key: cord-282372-nmii30mc
authors: Youk, Jeonghwan; Kim, Taewoo; Evans, Kelly V.; Jeong, Young-Il; Hur, Yongsuk; Hong, Seon Pyo; Kim, Je Hyoung; Yi, Kijong; Kim, Su Yeon; Na, Kwon Joong; Bleazard, Thomas; Kim, Ho Min; Ivory, Natasha; Mahbubani, Krishnaa T.; Saeb-Parsy, Kourosh; Kim, Young Tae; Koh, Gou Young; Choi, Byeong-Sun; Ju, Young Seok; Lee, Joo-Hyeon
title: Robust three-dimensional expansion of human adult alveolar stem cells and SARS-CoV-2 infection
date: 2020-07-10
journal: bioRxiv
DOI: 10.1101/2020.07.10.194498
sha: 
doc_id: 282372
cord_uid: nmii30mc

Severe acute respiratory syndrome-coronavirus 2 (SARS-CoV-2), which is the cause of a present global pandemic, infects human lung alveolar cells (hACs). Characterising the pathogenesis is crucial for developing vaccines and therapeutics. However, the lack of models mirroring the cellular physiology and pathology of hACs limits the study. Here, we develop a feeder-free, long-term three-dimensional (3D) culture technique for human alveolar type 2 (hAT2) cells, and investigate infection response to SARS-CoV-2. By imaging-based analysis and single-cell transcriptome profiling, we reveal rapid viral replication and the increased expression of interferon-associated genes and pro-inflammatory genes in infected hAT2 cells, indicating robust endogenous innate immune response. Further tracing of viral mutations acquired during transmission identifies full infection of individual cells effectively from a single viral entry. Our study provides deep insights into the pathogenesis of SARS-CoV-2, and the application of long-term 3D hAT2 cultures as models for respiratory diseases.

Several members of the family Coronaviridae are transmitted from animals to humans and cause severe respiratory diseases in affected individuals 1 . These include the severe acute respiratory syndrome (SARS) and the Middle East respiratory syndrome (MERS) coronavirus. Currently, Coronavirus disease 2019 , caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), is spreading globally 2 and more than 11.8 million confirmed cases with ~542K deaths have been reported worldwide as of 7 th Jul 2020 3 . The lung alveoli are the main target for these emerging viruses 4 .

To develop strategies for efficient prevention, diagnosis, and treatment, the characteristics of new viruses, including mechanisms of cell entry and transmission, kinetics in replication and transcription, host reactions and genome evolution, should be accurately understood in target tissues. Although basic molecular mechanisms in SARS-CoV-2 infection have been identified [5] [6] [7] [8] , most findings have been obtained from experiments using non-physiological cell lines 9 , model animals, such as transgenic mice expressing human angiotensin-converting enzyme 2 (ACE2) 10 , ferrets 11 and golden hamsters 12 , or from observation in clinical cohorts 13 and/or inference from in-silico computational methods [14] [15] [16] . As a consequence, we do not fully understand how SARS-CoV-2 affects human lung tissues in the physiological state.

Development of three-dimensional (3D) stem cell-derived organotypic culture models, conventionally called organoids, has enabled various physiologic and pathological studies using human-derived tissues in vitro [17] [18] [19] . Organoid models established from induced pluripotent stem cells (iPSCs), or adult stem cells in the human kidney, intestine, and airway have been used to investigate SARS-CoV-2 pathogenesis [20] [21] [22] [23] . Although human alveolar type 2 cells (hereafter referred to as hAT2s) are believed to be the ultimate target cells for SARS-CoV-2, their infection model has not previously been introduced.

We have established feeder-free, 3D hAT2 organoids (hereafter referred to as hAOs; definition of organoid is available at ref. 24 ) with defined factors which support molecular and functional identity of hAT2 cells over multiple passages, showing substantial improvements from the previous application of co-culture models [25] [26] [27] . Briefly, single-cell dissociated hAT2 cells were isolated by fluorescenceactivated cell sorting (FACS) for the hAT2 surface marker HTII-280 (CD31 -CD45 -EpCAM + HTII-280 + ) 25, 28 (Fig. 1a; Extended Data Fig. 1a) . Isolated HTII-280 + cells showed higher expression of AT2 cell marker SFTPC, while HTII-280cells revealed higher expressions of basal cell marker TP63 and secretory cell marker SCGB1A1 (Extended Data Fig. 1b) . We then plated HTII-280 + hAT2 cells into Matrigel for 3D cultures with our expansion medium, supplemented with CHIR99021, RSPO1 (Rspondin 1), FGF7, FGF10, EGF, NOG (Noggin), and SB431542, that are known to support the growth of human embryonic lung tip cells 29 . HTII-280cells were also cultured under conditions supporting human bronchial (airway) organoids (hereafter referred to as hBOs) that have previously been reported 30 . hAOs established from single hAT2 cells grew up to 4 weeks with heterogeneous morphology including budding-like and cystic-like structures consisting of mature AT2 cells expressing pro-SFTPC, HTII-280, and ABCA3, as well as exhibiting uptake of Lysotracker, a fluorescent dye that stains acidic organelles such as lamellar bodies 25 (Fig. 1b and 1c) . In contrast, hBOs grew quickly by day 14 with cystic-like structures consisting of a number of airway cell types, including KRT5 + TP63 + basal cells and SCGB1A1 + secretory cells, as previously reported 30 (Fig. 1d and 1e) . WNT activation was identified as an essential factor for hAO formation, because no colony formation was found in the absence of WNT activator CHIR99021 in culture (Extended Data Fig. 1c) . Importantly, our culture system allows the long-term expansion (>10 months) of hAT2s, although colony forming efficiency varied between tissue samples and reduced at later passages (Extended Data Fig. 1d ). Over passaging via single cells, hAT2s consistently formed organoids, exhibiting SFTPC expression following 9 months of continuous cultures, although growth began to slow, as evident by reduced organoid size and lower forming efficiencies (Extended Data Fig. 1d and 1e) . Alveolar type 1 cells (hAT1s) expressing HOPX and PDPN were also observed during early cultures, demonstrating differentiation capacity of hAT2 cells in our hAOs (Extended Data Fig. 1f) , although later passages exhibited loss of these cells.

The expressions of ACE2 and TMPRSS2, which are necessary for SARS-CoV-2 infection, were observed in the membrane and cytoplasm of hAO cells ( Fig. 1f and 1g; Extended Data Fig. 1g) . We next infected hAOs and hBOs with SARS-CoV-2 at a multiplicity of infection (MOI) of 1. The viral particles were prepared from a patient (known as KCDC03) who was diagnosed with COVID-19 on 26 th Jan, 2020, after traveling to Wuhan, China 31 . Vero cells were also infected as a positive control, although this was not directly comparable to our 3D models due to different technical procedures.

Infectious virus particles increased to significant titers in hAOs (Fig. 1h-1k; Extended Data Fig. 2) , reaching maximum levels within the 1st day post infection (dpi), suggesting that full infection occurs within 1 day from viral entry to hAOs. In hBOs, the increment of viral particles was observed as consistent with another study 32 , but their titers were <100 times lower than hAOs ( Fig. 1h and 1i; Extended Data Fig. 2 ). In line with viral particles, the amount of the viral RNA in hAOs and in its culture supernatant reached a plateau at 1 dpi ( Fig. 1j and 1k) . Although infected Vero cells exhibited significant cytopathic effects at 1 dpi, typically cell rounding, detachment, degeneration and syncytium formation 9 , SARS-CoV-2 infected hAOs and hBOs did not show prominent macroscopic pathologies up until 10 dpi.

Immunostaining for double-stranded viral RNA (dsRNA) and nucleocapsid protein (NP) of SARS-CoV-2 identified widespread viral infection in hAT2 cells co-expressing pro-SFTPC and ACE2 in hAOs ( Fig. 2a and 2b; Extended Data Fig. 3) . To further determine subcellular events at a higher resolution, transmission electron microscopic analysis was performed at 2 dpi ( Fig. 2c- 25 ), showed discernible viral particles in the cytoplasm. A fraction of cells in hAOs showed much higher viral burdens than other cells, with as many as 500 copies in the 100 nm section, implying that >10,000 SARS-CoV-2 particles per cell. Aggregated viral proteins, which appeared as electron-dense regions near nuclei 33 , were also detected (Fig. 2c) . Accordingly, several key pathogenic phenotypes were observed in the infected hAOs. Alveolar cells with enormous vacuoles were frequently observed (Fig. 2c, 2f, 2h, 2i) , similar to cytopathic signals in Zika virus 34 . Doublemembrane vesicles (DMVs), subcellular structures known as viral replication sites frequently seen in the early phase of infection 35, 36 , were observed in the vicinity of zippered endoplasmic reticulum in a small fraction of hAO cells ( Fig. 2j and 2k) . Viral particles were dispersed in the cytosol (Fig. 2e) or enclosed in the small vesicular structures ( Fig. 2f and 2i) . Diverse forms of viral secretion were also observed mainly through the apical surfaces of hAT2 cells ( Fig. 2g and 2l) . More ultrastructural pathologies are available at Extended Data Fig. 4 and the EMPIAR data archive (see Data availability).

From strand-specific deep RNA-sequencing, we explored gene expression changes in the infected hAOs. Indeed, a set of human genes were differentially expressed as infection progressed (i.e., 0, 1 and 3 dpi), although most genes showed good correlations (Extended Data Fig. 5a and Supplementary Table 1 ). Cytokeratin genes (including KRT16, KRT6A, KRT6B, and KRT6C), genes involved in keratinization (including SPRR1A), cytoskeleton (including S100A2) and cell-cell adhesion genes (including DSG3), were significantly reduced to ~2-3% in hAOs at 3 dpi (Fig 3a and 3b) . Many more genes were upregulated in the infected hAOs specifically at 3 dpi. In particular, transcription of a broad range of interferon-stimulated genes (ISGs), known to be typically activated by type I and III interferons 37 , were remarkably increased ( Fig. 3a and 3b) . These genes include interferon induced protein genes (such as IFI6, IFI27, IFI44, IFI44L), interferon induced transmembrane protein genes (such as IFITM1), interferon induced transmembrane proteins with tetratricopeptide repeats genes (IFIT1, IFIT2, IFIT3), 2'-5'-oligoadenylate synthetase genes (OAS1, OAS2), and miscellaneous genes known to be involved in innate cellular immunity (MX1, MX2, RSAD2, ISG15). These genes were expressed to >20 times higher levels in hAOs at 3 dpi than at 0 dpi. Many other known ISGs also showed moderate inductions (2-20 times) at 3 dpi, including BTS2 (~15 times), OAS3 (~11 times), HERC5 (~15 times), HERC6 (~12 times) and USP18 (~12 times). Antiviral functions are known for these ISGs 38 including (1) inhibition of virus entry (MX genes, IFITM genes), (2) inhibition of viral replication and translation (IFIT genes, OAS genes, ISG15, HERC5, HERC6, USP18) and (3) inhibition of viral egress (RSAD2 and BST2). Of note, given that immune cells are absent in our culture system, the innate immune response was completely autologous to alveolar cells, mimicking the initial phase of SARS-CoV-2 alveolar infection. In line with the notion, innate induction of some type I and type III interferons was observed. Of the 20 interferon genes, an interferon beta gene (IFNB1) and three interferon lambda genes (IFNL1, IFNL2, and IFNL3) showed significant transcriptional induction, although their absolute changes were not substantial (Fig. 3c) . The surface receptors of interferons were stably expressed in hAO cells without reference to viral infection (Fig.   3c ). Downstream signalling genes of the receptors were also upregulated such as STAT1 (~3.5 times), STAT2 (~2.5 times) and their associated genes IRF1 (~2.4 times) and IRF9 (~6.4 times). Of note, IRF1 is known to be specific to type I interferon responses 39 , while type I and type III ISGs are generally overlapping 40 .

In addition to ISGs, genes in the viral sensing pathway in cytosol showed increased expression in the infected hAOs at 3 dpi, for example, DDX58 (official gene name of RIG-1, from 1.9 to 25.0 TPM), IFIH1 (also known as MDA5, from 5.8 to 30.3 TPM), and TLR3 (Toll-like receptor 3, from 1.3 to 3.7 TPM), IRF7 (Interferon regulatory factor 7, from 4.4 to 29 TPM) and IL6 (0.6 to 2.5 at 1 dpi; proinflammatory factor).

Notably, these transcriptional changes were much stronger in hAOs than in hBOs. In the similar transcriptome profiling of the infected hBOs, the genes aforementioned were not significantly altered (Supplementary Table 2 and Extended Data Fig. 5b ). In addition, we identified few hBO specific differentially expressed genes (Extended Data Fig. 5c ). This finding implies cellular tropism of SARS-CoV-2 viral infection.

We further analysed the viral RNA sequences obtained from the infected models. In agreement with the plaque assay ( Fig. 1h and 1i) , relative transcription of SARS-CoV-2 genes plateaued by 1 dpi (Fig. 3d) , which is earlier than the host gene expression changes. Approximately 50% of the RNA sequencing reads were mappable to the SARS-CoV-2 genome in hAOs from 1 dpi (Fig. 3d) ,

indicating prevailing viral gene expression in infected hAO cells as observed in Vero cells 6 . Of note, the proportion of viral transcripts was much lower in the infected hBOs.

Transcripts from SARS-CoV-2 were not mapped uniformly to the viral genome sequence, but 3' genomic regions, where canonical subgenomic RNAs are located, showed much higher read-depth in all samples, consistent with the previous report 6 (Fig. 3e) . The vast majority of viral RNA sequences produced from the infected hAOs and hBOs was in the orientation of positive-sense RNA strands ( Fig. 3e; for example, 99.98% vs 0.02% for positive-and negative-sense RNAs, respectively, from hAO at 1 dpi). This is in good agreement with the nature of SARS-CoV-2, which is an enveloped, nonsegmented, and positive-sense RNA virus.

By cross-comparison of viral RNA sequences produced from a total of 11 infected hAOs (n=5) and hBCs (n=6), we identified 20 viral base substitutions (Supplementary Table 3 ). No mutation was at 100% variant allele fraction (VAF) and exclusive to an infected sample. Instead, sequence alterations showed a broad range of quasispecies heterogeneity in each culture (VAF ranges from 0.1% to 73.1%; Fig. 3f ), and a large proportion of the mutations (n=16; 80%) were shared by two or more infected models (by the cut-off threshold of 0.1%). Therefore, we speculate that most of these sequence changes were originally present in the pool of viral particles before their inoculation. Given the fact that these viral particles were prepared from one of the earliest COVID-19 patients, our finding suggests that mutations can accumulate in the viral genomes in a small number of rounds of viral transmissions, and appear with dramatic changes in quasispecies abundance. A substantially higher proportion of specific mutations in a sample may suggest a bottleneck in viral entry or stochasticity in viral replication.

To understand transcriptional changes of the infected hAOs at a single-cell resolution, we employed two 10X Genomics single-cell RNA-seq experiments for uninfected and infected hAOs at 3 dpi (to a throughput of 21.3 Gb and 28.5 Gb, respectively). We (Fig. 2) . The number of viral transcripts, however, was not uniformly distributed in all infected hAO cells, but enriched in Cluster 4 cells. Infected cells in Cluster 4 exhibited 13.7 times more viral UMI counts than cells in Cluster 3, on average ( Fig. 4c; 1 ,904 vs. 139 UMIs, respectively), despite cells in Cluster 4 containing relatively lower total UMI counts than ones in Cluster 3 (5,040 vs.

17,843 UMIs, respectively). When normalised with UMI counts for human genes, cells in Cluster 4 showed a >30 times higher viral RNA burden than cells in Cluster 3 (Fig 4d) . Interestingly, the infected cells in Cluster 4 showed reduced expression of canonical hAT2 marker genes, including SFTPB (Surfactant Protein B) and NKX2-1 (NK2 homeobox 1) ( Fig. 4e; Extended Data Fig. 6a ).

Compared with infected cells in Cluster 3, expression levels of ISGs, such as IFI44L and OAS3, were also highly reduced. Instead, these cells showed transcriptional induction of apoptosis mediator, GADD45B (growth arrest and DNA-damage-inducible, beta) and anti-apoptotic TNFAIP3 (tumor necrosis factor, alpha-induced protein 3), suggesting a catastrophic cellular pathway operating in a cell due to the extreme viral burdens.

Despite active protein expression ( Fig. 1f and 1g; Extended Data Fig. 1g) , we found 14 cells (0.4%) showing ACE2 transcripts, 579 cells (16.7%) expressing TMPRSS2 transcripts, and 4 cells (0.1%) coexpressing both in single-cell transcriptome sequencing (Fig. 4e) . These proportions are low at face value, but are consistent with a previous observation 41 . Although the previous report also suggested that ACE2 RNA expression can be stimulated as an infection-mediated response, particularly in human airway cells, such a trend was not observed in our dataset.

Finally, we statistically inferred the number of viral particles effectively entering each alveolar cell for infection. Although we incubated cells at an MOI of 1 on average, it is generally not known how many viral particles are necessary for effective infection of an alveolar cell. In an extreme scenario, one viral particle is sufficient. Alternatively, infection may be initiated with the entry of multiple viruses. We tracked the effective viral number of cellular entry using a mutation (NC_045512.2: 23,707C>U) as a viral barcode. From our sequencing, the mutation was estimated to be present at 4.3% VAF in the initial viral pool for innoculation. If the first scenario dominantly applies, the infected alveolar cells will ; Fig. 4f ). In a more sophisticated statistical analysis, infection by single viral entry is estimated as >2 times more frequent than by multiple viral entry (69% vs. 31%, respectively; Fig. 4g ). Our calculation indicates that a single viral particle is mainly responsible for SARS-CoV-2 infection in most alveolar cells, although multiple viral entry is also possible. It may also reflect the viral interference in SARS-CoV-2 alveolar infection.

In this study, we established conditions for optimised 3D long-term cultures of adult hAT2 cells, which provided an essential tool for studying initial intrinsic responses of SARS-CoV-2 infection. Single hAT2 cells were capable of self-constructing alveolus-like structures consisting of AT2 cell and differentiated AT1 cells. Mature hAT2 cells were maintained >10 months over multiple passages although self-renewal capacity and growth rate was reduced after 6 month in cultures. hAT1 cells were also lost from later culture, likely due to the persistent exposure to high WNT conditions allowing expansion of hAT2 cells over differentiation. Alteration of WNT activity in culture media (differentiation media) may enable to induce further AT1 cell differentiation.

hAOs showed remarkable phenotypic changes in the first few days after SARS-CoV-2 innoculation. The interferon response is the first line of host antiviral defense 40 . Contrary to a recent report 42 , we observed substantial ISGs in the alveolar cells induced by endogenously produced type I and III interferons. However, the induction of interferon responses was seen at 3 dpi in hAO models, 1-2 days later than the timing of viral amplification at 1 dpi. The timing of ISG induction may be earlier in vivo in concert with exogenous interferons from immune cells. For more physiological understanding, co-culturing SARS-CoV-2 infected hAO models with immune cells obtained from the same donor will be helpful.

In summary, our study highlights the power of feeder-free hAOs to elucidate the intrinsic responses of tissue damage including virus infection. Our data, including high-resolution electron microscopic images and the list of gene expression changes following infection, will be a great resource for the biomedical community to provide a deeper characterisation of SARS-CoV-2 infection specifically within adult hAT2 cells. We believe that our hAO models will enable more accurate and sophisticated analyses in the very near future, especially for studying the response of viral infection within vulnerable groups such as aged or diseased lungs, providing the opportunity to elucidate individual patient responses to viral infection. Furthermore, our models can be applied to other techniques, such as co-culture experiments with immune cells and robust in vitro screening of antiviral agents applicable to alveolar cells, in addition to being applicable for the study of the basic biology of alveolar cells as well as chronic disorders of the lung.

For the establishment of human lung organoid models, human distal lung parenchymal tissues from as with HTII-280 + , although with a few minor differences. Bronchial organoids (hBOs) were passaged every 21-28 days due to accelerated growth compared with alveolar organoids (hAOs), and were cultured in previously reported medium conditions 30 with the following concentration/factor edits; 100 ng/ml human FGF10, 10% R-SPONDIN-1, 10 µM SB431542 (instead of A83-01). 

Organoids were fixed and embedded in a paraffin block 43 . Pre-cut 7 µM paraffin sections were dewaxed and rehydrated (sequential immersion in xylene, 100% EtOH, 90% EtOH, 75% EtOH, distilled water) and either stained with hematoxylin and eosin (H&E) or immunostained. For antigen retrieval, slides were submerged into pre-heated citrate antigen retrieval buffer (10 mM sodium citrate, pH 6.0) and allowed to boil for 15 min. Slides were cooled in a buffer for 20 min, washed in running water for 3 min, and permeabilised with 0.3% Triton-X in PBS for 15 min. Cells were protected from light and imaged immediately using an EVOS cell imaging system.

Organoids were fixed in 4% paraformaldehyde (PFA) for 3 hrs at ice, and then dehydrated in PBS with 30% sucrose (v/v) (Sigma). Organoids were embedded with optimal cutting temperature (OCT) compound (Leica) and cut with 10 µM. Organoid section was blocked with 5% normal donkey serum in PBS 1% triton-X (Sigma). Sections were incubated with primary antibodies overnight at 4 °C, 

Freshly sorted HTII-280 + and HTII-280cells were lysed with TRIzol, and RNA was extracted. RNA was reverse transcribed using SuperScript IV (Thermo Fisher Scientific), and were assessed using the following Taqman probes; SFTPC (Hs00951326_g1), TP63 (Hs01114115_m1), SCGB1A1 (Hs00171092_m1).

Viral RNA samples were reverse-transcribed using SuperScript IV (Thermo Fisher Scientific). Viral N3 gene was targeted for qRT-PCR. Nucleotide sequences of the probes as below (CDC). 

Matrigel was sheared with the organoid media and frozen at -80 °C once. Thaw the solution and dilute by scale of 10. Each well containing Vero cells in 12 wells were infected with the diluted solution respectively at 37 °C, 5% CO2 for 1 hr. After infection, remove infection media and wash the Vero cells with PBS two times, mixed agar and Modified Eagle's Medium (Thermofisher) were poured on each well. When agar mixture was hardened, fix each well with 4% PFA for 3 days, and stain with crystal violet (Sigma). When there are individual spots, the original solution's viral titer was calculated.

Extracted cellular RNA was processed through Truseq Stranded Total RNA Gold kit, and cDNA library was sequenced 2 x 100 bp using Hiseq 2500. Fastq file was aligned to GRCh38 with virus sequence (NC 045512.2 from NCBI) using STAR 46 and normalized RNA expression was calculated using RSEM 47 . Differentially expressed genes are found from DEseq2 48 . We obtained enriched gene sets using in-house scripts. For mutation calling, we used Strelka2 49 , Varscan2 50 , and Samtools 51 , and then manually checked the position through IGV 52 .

Fastq file was aligned and each UMI count was calculated using Cell Ranger software provided by the manufacturer (10X Genomics). Cells with mitochondria RNA percent < 25%, total RNA number > 200 subsets were used for downstream analysis.

Starting from the 10X gene counts, we have normalized data as follows. The 10X data includes SARS-CoV-2 genome as an extra gene besides 19,941 human genes. For uninfected cells, such SARS-CoV-2 gene would have zero read count, while for infected cells, the gene could account for a large portion of total reads. One of the goals of normalisation is to remove technical difference such as sequencing coverage before comparing gene expression levels. As our interest is to compare expression levels between uninfected and infected cells on human genes, we have applied a normalisation method ('scater R package's 'logNormCounts' function) to human genes as a whole set.

The method calculated a scaling factor for each cell based on total human gene count, then scaled all genes before taking log-transformation. Using the same scaling factor learned during human gene normalisation, SARS-CoV-2 count was also normalized. For clustering, we combined single-cell data from infected and uninfected hAOs. Unsupervised clustering was performed using a shared nearest neighbor (SNN) based clustering algorithm in Seurat 53 . Contaminated cells (< 3%) were discarded. In house R scripts were used for more downstream analyses.

To assess whether alveolar cells tend to be infected by a single viral particle or multiple particles, we employed a likelihood approach. As a proof-of-concept, we assumed only two scenarios exist, one supporting a single viral entry and the other supporting double viral entry, then aimed to estimate the proportion of cells with a single viral particle ( ). Out data consists of the observed reference ( cell ) and variant ( ) read counts for each of the 547 reporting at least one read at the mutation site of NC_045512.2:23,707. Assuming a sequencing error rate (ε) of 0.1%, which will cover any Illumina sequencing errors or misalignment, the likelihood of data given the weight supporting a single virus scenario ( ) was computed as follows. 

We observed a distinct transcriptional feature of cells in Cluster 2 which express lower levels of canonical hAT2 marker genes, including SFTPC but detectable levels of airway marker genes, including SOX2, TP63, KRT5, and KRT17 (Extended Data Fig. 6b) . These expression patterns were not affected by virus infection. hAT2 cells expressing airway markers such as SOX2 were seen in chronic lung diseases such as lung cancer and idiopathic pulmonary fibrosis (IPF) 54, 55 , representing pathologic phenotypes of alveolar bronchiolization. A recent study also suggested the potential transition of hAT2 cells to KRT5 + basal-like cells in the context of IPF 27 . Given the fact that the hAOs used for our singlecell RNA sequencing study were derived from hAT2 cells isolated from adjacent normal counterparts of lung cancer and/or IPF, it is likely that this transcriptional feature reflects the cellular status of original tissues rather than virus-associated phenotype. This finding suggests that our hAO models maintain the pathophysiologic features of original tissues although we used apparently normal background regions for our hAO establishments. Further long-term tracing of changes in cellular identities and states in response to virus infection in hAO cells will be of significant interest to understand the progression of pathologic features and reparative mechanisms for developing therapeutic interventions.

Furthermore, from our scRNAseq analysis, most captured cells were hAT2 cells. It is likely that this might result from the enrichment of hAT2 cells in our hAOs (P2) and the nature of fragile hAT1 cells during the procedure of single-cell preparation for scRNAseq.

Supplementary Table 1 . RNA expression levels (TPM) of all genes in seven human alveolar organoid samples. 

The authors declare no competing interests.

All unique organoids generated in this study are available from Young Seok Ju or Joo-Hyeon Lee with a completed Materials Transfer Agreement.

Bulk RNA and single cell RNA sequencing datasets will be uploaded on the European Genome-Phenome Archive (EGA). Accession ID is not assigned yet. 

Human Coronavirus: Host-Pathogen Interaction

A new coronavirus associated with human respiratory disease in China

Health Organization Coranavirus disease (COVID-19) Situation Report -140

Integrated analyses of single-cell atlases reveal age, gender, and smoking status associations with cell type-specific expression of mediators of SARS-CoV-2 viral entry and highlights inflammatory programs in putative target cells

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is Blocked by a Clinically Proven Protease Inhibitor

The Architecture of SARS-CoV-2 Transcriptome

Structure, Function, and Antigenicity of the SARS-CoV-2 Spike Glycoprotein

Structural and Functional Basis of SARS-CoV-2 Entry by Using Human ACE2

Comparative tropism, replication kinetics, and cell damage profiling of SARS-CoV-2 and SARS-CoV with implications for clinical manifestations, transmissibility, and laboratory studies of COVID-19: an observational study

The pathogenicity of SARS-CoV-2 in hACE2 transgenic mice

Infection and Rapid Transmission of SARS-CoV-2 in Ferrets

Pathogenesis and transmission of SARS-CoV-2 in golden hamsters

Virological assessment of hospitalized patients with COVID-2019

The proximal origin of SARS-CoV-2

Phylogenetic network analysis of SARS-CoV-2 genomes

Structural basis of receptor recognition by SARS-CoV-2

Organoids as an in vitro model of human development and disease

Modelling Cryptosporidium infection in human small intestinal and lung organoids

In vitro expansion of human gastric epithelial stem cells and their responses to bacterial infection

Organoids of human airways to study infectivity and cytopathy of SARS-CoV-2

SARS-CoV-2 productively infects human gut enterocytes

Inhibition of SARS-CoV-2 Infections in Engineered Human Tissues Using Clinical-Grade Soluble Human ACE2

A Human Pluripotent Stem Cell-based Platform to Study SARS-CoV-2 Tropism and Model Virus Infection in Human Cells and Organoids

Lung organoids: current uses and future promise

Type 2 alveolar cells are stem cells in adult lung

Regeneration of the lung alveolus by an evolutionarily conserved epithelial progenitor

Human alveolar Type 2 epithelium transdifferentiates into metaplastic KRT5+ basal cells during alveolar repair

HTII-280, a biomarker specific to the apical plasma membrane of human lung alveolar type II cells

Human embryonic lung epithelial tips are multipotent progenitors that can be expanded in vitro as long-term self-renewing organoids

Long-term expanding human airway organoids for disease modeling

Identification of Coronavirus Isolated from a Patient in Korea with COVID-19

Generation of human bronchial organoids for SARS-CoV-2 research

SARS-coronavirus replication is supported by a reticulovesicular network of modified endoplasmic reticulum

Acute and persistent infection of human neural cell lines by human coronavirus OC43

Infectious bronchitis virus generates spherules from zippered endoplasmic reticulum membranes

SARS-coronavirus-2 replication in Vero E6 cells: replication kinetics, rapid adaptation and cytopathology

Early local immune defences in the respiratory tract

Interferon-stimulated genes: a complex web of host defenses

Differential Activation of the Transcription Factor IRF1 Underlies the Distinct Immune Responses Elicited by Type I and Type III Interferons

Type I and Type III Interferons -Induction, Signaling, Evasion, and Application to Combat COVID-19

SARS-CoV-2 Receptor ACE2 Is an Interferon-Stimulated Gene in Human Airway Epithelial Cells and Is Detected in Specific Cell Subsets across Tissues

Imbalanced Host Response to SARS-CoV-2 Drives Development of COVID-19

Lung stem cell differentiation in mice directed by endothelial cells via a BMP4-NFATc1-thrombospondin-1 axis

Human Intestinal Organoids Maintain Self-Renewal Capacity and Cellular Diversity in Niche-Inspired Culture Condition

Clinical diagnosis of early dengue infection by novel one-step multiplex realtime RT-PCR targeting NS1 gene

STAR: ultrafast universal RNA-seq aligner

RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome

Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2

Strelka: accurate somatic small-variant calling from sequenced tumornormal sample pairs

VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome sequencing

The Sequence Alignment/Map format and SAMtools

Integrative genomics viewer

Comprehensive Integration of Single-Cell Data

Bronchiolization of the alveoli in lung cancer: pathology, patterns of differentiation and oncogene expression

Single-cell RNA sequencing identifies diverse roles of epithelial cells in idiopathic pulmonary fibrosis

We thank Jinwook Choi