A human disease model of SARS-CoV-2-induced lung injury and immune responses with a microengineered organ chip

Coronavirus disease 2019 (COVID-19) is a global pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) that seriously endangers human health. There is an urgent need to build physiological relevant human models for deep understanding the complex organ-level disease processes and facilitating effective therapeutics for COVID-19. Here, we first report the use of microengineered alveolus chip to create a human disease model of lung injury and immune responses induced by native SARS-CoV-2 at organ-level. This biomimetic system is able to reconstitute the key features of human alveolar-capillary barrier by co-culture of alveolar epithelial and microvascular endothelial cells under microfluidic flow. The epithelial cells on chip showed higher susceptibility to SARS-CoV-2 infection than endothelial cells identified by viral spike protein expression. Transcriptional analysis showed distinct responses of two cell types to SARS-CoV-2 infection, including activated type I interferon (IFN-I) signaling pathway in epithelium and activated JAK-STAT signaling pathway in endothelium. Notably, in the presence of circulating immune cells, a series of alveolar pathological changes were observed, including the detachment of endothelial cells, recruitment of immune cells, and increased production of inflammatory cytokines (IL-6, IL-8, IL-1β and TNF-α). These new findings revealed a crucial role of immune cells in mediating lung injury and exacerbated inflammation. Treatment with antiviral compound remdesivir could suppress viral copy and alleviate the disruption of alveolar barrier integrity induced by viral infection. This bioengineered human organ chip system can closely mirror human-relevant lung pathogenesis and immune responses to SARS-CoV-2 infection, not possible by other in vitro models, which provides a promising and alternative platform for COVID-19 research and preclinical trials.


Introduction
Coronavirus disease 2019 (COVID-19) pandemic broke out in late 2019 and quickly became a global epidemic [1][2][3][4]. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), the causative virus of COVID-19, has infected over ten million individuals up to July 2020 according to the report from World Health Organization (WHO), and the number of patients and deaths are increasing globally. COVID-19 patients exhibit a broad spectrum of disease progression and multiple clinical features including fever, dry cough, and ground-glass opacities [5,6]. Human lung is the primary target for SARS-CoV-2 infection, which is characterized by the process ranging from mild syndrome to severe lung injury and multi-organ failure. Many severe cases of COVID-19 develop progressive respiratory failure, leading to death due to diffuse alveolar damage, inflammation and pneumonia [6][7][8][9]. Based on pathological features of COVID-19 by biopsy samples, inflammatory infiltration of mononuclear cells or lymphocytes could be observed in lung interstitial tissues or alveolar cavities [9,10]. Particularly, previous studies from severe patients have suggested that an excessive inflammatory cytokine storm induced by SARS-CoV-2 often result in aberrant immunopathology and lethal outcome. However, in-depth mechanism of pathogenesis of COVID-19 is still not clear.
Presently, SARS-CoV-2 infection is studied mostly relying on monolayer cultures of cell lines and primary tissue cells [11,12], human organoids [13,14], and animal models [15,16]. However, all of these models have their limitations. For example, monolayer cell cultures are oversimplified and cannot exhibit the complex structure and functions of human organ-specific microenvironments in vivo. Human organoids (e.g., lung organoids) can provide multiple cell types and more complex tissue-specific functions, but they cannot model organ-level features of lung, such as tissue-tissue interfaces, blood flow, cross-talk between epithelium and endothelium, and immune cell-host responses, which are essential for the pathological progression of viral infection in pulmonary tissue in vivo. In addition, animal models of SARS-CoV-2 infection have been established for validating therapeutics of drugs or vaccines [15,16]. However, given the species difference, their targeted organs or systemic responses to SARS-CoV-2 infection may be significant different from human individuals. Moreover, the expensive cost, time-consuming process and animal ethics should be seriously considered. As such, it is highly desirable to develop alternative preclinical models that can better reflect human-relevant organ pathophysiology and responses to accelerate SARS-CoV-2 research and candidate therapeutics for COVID-19.
Significant advances in bioengineered organs-on-chips technology have made it possible to reconstruct 3D human organotypic models in vitro by recapitulating the key functions of living organisms in microfluidic device, such as intestine, heart, liver, lung and kidney [17]. They are able to model human-relevant organ physiology, holding great potentials for disease studies and drug testing [18][19][20][21]. Here, we first report the establishment of a microengineered human pulmonary model of SARS-CoV-2 infection on a chip that allows to recapitulate the human relevant lung pathophysiology and immune responses associated with COVID-19 in vitro. The biomimetic lung alveolar chip consisted of two channels (alveolar lumen and vascular channels) sandwiched with extracellular matrix (ECM)-coated porous membrane that enables to the co-culture of human alveolus epithelial cells, pulmonary microvascular endothelial cells, and immune cells under

Characterization of microengineered human alveolus chip
Lung is the primary target organ in the progress of COVID-19. Human alveoli contained alveolar-capillary barrier is an important component unit of lung. Alveolar-capillary barrier is composed of alveolar epithelial and vascular endothelial cells interacting across a 3D ECM, which plays vital roles in gas exchange and prevention of external hazardous substances invasion or virus infection (Fig. 1A). Clinical and histopathological evidences suggested alveolar function is seriously damaged by SARS-CoV-2. In order to create the human lung infection model by SARS-CoV-2 in vitro, we initially design and construct a biomimetic human alveolus chip in a multilayer microfluidic device under dynamic culture conditions. The microfluidic device consists of two channels separated by a thin and porous PDMS membrane coated with ECM (Fig. 1B). The PDMS membrane allows to form bio-interface and is beneficial to the interactions of upper and bottom cell layers. The culture chamber permits the perfusion of media flow, which can facilitate nutrients exchange and waste exclusion.
In this work, human alveolar epithelial type II cell (AT II) line (HPAEpiC) and lung microvasculature cell line (HULEC-5a) were seeded on the upper and lower side of porous membrane, separately. These two types of cells were cultured for 3 days until confluent into monolayers under continuous media flow (50 μl/h) in upper and bottom channels, thus forming an alveolus epithelium-endothelium tissue interface. The integrity of formed tissue barrier was assessed by the expression of adherent junction proteins in both human epithelium and endothelium.
Immunostaining analysis showed that epithelial cells could form adherent junctions identified by E-cadherin, and endothelial cells formed conjunctions identified by VE-cadherin, respectively (supplementary Fig. S1). Furthermore, the integrity of barrier under different culture conditions was assessed by the diffusion rate of FITC-dextran between the two parallel channels (supplementary Fig.   S2). The barrier permeability was lower under fluid flow than that in static cultures, indicating the important role of flow in maintaining the function and integrity of alveolar-capillary barrier.

SARS-CoV-2 infection in human alveolus chip
It has been reported that SARS-CoV-2 uses ACE2 as a host receptor for cellular entry, and transmembrane serine proteinase 2 (TMPRSS2) for Spike protein priming [12,[22][23][24]. Prior to create the SARS-CoV-2 infection model based on human alveolus chip, we sought to identify the susceptibility of alveolar epithelial cells to this virus. Initially, we examined the expression of ACE2 and TMPRSS2 proteins in HPAEpiC and HULEC-5a cells, respectively ( Fig. 2A). The western blot data showed the positive expression of ACE2 and TMPRSS2 in both cell types [12,25], and the higher ACE2 expression in HPAEpiC than HULEC-5a cells. HPAEpiC cells were then infected with SARS-CoV-2 at a MOI of 10 in monolayer cultures. At 3 days post-infection, more than 20 % Spike protein-positive cells were observed (Fig. 2B). To further examine the ultrastructure of SARS-CoV-2-infected cells, transmission electron microscope (TEM) analysis of mock-or SARS-CoV-2-infected HPAEpiC cells were performed (Fig. 2C, D). The TEM micrographs showed that mock cells exhibited primary AT II cell-like morphological characteristics, including a small cellular size with square or round shape (Fig. 2C1), microvilli on free surface (Fig. 2C2) and lamellar bodies within cell body (Fig. 2C3) [26,27]. In infected cells, lots of viral particles were detected and distributed in clusters within cell bodies as shown in Fig. 2D, indicating the susceptibility of HPAEpiC cells to SARS-CoV-2 infection.
To model alveolar infection by SARS-CoV-2, virus was inoculated into the upper alveolus channel of chip at a MOI of 10, and cell were cultured for 3 days. The predominated expression of spike protein was observed in epithelial cells, demonstrating the viral infection and massive replication in alveolar epithelium ( Fig. 3A and B), but not in endothelial cells. In addition, there are not obvious changes in the organization of adherent junction proteins in HPAEpiC cells (E-cadherin) and HULEC-5a cells (VE-cadherin), as well as the confluent rate of epithelial cells and endothelial cells ( Fig. 3C and D). These results suggested that SARS-CoV-2 can infect and replicate primarily in the alveolar epithelial cells, but not in endothelial cells.

Transcriptional analysis of host cells responses to SARS-CoV-2 infection
To gain a global understanding of transcriptional responses to SARS-CoV-2 infection, we performed RNA-seq analysis of HPAEpiC and HULEC-5a cells following virus infection in the alveolus chip.
The ratio of virus-aligned reads over total reads in each sample was calculated to estimate the viral replication levels in these two cell types. The results showed that the ratio of viral reads was much higher in infected-HPAEpiC cells than HULEC-5a cells ( Fig. 4A and B), which are consistent with the immunostaining analysis (Fig. 3B). It revealed that human alveolar epithelial cells were more permissive to SARS-CoV-2 infection than microvascular endothelial cells, similar to the histopathological findings from autopsy reports [28]. Hierarchical clustering analysis showed SARS-CoV-2 infection elicited broad transcriptional changes in both HPAEpiC and HULEC-5a cells ( Fig. 4C). For identification of differentially expressed genes (DEGs), the cutoffs for the fold change and P value were set to 2.0 and 0.05, respectively. By combining the two data sets, we found the two cell types only shared 52 overlapping up-regulated DEGs (6.2% of total up-regulated DEGs) and only 43 overlapping down-regulated DEGs (5.5% of total down-regulated DEGs) (Fig. 4D). These

Immune responses in human alveolus chip following SARS-CoV-2 infection
The accumulation and extensive infiltration of immune cells observed in the lungs may contribute significantly to the pathogenesis in patients infected with respiratory viruses [29]. We next explored the roles of human circulating immune cells in the alveolar pathological process after SARS-CoV-2 infection. In this study, PBMCs were isolated from healthy human blood and infused into the lower Especially, viral infection caused a 10-fold increase in the level of IL-1β and IL-6 ( Fig. 5F and G).
We also examined the inflammatory cytokines secretion in culture media from the epithelial layer, which showed similar results to that from the vascular channel (supplementary Fig. S4). These results indicated that SARS-CoV-2 infection induced the recruitment of PBMCs, and aggravated inflammatory response, which are consistent with clinical findings. It appears this human organ chip could reflect human relevant pathological changes and host immune response to SARS-CoV-2, which are difficult to replicate in existing in vitro models.

Assessment of potential anti-viral therapeutics of remdesivir
To explore the potential therapeutics against SARS-CoV-2, we treated the virus infected human alveolus chip with remdesivir. Remdesivir is recognized as a promising antiviral compound against many RNA viruses (e.g., SARS, MERS-CoV), including SARS-CoV-2 [31][32][33]. Recent clinical trials showed remdesivir can shorten the disease course and reduce mortality of severe COVID-19 patients, and it has been approved by FDA in US [34]. In this study, an indicated dose of remdesivir (1μM) was added into the monolayer culture of HPAEpiC cells at 1h post-infection of SARS-CoV-2. After administration for 3 days, the culture supernatants were collected for virus titers determination by qRT-PCR. A marked decrease of virus titers was detected in the infected-HPAEpiC cells following remdesivir treatment (Fig. 6A). Furthermore, we tested the antiviral efficacy of remdesivir in the infected chip model with the addition of PBMCs in the vascular channel. As shown in Figure 6B and C, remdesivir treatment could restore the damage of epithelial layers and endothelial layer to some extent ( Fig. 6B and C). These results indicated the potential role of remdesivir in suppressing SARS-CoV-2 replication and alleviating the virus-induced injury of alveolar-capillary barrier. It also suggested early administration of remdesivir might be helpful for disease control and alleviating lung injury for COVID-19 patients.

Discussion:
In this study, we created a microengineered human disease Meanwhile, the activation of JAK-STAT signaling pathway was observed in pulmonary microvascular endothelial cells after viral infection. It is well known that cytokines (e.g. IL-6) are enable to activate JAK-STAT pathway and further regulate different cellular and immune processes [37,38]. Ruxolitinib, a JAK inhibitor could effectively relieve symptoms of patients with severe COVID-19 [39,40]. Our findings suggested the potential therapeutic target for COVID-19 treatment by targeting JAK-STAT signaling pathway in microvascular endothelium.
Clinically, severe COVID-19 patients often display inflammatory cytokine storm that is associated with excessive immune responses, which may aggravate respiratory failure and cause multi-organ damage. Circulating cytokines, including IL-1β, IL-6, IL-8 and TNF-α are significantly elevated in patients with severe COVID-19 [41,42]. Therefore, we detected these cytokines in the alveolus chip This human organ chip system provided a synthetic strategy to rebuild human organs and analyze pathological responses at organ level by flexibly varying system parameters, which is opening up new avenues for COVID-19 research and drug development.

Device fabrication
The
After one hour, cells were washed three times with PBS and kept in fresh medium for 3 days. At day HPAEpiC and HULEC-5a cells cultured on chip were fixed for immunofluorescence analysis or lysed for RNA-seq data analysis, respectively.

Permeability assay
The alveolar-capillary barrier permeability was assessed by detecting the FITC-dextran diffusion rate from the lower vessel layer to the upper alveolar channel. After a 3-day co-culture, the medium with FITC-dextran (40 kDa, 1 mg/mL) was then infused into the bottom channel of device. The media were collected from the upper channel at different time points (0 h, 1 h and 2 h) and the fluorescence intensity was measured using microplate system (ABI Vii 7).

Western blot analysis
Protein samples were separated on 10% SDS-PAGE and then transferred onto 0.

Immunostaining
HPAEpiC cells cultured on well plate were washed with PBS and fixed with 4% PFA at 4 overnight. Cells were then permeabilized with 0.2% Triton X-100 in PBS (PBST buffer) for 5 min and blocked with PBST buffer containing 5% normal goat serum for 30 minutes at room temperature.
Antibodies were diluted with PBST buffer. Cells were stained with corresponding primary antibodies at 4°C overnight and with secondary antibodies (supplementary Table S1) at room temperature for 1 hour. After staining with secondary antibodies, cell nuclei were counterstained with DAPI.
For immunofluorescent imaging of alveolus chip, cells were washed with PBS through the upper and bottom channels and fixed with 4% PFA. The fixed tissues were subjected to immunofluorescence staining by the same procedure as described above. All images were acquired using a confocal fluorescent microscope system (Carl zeiss LSM880). Image processing was done using ImageJ (NIH).

Analysis of inflammatory cytokines
To analyze the released cytokines, the media in the vessel channel were collected from each chip.
The concentrations of IL-6, IL-8, IL-1β, and TNF-α were measured using the corresponding human ELISA kits (Biolegend, USA) according to the manufacturer's instructions.

Real-time quantitative PCR (qRT-PCR)
Virus titers were determined by viral RNA detection using qRT-PCR. The culture supernatant for each condition was harvested for RNA extraction using the HP Viral RNA Kit (Roche, Cat no.

11858882001) according to the manufacturer's instructions. QRT-PCR was performed using One
Step

RNA extraction, library preparation and sequencing
HPAEpiC cells and HULEC-5a cells were collected separately from the chips, and total RNAs were extracted from samples using TRIzol (Invitrogen) following the methods by Chomczynski et al. [45].
DNA digestion was carried out after RNA extraction by DNaseI. RNA quality was determined by examining A260/A280 with NanodropTM OneCspectrophotometer (Thermo Fisher Scientific Inc).

RNA-seq data analysis
Raw sequencing data was first filtered by Trimmomatic (version 0.36), low-quality reads were discarded and the reads contaminated with adaptor sequences were trimmed. Clean Reads were further treated with in-house scripts to eliminate duplication bias introduced in library preparation and sequencing. In brief, clean reads were first clustered according to the UMI sequences, in which reads with the same UMI sequence were grouped into the same cluster, resulting in 65,536 clusters.
Reads in the same cluster were compared to each other by pairwise alignment, and then reads with sequence identity over 95% were extracted to a new sub-cluster. After all sub-clusters were generated, multiple sequence alignment was performed to get one consensus sequence for each sub-cluster.
After these steps, any errors and biases introduced by PCR amplification or sequencing were eliminated.

Statistical analyses
Data were collected in Excel (Microsoft). Differences between two groups were analyzed using a Student's t-test. Multiple group comparisons were performed using a one-way analysis of variance (ANOVA) followed by post-hoc tests. The bar graphs with error bars represent mean ± standard deviation (SD). Significance is indicated by asterisks: *, p < 0.05; **, p < 0.01; ***, p < 0.001.

Data availability
All relevant data are available in the manuscript or supporting information.