Abstract
The nasal epithelium is a plausible entry point for SARS-CoV-2, a site of pathogenesis and transmission, and may initiate the host response to SARS-CoV-2. Antiviral interferon responses are critical to outcome of SARS-CoV-2. Yet little is known about the interaction between SARS-CoV-2 and innate immunity in this tissue. Here we applied single-cell RNA sequencing and proteomics to a primary cell model of human primary nasal epithelium differentiated at air-liquid interface. SARS-CoV-2 demonstrated widespread tropism for nasal epithelial cell types. The host response was dominated by type I and III IFNs and interferon-stimulated gene products. Nevertheless, this response was notably delayed in onset compared to viral gene expression, and thus failed to impact substantially on SARS-CoV-2 replication. However, when provided prior to infection, recombinant IFNβ or IFNλ1 induced an efficient antiviral state that potently restricted SARS-CoV-2 viral replication, preserving epithelial barrier integrity. These data suggest nasal delivery of recombinant IFNs to be a potential chemoprophylactic strategy.
INTRODUCTION
SARS-CoV-2 is an emergent betacoronavirus responsible for coronavirus disease-19 (COVID-19)1. Since its identification in late 2019, global pandemic transmission of SARS-CoV-2 has resulted in over 105 million infections and approximately 2.9 million deaths. SARS-CoV-2 infects target cells via the entry receptor ACE22 leading to a spectrum of clinical outcomes, ranging from asymptomatic infection to death3. Although multiple host factors (e.g. age, sex, obesity) contribute to adverse clinical outcome4, the immune response also plays a decisive role, evidenced by the therapeutic benefit of immunomodulatory agents including corticosteroids5. Yet much remains to be understood about the immunopathogenesis of COVID-19. Identification of the cells hosting viral entry and characterisation of their response to infection is essential to understanding pathogenesis and improving therapy. The nasal epithelium is believed to be a key entry point of SARS-CoV-2. Nasal epithelial tropism and efficient viral shedding from the nasopharynx apparently contributes to the high transmissibility of SARS-CoV-26, as well as to pathologic features such as anosmia7. As an early viral target cell, nasal epithelial cells may also set the tone for the systemic immune response, potentially influencing disease outcome8. These factors emphasise the need to study host-virus interaction in human nasal cells. Ex vivo single-cell transcriptomic studies indicate that ciliated and/or goblet cells in the nasal mucosa express ACE2 and TMPRSS2, implicating them as probable SARS-CoV-2 target cells9,10. This has been confirmed by in vitro and in vivo studies demonstrating SARS-CoV-2 infection of human nasal epithelial cells11-13. Single-cell studies also revealed that nasal cells exhibit basal expression of an antiviral expression programme, characterised by induction of several interferon-stimulated genes (ISGs), suggesting that they may be primed to respond to viral infection9. Interestingly, ACE2 is also regulated by interferons (IFNs) 10,14, implying a complex relationship between IFN signalling and tropism. Type I and type III IFN (IFN-I/III) systems play a critical role in human antiviral innate immunity15 and have been recently implicated in defence against SARS-CoV-2: susceptibility to severe or life-threatening COVID-19 is associated with deleterious variants in IFNAR genes16,17 and IFN-I blocking autoantibodies18. In vitro, SARS-CoV-2 appears sensitive to the antiviral properties of IFN-I, at least in cell lines19,20, and this activity extends to in vivo model systems8. These findings motivate studies to improve understanding of the interaction between SARS-CoV-2 and the IFN-I system in primary human target cells, providing impetus to clinical trials of recombinant IFNs in treatment or prophylaxis of COVID-1921.
Organotypic cultures of primary human nasal epithelium differentiated at air-liquid interface (ALI) are a translationally-relevant primary cell model for studies of SARS-CoV-2 host-virus interaction11, with considerable potential to accelerate our understanding of pathogenesis. A small number of studies using this model demonstrate that SARS-CoV-2 replicates efficiently in human nasal cells11-13, yet important questions concerning cellular tropism and their innate immune response remain unresolved. Hou and colleagues report that only ciliated cells were permissive to SARS-CoV-2, despite expression of ACE2 and TMPRSS2 by all cell types12. They hypothesised that post-entry factors, such as innate immunity, might govern tropism. By contrast, Pizzorno and colleagues reported infection in all major cell types (ciliated, secretory and basal cells)13, consistent with prior indications from single cell RNA sequencing (scRNA-seq) data and studies in lower airway models22,23. While an IFN response to SARS-CoV-2 can be detected in nasal cells11,13, in apparent contrast to bronchial or alveolar epithelial cells24-26, the kinetics of induction and the antiviral function of IFNs in nasal epithelium has not been systematically characterised.
Here we employed a comprehensive range of techniques, including scRNA-seq and proteomics, in primary human nasal ALI cultures to define: (i) cellular tropism; (ii) the innate immune response to SARS-CoV-2; and (iii) the antiviral activity of IFN-I/III. We observed broad cellular tropism of SARS-CoV-2 for nasal epithelial cells, although ciliated cells were the most permissive. Nasal cells mounted a delayed IFN response that failed to exert control over viral replication. However, SARS-CoV-2 remained highly sensitive to IFN-restriction if exogenous IFN-I/III was applied prior to infection. These data enrich our understanding of the interaction of SARS-CoV-2 and the human IFN system at the earliest point of infection, with immediate therapeutic implications.
RESULTS
SARS-CoV-2 robustly infects primary differentiated nasal epithelial cultures
Primary nasal epithelial cultures were established from cryopreserved stocks from six adult donors, obtained prior to the SARS-CoV-2 pandemic. Cells were expanded, differentiated and then matured at ALI for 28 days, according to an established protocol27. We first sought to address their suitability as a model of SARS-CoV-2 infection. Visualisation of scRNA-seq data from two representative donors discerned five major populations of cells (Fig. 1A). These were ciliated, secretory and basal cells, alongside two rarer populations of FOXN4+ cells and ionocytes. Cells expressed characteristic markers (Fig. 1B) and corresponded broadly to ex vivo single-cell data from nasal biopsies9,10. Immunostaining verified the presence of major cell types in these cultures - including acetylated alpha-tubulin-positive (AAT) ciliated cells, mucin 5B-positive (MUC5B) secretory cells, and tumour protein 63-positive (TP63) basal cells (Fig. 1C) - with ciliated cells being the most frequent cell population. Consistent with published scRNA-seq data, mRNA for key SARS-CoV-2 entry receptors, ACE2 and TMPRSS2, was expressed, albeit at relatively low levels, alongside other genes implicated in possible alternate modes of SARS-CoV-2 entry such as FURIN and CTSL (Fig. S1)28. Robust expression of ACE2 and TMPRSS2 at the protein level was confirmed by immunoblotting of whole-cell lysates prepared from mature ALI cultures (Fig. 1D). To establish their permissiveness to infection, nasal ALI cultures were inoculated at the apical surface with a clinical SARS-CoV-2 isolate (England/2/20) at an approximate multiplicity of infection (MOI) 0.1 - consistent with other studies (0.1-0.5)11-13 - and monitored for infection over the next 72 h. Expression of SARS-CoV-2 nucleocapsid (N) gene and spike (S) protein increased significantly over time, indicative of viral replication (Fig. 1E-F). This was accompanied by the release of infectious viral particles, as determined by plaque assay of apical washes on Vero E6 cells, confirming productive infection (Fig. 1G). SARS-CoV-2 replication was accompanied by a progressive decline in epithelial barrier integrity starting from 48 hours post-infection (hpi), reflecting virus-induced epithelial dysfunction (Fig. 1H). These data established the suitability of the human nasal ALI model for modelling SARS-CoV-2 infection.
Evidence of broad cellular tropism of SARS-CoV-2
To revisit the question of whether individual cell types are more or less permissive to SARS-CoV-212,13, we first examined viral gene expression by scRNA-seq analysis at 24 hpi, selected to represent an early stage in the progress of infection. Of the viral transcripts, the N gene was most abundantly expressed (Fig. 2A). While all cell types expressed viral transcripts, there were notable differences both in the proportion of cells infected, and the relative abundance of different viral transcripts within these cells. A greater proportion of ciliated and secretory cells expressed viral transcripts compared to basal cells (Fig. 2B). However basal cells are located away from the apical surface; physical inaccessibility to apically-applied virus at this timepoint might at least partially account for this observation. To investigate tropism further, we undertook immunofluorescence analysis of viral spike (S) protein expression at 48 hpi. This analysis revealed broadly similar proportions of ciliated, secretory and basal cells expressing S protein. However, the mean pixel intensity of S protein was significantly greater in ciliated cells (Fig. 2C-D). To corroborate these findings, we undertook analysis of intracellular virion-like structures (VLSs) at 48 hpi by transmission electron microscopy (TEM, Fig 2E), focusing on ciliated and secretory cells. Intracellular VLSs were observed in both ciliated and secretory cells, predominantly towards the apical surface (Fig. 2E). Consistent with immunofluorescence analysis of S protein intensity, there was a significant increase in the number of VLSs per cell in ciliated compared with secretory cells (Fig. 2E). Collectively, these data suggested that the virus is capable of entering, and replicating in, all major nasal cell types.
Characterisation of nasal cell responses to SARS-CoV-2
Published ex vivo single-cell transcriptomic analyses report that nasal cells exhibit the basal expression of an innate antiviral programme, in the absence of viral infection, characterised by several ISGs9. This suggests that nasal epithelial cells may adopt a constitutive antiviral state. It was also hypothesised that cellular innate immunity might influence nasal cell tropism of SARS-CoV-212.
In epithelial cells, ISGs are predominantly regulated by paracrine signalling mediated by IFN-I/IIIs. IFN-I/IIIs are produced by virally infected cells, downstream of pattern-recognition receptor (PRR) signalling, alerting their neighbours to viral infection. Responding cells adopt an antiviral state as a consequence of the expression of hundreds of antiviral ISGs29. Receptors for both IFN-I (IFNAR) and IFN-III (IFNLR) are expressed by epithelial cells. Ligand binding to these receptors activates a shared JAK-STAT signalling pathway, culminating in ISG expression15,29. However, ISGs can also be regulated by IFN-independent mechanisms. These include the direct action of constitutively expressed transcription factors such as IRF3, activated downstream of PRRs29, or by constitutive ISG expression - a strategy employed by stem cells for antiviral defence30.
We examined scRNA-seq data to characterise the innate antiviral response of nasal cells, both at baseline and in response to SARS-CoV-2 infection at 24 hpi. For this analysis we distinguished cells in three experimental conditions: unexposed (mock-infected); SARS-CoV-2-exposed but uninfected (these cells would theoretically be exposed to IFNs and other paracrine signals, but not infected); and SARS-CoV-2-infected (as defined by expression of S transcripts). At baseline, expression of a composite ISG signature (Fig. 3A) and individual ISGs (Fig. 3B) was observed in unexposed cell types, consistent with ex vivo scRNA-seq findings9. Considering that viral infection occurred in all cell types, these data imply that relative differences in the basal ISG expression programmes did not substantially impact on SARS-CoV-2 permissiveness, at least under the infection conditions studied here. We next considered the involvement of paracrine IFNs. In SARS-CoV-2 infected cells, but not their uninfected neighbours, we observed an enrichment of the ISG signature (Fig. 3A-B). This finding is more suggestive of direct induction of ISGs in infected cells, rather than paracrine IFN-I/III signalling, at least at this timepoint. Consistent with this, we observed expression of IFN-I (IFNB, IFNK, IFNA5) and IFN-III (IFNL1-3) in only a small minority (0.1%) of infected ciliated cells, and very low levels in other nasal cell types (Fig. 3C). Considering that over 50% of ciliated cells expressed viral transcripts at this timepoint, these data suggest that SARS-CoV-2 efficiently avoided IFN induction in the vast majority of infected epithelial cells, consistent with data in non-nasal airway epithelial cells25,26. A range of other chemokines and cytokines were induced by infected epithelial cells, indicating the preserved capacity to detect viral infection (Fig. S2).
Kinetics of innate IFN-I/IFN-IIIs response to SARS-CoV-2
To investigate the kinetics of the IFN-I/III response, expression of IFN-I (IFNA1 and IFNB) and IFN-III (IFNL1) was examined by RT-PCR at 6, 24, 48 and 72 hpi (Fig. 4A). Induction of IFNL1 and IFNB was low at 24 hpi, consistent with scRNA-seq findings, but increased significantly by 48 and 72 hpi. IFNA1 was not induced, as observed in our scRNA-seq data. Compared to the timing of initiation of viral gene expression - which was detectable at 6 hpi and approached its maximum level by 24 hpi (Fig. 1E) - the induction of IFNs appeared delayed, as suggested by previous studies11,13. Infection was accompanied by progressive upregulation of proinflammatory cytokines such as IL1B, IL6 and TNF, consistent with initiation of a local inflammatory response (Fig. 4B). To confirm that the attenuated production of IFNL1 and IFNB at early times was not dependent on MOI or due to an intrinsic defect of these cells, infections were repeated at high MOI (~ 2). Infections at high MOI have been previously reported to enhance the relatively inefficient IFN-I induction in response to SARS-CoV-225. A preparation of Sendai virus (SeV) - containing a high proportion of immunostimulatory defective viral genomes (DVGs) - was used as a positive control. At 6 hpi, a time point at which IFNB and IFNL1 were robustly induced by SeV DVGs, there was no response to SARS-CoV-2 infection at high MOI (Fig. 4C), despite SARS-CoV-2 gene expression (Fig. 4C). Thus SARS-CoV-2 appears to efficiently evade PRR detection in nasal epithelial cells. To look for evidence of a paracrine response to IFN-I/III, we undertook analysis of expression of the ISGs RSAD2 and USP18 by RT-PCR and immunoblotting (Fig. 4D-E). This analysis revealed an increase in ISG expression at later times (48-72 hpi), corresponding to the expression of IFNB/IFNL1. This association is suggestive (although not causal evidence) of paracrine IFN-I/III signalling. Collectively, these results indicate that nasal epithelial cells are capable of expressing IFN-I/IIIs during SARS-CoV-2 infection, but that the response is delayed relative to viral replication.
IFN-signalling dominates the nasal host response to SARS-CoV-2 at the protein level
To validate and extend these findings, we undertook an unbiased assessment of the host response to SARS-CoV-2 infection by proteomics analysis. Whole-cell lysates were prepared from SARS-CoV-2 and mock-infected nasal ALI cultures from six donors at 72 hpi. Lysates were analysed by quantitative mass spectrometry (Supplementary Dataset S1, quality control data in Fig. S3). Overall, this analysis detected the differential expression (DE) of 180 proteins including viral proteins such as S, M, N, ORF1AB, ORF3A and ORF8 (Fig. 5A). The most highly increased host protein was Sorting Nexin 33 (SNX33), an endosomal protein that has not yet been implicated in the life cycle of SARS-CoV-2. Infected and uninfected cells clustered together by principal component analysis (Fig. 5B). Inspection of the DE proteins confirmed a robust host innate immune response, dominated by ISG products (Fig. 5A). Functional annotation revealed an enrichment of antiviral response and especially IFN-I signalling pathways (Fig. 5C, Table S1). These data are consistent with our earlier findings and contrary to prior reports in cell lines or human bronchial/tracheal epithelial cultures, where a robust endogenous IFN-I/III response to SARS-CoV-2 was not detected24-26. Key ISG proteins identified included IFITM1-3 and the OAS cluster (OAS1-3), the latter associated with genetic susceptibility to severe COVID-1917 (Fig. 5C). Significantly downregulated pathways were also identified, including TRIF-dependent toll-like receptor signalling, as well as RNA polymerase II transcription and endosomal transport (Table S2). This implied viral subversion of critical host functions, including host gene transcription, protein trafficking and viral sensing. Proteins involved in the maintenance of epithelial tight junctions were also downregulated, consistent with the loss of barrier integrity observed in earlier experiments (Fig. 1H).
Antiviral activity of IFN-I/III towards SARS-CoV-2 infection
Given the prominence of the IFN-I/III response in the proteome of SARS-CoV-2-infected cells, a key question was whether the IFN-I/III response had any impact on SARS-CoV-2 replication. To address this question, nasal ALI cultures were treated with the JAK inhibitor ruxolitinib (RUX). RUX antagonises signalling downstream of IFNAR and IFNLR, owing to the involvement of JAK1 in both signalling pathways. We reasoned that blocking paracrine IFN-I/III signalling would reveal its impact, if any, on SARS-CoV-2 replication. Cells were treated with 10 µM RUX (a dose optimised in prior experiments31) or vehicle control (DMSO) in the basal medium for 24 hours prior to infection. Nasal cultures were infected at the apical surface (MOI ~ 0.01) and inhibitors were refreshed every 24h. At 48 hpi, lysates were prepared and analysed by immunoblot. RUX treatment abolished expression of RSAD2 and USP18 in response to SARS-CoV-2 infection (Fig. 6A), indicating that induction of ISG protein products at this timepoint was largely dependent on paracrine IFN-I/III signalling, as previously suggested (Fig. 4D). However, viral S protein expression was not enhanced by RUX treatment (Fig. 6A), nor was release of infectious virus, as measured by plaque assay (Fig. 6B). These data suggested that the endogenous IFN-I/III response failed to impact SARS-CoV-2 replication up to 48 hpi, at least at the MOI tested. These data provided further evidence that SARS-CoV-2 triggers an endogenous paracrine IFN-I/III response in nasal cells, but showed that this response was insufficient to contain SARS-CoV-2 replication.
An important follow-up question was whether nasal cells could mount an antiviral state to SARS-CoV-2, providing IFN-I/III was delivered in a timely fashion. To address this, nasal ALI cultures were pre-treated with exogenous IFNβ or IFNλ1 for 16h to induce an antiviral state, subsequently infected with SARS-CoV-2 at MOI 0.01, and examined at 48 hpi. Analysis of infection by immunoblotting of whole-cell lysates for spike (S) protein expression or plaque assay of apical washes demonstrated a significant reduction in infection with either IFNβ or IFNλ1 pre-treatment (Fig. 6C-D). This was accompanied by robust induction of antiviral ISG products (Fig. 6C), and preservation of barrier integrity (Fig. S4). It is worth noting that the ISG expression induced in response to recombinant IFN-I/III was substantially greater than that induced by SARS-CoV-2 infection (i.e. endogenous IFN-I/III production, Fig. 6C). Thus exogenous IFN-I/III was capable of inducing in the nasal epithelium an antiviral state that potently inhibited SARS-CoV-2 infection, providing it was delivered (a) prior to infection, and (b) at sufficient concentration. This IFN-sensitivity of SARS-CoV-2 contrasts with the relative resistance of SARS-CoV19. These data suggest that mucosal delivery of IFNβ is a potential therapeutic strategy for SARS-CoV-2. In clinical practice, IFNs are unlikely to be used prior to infection, unless this is part of a prophylactic regimen. To determine the effectiveness of exogenously applied IFN-I once SARS-CoV-2 infection is underway, cells were treated with IFNβ at 6 or 24 hpi. In this experiment, IFNβ treatment at 6 hpi continued to impact on SARS-CoV-2 infection, whereas addition after 24 hpi had no effect (Fig. 6E-F). Interestingly, ISG induction was still observed in response to IFNβ treatment at 24 hpi, albeit at reduced magnitude (Fig. 6E). These data suggest that SARS-CoV-2 impairs, but does not abolish, JAK-STAT signalling in infected cells, implying that recombinant IFNs may have a therapeutic role in established SARS-CoV-2 infection, as recently shown in animal models8 and in early phase clinical trials32.
DISCUSSION
We report the most comprehensive characterisation of the human nasal epithelial response to SARS-CoV-2 to date, revealing a response dominated by IFN-I/IIIs and their downstream ISG products. While this is apparently insufficient to contain SARS-CoV-2, recombinant IFN-I/III treatment potently blocked SARS-CoV-2 replication, suggesting that mucosal delivery of IFNs could be a promising strategy for post-exposure prophylaxis.
The nasal mucosa is likely to be a main point of entry of SARS-CoV-2. Prior single-cell transcriptomic studies implied an abundance of target cells in the nasal mucosa and further suggested that they may be poised to mount an antiviral response9. Yet few studies to date have characterised SARS-CoV-2 replication in primary human differentiated nasal cells11-13, and ours is the first to analyse the host-virus interaction comprehensively, at single-cell resolution. Our findings indicate that the host response to SARS-CoV-2 in nasal epithelium is dominated by IFN-I/III, albeit this response is kinetically delayed. These results are consistent with evidence of IFN-I/III induction in nasal swabs from patients with COVID-1933. However, we found that blockade of the endogenous IFN response did not impact on SARS-CoV-2 infection. At first glance, this result may appear paradoxical. However, the probable explanation is the delay of induction of IFN-I/III relative to virus replication. IFN-I/III expression became detectable 36-48 h after the onset of SARS-CoV-2 replication - at a point at which infectious virion production was reaching its maximum, and over half of cells in the culture were already infected - suggesting the production of IFNs was too late to have a meaningful impact on the overall spread of infection. Consistent with this, there was evidence of downregulation of PRR signalling in the proteome of infected cells. We conclude that: (i) SARS-CoV-2 appears capable of evading PRR signalling in nasal epithelial cells in the early stages of viral replication; and (ii) the endogenous IFN-I/III response of nasal epithelial cells comes effectively “too little, too late” to inhibit SARS-CoV-2 in the nasal epithelium. The expression of various IFN evasion proteins34 and the sequestration of viral replication machinery within cytosolic vesicles35 underlie the capacity of SARS-CoV-2 to evade early detection by PRRs, as indicated by our proteomics data. However, an important question is what molecular patterns are responsible for IFN-I/III induction at later times. Future studies should address the relative contribution of host damage-associated molecular patterns versus viral pathogen-associated molecular patterns (e.g. defective viral genomes) accumulating during replication.
Our finding that exogenous IFN-I or IFN-III can potently induce an antiviral state in nasal cells is consistent with its apparent protective effects in patients16-18 and in early phase clinical trials32. The main limitation of our data in this nasal epithelial culture system is that it did not account for professional immune cells present in the nasal mucosa, for example plasmacytoid dendritic cells36, which are capable of more rapidly mounting an IFN-I/III response to SARS-CoV-237, potentially tipping the scales in favour of the host. We studied cells derived from adult donors, however it is possible that nasal cells from paediatric donors may behave differently in terms of their permissiveness to SARS-CoV-2 and/or the efficiency of their innate immune response. Furthermore, SARS-CoV-2 variants with mutations in the spike gene have emerged worldwide whilst we were undertaking the experiments described here; these variants may impact viral replication and/or host immunity, and should be included in future studies.
Nevertheless, our data, employing a variety of complementary methods, indicate that SARS-CoV-2 has a relatively broad tropism for nasal epithelial cells, confirming suggestions from prior scRNA-seq studies7,9 and other in vitro studies of primary nasal13 and tracheobronchial cells22. These findings contrast with the results of Hou and colleagues, who reported exclusive tropism of SARS-CoV-2 for ciliated cells in the airway12. It is not immediately clear how to reconcile these findings, given that secretory cells express relevant entry receptors12. Hou and colleagues used a fluorescent reporter virus, the tropism of which might have been slightly narrower than clinical isolates. It is also worth noting our findings show that while all cell types contained SARS-CoV-2 protein, the intensity of immunodetection was greater in ciliated cells, which also contained more virion-like structures per cell. This implies that although all cell types are permissive, there may also be quantitative differences in the efficiency of viral replication in different cell types. Hou and colleagues previously hypothesised that post-entry factors such as antiviral immunity might dictate permissiveness. We found no evidence to support this hypothesis, since while all nasal epithelial cells demonstrated a basal ISG signature - consistent with ex vivo nasal biopsy data9 - this did not vary substantially between cell types, and was apparently insufficient to mediate resistance to SARS-CoV-2.
The observation that IFN-I treatment prevents SARS-CoV-2 infection indicates that chemoprophylaxis with IFN-I may have therapeutic value. This approach has already been tested in a small clinical trial in China (although the absence of a control group makes it impossible to judge the efficacy of this approach38). Immunisation is the most tractable approach for large-scale primary prevention of COVID-19. However, owing to incomplete vaccine coverage and/or effectiveness against mildly symptomatic or asymptomatic infection, and the emergence of variants that may compromise vaccine efficacy, there will likely continue to be a need for targeted chemoprophylactic therapies to prevent transmission in specific circumstances. These include post-exposure prophylaxis of contacts - to avoid the need for self-isolation - as well as pre-exposure prophylaxis for certain high-risk encounters (e.g. in healthcare settings or prior to long-distance travel). Our data suggest that application of IFNβ to the nasal mucosa might have an important role to play in this setting and argue for urgent clinical assessment of this approach. In terms of the therapeutic efficacy of IFNβ in patients with COVID-19, our findings suggest that early administration will likely be a key factor determining its efficacy.
METHODS
Adult nasal airway epithelial cell culture at air-liquid interface (ALI)
Adult primary human nasal airway epithelial cells were derived from excess clinical material obtained during routine nasal surgical procedures, processed, differentiated and validated as previously described27 with the following modifications: cells were cryopreserved following expansion, then re-animated and differentiated at ALI in PneumaCult-ALI-S (Stemcell Technologies). The sex and age of donors are included in Table S3. Ethical approval for sample collection was provided (Research Ethics Committee Reference 17/NE/0361).
Viruses, cytokines and inhibitors
A clinical isolate of SARS-CoV-2 (England/2/2020) was obtained from Public Health England (PHE). The initial stock was propagated once in Vero E6 cells. The same viral stock was used for all experiments. Sendai virus (Cantell Strain) was obtained from Richard Randall (St Andrew’s University). For nasal ALI infections, apical poles were gently washed once with warm Dulbecco’s modified Eagle’s medium (DMEM; Gibco, USA) and then infected with 60 μL dilution of virus in DMEM, at a MOI of between 2 and 0.01 plaque-forming units per cell for 2 hours, when the virus-containing medium was removed. DMEM was used as inoculum for mock infection. Apical washes (in warm phosphate-buffered saline) were collected at different time-points and stored at −80°C for plaque assays. Plaque assays were undertaken in Vero E6 cells using a 1.2% (w/v) microcrystalline cellulose overlay (Sigma-Aldrich). Cytokines/inhibitors were used at the following concentrations: human recombinant IFNβ1 (1000 ng/mL; Avonex, Biogen Inc, USA); IFNλ1 (100 ng/mL; R&D Systems, USA); and Ruxolitinib (10 µM; Calbiochem, USA) alongside the appropriate dilution of DMSO vehicle. Treatment was applied through basolateral poles.
Single cell RNA sequencing (scRNA-seq)
For droplet-encapsulation cells were processed as described in the supplementary information (SI). Principal component (PC) analysis was undertaken to find the first 20 PCs, which were batch-adjusted using Harmony and used to generate the nearest-neighbour graph. Dimensionality reduction and embedding was performed using Uniform Manifold Approximation and Projection (UMAP), with the neighbourhood graph clustered using the Leiden algorithm. The Wilcoxon Rank Sum test was used to identify differentially expressed genes between clusters, and these were annotated based on expression of markers used in the literature to define populations. A published gene set was used to generate a list of type I ISGs (www.gsea-msigdb.org/gsea/msigdb/cards/GO_RESPONSE_TO_TYPE_I_INTERFERON). This was used to quantify the interferon response within identified cell populations using the AddModuleScore tool in Seurat. Infected cells were assigned as those cells with detectable SARS-CoV-2 spike RNA (‘covid—S’ transcript). Further details are included in the SI.
Proteome sample preparation
The protein concentration was determined by EZQ® protein quantification assay. Protein digestion was performed using the S-Trap™ sample preparation method and TMT-16 plex labelling was carried out as per the manufacturer’s instructions. Samples were cleaned using MacroSpin columns, and dried down prior to offline high performance liquid chromatography fractionation. Peptides were fractionated on a Basic Reverse Phase column on a Dionex Ultimate 3000 off-line LC system. A total of 18 fractions were collected, and each fraction was acidified and dried. Peptides were dissolved in 5% formic acid, and each sample was independently analysed on an Orbitrap Fusion Lumos Tribrid mass spectrometer, connected to an UltiMate 3000 RSLCnano System. All spectra were analysed using MaxQuant 1.6.10.43 and searched against SwissProt Homo sapiens and Trembl SARS-CoV-2 FASTA files. Reporter ion MS3 was used for quantification and the additional parameter of quantitation labels with 16 plex TMT on N-terminus or lysine was included. A protein and peptide false discovery rate (FDR) of less than 1% was employed in MaxQuant. Moderated t-tests, with patient accounted for in the linear model, was performed using Limma, where proteins with an adjusted P < 0.05 were considered as statistically significant. All analysis was performed using R. Raw data are present in Supplementary Dataset 1. A comprehensive description of the methods can be found in the SI.
Quantitative RT-PCR, immunoblotting and immunofluorescence
RT-PCR and immunoblot were done as previously described31, with further detail in the SI. Membranes were fixed in situ with 4% formaldehyde overnight at 4°C, before sectioning. Immunofluorescent staining was performed in accordance with published methods39, with modifications as summarised in the SI. Images were captured using a Nikon A1 confocal microscope (Nikon, Japan) and analysed via Fiji (Version 2.0) using the Cell Counter plugin. Primers and probes (Table S4) and antibodies (Table S5) are described in the SI.
Transmission electron microscopy (TEM)
ALI cultures were fixed with 2% glutaraldehyde (Sigma-Aldrich, MO, USA) with 0.1 M sodium cacodylate (pH 7.4) buffer overnight at 4°C. TEM resin processing was performed as described previously40. Additional details, including embedding, sectioning and analysis, are included in the SI.
Statistical analysis
Statistical analysis was performed and figures assembled using GraphPad Prism V8 (GraphPad Software, USA). Data are presented as mean ± SEM of individual donor values (derived typically from 2-3 independent repeat experiments per donor). The donor was used as the unit of experiment for statistical analysis purposes. Continuous data were normalised or log-transformed prior to analysis using parametric significance tests. Differences between two groups were compared using an unpaired, two-tailed Student’s t-test, whereas differences between more than two groups used ANOVA, with Dunnett’s post-test correction for multiple comparisons. A two-tailed alpha of < 0.05 was the threshold for statistical significance. Statistical analysis of proteomics and transcriptomics data sets is described in the relevant sections above.
Data availability
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium41 via the PRIDE partner repository42 with the dataset identifier PXD021026. Reviewers can access it through the Username: reviewer_pxd022523{at}ebi.ac.uk and the Password: UaEXYFKF. RNA sequencing data have been deposited to the European Genome-phenome archive (accession pending). Additional raw data are available on request from the corresponding author.
Author contributions
Conceived the study: MB, CW, CJAD with MH, SH, MT. Experimental design: MH, SH, MB, CW, MT, GR, CJAD. Nasal model development and patient material: IJH, BV, JS, JPG, SC, AJS, MB, CW. Virology data generation, analysis and interpretation: CFH, BJT, JSS, FG, AIG, TD, SH, CJAD. Proteomics data generation, analysis and interpretation: MED, SM, MT. Single-cell sequencing data generation, analysis and interpretation: RAB, ES, RH, JC, MH, GR. Supervised research: AJS, MH, SH, MB, CW, MT, GR, CJAD. Drafted the manuscript: CFH and CJAD with ES, BV, MED, TD, MT, GR. Revised the manuscript: RAB, MED, IJH, JSS, AJS, MH, SH, CW. Approved the manuscript for submission: all authors.
Acknowledgements
We acknowledge Public Health England for providing the SARS-CoV-2 isolate and Professor R Randall (St Andrew’s University) for providing Sendai virus and the Vero E6 cell line. We thank A Khan, M Glanville and S Dainty (Newcastle University Infectious Disease Facility) and A Laude (Bioimaging facility) for assistance. The work was part-funded by the Barbour Foundation (CJAD, SH, MB, MT, MH, CW), the UK-Coronavirus Immunology Consortium (CJAD, SH, MH) and the Medical Research Council SHIELD antimicrobial resistance consortium (JS, AJS). CFH is supported by a Medical Research Council studentship (MR/NO13840/1) and MB by an MRC Clinician Scientist Fellowship (MR/M008797/1). MT and SH are funded by a Wellcome Trust Investigator Awards (215542/Z/19/Z and 207556/Z/17/Z). CJAD and GR are Wellcome Trust Clinical Research Career Development Fellows (211153/Z/18/Z and 214539/Z/18/Z). MED is a Marie Sklodowska Curie Fellow within the European Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No. 890296. MH is funded by Wellcome (WT107931/Z/15/Z), The Lister Institute for Preventive Medicine and Newcastle NIHR Biomedical Research Centre (BRC). JPG, CW and BV were supported by the Medical Research Foundation (MRF Respiratory Diseases Research Award to JPG; Grant MRF-091-0001-RGGARNE) and Boehringer Ingelheim. TEM work was supported by a BBSRC Alert17 grant (BB/R013942/1). Professor Simpson is a National Institute for Health Research (NIHR) Senior Investigator. The views expressed in this article are those of the author(s) and not necessarily those of the NIHR, or the Department of Health and Social Care.