SARS-CoV-2 Causes Severe Epithelial Inflammation and Barrier Dysfunction

SARS-CoV-2 challenges health care systems and societies worldwide in unprecedented ways. Although numerous new studies have been conducted, research to better understand the molecular pathogen-host interactions are urgently needed.

IMPORTANCE SARS-CoV-2 challenges health care systems and societies worldwide in unprecedented ways. Although numerous new studies have been conducted, research to better understand the molecular pathogen-host interactions is urgently needed. For this, experimental models have to be developed and adapted. In the present study we used mono-cell culture systems, and we established a complex chip model, where epithelial and endothelial cells are cultured in close proximity. We demonstrate that epithelial cells can be infected with SARS-CoV-2, while the endothelium did not show any infection signs. Since SARS-CoV-2 is able to establish viremia, the link to thromboembolic events in severe COVID-19 courses is evident. However, whether the endothelial layer is damaged by the viral pathogens or whether other endothelial-independent homeostatic factors are induced by the virus is essential for understanding the disease development. Therefore, our study is important, as it demonstrates that the endothelial layer could not be infected by SARS-CoV-2 in our in vitro experiments, but we were able to show the destruction of the epithelial-endothelial barrier in our chip model. From our experiments, we can assume that virus-induced host factors disturbed the epithelial-endothelial barrier function and thereby promote viral spread.
T he novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a highly pathogenic virus causing severe respiratory infections, described as coronavirus disease-19  (1). Patients suffer from various symptoms such as fever, cough, shortness of breath, headache, muscle aches, and gastrointestinal symptoms. Hallmarks of severe COVID-19 courses are pneumonia, pulmonary edema, acute respiratory distress syndrome (ARDS), and multiple organ failure. In most patients, the disease has a mild course, but in some cases, e.g., elderly with comorbidities, the infection can develop into a lifethreatening condition. In particular, preexisting lung pathologies and systemic diseases such as diabetes predispose to severe infection courses described by George et al. (2).
Clinical studies revealed that the virus primarily replicates in the lung, which can cause severe lung damage up to necrotic destruction of large areas of the lung tissue (3). In autopsies of deceased COVID-19 patients, it has been observed that particularly in severe cases, viral particles can disseminate throughout the body (4,5). Additionally, systemic complications have been reported, such as the massive release of proinflammatory cytokines and thromboembolic events in various organs (6). Consequently, SARS-CoV-2 is regarded as a virus that primarily replicates in the nasopharyngeal area but can also cause a systemic infection that affects different organs with a high mortality rate.
Little is known about the initial infection process in the alveolar lung tissue, particularly about mechanisms that destroy the lung and mechanisms that allow the virus to affect different organs in the body. In vivo infection models that closely reflect the patient's situation are largely lacking, in part due to the different disease development in mice and humans. Furthermore, extensive safety measures are required to infect mice with SARS-CoV-2. Up to now, SARS-CoV-2 in vitro infection models have been mainly performed with human airway (nonalveolar) cells or nonhuman cell lines that naturally express the ACE2 viral receptor, such as the African green monkey Vero-76 cell line (7). These cells lack organ-and species-specific characteristics of human lung epithelial cells. For this purpose, cancerous human lung epithelial cells (Calu-3 cells) can, at least to some extent, reflect the response of the lung epithelium to viral infection (8).
In our study, we present a human-specific in vitro chip model composed of cells of human origin susceptible to a SARS-CoV-2 infection. This model was only recently developed in our lab (9). In the present study, it was modified by using SARS-CoV-2 permissive epithelial cells (Calu-3 cells). The epithelial and vascular cells (primarily isolated human umbilical vascular endothelial cells; HUVECs) were cocultured with macrophages (primarily isolated peripheral blood mononuclear cells; PBMCs), building a complex cell culture system. This composition is relevant not only for the gas exchange in the lung but also for an adequate immune response.
We were able to show that SARS-CoV-2 replicates in the epithelial layer while inducing an acute and robust inflammatory response followed by the destruction of the epithelial layer. Interestingly, in this infection scenario, the endothelial cells were not invaded by SARS-CoV-2 and did not propagate the virus, but nevertheless, the epithelial/endothelial barrier integrity was disrupted.
Mono-cell culture systems can be infected by SARS-CoV-2 and produce replication complexes at ER-derived membranes. First, we infected mono-cell culture systems with SARS-CoV-2 and compared the infection rate between Vero-76 cells and Calu-3 cells. It is already well known that Vero-76 cells can be efficiently infected by SARS-CoV-2 (7,13). To better mimic the situation in the human lung, we performed the infection in Calu-3 cells. Using transmission electron microscopy (TEM), we demonstrated that Vero-76 and Calu-3 cells efficiently propagate the virus (Fig. 1). Figure 1 illustrates infected Vero-76 (Fig. 1A) and Calu-3 cells (Fig. 1B) containing viral replication organelles. In the closeup panels, protein accumulation and generation of doublemembrane vesicles are visible in both cell types. In the middle panel (Fig. 1A), virion assembly in the endoplasmic reticulum-Golgi intermediate compartment (ERGIC) and a Golgi complex are imaged containing morphologically complete viral particles. Here, the particles are packed to be transported to the cellular surface for virus release. In both cell types, we see that some particles are still attached to the host cell membrane, whereas some viral particles are already fully released (lower right panels). These results indicate that SARS-CoV-2 induces replication complexes at ER-derived membranes, which were already shown for other types of coronaviruses (14) and which confirms the findings for Vero E6 cells (15).
In Fig. 2A, immunofluorescence measurements compare infected Vero-76 cells with infected Calu-3 cells, HUVECs, and macrophages. Most remarkably, in Vero-76 and Calu-3 cells, viral particles can be visualized to a similar extent using specific antibodies against SARS-CoV-2 spike proteins. These results are confirmed by Western blot analysis, demonstrating increased levels of SARS-CoV-2 spike protein in the cell lysates after 8 h and 24 h, which is missing in HUVECs and macrophages (Fig. 2B).
Additionally, we verified progeny virus particles by performing plaque assays from supernatants taken at 8 h and 24 h postinfection. Our results demonstrate increased replication during ongoing infection of Vero-76 and Calu-3 cells, but no viral propagation was found in HUVECs and macrophages (Fig. 2C). Measuring viral RNA loads in different infected host cell types 24 h postinfection, we found high RNA levels only in Vero-76 and Calu-3 cells (Fig. 2D). Thus, HUVEC mono-cell cultures and macrophages could not productively be infected by SARS-CoV-2 ( Fig. 2A to D).
We further analyzed the host response to the infection by measuring the cytokine mRNA expression of Calu-3 cells. Many inflammatory cytokines were significantly increased compared to control cells 24 h postinfection (Fig. 3). These results reflect the high cytokine levels found in COVID-19 patients (16), indicating that infected epithelial cells might contribute to the disease development in severe COVID-19 cases. Since different cytokines and chemokines are involved in the infection process, a robust immune response has been associated with a severe clinical course (17). Our results clearly show that the epithelial cell line Calu-3 can be efficiently infected by SARS-CoV-2, propagates the virus, and responds to the viral infection with a strong cytokine release.
SARS-CoV-2 infects epithelial cells within the human chip model, causing a strong IFN response. In the next step, we modified our human chip model (9) by seeding Calu-3 cells on the epithelial side, primarily isolated HUVECs on the endothelial side, and integrated PBMCs to represent the immune response. However, macrophages that are mainly involved in inflammatory responses (18) do not show productive viral replication (19) (Fig. 2A to D). The applied system is ventilated and perfused and can be infected with SARS-CoV-2 via the epithelial side. Using the viral particles isolated from the three COVID-19 patients' specimens, an infection by SARS-CoV-2 of the epithelial In general, the production of IFN is the most efficient way of fighting viral infections; e.g., secretion of type I IFN (IFN-a/b) exhibits direct antiviral effects by inhibiting viral replication (20) among many other interferon effects that promote the immune response to infection (21). Yet evasion strategies for different types of coronaviruses have been described. The viruses express factors and possess strategies to inhibit IFN induction/expression (22) or IFN signaling or to increase IFN resistance, which is reviewed by Kindler and Thiel (21). Consequently, SARS-CoV-2 is apparently able to cope with the interferon response of epithelial cells, which is reflected by our measurements, demonstrating efficient viral replication and persistence for up to 40 h despite a strong epithelial-mediated interferon response ( Fig. 3 and 4, Fig. S2).
In contrast, we did not detect viral propagation and did not measure an interferon response at the endothelial side of the biochip (Fig. 4C). We could not visualize viral components within endothelial cells either at 8 h (Fig. S2B), 28 h (Fig. 4B), or 40 h postinfection (Fig. S2D), demonstrating that the viral particles do not productively infect endothelial cells in the human chip model. Additionally, endothelial mono-cell culture systems could not be infected by SARS-CoV-2 ( Fig. 2A to D), confirming the cell-type specificity of the viral pathogens for lung epithelial cells. This is in line with in vivo studies that describe only a weak IFN response in the serum of COVID patients (23). In animal models with mouse-adapted SARS virus, a delayed onset of the IFN response resulting in immune dysregulation was described (24). The weak and delayed IFN levels in the serum are probably due to the host cell specificity of SARS-CoV-2, as lung epithelium represents the main infection focus, and endothelial cells are hardly infectible.
Taken together, these results indicate that endothelial cells are not the primary tar- Further, it is in line with the observation that endothelial layers of the alveolar capillaries of deceased COVID-19 patients were still intact, but epithelial tissue was found to be seriously damaged (5). Although the mechanism is not explored, the increased proinflammatory cytokine release might cause an endothelial dysfunction. In addition, comorbidities such as obesity and diabetes might render endothelial cells more susceptible to being infected, as recently described (27,28). There is clinical evidence for severe courses of COVID-19, in particular, when preexisting endothelial damage is suspected (29). Further studies are required to elaborate the impact of the endothelial phenotype and the infection conditions at which these cells are targeted by SARS-CoV-2. In the model used in this study, the pneumotropic features of SARS-CoV-2 could be confirmed based on viral uptake and replication in epithelial cells accompanied by an interferon response described in previous studies. The viral particles were not transferred to the neighboring endothelial layer, although the cells were cocultivated in close proximity.
SARS-CoV-2 disrupts the cellular barrier within the human chip model. Next, we analyzed the barrier function of the epithelial-endothelial cell layers in our biochip model. Many clinical case reports and studies describe that critically ill COVID-19 patients develop severe lung destruction and a systemic sepsis-like syndrome (30,31) that could be partly explained by a disrupted barrier function in the lung. From the immunofluorescence image (Fig. 4), we could observe some destruction of the epithelial or endothelial layer, in particular, 40 h postinfection ( Fig. S2C and D). To better visualize the cell layers during the infection, we performed scanning electron microscopy (SEM) analysis of the surface structures 28 h postinfection. On the epithelial side, we found dead cells and remnants attached to the cell layer. Dying cells are identified by shrinking, balling, disruption of the plasma membrane, and the loss of microvilli, which can be observed in the top and middle panel of the infected cells but to a much lesser extent in mock-treated cells (Fig. 5A, top panel). At the surface of dead cells of the infected epithelial layer (Fig. 5A, bottom panel) large numbers of uniformly sized spherical particles attached to the cell membrane could be found. Based on their shape and size (83.4 6 5.3 nm, n = 80) they can be considered to be SARS-CoV-2 virus particles.
It is well known that respiratory viral pathogens, such as the influenza virus (32), induce cell death (including apoptosis) in the respiratory epithelium. Previously, the 2003 SARS-CoV led to apoptotic cell death induced by membrane proteins via modulation of the Akt pathway (33). Additionally, prolonged stress of the endoplasmic reticulum (ER) was identified as a trigger for apoptosis (34). During the recent pandemic, it has been shown that the largest unique open reading frame (ORF) of the SARS-CoV-2 genome, ORF3a, is associated with a proapoptotic activity (35). These studies indicate that the induction of an apoptotic process in the course of SARS-CoV-2 infection is highly probable.
In contrast, on the endothelial side, we could not observe differences in the morphological appearance between SARS-CoV-2-infected and mock-treated cells. Here, the cell integrity of the cell layers appeared intact apart from shrinking artifacts due to the drying procedure (Fig. 5B). At high magnification (bottom panel), dead cell remnants show granular residues of the cytoplasm, indicating the loss of the plasma membrane, but no viral particles could be visualized.
To measure the epithelial and endothelial cell viability, we performed lactate dehydrogenase (LDH) assays on cells grown in the biochip (Fig. 5C). SARS-CoV-2 infection did not result in LDH release of epithelial or endothelial layers at 8 h postinfection. However, at 28 h postinfection, we observed in the infected epithelial layer a significantly enhanced release of LDH, whereas SARS-CoV-2 did not induce an enhanced LDH release in endothelial cells (Fig. 5C). However, after more extended infection periods (40 h; Fig. 5C), the barrier function on the endothelial side was also affected. These results suggest that even if the endothelial layer is not infected, the cell integrity gets disturbed, most likely by cytokines released by neighboring cells. In this respect, it is known that high cytokine/interferon levels can induce the disruption of the alveolar barrier function (36)(37)(38).
To further analyze the barrier integrity of the biochip system on a functional level, we performed permeability assays using fluorescein isothiocyanate (FITC)-dextran to investigate the endothelial and epithelial cell barrier integrity. We were able to show that SARS-CoV-2 infection did not affect tissue permeability at 8 h postinfection but resulted in significantly increased tissue permeability after 28 h postinfection (Fig. 5D). As a consequence of the disrupted barrier, we detected viral particles by performing plaque assays of the cell supernatants in the endothelial chamber in a time-dependent manner, with very low levels at 8 h postinfection and higher levels at 40 h postinfection (Fig. 5E). These results show that endothelial cells are affected by the viral infection at late time points and that the disturbed cell integrity results in translocation of the viral particles over the cellular barrier.

DISCUSSION
In this article, we present an in vitro chip model based on human cells that can be efficiently infected by SARS-CoV-2 (Fig. 6). The epithelial cells of the model were prone to SARS-CoV-2 infection and propagated viral particles with high viral loads. These findings are in line with clinical observations that by far the highest viral burdens are measured in lung tissue (3). This phenomenon can be explained by the cell tropism of SARS-CoV-2 to airway cells that contribute to the high shedding of viral particles in the respiratory system and the high infectivity of patients via aerosols. Nevertheless, due to the systemic symptoms in severe COVID-19 patients, it has been discussed whether other cell types besides the airway epithelium are targeted by SARS-CoV-2, as well. In particular, vascular complications, such as thrombotic events (39), could result from the dissemination and propagation of viral particles in the endothelial system. In our model system, we could not confirm viral invasion into endothelial cells, although the cells were cultured in close proximity to the infected epithelium.
These results are in line with other recent publications, indicating that endothelial cells are resistant to SARS-CoV-2 infection and suggesting that vascular dysfunction is caused by factors and mediators released by adjacent infected cells (25,26).
However, with increased time of infection (40 h), the endothelial cells become damaged, resulting in a decline in tissue barrier function. This effect is most likely mediated by the cytokine release of the infected neighboring epithelium. Cytokine release is known to disturb various cellular functions, such as protein biosynthesis and barrier integrity. Many stud- ies reveal that most severe cases of SARS-CoV-2 infections are not only due to enhanced viral burden, but are to a large extent due to aberrant immune responses (38). In our study, we present an infection model that could be further used to investigate several aspects of the SARS-CoV-2 infection. (i) At first, the cellular interaction can be analyzed in detail with increasing complexity. Here, the interaction between endothelial and epithelial cells and the role of different immune cells that can be integrated into the biochip could be elaborated. (ii) Another crucial aspect is preexisting damage, such as diabetic vascular changes or inflammatory foci that may promote a SARS-CoV- Consequently, our biochip model represents a valuable tool to study many aspects of SARS-CoV-2 infections.
For mono-cell culture studies with human monocyte-derived macrophages, monocytes were isolated from leukocyte concentrates obtained from freshly withdrawn peripheral blood of healthy male and female adult human donors (18 to 65 years old), which were provided by the Institute of Transfusion Medicine at the University Hospital Jena, Germany. The experimental protocol was approved by the ethics committee of the University Hospital Jena. All methods were performed in accordance with the relevant guidelines and regulations. Peripheral blood mononuclear cells (PBMC) were separated using dextran sedimentation, followed by centrifugation on lymphocyte separation medium (Histopaque-1077; Sigma- For the integration of PBMCs into the human chip model, cells were isolated and cultivated as previously described (9). Human umbilical vein endothelial cells (HUVECs) were isolated from anonymously acquired human umbilical cords according to the Declaration of Helsinki, "Ethical Principles for Medical Research Involving Human Subjects" (1964). After the cord veins were rinsed with 0.9% NaCl, endothelial cells were detached with collagenase (0.01%, 3 min at 37°C), suspended in M199/10% FCS, washed once (500 Â g, 6 min), and seeded on a cell culture flask coated with 0.2% gelatin. Then, 24 h later, full growth medium was added (M199, 17.5% FCS, 2.5% human serum, 7.5 mg ml 21 endothelial mitogen, 7.5 U ml 21 heparin, 680 mM glutamine, 100 mM vitamin C, 100 U ml 21 penicillin, 100 mg ml 21 streptomycin). HUVECs from the second passage were seeded on 30-mm dishes or on 90-mm dishes at a density of 27,500 cells/ cm 2 . Experiments were performed 72 h after seeding. For the cultivation of the human chip model, we used Calu-3 cells and macrophages at the epithelial side and HUVECs at the endothelial side. The multiorgan tissue flow (MOTiF) biochips were manufactured and obtained from microfluidic ChipShop GmbH (Jena, Germany), as explained previously (9).
The in vitro infection of the cell lines was performed as follows: cells were washed with PBS and either left uninfected (mock) or infected with SARS-CoV-2 with a multiplicity of infection (MOI) of 1 for 120 min in medium (EMEM with HEPES modification and 5 mM L-glutamine for Vero-76 cells, RPMI 1640 for Calu-3 cells, RPMI 1640 with GM-CSF and penicillin/streptomycin for macrophages, and endothelial cell medium for HUVECs) supplemented with 10% FCS. Subsequently, supernatants were removed, and cells were supplemented with fresh medium containing 10% FCS and further incubated at 37°C and 5% CO 2 .
For the infection of the human chip model, cells were washed with PBS once, followed by treatment with PBS (mock) at 37°C and RPMI (0.2% autologous human serum, 1 mM MgCl 2 , 0.9 mM CaCl 2 ) or infection with SARS-CoV-2 (1 MOI). After 90 min, incubation cells were washed and supplemented with medium. Afterward, cells were incubated for the indicated times at 37°C and 5% CO 2 .
Transmission electron microscopy. Confluent monolayers of Vero-76 cells or Calu-3 cells (9-cm petri dishes) were infected with SARS-CoV-2 (isolate 5159) using an MOI of 1. After 24 h, supernatants were removed, and samples were fixed with freshly prepared modified Karnovsky fixative consisting of 4% wt/vol paraformaldehyde and 2.5% vol/vol glutaraldehyde in 0.1 M sodium cacodylate buffer, pH 7.4, for 1 h at room temperature. After washing 3 times for 15 min each with 0.1 M sodium cacodylate buffer (pH 7.4), the cells were postfixed with 2% wt/vol osmium tetroxide for 1 h at room temperature. Subsequently, the cells were rewashed with 0.1 M sodium cacodylate buffer (pH 7.4), thoroughly scraped off the petri dishes, and pelleted by centrifuging at 600 Â g for 10 min. During the following dehydration in ascending ethanol series, poststaining with 1% wt/vol uranyl acetate was performed. Afterward, the pellets were embedded in epoxy resin (Araldite) and ultrathin-sectioned (70 nm) using a Then, 1 ml cDNA was added to 19 ml master mix (containing 10 ml SYBR green, 1.5ml forward primer [10 mM], 1.5 ml reverse primer [10 mM], and 6 ml RNase-free double-distilled water [ddH 2 O] per sample; for primer sequences, see Table S1), and the qRT-PCR was started using the following cycle conditions: 95°C for 2 min, followed by 40 cycles of 95°C for 5 sec and 60°C for 10 sec. The qRT-PCR cycle was ended by a stepwise temperature increase from 60°C to 95°C (1°C every 5 sec).
Detection of SARS-CoV-2 by using qRT-PCR. For the determination of SARS-CoV-2, we used the QIAamp viral RNA minikit (Qiagen, Hilden, Germany) according to manufacturer's guide. This was followed by a qRT-PCR purchased by RIDAgene (r-biopharm, Darmstadt, Germany) on Rotor-Gene Q (Qiagen, Hilden, Germany) to detect the E-gene of SARS-CoV-2. The RNA standard curve, prepared from the positive control of the RIDAgene (r-biopharm, Darmstadt, Germany) kit, cycle conditions were set as follows: 10 min at 58°C, 1 min at 95°C, and 45 cycles of 95°C for 15 sec and 60°C for 30 sec.
Statistical analysis. Statistical analysis was performed using Prism 8 (GraphPad Software). Statistical methods are described in the figure legends.
Data availability. The newly determined sequences are available in GenBank under accession numbers MW633322 to MW633324.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.2 MB. SUPPLEMENTAL FILE 2, PDF file, 0.8 MB.