As part of our overall field campaign, we collected fecal samples from 240 Rhinolophus pusillus bats, 20 each month, during May 2022–April 2023. We captured only Rhinolophus spp. bats and released other bat species immediately after capture. A trained veterinarian collected all the samples after anesthetizing the bats, and all the bats were released at the site of capture within 2–5 hours of capture.

We assessed sex, weight, and health of individual bats before collecting a fecal sample. For our sequencing studies, we selected samples from a mix of female, male, juvenile, and adult bats with different body condition scores. We chose bats with higher leukocyte counts if those data were available. We inactivated the samples by using TRIzol (Thermo Fisher Scientific, https://www.thermofisher.com) and selected a subset of 12 fecal samples from summer 2022 (Table), representing 2 roosting locations (sites 1 and 2) in Bandarban, for initial screening on an iSeq (Illumina, https://www.illumina.com). We then performed deep sequencing on those samples by using a NextSeq 500 (Illumina). We enriched libraries by using the Comprehensive Viral Research Panel (Twist Bioscience, https://www.twistbioscience.com) supplemented with a custom-designed Chiropteran virus enrichment panel containing 134,000 probes. All sequencing was performed in a US Biosafety Level 4 facility. We deposited data into the National Center for Biotechnology Information Sequence Read Archive (submission no. SUB15226189 and BioProject no. PRJNA1249517).

We identified coronavirus sequence reads in 3 samples: B3, B4, and B6. The strongest signal was in B4, which comprised 1.2% of total reads and 5.5% of classified reads in the entire B4 sample (Table; Appendix Figures 1–3). We characterized virome components by using KRAKEN2 (4) and RefSeq viral database version April 2023 (Illumina). The B4-derived coronavirus sequences had 76.7% BLASTn (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=blastn&PAGE_TYPE=BlastSearch&LINK_LOC=blasthome) identity to SARS-CoV-1 Tor2 (Appendix Figure 4), which was isolated in 2003 from a patient who traveled from Hong Kong, China, to Toronto, Ontario, Canada, and who was hospitalized with febrile respiratory illness (5).

We further compared B4 to Tor2 and relatives by creating similarity plots in SimPlot++ version 1.3 (https://github.com/Stephane-S/Simplot_PlusPlus), and creating maximum-likelihood phylogenetic trees aligning the whole genome, RNA-dependent RNA polymerase, and spike sequences by using MAFFT version 7.508 (https://mafft.cbrc.jp/alignment/software) and IQ-TREE version 2.3.6 (http://iqtree.cibiv.univie.ac.at) (Appendix Figures 4, 5) (6–11). We found 2 notable dropouts in the alignment to SARS-CoV-1 Tor2: a 1-kb gap at nonstructural protein (NSP) 2 and a 2.1-kb gap over most of the spike receptor-binding domain (RBD) (Appendix Figure 4). To obtain complete genome coverage, we designed 2 primer pairs for each gap and used those primers to generate and sequence amplicons (Appendix Table 1). Initial amplicon analysis using BLASTn provided GenBank accession no. KY417143.1, bat SARS-like coronavirus isolate RS4081, which shared 85%–94% identity over a 99% query length. The main difference was the spike region (85% identity), which had no hits for 214 nt. A BLASTx (https://blast.ncbi.nlm.nih.gov/Blast.cgi?PROGRAM=blastx&PAGE_TYPE=BlastSearch&LINK_LOC=blasthome) query of the spike-specific amplicon indicated 76.2% identity (182 mismatches and 8 gaps) over a 97% query length to GenBank accession no. QVN46559.1, a spike glycoprotein from bat SARS-like coronavirus Khosta-1. The top BLASTx hit for the NSP2-specific amplicon was protein sequence NP_828861.2, an NSP2 of SARS-CoV-1 Tor2, which had 72.5% identity over a 99% query length. Further phylogenetic analysis of whole genomes, spike, and RNA-dependent RNA polymerase supported a novel virus with close spike homology to Sarbecovirus spp. (Appendix Figure 5).

We used AlphaFold modeling (European Molecular Biology Laboratory, European Bioinformatics Institute, https://alphafold.ebi.ac.uk) to compare the RBD of B4 with SARS-CoV-1 Tor2 RBD (Appendix Figure 6). Folding indicated similar shape and functionality and exhibited nonsynonymous substitutions and insertions. Two insertions were asparagine dimers, located on an edge likely to interact with mammalian angiotensin converting enzyme 2 (ACE2), and 1 insertion was a threonine located on another edge, making the B4 RBD sample structurally close to a sample from a known zoonotic human outbreak.

Figure 2

Figure 2. Binding efficiency and fold change of SARS-CoV-1–like virus viruses in Rhinolophidae bats, Bandarban Region, Bangladesh. A) Heatmap depicting binding efficiency of receptor binding domain (RBD) sequences from B4 and regional...

The binding of virus RBD to the primary receptor ACE2 is necessary for spillover infection to occur. We used a synthetic Förster resonance energy transfer–based assay (12) to test the binding affinity of known RBDs and the B4-derived RBD from our genomic surveillance data (Appendix Figure 7). We used ACE2 receptors from a variety of sympatric mammals (Appendix Table 2), including species that might reside near our bat sampling sites, such as Rattus spp. rats, Leopard cats (Prionailurus bengalensis), and humans (13). We chose ACE2 of the Etruscan shrew (Suncus etruscus) , which had sequence available for protein derivation, as a representative Suncus species for testing. That species has not specifically been observed yet in Bandarban, but its close relative, the S. murinus shrew, is widespread there and throughout Bangladesh. Dissociation constants for the novel B4 and 10 other bat coronavirus RBDs showed moderate binding of B4 to several native sympatric animals (Figure 2).

We were also interested in the potential of the Tor2 homologous B4-derived virus to infect humans. Therefore, we evaluated binding affinity relative to the 2019 wild-type SARS-CoV-2 RBD (Figure 2). The B4-derived RBD demonstrated approximately one third the binding efficiency of the wild-type strain, which was similar to results for other tested bat coronaviruses not yet detected in humans.

In 2019, a Tor2 analog was described in a bat reservoir in Korea (14), indicating the viral homologue may be regionally widespread from Bangladesh to southern China and the Korean Peninsula. Abiotic stress including human land use is known to stress bat health and drive them closer to potential transitional hosts, a process implicated in spillover of other viruses (2,3).

We report a coronavirus in bats in Bangladesh that has high similarity to SARS-CoV-1 Tor2, isolated in 2003 from a febrile patient who had secondary exposure to a person who contracted coronavirus from an environmental source in southern China (5). The virus detected in Bandarban, Bangladesh, and sequenced and analyzed in this study shares identity with Tor2, except in the NSP2 and RBD genomic regions. The synthetically expressed RBD shows moderate binding affinity to ACE2 receptors of nearby species, suggesting potential for infection of co-occurring taxa within the host range. Additional study is needed to elucidate what drives host viral shedding and if spillovers are occurring that pose a public health risk. Describing the genetic diversity and transmission potential of this and other potentially zoonotic pathogens can aid in identifying sources and risk of future emerging spillovers.