Authors: Chien-Ting Wu , Peter V. Lidsky , Yinghong Xiao , Carlos Milla, Raul Andino, Peter K. Jackson  December 01, 2022DOI:https://doi.org/10.1016/j.cell.2022.11.030

Highlights

• •SARS-CoV-2 binds ACE2 on multicilia in airway epithelia immediately upon infection
• •Depleting motile cilia inhibits viral entry by SARS-CoV-2 and other respiratory viruses
• •SARS-CoV-2 activates PAK kinases to rearrange airway microvilli driving viral exit
• •Omicron variants accelerate cilia-dependent entry through the airway mucin barrier

Summary

How SARS-CoV-2 penetrates the airway barrier of mucus and periciliary mucins to infect nasal epithelium remains unclear. Using primary nasal epithelial organoid cultures, we found that the virus attaches to motile cilia via the ACE2 receptor. SARS-CoV-2 traverses the mucus layer, using motile cilia as tracks to access the cell body. Depleting cilia blocks infection for SARS-CoV-2 and other respiratory viruses. SARS-CoV-2 progeny attach to airway microvilli 24 h post-infection and trigger formation of apically extended and highly branched microvilli that organize viral egress from the microvilli back into the mucus layer, supporting a model of virus dispersion throughout airway tissue via mucociliary transport. Phosphoproteomics and kinase inhibition reveal that microvillar remodeling is regulated by p21-activated kinases (PAK). Importantly, Omicron variants bind with higher affinity to motile cilia and show accelerated viral entry. Our work suggests that motile cilia, microvilli, and mucociliary-dependent mucus flow are critical for efficient virus replication in nasal epithelia.

Graphical abstract

Introduction

SARS-CoV-2 caused the COVID-19 pandemic. After the viral spike glycoprotein (SP) binds to its host receptor, the angiotensin-converting enzyme 2 (ACE2) and the host transmembrane serine protease 2 (TMPRSS2) prime SP to facilitate viral fusion, and the virus enters the host cells and begins replication.

 Viral replication has been modeled in tissue culture. However, the primary site of SARS-Cov-2 replication is the upper respiratory tract, which is specialized to create a barrier to virus infection. Indeed, the mechanisms of virus cell entry, exit, and cell-to-cell spread in the airway epithelium are poorly understood.

Nasal airways are pseudostratified epithelia, including multiciliated apical epithelial, basal, and mucus-producing goblet cells. The mucus layer sits atop the ciliary brush, and an underlying periciliary layer (PC) keeps mucus away from the epithelia but permits ciliary beating.

 Mucus traps virus particles that are swept to the laryngopharynx by ciliated epithelial cells (CECs). CECs comprise 80% of nasal epithelium, and each express ∼300 cilia that show coordinated beating. Mucociliary clearance (MCC) eliminates infectious particles by coughing or swallowing. The underlying PCL mucin layer is critical to blocking viral entry. Small particles (∼25 nm) penetrate the PCL, but larger particles (∼100 nm) cannot. Depleting core PCL components increases influenza infection in various cells.

 Yet, many large, enveloped viruses (such as influenza and SARS-CoV-2) infect airway epithelial cells. How respiratory viruses bypass the innate barriers and cross back to spread remains unclear. Nasal and upper airway CECs are the main targets for initial COVID-19 infections. Many respiratory viruses, including SARS-CoV-1, influenza, parainfluenza (PIV), rhinovirus (RV), and respiratory syncytial virus (RSV), first infect airway CECs. SARS-CoV-2 receptors ACE2 and TMPRSS2 localize to cilia in CECs.

 Here, we report that SARS-CoV-2 infects nasal mucosa by a two-step process. First, SARS-CoV-2 particles bind the ACE2 receptor on the surface of airway cilia. Cilia then facilitate virus transport through the PCL mucin layer. Initially, only a few ciliated human nasal epithelial cells (HNEs) are productively infected. Within the next 24 h SARS-CoV-2 hijacks the host cell machinery to induce elongated and highly branched microvilli. This rearrangement requires protein kinases, including p21-activated kinases 1 and 4 (PAK1/4). Activating p21-activated kinases (PAK) enables the virus to exit across the PCL layer before lateral spread to other regions, potentially via mucociliary transport (MCT). Omicron variants dramatically accelerate spread via the ciliary transport/microvilli reprogramming pathway, which explains the increase in its attack rate compared to previous variants. This pathway is required for infection by other respiratory viruses, such as PIV and RSV. Understanding how cilia and microvilli are reprogrammed for virus entry and spread may identify candidate therapeutic targets to block airway replication of SARS-CoV-2 and other respiratory viruses.

Results

SARS-CoV-2 infection of nasal epithelium is a two-step process

To understand viral entry, we sought to identify which cell types are initially infected. We differentiated primary HNEs in air-liquid interface (ALI) cultures to form nasal epithelial organoids of ciliated, goblet, and basal cells (see Table S1 for characteristics of the nasal cell donors). These ALI-cultured HNEs recapitulated the morphology and function of normal human upper airway epithelium.

 After 28–30 days, fully differentiated nasal epithelial organoids were inoculated with the SARS-CoV-2 D614G variant at a MOI of 0.3 to assess spatial and temporal infection patterns and cell tropism within nasal epithelium. Infected cultures were fixed at various times and stained with antibodies to the SARS-CoV-2 nucleocapsid protein (NP), SP, and markers for ciliated HNEs (acetylated tubulin; ACTUB) or goblet cells (MUC5AC) (Figures 1A , 1B, and S1). NP and SP were only seen in ciliated HNEs at 6, 24, and 48 h post infection (hpi), indicating preferential, early infection of ciliated HNEs. SARS-CoV-2 infected few, if any, goblet cells even at 96 hpi. Thus, ciliated HNEs are the primary entry site of SARS-CoV-2 in nasal epithelia.

At 6 and 24 hpi, only ∼3% of ciliated HNEs were SARS-CoV-2 positive; by 48 hpi, positive HNEs increased to ∼80% (Figure 1A). We hypothesized that initial infection and entry are restricted to a few ciliated HENs, and then later, new virions spread laterally into neighboring nasal epithelium. The pattern did not change at different MOIs (MOI: 3, 0.3, and 0.03), suggesting a kinetic bottleneck for infection at viral entry (Figure 1C). These results support a two-step model in which SARS-CoV-2 crosses the airway mucosal barrier in a limited number of ciliated cells, followed by replication and rapid spread of progeny to neighboring cells (Figure 1D).

SARS-CoV-2 attaches to the cilia during initial stages of infection

ACE2 and TMPRSS2 are central to SARS-CoV-2 entry and localize to respiratory cilia. To determine if they also localize to motile cilia in ALI-cultured HNEs, we co-stained ACE2 and TMPRSS2 with ACTUB. Both localized to motile cilia (Figure S2A) in the ALI-cultured HNEs and human autopsy samples. We hypothesized that SARS-CoV-2 and other respiratory viruses attach to cilia via ACE2 to penetrate the PCL and enable the virus to traffic through the mucin layer.

To determine if the mucin network mesh blocks virus infections, we treated HNEs with StcE, a mucin-selective protease, to disrupt the structure of the mucin network mesh. This treatment reduced MUC1 levels in mucin network mesh and increased virus entry at 24 hpi (Figure 2A). Thus, the mucin network mesh blocks SARS-CoV-2 infection of HNE ALI cultures, explaining the delay in virus spread. To test if SARS-CoV-2 specifically binds cilia at the early step, we infected ALI-cultured HNEs with SARS-CoV-2 at an MOI of 0.3 and fixed cells at 6 hpi. Scanning (SEM) and transmission electron microscopy (TEM) confirmed that multiple virions attached to cilia (Figures 2B and 2C), suggesting that cilia are hijacked by SARS-CoV-2 to cross the nasal epithelium barrier. Using confocal immunofluorescence (IF) microscopy and TEM, we found that viral particles were attached to cilia near the mucus layer or to cilia and immersed in the PCL in ∼6% of the ciliated HNEs. An anti-SP monoclonal antibody that neutralizes SARS-CoV-2 inhibited attachment of SARS-CoV-2 to cilia (Figures 2D and S2B) and decreased infected cell numbers (Figure 2).

To examine SP binding to the ACE2 receptor during early infection in ciliated HNEs, we conjugated the recombinant SP binding domain (RBD) to fluorescent quantum dots to form QD₅₈₅-RBD pseudovirions. Quantum dots directly visualize RBD binding to ciliary ACE2. In time-lapse, live-cell microscopy, pseudovirion QD₅₈₅-RBD bound to motile beating cilia in ALI-cultured HNEs at 4 hpi. QD₅₈₅ alone did not (Figure S2C). QD₅₈₅-RBDs precisely tracked the ciliary beating in multiple time-lapse videos. This interaction was inhibited by competition with soluble ACE2 (Figure S2C), indicating QD₅₈₅-RBD binds to cilia via ACE2. Our results support that SARS-CoV-2 binds to a ciliary ACE2 receptor to facilitate cell entry.

Cilia facilitate infection by SARS-CoV-2 and other respiratory viruses

We hypothesized that depleting cilia would impede viral infection, so we blocked ciliary assembly by shRNA knockdown of CEP83, a protein critical for motile cilia formation in all ciliated cells, including ALI-cultured HNEs. Depleting CEP83 prevented cilia formation (Figure S2D) but had no effect on epithelial or goblet cell differentiation or expression of ACE2 and TMPRSS2 receptors (Figures S2E and S2F). Indeed, CEP83 downregulation inhibited SARS-CoV-2 infection of HNEs at 24 and 48 hpi at an MOI of 0.3 (Figure 2F). We observed no apparent toxicity in the cilia-depleted cultures.

We next examined whether the ciliary pathway is used by other respiratory viruses. The RSV and PIV receptors CX3CR1 and α2,3-linked sialic acid localize to motile cilia in human airway tissue sections. CX3CR1 colocalized to cilia in ALI-cultured NHE cells and human autopsy samples (Figure S2G). Depleting cilia with CEP83 shRNA inhibited both infections at 24 and 48 hpi in HNEs, indicating broad exploitation of cilia by respiratory viruses (Figures S2H and S2I).

Inhibiting ciliary trafficking attenuates viral uptake

Retrograde trafficking of ciliary proteins or receptors from the tip of cilia to the bottom is mediated by a ciliary dynein complex. We hypothesized that cilia facilitate virus movement from the tip or body of the cilia basally toward the cell body. We treated HNEs with ciliary dynein inhibitor, Ciliobrevin D, 3 h before SARS-CoV-2 infection and determined the number of infected cells at 24 and 48 hpi. The number of infected cells decreased, indicating that protein trafficking in cilia is important, but not strictly essential, during viral entry (Figure 2G). Next, we examined whether SARS-CoV-2 enters HNEs via endocytosis or cell-surface membrane fusion. We treated HNEs with a TMPRSS2 inhibitor, camostat mesylate, or hydroxychloroquine (HCQ), which increases the pH in endosomes, and thereby stops their acidification and maturation, for 2 h prior to SARS-CoV-2 infection and then determined the number of infected cells. HCQ had no effect on infection, indicating that endosome acidification is not required for entry in HNEs and that HCQ does not inhibit virus entry into airway cells. Treatment with TMPRSS2 inhibitor, linked to ACE2 entry, decreased the number of infected cells at 24 and 48 hpi (Figure 2G).

These data suggest two models. In model 1, SARS-Co-V2 binds to ACE2, and TMPRSS2 activates SP for ciliary membrane fusion at the cilium surface. Then, the RNA genome is released into the cilium and transported from the cilium into the cytosol by ciliary dynein (Figure 2H-i). Alternatively, an ACE2 and SARS-CoV-2 complex is transported from the tip of the cilia to the cell body by the ciliary dynein dependent, anterograde process. Then, the virus fuses with the cell membrane via an endocytic-independent process (Figure 2H-ii). Future tests will resolve the contributions of binding versus transport to viral uptake.

Newly produced SARS-CoV-2 particles co-localize with microvilli

Our data indicate that the percentage of SARS-CoV-2-infected cells increased from 24 to 48 hpi (Figure 1A). In contrast, the virus rapidly spreads without an evident lag phase in Vero cells. To better understand virus infection in nasal epithelium, we examined the egress of virus from the few initially infected cells. In Caco-2 and Vero cells, SARS-CoV-2 infection promotes actin-based filopodia protrusion with virus particles associated with these structures. Induction of virus-containing filopodia might be important for SARS-CoV-2 egress and cell-to-cell spread of progeny virions in HNE. Whether canonical microvilli exist on primary airway epithelial cells had been unclear. We used SEM and IF staining to systematically characterize microvilli-like structures in primary airway epithelial cells. Two classes of protrusions were noted on the apical surface of ciliated HNEs: long, wide motile cilia and stubby, dome-like microvilli (Figure S3A). IF staining of multiple core microvilli proteins (EZRIN, EBP50, SLK, and actin [marked by phalloidin]) revealed these proteins co-localize in microvilli-like structures in ALI-cultured HNEs and human airway tissue (Figures 3A , S3B, and S3C). This supports that these structures are canonical microvilli. Microvilli were not found on MUC5AC-positive goblet cells (Figure S3D).

We next sought to test if microvilli have a role in viral infection. We co-stained acetylated tubulin (ACTUB) and actin (phalloidin) with antibodies to SP or NP in infected HNEs at different times, using an MOI of 0.3 to track viral replication and egress. At 6 hpi, the only viral signal was in the cytoplasm, and the microvilli appeared stubby. From 24 hpi onward, NP and SP co-localized with the microvilli (Figures 3B and S3E). To determine if SARS-CoV-2-attached microvilli protrude from infected cell membranes, we co-stained for CEP164, a protein that anchors centrioles to the plasma membrane at the ciliary base and thus marks the apical surface, with spike protein SP and actin (phalloidin) (Figure S3F). Microvilli with attached virions were seen protruding from the membrane as early as 24 hpi. TEM and SEM found abundant viral particles specifically on microvilli rather than motile cilia, indicating that newly generated virus particles accumulate on microvillar structures (Figures 3C and 3D). We also noted that small viral vesicles clustered near or in microvilli and large viral-containing vesicles were in the cytoplasm near the apical side of the cells (Figure S3G). Next, we looked for viral binding to microvilli in infected animals. We infected K18-hACE2 mice with 6 × 10⁴ pfu SARS-CoV-2 (D614G variant) and collected nasal epithelium for immunohistochemistry at 48 hpi. Consistently, SARS-CoV-2 co-localized with microvilli at 48 hpi in vivo (Figure S3H), suggesting that SARS-CoV-2 interacts with microvilli to spread within the respiratory tract in infected animals.

SARS-CoV-2 regulates microvilli dynamics and promotes viral shedding by highly extended microvilli

Based on our initial microscopy, we hypothesized that SARS-CoV-2 infection regulates microvillar structure and function to enable viral egress. Several proteins organize microvilli, including the actin-binding protein ezrin (EZR). Phosphorylated EZR levels were increased in microvilli in infected HNEs (Figure 4A). Another microvillar core protein, EBP50, was increased on microvilli and in the cytoplasm of infected HNEs (Figure 4B). TEM showed that infection induced a dome-like structure protruding from the plasma membrane. Multiple SARS-CoV-2 particle-containing vesicles appeared on this dome-like structure (Figures 4C, left panel, and S4A). Infection also induced highly branched microvillar structures with many attached or budded viral particles at 48 hpi, compared to mock infection. Importantly, SARS-CoV-2 infection dramatically increased highly extended microvilli as early as 24 hpi as observed by IF staining (Figure 4D). These microvilli were accompanied by an accumulation of virus particles along its surface as seen by IF (Figure 4D; 48 and 96 hpi) and TEM (Figures 4E and S4B). These results support our model that SARS-CoV-2 infection modulates the activity, structure, and length of microvilli to facilitate viral egress and spread (Figure 4F).

We used inhibitors of microvillar core protein to determine if loss of microvilli inhibits infection. HNEs were incubated for 3 h with the EZR inhibitor NSC-668394 or the LOK/STK10 kinase inhibitor SB-633825. Treated cells were fixed and stained with phalloidin and anti-phospho-EZR (pEZR). NSC-668394 and the LOK inhibitor strongly decreased microvillar actin staining and levels of pEZR (Figure 4G). To determine if microvilli have a role in viral entry and/or exit, we treated HNEs before or after SARS-CoV-2 infection (Figure 4H) and counted infected cells at 24 and 48 hpi. The inhibitors had no effect within the first 24 hpi, indicating that they do not inhibit entry. In contrast, infected cell numbers decreased at 48 hpi even when inhibitors were administered after infection at 18 hpi (Figure 4I). SARS-CoV-2 particles attached to the cilia rather than microvilli during the viral entry. In contrast, many SARS-CoV-2 viral particles attached to the microvilli later during viral egress (Figure S4C). ACE2 and TMPRSS2 were not expressed in the microvilli (Figure S4D), suggesting that microvilli do not participate in SARS-CoV-2 viral entry. Our data suggest that SARS-CoV-2 hijacks the microvillar assembly pathway to facilitate viral egress by promoting formation of highly extended and branched microvilli.

Mucociliary transport assists in spread of SARS-CoV-2

We next hypothesized that, by associating with elongated microvilli, SARS-CoV-2 traffics to the surface mucus layer. In this model, progeny virions use mucus flow to traverse the airway tract to infect other airway cells. To test this, we used ALI cultures from patients with primary ciliary dyskinesia (PCD). PCD patients have little MCT. This rare autosomal recessive disorder results from functional defects that abrogate synchronous ciliary beating and clearance of pathoge. Nonetheless, airway cilia numbers and length are normal, but without MCT, clearance of many pathogens is compromised, and susceptibility to bacterial or viral pathogens is enhanced. Other ciliary functions, including trafficking of the intraflagellar transport complexes (IFT), the microtubule-motor driven transport and signaling system in cilia, are retained.

To determine if SARS-CoV-2 spreads by MCT, we cultured primary human nasal epithelial cells from three PCD patients and tracked infection (Table S2). Nasal epithelial cells from PCD patients have quantitatively normal cilia density and staining of IFT88 and ARL13B, two proteins essential for normal ciliary protein trafficking (Figure S5A). The epithelial cells had visible microvilli on their apical surfaces (Figure S5B). The numbers of goblet cells in PCD and healthy samples were similar (Figure S5C), as were levels of ACE2 and TMPRSS2 in nasal epithelial cells (Figure S5D). Percentages of SARS-CoV-2-infected cells at 24 hpi were similar in patients and healthy donors, suggesting that mutations in the dynein axonemal heavy chain do not affect initial infection rates (Figure 5A). However, at 48 hpi, PCD patients had fewer SARS-CoV-2-infected cells than healthy donors, suggesting MCT is important later in infection (Figure 5A). At low MOI (48 hpi), healthy samples showed long streaks of SP-positive epithelial cells that apparently traced viral infections along the flows in the mucus layer. In PCD nasal epithelial cells, SARS-CoV-2 only infected immediately surrounding cells, forming small local plaques (Figure 5B). Gentle pipetting to induce mechanical mucus flow increased the number of infected cells and infection area in PCD HNEs (Figure 5C). These results provide a mechanism for why individuals with PCD have no increased risk of infection or severe COVID-19 despite defective mucociliary clearance (Figure 5D).

We next sought to understand whether cell-to-cell contact is important for SARS-CoV-2 transfer. TEM and IF staining showed that cell boundaries and cell-cell junctions were intact, and no obvious viral particles were present in the intracellular space in SARS-CoV-2-infected nasal epithelium (Figures S5E and S5F). The respiratory epithelium comprises multiple layers of cells. However, until 48 hpi, only the uppermost cells faced to the lumen were infected by the virus, with very few cells in lower layers infected (Figures 1A and S1A). Thus, cell-to-cell contact may not be the only pathway for viral spread in nasal epithelium. Although the virus might spread through cell-to-cell contact in nasal epithelium, we believe spreading depends on mucus flow at the apical surface.

PAK1/4 regulate microvilli to facilitate SARS-CoV-2 budding and spreading

Given that regulatory kinases control cytoskeletal dynamics, including microvilli, we asked how infection activates host signaling pathways to modulate microvilli biogenesis, thereby facilitating viral egress. We used global phosphoproteomics to find protein kinases affected by SARS-CoV-2 infection. We infected ALI-cultured HNEs with SARS-CoV-2 at an MOI of 0.3 for 6, 24, 36, and 48 hpi in parallel with mock controls. Cells were harvested, and extracts were prepared for phosphoproteomic mass spectrometry and signaling analysis. We used a substrate-based kinase activity prediction model to determine the activity of kinases from these large-scale phosphoproteomic data. The model assumes that activity can be inferred through by measuring downstream phosphorylation events. Using kinase set enrichment analysis (KSEA), we assigned an enrichment score (ES) (weighted Kolmogorov-Smirnov statistic) to each kinase to reflect its activity in a manner analogous to that of gene set enrichment analysis (GSEA).These scores reflect a ranked statistical calculation of the most likely strong changes in phosphorylation, not a simple fold-change difference. Differential expression of phosphosites in infected HNEs at 6, 24, 36, and 48 hpi was calculated, compared to controls, and KSEA was performed based upon a kinase-substrate database created using PhosphoSitePlus and NetworKin.

Based on the phosphoproteomics and KSEA, we identified 72 kinases that are potentially highly activated after SARS-CoV-2 infection (Figure S6A). These activated kinases may serve a variety of functions important for the virus (endocytosis, viral replication, viral assembly, and others) at various times in the viral life cycle. To understand how SARS-CoV-2 modulates signaling pathways that facilitate the viral egress from microvilli, we focused on kinases related to microvilli or cytoskeleton regulation. We focused on five kinases, including cytoskeleton reorganizing p21-activated kinases 1 and 4 (PAK1 and PAK4), AKT serine/threonine kinase (AKT1/2), mitogen-activated protein kinase P38 Alpha (p38), mitogen-activated protein kinase 3 (ERK1), and Rho-associated coiled-coil containing protein kinase 1 (ROCK1).

We determined if these kinases are phosphorylated in infected cells. All five were activated in SP-positive HNEs (Figures 6A and S6B). Further, immunoblotting of infected HNE cell lysates confirmed these observations (Figure 6B). These results indicate that SARS-CoV-2 infection activates these five kinases in HNEs.

To understand whether these five kinases regulate microvilli structure, HNEs were pre-treated with kinase inhibitors and infected. Infected ciliated HNEs and cytotoxicity were quantified at 48 hpi (Figures S6C and S6F). PAK1, PAK4, CAMK4, and MAPKAPK2 inhibitors decreased infected cells. To determine which kinases regulate microvilli, we treated HNEs with kinase inhibitors for 3 h. Cells were fixed and stained for pEZR and phalloidin (Figures 6C and S6D). Only FRAX486, a PAK1-selective inhibitor, and LCH-7749944, a PAK4-selective inhibitor, significantly decreased expression of phalloidin and pEZR staining in the microvilli (Figure 6C) and changed the microvilli structure (Figure S6E), indicating a disruption in microvilli. Treatment with these drugs caused no significant cytotoxicity (as measured by adenylate kinase activity; Figure S6F). Interestingly, PAK1 and PAK4 regulate the actin cytoskeleton and filopodia.

We treated infected HNEs with specific PAK1 and PAK4 kinase inhibitors. Neither affected early viral cell entry. However, each reduced the number of SARS-CoV-2-positive cells during viral egress at 48 hpi (Figure 6D). PAK4 inhibition affected virus replication the most. We used siRNA to knock down EZR and PAK4 in HNEs and obtained similar results (Figures 6E–6G). Conversely, PAK4 kinases inhibitor and PAK4 siRNA showed little or no viral inhibition on ACE2-expressing A549 cells (Figures S6G and S6H). Thus, PAK1/4 signaling seems to regulate microvilli dynamics during SARS-CoV-2 egress in nasal epithelium.

We next investigated downstream targets of PAK1 and PAK4 to identify potential proteins that were phosphorylated at 36–48 hpi. Several phosphorylation sites were found among phosphoproteomic data hits. PAK1 kinase candidates and associated phosphorylation sites included ARHGEF2 (S151 and S174), BAIAP2 (S325), and FLNA (S2152). PAK4 kinase candidates included PAK4 (S181 and T207), MYH14 (T1503), and PTPN14 (S578) (Figure S6I). ARHGEF2, BAIAP2, MYH14, FLNA, PAK4, and PPP1R12A regulate actin cytoskeleton and filopodia.

To examine their localization, we co-stained the proteins with a microvillar marker in HNEs. BAIAP2 and PTPN14 localized to the microvilli, and ARHGEF2 and MYH14 localized to the base of the microvilli (Figure S6J). Thus, PAK1 and PAK4 may regulate microvilli by phosphorylating downstream targets during infection. Proteins involved in cytoskeleton and filopodia formation were phosphorylated by unidentified kinases after infection (Figure S6K). These included PPP1R12A, CDK16, and ANKRD35 localized to the base of microvilli, suggesting pathways in SARS-CoV-2-regulated microvilli elongation and viral egress (Figure S6L). The kinase activities induced by SARS-CoV-2 suggest that SARS-CoV-2 stimulates microvilli-mediated viral egress and spread via PAK1 and PAK4 signaling, suggesting potential therapeutic targets (Figure 6H). Finally, to confirm the effects of microvilli inhibitors in reducing virus titer in vivo, we applied SLK and PAK4 kinases inhibitors to K18-hACE2 transgenic mice by nasal spray. The PAK4 kinase inhibitor partially repressed infection (Figure S7A), but the SLK kinase inhibitor was less effective. One possible explanation is that human and mouse nasal cavities differ dramatically. In the mouse, ∼50% of the nasal cavity is olfactory epithelium (OE), and the rest is largely ciliated (respiratory) epithelium (RE). In humans, OE constitutes only ∼3% of the cells in the nasal cavity, and the rest are primarily RE. In the human, ciliated cells in the RE are the main target cell type for SARS-CoV-2 in the nasal cavity.

²

K18-Ace2 mice express the ACE2 receptor ubiquitously, which can contribute to a different profile of infected cells as compared to humans.

Omicron variants are more efficient in infecting nasal epithelia cultures

In 2022, Omicron BA.1 (B.1.1.529) rapidly replaced Delta (B.1.617.2) as the dominant circulating form.

Its SP protein has 37 amino acid substitutions, 15 in the receptor binding domain, suggesting that Omicron increased infectivity and transmission by altering virus-receptor binding. More recently, BA.4 and BA.5 displaced BA.1 and BA.2. Thus, Omicron variants may have a replication advantage in the upper respiratory tract from which the virus is transmitted from person to person. Intriguingly, the Omicron virus load from nasal and saliva samples is lower than previous variants.To characterize Omicron replication in the nasal epithelium, we determined the spatial and temporal infection pattern of D614G, Delta, and Omicron variants in HNE ALI cultures. D614G and Delta had similar characteristics (1%–3% positive cells at 6 and 24 hpi and 70%–80% at 48 hpi). In contrast, Omicron variants showed a notable increase of infected cells at early timepoints (∼10% at 6 hpi and ∼40% at 24 hpi) (Figures 7A and S7B). Interestingly, the viral titer of infectious D614G in Vero cells was higher than Delta and Omicron, implying that the infection mechanism varies in 2D and 3D cultures (Figure S7C). Omicron SP may bind more strongly to the ACE2 receptor.

We determined if Omicron variants are more likely to bind to motile cilia. By using time-lapse, live-cell microscopy, we determined that more pseudovirion QD585-Omicron RBDs (BA.1) attached to motile cilia than did D614G RBD in ALI-cultured HNEs at 4 hpi (Figure S7D). Thus, Omicron variants enter and replicate more efficiently in nasal epithelium than D614G and Delta, possibly due to an enhanced interaction with its receptor. To confirm that the entry of Omicron is higher than D614G virus, we used a competition assay with D614G and a chimeric virus harboring Omicron BA.1 SP in the WA1 backbone. Virus was engineered to express RFP and GFP proteins. IF showed that WA1-BA.1-spike-GFP is overwhelmingly superior to D614G-RFP for infecting cells (Figure S7E). Determinants within spike protein seem to be responsible for the increased replication fitness of the Omicron variants. The ability of Omicron to spread within the nasal epithelium without a lag phase suggests a mechanism for how Omicron variants displaced previous variants and became responsible for most global COVID infections.

Finally, we determined if SARS-CoV-2 variants used motile cilia and microvilli for entry and egress. Depletion of cilia with lentivirus expressing CEP83 shRNA inhibited Delta and Omicron infections at 48 hpi in HNEs (Figure 7B). SLK and PAK4 inhibitors also blocked Delta and Omicron viral egress at 48 hpi in HNEs (Figure 7C). Thus, inhibiting viral entry in airway cells is effective even for highly infectious variants.

Discussion

Here, we discovered that SARS-CoV-2 preferentially infects ciliated HNEs and not goblet or basal cells in ALI-cultured human nasal epithelium. We used an ALI airway organoid model with many nuances of the airway in vivo. Strikingly, this culture has a substantial kinetic delay (24–48 h), compared to tissue culture models, which we established as linked to airway barrier function. With EM and IF microscopy, we tracked the detailed steps required for viral entry in the airway. The virus attaches first to airway multicilia via the ACE2 receptor, and that binding and ciliary trafficking are critical for the virus to traverse the mucus-mucin protective barrier. Depleting cilia blocked SARS-Co-2 infection in ALI cultures, but depleting mucins with purified mucinase accelerated viral entry. At 24–48 h, viral numbers increased in parallel with a substantial rearrangement and expansion of microvilli in epithelial cells and activation of PAK1, PAK4, and SLK kinases. Inhibitors of these kinases blocked late viral spread, but not initial binding to cilia. PAK kinase inhibitors attenuated viral spread in mice. The importance of these distinct steps in virus entry is underscored by our finding that recent highly infectious variants of SARS-CoV-2 accelerate infection into the ALI model airway by more than 24 h. Other respiratory viruses require the airway multicilia, suggesting that targeting virus interactions with cilia or microvilli is a broad strategy for blocking airway entry and virus spread for new respiratory viruses.

Only a low percentage of SARS-CoV-2 infects ciliated HNEs at 24 hpi, even at high titers. We suspect the virus cannot penetrate the PCL of most cells and requires a previously unidentified gateway. The mucus and PCL in airway epithelium block entry of pathogens and particles over 40 nm.

Depleting MUC1 partially increased influenza infection, indicating the importance of the PCL as a physical barrier against viruses.

Using multiple imaging techniques, we showed that SARS-CoV-2 attaches with high efficiency to the distal ends of cilia during early infection (<6 h). Depleting cilia in nasal epithelial cells inhibited infection without affecting ACE2 and TMPRSS2 levels. We propose that cilia facilitate early viral attachment and passage through the PCL layer. Accordingly, the cilia could serve as a high-density binding lattice mediated by ACE2, forming an avidity trap to allow viral particles to find target cells via a multivalent binding “landing pad,” leading to super selectivity. Two non-exclusive mechanisms of entry are possible. First, SARS-CoV-2 interacts with ACE2 and TMPRSS2 at the cilia surface to trigger membrane fusion into cilia and release of the viral RNA genome, which could be transported from the cilium to the cytoplasm by ciliary dynein. Alternatively, cilia-attached SARS-CoV-2 could be transported along the surface of cilia by cycles of release and rebinding and bind to ACE2 on the epithelial cell body surface, directing the virus to undergo TMPRSS2-mediated membrane fusion. In this case, the virus-ACE2 complex could be transported by the dynein-dependent retrograde intraflagellar transport down the cilium to the epithelial cell apical surface.

SARS-CoV-2 egress is even less understood. We described the structure of microvilli in ciliated HNEs. SARS-CoV-2 is localized to and exits via the microvilli at 24 hpi, and SARS-CoV-2 infection promotes formation of dome-like structures at the base of microvilli and highly branched and highly extended microvilli over time. Although the mechanisms and consequences of these rearrangements are unknown, SARS-CoV-2 induced the highly extended microvilli to penetrate the PCL layer, and concatenated chains of virus formed on the extended microvillar structure. These viral chains appear to accumulate in the mucus layer and to use MCT to infect other cells. This hypothesis agrees with our findings in cells from PCD patients. PCD patients seem to have no increased risk of infection.

Disrupting microvilli with two distinct inhibitors had little to no effect on viral entry but severely inhibited viral exit. Thus, cilia may facilitate infection early on, and microvilli may mediate later stages of infection. Our study provides potential valuable targets for drug repurposing against COVID-19. A nasally delivered drug that transiently inhibits microvillar reprogramming in the upper respiratory tract may have the advantage of non-invasive delivery with less systemic toxicity or adverse effects.

Finally, using unbiased phosphoproteomic analysis, we found that SARS-CoV-2 regulates the cytoskeleton and microvilli by activating PAK1 and PAK4 to facilitate viral egress and spread. PAK1 and PAK4 kinase inhibitors inhibited SARS-CoV-2 cell egress and spread but did not affect entry. We identified several undescribed microvilli- or cytoskeleton-associated proteins as downstream targets of PAK1 and PAK4. They regulate filopodia or cytoskeleton and represent critical markers of viral reprogramming and possibly therapeutic target. Lastly, we identified potential microvilli- or filopodia-associated proteins that are phosphorylated during SARS-CoV-2 infection. Their roles remain to be explored.

Omicron is the dominant variant in the world circa 2022, but the mechanisms for its higher transmissibility are unclear. Our results indicate that, at similar MOIs (as measured in Vero cells), Omicron infected 40% of cells at 24 hpi, compared to 3%–5% for D614G and Delta. The number of cells infected by Omicron also increased at 6 hpi. The more efficient binding of the Omicron variant RBD to ACE2 and motile cilia provides a parsimonious explanation for the increased initial numbers of infected cells.

While cilia facilitate SARS-CoV-2 infection early on, microvilli mediate the later stages of viral infection, allowing particles to spread to other cells. We found several potential targets for antiviral drug therapies with new inhibitors or repurposing of existing inhibitors. Topical administration of kinase inhibitors to the nasal mucosa may modulate microvilli function and avoid undesirable systemic off-target effects for systemic treatment.

Motile cilia-regulated mucociliary clearance is important for the functioning of the human respiratory tract. Therefore, research has focused on how motile cilia beat, but little is known about how motile cilia directly function as a sensory organ and how signaling and airway remodeling are regulated. Specifically, there is very little understanding on whether ciliary proteins and receptors are specifically transported into and out of the motile cilia to regulate and maintain sensory function, and whether motile cilia are regulated by IFT and Bardet–Biedl syndrome complexes that regulate protein trafficking in primary cilia,

although clearly these pathways have an effect on airway function and morphology.

Limitations of the study

The sample sizes of HNEs were limited by tissue availability and the challenges of conducting experiments under BSL3. Samples from PCD patients were especially limited, as the estimated prevalence of PCD is 1:10,000 to 1:20,000 live-born children. Although patients with PCD share similar defective mucociliary clearance, mutations in >40 genes cause PCD with varying degrees of severity. Therefore, our findings are not necessarily representative of all PCD patients.