key: cord-0254860-hmoiqclt
authors: Leon, Juliette; Michelson, Daniel A.; Olejnik, Judith; Chowdhary, Kaitavjeet; Oh, Hyung Suk; Hume, Adam J.; Galván-Peña, Silvia; Zhu, Yangyang; Chen, Felicia; Vijaykumar, Brinda; Yang, Liang; Crestani, Elena; Yonker, Lael M.; Knipe, David M.; Mühlberger, Elke; Benoist, Christophe
title: A virus-specific monocyte inflammatory phenotype is induced by SARS-CoV2 at the immune-epithelial interface
date: 2021-09-29
journal: bioRxiv
DOI: 10.1101/2021.09.29.462202
sha: 6235df431c45989b44a6d8504ba24c106e0fa8c9
doc_id: 254860
cord_uid: hmoiqclt

Infection by SARS-CoV2 provokes a potentially fatal pneumonia with multiorgan failure, and high systemic inflammation. To gain mechanistic insight and ferret out the root of this immune dysregulation, we modeled by in vitro co-culture the interactions between infected epithelial cells and immunocytes. A strong response was induced in monocytes and B cells, with a SARS-CoV2-specific inflammatory gene cluster distinct from that seen in influenza-A or Ebola virus-infected co-cultures, and which reproduced deviations reported in blood or lung myeloid cells from COVID-19 patients. A substantial fraction of the effect could be reproduced after individual transfection of several SARS-CoV2 proteins (Spike and some non-structural proteins), mediated by soluble factors, but not via transcriptional induction. This response was greatly muted in monocytes from healthy children, perhaps a clue to the age-dependency of COVID-19. These results suggest that the inflammatory malfunction in COVID-19 is rooted in the earliest perturbations that SARS-CoV2 induces in epithelia.

Viral infections induce varied innate and inflammatory responses in the host. These responses help to control the viruses, but in some cases can become far more deleterious than the virus itself (1) . Infection with SARS-CoV2 (CoV2), the cause of the current COVID-19 pandemic, leads to an upper respiratory tract infection which, if not controlled by the innate and adaptive immune responses, can evolve into a lethal pneumonia. CoV2 infection is remarkable in its clinical heterogeneity, ranging from asymptomatic to fatal (2) , and several clinical characteristics demarcate the pathology associated with CoV2, when compared with other respiratory pathogens such as influenza A virus (IAV). First, critical COVID-19 is associated with multi-organ failure beyond the lungs and a concomitant severe vasculopathy (3) (4) (5) . Second, bacterial coinfection, a common complication in IAV infections (6, 7) , is rarely found in COVID-19, yet COVID-19 nonetheless adopts clinical aspects of bacterial sepsis (8) , with an over-effusive production of inflammatory cytokines (reviewed in (9) ). Finally, an important feature of COVID-19 is that children are usually spared from severe disease, showing asymptomatic or milder disease at the acute phase (10) (11) (12) (13) , even though viral loads are similar to adults (14) . Such an age imbalance is not seen in IAV infections.

Many studies have aimed to understand the molecular and immunological factors that drive these clinical phenotypes (15) (16) (17) (18) (19) . In severe COVID-19, profound alterations of the immune system have been described in myeloid cells (20, 21) , along with impaired interferon (IFN) responses (22) (23) (24) , impaired T cell functions (25) (26) (27) (28) , production of autoantibodies (29) and high circulating levels of inflammatory cytokines (17, 24, 30) . It is not obvious how to disentangle which of these manifestations causally partake in severe pathogenesis, and which are only bystander markers of the strong inflammation. Direct pathogenicity from virus-induced damage is unlikely to be a driver, as high viral loads can exist early in asymptomatic or mild disease (31, 32), pointing to a determining role of host factors. Abnormalities in the type I IFN pathway, resulting from genetic alterations (33) or from IFN-neutralizing autoantibodies (34-37), clearly have a causative or amplifying role in COVID-19, plausibly by allowing the virus to replicate unchecked during the early phases of infection, before adaptive immune defenses can be recruited. However, the response to CoV2 involves many cellular and molecular players, and it seems likely that additional pathways beyond type I IFN underlie both resistance and pathology. More generally, the question can be framed as understanding why the newly emerging coronaviruses, including MERS and SARS-CoV1, are so pathogenic, while others that have co-evolved with humans are not. A plausible virologic explanation is that their molecular structures are mostly novel to human immune systems, as the H1N1 IAV variant was during the 1918 influenza pandemic, such that toxicity derives from immunologic novelty. Another hypothesis, not mutually exclusive, is that these highly pathogenic coronaviruses are equipped to perturb immune responses, perturbations which in turn drive severe immunopathology. Coronaviruses have large genomes, encoding many non-structural proteins, some of which are thought to have immune-modulating capabilities (38) (39) (40) . They thus have the genetic leeway to evolve such strategies, their attempts at immune evasion potentially promoting particularly deleterious immunopathology.

To better understand the root factors leading to immune dysregulation in COVID-19, we designed an in vitro co-culture system in which immunocytes were exposed to epithelial cells infected with CoV2, then profiled by transcriptomics and flow cytometry. Epithelial CoV2 infection induced a strong, mixed inflammatory response in co-cultured monocytes resembling that of blood monocytes from COVID-19 patients. A large component of this response was not observed with two severe human pathogens used as comparators, IAV and Ebola virus (EBOV), and this response was strikingly muted in monocytes from children. Together, these results suggest that CoV2-infected epithelial cells elicit an early and specific pro-inflammatory response in monocytes, which may explain the severity of COVID-19.

To assess whether and how immunocytes are triggered by CoV2-infected cells, we established a co-culture model in which ex vivo blood immunocytes were placed in direct contact with virus-infected epithelial cells (Fig. 1A) . Because primary lung epithelial cells are difficult to expand and manipulate in such conditions, we chose as a surrogate epithelial cell, the human colorectal adenocarcinoma cell line Caco-2, which is permissive for CoV2 infection (41) and DNA transfection. Under our infection conditions, CoV2 nucleocapsid (N) expression was detected in approximately 50% of Caco-2 cells by flow cytometry and immunofluorescence (Fig. 1B) . Thirtyfive hours after CoV2 infection of the Caco-2 monolayer, unbound virus was removed and peripheral blood mononuclear cells (PBMC) from healthy donors (HD) were added to the cultures.

These were harvested 14 h later, and subpopulations were magnetically purified for transcriptome profiling by RNAseq (Fig. S1A ). In good part because of the experimental requirements of BSL4 biocontainment (e.g. lysates had to be heat-treated for biosafety), RNAseq data quality was lower than customary. Rather than the usual statistical tests, identification of differentially expressed genes relied on the convergence of two independent experiments, a third experiment being used for validation (see Methods).

We focused the analysis on CD14 + monocytes and B cells, which show perturbed transcriptomes in COVID-19 patients, and are both frontline sensors of infection. In purified monocytes, a robust response was observed, with at least 675 differentially expressed genes (DEG) (Fig. 1C , which displays transcripts of the reproducible DEGs, hereafter "CoV2 signature" - Table  S1 ). Immediately apparent were the induction of major pro-inflammatory cytokines and chemokines (IL1B, IL6, TNF, CCL3 and CCL4) and a substantial number of antiviral IFN stimulated genes (ISG; e.g. IFIT5, ISG20). Conversely, MHC class-II genes were significantly downregulated. Closer examination of cytokine-and chemokine-encoding genes revealed IL10 as the most induced cytokine transcript, along with the main pro-inflammatory trio (IL6, IL1B, TNF; 1D ). As analyzed further below, several of these traits evoked transcriptional changes in immunocytes of COVID-19 patients (15, 19) . Gene ontology analysis of these genesets (Fig. 1E , Table S2 ) revealed a complex set of functions: cytokines, innate signaling pathways, cell mobility and adherence, and antigen presentation, suggesting that exposure to CoV2-infected cells induces profound changes in monocyte physiology. In contrast, the direct transcriptomic effect of CoV2 in Caco-2 cells was very mild (see below) with none of the changes detected in monocytes.

Analysis of B cells from the same cultures also displayed numerous changes in this setting (Fig. S1B ). This response partially coincided with that of monocytes, but also included some components preferential or unique to either cell-type (Fig. 1F ). Some ISGs were induced in both, although induction of the antiviral response was strongest in monocytes (Fig. 1G) . Surprisingly, the cytokines and chemokines most strongly induced in monocytes were also induced in B cells (Fig. 1H ). Thus, the effects of CoV2 infection on neighboring cells were apparent in several celltypes.

Having observed a mixed inflammatory response to CoV2-infected epithelial cells in cocultured monocytes, we next asked whether it was specific to CoV2, by comparing monocyte responses to epithelial cells infected with either CoV2, IAV (another clinically significant respiratory pathogen of the orthomyxovirus family) or EBOV (a more distant non-respiratory virus of the filovirus family, with a highly lethal hemorrhagic course also associated with strong inflammation; Fig. 2A ). Epithelial infection levels were comparable between the three viruses (ranging from 30 to 80% in different IAV experiments, and ~80% for EBOV; Fig. S2A , S2B).

IAV and EBOV both induced sizeable numbers of DEGs in co-cultured monocytes (Fig. 2B), both viruses having roughly 50% stronger effects overall than CoV2). As for CoV2, the response to IAV infection in co-cultured B cells and monocytes was very similar (Fig. S2C ). Direct comparison of monocyte transcriptional changes induced by CoV2 and IAV revealed that most downregulated genes were shared between the two infections, while the upregulated genes consisted of both shared and virus-specific modules (Fig. 2C-F) . The CoV2-specific component included several of the pro-inflammatory cytokines described above, especially TNF and IL10;

IL1B was even downregulated in IAV-infected co-cultures (Fig. 2C ). On the other hand, IL6 and the granulocyte/monocyte stimulating factors CSF2 and CSF3 were equally induced by IAV and CoV2. A set of pro-inflammatory chemokines (CCL3, CCL4 and CCL19) were also upregulated preferentially by CoV2 infection (Fig. 2D ). The eosinophil chemotactic factor CCL24 was among the most strongly downregulated genes by both IAV and CoV2, suggesting that eosinophil recruitment is dampened in both infections. A substantial set of ISGs were induced at similar magnitudes by both viruses, but some ISGs also responded preferentially in the presence of CoV2 As discussed above, COVID-19 symptomatology includes several of the manifestations of sepsis, even in the absence of bacterial infection or obvious barrier breach. Furthermore, gene ontology analysis suggested that the CoV2 co-culture signature harbored elements of innate activation through Toll-like receptor (TLR) 4 activation (Fig. 1E ). To test this notion, PBMCs were incubated in parallel cultures with E. coli lipo-polysaccharide (LPS), a TLR2/4 ligand. The transcriptional signature of genes induced or repressed by LPS in monocytes super-imposed strongly with CoV2-imparted changes (Fig. 2F) . The LPS downregulated geneset was largely common to CoV2 and IAV infections, while the upregulated component of the LPS response was much more strongly influenced in CoV2-than in IAV-infected co-cultures (median FoldChange (FC) = 1.37 vs 0.95, chisq pval =< 0.0001 vs 0.24, respectively). Note that this intersection between the CoV2 co-culture signature and TLR activation was not merely due to dead epithelial cells released in the culture: the transcriptional changes elicited in the monocytes by exposure to lysed HEK cells (killed by freeze-thawing) bore no relation to effects of CoV2 or IAV-infected cells (Fig. S2D ).

Calculating an index for responsive genes confirmed that, across monocytes from seven different HDs, the CoV2-down signature was equally elicited in CoV2 and IAV co-cultures, but that the up signature was very specific to CoV2 (Fig. 2G ). To exclude that this ineffectiveness of IAV-infected Caco-2 cells to induce the full ISG set was due to suboptimal infection, we profiled monocytes in co-cultures with Caco-2 cells infected with a wide range of IAV multiplicity of infection (MOI). The CoV2-down signature was indeed most marked at an intermediate range ( Fig. S2E ), but the CoV2-up signature could not be significantly induced at any MOI (in contrast, ISG induction was essentially linear to infection dose).

The comparison of monocytes in EBOV-and CoV2-infected co-cultures largely reproduced the same themes (Fig. 2H ): some degree of shared effects, particularly for downregulated genes (quantitatively stronger for CoV2), comparable induction of some antiviral ISGs, but a preponderance of virus-specific inductions. As in the IAV comparison, the key inflammatory cytokines and chemokines TNF, IL1B, CCL3 and IL10 were uniquely induced by CoV2 (and even repressed by EBOV). In the EBOV co-cultures, IL6 transcripts were below the detection threshold. The LPS-induced signature showed branching into EBOV and CoV2 preferential induction, the latter being actually repressed in the EBOV co-cultures (Fig. 2H) .

Overall, these results are recapitulated in the heatmap of Fig. 2I , Table S4 , which also highlight the dichotomy between the two ISG-containing clusters (K2 and K4), only one of which was induced in all viral co-cultures (K4), while the other is highly specific to CoV2 co-cultures (K2; In sum, co-culture with CoV2-infected epithelial cells induces a complex response in monocytes, some of which is generic to all virus-infected cells, but most of which is quite specific to CoV2, in particular the pro-inflammatory moiety.

Given these specific effects of cells infected by CoV2, we then attempted to determine which viral proteins might be involved. As a screen, Caco-2 cells were transfected, in biological duplicates, with a panel of 27 plasmids encoding single viral proteins or GFP as a control (Fig.   3A ). Forty-eight hour later, these transfectants were co-cultured with HD PBMCs, and the monocytes were profiled by RNAseq after 14 h. Such transient transfections can be prone to technical artefacts from cell stress during transfection, plasmid DNA, or protein over-expression (42, 43) . Indeed, the RNAseq data were noisy, with substantial variation between biological replicates. We thus selected a set of differentially expressed genes whose overall variance in the dataset substantially exceeded inter-replicate variance, and with significant difference from GFPtransfected controls in at least one co-culture. We then cross-referenced these genes to transcripts of the CoV2-induced signature. Although some genes with variable expression in co-cultured monocytes showed no reproducible relation to effects in virus-infected co-cultures, two groups of genes (G2 and G6 in Fig. 3B ) had very strong overlap with the CoV2-up and CoV2-down genesets.

Several CoV2 proteins were able to upregulate G2 and downregulate G6 in the co-cultured monocytes (most clearly S, nsp5, 9 and 14) , while others had a moderate repressive effect (N, nsp12, orf9c). No notable level of cell death was induced by any of these plasmids, and the differential effects were reproducible in parallel experiments with independent plasmid preparations (not shown). Genes in G2 mostly corresponded to the pro-inflammatory and CoV2specific clusters K5 and K6 defined in Fig. 2 . On the other hand, the ISG component of the CoV2up signature did not belong to this group, but in a cluster with poor reproducibility and with little or no specific effects of S and nsp5 (Fig. S3A) . Thus, there was a disconnect between the proinflammatory and the ISG moieties of the CoV2-up signature: CoV2 proteins reproduced the inflammatory but not the ISG part.

Reciprocally, genes from G2 and G6 identified in the transfection co-cultures proved almost entirely shifted in virus-infected co-cultures (Fig. 3C ). Here again, most upregulated genes were not shared with IAV-infected co-cultures, but the downregulated signature was common to both ( Fig. 3C ).

For replication, we performed monocyte co-cultures with transfection into a second epithelial cell line (HEK). Comparable patterns were observed (Fig. S3B) , with dominant effects of S, nsp9 and nsp14, but opposite effects of orf8,9,10, which matched genes altered in both CoV2-infected and transfected Caco-2 co-cultures (Fig. S3B, top) . The effects of individual CoV2 proteins showed generally concordant distribution after transfection in both cell lines (Fig. 3D ).

Thus, it was possible to replicate some of the monocyte response to CoV2-infected cells by expression of single viral proteins, confirming that the observed signatures were not merely confounders of infected co-cultures, or induced by free viral RNA. Several proteins shared the same potential, implying that changes in monocytes were not due to viral proteins acting as specific triggers, but more likely through changes that they induced in the infected epithelial cells.

Active proteins settled into two groups, with diametrically opposite effects, which would presumably be balanced in the context of viral infection, but overall the virus best matched the S/nsp5/nsp14 group.

We then attempted to tackle the mechanistic pathway though which CoV2-infected or transduced Caco-2 cells elicit the CoV2 signature in healthy monocytes. We searched for candidate mediators by examining RNAseq profiles of CoV2-infected Caco-2 cells in our cultures. Few or no genes showed significant induction, except for viral proteins themselves (Fig. 4A ). In an attempt to bring out minor effects, we aligned the results of two independent culture experiments (each in biological duplicate), and observed no enrichment in the concordant segment of the graph (Fig.   4B ), suggesting that most of these low-significance signals were indeed noise. The few putatively reproducible changes in Fig. 4B did not show any bias in a previously published dataset of CoV2infected Caco-2 cells (44) (Fig. 4C) . Thus, in agreement with these authors, we conclude that CoV2 infection has surprisingly minor transcriptional effects in infected Caco-2 cells. We next generated RNAseq profiles from Caco-2 cells transfected with 27 individual viral genes, and searched for transcripts that would correlate, across all the transfectants, with the ability to induce the specific signature in co-cultured monocytes. Very few transcripts showed significant correlation, with a distribution of correlation coefficient similar to that observed with random label permutation (Fig. 4D ) and with no relationship between the most correlated transcripts and those putatively affected by CoV2 virus infection (Fig. 4E ). We concluded that CoV2 and its proteins were inducing the activating potential in Caco-2 cells via non-transcriptional means.

To determine if these CoV2-related effects were mediated by cell-to-cell contact or via soluble factor(s), we used a Transwell chamber to co-culture monocytes and HEK cells transfected with a selected set of viral genes ( patients (19) , whose contact with infected epithelia would most closely mimic our experimental configuration, gene expression signatures of alveolar macrophages from severe patients proved upregulated in our CoV2-co-cultured monocytes (Fig. 5B , left, chisq p<10 -4 ); some aligned with the swath equally affected in IAV-and CoV2-infected co-cultures (including ISGs like IFI27, ISG15), but the largest group belonged to the CoV2-specific quadrant (e.g. IL1B, TNF, CCL3, CD163, TIMP1, PLAC8). On the other hand, genes over-expressed in macrophages from HD lung were unaffected or downregulated in our datasets (Fig. 5B , right, chisq p=0.163). In blood monocytes (15) , chosen to assess a systemic spread of the effect, the index computed from the coculture CoV2-up signature showed a clear correspondence with disease severity (Fig. 5C ). In the other direction, the genes whose expression was up-or downregulated in blood monocytes from these severe COVID-19 patients relative to unexposed controls showed a strong bias in our coculture datasets (Fig. 5D , chisq p<0.006).

Given the described "sepsis without bacteria" clinical state of severe COVID-19 patients (3, 8) and the strong overlap between LPS-induced genes and our CoV2 co-culture signature, we asked if the CoV2 signature correlated with transcriptional alterations of the myeloid compartment in severe sepsis. Reyes et al. reported an expansion of a specific monocyte state (MS1) in patients with severe bacterial sepsis (45) , which was also upregulated in monocytes from COVID-19 patients (21) . Highlighting MS1 versus MS2 (classical MHC-II high monocytes) signature genes into the co-culture datasets revealed a significant enrichment in the CoV2-co-cultured monocytes but a strong downregulation from co-culture with IAV-infected cells (Fig. 5E , chisq p<10 -4 ). Thus, the in vitro co-culture CoV2 signature recapitulates the dysregulated myeloid state reported in severe COVID-19 patients, both at the local and systemic level, and overlaps with the bacterial sepsis monocyte profile.

Having observed a specific response to CoV2-infected cells in co-cultured monocytes which corresponded to signatures in blood monocytes in severe COVID-19, we hypothesized that these effects might be related to age-dependent course of disease and the mostly benign evolution of COVID-19 in children. To test this notion, we co-cultured PBMCs from healthy children (4-14 years of age) with mock-or CoV2-infected Caco-2 cells, and compared monocyte transcriptional responses to those observed with adult monocytes (2 independent BSL4 experiments). Analysis of the CoV2-up and -downregulated signatures derived from adults showed that the monocyte response to CoV2 was qualitatively conserved in children ( 

We developed an in vitro model of the initial encounter between immunocytes and CoV2infected epithelial cells to investigate the cause of CoV2-induced immune changes. CoV2-infected epithelial cells directly stimulated a mixed antiviral and inflammatory response in monocytes, with components that were unique to CoV2 when compared to influenza and Ebola viruses. Several Returning to the question of why CoV2 induces such a unique pathological response in some patients, our results offer several suggestive insights. Supporting our in vitro results, this unique inflammatory component has also been observed in transcriptomic analyses from severe COVID-19 patients (19, 46) , whereas it seemed absent in patients with severe IAV (47) . This component was shared between B cells and monocytes. The strong pro-inflammatory character of monocytes co-cultured with CoV2-infected cells, distinct from IAV and EBOV co-cultures and reminiscent in many ways of a TLR-driven response, suggests that CoV2 may deviate the immune response at its earliest stages, at the expense of effective antiviral immunity. Such an idea would nod toward a mechanistic underpinning of the "sepsis without bacteria" clinical picture of COVID-19 that has been reported (3, 8) . The induction of IL6, TNF and IL1B are easy to consider in this context, but IL10, classically considered an anti-inflammatory cytokine, is more puzzling. Like IL6, IL10 is strongly associated with COVID-19 severity (18, 48, 49) , and some prior reports suggest that it may paradoxically enhance inflammation in such settings (IL10 enhances endotoxemia (50) and induces IFNγ in patients with Crohn's disease (51) ). Later in the infection, this inappropriate initial polarization of the innate immune system may give rise to misfocused adaptive immune responses, such as the early germinal center exit of B cells and the poor T cell responses described in severe COVID-19 (18, 52, 53 (22) (23) (24) . In the co-cultures, ISGs were induced along with the pro-inflammatory response, which is not surprising given that the response to CoV2-infected cells is highly reminiscent of the response to TLR ligands, and that IFN induction is a consequence of activation by many TLRs (via TRIF and IRF3). It was interesting that only half of the ISGs induced by CoV2-infected cells were shared with EBOV or IAV infection. This cluster K4 included many key anti-viral ISGs, and we propose that these correspond to true IFNinduced responses elicited by all viruses. On the other hand, the CoV2-specific ISGs of cluster K2 may be induced independently of Type-1 IFN, e.g. by IFN or through other signaling pathways directly activated by infected epithelial cells. Thus, from the initial interaction between CoV2infected epithelial cells and monocytes, the stage is set to counterbalance an IFN response that is essential for viral clearance by a pro-inflammatory diversion. Table S3 ). (iv) Soluble mediators are at least partially involved, since cocultures with physical separation of the cells in Transwell reproduced these effects, but not simply by transcriptional induction of cytokine or chemokines, as evidenced by extensive profiling of the infected or transfected Caco-2 cells themselves. Integrating these threads, we suggest that CoV2 infection, and/or the expression of individual CoV2 proteins, causes the epithelial cells to display or release increased amounts of mediators that activate innate sensors in monocytes. Candidates include mediators such as HMGB1 (56), F-actin (57), or other cell-derived "damage-associated molecular patterns" (DAMPs (58)). Our hypothesis that host products from infected cells trigger monocytes complements recent reports of a molecular interaction between CoV2 proteins and TLR2 or C-type lectins on myeloid cells (59, 60) ; some of us have also reported that the S protein from SARS-CoV1, expressed in PBMCs via a herpes viral vector, can induce IL6 expression (40) .

This direct triggering by S may parallel the more general pro-inflammatory pathway induced by a variety of viral proteins, underlining the evolutionary importance of this response for highly pathogenic coronaviruses. Finally, Reyes et al. have shown that IL6 alone can induce in monocytes transcriptional changes (the "MS1" program) with similarities to deviations observed in COVID-19 or sepsis patients (21) , suggesting a causal role for IL6. This cannot be the case here (no IL6 was detected in infected or transfected Caco-2 cells), but it may be that IL6 acts as a feed-forward loop, induced by CoV2-infected cells and then further amplifying the deviation.

Finally, what to make of the muted response to CoV2-infected cells in monocytes from children, affecting both the ISG and pro-inflammatory components? The relative protection children enjoy from severe COVID-19 is one of the most unique aspects of CoV2 compared to other common respiratory viruses (11) (12) (13) . Although this is only a 2-point correlation, we speculate that the low responsiveness of their monocytes could be a key element of children's relative protection. Mechanistically, immunocytes from children may be less responsive due to a relative naiveté vis-à-vis prior inflammatory exposures (a relative absence of "trained immunity"), or the difference may reflect the systemic pro-inflammatory tone that develops with aging (61, 62) .

In conclusion, these results indicate that the dangerous inflammatory course followed by COVID-19 may be rooted in the very first immune interactions, with amplifying deviations that children are able to avoid. Modulating this inflammatory seed might prevent the subsequent exuberant and deleterious immune activation. Cells were passaged using 0.05% Trypsin-EDTA solution. Cultures were verified to be free of mycoplasma contamination using MycoAlert Mycoplasma Detection Kit (Lonza).

SARS-CoV2 stocks (isolate USA_WA1/2020, kindly provided by CDC's Principal Investigator Natalie Thornburg and the World Reference Center for Emerging Viruses and Arboviruses (WRCEVA)) and EBOV (isolate Mayinga, kindly provided by Heinz Feldmann, NIH NIAID Rocky Mountain laboratories) were grown in Vero E6 cells (ATCC CRL-1586) and cultured in DMEM supplemented with 2% FBS, penicillin (50 U/ml), and streptomycin (50 mg/ml). To remove confounding cytokines and other factors, viral stocks were purified by ultracentrifugation through a 20% sucrose cushion at 80,000 x g for 2 h at 4°C (1). Viral titers were determined in Vero E6 cells by tissue culture infectious dose 50 (TCID50) assay and calculated using the Spearman-Kärber algorithm.

All work with EBOV and CoV2 was performed in the biosafety level 4 (BSL-4) facility of the National Emerging Infectious Diseases Laboratories at Boston University, Boston, MA following approved SOPs and inactivation procedures.

Influenza A PR8-GFP virus (A/Puerto Rico/8/1934(H1N1), hereafter IAV) and Madin-Darby canine kidney (MDCK) cells were provided by D. Lingwood (Ragon Institute). IAV stock was made using MDCK cells and titer was determined by plaque assay as described previously (2) .

One day prior to infection, Caco-2 cells were seeded at a density of 10 5 cells per well of a 24-well tissue culture plate or 25.10 3 cells per well of a 96-well tissue-culture plate. Twenty-four hours later, cells were infected with CoV2 or EBOV at a nominal MOI of 10, with IAV at nominal MOI ranging from 0.1-10. After an adsorption period (2h for CoV2 and EBOV, 1h for IAV), the inoculum was removed, replaced with fresh media (2% FBS supplemented DMEM), and cells were incubated at 37˚C for 35 h prior to coculturing with PBMC.

Expression plasmids were kindly provided by D. Gordon 

To determine the response to LPS, PBMCs were cultured for 14 h, in parallel to HEK cocultures, with or without 1 ng/mL of LPS (LPS from E. coli O55:B5, Sigma, Cat# L2880) for 14 h. To assess the effects of co-culture with lysed cells, a freshly passaged single cell suspension of HEK cells at 2.5.10 5 cells/ml was frozen on dry ice, then thawed rapidly at 37°C. The freeze-thawed suspension was centrifuged at 500 x g to eliminate debris and added at a 10-fold dilution in PBMC cultures for 14 h.

Due to lack of flow cytometry sorting in the BSL-4 containment laboratory, different 

Low-input RNA-seq was performed following the standard ImmGen low-input protocol (www.immgen.org), from the 5μl of collected lysis buffer. For the viral co-cultures, magnetically isolated samples were centrifuged at maximum speed (17,000 x g) for 3 min at 4°C before loading, in order to pellet the cellular debris and magnetic particles, the cleaned RNA lysate remaining in the supernatant. Smart-seq2 libraries were prepared as described previously (4) with slight modifications. Briefly, total RNA was captured and purified on RNAClean XP beads (Beckman Coulter). Polyadenylated mRNA was then selected using an anchored oligo(dT) primer (50 -AAGCAGTGGTATCAACGCAGAGTACT30VN-30) and converted to cDNA via reverse transcription. First strand cDNA was subjected to limited PCR amplification followed by Tn5 transposon-based fragmentation using the Nextera XT DNA Library Preparation Kit (Illumina).

Samples were then PCR amplified for 12 cycles using barcoded primers such that each sample carries a specific combination of eight base Illumina P5 and P7 barcodes for subsequent pooling and sequencing. Paired-end sequencing was performed on an Illumina NextSeq 500 using 2 x 38bp reads with no further trimming.

Reads were aligned to the human genome (GENCODE GRCh38 primary assembly and Quality control: Samples with less than 1 million uniquely mapped reads were automatically excluded from normalization to mitigate the effect of poor-quality samples on normalized counts. Samples having fewer than 8,000 genes with over ten reads were also removed from the data. We screened for contamination by using known cell type specific transcripts (per ImmGen ULI RNAseq and microarray data). Finally, the RNA integrity for all samples was Differential gene expression and consensus CoV2 signature: We used an uncorrected t-test (comparing log-transformed expression values in monocytes of CoV2-vs mock-infected cocultures) to assess differential gene expression between the different groups from the normalized read counts dataset. Genes with a FC >2 or <0.5 and nominal p-value < 0.01 were selected for some analyses (Fig. S1, Fig. S2B Analysis of the transfectants co-culture: Datasets from co-cultured monocytes or B cells were pre-processed as above. For Caco-2 transfectants (Fig. 3) , QC-passing datasets were obtained for 27 single-protein datasets (+ GFP transfected controls), all in duplicates. We first selected (Fig. 3B , split into 7 empirically determined k-means clusters). The same procedure was followed for HEK transfectants (Fig. S3B ).

To search for transcripts in transfected Caco-2 cells associated with induction of the CoV2-Up signature in co-cultured monocytes, we first constructed a coculture effect vector by computing the mean log2 fold change vs GFP controls of genes belonging to the CoV2-Up signature in monocytes co-cultured with Caco-2 cells transfected with individual viral genes. Using matched transfectant samples, we computed a per-gene correlation (Spearman) between this vector and the log2 fold change in Caco-2 cells transfected with individual viral genes vs GFP controls. To assess the statistical significance of the resulting correlation coefficients, we repeated this procedure after permuting sample labels 100 times. As the correlation coefficients from the permutation were approximately normally distributed, p values were calculated using a two-tailed one-sample z-test using the mean and standard deviation of the coefficients from permuted data for each gene.

Geneset enrichment analysis: Enrichment of CoV2.up and CoV2.dn signature in Gene Ontology (GO) biological processes pathways and Reactome pathways was computed using

Fisher's exact test with BH-FDR correction, through the g:Profiler interface. Pathways with a false discovery rate (FDR) <1% were selected for further analyses and recorded in Dataset S2. In order to overcome the inherent redundancy of GO pathways, significant pathways were then interpreted and visualized as an enrichment network using Cytoscape (v3.8.2) (7) and its EnrichmentMap and AutoAnnotate modules. Briefly, Cytoscape allows collapsing of redundant significant pathways into a single biological theme using the Jaccard Overlap combined index (cutoff=0.5), thus simplifying the interpretation. After filtering noise, the network represents overlaps among the most enriched pathways (FDR 0.5%), in which similar pathways are automatically group into main biological themes. Regarding the type I IFN and IFN gamma signatures, data were downloaded from the Gene Expression Omnibus (GEO) (https://www.ncbi.nlm.nih.gov/geo/) from two human published datasets GSE142672 and GSE46599 (8) . To reduce noise, genes with a CV between biological replicates <0.7 and an expression level >20 in either comparison groups were selected.

Upregulated transcripts were defined, at an arbitrary threshold, as having a FC >1.5 and a t-test pvalue <0.05.

Overlap with COVID-19 patient profiling datasets: Signatures from in vivo myeloid population were extracted from published sources. The alveolar macrophage signature from severe COVID-19 patients was extracted from Liao et al. (9) , by merging the signature of their two predominant populations found in severe patients: FCN1 hi (group1) and FCN1 lo SPP1 + (group2).

Signature of the MS1 state in sepsis was directly downloaded from the supplementary tables of Reyes et al (10) . PBMC datasets from (11) were retrieved from the COVID19 cell atlas at https://www.covid19cellatlas.org/index.patient.html as an R dataset object which included the cellXgene matrix, and were used with the cell annotation provided by Wilk et al. The CD14 + monocyte cluster was extracted; indices for the CoV2-up signature genes were computed by averaging the pre-computed CoV2-up genes together per cell, and color-coded on the monocyte UMAP space for Fig.4C .

Unless specified otherwise, the data are presented as mean ± SD and tests of associations for different variables between infected/transfected and mock/GFP were computed using the nonparametric Mann Whitney test. Significance of signature overlaps into our dataset was assessed by Chi square test when computing one signature at a time (e.g. assessing one signature in an independent volcano plot), or by a Fisher's exact test with BH-FDR correction when using large curated GO signatures databases. Analyses and plots were done using S-Plus (v8.2.0), RStudio 

The data reported in this paper have been deposited in the Gene Expression Omnibus in at least one co-culture FC>2 or <0.5, p<0.01). The columns represent the different conditions, where only proteins for which both replicates passed the QC are shown. Annotations at the right ribbon show the overlap between these genes and part of the CoV2 signature that was up/down regulated in Caco-2. Figure S4 

T-cell-mediated immunopathology in viral infections

Clinical characteristics of coronavirus disease 2019 in China

Extrapulmonary manifestations of COVID-19

Multiorgan and Renal Tropism of SARS-CoV-2

Endothelial cell infection and endotheliitis in COVID-19

Viral-bacterial coinfection affects the presentation and alters the prognosis of severe community-acquired pneumonia

Increased incidence of co-infection in critically ill patients with influenza

SARS-CoV-2 and viral sepsis: observations and hypotheses

An aberrant inflammatory response in severe COVID-19

State-Level Data Report

SARS-CoV-2 Infection in Children

SARS-COV-2 infection in children and newborns: a systematic review

COVID-19 in children and adolescents in Europe: a multinational, multicentre cohort study

Virologic features of SARS-CoV-2 infection in children

A single-cell atlas of the peripheral immune response in patients with severe COVID-19

Multi-omics resolves a sharp disease-state shift between mild and moderate COVID-19

Systems biological assessment of immunity to mild versus severe COVID-19 infection in humans

Longitudinal analyses reveal immunological misfiring in severe COVID-19

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

Severe COVID-19 is marked by a dysregulated myeloid cell compartment

Plasma from patients with bacterial sepsis or severe COVID-19 induces suppressive myeloid cell production from hematopoietic progenitors in vitro

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

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

Untuned antiviral immunity in COVID-19 revealed by temporal type I/III interferon patterns and flu comparison

Functional exhaustion of antiviral lymphocytes in COVID-19 patients

Severely ill COVID-19 patients display impaired exhaustion features in SARS-CoV-2-reactive CD8 + T cells

Imbalance of regulatory and cytotoxic SARS-CoV-2-reactive CD4 + T cells in COVID-19

Profound Treg perturbations correlate with COVID-19 severity

Diverse functional autoantibodies in patients with COVID-19

Characterization of the inflammatory response to severe COVID-19 illness

Neutralizing Autoantibodies to Type I IFNs in >10% of Patients with Severe COVID-19 Pneumonia Hospitalized in Madrid, Spain

Neutralizing Autoantibodies to Type I Interferons in COVID-19 Convalescent Donor Plasma

Comparative host-coronavirus protein interaction networks reveal pan-viral disease mechanisms

The Global Phosphorylation Landscape of SARS-CoV-2 Infection

Expression of SARS coronavirus 1 spike protein from a herpesviral vector induces innate immune signaling and neutralizing antibody responses

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

Microarray analysis of gene expression profiles in cells transfected with nonviral vectors

Transcriptional effects of transfection: the potential for misinterpretation of gene expression data generated from transiently transfected cells

Transcriptomic profiling of SARS-CoV-2 infected human cell lines identifies HSP90 as target for COVID-19 therapy

An immune-cell signature of bacterial sepsis

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

Immunophenotyping of COVID-19 and influenza highlights the role of type I interferons in development of severe COVID-19

Longitudinal COVID-19 profiling associates IL-1RA and IL-10 with disease severity and RANTES with mild disease

A Potential Role of Interleukin 10 in COVID-19 Pathogenesis

Proinflammatory effects of IL-10 during human endotoxemia

Treatment of Crohn's disease with recombinant human interleukin 10 induces the proinflammatory cytokine interferon gamma

Loss of Bcl-6-Expressing T Follicular Helper Cells and Germinal Centers in COVID-19

Allelic variation in Class I HLA determines pre-existing memory responses to SARS-CoV-2 that shape the CD8 + T cell repertoire upon viral exposure

Single-Cell Sequencing of Peripheral Mononuclear Cells Reveals Distinct Immune Response Landscapes of COVID-19 and Influenza Patients

MDA5 governs the innate immune response to SARS-CoV-2 in lung epithelial cells

HMGB1 and RAGE in inflammation and cancer

Cytoskeletal Exposure in the Regulation of Immunity and Initiation of Tissue Repair

DAMP-sensing receptors in sterile inflammation and inflammatory diseases

TLR2 senses the SARS-CoV-2 envelope protein to produce inflammatory cytokines

SARS-CoV-2 exacerbates proinflammatory responses in myeloid cells through C-type lectin receptors and Tweety family member 2

Homeostasis, inflammation, and disease susceptibility

Drivers of age-related inflammation and strategies for healthspan extension

CD3 Alexa Fluor 700 (clone OKT3, BioLegend, #317340

1,000 cells were sorted directly into 5 μl TCL/ME lysis buffer. Assessment of infection rate in epithelial cells Flow cytometry

After blocking the non-specific binding sites by incubation in blocking buffer (PBS + 10% of donkey serum) for 1 h, cells were stained (2 h, RT) with either: anti-GFP AlexaFluor488 (clone FM264G, BioLegend, #338008, 1:200), rabbit anti-SARS-CoV nucleocapsid (N) protein (Rockland, #200-401-A50, 1:500, cross-reacts with the CoV2 nucleocapsid protein) or goat anti-EBOV VP35 protein (custom-made by Antagene, 1:200). For EBOV and CoV2, cells were stained 30 min with the secondary antibodies anti-rabbit and anti-goat AlexaFluor488, respectively

Scientific) for a minimum of 6 h at 4°C and removed from the BSL-4 laboratory

Ebolaviruses associated with differential pathogenicity induce distinct host responses in human macrophages

Influenza: propagation, quantification, and storage

Comparative host-coronavirus protein interaction networks reveal pan-viral disease mechanisms

Full-length RNA-seq from single cells using Smart-seq2

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

The architecture of SARS-CoV-2 transcriptome

Cytoscape: a software environment for integrated models of biomolecular interaction networks

Human MX2 is an interferon-induced post-entry inhibitor of HIV-1 infection

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

An immune-cell signature of bacterial sepsis

A single-cell atlas of the peripheral immune response in patients with severe COVID-19

FoldChange-FoldChange plot comparing the response of B cells versus monocytes in the context of IAV infection in Caco-2

Overlay of transcriptional signature derived from monocytes co-cultured with killed epithelial cells onto viral co-culture datasets. Left panel shows the gene expression changes induced in monocytes co-cultured with killed epithelial cells (freeze-thawed HEK, see methods) highlighting the differentially expressed genes (p<0.01, FC>2 or <0.5). Right panel shows these genes into the FoldChange-FoldChange

Changes in gene expression as a function of MOI of IAV infection in monocytes co-cultured with IAV infected Caco-2. Each gene is a dot. Genes belonging to the CoV2 down signature (left panel), the CoV2 up signature excluding the 2 ISG clusters (middle panel) or the two ISGs clusters

Dichotomy in the IFN response between IAV and CoV2 highlighted in a FoldChange-FoldChange plot comparing the response in monocytes co-cultured with IAV (y-axis) relative to CoV2 (x-axis)

highlighting: upregulated transcripts from GSE46599 (type I IFNtreated THP-1 cells for 24h), left panel, and GSE142672 (IFN gamma-treated blood monocytes for 24h

Heatmap of ISG transcript expression values in the Caco-2 transfectant dataset

Heatmap of the significantly differential expressed transcripts in monocytes co-cultured with transfected HEK (similar selection as for Caco-2: overall variance in the dataset substantially exceeded inter-replicate variance, and with significant difference from GFP