SARS-CoV-2-related bat viruses evade human intrinsic immunity but lack efficient transmission capacity

Abstract

Circulating bat coronaviruses represent a pandemic threat. However, our understanding of bat coronavirus pathogenesis and transmission potential is limited by the lack of phenotypically characterized strains. We created molecular clones for the two closest known relatives of SARS-CoV-2, BANAL-52 and BANAL-236. We demonstrated that BANAL-CoVs and SARS-CoV-2 have similar replication kinetics in human bronchial epithelial cells. However, BANAL-CoVs have impaired replication in human nasal epithelial cells and in the upper airway of mice. We also observed reduced pathogenesis in mice and diminished transmission in hamsters. Further, we observed that diverse bat coronaviruses evade interferon (IFN) and downregulate MHC-I. Collectively, our study demonstrates that despite high genetic similarity across bat coronaviruses, prediction of pandemic potential of a virus necessitates functional characterization. Finally, the restriction of bat coronavirus replication in the upper airway highlights that transmission potential and innate immune restriction can be uncoupled in this high-risk family of emerging viruses.

Editor summary:

Characterizing infection, pathogenesis, and transmission of BANAL-52 and BANAL-236 in primary respiratory cells, mice and hamsters shows how viruses closely related to SARS-CoV-2 present a threat for spill-over.

Introduction

Circulating animal coronaviruses (CoVs), particularly from bats, pose a substantial risk of zoonotic transmission events in humans. However, the vast majority of these spillover events do not result in sustained human-to-human transmission¹. Several SARS-CoV-2-related CoVs have been identified, predominantly in Rhinolophus bat species^2–10 but most have not been cultured, leaving their zoonotic potential unknown.

CoV genomes encode multiple non-structural, structural, and accessory proteins. The spike protein mediates viral entry and has two subunits (S1 and S2). The S1 subunit consists of N-terminal and C-terminal domains (NTD and CTD), and both can be used as receptor binding domains (RBDs)¹¹. Spike undergoes two proteolytic events, which prime and activate it. Priming occurs at the S1/S2 junction and can be mediated by host proteases including TMPRSS2, furin, and/or cathepsins. Activation of spike by cleavage within S2 promotes membrane fusion. After entry, the viral genome is directly translated and replicated to produce virus progeny¹². While spike is a major determinant for tropism and pathogenesis¹³, non-structural and accessory proteins also play a key role by modulating the host’s innate immune response^14,15.

Although hundreds of circulating bat CoVs have been identified¹⁶, just a few been successfully cultured^2,17–25. Recently, a cluster of bat CoVs termed BANAL-CoVs were sequenced from Rhinolophus bats². Three of these BANAL-CoVs are the closest known relatives to SARS-CoV-2. However, only BANAL-236 was successfully cultured. Other BANAL-CoVs, including BANAL-52, were sequenced but not isolated.²

A zoonotic pandemic in humans requires that a virus can successfully enter human cells, counteract innate immunity, evade pre-existing cross-reactive adaptive immunity, transmit efficiently between people, and cause disease. Like SARS-CoV-2, BANAL-CoVs use human ACE2 to enter cells². However, whether BANAL-CoVs can overcome human innate and adaptive immunity and undergo efficient transmission represents a crucial knowledge gap. Characterizing and defining barriers to efficient human infection is critical to better predict the zoonotic potential of bat CoVs and to enhance pandemic preparedness²⁶. In this study, we generated infectious molecular clones and reporter viruses of BANAL-236 and BANAL-52 to investigate their potential for immune evasion and transmission.

Results

Bat SARS-like viruses efficiently replicate in cell culture models.

BANAL-CoVs exhibit ~97% genetic identity with SARS-CoV-2 across the entire viral genome (Fig. 1A). Within the NTD of spike, BANAL-52 has ~98% identity while BANAL-236 has ~60% identity to SARS-CoV-2 (Extended Data Fig. 1). We synthesized molecular clones of BANAL-52 and BANAL-236 (Fig. 1B). Synonymous mutations were introduced to remove restriction sites as specified in Supplementary Table 2. We synthesized, subcloned, digested, ligated, and in vitro transcribed BANAL-CoV cDNA. (Extended Data Fig. 2A and 2B) ^27–29.

Figure 1. Molecular clones of BANAL-CoVs exhibit similar kinetics and drug susceptibility to SARS-CoV-2.

(A) Phylogenetic tree of sarbecoviruses based on whole genome sequences. (B) Illustration of the reverse genetics design for full-length BANAL-CoVs with and without mNeonGreen reporter. Created with BioRender.com. Viral growth curves of full-length bat sarbecoviruses in (C) Vero-ACE2/TMPRSS2 cells, (D) human bronchial epithelial cells in air liquid-interface, and (E) human nasal epithelial cells in air liquid-interface. Panel C is representative from two independent experiments with n=3 biological replicates. Panels D and E are pooled from two independent experiments. Panel D has 4 biological replicates at each timepoint, while panel E has 4 biological replicates at 1 hpi, and 6 biological replicates at 24, 48 and 72 hpi. Error bars illustrate standard deviation. Significance between viral titers was assessed by one-way ANOVA corrected for multiple comparisons, * indicates p<0.05, **** indicates p<0.0001 between SARS-2 and BANAL-CoVs. Exact p values for panel E are: for 24 hpi (SARS-2 vs BANAL-52 p=.0006, SARS-2 vs BANAL-236 p=.0183), for 48 hpi and 72 hpi (SARS-2 vs BANAL-52 and SARS-2 vs BANAL-236 p<.0001). (F-G) Inhibition of the three sarbecoviruses by remdesivir and nirmatrelvir assessed at 24 hpi (MOI 0.1) as determined by quantitative image analysis of percentage of cells expressing mNG. (H-J) Inhibition of the three sarbecoviruses by two doses of Pfizer vaccinated (H), Moderna vaccinated (I), or convalescent (J) pooled sera assessed at 24 hpi (MOI 0.1) as determined by quantitative image analysis of percentage of cells expressing mNG. Curves from panels F-J were made by non-linear regression analysis and shading reflects 95% confidence intervals. In panels F-J one out of two independent experiments performed in triplicate are shown. For all panels, data are represented by mean values +/− SD.

Passage one stocks from BANAL-52, and BANAL-236 and two control bat CoVs WIV1 and SHC014 exhibited titers above 10⁷ PFU/mL (Extended Data Fig. 2C)^18–19. This represents the first successful culture of BANAL-52. As our primary goal was to assess the pathogenesis and zoonotic potential of BANAL-CoVs, we directly compared BANAL-CoVs to the prototypic SARS-CoV-2/WA01, which was isolated shortly after the virus spilled over into humans. All four bat CoVs exhibited similar replication kinetics to SARS-CoV-2 in Vero-ACE2/TMPRSS2 cells (Fig. 1C).

To evaluate the replication kinetics of bat CoVs in a more physiologically relevant model, we infected primary human bronchial epithelial cells (HBECs) grown at air-liquid interface (ALI) with SARS-CoV-2, BANAL-52, and BANAL-236. Consistent with Vero-ACE2/TMPRSS2 results, we observed similar replication kinetics between SARS-CoV-2 and BANAL-CoVs (Fig. 1D). These results highlight that the two closest relatives of SARS-CoV-2 circulating in bats can infect primary human airway cells without further adaptation.

BANAL-CoVs inefficiently replicate in nasal epithelial cells.

We next examined the replication kinetics of BANAL-CoVs and SARS-CoV-2 in primary human nasal epithelial cells (HNECs) cultured at ALI. In contrast to HBECs, BANAL-CoV replication in HNECs was reduced relative to SARS-CoV-2 (Fig. 1E). At 24 h post-infection (hpi), we observed ~100 fold less replication of the BANAL-CoVs compared to SARS-CoV-2. At 48 hpi, we observed a decrease of ~1,000 and ~10,000 fold for BANAL-236 and BANAL-52, respectively, compared to SARS-CoV-2. These data suggest that BANAL-CoVs may replicate poorly in the human upper airway, potentially impacting transmission efficiency.

BANAL-CoVs are susceptible to SARS-CoV-2 small molecule inhibitors and neutralizing sera.

To facilitate high-throughput assessment of susceptibility of BANAL-CoVs to antiviral drugs, we generated infectious molecular clones of BANAL-CoVs expressing the fluorescent reporter mNeonGreen³⁰. (Fig. 1B). Viral titers were comparable in Vero-ACE2/TMPRSS2 cells between WT and mNG-expressing BANAL CoVs (Extended Data Fig. 2D). We confirmed mNeonGreen expression in infected Vero-ACE2/TMPRSS2 (Extended Data Fig. 2E). To address the sensitivity of BANAL-CoVs to SARS-CoV-2 antivirals, we compared the antiviral activity of FDA-approved drugs remdesivir and nirmatrelvir. Remdesivir had similar IC₅₀ values against SARS-CoV-2, BANAL-236, and BANAL-52 (Fig. 1F). Both BANAL-CoVs were susceptible to nirmatrelvir, although we observed slightly less potent IC₅₀ values (∼4 fold) against BANAL-CoVs relative to SARS-CoV-2 (Fig. 1G).

Next, we tested the ability of sera from people naturally infected with SARS-CoV-2 or vaccinated with Pfizer and Moderna mRNA vaccines to neutralize BANAL-CoVs. BANAL-CoVs were efficiently neutralized by sera from both previously infected and vaccinated individuals as measured by mNeonGreen reporter virus (Fig. 1H–1J) and virus-induced cell death (Extended Data Fig. 2F–2H). A subtle resistance of BANAL-52 compared to SARS-CoV-2 and BANAL-236 (less than 10-fold) was observed, but all three viruses were neutralized above a dilution 1:40. This suggests that existing antibody-based immunity to SARS-CoV-2 can confer protection against BANAL-CoVs.

Host protease dependence of BANAL-CoVs.

Next, we examined whether BANAL-CoVs possess an altered dependence on cellular proteases for cell entry. BANAL-CoVs and SARS-CoV-2 had similar susceptibility to the cathepsin inhibitor E64d in WT Vero-E6 cells, which lack TMPRSS2 expression (Extended Data Fig. 3A). The lack of a furin cleavage site (FCS) decreases sarbecovirus dependence on TMPRSS2³¹. We used camostat mesylate (TMPRSS2 inhibitor) and E64d (cathepsin inhibitor) to examine the impact on BANAL-CoVs and SARS-CoV-2 infection in Vero-ACE2/TMPRSS2 cells. Interestingly, BANAL-CoVs exhibited higher dependence on TMPRSS2 for entry demonstrated by the reduced camostat mesylate IC₅₀, ~20 fold lower for BANAL-52 (5 μM) and ~10 fold lower for BANAL-236 (13 μM) compared to SARS-CoV-2 (98 μM) (Extended Data Fig. 3B). BANAL-CoVs also exhibited modestly more dependence on the endocytic pathway for viral entry in Vero-ACE2/TMPRSS2 cells (Extended Data Fig. 3C). To assess protease dependence in physiologically relevant cells, we pre-treated HBECs with camostat mesylate and E64d. The three CoVs were similarly inhibited by camostat mesylate (Extended Data Fig. 3D) but not E64d in HBECs (Extended Data Fig. 3E). This demonstrates BANAL-CoVs can efficiently use proteases involved in both entry pathways.

Bat CoVs have enhanced evasion of IFN-mediated restriction in vitro.

Given the role of innate immunity in cross-species transmission and pathogenesis, we evaluated the ability of BANAL-CoVs to counteract human IFN-mediated antiviral defenses. All four bat CoVs possessed enhanced resistance to type I IFN mediated restriction in Vero-ACE2/TMPRSS2 cells, as measured by viral titers at 24 hpi (Fig. 2A). The bat CoVs, displayed at least ∼50-fold higher infectious titers than SARS-CoV-2 across all IFN-α concentrations tested. The IFN-α IC₅₀ for all bat CoVs was ~7- to 20-fold higher than SARS-CoV-2 in Vero-ACE2/TMPRSS2 cells (Fig. 2A).

Figure 2. Bat SARS-related viruses evade human innate immunity and downregulate MHC-I.

(A) Infectious viral titers from Vero-ACE2/TMPRSS2 cells pre-treated with IFN-α measured at 24 hpi. Concentrations of IFN used is specified in the figure panel, and inhibition of the five sarbecoviruses by IFN-α assessed at 24 hpi (MOI 0.1) as determined by plaque assay. Color-coded curves represent a non-linear regression least square fit of biological replicates (n=3) and 50% of viral inhibition with 95% confidence intervals (in parenthesis) of the fit are: SARS-CoV-2: 47 U/mL (40.39 – 53.68), BANAL-52: 391 U/mL (333.3 – 458.6), BANAL-236: 336 U/mL (289.7 – 390.7), WIV1: 4899 U/mL (2636 – 13512), SHC014: 1087 U/mL (822.5 – 1490). Panel A is representative of two independent experiments. (B-D) Infectious viral titers of SARS-CoV-2, BANAL-52 and BANAL-236 in HBECs pre-treated with (B) IFN-α, (C) IFN-γ, and (D) IFN-λ. Panels Band C represent pooled data from two independent experiments (biological repeats: n=5 for IFN pre-treated cells, n=6 for IFN untreated infected cells). Panel D is also pooled from two independent experiments with biological replicates n=5 for both treated and untreated cells. Mock treated transwells were used for multiple treatment groups but all within the same independent repeat. Error bars in B-D illustrate standard deviation. Significance between viral titers for panels B-D was assessed by one-way ANOVA corrected for multiple comparisons. * indicates p<0.05, *** indicates p<0.001, **** indicates p<0.0001. Exact p values for panel B are SARS-2 vs BANAL-52 p=.0001, SARS-2 vs BANAL-236 p<.0001. Exact p values for panel C are SARS-2 vs BANAL-52 p=.0003, SARS-2 vs BANAL-236 p<.0001. For panels A-D, data are represented by mean values +/− SD. (E) Representative histograms of SARS-2 spike (left) and MHC-I (right) expression from A549-ACE2/TMPRSS2 cells infected with the indicated viruses, measured by flow cytometry. Gray shaded histograms represent uninfected control. (F) Representative contour plots of A549-ACE2/TMPRSS2 cells infected with the indicated viruses measuring expression of spike (y-axis) and MHC-I (x-axis). Panels E-F are representative plots from one of two independent experiments, each with four technical replicates.

We next assessed the induction of representative IFN-stimulated genes (ISGs) in primary bronchial and nasal epithelial cells. We observed modestly more ISG15 and IFIT3 induction in HBECs with SARS-CoV-2 than BANAL-CoVs. In HNECs, we observed increased induction of ISG15, IFIT3, and RSAD2, which encodes viperin (Extended Data Fig. 4A and 4B).

We next pre-treated HBECs with IFN-α and found that BANAL-52 and BANAL-236 displayed a median of ∼10-fold higher viral peak titers at 48 hpi, relative to SARS-CoV-2 (Fig. 2B). Further, we pre-treated the HBECs cells with IFN-γ or IFN-λ and found a similar trend of enhanced resistance to IFN-γ when measuring viral peak titers (Fig. 2C). We observed no significant difference in viral titers with IFN-λ pre-treatment between SARS-CoV-2 and BANAL-CoVs (Fig. 2D, Extended Data Fig. 4C–4E).

In HNECs, in the absence of IFN, CoVs replication substantially differed at 48 hpi in HNECs (Extended Data Fig. 4F), therefore we normalized each data point of the IFN-α treated group to the median of the mock to assess for differences in viral replication (Extended Data Fig. 4G). Similar to what we observed in HBECs, BANAL-CoVs were inhibited less by IFN-α compared to SARS-CoV-2, suggesting BANAL-CoVs also possess resistance to IFN restriction in nasal epithelial cells. IFN-α pre-treatment did not differentially restrict SARS-CoV-2 and BANAL-CoVs pseudovirus infection in Vero-ACE2/TMPRSS2 cells. (Extended Data Fig. 4H). This suggests that the observed IFN resistance occurs post-entry in Vero-ACE2/TMPRSS2 cells. Together, these results demonstrate that full-length bat CoVs, BANAL-52, BANAL-236, WIV-1 and SHC014, possess an enhanced resistance from Type I and II IFN-mediated restriction in vitro.

BANAL-CoVs downregulate MHC-I similarly to SARS-CoV-2.

Major histocompatibility complex class (MHC-I) antigen presentation is critical for CD8+ T cell recognition and killing of infected cells. SARS-CoV-2 downregulates MHC-I to evade cellular immunity^32–34. To assess whether BANAL-CoVs downregulate MHC-I, we infected A549 cells overexpressing ACE2/TMPRSS2 (A549-ACE2/TMPRSS2). Infection levels of A549-ACE2/TMPRSS2 cells differed across the three viruses (Extended Data Fig. 4I–J, and Fig. 2E). Despite this, SARS-CoV-2 and BANAL-CoVs similarly downregulated MHC-I expression (Fig. 2E and 2F). These data suggest that BANAL-CoVs have an intrinsic capacity to downregulate human MHC-I, despite not having undergone adaptation in humans.

BANAL-CoVs are less pathogenic than SARS-CoV-2 in K18-hACE2 mice.

We next measured the pathogenic potential of BANAL-52 and BANAL-236 in K18-human ACE2 (hACE2) mice³⁵. SARS-CoV-2 caused substantial weight loss and 100% mortality in these mice while BANAL-236 caused ~15% peak weight loss and 20% mortality. We did not observe weight loss or mortality in mice infected with BANAL-52 (Fig. 3A-C). We observed ~50,000 fold decrease in infectious viral titers from the lungs of infected mice at 2 days post-infection (dpi) in BANAL-52, and ~10 fold less for BANAL-236 compared to SARS-CoV-2. At 4 dpi, we did not observe significant differences in viral titers across the three sarbecoviruses. At 7 dpi, we detected infectious virus from the lungs of BANAL-CoV infected mice, while all SARS-CoV-2 infected mice succumbed to infection by this time point (Fig. 3D).

Figure 3. BANAL-52, BANAL-236 and SARS-CoV-2 display a gradient of pathogenesis in K18-hACE2 mice.

(A) Design of infection experiments in K18-hACE2 mice. Created with BioRender.com. (B) Survival curve of K18-hACE2 mice after infection of 1×10⁶ PFU of each virus (PBS n=4, SARS-CoV-2 n=12, BANAL-52 n=11, BANAL-236 n=12). (C) Weight loss curve from panel B. N for each group is the same as panel B. Significance between the PBS and BANAL-236 group was assessed by two-way ANOVA corrected to multiple comparisons. Exact p values for PBS vs BANAL-236 are 4 dpi p=.0002, 5–8 dpi p<.0001, 12 dpi p=.0107. (D) Infectious viral load from infected mice lungs harvested at the indicated time-points. For 2 dpi, SARS-2 n=8, BANAL-52 n=10, BANAL-236 n=12. For 4 dpi, SARS-2 n=5, BANAL-52 n=6, BANAL-236 n= 8. For 7 dpi, BANAL-52 n=6, BANAL-236 n=6. Significance between virus titers were assessed by one-way ANOVA corrected to multiple comparisons at each timepoint. Exact p values for 2 dpi are SARS-2 vs BANAL-52 p<.0001, SARS-2 vs BANAL-236 p=.0002. For panels C-D, data are represented by mean values +/− SD. (E) Heatmap of ISG and cytokines expression from lungs of K18-mice infected with the three sarbecoviruses measured by qPCRs. For panel E, data is pooled from three independent repeats and values are from Fig. S5A. SARS-2 n=11, BANAL-52 n=12, BANAL-236 n=13. (F) Infectious viral titers from lungs of SARS-CoV-2, BANAL-52 and BANAL-236 in K18-hACE2 mice pre-treated with IFN-α 12 h pre-infection and measured at 2 dpi (n=8 for SARS-2 and BANAL-236, and BANAL-52 n=8 for mock treated mice and n=7 for IFN treated mice. Significance between mock treated and IFN-α treated mice was assessed by one-way ANOVA corrected for multiple comparisons. Exact p values for mock vs IFN treated viral titers are SARS-2 and BANAL-52 p<.0001, BANAL-236 p=.0002. (G) Graph of lung viral titers from mice pre-treated with IFN-α normalized to the median of untreated infected mice. Significance was assessed by one-way ANOVA corrected for multiple comparisons. (n=8 for SARS-2 and BANAL-236, and n=6 for BANAL-52). Exact p value for SARS-2 vs BANAL-236 p=.0057. (H) Comparison of infectious titers between lungs and nasal turbinates from infected mice at 2 dpi. Difference in viral titers were assessed by two-way ANOVA corrected for multiple comparisons. All p values of Lung vs NT titers for BANAL-52 and BANAL-236 were p<.0001. (I) Infectious viral titers from nasal turbinates of SARS-CoV-2, BANAL-52 and BANAL-236 in K18-hACE2 mice pre-treated with IFN-α and measured at 2 dpi. For panel H, n=8 for all groups. For panel I (n=8 for SARS-2 and BANAL-236, and BANAL-52 n=8 for mock treated mice and n=7 for IFN treated mice. All panels, except for E, show pooled data from two independent experiments and each dot represents one mouse. For panels F-G, data is represented by median, error bars illustrate standard deviation. For all panels, * indicates p<0.05, ** indicates p<0.01, *** indicates p<0.001, **** indicates p<0.0001.

We also measured ISG induction in lungs from mice infected with SARS-CoV-2, BANAL-52 and BANAL-236 at 2 dpi. ISG levels were similar between mice infected with SARS-CoV-2 and BANAL-236. However, ISG levels and inflammatory cytokine induction were significantly lower in BANAL-52 infected mice (Fig. 3E, Extended Data 5A). This correlated with a lower viral load in lungs from BANAL-52 infected mice (Fig. 3D). Taken together, our results show that BANAL-CoVs are infectious and replicate in mice expressing human ACE2. We demonstrate that there is a gradient of pathogenesis in K18-hACE2 mice with SARS-CoV-2 showing high, BANAL-236 intermediate, and BANAL-52 low pathogenicity.

BANAL-CoVs have resistance to IFN in vivo.

We next tested whether BANAL-CoVs have enhanced resistance to exogenous IFN in vivo, as we observed in vitro. K18-hACE2 mice were treated intranasally with IFN-α prior to challenge, and the median change in viral titers between mock and IFN-treated mice was calculated to compare the level of inhibition of each virus by IFN. The median of infectious titers between IFN-α treated vs. untreated mice showed a ~6,800-fold decrease for SARS-CoV-2, ~670-fold decrease for BANAL-52, and only ~40-fold decrease for BANAL-236 (Fig. 3F, G), demonstrating that BANAL-CoVs are less susceptible to exogenous IFN-α than SARS-CoV-2 in vivo.

BANAL-CoVs replicate inefficiently in the nasal cavity of mice.

We next assessed whether the efficiency of replication of BANAL-CoVs differs across regions of the mouse airway. SARS-CoV-2 was detected at similar titers between the lungs and nasal turbinates (Fig. 3H). In contrast, BANAL-CoV titers were significantly decreased in nasal turbinates (~1,000-fold) compared to lungs (Fig. 3H). Additionally, ISG and inflammatory cytokine induction in nasal turbinates from BANAL-52 and BANAL-236 infected mice were also significantly lower than in lungs, compared to SARS-CoV-2 infected mice (Extended Data Fig. 5A–C, and Fig. 3E). This decreased ISG induction is consistent with lower viral replication in nasal turbinates of BANAL-CoV-infected mice. We also observed a modest IFN resistance (less than 10-fold difference in the median decrease of IFN treated vs. mock) (Fig. 3I).

Syrian golden hamsters do not efficiently transmit BANAL-CoVs.

Syrian golden hamsters are a leading model of SARS-CoV-2 animal-to-animal transmission and moderate disease³⁶. We next asked whether BANAL-CoVs could infect and transmit between hamsters. BANAL-CoVs used hamster ACE2 for entry similar to SARS-CoV-2 (Extended Data Fig. 6A). Over the course of a four-day infection, SARS-CoV-2 caused the most weight loss, followed by BANAL-236. BANAL-52 did not cause weight loss relative to PBS-treated hamsters (Fig. 4A and 4B). Oropharyngeal viral loads revealed delayed replication kinetics for BANAL-CoVs compared to SARS-CoV-2 (Fig. 4C). SARS-CoV-2-infected hamsters had higher oropharyngeal viral loads at 1 dpi compared to BANAL-CoV-infected hamsters. Similar infectious viral titers were observed by 2 dpi in SARS-CoV-2- and BANAL-236-infected hamsters. BANAL-52 oropharyngeal infectious viral titers were lower than SARS-CoV-2 and BANAL-236-oropharyngeal titers, consistent with the limited weight loss.

Figure 4. BANAL-CoVs have impaired transmission capacity in a hamster model.

(A) Design of infection experiments with Syrian golden hamsters. Created with BioRender.com. (B) Weight loss curve of Syrian golden hamsters after infection of 1×10⁶ PFU of each virus. Significance between the PBS and BANAL-236 group was assessed by two-way ANOVA corrected for multiple comparisons. Exact p value for PBS vs BANAL-236 at 3 dpi is p=.0082. For panel B, uninfected hamsters for all days n=6, n of infected hamsters for 0–2 dpi n=10, for 3 and 4 dpi n=5. (C) Viral titers of longitudinal daily oral swabs of index hamsters. For panel C, n of hamsters for 0 dpi n=8, 1–2 dpi n=10, for 3 and 4 dpi n=5. SARS-2 at 2 dpi n=9. Exact p values at 1 dpi are SARS-2 vs BANAL-52 and SARS-2 vs BANAL-236 p<0.0001. At 2 dpi SARS-2 vs BANAL-52 p<0.0001, SARS-2 vs BANAL-236 p=.6240. (D-E) Viral titers at 2 dpi and 4 dpi from nasal turbinates (D) and lungs (E) of infected hamsters (n of hamsters = 5). Significance between groups was assessed by one-way ANOVA corrected to multiple comparisons. For panel D, exact p values for 2 dpi are SARS-2 vs BANAL-52 p=.0002, SARS-2 vs BANAL-236 p=.0245, BANAL-52 vs BANAL-236 p=.0286. At 4 dpi are SARS-2 vs BANAL-52 p=.0003, SARS-2 vs BANAL-236 p=.0162, BANAL-52 vs BANAL-236 p=.0765. For panel E, exact p values for 2 dpi are SARS-2 vs BANAL-52 p=.0004, SARS-2 vs BANAL-236 p=.0005, BANAL-52 vs BANAL-236 p=.9971. At 4 dpi are SARS-2 vs BANAL-52 p<0.0001, SARS-2 vs BANAL-236 p<0.0001, BANAL-52 vs BANAL-236 p<0.0001. (F) Viral titers of longitudinal daily oral swabs of contact hamsters. For panel F, n of hamsters is SARS-2 n =5, BANAL-52 and BANAL-236 n=6. (G) Viral titers at 2 dpi from lungs and nasal turbinates of contact hamsters. For panel G, n of hamsters is: SARS-2 n=5, BANAL-52 n=6, BANAL-236 n=12. (H-I) Viral titers from oral swabs (H) and lung homogenates (I) of infected hamsters using different PFU of the sarbecoviruses. For panels H and I, n of hamsters =5 for all groups except SARS-2 1×10³ (n=4), SARS-2 1×10⁵ (n=3), SARS-2 1×10⁶ (n=4). Panels from this figure show pooled data from two independent experiments and each dot represents one hamster. For all panels, error bars illustrate standard deviation, and * indicates p<0.05, ** indicates p<0.01, *** indicates p<0.001, **** indicates p<0.0001. For all panels except G, data are represented by mean values +/− SD. For panel G, data are represented by mean values +/− SEM.

In the nasal turbinates, we observed modestly lower viral loads for BANAL-CoVs relative to SARS-CoV-2 (Fig. 4D). In the lungs, we observed a substantial reduction in viral load for BANAL-CoVs relative to SARS-CoV-2 (Fig. 4E). The viral loads of BANAL-CoVs were higher in the nasal turbinates relative to the lungs, in contrast to SARS-CoV-2 (Fig. 4D–E). Further, SARS-CoV-2, but not BANAL-52 or BANAL-236, caused lung inflammation (Extended Data Fig. 6B). This is consistent with the low viral titers of BANAL-CoVs in the lungs of infected hamsters.

Next, we tested the transmission fitness of BANAL-CoVs in hamsters. At 2 dpi, an experimentally infected index hamster was co-housed with two naïve contact hamsters for 4 hours (h) (Fig. 4A). All contact hamsters (6/6) co-housed with SARS-CoV-2-infected index hamsters. However, none of the contact hamsters (0/6) co-housed with BANAL-52- or BANAL-236-infected index hamsters became infected when measuring viral titers from oral swabs (Fig. 4F). Infectious viral titers from the lungs and nasal turbinates of contact hamsters confirmed efficient transmission for SARS-CoV-2 (5/5) and the lack of transmission of BANAL-52 (0/6). Interestingly, one out of 12 contact hamsters co-housed with BANAL-236 index hamsters became infected. (Fig. 4G).

Finally, we assessed whether the impaired BANAL-CoV transmission in hamsters was dependent on the infectious dose (ID₅₀). Based on oral swabs, the calculated ID₅₀ of BANAL-236 and BANAL-52 were ~10² PFU and about ~10⁵ PFU, respectively. In contrast, the SARS-CoV-2 ID₅₀ was greater than 10¹ PFU (Fig. 4H). Notably, BANAL-CoVs required higher viral inoculum to productively infect the lower airway in hamsters (Fig. 4I). These results reveal a substantial difference in infection susceptibility between SARS-CoV-2 and BANAL-CoVs, potentially contributing to the reduced transmission fitness of these viruses in hamsters.

SARS-CoV-2 ΔFCS poorly replicates in HNECs and is outcompeted by WT virus in hamsters.

To assess how SARS-CoV-2 lacking an FCS (SARS-CoV-2 ΔFCS) (Extended Data Fig. 7A–C) growth compares to BANAL-CoVs, we infected Vero-ACE2/TMPRSS2 cells and HNECs with WT SARS-CoV-2, SARS-CoV-2 ΔFCS, WT BANAL-52, and WT BANAL-236. All four viruses replicated to similar titers in Vero-ACE2/TMPRSS2 cells (Fig. 5A). In contrast, SARS-CoV-2 ΔFCS and BANAL-CoVs had impaired replication in HNECs, relative to SARS-CoV-2 (Fig. 5B). This suggests that the FCS contributes to replication fitness in the upper airway of humans. Finally, we infected hamsters with SARS-CoV-2 ΔFCS (Fig. 5C). While SARS-CoV-2 ΔFCS titers were comparable to those of SARS-CoV-2, viral sequencing revealed outgrowth of SARS-CoV-2 with a functional FCS (from ~2% in the viral inoculum to ~92% at 2 dpi) (Extended Data Fig. 7C–D). These results demonstrate that FCS of SARS-CoV-2 promotes infection HNECs and hamsters.

Figure 5. The S1/S2 FCS promotes replication in HNECs and hamsters.

(A) Viral growth curve of the four sarbecoviruses in Vero-ACE2/TMPRSS2 cells (n=3 biological repeats for each time point). Panel is representative from 2 independent experiments. (B) Viral growth curve of sarbecoviruses in HNECs at ALI (n=2 biological repeats for1 hpi, n=4 for remaining timepoints). Panel is representative from 2 independent experiments. P values for SARS-2 vs SARS-2 ΔFCS, SARS-2 vs BANAL-52, and SARS-2 vs BANAL-236 were all p<0.0001 at 24, 48, and 72 hpi. (C) Viral titers at 2 dpi from lung homogenates and nasal turbinates of infected hamsters using 1×10⁶ of virus inoculum (n=3). Panel is pooled from 2 independent experiments. Significance between viral titers was assessed by one-way ANOVA corrected for multiple comparisons. P values for SARS-2 vs BANAL-236 and SARS-2 ΔFCS vs BANAL-236 were all p<0.0001 for both Lung Homogenates and Nasal Turbinates. For all panels data are represented by mean values +/− SD, and **** indicates p<0.0001 between SARS-2 against BANAL-CoVs and SARS-2 ΔFCS.

Discussion

We constructed infectious molecular clones and characterized the closest known relatives of SARS-CoV-2, BANAL-52 and BANAL-236, to determine their pathogenic and transmission potential. BANAL-CoVs replicate similarly to SARS-CoV-2 in cell lines and in primary HBECs. However, BANAL-CoV replication is restricted in primary HNECs. BANAL-CoVs can downregulate MHC-I similar to SARS-CoV-2 and are unexpectedly resistant to IFN in vitro and in vivo relative to SARS-CoV-2. Despite increased innate immune evasion, BANAL-CoVs replicate less efficiently with reduced pathogenesis in mice and hamsters and are poorly transmitted between hamsters.

The IFN resistance of BANAL-CoVs has important implications for pathogenesis and zoonosis. The enhanced resistance to IFN-mediated defense could reflect BANAL-CoV evolution in the face of constitutively expressed IFN-α, which is observed in some bat species³⁷. However, it is unclear if Rhinolophus bats that harbor sarbecoviruses constitutively express IFN-α. In addition to IFN resistance, BANAL-CoVs also downregulate MHC-I, as observed for SARS-CoV-2^32–34. This indicates that BANAL-CoVs have the potential to evade human innate and adaptive immunity, even though they have not been adapted to human hosts.

In both mice and hamsters, we observed a gradient of pathogenesis, with SARS-CoV-2 being the most pathogenic and BANAL-52 being the least pathogenic. The BANAL-236 pathogenesis observed here is consistent with a recent study in K18-hACE2 mice³⁸. Interestingly, the spike protein of the less pathogenic BANAL-52 is more similar to that of SARS-CoV-2. The NTD of spike from BANAL-236 is divergent from both SARS-CoV-2 and BANAL-52. It is possible that the NTD of BANAL-236 contributes to pathogenesis; however, differences outside of spike can also play a role^14,15,39. This demonstrates that there are important pathogenic determinants of CoVs beyond the FCS.

A striking observation is the transmission defect for BANAL-CoVs in hamsters relative to SARS-CoV-2. This transmission defect suggests that efficient viral replication in vitro is not sufficient for predicting efficient transmission. We envision several non-mutually exclusive explanations for this observation. First, the lack of transmission of BANAL-CoVs between hamsters may be due to reduced shedding of infectious virus from infected index hamsters. This is consistent with the lower BANAL-CoV viral loads in the upper and lower airways relative to SARS-CoV-2. Second, a higher infectious dose is needed to establish infection with BANAL-CoVs. This is consistent with an impaired replication of BANAL-CoVs in HNECs and mouse nasal turbinates. Third, it is possible that BANAL-CoV spikes exhibit differential host cell tropism at the site of entry, which may impact transmission efficiency. Notably, one contact hamster co-housed with BANAL-236 index animals in this study became infected. Thus, BANAL-236 is not completely devoid of transmission ability in this model and underscores the possibility of transmission in other animal hosts.

BANAL-CoVs lack the FCS between the S1/S2 domains of their spike sequences. SARS-CoV-2 mutants that lack an FCS are less pathogenic in mice and less transmissible in ferrets ^40,41. This is consistent with our findings that an FCS-deleted SARS-CoV-2 poorly replicates in HNECs and was swiftly out-competed by a small amount of contaminating wild-type virus in hamsters. Thus, the lack of an FCS in BANAL-CoVs may explain their inefficient nasal epithelium tropism, transmission fitness, and pathogenesis. Interestingly, a putative FCS at the S2’ region of spike (around residues 810–820) is conserved across SARS-CoV-2, BANAL-CoVs, and other related viruses⁴². Moreover, SARS-CoV also lacks an FCS at the S1/S2 junction, yet it can transmit between people and exhibits increased pathogenesis relative to SARS-CoV-2⁴³. This suggests that factors beyond the FCS play a role in sarbecovirus transmission and pathogenesis.

Our study has several limitations. First, due to biosafety concerns in generating chimeric viruses, we did not unambiguously identify the specific viral genes and amino acids that mediate BANAL-CoV pathogenesis and transmission. Specifically, we did not create chimeric BANAL-CoVs with a S1/S2 FCS to examine the effect on pathogenic or transmission potential. The development of biosafe tools to study novel and diverse bat CoVs with pandemic potential is needed. Second, our ultimate goal is to identify molecular features that predict pandemic potential of pre-emergent bat CoVs. However, this is intrinsically limited by animal models that may not perfectly extrapolate to humans. While we observe consistent patterns of pathogenesis in two animal models, no animal can faithfully predict the pathogenic and transmission potential of BANAL-CoVs in humans.

Although many novel sarbecovirus genomes have been discovered, the vast majority remain uncharacterized⁴⁴. We show that despite the ability to evade innate immunity and downregulate MHC-I, BANAL-CoVs exhibit reduced pathogenesis and transmission potential in rodent models. In addition, the sensitivity of BANAL-CoVs to existing direct-acting antivirals and susceptibility to SARS-CoV-2-elicited and vaccine-induced antibodies suggests that existing countermeasures would confer at least partial protection to emergence of BANAL-CoVs. In summary, this work highlights that the lack of functional characterization of diverse bat CoVs contributes to our inability to accurately predict CoV pandemic potential from nucleotide sequences. Future work is urgently needed to determine the viral properties that promote efficient sarbecovirus transmission and pathogenesis to enhance preparedness against the next coronavirus emergence.

Materials and methods

Ethics declaration.

Our research complies with relevant ethical regulations established by the institution. All experiments involving animals followed the Yale Institutional Animal Care and Use guidelines. All human sera was collected and pooled from de-identified and discarded human samples, therefore no informed consent was required.

Biosafety considerations

All experiments involving infectious virus were performed in the Yale University BSL-3/ABSL-3 facility. All experiments were approved by the Yale Institutional Animal Care and Use Committee, Yale recombinant DNA committee, Yale biological safety committee, the Yale BSL3 subcommittee, and Yale Environmental Health and Safety. All viruses were handled under the proper biosafety level guidelines as specified by the university committee. All personnel handling the viruses had HEPA-filtered respirators and adequate personal protective equipment. Expansion of bat CoVs stocks was minimized to avoid accumulation of mutations. BANAL-CoVs generated in this study had no amino acid differences relative to published sequences², with the exception of an inserted reporter. Per U.S. gain of function regulations, wild-type pathogens circulating or derived from nature are not enhanced potential pandemic pathogens.

Phylogenetic analysis

Viral sequences used in this study are specified in Supplementary Table 1. Full-length genome and spike protein alignments were performed in MEGA X and Jalview software, respectively. Both nucleotide and aminoacid alignment from sequencing of SARS-2 ΔFCS stock were made using Jalview software. The phylogenetic tree was created using maximum-likelihood method in MEGA X with 500 bootstraps. Animal silhouettes were obtained from PhyloPic (www.phylopic.org).

BANAL-52 and BANAL-236 reverse genetics design and cloning.

BANAL-52 and BANAL-236 genomic sequences were retrieved from GenBank, MZ937000.1 and MZ937003.2 respectively², and the genomic cDNA was synthesized using Twist BioScience. Full-length sequences of the molecular clones are included in Supplementary File 1. Briefly, the CoV genomes were divided into seven fragments and flanked by Type IIS restriction enzyme sites, as described previously for SARS-CoV-2²⁹. Fragments A-D were flanked by BsaI, while fragments E-G were flanked by Esp3I. Fragments A, D, E and F were cloned into pUC57, while fragments B, C, G and G-mNG were cloned into pCC1 by conventional PCR and Gibson assembly. At the time of this reverse genetics design, the BANAL-CoV genomic sequences in NCBI had incomplete 5’ and 3’ UTR sequences. Therefore, the 5’ and 3’ UTRs were designed based on a sarbecovirus alignment and according to SARS-CoV-2. The NCBI sequence of BANAL-236 was ultimately modified by the depositing authors². The updated UTR sequences matched our initial design. For G-mNG reporter fragments, a P2A peptide followed by the mNeonGreen open reading frame was inserted after ORF7a and without disruption of ORF7b³⁰. Finally, silent mutations were introduced to disrupt undesired restriction enzyme sites in both BANAL-52 and BANAL-236 (Supplementary Table 3). Next, every fragment was subcloned into the corresponding plasmid by Gibson assembly. After cloning, all fragments were confirmed by whole-plasmid sequencing (Plasmidsaurus).

Recovery of full-length BANAL-52 and BANAL-236.

Production of infectious BANAL-CoVs was performed similarly to SARS-CoV-2 and as described previously, with some minor modifications²⁹. Briefly, excision of the seven BANAL-CoV DNA fragments from the agarose gel was performed by using a Blue-Light transilluminator (Accuris). Next, equal molar amount of all seven viral fragments were used to perform the full-length ligation of the CoV DNA genome: A (0.305 μg), B (0.325 μg), C (0.375 μg), D (0.47 μg) were first ligated in a volume of 50 μl. Fragments E (0.375 μg), F (0.36 μg), G (0.3 μg), or G-mNG (0.3 μg) were ligated in a volume of 40 μl. These first reactions were incubated at 4°C for 16 h. After the initial ligation step, the two reactions were mixed, and the total volume was brought to 100 μl using T4 ligation buffer, T4 DNA ligase, and water. This final ligation reaction was incubated for 18 h at 4°C. The full-length CoV cDNA was purified employing the Genomic DNA Clean and Concentrator kit (Zymo Research #D4011). The ligation of full-length cDNA was confirmed using a 1% agarose gel.

Next, in vitro RNA transcription of the full-length CoV genome was performed as described previously, with additional modifications²⁹. The in vitro transcription reactions were performed by using the mMESSAGE mMACHINE T7 Transcription Kit (Thermo Scientific #AM1344) with a cap analog-to-GTP ratio of 1:1. For this reaction, 500 ng of each ligation product was used as the transcription template and incubated at 32°C for 16 h. After the transcription reaction, DNAse was added and incubated at 37°C for 15 min. Full-length RNA was purified by using the RNA Clean and Concentrator kit (Zymo Research #R1018). The completed transcription of full-length viral mRNA was confirmed using a 1% agarose.

To produce infectious virus, each CoV mRNA was introduced into BHK-21 cells expressing the SARS-CoV-2 nucleoprotein (BHK-N) by electroporation. 15 ug of each CoV mRNA transcript was added to 8 million BHK-N cells suspended in 390 uL of cold PBS. Electroporation procedure was performed by using the Electro Square Porator Device (BTX) with the following parameters: 860V, 5 pulses, 99 μsec, 1.1 sec. After the electroporation procedure, the cells were seeded into a T75 flask containing 1:2 ratio of BHK-N and wild type Vero-E6 cells. After 72 h post-electroporation, cytopathic effects were observed, and the supernatant (passage 0) was collected, centrifuged at 2,500 × g for 10 min, and filtered through a 0.22 μm filter. To propagate virus stocks, Vero-ACE2/TMPRSS2 cells were infected with 500 uL of clarified supernatant and incubated until cytopathic effects were observed. Supernatant was collected, centrifuged, and filtered as described above. Viral stocks were stored at −80°C.

Cells and viruses.

Vero-E6 cells and Vero-E6 cells overexpressing human ACE2 and TMPRSS2 (Vero-ACE2/TMPRSS2) were cultured in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 1% sodium pyruvate, 10% fetal bovine serum (FBS), and 5 μg/mL of puromycin at 37°C and 5% CO₂. A549-ACE2/TMPRSS2 cells were purchased from Invivogen (Cat. A549-hace2tpsa) and maintained in DMEM supplemented with 1% sodium pyruvate, 10% FBS, 100 μg/mL hygromycin, and 5 μg/mL puromycin at 37°C and 5% CO₂. BHK-N cells were transduced with a lentivirus encoding human codon-optimized SARS-CoV-2 N gene (Blue Heron Bio). SARS-CoV-2 N was subcloned via Gibson assembly (NEB) between the BsrGI and ClaI sites of pTRIPZ (Dharmacon), which was then packaged into VSV-G pseudotyped lentivirus particles into HEK-293T cells by standard methods. These particles were then used to transduce BHK-21 clone 15 cells⁴⁵. Productively transduced cells were selected in cell culture media containing 3 μg/mL puromycin (Invivogen) and doxycycline-inducible N gene expression was confirmed by Western blot. All cell lines were mycoplasma negative. We authenticated the Vero cells previously⁴⁶.

Wild type SARS-CoV-2/WA01 was generated by reverse genetics as described previously⁴⁷. SARS-CoV-2 mNG was kindly provided by Dr. Pei-Yong Shi. Infectious BANAL-52, BANAL-236, BANAL-52-mNG and BANAL-236-mNG were re-created in this study by reverse genetics. Full-length WIV1 and SHC014 were kindly provided by Dr. Ralph Baric^18,19. SARS-CoV-2 ΔFCS was obtained from SARS-CoV-2 hCOV-19/USA-WA1/2020 (BEI Resources Cat# NR-52281). This virus which was not derived from a molecular clone naturally acquired mutations in the FCS during tissue culture passage (Fig S7). To create working stocks for SARS-CoV-2/WA01, BANAL-52, BANAL-236, WIV1 and SHC014, Vero-ACE2/TMPRSS2 cells were infected at an MOI 0.1 for two days. For SARS-CoV-2 ΔFCS, Vero-E6 were used. After infection, the supernatant was clarified by centrifugation (2,500 × g for 10 min), filtered through a 0.22-micron filter and stored at −80°C. Viral titers were measured by plaque assay using Vero-ACE2/TMPRSS2 as described previously⁴⁸.

Virus sequencing and variant calling

BANAL-52 and BANAL-236 stocks were sequence confirmed as previously described⁴⁹. Briefly, total RNA was extracted and DNase treated prior to cDNA synthesis. The Illumina Nextera XT kit was used for library preparation. Individual libraries were quantified and pooled at equal concentrations for sequencing on the Illumina NovaSeq 6000 (paired-end 150) at the Yale Center for Genome Analysis. Each virus was sequenced in duplicate and negative (water) controls were included during library prep. The computational tool iVar (version 1.3.1) was used to align sequencing reads to reference genomes and generate consensus genomes at a minimum frequency threshold of 0.75 and minimum coverage of 10X, and to determine frequencies of intrahost single-nucleotide variants. Consensus genomes and frequencies of intrahost single-nucleotide variants were inspected to confirm successful cloning of the viruses.

SARS-CoV-2 NGS Sequencing.

SARS-CoV-2 sequencing was performed as previously described with some modifications⁵⁰. Briefly, RNA was purified from virus stocks and experimental samples using the Direct-Zol RNA Miniprep kit (Zymo Research, Cat #R2052). Sequencing libraries were prepared using ARTIC Network’s library preparation protocol (V5.3.2 primers). The Yale Center for Genome Analysis sequenced pooled libraries on the Illumina NovaSeq (paired-end 150).

SARS-CoV-2 ΔFCS sequencing analysis.

Raw NGS reads were first pooled across technical replicates. Bowtie2 (v2.5.1)⁵¹ was used to index the amplicon spanning the FCS (corresponding to ARTIC primers SARS-CoV-2_400_77_RIGHT_0 and SARS-CoV-2_400_77_LEFT_0) (bowtie2-build). Reads were aligned to the amplicon index (Bowtie2 option, --very-sensitive-local --no-mixed --no-discordant). We excluded from our analysis the amplicon derived from ARTIC primers V5.3.2 SARS-CoV-2_400_76_RIGHT_0 and SARS-CoV-2_400_76_LEFT_0, as the rightmost primer directly binds to, and thus overrepresents, the FCS. SAMtools (v1.18)⁵² was used to convert the alignment files from SAM (.sam) to BAM (.bam) format (option, view -bS) and extract mapped reads (option, view -b -F 4). Next, manual filtering was performed to remove mapped reads lacking two conserved 6-nt sites (GACTCA and CAATCC; positions 23,587–23,592 and 23,530–23,535 from Genbank ID MN908947.3, respectively) flanking the FCS. In order to calculate FCS abundance in each sample, the remaining reads were manually queried for the FCS sequence (TCTCCTCGGCGGGCACGT) without allowing for mismatches.

Air-liquid interface of human primary airway epithelial cells (HBECs and HNECs).

Primary HBECs and HNECs were purchased from Lonza and cultured in PneumaCult-Ex Plus Medium following the manufacturer’s instructions (STEMCELL Technologies) as previously described⁵³. To develop the air-liquid interface culture, 5×10⁴ cells HBECs or HNECs were cultured on transwell inserts with a 0.4-micron pore size (Costar, Corning Cat. #3470) and placed in 24-well cell culture plates. All wells were maintained in PneumaCult-Ex Plus Medium (STEMCELL Technologies Cat. # 05041) for the first three days, and then the media was switched to PneumaCult-ALI Medium (STEMCELL Technologies Cat. #05002) with ROCK inhibitor Y-27632 for another four days. During the first six days, fresh media was added (100 μl in the apical compartment / 500 μl in the basal compartment). Apical media was removed on day seven, while basal chambers were still maintained with 500 μl of PneumaCult-ALI Medium. HNECs were cultured at ALI for 21 days and HBECs for 28 days to allow for cell differentiation, which was visually determined by cilia movement (ciliated cells) and mucus-production (goblet cells). Media was replaced every 2 to 3 days (500 μl) during the differentiation phase.

Animal Experiments.

B6.Cg-Tg(K18-ACE2)2Prlmn/J (K18-humanACE2) mice were obtained from The Jackson Laboratory (#034870) and subsequently bred and housed at Yale University³⁵. Mice used for infection experiments were eight to fourteen weeks old. All mouse experiments used both male and female animals. Male Syrian golden hamsters (HSdHan:AURA) were obtained from Envigo. For infection, hamsters at the age of eight to ten weeks were used. All experiments were approved by the Yale Institutional Animal Care and Use Committee and in accordance with regulatory guidelines. Animal infection was defined by the presence of infectious virus from oropharyngeal swabs, lungs, and/or nasal turbinates. All animals were randomly assigned to given experimental condition.

Viral infections.

For Vero-ACE2/TMPRSS2, 4×10⁴ cells were seeded in a 96-well plate and infected at an MOI of 0.1. The culture media was removed, virus was added to each well in 30 μl total volume of DMEM and incubated at 37°C and 5% CO₂ for 1 h. The supernatant was removed and replaced with fresh DMEM with 2% FBS, and the cells were returned to 37°C.

For ALI cultures, the apical side of the HBECs and HNECs were gently rinsed three times with 100 μl of pre-warmed PBS without divalent cations (Gibco). 5×10⁵ PFUs of each virus (MOI of 0.5) were suspended in a total volume of 100 μl HBECs/HNECs culture media and added to the apical compartment of the Transwell^®. Cells were incubated at 37˚C and 5% CO₂ for 1 h, the unbound virus was removed, and each Transwell^® was washed three times with pre-warmed PBS.

Mice and hamsters were infected using 10⁶ PFU (titers determined on Vero-ACE2/TMPRSS2 cells) per animal. The virus was diluted in 50 μl of PBS and administered intranasally. For intranasal administration, 30% isoflurane was used to anesthetize mice and hamsters. All animals were randomly assigned prior to infection.

Assessment of SARS-CoV-2 inhibitor and antiviral efficacy on BANAL reporter viruses.

Vero-ACE2/TMPRSS2 were seeded and pre-treated for 4 h with the indicated concentration of remdesivir or nirmatrelvir in a 384-well plate using 4×10³ cells per well. Next, cells were infected with SARS-CoV-2-mNG, BANAL-52-mNG, or BANAL-236-mNG at 0.1 MOI using phenol-red free DMEM media with 5% FBS and incubated at 37°C. Infected cell frequencies were measured at 24 hpi by mNeonGreen expression using high-content imaging (Cytation 5, BioTek) as previously described^46,48.

Assessment of SARS-CoV-2 neutralizing sera effect on BANAL reporter viruses.

Moderna and Pfizer pooled vaccine sera was obtained from BEI resources (NRH-17846 and NRH-17727, respectively). Vaccine sera donors were not previously infected with SARS-CoV-2. Convalescent sera was pooled from ten de-identified patients that recovered from SARS-CoV-2 infection from the Yale New Haven Health System during March and April of 2020. Samples were pooled from sera known to be of high neutralizing titer against SARS-CoV-2. For SARS-CoV-2 vaccinated and convalescent sera experiments, the indicated dilution of sera was incubated with 4×10² PFU of SARS-CoV-2-mNG, BANAL-52-mNG or BANAL-236-mNG for 1 hour at 37°C. After this incubation period, 4×10³ of Vero-ACE2/TMPRSS2 cells were added to each well. Infected cell frequencies were measured at 24 hpi by mNeonGreen expression using high-content imaging (Cytation 5, BioTek) as previously described^46,48. Measurement of cell viability was done using CellTiter-Glo Luminescent assay (#G7571) from Promega according to the manufacturer’s (25 μl of the reagent per well) at 48 hpi.

Inhibitor assays.

Camostat mesylate (Bio-techne #3193) and E64d (Sigma-Aldrich E8640–1MG) were resuspended in DMSO and water respectively and diluted to the desired concentrations for viral inhibition assessment. Infection conditions were the same as the remdesivir and nirmatrelvir experiments. Cell viability evaluation after infection of Vero-ACE2/TMPRSS2 cells was measured using CellTiter-Glo Luminescent assay (# G7571) from Promega according to the manufacturer’s instructions at 48 hpi. Before infection, HBECs cultures were pre-treated for 48 hours by addition of each inhibitor to the basal media. Viral replication was measured 48 hours after infection.

IFN pre-treatments in cell culture and mice.

Universal IFNα (#11200–2), human IFNγ (#11500–2) and human IFNλ-IL29 (#11725–1) were purchased from PBL Biosciences. Vero-ACE2/TMPRSS2 cells were seeded at a density of 2×10⁴ cells per well and pretreated with the indicated concentrations of IFN for 18h before viral infection. For HBECs and HNECs, PneumaCult-ALI Medium was removed from the basal compartment and fresh media containing the indicated concentrations of IFN were added in the apical side (100 μl) and basolateral compartment (500 μl) for 18 h before infection. Fresh media without IFN was added before and after viral infection. Finally, for in vivo experiments, 10⁵ units of IFN (PBL #11200–2) or PBS control were intranasally administered to each mouse 12 h before viral infection.

MHC-I expression measurements by flow cytometry.

A549-ACE2/TMPRSS2 cells were seeded in a 24-well plate at a density of 3×10⁵ cells per well and infected for 40 h using an MOI of 0.1, then fixed for flow cytometry analysis as previously described³⁴. Briefly, cells were blocked using human BD Fc Block (1:100, BD Biosciences Cat # 564220) and Live/Dead Fixable Aqua (Thermo Fisher) for 15 min at room temperature. Cells were stained with APC anti-HLA-ABC (1:200, Thermofisher, Cat # 17–9983-42) for 20 min at room temperature. Cells were then washed with 2 mM EDTA-PBS and resuspended in 100 μL 2% PFA for 1 h at room temperature. Previously fixed cells were intracellularly stained by being permeabilized using eBioscience FoxP3/Transcription Factor Staining Buffer (Thermo Fisher) for 10 min at 4°C. Cells were subsequently washed and stained using AF488 anti-SARS-CoV-2 Spike S1 Subunit (R&D Systems, Cat # FAB105403GAPC) at 1:50 in permeabilization buffer for 30 min at 4°C. Cells were washed and resuspended in PBS containing 1% FBS and analyzed on an Attune NxT (Thermo Fisher) flow cytometer. Data analysis was done in FlowJo software (Tree Star).

RNA extractions and real-time quantitative PCRs.

Total RNA was isolated from lung or nasal turbinate samples from mice using the Direct-Zol RNA Miniprep kit (Zymo Research, Cat #R2052). For cDNA synthesis, 500 ng of freshly extracted RNA was used using the iScript cDNA Synthesis kit (Biorad, Cat #1708840). Finally, quantitative RT-PCR was performed using the iTaq Universal SYBR Green Supermix (Biorad, Cat #1725121). RNA levels of each gene were calculated using standard ddCT method and normalized relative to b-actin for mice. A complete list of primer sequences used in this study can be found in Supplementary Table 3.

Virus titer quantification.

Cell culture infected supernatants were obtained after infection and cleared of debris by centrifugation at 2,500 × g for 10 min. For infected HBECs and HNECs, 100 μl of media was placed on each Transwell^® and incubated for 25 min at 37°C. After incubation period, each Transwell^® was washed three times with incubation media to obtain infectious virus from each well.

For animal experiments, mice and hamsters were euthanized using 100% isoflurane. Lungs (right lobes) or nasal turbinates were placed in a bead homogenizer tube with 1 mL of DMEM with 2% FBS and 2% antibiotics/antimycotics (Gibco), homogenized 3 times for 30 seconds, and were cleared of debris by centrifugation (3,300 × g or 10 min). Daily oral swabs (Pruitan PurFlock Ultra 25–3206-U) were taken from hamsters and stored in 1 mL of DMEM with 2% FBS and 2% antibiotics/antimycotics (Gibco) at −80°C.

Infectious virus titers were determined by a plaque assay in Vero-ACE2/TMPRSS2 cells in DMEM supplemented with 2% FBS, and 0.6% Avicel RC-581⁴⁸. For BANAL-52 and BANAL-236, 1 μg/μL⁻¹ of TCPK treated trypsin (Thermo Scientific #20233) was added to the overlay media after the 1 h infection period. After 48 hpi, plaques were fixed using 10% Neutral Buffered Formalin for 30 min followed by staining for 20 min using 0.5% crystal violet in 20% ethanol. Plates were rinsed in water to visualize plaques, which were counted inside the BSL-3 facility.

Pseudovirus production and hamster ACE2 receptor usage assays.

VSV and lentiviral based pseudoviruses were produced as previously described^48,54,55. SARS-CoV-2, BANAL-52 and BANAL-236 spikes were synthesized and human codon-optimized (Twist Biosciences), and subsequently sub-cloned into the pCAGGs expression vector for pseudotyping.

Hamster and human ACE2 sequences were synthesized and subcloned into the pCMV vector for expression in cell culture (Twist Biosciences). For hamster ACE2 receptor usage experiments, HEK293T cells were transfected with GFP, human ACE2, or hamster ACE2 expressing plasmids in a 96-well plate at 20,000 cells per well. VSV pseudovirus infection was performed at 24 h post-transfection and entry efficiency using human or hamster ACE2s were normalized to GFP transfected cells at 24 hpi by using Renilla-Glo Luciferase Assay system (Promega Cat. E2750).

For IFN assays, Vero-ACE2/TMPRSS2 cells were pre-treated with 1000 U/mL of Universal IFNα (#11200–2) from PBL Biosciences for 18 h. Cells were then infected with lentiviral pseudoviruses (pHAGE2) and entry efficiency was measured at 48 hpi infection using Renilla-Glo Luciferase Assay system (Promega Cat. E2750).

Immunohistochemistry and pathological analysis.

Embedding, sectioning, and H&E staining of hamster lung tissue was performed by the Yale Pathology Tissue Services (YPTS). The slides were reviewed by a blinded pulmonary pathologist for immune cell infiltration and other related pathologies according to previously defined criteria⁵⁴.

Statistical analysis.

GraphPad Prism 9 was used for all statistical analyses in this study. Figure legends specify which tests were performed on each experiment. Across figure panels, * indicates p<0.05, ** indicates p<0.01,*** indicates p<0.001, **** indicates p<0.0001. No statistical analysis was performed to determine animal sample size, but groups are similar to those in previous publications⁵⁴. Additionally, data distribution was assumed to be normal but was not formally tested. Individual data points are shown in each panel when possible.

Data collection.

Data collection and analysis were not performed blind to the conditions of the experiments except Fig S6 which was analyzed by a blinded pathologist.

Extended Data

Extended Data Fig. 1. BANAL-236 spike NTD is divergent from SARS-2 and BANAL-52.

Protein alignment of BANAL-CoV and SARS-2 spikes. The S1 NTD and S1/S2 FCS are highlighted with red boxes.

Extended Data Fig. 2. BANAL-CoVs mNGs are susceptible to small molecule inhibitors and anti-SARS-2 sera.

(A) Representative image of an agarose gel showing all the BANAL-CoVs fragments synthesized and amplified before cloning them to their respective vectors. (B) Representative image of an agarose gel showing full-length genome ligation of BANAL-52 and BANAL-236 cDNA used for in vitro transcription and to produce infectious viruses. (C) Bat CoV titers in Vero-ACE2/TMPRSS2 cells. (D) BANAL-CoV and BANAL-CoV mNG titers in Vero-ACE2/TMPRSS2 cells. (E) Representative images of Vero-ACE2/TMPRSS2 cells infected with mNeonGreen viruses. (F-H) Inhibition of the three viruses by two doses of Pfizer (F) of Moderna (G) vaccinated pooled sera assessed at 48 hpi (MOI 0.1) as determined by cell viability measured by Cell-Titer Glo. Inhibition of the three viruses by convalescent pooled sera (H) assessed at 48 hpi (MOI 0.1) as determined by cell viability measured by Cell-Titer Glo. Color-coded cures represent a non-linear regression least square fit of biological replicates (n=3) and shading represents 95% confidence intervals of the fit. For panels F-H, data are represented by mean +/− SD. All figure panels are representative of two independent experiments.

Extended Data Fig. 3. BANAL-CoVs utilize similar entry factors than SARS-2.

(A-C) Inhibition of the three sarbecoviruses by (A) E64d assessed at 24 hpi in Vero-E6 as determined by % mNG expressing cells, (B) Camostat mesylate and (C) E64d assessed at 48 hpi in Vero-ACE2/TMPRSS2 cells as determined by cell viability measured by Cell-Titer Glo. Color-coded curves represent a non-linear regression least square fit of 4 biological replicates and shading represents 95% confidence intervals of the fit. Panels A-C are representative of two independent experiments. (D-E). Inhibition of the three sarbecoviruses by (D) camostat mesylate and (E) E64d in HBECs culture at 48 hpi. Error bars illustrate standard deviation. Significance between groups was assessed by two-way ANOVA corrected to multiple comparisons. For panel D, p values of mock and camostat treated cells were all p<0.0001 for SARS-2, BANAL-52 and BANAL-236. Panels D and E are pooled from two independent experiments with n=3 biological replicates each (n=6 total). Mock treated wells are same for panels D and E. For all panels, data are represented by mean +/− SD.

Extended Data Fig. 4. ISG induction in primary human airway cells after BANAL-CoV infection.

(A-B) mRNA fold change of ISG induction relative to uninfected cells measured by RT-qPCR from infected HBECs (A) or HNECs (B). Fold change was assessed by 2–ΔΔCt method using gapdh as a control. Significance of expression between viruses was assessed by one-way ANOVA. Panels A-B are pooled from two-independent experiments with a total of n=6 biological replicates. For panel A, exact p values for ISG15 SARS-2 vs BANAL-52 p=.0105 and SARS-2 vs BANAL.236 p=.0162. For IFIT3 SARS-2 vs BANAL-52 p=.0129 and SARS-2 vs BANAL.236 p=.0207. For panel B, exact p values for IFIT3 SARS-2 vs BANAL-52 p=.0125 and SARS-2 vs BANAL-236 p=.0386. For RSAD2 SARS-2 vs BANAL-52 p=.0449. (C-E) Reduction of infectious viral titers of IFN-pretreated HBECs normalized to the median of the mock group of SARS-CoV-2, BANAL-52 and BANAL-236 infection using (C) IFN-α, (D) IFN-γ, and (E) IFN-λ. Graphs represent pooled data from two independent experiments (n=5 biological replicates for IFN pre-treated cells, n=6 biological replicates for IFN untreated infected cells). Significance between viral titers was assessed by one-way ANOVA corrected for multiple comparisons. For panel C SARS-2 vs BANAL-52 and SARS-2 vs BANAL-236 p<0.0001. For panel D, SARS-2 vs BANAL-52 p=.0001 and SARS-2 vs BANAL-236 p<0.0001. For panel E, SARS-2 vs BANAL-52 p=.003 and SARS-2 vs BANAL-236 p=0.007. (F) Infectious viral titers and (G) normalization values of SARS-CoV-2 and BANAL-CoVs in HNECs pre-treated with IFN-α. Graphs F-G represent pooled data from two independent experiments (n=6). Significance between viruses in IFN normalized to mock values (second graph) was assessed by one-way ANOVA corrected for multiple comparisons. For panel F, Mock treated cells SARS-2 vs BANAL-52 p<0.0001 and SARS-2 vs BANAL-236 p=.0006. For IFN treated cells, p=.0350. For panel G, SARS-2 vs BANAL-52 p=.0006 and SARS-2 vs BANAL-236 p=.0045. (H) Normalization of pseudovirus infectivity following IFN-α pre-treatment in Vero-ACE2/TMPRSS2 cells for SARS-CoV-2 and BANAL-CoVs. One out of three independent experiments are shown. N= 6 biological replicates for SARS-2 and BANAL-52, and n=5 for BANAL-236. Data from panels C-H are represented by mean +/− SD. (I) Gating strategy for flow cytometry analysis. (J) Representative graphs of detection efficiency of A549-ACE2/TMPRSS2 SARS-2- spike positive cells by flow cytometry. N=4 biological replicates out of 2 independent experiments. Significance between viruses was assessed by one-way ANOVA corrected for multiple comparisons. P values for panel J of SARS-2 vs mock infected, SARS-2 vs BANAL-52, and SARS-2 vs BANAL-236 were all p<0.0001. For all panels, error bars illustrate standard deviation, and ** indicates p<0.01, *** indicates p<0.001, **** indicates p<0.0001.

Extended Data Fig. 5. BANAL-CoVs do not induce robust innate immune response in mouse nasal turbinates.

(A) mRNA fold change of ISGs and inflammatory cytokines from lungs of infected mice. (B) mRNA fold change of ISGs and inflammatory cytokines from nasal turbinates of infected mice. Panels A-B were measured by RT-qPCR at 2 dpi. Fold change was assessed by 2–ΔΔCt method using b-actin as a control. Significance of expression between viruses was assessed by one-way ANOVA corrected for multiple comparisons. * indicates p<0.05, ** indicates p<0.01, **** indicates p<0.0001. Exact p values for panel A are: ISG15 (BANAL-52 vs BANAL-236 p=.0374), Mx1 (SARS-2 vs BANAL-52 p=.0213, BANAL-52 vs BANAL-236 p=.0172), IFIT1 (SARS-2 vs BANAL-52 p=.0035, BANAL-52 vs BANAL-236 p=.0049). OAS1 (BANAL-52 vs BANAL-236 p=.0062), TNF (SARS-2 vs BANAL-52 p<0.0001, BANAL-52 vs BANAL-236 p=.0423), IL-6 (SARS-2 vs BANAL-52 p=.0166, BANAL-52 vs BANAL-236 p=.0109), CXCL10 (SARS-2 vs BANAL-52 p=.0047, BANAL-52 vs BANAL-236 p=.0116). For panel B, exact p values are: IFIT1 (SARS-2 vs BANAL-52 p=.0240), CXCL10 (SARS-2 vs BANAL-52 p=.04), CXCL19 (SARS-2 vs BANAL-52 p=.0118 and SARS-2 vs BANAL-236 p=.0123). (C) Heatmap of ISGs and cytokines expression from nasal turbinates of K18-mice infected with the three sarbecoviruses measured by qPCRs. Values are from panel S5B. For panel Ain this figure the n of mice is: SARS-2 n=11, BANAL-52 n=12, BANAL-236 n=13. For panels B and C the n of mice is: SARS-2 n=8, BANAL-52 n=8, BANAL-236 n=8. Figure is pooled from three-independent experiments.

Extended Data Fig. 6. BANAL-CoVs do not induce lung inflammation in hamsters.

(A) Measurement of hamster ACE2 usage by the three sarbecovirus spikes using lentiviral-based pseudoviruses (n= 4 biological replicates). Significance was assessed by two-way ANOVA. Data is represented by mean +/− SD. Graphs represent one out of two independent experiments. Dashed line represents limit of detection. (B) H&E staining from infected hamster lungs with each of the three sarbecoviruses or uninfected control. Images are representative from 5 hamsters after 2 dpi.

Extended Data Fig. 7. SARS-2 FCS is enriched upon infection in hamsters.

(A) Nucleotide and aminoacid alignment of 7 most abundant reads from SARS-CoV-2 ΔFCS stock (Reads 1–7) and reads that have mutations in the FCS region but were not between the top 20 most abundant reads (Reads 8–10). (B) Raw reads from the S1/S2 spike junction region highlighting total number reads and reads with FCS sequence. (C) Percent of reads that possess FCS from virus stock and experimental samples from hamster lungs after NGS sequencing. (D) Fold change of the number of reads containing the FCS sequence comparing SARS-2 ΔFCS stock vs hamster infected with SARS-2 ΔFCS at 2 dpi. For sequencing analyses, 4 different tubes of SARS-2 ΔFCS stock and 2 samples from hamsters infected with SARS-2 WT and SARS-2 ΔFCS were used. For panels B and C, data are represented by mean +/− SD. For panel D, individual values are shown.

Data and reagent availability statement.

All data related to this manuscript were included in the source data files. GenBank accession codes for the molecular clones of BANAL-52 and BANAL-236 are PP856573 and PP856574, respectively. Viral stocks and plasmids are available with a material transfer agreement (MTA) and confirmation of appropriate biosafety facilities and procedures. Since potential experiments of concern could be performed with these viral strains, we will require a list of planned experimental procedures and to be reviewed before the MTA is executed. Yale University will review and determine the MTA timeline. Correspondence should be to Akiko Iwasaki (akiko.iwasaki@yale.edu) and Craig B. Wilen (craig.wilen@yale.edu).

Code availability statement.

The NGS sequencing raw data and the code to reproduce the analysis of the SARS-CoV-2 FCS analysis from the Extended Data can be found on Zenodo at https://zenodo.org/records/11519958⁵⁶.