GoKawiilEthics statement
The use of human sera for this study was approved by the Scientific and Ethics Review Unit of the Kenya Medical Research Institute (protocol SSC 3426). Before the blood draw, donors gave individual consent for the use of their samples for research.
Construction of gene libraries
A schematic of the downstream analysis pipeline used for retrieving alphacoronavirus data and for generating the final spike protein library used in this study (n = 40) is provided in Fig. 1a. All publicly available alphacoronavirus genome sequences were retrieved from the Virus Pathogen Database and Analysis Resource platform hosted by the Bioinformatics Resource Center at the National Institute of Allergy and Infectious Diseases15. As of May 2021, the full database consists of 19,082 alphacoronavirus genomes, from which we extracted sequences of the whole spike protein-coding region to obtain a final database of 2,714 sequences. We constructed the spike protein-coding DNA sequence alignment using MAFFT (v.7.526)49,50 by integrating structural alignments of homologous spike protein structures queried from the UniProt Reference Clusters51. Maximum-likelihood phylogenetic reconstruction was performed using IQTREE (v.2.3.4)52 with 1,000 ultrafast bootstrap replicates (UFBoot) and 1,000 SH-like approximate likelihood ratio tests53 (Supplementary Fig. 20). We performed codon model reconstruction by determining the best-fit model using the model selection procedure implemented in IQTREE. Patristic distances (the sum of branch lengths along the shortest path) were then computed between all pairs of tips in R (v.4.4.1) using the ape package54, which informed the unbiased selection of the n = 40 spike protein-coding sequences by applying a greedy algorithm. In brief, let δ(i,j) denote the patristic distance between tips i and j on the reconstructed maximum-likelihood tree, with the branch length estimated in substitutions or sites, and let k be a previous number of tips to be selected. The greedy algorithm identifies the farthest pair arg max i,j δ(i,j) and initializes the final selection set S with these two tips. Next, for each subsequent selection step, it adds the tip x∉S that maximize its nearest-neighbour distance to the current set x = arg max x∉S min y∈S δ(i,j) and repeats the process until |S| = k. Although heuristic, this approach ensures that an optimal subset of tips (evolutionary units) is returned under the assumption of maximizing both minimum phylogenetic distance and phylogenetic diversity, as previously theoretically proposed and discussed14. We report Faith’s phylogenetic diversity55 of the induced minimal subtree and benchmarked against a 10,000 random panels of matching size. For the most divergent alphacoronaviruses, a RBD could not be readily identified for two unclassified viruses and viruses classified in the Soracivirus and Luchacovirus subgenera.
Plasmids used for pseudotyping
Selected alphacoronavirus spike protein-coding sequences were ordered from Biobasic as codon-optimized synthetic genes and subcloned in pcDNA3.1 with a HA tag in the C-terminal cytoplasmic tail. For the APN library, genes were ordered from GenScript and subcloned with a N-terminal V5 tag in the vector pCAGGS. For the ACE2 library and human DPP4 (GenBank: NP_001926), the ectodomains of the ORFs were subcloned with a N-terminal HA tag in pDisplay. CEACAM libraries were obtained from GenScript and subcloned in pcDNA3.1(+) with a C-terminal 8×His tag. Flag-tagged human TMPRSS2 (GenBank: NP_001128571.1) and human DPEP1 (GenBank: NP_001121613.1) were cloned into pcDNA3.1. GenBank accession numbers of the sequences used for the study are provided in Supplementary Table 1 for the alphacoronavirus spike protein library, Supplementary Table 2 for APN, Supplementary Table 3 for ACE2, Supplementary Table 7 for human CEACAM receptors, Supplementary Table 9 for the CcCoV local phylogeny spike protein library and Supplementary Table 10 for CEACAM6-like proteins of different mammalian species.
Cells
Mycoplasma-free HEK293T (human kidney), Calu3 (human lung), Caco2 (human colorectal adenocarcinoma) and HuH7 (human hepatoma) were cultured in Dulbecco’s modified Eagle medium (DMEM) supplemented with 10% FBS, 100 U ml–1 penicillin plus 100 µg ml–1 streptomycin (PenStrep) and 1 mM sodium pyruvate. Mycoplasma-free THP1 (human monocytes) and LCL (human B lymphocytes) cell lines were maintained in suspension in RPMI1640 medium supplemented with 10% FBS and PenStrep. All reagents for cell culture were purchased from Gibco. All cells were cultured in a humidified atmosphere at 37 °C, 5% CO 2 .
Pseudotype virus production
To pseudotype alphacoronavirus spike proteins, plasmids encoding their ORF were transfected in confluent HEK293T cells seeded in a 6-well plate using polyethylenimine (PEI, 5 µg ml–1). HIV-1-based lentiviral vectors coding for viral structural proteins (p8.91), particle packaging signals and a luciferase reporter gene (pCSFLW) were also included in the transfection mix56. The following day, the medium was replaced and pseudoviruses in the supernatant were collected 48 and 72 h after transfection and pooled together. Following the final collection, the supernatant was centrifuged for 10 min at 4,000 rpm to remove cellular debris. Finally, pseudoviruses were aliquoted and stored at −80 °C until further use. To verify spike protein incorporation, pseudoparticles were purified by ultracentrifugation at 23,000 rpm, 4 °C for 2 h using a 20% sucrose gradient. Supernatants were discarded and pellets were resuspended in PBS. Concentrated pseudoparticles were lysed by boiling with Laemmli (Bio-Rad) and spike protein expression was analysed by SDS–PAGE. Separated proteins were transferred onto a 0.45 µm nitrocellulose membrane (Cytiva), blocked in PBS supplemented with 0.05% Tween-20 and 5% (w/v) unskimmed milk powder and incubated with mouse monoclonal anti-HA (clone 6E2, Cell Signaling Technology, 1:5,000) and anti-p24 (clone 5, Abcam, 1:2,000) overnight at 4 °C. The following day, goat anti-mouse DyLight 680 (Invitrogen, 1:10,000) was used to probe primary antibodies, and signals were detected with an Odyssey DLx imaging system (Li-Cor Biosciences). Of note, for those spike proteins for which we saw entry (Fig. 1; HCoV-NL63, CcCoV-SD F3, CcCoV-A76, HCoV-229E, BtCoV-AT1A F41 and BtCoV-WA1087), protein expression in the purified pseudoparticle immunoblots (Extended Data Fig. 1) did not quantitatively correlate with entry signals, which indicated that it is difficult to establish a minimum threshold of spike protein incorporation needed for membrane fusion. For the six spike proteins that did not pseudotype (Supplementary Fig. 1), we attempted to substitute them for spike proteins that have been reported to pseudotype and that share at least 95% amino acid similarity (Supplementary Fig. 21). However, the only match was for PEDV-KDJ, which was replaced by PEDV-Colorado57. The selected HCoV-229E spike protein sequence pseudotyped to low levels and was not functional in our downstream applications (Supplementary Fig. 3a); therefore, we replaced it with a different strain of the same species. Original uncropped immunoblots are provided in the Zenodo repository (https://doi.org/10.5281/zenodo.17951484)58.
... continue reading