Binding behavior of spike protein and receptor binding domain of the SARS-CoV-2 virus at different environmental conditions

A novel coronavirus, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) was identified as the cause of the COVID-19 pandemic that originated in China in December 2019. Although extensive research has been performed on SARS-CoV-2, the binding behavior of spike (S) protein and receptor binding domain (RBD) of SARS-CoV-2 at different environmental conditions have yet to be studied. The objective of this study is to investigate the effect of temperature, fatty acids, ions, and protein concentration on the binding behavior and rates of association and dissociation between the S protein and RBD of SARS-CoV-2 and the hydrophobic aminopropylsilane (APS) biosensors using biolayer interferometry (BLI) validated with molecular dynamics simulation. Our results suggest three conditions—high ionic concentration, presence of hydrophobic fatty acids, and low temperature—favor the attachment of S protein and RBD to hydrophobic surfaces. Increasing the temperature within an hour from 0 to 25 °C results in S protein detachment, suggesting that freezing can cause structural changes in the S protein, affecting its binding kinetics at higher temperature. At all the conditions, RBD exhibits lower dissociation capabilities than the full-length S trimer protein, indicating that the separated RBD formed stronger attachment to hydrophobic surfaces compared to when it was included in the S protein.


Scientific Reports
| (2022) 12:789 | https://doi.org/10.1038/s41598-021-04673-y www.nature.com/scientificreports/ bottle-top filter membrane prior to loading on a Ni-IMAC FF Sepharose column. The Ni-charged IMAC column was equilibrated with 50 mM sodium phosphate with 300 mM NaCl and 20 mM imidazole, pH 7.4. The clarified feed was loaded onto the column at a linear flow rate of 90 cm/h. The column was washed with 2 column volumes (CVs) of equilibration buffer followed by a wash with 2 CVs of equilibration buffer containing 50 mM imidazole. The bound RBD-8His was eluted with 3 CVs of 50 mM sodium phosphate, 300 mM NaCl buffer containing 250 mM imidazole. All steps except for the sample load were done at linear velocity of 150 cm/h. The imidazole removal from the elution pools was performed using desalting column (HiPrep 26/10 Desalting, Cytiva). Purity assessment of purified proteins was performed by analytical size exclusion chromatography (YMC Pack Diol 200 sizing column) using UV 280 detection.

Preparation of samples containing SARS-CoV-2 spike protein. The original IMAC (Immobilized
Metal Affinity Chromatography) purified, ultrafiltration (UF) concentrated, and diluted by diafiltration (DF) into phosphate-buffered saline (PBS) at a concentration of 0.24 mg/mL SARS-CoV-2 spike (S) Streptavidin and His-tagged recombinant protein, is a highly glycosylated trimer (3 × 158 kDa) with a final weight of about 600 kDa. The S protein was resuspended in PBS to prepare 1:2, 1:5, and 1:10 dilutions at 25 °C. Oleic acid is a common fatty acid naturally occurring in animals and vegetables and is easily accessible. Therefore, oleic acid served as the fatty acid simulant in this study. Oleic acid (Spectrum) was mixed with S protein at 1:1 and 4:1 ratio, respectively, in Eppendorf tubes at 25 °C. The 2 × S protein dilution in PBS was transferred into four separate Eppendorf tubes and (1) placed in ice water to cool to 0 °C, (2) inside a refrigerator to reach 9 °C, (3) at room temperature of 25 °C, and (4) heated up to 37 °C on a heat block. One of the samples at 0 °C was left at the room temperature of 25 °C immediately after it was removed from the ice water.

Preparation of samples containing SARS-CoV-2 RBD (receptor binding domain). The purified,
original Fc and His-tagged recombinant RBD of SARS-CoV-2 is the binding portion of a single S unit with a size of about 58 kDa. The concentration of the original RBD was 1.822 mg/mL. RBD was diluted in PBS to prepare 1:2, 1:5, and 1:10 dilutions at 25 °C. The four 2 × RBD dilutions were each exposed to different temperatures at 0 °C, 9 °C, 25 °C, and 37 °C similarly to the S protein.

Determination of the basic kinetics for SARS-CoV-2 spike protein and RBD.
Bio-layer interferometry (BLI) was performed to study the basic kinetics of each of the prepared samples of S protein and RBD of SARS-CoV-2 mixed with various substances or exposed to different temperatures. The major instrument used in the analysis was the personal assay BLItz system (ForteBio) with APS biosensors (ForteBio). The APS biosensors were hydrated in PBS in 96 wells plates for at least 10 min before use. Each run started with a baseline step of 30 s with 4 µL PBS as the buffer. After the baseline step, 4 µL S protein or RBD sample at different concentrations or mixed with different substances were added to the drop holder and allowed to associate with the hydrophobic APS biosensor surfaces for 300 s. Following association, APS sensors with attached S protein or RBD were exposed to 4 µL PBS or Milli-Q (MQ) water for 300 s and the rate of dissociation from the APS surfaces was measured. Each sample was analyzed by the BLI at least twice to ensure reproducibility of the results. An APS biosensor and PBS were used in the same manner before each experiment to serve as the reference correction for each assay. The binding curves were generated and fitted to the local model using the BLItz Pro 1.3 Software (ForteBio). The binding affinity, association rate, and dissociation rate were calculated and displayed in the BLItz Pro 1.3 Software (ForteBio). The kinetic constants were shown on the graphs as mean ± standard deviation.
Statistical analysis. Microsoft Excel 2016 was used to calculate the mean and standard deviation of rates of association and dissociation and to illustrate the data. JMP Pro 16 was used for statistical significance analysis. The one-way Analysis of Variance (ANOVA) was applied to determine the statistical significance among three or more groups. The pooled t-test was used to determine the statistical significance between any two groups. Groups having p-values of less than 0.05 were considered statistically significant differences.

Results and discussion
S protein diluted in PBS dissociating in PBS and in water at 25 °C; 1.569 µM S protein dissociating in water at five temperatures. Based on the concentration and the molecular weight of the original S protein, the molar concentrations for the 1:1 (1x), 1:2 (2x), 1:5 (5x), and 1:10 (10x) PBS diluted S protein were computed in the BLI system to be 3.137 µM, 1.569 µM, 0.6275 µM, and 0.3137 µM. The first set of basic kinetics analysis was performed with the APS sensors dissociating in PBS at room temperature of 25 °C. The strongest binding was observed with 3.137 µM S protein forming a binding layer of 3.39 nm, and the most diluted S protein showed the thinnest binding layer on the sensor of 2.50 nm out of the four dilutions (Fig. 1a). The analysis indicates that the binding activities were stronger with a higher concentration of S protein present in the sample. At the end of the association step, the thickness of the binding layer for 3.  Fig. 1b show that decreasing S protein concentration from 3.137 µM to 0.6275 µM led to increasing association constants, as the less concentrated S protein bound to the hydrophobic surface at a higher rate; the difference between the ka at   Figure 1c shows some but still low dissociation in the presence of ions in PBS from the hydrophobic surface only at the lower protein concentration of 0.3137 µM. All the other three S protein concentrations had minimal to no detachment, meaning that the S protein was not able to resuspend into the ionic environment once it bound to the hydrophobic surface. However, the kd for the four S protein concentrations was not statistically different (p = 0.1843). The basic kinetics analysis was repeated on 3.137 µM, 1.569 µM, 0.6275 µM, and 0.3137 µM S protein with performing the dissociation step with purified water instead of PBS to investigate the role of ions in S protein detachment. The binding curves shown in Fig. 1d shared the similar features as Fig. 1a, with the most concentrated S protein generating the thickest binding layer throughout the analysis. At the end of association, the binding layers for 3.137 µM, 1.569 µM, 0.6275 µM, and 0.3137 µM S protein were 3.29 nm, 3.16 nm, 2.84 nm, and 2.54 nm, respectively. The steep increase in the binding layer indicates that the majority of the binding occurred at the beginning of the association step (Fig. 1d). The ka of 3.137 µM, 1.569 µM, 0.6275 µM, and 0.3137 µM S protein dissociating in water were 1.01E5 Ms −1 , 8.36E4 Ms −1 , 1.95E5 Ms −1 , and 2.43E5 Ms −1 , respectively. From Fig. 1e, the association pattern is similar to the previous behavior in the ionic environment as the least concentrated S protein shows the highest association rate, and the difference between the ka of four concentrations was also statistically significant (p = 0.0011). Although the ka of the 1.569 µM S protein was lower than the 3.137 µM S protein, the difference was small (within one magnitude) and was not statistically significant (p = 0.3089). Therefore, the same observation as in Fig. 1b  To investigate the effect of temperature on the binding of S protein to the hydrophobic APS surface, 1.569 µM S protein at four temperatures-0 °C, 9 °C, 25 °C, and 37 °C-was studied. In addition, 1.569 µM S protein that was previously kept at 0 °C was transferred to room temperature of 25 °C and was compared with the S protein at the same concentration without freezing. This set of experiments was designed to answer the question whether freezing can induce change on the binding of S protein. From Fig. 1g, the binding curves of 1.569 µM S protein at 37 °C and 9 °C showed similar behaviors, as they both bound quickly at the beginning of the association step and the binding became much slower during the rest of the period. On the other hand, the binding of 1.569 µM S protein at 0 °C and 25 °C was slower at the beginning and faster for the rest of the association step, resulting in a continuously increasing curve. The binding curve of the 0 °C to 25 °C 1.569 µM S protein had the similar shape as 37 °C and 9 °C, but with weaker binding activities overall. Although behaviors during association were not all the same for the five temperatures, all the binding distance remained relatively constant during the dissociation step, indicating minimal resuspension of S protein into water from the APS sensor surface. At the end of the association, the binding layers for 1.569 µM S protein at 0 °C, 9 °C, 25 °C, 0 °C to 25 °C, and 37 °C were 3.46 nm, 3.17 nm, 3.16 nm, 2.16 nm, and 3.41 nm, respectively. The ka of 1.569 µM S protein at 0 °C, 9 °C, 25 °C, 0 °C to 25 °C, and 37 °C were 8.96E4 Ms −1 , 1.78E5 Ms −1 , 8.36E4 Ms −1 , 1.50E5 Ms −1 , and 1.34E5 Ms −1 , respectively. From Fig. 1h, 1.569 µM S protein at 0 °C and 25 °C shows a lower rate of association while the 9 °C test shows the highest association. The ka among the S protein at four temperatures-0 °C, 9 °C, 25 °C, and 37 °C-was statistically significantly different (p = 0.006), indicating that temperature impacts the rate of association of 1.569 µM S protein. The ka at 9 °C was significantly higher than that at 0 °C (p = 0.004) and 25 °C (p = 0.007). The highest ka at 9 °C in our study is consistent with the results of Shi et al. stating that the number of COVID-19 cases was the highest when the temperature was at around 10 °C and can be correlated to the higher numbers of infectious cases found in their study 50 . The kd of 1.569 µM S protein at 0 °C, 9 °C, 25 °C, 0 °C to 25 °C, and 37 °C were 2.26E−7 s −1 , 6.44E−5 s −1 , 1.31E−4 s −1 , 6.27E−4 s −1 , and 4.52E−5 s −1 , respectively. 1.569 µM S protein at 0 °C, 9 °C, 25 °C, and 37 °C shows significantly different (p = 0.0066) but low rate of dissociation, indicating strong attachment of the S protein to the hydrophobic surface, especially at 0 °C (Fig. 1i). The kd at 0 °C was significantly lower than those at the higher temperatures of 9 °C (p = 0.0433), 25 °C (p = 0.0126), and 37 °C (p = 0.0043). The low kd value at 0 °C shows strong attachment between the S protein and the hydrophobic surface, indicating high stability. This finding is in agreement with the conclusion of Chin et al. that the virus was highly stable at low temperature at around 4 °C 51 . These results suggest that the S protein is less likely to resuspend from the hydrophobic surface, such as the high amount of fat generated during meat processing, at 0 °C due to the strong attachment. At temperature as low as 0 °C which is commonly found in the chiller rooms of meat processing plants, S protein is more likely to stay entrained in the airflow for longer periods of time once they form the strong attachment to the aerosolized fats. Figure 1i also shows that exposing the 1.569 µM S protein to a wide temperature range from 0 to 25 °C resulted in strong protein detachment. Comparing the two 1.569 µM S protein at 25 °C, the one with previous freezing at 0 °C had higher rate of dissociation than the one without freezing, yet the difference was not significant (p = 0.1407). The previously frozen 25 °C S protein also had significantly (p = 0.0183) higher rate of association as shown in Fig. 1h. This indicates that freezing can cause change in S protein that affects its binding kinetics at higher temperatures. Another study stated that loss of lipids can destabilize the trimeric S protein and prevent receptor binding, however, the infectivity of released virus can be impacted by the nature of lipids that are present at the site of infection 56 . Our results show that the 1.569 µM S protein in oleic acid had smaller kd due to stronger attachment to the hydrophobic APS surface compared to the 0.6275 µM S protein in oleic acid, indicating that the S protein at higher concentrations may undergo conformational changes in fatty acids. Our results suggest that, in addition to the ligands already studied by Shoemark et al. and Carrique et al., oleic acid can also affect the attachment and detachment of S protein from the hydrophobic surface, which is in agreement with their findings, and this impact is concentration dependent. Future studies need to be conducted to investigate the scope of this impact.   Figure 3d shows that all the four RBD dilutions bound rapidly to the APS sensor at the beginning of the association step, and the binding rate became slower for the rest of the step. The binding curve for 26.79 µM RBD barely increased after the rapid binding period, while the other three RBD dilutions continuously bound at a slow rate. At the end of the association, the thickness of the binding layers for 26.79 µM, 13.4 µM, 5.359 µM, and 2.679 µM RBD dilutions were 2.96 nm, 4.29 nm, 3.73 nm, and 2.58 nm, respectively. The rapid increase at the start of the dissociation step was due to random errors in the BLItz system and should not be counted as part of the binding layer. During the dissociation step, 26.79 µM, 13.4 µM, and 5.359 µM RBD dilutions continuously bound to the APS sensor, the binding layers then decreased during the rest of the dissociation step as part of it re-suspended into water. The binding for 2.679 µM RBD dilution increased steadily during the entire dissociation step.  Figure 3e demonstrates a similar trend as in Fig. 3b that decreasing RBD concentrations corresponded to increasing association, and the difference was statistically significant (p = 0.0002). The same uniformly minimal or no dissociation from the hydrophobic surface was shown in Fig. 3f. Compared to the S protein, the ionic environment provided with PBS had less impact on both the ka and kd of the RBD at all levels of protein concentrations.

RBD diluted in PBS dissociating in PBS and in water at
The 13.4 µM RBD was studied at five temperature conditions-0 °C, 9 °C, 25 °C, 0 °C to 25 °C, and 37 °C. The 13.4 µM RBD at 37 °C showed stronger binding activities compared to the other four temperatures (Fig. 3g). All  Figure 3i shows uniformly minimal or no dissociation from the hydrophobic surface at all the tested temperatures. Compared to S protein where temperature impacted the kd significantly and a wide temperature change from 0 to 25 °C induced strong protein detachment, changing the temperature of the environments had less effects on the resuspension of RBD.
S protein without tags diluted in PBS, dissociating in PBS at 25 °C. The recombinant S protein was further purified to remove Streptavidin and His tags and underwent BLI to confirm that the interferometry results in the previous sections were not significantly affected by the tags and can be applied to S protein without tags. The truncated S protein was diluted in PBS to obtain the same molar concentrations used for the recombinant S protein of 3.137 µM, 1.569 µM, 0.6275 µM, and 0.3137 µM. The baseline and dissociation steps were established with PBS. The BLI analysis of the four dilutions was performed at a constant room temperature of 25 °C. The thickest binding layer was formed with the most concentrated no-tagged S protein at 3.137 µM, and the thinnest binding layer was found with the least concentrated no-tagged S protein of 0.3137 µM (Fig. 4a). These results are in agreement with the S protein binding curves in Fig. 1a. The interferometry results suggest the similar positive correlation between the concentrations of S protein and the amounts of ligands binding to the APS sensors, meaning that increasing S protein concentration in the solution leads to stronger binding. At the end of the association, the thickness of the binding layers for 3.137 µM, 1.569 µM, 0.6275 µM, and 0.3137 µM no-tagged S protein were 2.61 nm, 2.34 nm, 2.09 nm, and 1.99 nm, respectively. The shape of the binding curves in Fig. 4a was similar to Fig. 1a, with rapid ligand binding to the sensor at the beginning of the association and minimum to none change for the rest of the analysis. The ka of 3.137 µM, 1.569 µM, 0.6275 µM, and 0.3137 µM no-tagged S protein were 1.37E5 Ms −1 , 1.40E5 Ms −1 , 1.79E5 Ms −1 , and 1.76E5 Ms −1 , respectively. Comparing with Fig. 1b, the no-tagged and tagged S protein had ka values at the same magnitude of 1E5 Ms −1 . The ka values for all four concentrations varied within a small range, with the more concentrated S protein having higher ka values; the difference of ka among the four concentrations was statistically significant (p = 0.032). The kd of 3.137 µM, 1.569 µM, 0.6275 µM, and 0.3137 µM S protein were 6.82E−5 s −1 , 1.38E−4 s −1 , 3.30E−4 s −1 , and 3.28E−4 s −1 , respectively. Compared to the recombinant S protein, the kd for the no-tagged S protein was in general two magnitudes higher, with the smallest kd recorded with the highest concentration of 3.137 µM (Fig. 4c). The difference of kd among the four tested concentrations was statistically significant (p = 0.014). Although higher kd was obtained with the no-tagged S protein, the kd was still low, indicating minimum detachment from the hydrophobic APS surface at all tested protein concentrations. When comparing the ka and kd of the no-tagged and recombinant S protein at the same concentration of 3.137 µM, the difference was neither statistically significant for the ka (p = 0.8304) nor for the kd (p = 0.2947). Since the no-tagged S protein shared similar ka and www.nature.com/scientificreports/ higher but still small kd as the recombinant S protein, it is proved that the tags on the recombinant S protein did not significantly affect the binding affinities to the APS sensors, and the interferometry results of recombinant S protein can apply to no-tagged S protein.

Protein-ligand interaction verification.
The docking pose of the APS ligand to the RBD shows the formation of multiple non-covalent bonds (Fig. 5a). One hydrogen bond and one salt bridge were found between APS and GLU484. The distance of the hydrogen bond was 2.78 Å and the donor-hydrogen-acceptor angle was 171.2° (Fig. 5b). Two hydrophobic bonds were formed with residues PHE490 and LEU492. Two polar interactions were formed with residues GLN493 and SER494. An ionic bond was formed at residue LYS452. This single APS gave a binding affinity of − 1.834 kcal/mol due to the limited binding capacity of the tiny ligands. However, the complex of ligands and proteins showed good close contacts, suggesting that immobilized APS arrays can produce strong binding interactions with RBD. The presence of hydrophobic and ionic residues in the interaction was also in agreement with the results of interferometry of BLI.

Molecular dynamics (MD) simulation of RBD with fatty acids.
To understand the effect of fatty acids on RBD at certain temperatures and ionic concentrations in the aqueous solvents, all-atom MD simulations in explicit solvent models were performed (Fig. 6a). In the solvent model, 0.1 M Na + and Cl − ions were added. Four levels of fatty acid concentrations were tested (0 mM, 7.5 mM, 15 mM and 30 mM) under temperature of 300 K and pressure of 1 atm. In Fig. 6b, RBD protein α-carbon root-mean-square deviation (RMSD) values were monitored throughout the simulation. RMSD is related to the stability of protein structure, and high RMSD values (> 3 Å) indicate that the structure of the protein may have changed during the simulation. In Table 1, the RMSD was increased with the increase of fatty acid concentration. In the control group, the RMSD was 2.08 ± 0.31 Å, suggesting a stable simulation. While, in the group of 30 mM fatty acid, the RMSD increased to 61.40 ± 12.33 Å.
In high concentrations of fatty acids, the RBD structure was changed. The structural change occurred in the first 1 ns, and then the fluctuations decreased to a low level (Fig. 6b). The RMSD value of 66.07 ± 0.07 Å for the second half of the 30 mM simulation also confirmed this observation. The standard deviation decreased from 12.33 to 0.07 Å. The same results were found in simulations of the other concentrations. These results confirmed that all four groups of MD simulations ended up at equilibrium. In Fig. 6c,d, the fatty acid molecules were aggregated after MD simulation. Most of the fatty acid molecules were attached to the hydrophobic surface. The hydrophobic surface area of the RBD increased with the increase of fatty acids. In the control group, hydrophobic surface area was 1812.358 Å 2 , while in the group of 30 mM fatty acid, the hydrophobic surface area was increased to 2053.426 Å 2 . Hydrophobic residues inside the protein came into contact with fatty acids, thereby increasing the hydrophobic surface of the RBD. Meanwhile, the hydrophobic region was occupied by fatty acids, which may reduce the binding ability of RBD proteins.

Conclusion
In this study, three conditions are identified to enhance the attachment of the purified S protein and its RBD to hydrophobic surfaces: high ionic concentration, presence of hydrophobic fatty acids, and low temperature. The S protein exposed to a wide temperature change from 0 °C to 25 °C within one hour results in S protein detachment, suggesting that freezing can cause structural changes in the S protein that affect its binding kinetics after it is recovered at higher temperature. This rapid change of temperature within an hour was applied to simulate the sudden temperature drop which is common in meat processing plants when the workers move between warmer and colder locations in the facility, for example, from the 25 °C breakroom to the chiller or fabrication www.nature.com/scientificreports/ rooms where temperature is kept low, usually < 12 °C, and even lower due to the presence of dry ice containers to keep the products safe during processing. As virus aerosols can also be transported with the airflow through the openings between these locations, they become exposed to the different temperatures. At all the conditions, RBD exhibits lower dissociation capabilities than the full-length S trimer protein, indicating that the separated RBD formed stronger attachment to hydrophobic surfaces compared to when it was included in the S protein. The interaction between RBD of S protein and APS ligand was verified via molecular docking. MD simulation further revealed that the presence of fatty acid molecules has the potential to increase the hydrophobic surface area of RBD, changing its binding ability. The findings of this study implied that certain environmental conditions-low temperature, high humidity, and presence of fatty acids-that are typical in critical infrastructures such as meat processing plants enhance the binding by the S protein and RBD of the SARS-CoV-2 to hydrophobic surfaces. Under such conditions, SARS-CoV-2 is harder to be removed through typical sanitation procedures such as ventilation and hosing due to the enhanced attachment. The findings also suggested that the environmental  www.nature.com/scientificreports/ conditions affect the transmission of SARS-CoV-2. With the presence of fat particles in the air, the binding can form between SARS-CoV-2 Spike and fat aerosols, which are entrained in the ventilated airflow and can travel for a longer distance, increasing the chances of airborne transmission of the virus. The enhanced attachment of the virus to equipment surfaces and workers' clothes makes sanitation challenging and can lead to longer residence time of the virus and impose higher risks to contact transmission. This study helps recommend necessary modifications to sanitation and cleaning procedures in meat processing plants. For example, hosing the floors and workbenches with warm water, heating the surface temporarily before cleaning, and modifying the mode of ventilation to reach a lower humidity can potentially increase the efficiency of removing SARS-CoV-2 from the facilities and providing a safer and cleaner environment to protect workers. Future studies can explore higher S protein concentrations and intermediate temperatures between 0 and 37 °C to further delineate the binding kinetics of the virus proteins. The MD simulation can be performed in the future on S protein and RBD at 0 °C to assess any structural changes in low temperature environments.

Data availability
All data generated or analyzed during this study are included in this published article [and its Supplementary information files].