key: cord-1041995-4et7ucwu
authors: Kalocsay, Marian; Maliga, Zoltan; Nirmal, Ajit J.; Eisert, Robyn J.; Bradshaw, Gary A.; Solomon, Isaac H; Chen, Yu-An; Pelletier, Roxanne J.; Jacobson, Connor A.; Mintseris, Julian; Padera, Robert F.; Martinot, Amanda J.; Barouch, Dan H.; Santagata, Sandro; Sorger, Peter K.
title: Multiplexed proteomics and imaging of resolving and lethal SARS-CoV-2 infection in the lung
date: 2020-10-15
journal: bioRxiv
DOI: 10.1101/2020.10.14.339952
sha: 0e20a7f06ff1b4bf4a11a5286ccd059c0b9cc9b8
doc_id: 1041995
cord_uid: 4et7ucwu

Normal tissue physiology and repair depends on communication with the immune system. Understanding this communication at the molecular level in intact tissue requires new methods. The consequences of SARS-CoV-2 infection, which can result in acute respiratory distress, thrombosis and death, has been studied primarily in accessible liquid specimens such as blood, sputum and bronchoalveolar lavage, all of which are peripheral to the primary site of infection in the lung. Here, we describe the combined use of multiplexed deep proteomics with multiplexed imaging to profile infection and its sequelae directly in fixed lung tissue specimens obtained from necropsy of infected animals and autopsy of human decedents. We characterize multiple steps in disease response from cytokine accumulation and protein phosphorylation to activation of receptors, changes in signaling pathways, and crosslinking of fibrin to form clots. Our data reveal significant differences between naturally resolving SARS-CoV-2 infection in rhesus macaques and lethal COVID-19 in humans. The approach we describe is broadly applicable to other tissues and diseases. Summary Proteomics of infected tissue reveals differences in inflammatory and thrombotic responses between resolving and lethal COVID-19.

The great majority of tissue specimens acquired via biopsy, necropsy and autopsy in research and clinical settings are subjected to formalin-fixation and paraffin-embedding (FFPE). Molecular analysis of FFPE specimens is challenging but the use of fixed tissue has many advantages: FFPE samples are routine in diagnostic pathology and many archival samples are available, fixation better preserves tissue architecture than freezing, fixed tissues can be stored and shared easily, and use of thin sections (typically 5µm thick) conserves precious samples. In the context of COVID-19, fixation has the additional advantage that formaldehyde inactivates virus. Discovery proteomics of FFPE tissue is particularly difficult 1 due to chemical modification of peptides and inhibition of sample proteolysis by crosslinking. A decade ago it was demonstrated that performing mass spectrometry (MS) on FFPE specimens is feasible 2 and more advanced MS has recently been used to profile unfixed 3 

tumor banks 4 and also to assay sub-regions of a single specimen 5 .

host proteases such as transmembrane serine protease 2 (TMPRSS2) and many of these cells are found in the respiratory tract (e.g. type II pneumocytes) 6 . Infection can cause uncontrolled activation of the immune system resulting in severe pulmonary injury and even death 7 , particularly in individuals with co-morbidities such as obesity, diabetes and hypertension. Study of the underlying molecular mechanisms has focused on bulk and single-cell RNA sequencing of relatively easy to access specimens such as blood, throat swabs and bronchoalveolar lavage 8 and proteomics of serum 9 and cell lines 10, 11 . In humans, these data reveal strong but aberrant induction of interferon and inflammatory signaling 8, 12, 13 , upregulation of complement cascades 14, 15 , and activation of platelets 16 .

In this paper we analyze detailed molecular and cell-level data on signaling pathways activated by SARS-CoV-2 infection using lung tissue obtained via necropsy of infected non-human primates (NHPs; rhesus macaques) 17 in which infection naturally resolves and autopsy of humans who died from COVID-19. Acquiring these data required development of a proteomic method for protein extraction from multiple FFPE samples in a time series 18, 19 combined with multiplexed imaging 20 of the same specimen. We accomplished this by developing a denaturing protein extraction protocol for FFPE samples involving thermal reversion of formaldehyde crosslinks. The extraction proteins were analyzed using stable-isotope tandem mass tag (TMT) MS in which individual sample digests are labelled with TMT, mixed together and then analyzed in a single MS run 21 . Key advantages of this approach are high accuracy of quantification (10% differences in protein abundance can be reliably detected) 18 with few missing values across samples 22 . In the current work, capture of phospho-peptides on Fe-NTA 23 following sample digestion and TMT labeling enabled quantification of protein phosphorylation. In human specimens, viral load could be measured by MS based on peptides derived from SARS-CoV-2 spike (S) and nucleocapsid (N) proteins; these peptides were not detected in rhesus macaque specimens, which contained substantially less virus. However, viral burden could be measured in animals using multiplexed cyclic immunofluorescence (CyCIF) imaging 20 .

The combined use of proteomics and imaging made it possible to characterize multiple steps in disease response from early cytokine accumulation and protein phosphorylation to activation of downstream receptor pathways, changes in cell state, and crosslinking of fibrin to form clots. Our data demonstrate interferon induction in NHPs that correlates positively with viral burden. In contrast, high viral burden with little interferon activation is possible in humans who died from COVID-19. Significant upregulation of cytokines, spatially widespread infiltration of macrophages and other immune cells, and thrombosis are observed in human decedents but are limited in animals to regions of lung injury. These and related findings exploit the strengths of multiplexed proteomics and imaging of intact tissue to provide a detailed molecular picture of naturally resolving SARS-CoV-2 infection and of lethal COVID-19.

We analyzed a cohort of SARS-CoV-2 infected adult NHPs, from which multiple specimens were collected 2, 4, or 14 days after infection as well as three animals re-challenged with virus 35 days after initial infection and necropsied 14 days later (Fig. 1a) . NHPs are widely used to study viruses and vaccines 17,24-26 and we and others have shown that SARS-CoV-2 infection results in robust viral replication in the upper and lower respiratory tracts followed by disease resolution within 7-14 days 17, 27 .

Multiple lung lobes were sampled from each animal and imaging showed that viral burden varied with days post infection and lung location 17 . This variation was particularly evident on day 2, at which point specimens with high and low viral loads were acquired from a single lung (Fig. 1b,c) . For analysis by quantitative MS, nine specimens from infected and three from re-challenged rhesus macaques (n=8 animals) along with three specimens from an uninfected control animal were TMT-labelled and analyzed as a single mixed sample. Key findings were confirmed by repeat analysis of six additional day 2 or 4 post-infection specimens from the same animals (Extended Data Fig. 1a ). Over ~6,700 proteins were quantified per specimen and both principal component analysis (PCA) and two-way hierarchical clustering of MS data demonstrated separation by day of infection and viral load (Fig. 1d ,e, Supplementary Tables 1,2). Interferon inducible genes (ISGs) co-clustered, as did proteins normally found in stoichiometric complexes such as subunits of the proteasome (Extended Data Fig. 1b,c) .

These data are consistent with retention of information on cellular pathways and macromolecular complexes during MS of FFPE samples.

We also analyzed 13 human FFPE lung specimens collected at autopsy from three SARS-CoV-2 infected and three uninfected decedents (Supplementary Table 3) 28 . A similar number of proteins was quantified as in NHPs (Extended Data Fig. 1d and Supplementary Table 4 ) and co-clustering of ISGs (e.g. MX1), fibrins and clotting factors (e.g. Factors V, XIIIb and plasmin) was observed (Fig. 1f ). The decedent with the highest viral levels of S and N viral proteins also had elevated levels of the angiotensin processing enzyme (CPA6), potentially reflecting binding of SARS-CoV-2 to angiotensinconverting enzyme-2 (ACE2) 29 . Reactome pathway analysis of rhesus macaque and human MS data Table 3 ). This complicates comparison of COVID-19 patients with each other and with controls. Nonetheless, in both monkey and human specimens, cluster analysis separated controls from SARS-CoV-2 infected specimens and pathway analysis revealed similar up and down-regulated pathways.

In NHPs, Gene Set Enrichment Analysis (GSEA; the term "gene sets" is used for simplicity but proteomic data were analyzed) revealed significant (FDR <0.05) upregulation of interferon type I and II (IFNα and IFNγ) pathways 2 and 4 days post infection compared to uninfected controls ( Fig. 2a and Extended Data Fig. 3a,b) . By day 14, when virus was undetectable 17 , IFN programs had largely returned to pre-infection levels. Activation of interferon pathways was confirmed by upregulation of multiple IFN-stimulated genes (ISGs) at days 2 and 4, with substantially stronger upregulation when viral load was high (compare NHP specimens 2a, b with 2c, d; Fig. 2b ,c). In humans, the same IFN pathways were upregulated by viral infection, as were many of the same ISG proteins (Fig. 2b , right panel). In day 14 and re-challenged NHPs, the levels of ISGs had largely returned to normal but PCA and GSEA revealed differences relative to control animals; these differences largely involved pathways also upregulated at days 2 and 4 post infection.

Multiple significant changes observed in proteomic data were confirmed by immunofluorescence (IF). In NHPs, imaging showed that MX1 levels peaked at day 2 ( Fig. 2d ) and MX1 levels measured by IF and MS correlated well across all specimens analyzed (Fig. 2e) . MX1 staining was observed in multiple types of immune cells, and punctate intracellular staining was consistent with assembly of MX1 into higher-order structures (Fig. 2f) 30 . MS also quantified changes in the phosphorylation of multiple proteins involved in IFN transduction and responses including IFIT3 31 , SCAMP2 32 , and IRF2BP1 (see Supplementary Table 5 for all quantified phosphosites). For IFIT3 and IRF2BP1, both protein and phospho-site levels increase in parallel and in other cases, SCAMP2 for example, the stoichiometry of phosphorylation increased in specimens with high viral burden; both patterns of modification reflect known regulation of IFN mediators (Fig. 2g ). By imaging, IFIT3 levels were elevated in day 2 and 4 specimens as in proteomic data and exhibited a diffuse cytoplasmic pattern in epithelial, endothelial and some immune cells (Fig. 2f) . In specimens from re-challenged animals, which are resistant to reinfection 17 , no significant activation of IFN pathways or induction of ISGs was observed. However, GSEA revealed enrichment for plasma and NK cells (Extended Data Fig. 3c ), which is consistent with an antigen-specific and other anti-viral responses 33, 34 . Overall these data demonstrate a classical IFNmediated response to viral infection in NHPs that is proportional to viral burden. In human specimens, GSEA also revealed significant enrichment for IFNα and γ responses (Fig. 2h ), but induction of ISGs was not obviously related to viral load (Fig. 2i,k) : for example, decedent 3 with the highest viral load exhibited minimal ISG induction at death. Although we cannot exclude the confounding effects of comorbidities in this small sample of decedents, our proteomic and imaging data are consistent with previous reports of an impaired or dysregulated IFN response in COVID-19 patients 8, 35 .

Additional, significantly upregulated pathways in human decedents and high-virus NHP specimens included the unfolded protein response and mTORC1 signaling (Fig. 2a, Extended Data Fig.   3d ). The former is consistent with the hypothesis that SARS virus infection induces proteostatic stress 36 and might be considered for therapeutic intervention 37 . mTORC1 activation has been linked to insulin resistance and induction of type II diabetes in hepatitis C virus infection 38 . A similar association has also been suggested for COVID-19: new-onset diabetes is observed with COVID-19 39 and diabetes is a known co-morbidity 40 . Our data are consistent with a mechanistic link between SARS-CoV-2 infection and the etiology of type II diabetes 41 .

Severe COVID-19 is associated with dramatic increases in cytokine levels 13 and cytokine storms are postulated to play an important role in disease pathology 42 . Measuring cytokines in situ in tissue has historically been difficult, but we quantified 35 cytokines in human and 13 in macaque samples by MS. In one or more human decedents we found more cytokines significantly upregulated (11; P < 0.05) or downregulated (5) than in NHPs (Fig. 3a ,b and Extended Data Fig. 4a ). Specimens from human patient 2 (H2) exhibited a particularly strong induction of multiple cytokines: IL-1β, CXCL1/5/10 and CCL19/20 were all significantly upregulated (P < 0.05) as were protein components of the NLRP3 pathway ( Fig. 3a,b) . The NLRP3 inflammasome is a key component of the innate immune system that regulates secretion of pro-inflammatory cytokines IL-1β and IL-18 in response to infection and cellular damage 43 . Upregulation of IL1RAP (Fig. 3a , bar plot), a co-receptor for the IL1 receptor type 1, suggests enhanced responsiveness to these cytokines. ssGSEA also revealed enrichment of pathways regulated by IL-10 and IL-1 along with TNFα signaling via the NF-κB pathway (Fig. 3c ). IL-6 and IL-10, which are known to play a role in severe COVID-19 44, 45 , were not quantified in our proteomic data, but upregulation of IL-6 and IL-10 response genes was evident by ssGSEA in two of three patients (Fig. 3d) . We surmise that both interleukins escaped detection in our MS runs but might be quantifiable using targeted MS. While dysregulation of NLRP3 46 and NF-κB 35 has previously been described in severe COVID-19 cases based on whole-blood transcriptomics and cytokine profiling, our data demonstrate that this occurs in infected lung tissue and is associated with changes in the levels of cytokines, receptors and downstream response proteins at the site of infection.

Immune cells shape and respond to the cytokine microenvironment. When ssGSEA was performed on human proteomics data, we observed enrichment for multiple myeloid cell lineages with Table 3 ). In NHPs, immune signatures were upregulated primarily at day 2 in virus-high samples ("2c-d", Extended Data Fig. 4c ) as was an M1-like macrophage signature (FDR <0.05, Fig. 3d, upper panel) . Imaging of serial sections confirmed recruitment of CD16 + CD68 + macrophages to regions of tissue damage (consolidated lung regions, Fig. 3f ,g) 17 consistent with a classic inflammatory phenotype. Together these data demonstrate strong activation of innate immune responses in both human and NHP specimens, with considerable diversity among human decedents.

In contrast to other respiratory viruses, SARS-CoV-2 infection is associated with activation of complement cascades 48 and microvascular thrombosis 49 . By ssGSEA, activation of both the classical and alternate complement pathways was detected in H1 and H2 (but not in H3); the lectin pathway was downregulated in all patients as compared to controls (Fig. 4a,e) . By MS, we quantified levels of fibrin Q256-K437 crosslinked peptides (Fig. 4b) , a major form of fibrin α-chain crosslinking previously characterized by MS using purified components 50 . Q256-K437 crosslinking was proportional to α-chain abundance (Fig. 4c) . These data demonstrate factor XIII transglutaminase activation, the endpoint of the blood clotting cascade. Imaging also confirmed the presence of thrombi within the lung microvasculature (Fig. 4d) . In H2 we observed upregulation of proteins involved in fibrin activation (e.g. prothrombin, thrombin, tissue factor, factor XI) as well as inhibition of clotting suggesting homeostatic resolution of thrombi. Control patient HU2 died of an acute myocardial infarction and exhibited significant enrichment for pathways related to activation of platelets, but neither to fibrin nor complement; the thrombotic signatures in HU2 were therefore distinct from those of COVID-19 decedents (Fig. 4e) .

Unexpectedly, in NHP samples, we observed significant downregulation (FDR <0.05) of complement, fibrin, and platelet-related signatures early in infection with recovery by day 14. Capillary thrombi were limited in extent and most apparent in day 2 specimens (Fig. 4f,g) . To better understand this phenomenon, we performed spatial transcriptomic analysis with the GeoMx DSP platform 51 on uninfected lung (22 regions) or infected lung with different levels of inflammation (16 high, 14 mid, and 11 low from the specimen in Fig. 3f ). Regions of lung with high viral load but lacking overt tissue damage exhibited activation of complement pathways, whereas consolidated regions had upregulation of platelet aggregation, fibrin related pathways and downregulation of fibrinolysis (Fig. 4h) . We therefore propose that in NHPs complement pathways and platelets are locally active during acute viral replication and consequent tissue consolidation and damage, as they are in humans. As disease resolves, homeostatic regulation decreases clotting and in NHPs this manifests as a general downregulation of signatures for complement, fibrin and platelets.

In NHPs, SARS-CoV-2 infection is characterized by acute IFN induction proportional to viral burden as well as local induction of thrombotic programs. As virus is cleared over a two-week period, lung physiology returns to normal 17 . In contrast, severe COVID-19 in humans causes IFN pathway dysregulation, accumulation of immune cells 8, 13, 52, 53 , microvascular injury, and thrombosis 54 and we detect these pathologies by tissue proteomics and imaging. Significant increases in cytokine levels are evident in some human decedents but IFN induction is not in proportion to viral burden, with the decedent whose viral burden was highest exhibiting the weakest IFN response. The immune infiltrates, including macrophage activation states, also differ among decedents. Strong upregulation of pathways involving complement, fibrin processing and platelet activation is associated with extensive fibrin crosslinking and formation of thrombi. Because SARS-CoV-2 responses in NHPs are similar to those previously described for other respiratory viruses 55 , we propose that observed differences between infected NHPs and humans primarily arise from differences between naturally resolving infection and end-stage disease in the presence of co-morbidities.

Parallel analysis of fixed lung tissue by multiplexed mass spectrometry is a powerful counterpart to more common transcriptional and cytokine profiling of bronchoalveolar lavage and blood specimens in characterizing infection. The data in this paper highlight the ability of proteomics, imaging, and spatial transcriptomics of the same specimens to trace successive steps in diverse regulatory cascades. We quantify changes in cytokine levels, cognate receptors, signaling proteins, and downstream consequences of enzyme activity (e.g. phosphorylation, fibrin crosslinking), cell state and function. To our knowledge, no other approach involving readily available histological specimens provides a similarly integrated picture of a disease process; moreover, it is applicable to any fixed tissue and thus, to a wide variety of normal and abnormal physiologies. between -1 to 1; dendrogram represents hierarchical clustering using Euclidean distance. Reactome is abbreviated to RO in signature names. f, Analysis for NHP samples as in a. g, Carstairs' stain of capillary thromboses in day 2 NHP specimen (2c); arrow denotes fibrin deposition along hyaline membrane. h, ssGSEA of spatial RNA transcription data (GeoMx) data from uninfected NHP animals and a virus-high day2 specimen. Morphological regions R1-R3 are shown in Fig. 2f .

Tissue from de-waxed formalin-fixed paraffin-embedded (FFPE) samples was scraped off slides using a fresh razor blade for each slide. Visible major blood vessels were avoided. The samples were transferred to individual PCR tubes with 100µL lysis buffer (50mM Tris pH8.5, 2% SDS, 150mM NaCl). Samples were transferred to clean 1.5mL microcentrifuge tubes and precipitated with a 1:1 volume ratio of 55% ice-cold trichloroacetic acid (TCA) during incubation on ice for 15 minutes. Samples were then centrifuged at ~20,000 x g at 4 o C for 10 minutes and the supernatant was aspirated. The pellets were washed with cold acetone (-20 o C), which was immediately aspirated, and washed three more times with -20 o C cold acetone followed by centrifugation at ~20,000 x g at 4 o C for 10 minutes. digests were centrifuged at ~20,000 x g for 5 minutes and supernatants were transferred to clean microcentrifuge tubes. Acetonitrile (ACN) was added to each sample to a concentration of 30% and peptides were labeled with TMTpro reagents (Thermo Fisher Scientific, A44520) for 1 hour at room temperature with brief mixing by vortexing every 10 minutes. After labeling, a ratio check was performed by pooling 2.5µL of each reaction and label incorporation (>95%) and ratios were determined by mass spectrometry. The reactions were then quenched with a final concentration of 0.5% hydroxylamine for 15 minutes, samples mixed into a single 16-plex, ratios determined by MS analysis again and the sample acidified with formic acid (FA) to a final concentration of 2% v/v. Labeled samples were pooled to obtain on average a 1:1 ratio of TMT-labeled peptides The peptide pool was dried by SpeedVac, reconstituted in 1% FA with 0.1% trifluoroacetic acid (TFA) and desalted using a Sep-Pak C18 cartridge (200mg, Waters, WAT054925) to remove salt. The cartridge was pre-conditioned with 1% FA with 0.1% TFA and the sample was slowly loaded by gravity flow followed by 2 washes with aqueous 1% FA.

Peptides were eluted with 85% ACN with 15% aqueous 1% FA.

Once dried by SpeedVac, samples were phospho-enriched using High-Select™ Fe-NTA Phosphopeptide Enrichment cartridges (Thermo, A32992). The flow-through was retained for further analysis. The phospho-enriched sample was desalted by C18 Empore Extraction Disks (Fisher Scientific, 13-110-019) solid-phase exchange (Stage Tip) and dried. The final phospho-enriched peptide sample was reconstituted in aqueous 1% FA without ACN for MS analysis.

The retained flow-through was purified by Sep-Pak as above again with binding to the column in 1% FA without TFA and an intermediate washing step of 1% FA with 5% acetonitrile to remove unincorporated TMTpro labels. Alkaline reversed-phase peptide fractionation was performed on an Agilent 1200 system using a flow rate of 600µL/minute and 18-45% Buffer B over 75 minutes. (Buffer A is 10mM ammonium bicarbonate pH 8 with 5% ACN, Buffer B is 10% 10mM ammonium bicarbonate pH 8 with 90% ACN). A total of 96 fractions was collected and combined into 24 samples, which were dried, desalted by stage tipping, and dried. The fractions were reconstituted in aqueous 1% FA with 3% ACN prior to MS analysis.

All samples were analyzed using an Orbitrap Lumos or Eclipse Tribrid MS with an EASY-nLC 1200 system (Thermo Fisher Scientific) with a capillary column (length: 30cm, inner diameter: 75mm) packed with 2.6µm Accucore C18 resin and heated to 60 o C. Phospho-enriched peptides were analyzed with multi-stage activation to account for phospho neutral loss (-97.977/z) and gradients ranging from 3- The Orbitrap resolution was set to 50,000 (dimensionless units) and the maximum injection time varied with up to 650ms. Further details on TMTpro multiplexing, LC and MS parameters and settings were identical to those recently described. 57 An in-house developed software suite was used to convert recorded .raw files to mzXML format, correct monoisotopic m/z measurements, and perform a post-search calibration. Peptide-spectrum matches used a Sequest (v.28, rev.12) based search engine. Searches were performed using a mass tolerance of 20ppm for precursors and a fragment ion tolerance of 0.9Da against databases with the Macaca mulatta reference proteome (UniProt) with additional Macaca mulatta proteins or a human reference proteome (2018, UniProt) and SARS-CoV-2 proteins (UniProt). Search databases contained common contaminant proteins and reversed sequences as decoys. Search results were filtered by linear discriminant analysis (LDA) against reversed peptide sequences with a false discovery rate (FDR) set to <1% and a protein FDR <1% for collapsed protein groups. Search parameters accounted for oxidized methionine (+15.9949Da) dynamically with TMTpro label (+304.2071Da) on peptide N-termini and lysine as static modification and in the case of phospho-peptide samples dynamic phospho-Ser/Thr/Tyr (+79.9663Da). A modified Ascore algorithm was applied to assess the confidence of phosphorylation-site position assignment. 58 Phosphorylation localized to individual sites required Ascore values >13 (p ≤0.05) for confident localization. Site localization score (Max Score) for phosphopeptides quantified is reported in Supplementary Table 5 . MS 3 was used to quantify relative TMTpro signal to noise (s/n), peptide quant data were filtered for an isolation specificity of greater than 70% and a sum s/n of greater than 200 over all sixteen channels for each peptide. Relative protein quantification was achieved by calculating the summed s/n for all peptides mapping to a given protein.

Channels were normalized for equal total protein content by summed s/n for all proteins in a given channel. Details of the TMT intensity quantification method and further search parameters applied were described previously. 59 Proteomics raw data and search results were deposited in the PRIDE archive 56 and can be accessed under ProteomeXchange accession numbers: PXD021453, PXD021456 and PXD021459 (note to reviewers: datasets will be publicly available when paper is published, per PRIDE policy. Data are available during the review phases at https://www.ebi.ac.uk/pride/login with Username: reviewer_pxd021453@ebi.ac.uk , Password:

LPCv6VLi ; Username: reviewer_pxd021456@ebi.ac.uk , Password: mJoqEZJU and Username: reviewer_pxd021459@ebi.ac.uk , Password: 1ruKkhd5 ).

To identify transglutaminase cross-linked fibrin peptides, all MS data were searched with the PIXL search engine, 60 specifying the Lys -Gln residue linkage alongside regular peptide search.

Fibrinogen and fibronectin proteins were included in cross-linked search, along with serum albumin, alpha-2-macroglobulin and alpha-2-antiplasmin -all proteins previously implicated with transglutaminase activity. 61 Precursor tolerance was set to 15ppm and fragment ion tolerance to 0.4Th. Due to the low number of cross-linked peptides in the sample, they were filtered to 1% FDR using LDA alongside regular peptides.

Relative summed s/n TMT data for 16-plex experiments were scaled (0-100%) for each protein and analyzed by principal component analysis (PCA) and 2-way hierarchical clustering (Ward) using the JMP Pro 15 software package. DESeq2 62 was also used for differential expression analysis. A corrected P-value cut-off of 0.05 was used to assess significant genes that were up-regulated or downregulated after application of the Benjamini-Hochberg procedure. 63

A compendium of biological and immunological signatures was identified from publicly available databases or published manuscripts for performing enrichment analysis. In order to perform gene set enrichment analysis, two previously published methods (Gene Set Enrichment Analysis (GSEA) 64 and Single-sample GSEA (ssGSEA) were primarily used. The R package clusterProfiler 65 was used to perform GSEA analysis and the R package GSVA 66 was used to perform ssGSEA analysis which calculates the degree to which genes in a particular gene set are coordinately up-or down-regulated within a sample. The interferon, complement, platelet, fibrin, IL1, IL1R, NF-KB, IL10, MTORC and insulin related signatures were curated from MSigDB 67 and immune cell related signatures were manually curated from published studies [68] [69] [70] .

Tissue-based Cyclic Immunofluorescence (t-CyCIF) was performed as previously described 20 with antibody validation as described 71 . The method involves iterative cycles of antibody incubation, imaging, and fluorophore inactivation. Briefly, FFPE specimens on glass slides were de-waxed and antigen retrieval was performed on a Leica Bond RX. Each section of tissue was stained and imaged with fluorescently conjugated antibodies or unconjugated primary and conjugated secondary antibody (Table S6) . Slides were scanned with a RareCyte CyteFinder scanner equipped with a 20x, 0.75NA objective. The raw images were corrected for uneven illumination using the BaSiC tool 72 and images from multiple rounds of scanning of the same tissue were stitched and aligned using ASHLAR. 73 . Cellnucleus pixel probability maps were first generated using a trained U-Net 74 model and then single cells were segmented using marker-controlled watershed. Mean fluorescence intensities of each marker for each cell were computed to determine cell types. In Fig. 2-4 selected CyCIF channels are shown

NanoString GeoMx was performed as previously described 75 at NanoString Inc. as part of their DSP Technology Access Program. 5μm sections from high-virus day 2 and uninfected NHP tissue blocks were used. Immunofluorescence with a DNA stain and antibodies against PANCK, CD45 were used to visualize tissue, with area of interest (AOI) guided by staining of SARS-N and DNA staining in an adjacent stained tissue section. 67 AOIs representing five morphological sites (uninfected/control (24) , airway (2), two virus positive regions (25) with differential levels of immune infiltration as well as a virus-negative region of consolidated region (16)) were extracted and assayed. The sequencing and probe quality-controlled data received directly from NanoString were then used for pathway analysis. 

Proteomic developments in the analysis of formalin-fixed tissue

Toward deciphering proteomes of formalin-fixed paraffin-embedded (FFPE) tissues

System-wide Clinical Proteomics of Breast Cancer Reveals Global Remodeling of Tissue Homeostasis

Multi-level Proteomics Identifies CT45 as a Chemosensitivity Mediator and Immunotherapy Target in Ovarian Cancer

A streamlined mass spectrometry-based proteomics workflow for large-scale FFPE tissue analysis

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

Acute Respiratory Distress Syndrome

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

Proteomic and Metabolomic Characterization of COVID-19

A SARS-CoV-2 protein interaction map reveals targets for drug repurposing

The Global Phosphorylation Landscape of SARS-CoV-2 Infection

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Clinical and immunological features of severe and moderate coronavirus disease 2019

Single-cell landscape of bronchoalveolar immune cells in patients with COVID-19

Pathological inflammation in patients with COVID-19: a key role for monocytes and macrophages

Endotheliopathy in COVID-19-associated coagulopathy: evidence from a singlecentre, cross-sectional study

SARS-CoV-2 infection protects against rechallenge in rhesus macaques

A Review on Quantitative Multiplexed Proteomics. Chembiochem Eur

TMTpro: Design, Synthesis, and Initial Evaluation of a Proline-Based Isobaric 16-Plex Tandem Mass Tag Reagent Set

Highly multiplexed immunofluorescence imaging of human tissues and tumors using t-CyCIF and conventional optical microscopes

MultiNotch MS3 Enables Accurate, Sensitive, and Multiplexed Detection of Differential Expression across Cancer Cell Line Proteomes

Review, evaluation, and discussion of the challenges of missing value imputation for mass spectrometry-based label-free global proteomics

SL-TMT: A Streamlined Protocol for Quantitative (Phospho)proteome Profiling using TMT-SPS-MS3

Comparative pathology of rhesus macaque and common marmoset animal models with Middle East respiratory syndrome coronavirus

Clinical benefit of remdesivir in rhesus macaques infected with SARS-CoV-2

DNA vaccine protection against SARS-CoV-2 in rhesus macaques

Respiratory disease in rhesus macaques inoculated with SARS-CoV-2

In situ detection of SARS-CoV-2 in lungs and airways of patients with COVID-19

Characterization of Carboxypeptidase A6, an Extracellular Matrix Peptidase

Human MxA protein: an interferon-induced dynamin-like GTPase with broad antiviral activity

IFN-Induced TPR Protein IFIT3 Potentiates Antiviral Signaling by Bridging MAVS and TBK1

Function of the t-SNARE SNAP-23 and secretory carrier membrane proteins (SCAMPs) in exocytosis in mast cells

Human natural killer cells mediate adaptive immunity to viral antigens

Antigen-specific NK cell memory in rhesus macaques

Impaired type I interferon activity and inflammatory responses in severe COVID-19 patients

Coronavirus Infection Modulates the Unfolded Protein Response and Mediates Sustained Translational Repression

Pathological Aspects of COVID-19 as a Conformational Disease and the Use of Pharmacological Chaperones as a Potential Therapeutic Strategy

Hepatitis C Virus Activates the mTOR/S6K1 Signaling Pathway in Inhibiting IRS-1 Function for Insulin Resistance

New-Onset Diabetes in Covid-19

Diabetes in COVID-19: Prevalence, pathophysiology, prognosis and practical considerations

Practical recommendations for the management of diabetes in patients with COVID-19

SARS-CoV-2: a storm is raging

The NLRP3 Inflammasome: An Overview of Mechanisms of Activation and Regulation

Reduction and Functional Exhaustion of T Cells in Patients With Coronavirus Disease 2019 (COVID-19)

Characteristics of lymphocyte subsets and cytokines in peripheral blood of 123 hospitalized patients with 2019 novel coronavirus pneumonia (NCP)

Severe COVID-19: NLRP3 Inflammasome Dysregulated

Profiling tumor infiltrating immune cells with CIBERSORT

Genomic characterisation and epidemiology of 2019 novel coronavirus: implications for virus origins and receptor binding

Pulmonary post-mortem findings in a series of COVID-19 cases from northern Italy: a two-centre descriptive study

Identification of Respective Lysine Donor and Glutamine Acceptor Sites Involved in Factor XIIIa-catalyzed Fibrin α Chain Cross-linking

Spatial Analysis of RNA in FFPE Tissue

Heightened Innate Immune Responses in the Respiratory Tract of COVID-19

Dysregulation of immune response in patients with COVID-19 in Wuhan, China

Complement associated microvascular injury and thrombosis in the pathogenesis of severe COVID-19 infection: A report of five cases

The role of IFN in respiratory syncytial virus pathogenesis

The PRIDE database and related tools and resources in 2019: improving support for quantification data

TMTpro reagents: a set of isobaric labeling mass tags enables simultaneous proteomewide measurements across 16 samples

A probability-based approach for high-throughput protein phosphorylation analysis and site localization

Quantitative mass spectrometry-based multiplexing compares the abundance of 5000 S. cerevisiae proteins across 10 carbon sources

High-density chemical cross-linking for modeling protein interactions

Mass spectrometry-based molecular mapping of native FXIIIa cross-links in insoluble fibrin clots

Moderated estimation of fold change and dispersion for RNAseq data with DESeq2

Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing

Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles

clusterProfiler: an R Package for Comparing Biological Themes Among Gene Clusters

GSVA: gene set variation analysis for microarray and RNA-Seq data

Molecular signatures database (MSigDB) 3.0

Dissecting the multicellular ecosystem of metastatic melanoma by single-cell RNAseq

Immune Cell Gene Signatures for Profiling the Microenvironment of Solid Tumors

Robust enumeration of cell subsets from tissue expression profiles

Qualifying antibodies for image-based immune profiling and multiplexed tissue imaging

A BaSiC tool for background and shading correction of optical microscopy images

Highly multiplexed immunofluorescence images and single-cell data of immune markers in tonsil and lung cancer

Convolutional Networks for Biomedical Image Segmentation. in Medical Image Computing and Computer-Assisted Intervention -MICCAI 2015

Enabling pathway analysis of RNA expression in formalin-fixed paraffin embedded tissues with the GeoMx DSP Platform

We thank David Weinstock, Raquel Arias-Camison, Linda Wrijil, Alyce Chen, and the Gygi Lab for scientific advice and technical assistance and NanoString Inc. for GeoMx DSP data acquisition via its Technology Access Program. Funding: This work was funded by NCI grants U54-CA225088, R35-