Organotypic human ex vivo models for coronavirus disease 2019 research and drug development

Since the discovery of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) in late 2019, intense research efforts on an unprecedented scale have focused on the study of viral entry mechanisms and adaptive immunity. While the identification of angiotensin-converting enzyme 2 (ACE2) and other co-receptors has elucidated the molecular and structural basis for viral entry, the pathobiological mechanisms of SARS-CoV-2 in human tissues are less understood. Recent advances in bioengineering have opened opportunities for the use of organotypic human tissue models to investigate host–virus interactions and test antiviral drug candidates in a physiological context. Although it is too early to accurately quantify the added value of these systems compared with conventional cell systems, it can be assumed that these advanced three-dimensional (3D) models contribute toward improved result translation. This mini-review summarizes recent work to study SARS-CoV-2 infection in human 3D tissue models with an emphasis on the pharmacological tools that have been developed to understand and prevent viral entry and replication.


Introduction
The coronavirus disease 2019 (COVID- 19) pandemic has resulted in an unprecedented shift in research activities aimed at collectively combining efforts to develop treatments for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) and minimize its spread in the community. This has led to the uncoordinated use of different cell and tissue models to study SARS-CoV-2 infection and treatment modalities [1]. The choice of model system has important implications in the context of functional pharmacology, where lead compounds encounter a bottleneck in drug discovery at the preclinical or clinical phase because of a lack of therapeutic value. What complicates the matter further is that although COVID-19 is primarily a respiratory disease, it is now clear that it can also result in a multitude of extrapulmonary manifestations, including liver and renal injury, endothelial damage, neurologic symptoms, and beta-cell dysfunction [2].
The expression of SARS-CoV-2 receptors and entry factors has been studied across healthy human tissues [3], but the effects of infection on host cell molecular signatures are limited to the airway epithelium, lung, and intestine and are mostly based on biopsy samples. To allow for longitudinal profiling, the establishment of an accessible panel of human organotypic threedimensional (3D) tissue models provides a promising tool kit to study the tissue-specific pathobiology of viral infections. Previously, 3D models have been used to study a multitude of enteric infections, including coxsackie B1 virus [4], norovirus [5], rotavirus [6] and enterohemorrhagic Escherichia coli [7], hepatitis viruses [8,9], and respiratory viruses, such as respiratory syncytial virus [10], influenza virus [11,12], and parainfluenza virus 3 [13]. The choice of model system depends on several factors, but perhaps the biggest differences are found when investigating the pharmacology of conventional prophylactics versus vaccines. While prophylactics largely depend on the nature of the selected cell system, vaccines rely on an understanding of the virus itself. In the case of COVID-19, which can progress to affect multiple organs, further consideration should be given to the stage of infection (i.e. initial infection vs. multiorgan failure).
To study the complex pathophysiology of SARS-CoV-2 infection in human tissues, cell lines and animal models have important limitations because of species differences in virus tropism and cell lineespecific peculiarities of the host cell response. For instance, since their isolation in the late 1960s, African green monkey kidney cells (Vero) have routinely been used as host cells for virus propagation and the testing of antiviral drug candidates. However, Vero cells lack the capacity to produce interferons, which is an important consideration, particularly for the study of interferon susceptible viruses, such as SARS-CoV-2. Such caveats necessitate the use of other emerging culture models, such as stem cellederived organoids and 3D primary tissue models, that may better represent the pathophysiology of infection and increase the fidelity of result translation ( Table 1). For methodological details, as well as critical discussions of the advantages and limitations of these culture systems, we refer the interested reader to recent comprehensive reviews [14e17].

3D human airway models
Patients with severe COVID-19 can develop acute respiratory distress syndrome characterized by respiratory failure with hypoxemia and acute bilateral pulmonary infiltrates and histopathologic patterns of diffuse alveolar damage and pulmonary microthrombi [18]. Several distal lung epithelial models have been developed that can be infected with SARS-CoV-2 and allow to study viral effects in the distal gas-exchange region of the lungs (Figure 1a).
Human primary alveolar organoids have been generated from cells isolated from normal human distal lung tissue co-cultured with MRC-5 human lung fibroblasts [19]. These organoids were susceptible to SARS-CoV-2 infection and showed robust induction of host response genes, including various cytokines, as well as cell autonomous and nonecell autonomous apoptosis, indicative of alveolar injury. At physiological concentrations, remdesivir significantly reduced viral load, whereas no effects were observed for hydroxychloroquine.
Human alveolar type II (AT2) cells derived from primary human lung tissue have also been used to investigate the host cell response to SARS-CoV-2 infection in long-term 3D culture [20]. These studies revealed rapid viral replication within a few days postinfection paralleled by a strong induction of interferon response gene signatures, indicative of a robust endogenous innate immune response. Furthermore, single-cell sequencing showed that although uninfected AT2 cells were relatively homogeneous, SARS-CoV-2 infection resulted in the subdivision of cells into distinct clusters, of which some were characterized by a loss of AT2 identity and compromised alveolar function. Distal lung organoids have also been generated from adult AT2 or keratin 5epositive (KRT5þ) basal cells [21]. Alveolar organoids differentiated into alveolar type I (AT1) cells, whereas basal cell organoids became luminal lined with differentiated club and ciliated cells within bronchiolar structures. Importantly, distal lung organoids with apical-out polarity were susceptible to infection, and club cells were identified as a novel SARS-CoV-2 target population.
Lung organoids generated by differentiation of primary stem cells derived from tissue sections enabled the modeling of both distal and proximal lung epithelium [22]. Notably, proximal epithelial cells were more permissive to SARS-CoV-2 infection, whereas distal epithelial cells showed increased host cell immune response, highlighting the importance of spatially refined tissue modeling.
Alternatively, human pluripotent stem cells (hPSCs) can be used to generate organoids that recapitulate the cellular complexity of the human distal lung, including AT1 and AT2 cells, stromal cells, pulmonary neuroendocrine cells, as well as airway epithelial cells [23]. Notably, this model allowed for sufficient throughput to screen >1200 US Food and Drug Administration FDA)-approved drugs for their ability to inhibit SARS-CoV-2 entry, and the authors identified candidate molecules, including imatinib and chloroquine, which inhibited infectivity at subtoxic concentrations. hPSC-derived AT2 cells can furthermore be cultivated as monolayers with apical-basal polarization and barrier integrity on filters in aire liquid interface (ALI) cultures [24]. The presence of virus particles in the vicinity of lamellar bodies suggests that secretion of lung surfactant may be dysregulated by SARS-CoV-2 infection. As in cultures from primary cells, infection increased inflammatory signaling and loss of mature alveolar phenotypes. It is, however, important to note that these culture models lacked immune cell components.
To overcome this limitation, lung and macrophage coculture systems were developed based on directed differentiation of hPSCs [25]. Importantly, in this model, both M1 and M2 macrophages inhibited SARS-CoV-2 infection, but only M1 macrophages increased the production of inflammatory cytokines that facilitated lung cell injury. Inhibition of viral entry improved M2mediated viral clearance and resulted in significant protection of lung cells, suggesting that blockade of viral entry or stimulation of M2 macrophages are potential strategies for therapeutic intervention.
Taken together, both organoids and ALI cultures can be used as tools to study SARS-CoV-2 infection of pulmonary tissues. In particular, human primary models faithfully recapitulate the molecular and cellular phenotypes of distal and proximal lung epithelium. However, they lack the ability to be stably propagated and are consequently more suitable for mechanistic studies or target validations rather than large-scale screens. A major drawback of most current models remains the relatively low cellular complexity, particularly the absence of immune cells, which impairs our Table 1 Overview of 3D human tissue models for SARS-CoV-2 infection.

3D human liver models
There is considerable interest in organotypic liver models to study viruseliver cell interactions, as well as the pharmacokinetics and toxicity of candidate drugs. This is reinforced by the observation that liver injury has been detected in up to 60% of severely ill COVID-19 patients [26], suggesting that liver models could be useful to study extrapulmonary disease biology ex vivo.
Hepatic effects of SARS-CoV-2 have been studied in stem cellederived organoids and in organotypic 3D cultures of mature human liver cells. In stem celle derived liver organoids, SARS-CoV-2 infection could be detected in albumin-positive hepatocyte-like cells [27], as well as in cholangiocytes in human liver ductal organoids based on bile ductederived progenitor cultures [28]. Interestingly, in the latter, the authors demonstrated that infection impaired the barrier and bile acid transporting functions of cholangiocytes by modulating the expression of tight junction and bile acid transport genes, thus providing a possible link between SARS-CoV-2 and cholestatic liver injury (Figure 1b).
Using a well-established primary human hepatocyte 3D spheroid model in which liver cells retain their transcriptomic, proteomic, and metabolomic configuration [29e31], it was shown that mature human hepatocytes are permissive to SARS-CoV-2 infection despite expressing only low levels of angiotensin-converting enzyme 2 (ACE2) [32]. Interestingly, exposure to proinflammatory cytokines, particularly class I interferons, resulted in the induction of ACE2 expression and increased liver cell infectivity [33]. However, whether the full-length functional or truncated nonfunctional ACE2 isoform [34,35] becomes induced in human liver remains to be determined. SARS-CoV-2 infection was moreover found to modulate the host cell response of hepatocytes to inflammatory signaling, specifically altering the expression of coagulation factors and genes involved in platelet activation, which might at least in part explain the coagulation defects observed in COVID-19 patients [33].
The development of a model that recapitulates the pathophysiology of SARS-CoV-2 infection ex vivo opened the possibility to test candidate molecules for COVID-19 drug repurposing. Baricitinib, an oral inhibitor of JAK/STAT signaling approved for the treatment of rheumatoid arthritis, had been suggested by artificial intelligenceebased knowledge graphs as a candidate therapy for COVID-19 by virtue of inhibiting both excessive cytokine signaling and numb-associated kinases that regulate receptor-mediated endocytosis and viral entry at nanomolar levels [36]. Importantly, baricitinib significantly reversed virus-induced host cell  transcriptomic alterations and reduced infectivity, as well as viral load in 3D liver spheroids at therapeutically relevant concentrations (Figure 1cee). More importantly and in alignment with these ex vivo findings, baricitinb significantly reduced mortality in observational trials of elderly patients in Italy and Spain [33]. The promising clinical effects have since been corroborated by a large randomized controlled trial [37] and have resulted in the FDA emergency use authorization of baricitinib for COVID-19.
Intestinal ex vivo models The manifestation of gastrointestinal (GI) symptoms in a subset of SARS-CoV-2 patients [38,39] and the high expression level of ACE2 in enterocytes [40] have fueled studies into the pathobiology of SARS-CoV-2 in human intestinal tissues. In contrast to lung and liver, ex vivo studies of SARS-CoV-2 infection of intestinal cells have only been presented in stem cellederived organoid cultures. In small intestinal organoids derived from primary gut epithelial stem cells, SARS-CoV-2 productively infected the enterocyte lineage (Figure 1f), whereas enteroendocrine cells and goblet cells were not infected [41]. Infections of ACE2-positive enterocytes were promoted by the mucosa-specific serine proteases TMPRSS2 and TMPRSS4 [42]. Interestingly, TMPRSS2 is expressed on goblet cells, which are themselves resistant to infection but promotes infection of neighboring enterocytes in trans. Infectious viral particles were detected in stool samples of a patient with diarrheal COVID-19, and viral replication was found to be higher in human colonoids compared with enteroids [43].
Infection, active replication, and viral spread of SARS-CoV-2 also occurred in intestinal organoids derived from hPSC that recapitulated the cellular heterogeneity of the human intestinal lining (Figure 1g). In this model, the authors observed infection of enterocytes as well as Paneth cells, suggesting potential disruption of hormonal secretion and perturbed local immune defense [44]. The model was successfully used to confirm the antiviral effects of remdesivir ( Figure 1h) and the peptidic pan-coronavirus fusion inhibitor EK1, but not of the putative TMPRSS2 inhibitor famotidine. In addition, following a high-throughput screen of FDAapproved drugs for antiviral action, imatinib, mycophenolic acid, and quinacrine dihydrochloride were confirmed to inhibit SARS-CoV-2 infectivity in hPSCderived colonic organoids [23]. In short, these results suggest that both primary gut stem cellederived and iPSC-derived organoids provide suitable tools to study SARS-CoV-2 cellular tropism and infection biology and validate antiviral drug effects.

3D renal organoids
Acute kidney injury constitutes a common complication in severely ill COVID-19 patients, occurring in 0.5e7% of cases and in 2.9e23% of intensive care unit patients [45]. The SARS-CoV-2 entry factors ACE2 and TMPRSS2 are expressed in podocytes and proximal tubule cells, suggesting that these are the primary target cell types. Notably, upon ex vivo culture, ACE2 expression was twice as high in 3D proximal tubule organoids compared with conventional 2D culture, suggesting that 3D cultured cells might be more readily infectable in organotypic culture configurations [46].
Infectivity with SARS-CoV-2 was also modeled using hPSC-derived kidney organoids that featured clear tubular structures and podocyte-like cells (Figure 1i) with molecular phenotypes resembling secondtrimester human fetal kidneys [47,48]. Importantly, viral entry in kidney organoids could be blocked dose dependently by human, but not mouse soluble recombinant ACE2 [48]. Notably, the effect of human soluble recombinant ACE2 was additive with the intracellular viral replication inhibitor remdesivir, suggesting that combinatorial therapies of agents with mechanistically distinct targets might provide promising strategies for COVID-19 [49].

Conclusions
Although conventional monolayer cultures of cell lines were the predominant in vitro models to study virus biology and antiviral drug action during previous coronavirus outbreaks of MERS and SARS-CoV-1, these cell systems were mostly superseded by organotypic 3D human culture systems of various pulmonary and extrapulmonary tissues during the COVID-19 pandemic, developments, which were made possible by the rapid advancements in primary tissue and organoid culture methods in the last decade. While the added value of these changes is difficult to accurately quantify, these physiological systems have demonstrated their translational value for the identification of potential targets and the development and testing of antiviral candidates in phenotypically appropriate contexts. On the basis of these experiences, further investments into human tissue engineering, as well as initiatives to improve the benchmarking, standardization, and accessibility to such models thus seem warranted to increase the preparedness for current and future public health emergencies.

Conflict of interest statement
V.M.L. is the CEO and shareholder of HepaPredict AB and co-founder and shareholder of PersoMedix AB. In addition, V.M.L. discloses consultancy work for Enginzyme AB. The other authors declare no conflicts of interest.