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

Summary There is pressing urgency to understand the pathogenesis of the severe acute respiratory syndrome coronavirus clade 2 (SARS-CoV-2), which causes the disease COVID-19. SARS-CoV-2 spike (S) protein binds angiotensin-converting enzyme 2 (ACE2), and in concert with host proteases, principally transmembrane serine protease 2 (TMPRSS2), promotes cellular entry. The cell subsets targeted by SARS-CoV-2 in host tissues and the factors that regulate ACE2 expression remain unknown. Here, we leverage human, non-human primate, and mouse single-cell RNA-sequencing (scRNA-seq) datasets across health and disease to uncover putative targets of SARS-CoV-2 among tissue-resident cell subsets. We identify ACE2 and TMPRSS2 co-expressing cells within lung type II pneumocytes, ileal absorptive enterocytes, and nasal goblet secretory cells. Strikingly, we discovered that ACE2 is a human interferon-stimulated gene (ISG) in vitro using airway epithelial cells and extend our findings to in vivo viral infections. Our data suggest that SARS-CoV-2 could exploit species-specific interferon-driven upregulation of ACE2, a tissue-protective mediator during lung injury, to enhance infection.


In Brief
Analysis of single-cell RNA-seq datasets from human, non-human primate, and mouse barrier tissues identifies putative cellular targets of SARS-CoV-2 on the basis of ACE2 and TMPRSS2 expression. ACE2 represents a previously unappreciated interferon-stimulated gene in human, but not mouse, epithelial tissues, identifying anti-viral induction of a host tissue-protective mechanism, but also a potential means for viral exploitation of the host response.

INTRODUCTION
Human coronaviruses (CoVs) are single-stranded positive-sense RNA viruses that can cause mild to severe respiratory disease (Fung and Liu, 2019). Over the past two decades, zoonotic transmission events have led to the emergence of two highly patho-genic CoVs: severe acute respiratory syndrome (SARS)-CoV and Middle East respiratory syndrome (MERS)-CoV. SARS-CoV-2, which causes the disease known as COVID-19, was first reported in late 2019 (Coronaviridae Study Group of the International Committee on Taxonomy of, 2020;Lu et al., 2020;Paules et al., 2020). COVID-19 is characterized by pneumonia, fever, cough, and occasional diarrhea (Guan et al., 2020;Holshue et al., 2020;Huang et al., 2020), and SARS-CoV-2 RNA has been reliably detected in nasopharyngeal swabs, sputum, and stool samples Wö lfel et al., 2020;Zou et al., 2020). As of April 19, 2020, SARS-CoV-2 continues to spread worldwide, and there are over 2,401,379 confirmed cases, 165,044 deaths, and 623,903 recovered individuals in 185 countries and regions (Dong et al., 2020a). Early models of COVID-19 transmission dynamics estimate one infectious individual infects slightly over two individuals; travel restrictions reduce that spread to one in-dividual, although these figures might evolve as more accurate epidemiological data become available (Kucharski et al., 2020).
Work during the first SARS-CoV epidemic identified the human host factor angiotensin-converting enzyme 2 (ACE2) as the receptor for SARS-CoV (Li et al., 2003). SARS-CoV-2 spike (S) protein has been experimentally shown to bind ACE2 on host cells with significantly higher affinity than SARS-CoV-S (Hoffmann et al., 2020;Wrapp et al., 2020). The main host protease that mediates S protein activation on primary target cells and initial viral entry is the type II transmembrane serine protease TMPRSS2 (Glowacka et al., 2011;Hoffmann et al., 2020;Iwata-Yoshikawa et al., 2019;Matsuyama et al., 2010;Shulla et al., 2011;Walls et al., 2020). Other host proteases, such as furin, have also been suggested to promote the pathogenesis of this pandemic SARS-CoV-2 clade, but when and where they process S protein remains to be determined (Bö ttcher-Friebertshä user et al., 2013;Bugge et al., 2009;Coutard et al., 2020;Walls et al., 2020). Binding of SARS-CoV-S to ACE2 results in receptor-mediated internalization (Grove and Marsh, 2011;Kuba et al., 2005). Importantly, ACE2 functions as a key tissue-protective component during severe acute lung injury Kuba et al., 2005).
A tissue-level basis for understanding SARS-CoV tropism was proposed based on ACE2 histological staining and expression in human epithelia of the lung and small intestine (Hamming et al., 2004;Harmer et al., 2002;Jonsdottir and Dijkman, 2016). However, unlike the specific expression of CDHR3 (the rhinovirus-C receptor), which is resolved to ciliated epithelial cells of the upper airway (Griggs et al., 2017), the specific cell subsets within each tissue that express ACE2 remain unknown. Identifying the cell subsets targeted by SARS-CoV-2 (ACE2 + ) and those at greatest risk of direct infection (ACE2 + TMPRSS2 + ) is critical for understanding and modulating host defense mechanisms and viral pathogenesis.
Although the host response to SARS-CoV highlighted a role for IFNs, most studies assessed the effect of IFN restriction in cell lines that might not fully recapitulate the repertoire of ISGs present in primary human target cells (Bailey et al., 2014;de Lang et al., 2006;Sainz et al., 2004;Zheng et al., 2004). One study of SARS-CoV suggested the timing of the type I IFN response was critical in vivo (Channappanavar et al., 2016). Clinical therapy using approved IFNs has been attempted for SARS-CoV, MERS-CoV, and SARS-CoV-2 in the absence of a controlled trial to mixed effect, resulting in anecdotal evidence suggesting either rapid improvement or worsening of symptoms (Dong et al., 2020b;Lei et al., 2020;Li and De Clercq, 2020). Elucidating tissue-and cell-type-specific ISGs and their activity is essential for understanding the role of IFNs in host defense during human SARS-CoV-2 infection.
Massively parallel single-cell RNA-sequencing (scRNA-seq) is transforming our ability to comprehensively map the cell types, subsets, and states present during health and disease in barrier tissues (Ordovas-Montanes et al., 2020;Ordovas-Montanes et al., 2018;Smillie et al., 2019). This has been particularly evident in the elucidation of novel human epithelial and stromal cell subsets and states Regev et al., 2017;Ruiz García et al., 2019;Schiller et al., 2019;Smillie et al., 2019;Vieira Braga et al., 2019). Recently, scRNA-seq has been applied to better understand the cellular variation present during viral infection in vitro and in vivo (Russell et al., 2018;Steuerman et al., 2018). Global single-cell profiling efforts such as the Human Cell Atlas (HCA) initiative are ideally poised to rapidly share critical data and enhance our understanding of disease during emergent public health challenges (Sungnak et al., 2020).
Here, using published and unpublished datasets (all from non-SARS-CoV-2-infected samples), we analyze human, NHP, and mouse tissues that have been clinically identified to harbor virus in patients exhibiting COVID-19 symptoms. We provide a cautionary note on the interpretation of the scRNA-seq data presented below, given that many factors such as dissociation, profiling method, and sequencing depth can influence results (STAR Methods). Here, we focus our analysis and discussion on the specific subsets where ACE2 and TMPRSS2 are enriched and on relative comparisons within each dataset, rather than between datasets or equivalence to absolute numbers of total cells. Across several studies of human and NHP tissues, we found ISGs upregulated in ACE2-expressing cells.
Strikingly, by treating primary human upper airway basal cells with distinct types of inflammatory cytokines, we demonstrate that IFN-a drives ACE2 expression. Human influenza infection also induces broader expression of ACE2 in upper airway epithelial cells and is corroborated by publicly available databases. Overall, our data provide motivation to better understand the trade-offs of antiviral and/or IFN therapy in humans infected with SARS-CoV-2 in order to balance host restriction, tissue tolerance, and viral enhancement mechanisms (Davidson et al., 2015;Fung and Liu, 2019;Imai et al., 2005;Iwasaki et al., 2017;Kuba et al., 2005;Lei et al., 2020;Medzhitov et al., 2012;Zou et al., 2014). Importantly, although our findings identify similar cell subsets enriched for Ace2 in mice, neither in vitro nor in vivo IFN-stimulation nor in vivo viral challenge substantially alter Ace2 expression levels. The dynamic, species-specific and multifaceted role of IFN raises implications for pre-clinical COVID-19 disease modeling.

Lung Epithelial Cell Expression of Host Factors Used by SARS-CoV-2 in Non-Human Primates and Humans
To investigate which cells within human and NHP tissues represent likely SARS-CoV-2 targets, we analyzed new and existing scRNA-seq datasets to assess which cell types express ACE2, alone or with TMPRSS2. In a previously unpublished dataset consisting of NHP (Macaca mulatta) lung tissue collected after necropsy of healthy adult animals and analyzed by using Seq-Well v1 , we recovered at least 17 distinct major cell types, including various lymphoid, myeloid, and stromal populations (Figures 1A-1C; Table S1; STAR Methods). ACE2 and TMPRSS2 were primarily expressed in epithelial cells, with 6.7% of type II pneumocytes expressing ACE2 and 3.8% co-expressing ACE2 and TMPRSS2 ( Figures 1B and 1C).
Next, we compared ACE2 + with ACE2 À type II pneumocytes to explore broader gene programs that differentiate putative SARS-CoV-2 target cells from cells of a similar phenotype and ontogeny ( Figure 1D; Table S1). Among genes significantly upregulated in ACE2 + type II pneumocytes, we observed IFNGR2 (false discovery rate [FDR]-adjusted p = 0.022), a receptor for type II IFNs. Notably, previous work has demonstrated limited anti-viral potency of IFN-g for SARS-associated coronaviruses, compared with that of type I IFNs, at least in vitro (Sainz et al., 2004;Zheng et al., 2004). Other co-regulated genes of potential interest include TRIM27 (FDR-adjusted p = 0.025), as well as NT5DC1 (FDR-adjusted p = 0.003) and ARL6IP1 (FDR-adjusted p = 0.047), which were upregulated in the A549 adenocarcinoma alveolar basal epithelial cell line after exposure to IFN-a and IFN-g for 6 h (Sanda et al., 2006). We found IFNAR1 consistently expressed among both ACE2 + type II pneumocytes and ACE2 + TMPRSS2 + co-expressing type II pneumocytes, but its level of upregulation compared with all remaining pneumocytes did not meet statistical significance (FDR-adjusted p = 0.11). This analysis finds ACE2 + cells enriched within a rare fraction of secretory cells in NHPs and that ACE2 expression is co-regulated with genes involved in IFN responses.
To assess whether the findings from NHP lung cells were similarly present in humans, we analyzed a previously unpublished scRNA-seq dataset derived from surgical resections of fibrotic lung tissue collected with Seq-Well S 3 . Unsupervised analysis identified multiple cell types and subtypes of immune cells (Figures 2A-2C; STAR Methods), as defined by the genes displayed in Figure 2C (full lists available in Table S2). Here, we found that ACE2 and TMPRSS2 were primarily expressed within type II pneumocytes and ciliated cells, in line with our analysis of the NHP-derived cells (Figures 1 and 2A, (A) Schematic of protocol for isolation of human lung tissue from surgical excess, creation of scRNA-seq libraries by using Seq-Well S 3 , and computational analysis to identify cell types by using unbiased methods. Shown on the right is a UMAP projection of 18,915 cells across 8 donors (n = 3 TB + HIV + ; n = 3 TB + ; n = 2 non-infected patients). Cells represented by points, colored according to cell type (see STAR Methods). (B) UMAP projection as in (A), points colored by detection of ACE2 (top) or TMPRSS2 (bottom). Color coding is as follows: black, RNA positive; blue, RNA negative. (C) Dot plot of 2 defining genes for each cell type (FDR-adjusted p < 0.001), and ACE2 and TMPRSS2; dot size represents fraction of cells within cell type expressing a given gene, and color intensity represents binned count-based expression amount (log(scaled UMI+1)) among expressing cells. All cluster-defining genes are provided in Table S2. Red arrow indicates cell types with largest proportion of ACE2 + TMPRSS2 + cells. (D) Volcano plot identifying significantly upregulated genes in ACE2 + TMPRSS2 + pneumocytes compared with all remaining pneumocytes. Red points represent genes with a FDR-adjusted p < 0.05, and log 2 (fold change) >1.5. Text highlighting specific genes; the full list is available in Table S2. (E) Expression of ACE2 across human donors by HIV and TB status (p = 0.009 by likelihood-ratio test). See also Table S2. Figure 3. NHP and Human Ileal Absorptive Enterocytes Co-Express ACE2 and TMPRSS2 (A) Expression ACE2 across diverse tissues in healthy NHPs (n = 3 animals; 52,858 cells). (B) Schematic of protocol for isolation of NHP ileum (n = 5) at necropsy for scRNA-seq using Seq-Well v1, and computational pipeline to identify cell types by using unbiased methods. Shown on the right is a UMAP projection of 4,515 cells colored by cell type. (C) Dot plot of 2 defining genes for each cell type, with ACE2 and TMPRSS2. Dot size represents fraction of cells within cell type expressing a given gene, and color intensity represents binned count-based expression amounts (log(scaled UMI+1)) among expressing cells. All cluster defining genes are provided in Table S4. Red arrow indicates cell type with largest proportion of ACE2 + TMPRSS2 + cells. (D) Schematic of protocol for isolation of human ileal cells from endoscopic pinch biopsies in non-inflamed regions (n = 13). Shown on the right is a tSNE plot of 13,689 epithelial cells selected from original dataset generated by 10x 3 0 v2 (see Figure S2), colored by cellular subsets.
(legend continued on next page) 2B). In type II pneumocytes (identified by unique expression of surfactant proteins SFTPC, SFTPB, and SFTPA1), we found 1.4% of cells expressing ACE2 (FDR-adjusted p = 1.35EÀ21), 34.2% expressing TMPRSS2 (FDR-adjusted p < 1EÀ300), and 0.8% co-expressing both. In ciliated cells, we found 7% were ACE2 + (FDR-adjusted p = 5EÀ64), 24.6% were TMPRSS2 + (FDR-adjusted p = 3.8EÀ30), and 5.3% co-expressed both. As above, to assess for cellular pathways significantly co-expressed within putative target cells for SARS-CoV-2, we computed differentially expressed genes between ACE2 + TMPRSS2 + type II pneumocytes and all other type II pneumocytes ( Figures  2C and 2D; Table S2). We found significant enrichment of BATF among ACE2 + TMPRSS2 + cells (FDR-adjusted p = 3.25EÀ7), which has been demonstrated previously to be upregulated by type I and type II IFNs (Murphy et al., 2013). Of note, we also observed TRIM28 co-expressed with ACE2 and TMPRSS2 among type II pneumocytes in this dataset (FDRadjusted p = 2.34EÀ9), which might play a role in potentiating an IFN response in lung epithelial cells (Krischuns et al., 2018). Within this cohort of donors, 3 individuals were human immunodeficiency virus (HIV) + and diagnosed with active tuberculosis, 3 donors had active tuberculosis and were HIV À , and 2 were negative for both pathogens. Surprisingly, we found that all of the ACE2 + cells across all cell types were derived from HIV + Mycobacterium tuberculosis (Mtb) + donors despite approximately equivalent recovery of epithelial cell types from all donors (likelihood-ratio test, p = 0.009) ( Figure 2E). Given limited cell and patient numbers combined with potential sampling biases, we caution that this observation requires much broader cohorts to validate a potential role for co-infections; still, we note our observation is suggestive of a role for chronic IFNs in the induction of ACE2, given that HIV infection is associated with persistent upregulation of ISGs, and we observed elevated amounts of IF-NAR2, IFI30, and IKBKB (Utay and Douek, 2016) (FDR-adjusted p = 1.1EÀ6, 8.8EÀ9, 1.57EÀ7, respectively; HIV + versus HIV À epithelial cells).
Next, using a previously unpublished scRNA-seq dataset consisting of granuloma and adjacent, uninvolved lung samples from Mtb-infected NHPs (Macaca fascicularis) collected with Seq-Well S 3 , we identified subsets of epithelial cells expressing ACE2 and TMPRSS2 ( Figure S1; Table S3; STAR Methods). The majority of ACE2 + TMPRSS2 + cells were, once again, type II pneumocytes (22%) and type I pneumocytes (9.7%) and were largely enriched within granulomatous regions compared with those in adjacent uninvolved lung (Figures S1B and S1C) (p = 0.006, Fisher Exact Test). ACE2 + TMPRSS2 + type II pneumocytes expressed significantly higher amounts of antimicrobial effectors such as LCN2 compared with remaining type II pneumocytes ( Figure S1D). Cells with club cell/secretory, type I pneumocyte, and ciliated cell types also contained some ACE2 + TMPRSS2 + cells, but we did not have sufficient power to detect significantly differentially expressed genes between these cells and other cells within those clusters. Altogether, we identify ACE2 + TMPRSS2 + cells in lower airways of humans and NHPs with consistent cellular phenotypes and evidence supporting a potential role for IFN-associated inflammation in upregulation of ACE2.
Ileal Absorptive Enterocytes Express Host Factors Used by SARS-CoV-2 Next, we examined several other tissues for ACE2-expressing cells on the basis of the location of hallmark symptoms of COVID-19, focusing on the gastrointestinal tract due to reports of clinical symptoms and viral shedding . Leveraging a previously unpublished scRNA-seq atlas of NHP (M. mulatta) tissues collected with Seq-Well v1, we observed that the majority of ACE2 + cells reside in the small intestine, principally within the ileum, jejunum, and, to a lesser extent, the liver and colon ( Figure 3A; STAR Methods). Critically, we note that, in this experiment, the dissociation method used on each tissue was optimized to preserve immune cell recovery, and therefore under-sampled stromal and epithelial populations, as well as neurons from the brain. Within the ileum, we identified ACE2 + cells as absorptive enterocytes on the basis of specific expression of ACE2 within cells defined by APOA1, SI, FABP6, and EN-PEP, among others, by a likelihood-ratio test (Figures 3B and 3C) (p < 1EÀ300, 62% of all absorptive enterocytes; see Table S4). All other epithelial subtypes expressed ACE2 to a lesser extent, and variably co-expressed ACE2 with TMPRSS2 (see Table S4 for full statistics).
Persistent viral RNA in rectal swabs has been detected in pediatric infection, even after negative nasopharyngeal tests . In an additional dataset consisting of endoscopic biopsies from the terminal ileum of a human pediatric cohort (n = 13 donors, ranging in age from 10 to 18 years old), collected with 10X 3 0 v2, we confirmed a large abundance of ACE2 + cells with selective expression within absorptive enterocytes (29.7% ACE2 + , FDR-adjusted p = 2.46EÀ100) (Figures 3D and 3E; Table  S5; STAR Methods). Furthermore, we identified a subset (888 cells, $6.5% of all epithelial cells) that co-express both genes ( Figures S2A-S2C). We performed differential expression testing and GO-term enrichment using these cells relative to matched non-expressers to highlight putative biological functions enriched within them, such as metabolic processes and catalytic activity, and to identify shared phenotypes of ACE2 + TMPRSS2 + ileal cells across both human and NHP cohorts (Table S5). We speculate that viral targeting of these cells, taken from patients without overt clinical viral infection, might help explain intestinal symptoms. Finally, we compared ileal absorptive enterocytes from healthy NHPs and NHPs infected with simian-human immunodeficiency virus (SHIV) and then treated for 6 months with antiretroviral therapy (animal and infection characteristics published in Colonna et al., 2018) (STAR Methods). We found significant upregulation of ACE2, STAT1, and IFI6 within the absorptive (E). Dot plot of 2 defining genes for each cell type, with ACE2 and TMPRSS2. Dot size represents fraction of cells within cell type expressing a given gene, and color intensity represents binned count-based expression amounts (log(scaled UMI+1)) among expressing cells. All cluster defining genes are provided in Table  S5. Red arrow indicates cell type with largest proportion of ACE2 + TMPRSS2 + cells. (F). Expression of ACE2 (left) and TMPRSS2 (right) among all epithelial subsets from human donors. See also Figure S2 and Tables S4 and S5.

Upper Airway Expression of Host Factors Used by SARS-CoV-2
To identify potential viral target cells in nasal and sinus tissue, two regions that are frequently primary sites of exposure for coronaviruses, we analyzed existing scRNA-seq datasets from the human upper airway (inferior turbinate and ethmoid sinus mucosa) across a spectrum of healthy donors and individuals with allergic inflammation due to chronic rhinosinusitis (CRS) collected with Seq-Well v1 ( Figure  . We speculate that these cells, taken from clinically non-virally infected patients, yet constantly exposed to environmental viruses, might provide one of the earliest locations for coronaviruses to infect before spreading to other tissues. We observed significant enrichment of ACE2 expression in apical epithelial cells and, to a lesser extent, ciliated cells compared with all cell types recovered from surgically resected mucosa (1% of apical epithelial cells, FDR-adjusted p = 4.55EÀ6, n.s. in ciliated cells) ( Figure 4B; Table S6).
To better map putative SARS-CoV-2 targets among epithelial subsets, we employed a finer-grained clustering method applied to both ethmoid sinus surgical specimens and scrapings from the inferior turbinate and ethmoid sinus ( Figures 4C-4F). Once again, we observed selective expression of ACE2 within a minority of cell types, with 1.3% of all secretory cells expressing ACE2 ( Figure 4C) (FDR-adjusted p = 0.00023), specifically sub-clusters 7 and 13, which represent two varieties of secretory epithelial cell (Figures 4C, 4F, and 4G). Cluster 7 secretory cells are marked by S100P, LYPD2, PSCA, CEACAM5, and STEAP4; encompass some MUC5AC goblet cells; and contain the most significantly enriched ACE2 and TMPRSS2 expression (4% express ACE2, FDR-adjusted p = 7.32EÀ28; 28% express TMPRSS2, FDRadjusted p = 2.15EÀ132; Table S6). We next explicitly gated cells by their TMPRSS2 and ACE2 expression, identifying a rare subset that co-expresses both, the majority of which fall within the ''Secretory Cluster 7'' cell type ( Figures 4E and 4F) (30 cells, $0.3% of all upper airway secretory cells, 1.6% of goblet ''Secretory Cluster 7''). These findings are aligned with concurrent work by the HCA Lung Biological Network on human nasal scRNA-seq data, which identified nasal secretory cells to be enriched for ACE2 and TMPRSS2 expression (Sungnak et al., 2020).
Although we identified co-expression of ACE2 and TMPRSS2 in few airway cells overall, we detected ACE2 and TMPRSS2 single-and double-positive cells in over 20 donors and thus posit that these genes are enriched in secretory cells and are not a product of individual-patient-driven variability ( Figure S3A). Inferior turbinate scrapings collected on Seq-Well S 3 , which increases the resolution of lower-abundance transcripts compared with Seq-Well v1, revealed consistent and specific expression restricted to goblet secretory cells, but at a greater detection frequency in samples from the same donors (Figure S3B) (ACE2 + from 4.7% v1 to 9.8% S 3 ; ACE2 + TMPRSS2 + from 1.9% v1 to 4% S 3 ) . Using the gated ACE2 + TMPRSS2 + cells, we tested for differentially expressed genes compared to the remaining secretory epithelial cells (full results provided in Table S6). Notably, we observed significant upregulation of ADAR, GBP2, OAS1, JAK1, and DUOX2 (FDR adjusted, all p < 0.02) within ACE2 + TMPRSS2 + cells, potentially indicative of IFN signaling ( Figure 4G). Almost all ''Secretory Cluster 7'' cells were from inferior turbinate scrapings of healthy and allergically inflamed individuals, few cells were from the ethmoid sinus tissue of patients with chronic rhinosinusitis without nasal polyps, and no cells were detected in polyp tissue (Figure 4H). Gene Ontology (GO) analysis of enriched genes in double-positive cells include processes related to intracellular cytoskeleton and macromolecular localization and catabolism, potentially involved in viral particle entry, packaging, and exocytosis (Fung and Liu, 2019).
We next utilized IFN-inducible gene sets of relevance to human airway epithelial cells, which we derived from a prior study by performing differential expression on a published dataset intensity represents binned count-based expression amounts (log(scaled UMI+1)) among expressing cells (see Table S6 for statistics by subset). Red arrow indicates cell types with largest proportion of ACE2 + TMPRSS2 + cells.  (D), identifying epithelial cells co-expressing ACE2 and TMPRSS2 (30 cells, black points). (F) tSNE as in (D), colored by detailed cell types with higher granularity, as in (C). (G) Individual differentially expressed genes between ACE2 + TMPRSS2 + cells and all other secretory epithelial cells (see Table S6 for full gene list with statistics). Bonferroni-adjusted likelihood-ratio test p < 0.02 for all genes displayed. (H) Stacked bar plot of each subset of epithelial cells among all epithelial cells by donor (each bar) and sampling location (noted below graph) (unpaired t test p < 0.00035 for Secretory Goblet 7 inferior turbinate versus ethmoid sinus; see Table S6 for raw values). (I) Violin plot of cell clusters in respiratory epithelial cells (from Figures 4C and 4F) ordered by average expression of IFN-a-induced gene signatures, presented as a gene module score; non-normal distribution by Lilliefors test, Mann-Whitney U-test p = 2.2EÀ16, 1.21 effect size, IFN-a signature for Secretory Goblet Cluster 7 versus all epithelial cells. Arrow indicates cluster containing majority ACE2 + TMPRSS2 + cells. See also Figure S3 and Table S6. where air-liquid interface cultures from primary human nasal epithelial cells were treated with IFN-aA/D, IFN-b1a, IFN-g, IL-4, or IL-13 (Giovannini-Chami et al., 2012;Ordovas-Montanes et al., 2018). Using these gene lists, we scored the human nasal epithelial cells analyzed by scRNA-seq described in Figures 4C and 4F and found significant concomitant upregulation of the IFN-a-stimulated gene set within ACE2 + TMPRSS2 + secretory goblet cluster 7 ( Figure 4I).

Type I Interferon IFN-a Drives ACE2 Expression in Primary Human Nasal Epithelial Cells
The meta-analysis described above consistently identified an association between ACE2 expression and canonical ISGs or components of the IFN-signaling pathway. This prompted us to investigate whether IFNs might play an active role in regulating ACE2 expression levels in specific target cell subsets, thus potentially allowing for a tissue-protective host response or increased viral binding of SARS-CoV-2 through ACE2. Our initial literature search indicated that IFN-g and IL-4 downregulate the SARS-CoV receptor ACE2 in Vero E6 cells (African green monkey kidney epithelial cells [de Lang et al., 2006]), appearing to invalidate this hypothesis. Relatedly, in vitro stimulation of A549 cells, a commonly used cell line model for lung epithelia, with IFN-a, IFN-g, and IFN-a+IFN-g for 24 h did not identify ACE2 as an ISG (Russell et al., 2018). This is potentially explained by recent work that aimed to understand SARS-CoV-2 receptor usage by performing screening studies within cell line models and found that A549 cells did not express ACE2 and therefore represents a poor model to understand regulation of this gene (Letko et al., 2020). While conducting experiments to directly test the hypothesis that ACE2 is an ISG, we noted in our own gene lists used for scoring from Ordovas-Montanes et al., 2018 and in a supplementary extended table available from Giovannini-Chami et al., 2012 that ACE2 was in upregulated gene lists after exposure to Type I IFN.
We directly tested whether IFN-a induces ACE2 in primary human upper airway epithelial cells in greater detail. We cultured human primary basal (stem and progenitors) epithelial cells to confluence and treated them with increasing doses (0.1-10 ng/ mL) of IFN-a2, IFN-g, IL-4, IL-13, IL-17A, or IL-1b for 12 h and then performed bulk RNA-seq ( Figure S3C). Only IFN-a2 and IFN-g led to upregulation of ACE2 over the time period tested, and compared with all other cytokines, IFN-a2 lead to greater and more significant upregulation over all doses tested (Figure S3D,Wilcoxon test: IFN-a2 FDR-adjusted p = 4.1EÀ07; IFN-g p = 9.3E-03, Figures S3E and S3F, all statistical tests compared with 0 ng/mL dose). We confirmed substantial and dose-dependent induction of canonical members of the interferon response after IFN-a2 and IFN-g ( Figures S3G and S3H). Conversely, we found that IFN-g, relative to IFN-a2, induced potent upregulation of GBP5, a GTPase-like protein thought to act as a viral restriction factor through inhibiting furin-mediated protease activity, which could limit viral processing from infected cells, whereas IFN-a2 more robustly induced IFITM1 To further extend and substantiate these findings, as above, we stimulated primary mouse tracheal basal cells, the commonly used human bronchial cell line BEAS-2B, and upper airway basal cells from two human donors . We confirmed appropriate induction of an IFN response in each cell type by performing differential expression testing between untreated cells and IFN-treated cells for each condition (Table S7). Within each cell type, stimulation with IFN-a2, IFN-g, or IFN-b resulted in dose-dependent upregulation of canonical ISGs, including STAT1/Stat1, BST2/Bst2, XAF1/Xaf1, IFI35/Ifi35, MX1/Mx1, and GBP2/Gbp2. Notably, Ace2 expression was not robustly induced in basal cells derived from healthy mouse trachea under any interferon stimulation condition ( Figure 5A). The magnitude of ACE2 upregulation was diminished in BEAS-2B cells compared to that in our original findings in primary human upper airway epithelial cells, but reached statistical significance compared with that of the untreated condition after IFN-g exposure ( Figure 5B). In primary basal cells derived from healthy nasal mucosa, we confirmed significant induction of ACE2 after IFN-a2 stimulation and, to a lesser extent, after stimulation with IFN-g (IFN-a2-stimulated: both Bonferroni-adjusted p < 0.001; IFN-g-stimulated: both Bonferroni-adjusted p < 0.05) ( Figures  5C and 5D). Expression of ACE2 was significantly correlated with expression of STAT1 in all human cell types, with a larger effect size and correlation coefficient in primary human basal cells . These experiments support a relationship between induction of the canonical IFN response, including key transcription factors and transcriptional regulation of the ACE2 locus. Finally, among primary human samples, we confirmed the dose-dependence of ACE2 upregulation after IFN-a2 or IFN-g treatment and significant induction of ACE2 after IFN-a2 stimulation at concentrations as low as 0.1-0.5 ng/mL .
Next, using a publicly available resource (interferome.org) that hosts genomic and transcriptomic data from cells or tissues treated with IFN, we queried ACE2 expression within human and mouse cells, searching for datasets with a log 2 -fold-change of >1 or < À1 compared with untreated samples, including all IFN types (Rusinova et al., 2013). We recovered 21 datasets spanning 8 distinct primary tissues or cell lines with non-trivial changes in ACE2 expression after both type I and type II IFN treatment ( Figure S4A). We observed substantial upregulation of ACE2 in primary skin and primary bronchial cells treated with either type I or type II IFN (> 5-fold upregulation compared with that in untreated cells), in strong support of our in vitro data ( Figures 5C, 5D, 5G-5L, and S3D-S3F). Immune cell types, such as CD4 T cells and macrophages, were noticeably absent from datasets with a significant change in ACE2 expression after IFN stimulation or were even found to downregulate ACE2 (e.g., primary CD4 T cells + type I IFN) (Figure S4A, and in our analysis of scRNA-seq peripheral blood mononuclear cell data from Butler et al., (2018); data not shown).
Given that the majority of cells robustly upregulating ACE2 were epithelial, this observation potentially explains why previous analyses to define canonical ISGs within immune populations did not identify ACE2 as an induced gene. Furthermore, using both Transcription Factor database (TRANSFAC) data hosted by the interferome database, as well as chromatin immunoprecipitation sequencing (ChIP-seq) data (provided by the ENCODE Factorbook repository), we found evidence for STAT1, STAT3, IRF8, and IRF1 binding sites within À1500-ll 500 bp of the transcription start site of ACE2 (all in human studies, Figure S4B) (Gerstein et al., 2012;Matys et al., 2003;Wang et al., 2012;Wang et al., 2013). This finding is supportive of our current hypothesis that ACE2 represents a previously unappreciated ISG in epithelial cells within barrier tissues.
Given minimal upregulation of Ace2 among primary mouse basal cells in vitro, we were curious as to whether Ace2 represented a murine ISG in vivo. We treated two mice intranasally with saline and two mice intranasally with 10,000 units of IFN-a (Guerrero-Plata et al., 2005). After 12 h, we isolated the nasal mucosa, consisting of both respiratory and olfactory epithelium, with underlying lamina propria, and performed scRNA-seq using Seq-Well S 3 ( Figure S5A). We collected from both tissue sites because of early reports of anosmia in COVID-19 (Lechien et al., 2020). We recovered 11,358 single cells, including epithelial, stromal, neuronal, and immune cell types, generating the largest single- . Abbreviation is as follows: TP10K, transcripts per 10,000 reads. ***p < 0.001, **p < 0.01, *p < 0.05, Bonferroni-corrected post hoc testing compared with 0 ng/mL condition. See also Figures S3 and S4 and Table S7. cell atlas of mouse respiratory and olfactory mucosa to date (Figures 6A and S5B). We annotated all 36 clusters, focusing our attention on epithelial cell clusters, given that we noted enrichment for Ace2 and Tmprss2 within epithelial cell subsets, consistent with our human and NHP results (Table S8). Specifically, we found Ace2 enriched within olfactory epithelial gland cells, Muc5b + Scgb1c1 + goblet cells, basal epithelial cells, and myofibroblasts/pericytes (Bonferroni-corrected p < 0.01) ( Figures 6B   and S5B) (Brann et al., 2020;Dear et al., 1991;Montoro et al., 2018;Tepe et al., 2018). Notably, Furin was enriched within olfactory epithelial gland cells (Table S8). Next, we asked whether a 12 h stimulation with IFN-a would upregulate Ace2 in vivo.
Focusing on basal epithelial cells, which contain the highest abundance of Ace2 + cells, we found that despite robust upregulation of canonical murine ISGs, Ace2 expression was only slightly elevated after IFN-a treatment ( Figures 6C, 6D, S5C, and S5D).   Figure S5 and Tables S6 and S8. This observation was supported by analysis of scRNA-seq data from 5,558 epithelial cells from the lungs of mice 3-6 days after intranasal infection with murine gamma herpesvirus-68 (MHV68) ( Figure S5E). Here, we found significant enrichment of Ace2 + cells within type II pneumocytes, in line with our data from NHP and human lungs ( Figures S5F). We did not observe changes in Ace2 expression among viral-transcript-positive cells or ''bystander'' type II pneumocytes (those without detectable cell-associated viral RNA in MHV68-infected animals), nor did we see significant alterations in Ace2 + cell abundance among MHV68-infected mice lacking IFN-gR ( Figure S5G and S5H).
These observations were in agreement with our in vitro murine basal cell assay ( Figure 5A and 5E).
Finally, we sought to validate our hypothesis that ACE2 is upregulated in human epithelial cells during upper airway viral infections, which are known to induce a robust IFN response (Bailey et al., 2014;Everitt et al., 2012;Iwasaki and Pillai, 2014;Jewell et al., 2010;Russell et al., 2018;Steuerman et al., 2018). We reanalyzed a publicly available dataset of RNA-seq from human lung explants isolated after surgical resections that were infected with influenza A virus ex vivo for 24 h. Here, we found that ACE2 expression was significantly correlated with that of SFTPC, supporting our hypothesis that ACE2 is expressed within type II pneumocytes ( Figures 1C, 2C, S5I, and S5J) (Matos et al., 2019). Furthermore, although the abundance of SFTPC was not significantly altered by influenza A virus infection, ACE2 expression was significantly upregulated after viral exposure (p = 0.0054, ratio paired t test) (Figures S5K and S5L). This suggests that influenza A virus infection increases ACE2 expression. Nevertheless, these population-level analyses are not able to definitively resolve specific cell subsets of relevance, nor whether they are directly infected cells or bystanders of infection.
In order to address these questions, we leveraged an ongoing scRNA-seq study of nasal washes from 18 individuals with confirmed influenza A virus or influenza B virus infection or healthy controls collected with Seq-Well v1, which yielded 35,840 cells resolved into 17 distinct cell types ( Figure 6E; STAR Methods) (Cao et al., 2020). We investigated the cell types with greatest enrichment for ACE2 and TMPRSS2 in non-infected controls and individuals with influenza A and B. Strikingly, ACE2 was most upregulated in samples from influenza-virus-infected individuals within bystander goblet or squamous cells not directly infected by virus ( Figures 6F and 6G). ACE2 + TMPRSS2 + goblet cells during influenza infection exhibited enrichment for canonical ISGs such as the CXCL9/CXCL10/CXCL11 gene cluster; correspondence with ACE2 + TMPRSS2 + goblet cells in healthy and allergic nasal scrapings; and a shared overlap in ISGs including GBP2, ZNFX1, ADAR, and ACE2 (significantly differentially expressed gene lists) (Table S6). Together, our data suggest that ACE2 is an ISG in vitro and in vivo in human primary upper airway epithelial basal cells, but that the murine homolog Ace2 is not in airway epithelial basal cells or pulmonary epithelial cells in vitro or in vivo. Collectively, our findings suggest that careful considerations of animal and cellular models will be needed for assessing therapeutic interventions targeting the IFN system when studying ACE2/Ace2-associated biology.
Finally, because our in vivo and in vitro work indicate that IFN might promote human cellular targets for SARS-CoV-2 infection in the human upper airway by inducing ACE2, we attempted to extend our transcriptomic data on IFN-driven expression of ACE2 to protein-level induction of ACE2. As testing of various commercially available polyclonal antibody preparations found broad evidence for non-specific or inconclusive staining in histological immunofluorescent based readouts (data not shown), we assessed whether IFN-g-stimulated human bronchial air-liquid interface cultures induced ACE2 within 24 h. Our results show that cells from one patient robustly induced ACE2 (+2.02x), cells from another mildly induced ACE2 (+1.21x) and two patient's cells showed minor changes (+/À1.12x) ( Figure S5M). We provide a note of caution as these cells were derived from asthmatic patients, and the overall changes did not reach significance. Furthermore, we could not determine cell surface localization of ACE2 but do note that these results align with our transcriptomic data.

DISCUSSION
Here, we utilize scRNA-seq across various barrier tissues and model organisms to identify the potential initial cellular targets of SARS-CoV-2 infection. To review the data presented: (1) we found that expression of the cellular entry receptor for SARS-CoV-2, ACE2, is primarily restricted to type II pneumocytes in the lung, absorptive enterocytes within the gut, and goblet secretory cells of the nasal mucosa; (2) ACE2 and TMPRSS2 co-expression in respiratory tissues is consistently found only among a rare subset of epithelial cells; (3) we observed similarities in the cellular identities and frequencies of putative SARS-CoV-2 target cells across human and NHP cohorts; (4) we observe increased expression of ACE2 during SHIV and TB infection of NHPs, and HIV/TB co-infection and influenza infection of humans compared with that in matched controls but caution that none of the datasets presented here were designed to answer this specific query. Specific targeting of these cell subsets has only been described for a handful of viruses, including the following: goblet cells by human adenovirus-5p and enterovirus 71, type II pneumocytes by H5N1 avian influenza, and absorptive enterocytes by rotavirus (Fleming et al., 2014;Good et al., 2019;Holly and Smith, 2018;Weinheimer et al., 2012).
Additionally, we provide an overall note of caution when interpreting scRNA-seq data for low abundance transcripts like ACE2 and TMPRSS2 because detection inefficiencies might result in an underestimation of the actual frequencies of ACE2 + or ACE2 + TMPRSS2 + cells in a tissue. Moreover, the protein amounts of each might differ from their mRNA abundances (Genshaft et al., 2016;Jovanovic et al., 2015;Rabani et al., 2011;Shalek et al., 2013). We also present datasets separately, given that each study differed in its methods of tissue processing and collection, which can influence the frequency of recovered cell subsets (STAR Methods). We provide Table S9 as a summary of ACE2 + and ACE2 + TMPRSS2 + cells across various datasets. Moreover, we present Figure S6, which describes statistical modeling and power calculations underlying detection and dropout of ACE2, to help guide interpretation of these data. This includes an examination of the probability to detect a lowly expressed transcript like ACE2 within a cell, as well as upper bound estimates on the percentage of positive cells within a cluster, considering the effects of transcript counts, sequencing depth, and cell numbers in these calculations (STAR Methods).
Whether ACE2 and TMPRSS2 are needed on the same cell or soluble proteases can activate SARS-CoV-2 S protein to invade ACE2 single-positive cells is an area of active inquiry (Coutard et al., 2020;Letko et al., 2020). Importantly, rapidly evolving literature has identified that SARS-CoV-2-S might have a furin cleavage site, leading to a broader set of host proteases that could mediate S protein activation (Bugge et al., 2009;Coutard et al., 2020;Walls et al., 2020). However, because an active S protein has a finite lifetime to find a target cell membrane, the timing and cellular location of S protein activation is key to consider. Activation events proximal to the plasma membrane have been shown to be most effective for SARS-CoV entry (Shulla et al., 2011).
Our study finds that type I IFNs, and to a lesser extent type II IFNs, upregulate ACE2. This is based on several lines of evidence: (1) we identified a human goblet secretory cell subset in upper airway nasal epithelium enriched for ACE2 expression to have the highest IFN-a-induced gene signature; (2) we found that IFN-a, and to a lesser extent IFN-b or IFN-g, induced ACE2 expression in a published dataset of air-liquid interface cultures derived from human nasal epithelial cells (Giovannini-Chami et al., 2012;Ordovas-Montanes et al., 2018); (3) we extended our search through the Interferome database (Rusinova et al., 2013) and found that, in epithelial barrier tissues, type I IFNs upregulate ACE2 in multiple studies, especially in primary bronchial cells and keratinocytes (Rusinova et al., 2013); (4) we found two STAT1 binding sites in the promoter of ACE2; (5) in our unpublished atlas of SHIV-infected macaques, known to have elevated amounts of chronic IFN signaling, we found ACE2 upregulation in absorptive enterocytes; (6) we directly provided evidence for IFN-a, and to some extent IFN-g, inducing ACE2 expression in primary human upper airway basal cells; and (7) influenza infection in humans, a known inducer of the IFN pathway, leads to increased ACE2 expression in goblet secretory cells of the nasal epithelium (Cao et al., 2020).
Altogether, our own and publicly available data highlight that ACE2 might have been missed as a canonical ISG because of its notable absence in peripheral blood mononuclear cell datasets and in lung-derived transformed cell lines such as the A549 cell line (Butler et al., 2018;Letko et al., 2020;Rusinova et al., 2013). Importantly, other groups have independently analyzed publicly available datasets, some referenced in our work, and observed ACE2's behavior as an ISG (Wang and Cheng, 2020). Furthermore, we found weak IFN-or virally driven induction of Ace2 in murine cells and tissues. This highlights the importance of studying primary human epithelial cells and the careful consideration of appropriately selected gene lists and in vitro models of in vivo cellular systems for understanding human biology (Jonsdottir and Dijkman, 2016;Mead and Karp, 2019;Regev et al., 2017).
As SARS-CoV-S leads to ACE2-receptor-mediated internalization, the host IFN response could thus promote the ability for SARS-CoV and SARS-CoV-2 to maintain cellular targets in neighboring human upper airway epithelial cells. Altogether along with a study of HCoV-OC43, which co-opts IFN-inducible transmembrane 2 (IFITM2) and IFITM3 to promote viral entry, this adds to the growing evidence that coronaviruses, as well as other viruses, have evolved to leverage features of the human IFN pathway (Fung and Liu, 2019;Mar et al., 2018;Zhao et al., 2014). Whether type I IFNs are net protective or detrimental to the host might depend on the stage of infection; cell subsets in question; the SARS viral clade (Channappanavar et al., 2016;Channappanavar et al., 2019;Channappanavar and Perlman, 2017;Davidson et al., 2015); and other factors such as co-infection, age, gender, and co-morbidities, among others. Understanding the specific host restriction factors targeting SARS-CoV-2 and identifying specific drivers of these genes in the absence of ACE2 upregulation might provide strategies to dissociate the dual roles of IFN in certain coronavirus infections. Whether IFNs upregulate ACE2 in putative target cell subsets in vivo will be of significant interest to define in future work once current COVID-19-related restrictions on basic scientific inquiry are lifted (Qian et al., 2013).
ACE2 is a central component of the renin-angiotensin system, which has emerged as a key regulator of sterile-or microbially induced lung pathology . In brief, ACE cleaves angiotensin I to generate angiotensin II (Skeggs et al., 1980). Angiotensin II then acts to drive acute lung injury through various mechanisms, including increased vascular permeability . Amounts of angiotensin II in humans and mice are elevated during influenza infection, and ACE2 exerts tissue-protective functions by reducing amounts of angiotensin II . Binding of SARS-CoV-S to mouse ACE2 in vivo reduced ACE2 expression leading to acute acid-aspirationinduced lung failure . Depending on the questions asked in future work, there are mouse models available on the basis of transgenic expression of human ACE2 (required for overt infectious pathology of SARS-CoV in mice), there are established NHP models available of SARS-CoV infection in M. fascicularis and C. aethiops, and early reports suggest symptomatic infection in M. mulatta and M. fascicularis models for SARS-CoV-2 (Bao et al., 2020;McCray et al., 2007;Munster et al., 2020;Rockx et al., 2020;Smits et al., 2011). For example, examining the efficacy of recombinant human ACE2 to act as a decoy receptor or the effect of ''ACE inhibitors'' in patients with, or at risk for, COVID-19 will require careful experimentation in appropriate models together with well-controlled clinical trials (Hofmann et al., 2004;Monteil et al., 2020;Vaduganathan et al., 2020).
IFN responses that induce ISGs are essential for host antiviral defense in mice, NHPs, and humans (Bailey et al., 2014;Dupuis et al., 2003;Everitt et al., 2012). Canonical ISGs function by directly restricting viruses and reducing burden (Schneider et al., 2014). More recently, disease tolerance to equivalent pathogen burden by factors that increase the ability of the host to tolerate tissue damage has been identified as part of a combined host defense strategy (Iwasaki et al., 2017;Iwasaki and Pillai, 2014;Medzhitov et al., 2012;Schneider and Ayres, 2008). Disease tolerance factors in the lung include IL-22 and amphiregulin (Iwasaki et al., 2017). During acute infection in the respiratory system, ACE2 is critical for early tissue tolerance responses to respiratory infection, including H5N1 influenza (Huang et al., 2014;Zou et al., 2014). However, our discovery that ACE2 is an ISG in human epithelial cells, along with SARS-CoV-2 utilizing host ACE2 ll to gain entry to cells, suggests that SARS-CoV and SARS-CoV-2 might exploit the ACE2-mediated tissue-protective response to provide further cellular targets for entry. This potential strategy employed by SARS-CoV-2 could present a unique challenge for the human host and is distinct from HCoV-OC43, which targets the two restriction factors IFITM2 and IFITM3 . Our study provides motivation to understand the specific role and balance of type I and type II IFNs, as well as type III IFNs, in tissue protection during, and host restriction of, SARS-CoV-2 infection. Key experiments to understand ACE2 as an ISG in tissue protection or genuine tolerance will require the appropriate mouse, NHP, or other model in BSL3 or BSL4 facilities to execute SARS-CoV-2 viral infections and measure host tissue health along with viral loads. Further work will also be needed to understand how co-infections, as well as other host factors, might affect both the susceptibility to, and dynamics of, host SARS-CoV-2 infection. Moreover, carefully controlled clinical trials will be essential to determine the overall effects of different IFNs (Prokunina-Olsson et al., 2020).
Altogether, we anticipate that comprehensive characterization of the putative cellular targets of SARS-CoV-2 will be critical to understand basic mechanisms of viral tropism and disease pathophysiology, inform differential susceptibility among vulnerable populations, and potentially suggest unanticipated targets for drug inhibitors of viral infection. The cellular targets we nominate will need to be confirmed by specific reagents for SARS-CoV-2, as done for SARS-CoV (Ding et al., 2004). Furthermore, the transcriptional response to the virus will need to be rigorously characterized in appropriate in vitro and in vivo model systems (Blanco-Melo et al., 2020). We provide gene lists associated with target cells in specific tissues and diseases to aid the community in understanding this emergent disease. A concurrent HCA Lung Biological Network study assessing ACE2 and TMPRSS2 across more tissues also identified enrichment in nasal goblet and ciliated cells (Sungnak et al., 2020). Other studies are considering additional tissues; co-variates such as age, sex, and co-infection state; and represent a large coordinated international effort to the ongoing crisis (Pinto et al., 2020). One study in particular identified upregulation of ACE2 by respiratory viruses and TMPRSS2 by IL-13 in a pediatric cohort, suggesting further links to how underlying allergic conditions or co-infections might modulate these two SARS-CoV-2related host factors (Sajuthi et al., 2020).
During the preparation of this manuscript, several papers have been posted to bioRxiv assessing patterns of ACE2 + and TMPRSS2 + cells in barrier tissues (Brann et al., 2020;Lukassen et al., 2020;Qi et al., 2020;Wu et al., 2020;Zhang et al., 2020). At a high level, these studies are largely in agreement with our report. Furthermore, another study appeared on medRxiv profiling bronchoalveolar lavage fluid from 3 severe and 3 mild COVID-19 patients, though they were unable to profile sufficient numbers of epithelial cells (Liao et al., 2020).
Our study highlights the power of scRNA-seq datasets, both existing and novel, to derive hypotheses relevant to human disease that might differ from paradigms established by using cell lines. Further work will be critical to determine how SARS-CoV-2 influences temporal dynamics of host responses at single-cell resolution and which host factors might affect this (Kazer et al., 2020). Given the unappreciated complexities of host-pathogen interactions between humans and SARS-CoV-2, the best measures to combat this pandemic continue to be surveillance and avoidance-especially given that a deep understanding of the full spectrum of resistance and tolerance mechanisms will require the concerted efforts of scientists around the globe (Amanat et al., 2020;Chu et al., 2020;Hadfield et al., 2018). Here, we seek to share our initial findings and data so that other groups might build on this discovery of ACE2 as an ISG and further consider the careful balance between tissue tolerance and viral infection needed at the human airway epithelium.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: Merck, Honeycomb Biotechnologies, Cellarity, Cogen Therapeutics, Orche Bio, and Dahlia Biosciences. L.S.K. is on the SAB for HiFiBio; she reports research funding from Kymab Limited, Bristol Meyers Squibb, Magenta Therapeutics, BlueBird Bio, and Regeneron Pharmaceuticals and consulting fees from Equillium, FortySeven, Inc, Novartis, Inc, EMD Serono, Gilead Sciences, and Takeda Pharmaceuticals. A.S. is an employee of Johnson and Johnson. N.K. is an inventor on a patent using thyroid hormone mimetics in acute lung injury that is now being considered for intervention in COVID-19 patients. J.L. is a scientific consultant for 10X Genomics, Inc. O.R.R, is a co-inventor on patent applications filed by the Broad Institute to inventions relating to single-cell genomics applications, such as in PCT/US2018/060860 and US Pro- Coronaviridae Study Group of the International Committee on Taxonomy of V. (2020). The species Severe acute respiratory syndrome-related coronavirus: classifying 2019-nCoV and naming it SARS-CoV-2. Nat. Microbiol. https:// doi.org/10.1038/s41564-020-0695-z.

Lead Contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by Dr. Jose Ordovas-Montanes (jose.ordovas-montanes@childrens.harvard.edu).

Materials Availability
This study did not generate new unique reagents.

Data and Code Availability
In Table S9, we provide a guide to all datasets analyzed in this paper as well as links to each individual dataset for download with the main landing page here: https://singlecell.broadinstitute.org/single_cell?scpbr=the-alexandria-project. To download the data from the portal, follow the link to the visualization page, sign in a free account in the portal using a Google apps enabled email address, All data included in the present study can be visualized using the following web viewer: https://singlecell.broadinstitute.org/single_cell?scpbr=the-alexandria-project. As we gain further insight and feedback from our own groups, collaborators, and investigators, we will continue to provide updates on our resource websites, including the utility of in vitro systems, such as organoids (Mead et al., 2018), for the study of SARS-CoV-2: http://shaleklab.com/resource/covid-19-resources/ and www.ordovasmontaneslab.com/covid-19-resources/. We also note that there are several ongoing efforts unified together through the HCA Lung Biological Network group that we will reference and to which we will link as they become available.
No custom code was used to analyze these data and all methods and packages used are cited in the Method Details section.

EXPERIMENTAL MODEL AND SUBJECT DETAILS Human Intestinal Biopsies
For human intestinal biopsies from the terminal ileum, the subjects were enrolled on a multi-center clinical study, which was approved by the Institutional Review Board at Boston Children's Hospital (protocol number: IRB-P00030890). Full information related to subject age/developmental stage and sex found in metadata associated with provided raw datasets.

Human Lungs, Surgical Excess
Samples were obtained through indicated lung lobe resection or diagnostic procedures in collaboration with clinicians at the Department of Cardiothoracic Surgery at Inkosi Albert Luthuli Central Hospital in Durban, South Africa. Informed consent was obtained from each participant. The study protocol was approved by the University of KwaZulu-Natal Institutional Review Board (approval BE024/ 09). Full information related to subject age/developmental stage and sex found in metadata associated with provided raw datasets.

Human Nasal Polyps and Scrapings
For inferior turbinate nasal scrapings, polyp scrapings, and ethmoid sinus surgical tissue samples, the Partners HealthCare Institutional Review Board (Boston, Massachusetts), approved the study and all subjects provided written informed consent . Full information related to subject age/developmental stage and sex found in metadata associated with provided raw datasets.

Human Nasal Washes, Healthy and Influenza Infected
The Institutional Review Board of the University of Massachusetts Medical School (Worcester, Massachusetts) approved the study and all subjects provided written informed consent.

Cell Culture of Primary Basal Cells and Cell Lines
Human basal cells from non-polyp surgical resections from ethmoid sinus, BEAS-2B cells (ATCC), or mouse tracheal basal cells were placed into culture at a number of 10,000 cells seeded at passage 3 and cultured at confluence in 96 well flat-bottom collagen-coated tissue culture plates (Corning 3596) for 48 h in Pneumacult-Ex serum-free media (StemCell Technologies, Inc.). All cells were incubated at 37 C and 5% CO2.

Non-Human Primates (M. mulatta)
Healthy and SHIV-infected non-human primate (M. mulatta) work was conducted at the Washington National Primate Research Center (WaNPRC), an AAALAC accredited program, in accordance with the regulations detailed in the U.S. Department of Agriculture Animal Welfare Act and in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. It was approved by University of Washington Institutional Animal Care and Use Committee. Expanded cohort characteristics described previously (Colonna et al., 2018). Full information related to subject age/developmental stage and sex found in metadata associated with provided raw datasets.

Non-Human Primates (M. fascicularis)
Tissues from Mycobacterium tuberculosis-infected non-human primates (M. fascicularis) were conducted at the University of Pittsburgh School of Medicine, an AAALAC accredited program, in accordance with the regulations detailed in the U.S. Department of Agriculture Animal Welfare Act and in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. Full information related to subject age/developmental stage and sex found in metadata associated with provided raw datasets.
Mouse Nasal and Olfactory Epithelium and Tracheal Cells C57BL/6J mice purchased from Jackson laboratory (Bar Harbor, ME, USA) were maintained within Ragon Institute's HPPF barrier facility and all experiments were conducted with institutional IACUC approval. In this study, mice were 8-10 weeks of age, representing male and female animals.
Mouse Lungs, MHV68 Infection C57BL/6 mice were purchased from Charles River Laboratories (Sulzfeld, Germany). IFNgRÀ/À mice on C57BL/6 background (C57BL/6, IFNgR À/À B6.129S7-Ifngr1 tm1Agt /J) were originally obtained from the Jackson Laboratory (Bar Harbor, ME, USA) and subsequently bred and propagated under SPF conditions at the Helmholtz Zentrum Mü nchen. Animals with different genotypes were kept in the same animal room for the time of the experiment including an adaptation period prior to the start of the experiment. All animal experiments were in compliance with the German Animal Welfare Act (German Federal Law x8 Abs. 1 TierSchG), and the protocols were approved by the local Animal Care and Use Committee.

METHOD DETAILS
Methods of Sample Collection and Tissue Preparation for Single-Cell RNA-Seq NHP Ileum, Jejunum, Colon, Liver, Tonsil, Thymus, and Lung Tissue Animals were perfused with 0.5 L of PBS/kg immediately following euthanasia, tissues were isolated and placed in RPMI + 10% FBS and kept on ice until dissociation. Tissue sections were digested by mincing and incubating with collagenase IV (Life Technologies) and DNase I (Roche) at 37 C for 1 h with agitation. Digested tissue was passed through a 100 mm metal strainer, cells were pelleted by centrifugation at 300 g, rinsed with RPMI + 10% FBS, counted, and prepared as a single cell suspension for scRNA-seq using Seq-Well v1 (see below). NHP Lymphoid Organs, Bone Marrow, PBMCs All lymph nodes, spleen, and bone marrow were ground through a metal strainer, transferred to a conical in RPMI + 10% FBS, and pelleted by centrifugation at 400 g x 10 min. LN-derived cells were resuspended in RPMI + 10% FBS, counted and prepared as a single cell suspension. Spleen, bone marrow, and PBMCs were subjected to ACK lysis for 10 min at room temperature, quenched with RPMI + 10% FBS. PBMCs and bone marrow derived cells were purified over a ficoll gradient (GE Healthcare) by centrifuging at 400 g for 20 min at room temperature with no brake. Cells were then resuspended in RPMI + 10% FBS, counted, and diluted for scRNA-seq using Seq-Well v1 (see below). NHP Tuberculosis Infected Lung and Granuloma Ten Mycobacterium tuberculosis infected (Martin et al., 2017) adult non-human primates (M. fascicularis) were included in this study. A piece of lung tissue (without any grossly visible pathology) and 4 individual TB lung granulomas per animal were excised at necropsy and enzymatically dissociated using the GentleMacs system (Tumor dissociation kit, human; Miltenyi Biotec). Single cell suspensions were resuspended in RPMI + 10% FBS, counted and diluted for scRNA-seq using Seq-Well S 3 (see below). Human Lung Tissue Surgical samples from diseased lung tissue (n = 3 TB + HIV + ; n = 3 TB + ; n = 2 non-infected patients) were processed as described in (Ardain et al., 2019). Briefly, each sample was collected into cold RP-10 (RPMI (Sigma-Aldrich) + 10% FBS), minced, and incubated for 25-30 min at 37 C with digestion buffer containing collagenase D (Sigma-Aldrich), DNase I (Sigma-Aldrich) in RPMI 1640 (Sigma-Aldrich) with 10% FBS (Hyclone). Following incubation, samples were homogenized using a GentleMACS, filtered using a 70 mm metal strainer, and pelleted by centrifugation at 400 g for 5 min. After obtaining the pellet, cells were resuspended in RP-10, passed through another 70mm strainer (Corning), stained with trypan blue, and then counted and diluted for scRNA-seq using Seq-Well S 3 (see below).

Human Ileum
Single-cell suspensions were collected from biopsies as described (Smillie et al., 2019). Briefly, biopsies were rinsed in cold PBS, the epithelial layer was separated from the underlying lamina propria by end over end rotation for 15 min. The lamina propria and epithelial fractions were digested separately, using Liberase TM (Roche) and DNase I (Roche) for the lamina propria, and TrypLE (Thermo-Fisher) for the epithelial fraction. Following digestion, cells were pelleted by centrifugation, subjected to ACK lysis for 3 min, and filtered through a 40 mm strainer. Following centrifugation, the cells were counted and prepared as a single cell suspension for scRNA-seq using 10X 3 0 v2 (10X Genomics). Nasal Mucosa and Nasal Scrapings Surgical samples from ethmoid sinus and nasal scraping of the inferior turbinate were processed as described (Ordovas-Montanes et al., 2018). Briefly, each sample was collected into cold RPMI (Corning), minced and incubated for 30 min (15 min for nasal Pearson's r = À0.31, non-significant). Further, when comparing ACE2+ cells to ACE2-cells within a given tissue, we did not observe a positive correlation between read depth and ACE2 status (i.e., mean ± standard error of the mean, SEM, reads among all lung cells = 28,512 ± 344; mean ± SEM reads among ACE2+ lung cells = 28,553 ± 2,988; mean ± SEM reads among all ileum cells = 14,864 ± 288; mean ± SEM reads among ACE2+ ileum cells = 10,591 ± 441, full statistics on cell depth among ACE2+ cells compared to ACE2-cells of the same cell type can be found in Table S9). Thus, we can be confident that the observed differences in ACE2+ proportions across clusters are not driven by differences in sequencing depth. Finally, we investigated how observed differences in ACE2+ proportions across clusters might be affected by cell sampling. Using the proportion of ACE2+ cells in a ''typical'' cluster annotated as being ACE2 positive (i.e., 6.8% in non-human primate type II pneumocytes, Figure 1), we calculated the cluster sizes needed to be confident that the probability of observing zero to a few positive cells is unlikely to have arisen by random chance (probabilities calculated under a negative binomial distribution with parameter p = 0.068, Figure S6E). We found that as cluster sizes approach and exceed 100 cells, the probability of observing zero to a few positive cells rapidly approaches zero, if we assume 6.8% of cells are positive. Further, to examine our confidence in estimating an approximate upper bound (ignoring the impact of protocol inefficiencies discussed above) for the fraction of cells positive in a cluster as a function of the number of cells in that cluster, we also calculated the probability of observing zero (and its complement, probability of observing at least 1) ACE2+ cells as a function of cluster size across true positive proportions ranging from 0.1% to 10% (probabilities calculated under a negative binomial distribution with parameter p = 0.001 to 0.1, representing hypothetical proportions of ACE2+ cells Figure S6F). Given our typical cluster sizes (on the order of hundreds of cells, exact values provided in Table S9), we find that for us to observe 0 ACE2+ cells in a cluster due to sampling artifacts, the fraction of true positives must be $1% or less. Thus, these complementary approaches demonstrate that our observed variations in ACE2+ cell proportions across clusters likely reflect underlying biological differences, rather than random chance.

Statistical Testing
Parameters such as sample size, number of replicates, number of independent experiments, measures of center, dispersion, and precision (mean ± SEM) and statistical significances are reported in Figures and Figure Legends. A p value less than 0.05 was considered significant. Where appropriate, a Bonferroni or FDR correction was used to account for multiple tests, alternative correction methods are noted in the figure legends or Methods. All statistical tests corresponding to differential gene expression are described above and completed using R language for Statistical Computing.