A humanized mouse model of chronic COVID-19 to evaluate disease mechanisms and treatment options

Coronavirus-associated acute respiratory disease, called coronavirus disease 2019 (COVID-19) is an infectious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). More than 90 million people have been infected with SARS-CoV-2 and more than 2 million people have died of complications due to COVID-19 worldwide. COVID-19, in its severe form, presents with an uncontrolled, hyperactive immune response and severe immunological injury or organ damage that accounts for morbidity and mortality. Even in the absence of complications, COVID-19 can last for several months with lingering effects of an overactive immune system. Dysregulated myeloid and lymphocyte compartments have been implicated in lung immunopathology. Currently, there are limited clinically-tested treatments of COVID-19 with disparities in the apparent efficacy in patients. Accurate model systems are essential to rapidly evaluate promising discoveries but most currently available in mice, ferrets and hamsters do not recapitulate sustained immunopathology described in COVID19 patients. Here, we present a comprehensively humanized mouse COVID-19 model that faithfully recapitulates the innate and adaptive human immune responses during infection with SARS-CoV-2 by adapting recombinant adeno-associated virus (AAV)-driven gene therapy to deliver human ACE2 to the lungs of MISTRG6 mice. Our unique model allows for the first time the study of chronic disease due to infection with SARS-CoV-2 in the context of patient-derived antibodies to characterize in real time the potential culprits of the observed human driving immunopathology; most importantly this model provides a live view into the aberrant macrophage response that is thought to be the effector of disease morbidity and ARDS in patients. Application of therapeutics such as patient-derived antibodies and steroids to our model allowed separation of the two aspects of the immune response, infectious viral clearance and immunopathology. Inflammatory cells seeded early in infection drove immune-pathology later, but this very same early anti-viral response was also crucial to contain infection.

COVID19 is a remarkably heterogenous disease with few therapeutic options, amplifying the urgency to better understand mechanisms of immunopathology and immune-protection in this devastating disease. Disparities in the apparent efficacy of appropriate treatments of COVID-19 in patients renders accurate model systems essential to rapidly and comprehensively evaluate promising discoveries. Here we present a comprehensively humanized COVID-19 model and evaluate its potential to faithfully model the innate and adaptive human immune system during infection with SARS-CoV-2.
Mice with a human immune system (humanized mice) serve as invaluable tools to study the development and function of the human immune system in vivo. Humanized mice are generated by transplantation of human hematopoietic stem and progenitor cells (HSPCs) into various strains of immune-compromised mice that thereby allow xeno-graftment 5,6 . The MISTRG6 mouse model was engineered by a human/mouse homolog gene-replacement strategy to provide physiological factors with regard to quantity, location and time and enable essentially all classes of human hematopoietic cells to develop in mice. MISTRG6 (acronym for genes replaced) mice encode humanized M-CSF (enabling monocytes and tissue macrophage development) 7 , GM-CSF/IL-3 (to provide lung alveolar macrophages) 8 , SIRPa (establish macrophage tolerance to human cells ) 9 , ThPO (hematopoiesis and platelets) 10 , and IL6 (better engraftment allowing study of adult human patients and improved antigen-specific antibody responses as well as human IL-6 per se 11 ) 2512,13, in a Rag2/Gamma common chain deleted background. MISTRG6 mice have a comprehensive immune system relatively comparable to humans in strong contrast to other commonly used humanized mouse models, which either lack relevant human myeloid cells, specifically monocytes and alveolar macrophages, or support human myelopoiesis but at the expense of human hematopoietic stem cell maintenance 14,15 . Of particular importance for COVID-19 research, MISTRG mice, express physiologic levels of GM-CSF ,essential to repopulation of the murine lung with human alveolar macrophages 16 ; and M-CSF which enable tissue macrophage and blood monocyte development 7 , which together in humans are thought to contribute to COVID-19 severity. Moreover, the human myeloid cells secrete IL-15 which directs the robust development of human NK cells that are also implicated in COVID19 pathogenesis 17 . By adapting recombinant adeno-associated virus (AAV)-driven gene therapy to deliver human ACE2 to the lungs 18 , which allows infection with SARS-CoV-2 of MISTRG6 mice engrafted with a human hematopoietic stem and progenitor cells, we created a humanized mouse model of COVID19 that recapitulates the distribution and function of the human innate and adaptive immune system and is amenable to the mechanistic study of COVID 19 and its myriad of complications 12,19 .

MISTRG6 humanized mice that transiently express hACE2 can be infected with SARS-CoV-2.
SARS-CoV-2 does not infect standard laboratory mice due to differences between mouse and human ACE2 receptor that limit viral entry 20,21, . Introduction of human ACE2 into the murine host lung via recombinant adeno-associated virus (AAV)-driven gene therapy enables SARS-CoV-2 infection; however, in a standard laboratory mouse AAV-mediated hACE2 expression affords only acute, transient infection with SARS-CoV-2 . We hypothesized that a functional human immune system would confer much of the chronicity and pathology seen in patients onto a small animal model 22,23 .
We successfully delivered AAV-hACE2 to lung tissues 18 of immune-reconstituted MISTRG6 mice (MISTRG6-hACE2) (Fig 1A). MISTRG6-hACE2 mice were then infected with SARS-CoV-2 which yielded comparable viral RNA levels and viral titers ( Fig.1B-C) 2 as described for wild type mice early in infection but with higher viral titers which were sustained for at least 7 days as well as chronically high levels of viral RNA for at least 28 days post infection (dpi). In contrast, MISTRG6 mice lacking AAV-hACE2 expression did not have detectable viral titers even early in infection, confirming the necessity of human ACE2 for infection (Fig. 1C). Laboratory mice do not normally develop severe, chronic disease 18,24,25 . In contrast, the presence of human immune cells in MISTRG6-hACE2 mice caused more severe disease with severe weight loss and a chronic inability to restore body weight for at least 28 dpi (Fig. 1D). MISTRG6 humanized mice also exhibit more severe lung pathology compared to published wild type mice, and other animal models 18,[25][26][27][28] . This lung pathology was characterized by infiltrating monocytes and macrophages and unresolved diffuse alveolar damage, reminiscent of human patients 4,[29][30][31] (Fig. 1F-S1A) . Strikingly, severe lung pathology persisted for at least 28dpi ( Fig   S1A), by which time signs of fibrosis manifested (Fig. 1G). Of note, MISTRG6-hACE2 mice that were infected but not engrafted with human HSPCs (so lack human immune cells), had reduced viral titers ( Fig. 1B-C). Thus, these data suggest the human immune cells contribute to worsened pathology and amplification of viral infection.

Immune landscape in MISTRG6-hACE2 humanized mice infected with SARS-CoV2 is characterized by inflammatory macrophages and monocytes.
We next characterized human immune cells in SARS-CoV2 infected and uninfected control MISTRG6-hACE2 humanized mice by flow cytometry to better evaluate the immunological landscape that drives immunopathology and viral clearance ( Fig S2). The degree of humanization, measured by the ratio of human immune cells among total immune cells, was comparable in blood and lung between uninfected and infected mice, but SARS-CoV2 infection drastically increased recruitment of human immune cells to the lower respiratory tract and lung parenchyma, as assessed by the number of human cells in whole lung homogenates and bronchiolar lavage (BAL) ( Fig. 2A Intermediate and non-classical monocytes infiltrated infected lungs at high frequencies as early as 2dpi, peaking at 4dpi. Increased numbers of macrophages were observed at 4dpi, which remained high until 28dpi, the last time point measured (Fig. 2D). By 2dpi, the macrophage compartment was already enriched for inflammatory and monocyte-derived macrophages, which outnumbered alveolar macrophages, suggesting that macrophages seeded early from the circulation are the long term contributors of immunopathology ( Fig. 2E-G). To our surprise, plasmacytoid dendritic cells (pDCs), known for their contribution to early antiviral response and main producers of type I interferon alpha (IFNa), were particularly enriched but substantially later in infection (14dpi in Fig. 2H). This coincided with a chronically sustained interferon response coupled with sustained inflammatory macrophages, which is in line with observations in patients with severe COVID19 immunopathology that is believed to be at least partially driven by inflammatory macrophages and accompanied by a sustained type I interferon response 33,34 .

COVID-19 in MISTRG6-hACE2 humanized mice presents with systemic T cell lymphopenia.
A notable characteristic of human COVID-19 is profound T cell lymphopenia and this is strikingly recapitulated in our humanized mice. Infected MISTRG6-hACE2 humanized mice also presented with lymphopenia characterized by a profound loss of T cells, especially CD8+ T cells, in blood ( Fig. 3A;   Fig. S3A-B) and spleens (Fig. 3B, Fig. S3B-C) as observed in COVID-19 patients 35 Moreover, in the adaptive arm of the anti-viral response in the lung of MISTRG6-hACE2 humanized mice, T cells, the main producers of IFN-gamma, displayed markers of activation such as HLA-DR, CXCR3 (Fig. 3C), ICOS and PD1 (Fig. 3D) as reported in COVID-19 patients [35][36][37] . T cell populations comprised of both TCRalpha/beta T cells (Tab) that were enriched for CD4+ T cells, and TCRgamma/delta T cells (Tgd) cells in both uninfected and infected mice (Fig. S3D). Yet, during the course of infection, as the circulating T cell numbers decline, the total number of T cells in lungs increased with higher representation of Tgd cells compared with uninfected mice (Fig. 3E). The early T cell response in infected lungs corresponded to an increase in both resident and infiltrating Tgd cells (as assed by scRNA seq; cluster 4 in Fig 4C) as well as bystander activated memory Tab cells as determined by the lung transcriptional profile (Fig. S4A), B cell numbers also gradually increased in response to infection over several weeks, peaking late at 28dpi in lungs and BAL ( Fig. 3G-H). Although the germinal center B cell response has been reported to be suboptimal in reconstituted MISTRG6 humanized mice 13

Transcriptional landscape of SARS-CoV-2 infection is marked by sustained interferon response and SLE like features, reminiscent of COVID-19 patients.
Next, we evaluated the transcriptional landscape in uninfected and SARS-CoV-2 infected lungs of humanized mice at multiple time points (2, 4, 7, 14, 28 dpi). Mapping of transcripts to the human genome or the mouse genome separately identified 285 human genes and 516 mouse genes that were over-represented in infected lungs (Fig. 4A, Table S1). Although there was marked heterogeneity in the strength of the response, the anti-viral response was strikingly sustained throughout the course of infection long after infectious virus was apparently cleared, suggesting that early anti-viral responses were either maintained or amplified late in infection (Fig. 4A). Corresponding pathway analysis of these differentially expressed mouse genes (DEGs) using multiple platforms (Gene Ontology (GO), Gene Set Enrichment analysis (GSEA), Ingenuity) identified cellular response to interferons, cytokine production, ribonuclease activity and neutrophil activation as top biological processes that are induced during SARS-CoV-2 infection in humanized mice (Table S2). A similar analysis showed human genes that were enriched for extracellular matrix assembly, opsonization, complement activation with a focus on immune phenotypes in monocytes, activated T cells and B cells, which further corroborated our findings from flow cytometric analysis (Table S2). Similarly, circuitry of monocytes, macrophages, activated T cells and in particular extrafollicular B cells has been described in humans with SARS-CoV-2 pneumonia 43 .
We performed single cell RNA sequencing using the 10X Genomics platform to better evaluate the  43 , alveolar macrophages in infected MISTRG6-hACE2 humanized mice were the main producers of T cell chemoattractants such as CXCL10 (Fig. 4E).
Strikingly, the majority of the human and mouse DEGs in lungs were interferon responsive genes (Fig.   4B). Although type I and type III interferon expression perse could not be detected, interferon responsive genes were sustained at high levels throughout infection persisting even as late as 14dpi and 28dpi, recapitulating the interferon-dependent phenotypes identified in COVID19 patients ( Fig. 4A- Table S3) 30,33,44,45 . IFN signaling coinciding with the onset of lung recovery (7-9 dpi) in influenza infection has been shown to prevent epithelial cell proliferation and differentiation, hence interfering with lung repair 46 . In line with known effects of interferons on lung tissue recovery during influenza infection 46 , our findings suggest that sustained type I IFN signaling may also contribute to sustained lung tissue injury in COVID19 as supported by histopathological assessment of 14 and 28 dpi lungs of the infected humanized mice (Fig. 1F). Human and mouse DEGs were also enriched for type II responsive genes. Type II interferon IFNg was mainly produced by T cells (Fig. 4C,E) starting as early as 2dpi and was sustained until 28dpi. (Fig. 4F). In addition, levels of various pro-inflammatory cytokines (IL6, IL8,TNF, IL1B) were elevated and peaked late in infection (IL6 and IL8 at 14dpi) after viral clearance, further supporting that delayed immune-resolution is a characteristic of humanized COVID-19 ( Fig 4F). Inflammatory cytokine signature, particularly elevated IL6 and IL8 but not TNF, closely correlated with severity of COVID19 in patients 47,48 and is notably recapitulated in humanized COVID-19.
We focused analysis on the genes that typify bystander activation of memory T cells 49 ; this suggested that early in infection, T cell activation in 2dpi and 4dpi lungs may be an antigen-independent, interferon driven response (Fig. S4A). We sought to identify the origin of genes upregulated in patients by validating their expression in our infected mice and then identifying their cellular source. This showed that the B cell response in humanized lungs was particularly enriched for genes that are upregulated in moderate and severe COVID19 compared with healthy donors (Fig. S4B-C). Strikingly, genes differentially expressed at 28dpi (our latest timepoint) were enriched for extrafollicular differentiation of B cells and also presented with features of systemic lupus erythematosus (SLE) pathways ( (Fig. 4G).
Lack of germinal center formation in the spleens and lymph nodes of patients that have succumbed to COVID-19 coupled with extrafollicular B cell responses have been correlated with morbidity and poor clinical outcomes in COVID19 patients 40,50 . In line with these observations, unbiased pathway analysis of the 28dpi lung transcriptome (Fig. S4B) and a more focused look at the SLE gene signatures 51 identified a particular enrichment of SLE-like extrafollicular responses in lungs of infected mice by 28dpi ( Fig. 4D). Furthermore, B cell responses at 28dpi exhibited the features of bystander responses previously characterized in influenza infection in humans 52 (Fig. S4D), further suggesting a highly inflammatory, bystander B cell response in humanized COVID19 36,37 . Taken together, our transcriptome analysis of infected humanized lungs identifies monocyte derived-macrophages at the center of early anti-SARS-Cov-2 response that maintain an interferon dependent response that is further amplified later in infection. Bystander T cell activation and SLE-like features of B cells suggest that interferons not only shape innate immunity but also impact the nature of the adaptive immune response induced by SARS-CoV-2.

Human monoclonal recombinant antibodies as prophylactic and therapeutic interventions impact disease outcome.
We wanted to test whether MISTRG6-hACE2 mice could be used to evaluate patient-derived human antibodies as modulators of infection. Convalescent plasma samples from the top 30 neutralizers in a cohort of 148 individulas were pooled to create a mixture with an NT50 titer of 1597 against HIV-1 pseudotyped with SARS-CoV-2 S protein 23 . MISTRG6-hACE2 mice were treated with the mixed plasma 8 hours before infection with SARS-CoV-2 (Fig. 5A). The treated mice had significantly lower viral titers in lungs at 4dpi and therefore the plasma was only partially effective ( Fig 5A). Yet, prophylactic convalescent plasma did not prevent human immune cell infiltration, particularly inflammatory macrophages, to the lungs (  (Fig. 5I). Next, we tested whether MISTRG6 mice could be used to model therapeutic mAb therapy. While therapeutic treatment of mAbs similarly prompted infectious viral clearance at both early (11hpi) and late time points (36hpi; Fig. 5J), by contrast to prophylaxis, mAbs failed to prevent immune infiltration in lungs (Fig. 5K). Humanized mice treated with both mAbs early (11h) post infection had fewer immune cells in BAL at 4dpi compared to untreated and late (35h)-treated mAb groups, suggesting that the immune-infiltrate and inflammatory responses are attenuated when mice are treated with mAbs early but less so when treated later in infection (Fig. 5L). Although neither therapeutic intervention prevented weight loss, early treatment prevented systemic T cell lymphopenia ( Fig. 5M, S5G). By contrast, later administration of neutralizing mAbs showed little effect, and a similar infiltration profile as untreated mice at 4dpi with enrichment in inflammatory macrophages and monocytes (Fig S5H,I). These findings highlight clear efficacy of mAb treatment in controlling viral infection and viral titers but they underline the need for early treatment particularly in controlling the immunopathology, as has been noted clinically 22,53,54 Accurate timing of corticosteroids is necessary to balance viral clearance and prevent immunopathology.
Our transcriptome analysis revealed glucocorticoids as possible upstream regulators of DEGs that are induced in infected lungs (Fig 4A-B). Moreover, given that dexamethasone has been so far the only therapeutic treatment that has impacted recovery and reduced mortality in a major way when given in patients with severe disease 56 , we hypothesized that dexamethasone treatment in humanized mice may favorably impact immunopathology in mice infected with SARS-CoV-2. To test this hypothesis, we treated mice with dexamethasone for 3 days starting at 7dpi once the immune infiltration is established but viral titers were significantly reduced in the lungs (Fig. S6A). Indeed, mice treated with dexamethasone close to viral clearance (7dpi), recovered rapidly in weight by 14dpi and returned to weight gain comparable to their uninfected counterparts (Fig. 6A,1D). Dexamethasone treatment reduced human immune infiltrate and reversed many aspects of immune-activation (Fig. 6B). Mouse neutrophils in BAL were fewer in dexamethasone treated mice ( Fig S6B). Macrophages, in particular inflammatory macrophages, were largely absent in the lungs of dexamethasone treated mice (Fig 6C-F). Dexamethasone treatment also blocked accumulation of pDCs and reduced T cell activation in lungs at both 14dpi and 28dpi (Fig 6J-L and S6E). Interestingly, dexamethasone treatment also blocked IgG specific B cell response as IgG+ B cells but not IgM+ B (Fig S6C). Relative contrubitions of these B cells to immune-pathology vs immune-protection remain open questions. It was notable that lack of immune cells in dexamethasone treated lungs also correlated with lowered viral RNA levels by 28dpi ( Fig S6G). Given that the immune infiltrate is established early in mice (by 4dpi), we were prompted to investigate the timing of dexamethasone mediated control of immunopathology for COVID19. We therefore treated mice with dexamethasone for 3 days starting at 3dpi once the immune infiltration is established. In stark contrast to late dexamethasone treatment, early dexamethasone treated mice became moribund by 7dpi with rapidly declining weights compared with untreated mice (Fig. 6F).
Dexamethasone-treated mice had significantly fewer immune cells infiltrating the lungs and in particular lacked inflammatory macrophages (Fig. 6K, S6G). Importantly, the disabled antiviral response in these mice led to significantly higher viral load in the lungs (Fig. 6L). These findings upon early dexamethasone treatment highlight the importance of the early antiviral response to contain viral infection treatment. The timing of dexamethasone treatments was instrumental in showcasing the necessity of the early anti-viral response in containing infection and the role of immune cells later in disease pathology.

Discussion:
Accurate model systems that rapidly and comprehensively characterize COVID19 are and should remain pivotal in the development of promising discoveries. Here, we reveal a humanized mouse COVID-19 model that combines vector-based delivery of human ACE2 and a comprehensive human immune system that recapitulated both the innate and adaptive human immune systems during infection with SARS-CoV-2. Our unique model allowed for the first time chronic disease with SARS-CoV-2 infection in the context of patient-derived antibodies and characterized the potential players for immunopathology (Table S4), in particular the aberrant macrophage response that is thought to be the effector of disease morbidity and ARDS in patients 57 .
Our findings document that gross disease parameters such as weight loss and viral load were driven by the human immune system in our model, which suggested that human immune cells contribute uniquely to the pathology of human SARS-CoV2 infection. Sustained viral RNA and gross clinical features including failure to recover body weight even at very late time points post infection,are unique among animal models to our MISTRG6 humanized mice, with human immune cells and human ACE2 expression. Chronic disease manifestations were reflected in histopathological assessment of infected humanized lungs late in infection (14dpi and 28dpi). Significant cellular infiltrates, thickened septa and collagen deposition in lungs at 28dpi point to lack of recovery and fibrosis in infected humanized lungs long after infectious virus is cleared, recapitulating what is observed in severe human COVID19. To our knowledge, this is the only disease model that recapitulates chronic weight loss, sustained high viral RNA and chronic histopathology with pulmonary fibrosis seen in human patients 4,29-31,58 and has not been observed in any of the prior animal (mice, ferrets, hamsters) models of COVID19 (Table S4) 18,[25][26][27][28] . Nonetheless,chronic, humanized COVID-19 is not a lethal disease, which will make interesting further investigation of variables such as age, pre-existing health conditions and co-morbidities that contribute to high case fatality rate in humans. We created a humanized mouse model of COVID19 that recapitulates the distribution and function of the human innate and adaptive immune system amenable to the mechanistic study of COVID 19 and its myriad of complications (Table S4). MISTRG6-hACE2 mice could be used to study two aspects of the immune response, infectious viral clearance and immunopathology recapitulated in our model. We Our humanized mouse model of COVID19 is uniquely adapted to reflect patient heterogeneity but also provides consistency in a highly reproducible mouse model. Transcriptome analysis revealed differences between individual animals in the strength of the inflammatory response, which may in part help explain the variable outcome observed in disease morbidity, and mortality in human SARS-CoV-2 infection 59 . Yet, regardless of such heterogeneity, sustained interferon response, as has been postulated in humans 30,33,44 , was a common theme that shaped both the early anti-viral innate response as well as the late adaptive immune response in humanized COVID-19.
Emerging patient data detail more debilitating effects of COVID-19 in certain patient groups even in absence of previously described high risk criteria (age, pre-existing health conditions etc.). Although socio-economic factors are likely responsible for some or perhaps all of these effects, perturbation of our system should allow testing of the genuinely medical effects. Our humanized mouse system can be completely personalized by matching patient HSPCs with antibodies and medical history, allowing researchers to test novel therapeutics and other immunomodulatory agents to address conflicting reports in pre-clinical models and to predict efficacy in patients.

Acknowledgements:
The generation of the original

Data Avaliability:
All data that support the findings of this study are available within the paper and its Supplementary Information files. All 10x Genomics single cell RNA sequencing and bulk RNA sequencing data that support the findings of this study will be deposited in the Gene Expression Omnibus (GEO) repository with an accession code to be determined.

Competing financial interests
The authors declare no competing financial interests.   A. Humanization measured by ratio of human CD45+ cells in total CD45+ cells (mouse and human CD45+ combined) in blood, lung and BAL of uninfected and infected MISTRG6-hACE2 mice. N=4-12. B. Human immune cell numbers in lungs and BAL of uninfected and infected mice at 2,4,7,14,28 dpi. N=4-12.   C. t-distributed stochastic neighbor embedding (t-SNE) plot with clustering results of single cell RNA sequencing of human immune cells from lungs at 4dpi. Single cell suspensions from whole infected lung at 4dpi were processed and sequenced. There were 421 cells identified as human immune cells. D. Expression of cluster identifying genes in human immune cells described in C. E. Cluster distribution and expression of human inflammatory cytokines for clusters described in C-D. F. Normalized counts for inflammatory cytokines implicated in COVID19 patients. Counts were reported separately for human(red) and mouse(blue) cytokine genes. G. Heatmap of genes that are implicated in SLE like B cells based on GSE10325 (Hutcheson et al., 2007) in infected lungs of MISTRG6 mice at 2,4,7.14.28 dpi. Row min and max of transformed values, calculated by subtracting row mean and diving by STD for each gene across all samples, are visualized.       F. Schematic of experimental design of SARS-Cov2 infected MISTRG6-hACE2 mice either treated with dexamethasone on days 3,4,5 dpi or left untreated. G. CD206 and CD68 expression in lung human immune cells in mice treated with dexamethasone or left untreated at 7dpi. CD206hi+ CD68+ cells are alveolar macrophages. N=4-6.
Supplementary Table Legends:   Table S1. Differentiyally regulated genes (DEGs) in SARS-CoV-2 infected lungs. Normalized counts, foldchanges in infected lungs compared with respect to uninfected mice and adjusted p values are presented for 516 mouse and 285 human differentially regulated genes.

AAV-hACE2 administration
AAV9 encoding hACE2 was purchased from Vector Biolabs (AAV9-CMV-hACE2). Animals were anaesthetized using isoflurane . The rostral neck was shaved and disinfected. A 5-mm incision was made, and the trachea was visualized. Using a 32-G insulin syringe, a 50-µl injection dose of 10 11 genomic copies per milliliter of AAV-CMV-hACE2 was injected into the trachea. The incision was closed with 4-0 Vicryl suture and/or 3M Vetbond tissue adhesive. Following administration of analgesic animals were placed in a heated cage until full recovery. Mice were then moved to BSL-3 facilities for acclimation.

Viral titers
Mice were euthanized in 100% isoflurane. Approximately half of the right lung lobe was placed in a bead homogenizer tube with 1 ml of PBS + 2% FBS.

Transplantation of human CD34+ hematopoietic progenitor cells into mice.
Fetal liver samples were cut in small fragments, treated for 45 min at 37 °C with collagenase D (Roche, 200 μg/ml), and prepared into a cell suspension. Human CD34+ cells were purified by performing density gradient centrifugation (Lymphocyte Separation Medium, MP Biomedicals), followed by positive immunomagnetic selection with EasySep ™ Human CD34 Positive Selection Kit (Stemcell). For intrahepatic engraftment, newborn 1-3 day-old pups were injected with 20,000 fetal liver CD34+ cells in 20 μl of PBS were injected into the liver with a 22-gauge needle (Hamilton Company). All use of human materials was approved by the Yale University Human Investigation Committee.

Isolation of cells and flow cytometry
All mice were analyzed at approximately 9-11 weeks of age. Single cell suspensions were prepared from blood, spleen BAL and lung. Mice were euthanized with 100% isoflurane. BAL was performed using standard methods with a 22G Catheter (BD). Blood was collected either retro-orbitally or via cardiac puncture following euthanasia. BAL was performed using standard methods with a 22G Catheter (BD) 61 . Lungs were harvested, minced and incubated in a digestion cocktail containing 1 mg/ml collagenase D (Sigma) and 30 µg/ml DNase I (Sigma-Aldrich) in RPMI at 37°C for 20 min. Tissue was then filtered through a 70-µm filter. Cells were treated with ammonium-chloride-potassium buffer and resuspended in PBS with 1% FBS. Mononuclear cells were incubated on ice with human (BD) and mouse (BioxCell, BE0307) Fc block for 10 min. After washing, primary antibody staining was performed at 4C for 20 min. After washing with PBS, cells were fixed using 4% paraformaldehyde. For intracellular staining, cells were washed with BD permeabilization buffer and stained in the same buffer for 45 min at room temperature. Samples were analyzed on an LSRII flow cytometer (BDBiosciences). Data were analyzed using FlowJo software.

Antibodies
Antibodies against the following antigens were used: Nussenzweig as has been previously described 23 .
Bulk whole tissue lung RNA-sequencing RNA isolated from homogenized lung tissue used for viral RNA analysis was also used for whole tissue transcriptome analysis. Libraries were made with the help of the Yale Center for Genomic Analysis. Differential expression analysis was also performed with DESeq2. For IFN-stimulated gene identification, http://www.interferome.org was used with parameters -In Vivo, -Mus musculus or Homo sapiens -fold change up 2 and down 2.

Single Cell RNA Sequencing 10X Genomics
Single cell suspensions from digested lungs were processed for droplet based scRNA-seq and 10000 cells were encapsulated into droplets using 10X Chromium GEM technololgy. Libraries were prepared in house using Chromium Next GEM Single Cell 3ʹ Reagent Kits v3.1 (10X Genomics).
scRNA-seq libraries were sequenced using Nova-Seq. Raw sequencing reads were processed with Cell Ranger 3.1.0 using a human-mouse combined reference to generate a gene cell count matrix. To