key: cord-1012328-avdp3fip
authors: Sposito, Benedetta; Broggi, Achille; Pandolfi, Laura; Crotta, Stefania; Clementi, Nicola; Ferrarese, Roberto; Sisti, Sofia; Criscuolo, Elena; Spreafico, Roberto; Long, Jaclyn M.; Ambrosi, Alessandro; Liu, Enju; Frangipane, Vanessa; Saracino, Laura; Bozzini, Sara; Marongiu, Laura; Facchini, Fabio A.; Bottazzi, Andrea; Fossali, Tommaso; Colombo, Riccardo; Clementi, Massimo; Tagliabue, Elena; Chou, Janet; Pontiroli, Antonio E.; Meloni, Federica; Wack, Andreas; Mancini, Nicasio; Zanoni, Ivan
title: The interferon landscape along the respiratory tract impacts the severity of COVID-19
date: 2021-08-19
journal: Cell
DOI: 10.1016/j.cell.2021.08.016
sha: 725af2874c52e117207834151bccf38e89fc7249
doc_id: 1012328
cord_uid: avdp3fip

Severe COVID-19 is characterized by overproduction of immune mediators, but the role of interferons (IFNs) of the type I (IFN-I) or type III (IFN-III) families remains debated. We scrutinized the production of IFNs along the respiratory tract of COVID-19 patients and found that high levels of IFN-III, and to a lesser extent IFN-I, characterize the upper airways of patients with high viral burden but reduced disease risk or severity. Production of specific IFN-III, but not IFN-I, members, denotes patients with a mild pathology and efficiently drives the transcription of genes that protect against SARS-CoV-2. In contrast, compared to subjects with other infectious or non-infectious lung pathologies, IFNs are over-represented in the lower airways of patients with severe COVID-19 that exhibit gene pathways associated with increased apoptosis and decreased proliferation. Our data demonstrate a dynamic production of IFNs in SARS-CoV-2-infected patients and show IFNs play opposing roles at distinct anatomical sites.

Since the outbreak of the coronavirus disease 2019 in late 2019, the novel, severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has infected over 188 million people globally and caused more than 4 million deaths as of July 2021. SARS-CoV-2 infection can lead to acute respiratory distress syndrome (ARDS) characterized by elevated levels of pro-inflammatory cytokines in the bloodstream (Guan et al., 2020; Lee et al., 2020; Lucas et al., 2020; Zhou et al., 2020a) . Mouse models and retrospective human studies suggest that severity and death following SARS-CoV-2 encounter is correlated with exaggerated inflammation rather than viral load (Bergamaschi et al., 2021; Guan et al., 2020; Karki et al., 2021; Lee et al., 2020; Lucas et al., 2020; Ruan et al., 2020; Winkler et al., 2020; Zhou et al., 2020a) . Nevertheless, how a balance between the benefits (restricting viral replication and spread) and risks (inducing a cytokine storm) of efficient immune cell activation is achieved during COVID-19 remains a mystery.

Of the many inflammatory mediators produced upon infection with SARS-CoV-2, interferons (IFNs) have attracted much attention since the inception of the pandemic. IFNs belong to three major families: IFN-I (mainly represented by IFN-αs and IFN-β), IFN-II (IFN-γ), and IFN-III (IFN-λ1-4). Upregulation of IFN-II in patients with severe COVID-19 (Karki et al., 2021; Lucas et al., 2020) is associated with increased PANoptosis, which exacerbates pathology and death (Karki et al., 2021) . In contrast, the roles of IFN-I and IFN-III during SARS-CoV-2 infection have been a matter of debate. Indeed, IFN-I and IFN-III exert potent antiviral functions via the induction of IFN-stimulated genes (ISGs) . Several studies showed that SARS-CoV-2, compared to other viruses, boosts the production of inflammatory mediators, while delaying and/or dampening anti-viral IFN responses in patients with severe COVID-19 (Blanco-Melo et al., 2020; Galani et al., 2021; Hadjadj et al., 2020; Mudd et al., 2020) . Nevertheless, regulation of IFN-I and IFN-III production following infection with SARS-CoV-2 appears to be more complex. In fact, analyses of the nasopharyngeal swabs (Cheemarla et al., 2021; Lieberman et al., 2020; Ziegler et al., 2021) , the bronchoalveolar lavage fluid (BALF) (Zhou et al., 2020b) , or the peripheral blood monocytes (Lee et al., 2020) of COVID-19 patients have revealed potent ISG induction. Production of IFNs is also sustained in the blood of a longitudinal cohort of severe COVID-19 patients compared to subjects with a mild illness (Lucas et al., 2020) .

Aside from the challenge of understanding the pattern of expression of IFNs, a major unanswered question is whether IFNs serve protective or detrimental functions in . Recent studies show that J o u r n a l P r e -p r o o f patients with severe COVID-19 have defective IFN responses Combes et al., 2021; Laing et al., 2020; Pairo-Castineira et al., 2021; Wang et al., 2021; Zhang et al., 2020) . Other studies, however, report that heightened and prolonged production of IFNs in patients infected with SARS-CoV-2 is correlated with negative clinical outcomes (Lee et al., 2020; Lucas et al., 2020) . We and others have also recently demonstrated that the production of IFN-III, and to a lesser extent of IFN-I, impairs lung function and may trigger a severe disease in mouse models of lung viral infections (Broggi et al., 2020a; Major et al., 2020) .

Thus, it is urgent to fully unravel the role of IFNs in the pathogenesis of COVID-19.

To define how IFN production impacts the progression of COVID-19, here we analyzed the pattern and level of expression of IFNs and the transcriptional programs associated with the IFN landscape in the upper or lower respiratory tract of COVID-19 patients, subjects with infectious and non-infectious lung diseases and healthy controls.

We initially analyzed IFN gene expression in nasopharyngeal swabs derived from SARS-CoV-2 positive and negative subjects (Table S1 and Supplemental Figure 1A -C) and found that, in subjects positive for SARS-CoV-2, IFNL1 and IFNL2,3 (amongst type III interferons) and IFNB1 and IFNA2 (amongst IFN-I) were significantly upregulated (Figure 1A-F). As controls, IL1B and IL6 were also analyzed in the same cohort of subjects (Supplemental Figure 1D , E). To account for the bimodal distribution of cytokine gene expression, we transformed gene expression data in discrete variables (expressed or undetected) and obtained results similar to what we observed with continuous gene expression (Supplemental Figure 1F-M) .

Next, we examined the distribution of IFN levels relative to the viral load. Of the IFN-III family members, IFN-λ1 and IFN-λ2,3 positively correlated with viral load (Figure 1G-I) . Amongst IFN-I, IFN-β and IFN-α4 also showed a positive correlation with the viral load ( Figure 1J-L) . Transcript levels of the proinflammatory cytokines IL1B and IL6 were also positively correlated with the viral load (Supplemental Figure 1N-O) . Next, we divided our patient cohort into terciles based on the viral load (Table S1) , and analyzed gene expression J o u r n a l P r e -p r o o f using discrete variables. These analyses confirmed that IFN-λ1, IFN-λ2,3, IFN-β, IL-1β and IL-6 were preferentially expressed in high, compared to low, viral load samples (Supplemental Figure 1P-W) .

We, then, evaluated how IFN gene expression relates to the age of patients, a key determinant of severity and lethality of COVID-19 (McPadden et al., 2020; Williamson et al., 2020) . Our analyses demonstrated that IFN-III and IFN-I expression is significantly associated with the viral load for the < 70 cohort ( Figure 1M-R) . In contrast, IFN expression in the ≥70 patient cohort either completely lost association with the viral load and/or showed a significantly lower correlation coefficient compared to the < 70 cohort (Figure 1M -R). IL-1β and IL-6 maintained their association with the viral load independent of age and were not significantly different in the two age cohorts (Supplemental Figure 1X , Y). When we analyzed gene expression as a discrete variable, we found that response patterns to viral load were significantly different between elderly (≥70) and younger (<70) patients for IFN-λ2,3 and IFN-α4 (Supplemental Figure 1Z -AE and Table S2 ). This analysis also showed that only younger patients have a dose-response relationship between IFN gene expression and viral load. In contrast to IFNs, no difference in the dose-response relationship between IL-1β and IL-6 expression and viral load was observed between age groups (Supplemental Figure 1AF , AG, and Table S2 ). These results indicate that in COVID-19 patients the production of IFNs correlates with the viral load in the upper respiratory tract and that elderly patients, who are at risk of developing severe disease, have dysregulated IFN induction which correlates more loosely with the viral load compared to younger patients.

To explore the link between IFN production and disease severity, we analyzed nasopharyngeal swabs from a subset of patients with known clinical follow-up. Disease severity was assessed as follows: patients with mild disease manifestations discharged from the emergency room without being hospitalized (home-isolated, HI); severe patients that required hospitalization (HOSP); and critically ill patients admitted to the intensive care unit (ICU) ( Table S3) . When gene expression levels were plotted against the viral load in HI vs. HOSP/ICU (Figure 2A-H) , patients with a mild disease showed a positive correlation with expression of several members of the IFN-III family (Figure 2A-C) . In HOSP/ICU patients this correlation was lost for IFN-λ2,3 and was significantly reduced for IFN-λ1 compared to HI patients (Figure 2A, B) . In contrast to IFN-III, the positive J o u r n a l P r e -p r o o f correlation between IL6 levels and viral load was maintained only for HOSP/ICU patients ( Figure 2H ). When members of the IFN-I family, or IL1B expression, were analyzed, no positive correlation was found in either hospitalized or HI patients (Figure 2D-G) . To control for possible differences due to random sampling, we assessed how the viral load varies based on the day from symptom onset in patients with different disease severity (Table S3 ) and found no significant difference ( Figure 2I ).

To further investigate the distribution of IFN-III production in subjects with mild, severe or critical illness, we performed K-mean clustering based on the expression of IFN-I, IFN-III and IL-1β. Our results reveal that To gain more insight into the transcriptional programs linked to expression of specific IFN members, we used targeted RNA sequencing (RNA-seq) to examine the swabs of a subset of COVID-19 patients (Table   S4) . We found that IFN-λ1 and IFN-λ3 (now distinguishable from IFN-λ2 because of sequencing) segregated with subjects with mild COVID-19 and a high viral load compared to healthy controls or more severely ill COVID-19 patients ( Figure 3A) . IFN-γ was expressed in patients with mild and severe COVID-19, while IFN-I as well as IFN-λ2 were mostly associated with critical, and to a lesser extent with severe, patients ( Figure 3A) .

When gene set enrichment analysis was performed, the IFN responses were the most significantly enriched in subjects with mild, compared to severe or critical, COVID-19 ( Figure 3B and Supplemental Figure 3A , B).

When compared to swabs from SARS-CoV-2 negative subjects, only patients with mild and severe, but not critical, COVID-19 were enriched in IFN responses (Supplemental Figure 3C ). To determine whether the pattern of IFNs found in HI patients drove a protective response against SARS-CoV-2, we tested expression of J o u r n a l P r e -p r o o f >50 ISGs that directly restrain SARS-CoV-2 infection . RNA-seq data demonstrate that only patients with mild manifestations efficiently upregulated this set of protective ISGs ( Figure 3C and Supplemental Figure 3D , E). and that this set of ISGs was significantly enriched compared to controls (Supplemental Figure 3F ).

Due to the high sequence identity of the IFN-III family members (Broggi et al., 2020b) , we, next, compared the capacity of IFN-λ1, IFN-λ2, and IFN-λ3 to induce specific ISGs. We stimulated human bronchial epithelial cells (hBECs) with different type III IFNs and found that IFN-λ1 induces and sustains the transcription of several ISGs more efficiently than IFN-λ2, and, to some extent, than IFN-λ3 ( Figure 3D-G) . Overall, our data demonstrate that specific members of the IFN families associate with mild or severe COVID-19. Also, that the landscape of IFNs determines the ISGs induced in the upper airways and that IFN-λ1 is uniquely capable of inducing potent anti-SARS-CoV-2 ISGs in patients with mild COVID-19.

A detailed analysis of the IFNs produced in the lower airways of SARS-CoV-2 infected subjects is lacking. We, thus, analyzed BALF samples derived from COVID-19 hospitalized patients, including ICUadmitted subjects, and, as controls, samples derived from patients with non-infectious lung pathologies (sex and age distribution reported in Table S5 We next compared the expression of IFNs between the lower and upper airways of COVID-19 patients with similar disease severity. Sex and age were distributed as reported in Table S5 . We found that, except for IFN-λ1, levels of IFNs in severe-to-critical patients were higher in the lower compared to the upper airways ( Figure 4I-N We, next, performed RNA-seq of the BALF of a subset of ICU-isolated patients and of subjects with non-infectious lung pathologies (Table S6) . Gene set enrichment analysis confirmed that IFN responses characterize COVID-19 patients, compared to non-microbially infected patients ( Figure 5A , B). In keeping with the capacity of IFNs to increase apoptosis and facilitate lung tissue damage (Broggi et al., 2020a; Major et al., 2020) , gene enrichment also revealed that the p53 pathway is significantly upregulated in COVID-19 patients ( Figure 5A, C) . Notably, the IFNs landscape in the upper and lower airways of critical patients was strikingly similar ( Figure 5D) . Also, the induction of ISGs that protect against SARS-CoV2 was significantly decreased in the lower airways of critical COVID-19 patients, compared with the upper airways of patients with milder, as well as similar, disease severity (Supplemental Figure 5A -C). The gene signatures in the upper airways of mildly ill patients, compared with either the upper or lower airways of critical patients, were enriched for pathways associated with the induction of ISGs and other inflammatory pathways, ( Figure 5E ). In keeping with the capacity of IFNs to dampen cell proliferation and delay tissue repair (Broggi et al., 2020a; Major et al., 2020) , gene programs linked to proliferation were significantly downmodulated in the lower airways of critical patients compared to the upper airways of subjects with a mild disease (Supplemental Figure 5D ).

Overall, these data demonstrate that a unique IFN signature characterizes severe-to-critical COVID-19 patients along the respiratory tract and that the induction of unique set of IFNs is coupled with either protective ISGs, or gene programs associated with apoptosis and reduced proliferation. patients. However, whether the relative distribution of the IFN members, as measured by mRNA transcripts, correlates with their protein levels remains unknown. We thus assessed protein levels of IFNs and other inflammatory cytokines in the BALF of subjects infected with COVID-19 compared to the BALFs of patients with ARDS not driven by SARS-CoV-2 or patients with non-infectious lung involvement including fibrosis, sarcoidosis, or lung transplant (going forward, referred to as "controls") ( Table S7) When we compared the protein levels in the BALF and in the plasma of a subset of COVID-19 patients (STAR Methods), no correlation between these levels for any protein analyzed was found ( Figure 6E -J and Supplemental Figure 6I ), confirming at the protein level the transcriptional differences recently highlighted between the peripheral blood and the lungs of COVID-19 patients (Overholt et al., 2021) .

When we performed unbiased K-means clustering of the protein analyzed, we found that COVID-19 patients were significantly enriched in cluster 3 which is characterized by a unique signature of IFNs (which encompasses all three IFN families) and IL-10 production ( Figure 6K -O and STAR Methods). Many proinflammatory cytokines are also upregulated in cluster 2 which is enriched in patients who have ARDS that is not driven by SARS-CoV-2 ( Figure 6L and Supplemental Figure 6J) ; most of these patients also express IFN-λ1, but not other IFNs. Control patients were, in contrast, enriched in cluster 1, characterized by low proinflammatory cytokine and IFN responses ( Figure 6L and Supplemental Figure 6K ).

Overall, these data demonstrate that COVID-19 patients are characterized by a unique IFN signature in the lower airways relative to patients with ARDS of different etiology.

J o u r n a l P r e -p r o o f Based on the heterogenous induction of IFNs along the respiratory tract of COVID-19 patients with different disease severity, we hypothesized that different populations of cells contribute to production of specific IFNs by activating discrete pattern recognition receptors (PRRs). Our finding that the mRNA for IFNL1 is absent in the lower airways of COVID-19 patients ( Figure 4A ), but protein levels for IFN-λ1 are present at the same anatomical site ( Figure 6A ) suggests that cells that actively produce the mRNA for IFNL1 are underrepresented in the BALF. However, IFNL1 is one of the most upregulated genes in the upper airways, supporting the hypothesis that the cells that produce it are highly represented in the swabs. We, thus, explored the cellular composition of the swabs and BALFs by deconvoluting our bulk RNAseq data ( Figure 7A -C and Supplemental Figure 7A , B). We found that the epithelial compartment, represented by several epithelial cell lineages, is more represented than the hematopoietic compartment in swabs from SARS-CoV-2 negative and positive subjects ( Figure 7A , C and Supplemental Figure 7A ). In contrast, BALFs from both SARS-CoV-2 negative and positive patients present very diversified hematopoietic populations ( Figure 7B and Supplemental Figure 7B ) that are more represented than epithelial cells ( Figure 7C) .

We, thus, explored how epithelial cells, or phagocytes, differentially contribute to the production of IFNs during SARS-CoV-2 encounter ( Figure 7D ). We confirmed that polarized hBECs of healthy individuals are sensitive to SARS-CoV-2 infection (Supplemental Figure 7C ) and respond by expressing IFNs (Figure 7E-H) and pro-inflammatory cytokines (Supplemental Figure 7D, E) . Notably, hBECs infected with SARS-CoV-2 mostly produced IFN-λ1 compared to other IFNs ( Figure 7E-H) . Among human phagocytes, plasmacytoid (p)DCs respond to SARS-CoV-2 by producing mainly IFN-I (Onodi et al., 2021) . Based on the potent induction of IFN-III in patients with mild COVID-19, we focused our attention on conventional (c)DCs that we recently described as major producers of IFN-III in the lungs of mice (Broggi et al., 2020a) . Human cDCs isolated from the blood of healthy donors did not produce IFNs or other inflammatory cytokines when exposed to SARS-CoV-2 in vitro (data not shown). To test the possible involvement of cDCs during COVID-19, we infected a human lung epithelial cell (hLEC) line with SARS-CoV-2 and exposed cDCs to the supernatant of these cells.

We found that only cDCs exposed to the supernatant of virally-infected hLECs upregulated the expression of IFN-λ2,3 (but not IFN-λ1), members of the IFN-I family, as well as IL-1B and IL-6, ( 

To identify the PRRs involved in the production of IFNs by either human epithelial cells or cDCs, we tested different PRR ligands ( Figure 7M ). In keeping with a central role of the RIG-I/MDA-5 pathway in epithelial cells for sensing SARS-CoV-2 Wu et al., 2021; , stimulation of the RIG-I pathway, and to a lesser extent of TLR3, in epithelial cells potently induced the transcripts of IFN-III and IFN-I, but not of other pro-inflammatory mediators (Supplemental Figure 7H , and STAR Methods). The analysis of protein levels confirmed the transcriptional data ( Figure 7N , Supplemental Figure 7I , J, and STAR Methods). In keeping with SARS-CoV-2 infection, epithelial cells were more potent producers of IFN-λ1 compared to IFN-λ2,3 upon stimulation of TLR3, RIG-I and MDA-5 pathways ( Figure 7N , Supplemental Figure 7I ).

We next evaluated the response of cDCs. As a comparison, we also treated bulk peripheral blood mononuclear cells (PBMCs), monocytes isolated from PBMCs, as well as monocyte-derived DCs (moDCs).

While PBMCs were particularly able to produce IFN-II in response to viral and bacterial ligands, cDCs were uniquely capable of producing very high levels of IFN-λ2,3, and to a lesser extent of IFN-λ1, solely in response to TLR3 stimulation ( Figure 7O , Supplemental Figure 7K -M, and STAR Methods). Monocytes and moDCs were poor producers of IFNs in response to all the stimuli tested. When these analyses were extended to other inflammatory mediators, each cell type revealed a unique pattern of protein production (Supplemental Figure   7N , and STAR Methods), underscoring the complexity and cell-specificity of the inflammatory response.

Collectively, these data demonstrate that epithelial cells preferentially produce IFN-λ1 upon SARS-CoV-2 infection and suggest that IFN production is driven via RIG-I/MDA-5 or TLR3 stimulation. Also, that cDCs only respond to the supernatant of SARS-CoV-2-infected cells, and that TLR3 is the major driver of IFN-III production by human cDCs.

COVID-19 has caused millions of deaths, and has had devastating societal and economic effects.

Notwithstanding the efficacy of the COVID-19 vaccines, a better understanding of the molecular underpinnings that drive the severe disease in patients infected with the SARS-CoV-2 virus is imperative to implement effective additional prophylactic and/or therapeutic interventions. IFN-I and IFN-III are potent anti-viral cytokines and the potential of using clinical grade recombinant IFN-I or IFN-III as therapeutics has raised much J o u r n a l P r e -p r o o f hope and interest (Prokunina-Olsson et al., 2020) . To date, though, opposing evidence has complicated our view of the role played by members of the IFN-I and IFN-III families during SARS-CoV-2 infection.

We found that in the upper airways of patients with mild manifestations, the presence of IFN-λ1 and IFN-λ3, but not IFN-λ2 or IFN-I, was associated with the induction of ISGs known to efficiently contain SARS-CoV-2. Our data also demonstrated that critically ill patients express high levels of compared to subjects with a mild disease or healthy controls. These patients show a reduced induction of protective ISGs and, in general, of IFN responses. Two non-mutually exclusive explanations for this behavior may be that: i) the pattern of IFN expression of critically ill patients is less capable of inducing the protective ISGs; ii) other factors, such as the production of specific antibodies that block ISG induction (Combes et al., 2021) , or viral adaptation to evade control by IFN-I (Lei et al., 2020; Xia et al., 2020) , restrain the capacity of this set of IFNs to mount a strong response.

The present in-depth analysis shows not only that high viral loads of SARS-CoV-2 induce the efficient production of IFN-III in the upper airways of younger and/or milder patients, but also that severely ill COVID-19 patients are characterized by the highest levels of IFNs (at the mRNA as well as protein levels) in the lower airways. These data support the hypothesis that IFNs have opposing roles along the respiratory tract, and reconcile some of the seemingly contradictory findings on IFNs in COVID-19 patients. Efficient initiation of IFN production in the upper airways can lead to a more rapid elimination of the virus and may limit viral spread to the lower airways, as suggested by studies that report defects in IFN signaling of severe COVID-19 patients Pairo-Castineira et al., 2021; Wang et al., 2021; Zhang et al., 2020) . On the other hand, when the virus escapes immune control in the upper airways, the IFN production that is potently boosted in the lungs likely contributes to the cytokine storm and associated tissue damage, that are typical of patients with severe-to-critical COVID-19, characterized by reduced proliferation and increased pro-apoptotic p53 transcriptional signatures.

Another novel finding in the present study is that the type of IFN produced in response to different PRR pathways varies according to cell types. In keeping with ACE2 + cells being the primary cells infected by SARS-CoV-2, we measured a potent immune response in human bronchial epithelial cells, but not in cDCs infected with SARS-CoV-2. Nevertheless, we found that cDCs efficiently express specific members of the IFN-III and IFN-I families when exposed to the supernatant of lung epithelial cells previously infected with SARS-CoV-2, or J o u r n a l P r e -p r o o f in response to dsRNA. These data suggest that cDCs, despite not responding directly to SARS-CoV-2 infection, may play fundamental roles in recognizing intermediates of viral replication and/or DAMPs released by dying cells.

Finally, our findings highlight the importance of the timing of production and/or administration of IFNs during COVID-19 and suggest that early administration (before infection or early after symptom onset) of specific recombinant IFN-III may be an effective therapeutic intervention, and that targeting the upper airways, while avoiding systemic administration as previously proposed (Park and Iwasaki, 2020) , represents the best way to exploit the anti-viral activities of IFNs.

In conclusion, our data define the anatomical map of inter and intra-family production of IFNs during COVID-19, and highlight how IFN production is linked to the different clinical outcomes, based on the location of the IFN response. Our findings reconcile a large portion of the literature on IFNs, and further stress the key role played by IFN-III, compared to IFN-I, at mucosal surfaces during life-threatening viral infections. These findings will be fundamental for designing appropriate pharmacological interventions to prevent infection with SARS-CoV-2 or to dampen the severity of COVID-19 and will help to better understand how the IFN landscape affects human immune responses to respiratory viral infections.

Limitations of the Study. Our findings shed new light on the nature of the IFNs and on the molecular pathways that drive intrinsic immunity. The capacity of lung epithelial cells to recognize and respond to viral components is confounded by the presence of SARS-CoV-2 effector proteins which block immune recognition and IFN production (Banerjee et al., 2020; Konno et al., 2020; Lei et al., 2020; Wu et al., 2021) . We show that high viral load in the upper airways of COVID-19 patients induces a potent immune response and that viral loads are not correlated per se with disease severity. High viral loads in the upper airways may therefore be associated to a protective immune response in young individuals, whilst eliciting a dysregulated inflammatory response in older patients, as observed in our study. Nevertheless, additional studies are needed to directly link specific IFNs to particular cell types and, above all, to specific protective or detrimental immune cell functions. As an example, our data suggest that cDCs do not directly sense SARS-CoV-2. Intriguingly, a recent report showed that specific cDC subtypes may instead respond to SARS-CoV-2 encounter (Marongiu et al., 2021) , but the capacity of these subtypes to produce specific IFNs remains an open question. Furthermore, understanding J o u r n a l P r e -p r o o f the specific contribution of different PRRs to the IFN response elicited in patients infected with SARS-CoV-2 also requires further analyses.

We thank the members of the Zanoni and Mancini laboratories for thoughtful discussion and comments on the project. IZ is supported by NIH grants 2R01AI121066, 5R01DK115217, NIAID-DAIT-NIHAI201700100, Lloyd Figure S1 and Table S1. Figure S4 and Table   S5 . hospitalized (Swab HOSP, 6) or home-isolated (Swab HI, 4) and from SARS-CoV-2 negative (Swab NEG, 2) patients, and BALF from SARS-CoV-2 positive patients (BALF POS, 7) and from patients with non-infectious lung pathologies (BALF NEG CTRL, 3) was performed. Data was deconvoluted based on publicly available scRNAseq datasets (Ziegler et al., 2021) using CIBERSORTx (Newman et al., 2019) 

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Ivan Zanoni (ivan.zanoni@childrens.harvard.edu).

This study did not generate new unique reagents.

 Targeted transcriptomics data have been deposited at GEO and are publicly available as of the date of publication. Accession numbers are listed in the Key Resources Table.  Gene expression matrix from targeted transcriptomics, Gene expression matrix from qPCR experiments, cytokine expression matrix from multiplex analysis of BALF, Plasma and supernatants of phagocytes are deposited at Mendeley and are publicly available as of the date of publication.

The DOI is listed in the Key Resources Table.  The code used to analyze the data is available upon request to the corresponding authors.

Nasopharyngeal swabs were collected using FLOQSwabs® (COPAN Cat#306C) in UTM® Universal Transport Table S1 , S3, S4, S5, S6 for patient information. All samples were stored at −80°C until processing. 500 μl of each BALF and swab sample were lysed and used for RNA extraction (see

Clinical metadata were obtained from the COVID-BioB clinical database of the IRCCS San Raffaele Hospital.

The study was approved by the Ethics Committee of San Raffaele Hospital (protocol 34/int/2020). All of these patients signed an informed consent form. Our research was in compliance to the Declaration of Helsinki.

BALF from 29 SARS-CoV-2 positive patients hospitalized in the Intensive Care Unit (ICU) at Luigi Sacco Hospital (Milan, Italy) were collected from September to November 2020. The total volume for each lavage was 120ml. Blood from 17 of these patients was also collected on the same day. BALF from patients affected by ARDS (9 in total, 5 of which were diagnosed H1N1 influenza A virus) were collected from February 2014 to March 2018. Samples from: lung fibrosis patients (10) were collected from May 2018 to September 2020; sarcoidosis patients (10) were collected from August to July 2020; lung transplant patients (10) were collected from January 2018 to September 2020 by IRCCS Policlinico San Matteo Foundation (Pavia, Italy). The total volume for each lavage was 150ml. None of the patients affected by lung fibrosis, sarcoidosis or that received lung transplant was diagnosed a respiratory viral or bacterial infection. See Table S7 for patient information.

Research and data collection protocols were approved by the Institutional Review Boards (Comitato Etico di Area 1) (protocol 20100005334) and by IRCCS Policlinico San Matteo Foundation Hospital (protocol 20200046007). All patients signed an informed consent form. Our research was in compliance to the Declaration of Helsinki.

Human phagocytes were isolated from collars of blood received from Boston Children's Hospital blood donor center for in vitro stimulations and from San Raffaele Hospital blood donor center for SARS-CoV-2 infections.

Briefly, blood was diluted 1:2 in PBS and PBMCs were isolated using a Histopaque (Sigma Cat# 1077-1) Table) hBECs were expanded in a T-75 flask to 60% confluence and then trypsinized and seeded either on 48 well plates (2x10 5 cells/well) for IFN stimulations or (3x10 4 cells/transwell) onto 0.4 μm pore size clear polyester membranes (Corning Cat# 3470) coated with a collagen solution for PRR agonist stimulations and SARS-CoV-2 infections.

For hBECs infection experiments with SARS-CoV-2, the isolate England/02/2020 (GISAID accession ID:

experiments with SARS-CoV-2 the isolate hCoV-19/Italy/UniSR1/2020 (GISAID accession ID: EPI_ISL_413489) was propagated and titrated in Vero E6 cells (ATCC Cat# CRL-1586). All infection experiments were performed in a biosafety level-3 (BLS-3) laboratory.

The viral load was inferred on nasopharyngeal swabs through cycle threshold (Ct) determination with Cobas® SARS-CoV-2 Test (Roche Cat# P/N 09175431190), a real-time PCR dual assay targeting ORF-1a/b and Egene regions on SARS-CoV-2 genome. The mean between ORF-1a/b and E Ct was used as an indirect measure of the viral load. Non-infectious plasmid DNA containing a specific SARS-CoV-2 sequence and a pan-Sarbecovirus sequence is used in the test as positive control. A non-Sarbecovirus related RNA construct is used as internal control. The test is designed to be performed on the automated Cobas® 6800 Systems under Emergency Use Authorization (EUA). The test is available as a CE-IVD test for countries accepting the CE-mark. 

IFN stimulations were performed one day after seeding by treating cells with 2ng/ml IFN-λ1, IFN-λ2 and IFN-λ3 for 4 and 24 hours. Cell lysates were processed for RNA extraction as described below. For PRR agonist stimulations and SARS-CoV-2 infections cells were grown in submersion until confluent, and then exposed to air to establish an air-liquid interface (ALI). At ALI day 15, cells were stimulated with LPS (100 ng/ml), R848 (10 μg/ml), CpG(C) (1 μM), Poly (I:C) (50 μg/ml), Poly (I:C) (1 μg/10 6 cells) + Lipofectamine, 3p-hpRNA/LyoVec (100 ng/ml), and cGAMP (10 μg /ml). Supernatants and cell lysates were collected 24 hours post treatment.

Supernatants were processed with LEGENDplex TM (BioLegend Cat# 740390) according to manufacturer's instructions and read by flow cytometry. Lysates were processed for RNA extraction as described below. For SARS-CoV-2 infections on day 15 of ALI cells were washed apically with PBS and infected at a multiplicity of infection (MOI) of 10 -1 for 30 minutes at 37C. The inoculum was then removed, and cell lysates were collected at 24 or 48 hours post infection for RNA extraction as described below.

BALF specimens from COVID-19 patients were managed in a biosafety level 3 laboratory until viral inactivation with a 0.2% SDS and 0.1% Tween-20 solution and heating at 65 °C for 15 min. Cell-free BALF supernatants J o u r n a l P r e -p r o o f were stored at − 20 °C until analysis. Blood was centrifuged at 400g for 10 minutes without brake and plasma was stored at − 20 °C until analysis. Samples were processed with LEGENDplex TM (BioLegend Cat# 740390) according to manufacturer's instructions and read by flow cytometry.

PBMCs, monocytes, moDCs and cDCs were stimulated with LPS (100 ng/ml), R848 (10 μg/ml), CpG(C) (1 μM), Poly (I:C) (50 μg/ml), 3p-hpRNA/LyoVec (2.5 μg/ml), and cGAMP (10 μg/ml). Supernatants were collected 24 hours post treatment and stored at − 20 °C until analysis. cDCs were also stimulated with conditioned media from hLECs. hLECs were infected or not with SARS-CoV-2 at an MOI of 10 -1 and supernatant was collected 72 hours post infection. Cell lysates were collected 24 and 48 hours after treatment for RNA extraction as described below.

RNA was extracted from nasopharyngeal swabs, BALFs, hBECs (stimulated with PRR agonists, with IFNs and infected with SARS-CoV-2) lysates and cDCs (stimulated with supernatant from SARS-CoV-2 infected hLECs) using Pure Link RNA Micro Scale kit (Invitrogen Cat# 12183016) according to manufacturer's instruction, including in-column DNase treatment. Reverse transcription was performed on all samples except IFN-treated hBECs using SuperScript TM III First-Strand Synthesis System (Invitrogen Cat# 18080051) according to manufacturer's instruction. qRT-PCR analysis was then carried out with Taqman TM Fast Advanced Master Mix (Applied Biosystems Cat#4444963) by using specific Taqman TM Gene Expression Assays from Thermo Fisher. 

For targeted transcriptome sequencing, RNA (15ng) isolated from clinical samples described in Table 4 and Table 6 was retro-transcribed to cDNA using SuperScript VILO cDNA Synthesis Kit (Invitrogen Cat# 11754-05). Barcoded libraries were prepared using the Ion AmpliSeq Transcriptome Human Gene Expression Kit (Ion Torrent Cat# A26325) as per the manufacturer's protocol and sequenced using an Ion S5 system (Ion Torrent Cat# A27212). Differential gene expression analysis was performed using the Transcriptome Analysis Console (TAC) software with the ampliSeqRNA plugin (ThermoFisher Scientific).

We used CIBERSORTx (Newman et al., 2019) to estimate the abundances of epithelial end hematopoietic cell types using using bulk gene expression data as an input and scRNAseq signature matrices from single-cell RNA sequencing data to provide the reference gene expression profiles of pure cell populations. The scRNAseq signature matrix used to deconvolute RNAseq dataset from swabs or BALFs were derived from (Wauters et al., 2021; Ziegler et al., 2021) .

Gene set enrichment analysis and enrichment plot were generated in R using the Fast Gene Set Enrichment Analysis package (fGSEA) (Korotkevich et al., 2021) . Heatmaps were generated in R and visualized with the ComplexHeatmap package (Gu et al., 2016) . Clustering analysis was performed using Euclidean distances on individual z-scores. Code available upon request.

One-way ANOVA with Turkey's post-hoc test was used to compare continuous variables among multiple groups. Kruskal-Wallis test with Dunn's post-hoc test or Multiple Mann-Whitney tests with Holm-Šídák method were used instead when data did not meet the normality assumption. Fisher's exact test was used to compare categorical variables. Spearman correlation analysis was used to examine the degree of association between J o u r n a l P r e -p r o o f two continuous variables. To establish the appropriate test, normal distribution and variance similarity were assessed with the D'Agostino-Pearson omnibus normality test.

Cluster analysis with unbiased K-mean methods based on the expression of IFN-I, IFN-III and the proinflammatory cytokine IL-1β were used to classify a subset of COVID-19 patients into 3 exclusive clusters.

Cluster analysis with unbiased K-mean methods based on the expression of Interferons and pro-inflammatory cytokines in the BALF were used to classify COVID-19 patients, non-COVID-19 ARDS patients, and controls into 3 exclusive clusters. Heatmaps and K-mean clustering were generated in R and visualized with the ComplexHeatmap package. Clustering analysis was performed using Euclidean distances. Estimated (K) value was selected based on the elbow point cluster number. Logistic regression models were performed to estimate the association of gene expression as binary outcome within viral load terciles (defined by mean viral RNA CT <20, >20 and <30, > 30), and clusters (cluster 1, cluster 2 and cluster 3). Interaction between viral load terciles and age groups (≥70 years vs <70 years) were tested to detect significant difference between elder patients and young patients in their gene expression response to different levels of viral load. All statistical analyses were two-sided and performed using Prism9 (Graphpad) software or SAS version 9.4 (SAS Institute). All statistical analyses are indicated in figure legends. Throughout the paper significant is defined as follows : ns, not significant (P>0.05); *P<0.05, **P<0.01, ***P<0.001, and ****P<0.0001. Genes   MAX  ST3GAL4  GSDMD  NAPA  STAT1  SPATS2L  B4GALT5  ELF1  RAB27A  IFIT1  IFIT5  IFIT3  DDX60  STAT2  BST2  FAM46C  UBD  SUSD3  REC8  ETV6  FAM46A  CNP  USP18  ARNTL  CLEC4D  GBP3  NRN1  TAGAP  PIK3AP1  MYD88  ISG15  MLKL  RSAD2  CCND3  SPATA13  ISG20  ZBP1  IFITM2  IFITM3  C9orf91  RGS22  FAM134B  FGD2  FZD5  CASP7  APOL4  ERLIN1  MSR1  NT5C3  HSPA8  LY6E  GNB4  TRIM21  LOC152225  PHF15  RAB39A  GBA3  ANGPTL4  UPP2  FNDC4  DNAJC6  APOL2  CRP  IL11  IL4I1 severity_grade viral_load_tertile Z−score Genes   IFNG   IL28B   IFNK   IL29   IFNA7   IFNA6   IFNB1   IL28A   IFNA13   IFNA4   IFNA22P   IFNA17   IFNA10   IFNA5   IFNA21   IFNA1   IFNA14   IFNA16   IFNA8   IFNA2   IFNW1   severity_grade   viral_load_tertile interferons Z−score GSDMD  NAPA  STAT1  SPATS2L  B4GALT5  ELF1  RAB27A  IFIT1  IFIT5  IFIT3  DDX60  STAT2  BST2  FAM46C  UBD  SUSD3  REC8  ETV6  FAM46A  CNP  USP18  ARNTL  CLEC4D  GBP3  NRN1  TAGAP  PIK3AP1  MYD88  ISG15  MLKL  RSAD2  CCND3  SPATA13  ISG20  ZBP1  IFITM2  IFITM3  C9orf91  RGS22  FAM134B  FGD2  FZD5  CASP7  APOL4  ERLIN1  MSR1  NT5C3  HSPA8  LY6E  GNB4  TRIM21  LOC152225  PHF15  RAB39A  GBA3  ANGPTL4  UPP2  FNDC4  DNAJC6  APOL2  CRP  IL11 Genes   IFNG   IL28B   IFNK   IL29   IFNA7   IFNA6   IFNB1   IL28A   IFNA13   IFNA4   IFNA22P   IFNA17   IFNA10   IFNA5   IFNA21   IFNA1   IFNA14   IFNA16   IFNA8   IFNA2   IFNW1   severity_grade   viral_load_tertile interferons Z−score IFNG   IFNL3   IFNK   IFNL1   IFNA7   IFNA6   IFNB1   IFNL2   IFNA13   IFNA4   IFNA22P   IFNA17   IFNA10   IFNA5   IFNA21   IFNA1   IFNA14   IFNA16   IFNA8   IFNA2   IFNW1   4 Genes   IFNA8   IFNA16   IFNA2   IFNW1   IFNA1   IFNA21   IFNA14   IFNA6   IFNB1   IFNA10   IFNA5   IFNA17   IFNA13   IFNA4   IFNA22P   IL28A   IFNK   IFNA7   IL28B   IL29 Genes   IFNA8   IFNA16   IFNA2   IFNW1   IFNA1   IFNA21   IFNA14   IFNA6   IFNB1   IFNA10   IFNA5   IFNA17   IFNA13   IFNA4   IFNA22P   IL28A   IFNK   IFNA7   IL28B   IL29 

SARS-CoV-2 Disrupts Splicing, Translation, and Protein Trafficking to Suppress Host Defenses

Longitudinal analysis reveals that delayed bystander CD8+ T cell activation and early immune pathology distinguish severe COVID-19 from mild disease

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

Type III interferons disrupt the lung epithelial barrier upon viral recognition

Type III interferons: Balancing tissue tolerance and resistance to pathogen invasion

Dynamic innate immune response determines susceptibility to SARS-CoV-2 infection and early replication kinetics

Global absence and targeting of protective immune states in severe COVID-19

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

Complex heatmaps reveal patterns and correlations in multidimensional genomic data

Clinical Characteristics of Coronavirus Disease 2019 in China

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

Synergism of TNF-alpha and IFN-gamma Triggers Inflammatory Cell Death, Tissue Damage, and Mortality in SARS-CoV-2 Infection and Cytokine Shock Syndromes

SARS-CoV-2 ORF3b Is a Potent Interferon Antagonist Whose Activity Is Increased by a Naturally Occurring Elongation Variant

A dynamic COVID-19 immune signature includes associations with poor prognosis

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

Activation and evasion of type I interferon responses by SARS-CoV-2

In vivo antiviral host transcriptional response to SARS-CoV-2 by viral load, sex, and age

ISG15-dependent activation of the sensor MDA5 is antagonized by the SARS-CoV-2 papain-like protease to evade host innate immunity

Longitudinal analyses reveal immunological misfiring in severe COVID-19

Type I and III interferons disrupt lung epithelial repair during recovery from viral infection

Maturation signatures of conventional dendritic cell subtypes in COVID-19 suggest direct viral sensing

Functional landscape of SARS-CoV-2 cellular restriction

Distinct inflammatory profiles distinguish COVID-19 from influenza with limited contributions from cytokine storm

Determining cell type abundance and expression from bulk tissues with digital cytometry

SARS-CoV-2 induces human plasmacytoid predendritic cell diversification via UNC93B and IRAK4

Dissecting the common and compartmentspecific features of COVID-19 severity in the lung and periphery with single-cell resolution. iScience 24, 102738

Type I and Type III Interferons -Induction, Signaling, Evasion, and Application to Combat COVID-19

COVID-19 and emerging viral infections: The case for interferon lambda

Clinical predictors of mortality due to COVID-19 based on an analysis of data of 150 patients from Wuhan, China

Discriminating mild from critical COVID-19 by innate and adaptive immune single-cell profiling of bronchoalveolar lavages

Factors associated with COVID-19-related death using OpenSAFELY

SARS-CoV-2 infection of human ACE2-transgenic mice causes severe lung inflammation and impaired function

ORF9b inhibits RIG-I-MAVS antiviral signaling by interrupting K63-linked ubiquitination of NEMO

Evasion of Type I Interferon by SARS-CoV-2

Inborn errors of type I IFN immunity in patients with life-threatening COVID-19

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

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

Impaired local intrinsic immunity to SARS-CoV-2 infection in severe COVID-19. bioRxiv