The structure of tuftsin was found in PubChem (https://pubchem.ncbi.nlm.nih.gov/). The 3D structure of tuftsin was built using Chem3D. Afterward, the target proteins corresponding to tuftsin screened from the Pharmmapper database and PubMed database were standardized in UniProt (http://www.uniprot.org/). Finally, Cytoscape 3.8.2 was used to determine the drug-target network. COVID-19-related genes were mined from the GeneCards database. All of the disease gene targets were normalized with R software using the Bioconductor package when redundancy was deleted (Yu et al., 2012).

Screening for drug-disease crossover genes was performed. Based on previous steps, two sets of target lists were prepared: drug targets and disease-related genes. The crossover genes were filtered with R software using the Venn Diagram package. The STRING 11.5 database (http://string-db.org/) was used to analyse the intersecting protein–protein interactions (PPIs), and the common targets were counted with R software.

The proteins with overlapping expression patterns were evaluated by bioinformatics annotation with R software using the Bioconductor package, including a panther classification system (http://www.pantherdb.org/), a gene ontology (GO) annotation database website (http://www.geneontology.org), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis (http://www.genome.jp/kegg/). A p < 0.05 was considered statistically significant.

The flexible docking process between tuftsin and target proteins was conducted by softwere Discovery Studio 2021 (DS). Briefly, the crystallographic structures of human ACE2 (PDB ID: 1R42) and human NRP1 (PDB ID: 2QQ1) with high resolution were prepared using the Prepare Protein and Minimization module of DS. The active binding site of each protein was defined based on the most representative features of the SARS-CoV-2 interface. Tuftsin was docked into the active binding site of ACE2 and NRP1 using the molecular docking module in DTS.

The recombinant human ACE2 protein (Novoprotein, Beijing, China) and recombinant human NRP1 protein were used for surface plasmon resonance (SPR) analysis using a Biacore T200 instrument (Biacore, Uppsala, Sweden). Each target was immobilized onto flow cells in a CM5 sensor chip (GE Healthcare) via the amine-coupling method. Briefly, ACE2 and NRP1 were diluted in 10 mM pH 4.5 acetate to 20 μg/ml. Then, the protein solutions were injected individually on the carboxyl-modified sensor surface to form amine bonds. Both ACE2 and NRP1 immobilized levels were approximately 10,000 RU. Binding analyses were carried out at 25°C and a flow rate of 30 µl/min. Tuftsin diluted in running buffer (1×PBS, 0.05% Tween 20 and 5% dimethyl sulfoxide, pH 7.4) was run over each target at gradient concentrations. An empty flow cell without any immobilized protein was used as a deducted reference. The binding curves were analysed using the steady state affinity model supplied with Biacore Evaluation Software (GE Healthcare).

For the competition binding experiment, the SARS-CoV-2 S1 protein was immobilized on the CM5 sensor chip via the amine-coupling method. 5 nM ACE2 was injected for negative control. Tuftsin was diluted into a series of solutions with gradient concentrations and fixed with 5 nM ACE2, and then the solutions were injected into the chip. The blocking efficacy was evaluated by comparison of response units with and without tuftsin incubation.

Partial results were analysed using chi-square test with SPSS software and R 4.1.0.

The 2D structure of tuftsin was obtained from the PubChem database (Compound CID: 156080), and the most stable 3D structure was built based on the 2D structure through a molecular simulation assay (Figure 1A). In addition to the reported receptors of tuftsin, the potential targets of tuftsin in human body were predicted through the PharmMapper database. As a result, 284 targets of tuftsin were selected (Figure 1B and Supplementary Data S1), and 2,572 disease-associated genes of COVID-19 were excavated from the GeneCards database (Supplementary Data S2). It is surprised to find that there exists 91 intersecting proteins between tuftsin targets and COVID-19-associated gene-coded proteins through intersection analysis, and the overlapping proteins account for nearly one-third of tuftsin targets (Figure 1C). The protein–protein interaction network of the overlapping proteins was established, and the result indicated that JAK2, STAT1 and AKT1 are core molecules in the network (Figure 1D). The 91 intersecting genes were further studied by enrichment analysis. GO and KEGG annotation revealed that the expressed tuftsin-COVID-19 crossover proteins are mainly associated with immune functions such as neutrophil activation, neutrophil-mediated immunity and cytokine receptor binding, and the COVID-19 pathway is most significantly enriched. In addition, many target genes were strongly associated with some immune pathways, such as Th17 cell differentiation, the IL-17 signaling pathway and the immune checkpoint pathway (Figure 1E). In the COVID-19 pathway, the SARS-CoV-2 receptors ACE2 and NRP1, some immune molecules such as IL-2, STAT1, and some complement molecules are the targets of tuftsin (Figure 1F). These results suggest that tuftsin is a promising candidate against COVID-19 owing to its multifaceted pharmacological activities.

Having proved by the above pharmacology analysis of tuftsin, it is possible that ACE2 is a target of tuftsin. Therefore, the potential binding sites and binding affinity of tuftsin with the SARS-CoV-2 receptors ACE2 and NRP1 were further determined by molecular docking analysis. The interaction interface of SARS-CoV-2 S1 receptor-binding domain (RBD) with ACE2 was defined as the active sites of ACE2. These interface sites in ACE2 include Q24, M82, N330, and R393 residues, which are mainly located in the N-terminal peptidase domain of ACE2 (Lan et al., 2020). The docking region of ACE2 is a sphere containing the defined ACE2 active sites (Supplementary Figure S1A). The docking results showed that the affinity of tuftsin with ACE2 was −6.9 kcal/mol, demonstrating that they could combine spontaneously (Figure 2A). Furthermore, tuftsin could form strong hydrogen bonds to Ser47 and Asp67, hydrogen bonds to His345, Asp67 and Asn51, and salt bridges to Asp67 residues of ACE2 (Figure 2B). In addition, the bond lengths of the interacted sites are that 2.110 Å of Ser47, 1.854 Å of Asp67, 2.814 Å of His345 and 2.425 Å of Asn51 (Figure 2C). It is worth mentioning that the binding sites were adjacent to the interactional sites of S1-RBD and ACE2 (Lan et al., 2020), indicating that tuftsin could inhibit S1 binding to ACE2 by covering their binding sites. It is reported that the extracellular b1b2 domain of NRP1 combines with S1 CendR peptides (Plein et al., 2014; Daly et al., 2020), therefore the b1b2 domain of NRP1 was prepared for docking. The interactional sites of S1-RBD with b1b2 proteins were defined as the active sites of NRP1 b1b2 domain, including D320, E348, Y353 residues and so on (Daly et al., 2020). Subsequently, the docking region was a sphere containing these defined active sites in the b1b2 domain of NRP1 (Supplementary Figure S1B). The docking results showed that tuftsin and NRP1 b1b2 domain have a high binding affinity of −8.1 kcal/mol. In addition, tuftsin fits solidly into a binding pocket on NRP1 b1b2 domain (Figure 2D). Furthermore, tuftsin could form a salt bridge to Lys397 and a carbon-hydrogen bond to Pro398 residues of NRP1 (Figure 2E), which are near the interactional sites of S1-RBD with NRP1 b1b2 domain. The shortest bond lengths of these interacted sites are that 1.833 Å of Lys397 and 2.394 Å of Pro398 (Figure 2F). Moreover, the binding region of tuftsin and NRP1 overlapped with the binding area of NRP1 with S1-RBD in space. Collectively, these results demonstrated that tuftsin could bind ACE2 and NRP1, and inhibit the SARS-CoV-2 S1 binding to ACE2 and NRP1 by covering their interactional sites.

The interactions of tuftsin with ACE2 and NRP1 were further evaluated by real-time biomolecular interaction analysis with SPR. The steady state affinity of the binding reaction were determined by injecting different concentrations of tuftsin over recombinant human ACE2 immobilized on one half of the chip surface, and over recombinant human NRP1 immobilized on another half of the chip surface. The results showed that tuftsin can bind ACE2 with an equilibrium dissociation constant (K _D) of 460 μmol/L (Figure 3A). Moreover, the K _D fitting curve of tuftsin with ACE2 is gradually stable when the concentrations of tuftsin increased gradually, indicating that the interaction of tuftsin and ACE2 is specific (Figure 3A). Tuftsin can also bind NRP1 strongly with a higher binding affinity with K _D value calculated as 10.65 μmol/L (Figure 3B). The K _D fitting curve of tuftsin with NRP1 is gradually stable when the concentrations of tuftsin increased gradually, conforming to the characteristic of specific binding (Figure 3B). As SPR is the gold standard for detecting drug-target interactions, these results demonstrate that tuftsin binds ACE2 and NRP1 directly and specifically with ideal affinities, and accord with the accuracy of the previous results of bioinformatics analyses and molecular docking assays.

Whether tuftsin could affect the binding of S1 protein with ACE2 was further determined by SPR-based competition assay. The binding affinity of the S1 protein with ACE2 was firstly determined by SPR assay, which unsurprisingly showed a high affinity. A suitable concentration ACE2 solution was injected over the immobilized SARS-CoV-2 S1 protein as a control. A series of gradient concentrations of tuftsin solutions containing equal concentrations of ACE2 were injected over the immobilized SARS-CoV-2 S1 protein for comparison. As a result, 9 μmol/L tuftsin has a mild inhibitory effect. It is worth noting that the addition of 156 μmol/L tuftsin significantly attenuates the response signal about two-thirds compared to that of ACE2 alone over the immobilized S1. Notably, a substantial decrease in the response signal was observed with increasing concentrations of tuftsin. The response single is declined approximately to zero when the added concentration of tuftsin was 625 μmol/L. This result indicates that the interaction between S1 and ACE2 is completely abrogated in the presence of 625 μmol/L tuftsin (Figure 4). The experiment was repeated three times independently. In conclusion, the competition binding experiment revealed that tuftsin effectively impairs the binding of SARS-CoV-2 S1 to ACE2 in a dose-dependent manner.