1. Medicine
  2. Stem Cells and Regenerative Medicine
Download icon

Adult stem cell-derived complete lung organoid models emulate lung disease in COVID-19

  1. Courtney Tindle
  2. MacKenzie Fuller
  3. Ayden Fonseca
  4. Sahar Taheri
  5. Stella-Rita Ibeawuchi
  6. Nathan Beutler
  7. Gajanan Dattatray Katkar
  8. Amanraj Claire
  9. Vanessa Castillo
  10. Moises Hernandez
  11. Hana Russo
  12. Jason Duran
  13. Laura E Crotty Alexander
  14. Ann Tipps
  15. Grace Lin
  16. Patricia A Thistlethwaite
  17. Ranajoy Chattopadhyay  Is a corresponding author
  18. Thomas F Rogers  Is a corresponding author
  19. Debashis Sahoo  Is a corresponding author
  20. Pradipta Ghosh  Is a corresponding author
  21. Soumita Das  Is a corresponding author
  1. Department of Cellular and Molecular Medicine, University of California San Diego, United States
  2. HUMANOID CoRE, University of California San Diego, United States
  3. Department of Computer Science and Engineering, Jacobs School of Engineering, University of California San Diego, United States
  4. Department of Pathology, University of California San Diego, United States
  5. Department of Immunology and Microbiology, The Scripps Research Institute, United States
  6. Division of Cardiothoracic Surgery, University of California San Diego, United States
  7. Division of Cardiology, Department of Internal Medicine, UC San Diego Medical Center, United States
  8. Pulmonary Critical Care Section, Veterans Affairs (VA) San Diego Healthcare System, United States
  9. Division of Pulmonary and Critical Care, Department of Medicine, University of California, San Diego, United States
  10. Cell Applications Inc., United States
  11. Division of Infectious Diseases, Department of Medicine, University of California, San Diego, United States
  12. Department of Pediatrics, University of California, San Diego, United States
  13. Department of Medicine, University of California, San Diego, United States
Tools and Resources
  • Cited 0
  • Views 798
  • Annotations
Cite this article as: eLife 2021;10:e66417 doi: 10.7554/eLife.66417

Abstract

Background:

SARS-CoV-2, the virus responsible for COVID-19, causes widespread damage in the lungs in the setting of an overzealous immune response whose origin remains unclear.

Methods:

We present a scalable, propagable, personalized, cost-effective adult stem cell-derived human lung organoid model that is complete with both proximal and distal airway epithelia. Monolayers derived from adult lung organoids (ALOs), primary airway cells, or hiPSC-derived alveolar type II (AT2) pneumocytes were infected with SARS-CoV-2 to create in vitro lung models of COVID-19.

Results:

Infected ALO monolayers best recapitulated the transcriptomic signatures in diverse cohorts of COVID-19 patient-derived respiratory samples. The airway (proximal) cells were critical for sustained viral infection, whereas distal alveolar differentiation (AT2→AT1) was critical for mounting the overzealous host immune response in fatal disease; ALO monolayers with well-mixed proximodistal airway components recapitulated both.

Conclusions:

Findings validate a human lung model of COVID-19, which can be immediately utilized to investigate COVID-19 pathogenesis and vet new therapies and vaccines.

Funding:

This work was supported by the National Institutes for Health (NIH) grants 1R01DK107585-01A1, 3R01DK107585-05S1 (to SD); R01-AI141630, CA100768 and CA160911 (to PG) and R01-AI 155696 (to PG, DS and SD); R00-CA151673 and R01-GM138385 (to DS), R01- HL32225 (to PT), UCOP-R00RG2642 (to SD and PG), UCOP-R01RG3780 (to P.G. and D.S) and a pilot award from the Sanford Stem Cell Clinical Center at UC San Diego Health (P.G, S.D, D.S). GDK was supported through The American Association of Immunologists Intersect Fellowship Program for Computational Scientists and Immunologists. L.C.A's salary was supported in part by the VA San Diego Healthcare System. This manuscript includes data generated at the UC San Diego Institute of Genomic Medicine (IGC) using an Illumina NovaSeq 6000 that was purchased with funding from a National Institutes of Health SIG grant (#S10 OD026929).

Introduction

SARS-CoV-2, the virus responsible for COVID-19, causes widespread inflammation and injury in the lungs, giving rise to diffuse alveolar damage (DAD) (Andrea Valeria Arrossi and Farver, 2020; Damiani et al., 2021; Borczuk et al., 2020; Li et al., 2021; Roden, 2020), featuring marked infection and viral burden leading to apoptosis of alveolar pneumocytes (Hussman, 2020), along with pulmonary edema (Bratic and Larsson, 2013; Carsana et al., 2020). DAD leads to poor gas exchange and, ultimately, respiratory failure; the latter appears to be the final common mechanism of death in most patients with severe COVID-19 infection. How the virus causes so much damage remains unclear. A particular challenge is to understand the out-of-control immune reaction to the SARS-CoV-2 infection known as a cytokine storm, which has been implicated in many of the deaths from COVID-19. Although rapidly developed preclinical animal models have recapitulated some of the pathognomonic aspects of infection, for example, induction of disease, and transmission, and even viral shedding in the upper and lower respiratory tract, many failed to develop severe clinical symptoms (Lakdawala and Menachery, 2020). Thus, the need for preclinical models remains both urgent and unmet.

To address this need, several groups have attempted to develop human preclinical COVID-19 lung models, all within the last few months (Duan et al., 2020; Mulay et al., 2020; Salahudeen et al., 2020). While a head-to-head comparison of the key characteristics of each model can be found in Table 1, what is particularly noteworthy is that most of the models do not recapitulate the heterogeneous epithelial cellularity of both proximal and distal airways, that is, airway epithelia, basal cells, secretory club cells, and alveolar pneumocytes. Although induced pluripotent stem cells (iPSC)-derived AT2 cells be differentiated into proximal and distal cell types, including AT1, ciliated, and club cells (Kawakita et al., 2020; Dye et al., 2015; Huang et al., 2020), these iPSC-derived models lack propagability and cannot be reproducibly generated for biobanking; nor can they be scaled up in cost-effective ways for use in drug screens. More specifically, adult lung organoid models that can be grown in a sustainable mode and are complete with proximo-distal epithelia are yet to emerge. Besides the approaches described so far, there are a few more approaches used for modeling COVID-19: (i) 3D organoids from bronchospheres and tracheospheres have been established before (Hild and Jaffe, 2016; Rock et al., 2009; Tadokoro et al., 2016) and are now used in apical-out cultures for infection with SARS-COV-2 (Suzuki, 2020); (ii) the most common model used for drug screening is the air-liquid interphase (ALI model) in which pseudo-stratified primary bronchial or small airway epithelial cells are used to recreate the multilayered mucociliary epithelium (Mou et al., 2016; Randell et al., 2011); (iii) several groups have also generated 3D airway models from iPSCs or tissue-resident stem cells (Dye et al., 2015; Wong et al., 2012; Ghaedi et al., 2013; Konishi et al., 2016; McCauley et al., 2017; Miller et al., 2019); (iv) others have generated AT2 cells from iPSCs using closely overlapping protocols of sequential differentiation starting with definitive endoderm, anterior foregut endoderm, and distal alveolar expression (Gotoh et al., 2014; Jacob et al., 2017; Jacob et al., 2019; Yamamoto et al., 2017; Chen et al., 2017; Huang et al., 2014); and (v) finally, long-term in vitro culture conditions for pseudo-stratified airway epithelium organoids, derived from healthy and diseased adult humans suitable to assess virus infectivity (Sachs et al., 2019; van der Vaart and Clevers, 2021; Zhou et al., 2018), have been pioneered; unfortunately, these airway organoids expressed virtually no lung mesenchyme or alveolar signature. What remains unclear is if any of these models accurately recapitulate the immunopathological phenotype that is seen in the lungs in COVID-19.

Table 1
A comparison of current versus existing lung organoid models available for modeling COVID-19.
AuthorSource of stem cellsPropagabilityCell typesSARS-COV-2 infectionDemonstrated reproducibility using more than one patientCost-effective (use of conditioned media)Notes
AT1AT2ClubBasalCiliatedGoblet
Zhou et alPMID: 29891677Small pieces of normal lung tissue adjacent to the diseased tissue from patients undergoing surgical resection for clinical conditions.Long term culture > 1 yInfection with H1N1 pandemic Influenza virusProximal differentiation (PD) of human Adult Stem Cell-derived airway organoid (AO) culture. Differentiation conditions (PneumaCult-ALI medium) increase ciliated cells. Serine proteases known to be important for productive viral infection, were elevated after PD.
Sachs et alPMID: 30643021Generation of normal and tumor organoids from resected surplus lung tissue of patients with lung cancers.long term culture for over 1 yearNot clearly mentionedairway organoid (AO) expressed no mesenchyme or alveolar transcripts. Strongly enriched for bulk lung and small airway epithelial signature limited to basal, club, and ciliated cells Withdrawal of R-spondin terminated AO expansion after 3–4 passages similar to the withdrawal of FGFs
Duan et alPMID: 32839764hPSC derived lung cells and macrophagesLowSARS-CoV-2 infection mediated damage onset by macrophages.Co-culture of lung cells and macrophages. Protocol followed enables alveolar differentiation process, although described presence of almost all lung cell types.
Salahudeen et alPMID: 33238290Cells sorted from human peripheral lung tissues.Distal Lung organoid with possibility of long-term cultureFrom differentiation of AT2After diff of basal cellsInfection and presence of dsRNA and nucleocapsidNo RNA seq of infected samples to compare with COVID Differentiation to different cell types SARS CoV2 infection in apical-out organoids (not polarized monolayers). The combination of EGF and the Noggin was optimal, without any additional growth-promoting effects of either WNT3A or R-spondin
Han et alPMID: 33116299hPSC-derived lung organoidsOrganoids were generated by 50 days of differentiationSARS-CoV-2 and SARS-CoV-2-Pseudo-Entry Viruses.AT1, AT2, stromal cells, low number of pulmonary neuroendocrine cells, proliferating cells, and airway epithelial cells were reported. Mostly AT2 based ACE2 receptor was used for virus infection. High throughput screen using hPSC-derived lung organoids identified FDA-approved drug candidates, including imatinib and mycophenolic acid, as inhibitors of SARS-CoV-2 entry.
Youk et alPMID: 33142113Adult alveolar stem cells isolated from distal lung parenchymal tissues by collagenase, dispase and sortingMultiple passages upto 10 monthsFrom AT2; Lost in higher passagesIn the organoid formSingle cell transcriptomic profiling identified two clusters and type I interferon signal pathway are highly elevated at three dpi
Mulay et alPMID: 32637946doi.org/10.1101/2020.06.29.174623Alv organoids with distal lung epithelial cells with lung fibroblast cellsIn the organoid formInfection of AT2 cells trigger apoptosis that may contribute to alveolar injury. Alteration of innate immune response genes from AT2 cells
Proximal airway ALI with heterogenous cellsInfection of ciliated and goblet cells Two separate models for SARS-CoV2 infection
Huang JPMID:32979316iPSC derived AT2 cell ALI modelBulk RNA seq after day 1 and day four infection. The infection induces rapid inflammatory responses.
Abo et alPMID: 32577635doi: 10.1101/2020.06.03.132639iPSC derived basal cells as oranoids or 2D ALIiPSCs transcripts match human lung better than cancer cell lines. iPSC AT2 cells express host genes mportant for SARS-CoV-2 infection.
iPSC AT2 cells as organoids or 2D ALI
Rock et alPMID: 19625615Bronchospheres were isolated from human lung tissue.Bronchospheres derived from human lung can act as stem cells and can differentiate into other cell types.
Lamers et alPMID: 33283287Lung organoids derived from fetal Lung epithelial bud tips and differentiated ALI model.14 passagesDetected SCGB3A2(ATII/club marker)2 subjects were mentionedOrganoid model derived from fetal lung bud tip tissue consists primarily of SOX2+ SOX9+ progenitor cells. Differentiation under ALI conditions is necessary to achieve mature alveolar epithelium. ALI model was found to contain mostly ATII and ATI cells, with small basal and rare neuroendocrine populations. SFTPC + Alveolar type II like cells were most readily infected by SARS-CoV-2. The infectious virus titer is much higher (five log) compared to other established model.
Suzuki et aldoi: https://doi.org/10.1101/2020.05.25.115600Commercially available adult HBEpC cells were used to generate human bronchial organoids.In the organoid formOrganoids derived from HBEpC cells undergo differentiation process to achieve mature phenotype. Organoids are lacking distal epithelial cell types SARS-CoV-2 infection was performed on organoids and only the basolateral region came in to contact with the virus. Treatment with a TMPRSS2 inhibitor prior to infection demonstrated a reduction in infectivity.
Tiwari et alPMID: 33631122Differentiated human iPSCs into lung organoids.80 daysIn the organoid formOrganoids originated from iPSC cells and have proximal and distal epithelial cells. Infected organoids with SARS-CoV-2 and pseudovirus. SARS-CoV-2 pseudovirus entry was blocked by viral entry inhibitors.
Tindle et al [Current study]Deep lung tissue sections surgically obtained from patients undergoing lobe resections for lung cancers.RNA Seq and cross-validation of COVID-19 model. Single model with all the cells types and infection of SARS-CoV2 in the 2D form with Apical accessibility that close to physiologic state.
  1. ACE2: angiotensin-converting enzyme II; ALI: air-liquid interphase; TMPRSS2: transmembrane serine protease 2.

  2. Blue color cells denote the presence of the features.

  3. Red color cells denote the absence of the features.

  4. Grey color cells denote information not found.

We present a rigorous transdisciplinary approach that systematically assesses an adult lung organoid model that is propagable, personalized, and complete with both proximal airway and distal alveolar cell types against existing models that are incomplete, and we cross-validate them all against COVID-19 patient-derived respiratory samples. Findings surprisingly show that cellular crosstalk between both proximal and distal components is necessary to emulate how SARS-CoV-2 causes diffuse alveolar pneumocyte damage; the proximal airway mounts a sustained viral infection, but it is the distal alveolar pneumocytes that mount the overzealous host response that has been implicated in a fatal disease.

Results

A rationalized approach for creating and validating acute lung injury in COVID-19

To determine which cell types in the lungs might be most readily infected, we began by analyzing a human lung single-cell sequencing dataset (GSE132914) for the levels of expression of angiotensin-converting enzyme II (ACE2) and transmembrane serine protease 2 (TMPRSS2), the two receptors that have been shown to be the primary sites of entry for the SARS-CoV-2 (Hoffmann et al., 2020). The dataset was queried with widely accepted markers of all the major cell types (see Table 2). Alveolar epithelial type 2 (AT2), ciliated and club cells emerged as the cells with the highest expression of both receptors (Figure 1A, Figure 1—figure supplement 1A). These observations are consistent with published studies demonstrating that ACE2 is indeed expressed highest in AT2 and ciliated cells (Mulay et al., 2020; Zhao et al., 2020; Jia et al., 2005). In a cohort of deceased COVID-19 patients, we observed by H&E (Figure 1—figure supplement 1B) that gas-exchanging flattened AT1 pneumocytes are virtually replaced by cuboidal cells that were subsequently confirmed to be AT2-like cells via immunofluorescent staining with the AT2-specific marker, surfactant protein C (SFTPC; Figure 1B, upper panel, Figure 1—figure supplement 1C, top). We also confirmed that club cells express ACE2 (Figure 1—figure supplement 1C, bottom), underscoring the importance of preserving these cells in any ideal lung model of COVID-19. When we analyzed the lungs of deceased COVID-19 patients, the presence of SARS-COV-2 in alveolar pneumocytes was also confirmed, as determined by the colocalization of viral nucleocapsid protein with SFTPC (Figure 1B, lower panel, Figure 1—figure supplement 1D). Immunohistochemistry studies further showed the presence of SARS-COV-2 virus in alveolar pneumocytes and in alveolar immune cells (Figure 1—figure supplement 1E). These findings are consistent with the gathering consensus that alveolar pneumocytes support the interaction between the epithelial cells and inflammatory cells recruited to the lung; via mechanisms that remain unclear, they are generally believed to contribute to the development of acute lung injury and acute respiratory distress syndrome (ARDS), the severe hypoxemic respiratory failure during COVID-19 (Hou et al., 2020; Spagnolo et al., 2020). Because prior work has demonstrated that SARS-CoV-2 infectivity in patient-derived airway cells is highest in the proximal airway epithelium compared to the distal alveolar pneumocytes (AT1 and AT2) (Hou et al., 2020), and yet, it is the AT2 pneumocytes that harbor the virus, and the AT1 pneumocytes that are ultimately destroyed during DAD, we hypothesized that both proximal airway and distal (alveolar pneumocyte) components might play distinct roles in the respiratory system to mount the so-called viral infectivity and host immune response phases of the clinical symptoms observed in COVID-19 (Chen and Li, 2020).

Table 2
Markers used to identify various cell types in the lung.
Cell typeMarkers
AT1AQP5*$, PDPN*$$, Carboxypeptidase M, CAV-1, CAV-2, HTI56, HOPX, P2R × 4*$$, Na+/K + ATPase$, TIMP3*++, SEMA3F PDPN* AQP5* P2R × 4* TIMP3* SERPINE*
AT2ABCA3*$$, CC10 (SCGB1A1*)+, CD44v6, Cx32, gp600++, ICAM-1++, KL-6, LAMP3*$$, MUC1, SFTPA1*$$, SFTPB*$, SFTPC*+, SFTPD*, SERPINE1
ClubCC10 (SCGB1A1*)+, CYP2F2*, ITAG6*$$, SCGB3A2*$$, SFTPA1*$$, SFTPB*$, SFTPD*
GobletCDX-2*, MUC5AC*, MUC5B*, TFF3*, UEA1+
CiliatedACT (ACTG2*)$, BTub4 (TUBB4A*), FOXA3*++, FOXJ1*, SNTN*
BasalCD44v6 (CD44*), ITGA6*$$, KRT5*$, KRT13*, KRT14*, p63 (CKAP4*), p75 (NGFR*)$$
Generic Lung LineageCx43 (GJA1*), TTF-1 (TTF1*; Greatest in AT2 & Club), EpCAM (EPCAM*)
  1. *Markers used for single-cell gating (Figure 1A).

  2. $ denotes markers used in this work for Immunofluorescence (IF).

  3. $$ denotes markers used in this work for qPCR.

  4. + denotes markers used in both IF and qPCR.

  5. ++ denotes obscure markers (Not a lot of research relative to lung).

Figure 1 with 1 supplement see all
A rationalized approach to building and validating human preclinical models of COVID-19.

A) Whisker plots display relative levels of angiotensin-converting enzyme II (ACE2) expression in various cell types in the normal human lung. The cell types were annotated within a publicly available single-cell sequencing dataset (GSE132914) using genes listed in Table 1. p-values were analyzed by one-way ANOVA and Tukey’s post hoc test. (B) Formalin-fixed paraffin-embedded sections of the human lung from normal and deceased COVID-19 patients were stained for SFTPC, alone or in combination with nucleocapsid protein and analyzed by confocal immunofluorescence. Representative images are shown. Scale bar = 20 µm. (C) Schematic showing key steps generating an adult stem cell-derived, propagable, lung organoid model, complete with proximal and distal airway components for modeling COVID-19-in-a-dish. See Materials and methods for details regarding culture conditions. (D) A transcriptome-based approach is used for cross-validation of in vitro lung models of SARS-CoV-2 infection (left) versus the human disease, COVID-19 (right), looking for a match in gene expression signatures.

Because no existing lung model provides such proximodistal cellular representation (Table 1), and hence, may not recapitulate with accuracy the clinical phases of COVID-19, we first sought to develop a lung model that is complete with both proximal and distal airway epithelia using adult stem cells that were isolated from deep lung biopsies (i.e., sufficient to reach the bronchial tree). Lung organoids were generated using the work flow outlined in Figure 1C and a detailed protocol that had key modifications from previously published (Sachs et al., 2019; Zhou et al., 2018) methodologies (see Materials and methods). Organoids grown in 3D cultures were subsequently dissociated into single cells to create 2D monolayers (either maintained submerged in media or used in ALI model) for SARS-CoV-2 infection, followed by RNA seq analysis. Primary airway epithelial cells and hiPSC-derived alveolar type II (AT2) pneumocytes were used as additional models (Figure 1D, left panel). Each of these transcriptomic datasets was subsequently used to cross-validate our ex vivo lung models of SARS-CoV-2 infection with the human COVID-19 autopsy lung specimens (Figure 1D, right panel) to objectively vet each model for their ability to accurately recapitulate the gene expression signatures in the patient-derived lungs.

Creation of a lung organoid model, complete with both proximal and distal airway epithelia

Three lung organoid lines were developed from deep lung biopsies obtained from the normal regions of lung lobes surgically resected for lung cancer; both genders, smokers and non-smokers, were represented (Figure 2—figure supplement 1A; Table 3). Three different types of media were compared (Figure 2—figure supplement 1B); the composition of these media was inspired either by their ability to support adult-stem cell-derived mixed epithelial cellularity in other organs (like the gastrointestinal [GI] tract [Miyoshi and Stappenbeck, 2013; Sato et al., 2009; Sayed et al., 2020c]) or rationalized based on published growth conditions for proximal and distal airway components (Gotoh et al., 2014; Sachs et al., 2019; van der Vaart and Clevers, 2021). A growth condition that included conditioned media from L-WRN cells that express Wnt3, R-spondin, and Noggin, supplemented with recombinant growth factors, which we named as ‘lung organoid expansion media,’ emerged as superior compared to alveolosphere media-I and II (Jacob et al., 2019; Yamamoto et al., 2017) (details in Materials and methods), based on its ability to consistently and reproducibly support the best morphology and growth characteristics across multiple attempts to isolate organoids from lung tissue samples. Three adult lung organoid lines (ALO1-3) were developed using the expansion media, monitored for their growth characteristics by brightfield microscopy and cultured with similar phenotypes until P10 and beyond (Figure 2—figure supplement 1C and D). The 3D morphology of the lung organoid was also assessed by H&E staining of slices cut from formalin-fixed paraffin-embedded (FFPE) cell blocks of HistoGel-emb`edded ALO1-3 (Figure 2—figure supplement 1E).

Table 3
Characteristics of patients enrolled into this study for obtaining lung tissues to serve as source of stem cells to generate lung organoids.
NameDate of surgeryAgeSexSmoking historyReason for surgeryHistology
ALO14/17/202064MaleCurrent, chronic smokerPacks/day: 0.50Years: 53Pack years: 26.5Lung carcinomaInvasive squamous cell carcinoma, non-keratinizing
ALO24/17/202059MaleNon-smokerLung carcinomaInvasive adenocarcinoma
ALO37/7/202046FemaleNon-smokerLeft lower lobe noduleInvasive adenocarcinoma

To determine if all the six major lung epithelial cells (illustrated in Figure 2A) are present in the organoids, we analyzed various cell-type markers by qRT-PCR (Figure 2B–H and Figure 2—figure supplement 2A-H). All three ALO lines had a comparable level of AT2 cell surfactant markers (compared against hiPSC-derived AT2 cells as positive control) and a significant amount of AT1, as determined using the marker AQP5. ALOs also contained basal cells (as determined by the marker ITGA6, p75/NGFR, TP63), ciliated cells (as determined by the marker FOXJ1), and club cells (as determined by the marker SCGB1A1). As expected, the primary normal human bronchial epithelial cells (NHBE) had significantly higher expression of basal cell markers than the ALO lines (hence, served as a positive control), but they lacked stemness and club cells (hence, served as a negative control).

Figure 2 with 5 supplements see all
Adult stem cell-derived lung organoids are propagatable models with both proximal and distal airway components.

(A) Schematic lists the various markers used here for qPCR and immunofluorescence to confirm the presence of all cell types in the 3D lung organoids here and in 2D monolayers later (in Figure 3). (B–H) Bar graphs display the relative abundance of various cell-type markers (normalized to 18S) in adult lung organoids (ALO), compared to the airway ( normal human bronchial epithelial cell [NHBE]) and/or alveolar (AT2) control cells, as appropriate. p-values were analyzed by one-way ANOVA. Error bars denote SEM; n = 3–6 datasets from three independent ALOs and representing early and late passages. See also Figure 2—figure supplement 2 for individual ALOs. (I, J). H&E-stained cell blocks were prepared using HistoGel (I). Slides were stained for the indicated markers and visualized by confocal immunofluorescence microscopy. Representative images are shown in (J). Scale bar = 50 µm. (K) 3D organoids grown in 8-well chamber slides were fixed, immunostained, and visualized by confocal microscopy as in (J). Scale bar = 50 µm. See also Figure 2—figure supplement 2. Top row (ACE2/KRT5-stained organoids) displays the single and merged panels as max projections of z-stacks (top) and a single optical section (bottom) of a selected area. For the remaining rows, the single (red/green) channel images are max projections of z-stacks; however, merged panels are optical sections to visualize the centers of the organoids. All immunofluorescence images showcased in this figure were obtained from ALO lines within passage #3–6. See also Figure 2—figure supplements 35 for additional evidence of mixed cellularity of ALO models, their similarity to lung tissue of origin, and stability of cellular composition during early (#1–8) and late (#8–15) passages, as determined by qPCR and flow cytometry.

The presence of all cell types was also confirmed by assessing protein expression of various cell types within organoids grown in 3D cultures. Two different approaches were used—(i) slices cut from FFPE cell blocks of HistoGel-embedded ALO lines (Figure 2I and J) or (ii) ALO lines grown in 8-well chamber slides were fixed in Matrigel (Figure 2K), stained, and assessed by confocal microscopy. Such staining not only confirmed the presence of more than one cell type (i.e., mixed cellularity) of proximal (basal-KRT5) and distal (AT1/AT2 markers) within the same ALO line, but also, in some instances, demonstrated the presence of mixed cellularity within the same 3D structure. For example, AT2 and basal cells, marked by SFTPB and KRT5, respectively, were found in the same 3D structure (Figure 2J, interrupted curved lines). Similarly, ciliated cells and goblet cells stained by Ac-Tub and Muc5AC, respectively, were found to coexist within the same structure (Figure 2J, interrupted box; Figure 2K, arrow). Intriguingly, we also detected 3D structures that co-stained for CC10 and SFTPC (Figure 2J, bottom panel) indicative of mixed populations of club and AT2 cells. Besides the organoids with heterogeneous makeup, each ALO line also showed homotypic organoid structures that were relatively enriched in one cell type (Figure 2J, arrowheads pointing to two adjacent structures that are either KRT5- or SFTPB-positive). Regardless of their homotypic or heterotypic cellular organization into 3D structures, the presence of mixed cellularity was documented in all three ALO lines (see multiple additional examples in Figure 2—figure supplement 2I). It is noteworthy that the coexistence of proximal and distal epithelial cells in lung organoids has been achieved in one another instance prior; Lamers et al. showed such mixed cellular composition in fetal lung bud tip-derived organoids Lamers et al., 2021. However, their model lacked ciliated and goblet cells (Lamers et al., 2021), something that we could readily detect in our 3D organoids.

Finally, using qRT-PCR of various cell-type markers as a measure, we confirmed that the ALO models overall recapitulated the cell-type composition in the adult lung tissues from which they were derived (Figure 2—figure supplement 3) and retained such composition in later passages without significant notable changes in any particular cell type (Figure 2—figure supplement 4). The mixed proximal and distal cellular composition of the ALO models and their degree of stability during in vitro culture was also confirmed by flow cytometry (Figure 2—figure supplement 5).

Organoid cellularity resembles tissue sources in 3D cultures but differentiates in 2D cultures

To model respiratory infections such as COVID-19, it is necessary for pathogens to be able to access the apical surface. It is possible to microinject into the lumens of 3D organoids, as done previously with pathogens in the case of gut organoids (Engevik et al., 2015; Forbester et al., 2015; Leslie et al., 2015; Williamson et al., 2018), or FITC-dextran in the case of lung organoids (Porotto et al., 2019), or carry out infection in apical-out 3D lung organoids with basal cells (Salahudeen et al., 2020). However, the majority of the researchers have gained apical access by dissociating 3D organoids into single cells and plating them as 2D-monolayers (Duan et al., 2020; Mulay et al., 2020; Huang et al., 2020; Sachs et al., 2019; Zhou et al., 2018; Han et al., 2020a; Han et al., 2020b. As in any epithelium, the differentiation of airway epithelial cells relies upon dimensionality (apicobasal polarity); because the loss of dimensionality can have a major impact on cellular proportions and impact disease-modeling in unpredictable ways, we assessed the impact of the 3D-to-2D conversion on cellularity by RNA seq analyses. Two commonly encountered methods of growth in 2D monolayers were tested: (i) monolayers polarized on trans-well inserts but submerged in growth media (Figure 3A and Figure 3—figure supplement 1A-D) and (ii) monolayers were grown at the air-liquid interface (popularly known as the ‘ALI model’; Prytherch et al., 2011; Dvorak et al., 2011) for 21 days to differentiate into the mucociliary epithelium (Figure 3A and Figure 3—figure supplement 1E-G). The submerged 2D monolayers had several regions of organized vacuolated-appearing spots (Figure 3—figure supplement 1D, arrow), presumably due to morphogenesis and cellular organization even in 2D. Consistent with this morphological appearance, the epithelial barrier formed in the submerged condition was leakier, as determined by relatively lower transepithelial electrical resistance (TEER; Figure 3—figure supplement 1B) and the flux of FITC-dextran from apical to basolateral chambers (Figure 3—figure supplement 1C), and corroborated by morphological assessment by confocal immunofluorescence of localization of occludin, a bona fide TJ marker. We chose occludin because it is a shared and constant marker throughout the airway that stabilizes claudins and regulates their turnover McGowan, 2014 and plays an important role in maintaining the integrity of the lung epithelial barrier Liu et al., 2014. Junction-localized occludin was patchy in the monolayer, despite the fact that the monolayer was otherwise intact, as determined by phalloidin staining (Figure 3—figure supplement 1H and I). Our finding that ALO 3D organoids differentiating into monolayers in submerged cultures (where alveolar differentiation and cell flattening happens dynamically as progenitor cells give rise to AT1/2 cells) are leaky is in keeping with prior work demonstrating that the TJs are rapidly remodeled as alveolar cells mature Schlingmann et al., 2015; Yang et al., 2016. By contrast, and as expected Rayner et al., 2019, the ALI monolayers formed a more effective epithelial barrier, as determined by TEER (Figure 3—figure supplement 1F), and appeared to be progressively hazier with time after air-lift, likely due to the accumulation of secreted mucin (Figure 3—figure supplement 1G).

Figure 3 with 3 supplements see all
Monolayers derived from lung organoids differentiate into proximal and distal airway components.

(A, B) Samples collected at various steps of lung organoid isolation and expansion in culture, and from the two types of monolayers prepared using the lung organoids were analyzed by bulk RNA seq and the datasets were compared for % cellular composition using the deconvolution method, CYBERSORTx. Schematic in (A) shows the workflow steps, and bar plots in (B) show the relative proportion of various lung cell types. (C, D) hiPSC-derived AT2 cells and alveolospheres (C) were plated as monolayers and analyzed by RNA seq. Bar plots in (D) show % cellular composition. (E, F) Submerged adult lung organoids (ALO) monolayers in transwells (E) or monolayers were grown as air-liquid interphase (ALI) models (F) were fixed and stained for the indicated markers and visualized by confocal immunofluorescence microscopy. The representative max projected z-stack images (left) and the corresponding orthogonal images (right) are displayed. Arrows in (E) indicate AT2 cells; arrowheads in (E) indicate club cells; asterisk in (F) indicates bundles of cilia standing perpendicular to the plane of the ALI monolayers; arrowheads in (F) indicate bundles of cilia running parallel to the plane of the ALI monolayers. Scale bar = 20 µm. (G) Monolayers of ALO1-3 were challenged with SARS-CoV-2 for indicated time points prior to fixation and staining for KRT5, SARS-COV2 viral nucleocapsid protein and DAPI and visualized by confocal microscopy. A montage of representative images are shown, displaying reticulovesicular network patterns and various cytopathic effects. Scale bar = 15 µm. (H) Monolayers of ALO, hiPSC-derived AT2 cells, and other alternative models (see Figure 3—figure supplements 12) were infected or not with SARS-CoV-2 and analyzed for infectivity by qPCR (targeted amplification of viral envelope, E gene). See also Figure 3—figure supplement 3B, C for comparison of the degree of peak viral amplification across various models. (I) ALO monolayers pretreated for 4 hr with either vehicle (DMSO) control or EIDD-parent (NHC) or its metabolite EIDD-2801/MK-4482 were infected with SARS-CoV-2 and assessed at 48 hpi for infectivity as in (H). Line graphs display the relative expression of E gene. Error bars display SEM. p value **<0.01; ***<0.001.

RNA seq datasets were analyzed using the same set of cell markers, as we used in Figure 1A (listed in Table 2). Consistent with our morphological, gene expression, and FACS-based studies showcased earlier (Figure 2 and Figure 2—figure supplements 25), cell-type deconvolution of our transcriptomic dataset using CIBERSORTx (https://cibersortx.stanford.edu/runcibersortx.php) confirmed that cellular composition in the human lung tissues was reflected in the 3D ALO models and that such composition was also relatively well-preserved over several passages (Figure 3B, left); both showed a mixed population of simulated alveolar, basal, club, ciliated, and goblet cells. When 3D organoids were dissociated and plated as 2D monolayers on transwells, the AT2 signatures were virtually abolished with a concomitant and prominent emergence of AT1 signatures, suggesting that growth in 2D monolayers favors differentiation of AT2 cells into AT1 cells Shami and Evans, 2015 (Figure 3B, middle). A compensatory reduction in proportion was also observed for the club, goblet, and ciliated cells. The same organoids, when grown in long-term 2D culture conditions in the ALI model, showed a strikingly opposite pattern; alveolar signatures were almost entirely replaced by a concomitant increase in ciliated and goblet cells (Figure 3B, right). These findings are consistent with the well-established notion that ALI conditions favor growth as pseudo-stratified mucociliary epithelium Prytherch et al., 2011; Dvorak et al., 2011. As an alternative model for use as monolayers for viral infection, we developed hiPSC-derived AT2 cells and alveolospheres (Figure 3C), using established protocols Huang et al., 2020. Because they were grown in the presence of CHIR99021 (an aminopyrimidine derivate that is a selective and potent Wnt agonist) Jacob et al., 2019; Yamamoto et al., 2017; Abdelwahab et al., 2019, which probably inhibits the AT2→AT1 differentiation, these monolayers were enriched for AT2 and devoid of AT1 cells (Figure 3D).

The multicellularity of lung organoid monolayers was also confirmed by immunofluorescence staining and confocal microscopy of the submerged and ALI monolayers, followed by the visualization of cell markers in either max-projected z-stacks (Figure 3E, left) or orthogonal views of the same (Figure 3E, right). As expected, markers for the same cell type (i.e., SFTPB and SFTPC, both AT2 markers) colocalize, but markers for different cell types do not. Submerged monolayers showed the prominent presence of both AT1 (AQP5-positive) and AT2 cells. Compared to the 3D organoids, the 2D organoid cultures, especially the ALI model, showed a significant increase in ciliated structures, as determined by acetylated tubulin (compare Ac Tub-stained panels in Figure 2J and K with Figure 3E and F). The observed progressive prominence of ciliary structures from 3D to 2D models is in keeping with the fact that 3D ALOs that are yet to form lumen represent the least differentiated state, whereas 2D submerged monolayers are intermediate and the 2D ALI monolayers are maximally differentiated; differentiation is known to establish apicobasal polarity, which is essential for the emergence of mature cilia on the apical surface. This increase in ciliated epithelium was associated with a concomitant decrease in KRT5-stained basal cells (Figure 3F). Such loss of the basal cell marker KRT5 between submerged monolayers and the ALI model can be attributed to and the expected conversion of basal cells to other cell types (i.e., ciliated cells) Gras et al., 2017; Khelloufi et al., 2018. The presence of AT2 cells, scattered amidst the ciliated cells in these ALI monolayers, was confirmed by co-staining them for SFTPC and Ac-Tub (Figure 3—figure supplement 1J).

Finally, we sought to confirm that the epithelial barrier that is formed by the submerged monolayers derived from ALO is responsive to infections. To this end, we simulated infection by challenging ALO monolayers with LPS. Compared to unchallenged controls, the integrity of the barrier was impaired by LPS, as indicated by a significant drop in the TEER (Figure 3—figure supplement 1K and L), which is in keeping with the known disruptive role of LPS on the respiratory epithelium Kalsi et al., 2020.

Taken together, the immunofluorescence images are in agreement with the RNA seq dataset; both demonstrate that the short-term submerged monolayer favors distal differentiation (AT2→AT1), whereas the 21-day ALI model favors proximal mucociliary differentiation. It is noteworthy that these distinct differentiation phenotypes originated from the same 3D organoids despite the seeding of cells in the same basic media composition (i.e., PneumaCult) prior to switching over to an ALI maintenance media for the prolonged growth at ALI; the latter is a well-described methodology that promotes differentiation into ciliated and goblet cells Rayner et al., 2019.

Differentiated 2D monolayers show that SARS-CoV-2 infectivity is higher in proximal than distal epithelia

Because the lung organoids with complete proximodistal cellularity could be differentiated into either distal-predominant monolayers in submerged short-term cultures or proximal-predominant monolayers in long-term ALI cultures, this provided us with an opportunity to model the respiratory tract and assess the impact of the virus along the entire proximal-to-distal gradient. We first asked if ALO monolayers are permissive to SARS-CoV-2 entry and replication and support sustained viral infection. Confocal imaging of infected ALO monolayers with anti-SARS-COV-2 nucleocapsid protein antibody showed that submerged ALO monolayers did indeed show progressive changes during the 48–72 hr window after infection (Figure 3G): by 48 hpi, we observed the formation of ‘reticulovesicular patterns’ that are indicative of viral replication within modified host endoplasmic reticulum (Knoops et al., 2008; Figure 3G, left), and by 72 hpi we observed focal cytopathic effect (CPE) (Kaye, 2006) such as cell-rounding, detachment, and bursting of virions (Figure 3G, right, Figure 3—figure supplement 3A).

We next asked how viral infectivity varies in the various lung models. Because multiple groups have shown the importance of the ciliated airway cells for infectivity (i.e., viral entry, replication, and apical release [Hou et al., 2020; Milewska et al., 2020; Zhu et al., 2020; Hui et al., 2020]), as positive controls, we infected monolayers of human airway epithelia (see the legend, Figure 3—figure supplement 2A-D. AT2 cells, which express high levels of viral entry receptors ACE2 and TMPRSS2 (Figure 1A, Figure 1—figure supplement 1A), have been shown to be proficient in the viral entry but are least amenable to sustained viral release and infectivity (Hou et al., 2020; Hui et al., 2020). To this end, we infected monolayers of hiPSC-derived homogeneous cultures of AT2 cells as secondary controls (see the legend, Figure 3—figure supplement 2E-G). Infection was carried out using the Washington strain of SARS-CoV-2, USA-WA1/2020 (BEI Resources NR-52281 Rogers et al., 2020). As expected, the 2D lung monolayers we generated, both the submerged and the ALI models, were readily infected with SARS-CoV-2 (Figure 3—figure supplement 3B), as determined by the presence of the viral envelope gene (E gene; Figure 3H); however, the kinetics of viral amplification differed. When expressed as levels of E gene normalized to the peak values in each model (Figure 3H), the kinetics of the ALI monolayer model mirrored that of the primary airway epithelial monolayers; both showed slow beginning (0–48 hpi) followed by an exponential increase in E gene levels from 48 to 72 hpi. The submerged monolayer model showed sustained viral infection during the 48–72 hpi window (Figure 3—figure supplement 3B, left). In the case of AT2 cells, the 48–72 hpi window was notably missing in monolayers of hiPSC-derived AT2 cells (Figure 3H and Figure 3—figure supplement 3B, right). When we specifically analyzed the kinetics of viral E gene expression during the late phase (48–72 hpi window), we found that proximal airway models (human bronchial airway epi [HBEpC]) showed high levels of sustained infectivity than distal models (human small airway epi [HSAEpC] and AT2) to viral replication (Figure 3—figure supplement 3C); the ALO monolayers showed intermediate sustained infectivity (albeit with variability). All models showed extensive cell death and detachment by 96 hr and, hence, were not analyzed. Finally, using the E gene as a readout, we asked if ALO models could be used as platforms for preclinical drug screens. As a proof of concept, we tested the efficacy of nucleoside analog N4-hydroxycytidine (NHC; EIDD-parent) and its derivative pro-drug, EIDD-2801; both have been shown to inhibit viral replication, in vitro and in SARS-CoV-2-challenged ferrets (Cox et al., 2021; Sheahan et al., 2020). ALO monolayers plated in 384-wells were pretreated for 4 hr with the compounds or DMSO (control) prior to infection and assessed at 48 hpi for the abundance of E gene in the monolayers. Both compounds effectively reduced the viral titer in a dose-dependent manner (Figure 3I), and the pro-drug derivative showed a better efficacy, as shown previously.

Taken together, these findings show that sustained viral infectivity is best simulated in monolayers that resemble the proximal mucociliary epithelium, that is, 2D monolayers of lung organoids grown as ALI models and the primary airway epithelia. Because prior studies conducted in patient-derived airway cells (Hou et al., 2020) mirror what we see in our monolayers, we conclude that proximal airway cells within our mixed-cellular model appear to be sufficient to model viral infectivity in COVID-19. Findings also validate optimized protocols for the adaptation of ALO monolayers in miniaturized 384-well formats for use in high-throughput drug screens.

Differentiated 2D monolayers show that host immune response is higher in distal than proximal epithelia

Next, we asked if the newly generated lung models accurately recapitulate the host immune response in COVID-19. To this end, we analyzed the infected ALO monolayers (both the submerged and ALI variants) as well as the airway epithelial (HSAEpC) and AT2 monolayers by RNA seq and compared them all against the transcriptome profile of lungs from deceased COVID-19 patients. We did this analysis in two steps of reciprocal comparisons: (i) The actual human disease-derived gene signature was assessed for its ability to distinguish infected from uninfected disease models (in Figure 4). (ii) The ALO model-derived gene signature was assessed for its ability to distinguish healthy from diseased patient samples (in Figure 5). A publicly available dataset (GSE151764) Nienhold et al., 2020, comprising lung transcriptomes from victims deceased either due to noninfectious causes (controls) or due to COVID-19, was first analyzed for differentially expressed genes (DEGs; Figure 4A and B). This cohort was chosen as a test cohort over others because it was the largest one available at the time of this study with appropriate postmortem control samples. DEGs showed an immunophenotype that was consistent with what is expected in viral infections (Figure 4C, Table 4, and Figure 4—figure supplement 1) and showed overrepresentation of pathways such as interferon, immune, and cytokine signaling (Figure 4D, Table 5, and Figure 4—figure supplement 2). DEG signatures and reactome pathways that were enriched in the test cohort were fairly representative of the host immune response observed in patient-derived respiratory samples in multiple other validation cohorts; the signature derived from the test cohort could consistently classify control (normal) samples from COVID-19-samples (receiver operating characteristics area under the curve [ROC AUC] 0.89–1.00 across the board; Figure 4E). The most notable finding is that the patient-derived signature was able to perfectly classify the EpCAM-sorted epithelial fractions from the bronchoalveolar lavage fluids of infected and healthy subjects (ROC AUC 1.00; GSE145926-Epithelium Liao et al., 2020), suggesting that the respiratory epithelium is a major site where the host immune response is detected in COVID-19. When compared to existing organoid models of COVID-19, we found that the patient-derived COVID-19-lung signature was able to perfectly classify infected vs. uninfected late passages (>50) of hiPSC-derived AT1/2 monolayers (GSE155241) Han et al., 2020a and infected vs. uninfected liver and pancreatic organoids (Figure 4F). The COVID-19-lung signatures failed to classify commonly used respiratory models, for example, A549 cells and bronchial organoids, as well as intestinal organoids (Figure 4F). A similar analysis on our own lung models revealed that the COVID-19-lung signature was induced in submerged monolayers with distal-predominant AT2→AT1 differentiation, but not in the proximal-predominant ALI model (ROC AUC 1.00 and 0.50, respectively; Figure 4G). The ALI model and the small airway epithelia, both models that mimic the airway epithelia (and lack alveolar pneumocytes; see Figure 3B), failed to mount the patient-derived immune signatures (Figure 4H, left). These findings suggested that the presence of alveolar pneumocytes is critical for emulating host response. To our surprise, induction of the COVID-19-lung signature also failed in hiPSC-derived AT2 monolayers (Figure 4H, right), indicating that AT2 cells are unlikely to be the source of such host response. These findings indicate that both proximal airway and AT2 cells, when alone, are insufficient to induce the host immune response that is encountered in the lungs of COVID-19 patient.

Figure 4 with 2 supplements see all
Gene expression patterns in the lungs of patients with COVID-19 (actual disease) are recapitulated in lung organoid monolayers infected with SARS-CoV-2 (disease model).

(A–C) Publicly available RNA seq datasets (GSE151764) of lung autopsies from patients who were deceased due to COVID-19 or noninfectious causes (healthy normal control) were analyzed for differential expression of genes (B). The differentially expressed genes (DEGs) are displayed as a heatmap labeled with selected genes in (C). See also Figure 4—figure supplement 1 for the same heatmap with all genes labeled. (D) Reactome-pathway analysis shows the major pathways up- or downregulated in the COVID-19-afflicted lungs. See also Figure 4—figure supplement 2 for visualization as hierarchical ReacFoam. (E) Bar plots display the ability of the DEGs in the test cohort (GSE151764) to classify human COVID-19 respiratory samples from four other independent cohorts. (F) Bar plots display the ability of the DEGs in the test cohort (GSE151764) to classify published in vitro models for SARS-CoV-2 infection where RNA seq datasets were either generated in this work or publicly available. (G, H) Bar (top) and violin (bottom) plots compare the relative accuracy of disease modeling in four in vitro models used in the current work, as determined by the induction of COVID-19 lung signatures in each model. (G) Monolayer (left) and air-liquid interphase (ALI) models (right) prepared using adult lung organoids (ALOs). (H) Primary human small airway epithelium (left) and hiPSC-derived AT2 monolayers (right). Table 6 lists details regarding the patient cohorts/tissue or cell types represented in each transcriptomic dataset.

Figure 5 with 3 supplements see all
Genes and pathways induced in the SARS-CoV-2-infected lung organoid monolayers (disease model) are induced also in the lungs of COVID-19 patients (actual disease).

(A–C) Adult lung organoid monolayers infected or not with SARS-CoV-2 were analyzed by RNA seq and differential expression analysis. Differentially expressed genes (DEGs; B) are displayed as a heatmap in (C). While only selected genes are labeled in panel (C) (which represent overlapping DEGs between our organoid model and publicly available COVID-19 lung dataset, GSE151764), the same heatmap is presented in Figure 5—figure supplement 1 with all genes labeled. (D) Reactome-pathway analysis shows the major pathways upregulated in SARS-CoV-2-infected lung organoid monolayers. See also Figure 5—figure supplement 2 for visualization as hierarchical ReacFoam. (E) A Venn diagram showing overlaps in DEGs between model (current work; B) and disease (COVID-19 lung dataset, GSE151764; Figure 4). (F) Bar plots display the ability of the DEGs in infected lung monolayers to classify human normal vs. COVID-19 respiratory samples from five independent cohorts. (G–I) Bar (top) and violin (bottom) plots compare the accuracy of disease modeling in three publicly available human lung datasets, as determined by the significant induction of the DEGs that were identified in the SARS-CoV-2-challenged monolayers. See also Table 6, which enlists details regarding the patient cohorts/tissue or cell types represented in each transcriptomic dataset.

Table 4
Upregulated genes and pathways: healthy vs COVID-19 lung (GSE151764).
Genes
BRCA2XAGE1BCDK1SNAI2CXCL11
CYBBCCR5GBP1IFITM1IFI27
KRT5CCR2HLA-GGZMBIFI35
C1QBALOX15BIDO1CD163TDO2
FCGR1ACMKLR1ISG20CD38GZMA
IL10MX1LAG3BST2OAS3
IL6TNFRSF17MAD2L1BUB1POU2AF1
CD44CCR1CXCL9CCL20CXCL13
CD276CXCR3MKI67CCNB2GNLY
DMBT1SLAMF8IFIT2TNFSF18IFIT3
DDX58IL21IFIT1ISG15TOP2A
TNFAIP8FOXM1CXCL10CDKN3LILRB1
LAMP3IFIH1IRF4C1QAHERC6
KIAA0101IFI6PSMB9OAS1TNFSF13B
MELKPDCD1LG2CCL18OAS2IFI44L
PathwaysSTAT1
Namep-valueFDR
Interferon signaling1.11E-161.11E-14
Interferon alpha/beta signaling1.11E-161.11E-14
Cytokine signaling in immune system1.11E-161.11E-14
Immune ssystem1.11E-161.11E-14
Interleukin-10 signaling9.85E-137.88E-11
Interferon gamma signaling9.26E-126.11E-10
Chemokine receptors bind chemokines1.08E-106.17E-09
Signaling by interleukins6.81E-093.41E-07
Insulin-like growth factor-2 mRNA binding proteins (IGF2BPs/IMPs/VICKZs) bind RNA1.27E-070.000005581122619
Antiviral mechanism by IFN-stimulated genes0.0000019330583490.00007732233398
CD163 mediating an anti-inflammatory response0.0000077986761690.0002807523421
OAS antiviral response0.000010208709970.0003368874291
Peptide ligand-binding receptors0.000017140576870.0005142173062
Interleukin-4 and Interleukin-13 signaling0.00010149486610.002841856252
Cyclin A/B1/B2-associated events during G2/M transition0.00018878164650.00490832281
G0 and early G10.00036071218380.009017804596
Interleukin-6 signaling0.00046566784440.01071036042
ISG15 antiviral mechanism0.00083139919880.01745938317
Regulation of APC/C activators between G1/S and early anaphase0.00083139919880.01745938317
Polo-like kinase-mediated events0.0011105065130.02221013026
APC/C-mediated degradation of cell cycle proteins0.0013081035810.02354586446
Regulation of mitotic cell cycle0.0013081035810.02354586446
G2/M DNA replication checkpoint0.0017501563320.02975265764
Class A/1 (rhodopsin-like receptors)0.0023550630450.03537666782
Interleukin-6 family signaling0.0023584445210.03537666782
TNFs bind their physiological receptors0.0023584445210.03537666782
Table 5
Downregulated genes and pathways: healthy vs COVID-19 lung (GSE151764).
Genes
CX3CR1JAMLKLRB1GRAP2CD226
ARG1CX3CR1LY9MMP9CD160
MPOHLA-DQB2CCL17RORCFOXP3
IL2TNFRSF9CCL22CCR4CRTAM
BCL2CXCR5TCF7IRS1CCR6
CA4CD1CCXCR4ITKCEACAM8
IGF1RCD69CD83KLRG1PTGS2
Pathways
Namep-valueFDR
Chemokine receptors bind chemokines2.85E-114.98E-09
Immune system1.25E-101.09E-08
Interleukin-4 and interleukin-13 signaling2.82E-091.64E-07
RUNX1 and FOXP3 control the development of regulatory T lymphocytes (Tregs)4.31E-070.00001853717999
Peptide ligand-binding receptors6.71E-070.00002348305743
Signaling by Interleukins0.0000015036584930.0000436060963
Cytokine signaling in Immune system0.000026065058550.0006516264636
Dectin-1-mediated noncanonical NF-kB signaling0.000086405432150.001814514075
Immunoregulatory interactions between a lymphoid and a non-lymphoid cell0.00010833886750.002058438482
Class A/1 (rhodopsin-like receptors)0.00018330488280.003116183008
Interleukin-10 signaling0.00023669619340.0035504429
RUNX3 regulates immune response and cell migration0.00057918141130.007747184934
Extra-nuclear estrogen signaling0.00059593730260.007747184934
BH3-only proteins associate with and inactivate anti-apoptotic BCL-2 members0.00069925475230.008391057028
CLEC7A (Dectin-1) signaling0.00082280351450.00905083866
Generation of second messenger molecules0.0011719919080.01171991908
Innate immune system0.0016764040920.01572360367
GPCR ligand binding0.0017470670740.01572360367
Adaptive immune system0.0020598359910.01853852391
Estrogen-dependent nuclear events downstream of ESR-membrane signaling0.004670055830.03736044664
C-type lectin receptors (CLRs)0.005458044950.0436643596
Transcriptional regulation by RUNX30.0081243325990.05687032819
BMAL1:CLOCK, NPAS2 activates circadian gene expression0.0095182727090.06662790896
ESR-mediated signaling0.012073762370.08451633662
Transcriptional regulation by RUNX10.012881563710.08786708747
TCR signaling0.014644514580.08786708747

Next, we analyzed the datasets from our ALO monolayers for DEGs when challenged with SARS-COV-2 (Figure 5A,B). Genes and pathways upregulated in the infected lung organoid-derived monolayer models (Figure 5—figure supplements 12) overlapped significantly with those that were upregulated in the COVID-19 lung signature (compare Figure 4C,D with Figure 5C,D, Table 6, Table 7, Table 8). We observed only a partial overlap (ranging from ~22–55% across various human datasets; Figure 5—figure supplement 3) in upregulated genes and no overlaps among downregulated genes between model and disease (COVID-19; Figure 5E). Because the degree of overlap was even lesser (ranging from ~10 to 25% across various human datasets; Figure 5—figure supplement 3) in the case of another publicly released model (GSE160435) (Mulay et al., 2020), these discrepancies between the model and the actual disease likely reflect the missing stromal and immune components in our organoid monolayers. Regardless of these missing components, the model-derived DEG signature was sufficient to consistently and accurately classify diverse cohorts of patient-derived respiratory samples (ROC AUC ranging from 0.88 to 1.00; Figure 5F); the model-derived DEG signature was significantly induced in COVID-19 samples compared to normal controls (Figure 5G,H). Most importantly, the model-derived DEG signature was significantly induced in the epithelial cells recovered from bronchoalveolar lavage (Figure 5I).

Table 6
The list of GSE numbers used in the figures.
GSE#Cell type/tissueReferencesFigure
GSE132914Tissue from idiopathic pulmonary fibrosis subjects and donor controlsPMID:32991815Figure 1A
GSE151764COVID-19 and normal lung tissue post-mortemPMID:33033248Figure 4A–E, Figure 5E–G
GSE155241hPSC lung organoids and colon organoids infected with SARS-CoV-2PMID:33116299Figure 4E,F, Figure 6D
GSE156063Upper airway of COVID-19 patients and other acute respiratory illnessesPMID:33203890Figure 4E, Figure 5F,H
GSE147507A549 cells and bulk lungPMID:32416070; PMID:33782412Figure 4E,F, Figure 5F
GSE145926Bronchoalveolar lavage fluid (BALF) immune cells from COVID-19 and healthy subjectsPMID:32398875Figure 4E, Figure 5F,I
GSE150819Human bronchial organoidsFrom commercially available HBEpCFigure 4F, Figure 6C
GSE149312Intestinal organoids infected with SARS-CoV or SARS-CoV-2PMID:32358202Figure 4F
GSE151803hPSC-derived pancreatic and lung organoids infected with SARS-CoV-2No publication yetFigure 4F
GSE153940Vero E6 control or SARS-CoV-2-infected cellsPMID:32707573Figure 6B
GSE153218SARS-CoV-2-infected bronchoalveolar cells derived from organoids grown using progenitor cells from human fetal lung but tip (LBT).PMID:33283287Figure 6H
Table 7
Upregulated genes and pathways: uninfected vs infected (48 hpi) lung organoid monolayers.
Genes
IFI35EPSTI1AMIGO2IFITM2
SLC4A11CMPK2WARS1FAAP100
APOL1OASLIFI27ISG15
OAS3IFI44LCD14SLC35F6
IFIT3IFI44SAMD9L
IFIT2PARP9SRP9P1
Pathways
Namep-valueFDR
Interferon signaling1.11E-164.22E-15
Interferon alpha/beta signaling1.11E-164.22E-15
Cytokine signaling in Immune system1.15E-102.89E-09
Immune system0.0000025401148790.00004826218271
OAS antiviral response0.00047645456630.007146818495
Antiviral mechanism by IFN-stimulated genes0.0010333472610.01240016713
Interferon gamma signaling0.0018896946190.02078664081
Transfer of LPS from LBP carrier to CD140.0063187722450.05686895021
TRIF-mediated programmed cell death0.020912675860.1656073329
MyD88 deficiency (TLR2/4)0.037337482710.1656073329
IRAK2-mediated activation of TAK1 complex upon TLR7/8 or 9 stimulation0.037337482710.1656073329
TRAF6-mediated induction of TAK1 complex within TLR4 complex0.039371738120.1656073329
IRAK4 deficiency (TLR2/4)0.039371738120.1656073329
Activation of IRF3/IRF7 mediated by TBK1/IKK epsilon0.041401833220.1656073329
Caspase activation via death receptors in the presence of ligand0.041401833220.1656073329
IKK complex recruitment mediated by RIP10.049480774760.1855013265
Table 8
Downregulated genes and pathways: uninfected vs. infected (48 hpi) lung organoid monolayers.
AC093392.1ARHGAP19HLA-VRN7SL718P
MT-TVAC138969.3AC016766.1
Pathways
Namep-valueFDR
rRNA processing in the mitochondrion0.018927312460.08366120773
tRNA processing in the mitochondrion0.021271491050.08366120773
Mitochondrial translation termination0.043991554460.08366120773
Mitochondrial translation elongation0.043991554460.08366120773
Mitochondrial translation initiation0.044909217620.08366120773
Mitochondrial translation0.047657678440.08366120773

Taken together, these cross-validation studies from disease to model (Figure 4) and vice versa (Figure 5) provide an objective assessment of the match between the host response in COVID-19 lungs and our submerged ALO monolayers. Such a match was not seen in the case of the other models, for example, the proximal airway-mimic ALI model, HSAEpC monolayer, or hiPSC-derived AT2 models. Because the submerged ALO monolayers contained both proximal airway epithelia (basal cells) and promoted AT2→AT1 differentiation, findings demonstrate that mixed cellular monolayers can mimic the host response in COVID-19. A subtractive analysis revealed that the cell type that is shared between models that showed induction of host response signatures [i.e., ALO submerged monolayers and GSE155241 (Han et al., 2020a; Figure 5F)] but is absent in models that do not show such response (hu bronchial organoids, small airway epi, ALI-model of ALO) is AT1. We conclude that distal differentiation from AT2→AT1, a complex process that comprises distinct intermediates (Choi et al., 2020), is essential for modeling the host immune response in COVID-19. Further experimental evidence is needed to directly confirm if and which intermediate states during the differentiation of AT2 to AT1 are essential for the immune response to COVID19.

Both proximal and distal airway epithelia are required to mount the overzealous host response in COVID-19

We next asked which model best simulated the overzealous host immune response that has been widely implicated in fatal COVID-19 (Lowery et al., 2021; Lucas et al., 2020; Schultze and Aschenbrenner, 2021). To this end, we relied upon a recently described artificial intelligence (AI)-guided definition of the nature of the overzealous response in fatal COVID-19 (Sahoo et al., 2021). Using ACE2 as a seed gene, a 166-gene signature was identified and validated as an invariant immune response that was shared among all respiratory viral pandemics, including COVID-19 (Figure 6A). A subset of 20 genes within the 166-gene signature was subsequently identified as a determinant of disease severity/fatality; these 20 genes represented translational arrest, senescence, and apoptosis. These two signatures, referred to as ViP (166-gene) and severe ViP (20-gene) signatures, were used as a computational framework to first vet existing SARS-CoV-2 infection models that have been commonly used for therapeutic screens (Figure 6B–D). Surprisingly, we found that each model fell short in one way or another. For example, the Vero E6, which is a commonly used cultured cell model, showed a completely opposite response; instead of being induced, both the 166-gene and 20-gene ViP signatures were suppressed in infected Vero E6 monolayers (Figure 6B). Similarly, neither ViP signature was induced in the case of SARS-CoV-2-challenged human bronchial organoids (Suzuki, 2020) (Figure 6C). Finally, in the case of the hiPSC-derived AT1/2 organoids, which recapitulated the COVID-19-lung derived immune signatures (in Figure 4F), the 166-gene ViP signature was induced significantly (Figure 6D, top), but the 20-gene severity signature was not (Figure 6D, bottom). These findings show that none of the existing models capture the overzealous host immune response that has been implicated in a fatality.

Both proximal and distal airway components are required to model the overzealous host response in COVID-19.

(A) Schematic summarizing the immune signatures identified based on ACE2-equivalent gene induction observed invariably in any respiratory viral pandemic. The 166-gene ViP signature captures the cytokine storm in COVID-19, whereas the 20-gene subset severe ViP signature is indicative of disease severity/fatality. (B–D) Publicly available RNA seq datasets from commonly used lung models, Vero E6 (B), human bronchial organoids (C), and hPSC-derived AT1/2 cell-predominant lung organoids are classified using the 166-gene ViP signature (top row) and 20-gene severity signature (bottom row). (E–G) RNA seq datasets generated in this work using either human small airway epithelial cells (E), adult lung organoids as submerged or air-liquid interphase (ALI) models (left and right, respectively, in F) and hiPSC-derived AT2 cells (G) were analyzed and visualized as in (B–D). (H) Publicly available RNA seq datasets from fetal lung organoid monolayers (Lamers et al., 2021) infected or not with SARS-CoV-2 were analyzed as in (B–D) for the ability of ViP signatures to classify infected (I) from uninfected (U) samples. Receiver operating characteristics area under the curve (ROC AUC) in all figure panels indicate the performance of a classification model using the ViP signatures. (I) Summary of findings in this work, its relationship to the observed clinical phases in COVID-19, and key aspects of modeling the same. Table 6 lists details regarding the patient cohorts/tissue or cell types represented in each transcriptomic dataset.

Our lung models showed that both the 166- and 20-gene ViP signatures were induced significantly in the submerged ALO-derived monolayers that had distal differentiation (Figure 6E, left), but not in the proximal-mimic ALI model (Figure 6E, right). Neither signatures were induced in monolayers of small airway epithelial cells (Figure 6F) or hiPSC-derived AT2 cells (Figure 6G). Finally, we analyzed a recently published lung organoid model that supports robust SARS-CoV-2 infection; this model was generated using multipotent SOX2+ SOX9+ lung bud tip (LBT) progenitor cells that were isolated from the canalicular stage of human fetal lungs (~16–17 weeks post-conception) (Lamers et al., 2021). Despite mixed cellularity (proximal and distal), this fetal lung organoid model failed to induce the ViP signatures (Sahoo et al., 2021) (Figure 6H). These findings indicate that despite having mixed cellular composition and the added advantage of being able to support robust viral replication (achieving ~5 log-fold increase in titers), the model lacks the signature host response that is seen in all human samples of COVID-19.

Taken together with our infectivity analyses, these findings demonstrate that although the proximal airway epithelia and AT2 cells may be infected, and as described by others (Dye et al., 2015; Hou et al., 2020), may be vital for mounting a viral response and for disease transmission, these cells alone cannot mount the overzealous host immune response that is associated with the fatal disease. Similarly, even though the alveolar pneumocytes, AT1 and AT2 cells, are sufficient to mount the host immune response, in the absence of proximal airway components, they too are insufficient to recapitulate the severe ViP signature that is characterized by cellular senescence and apoptosis. However, when both proximal and distal components are present, that is, basal, ciliated, and AT1 cells, the model mimicked the overzealous host immune response in COVID-19 (Figure 6I).

Discussion

The most important discovery we report here is the creation of adult lung organoids that are complete with both proximal airway and distal alveolar epithelia; these organoids can not only be stably propagated and expanded in 3D cultures but also used as monolayers of mixed cellularity for modeling viral and host immune responses during respiratory viral pandemics. Furthermore, an objective analysis of this model and other existing SARS-CoV-2-infected lung models against patient-lung-derived transcriptomes showed that the model that most closely emulates the elements of viral infectivity, lung injury, and inflammation in COVID-19 is one that contained both proximal and distal alveolar signatures (Figure 6H), whereas the presence of just one or the other fell short.

There are three important impacts of this work. First, the successful creation of adult human lung organoids that are complete with both proximal and distal signatures has not been accomplished before. Previous works show the successful use of airway basal cells for organoid creationtion, but ensuring the completeness of the model with all other lung cells has been challenging to create (Nikolić and Rawlins, 2017). The multicellularity of the lung has been a daunting challenge that many experts have tried to recreate in vitro; in fact, the demand for perfecting such a model has always remained high, not just in the current context of the COVID-19 pandemic but also with the potential of future pandemics. We have provided evidence that the organoids that were created using our methodology retain proximal and distal cellularity throughout multiple passages and even within the same organoid. Although a systematic design of experiment (DoE) approach (Bukys et al., 2020) was not involved in getting to this desirable goal, a rationalized approach was taken. For example, a Wnt/R-spondin/Noggin-containing conditioned media was used as a source of the so-called ‘niche factors’ for any organoid growth (Sato and Clevers, 2015). This was supplemented with recombinant FGF7/10; FGF7 is known to help cell proliferation and differentiation and is required for normal branching morphogenesis (Padela et al., 2008), whereas FGF10 helps in cell maturation (Rabata et al., 2020) and in alveolar regeneration upon injury (Yuan et al., 2019). Together, they are likely to have directed the differentiation toward distal lung lineages (hence, the preservation of alveolar signatures). The presence of both distal alveolar and proximal ciliated cells was critical: proximal cells were required to recreate sustained viral infectivity, and the distal alveolar pneumocytes, in particular, the ability of AT2 cells to differentiate into AT1 pneumocytes was essential to recreate the host response. It is possible that the response is mediated by a distinct AT2-lineage population, that is, damage-associated transient progenitors (DATPs), which arise as intermediates during AT2→AT1 differentiation upon injury-induced alveolar regeneration (Choi et al., 2020). Although somewhat unexpected, the role of AT1 pneumocytes in mounting innate immune responses has been documented before in the context of bacterial pneumonia (Yamamoto et al., 2012; Wong and Johnson, 2013). In work (Huang et al., 2020) that was published during the preparation of this article, authors used long-term ALI models of hiPSC-derived AT2 monolayers (in growth conditions that inhibit AT2→AT1 differentiation, as we did here for our AT2 model) and showed that SARS-CoV-2 induces iAT2-intrinsic cytotoxicity and inflammatory response, but failed to induce type 1 interferon pathways (IFN α/β). It is possible that prolonged culture of iAT2 pneumocytes gives rise to some DATPs but cannot robustly do so in the presence of inhibitors of AT1 differentiation. This (spatially segregated viral and host immune response) is a common theme across many lung infections (including bacterial pneumonia and other viral pandemics (Hou et al., 2020; Taubenberger and Morens, 2008; Weinheimer et al., 2012; Chan et al., 2013) and hence, this mixed cellularity model is appropriate for use in modeling diverse lung infections and respiratory pandemics to come.

Second, among all the established lung models so far, ours features four key properties that are desirable whenever disease models are being considered for their use in HTP modes for rapid screening of candidate therapeutics and vaccines: (i) reproducibility, propagability, and scalability; (ii) cost-effectiveness; (iii) personalization; and (iv) modularity, with the potential to add other immune and nonimmune cell types to our multicellular complex lung model. We showed that the protocol we have optimized supports isolation, expansion, and propagability at least up to 12–15 passages (at the time of submission of this work), with documented retention of proximal and distal airway components up to P8 (by RNA seq). We noted some variability of cell types between patient to patient, and between early and late passages of ALOs, which is probably because of the heterogeneity of organoids isolated from patient’s lung specimens. Feasibility has also been established for scaling up and optimizing the conditions for them to be used in miniaturized 384-well infectivity assays. We also showed that the protocols for generating lung organoids could be reproduced in both genders, and regardless of the donor’s smoking status, consistency in outcome and growth characteristics was observed across all isolation attempts. The ALOs are also cost-effective; the need for exclusive reliance on recombinant growth factors was replaced at least in part with conditioned media from a commonly used cell line (L-WRN/ ATCC CRL-2647 cells). Such media has a batch-to-batch stable cocktail of Wnt, R-Spondin, and Noggin, and has been shown to improve reproducibility in the context of GI organoids in independent laboratories (VanDussen et al., 2019). By that token, our culture conditions may have also improved reproducibility. The major disadvantage, however, remains that the composition of the media is undefined. Because the model is propagable, repeated iPSC-reprogramming (another expensive step) is also eliminated, further cutting costs compared to many other models. As for personalization, our model is derived from adult lung stem cells from deep lung biopsies; each organoid line was established from one patient. By avoiding iPSCs or expanded potential stem cells (EPSCs), this model not only captures genetics but also retains organ-specific epigenetic programming in the lung, and hence, potentially additional programming that may occur in disease (such as in the setting of chronic infection, injury, inflammation, somatic mutations, etc.). The ability to replicate donor phenotype and genotype in vitro allows for potential use as preclinical human models for phase ‘0’ clinical trials. As for modularity, by showing that the 3D lung organoids could be used as polarized monolayers on transwells to allow infectious agents to access the apical surface (in this case, SARS-CoV-2), we demonstrate that the organoids have the potential to be reverse-engineered with additional components in a physiologically relevant spatially segregated manner: for example, immune and stromal cells can be placed in the lower chamber to model complex lung diseases that are yet to be modeled and have no cure (e.g., idiopathic pulmonary fibrosis, etc.).

Third, the value of the ALO models is further enhanced due to the availability of companion readouts/biomarkers (e.g., ViP signatures in the case of respiratory viral pandemics, or monitoring the E gene, or viral shedding, etc.) that can rapidly and objectively vet treatment efficacy based on set therapeutic goals. Of these readouts, the host response, as assessed by ViP signatures, is a key vantage point because an overzealous host response is what is known to cause fatality. Recently, a systematic review of the existing preclinical animal models revealed that most of the animal models of COVID-19 recapitulated mild patterns of human COVID-19; no severe illness associated with mortality was observed, suggesting a wide gap between COVID-19 in humans (Spagnolo et al., 2020) and animal models (Ehaideb et al., 2020). It is noteworthy that alternative models that effectively support viral replication, such as the proximal airway epithelium or iPSC-derived AT2 cells (analyzed in this work) or a fetal lung bud tip-derived organoid model recently described by others (Lamers et al., 2021), do not recapitulate the host response in COVID-19. The lung model we present here is distinct from all currently available other models (see Table 1) because of the confirmed presence of both proximal and distal airway cell types over successive passages, which is yet to be accomplished for adult lung organoid models. Another distinguishing feature of our model is the way we rigorously validated its usefulness in modeling COVID-19 via computational approaches. We confirmed, based on the gene expression changes upon SARS-CoV-2-challenge, that our model most closely recapitulates the human disease, that is, COVID-19. Analyses also pinpointed the importance of two factors that were critical in modeling COVID-19: (i) adult source and (ii) model completeness, with both proximal and distal airway cells. We conclude that the model revealed here, in conjunction with the ViP signatures described earlier (Sahoo et al., 2021), could serve as a preclinical model with companion diagnostics to identify drugs that target both the viral and host response in pandemics.

Limitations of the study

Our adult stem-cell-derived lung organoids, complete with all epithelial cell types, can model COVID-19, but remains a simplified/rudimentary version compared to the adult human organ. While we successfully demonstrated the proximo-distal mixed cellular composition of the ALOs using four different approaches (flow cytometry, RNA seq, confocal immunofluorescence, and targeted qPCR) and showed that such mixed cellularity is preserved during prolonged culture, the exact cellular proportion was not assessed here. Single-cell sequencing and multiplexed profiling by flow cytometry are some of the approaches that can provide such in-depth characterization to assess cellular composition at baseline and track how such composition changes upon infection and injury. For instance, although the epithelial contributions to the host response are important, they alone cannot account for the response of the immune cells and the nonimmune stromal cells, and their crosstalk with the epithelium. Given that epithelial inflammation and damage is propagated by vicious forward-feedback loops of multicellular crosstalk, it is entirely possible that the epithelial signatures induced in infected ALO-derived monolayers are also only a fraction of the actual epithelial response mounted in vivo. Regardless of the missing components, what appears to be the case is that we already have a model that recapitulates approximately a quarter to half of the genes that are induced across diverse COVID-19-infected patient samples. This model can be further improved by the simultaneous addition of endothelial cells and immune cells to better understand the pathophysiological basis for DAD, microangiopathy, and even organizing fibrosis with loss of lung capacity that has been observed in many patients (Spagnolo et al., 2020); these insights should be valuable to fight some of the long-term sequelae of COVID-19. Future work with flow cytometry and cell sorting of our lung organoids would help us understand each cell type’s role in viral pathogenesis. Larger living biobanks of genotyped and phenotyped ALOs, representing donors of different age, ethnicity, predisposing conditions, and coexisting comorbidities, will advance our understanding of why SARS-CoV-2 and possibly other infectious agents may trigger different disease course in different hosts. Although we provide proof-of-concept studies in low-throughput mode demonstrating the usefulness of the ALOs as human preclinical models for screening therapeutics in phase ‘0’ trials, optimization for the same to be adapted in HTP mode was not attempted here.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
AntibodyAnti-ACE2 (mouse monoclonal)Santa CruzCat# sc390851RRID::AB_2861379IF (1:100)
AntibodyAnti-human ACE2 (rat monoclonal)BioLegendCat# 375802RRID::AB_2860959IF (1:50)
AntibodyAnti-acetylated ɑ tubulin (mouse monoclonal)Santa CruzCat# sc23950RRID::AB_628409IF (1:500)FC (1:8000)
AntibodyAnti-AQP5 (mouse monoclonal)Santa CruzCat# sc514022RRID::AB_2891066IF (1:100)FC (1:800)
AntibodyAnti-CC10 (mouse monoclonal)Santa CruzCat# sc365992RRID::AB_10915481IF (1:100)
OtherDAPIInvitrogenCat# D1306RRID::AB_2629482IF (1:500)
AntibodyRecombinant anti-cytokeratin 5 (rabbit monoclonal)AbcamCat# ab52635RRID::AB_869890IF (1:100)FC (1:8000)
AntibodyRecombinant anti-mucin 5AC (rabbit monoclonal)AbcamCat# ab229451RRID::AB_2891067IF (1:150)FC (1:800)
AntibodyAnti-sodium potassium ATPase (rabbit monoclonal)AbcamCat# ab76020RRID::AB_1310695IF (1:400)
AntibodyAnti-occludin (mouse monoclonal)Thermo FisherCat# OC-3F10RRID::AB_2533101IF (1:500)
OtherPhalloidin, Alexa Fluor 594InvitrogenCat# A12381RRID:AB_2315633IF (1:500)
OtherPropidium iodideInvitrogenV13241FC (1:100)
AntibodySARS-CoV/SARS-CoV-2 nucleocapsid antibody (mouse monoclonal)Sino BiologicalCat# 40143-MM05RRID::AB_2827977IF (1:250)IHC (1:500)
AntibodyAnti-SARS spike glycoprotein (mouse monoclonal)AbcamCat# ab273433RRID::AB_2891068IHC (1:250)
Antibodyanti-SP-B (mouse monoclonal)Santa CruzCat# sc133143RRID::AB_2285686IF (1:100)FC (1:8000)
AntibodyAnti-prosurfactant protein C (rabbit polyclonal)AbcamCat# ab90716RRID::AB_10674024IF (1:150)
AntibodyGoat anti-rat IgG H&L secondary antibody, Alexa Flour 594InvitrogenCat# A-11007RRID:AB_10561522IF (1:500)
AntibodyGoat anti-rabbit IgG H&L secondary antibody, Alexa Fluor 594InvitrogenCat# A-11012RRID:AB_2534079IF (1:500)
AntibodyGoat anti-mouse IgG H&L secondary antibody, Alexa Fluor 488InvitrogenCat# A-11011RRID:AB_143157IF (1:500)FC (1:1000)
AntibodyGoat anti-rabbit IgG H&L secondary antibody, Alexa Fluor 488AbcamCat# ab150077RRID:AB_2630356FC (1:1000)
OtherCountess II Automated Cell CounterThermo Fisher ScientificAMQAX1000Section‘The preparation of lung organoid-derived monolayers’
OtherEpithelial Volt-Ohm (TEER) MeterMilliporeMERS00002Section ‘Permeability of lung monolayer using FITC-dextran’
OtherLeica TCS SPE ConfocalLeica MicrosystemsTCS SPESection‘Immunofluorescence’
OtherPower Pressure Cooker XLTristar ProductsSection‘Immunohistochemistry’
OtherCanon Rebel XS DLSRCanonFigure 2—figure supplement 1
OtherMiniAmp Plus Thermal CyclerApplied BiosystemsCat# A37835Section‘Quantitative (q)RT-PCR’
OtherQuantStudio5Applied BiosystemsCat# A28140 RRID:SCR_020240Section‘Quantitative (q)RT-PCR’
OtherLight Microscope (brightfield images)Carl Zeiss LLCAxio Observer, Inverted; 491917-0001-000Figure 2—figure supplement 1
OtherSpark 20 M Multimode Microplate ReaderTecanSection‘Permeability of lung monolayer using FITC-dextran’
OtherGuava easyCyte Benchtop Flow CytometerMilliporeGuava easyCyte 62LSection‘The characterization of lung cell types using flow cytometry’
Software, algorithmImageJImageJRRID:SCR_003070
Software, algorithmGraphPad PrismGraphPad PrismRRID:SCR_002798
Software, algorithmLAS AF SoftwareLAS AF Software
Software, algorithmQuantStudio Design & Analysis SoftwareQuantStudio Design & Analysis Software
Software, algorithmCIBERSORTxCIBERSORTx
Software, algorithmFlowJoFlowJo V10, BD BioSciencesRRID:SCR_008520
Chemical compound, drugZinc formalinFisher ScientificCat# 23-313096
Chemical compound, drugXyleneVWRCat# XX0060-4
Chemical compound, drugHematoxylinSigma-Aldrich IncCat# MHS1
Chemical compound, drugEthanolKoptecCat# UN1170
Chemical compound, drugSodium citrateSigma-AldrichCat# W302600
Chemical compound, drugDAB (10×)Thermo FisherCat# 1855920(1:10)
Chemical compound, drugStable peroxidase substrate buffer (10×)Thermo FisherCat# 34062(1:10)
Chemical compound, drug3%hydrogen peroxideTargetCat# 245-07-3628
Chemical compound, drugHorse serumVector LabsCat# 30022
Commercial assay or kitHRP Horse Anti-Rabbit IgG Polymer Detection KitVector LaboratoriesCat# MP-7401
Chemical compound, drugParaformaldehyde 16% Solution, EM GradeElectron Microscopy SciencesCat# 15710
Chemical compound, drug100%methanolSupelcoCat# MX0485
Chemical compound, drugGlycineFisher ScientificCat# BP381-5
Chemical compound, drugBovine serum albuminSigma-AldrichCat# A9647-100G
Chemical compound, drugTriton-X 100Sigma-AldrichCat# X100-500ML
Chemical compound, drugProLong GlassInvitrogenCat# P36984
Chemical compound, drugNail Polish (Rapid Dry)Electron Microscopy SciencesCat# 72180
Chemical compound, drugGill Modified Hematoxylin (Solution II)Millipore SigmaCat# 65066-85
Chemical compound, drugHistoGelThermo ScientificCat# HG4000012
Chemical compound, drugTrypLE SelectThermo ScientificCat# 12563-011
Chemical compound, drugAdvanced DMEM/F-12Thermo ScientificCat# 12634-010
Chemical compound, drugHEPES bufferLife TechnologiesCat# 15630080
Chemical compound, drugGlutamaxThermo ScientificCat# 35050-061
Chemical compound, drugPenicillin-streptomycinThermo ScientificCat# 15140-122
Chemical compound, drugCollagenase type IThermo ScientificCat# 17100-017
Chemical compound, drugMatrigelCorningCat# 354234
Chemical compound, drugB-27Thermo ScientificCat# 17504044
Chemical compound, drugN-acetyl-L-cysteineSigma-AldrichCat# A9165
Chemical compound, drugNicotinamideSigma-AldrichCat# N0636
Chemical compound, drugFGF-7 (KGF)PeproTechCat# 100-19-50ug
Chemical compound, drugFGF10PeproTechCat# 100-26-50ug
Chemical compound, drugA-83-01Bio-Techne Sales Corp.Cat# 2939/50
Chemical compound, drugSB202190Sigma-AldrichCat# S7067-25MG
Chemical compound, drugY-27632R&D SystemsCat# 1254/50
Chemical compound, drugDPBSThermo ScientificCat# 14190-144
Chemical compound, drugUltrapure WaterInvitrogenCat# 10977-015
Chemical compound, drugEDTAThermo ScientificCat# AM9260G
Chemical compound, drugHydrocortisoneSTEMCELL TechnologiesCat# 7925
Chemical compound, drugHeparinSigma-AldrichCat# H3149
OtherPneumaCult Ex-Plus MediumSTEMCELL TechnologiesCat# 5040Section‘The preparation of lung organoid-derived monolayers’
OtherPneumaCult ALI MediumSTEMCELL TechnologiesCat# 5001Section‘ALImodel of lung organoids’
Chemical compound, drugGoat serumVector LaboratoriesCat# MP-7401
Chemical compound, drugFetal bovine serumSigma-AldrichCat# F2442-500ML
Chemical compound, drugAnimal Component-Free Cell Dissociation KitSTEMCELL TechnologiesCat# 5426
Chemical compound, drugRed Blood Cell Lysis BufferInvitrogenCat# 00-4333-57
Chemical compound, drugCell Recovery SolutionCorningCat# 354253
Chemical compound, drugSodium azideFisher ScientificCat# S227I-100
Chemical compound, drugCyto-Fast Fix/Perm Buffer SetBioLegendCat# 426803
Chemical compound, drugFITC-dextranSigma-AldrichCat# FD10S
Commercial assay or kitQuick-RNA MicroPrep KitZymo ResearchCat# R1051
Commercial assay or kitQuick-RNA MiniPrep KitZymo ResearchCat#R1054
Chemical compound, drugEthyl alcohol, pureSigma-AldrichCat# E7023
Chemical compound, drugTRI ReagentZymo ResearchCat# R2050-1-200
Sequence-based reagent2x SYBR Green qPCR Master MixBimakeCat# B21203
Sequence-based reagentqScript cDNA SuperMixQuanta BiosciencesCat# 95048
Sequence-based reagentApplied Biosystems TaqMan Fast Advanced Master MixThermo ScientificCat# 4444557
Sequence-based reagent18S, Hs99999901_s1Thermo ScientificCat# 4331182
Sequence-based reagentE_Sarbeco_F1 Forward PrimerIDTCat# 10006888
Sequence-based reagentE_Sarbeco_R2 Reverse PrimerIDTCat# 10006890
Sequence-based reagentE_Sarbeco_P1 ProbeIDTCat# 10006892
Other12-well Tissue Culture PlateCytoOneCat# CC7682-7512Section‘Isolation and culture of human whole lung-derived organoids’
OtherTranswell Inserts (6.5 mm, 0.4µm pore size)CorningCat# 3470Section‘The preparation of lung organoid-derived monolayers’
OtherMicroscope Cover Glass (#1 Thickness) 24 × 50 mmVWRCat# 16004-098Section‘Immunofluorescence’
OtherMicroscope Cover Glass (#1 Thickness) 25 mm diameterChemglass Life SciencesCat# CLS-1760-025Section‘Immunofluorescence’
OtherMillicell EZ Slide 8-Well ChamberMillipore SigmaCat# PEZGS0816Section‘Immunofluorescence’
OtherTrypan Blue StainInvitrogenCat# T10282(1:2)
Other70µm Cell StrainerThermo Fisher ScientificCat# 22-363-548Section‘The preparation of lung organoid-derived monolayers’
Other100 µm Cell StrainerCorningCat# 352360Section‘Isolation and culture of human whole lung-derived organoids’
OtherNoyes Spring Scissors – AngledFine Science ToolsCat# 15013-12Section‘Isolation and culture of human whole lung-derived organoids’

Detailed methods

Collection of human lung specimens for organoid isolation

Request a detailed protocol

To generate adult healthy lung organoids, fresh biopsy bites were prospectively collected after surgical resection of the lung by the cardiothoracic surgeon. Before collection of the lung specimens, each tissue was sent to a gross anatomy room where a pathologist cataloged the area of focus, and the extra specimens were routed to the research lab in Human Transport Media (HTM, Advanced DMEM/F-12, 10 mM HEPES, 1× Glutamax, 1× penicillin-streptomycin, 5 μM Y-27632) for cell isolation. Deidentified lung tissues obtained during surgical resection, which were deemed excess by clinical pathologists, were collected using an approved human research protocol (IRB# 101590; PI: Thistlethwaite). Isolation and biobanking of organoids from these lung tissues were carried out using an approved human research protocol (IRB# 190105: PI Ghosh and Das) that covers human subject research at the UC San Diego HUMANOID Center of Research Excellence (CoRE). For all the deidentified human subjects, information, including age, gender, and previous history of the disease, was collected from the chart following the rules of HIPAA and described in Table 3.

A portion of the same lung tissue specimen was fixed in 10% zinc-formalin for at least 24 hr followed by submersion in 70% EtOH until embedding in FFPE blocks.

Autopsy procedures for lung tissue collection from COVID-19-positive human subjects

Request a detailed protocol

The lung specimens from COVID-19-positive human subjects were collected through autopsy (the study was IRB exempt). All donations to this trial were obtained after telephone consent followed by written email confirmation by the next of kin/power of attorney per California state law (no in-person visitation could be allowed into our COVID-19 ICU during the pandemic). The team member followed the CDC guidelines for COVID19 and the autopsy procedures (CAP, 2020; CDC, 2020). Lung specimens were collected in 10% zinc-formalin and stored for 72 hr before processing for histology. Patient characteristics are listed in Table 3.

Autopsy #2 was a standard autopsy performed by anatomical pathology in the BSL3 autopsy suite. The patient expired and his family consented for autopsy. After 48 hr, the lungs were removed and immersion fixed whole in 10% formalin for 48 hr and then processed further. Lungs were only partially fixed at this time (about 50% fixed in thicker segments) and were sectioned into small 2–4 cm chunks and immersed in 10% formalin for further investigation.

Autopsy #4 and #5 were collected from rapid postmortem lung biopsies. The procedure was performed in the Jacobs Medical Center ICU (all of the ICU rooms have a pressure-negative environment, with air exhausted through HEPA filters [Biosafety Level 3 (BSL3)] for isolation of SARS-CoV-2 virus). Biopsies were performed 2–4 hr after patient expiration. The ventilator was shut off to reduce the aerosolization of viral particles at least 1 hr after the loss of pulse and before sample collection. Every team member had personal protective equipment in accordance with the university policies for procedures on patients with COVID-19 (N95 mask+ surgical mask, hairnet, full face shield, surgical gowns, double surgical gloves, booties). Lung biopsies were obtained after L-thoracotomy in the fifth intercostal space by the cardiothoracic surgery team. Samples were taken from the left upper lobe (LUL) and left lower lobe (LLL) and then sectioned further.

Isolation and culture of human whole lung-derived organoids

Request a detailed protocol

A previously published protocol was modified to isolate lung organoids from three human subjects (Sachs et al., 2019; Zhou et al., 2018). Briefly, normal human lung specimens were washed with PBS/4× penicillin-streptomycin and minced with surgical scissors. Tissue fragments were resuspended in 10 ml of wash buffer (Advanced DMEM/F-12, 10 mM HEPES, 1× Glutamax, 1× penicillin-streptomycin) containing 2 mg/ml Collagenase Type I (Thermo Fisher, USA) and incubated at 37°C for approximately 1 hr. During incubation, tissue pieces were sheared every 10 min with a 10 ml serological pipette and examined under a light microscope to monitor the progress of digestion. When 80–100% of single cells were released from connective tissue, the digestion buffer was neutralized with 10 ml wash buffer with added 2% fetal bovine serum; the suspension was passed through a 100 µm cell strainer and centrifuged at 200 rcf. Remaining erythrocytes were lysed in 2 ml red blood cell lysis buffer (Invitrogen) at room temperature for 5 min, followed by the addition of 10 ml of wash buffer and centrifugation at 200 rcf. Cell pellets were resuspended in cold Matrigel (Corning, USA) and seeded in 25 µl droplets on a 12-well tissue culture plate. The plate was inverted and incubated at 37°C for 10 min to allow complete polymerization of the Matrigel before the addition of 1 ml Lung Expansion Medium per well. Lung expansion media was prepared by modifying a media that was optimized previously for growing GI-organoids (50% conditioned media, prepared from L-WRN cells with Wnt3a, R-spondin, and Noggin, ATCC-CRL-3276) (Sayed et al., 2020c; Ghosh et al., 2020; Sayed et al., 2020a; Sayed et al., 2020b) with a proprietary cocktail from the HUMANOID CoRE containing B27, TGF-β receptor inhibitor, antioxidants, p38 MAPK inhibitor, FGF 7, FGF 10, and ROCK inhibitor. The lung expansion media was compared to alveolosphere media I (IMDM and F12 as the basal medium with B27, low concentration of KGF, monothioglycerol, GSK3 inhibitor, ascorbic acid, dexamethasone, IBMX, cAMP, and ROCK inhibitor) and II (F12 as the basal medium with added CaCl2, B27, low concentration of KGF, GSK3 inhibitor, TGF-β receptor inhibitor dexamethasone, IBMX, cAMP, and ROCK inhibitor) modified from previously published literature (Jacob et al., 2019; Yamamoto et al., 2017). Neither alvelosphere media contain any added Wnt3a, R-spondin, and Noggin. The composition of these media was developed either by fundamentals of adult stem cell-derived mixed epithelial cellularity in other organs (like the GI tract [Miyoshi and Stappenbeck, 2013; Sato et al., 2009; Sayed et al., 2020c]) or rationalized based on published growth conditions for proximal and distal airway components (Gotoh et al., 2014; Sachs et al., 2019; van der Vaart and Clevers, 2021). Organoids were maintained in a humidified incubator at 37°C/5% CO2, with a complete media change performed every 3 days. After the organoids reached confluency between 7 and 10 days, organoids were collected in PBS/0.5 mM EDTA and centrifuged at 200 rcf for 5 min. Organoids were dissociated in 1 ml trypLE Select (Gibco, USA) per well at 37°C for 4–5 min and mechanically sheared. Wash buffer was added at a 1:5, trypLE to wash buffer ratio. The cell suspension was subsequently centrifuged, resuspended in Matrigel, and seeded at a 1:5 ratio. Lung organoids were biobanked and passage 3–8 cells were used for experiments. Subculture was performed every 7–10 days.

The preparation of lung organoid-derived monolayers

Request a detailed protocol

Lung organoid-derived monolayers were prepared using a modified protocol of GI organoid-derived monolayers (Sayed et al., 2020c; Ghosh et al., 2020; Sayed et al., 2020a; Sayed et al., 2020b). Briefly, transwell inserts (6.5 mm diameter, 0.4 µm pore size, Corning) were coated in Matrigel diluted in cold PBS at a 1:40 ratio and incubated for 1 hr at room temperature. Confluent organoids were collected in PBS/EDTA on day 7 and dissociated into single cells in trypLE for 6–7 min at 37°C. Following enzymatic digestion, the cell suspension was mechanically sheared through vigorous pipetting with a 1000 µl pipette and neutralized with wash buffer. The suspension was centrifuged, resuspended in PneumaCult Ex-Plus Medium (StemCell, Canada), and passed through a 70 µm cell strainer. The coating solution was aspirated, and cells were seeded onto the apical membrane at 1.8E5 cells per transwell with 200 µl PneumaCult Ex-Plus media. 700 µl of PneumaCult Ex-Plus was added to the basal chamber. Cells were cultured over the course of 2–4 days. A media change of both the apical and basal chambers was performed every 24 hr.

ALI model of lung organoids

Organoids were dissociated into single cells and expanded in T-75 flasks in PneumaCult Ex-Plus Medium until confluency was reached. Cells were dissociated in ACF Enzymatic Dissociation Solution (StemCell) for 6–7 min at 37°C and neutralized in equal volume ACF Enzyme Inhibition Solution (StemCell). Cells were seeded in the apical chamber of transwells at 3.3E4 cells per transwell in 200 µl of PneumaCult Ex-Plus Medium. 500 µl of PneumaCult Ex-Plus was added to the basal chamber. Media in both chambers was changed every other day until confluency was reached (~4 days). The media was completely removed from the apical chamber, and media in the basal chamber was replaced with ALI Maintenance Medium (StemCell). The media in the basal chamber was changed every 2 days. The apical chamber was washed with warm PBS every 5–7 days to remove accumulated mucus. Cells were cultured under ALI conditions for 21+ days until they completed differentiation into a pseudostratified mucociliary epithelium. To assess the integrity of the epithelial barrier, TEER was measured with an Epithelial Volt-Ohm Meter (Millicell, USA). The media was removed from the basal chamber, and wash media was added to both chambers. Cultures were equilibrated to 37°C before TEER values were measured. Final values were expressed as Ω·cm2 units and were calculated by multiplying the growth area of the membrane by the raw TEER value.

The culture of primary airway epithelial cells and iPSC-derived alveolar epithelial cells

Request a detailed protocol

Primary NHBE cells were obtained from Lonza and grown according to instructions. NHBE cells were cultured in T25 cell culture tissue flasks with PneumaCult-Ex Plus media (StemCell). Cells were seeded at ~100,000 cells/T25 flask and incubated at 37°C, 5% CO2. Once cells reached 70–80% confluency, they were dissociated using 0.25% Trypsin in dissociation media and plated in 24-well transwells (Corning). Primary human bronchial epithelial cells (HBEpC) and small airway epithelial cells (HSAEpC) were obtained from Cell Applications Inc The HBEpC and HSAEpC were cultured in human bronchial/tracheal epithelial cell media and small airway epithelial cell media, respectively, following the instructions of Cell Application.

Human iPSC-derived alveolar epithelial type 2 cells (iHAEpC2) were obtained from Cell Applications Inc and cultured in growth media (i536K-05, Cell Applications Inc) according to the manufacturer’s instructions. All the primary cells were used within early passages (5–6) to avoid any gradual disintegration of the airway epithelium with columnar epithelial structure and epithelial integrity.

The infection with SARS-Cov2

Request a detailed protocol

Lung organoid-derived monolayers or primary airway epithelial cells either in 384-well plates or in transwells were washed twice with antibiotic-free lung wash media. 1E5 PFU of SARS-CoV-2 strain USA-WA1/2020 (BEI Resources NR-52281) in complete DMEM was added to the apical side of the transwell and allowed to incubate for 24, 48, 72, and 96 hr at 34°C and 5% CO2. After incubation, the media was removed from the basal side of the transwell. The apical side of the transwells was then washed twice with (antibiotic-free lung wash media) and then twice with PBS. TRIzol Reagent (Thermo Fisher 15596026) was added to the well and incubated at 34°C and 5% CO2 for 10 min. The TRIzol Reagent was removed and stored at –80°C for RNA analysis.

RNA isolation

Organoids and monolayers used for lung cell-type studies were lysed using RNA lysis buffer followed by RNA extraction per Zymo Research Quick-RNA MicroPrep Kit instructions. Tissue samples and monolayers in SARS-CoV2 studies were lysed in TRI-Reagent and RNA was extracted using Zymo Research Direct-zol RNA Miniprep.

Quantitative (q)RT-PCR

Request a detailed protocol

Organoid and monolayer cell-type gene expression was measured by qRT-PCR using 2x SYBR Green qPCR Master Mix. cDNA was amplified with gene-specific primer/probe set for the lung cell type markers and qScript cDNA SuperMix (5×). qRT-PCR was performed with the Applied Biosystems QuantStudio 5 Real-Time PCR System. Cycling parameters were as follows: 95°C for 20 s, followed by 40 cycles of 1 s at 95°C and 20 s at 60°C. All samples were assayed in triplicate and eukaryotic 18S ribosomal RNA was used as a reference.

Cell typesMarkerPrimer sequence
Basal cellsITGA6, NGFR, TP63ITGA6 F ′CGAAACCAAGGTTCTGAGCCC′ITGA6 R ′CTTGGATCTCCACTGAGGCAGT′NGFR F′ CCTCATCCCTGTCTATTGCTCCNGFR R′ GTTGGCTCCTTGCTTGTTCTGCTP63 F′ CAGGAAGACAGAGTGTGCTGGTTP63 R′ AATTGGACGGCGGTTCATCCCT
GobletMuc5ACMuc5AC F ′GGAACTGTGGGGACAGCTCTT′Muc5AC R ′GTCACATTCCTCAGCGAGGTC′
CiliaFoxJ1FoxJ1 F ′ACTCGTATGCCACGCTCATCTG′’FoxJ1 R ′GAGACAGGTTGTGGCGGATTGA′
Club cellSCGB1A1SCGB1A1 F ′CAAAAGCCCAGAGAAAGCATC′SCGB1A1 R ′CAGTTGGGGATCTTCAGCTTC′
Alveolar type 1AQP5, PDPNAQP5 F ′TACGGTGTGGCACCGCTCAATG′AQP5 R ′AGTCAGTGGAGGCGAAGATGCA′PDPN F ′GTGCCGAAGATGATGTGGTGAC′PDPN R ′GGACTGTGCTTTCTGAAGTTGGC′
Alveolar type 2SFTPA1, SFTPCSFTPA1 F ′CACCTGGAGAAATGCCATGTCC′SFTPA1 R ′AAGTCGTGGAGTGTGGCTTGGA′SFTPC F ′GTCCTCATCGTCGTGGTGATTG′SFTPC R ′AGAAGGTGGCAGTGGTAACCAG′

Assessment of SARS-CoV-2 infectivity test

Request a detailed protocol

Assessment of SARS-CoV-2 infectivity test was determined by qPCR using TaqMan assays and TaqMan Universal PCR Master Mix as done before (Corman et al., 2020; Lamers et al., 2020). cDNA was amplified with gene-specific primer/probe set for the E gene and QPCR was performed with the Applied Biosystems QuantStudio 3 Real-Time PCR System. The specific TaqMan primer/probe set for E gene are as follows: forward 5′-ACAGGTACGTTAATAGTTAATAGCGT-3′ (IDT, Cat# 10006888); reverse 5′-ATATTGCAGCAGTACGCACACA-3′; probe 5′-FAM-ACACTAGCCATCCTTACTGCGCTTCG-BBQ-3′ and 18S rRNA. Cycling parameters were as follows: 95°C for 20 s, followed by 40 cycles of 1 s at 95°C and 20 s at 60°C. All samples were assayed in triplicate, and gene eukaryotic 18S ribosomal RNA was used as a reference.

Immunofluorescence

Request a detailed protocol

Organoids and lung organoid-derived monolayers in plates or in 8-well chamber slides were fixed by either (i) 4% PFA at room temperature for 30 min and quenched with 30 mM glycine for 5 min, (ii) ice-cold 100% methanol at –20°C for 20 min, and (iii) ice-cold 100% methanol on ice for 20 min. Subsequently, samples were permeabilized and blocked for 2 hr using an in-house blocking buffer (2 mg/ml BSA and 0.1% Triton X-100 in PBS); as described previously (Lopez-Sanchez et al., 2014). Primary antibodies were diluted in blocking buffer and allowed to incubate overnight at 4°C; secondary antibodies were diluted in blocking buffer and allowed to incubate for 2 hr in the dark. Antibody dilutions are listed in the key resources table. ProLong Glass was used as a mounting medium. #1 Thick Coverslips were applied to slides and sealed. Samples were stored at 4°C until imaged. FFPE-embedded organoid and lung tissue sections underwent antigen retrieval as previously described in Materials and methods for immunohistochemistry staining. After antigen retrieval and cooling in DI water, samples were permeabilized and blocked in blocking buffer and treated as mentioned above for immunofluorescence. Images were acquired at room temperature with Leica TCS SPE confocal and with DMI4000 B microscope using the Leica LAS-AF Software. Images were taken with a 40× oil-immersion objective using 405-, 488-, 561 nm laser lines for excitation. z-stack images were acquired by successive z-slices of 1 µm in the desired confocal channels. Fields of view that were representative and/or of interest were determined by randomly imaging three different fields. z-slices of a z-stack were overlaid to create maximum intensity projection images; all images were processed using FIJI (ImageJ) software.

Embedding of organoids in HistoGel

Request a detailed protocol

Organoids were seeded on a layer of Matrigel in 6-well plates and grown for 7–8 days. Once mature, organoids were fixed in 4% PFA at room temperature for 30 min and quenched with 30 mM glycine for 5 min. Organoids were gently washed with PBS and harvested using a cell scraper. Organoids were resuspended in PBS using wide-bore 1000 µl pipette tips. Organoids were stained using Gill’s hematoxylin for 5 min for easier FFPE embedding and sectioning visualization. Hematoxylin-stained organoids were gently washed in PBS and centrifuged and excess hematoxylin was aspirated. Organoids were resuspended in 65°C HistoGel and centrifuged at 65°C for 5 min. HistoGel-embedded organoid pellets were allowed to cool to room temperature and stored in 70% ethanol at 4°C until ready for FFPE embedding by LJI Histology Core. Successive FFPE-embedded organoid sections were cut at a 4 µm thickness and fixed on to microscope slides.

Immunohistochemistry

Request a detailed protocol

For SARS CoV-nucleoprotein (np) antigen retrieval, slides were immersed in Tris-EDTA buffer (pH 9.0) and boiled for 10 min at 100°C inside a pressure cooker. Endogenous peroxidase activity was blocked by incubation with 3% H2O2 for 10 min. To block nonspecific protein binding, 2.5% goat serum was added. Tissues were then incubated with a rabbit SARS CoV-NP antibody (Sino Biological, see key resource table) for 1.5 hr at room temperature in a humidified chamber and then rinsed with TBS or PBS three times, for 5 min each. Sections were incubated with horse anti-rabbit IgG secondary antibodies for 30 min at room temperature and then washed with TBS or PBS three times for 5 min each. Sections were incubated with DAB and counterstained with hematoxylin for 30 s.

Permeability of lung monolayer using FITC-dextran

Request a detailed protocol

Adult lung monolayers were grown for 2 days in PneumaCult Ex-Plus media on transwell inserts (6.5 mm diameter, 0.4 µm pore size, Corning). TEER was monitored with an Epithelial Volt-Ohm Meter (Millicell). On the second day of growth, FITC-dextran (10 kD) was added at a 1:50 dilution in lung wash media. The basolateral side of the insert was changed to lung wash media only. After 30 min of incubation with FITC-dextran, 50 µl of the basolateral supernatant was transferred to an opaque welled 96-well plate. Fluorescence was measured using a TECAN plate reader.

The characterization of lung cell types using flow cytometry

Request a detailed protocol

Lung organoids were dissociated into single cells via trypLE digestion and strained with a 30 µm filter (Miltenyi Biotec, Germany). Approximately 2.5E5 cells for each sample were fixed and permeabilized at room temperature in Cyto-Fast Fix Perm buffer (BioLegend, USA) for 20 min. The samples were subsequently washed with Cyto-Fast Perm Wash solution (BioLegend) and incubated with lung epithelial cell-type markers for 30 min. Following primary antibody incubation, the samples were washed and incubated with propidium iodide (Invitrogen) and Alexa Flour 488 secondary antibodies (Invitrogen) for 30 min. Samples were resuspended in FACS buffer (PBS, 5% FBS, 2 mM sodium azide). Flow cytometry was performed using Guava easyCyte Benchtop Flow Cytometer (Millipore) and data was analyzed using InCyte (version 3.3) and FlowJo X v10 software.

RNA sequencing

Request a detailed protocol

RNA sequencing libraries were generated using the Illumina TruSeq Stranded Total RNA Library Prep Gold with TruSeq Unique Dual Indexes (Illumina, San Diego, CA). Samples were processed following the manufacturer’s instructions, except modifying RNA shear time to 5 min. The resulting libraries were multiplexed and sequenced with 100 basepair (bp) paired-end (PE100) to a depth of approximately 25–40 million reads per sample on an Illumina NovaSeq 6000 by the Institute of Genomic Medicine (IGM) at the University of California San Diego. Samples were demultiplexed using bcl2fastq v2.20 Conversion Software (Illumina). RNAseq data was processed using kallisto (version 0.45.0) and human genome GRCh38 Ensembl version 94 annotation (Homo_sapiens GRCh38.94 chr_patch_hapl_scaff.gtf). Gene-level transcripts per million (TPM) values and gene annotations were computed using tximport and biomaRt R package. A custom Python script was used to organize the data and log reduced using log2(TPM +1). The raw data and processed data are deposited in Gene Expression Omnibus under accession nos. GSE157055 and GSE157057.

Data collection and annotation

Request a detailed protocol

Publicly available COVID-19 gene expression databases were downloaded from the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus website (GEO) (Edgar et al., 2002; Barrett et al., 2005; Barrett et al., 2013). If the dataset is not normalized, RMA (Robust Multichip Average) (Irizarry et al., 2003a; Irizarry et al., 2003b) is used for microarrays and TPM (Li and Dewey, 2011; Pachter, 2011) is used for RNA seq data for normalization. We used log2(TPM +1) to compute the final log-reduced expression values for RNA seq data. Accession numbers for these crowdsourced datasets are provided in the figures and article. All of the above datasets were processed using the Hegemon data analysis framework (Dalerba et al., 2011; Dalerba et al., 2016; Volkmer et al., 2012).

Analysis of RNA seq datasets

Request a detailed protocol

DESeq2 (Love et al., 2014) was applied to uninfected and infected samples to identify up- and downregulated genes. A gene signature score is computed using both the up- and downregulated genes that are used to order the sample. To compute the gene signature score, first, the genes present in this list were normalized according to a modified Z-score approach centered around StepMiner threshold (formula = (expr -SThr)/3*stddev). The normalized expression values for every probeset for all the genes were added or subtracted (depending on up and downregulated genes) together to create the final score. The samples were ordered based on the final gene signature score. The gene signature score is used to classify sample categories, and the performance of the multiclass classification is measured by ROC-AUC values. A color-coded bar plot is combined with a violin plot to visualize the gene signature-based classification. All statistical tests were performed using R version 3.2.3 (2015-12-10). Standard t-tests were performed using Python scipy.stats.ttest_ind package (version 0.19.0) with Welch’s two-sample t-test (unpaired, unequal variance (equal_var = False), and unequal sample size) parameters. Multiple hypothesis correction was performed by adjusting p-values with statsmodels.stats.multitest.multipletests (fdr_bh: Benjamini/Hochberg principles). The results were independently validated with R statistical software (R version 3.6.1; 2019-07-05). Pathway analysis of gene lists was carried out via the Reactome database and algorithm (Fabregat et al., 2018). Reactome identifies signaling and metabolic molecules and organizes their relations into biological pathways and processes. Violin, swarm, and bubble plots were created using Python seaborn package version 0.10.1.

Single-Cell RNA seq data analysis

Request a detailed protocol

Single-Cell RNA seq data from GSE145926 was downloaded from GEO in the HDF5 Feature Barcode Matrix Format. The filtered barcode data matrix was processed using Seurat v3 R package (Stuart et al., 2019) and Scanpy Python package (Wolf et al., 2018). Pseudo bulk analysis of GSE145926 data was performed by adding counts from the different cell subtypes and normalized using log2(CPM + 1). Epithelial cells were identified using SFTPA1, SFTPB, AGER, AQP4, SFTPC, SCGB3A2, KRT5, CYP2F1, CCDC153, and TPPP3 genes using SCINA algorithm (Zhang et al., 2019). Pseudo bulk datasets were prepared by adding counts from the selected cells and normalized using log(CPM + 1).

Assessment of cell-type proportions

Request a detailed protocol

CIBERSORTx (https://cibersortx.stanford.edu/runcibersortx.php) was used for cell-type deconvolution of our dataset (which was normalized by CPM). As reference samples, we first used the single-cell RNA seq dataset (GSE132914) from Gene Expression Omnibus (GEO). Next, we analyzed the bulk RNA seq datasets for the identification of cell types of interest using relevant gene markers (see Table 2): AT1 cells (PDPN, AQP5, P2RX4, TIMP3, SERPINE1), AT2 cells (SFTPA1, SFTPB, SFTPC, SFTPD, SCGB1A1, ABCA3, LAMP3), BASAL cells (CD44, KRT5, KRT13, KRT14, CKAP4, NGFR, ITGA6), CLUB cells (SCGB1A1, SCGB3A2, SFTPA1, SFTPB, SFTPD, ITGA6, CYP2F1), GOBLET cells (CDX2, MUC5AC, MUC5B, TFF3), ciliated cells (ACTG2, TUBB4A, FOXA3, FOXJ1, SNTN), and generic lung lineage cells (GJA1, TTF1, EPCAM) were identified using SCINA algorithm. Then, normalized pseudo counts were obtained with the CPM normalization method. The cell-type signature matrix was derived from the single-cell RNA seq dataset, cell types, and gene markers of interest. It was constructed by taking an average from gene expression levels for each of the markers across each cell type.

Statistical analysis

Request a detailed protocol

All experiments were repeated at least three times, and results were presented either as one representative experiment or as average ± SEM. Statistical significance between datasets with three or more experimental groups was determined using one-way ANOVA including a Tukey’s test for multiple comparisons. For all tests, a p-value of 0.05 was used as the cutoff to determine significance (*p<0.05, **p<0.01, ***p<0.001, and ****p<0.0001). All experiments were repeated a least three times, and p-values are indicated in each figure. All statistical analyses were performed using GraphPad Prism 6.1. A part of the statistical tests was performed using R version 3.2.3 (2015-12-10). Standard t-tests were performed using Python scipy.stats.ttest_ind package (version 0.19.0).

Data availability

Sequencing data have been deposited in GEO under accession codes GSE157055, and GSE157057.

The following data sets were generated
    1. Sahoo D
    2. Das S
    3. Ghosh P
    (2020) NCBI Gene Expression Omnibus
    ID GSE157057. Human lung organoid for modeling infection and disease conditions.

References

  1. Book
    1. McGowan SE
    (2014)
    chapter 4 - the formation of pulmonary alveoli
    In: Pinkerton KE, editors. Lung: Development, Aging and the Environment (2nd edition). Academic Press. pp. 65–84.
  2. Book
    1. Shami SG
    2. Evans MJ
    (2015) Kinetics of pulmonary cells
    In: Parent RA, editors. (EM ID= Biology of the Normal Lung’’c1fd9724-9ce9-40eb-8620-4combiolofve34dthe. Academic Press. pp. 1–164.
    https://doi.org/10.1016/B978-0-12-404577-4.15002-7

Decision letter

  1. Milica Radisic
    Reviewing Editor; University of Toronto, Canada
  2. Jos W Van der Meer
    Senior Editor; Radboud University Medical Centre, Netherlands
  3. Milica Radisic
    Reviewer; University of Toronto, Canada
  4. Hans Clevers
    Reviewer; University Medical Centre Utrecht, Netherlands

Our editorial process produces two outputs: i) public reviews designed to be posted alongside the preprint for the benefit of readers; ii) feedback on the manuscript for the authors, including requests for revisions, shown below. We also include an acceptance summary that explains what the editors found interesting or important about the work.

Acceptance summary:

Comparing with previous systems, the new organoid culture that you describe can maintain both proximal and distal cell types in adult lungs, and this ratio appears to be relatively stable in long term cultures. These organoids can also be used to mimic infectivity and immune responses characteristic of Covid19.

Decision letter after peer review:

Thank you for submitting your article "Adult Stem Cell-derived Complete Lung Organoid Models Emulate Lung Disease in COVID-19" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Milica Radisic as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by Jos van der Meer as the Senior Editor. The following individual involved in review of your submission has agreed to reveal their identity: Hans Clevers (Reviewer #2).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

1) All reviewers agree that the cell phenotypes and stability of the organoids in culture should be better characterized with time in culture using a number of new markers that are listed in the reviewer reports, as well as additional methods such as immunostaining and flow. Reviewers have raised concerns that organoids are not stable during culture.

2) Claims should be tampered, in particular related to high throughput drug screening since only few drugs were tested.

3) Discrepancies with transcriptomic signatures of the infected human lungs need to be more carefully considered.

Reviewer #1 (Recommendations for the authors):

The manuscript by Tindle et al. describes generation of adult lung organoids (ALO) from human lung biopsies and their use to study the changes in gene expression as a result of SARS-CoV-2 infection. The main advantage of the use of organoids is the ability to generate many cell types that make up the lung. In this particular case the authors report the presence of AT1, AT2 cells, Basal cells, Goblet cells, Ciliated cells and Club cells. The authors were able to cultivate the cells at the air-liquid interface and to establish cultures of predominately proximal and predominately distal airway cells. The main finding is that proximal cells are more prone to viral infection, while distal cells are governing the exuberant inflammatory response, with both cells required for the exuberant response to occur. A useful information provided by the paper is the analysis gene signatures of various cellular models, in comparison to the infected human lung.

1. Although cellular complexity is notable compared to some other models, it is important to more precisely benchmark how this compares to the percentages of cells in the actual lung. Figure 3 shows percentages of cells based on RNA sequencing, however a more precise enumeration might be provided by flow cytometry provided that the cells can be accurately dissociated.

2. In general, organoid models lack functional readouts such as permeability or barrier function. Were authors able to establish and enumerate the differences in barrier function in the monolayer model in the transwell system?

3. It is slightly discouraging that the model captures only 7 upregulated genes of the 76 reported in covid19 patient lungs and no downregulated genes. It also upregulates 15 genes that are not reported in the patient lungs (Figure 5E). These discrepancies should be better discussed.

Reviewer #2 (Recommendations for the authors):

The manuscript "Adult Stem Cell-derived Complete Lung Organoid Models Emulate Lung Disease in COVID-19" by Das and colleagues introduces a new model system of airway epithelium derived from adult lung organoids (ALO) to be utilised for the study of COVID-19-related processes. In this manuscript two main novelties are claimed: the development of a new model system which represents both proximal as distal airway epithelium and a computationally acquired gene signature that identifies SARS-CoV-2-infected individuals. While interesting data are presented, the novelty claim is questionable and the data is not always convincing.

1. The manuscript claims a novel model system of distal and proximal airway epithelium. The authors however fail to discuss a recently published COVID-organoid study which reports similar observations. While their novelty claim could still hold in some areas, the authors do need to discuss the following manuscript (https://doi.org/10.15252/embj.2020105912).

2. Expression patterns in Figure 2B are hard to interpret when no tissue control is used. A positive control of lung tissue should be used to make valid conclusions.

3. The authors claim the presence of multiple cell types within single ALO. Data are not strong.

a. Co-staining of KRT5 and SFTPB (as shown in figure 2J) is difficult to explain, since these markers are expressed in two different cell types. The lower panel in the same figure does show two separate populations, but the images shown in the second panel do not.

b. While the authors claim the presence of proximal and distal cells in ALOs, the images show 100% of SFTPC+ cells or 100% KRT5+ cells or 100% Na/K-ATPase cells. This rather shows two different types of ALO within one culture, then mixed cell populations within a single ALO.

c. The authors claim the presence of ciliated cells by staining of acetylated α tubulin in figure 2K-J and figure 3E. The images shown however are hardly showing specific ciliated cell staining and more importantly the cilia present on the cell membrane. The authors should include much higher quality images that clearly show cilia or other markers which are exclusive for ciliated cells.

d. In figure 3H, the authors show MUC5B positive cells in their monolayer cultures. These goblet cells seem to comprise quite a significant proportion of the culture. Their graph in 3B however shows no goblet cell contribution.

The claim that ALO consists of a mix of defined cell types is essentially based on deconvoluted bulk RNAseq data. Such a claim is not conclusive.

4. Authors claim that ALO maintain cell type ratios similar to lung tissue over several passages. Their data in figure 3B however show a drastic change in the composition of the ALOs. The shift from AT2 to AT1 is explained to occur when the ALO are cultured as submerged monolayers. However, the shift is already visible at later passages in 'standard' ALO culture. Moreover, when comparing cell types between passage 1 and 8, the composition changes dramatically. The cultures have lost basal cells, ALO1 has lost goblet cells, ALO2 has lost ciliated cells. In addition, ALO1 composition differs from ALO2 at passage 2 which doesn't support the statement that the cultures are stable/robust.

5. While the authors claim that proximal airway cultures are infected and responsible for maintaining virus replication, there are no virus entry marker-annotated cells in the composition graph of figure 1B.

6. Similar to point 3, the authors claim the development of the first combined organoid culture of proximal and distal cell types, as supported by deconvoluted bulk RNAseq. The hiPSC-derived AT2 cultures however also show the presence of both distal and proximal cell types.

7. Figure 3H shows a graph presenting the infectivity relative to peak. This axis makes comparison of infectivity impossible. The authors may want to include suppl Figure 5A-B as main figure and exclude 3H from the manuscript. Moreover, viral E-gene qPCR should be complemented by viral titer measurements to verify findings for live virus. The analysis of SARS-CoV-2 infectivity is very limited when only showing a few infected cells and viral E-gene graphs. Suppl Figure 5C also only shows one or two infected cells in all samples which does convince as proof of infectivity.

8. The authors provide data for a single drug to indicate the possibility of high-throughput screening of drugs for COVID-19 treatment. This does not say much about the throughput of the assay.

9. Figure 5C compares uninfected and infected monolayer cultures for genes that are identified in the patient cohort. While the authors claim comparable patterns between the cultures and the patients, the heatmap can not be directly read. The authors should include more details in the legends and combine the heatmaps in 4C and 5C for direct comparison. Currently, the colours represent raw z-scores which do not indicate transcript read numbers but relative differences of the expression of the indicated genes in the samples analysed. These numbers could differ extensively between organoids and patients.

10. In addition to point 9, the authors show that there is very limited overlap between significantly differentially expressed genes in monolayers and patients. While the authors believe this is mostly due to the lack of mesenchymal and immune components, this indicates that the monolayer itself is not important for the observed infection signature. This makes the claim that the monolayers represent infectivity in patients questionable.

11. Description in the discussion of growth factors supplied in the medium like FGF7 and FGF10 are not novel. Sachs et al. 2019 (https://doi.org/10.15252/embj.2018100300) already described these growth factors in the culture medium of airway organoids.

12. In the discussion, the authors describe the advantages of their model system including reproducibility, retainment of genetics and use for infection studies. These claims are however not novel since previous papers from multiple groups have reported similar organoid-based models for SARS-CoV-2 research.

Reviewer #3 (Recommendations for the authors):

In order to further improve this manuscript we recommend that the authors address the following points:

Figure 2.

1) In Figure 2H. the authors use NGFR as a generic stem cell marker. However, previous publications have shown that NGFR is a basal cell marker (Rock et al., 2009). We suggest that the authors to clarify what is the stem cell population they are attempting to mark here.

2) In all of Figure 2, but particularly, Figure 2K and 2J, we recommend the authors specify which passage number these organoids are from and show some simple quantification about different lineages to help the readers understand if different lineage remain stable between early and late passages.

3) None of the organoid immunofluorescence images in Figure 2 demonstrate that individual organoids contain both airway and alveolar lineages. In particular, SFTPB is presented as a alveolar type 2 cell marker, but it is well-established that this protein is also expressed in club cells (see: https://hlca.ds.czbiohub.org). It is not therefore useful for co-staining with basal cell markers to establish that individual organoids contain both airway and alveolar cell lineages. Mixed airway and alveolar organoid lineage is not yet convincingly demonstrated.

4) The Ac-Tub staining is not convincing as the cilia structures are not visible at the magnification presented.

5) The authors need to comment on the fate of the CC10/SFTPC double positive cell shown in Figure 2J bottom panel. Previous efforts to find SFTPC/CC10 co-expressing cells in human lungs in vivo have been unsuccessful (See: https://hlca.ds.czbiohub.org).

6) We recommend the authors also perform some simple polarity marker gene staining, such as ECAD and ZO-1, to confirm the apical-basal polarity of the organoids. This would help to justify the need of generating monolayer of epithelium for COVID infection in following experiments.

Figure 3.

1) The authors need to discuss in Figure 3B the extent of the variation of cell lineages observed in the self-renewing 3-D organoid cultures between passages 1 and 8. These changes suggest that the organoids do not have a stable cellular phenotype over time and cast serious doubt on their ability to reproducibly model lung infections, or other diseases. Similarly, in the submerged cultures do the 5 different bars represent the results of 5 different experiments using ALO1? Which passage was transferred to 2D culture? And can the authors comment on the reproducibility of their data?

2) For the same figure, the authors need to explain what are the cell populations called the 'general lung lineage (GLL)' and the 'viral entry marker (VEM)'. Does the VEM population include AT2, club cells and ciliated cells?

3) Figure 3E and Figure 3B (middle panel) show quite contradictory results regarding different proportion of cell lineages existing in the organoid system. Figure 3B middle panel seems to suggest a dominant AT1 and airway lineage, but few AT2 lineage cells. However, in Figure 3E, the authors seemed to suggest AT2 is also quite prevailing. Is this another example of variability?

4) We recommend the authors to also check AT1 and AT2 marker genes in the ALI culture, as a recent preprint has suggested that alveolar cell fates can also be maintained in ALI culture (Abo et al., 2020).

5) The authors need to address if the 'not sustained viral release and infectivity' of iAT2 cells was due to the virus quickly infecting, and killing, the AT2 cells within the first 48 hours, which could be another explanation for the kinetics observed in Figure 3H.

Figures 4-6

The approach in Figures 4-6 of characterizing the COVID-infected lungs at a transcriptional level, deriving transcriptional signatures and testing to what extent these are replicated in the various organoid models is innovative and highly commendable. The authors state that these data strongly suggest that the mixed organoids presented are a better culture model for COVID infection than previous organoid experiments. The spread of the data, e.g. in 6E, may also reflect the variability of this culture system. However, this reviewer is not sufficiently bioinformatically-literate to comment in detail on this aspect of the manuscript.

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Adult Stem Cell-derived Complete Lung Organoid Models Emulate Lung Disease in COVID-19" for further consideration by eLife. Your revised article has been evaluated by Jos van der Meer (Senior Editor) and a Reviewing Editor.

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

The Reviewers found that your revised manuscript is improved, however immunostaining is still not found to be completely conclusive, for example immunostaining of ciliated cells and others would help.

We suggest that the authors should alter the text in response to the points raised in the second round of reviews to tamper the claims according to the suggestions of the Reviewers listed below.

Reviewer #2 (Recommendations for the authors):

The revisions on the manuscript "Adult Stem Cell-derived Complete Lung Organoid Models Emulate Lung Disease in COVID-19" by the group of Prof. Das have been extensive, as is the rebuttal letter.

Overall, the authors have answered most of the questions about COVID-19 in an adequate manner. However, major issues regarding the characterisation of the organoid model remain. In general, the authors still present contradicting data about the cell type composition of the model. This also casts indirect doubt on the COVID experiments.

Below, I go through the rebuttal letter point-by-point

Comment # 1

All reviewers agree that the cell phenotypes and stability of the organoids in culture should be better characterized with time in culture using a number of new markers that are listed in the reviewer reports, as well as additional methods such as immunostaining and flow. Reviewers have raised concerns that organoids are not stable during culture.

Response to rebuttal

The authors have included additional analyses to verify the stability of their model system. These additional analyses however are not overlapping with existing data. Discussed by method:

- The authors have used flow cytometry to identify cell compositions in the organoid models in early and late passages. While the authors indeed show (Figure 2S5) that the percentage of cell types based on key markers is stable in passages, the data itself is not convincing. The authors describe a model that consists of 96% basal cells, 60% goblet or secretory cells, 90% ciliated cells, 48% AT1 cells and 90% AT2 cells. While the authors already discuss that these percentages add up to more that 100%, the sum of the cell types is even 384% over 5 cell types. The authors claim this is due to overlap of marker genes between cell types and reference a manuscript. The manuscript however does not cover this overlap but actually shows that these markers can be used to separate the cell types. Apart from this, the same antibodies when used in immunostaining (Figure 3E-F) do not show these high percentages of cell positivity for the given markers. While in the FACS analysis the monolayers consist of 92% SFTPB+ cells, the immunostaining shows 10-20% positivity. Both methods are based on protein levels and should therefore show similar phenotypes. Secondly, the authors' claim of lowly expressed marker genes in all cell types this could be resolved by using a more stringent gate in the flow cytometer, only gating the real positives or by addition of a second marker in another channel. The first approach seems to already work for AQP5+ population in which there is a second population that is more positive and therefore potentially true AT1 cells. This result is also in line with the immunostaining in Figure 3E.

Anyway, the markers used are known to be highly specific to each of the cell types. Thus, this part of the manuscript remains very confusing.

– While the qRT-PCR data of the comparison between tissue and ALOs is convincing (Figure 2S3) and indeed underlines the representative nature of the model, the comparison between passages is slightly discouraging (Figure 2S4). While the authors state that cell type composition is stable during culture based on the qRT-PCR results, the data show some large differences between early and late passages. Loss of SCGB1A1 and increase in FOXJ1 in culture in all ALOs is visible as well as some varying levels of other markers in one or more lines does not comply with the stability stated in the text. The authors should still state this variation somewhere in the text.

– Immunostaining of ciliated cells and others would really help. The authors however have not included immunostainings of cell types in different passages to verify the qRT-PCR results. These immunostainings could be quantified to show cell compositions in the ALOs in 3D and 2D.

In all, the authors should use qRT-PCR to determine presence of all cell types, and combine this with immunostainings to determine cell compositions of the models. The authors still show the CIBERSORTx method which confuses the reader since it contradicts data shown in other figures. The authors explain in their rebuttal letter that the method was only used as corroborative, yet the data contradicts their own data.

Comment #2

Claims should be tampered, in particular related to high throughput (HTP) drug screening since only few drugs were tested.

Response to rebuttal

The authors explain in great depth that they did not imply that HTP were performed. This can however be inferred from the text even after changing the text as given in the rebuttal. The finding that the ALOs can be used in a 384-well format is important but the authors should add some more in-depth detailed description that this only generates a possibility for potential HTP screens instead of inferring HTP can be done on these models. The description of a "proof-of-concept for HTP" confuses. This term implies that the authors have used an initial HTP drug screen to validate. This can be changed to "initial protocol for building a platform on which HTP drug screens could be performed"?

Comment #3

Discrepancies with transcriptomic signatures of the infected human lungs need to be more carefully considered.

Response to rebuttal

The authors have included more clinical datasets and comparisons between other published model systems. It remains important to mention the discrepancies between clinical samples and the model. This is altered in the text sufficiently. This comment is therefore satisfactorily answered in the revised version of the manuscript.

Personal reviewer comments

Comment 1

The manuscript claims a novel model system of distal and proximal airway epithelium. The authors however fail to discuss a recently published COVID-organoid study which reports similar observations. While their novelty claim could still hold in some areas, the authors do need to discuss the recent manuscripts.

Response to rebuttal

The authors have extended their table including all the existing models with the recent published model systems. I agree that keeping up with COVID-19 model systems is hard in these times. The inclusion into the tables has, however, not led to the authors changing their claim of a novel model system of bronchiolar and alveolar potential. In line 102-104 the authors still write "…what is particularly noteworthy is that none recapitulate the heterogeneous epithelial cellularity of both proximal and distal airways…" which should be altered since Lamers et al. did exactly this. Moreover, in the rebuttal letter, the authors also mention that iPSC-derived AT2 cultures can contain both proximal and distal cell types.

The comparison with the currently published method of Lamers et al. 2021 is now in the manuscript and could be highlighted in the discussion or at the end of the paragraph named "Creation of a lung organoid model, complete with both proximal and distal airway epithelia" by adding a few lines comparing the cell types present in this model (ciliated and goblet) which was not shown in Lamers et al. 2021 (of at least this is convincingly demonstrated)

Comment 2.

Expression patterns in Figure 2B are hard to interpret when no tissue control is used. A positive control of lung tissue should be used to make valid conclusions.

Response to rebuttal

The authors have answered this comment by adding Figure 2S3

Comment 3.

The authors claim the presence of multiple cell types within single ALO. Data are not strong.

3a. Co-staining of KRT5 and SFTPB (as shown in figure 2J) is difficult to explain, since these markers are expressed in two different cell types. The lower panel in the same figure does show two separate populations, but the images shown in the second panel do not.

and

3b. While the authors claim the presence of proximal and distal cells in ALOs, the images show 100% of SFTPC+ cells or 100% KRT5+ cells or 100% Na/K-ATPase cells. This rather shows two different types of ALO within one culture, then mixed cell populations within a single ALO.

Response to rebuttal

By changing the images and the text, the authors made clear that the ALOs are either composed of multiple cell types or of a singular cell type. Therefore, this comment is sufficiently answered.

3c. The authors claim the presence of ciliated cells by staining of acetylated α tubulin in figure 2K-J and figure 3E. The images shown however are hardly showing specific ciliated cell staining and more importantly the cilia as present on the cell membrane.

Response to rebuttal

By addition of the higher resolution images, the authors answered this comment sufficiently.

3d. In figure 3H, the authors show MUC5B positive cells in their monolayer cultures. These goblet cells seem to comprise quite a significant proportion of the culture. Their graph in 3B however shows no goblet cell contribution. The claim that ALO consists of a mix of defined cell types is essentially based on deconvoluted bulk RNAseq data. Such a claim is not conclusive.

Response to rebuttal

This comment comes back to earlier described methods in the editor comment section. To shortly summarise, the CYBERSORTx data confuses the reader and is contradictory to the immunostainings, qRT-PCR and flow cytometry data.

Comment 4.

Authors claim that ALO maintain cell type ratios similar to lung tissue over several passages. Their data in figure 3B however show a drastic change in the composition of the ALOs. The shift from AT2 to AT1 is explained to occur when the ALO are cultured as submerged monolayers. However, the shift is already visible at later passages in 'standard' ALO culture. Moreover, when comparing cell types between passage 1 and 8, the composition changes dramatically. The cultures have lost basal cells, ALO1 has lost goblet cells, ALO2 has lost ciliated cells. In addition, ALO1 composition differs from ALO2 at passage 2 which doesn't support the statement that the cultures are stable/robust.

Response to rebuttal

See earlier comments about using CYBERSORTx and flow cytometry data.

Comment 5.

While the authors claim that proximal airway cultures are infected and responsible for maintaining virus replication, there are no virus entry marker-annotated cells in the composition graph of figure 1B.

Response to rebuttal

The comment was aimed at the proximal ALI culture cell composition graph in original figure 3B. This graph did not indicate a VEM population. Still this culture was infected by the virus, without showing a VEM population. The authors have removed the viral entry marker cells from their data representation. This was due to bioinformatic difficulty and VEM cells being a separate cell type in their analyses.

Comment 6

Similar to point 3, the authors claim the development of the first combined organoid culture of proximal and distal cell types, as supported by deconvoluted bulk RNAseq. The hiPSC-derived AT2 cultures however also show the presence of both distal and proximal cell types.

Response to rebuttal

In the rebuttal letter the authors mention the manuscripts describing the presence of proximal cell types in a AT2 differentiation protocol of iPSCs. This once more underlines that their statement of the first culture of proximal and distal cell types is incorrect.

Comment 7

Figure 3H shows a graph presenting the infectivity relative to peak. This axis makes comparison of infectivity impossible. The authors may want to include suppl Figure 5A-B as main figure and exclude 3H from the manuscript. Moreover, viral E-gene qPCR should be complemented by viral titer measurements to verify findings for live virus. The analysis of SARS-CoV-2 infectivity is very limited when only showing a few infected cells and viral E-gene graphs. Suppl Figure 5C also only shows one or two infected cells in all samples which does convince as proof of infectivity.

Response to rebuttal

The authors provide an extensive reasoning behind the use of the legend in figure 3H. The sustained infectivity the authors want to bring across is sufficiently depicted in figure 3H. Figure 3H might confuse the readers that the submerged culture are more heavily infected than ALI cultures. Moreover, this would allow the authors to already claim their permissiveness for infectivity in proximal airways without having to only look at 48-72h timeframe. "Permissive" in general refers to infection at an early stage which is shown in figure 3S3B and not figure 3H.

Comment 8

The authors provide data for a single drug to indicate the possibility of high-throughput screening of drugs for COVID-19 treatment. This does not say much about the throughput of the assay.

Response to rebuttal

This comment has been answered and once more commented in editor comments.

Comment 9

Figure 5C compares uninfected and infected monolayer cultures for genes that are identified in the patient cohort. While the authors claim comparable patterns between the cultures and the patients, the heatmap can not be directly read. The authors should include more details in the legends and combine the heatmaps in 4C and 5C for direct comparison. Currently, the colors represent raw z-scores which do not indicate transcript read numbers but relative differences of the expression of the indicated genes in the samples analysed. These numbers could differ extensively between organoids and patients.

Response to rebuttal

The authors have added some extra lines on the method used to generate these heatmaps. It now becomes clear that these datasets can not be directly compared within one heatmap.

Comment 10

In addition to point 9, the authors show that there is very limited overlap between significantly differentially expressed genes in monolayers and patients. While the authors believe this is mostly due to the lack of mesenchymal and immune components, this indicates that the monolayer itself is not important for the observed infection signature. This makes the claim that the monolayers represent infectivity in patients questionable.

Response to rebuttal

This has been discussed in the editor's section of the revision above

Comment 11.

Description in the discussion of growth factors supplied in the medium like FGF7 and FGF10 are not novel. Sachs et al. 2019 (https://doi.org/10.15252/embj.2018100300) already described these growth factors in the culture medium of airway organoids.

Response to rebuttal

The authors have sufficiently answered this comment.

Comment 12.

In the discussion, the authors describe the advantages of their model system including

reproducibility, retainment of genetics and use for infection studies. These claims are however not novel since previous papers from multiple groups have reported similar organoid-based models for SARS-CoV-2 research.

Response to rebuttal

With the comments above and the answers from the authors, this comment will be answered. The authors should better define their novelty by explaining the origin of the organoids model (adult stem cells) and their improvement in COVID-19 modelling.

Reviewer #3 (Recommendations for the authors):

The authors established a new 3D adult lung organoid system. Comparing with previous systems, the new organoid culture can maintain both proximal and distal cell types in adult lungs, and this ratio appears to be relatively stable in long term cultures. The 3D organoids can subsequently be dissociated and re-plated into 2D submerge culture, which exhibited promising features for modelling COVID19 infection. The authors further used bioinformatics analysis to show that the COVID19 infection using the 2D submerge culture was able to recapitulate both the infectivity and the immune responses in COVID patients.

In the revision processes, the authors have provided more supporting data to show the co-existence of proximal and distal lung cell types in the long-term culture and addressed most of the previous reviewers' comments. The few comments that we recommend the authors to further address are as follows:

1) The author added the Act-TUB staining for 2D culture in Figure 3, which looks convincing, however, for the 3D organoids, Act-TUB staining in Figure 2J still doesn't look like any cilia structure.

2) In Page 6 text line 189, the author mentioned a 'stem cell' population, which is TP63 positive. This is separate to the basal cells on line 188. Whereas in Figure 2-Supp4 the authors have acknowledged TP63 as a basal cell marker. We recommend the authors to make this consistent as basal cells which are well-established to be the airway stem cells (Rock et al., 2009).

3) The authors still make a strong claim about the co-existence of proximal (airway) and distal (alveolar) cell populations in a single 3D organoid (line 198). However, the staining shown in Figure 2 can only infer the existence of either proximal or distal cell populations in a single 3D organoid, given SFTPB is not a specific marker for alveolar lineage. Additionally, this strong claim wasn't adding much value to the manuscript as the authors only need to show proximal-distal cells exist in the overall population as 3D culture, given only the 2D submerged culture was used for COVID infection. We recommend the author not to mention this claim as even the revised data do not support it.

In addition, the same paragraph describes BASCs in the organoids. We cannot cite a reference to refute the existence of BASC cells in human lungs. However, in now more than 10 years of people looking for them a convincing demonstration that BASCs exist in human lungs is still missing. Human distal airways have a very different organisation to those of mice (respiratory bronchioles). BASC cells have not been found in them. We recommend that you do not highlight human BASCs in the organoids given the lack of credible evidence that they exist in vivo.

4) The authors need to clarify how the 5 replicates for submerged culture and 2 replicates for ALI culture were done in Figure 2B middle panel. Were they from the same ALO line or from different ALO lines? What was the passage number used here? These will help the readers to have an idea about how reproducible the system is.

https://doi.org/10.7554/eLife.66417.sa1

Author response

Essential revisions:

1) All reviewers agree that the cell phenotypes and stability of the organoids in culture should be better characterized with time in culture using a number of new markers that are listed in the reviewer reports, as well as additional methods such as immunostaining and flow. Reviewers have raised concerns that organoids are not stable during culture.

We also agree that cell phenotypes and stability of organoids in culture are two very claims that should be rigorously backed up with evidence. We recognize that stability in culture is especially important from a QC standpoint if this particular lung organoid model was going to be used in other studies, by us and others, to unravel disease mechanism(s) and/or for the preclinical drug screening.

Action(s) taken: To address these important issues, we have performed the following new experiments:

A) Immunostaining and Flow cytometry with early and late passaged lung organoids from three different patients (Figure 2—figure supplement 5). The complete lung organoid showed the presence of distal and proximal cell markers and there is not much difference in the percentage of positive cells in these cultures between an early and late passage (shown in the table in Figure 2-figure supplement 5D). For these studies, we have carefully dissociated the lung organoid to single cells and each antibody for the cell markers are optimized following the staining index and using proper isotype control antibody to rule out the non-specific binding.

The strategy for gating of the % positive cells and the corresponding isotype IgG controls are also shown in Figure 2-figure supplement 5.

One issue that is worth noting is that individual markers do not add up to 100% and show a higher percentage which is because it is well known that lung cell types share markers, as shown by others by FACS1. What we can report is that during immunostaining, we tested a serial dilution of antibodies to ensure that only minimal conc. of each antibody were used to detect the cell types (in addition to the isotype controls). Thus, the rigor in the analyses (alongside the IF data showcases earlier in the intact organoids) gives us confidence that what we see is believable.

B) qRT-PCR: We performed qPCR on the organoids and the lung tissue specimens from which they were derived to compare their cell type compositions (Figure 2-figure supplement 3). Our analyses showed that the tissue specimens and the organoids have a comparable amount of cell type markers.

We also performed qPCR studies on all three adult lung organoid lines (ALO1-3) from early (below passage 8) and late (above passage 8) passages and determined that the cell types remain stable in the culture. These data are included as a new figure (Figure 2-figure supplement 4).

C) Immunostaining: We added new staining data as requested by Reviewer #1 (higher resolution images of the Acetylated Tubulin-stained structures indicative of the presence of abundant ciliated cells (Figure 3F; maxprojected Z stack and orthogonal views).

The altered text reads as follows:

“Finally, using qRT-PCR of various cell-type markers as a measure, we confirmed that the ALO models overall recapitulated the cell type composition in the adult lung tissues from which they were derived (Figure 2—figure supplement 3) and retained such composition in later passages without significant notable changes in any particular cell type (Figure 2—figure supplement 4). The mixed proximal and distal cellular composition of the ALO models and their degree of stability during in vitro culture was confirmed also by flow cytometry (Figure 2—figure supplement 5).”

2) Claims should be tampered, in particular related to high throughput drug screening since only few drugs were tested.

It was never our intention to claim that the model’s usefulness was somehow validated for use in HTP drug screens. What we had intended to state was that we successfully optimized the use of ALO monolayers (cell #, plating conditions to achieve an intact monolayer, matrigel coating, timing of infection, cell viability, viral E gene detection and IF and gene expression analyses; etc) for infectivity in 384-well miniaturized format. This itself is an important step that sets us up for conducting HTP screens. It is unfortunate that the way it was written it may have come across that we have somehow done such screen already. We regret that.

Action(s) taken: We found 4 places within the text where the term ‘HTP’ was used. Here is how we have made changes that to tamper down the claim.

On Page 10:

Original sentence: “Findings also provide proof-of-concept that ALO monolayers may serve as effective models for use in HTP therapeutic screen”.

Changed to read: “Findings also provide proof-of-concept that ALO monolayers may be adapted in miniatured formats for use in 384-well plates for high-throughput (HTP) drug screens”.

On Page 14:

Original sentence: “Second, among all the established lung models so far, ours features 4 key properties that are desirable whenever disease models are being considered for their use in HTP modes for rapid screening of candidate therapeutics and vaccines……”

Changes made: None, because the use of HTP in this sentence was referring to desirable properties of any model.

On Page 14:

Original sentence: “Feasibility has also been established for scaling up for use in 384-well HTP assays.”

Changes made: “Feasibility has also been established for scaling up and optimizing the conditions for them to be used in miniaturized 384-well infectivity assays.”

On Page 15:

Original sentence: “Although we provide proof-of-concept studies in low throughput mode demonstrating the usefulness of the ALOs as human pre-clinical models for screening therapeutics in Phase ‘0’ trials, optimization for the same to be adapted in HTP mode was not attempted here.”

Changes made: None. This sentence was meant to accurately describe the study limitation, i.e., that HTP studies were not attempted here.

3) Discrepancies with transcriptomic signatures of the infected human lungs need to be more carefully considered.

In this comment, the Editor has summarized what appears to be a common criticism from both the Reviewers, i.e., the lower number of overlapping genes between human COVID-19-affected lung tissue and our lung organoid monolayers infected with SARS CoV2. To recap, this is specifically referring to the figure panel 5E in the original submission, in which we used a venn diagram to show that UP-regulated DEGs in our model overlap a ~one third of the time (7/22 genes) with UP-regulated DEGs in the infected patient lung, whereas there are no overlaps in DOWN-regulated DEGs.

We never addressed this apparent discrepancy in our original submission, and retrospectively, as regret such an error of omission.

Actions taken:

In this revised submission, we have performed several new analyses of 3 publicly available lung models of COVID-19 (ours, GSE160435, and GSE153218) against the following new patient-derived COVID19 datasets:

1) GSE151764- post-mortem COVID-19 and normal lung tissues

2) GSE156063- upper airway samples from patients with COVID-19

3) GSE145926- epi -- bronchoalveolar lavage fluid (BALF) cells from patients with varying degrees of COVID-19 severity

4) GSE157526-tracheal-bronchial cells infected with SARS-Cov2.

(a new Table of datasets (Table 8) has been included for the convenience of the reviewers and readers)

What did we do?: We set out to compare the DEGs from each of the 3 models against patient-derived datasets. GSE160435 (PMID: 32637946-preprint) is a model in which differentiated air-liquid interface from 3D organoid cultures of the alveolar epithelium were infected with SARS-CoV2. GSE153218 (PMID: 33283287; EMBO J) is 3D lung organoids derived from fetal (16-17 wk) lung bud tips.

What did we find?: As displayed in Figure 5—figure supplement 3:

– Our model: showed 22-54% overlaps in UP-regulated DEGs, and no overlaps with DOWN-regulated DEGs. (Panel A)

– GSE160435: showed 10-25% overlaps in UP-regulated DEGs, and no overlaps with DOWN-regulated DEGs. (Panel B)

– Our model and GSE160435 had a 50% overlap among the UP-regulated DEGs (Panel C)

– GSE153218: showed only 3 DEGs even for p-adj=0.5; lfc=0.0 cut-off values. Out of 17347 with nonzero total read count

adjusted p-value < 0.5

LFC > 0 (up) : 1, 0.0058% == CYP4A11__chr1

LFC < 0 (down) : 2, 0.012% == SARS-CoV-2, SARS1-HKU-39849 outliers [1] : 0, 0% low counts [2] : 0, 0%

(mean count < 0)

In the absence of DEGs our ability compares this third model against human samples was impaired. We also carried out (as requested by Reviewer #1) further analyses on GSE153218, which we have showcased later in the rebuttal (see Response to Reviewer #2, Comment #1). Briefly, this third model did not show the telltale signatures of host immune response to viral infection and hence, in the absence of those signature and absence of significant DEGs, this third model was excluded from the analysis in Figure 5—figure supplement 3.

What do we make of these new findings? We believe that although the epithelial contributions to the host response are important, it alone cannot account for the complete host response because the response of the immune and non-immune stromal cells, and their crosstalk with the epitheliual are missing from these minimalistic single component models. Given that inflammation is propagated by forward feedback loops of multi-compartment crosstalk, we believe that the epithelial signatures induced in vitro are only partially capturing the response that is likely to exist in vivo. Regardless of the missing components, what appears to be the case is that we have a model that recapitulates a one quarter to one half of the UP-regulated genes in COVID-19 despite cohort heterogeneity.

How have we modified the text to reflect these analyses?

In this version of the manuscript we have edited two sections.

On page 11: Results and Discussion section:

“Next, we analyzed the datasets from our ALO monolayers for differentially expressed genes (DEGs) when challenged with SARS-COV-2 (Figure 5A-B). Genes and pathways upregulated in the infected lung organoid-derived monolayer models (Figure 5—figure supplement 1-2) overlapped significantly with those that were upregulated in the COVID-19 lung signature (compare Figure 4C-D with 5C-D, Table 6-7). We observed only a partial overlap (ranging from ~22-55% across various human datasets; Figure 5—figure supplement 3) in upregulated genes and no overlaps among downregulated genes between model and disease (COVID-19) (Figure 5E). Because the degree of overlap was even lesser (ranging from ~10-25% across various human datasets; Figure 5—figure supplement 3) in the case of another publicly released model (GSE160435)2, these discrepancies between model and the actual disease likely reflects the missing stromal and immune components in our organoid monolayers.”

On page 15: Study limitations.

“Limitations of the study

Our adult stem-cell-derived lung organoids, complete with all epithelial cell types, can model COVID-19, but still remains a simplified/rudimentary version compared to the adult human organ. For instance, although the epithelial contributions to the host response are important, it alone cannot account for the response of the immune cells and of the non-immune stromal cells, and their crosstalk with the epithelium. Given that epithelial inflammation and damage is propagated by vicious forward-feedback loops of multicellular crosstalk, it is entirely possible that the epithelial signatures induced in infected ALO-derived monolayers are also only a fraction of the actual epithelial response mounted in vivo. Regardless of the missing components, what appears to be the case is that we already have a model that recapitulates a ¼ th to ½ of the genes that are induced across diverse COVID-19 infected patient samples. This model can be further improved by the simultaneous addition of endothelial cells and immune cells to better understand the pathophysiologic basis for DAD, microangiopathy, and even organizing fibrosis with loss of lung capacity that has been observed in many patients3; these insights should be valuable to fight some of the long-term sequelae of COVID-19.”

Reviewer #1 (Recommendations for the authors):

The manuscript by Tindle et al. describes generation of adult lung organoids (ALO) from human lung biopsies and their use to study the changes in gene expression as a result of SARS-CoV-2 infection. The main advantage of the use of organoids is the ability to generate many cell types that make up the lung. In this particular case the authors report the presence of AT1, AT2 cells, Basal cells, Goblet cells, Ciliated cells and Club cells. The authors were able to cultivate the cells at the air-liquid interface and to establish cultures of predominately proximal and predominately distal airway cells. The main finding is that proximal cells are more prone to viral infection, while distal cells are governing the exuberant inflammatory response, with both cells required for the exuberant response to occur. A useful information provided by the paper is the analysis gene signatures of various cellular models, in comparison to the infected human lung.

We appreciate the accurate recap of the seminal findings and that the reviewer felt the computational analysis of gene signatures between models and disease as an useful information.

1. Although cellular complexity is notable compared to some other models, it is important to more precisely benchmark how this compares to the percentages of cells in the actual lung. Figure 3 shows percentages of cells based on RNA sequencing, however a more precise enumeration might be provided by flow cytometry provided that the cells can be accurately dissociated.

This comment has two parts:

i) In the first part, the reviewer asks for a comparison between ALO models and the parent tissue of origin.

ii) In the second part, the reviewer asks for a better enumeration of cell types and confirmation of mixed cellularity in ALO models after dissociating the organoids and assessing by immunostaining and flow cytometry (if such careful dissociation is feasible). In Figure 3B of the original manuscript we used CIBERSORTX, a machine learning method that extends this framework and infer cell-type-specific gene expression profiles without physical cell isolation. This method depends on the feeding of the data with specific markers that used to designate the cell types. As the lung markers are shared between different cell populations, we expected this computational method will be more corroborative. Also it is well known that lung is complex model and cellular plasticity is a major features4,5. We agree with the reviewer that beyond the computational approach, experimental validation is needed to confirm the mixed cellularity of the organoids.

Actions taken:

Please see response #1 in the Essential revisions.

Briefly, we have carried out what was asked of us, and have added new findings (in a total of 3 figures, Figure 2—figure supplement 3, 4, and 5). These new data (described in detail above on Page #9) show 6 different cell types by qRT-PCR and immunostaining/FACS. For the convenience of the reviewer, the edited text in the revised manuscript is also presented in within the response #1 in this rebuttal document.

2. In general, organoid models lack functional readouts such as permeability or barrier function. Were authors able to establish and enumerate the differences in barrier function in the monolayer model in the transwell system?

We agree that this was an important point, one that we only addressed partially earlier. In the original submission, we had only shown Transepithelial Electrical Resistance (TEER) in ALI monolayers derived from ALOs (Figure S7 in the original submission; currently Figure 3—figure supplement 1E). Hence, epithelial barrier in transwells was not evaluated, which we have now rectified in this revised submission.

Actions taken:

Expts conducted: We have dissociated the organoids to single cells and added to transwells. Following differentiation, we have performed the functional readout of barrier function as measured by the Transepithelial Electrical Resistance (TEER) and the permeability using the FITC-Dextran (10 kD). We have added the new TEER data from two different lung monolayer models in Figure 3-figure supplement 1B and the permeability using FITC-dextran in the same monolayers in Figure 3-figure supplement 1C. The impact of LPS on the TEER was also studied (Figure 3- figure supplement 1K-L).

Key Findings:

– Our data with TEER and permeability correlate with each other, and both show that the ALO-derived submerged monolayers can form an epithelial barrier.

– But these submerged monolayers were leakier than ALI models derived from the same organoids (which was expected) and compared to previously published TEER (400-1000 ohm-cm2) of human bronchial epithelial cells (NHBE), as shown in the published literature6.

– Occludin was visualized in patches, despite intactness of the monolayer (as determined by Phalloidin), indicative of leaky areas. Figure 3—figure supplement 1F-G. We chose to stain for Occludin because it is an important regulator of tight junction stability and function, is under the transcriptional control of TTF1/NKX2.17. While there are numerous types of Claudins in the lung epithelial cells (Claudins 1-3-4-5-78-10), Occludin, however, is a more shared and constant marker throughout the airway whose role is to stabilize claudins and regulate their turnover8.

Interpretation: The submerged ALO-derived monolayers were able to form an epithelial barrier; it is however, leaky. We believe that this leakiness is likely due to the dynamic changes in proximal-distal cell ratios as ALO3D organoids and differentiated into ALO-monolayers and the persistence of progenitors at various stages of differentiation to AT1 cells. Our findings are in keeping with prior work demonstrating that as columnar wedge-shaped progenitors flatten to become AT1 cells, apical tight junctions (TJ) are maintained whereas lateral junctions are lost9. As alveolar differentiation takes place in the submerged monolayers, we not only expect shifting cellular ratios, but also changes in the types of Claudins. For example, some Claudins make the barrier tighter (Claudin-18) and others than make the barrier leakier (Claudins-3/4)10, Some of the claudins are alv specific (Claudin -18) whereas others are present in the bronchial passage (Claudin-1) and some that are present throughout (Claudin-4/7). We speculate that claudin-3 being more abundant on AT2 than AT1 cells and increases the alveolar permeability, and the permeability of our ALO-derived submerged monolayers.

Changes made to the manuscript text: To succinctly described the findings, and yet demonstrate restraint in avoiding speculative statements and/or not destroy the flow of the manuscript or distract readers with extensive review of the literature, we have made the following edits in the manuscript:

The epithelial barrier was leakier, as determined by relatively lower trans-epithelial electrical resistance (TEER; Figure 3—figure supplement 1B) and the flux of FITC-dextran from apical to basolateral chambers (Figure 3—figure supplement 1C), and corroborated by morphological assessment by confocal immunofluorescence of localization of occludin, a bona-fide TJ marker. We chose occludin because it is a shared and constant marker throughout the airway that stabilizes claudins and regulates their turnover8 and plays an important role in maintaining the integrity of the lung epithelial barrier11. Junction-localized occludin was patchy in the monolayer, despite the fact that the monolayer was otherwise intact, as determined by phalloidin staining (Figure 3- figure supplement 1H-I). Our finding, that ALO 3D organoids differentiating into monolayers in submerged cultures (where alveolar differentiation and cell-flattening happens dynamically as progenitor cells give rise to AT1/2 cells) are leaky is in keeping with prior work demonstrating that the TJs are rapidly remodeled as alveolar cells mature9,10. By contrast, and as expected6, the ALI-monolayers formed a more effective epithelial barrier, as determined by TEER (Figure 3—figure supplement 1F) and appeared to be progressively hazier with time after air-lift, likely due to the accumulation of secreted mucin (Figure 3—figure supplement 1G).”

3. It is slightly discouraging that the model captures only 7 upregulated genes of the 76 reported in covid19 patient lungs and no downregulated genes. It also upregulates 15 genes that are not reported in the patient lungs (Figure 5E). These discrepancies should be better discussed.

We agree that this is an important point. Because this was a major “Essential Revisions” request from the Editors, to avoid duplication of text, we kindly refer the reviewer to our Response # 3 in the Essential revisons.

Reviewer #2 (Recommendations for the authors):

The manuscript "Adult Stem Cell-derived Complete Lung Organoid Models Emulate Lung Disease in COVID-19" by Das and colleagues introduces a new model system of airway epithelium derived from adult lung organoids (ALO) to be utilised for the study of COVID-19-related processes. In this manuscript two main novelties are claimed: the development of a new model system which represents both proximal as distal airway epithelium and a computationally acquired gene signature that identifies SARS-CoV-2-infected individuals. While interesting data are presented, the novelty claim is questionable and the data is not always convincing.

1. The manuscript claims a novel model system of distal and proximal airway epithelium. The authors however fail to discuss a recently published COVID-organoid study which reports similar observations. While their novelty claim could still hold in some areas, the authors do need to discuss the following manuscript (https://doi.org/10.15252/embj.2020105912).

Within just weeks of submission of our paper, we ourselves recognized that the Table 1 (which listed all existing models and compared ours against them) was incomplete, because a new paper describing new models emerged every almost every other week. Consequently, we had no way of discussing these newer publications/citing them in our paper. The manuscript in question12 (Lamers et al., EMBO J, 2021) is one of them because it was published after our work was submitted, and hence, we had missed citing it.

Actions taken:

In this revised version of our manuscript, we have now cited this work and also added it (and 3 other models, as requested by Reviewer #3) to an updated Table 1.

As for the major differences between ours and the model described by Lamers et al., EMBO J (2021)40:e105912 are the following:

i) In their system, culture conditions were established to support long-term self-renewal of multipotent sox2+SOX9+ lung bud tip progenitor cells, which in vivo differentiate into both airway and alveolar cells. These lung bud tip organoids (LBT) are from canalicular stage human fetal lungs 16–17 wk postconception weeks and two such lines were used to establish the model. By contrast, our lung organoid model is from the adult lung specimens collected from 3 different donors that represent both the genders, smokers and non-smokers as illustrated in Figure 2 Figure supplement 1A. These fetal vs adult tissue source of stem cells is a key distinguishing feature that we have listed in a revised Table 1.

ii) The bronchio-alveolar like model presented here shows a robust increase in infectious virus titers over time of ~ 5 logs, whereas other recently developed adult-derived 3D alveolar organoid models show a more limited increase in viral titers (~ 1–2 logs). Our model, consistent with other adult models, has a limited increase in viral titers.

iii) As mentioned in our Response # 3 to the “Essential revisions” of this rebuttal, we did not see significant DEGs induced/repressed in the brochoalveolar lung datasets infected or not with SARS-CoV-2. The raw metrics for this analysis is presented above. This is unusual given that all other publicly released datasets show gene expression changes that have a varying degree of overlaps with numerous patient-derived samples. We addressed this issue in great detail in our Response #3. To avoid duplication of entire text, we kindly refer the reviewer to that section.

iv) In the absence of DEGs, which impaired our ability to proceed with cross-comparison against human samples, we went ahead and conducted additional analyses to see if the fetal lung derived model that is proficient in viral infectivity and permissive to massive replication may do so because it lacks significant host immune response. In fact, that is exactly what we found. Analyses of the dataset showed that the feta lung organoid-derived bronchoalveolar monolayers failed to mount the host immune response. This is rather unusual, because this signature has been widely validated prospectively in all adult COVID-19 human samples that have emerged to date, i.e., a total of 727 samples (Sahoo et al., eBioMedicine, 2021)13. We have included these new analyses in Figure 6H and edited the legend to reflect this update.

It is important for us to clarify that in the same work (Lamers et al., EMBO J (2021) 40:e105912), small airway epithelial cells were infected (as positive controls). We analyzed these datasets and found that they showed an induction of the ViP signatures. This data is presented Author response image 1, and not included in the manuscript.

Author response image 1
Publicly available RNA seq datasets (GSE153218) from Small Airway Epi (SAEp) monolayers12 infected or not with SARS-CoV-2 were analyzed for the ability of ViP signatures to classify infected (Inf) from uninfected (Uninf) samples.

ROC AUC indicate the performance of a classification model using the ViP signatures. Unlike the brochoalveolar monolayers (see Figure 6H in the revised manuscript) derived from fetal lung organoids in the same work, SAEp monolayers successfully induced the ViP signatures because the signatures were induced in infected monolayers.

Interpretation: From these new analyses we conclude that the fetal lung organoids are more permissive to viral infection and replication, but do not mount the host immune response that is seen in COVID-19, which largely affects adults.

Changes in the revised manuscript: several changes were made. For the convenience of the reviewer we have copied and pasted the text (annotated with page #).

New Figure 6H and legend: New panel added showing the 166-gene ViP signature and 20-gene sViP signature based classification of infected vs uninfected samples.

On Page #12-13:

“Our lung models showed that both the 166- and 20-gene ViP signatures were induced significantly in the submerged ALO-derived monolayers that had distal differentiation (Figure 6E; left), but not in the proximal-mimic ALI model (Figure 6E; right). Neither signatures were induced in monolayers of small airway epithelial cells (Figure 6F) or hiPSC-derived AT2 cells (Figure 6G). Finally, we analyzed a recently published lung organoid model that supports robust SARS-CoV-2 infection; this model was generated using multipotent sox2+SOX9+ lung bud tip (LBT) progenitor cells that were isolated from the canalicular stage of human fetal lungs (~16–17 wk post-conception)12. Despite mixed cellularity (proximal and distal), this fetal lung organoid model failed to induce the ViP signatures (Fig 6H). These findings indicate that despite having mixed cellular composition and the added advantage of being permissive to robust viral replication (achieving ~ 5 log-fold increase in titers), the model lacks the signature host response that is seen in all human samples of COVID-1913.”

On Page #15:

“Third, the value of the ALO models is further enhanced due to the availability of companion readouts/ biomarkers (e.g., ViP signatures in the case of respiratory viral pandemics, or monitoring the E gene, or viral shedding, etc.) that can rapidly and objectively vet treatment efficacy based on set therapeutic goals. Of these readouts, the host response, as assessed by ViP signatures, is a key vantage point because an overzealous host response is what is known to cause fatality. Recently, a systematic review of the existing pre-clinical animal models revealed that most of the animal models of COVID-19 recapitulated mild patterns of human COVID-19; no severe illness associated with mortality was observed, suggesting a wide gap between COVID-19 in humans 3 and animal models 14. It is noteworthy that alternative models that effectively support viral replication, such as the proximal airway epithelium or iPSC-derived AT2 cells (analyzed in this work) or a fetal lung bud tipderived organoid model recently described by others12, do not recapitulate the host response in COVID-19. The model revealed here, in conjunction with the ViP signatures described earlier 15, could serve as a pre-clinical model with companion diagnostics to identify drugs that target both the viral and host response in pandemics.”

In conclusion, we hope that we have discussed the two models accurately and presented concrete evidence to show how the adult lung organoid model described here is novel or different from the fetal model described by Lamers et al. EMBO J, 2021.

2. Expression patterns in Figure 2B are hard to interpret when no tissue control is used. A positive control of lung tissue should be used to make valid conclusions.

We agree.

Actions taken: In this revised submission we performed qPCR on the organoids and the lung tissue specimens from which they were derived to compare their cell type compositions (Figure 2-figure supplement 3). Our analyses showed that the tissue specimens and the organoids have a comparable amount of cell type markers.

3. The authors claim the presence of multiple cell types within single ALO. Data are not strong.

a. Co-staining of KRT5 and SFTPB (as shown in figure 2J) is difficult to explain, since these markers are expressed in two different cell types. The lower panel in the same figure does show two separate populations, but the images shown in the second panel do not.

We agree that the way the images are presented, it can be confusing. For the convenience of the reviewer(s) and the editors, we have pasted the original figure in question within this rebuttal (please see Author response image 2). While the lower panel was meant to show more cell type specific organization within a single organoid structure, the upper panel was meant to show that the cell types mixed/interleaved with each other. We agree that although there are distinct red and green nonoverlapping cell types, in some areas (due to thickness of the cut and max projected zstacks), the merged panel shows ‘yellow’ pixels that make it look like KRT5 and SFTPB are overlapping with each other. The second interpretation can be a very unfortunate conclusion because it raises concerns about antibody specificity, controls, technical rigor, and overall concerns regarding one of the central claims (i.e., mixed cellularity of the organoids).

Author response image 2

Within this rebuttal we have provided additional examples (Author response image 3), one of which was used to replace the upper panel:

Author response image 3

b. While the authors claim the presence of proximal and distal cells in ALOs, the images show 100% of SFTPC+ cells or 100% KRT5+ cells or 100% Na/K-ATPase cells. This rather shows two different types of ALO within one culture, then mixed cell populations within a single ALO.

This is indeed an important point, which we addressed, but incompletely. To be clear, we never intended to claim that each organoid structure in each ALO line has all cell components. That was simply not our intention, and if that is how it came across, we regret the description presented in the original submission. We simply wanted to state that both proximal and distal airway components are present in the same line of organoids in culture; at times the structures were heterogeneous, at times they were homogeneous in their cell composition. For example, in this particular field of SFTPB/KRT5 stained structures, we see a single large organoid with heterogeneous cellularity, and two smaller structures that are almost exclusively and homogeneously comprised of one or the other (arrows in Author response image 4).

Author response image 4

Figure 2J: the example showcased here was usedOn page #7: The text was updated to explicitly discuss/clarify this issue. For the convenience of the reviewer we have copied and pasted the text.

“The presence of all cell types was also confirmed by assessing protein expression of various cell types within organoids grown in 3D cultures. Two different approaches were used—(i) slices cut from FFPE cell blocks of HistoGel-embedded ALO lines (Figure 2I-J) or (ii) ALO lines grown in 8-well chamber slides were fixed in matrigel (Figure 2K), stained, and assessed by confocal microscopy. Such staining not only confirmed the presence of all cell types in each ALO line but also demonstrated the presence of more than one cell type (i.e., mixed cellularity) of proximal (basal-KRT5) and distal (AT1/AT2 markers) within the same organoid structure. For example, AT2 and basal cells, marked by SFTPB and KRT5, respectively, were found in the same 3D-structure (Figure 2J, interrupted curved lines). Similarly, ciliated cells and goblet cells stained by Ac-Tub and Muc5, respectively, were found to coexist within the same structure (Figure 2J, interrupted box; Figure 2K, arrow). Besides the organoids with heterogeneous makeup, each ALO line also showed homotypic organoid structures that were relatively enriched in one cell type (Figure 2J, arrowheads pointing to two adjacent structures that are either KRT5- or SFTPB-positive). Regardless of their homotypic or heterotypic cellular organization into 3D-structures, the presence of mixed cellularity was documented in all three ALO lines (see multiple additional examples in Figure 2—figure supplement 2I).”

c. The authors claim the presence of ciliated cells by staining of acetylated α tubulin in figure 2K-J and figure 3E. The images shown however are hardly showing specific ciliated cell staining and more importantly the cilia present on the cell membrane. The authors should include much higher quality images that clearly show cilia or other markers which are exclusive for ciliated cells.

We agree that it is incredibly difficult to appreciate finer structures when the original images are reduced to post-stamp sized panels to assemble into figures, and the figures had to be reduced in size for uploading.

Actions taken: We have performed new experiments and images of higher quality image of the cilia is now added in Figure 3F.

d. In figure 3H, the authors show MUC5B positive cells in their monolayer cultures. These goblet cells seem to comprise quite a significant proportion of the culture. Their graph in 3B however shows no goblet cell contribution.

The claim that ALO consists of a mix of defined cell types is essentially based on deconvoluted bulk RNAseq data. Such a claim is not conclusive.

The reviewer brings up an important point that there is an apparent discrepancy between Figure 3B (which is RNA seq-based data) and Figure 3E (we believe that the reviewer meant 3E, because 3H is infectivity graphs), which is immunofluorescence staining. While the RNA Seq data did not show much “goblet” cell fraction, the immunostaining approach showed significant amount. This is an unfortunate outcome of trying to provide multiple lines of evidence through diverse approaches, but not clarifying what might be the strengths and weaknesses of each approach.

Limitations of RNA seq-based claims in Figure 3B: CIBERSORTx is a machine learning method that extends this framework and infer cell-type-specific gene expression profiles without physical cell isolation. This method depends on the feeding of the data with specific markers that are used to designate the cell types. Because many lung cell markers are shared between different cell types, we expect this computational method will be more of a corroborative evidence and not the overriding evidence. For example, CIBERSORTX has indicated that % cellularity in submerged has mostly AT1/2, and that corroborates with the IF panels.

Alternative explanations: Because 3B and 3E compare mRNA and protein markers of goblet cells, it is possible that the mRNA and protein abundance do not track each other well when it comes to those markers. Alternatively, it is possible that the expression level of the markers associated with goblet cells are low compared to the other cell types; therefore the goblet cell fraction is falsely underrepresented in the CYBERSORTx analyses.

Actions taken: We understand that rigorous and complementary approaches are better in both transcriptional and translational level. In this revised submission we added additional evidence (qRT-PCR, IF and FACS), as discussed in the Response #1 of “Essential Revisions” to determine different cellular proportion present in the lung organoid.

4. Authors claim that ALO maintain cell type ratios similar to lung tissue over several passages. Their data in figure 3B however show a drastic change in the composition of the ALOs. The shift from AT2 to AT1 is explained to occur when the ALO are cultured as submerged monolayers. However, the shift is already visible at later passages in 'standard' ALO culture. Moreover, when comparing cell types between passage 1 and 8, the composition changes dramatically. The cultures have lost basal cells, ALO1 has lost goblet cells, ALO2 has lost ciliated cells. In addition, ALO1 composition differs from ALO2 at passage 2 which doesn't support the statement that the cultures are stable/robust.

We agree. There are two things worth mentioning in this regard. In the original submission, viral entry markers (VEMs; e.g., ACE2, TMPRSS2, etc) were used as a part of the CYBERSPRTx analysis, which has been shown to be present on many cell types. This caused the cellular proportion analysis by CYBERSORTx less useful or interpretable. Second, as we have highlighted above, CYBERSORTx is imperfect; it was meant to be used as a corroborating evidence, but not the evidence.

Actions taken: The reason is, as the reviewers have themselves pointed out, this is a poor-man’s way to obtain cellularity information in the bulk transcriptome. The evidence that there is mixed cellularity in each line is drawn from – IF and qPCR studies. In this revised submission, we have now provided additional evidence about the nature and extent of drift for ALO 1-2-3 from early passages (1-8) and late passages (above 8) in Figure 2-figure supplement 3. We also compared the organoids with the tissue specimens to compare the cell types (Figure 2-figure supplement 3).

5. While the authors claim that proximal airway cultures are infected and responsible for maintaining virus replication, there are no virus entry marker-annotated cells in the composition graph of figure 1B.

We are unclear on what the reviewer is referring to in this comment and what he asks to see as part of revisions. Figure 1A is a box plot of the abundance of ACE2 (the viral entry marker) among various lung epithelial cell types. Because AT2 cells had high ACE2 transcripts, we then went on to show in human lungs (Figure 1B) the presence of SFTPC+vs AT2 cells in healthy lungs (top) and SARS-CoV-2 laden AT2 cells in the lung tissue of a patient with fatal COVID-19. If by viral entry marker in 1B the reviewer asks to see ACE2, we did not do that because there have been far too many groups who have shown ACE2+ve AT2 cells.

Alternatively, it is possible that the reviewer meant to refer to Figure 3B, but by mistake typed Figure 1B. If that is so, he might be wondering why viral entry marker (VEM)-annotated cells were not seen in the CYBERSORTx graphs. We can clarify why. As shown in Table 2 (that accompanied Figure 3B and listed the cell type markers used in the RNA Seq analyses), we used many more markers to ‘gate’ AT2 and other cell types, all of which carry VEMs. Hence, the proportion of VEM-containing cell types was all accounted for within the % distribution of all other cell types.

Actions taken: During this revised submission, we have corrected an error that we made in our judgement in the initial submission. We gated the RNA se data with lung cell type markers, and also included VEM as another ‘cell type’, which, retrospectively was not the right thing to do. We have re-done the CYBERSORTx analysis and replaced Figure 3B after excluding VEMs. For the convenience of the reviewers, we present here the original and the new versions side by side.

6. Similar to point 3, the authors claim the development of the first combined organoid culture of proximal and distal cell types, as supported by deconvoluted bulk RNAseq. The hiPSC-derived AT2 cultures however also show the presence of both distal and proximal cell types.

The fact that iPSC-derived AT2 cells were found to have other cell types markers by CYBERSORTx is not unexpected because it is well known that iPSC-derived AT2 be differentiated to other cell types, including AT1 and ciliated and club cells16-18. and/or express other markers that are commonly found on clara/club cells19.

7. Figure 3H shows a graph presenting the infectivity relative to peak. This axis makes comparison of infectivity impossible. The authors may want to include suppl Figure 5A-B as main figure and exclude 3H from the manuscript. Moreover, viral E-gene qPCR should be complemented by viral titer measurements to verify findings for live virus. The analysis of SARS-CoV-2 infectivity is very limited when only showing a few infected cells and viral E-gene graphs. Suppl Figure 5C also only shows one or two infected cells in all samples which does convince as proof of infectivity.

This comment has several parts, all of which relate to the veracity of the proof of infectivity, which in turn reflects the utility of the model to serve as a platform for screening drugs. We agree that this is an important point.

In the original submission, we provided the following proofs of infectivity and appropriate host response to the same:

1) Direct visualization of infectivity and examples of cellular cytopathic changes, evidence of intracellular packaging of viral particles (two montages of IF images in Figure 3G and Figure 3—figure supplement 2A).

2) E gene amplification by qPCR, at various time points after infection: Figure 3H and Figure 3—figure supplement 2B-C.

3) Impact of directly acting a previously confirmed anti-viral agent with anti-SARS-CoV-2 activity: Figure 3I.

4) Recapitulation of host response (gene signature induction) upon SARS-CoV-2 infection, where most other models described to date, fail or underperform: Figures 4-6.

The reviewer asks us to remove Figure 3H from the manuscript. His rationale for that suggestion is that this figure does not allow one to compare infectivity. We agree.

The reviewer believes that the Figure 3—figure supplement 3B-C is instead much more informative for comparing infectivity. We Agree.

We chose to show the sustained nature of the It was not our intention to highlight the degree of infectivity, rather the sustained nature of infectivity because one of the major distinguishing features (and, hence, in our mind, a novelty) of our model is that it recapitulated the nature of the host response to infection that is observed in human airway/lung samples and in numerous cohorts of patients with COVID-19 (> 700 datasets, which were used to prospectively validate the ViP signatures).

As mentioned above (Response #1 to this reviewer), when we compared our adult lung organoid model against the fetal lung organoid model (Lamers et al., EMBO J. Mar 2021), modeling COVID-19 not just requires permissiveness to viral infection, but infectivity with proportionate host immune response.

Actions taken:

We would like to retain Figure 3H in the main figure because we believe that the sustained infectivity of the model is a major aspect of modeling COVID-19. If the reviewer feels very strongly, and insists we change it, we can reconsider. We have, however, modified the legend for Figure 3H and refer readers to Figure 3—figure supplement 3B-C for comparing the peak viral amplification across various models. Because eLife allows supplementary figures to be interleaved within the main figures, should this manuscript be accepted for publication, we believe that the readers will readily have access to all data, regardless of where we insert it.

8. The authors provide data for a single drug to indicate the possibility of high-throughput screening of drugs for COVID-19 treatment. This does not say much about the throughput of the assay.

We agree. It was never our intention to claim that the model’s usefulness was somehow validated for use in HTP drug screens. What we had intended to state was that we successfully optimized the use of ALO monolayers (cell #, plating conditions to achieve an intact monolayer, matrigel coating, timing of infection, cell viability, viral E gene detection and IF and gene expression analyses; etc) for infectivity in 384 well miniaturized format. This itself is an important step that sets us up for conducting HTP screens. It is unfortunate that the way it was written it may have come across that we have somehow done such screen already. We regret that.

Actions taken: There were a total of 4 places in the manuscript where we used the term HTP. As part of Response #2 to “Essential Revisions”, we have eliminated and/or clarified each sentence. To avoid duplication of text, we kindly refer the reviewer to refer to this section in our Response #2 to Editors.

9. Figure 5C compares uninfected and infected monolayer cultures for genes that are identified in the patient cohort. While the authors claim comparable patterns between the cultures and the patients, the heatmap can not be directly read. The authors should include more details in the legends and combine the heatmaps in 4C and 5C for direct comparison. Currently, the colours represent raw z-scores which do not indicate transcript read numbers but relative differences of the expression of the indicated genes in the samples analysed. These numbers could differ extensively between organoids and patients.

We found this comment/suggestion difficult to mitigate, and perhaps reflects our inability to clearly explain what was done and why. Figure 4 and Figure 5 are dedicated to cross-validation of the actual COVID19 disease versus various SARS-CoV-2-infected in vitro models of the same (including our newly developed ALO model). As is displayed through workflow schematics in Figure 4, we first extract DEGs from the actual human disease (lungs of healthy vs COVID-19 patients; 4A-C) and tests the conservation of those gene sets in other human datasets (4E) before moving on to testing their ability to classify infected vs uninfected in vitro models (4F-G). In Figure 5, we extract DEGs from the infected vs uninfected ALO-model (5A-C) prior to their use to classify numerous human diseased samples (5F-I).

Thus, Figure 4C and Figure 5C represent heatmaps of DEGs in two different datasets from different technological platforms (Ion Torrent and Illumina NovaSeq). Also, read numbers are not comparable across these two datasets because of different tissue types and different RNASeq processing pipelines ( e.g., two different genome builds were used as a reference for each dataset, our organoid Model used hg38 while the other one used hg19). Therefore, we cannot combine the heatmap of Figure 4C and 5C.

Actions taken: We have inserted an additional sentence on Page #10 in “results” section to clarify what was being done. For the convenience of the reviewer, we have copied and pasted that edited piece of text below and indicated with highlight the newly added sentence. We have also edited the legends for figure 4 and 5 to improve clarity. We hope the reviewer finds it a bit more accessible

“Next, we asked if the newly generated lung models accurately recapitulate the host immune response in COVID-19. To this end, we analyzed the infected ALO monolayers (both the submerged and ALI variants) as well as the airway epithelial (HSAEpC) and AT2 monolayers by RNA seq and compared them all against the transcriptome profile of lungs from deceased COVID-19 patients. We did this analysis in two steps of reciprocal comparisons: (i) First, the actual human disease-derived gene signature was assessed for its ability to distinguish infected from uninfected disease models (in Figure 4). (ii) Second, the ALO model-derived gene signature was assessed for its ability to distinguish healthy from diseased patient samples (in Figure 5).”

10. In addition to point 9, the authors show that there is very limited overlap between significantly differentially expressed genes in monolayers and patients. While the authors believe this is mostly due to the lack of mesenchymal and immune components, this indicates that the monolayer itself is not important for the observed infection signature. This makes the claim that the monolayers represent infectivity in patients questionable.

We agree that this is a very important point, raised also by other reviewers, and was one of the important “Essential Revisions” suggested by the Editors.

Action taken: We have now extensively addressed this very important issue with the addition of new analyses and numerous figure panels. To avoid duplication of large passage of text with the rebuttal, we kindly refer the reviewer to Response #3 to Editor’s list of “Essential Revisions”.

11. Description in the discussion of growth factors supplied in the medium like FGF7 and FGF10 are not novel. Sachs et al. 2019 (https://doi.org/10.15252/embj.2018100300) already described these growth factors in the culture medium of airway organoids.

We agree. It was never our intention to stake claims that our media composition is novel. We have cited the Sachs et al. 2019 paper. Our media is a modified version of the media mentioned here and all the specifics are already added in “Methods”.

12. In the discussion, the authors describe the advantages of their model system including reproducibility, retainment of genetics and use for infection studies. These claims are however not novel since previous papers from multiple groups have reported similar organoid-based models for SARS-CoV-2 research.

In our updated Table 1 of this revised submission, we have tried to cite everything that has been tried to date and released publicly, and analyzed any transcriptomic datasets that were publicly available for each of those models. The reviewer is right that there are several papers that describe the development of organoid models, either iPSC-derived or adult stem-cell derived. We highlight in the Table and in the Discussion section (which has been modified to reflect the same) that the current model is unique and novel due to two major points.

i) The presence of the major cell types of proximal and distal region of adult lung.

ii) The infection with SARS-CoV-2 induces the expected host immune response that is observed in the actual disease.

These features, and the fact that they can be adapted easily to miniaturized 384-well formats for infection and drug Rx make them promising models for semi-HTP drug screening. As showcased in our response to Editors’ general comments, others groups have already independently used the model and reproduced its ability to serve as disease model.

Reviewer #3 (Recommendations for the authors):

In order to further improve this manuscript we recommend that the authors address the following points:

Figure 2.

1) In Figure 2H. the authors use NGFR as a generic stem cell marker. However, previous publications have shown that NGFR is a basal cell marker (Rock et al., 2009). We suggest that the authors to clarify what is the stem cell population they are attempting to mark here.

We understand the concern that NGFR is a generic stem cell marker and specially for human oral keratinocyte stem/progenitor cell/epidermal stem cell and reported in the human cornea and epidermis. Previous report in lung has shown that following cell damage by chemical agents (naphthalene) or viral infection there are rapid changes in the proliferation of the basal cells and these cells quickly regenerate the epithelium and restore barrier function. Some of the Club secretory cells can undergo reprogramming to become Krt5+ Trp63+ basal cells and can function as stem cells in vivo26,27.

Another report has stated that, a small subpopulation of Scgb1a1+ Club cells in the distal bronchioles can also express Sftpc28 (known as dual-positive, bronchioalveolar stem cells or BASCs) using factors made by lung endothelial cells29,30. BASCs have the potential to differentiate into both AEC2s and airway cells.

Actions taken: In this revised version of the manuscript, we have performed new experiments with our organoid model and compared the levels of expression of Scgb1a1, Sftpc, and Trp63/TP63, added in Figure 2—figure supplement 3-4.

2) In all of Figure 2, but particularly, Figure 2K and 2J, we recommend the authors specify which passage number these organoids are from and show some simple quantification about different lineages to help the readers understand if different lineage remain stable between early and late passages.

For Figure 2K and 2J we have used cells from passages #3-6.

But, we agree with the other two reviewers and the editors that the characterization of cell composition stability of ALOs from early to late passages is something that needed to be more formally assessed and addressed in the manuscript, beyond just clarifying which figure used what passage of organoids.

Actions taken: In this revised manuscript we have performed and added new experiments assessing early and late passages of ALO1-2-3 by qRT-PCR (for cell type gene expression) and by immunostaining followed by Flowcytometry (cell marker protein expression). To avoid duplication of large text passages within this rebuttal, we kindly refer this reviewer to our detailed response to the Response #1 to Editor’s list of Essential Reviews. We conclude that lung organoids from early passages # 3-8 are not significantly different from late passages #9-15.

3) None of the organoid immunofluorescence images in Figure 2 demonstrate that individual organoids contain both airway and alveolar lineages. In particular, SFTPB is presented as a alveolar type 2 cell marker, but it is well-established that this protein is also expressed in club cells (see: https://hlca.ds.czbiohub.org). It is not therefore useful for co-staining with basal cell markers to establish that individual organoids contain both airway and alveolar cell lineages. Mixed airway and alveolar organoid lineage is not yet convincingly demonstrated.

Whether or not each individual 3D organoid structure has mixed cell types is a claim that we agree that we are not in any position to make. Part of that is because, as the reviewer states, lung epithelial cells are notoriously known to share markers among cell types. There is no good way to tell.

What we can convincing state, and have added many supporting evidence in this revised submission, is that organoids are either homotypic or heterotypic in composition (see Figure 2J, second row from the top). We have addressed this particular issue in our Response #3a-b to Reviewer #2 of this rebuttal. We have also carried out cell type composition studies (qPCR and FACS) (Figure 2—figure supplement 3-5).

Regardless of whether the 3D structures assemble into homogeneous or heterogeneous cell types, what is clear is that each ALO line overall has both proximal and distal cell types. Thus, while we agree with the premise of the critique, but respectfully disagree with the last statement.

4) The Ac-Tub staining is not convincing as the cilia structures are not visible at the magnification presented.

Agree. New Figure 3F has been added with higher magnification.

5) The authors need to comment on the fate of the CC10/SFTPC double positive cell shown in Figure 2J bottom panel. Previous efforts to find SFTPC/CC10 co-expressing cells in human lungs in vivo have been unsuccessful (See: https://hlca.ds.czbiohub.org).

We thank the reviewer for pointing out this error of omission, in that we had not addressed what these structures are. We believe that the SFTPC/CC10 double stained structure represent multipotent stem cells termed bronchioalveolar stem cells (BASCs) which have been found to be located at the bronchioalveolar-duct junctions (BADJs)31,32. BASCs coexpress club cell maker secretoglobin 1a1 (Scgb1a1 or CC10) and AT2 cell maker surfactant protein C (Sftpc or SPC).

The reviewer points to a github CZI link that appears to be a single cell RNA Seq based lung cell atlas. In doing so, he/she is requesting that the failure to detect CC10/SFTPC double positive (for RNA expression) cells in single cell sequencing studies of the human lung be addressed in light of our organoids showing the double-stained (for protein) 3D lung organoid structures in 2J. Besides the obvious differences in comparing such distinct approaches and samples, it is important to note that single cell studies have significant limitations too (high ‘dropouts’ due to degrees of sparsity and technical limitations). As highlighted and summarized in this leading biotechnology article single cell seq studies are still evolving and must get better.

Actions taken: We have inserted a sentence in the “results” section on Page #7 about the double-stained structures. While we appreciate that the reviewer pointed out the scSeq studies, we decline to comment on the fact that they did not find such BASCs. We felt that going into such comparison of two techniques, looking at RNA vs protein would confuse and distract the readers (or give the notion that what we see is somehow contradictory) in an otherwise descriptive study.

6) We recommend the authors also perform some simple polarity marker gene staining, such as ECAD and ZO-1, to confirm the apical-basal polarity of the organoids. This would help to justify the need of generating monolayer of epithelium for COVID infection in following experiments.

We agree. Reviewer #1 also asked a similar question. In order to avoid duplicating large chunks of text, we kindly refer the reviewer to Response to Comment #2 from Reviewer #1. We detailed in our response the experiments conducted during revision, our findings and interpretations, our choice of markers and the actions taken (revision to text and figures).

Figure 3.

1) The authors need to discuss in Figure 3B the extent of the variation of cell lineages observed in the self-renewing 3-D organoid cultures between passages 1 and 8. These changes suggest that the organoids do not have a stable cellular phenotype over time and cast serious doubt on their ability to reproducibly model lung infections, or other diseases. Similarly, in the submerged cultures do the 5 different bars represent the results of 5 different experiments using ALO1? Which passage was transferred to 2D culture? And can the authors comment on the reproducibility of their data?

We agree that stability of organoids in culture conditions through later passages is an important point. To answer this reviewer’s question directly, the 5 sequenced samples in Figure 3B represent ALO1-2, passages #3-6. 2D cultures were carried out at the time of manuscript submission from passages #3-8, but since then we have characterized passages #1-15.

Actions taken: In this revised submission, we have carried out numerous studies and added them to the manuscript to address this issue squarely through dedicated figures. More specifically, the stability of cellular composition in long term culture was assessed using two methods (qPCR and FACS) spanning early (1-8) and late (9-15) passages (Figure 2—figure supplement 4-5). We hope that this additional evidence helps mitigate any concerns this reviewer has about the drift of ALOs during long term culture.

2) For the same figure, the authors need to explain what are the cell populations called the 'general lung lineage (GLL)' and the 'viral entry marker (VEM)'. Does the VEM population include AT2, club cells and ciliated cells?

We understand reviewer's concern. This was a shared concern of at least one another reviewer (Reviewer #2). We agree that including VEM in the gating strategy was a bad idea because these are not cell types and were likely skewing the overall results because the fraction would include AT2 and club cells and other cells that have some of these VEMs.

Actions taken: In the current version we have simplified the gating strategy and added only the 6 cell types (basal, AT1, AT2, cilia, club cells, goblet cells), that is the focus of the current paper and removed the GLL and VEM markers.

3) Figure 3E and Figure 3B (middle panel) show quite contradictory results regarding different proportion of cell lineages existing in the organoid system. Figure 3B middle panel seems to suggest a dominant AT1 and airway lineage, but few AT2 lineage cells. However, in Figure 3E, the authors seemed to suggest AT2 is also quite prevailing. Is this another example of variability?

We recognize (based on the comments from other reviewers and this reviewer) that the CYBRSORTx analysis from the RNA seq has limitations; it is, afterall, a machine learning approach that produces data based on our knowhow of what lung cell markers are ‘fed’ into the application as a starting point. Therefore, CYBERSORTx is, at best, more of a corroborative evidence, but not ‘the’ evidence.

Actions taken: In this revised submission, we have carried out numerous other studies and added them to the manuscript to support with evidence the following 3 claims:

1) Multicellular composition of ALO lines (qPCR and FACS)- Figure 2—figure supplement 3-5.

2) Comparison of ALO lines against the adult lung tissue from which they originated (qPCR)- Figure 2—figure supplement 3.

3) Stability of cellular composition in long term culture (qPCR and FACS)-- Figure 2—figure supplement 4-5.

4) IF stained 3D organoids with evidence of heterogeneous cellular composition in single 3D structures: Figure 2J. Figure 2—figure supplement 2.

5) IF stained 2D monolayers with evidence of heterogeneous cellular composition in monolayers: Figure 3. Figure 3—figure supplement 2.

4) We recommend the authors to also check AT1 and AT2 marker genes in the ALI culture, as a recent preprint has suggested that alveolar cell fates can also be maintained in ALI culture (Abo et al., 2020).

We agree.

Actions taken: In the revised version of this manuscript, we have carried out additional staining on ALI monolayers and included a new panel Figure 3—figure supplement 1J. For the convenience of the reviewer, we have copied and pasted that panel below. The figure legend and the text has been accordingly modified citing this additional panel.

5) The authors need to address if the 'not sustained viral release and infectivity' of iAT2 cells was due to the virus quickly infecting, and killing, the AT2 cells within the first 48 hours, which could be another explanation for the kinetics observed in Figure 3H.

We agree that this is a plausible explanation, and is certainly consistent with the diffuse alveolar damage (DAD) that is pathognomonic of the COVID-19 affected acute lung injury, which leads to ARDS. We have included a sentence to entertain this possibility in the text.

Figures 4-6

The approach in Figures 4-6 of characterizing the COVID-infected lungs at a transcriptional level, deriving transcriptional signatures and testing to what extent these are replicated in the various organoid models is innovative and highly commendable. The authors state that these data strongly suggest that the mixed organoids presented are a better culture model for COVID infection than previous organoid experiments. The spread of the data, e.g. in 6E, may also reflect the variability of this culture system. However, this reviewer is not sufficiently bioinformatically-literate to comment in detail on this aspect of the manuscript.

We thank the reviewer for appreciating the degree of diligence that we demonstrated in confirming if the disease model is accurately able to not just model infectivity, but the host immune response that is observed in the actual disease, and is believed to be the cause of fatality. Because we do not want readers to feel left out if they do not find themselves as familiar with computational approaches, we have modified the results section with the intent to explain what was done and why. We also added additional datasets (prospectively) of models released since the submission of this manuscript (Fig 6H, new) and made head to head comparisons of other models versus disease (Fig 5- Figure Supplement 3).

[Editors' note: further revisions were suggested prior to acceptance, as described below.]

The manuscript has been improved but there are some remaining issues that need to be addressed, as outlined below:

The Reviewers found that your revised manuscript is improved, however immunostaining is still not found to be completely conclusive, for example immunostaining of ciliated cells and others would help.

We suggest that the authors should alter the text in response to the points raised in the second round of reviews to tamper the claims according to the suggestions of the Reviewers listed below.

We are pleased to see that all three reviewers found that the revised manuscript is improved and that we have adequately addressed their major concerns. As for the two major points summarized by the Editors, we have addressed those in the following way:

i) “immunostaining of ciliated cells…is not completely conclusive”: This point refers to the patterns of ciliary structures. We have addressed this with edits to the text to explain why 2D submerged monolayers but not 3D structures had prominent apical ciliary structures, which were further augmented in the 2D-ALI model. These progressive changes in staining patterns are consistent with the fact that differentiation and apicobasal polarity is minimal in 3D structures that are yet to form lumen (Figure 2J-K), is intermediate in the submerged 2D model, and maximal in the 2D ALI model (Figure 3F). Revised text now calls this pattern to attention (on Page 9). A detailed response to this point is on Page 13 of this rebuttal document (Response to Comment 1 from Reviewer #3).

ii) “alter the text …..to tamper the claims”: We have now done that throughout the text. For the convenience of the editors and the reviewers, we have copied and pasted the relevant edited text within the body of this rebuttal.

Reviewer #2 (Recommendations for the authors):

The revisions on the manuscript "Adult Stem Cell-derived Complete Lung Organoid Models Emulate Lung Disease in COVID-19" by the group of Prof. Das have been extensive, as is the rebuttal letter.

Overall, the authors have answered most of the questions about COVID-19 in an adequate manner. However, major issues regarding the characterisation of the organoid model remain. In general, the authors still present contradicting data about the cell type composition of the model. This also casts indirect doubt on the COVID experiments.

Below, I go through the rebuttal letter point-by-point

Comment # 1

All reviewers agree that the cell phenotypes and stability of the organoids in culture should be better characterized with time in culture using a number of new markers that are listed in the reviewer reports, as well as additional methods such as immunostaining and flow. Reviewers have raised concerns that organoids are not stable during culture.

Response to rebuttal

The authors have included additional analyses to verify the stability of their model system. These additional analyses however are not overlapping with existing data. Discussed by method:

– The authors have used flow cytometry to identify cell compositions in the organoid models in early and late passages. While the authors indeed show (Figure 2S5) that the percentage of cell types based on key markers is stable in passages, the data itself is not convincing. The authors describe a model that consists of 96% basal cells, 60% goblet or secretory cells, 90% ciliated cells, 48% AT1 cells and 90% AT2 cells. While the authors already discuss that these percentages add up to more that 100%, the sum of the cell types is even 384% over 5 cell types. The authors claim this is due to overlap of marker genes between cell types and reference a manuscript. The manuscript however does not cover this overlap but actually shows that these markers can be used to separate the cell types. Apart from this, the same antibodies when used in immunostaining (Figure 3E-F) do not show these high percentages of cell positivity for the given markers. While in the FACS analysis the monolayers consist of 92% SFTPB+ cells, the immunostaining shows 10-20% positivity. Both methods are based on protein levels and should therefore show similar phenotypes. Secondly, the authors' claim of lowly expressed marker genes in all cell types this could be resolved by using a more stringent gate in the flow cytometer, only gating the real positives or by addition of a second marker in another channel. The first approach seems to already work for AQP5+ population in which there is a second population that is more positive and therefore potentially true AT1 cells. This result is also in line with the immunostaining in Figure 3E.

Anyway, the markers used are known to be highly specific to each of the cell types. Thus, this part of the manuscript remains very confusing.

– While the qRT-PCR data of the comparison between tissue and ALOs is convincing (Figure 2S3) and indeed underlines the representative nature of the model, the comparison between passages is slightly discouraging (Figure 2S4). While the authors state that cell type composition is stable during culture based on the qRT-PCR results, the data show some large differences between early and late passages. Loss of SCGB1A1 and increase in FOXJ1 in culture in all ALOs is visible as well as some varying levels of other markers in one or more lines does not comply with the stability stated in the text. The authors should still state this variation somewhere in the text.

– Immunostaining of ciliated cells and others would really help. The authors however have not included immunostainings of cell types in different passages to verify the qRT-PCR results. These immunostainings could be quantified to show cell compositions in the ALOs in 3D and 2D.

In all, the authors should use qRT-PCR to determine presence of all cell types, and combine this with immunostainings to determine cell compositions of the models. The authors still show the CIBERSORTx method which confuses the reader since it contradicts data shown in other figures. The authors explain in their rebuttal letter that the method was only used as corroborative, yet the data contradicts their own data.

We understand the reviewer’s concern that the 4 methodologies used here are not 100% in agreement with each other when it comes to the proportion of cell types in the ALO models. In fact, this particular issue is brought up by this reviewer in two more comments later within this document (Comments #3d and 4), criticizing the FACS and CYBERSORTx approaches. We have attempted to answer all of those questions in this response.

To recap for the readers, our intended use of these approaches was always to rigorously test the presence of a mixed population of proximal and distal (AT2/AT1) cells in the organoid lines (a major claim in the manuscript), despite passaging in culture. It was never our intention of using them to draw conclusions regarding the exact proportion of each cell type. In fact, we have not claimed anything regarding the relative proportions of cells anywhere in the manuscript.

The reason is that each of the 4 different approaches used here has its own set of strengths (PROS) and limitations (CONS):

Author response table 1
ApproachPROSCONSMajor conclusionCaution
FACS of dispersed cells from organoidsHighthroughputanalysisAnalyzes protein, nottranscriptsAb-related artifactsHas the potential to introduce artifactsduring dissociationMixed cellularity was confirmed in all 3 ALO lines.Mixed cellularity is retained despite the passageThis methodology, standalone, is not appropriate to draw conclusions regarding the absolute proportions of each cell type because of shared markers, and antibody limitations
Targeted qPCRHighly sensitive andspecificLow-throughputanalysis that only measures transcript, but does not inform about protein translationMixed cellularity was confirmed in all 3 ALO lines.Mixed cellularity is retained despite passagethis methodology, standalone, is not sufficient to draw conclusions regarding the proportions of each cell type because of shared transcripts between cell types.
RNASeq>deconvolution using CYBERSORTxhighthroughput analysisResults are as good as our collective knowledge of cell type markers, many of which are sharedMixed cellularity was confirmed in all 3 ALO lines.Mixed cellularity is retained despite the passagethis methodology, standalone, is not sufficient to draw conclusions regarding the proportions of each cell type because of shared transcripts between cell types
In situ detection of protein by immunostaining of3D/2D organoidsDetection of protein (not just transcript) and with fewerartifactsbecause of in situ analysisLow-throughput qualitative analysisWe prioritized this methodology over others and used two different approaches (FFPE samples after embedding in Histogel Figure 2I-J and direct fixation with PFA/methanol, Figure 2K) to reduce the fixation related aritfacts of any one particular methodology.Although this approach was the best way to show mixed cellularity in each line, and at times, within the same 3D organoid structure, it is low throughput and qualitative (not quantitative), and hence, not suitable to be used for serial imaging of markers to estimate % cellularity/composition.

What is important to note that all methodologies used, regardless of whether they detected mRNA or protein, confirmed that both proximal and distal epithelial cells were present.

No conclusions were drawn at any point in the manuscript regarding the relative proportion of each cell type.

While we agree that it would be nicer to have confocal imaging on all lines from all passages, this is the lowest throughput methodology and doing this is not feasible.

Action taken: We have revised the “Limitations” section of the discussion (Page 17) to explicitly state that the type of evidence presented only supports the claim of mixed proximo-distal cellular composition, but does not reveal the absolute proportion of such cellularity. For the convenience of the reviewer and editor, we have copied and pasted the altered text below:

“While we successfully demonstrated the proximo-distal mixed cellular composition of the ALOs using four different approaches (flow cytometry, RNA seq, confocal immunofluorescence and targeted qPCR), and showed that such mixed cellularity is preserved during prolonged culture, the exact cellular proportion was not assessed here. Single cell sequencing and multiplexed profiling by flow cytometry are some of the approaches that can provide such in-depth characterization to assess cellular composition at baseline and track how such composition changes upon infection and injury.”

And again on Page 16:

“We noted some variability of cell types between patient to patient, and between early and late passages of ALOs, which is probably because of the heterogeneity of organoids isolated from patient’s lung specimens”.

Comment #2

Claims should be tampered, in particular related to high throughput (HTP) drug screening since only few drugs were tested.

Response to rebuttal

The authors explain in great depth that they did not imply that HTP were performed. This can however be inferred from the text even after changing the text as given in the rebuttal. The finding that the ALOs can be used in a 384-well format is important but the authors should add some more in-depth detailed description that this only generates a possibility for potential HTP screens instead of inferring HTP can be done on these models. The description of a "proof-of-concept for HTP" confuses. This term implies that the authors have used an initial HTP drug screen to validate. This can be changed to "initial protocol for building a platform on which HTP drug screens could be performed"?

We understand the concern. We are happy to share with the reviewer that our method of monolayers preparation for drug screenings is adapted by other groups and the publication is recently released in BioRxiv (PMID: 34159337).

Action taken: To satisfy the reviewer, we have modified the text in the result on Page 11.

“Findings also validate optimized protocols for the adaptation of ALO monolayers in miniaturized 384-well formats for use in high throughput drug screens”.

Comment #3

Discrepancies with transcriptomic signatures of the infected human lungs need to be more carefully considered.

Response to rebuttal

The authors have included more clinical datasets and comparisons between other published model systems. It remains important to mention the discrepancies between clinical samples and the model. This is altered in the text sufficiently. This comment is therefore satisfactorily answered in the revised version of the manuscript.

We are pleased to see that the reviewer thinks this comment was addressed satisfactorily.

Comment 1

The manuscript claims a novel model system of distal and proximal airway epithelium. The authors however fail to discuss a recently published COVID-organoid study which reports similar observations. While their novelty claim could still hold in some areas, the authors do need to discuss the recent manuscripts.

Response to rebuttal

The authors have extended their table including all the existing models with the recent published model systems. I agree that keeping up with COVID-19 model systems is hard in these times. The inclusion into the tables has, however, not led to the authors changing their claim of a novel model system of bronchiolar and alveolar potential. In line 102-104 the authors still write "…what is particularly noteworthy is that none recapitulate the heterogeneous epithelial cellularity of both proximal and distal airways…" which should be altered since Lamers et al. did exactly this. Moreover, in the rebuttal letter, the authors also mention that iPSC-derived AT2 cultures can contain both proximal and distal cell types.

The comparison with the currently published method of Lamers et al. 2021 is now in the manuscript and could be highlighted in the discussion or at the end of the paragraph named "Creation of a lung organoid model, complete with both proximal and distal airway epithelia" by adding a few lines comparing the cell types present in this model (ciliated and goblet) which was not shown in Lamers et al. 2021 (of at least this is convincingly demonstrated)

The reviewer is right in that the model developed by Lamers et al., has mixed cellularity. We should have defined knowledge gap with more specificity.

Action taken: We have edited the following two sentences and added a new sentence in the introduction on Page 4 to sharpen the scope and context, and accurately present what has been done prior to this work.

“While a head-to-head comparison of the key characteristics of each model can be found in Table 1, what is particularly noteworthy is that most of the models lack the heterogeneous epithelial cellularity of both proximal and distal airways, i.e., airway epithelia, basal cells, secretory club cells and alveolar pneumocytes.”.

“Also, iPSC-derived AT2 cells be differentiated to proximal and distal cell types, including AT1 and ciliated and club cells13-15 but the models derived from iPSCs lack propagability and/or cannot be reproducibly generated for biobanking; nor can they be scaled up in cost-effective ways for use in drug screens”.

“More specifically, adult lung organoid models that can be grown in a sustainable mode and are complete with proximo-distal epithelia are yet to emerge.”

We also added the work from Lamers et al. (EMBOJ, 2021) in the result section on Page 7.

“It is noteworthy that the co-existence of proximal and distal epithelial cells in lung organoids has been achieved in one another instance prior; Lamers et al., showed such mixed cellular composition in fetal lung bud tip-derived organoids45. However, their model lacked ciliated and goblet cells45, something that we could readily detect in our 3D organoids.”

Comment 2.

Expression patterns in Figure 2B are hard to interpret when no tissue control is used. A positive control of lung tissue should be used to make valid conclusions.

Response to rebuttal

The authors have answered this comment by adding Figure 2S3

We are pleased to see that the reviewer thinks this comment was addressed satisfactorily.

Comment 3.

The authors claim the presence of multiple cell types within single ALO. Data are not strong.

3a. Co-staining of KRT5 and SFTPB (as shown in figure 2J) is difficult to explain, since these markers are expressed in two different cell types. The lower panel in the same figure does show two separate populations, but the images shown in the second panel do not.

and

3b. While the authors claim the presence of proximal and distal cells in ALOs, the images show 100% of SFTPC+ cells or 100% KRT5+ cells or 100% Na/K-ATPase cells. This rather shows two different types of ALO within one culture, then mixed cell populations within a single ALO.

Response to rebuttal

By changing the images and the text, the authors made clear that the ALOs are either composed of multiple cell types or of a singular cell type. Therefore, this comment is sufficiently answered.

We are pleased to see that the reviewer thinks this comment was addressed satisfactorily.

3c. The authors claim the presence of ciliated cells by staining of acetylated α tubulin in figure 2K-J and figure 3E. The images shown however are hardly showing specific ciliated cell staining and more importantly the cilia as present on the cell membrane.

Response to rebuttal

By addition of the higher resolution images, the authors answered this comment sufficiently.

We are pleased to see that the reviewer thinks this comment was addressed satisfactorily.

3d. In figure 3H, the authors show MUC5B positive cells in their monolayer cultures. These goblet cells seem to comprise quite a significant proportion of the culture. Their graph in 3B however shows no goblet cell contribution. The claim that ALO consists of a mix of defined cell types is essentially based on deconvoluted bulk RNAseq data. Such a claim is not conclusive.

Response to rebuttal

This comment comes back to earlier described methods in the editor comment section. To shortly summarise, the CYBERSORTx data confuses the reader and is contradictory to the immunostainings, qRT-PCR and flow cytometry data.

We have provided detailed answer to this point in our response to Comment #1.

Comment 4.

Authors claim that ALO maintain cell type ratios similar to lung tissue over several passages. Their data in figure 3B however show a drastic change in the composition of the ALOs. The shift from AT2 to AT1 is explained to occur when the ALO are cultured as submerged monolayers. However, the shift is already visible at later passages in 'standard' ALO culture. Moreover, when comparing cell types between passage 1 and 8, the composition changes dramatically. The cultures have lost basal cells, ALO1 has lost goblet cells, ALO2 has lost ciliated cells. In addition, ALO1 composition differs from ALO2 at passage 2 which doesn't support the statement that the cultures are stable/robust.

Response to rebuttal

See earlier comments about using CYBERSORTx and flow cytometry data.

We have already addressed it in comment 1 in the Editors list of Essential Revisions and in comment 3 above. To avoid repetition, please see the response to Comment 1 above.

Comment 5.

While the authors claim that proximal airway cultures are infected and responsible for maintaining virus replication, there are no virus entry marker-annotated cells in the composition graph of figure 1B.

Response to rebuttal

The comment was aimed at the proximal ALI culture cell composition graph in original figure 3B. This graph did not indicate a VEM population. Still this culture was infected by the virus, without showing a VEM population. The authors have removed the viral entry marker cells from their data representation. This was due to bioinformatic difficulty and VEM cells being a separate cell type in their analyses.

This mistake was also picked up by reviewer #3. We thank the reviewer for closely reading the manuscript.

Action taken: We fixed the legend in Figure 3 by removing GLL and VEM.

Comment 6

Similar to point 3, the authors claim the development of the first combined organoid culture of proximal and distal cell types, as supported by deconvoluted bulk RNAseq. The hiPSC-derived AT2 cultures however also show the presence of both distal and proximal cell types.

Response to rebuttal

In the rebuttal letter the authors mention the manuscripts describing the presence of proximal cell types in a AT2 differentiation protocol of iPSCs. This once more underlines that their statement of the first culture of proximal and distal cell types is incorrect.

This sentence is an unfortunate misunderstanding. We understand that our statement on iPSCderived AT2 cells being able to differentiate also into proximal cell types can be confusing to the readers.

Action taken: Therefore, we have edited the following line to compare the difference between our model with the iPSC-derived models in the introduction on Page 10.

“Also, iPSC-derived AT2 cells be differentiated to proximal and distal cell types, including AT1 and ciliated and club cells13-15 but the models derived from iPSCs lack propagability and/or cannot be reproducibly generated for biobanking; nor can they be scaled up in cost-effective ways for use in drug screens”.

Comment 7

Figure 3H shows a graph presenting the infectivity relative to peak. This axis makes comparison of infectivity impossible. The authors may want to include suppl Figure 5A-B as main figure and exclude 3H from the manuscript. Moreover, viral E-gene qPCR should be complemented by viral titer measurements to verify findings for live virus. The analysis of SARS-CoV-2 infectivity is very limited when only showing a few infected cells and viral E-gene graphs. Suppl Figure 5C also only shows one or two infected cells in all samples which does convince as proof of infectivity.

Response to rebuttal

The authors provide an extensive reasoning behind the use of the legend in figure 3H. The sustained infectivity the authors want to bring across is sufficiently depicted in figure 3H. Figure 3H might confuse the readers that the submerged culture are more heavily infected than ALI cultures. Moreover, this would allow the authors to already claim their permissiveness for infectivity in proximal airways without having to only look at 48-72h timeframe. "Permissive" in general refers to infection at an early stage which is shown in figure 3S3B and not figure 3H.

We are pleased to see that the reviewer agrees–“The sustained infectivity the authors want to bring across is sufficiently depicted in figure 3H.” The reviewer opposes the use of the word “permissive/permissiveness” because he believes that this word suggests early events in viral infection, whereas we are investigating and drawing conclusions about sustained infectivity. We agree.

Action taken: We had used the word “permissive” in 3 places within the text. The first two times it is used in the context of the current model. The third instance is when describing the fetal lung model developed by Lamars et al., EMBO J.

We edited each of those as follows.

First (Page 10): “We first asked if ALO monolayers are permissive to SARS-CoV-2 entry and replication and support sustained viral infection.”

Second (Page 10): “When we specifically analyzed the kinetics of viral E gene expression during the late phase (48-72 hpi window), we found that proximal airway models [human Bronchial airway Epi (HBEpC)] showed high levels of sustained infectivity than distal models [human Small Airway Epi (HSAEpC) and AT2] to viral replication (Figure 3—figure supplement 3C); the ALO monolayers showed intermediate sustained infectivity (albeit with variability).

Third (Page 13): “These findings indicate that despite having mixed cellular composition and the added advantage of being able to support robust viral replication (achieving ~ 5 log-fold increase in titers), the model lacks the signature host response that is seen in all human samples of COVID-19.”

Comment 8

The authors provide data for a single drug to indicate the possibility of high-throughput screening of drugs for COVID-19 treatment. This does not say much about the throughput of the assay.

Response to rebuttal

This comment has been answered and once more commented in editor comments.

We are pleased to see that the reviewer thinks this comment was addressed satisfactorily.

Comment 9

Figure 5C compares uninfected and infected monolayer cultures for genes that are identified in the patient cohort. While the authors claim comparable patterns between the cultures and the patients, the heatmap can not be directly read. The authors should include more details in the legends and combine the heatmaps in 4C and 5C for direct comparison. Currently, the colors represent raw z-scores which do not indicate transcript read numbers but relative differences of the expression of the indicated genes in the samples analysed. These numbers could differ extensively between organoids and patients.

Response to rebuttal

The authors have added some extra lines on the method used to generate these heatmaps. It now becomes clear that these datasets can not be directly compared within one heatmap.

We are pleased to see that the reviewer now understands why the previously made request for directly comparing distinct datasets within one heatmap is not appropriate and hence, not done.

Comment 10

In addition to point 9, the authors show that there is very limited overlap between significantly differentially expressed genes in monolayers and patients. While the authors believe this is mostly due to the lack of mesenchymal and immune components, this indicates that the monolayer itself is not important for the observed infection signature. This makes the claim that the monolayers represent infectivity in patients questionable.

Response to rebuttal

This has been discussed in the editor's section of the revision above

We presented objective analyses of all available models (including ours) and human disease tissue. Our analyses showed that compared to all other models, our model most closely recapitulated the human disease. More specifically, despite missing immune cell and stromal cell components, the epithelial cells alone captured ~ 25-50% of the gene expression changes (Differentially upregulated genes). This is a level of rigor seldom met by any disease modeling work.

In light of this objective evidence, we respectfully disagree with the reviewer’s skepticism that the epithelial monolayers do not represent infectivity in patients.

It was never our intention to question the importance of the other cell types in any disease model (and we explicitly discussed that as a study limitation). However, we believe that understanding the epithelial cell response to infection is critical to understand how the first responders in our innate immune defense system may respond to the infection. Modeling such epithelial response is very important before we go for the complex model after adding mesenchymal and immune cells.

Action taken: We also discussed the findings in the section titled “limitations of the study” on Page 17.

“ Our adult stem-cell-derived lung organoids, complete with all epithelial cell types, can model COVID-19, but remains a simplified/rudimentary version compared to the adult human organ. For instance, although the epithelial contributions to the host response are important, it alone cannot account for the response of the immune cells and the non-immune stromal cells, and their crosstalk with the epithelium. Given that epithelial inflammation and damage is propagated by vicious forward-feedback loops of multicellular crosstalk, it is entirely possible that the epithelial signatures induced in infected ALO-derived monolayers are also only a fraction of the actual epithelial response mounted in vivo. Regardless of the missing components, what appears to be the case is that we already have a model that recapitulates approximately a quarter to half of the genes that are induced across diverse COVID-19 infected patient samples. This model can be further improved by the simultaneous addition of endothelial cells and immune cells to better understand the pathophysiologic basis for DAD, microangiopathy, and even organizing fibrosis with loss of lung capacity that has been observed in many patients40; these insights should be valuable to fight some of the long-term sequelae of COVID-19. ”.

Comment 11.

Description in the discussion of growth factors supplied in the medium like FGF7 and FGF10 are not novel. Sachs et al. 2019 (https://doi.org/10.15252/embj.2018100300) already described these growth factors in the culture medium of airway organoids.

Response to rebuttal

The authors have sufficiently answered this comment.

We are pleased to see that the reviewer thinks this comment was addressed satisfactorily.

Comment 12.

In the discussion, the authors describe the advantages of their model system including

reproducibility, retainment of genetics and use for infection studies. These claims are however not novel since previous papers from multiple groups have reported similar organoid-based models for SARS-CoV-2 research.

Response to rebuttal

With the comments above and the answers from the authors, this comment will be answered. The authors should better define their novelty by explaining the origin of the organoids model (adult stem cells) and their improvement in COVID-19 modelling.

We are pleased to see that the reviewer felt that our answers regarding the novelty of the model are satisfactorily answered. We appreciate the suggestion that we should improve that description of novelty.

Action taken: In this revised submission, we have expanded the last paragraph in the discussion on Page 17 (just before the paragraph on study limitations).

“The lung model we present here is distinct from all currently available other models (see Table 1) because of the confirmed presence of both proximal and distal airway cell types over successive passages, which is yet to be accomplished for adult lung organoid models. Another distinguishing feature of our model is the way we rigorously validated its usefulness in modeling COVID-19 via computational approaches. We confirmed, based on the gene expression changes upon SARS-CoV-2-challenge, that our model most closely recapitulates the human disease, i.e., Covid-19. Analyses also pinpointed the importance of two factors that were critical in modeling COVID-19: (1) adult source, and (2) model completeness, with both proximal and distal airway cells.”.

Reviewer #3 (Recommendations for the authors):

The authors established a new 3D adult lung organoid system. Comparing with previous systems, the new organoid culture can maintain both proximal and distal cell types in adult lungs, and this ratio appears to be relatively stable in long term cultures. The 3D organoids can subsequently be dissociated and re-plated into 2D submerge culture, which exhibited promising features for modelling COVID19 infection. The authors further used bioinformatics analysis to show that the COVID19 infection using the 2D submerge culture was able to recapitulate both the infectivity and the immune responses in COVID patients.

In the revision processes, the authors have provided more supporting data to show the co-existence of proximal and distal lung cell types in the long-term culture and addressed most of the previous reviewers' comments. The few comments that we recommend the authors to further address are as follows:

We are pleased to see that this reviewer found our revisions strengthened the major claim in this manuscript, i.e., co-existence of proximal and distal cell types in long term culture.

The reviewer also noted that although most of the previous reviewers’ comments were answered, he/she recommends that we address 4 major and 3 minor points. We have done that below. We hope that the changes we made address the remaining concerns adequately.

1) The author added the Act-TUB staining for 2D culture in Figure 3, which looks convincing, however, for the 3D organoids, Act-TUB staining in Figure 2J still doesn't look like any cilia structure.

In this comment, the reviewer is comparing Acetylated Tubulin staining patterns in Figure 2, which shows 3D organoids and Figure 3, which shows 2D organoids (both submerged monolayers or in the ALI model). Our claim was that Ac Tub positive ciliary structures are seen in 2D organoids, most prominently in the more differentiated ALI model. The cilia in ALI has been shown in a magnified view in Figure 3 Figure supplement 1J. We do not see such prominent ciliary structures in 3D organoids; it is possible that these 3D organoids do not show the ciliated structures simply because they are not differentiated sufficiently (e.g., no lumen, and hence, we suspect that apicobasal polarity is yet to be established, which is essential for apical cilia formation). Others have demonstrated cilia formation in ALI models of 3D growth (PMID: 27713058), where the size of the organoid structure is substantially larger and lumen formation is associated with apicobasal polarity and cilia formation.

Action taken: To ensure that our findings, figures, and description in text avoid misleading readers, we have added the following sentences on Page 9 to address this issue with greater clarity:

“Compared to the 3D organoids, the 2D organoid cultures, especially the ALI model, showed a significant increase in ciliated structures, as determined by Acetylated Tubulin (compare Ac Tub stained panels in Figure 2J-K with Figure 3E-F)”. The observed progressive prominence of ciliary structures from 3D to 2D models is in keeping with the fact that 3D ALOs that are yet to form lumen represent the least differentiated state, whereas 2D submerged monolayers are intermediate and the 2D-ALI monolayers are maximally differentiated; differentiation is known to establish apicobasal polarity, which is essential for the emergence of mature cilia on the apical surface”.

2) In Page 6 text line 189, the author mentioned a 'stem cell' population, which is TP63 positive. This is separate to the basal cells on line 188. Whereas in Figure 2-Supp4 the authors have acknowledged TP63 as a basal cell marker. We recommend the authors to make this consistent as basal cells which are well-established to be the airway stem cells (Rock et al., 2009).

We agree with the suggestion. We also believe that TP63 should be represented as a basal cell marker and hence, in the original version of the manuscript, we had indicated the same. However, upon suggestion of this reviewer, we analyzed another different marker (p75/NGFR) and had added that finding. We are glad to see that the reviewer now agrees that TP63 is a bona fide marker of airway stem cells.

Action taken: In this revised submission, we have edited it throughout the manuscript to indicate that TP63 is a marker of basal cells, which are well established to be airway stem cells.

3) The authors still make a strong claim about the co-existence of proximal (airway) and distal (alveolar) cell populations in a single 3D organoid (line 198). However, the staining shown in Figure 2 can only infer the existence of either proximal or distal cell populations in a single 3D organoid, given SFTPB is not a specific marker for alveolar lineage. Additionally, this strong claim wasn't adding much value to the manuscript as the authors only need to show proximal-distal cells exist in the overall population as 3D culture, given only the 2D submerged culture was used for COVID infection. We recommend the author not to mention this claim as even the revised data do not support it.

In addition, the same paragraph describes BASCs in the organoids. We cannot cite a reference to refute the existence of BASC cells in human lungs. However, in now more than 10 years of people looking for them a convincing demonstration that BASCs exist in human lungs is still missing. Human distal airways have a very different organisation to those of mice (respiratory bronchioles). BASC cells have not been found in them. We recommend that you do not highlight human BASCs in the organoids given the lack of credible evidence that they exist in vivo.

This comment has two parts and we have broken down the parts below and responded to each separately:

i) First, the reviewer says that while it is convincing that the same ALO line has a mixed population of proximal and distal airway epithelial cells, such a mixed population is not there in the same 3D organoid structure does not. The reviewer acknowledges that because infections are all carried out in 2D monolayers of a mixed cell population, the claim that a single 3D structure has mixed cell types is unnecessary for the major point that we are trying to make.

We agree that the presence of mixed cellularity in a single 3D organoid structure is not essential for supporting the major claims in this manuscript which use 2D mixed cellular model for SARS-CoV-2 infection and COVID-19 modeling.

However, we respectfully disagree with the reviewer that single organoid structures do not have a mixture of cells. We believe that this is an important point in the characterization of the ALO model because the fact that ALOs can achieve both homotypic and heterotypic cellular composition may influence how these models may behave in 3D-ALI or whether they are appropriate for use in 3D models of infection with microinjections.

Action taken:

To ensure we state the findings, but not overstate it, we modified the sentence to strengthen the claim that this reviewer acknowledges, i.e., that mixed proximal and distal cellularity was there in each ALO line, and soften the claim that such cellularity was detected also in the same 3D structure.

“The presence of all cell types was also confirmed by assessing protein expression of various cell types within organoids grown in 3D cultures. Two different approaches were used—(i) slices cut from FFPE cell blocks of HistoGel-embedded ALO lines (Figure 2I-J) or (ii) ALO lines grown in 8-well chamber slides were fixed in matrigel (Figure 2K), stained, and assessed by confocal microscopy. Such staining not only confirmed the presence of more than one cell type (i.e., mixed cellularity) of proximal (basal-KRT5) and distal (AT1/AT2 markers) within the same ALO line, but also, in some instances, demonstrated the presence of mixed cellularity within the same 3D structure. For example, AT2 and basal cells, marked by SFTPB and KRT5, respectively, were found in the same 3D-structure (Figure 2J, interrupted curved lines). Similarly, ciliated cells and goblet cells stained by Ac-Tub and Muc5AC, respectively, were found to coexist within the same structure (Figure 2J, interrupted box; Figure 2K, arrow). Intriguingly, we also detected 3D structures that co-stained for CC10 and SFTPC (Figure 2J, bottom panel) indicative of mixed populations of club and AT2 cells. Besides the organoids with heterogeneous makeup, each ALO line also showed homotypic organoid structures that were relatively enriched in one cell type (Figure 2J, arrowheads pointing to two adjacent structures that are either KRT5- or SFTPB-positive). Regardless of their homotypic or heterotypic cellular organization into 3D-structures, the presence of mixed cellularity was documented in all three ALO lines (see multiple additional examples in Figure 2—figure supplement 2I)”.

ii) Second, the reviewer asks us to remove the discussion and citation surrounding BASCs, which have not been found in the human lung.

Action taken: We completely agree with the reviewer and we removed the line on Page 7 with the associated reference on BASC.

We have also edited the following line on Page 7.

“Intriguingly, we also detected 3D structures that co-stained for CC10 and SFTPC (Figure 2J, bottom panel) indicative of mixed populations of club and AT2 cells”.

4) The authors need to clarify how the 5 replicates for submerged culture and 2 replicates for ALI culture were done in Figure 2B middle panel. Were they from the same ALO line or from different ALO lines? What was the passage number used here? These will help the readers to have an idea about how reproducible the system is.

We believe that the reviewer meant to say Figure 3B instead of 2B because in Figure 2B there are no submerged or ALI cultures. Instead, the description of panels provided here matches the Figure panel 3B where % of the cellular composition is predicted using the CIBERSORTx.

For the submerged monolayers, ALO1 and ALO2 were used where we have the following 5 samples (i and ii) from ALO1 passage 3, (iii and iv) from ALO2 passage 3 and (v) from ALO1 passage 8. The ALI model is generated from ALO1 passage 3 and passage 8.

Action taken: In this revised submission, we have included these details also in Figure 3 panel B.

https://doi.org/10.7554/eLife.66417.sa2

Article and author information

Author details

  1. Courtney Tindle

    1. Department of Cellular and Molecular Medicine, University of California San Diego, San Diego, United States
    2. HUMANOID CoRE, University of California San Diego, San Diego, United States
    Contribution
    Conceptualization, Conducted experiments on adult lung-derived 3D-organoids and 2D-monolayers, Data curation, Formal analysis, Investigation, Methodology, Writing – original draft
    Contributed equally with
    MacKenzie Fuller, Ayden Fonseca, Sahar Taheri, Pradipta Ghosh and Soumita Das
    Competing interests
    None
  2. MacKenzie Fuller

    1. Department of Cellular and Molecular Medicine, University of California San Diego, San Diego, United States
    2. HUMANOID CoRE, University of California San Diego, San Diego, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – review and editing
    Contributed equally with
    Courtney Tindle, Ayden Fonseca, Sahar Taheri, Pradipta Ghosh and Soumita Das
    Competing interests
    none
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6781-8710
  3. Ayden Fonseca

    1. Department of Cellular and Molecular Medicine, University of California San Diego, San Diego, United States
    2. HUMANOID CoRE, University of California San Diego, San Diego, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing – review and editing
    Contributed equally with
    Courtney Tindle, MacKenzie Fuller, Sahar Taheri, Pradipta Ghosh and Soumita Das
    Competing interests
    None
  4. Sahar Taheri

    Department of Computer Science and Engineering, Jacobs School of Engineering, University of California San Diego, San Diego, United States
    Contribution
    Data curation, Formal analysis, Methodology, Project administration
    Contributed equally with
    Courtney Tindle, MacKenzie Fuller, Ayden Fonseca, Pradipta Ghosh and Soumita Das
    Competing interests
    None
  5. Stella-Rita Ibeawuchi

    Department of Pathology, University of California San Diego, San Diego, United States
    Contribution
    Data curation, Formal analysis, Investigation, Methodology
    Competing interests
    None
  6. Nathan Beutler

    Department of Immunology and Microbiology, The Scripps Research Institute, La Jolla, United States
    Contribution
    Data curation, Formal analysis, Investigation, Methodology
    Competing interests
    None
  7. Gajanan Dattatray Katkar

    Department of Cellular and Molecular Medicine, University of California San Diego, San Diego, United States
    Contribution
    Data curation, Formal analysis, Investigation, Methodology
    Competing interests
    None
  8. Amanraj Claire

    1. Department of Cellular and Molecular Medicine, University of California San Diego, San Diego, United States
    2. HUMANOID CoRE, University of California San Diego, San Diego, United States
    Contribution
    Data curation, Formal analysis, Investigation, Methodology
    Competing interests
    None
  9. Vanessa Castillo

    Department of Cellular and Molecular Medicine, University of California San Diego, San Diego, United States
    Contribution
    Data curation, Formal analysis, Investigation, Methodology
    Competing interests
    None
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4182-8846
  10. Moises Hernandez

    Division of Cardiothoracic Surgery, University of California San Diego, San Diego, United States
    Contribution
    Data curation, Methodology, Resources
    Competing interests
    None
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7651-2673
  11. Hana Russo

    Department of Pathology, University of California San Diego, San Diego, United States
    Contribution
    Methodology, Resources
    Competing interests
    None
  12. Jason Duran

    Division of Cardiology, Department of Internal Medicine, UC San Diego Medical Center, San Diego, United States
    Contribution
    Methodology, Resources
    Competing interests
    None
  13. Laura E Crotty Alexander

    1. Pulmonary Critical Care Section, Veterans Affairs (VA) San Diego Healthcare System, La Jolla, United States
    2. Division of Pulmonary and Critical Care, Department of Medicine, University of California, San Diego, La Jolla, CA, United States
    Contribution
    Methodology, Resources, Writing – review and editing
    Competing interests
    None
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5091-2660
  14. Ann Tipps

    Department of Pathology, University of California San Diego, San Diego, United States
    Contribution
    Methodology, Resources
    Competing interests
    None
  15. Grace Lin

    Department of Pathology, University of California San Diego, San Diego, United States
    Contribution
    Methodology, Resources
    Competing interests
    None
  16. Patricia A Thistlethwaite

    Division of Cardiothoracic Surgery, University of California San Diego, San Diego, United States
    Contribution
    Methodology, Resources, Writing – review and editing
    Competing interests
    None
  17. Ranajoy Chattopadhyay

    1. Department of Cellular and Molecular Medicine, University of California San Diego, San Diego, United States
    2. HUMANOID CoRE, University of California San Diego, San Diego, United States
    3. Cell Applications Inc., La Jolla, CA, United States
    Contribution
    Methodology, Resources, Writing – review and editing
    For correspondence
    rachatto72@gmail.com
    Competing interests
    None
  18. Thomas F Rogers

    1. Department of Immunology and Microbiology, The Scripps Research Institute, La Jolla, United States
    2. Division of Infectious Diseases, Department of Medicine, University of California, San Diego, La Jolla, United States
    3. Department of Immunology and Microbiology, The Scripps Research Institute, La Jolla, United States
    Contribution
    Methodology, Resources
    For correspondence
    trogers@health.ucsd.edu
    Competing interests
    None
  19. Debashis Sahoo

    1. Department of Computer Science and Engineering, Jacobs School of Engineering, University of California San Diego, San Diego, United States
    2. Department of Pediatrics, University of California, San Diego, La Jolla, CA, United States
    Contribution
    Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Resources, Software, Supervision, Writing – original draft, Writing – review and editing
    For correspondence
    dsahoo@ucsd.edu
    Competing interests
    None
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2329-8228
  20. Pradipta Ghosh

    1. Department of Cellular and Molecular Medicine, University of California San Diego, San Diego, United States
    2. HUMANOID CoRE, University of California San Diego, San Diego, United States
    3. Department of Medicine, University of California, San Diego, La Jolla, CA, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing – original draft, Writing – review and editing
    Contributed equally with
    Courtney Tindle, MacKenzie Fuller, Ayden Fonseca, Sahar Taheri and Soumita Das
    For correspondence
    prghosh@ucsd.edu
    Competing interests
    None
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-8917-3201
  21. Soumita Das

    1. HUMANOID CoRE, University of California San Diego, San Diego, United States
    2. Department of Pathology, University of California San Diego, San Diego, United States
    Contribution
    Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Writing – original draft, Writing – review and editing
    Contributed equally with
    Courtney Tindle, MacKenzie Fuller, Ayden Fonseca, Sahar Taheri and Pradipta Ghosh
    For correspondence
    sodas@ucsd.edu
    Competing interests
    None
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-3895-3643

Funding

National Institute of Diabetes and Digestive and Kidney Diseases (3R01DK107585-05S1)

  • Soumita Das

National Institute of Diabetes and Digestive and Kidney Diseases (1R01DK107585-01A1)

  • Soumita Das

National Institute of Allergy and Infectious Diseases (R01-AI 155696)

  • Debashis Sahoo
  • Pradipta Ghosh
  • Soumita Das

National Institute of Allergy and Infectious Diseases (R01-AI141630)

  • Pradipta Ghosh

National Cancer Institute (CA100768)

  • Pradipta Ghosh

National Cancer Institute (CA160911)

  • Pradipta Ghosh

National Institute of General Medical Sciences (R01-GM138385)

  • Debashis Sahoo

National Heart, Lung, and Blood Institute (R01- HL32225)

  • Patricia A Thistlethwaite

University of California, San Diego (UCOP-R01RG3780)

  • Debashis Sahoo
  • Pradipta Ghosh
  • Soumita Das

University of California, San Diego (UCOP-R00RG2642)

  • Pradipta Ghosh
  • Soumita Das

National Heart, Lung, and Blood Institute (R01-HL137052)

  • Laura E Crotty Alexander

Department of Veterans Affairs Merit Award (1I01BX004767)

  • Laura E Crotty Alexander

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

The authors thank Victor Pretorius, Rachel White, and Jen Bigbee (Department of Cardiothoracic Surgery, UC San Diego) who assisted with thoracotomies during rapid autopsies.

Ethics

The human subject research was performed following the approved protocol at University of California San Diego. Deidentified lung tissues obtained during surgical resection, that were deemed excess by clinical pathologists, were collected using an approved human research protocol (IRB# 101590; PI: Thistlethwaite). Isolation and biobanking of organoids from these lung tissues were carried out using an approved human research protocol (IRB# 190105: PI Ghosh and Das) that covers human subject research at the UC San Diego HUMANOID Center of Research Excellence (CoRE). For all the deidentified human subjects, information including age, gender, and previous history of the disease, was collected from the chart following the rules of HIPAA.

Senior Editor

  1. Jos W Van der Meer, Radboud University Medical Centre, Netherlands

Reviewing Editor

  1. Milica Radisic, University of Toronto, Canada

Reviewers

  1. Milica Radisic, University of Toronto, Canada
  2. Hans Clevers, University Medical Centre Utrecht, Netherlands

Publication history

  1. Received: January 10, 2021
  2. Accepted: August 11, 2021
  3. Accepted Manuscript published: August 13, 2021 (version 1)
  4. Version of Record published: September 24, 2021 (version 2)

Copyright

© 2021, Tindle et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 798
    Page views
  • 237
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

  1. Further reading

Further reading

    1. Cancer Biology
    2. Medicine
    Maria Ochoa de Olza
    Insight

    Organ-on-chip approaches could help researchers to better predict the toxicity of cancer immunotherapy drugs.

    1. Medicine
    2. Microbiology and Infectious Disease
    Alexander O Pasternak et al.
    Research Article Updated

    Background:

    It remains unclear whether combination antiretroviral therapy (ART) regimens differ in their ability to fully suppress human immunodeficiency virus (HIV) replication. Here, we report the results of two cross-sectional studies that compared levels of cell-associated (CA) HIV markers between individuals receiving suppressive ART containing either a non-nucleoside reverse transcriptase inhibitor (NNRTI) or a protease inhibitor (PI).

    Methods:

    CA HIV unspliced RNA and total HIV DNA were quantified in two cohorts (n = 100, n = 124) of individuals treated with triple ART regimens consisting of two nucleoside reverse transcriptase inhibitors (NRTIs) plus either an NNRTI or a PI. To compare CA HIV RNA and DNA levels between the regimens, we built multivariable models adjusting for age, gender, current and nadir CD4+ count, plasma viral load zenith, duration of virological suppression, NRTI backbone composition, low-level plasma HIV RNA detectability, and electronically measured adherence to ART.

    Results:

    In both cohorts, levels of CA HIV RNA and DNA strongly correlated (rho = 0.70 and rho = 0.54) and both markers were lower in NNRTI-treated than in PI-treated individuals. In the multivariable analysis, CA RNA in both cohorts remained significantly reduced in NNRTI-treated individuals (padj = 0.02 in both cohorts), with a similar but weaker association between the ART regimen and total HIV DNA (padj = 0.048 and padj = 0.10). No differences in CA HIV RNA or DNA levels were observed between individual NNRTIs or individual PIs, but CA HIV RNA was lower in individuals treated with either nevirapine or efavirenz, compared to PI-treated individuals.

    Conclusions:

    All current classes of antiretroviral drugs only prevent infection of new cells but do not inhibit HIV RNA transcription in long-lived reservoir cells. Therefore, these differences in CA HIV RNA and DNA levels by treatment regimen suggest that NNRTIs are more potent in suppressing HIV residual replication than PIs, which may result in a smaller viral reservoir size.

    Funding:

    This work was supported by ZonMw (09120011910035) and FP7 Health (305522).