome. The slide was then imaged. A pseudo-coloured image was created. Original monochrome images were analyzed using the coloc 2 sub-routine in FIJI and the Pearson correlation coefficient was calculated52.
hDPP4-transgenic mouse model for MERS-CoV infection
The in vivo anti-MERS-CoV activity of H84T-BanLec was evaluated in an established hDPP4-transgenic mouse model as we described previously23,24. Briefly, hDPP4 mice, aged 5-8 weeks, were obtained from the HKU Centre for Comparative Medicine Research and divided into different groups to receive H84T-BanLec or sham (PBS) treatment. After anaesthesia, the mice were intranasally inoculated with 20μl of virus suspension containing 500PFU of MERS-CoV (n=10 per group). Each mouse in the H84T-BanLec group was treated with intraperitoneal (15 mg/kg/animal/day every 24h) H84T-BanLec starting at 6 hours before virus challenge until 3dpi. The control mice were treated with intraperitoneal PBS starting at 6h before virus challenge daily until 3dpi. Five animals per group were sacrificed at 4dpi for viral load quantitation by qRT-PCR, virus titer quantitation by plaque assay, and histopathological studies as described previously23, 24. The survival rates and body weight changes of the remaining 5 animals per group were observed until 14dpi or death as described previously. Immunofluorescence staining for MERS-CoV nucleocapsid protein expression was performed as we described previously24.
Golden Syrian hamster model for SARS-CoV-2 infection
The in vivo anti-SARS-CoV-2 activity of H84T-BanLec was evaluated in an established golden Syrian hamster as we described previously25,45. Briefly, male and female golden Syrian hamsters, aged 6-10 weeks, were obtained from the Chinese University of Hong Kong Laboratory Animal Service Centre through the HKU Centre for Comparative Medicine Research. The hamsters were divided into different groups to receive H84T-BanLec or sham (PBS) treatment and their clinical scores were recorded as described previously53. At 0dpi, each hamster was intranasally inoculated with 100µL of DMEM containing 105 PFU of SARS-CoV-2 (HKU-001a strain) under intraperitoneal ketamine (200mg/kg) and xylazine (10mg/kg) anaesthesia. The H84-BanLec group hamsters were treated with intraperitoneal (15.0mg/kg/day every 24h) or intranasal (1.5 mg/kg/animal/day every 24h) H84T-BanLec starting at 6h before virus challenge (pre-challenge) or 24h after virus challenge (post-challenge) until 3dpi (n=10 per group). The control hamsters were treated with intraperitoneal or intranasal PBS starting at 6h before virus challenge daily until 3dpi (n=10 per group). Five animals per group were sacrificed at 4dpi for viral load quantitation by qRT-PCR, virus titer quantitation by plaque assay, and histopathological studies as described previously13. The survival rates and body weight changes of the remaining 5 animals per group were observed until 14dpi or death. Immunofluorescence staining for SARS-CoV-2 nucleocapsid protein expression was performed as we described previously13.
Time-of-drug-addition assay was performed as we described previously to determine the phase(s) of the SARS-CoV-2 replication cycle targeted by H84T-BanLec26. Briefly, VeroE6 cells were infected with SARS-CoV-2 (MOI=3.0) and 20nM H84T-BanLec was added pre-infection (-1h), at the time of infection (0h), pre-incubated with SARS-CoV-2 pre-infection, or at 1h post-infection. Intracellular viral genome copy numbers of the corresponding time points were determined at 8h post-infection by qRT-PCR and compared with PBS-treated controls.
The ex vivo human lung tissue culture experiments and the animal experiments were approved by the Institutional Review Board of HKU/Hospital Authority Hong Kong West Cluster and the HKU Committee on the Use of Live Animals in Teaching and Research, respectively. The autopsy of the deceased COVID-19 patient was exempt from institutional review board approval.
All data were analysed with GraphPad Prism software (GraphPad Software, Inc). One-way ANOVA or student’s t-test was used to determine significant differences in viral loads and titers, and Kaplan-Meier survival curves were analysed by the log rank test. P<0.05 was considered statistically significant.
Conjugation of SARS-CoV-2 spike trimer and H84T-BanLec through histidine residues of His-tagged protein
A maleimide-Poly(ethylene glycol) (PEG) linker was attached to a 3-aminopropyltriethoxysilane (APTES)-coated AFM cantilever (or silicon nitride surface for H84T-BanLec immobilization) by incubating the cantilevers for 2h in 500µL of chloroform containing 1mg of maleimide-PEG-N-hydroxysuccinimide (NHS) (Polypure, Oslo, Norway) and 30µl of triethylamine. After washing 3 times with chloroform and drying with nitrogen gas, the cantilevers were immersed for 2h in a mixture of 100µL of 2mM thiol-trisNTA, 2µL of 100mM EDTA (pH 7.5), 5µL of 1M HEPES (pH 7.5), 2µl of 100mM tris(carboxyethyl)phosphine (TCEP) hydrochloride, and 2.5µL of 1M HEPES (pH 9.6) buffer, and subsequently washed with HEPES-buffer saline (HBS). Thereafter, the cantilevers were incubated for 4h in a mixture of 4µL of 5mM NiCl2 and 100µL of 0.2µM His-tagged Spike trimer. After washing with HBS, the cantilevers were stored in HBS at 4°C54.
Single-Molecule Force Spectroscopy (SMFS) measurement
Force distance measurements were performed at room temperature (~25ºC) using cantilevers with 0.01N/m nominal spring constants (MSCT, Bruker) in TBS buffer containing 1mM CaCl2, and 0.1% Tween-20. Precise spring constant values of AFM cantilevers were determined by measuring the thermally driven mean-square bending of the cantilever using the equipartition theorem in an ambient environment55. The deflection sensitivity was calculated from the slope of the force-distance curves recorded on a bare silicon substrate. Determined spring constants ranged from 0.008-0.015N/m. Force-distance curves were acquired by recording at least 1000 curves with vertical sweep rates between 0.5s and 10s and at a z-range of typically 500-1000nm, resulting in a loading rate from 10-10,000pN/s, using a commercial AFM (Keysight Technologies, USA). The relationship between experimentally measured unbinding forces and parameters depicting the interaction potential were deciphered using the kinetic models of Bell, and Evans and Ritchie32,33.
Length data analysis
Software for data analysis was written in Matlab (MathWorks, Inc.). Force curves (n=4000, 2 different tips) were analysed for the calculation of the distance and for the construction of experimental probability density function (PDF) of distance. The distance between two unbinding events was calculated with the correction of the cantilever deflection. Thereafter, for each length difference and the corresponding standard deviation given by the pixel mash of the data recordings, single Gaussians with unitary area were constructed, summed up, and normalized to calculate the experimental probability density function. This data presentation is advantageous over conventional histograms, as the data accuracy is taken into account and binning artefacts can be avoided.
Surface Plasmon Resonance (SPR) measurement
SPR was used to study the kinetics of binding and dissociation of H84T-BanLec to SARS-CoV-2 spike protein in real time. SARS-CoV-2 spike protein was immobilized onto a Sensor Chip NTA (cytiva) via its His6-tag after washing the chip for at least 3min with 350mM EDTA and activation by applying a solution of 0.5mM NiCl2 for 1min. 50nM SARS-CoV-2 spike protein was injected multiple times to generate a stable surface. For the determination of kinetic and equilibrium constants, H84T-BanLec was injected at different concentrations (10-200nM) and subsequently removed with buffer. Running buffer was TBS, pH 7.4, containing 0.1% Tween-20, 1% BSA, and 1mM Ca2+. The resulting experimental binding curves were fitted using the “bivalent analyte model”, assuming two-step binding of H84T-BanLec to SARS-CoV-2 spike protein.
Cell culture experiments
All force-distance curves were recorded at room temperature by using a PicoPlus 5500 AFM setup (Agilent Technologies, Chandler, AZ, USA) on living cells with the assistance of a CCD camera for localization of the cantilever tip on selected cells. The optical system of the AFM was focused on the cantilever tip, while the sample plate with the Petri dish was moved upwards by the step motor. Before the cells on the dish reached the focus, the piezo tube of the AFM was started to scan in the z-axis with a scanning range of 3µm and at a scanning frequency of 1Hz. The sample plate was moved upwards by the step motor using manual control with 1µm per step. Due to the resistance of the liquid, a gap between the approaching curve and the retraction curve appeared, when the AFM tip was close to the sample surface. About 2µm before the AFM tip touched the sample surface, the approaching curve was no longer parallel to the retraction curve. With this signal, the movement of the step motor was stopped. Further approaching was accomplished by gradually changing the voltage on the piezo tube. With this approaching method, the indentation force of the first contact between the AFM tip and the sample surface was controlled to be less than 30pN.
The functionalized cantilever (pegylated with SARS-CoV-2 spike protein) with a nominal spring constant of 0.01N/m was moved downward to the cell surface and moved upward after the deflection of the cantilever reached the force limit. The deflection (z) of the cantilever was monitored by a laser beam on the cantilever surface and plotted versus the Z-position of the scanner, from which the force (F) was determined according to Hook’s law (F =kz, with k being the cantilever spring constant). When the tip tethered molecule bound to ACE2 on the cell surface, a pulling force developed during the upward movement of the cantilever causing the cantilever to bend downwards. At a critical force, i.e. the unbinding force, the tip tethered spike protein detached from ACE2, and the cantilever jumped back to its neutral position. The sweep range was fixed at 3000nm and the sweep rate was set at 1 Hz. On each cell, at least 100 force-distance cycles with 2000 data points per cycle and typical force limit of about 30pN were recorded.
Structural data analysis
The unbinding event was identified by local maximum analysis using a signal-to-noise threshold of 2. The binding activity was calculated from the fraction of curves showing unbinding events. Two-tailed Student’s t-test was performed for statistical analysis.
High-speed atomic force microscopy (HS-AFM) imaging
10nM of purified SARS-CoV-2 spike protein was suspended in imaging buffer (10mM Hepes, pH 7.4, 140mM NaCl, 5mM KCl, 1mM CaCl, 1mM MgCl), containing 5µM NiCl2, of which 1.5µL were applied to freshly cleaved mica disc (2mm diameter). After 3min, the surface was rinsed with ~15µL imaging buffer (without drying) and the sample was mounted in the imaging chamber of the HS-AFM (RIBM, Japan). Similarly, H84T-BanLec stock was diluted to 2µM in imaging buffer and 1.5µL of the diluted solution was applied to mica for 3 min, rinsed with 15µL imaging buffer, and imaged using HS-AFM. For following the complexation of H84T-BanLec and SARS-CoV-2 spike protein, the SARS-CoV-2 spike trimer was deposited on mic as descrbed above, before 2µM H84T-BanLec was added to the imaging chamber. For imaging, we used ultra-short cantilevers USC-F1.2-k0.15 (NanoWorld) with nominal spring constant 0.15N/m, resonance frequency of ~500kHz, and quality factor of ~2. During image acquisition, the amplitude was set to 85-90% of the free amplitude (~3nm) and kept constant using a feedback loop.
HS-AFM image processing and volume measurement
Horizontal scars, which occurred due to feedback instabilities or particles sticking to the AFM tip, were selected and removed by Laplacian background substitution. A height threshold mask was used for selecting the background prior to correction of scanning artefacts. Next, Gaussian filter was applied to the images. For the volume measurement, the protein surface was selected using a height threshold mask defined from a minimum height of 0.25-0.35nm to the maximum height of the protein structure. Image processing was done using Gwyddion 2.55.
Three-dimensional structural modelling
All structural models of the SARS-CoV-2 spike trimer were made from the fully glycosylated 3D model of the SARS-CoV-2 spike trimer, based on the cryo-EM structure of Walls et al. (PDB: 6VYB)56. Curved distances over the surface of the protein between glycans were calculated using the distance field algorithm57. This algorithm determines the shortest distance between the centres of geometry of selected glycans over a grid with grid-spacing of 0.5nm, avoiding any grid points that are within 0.5nm of any protein atom. The path was further smoothened by averaging over 5 neighbouring path points. Models of the complex of spike with H84T-BanLec were created using the structure of BanLec (PDB: 4PIK)2. A superposition of the bound carbohydrates of chain A of this model with the oligomannosidic glycan at N234 of the SARS-CoV-2 spike was performed to create models of the H84TBanLec/SARS-CoV-2 spike complex.