key: cord-0871164-n3xj8p78 authors: Szabo, P. A.; Dogra, P.; Gray, J. I.; Wells, S. B.; Connors, T. J.; Weisberg, S. P.; Krupska, I.; Matsumoto, R.; Poon, M. M. L.; Idzikowski, E.; Morris, S. E.; Chloe, P.; Yates, A. J.; Ku, A.; Chait, M.; Davis-Porada, J. M.; Zhou, J.; Steinle, M.; Mackay, S.; Saqi, A.; Baldwin, M. R.; Sims, P. A.; Farber, D. L. title: Analysis of respiratory and systemic immune responses in COVID-19 reveals mechanisms of disease pathogenesis date: 2020-10-19 journal: medRxiv : the preprint server for health sciences DOI: 10.1101/2020.10.15.20208041 sha: 4653add0c5b4ed173bb4eed3afb6598a2dc6f565 doc_id: 871164 cord_uid: n3xj8p78 Immune responses to respiratory viruses like SARS-CoV-2 originate and function in the lung, yet assessments of human immunity are often limited to blood. Here, we conducted longitudinal, high-dimensional profiling of paired airway and blood samples from patients with severe COVID-19, revealing immune processes in the respiratory tract linked to disease pathogenesis. Survival from severe disease was associated with increased CD4+T cells and decreased monocyte/macrophage frequencies in the airway, but not in blood. Airway T cells and macrophages exhibited tissue-resident phenotypes and activation signatures, including high level expression and secretion of monocyte chemoattractants CCL2 and CCL3 by airway macrophages. By contrast, monocytes in blood expressed the CCL2-receptor CCR2 and aberrant CD163+ and immature phenotypes. Extensive accumulation of CD163+monocyte/macrophages within alveolar spaces in COVID-19 lung autopsies suggested recruitment from circulation. Our findings provide evidence that COVID-19 pathogenesis is driven by respiratory immunity, and rationale for site-specific treatment and prevention strategies. The novel respiratory virus SARS-CoV-2 has resulted in devastating impacts to the 2 world's population, both as a result of morbidity and mortality caused by COVID-19, as well as 3 the life-altering measures implemented to mitigate spread. While the majority of infected 4 individuals (>90%) develop a self-limiting disease and recover, approximately 5-10% of 5 individuals develop severe respiratory disease marked by lung infiltrates and reduced oxygen 6 saturation, which can progress to acute respiratory distress syndrome (ARDS), multi-organ Mononuclear cells from paired airway and blood samples were isolated by centrifugation 1 through ficoll (see methods), stained using a 34 marker panel containing antibodies specific for 2 major lineage determinants and markers for differentiation, tissue residence, activation, and 3 function (see methods), and analyzed by spectral flow cytometry (gating strategy for 4 mononuclear cells shown in Figure S1 ). Principal component analysis (PCA) of mean marker 5 expression for each sample showed distinct clustering of airway and blood samples by site, but 6 not by outcome or by patient ( Figure 1B , Table S4 ). Computational analysis of flow cytometry 7 data visualized by uniform manifold approximation and projection (UMAP) embedding (see 8 methods) showed distinct separation of the major lineages into monocytes/macrophages, CD4 + T 9 cells, CD8 + T cells, B cells, and innate lymphoid cells (predominantly NK cells) for all samples 10 ( Figure 1C , Figure S2A ). Compiled data for each timepoint revealed distinct immune cell 11 composition in airway compared to blood ( Figure 1D ). Airway samples had predominant 12 frequencies of monocytes/macrophages (40-90%), lower T cell frequencies, and very low-to- 13 negligible frequencies of B cells and ILCs, while blood contained higher lymphocyte frequencies 14 with monocytes comprising ~50% of all non-neutrophil leukocytes ( Figure 1D ). Similar immune 15 cell compositions were confirmed by scRNAseq analysis of airway and blood from four 16 individuals ( Figure S2B ). These results show distinct immune cell profiles in airway compared to 17 blood across all patients and timepoints analyzed. 18 We investigated whether the airway or blood immune cell composition differentiated 19 between patients or correlated with overall survival. Hierarchical clustering of aggregated immune cell composition for individual patient samples also showed distinct composition in 5 airways, which did not correspond to blood either in frequency or in changes over time ( Figure 6 1F, Figure S3A ). However, examination of specific lineages showed significant associations with 7 outcome and correlation with age. Notably, there was a significant decrease in frequencies of 8 airway monocytes/macrophages and an increase in airway CD4 + T cells in patients who survived 9 the disease versus those who succumbed ( Figure 1G , left, Figure S3B ), while the frequency of 10 the corresponding blood immune cell subsets did not significantly differ between patients based 11 on outcome, nor did they correlate with age ( Figure 1G , right, Figure S3B ). Accordingly, 12 clustering the longitudinal patterns of cell type frequencies using K-means further suggests that 13 airway immune cell trajectories are a better indicator of clinical outcome than their blood 14 counterparts ( Figure S3C ). Together, these results show that airways exhibit an immune cell 15 composition distinct from blood, and that the dynamics of airway T cells and 16 monocyte/macrophages are significantly associated with outcome, suggesting key roles for these 17 cell types in disease pathogenesis. 18 19 Tissue resident memory T cells are the major T cell subset in airways 20 The subset composition and transcriptional profile of airway T cells in comparison to 21 those in blood was further examined through high-dimensional, single cell approaches. Multiple 22 markers of T cell differentiation were used to distinguish naïve and memory populations 23 8 (CD45RA, CCR7, CD95, CD27), activation (HLA-DR, PD-1), functional subsets (FOXP3/CD25 1 for Tregs, CXCR5/PD-1 for Tfh-like, TCRGD for  T cells), specific states of senescence or 2 terminal differentiation (CD57, KLRG1), and tissue residence (CD69, CD103). We used UMAP 3 embedding to visualize expression of these multiple markers by airway and blood T cells, 4 showing increased expression of CD69, CD103, PD-1, and HLA-DR in the airways and 5 increased CCR7, CD45RA, and CD127 expression in the blood ( Figure S4A and S4B). 6 Phenograph clustering based on marker expression by CD4 + and CD8 + T cells yielded 27 7 clusters, which were coalesced into 15 clusters denoting biological subsets or 8 functional/activation states in airway and blood (Figure 2A , 2B). 9 There were significant qualitative and quantitative differences in T cell subset 10 composition and activation state between airway and blood. In particular, airway contained CD4 + 11 and CD8 + TRM cells (CD69 + CD103 +/-) along with activated TRM subsets expressing elevated 12 levels of HLA-DR and PD-1, and reduced levels of CD127 compared to non-activated TRM 13 ( Figure 2A, 2B) . TRM cells, regardless of activation state, were largely confined to the airways 14 and not significantly present in blood ( Figure 2B , 2C), consistent with virus-responding T cells 15 located at the site of infection. Innate-like  cells were also present in higher frequencies in 16 airways compared to blood ( Figure 2C ). Circulating TEM cells (CD69 -CD103 -) were present in 17 both sites, with non-activated CD8 + TEM enriched in the blood ( Figure 2B, 2C ). Blood also 18 contained higher frequencies of naïve CD4 + and CD8 + T cells, CD4 + TCM cells ( Figure 2B , 2C). 19 Between patients, there was variability in the proportions of the major subsets represented; most 20 patients (9/13) had predominant CD8 + TRM in airways, while 3/13 patients had higher 21 frequencies of CD4 + TRM in airways ( Figure 2D ). Together, these analyses indicate that both Table S5 ). Airway T cells also showed upregulated expression of genes encoding key cytokines 13 and chemokines, including IFNG, CCL2, and CCL4 ( Figure 2E , F), consistent with an activated 14 and pro-inflammatory state. By contrast, blood T cells exhibited higher expression of genes context of relatively quiescent blood T cells, suggesting that the protective T cell response is 19 targeted to the respiratory environment. The vast majority of B cells profiled in COVID-19 patients were from the blood; 1 however, there was a small but detectable population in the airways ( Figure 1C, 1D) . Comparing 2 B cell profiles by flow cytometry analysis revealed differential expression of key B cell markers 3 delineating specific B cells subsets in the airways and blood ( Figure S5A , S5B). In particular, 4 airway B cells exhibited increased expression of CD69, a marker expressed by human tissue 5 resident B cells (Weisel et al., 2020) , and activation markers CD86 and CD95 ( Figure S5A providing further support for spatial segregation of adaptive immunity. Airway monocytes/macrophages exhibit activation and inflammatory profiles 12 We applied similar high-dimensional flow cytometry and scRNA-seq analysis to the 13 monocyte/macrophage populations in paired airway and blood samples from COVID-19 14 patients. Phenotypic profiling of airway and blood samples defined a major 15 monocyte/macrophage population (see Figure 1 ), which largely segregated by site ( Figure 3A ). Expression of markers HLA-DR, CD11c, and CD16 distinguished airway from blood 17 monocyte/macrophages, while those in blood expressed higher levels of CD14 and CD163 18 ( Figure 3A, B) . There was no difference in the relative expression of CD64 and CD86 between 19 airway and blood monocyte/macrophages ( Figure 3B ). Phenograph clustering of 20 monocytes/macrophages identified 20 clusters, which were coalesced into 6 clusters classified by Figure 3C ). Non-classical monocytes/macrophages (both non-23 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . activated and activated) and activated classical monocytes/macrophages were enriched in the 1 airway ( Figure 3D , E). By contrast, classical and intermediate monocytes/macrophages without 2 activation markers were increased in the blood ( Figure 3D , E). These data indicated increased 3 activation of monocyte/macrophage lineages in the airway compared to blood. 4 We further investigated the subset delineation, differentiation, and functional state of 5 monocyte/macrophages by scRNA-seq. Transcriptionally, airway monocyte/macrophages 6 exhibited certain shared and distinct gene expression patterns compared to blood counterparts, 7 which were consistent across individuals ( Figure 4A , Table S6 ). There was comparable Compartmentalized production of cytokines and chemokines in airway and blood 16 We further assessed inflammation in both sites by direct examination of cytokine and 17 chemokine content in airway supernatants and plasma samples from an early (day 1) and later 18 (days 3-7) timepoint for each patient (Table S2) . We used a microfluidic chip multiplexed 19 secretome proteomic platform for assessment of soluble mediators from each site with high 20 sensitivity from small volumes (see methods)(Farhadian et al., 2020). Overall, we found major, 21 significant differences in the cytokine and chemokine protein content in the airway compared to 22 plasma, but no significant differences between the two timepoints within a site ( Figure 5A , 23 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . Figure S7A ,B). Analytes significantly elevated in airways compared to blood include 1 monocyte/macrophage chemoattractants MCP-1 (CCL2), MIP-1CCL3and MIP1CCL4 2 in all samples, as well as granzyme B, IL-7, and TNF- associated with T cells and homeostasis 3 ( Figure 5A , 5B, Figure S7B ). By contrast, in the blood, MCP-1 (CCL2), MIP-1CCL3, 4 granzyme B, TNF- and IL-7 were undetectable, while MIP-1 (CCL4was present at variable 5 levels across patients ( Figure 5A, B) . Both blood and airways contained low and/or variable 6 levels of molecules associated with T cell effector function (perforin, IFN-, IL-17, and IL-2), 7 additional innate cytokines (IL-6 and IL-8), and TGF-while none of the analytes measured 8 were uniquely expressed by blood and not found in airways ( Figure 5A , 5B, Figure S7B ). Together, these results show compartmentalized production of pro-inflammatory chemokines 10 and cytokines in the airway with a subset of these detected in blood, suggesting that systemic 11 cytokines may derive from inflammatory processes originating at the infection site. To define the cellular origin of the chemokines and cytokines detected in each 13 compartment, we analyzed transcript expression for each of the analytes from Figure 5B by 14 scRNA-seq. Overall, transcript expression of prominent cytokines/chemokines largely correlated 15 to the protein data; airway myeloid cells expressed high levels of CCL2, CCL3 and CCL4 16 transcripts corresponding to the high levels of the respective proteins in airways, while blood 17 myeloid cells expressed lower or undetectable levels of these transcripts ( Figure 5C ). Airway and 18 blood myeloid cells also expressed CXCL8 and TGFB1, consistent with the protein data ( Figure 19 5C). In the airways, T cells expressed GZMB, CXCL8, CCL4, PRF, IFNG, and TGFB1 20 transcripts, which were expressed by blood T cells at lower or variable levels ( Figure 5C ). Airway epithelial cells expressed predominantly CXCL8 transcripts, as well as lower levels of 22 transcripts for IL-7 and several chemokines ( Figure 5C ). Overall, these results demonstrate 23 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . 14 compartmentalized secretion of monocyte/macrophage-derived chemokines and inflammatory 1 mediators in the airways with potential roles for recruiting immune cells to the lung that may 2 contribute to lung inflammation and tissue damage. 3 4 COVID-19-induced features of airway and blood immune cells 5 To assess COVID-19-related alterations in airway and blood immune cells that could 6 potentially contribute to disease pathogenesis, we obtained baseline controls of blood from 7 uninfected, healthy adults, and airway washes from lungs of SARS-CoV-2-negative organ To more closely examine site-specific differences between immune cells in healthy and 22 COVID-19 individuals, we analyzed T cell and monocyte/macrophage populations separately. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. Figure 6C ). Comparing T cells in healthy and COVID-19 samples, there were increased 7 frequencies of CD69 + CD103 + TRM and T cells expressing activation markers HLA-DR and PD- 8 1 in the airways of COVID-19 patients compared to uninfected individuals; these markers were 9 not expressed significantly by blood T cells in COVID-19 nor in healthy controls ( Figure 6D ). The compartmentalized activation of T cells in airways in COVID-19 provides further evidence 11 for dynamic T cell immunity at the infection site. For monocytes/macrophages, UMAP embeddings revealed compartmentalized profiles 13 between healthy airway and blood, but considerable overlap of monocyte/macrophage profiles 14 between COVID-19 airway and blood ( Figure 6E ). Accordingly, Minkowski distance 15 calculations confirmed that healthy airway and blood monocyte/macrophage subsets were (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . significantly reduced proportions of HLA-DR hi and CD86 hi cells ( Figure 6F ), consistent with 1 recent findings regarding blood monocyte profiles in severe COVID-19 and suggestive of an 2 immature phenotype (Schulte-Schrepping et al., 2020). Taken together, these results indicate 3 profound alterations in blood monocytes in COVID-19, which also share similar features with 4 airway macrophages, suggesting that airway resident myeloid cells in severe COVID-19 may 5 derive, in part, from these circulating precursors and that interactions between airway and blood 6 myeloid cells may contribute to disease pathology. 7 Accumulation of CD163 + cells in the lungs of severe COVID-19 patients 8 We hypothesized that the production of monocyte-chemoattractant chemokines by airway 9 monocyte/macrophages along with the elevated levels of CD163 + monocytes in COVID-19 10 blood may result in their dysregulated infiltration into the lung. We therefore examined immune (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . similarly increased ( Figure 7B ). We assessed expression of genes associated with cell cycle or 1 proliferation (Ki67, TOP2A, UBE2C) in monocyte/macrophage populations in the airway or 2 blood by scRNA-seq, revealing no significant expression of these markers ( Figure 7C ). Together 3 with the high-dimensional analysis of airway immune cells, these findings implicate the 4 recruitment of immature monocytes from the periphery into the lung, where they subsequently 5 become highly pro-inflammatory and drive the pathogenesis of severe COVID-19. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Characterizing respiratory immune responses in situ in the context of circulating immune cell 11 populations is needed to dissect mechanisms of disease pathogenesis to combat this pandemic. In this study, we obtained paired respiratory and blood samples from patients with severe (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . chemokines, which direct recruitment of multiple immune cell types. Our findings rather suggest 1 that pro-inflammatory cytokines emanating from the respiratory tract recruit circulating 2 inflammatory monocytes to the lungs and perpetuate lung damage. Immunofluorescence imaging 3 of lungs from severe COVID-19 patients shows a striking increase in CD163-expressing 4 monocytes/macrophages within the damaged lung tissue that lack proliferative signatures and 5 therefore likely derive from recruitment. These cells specifically accumulate in the alveolar 6 spaces of the lungs, a key site for blood gas exchange, suggesting their involvement in diffuse Our finding that increased proportions of airway T cells are associated with better 21 outcome and younger age suggests that promoting lung-localized immune responses is an 22 important consideration for vaccine design. In mouse models, intranasal administration of the 23 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . 1 We wish to express our gratitude to the Medical ICU nurse champions, Cora Garcellano, Tenzin 2 Drukdak, Harriet Avila Raymundo, Lori Wagner, and Ricky Lee, who led the efforts to obtain (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . P.A.Sz., S.W., J.G., P.D. processed samples, designed and optimized high-dimensional flow 1 cytometry panels, analyzed data, made figures, and wrote the manuscript. P.A.Sz. and S.W. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . The 20 phenograph clusters were collapsed into 6 cellular subsets based on common myeloid (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . and healthy controls across airway and blood (upper two panels). Lower panels indicate 23 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . ILC, B cells, T:myeloid and CD4:CD8 content in airway (left) and blood (right) with outcome 1 (deceased or survived) and correlation with age. Statistical significance was calculated using 2 Mann-Whitney U-tests (box-plots) or Pearson correlations (scatter plots) and indicated by ***, p 3 ≤ 0.001; **, p ≤ 0.01; *, p ≤ 0.05. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. For figures, p-value < 0.05 = *, p-value < 0.01 = ** and p-value < 0.001 = **. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. This study did not generate new unique reagents. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. Donors were free of cancer, chronic diseases, seronegative for hepatitis B, C, and HIV, and 6 negative for SARS-CoV-2 by PCR (Table S7 ). Use of organ donor tissues does not qualify as 7 "human subjects" research, as confirmed by the Columbia University IRB as tissue samples were 8 obtained from brain-dead (deceased) individuals. Whole blood collected in heparinized vacutainers was centrifuged at 400 x g for 10 min at room 12 temperature (RT) to isolate plasma, which was then stored at −80 °C for subsequent analysis. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . To collect airway supernatants, DPBS was added 1:1 directly to airway samples and centrifuged 1 at 400 x g for 10 min at RT. The resulting supernatants were stored at −80 °C for subsequent 2 analysis. To isolate airway MNCs, samples were treated with Benzonase (Millipore Sigma), 3 purified through 100 µm filters, and centrifuged on a density gradient using Ficoll-Paque PLUS. Table S3 . 20 Non-diseased lungs were obtained from deceased organ donors as described above. Airway 21 washes were obtained by flushing out the major airway with 60 mL saline as described (Snyder 22 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . incubated on ice for 20 min. For intracellular staining, surface stained cells were fixed for 25min 1 at RT in fixing buffer (Invitrogen cat# 00-5123-43), followed by staining in permeabilization 2 buffer (Invitrogen cat# 00-8333-56) at RT for 30 min. Cells were washed and data was collected 3 on 5-lazer Cytek ® Aurora machine (Cytek Bio). Highly-multiplexed CodePlex chip secretome proteomics 6 Cryopreserved tracheal washes and plasma were thawed at room temperature for 30-60 minutes (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . 14 Flow cytometry data was pre-gated to exclude any doublets, dead cells and CD66b + granulocytes 15 using FlowJo v 10.7 (Tree Star) ( Figure S1 ). Cleaned data was exported as .fcs files with 16 compensated parameters and analyzed further and visualized using a Python (v3.7) (Python 17 Software Foundation. Python Language Reference, version 2.7.) computational pipeline. In brief, 18 first the data was filtered to remove any noise using quantile gates; events that fell below 0.01% 19 of marker expression intensity were removed from the sample. Following initial filtering, data 20 from COVID-19 and healthy samples was merged after subsetting 70,000 events from each 21 sample. Any sample with fewer than 1000 events was removed from further analysis. The 22 merged dataset as was transformed using arcsinh function from Python numpy library(van der 23 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. remove any residual granulocyte contamination identified as clusters of CD45 lo CD66b + cells. 6 The resulting "no neutrophil dataset" dataset was split into COVID-19 and healthy samples and 7 used for downstream analysis. 8 For further analysis we downsampled the "no neutrophil COVID-19 dataset" to include 9 20,000 events from each of 141 longitudinal samples and was used to run PCA analysis at 10 sample level using mean expression of markers in each sample, PCA loadings provided in Table 11 S3. We ran UMAP dimensionality reduction (k = 60) on this dataset using 14 lineage-defining all time points and used for hierarchical clustering of samples using "ward" method and 18 "jensenshannon" metric. 19 For lineage specific analysis, we ran UMAP dimensionality reduction and subsequent Figure S4A . B cell markers used are shown in Figure S5A and monocyte/macrophages shown 23 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . 40 in Figure 3A . Major cell subset clusters were identified and functionally similar subsets were 1 coalesced and manually annotated. Heatmaps were generated for average marker expression in 2 each cluster. Data are presented as row normalized expression of marker across all clusters. For analysis of COVID-19 and healthy samples, paired blood and airway samples across 4 all timepoints from COVID-19 donors were downsampled to 5,000 events, and each healthy 5 airway and blood sample was downsampled to 30,000 events and merged to create a reduced "no 6 neutrophil Healthy + COVID-19 dataset". UMAP dimensionality reduction and identification of 7 major cell lineages was done as described above for the COVID-19 dataset. To evaluate outcome was robust to this definition of distance, giving identical results when performed using 23 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . log-or logit-transformed frequencies. Classification performance was defined as the percentage 1 of donors that were assigned to the correct outcome cluster (i.e. deceased or survived). We Processing of scRNA-seq Data 9 We used kallisto v0.46.2 in "BUS" mode to pseudo-align the raw reads for each sample to a 20 We merged the scRNA-seq data from all of the airway samples and identified likely markers of 21 specific subpopulations using the previously described drop-out score method for finding genes (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . ). Next, we computed a cell-by-cell Spearman's rank correlation matrix using 1 these putative marker genes. Using this matrix, we constructed a k-nearest neighbor's graph 2 (k=20) as input for Louvain community detection as implemented in Phenograph (Levine et al., 3 2015). To associate the resulting clusters with major cell populations in the airway, we examined 4 the statistical enrichment of the following marker genes in each cluster using the binomial test as HBA2, HBB). We identified clusters as likely multiplets based on co-expression of multiple 11 marker sets (e.g. clusters enriched in both CD14 and CD3D were marked as likely T cell / 12 myeloid cell multiplets). All of the cells in these clusters were marked as multiplets. 13 In the main text, we present focused analyses on myeloid cells, T cells, and 14 epithelial/club/goblet cells from the airway. To further refine our annotation, we re-clustered the 15 cells annotated as each of these three cell types separately using the methods described above. 16 We then re-analyzed the enrichment of cell type-specific markers in the resulting new clusters. As expected, this focused re-analysis of each of these three major populations identified 18 additional putative multiplet clusters and cells that we likely mis-clustered in the initial merged 19 analysis. We conducted two rounds of re-clustering for each of these three major cell types to 20 produce a refined annotation. The top of Figure S2B shows a gene expression heatmap for key 21 markers genes in the merged airway data set colored by patient and cell type annotation. We 22 repeated the above procedure for the merged blood scRNA-seq profiles including a focused re-23 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . clustering analysis of the cells that we originally annotated as myeloid and T cells for refinement. 1 The bottom of Figure S2B shows a gene expression heatmap for key markers genes in the 2 merged blood data set colored by patient and cell type annotation. 5 We generated merged UMAP embeddings for the blood and airway T cells ( Figure 2E , F) and 6 the blood and airway myeloid cells (Figure 4) . In each case, we first identified genes that were 7 likely to contaminate either the myeloid or T cell profiles in either the blood or airway to avoid 8 including them in any of our downstream clustering, visualization, or differential expression 9 analysis. We conducted pairwise differential expression analysis between all of the cells 10 annotated as a cell type-of-interest (e.g. myeloid) and each group of cells with a different 11 annotation for the blood and airway from each patient separately. For each pairwise comparison, 12 we randomly subsampled the two groups of cells to the same cell number. Next, we randomly 13 subsampled the molecular counts for cells in the two groups such that they have the same 14 average number of counts per cell. We then generated a merged count matrix for the two groups for false discovery using the Benjamini-Hochberg Procedure with the function multipletests from 20 the Python package statsmodels. Using the results of pairwise differential expression analysis, 21 we generated a blacklist of genes for a given cell type by taking any gene with at least 10-fold 22 enrichment in a different cell type with FDR<0.001 in at least two patients. We removed all 23 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . genes with any enrichment in the cell type-of-interest with FDR<0.001 in any patient to avoid 1 eliminating patient-specific markers of the cell type-of-interest. The final blacklists for blood and 2 airway myeloid and T cells appear in Table S8 . 3 For both myeloid and T cells, we took all of the cells in the data set that we had annotated 4 as each of these two cell types and used the drop-out score method described above to generate a 5 list of putative, highly variable marker genes for each patient. Next, we generated a merged 6 count matrix across all patients for a given cell type, which we normalized using the pooling Harmony-corrected principal components. These embeddings appear in Figures 2 and 4 . 18 For the differential expression analysis between blood and airway myeloid cells and 19 between blood and airway T cells, we used the Mann-Whitney U-test approach described above. 20 We removed genes on the blacklists described above for each cell type prior to subsampling, 21 normalization, and statistical testing. We also restricted this analysis to protein-coding genes and 22 removed all T cell receptor and immunoglobulin variable regions. We performed differential 23 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . expression separately on each pair of matched airway and blood samples (there are 12 patient 1 time points for which we have matched samples). Stringent criteria were used to select the 2 differentially expressed genes displayed in the heatmaps in Figures 2 and 4 . For the myeloid cell 3 heatmap in Figure 4 , a gene had to be differentially expressed with a fold-change of at least 4 in 4 either direction and FDR<0.001 in at least 9 of the 12 matched sample pairs. For the T cell 5 heatmap in Figure 2 , we applied the first two criteria, but required them in only 6 of the 12 6 matched sample pairs. Results for all of the pairwise differential expression analyses comparing 7 airway and blood T and myeloid cells can be found in Table S5 and Table S6 , respectively. Lung tissue imaging analysis 19 Tissue segmentation was performed using inForm software on 10-30 representative fields (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . Differences in mean between two sample groups were compared using Mann-Whitney U test, 10 multiple group comparisons were done using ANOVA followed by Tukey's HSD post-test and (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . 1 Table S1 . Clinical information for COVID-19 patients in this study. 2 Table S2 . Assays performed on the samples from individual COVID-19 patients. 3 Table S3 . Summary of sample details for scRNA-seq analysis. 4 Table S4 . PCA loadings of markers for PC1 and PC2. 5 Table S5 . Differential gene expression by T cells in airway versus blood for each sample by Table S7 . Deceased donors for control airway and COVID-19 lung autopsy samples 10 Table S8 . Blacklisted genes for a given cell type for the scRNAseq analysis. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . Bioinformatics (Oxford, England) 35, 4472-4473. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. . (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. IFNG IL10 IL13 IL15 IL17A IL2 IL4 IL5 IL6 IL7 CXCL8 IL9 CXCL10 CCL2 CCL3 CCL4 PRF1 TNFRSF9 TGFB1 TNF C CSF2 GZMB IFNG IL10 IL13 IL15 IL17A IL2 IL4 IL5 IL6 IL7 CXCL8 IL9 CXCL10 CCL2 CCL3 CCL4 PRF1 TNFRSF9 TGFB1 TNF Prf TGF-β1 TNF-β log 10 (x+1) pg/ml All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted October 19, 2020. All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. Diverse macrophage populations 2 mediate acute lung inflammation and resolution Sustained protective immunity against Bordetella pertussis nasal colonization by intranasal 7 immunization with a vaccine-adjuvant combination that induces IL-17-secreting TRM cells Flow Cytometry Reveals Similarities Between Lung Macrophages in Humans and Mice Near-optimal probabilistic RNA-16 seq quantification Regulation of scavenger receptor CD163 expression in human monocytes and macrophages by 20 pro-and antiinflammatory stimuli Tissue-Resident T Cells and Other Resident 1 Deep immune profiling of COVID-19 patients reveals distinct immunotypes with therapeutic implications Adaptation and Validation of a Pediatric 8 Sequential Organ Failure Assessment Score and Evaluation of the Sepsis-3 Definitions in 9 Critically Ill Children UMAP: Uniform Manifold Approximation and Projection for 12 Dimension Reduction UMAP: Uniform Manifold 15 Approximation and Projection Modular and efficient pre-processing of single-cell RNA-seq. bioRxiv The barcode, UMI, set format and BUStools Sex differences in immune responses that underlie 11 COVID-19 disease outcomes Cutting edge: tissue-retentive lung memory CD4 T cells mediate optimal protection to 15 respiratory virus infection Endothelial cells are central orchestrators 19 of cytokine amplification during influenza virus infection Myeloid cells in sepsis-acquired 2 immunodeficiency SARS-CoV-2-triggered 6 neutrophil extracellular traps mediate COVID-19 pathology Tissue-Resident Memory T Cells Mediate Immune Homeostasis in the Human Pancreas through the PD-1/PD-L1 Pathway Antibody responses to SARS-CoV2 are distinct in 16 children with MIS-C compared to adults with COVID-19 Comprehensive analysis of B cell compartments across the human 20 body reveals novel subsets and a gut resident memory phenotype