Long-Term Modeling of SARS-CoV-2 Infection of In Vitro Cultured Polarized Human Airway Epithelium

The pandemic of coronavirus disease 2019 (COVID-19), which is caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), has led to >35 million confirmed cases and >1 million fatalities worldwide. SARS-CoV-2 mainly replicates in human airway epithelia in COVID-19 patients. In this study, we used in vitro cultures of polarized human bronchial airway epithelium to model SARS-CoV-2 replication for a period of 21 to 51 days. We discovered that in vitro airway epithelial cultures endure a long-lasting SARS-CoV-2 propagation with recurrent peaks of progeny virus release at an interval of approximately 7 to 10 days. Our study also revealed that SARS-CoV-2 infection causes airway epithelia damage with disruption of tight junction function and loss of cilia. Importantly, SARS-CoV-2 exhibits a polarity of infection in airway epithelium only from the apical membrane; it infects ciliated and goblet cells but not basal and club cells. Furthermore, the productive infection of SARS-CoV-2 requires a high viral load of over 2.5 × 105 virions per cm2 of epithelium. Our study highlights that the proliferation of airway basal cells and regeneration of airway epithelium may contribute to the recurrent infections.

stranded RNA virus, and belongs to the genus Betacoronavirus of the family Coronaviridae (1)(2)(3)6). The clinical syndrome of COVID-19 is characterized by various degrees of severity, ranging from a mild upper respiratory illness (7) to severe interstitial pneumonia and acute respiratory distress syndrome (ARDS), a life-threatening lung injury that allows fluid to leak into the lung (5,(8)(9)(10). Compared to the approximately 34% fatality rate of the Middle East respiratory syndrome (MERS) and ϳ10% fatality rate of the severe acute respiratory syndrome (SARS) (11), COVID-19 has a lower fatality rate, ranging from 0.7% to 5.7% in the United States (12); however, it spreads more efficiently than SARS and MERS (13,14), making it difficult to contain. COVID-19 has become a pandemic (15), which has led to over 35.6 million confirmed cases and Ͼ1 million fatalities worldwide as of 5 October 2020.
While SARS-CoV-2 viral RNA can be detected in nasal swabs, nasopharyngeal aspirates, and bronchoalveolar lavage fluids throughout the airways (1-3, 16, 17), how the virus infects epithelial cells at different levels of the respiratory tree and the underlying pathogenesis remain unclear. Thus, a comprehensive understanding of how SARS-CoV-2 replicates and causes pathogenesis in its native host, the epithelial cells lining different levels of the airways, is essential to devising therapeutic and prevention strategies to counteract COVID-19. Primary human nasal, trachea, and bronchial epithelial cells can be cultured and differentiated at an air-liquid interface (ALI), forming a pseudostratified mucociliary airway epithelium that is composed of ciliated cells, goblet cells, club cells, and basal cells with an arrangement closely reflective of an in vivo cellular organization (18,19). This in vitro model of human airway epithelium (HAE) cultured at an ALI (HAE-ALI) closely recapitulates many important characteristics of respiratory virus-host cell interactions seen in the infected upper and lower airways in vivo and has been used to study many human respiratory viruses (20)(21)(22)(23)(24)(25)(26)(27)(28)(29), including SARS-CoV (30,31). Primary HAE-ALI can be infected by SARS-CoV-2 (32,33), resulting in epithelial damage (34), and they can be used for virus isolation (1,16). Notably, differentiation at an ALI results in a drastic increase in expression and the polar presentation of the viral receptor angiotensin-converting enzyme 2 (ACE2) (2,35) on the apical membrane (30,31). Thus, HAE-ALI is an optimal cell culture model to study SARS-CoV-2 infection in vitro.
In this study, we generated HAE-ALI cultures directly from primary bronchial epithelial cells without propagation prior to differentiation at an ALI. We used these cultures to model SARS-CoV-2 infection for a long period of 21 to 51 days, focusing on the viral replication kinetics, the dose dependency of viral infection, epithelial damage, and the permissive subpopulation of the epithelial cell types. Additionally, we showed that SARS-CoV-2 favors apical infection of HAE-ALI, confirming the polar infection of SARS-CoV-2 in human airway epithelia. While SARS-CoV-2 efficiently infected HAE-ALI through the apical side at a viral load as low as a multiplicity of infection (MOI) of 0.002 plaque-forming units (PFU) per cell, viral replication at an MOI lower than this threshold was not detected. Notably, SARS-CoV-2 infection presented as an enduring infection in HAE-ALI with recurrent peaks of virus released from the infected ciliated and goblet cells, while the airway basal cells and club cells were nonpermissive.

RESULTS
SARS-CoV-2 infection of human airway epithelia presents a long-lasting infection and causes epithelial damage. SARS-CoV-2 primarily infects human airway epithelial cells of the respiratory tracts and lungs of COVID-19 patients (36,37). SARS-CoV-2 infections in the in vitro model of well-differentiated HAE-ALI and organoids have been reported (34,36,38,39). However, these studies were focused on short-term virus replication and cytopathic effects as they were carried out in a time frame of less than 1 week postinfection. Since airway epithelia capably repair, regenerate, and remodel themselves (40), we hypothesized that a long-term monitoring of SARS-CoV-2 infection in HAE-ALI might reveal unknown important features that were missed in prior studies.
To this end, we first chose two HAE-ALI cultures, B4-20 and B9-20 (HAE-ALI B4-20 and HAE-ALI B9-20 ), which were generated from primary bronchial epithelial cells freshly isolated from two donors. The initial study was performed with the infection of SARS-CoV-2 at an MOI of 2 PFU/cell. We collected the apical washes on a daily basis for continued monitoring of virus replication through titration for infectious virions with a plaque assay in Vero-E6 cells. We also periodically performed confocal microscopy analyses of the infected HAE with immunofluorescence assays. As expected, we observed rapid virus release from the infected HAE-ALI cultures, which reached a peak of 9 ϫ 10 5 PFU/ml at 2 days postinfection (dpi). The apical virus release from HAE-ALI B4-20 remained at the peak for 3 days and then decreased to a level less than 800 PFU/ml at 7 dpi (Fig. 1, B4 -20). The infection of HAE-ALI B9-20 presented a similar trend, with the peak at 7.5 ϫ 10 5 PFU/ml from 2 to 6 dpi, which dropped to 3 ϫ 10 3 PFU/ml at 9 dpi ( Fig. 1, B9 -20). However, at later time points, continued study revealed virus release kinetics with at least two peaks during the course of 3 weeks from both infections. In the infection of HAE-ALI B4-20 , virus release started to increase again from 8 dpi and reached a peak of 7.6 ϫ 10 5 PFU/ml at 12 dpi. The apical release of virus then dropped to 3 ϫ 10 4 PFU/ml at 14 dpi, followed by another peak of 7 ϫ 10 5 PFU/ml at 17 dpi ( Fig. 1, B4 -20). Notably, while the decrease of the virus released from the first peak of infected HAE-ALI B9-20 lagged behind that of the infected HAE-ALI B4-20 by 2 days, it demonstrated a second peak at 11 to 13 dpi ( Fig. 1, B9 -20) with viral shedding at 7.6 ϫ 10 5 PFU/ml. We reasoned that this was due to the donor variation, which affects the extent of differentiation and the subpopulation ratio of epithelial cell types, but not the properties permissive to SARS-CoV-2 infection.
We next extended the study to another HAE-ALI culture derived from a different donor, the HAE-ALI B3-20 . Infection was conducted with a 10-fold-reduced virus inoculum (MOI of 0.2 PFU/cell). We observed similar replication kinetics with two virus release peaks (see Fig. S1A in the supplemental material). Of note, even though a reduced MOI was applied to HAE-ALI B3-20 , we recorded a higher viral shedding in the first peak (4 ϫ 10 6 PFU/ml) than that from HAE-ALI B4-20 and HAE-ALI B9-20 , while those in the second peaks of the infections in the three cultures were approximately at a similar level (5 ϫ 10 5 PFU/ml). SARS-CoV-2 infection of HAE-ALI B3-20 significantly reduced the transepithelial electrical resistance (TEER) value, which is a hallmark of epithelial integrity, starting at 1 dpi (Fig. S1B) and resulted in dispersed zonula occludens-1 (ZO-1) expression and reduced ␤-tubulin IV staining that suggested the loss of cilia ( Fig. S1C and D), which will be further discussed below. SARS-CoV2-infection in HAE-ALI was visualized by immunostaining for the expression of viral nucleocapsid protein (NP) in the infected cells. The analyses revealed the relative increases of NP-positive (NPϩ) cells, aligned roughly with the apical virus release kinetics ( Fig. 2 and Fig. S2, NP). Of note, infected HAE-ALI showed poor staining of ZO-1, which started at 1 dpi and remained throughout the infection, indicating rapid epithelial damage caused by the infection as the tight junctions of the epithelia were destroyed ( Fig. 2A, Fig. S1C, and Fig. S2A, ZO-1). The infected HAE-ALI also showed a partial loss of cilia, indicated by immunostaining with anti-␤-tubulin IV, which also started at 1 dpi and remained at a similar level throughout the course of infection (Fig. 2B, Fig. S1D, and Fig. S2B).
To examine the infected epithelia in greater detail, we performed Z-stacked imaging of the infected HAE-ALI B9-20 at 15 dpi. The images showed a percentage of ϳ10% NPϩ . Of note, most of the NPϩ cells remained ␤-tubulin IV stained, suggesting that ciliated cells represent the major cell type in HAE permissive to SARS-CoV-2. We also noticed that there were fewer cells present in the areas where NP1 staining was positive compared to the mock infection, as determined by the number of the nuclei in the imaged area, indicating cell loss (death) of the infected epithelia (Fig. 3, DAPI, SARS-CoV-2 versus Mock).
Taken all together, these results demonstrated that SARS-CoV-2 infection of HAE-ALI represents a long-lasting process with multiple peaks of virus infections (apical virus release and NP-expressing cells) and that the infection degrades two hallmarks of the airway epithelia, tight junctions and ciliary expression.

SARS-CoV-2 infection of HAE presents multiple peaks and requires a high viral load.
To further examine the recurrent peaks of virus release from the infections and the barrier dysfunction of SARS-CoV-2-infected HAE, we performed a longer monitoring period of 31 days for the infection of HAE-ALI B4-20 at an MOI from 0.2 to 2 ϫ 10 Ϫ5 (Fig. 4 and 5). We also infected HAE-ALI L209 , which was polarized on the large MilliCell insert  Viral shedding to the culture medium in the basolateral chamber across the supportive membrane was also continuously monitored in the experiments. The infectious virions in the medium were detected in the early time points after the infection was initiated, but at a level of 2 to 3 orders of magnitude less than that in the apical washes ( Fig. 4A, Fig. S1A, and Fig. S3A), suggesting that the progeny of the SARS-CoV-2 are predominately released from the apical membrane of the infected HAE. We concluded that the trace of detected viruses in the basolateral medium must come from the leakage of the apically secreted viruses across the supportive semipermeable membrane due to epithelial damage caused from SARS-CoV-2 infection. Interestingly, no infectious virions were found in the basal medium when the peaks of the viral progeny in the apical washes reappeared at the late time points, even though virus burdens were at similar levels at these peaks ( Fig. 4A and C and Fig. S1A). These observations suggest the regeneration of the destructive mucosal lesions occurs during the SARS-CoV-2 infection and that such repair is sufficient to prevent the viral shedding to the basolateral chamber, although the repair did not lead to the recovery of the TEER of the infected HAE ( Fig. 4B and D). A different observation came from the infection of HAE-ALI L209 , in which the epithelial cells from donor L209 were cultured and polarized in the large Millicell inserts. Traces of viral shedding in the basal medium were also found at the late time points, and their levels changed responding to each peak of the virus replication (Fig. S3A). We reasoned that this is due to the inefficient epithelium repair in the infected HAE-ALI L209 , which was affected by either donor variation or the different culture formats. We speculate that the repair of the airway epithelium through basal cell differentiation stops most potential leakage of the virus but is not sufficient to recover full airway function.
The recurrent peaks of virus progeny released during the course of infection were further displayed in infections of HAE-ALI B4-20 at lower MOIs of 0.02 and 0.002 (Fig. 4C), which show three almost identical peaks at 3, 13, and 26 dpi over the course of 30 days. Of note, the third peak became obvious at 26 dpi from these low-MOI infections compared to that at 30 dpi from the higher MOI of 0.2, but the amounts of infectious virions released from those peaks remained at roughly the same level (ϳ1 ϫ 10 6 PFU/ml). The epithelial damage was indicated by the decrease of TEER beginning at 2 and 3 dpi, respectively (Fig. 4D), as well as revealed by the dispersed ZO-1 expression and loss of cilia (Fig. 5A, MOIs ϭ 0.02 and 0.002).
We then carried out the infection at the much lower MOIs of 2 ϫ 10 Ϫ4 and 2 ϫ 10 Ϫ5 over a course of 3 weeks. Surprisingly, we found that HAE-ALI B4-20 cultures were not productively infected by SARS-CoV-2, as evidenced by no NPϩ cells at 30 dpi (Fig. 5, MOI ϭ 2 ϫ 10 Ϫ4 and 2 ϫ 10 Ϫ5 ). No infectious virions released from the apical side in these low-MOI conditions were detectable. To verify this result, we performed infections in HAE-ALI B9-20 at MOIs of 2 ϫ 10 Ϫ4 and 2 ϫ 10 Ϫ5 . The results reproduced the same observations of no productive infection in the cultures derived from a different lung donor (Fig. S5). This is in contrast to the SARS-CoV-2 infection in Vero-E6 cells. At MOIs of 2 ϫ 10 Ϫ4 and 2 ϫ 10 Ϫ5 , we did not observe an obvious loss of cilia in both infected HAE-ALI B4-20 and HAE-ALI B9-20 ( Fig. 5B and Fig. S5B); however, we observed a cytoplasmic expression and a weak junction expression of ZO-1 at 30 dpi (Fig. 5A) and 21 dpi (Fig. S5A) for infected HAE-ALI B4-20 and HAE-ALI B9-20 , respectively. These results demonstrate that a high viral load (at least Ͼ100 PFU [ϳ8.2 ϫ 10 4 viral genome copies {vgc}]) to an epithelium of 0.33 cm 2 , which contains ϳ5 ϫ 10 5 epithelial cells, is necessary to initiate a productive infection.
Ciliated and goblet cells are permissive to SARS-CoV-2 but not the basal and club cells. We next examined SARS-CoV-2 infection in which a high MOI of 2 was applied to the basolateral side of HAE-ALI B4-20 . The results showed there were no detectable infectious virions released from both the apical and basolateral sides (Fig. 6A). There were no signs of epithelial impairment observed as well. The TEER of the infected HAE displayed no significant changes over the course of 23 days (Fig. 6B). Immunofluorescence analyses revealed well-preserved tight junctions and the rich cilium expression (Fig. 6C and D). Importantly, NPϩ cells were not detected for as long as 23 dpi. Similar results were verified in infection of HAE-ALI B9-20 over an infection course of 3 weeks. These results demonstrated that SARS-CoV-2 does not infect epithelial cells from the basolateral side.
To determine the permissive epithelial cell types in the infected HAE-ALI, we carefully examined the infected cells by immunofluorescence assays using various epithelial cell markers. Cells were dissociated from the supportive membranes of the infected Transwell inserts and cytospun onto slides for imaging. Costaining of a specific cell marker and the viral NP expression visualized the cell types permissive for SARS-CoV-2 infection. The results, as the representative images shown in Fig. 7, demonstrated that the majority of cell populations in the HAE-ALI were basal cells, which expressed cytokeratin 5 (CKRT5ϩ) (41), and ciliated cells with positive anti-␤-tubulin IV staining. Consistent with previous imaging results (Fig. 3), most of the NPϩ cells were also positive with anti-␤-tubulin IV staining (Fig. 7A), whereas almost all the CKRT5ϩ basal cells were negative for anti-NP staining (Fig. 7C). Probing secretoglobin family 1A member 1 (SCGB1A1) expression for club cells and mucin 5AC (MUC5AC) expression for goblet cells (41) revealed that the secretory cells were less abundant subpopulations in the infected HAE-ALI cultures. While we could not locate any club cells stained positively for NP expression (Fig. 7D), we found some NPϩ/MUC5ACϩ goblet cells (Fig. 7B). Importantly, we observed that in SARS-CoV-2-infected HAE-ALI, a subset of CKRT5ϩ basal cells are found associated with the expression of Ki67, but not in mock-infected HAE-ALI (Fig. 8B). As Ki67 is a marker of cell proliferation (42), this result suggested that SARS-CoV-2 infection pushes basal cells toward proliferation.
Taking these lines of evidence together, our results confirm that SARS-CoV-2 mainly infects ciliated cells of HAE, as well as goblet cells, despite the lower abundance of goblet cells in HAE-ALI cultures. Our study suggests that basal and club cells are not permissive to SARS-CoV-2.

DISCUSSION
In this study, we modeled SARS-CoV-2 infection in HAE-ALI cultures generated from primary bronchial epithelial cells directly isolated from 4 lungs of independent donors. The infections were conducted at various MOIs, from both apical and basolateral sides, and for a long period from 21 to 51 days. Our studies demonstrated that the SARS-CoV-2 infection of HAE results in multiple replication peaks of virus progeny at MOIs from 2 to as low as 0.002, although the length of peak emergence time varied from ϳ7 to 10 days. The most striking result we obtained is the resistance of HAE to SARS-CoV-2 infection at an MOI of 2 ϫ 10 Ϫ4 PFU/cell (ϳ300 PFU/cm 2 of epithelium), which is in contrast to the infection in Vero-E6 cells at the same MOI or lower. Our studies also revealed that the basal (CKRT5ϩ) cells and club (SCGB1A1ϩ) cells are not permissive, whereas ciliated cells (␤-tubulin IVϩ) and goblet (MUC5ACϩ) cells form the primary body of permissive cells in the SARS-CoV-2-infected HAE.

SARS-CoV-2 Infection in Human Airway Epithelium
® that ciliated cells were the predominately infected cell type (38). With the single epithelial cell suspension recovered from the infected inserts, we prepared cytospin slides to investigate in detail the cell types permissive to the infection. It is curious that secretory goblet and club cells behave in opposite ways to SARS-CoV-2 infection, since  We observed that SARS-CoV-2 was unable to infect epithelial cells from the basolateral side, where CKRT5ϩ basal cells reside. The airway basal cells are the epithelial cell type not presenting on the surface of the airway lumen; thus, they are not accessible to the virus on the apical side. However, when the infection commences and the epithelial damage occurs, the destructive mucosal lesions (and the death of the infected ciliated and goblet cells) would allow the virus to gain access to the basal cells (Fig. 8C). Indeed, the detectable virus shedding to the basolateral chamber indicates a possible window to expose the basal cells to SARS-CoV-2. Notably, these time points also represent the peaks of release of virus progeny. However, none of the CKRT5ϩ and NPϩ cells were found in SARS-COV-2-infected HAE. The nonpermissive nature of basal cells to SARS-CoV-2 is likely due to the negligible expression of TMPRSS-2 (43), since it expresses ACE2 (Fig. S6C) (46,47). Of note, we observed a subset of basal cells proliferating after SARS-CoV-2 infection, indicating that the basal cells should play an important role in repairing the epithelium lesions caused by viral infection. At the airway epithelial cellular level, the tight-junction-associated proteins, such as ZO-1, occludin, and claudins, play a central part in the epithelial cytoprotection by maintaining a physical selective barrier between external and internal environments (40). The tight junction proteins are highly labile structures whose formation and structure may be very rapidly altered after airway injury, for example, airway inflammation. Proinflammatory cytokines have a drastic effect on tight junction expression and barrier functions, which significantly alter the epithelial barrier permeability (48)(49)(50). SARS-CoV-2 infection distorts the ZO-1 expression, and thereafter causes barrier dysfunction (TEER decrease). The infection not only alters the ZO-1 expression of infected (NP1ϩ cells) but also uninfected cells (NP-negative [NPϪ] cells) (Fig. 3). This is also true for the loss of cilia. We believe that SARS-CoV-2 infection produces inflammatory cytokines as an innate immunity response upon virus infection (51), which either disturbs ZO-1 and tubulin expression or alters their structures. The innate immunity response may also limit virus infection at the front line. In fact, SARS-CoV-2 requires a high viral load (Ͼ300 PFU/cm 2 of HAE) to initiate a productive infection (Fig. 4). Of note, the infectious titer (PFU) was determined by plaque assay in Vero-E6 cells, which are interferon deficient (52). We determined that 1 PFU of SARS-CoV-2 in Vero-E6 cells has a particle (viral genome copy) number of 820, suggesting that a load of 2.46 ϫ 10 5 particles is required to productively infect 1 cm 2 of the airway epithelium, which is much higher than the small DNA parvovirus human bocavirus 1 (HBoV1) we studied (53). HBoV1 can infect HAE at an MOI of as low as 0.001 genome copies per cell, which equals 1.5 ϫ 10 3 particles per 1 cm 2 of the airway epithelium. Apparently, whether and how strong an innate immunity response is induced during SARS-CoV-2 infection of HAE-ALI cultures warrant further investigation.
Airway epithelium repair or regeneration is critical for the maintenance of the barrier function and the limitation of airway hyperreactivity (54). In a biopsy specimen study of fresh tracheas and lungs from five deceased COVID-19 patients, it was found that the epithelium was severely damaged in some parts of the trachea, and extensive basal cell proliferation was observed in the trachea, where ciliated cells were damaged, as well as in the intrapulmonary airways (37). These data support our conclusion that basal cells are not permissive to SARS-CoV-2. As a response to these previous findings, our study observed that a subset of proliferating basal cells in the SARS-CoV-2 infected HAE-ALI, but not in the mock-infected HAE-ALI (Fig. 8B). Thus, we hypothesize that SARS-CoV-2 infection induces basal cell proliferation, which accounts for the observed long-lasting infections with recurrent peaks of viral replication and warrants future investigation. In addition, we speculate that the apical virus release peaks might correlate with the progress of regeneration in the damaged epithelia, despite the finding that the recurrent infection of SARS-CoV-2 in newly differentiated permissive cells prevents the epithelium from being fully repaired, which was indicated by the dispersed ZO-1 expression over the course of infection.
Overall, we propose a model of SARS-CoV-2-infection of HAE (Fig. 8C): SARS-CoV-2 selectively infects ciliated and goblet cells on the surface of the airway lumen (the apical side of HAE). Upon invading these cells, SARS-CoV-2 replicates and produces infectious virions, which eventually leads to cell death and epithelial damage. Upon the destructive lesions, airway epithelium has the capacity to progressively repair and regenerate itself. Thus, basal cells (possibly also including club cells) proliferate and differentiate to ciliated cells or goblet cells to fill up the areas that have lost ciliated or goblet cells. Then, the virus released from the last round of infection infects newly regenerated ciliated or goblet cells (repaired epithelia), followed by the second round of active replication and virus production. Therefore, airway epithelial regeneration confers a persistent, cyclically peaked infection of SARS-CoV-2 in human epithelia. Primary airway epithelium cultured at an air-liquid interface (HAE-ALI) (Fig. 8A). The primary HAE-ALI cultures HAE-ALI B3-20 , HAE-ALI B4-20 , and HAE-ALI B9-20 were provided by the Cells and Tissue Core of the Center for Gene Therapy, University of Iowa (18,(55)(56)(57). These polarized HAE-ALI cultures were derived from three independent donors. The freshly isolated human bronchial epithelial cells from the donor tissues were seeded onto collagen-coated, semipermeable polycarbonate membrane inserts (0.33 cm 2 , 0.4-m pore size, Costar Transwell, catalog no. 3413, Corning, New York), and grown at an ALI as previously described (58). The cultures were maintained in USG medium containing 2% Ultroser G (USG) serum substitute (Pall BioSepra, France). After 3 to 4 weeks of culture at an ALI, the polarized culture was fully differentiated. The polarity of the HAE was determined for the TEER using an epithelial volt-ohm meter (Millipore). A value of 1,000 ⍀·cm 2 or higher was chosen for SARS-CoV-2 infection as we previously used for HBoV1 infection (50,59). HAE-ALI L209 cultures on 1.1 cm 2 Millicell-PCF (Millipore, Billerica, MA) were provided by Dr. Matthias Salathe, which were generated following a published method (60) using primary airway bronchial epithelial cells isolated from the lung of a donor (L209).

Virus infection, sample collection, and titration. (i) Virus infection.
For apical infection, welldifferentiated primary HAE-ALI in Transwell inserts (0.33 cm 2 ; Costar) or in Millicell inserts (1.1 cm 2 ; Millipore) were inoculated with 100 l or 300 l of SARS-CoV-2 at various MOIs, as indicated in each figure legend, applied to the apical chamber. For basolateral infection, HAE-ALI cultures in Transwell inserts were inoculated with SARS-CoV-2 diluted in 500 l of USG medium added to the basolateral chamber. The infected HAE-ALI cultures were incubated at 37°C and 5% CO 2 for 1 h followed by aspiration of the virus from the apical or basolateral chamber and washing of the cells with D-PBS three times (the last wash was saved and used for plaque assay, which was presented as the virus residue right after infection at the day 0 postinfection [0 dpi]). The HAE-ALI cultures were then further cultured at 37°C and 5% CO 2 .
(ii) Viral sample collection. Viral samples were collected from both the apical wash of the epithelium surface and the culture medium in basolateral chamber at multiple time points. In brief, 100 l (or 300 l) of D-PBS was added to the apical chamber for a short incubation of 30 min at 37°C and 5% CO 2 . Thereafter, this apical wash was recovered carefully from the apical chamber without disturbing the culture. To quantitate the viruses released from the basal membrane to the culture medium, 100 l of medium was collected from each basolateral chamber. The infectious titers in the collected samples were determined by plaque assays in Vero-E6 cells.
(iii) Plaque assays. Vero-E6 cells were seeded in 24-well plates at a density of ϳ0.5 ϫ 10 6 cells and grown to confluence the second day. Virus (apical washes or basolateral media) was serially diluted 10-fold in D-PBS. Two hundred microliters of the diluent was added to each well and incubated for 1 h on a rocking rotator. After the virus diluent was removed, ϳ0.5 ml of overlay medium (1% methylcellulose [Sigma, catalog no. M0387] in DMEM with 5% FBS) was added to each well. The plates were incubated at 37°C under 5% CO 2 for 4 days. After the methylcellulose overlays were removed, the cells were fixed using the 10% formaldehyde solution for 30 min and stained with 1% crystal violet solution followed by extensive washing using distilled water. Plaques in each well were manually counted and multiplied by the dilution factor to determine the virus titer at the unit of plaque-forming units per milliliter.
(iv) Reverse transcription and quantitative PCR (RT-qPCR). To eliminate free viral RNA in the samples, 25 units of Benzonase (Sigma) was added to 100 l of the virus samples for 30 min (61). The nuclease-treated samples were used for viral RNA extraction using the Viral RNA extraction kit (Quick-RNA Viral kit) (catalog no. R1035; Zymo Research) following the manufacturer's instructions. Moloney murine leukemia virus (M-MLV) reverse transcriptase (catalog no. M368A; Promega) was used to reverse transcribe viral RNA with the reverse PCR primer according to the manufacturer's instructions. A 2.5-l portion of the cDNA was quantified by TaqMan qPCR in a reaction of 25 l to determine the number of viral genome copies (vgc) using the CDC 2019-nCoV_N1 set of primers and probe, which were synthesized at IDT (Coralville, IA). The plasmid pcDNA6B-(SARS-CoV-2)N, which contains the SARS-CoV-2 NP gene (nucleotides [nt] 998 to 2244), was used as a reference control (1 vgc ϭ 7 ϫ 10 Ϫ12 g) to establish a standard curve for absolute quantification on an Applied Biosystems 7500 Fast system (Foster City, CA).
Immunofluorescence confocal microscopy. (i) Immunofluorescence assay. For analysis of the SARS-CoV-2 infection in the HAE grown on the supportive membranes of the Transwell inserts, we cut off the membranes from the inserts and fixed them with 4% paraformaldehyde in PBS at 4°C overnight. The fixed membrane was washed in PBS for 5 min three times and then split into 4 (for 0.33-cm 2 membrane) or 8 (for 1.01-cm 2 membrane) pieces for whole-mount immunostaining. For cell marker analysis, we dissociated the cells off the supportive membranes of the Transwell inserts by incubation with Accutase (Innovative Cell Technologies, Inc., San Diego, CA). After incubation for 1 h at 37°C, cells were completely detached from the membrane and well separated. Cells were collected and then cytocentrifuged at 1,800 rpm for 3 min onto slides using a Shandon Cytospin 3 cytocentrifuge. After the slides were cytospun, they were fixed overnight in 4% paraformaldehyde at 4°C.
The fixed HAE or dissociated cells were permeabilized with 0.2% Triton X-100 for 15 min at room temperature. Then, the slide was incubated with primary antibody in PBS with 2% FBS for 1 h at 37°C. After the membrane was washing, it was incubated with fluorescein isothiocyanate-and rhodamineconjugated secondary antibodies, followed by staining of the nuclei with DAPI (4=,6-diamidino-2phenylindole).
(ii) Confocal microscopy. The cells were then visualized using a Leica TCS SPE confocal microscope at the Confocal Core Facility of the University of Kansas Medical Center. Images were processed with the Leica Application Suite X software.
Transepithelial electrical resistance (TEER). One hundred microliters of D-PBS was added to the apical chamber to determine the TEER using a Millicell ERS-2 volt-ohm meter (MilliporeSigma, Burlington, MA) following a previously used method (58).
Statistics. Virus release kinetics were determined with the means and standard deviations obtained from at least three independent HAE-ALI B3-20, HAE-ALI B3-20 , and HAE-ALI B9-20 and from duplicated HAE-ALI L209 by using GraphPad Prism version 8.0. Error bars represent means and standard deviations (SD). Statistical significance (P value) was determined by using unpaired (Student) t test for comparison of two groups. grant YAN19XX0 from the Cystic Fibrosis Foundation. The confocal core is supported, in part, by NIH/NIGMS COBRE grant P30GM122731.