The human airway and nasal organoid culture system was previously established by Sachs et al. and Rajan et al. [36,37]. In this study, we made specific modifications to the system and successfully generated respiratory organoid models, including 3D cultures of hNOs and hLOs (Figure S1a–S1c). The respiratory organoid culture medium provided optimal conditions for robust growth and long-term propagation of human respiratory organoids, allowing continuous passaging for up to 6 months (Figure S1b, c).

The presence of diverse respiratory epithelial cell types in hNOs and hLOs was detected by the expression of specific cell markers, including KRT5 in basal cells, MUC 5AC in goblet cells, Acetylated α Tubulin (ACCTUB) in ciliated cells, CC10 in club cells and SFTPC in alveolar type II (AT2) cells (Figure S1d, e). Notably, the hLOs also expressed the ACE2 receptor, which is shown to be crucial for infection of certain coronaviruses (Figure S1e). These findings demonstrate that the organoid models we generated had a high degree of similarity in cellular composition and characteristics with corresponding human tissues.

Using the organoid models generated above, we first assessed infection characteristics of five HCoVs (SARS-CoV-2 original strain, Omicron BA.1, MERS-CoV, HCoV-229E, and HCoV-OC43) using, with hNOs representing the upper respiratory tract, and hLOs representing the lower respiratory tract. A comprehensive evaluation system for coronavirus pathogenicity was constructed based on five critical indicators: cell tropism, virus replication activity, invasion preference, host response, and cell apoptosis & death. We then used this evaluation system and assessed the interspecies infection risk of three animal-borne coronaviruses, bat coronavirus WIV1 and rRsSHC014S [38], and pangolin coronavirus MpCoV-GX [25] from various aspects (Figure 1(a)). At last, we tested the efficacy and efficiency of CBD upon viral infection in the organoid models and proposed a potential prevention and treatment strategy for diverse coronaviruses. Figure 1.

Cell tropism of HCoVs in human respiratory organoids. (a) Scheme of this study. (b) Cell tropism of five HCoVs in hNOs. (c) Cell tropism of five HCoVs in hLOs. hNOs and hLOs were infected with five HCoVs respectively (MOI = 1). Organoid samples were harvested at 48 hpi (n = 3). Immunofluorescence staining was performed. Cell nuclei (DAPI, blue), virus (NP, red), and epithelial cell markers (ciliated cells: ACCTUB, club cells: CC10, goblet cells: MUC5AC, alveolar type II cells: SFTPC, green) were stained. Scale bar = 50μm.

The cellular tropism of a virus determines its infection and transmission route, and is closely associated with disease severity and host response. Our data indicated that SARS-CoV-2 original strain, Omicron BA.1, MERS-CoV, and HCoV-OC43 were capable of infecting club, goblet, ciliated cells in hNOs, and club, goblet, ciliated, and AT2 cells in hLOs. By comparison, HCoV-229E failed to infect ciliated cells in both hNOs and hLOs (Figure 1(b,c), Table 1). Table 1.

Comprehensive weighted score.

Based on the minimal variations observed across different viral strains, we assigned 10% of the weightage to the indicator of cellular tropism and scores were given accordingly for different viruses in the evaluation system (Supplementary text). For example, SARS-CoV-2 original strain, Omicron BA.1, MERS-CoV and HCoV-OC43 received an identical score in this regard, while HCoV-229E obtained a slightly lower score (Supplementary text, Table 1, Figure S5a).

The infectivity of various viruses towards human organoids displays significant variability, influenced by factors such as differences in virus receptor and entry co-factor expression, host innate immunity, and microenvironmental disparities. We assessed the viral replication and release of five HCoVs in organoids, with viral load within the organoids indicating replication ability. MERS-CoV exhibited the most extensive tissue tropism and replication capabilities as indicated by efficiently infecting and releasing in both hNOs and hLOs. Notably, its replication level was higher in the lower respiratory tract region of the lungs. While SARS-CoV-2 original strain infected both hNOs and hLOs, it displayed weaker release in the former and a stronger preference for the latter. Conversely, Omicron BA.1 showed weaker infection capacity in both types of organoids but more robust virus release capability in hNOs. HCoV-229E demonstrated a preference for infecting and releasing only in the upper respiratory tract region of the lungs and HCoV-OC43 can infect cells in both regions but has weaker virus release ability in hNOs (Figure 2(a)). These findings provide evidence for the idea that tissue tropism correlates directly with virus pathogenicity. Viruses that preferentially target upper respiratory tracts tend to exhibit relatively weaker pathogenicity compared to those targeting lower respiratory tracts. Figure 2.

Replication and invasion variability of HCoVs in human respiratory organoids. (a) Five HCoVs show distinct variability in hNOs and hLOs. hNOs and hLOs were infected with five HCoVs respectively (MOI = 0.1). Samples were harvested at 3, 24, 48 and 72 hpi. Viral load in culture supernatant or within cells was quantified by qRT-PCR on total viral RNA (NP gene as target) (n = 3). Dashed line in supernatant plot: effective release, virus amplification more than 100-fold; dashed line in intracellular plot: effective replication, virus amplification more than 10-fold. (b and c) Different HCoVs exhibit distinct invasion mechanisms in hNOs and hLOs. Organoids were pre-treated with 100 μM E-64D or camostat mesylate for 1 h (camostat mesylate, a TMPRSS2 inhibitor; E64-D, a CTSL/B inhibitor) before infected with five HCoVs respectively (MOI = 1). Organoid cell samples were collected at 24 hpi. Viral load in cells was quantified by qRT-PCR on total viral RNA (NP gene as target) (n = 3). Data are the mean ± SEM. Statistical significance is analysed by Student’s t-test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).

Coronaviruses are enveloped viruses that predominantly invade host cells through two distinct pathways, direct membrane fusion and endocytosis. The viral spike protein is cleaved by TMPRSS2 for the former pathway, while proteases such as cathepsin B (CTSB) and cathepsin L (CTSL) cleave the viral spike protein and facilitate fusion with endosomal membranes for the latter pathway. However, it is unclear whether coronaviruses with different pathogenicity employ preferential invasion pathways into host cells.

To answer this question, we used two inhibitors, membrane inhibitor camostat mesylate and endocytosis inhibitor E64-D to block individual virus entry pathway. In hNOs, treatment with camostat mesylate but not E64-D, led to a significant decrease in viral loads of the two SARS-CoV-2 strains and MERS-CoV while virus load of HCoV-229E and HCoV-OC43 decreased significantly after treatment with both camostat mesylate and E64-D (Figure 2(b)). Similarly, in hLOs, the viral loads of the SARS-CoV-2 original strain and MERS-CoV were significantly diminished following treatment with camostat mesylate but not E64-D. Additionally, both Omicron BA.1 and HCoV-OC43 exhibited a substantial reduction in viral loads after treatment with camostat mesylate and E64-D. By contrast, HCoV-229E was only affected by E64-D treatment (Figure 2(c)). These findings suggest that highly pathogenic viruses tend to utilize the direct membrane fusion pathway for cell invasion in both the upper and lower respiratory tracts, whereas low pathogenic viruses prefer the endocytic pathway.

Therefore, we considered viral invasion preference as the second indicator within the comprehensive evaluation system, accounting for a weight of 15%. This indicator encompassed both tissue tropism and invasion pathway preference. According to the scoring criteria (Supplementary text), the ranking of the five HCoVs, from highest to lowest, was as follows: MERS-CoV, HCoV-OC43, SARS-CoV-2 original strain, HCoV-229E, Omicron BA.1 (Supplementary text, Table 1, Figure S5a).

To probe variations in the replication activity among different HCoVs, we quantified virus subgenomic RNA (sgRNA) levels in organoid cells by qRT-PCR and virus titres in culture supernatants by plaque assays. Due to the low overall infection rate, sgRNA expression levels are generally low in organoid cells. Notably, Omicron BA.1 was detectable in both hNOs and hLOs as early as 3 hpi, indicating its rapid release, which was consistent with findings observed in other cell models [39] (Figure 3(a,b)). In hNOs, both HCoV-229E and -OC43 showed viral titres higher than Omicron BA.1 in the supernatant, in agreement with higher levels of actively replicating viruses within organoid cells. HCoV-OC43 was detectable at 3 hpi, while HCoVs-229E maintained constantly high viral titres. In contrast, both SARS-CoV-2 original strain and MERS-CoV displayed relatively slow replication rates with viral titres detectable only after 48 hpi (Figure 3(c)). For hLOs, SARS-CoV-2 original strain demonstrated the highest viral titre and the most rapid replication rate, with MERS-CoV following closely behind (Figure 3(d)). Both viruses can constantly sustain high viral titres, underscoring their pronounced tissue affinity for the lower respiratory tract. In comparison, the other three HCoVs displayed lower viral titres (Figure 3(d)). The overall pattern indicates that highly pathogenic coronaviruses tend to infect hLOs, while low pathogenic coronaviruses tend to infect hNOs. Figure 3.

Human respiratory organoids support replication of HCoVs. (a) Differential sgRNA replication activity of five HCoVs in hNO cells. (b) Differential sgRNA replication activity of five HCoVs in hLO cells. (c) Variation in the viral titres of five HCoVs in the culture supernatant of hNO cells. Dashed line: the lowest limit of detection threshold. (d) Variation in the viral titres of five HCoVs in the culture supernatant of hLO cells. Dashed line: the lowest limit of detection threshold. hNOs and hLOs were infected with five HCoVs respectively (MOI = 0.1). Samples were harvested at 3, 24, 48 and 72 hpi (n = 3). Viral load in organoid cells was quantified by qRT-PCR and viral load in culture supernatant was quantified by plaque assay. Error bars represent the mean ± SEM.

Based on the strength of viral replication as well as viral titre rankings across all five viruses, we assigned the third indicator as virus replication activity, which accounted for 20% of the weight, and calculated a composite index (Supplementary text). As regards to the virus replication activity, the five viruses ranked in descending order as follows: HCoV-OC43, HCoV-229E, SARS-CoV-2 original strain, MERS-CoV, Omicron BA.1 (Supplementary text, Table 1, Figure S5a).

Upon coronavirus infection, organoids promptly exhibit immediate non-specific host responses. The innate immunity plays a crucial role as an early defense mechanism by swiftly inducing interferon responses. We categorized the host response into two parts: initiation of interferon expression and acute local inflammation marked by interleukin expression as well as widespread inflammation associated with cytokine and chemokine expression. We evaluated representative host factors in each part and found that organoids exhibit differential responses to various HCoVs (Figure 4(a)). SARS-CoV-2 original strain, MERS-CoV, and Omicron BA.1 elicited much robust host responses characterized by an initially weak then increasing or continuously increasing interferon response pattern. Moreover, these three viruses induced higher levels of overall inflammatory cytokines/chemokines expression, suggesting a more pronounced inflammatory response compared to the other two viruses (Figure 4(a)). Figure 4.

Host responses to HCoV infection in hLOs. (a) Induction of antiviral and inflammatory responses by infection of five HCoVs (MOI = 0.1). Organoid cell samples were harvested at 24, 48 and 72 hpi (n = 3). Gene expression was quantified by qRT-PCR. (b) Proportion of infected cells in hLOs after viral infection (MOI = 1). (c) Cell death and apoptosis rates induced by virus infection (MOI = 1). Organoid samples were digested into single cells, stained with apoptosis and death probe, fixed, and labelled with virus NP antibodies for flow cytometry analysis (n = 3). Data are the mean ± SEM. Statistical significance is analysed by Student’s t-test (*p < 0.05; **p < 0.01; ***p < 0.001).

Host response, which plays a pivotal role in virus pathogenicity, served as the fourth evaluation indicator and carried a substantial weight of 30%. The strength of host responses was evaluated by analysing the trend of the response and calculating the area under the curve (AUC) using relative gene expression levels as indicators. Scores were ranked in descending order as follows: MERS-CoV, SARS-CoV-2 original strain, Omicron BA.1, HCoV-229E, -OC43 (Supplementary text, Table 1, Figure S5a).

To investigate the cell apoptotic and death responses induced by different coronaviruses, we employed the Apotracker™ Green probe to indicate cell apoptosis and the Zombie Aqua^TM probe to detect cell death. At 24 hpi, infection of SARS-CoV-2 original strain significantly increased cell death, while HCoV-229E infection led to a marked increase in both cell apoptosis and death. In contrast, HCoV-OC43 infection resulted primarily in a significant increase in cell apoptosis. Notably, at 72 hpi, no significant changes in apoptosis or death were observed for most viruses, except for Omicron BA.1, which was associated with suppressed cell apoptosis (Figure 4(b,c), S3). Therefore, regarding indicator of host cell apoptosis and death, which accounted for 25% of the evaluation, we assigned identical scores to SARS-CoV-2 original strain, Omicron BA.1, and MERS-CoV, and lower scores to HCoV-229E and -OC43 (Supplementary text, Table 1, Figure S5a).

Based on the five criteria mentioned above, we developed the comprehensive coronavirus evaluation system and calculated the total weighted score for individual HCoV (Supplementary text, Table 1, Figure S5a). The five HCoVs were ranked in descending order as follows: MERS-CoV, SARS-CoV-2 original strain, HCoV-OC43, Omicron BA.1 and HCoV-229E (Figure S5c). These results align well with clinical data for the infection capacity of five human coronaviruses, as a proof of concept for the reliability of using this system to predict clinical outcomes following coronavirus infection.

With the comprehensive coronavirus evaluation system, we further examined the interspecies infection and pathogenicity capacity of three animal-borne SARSr-CoVs. With regard to cell tropism, WIV1 and rRsSHC014S were capable of infecting all tested human respiratory epithelial cell types, whereas MpCoV-GX failed to infect goblet cells and AT2 cells in hLOs (Figure S2a,b, Table S1).

To investigate tissue tropism of these viruses, we infected both the respiratory organoids and intestine organoids with three animal-borne SARSr-CoVs. WIV1 demonstrated high replication in hLOs, human small intestinal organoids (hIOs), and human colonic organoids (hCOs), but successful virus release was only observed in the hCOs. rRsSHC014S replicated in hNOs but failed to release in all four types of organoids. MpCoV-GX displayed relatively low replication capacity in all four types of organoids but released only in hLOs (Figure 5(a)). These findings suggest that WIV1 has a broader tissue tropism than the other two viruses. rRsSHC014S shows a preference for infecting upper respiratory tissue. MpCoV-GX exhibits weaker overall replication ability with a stronger preference for gastrointestinal tissues. Overall, the three animal SARSr-CoVs exhibited lower infection capacity to respiratory and digestive tract compared to SARS-CoV-2 original strain (Figure 2(a), S4). Figure 5.

Tissue Tropism and Invasion Mechanism Preferences of animal-borne coronaviruses. (a) Replication levels of three animal-borne coronaviruses in hNOs, hLOs, hIOs and hCOs (MOI = 0.1, n = 3). Dashed line in supernatant plot: effective release, virus amplification more than 100-fold; dashed line in intracellular plot: effective replication, virus amplification more than 10-fold. (b) Impact of protease inhibitors on three animal-borne coronaviruses in hNOs and hLOs (MOI = 1, n = 3). (c) SgRNA levels of three animal-borne coronaviruses in hNO and hLO cells (MOI = 0.1, n = 3). (d) Virus titres of three animal-borne coronaviruses in hNO and hLO culture supernatants (MOI = 0.1, n = 3). Dashed line: the lowest limit of detection threshold. Data are the mean ± SEM. Statistical significance is analysed by Student’s t-test (**p < 0.01; ****p < 0.0001).

As for invasion preference, treatment with camostat mesylate resulted in a significant reduction in WIV1 viral loads in hNOs and hLOs. In contrast, rRsSHC014S exhibited comparable reliance on both direct membrane fusion and endocytic pathways in hNOs but no significant inhibitory effect was observed in hLOs. MpCoV-GX showed no significant changes upon inhibitor treatment in both hNOs and hLOs, which may be attributed to its low levels of viral replication (Figure 5(b)).

We next examined the replication activity of the three animal-borne SARSr-CoVs. Within the organoid cells, both WIV1 and rRsSHC014S displayed higher sgRNA replication levels compared to MpCoV-GX in hNOs. WIV1 and MpCoV-GX exhibited stronger sgRNA replication activity in hLOs than rRsSHC014S (Figure 5(c)). In the supernatant of organoid culture, WIV1 and rRsSHC014S were detectable in hNOs only at 48 hpi, whereas in hLOs, they were readily detected from 24 hpi to 72 hpi, indicating that WIV1 and rRsSHC014S replicated more rapidly and demonstrated greater infectivity in the lower respiratory tract. Live virus titres of MpCoV-GX were undetectable in both hNOs and hLOs (Figure 5(d)).

Host response assays revealed that, during the early stages of infection, rRsSHC014S and MpCoV-GX induced a robust upregulation of IFNλ1 expression, while WIV1 infection resulted in downregulation of IFNB expression. In the mid-stage of infection, MpCoV-GX continued to upregulate IFNλ1 expression, whereas WIV1 infection sustained the downregulation of IFNB expression. Moreover, MpCoV-GX infection triggered the upregulation of tumour necrosis factor-α (TNFα), C-X-C motif chemokine ligand 10 (CXCL10), and monocyte chemotactic protein 1 (MCP1) expression. As the infection progressed to the late stage, while IFNB downregulation and IL1β upregulation persisted with WIV1 infection, CXCL10 expression was significantly reduced, indicating controlled inflammation. rRsSHC014S infection was associated with downregulation of most genes, suggesting that the infection was under control without a remarkable host response. MpCoV-GX infection induced an upregulation of certain inflammatory cytokines and chemokines indicating its potential pathogenicity (Figure 6(a)). Figure 6.

Host responses to animal-borne coronavirus infection in hLOs. (a) Induction of antiviral and inflammatory responses by infection of three animal-borne coronaviruses (MOI = 0.1). Organoid cell samples were harvested at 24, 48 and 72 hpi (n = 3). Gene expression was quantified by qRT-PCR. (b) Proportion of infected cells in hLOs after viral infection (MOI = 1). (c) Cell death and apoptosis rates induced by infection (MOI = 1). Organoid samples were digested into single cells, stained with apoptosis and death probe, fixed, and labelled with virus NP antibodies for flow cytometry analysis (n = 3). Data are the mean ± SEM. Statistical significance is analysed by Student’s t-test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001).

As regards to virus-induced cell death, varying degrees of host cell apoptosis or death were detected with different animal coronaviruses. At 24 hpi, all three coronaviruses induced significant cell apoptosis and death. By 72 hpi, however, the cell viability largely recovered and cells infected by rRsSHC014S showed a significant decrease in apoptosis (Figure 6(b,c)).

Finally, we applied the comprehensive coronavirus evaluation system to score the infection characteristics of three animal-borne coronaviruses and compared their scores to those of the five HCoVs (Supplementary text, Table 1, Figure S5b,c). In terms of infection capacity, the viruses were ranked in descending order as follows: MERS-CoV, SARS-CoV-2 original strain, HCoV-OC43, Omicron BA.1, HCoV-229E, WIV1, rRsSHC014S, MpCoV-GX (Figure S5c).

By use of the organoid models, we tested the direct impact and antiviral effect of the cannabis extract CBD on hLOs. The results showed that 10 μM CBD did not have any negative impact on organoid formation (Figure S6a). And no significant changes were observed in the expression levels of coronavirus receptors and proteases, apart from an increased expression of the HCoV-229E receptor aminopeptidase N (APN) (Figure S6b). To assess the antiviral effectiveness of CBD against the evolving SARS-CoV-2 virus, we utilized the newly emerged variant, Omicron BA.5. In the supernatant of hLO culture, CBD prominently suppressed the levels of SARS-CoV-2 original strain, Omicron BA.5, MERS-CoV, and HCoV-OC43 (Figure 7(a)). Within hLO cells, replication of Omicron BA.5, MERS-CoV, HCoV-OC43, and WIV1 was significantly inhibited by CBD, as indicated by viral load (Figure 7(b)) as well as sgRNA levels (Figure S7). However, CBD did not show inhibitory effects on HCoV-229E replication, possibly due to the enhanced expression of its receptor, APN (Figure S6b). Additionally, the inhibitory effects of CBD on rRsSHC014S and MpCoV-GX were not as apparent, presumably due to the low infection levels of these two viruses in hLOs (Figure 7(a,b)). Figure 7.

CBD as a potential broad-spectrum inhibitor of coronaviruses (a) Changes in viral load in hLO culture supernatant after CBD treatment. (b) Changes in viral load in hLO cells after CBD treatment. (c) Changes in ISG response in hLOs after CBD treatment. (d) Changes in inflammatory response in hLO after CBD treatment. HLOs were pre-treated with 10 μM CBD for 12 h before infection with eight coronaviruses respectively (MOI = 1). After infection, the hLOs were cultured with 10 μM CBD for 24 h (n = 3) before harvest. Viral load or related gene expression was then quantified by qRT-PCR. Data are the mean ± SEM. Statistical significance is analysed by Student’s t-test (*p < 0.05; **p < 0.01; ***p < 0.001; ****p < 0.0001). In figure c and d, pairwise comparisons were performed between the CBD-treated and DMSO-treated groups for each set of experiments.

Upon viral infection, CBD significantly upregulated the expression of interferon-stimulated genes, including IFIT3, ISG15, and SOCS1 (Figure 7(c)). Additionally, CBD effectively reduced the release of inflammatory cytokine TNFα (Figure 7(d)). To test whether this effect is attributed to CBD's ability to enhance ER stress response, we examined ER stress-related gene expression in hLOs after CBD treatment and revealed only a mild influence on ER stress by CBD (Figure S8). In summary, our results supported the general antiviral effect of CBD possibly independent from its role in regulating ER stress.