key: cord-0793268-m140g2vt authors: van der Sluis, Renée M.; Cham, Lamin B.; Oliver, Albert Gris; Gammelgaard, Kristine R.; Pedersen, Jesper G.; Idorn, Manja; Ahmadov, Ulvi; Hernandez, Sabine Sanches; Cémalovic, Ena; Godsk, Stine H.; Thyrsted, Jacob; Gunst, Jesper D.; Nielsen, Silke D.; Jørgensen, Janni J.; Bjerg, Tobias Wang; Laustsen, Anders; Reinert, Line S.; Olagnier, David; Bak, Rasmus O.; Kjolby, Mads; Holm, Christian K.; Tolstrup, Martin; Paludan, Søren R.; Kristensen, Lasse S.; Søgaard, Ole S.; Jakobsen, Martin R. title: Distinct SARS-CoV-2 sensing pathways in pDCs driving TLR7-antiviral vs. TLR2-immunopathological responses in COVID-19 date: 2021-11-24 journal: bioRxiv DOI: 10.1101/2021.11.23.469755 sha: 29f954b5ea582a375476c88cdf021fc01510dea3 doc_id: 793268 cord_uid: m140g2vt Understanding the molecular pathways driving the acute antiviral and inflammatory response to SARS-CoV-2 infection is critical for developing treatments for severe COVID-19. Here we show that in COVID-19 patients, circulating plasmacytoid dendritic cells (pDCs) decline early after symptom onset and this correlated with COVID-19 disease severity. This transient depletion coincides with decreased expression of antiviral type I IFNα and the systemic inflammatory cytokines CXCL10 and IL-6. Importantly, COVID-19 disease severity correlated with decreased pDC frequency in peripheral blood. Using an in vitro stem cell-based human pDC model, we demonstrate that pDCs directly sense SARS-CoV-2 and in response produce multiple antiviral (IFNα and IFNλ1) and inflammatory (IL-6, IL-8, CXCL10) cytokines. This immune response is sufficient to protect epithelial cells from de novo SARS-CoV-2 infection. Targeted deletion of specific sensing pathways identified TLR7-MyD88 signaling as being crucial for production of the antiviral IFNs, whereas TLR2 is responsible for the inflammatory IL-6 response. Surprisingly, we found that SARS-CoV-2 engages the neuropilin-1 receptor on pDCs to mitigate the antiviral IFNs but not the IL-6 response. These results demonstrate distinct sensing pathways used by pDCs to elicit antiviral vs. immunopathological responses to SARS-CoV-2 and suggest that targeting neuropilin-1 on pDCs may be clinically relevant for mounting TLR7-mediated antiviral protection. One Sentence Summary pDCs sense SARS-CoV-2 and elicit antiviral protection of lung epithelial cells through TLR7, while recognition of TLR2 elicits an IL-6 inflammatory response associated with immunopathology. Graphical abstract: SARS-CoV-2 sensing by plasmacytoid dendritic cells. SARS-CoV-2 is internalized by pDCs via a yet unknown endocytic mechanism. The intracellular TLR7 sensor detects viral RNA and induces a signaling cascade involving MyD88-IRAK4-TRAF6 (1) to induce CXCL10 and, via IRF7 phosphorylation and translocation, inducing type I and III Interferons (2). Once secreted, type I and III IFNs initiate autocrine and paracrine signals that induce the expression of IFN stimulated genes (ISGs), thereby facilitating an antiviral response that can protect the cell against infection. However, SARS-CoV-2, has the intrinsic property to facilitate CD304 signaling, potentially by interfering with IRF7 nuclear translocation, thereby inhibiting type I IFNα production and thus reducing the antiviral response generated by the pDC (4). Furthermore, the SARS-CoV-2 envelope (E) glycoprotein is sensed by the extracellular TLR2/6 heterodimer and this facilitates production of the inflammatory IL-6 cytokine (5). Illustration was created with BioRender.com The severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) has, since its first appearance in 2019, resulted in a devastating pandemic of coronavirus disease 2019 that prevails mid 2021 (1, 2). The severity of COVID-19 is highly variable between individuals and a great effort is made to understand why some people develop mild disease whilst others 5 require hospitalization (3, 4) . A reported driver of disease severity is the imbalanced induction of an immune response consisting of a broad range of inflammatory cytokines combined with a delayed induction of antiviral interferons (IFNs) (5) (6) (7) . Factors associated with severe disease are inborn errors in the Toll-like receptor (TLR)3 and interferon regulatory factor (IRF)7-dependent type I IFN production and the presence of auto-antibodies against type I IFNs (8, 9) . This indicates 10 that sufficient amounts of IFNs are essential for controlling the infection. Yet, it remains unclear which immune cells detect SARS-CoV-2 and initiate the inflammatory response. Alveolar macrophages seem incapable of sensing SARS-CoV-2 (10) and whether in vitro generated macrophages and myeloid dendritic cells (DCs) are able to elicit production of pro-inflammatory and antiviral cytokines in response to SARS-CoV-2 is currently unclear (11, 12) ; and one study 15 suggests that lung epithelial cells are needed for the macrophages to produce antiviral cytokines (13) . Importantly, lung epithelial cells can however detect SARS-CoV2 and produce type I IFNβ and type III IFNλ1, but only after initiation of virus replication (14, 15) . Plasmacytoid DCs (pDCs) are an autonomous cell type and major producers of type I IFNα, making them pivotal for the human immune system to control viral infections (16) . Clinical studies 20 reveal that severe COVID-19 cases have a reduction in circulating pDCs as well as minimal influx of pDCs into the lungs compared to patients with moderate disease and healthy controls (6, (17) (18) (19) (20) (21) (22) . These severe cases of COVID-19 also exhibited reduced type I IFNα, type III IFNλ and interleukin (IL-)3 levels in plasma, of which IL-3 is known to be important for pDC function (19) . Whether disease severity is due to the lack of pDCs in the lungs or due to dysfunctional cytokine production by the pDCs, remains unclear. Furthermore, the mechanism of how pDCs may sense SARS-CoV-2 has not been resolved. Generally, cytokine production from pDCs is triggered upon the innate detection of viral 5 components via various extra-and intra-cellular receptors also known as pattern recognition receptors (PRRs). In particular, the Toll-like receptors (TLRs) and retinoic acid-inducible gene I (RIG-I)-like receptors (RLR) are the major receptor classes responsible for sensing RNA virus infection and triggering antiviral IFN production (23). In the present study, we explored via which molecular mechanism human pDCs sense SARS-CoV-2, by using a CRISPR-editing approach to 10 screen for several innate immune sensor pathways that are required for the production of antiviral IFNs and inflammatory cytokines upon viral sensing. 15 To investigate the impact of SARS-CoV-2 infection on frequency and phenotype of circulating pDCs, we collected blood samples from patients hospitalized for COVID-19 at hospital admission (day 1) and 5 days after admission. We categorized patients according to symptom duration, which was defined as time from onset of the first self-reported COVID-19 symptom to date of hospital 20 admission (0 to 4; 5 to 8; 9 to 12; and ≥13 days). When we compared the percentage of pDCs out of total PBMCs across the different symptom duration categories, we found a significant lower frequency of circulating pDCs among patients with symptom duration between 5 to 12 days as compared to those with short symptom duration (0 to 4 days) and long symptom duration (≥13 days) ( Figure 1A , EV1A), which was not observed for the myeloid DC subset ( Figure 1B) . We next explored the changes of pDC percentage over time within each patient and observed a significant decrease in pDC frequency and numbers after 5 days of hospitalization ( Figure 1C , EV1B), which was not seen in the myeloid DC subset ( Figure 1D ). Compared to healthy controls 5 (HC), pDC frequency and counts in peripheral blood were decreased in COVID-19 patients ( Figure EV1C-D) . We then evaluated whether the reduction in circulating pDCs was associated with systemic inflammatory cytokine levels. Except for IL-8, the plasma concentration of multiple pro-inflammatory cytokines, in particularly IFNα2a, CXCL10 and IL-6, changed significantly between D1 and D5 ( Figure 1E -J). To determine the association between pDC frequency (in total 10 PBMC) and disease severity, we next categorized patients into i) Hospitalized and no oxygen supplementation required, ii) Hospitalized and nasal oxygen supplementation required, and iii) Hospitalized and high flow oxygen supplementation required. The pDC frequency was significantly lower among patients who required high flow oxygen supplementation as compared to the group that did not require oxygen supplementation ( Figure 1K ) and a correlation between 15 decreased pDC frequency and disease severity was observed ( Figure 1L ). These findings suggest that COVID-19 disease can be associated with a decrease in the percentage of pDCs in peripheral blood and that pDCs are a driver of the inflammatory signals observed during acute SARS-CoV-2 infection. Human pDCs sense SARS-CoV-2 but are refractory to infection. Studying viral sensing by human pDCs is hampered by the limited amount of pDCs that can be obtained from peripheral blood and their incapability to be genetically modified. To overcome this and enable the investigation of potential pDC sensing mechanisms of SARS-CoV-2, we adopted a cellular platform designed to generate human primary pDCs ex vivo using hematopoietic stem and progenitor cells (HSPC) from healthy individuals (Appendix Figure S1 ) (24, 25). The HSPCderived pDCs, produced from multiple healthy donors, were exposed to two different SARS-CoV-2 isolates at 0.1 multiplicity of infection (MOI, as determined by limiting dilution on VeroE6-5 TMPRSS2 cells); the Freiburg isolate (FR2020) which is an early 2020 Wuhan-like strain and the SARS-CoV-2 alpha variant (B.1.1.7). Type I IFNα and CXCL10 production was assessed longitudinally and found to be induced by both variants (Figure 2A -B) with a trend towards a more rapid type I IFNα induction observed for the SARS-CoV-2 alpha variant. To further characterize how pDCs sense SARS-CoV-2, we continued with the SARS-CoV-2 Freiburg isolate. First, pDCs 10 were exposed to TLR agonists or SARS-CoV-2 and after 24hrs cell culture supernatants and pDCs were collected to assess a broader range of inflammatory cytokines, at both the protein and mRNA levels respectively. Following SARS-CoV-2 exposure, we observed increased production of type I IFNα and type III IFNλ1, but not type I IFNβ and type II IFNγ ( Figure 2C -F), resembling the TLR7 agonist response. The cytokines IL-6, IL-8 and CXCL10 were likewise induced, with 15 SARS-CoV-2 inducing higher IL-6 levels than the TLR7 agonist ( Figure 2G -I). Tumor necrosis factor (TNF)α was only marginally induced in pDCs from some donors challenged with SARS-CoV-2 ( Figure 2J ), a pattern resembling neither TLR3 nor TLR7 agonists. Next, we evaluated the type I IFNα expression pattern relative to viral titer and duration of exposure. A viral MOI of 1, resulted in a strong type I IFNα response on both RNA and protein levels ( Figure 2K-L) , and a 20 clear positive correlation between type I IFNα induction and exposure time was observed ( Figure 2M ). A similar pattern was observed for type III IFNλ1 and multiple inflammatory cytokines ( Figure EV2 ). The HSPC-pDC cytokine responses to SARS-CoV-2 were very similar to what was obtained when using freshly isolated human pDCs from peripheral blood ( Figure EV3 ). Next, we assessed whether SARS-CoV-2 was able to replicate in pDCs. However, no viral products indicative of SARS-CoV-2 replication were detected in pDCs (Appendix Figure S2 ), which is supported by others (26). Overall, these results demonstrate that pDCs are capable of sensing SARS-CoV-2 and in response produce type I IFNα and numerous inflammatory cytokines that are 5 important to the cytokine storm observed in people suffering from severe COVID-19 disease (5- Detection of SARS-CoV-2 in pDCs facilitates a protective antiviral response through a broad inflammatory gene signature. 10 A hallmark of antiviral activity is protection of target cells against the pathogen. To investigate if pDC-secreted cytokines protected cells from SARS-CoV-2 infection, we next exploited two different lung epithelial cell types -A549 hACE2 and Calu-3 -and exposed them to cell culture supernatant from pDCs that were either cultured as normal or exposed to SARS-CoV-2, followed by virus inoculation. Pre-treatment with supernatant from SARS-CoV-2-exposed pDCs reduced 15 virus replication in both cell lines in a dose dependent manner ( Figure EV4A -B). Blocking type I IFN signaling enhanced virus replication for all pDC donors tested, but this did not reach statistical significance, indicating that protection was mediated partially by type I IFNs ( Figure EV4C ). Overall, these data indicate that cytokines produced by pDCs in response to SARS-CoV-2 can protect lung cells from infection by reducing virus replication and thereby limit viral spread. To broader investigate the nature and timing of SARS-CoV-2-induced antiviral responses in pDCs we next profiled 789 selected genes covering major immunological pathways (Table S1 ) using the NanoString nCounter technology (27). We profiled the selected genes 4, 24 and 48 hrs after SARS-CoV-2 infection in two individual donors found to be high (D high ) and low (D low ) responders in terms of type I IFNα production (see Figure 2L , where triangles denote D high and squares denote D low ). There was a large overlap between the two donors in gene expression detected above background levels ( Figure EV5A ) and while multiple genes were induced as early as 4 hrs post SARS-CoV-2 exposure, the immunological response seemed stronger after 48 hrs ( Figure 3A -B, 5 Figure EV5B -D, Table S2, Table S3 ). Interestingly, IL-6, CXCL10, CCL2 and CCL8 were among the most upregulated genes in both donors after both 4 and 48 hrs ( Figure EV5E -F). When comparing the expression of the most strongly induced genes (fold change > 2, relative to mock treated cells) after 48 hours of infection (104 genes in D high and 66 genes in D low ) with the earlier time points, it became apparent that distinct gene sets behaved differently ( Figure 3C -D). For 10 instance, some genes peaked at 4 hrs or 24 hrs, some clearly peaked at 48 hrs (cluster 4 Figure 3C and cluster 5 Figure 3D ), while others had a clear biphasic expression format (induced early, disappearing after 24 hrs and then re-induced after 48 hrs; cluster 6 Figure 3C and cluster 2 Figure 3D ). These gene clusters included pathways involved in the pDCs' anti-viral response, and pathways representing more general innate immune activation (Appendix Figure S3 and S4). To 15 confirm the intriguing biphasic gene induction, CXCL10 gene expression was selected for further analysis in multiple pDC donors. RT-qPCR analysis revealed that the wave pattern of gene induction was specific for SARS-CoV-2 sensing and did not follow a TLR7 or TLR3 agonistinduced pattern ( Figure 3E ), confirming our previous observations. Although the NanoString analysis would ideally have been performed using more donors, the data do demonstrate that 20 SARS-CoV-2 activates different steps at different time points in the pDCs' viral sensory pathways, indicating a multifaceted sensory mechanism, where antiviral type I IFNα is found in the early phase, succeeded by excessive inflammatory cytokines at later times. This reflects in part the pathogenesis of COVID-19 (7). SARS-CoV-2 -a single stranded RNA virus -may potentially be sensed by the endosomal TLR-5 MyD88 (Myeloid differentiation primary response 88) pathway (1, 2, 16). To evaluate this in detail, we first generated MyD88 knockout (MyD88 KO ) pDCs using CRISPR/Cas9. As a control, we included cells targeted with CRISPR at the inert (safe-harbor) genomic locus AAVS1 (AAVS1 KO ). MyD88 knockout was confirmed by protein expression ( Figure 4A ), inference of CRISPR edits (ICE) analysis ( Figure 4B ), as well as lack of type I IFNα and cytokine induction in 10 response to TLR7 agonist stimulation (Appendix Figure S5A -E). Of note, knockout of MyD88 did not affect the pDC phenotype (Appendix Figure S5F ). When exposed to SARS-CoV-2, we observed that MyD88 KO pDCs were severely impaired in the induction of CXCL10 and type I IFNα, as compared to the AAVS1 KO control pDCs ( Figure 4C can also bind regions found in secondary RNA structures such as loops and bulges (31). As pDCs have been reported to express TLR3, albeit at lower levels than classical myeloid DCs (29), we 20 next generated and validated pDCs with a TRIF KO and a double TRIF+MyD88 KO (Appendix Figure S6 ). Disrupting TRIF signaling impaired agonist-induced IFNλ1 production in response to TLR3 agonist (Appendix Figure S6D ) but type I IFNα was still produced in response to SARS-CoV-2 exposure ( Figure 4E ). Next, we tested the RIG-I pathway, which is both expressed and further upregulated in pDCs upon TLR stimulation and type I IFN signaling (28, 30). Disrupting RIG-I signaling showed a similar response pattern as observed for the TRIF KO pDCs exposed to SARS-CoV-2, indicating this pathway is not necessary for the sensing of SARS-CoV-2 and subsequent type I IFNα production by pDCs ( Figure 4F , Appendix Figure S6 ). Altogether, our 5 results indicate that pDCs primarily sense SARS-CoV-2 and induce antiviral cytokine production via a MyD88 controlled pathway. The initial observations of knocking out different signal components in RNA sensing pathways 10 prompted us to narrow down the TLR responsible for sensing SARS-CoV-2 and controlling the induction of cytokines. We first selected to generated pDCs with TLR3 KO or TLR7 KO pathways. Disrupting these two pattern recognition receptors ( Figure 5A -C and Appendix S7) clearly demonstrated that TLR3 was not involved in the production of type I IFNα and CXCL10 post sensing of SARS-CoV-2 (Appendix Figure S7 ). However, TLR7 knockout completely abolished 15 type I IFNα and showed a trend toward impaired CXCL10 production in response to SARS-CoV-2 exposure, as compared to AAVS1 KO control pDCs ( Figure 5B -E). Disruption of TLR8, another intracellular viral RNA sensor, with and without TLR7 KO , confirmed that type I IFNα production in response to SARS-CoV-2 was solely driven by TLR7 (Appendix Figure S8A -D). We also explored the effect of inhibition of the Interleukin 1 Receptor Associated Kinase 4 (IRAK4), as it 20 has previously been shown to be important for SARS-CoV-2-induced cytokine induction in pDCs (26). We observed that pDCs treated with IRAK4i prior to viral exposure displayed significantly reduced type I IFNα and CXCL10 protein production, without major effects on cell viability ( Figure 5F and S9). Remarkably, we next observed that SARS-CoV-2-induced IL-6 production was completely unaffected by the disruption of the TLR7 and TLR8 sensing pathway ( Figure 5G Appendix Figure S8E -F), suggesting a parallel endosomal-and viral RNA-independent sensing mechanism. Multiple studies have shown that elevated levels of IL-6 in COVID-19 patients are associated with 5 disease severity (5, 6, 32, 33) and thus we next focused on determining what mechanism was responsible for the IL-6 production by pDCs. As murine bone marrow-derived macrophages and human PBMCs can utilize TLR2 to detect SARS-CoV-2 envelope protein (12) we hypothesized that this TLR could be engaged by human pDCs to sense SARS-CoV-2 and produce IL-6. First, we generated TLR2 KO pDCs ( Figure 6A -B and Appendix Figure S10A ) and observed that 10 disruption of TLR2 did not affect SARS-CoV-2-mediated type I IFNα production ( Figure 6C ), but did significantly impair IL-6 production ( Figure 6D ). Using recombinant glycoproteins of SARS-CoV-2 we next showed that TLR2 sensing and IL6 production was triggered by the envelope protein but not the viral spike protein ( Figure 6E ). However, neither S or E protein induced the production of type I IFNα ( Figure 6F ). Importantly, TLR2 is known to form heterodimers with 15 either TLR1 or TLR6 (34) suggesting that these receptors could also be involved in SARS-CoV-2 envelope protein sensing. Here, TLR6 KO pDCs but not TLR1 KO pDCs displayed a disrupted IL-6 production in response to SARS-CoV-2 exposure (Appendix Figure S10 ), indicating that pDCs produce IL-6 in response to a TLR2 and TLR2/6-mediated sensing of SARS-CoV-2 glycoproteins. These observations were also confirmed in peripheral blood isolated pDCs ( Figure 6G -H). A few papers have suggested that SARS-CoV-2 can bind to neuropilin-1/CD304/BDCA-4 as alternative to ACE2 for viral entry (35, 36) . However, ACE2 is not expressed on pDCs (See Appendix Figure S2E -F) (21, 26), but interestingly; neuropilin-1 is one of the phenotypic markers for pDCs and often highly expressed on these cells. Neuropilin-1 has been reported to have a functional role in pDCs by reducing type I IFNα production (37, 38). Thus, we hypothesize that 5 SARS-CoV-2 engagement with CD304 on the pDCs' cell surface, may interfere with the immunological responses following viral sensing. To test this, we first evaluated if CD304 receptor engagement, would affect IFNα secretion in our experiments. Exposing pDCs to CD304-specific antibody prior to stimulation with TLR7 agonist, reduced production of IFNα in all donors tested, both stem cell-generated pDCs and blood isolated pDCs ( Figure 7A ). Next, we generated 10 CD304 KO pDCs from multiple donors ( Figure 7B ). After exposure to SARS-CoV-2 infection, the CD304 KO cells showed a strong increase in type I IFNα secretion (up 4.5 fold) at multiple time points ( Figure 7C ). This clearly indicates that viral engagement with surface neuropilin-1 on the pDC impairs the type I IFNα production by pDCs. Notably, secretion of both pro-inflammatory cytokines CXCL10 and IL-6 production upon SARS-CoV-2 sensing was unaffected in the 15 CD304 KO pDCs ( Figure 7C -D). This illustrates a novel potential immune evasion strategy of SARS-CoV-2 to reduce the pDCs' type I antiviral IFNα production without affecting the immunopathological pro-inflammatory responses upon infection. Using CRISPR/Cas9-editing of human stem cell-derived pDCs we here demonstrated that pDCs sense SARS-CoV-2 and produce different pro-inflammatory cytokines in response to viral exposure (see graphical abstract). The viral E glycoprotein is recognized by the extracellular TLR2/6 heterodimer, leading to production of the pro-inflammatory IL-6 cytokine. The intracellular TLR7-MyD88-IRAK4 pathway facilitates the production of CXCL10 and antiviral type I IFNα, of which the latter can protect lung epithelial cells from de novo SARS-CoV-2 infection. Removing expression of the NRP1/CD304 receptor from the pDCs' cell surface, alleviates the SARS-CoV-2-induced inhibition on the antiviral response and enhances type I IFNα 5 production, indicating that SARS-CoV-2 utilizes an intrinsic immune evasion strategy that mitigates antiviral IFN production. COVID-19 severity is associated with the excessive production of inflammatory cytokines, also 10 described as a 'cytokine storm', yet which cells produce these cytokines succeeding SARS-CoV-2 infection is not fully understood. Our findings show that pDCs, an immune cell type important for the host defense against many viruses, efficiently detect SARS-CoV-2 by a multi-faceted sensing mechanism and in response produce inflammatory and antiviral cytokines, including type I IFNα and IL-6. Since SARS-CoV-2 emerged, multiple studies have suggested that different cell types as well as diverging sensing pathways to be responsible for the control of the viral infection and the increased levels of inflammatory cytokines observed in patients. One of the challenges by exploring the antiviral response of pDCs is the limited number of cells to collect from blood and the notorious difficulties to genetically manipulate these cells. This can partly be overcome by collecting pDCs 20 from patients with genetic disorders (26) or by studying mice. However, some TLR pathways have been reported to either being nonfunctional or controversial in mice models (39, 40) . In the present study, using a stem cell-based human pDC model in combination with CRISPR technology to knockout multiple TLRs and signaling factors, we demonstrated that TLR7 is critical for the inflammatory signal induced by SARS-CoV-2 infection. Unexpectedly, reduction of the inflammatory cytokine IL-6 was solely dependent on the TLR2 pathway whereas TLR7-MyD88 was responsible for the remaining inflammatory cytokines. Highly pathogenic coronaviruses, similar to other viruses, have multiple strategies to interfere with 5 the host's immune response and efficient immune evasion is associated with pathogenicity (23). Therefore, a detailed understanding of SARS-CoV-2's immune evasion strategies is critical for the development of antiviral therapeutics. Our data indicate that SARS-CoV-2 utilizes neuropilin-1 not only as alternative receptor to ACE2 for viral entry, but also to mitigate the production of type I IFNα by pDCs, thereby reducing the host's innate antiviral immune response. The molecular 10 mechanisms leading to the reduce type I IFNα seen when CD304 is present on pDCs will need further investigations. As pDCs support both the rapid type I IFNα secretion and IL-6 production, this suggests that these cells may have a double-edged function during COVID-19 pathogenesis. Without active pDCs in the lungs, antiviral protection may not be mounted, whereas sustained pDC activation could 15 exacerbate lung inflammation via IL-6 production. Blocking IL-6 responses may not necessarily be successful clinically but therapy with antagonists that specifically impair TLR2, and not TLR7, or therapeutics targeting the viral E glycoprotein could potentially be a scenario to direct immune cells, such as pDCs, to mount a stronger type I and III IFN response that could mitigate disease pathogenesis. In conclusion, our study provides evidence that circulating pDCs could be a potential therapeutic target to maintain desired antiviral IFN levels allowing for the mitigation of COVID-19 severity. HSPC-pDCs were generated as described previously (24, 25). In brief, CD34 + HSPCs were purified from human umbilical cord blood (CB) acquired from healthy donors under informed consent from the Department of Gynecology and Obstetrics, Aarhus University Hospital, Aarhus. (DMEM supplemented with 5% (v/v) hiFCS, 2mM L-glutamine, 100 U/mL penicillin, and 100 µg/mL streptomycin), supplemented with 10 μg/mL blasticidin (Invivogen) to maintain TMPRSS2 expression. All cells were cultured at 37°C and 5% CO2. 15 Air-liquid interface (ALI) epithelium model ALI cells were generated and cultured as described previously (10, 42) . In brief, primary nasal cells were isolated using a nasal brush (Dent-O-Care). Cells were cultured as a monolayer in tissue culture flasks coated with 0.1 mg/mL Bovine type I collagen solution (Sigma-Aldrich). At passage two, cells were seeded at 2-3 × 10^4 cells on 6.5 mm Transwell membranes (Corning) coated with 20 30 μg/mL Bovine type I collagen solution and cultured in 2x P/S (200 U/mL Pen/Strep DMEMlow glycose (Sigma-Aldrich) mixed 1:1 (v/v) with 2x Monolayer medium (Airway Epithelium Cell Basal Medium, PromoCell, supplemented with 2 packs of Airway Epithelial Cell Growth Medium Supplement, PromoCell, without triiodothyronine + 1 mL of 1.5 mg/mL BSA). When cultures reached confluency, Air-liquid interface (ALI) is introduced and medium is changed to ALI medium (Pneumacult ALI medium kit (StemCell) + ALI medium supplement (StemCell) + 100 U/mL Pen/strep) supplemented with 0.48 μg/mL of hydrocortisone (StemCell) and 4 μg/mL heparin (StemCell). Cells were allowed to differentiate for at least 21 days, as verified by extensive 5 cilia beating and mucus covering, prior to experiment initiation. The study population was derived from a cohort of PCR-confirmed hospitalized COVID-19 patients who were enrolled in a clinical trial (43). Individuals for whom there were no 10 cryopreserved peripheral blood mononuclear cells (PBMCs) at baseline, who were pregnant, breastfeeding or had serum total bilirubin x3 above upper limit of normal were excluded from the study. Peripheral blood was collected at time of hospitalization (day 1) and after 5 days. Peripheral blood mononuclear cells (PBMCs) and plasma were separated using ficoll gradient centrifugation, aliquoted, and stored in liquid nitrogen. Self-reported earliest symptom experience was used as Cells were washed with PBS and stored as pellets at -80˚C until further analysis by RT-qPCR. 15 In some experiments, the SARS-CoV-2 envelope (E) protein (ABclonal RP01263) or the SARS-CoV-2 spike (S) protein (ABclonal RP01283LQ) was added to pDCs at a final concentration of 1 μg/mL. The IRAK4 inhibitor (Pf06650833, Sigma-Aldrich PZ0327) was used at a final concentration of 10 μM. VeroE6 cells constitutively produce low level IL-6 independently of SARS-CoV-2 propagation. Thus to discriminate between de novo IL-6 production by pDCs upon 20 SARS-CoV-2 exposure, mock Vero-virus conditions were run in parallel and the IL-6 signal was subtracted from the actual infection samples, to properly determine IL-6 production by pDCs. In some experiments, CD304 targeting antibodies were used. Here, pDCs were incubated with anti-CD304 (Purified anti-human CD304 (Neuropilin-1), clone 12C2, BioLegend Cat#354502) or isotype control (Ultra-LEAF Purified mouse IgG2a, clone MOPC-173, BioLegend Cat#400264) antibody for 15 minutes prior to stimulation with TLR7 (2.5 μg/mL R837) agonist for 4 hrs in 200 uL, after which the culture volume was topped up to 1 mL. The final concentration 5 (after topping up the culture volume) of each antibody was 10 μg/mL. The final concentration (after topping up the culture volume) of each antibody was 10 μg/mL. To determine the amount of infectious virus in cell culture supernatant or generated virus stocks, a limiting dilution assay was performed. 2×10 4 VeroE6-TMPRRS2 cells were seeded in 50 μL DMEM5 in a 96 well plate. The next day, samples were thawed and 5x diluted, followed by 10-5 fold serial dilution using DMEM5, and 50 uL of each dilution was added to the cells. Final dilution range covered 10 -1 -10 -11 in quadruplicate for supernatants or octuplicate for virus stocks. Each well was evaluated for cytopathic effect (CPE) by eye using standard microscopy, and the tissue culture infectious dose 50 (TCID50/mL) was calculated using the Reed and Muench method (44) . To convert to the mean number of plaque forming units (pfu)/mL, the TCID50/mL was multiplied 10 by factor 0.7 (ATCC -Converting TCID[50] to plaque forming units (PFU)). Additionally, cells were fixed by adding 10% Formalin (Sigma-Aldrich) at a 1:1 (v/v) ratio, stained with crystal violet solution (Sigma-Aldrich) and stored at room temperature. Reverse transcriptase-quantitative PCR (RT-qPCR) 15 To determine expression levels of the human IFNa2a, TNFa, CXCL10, IFNL1, GAPDH, ACE2, TMPRSS2 and SARS-CoV-2 N1 gene, RNA was purified from cells using the RNeasy mini kit Technology, cat#4283) was used. For TRIF, r-a-hTRIF (98 kDa, Cell Signaling Technology, cat#4596) was used, and for RIG-I, r-a-hRIG-I (clone D14G6, 102 kDa, Cell Signaling Technology, cat#3743) was used. Each membrane was re-used to for the loading control vinculin (mouse-a-hVCL, clone hVIN-1, 116 kDa, Sigma-Aldrich, cat#V9131). As secondary antibodies, peroxidase-conjugated donkey-anti-rabbit and donkey-anti-mouse was used (Jackson Immuno Research 711-036-152 and 715-036-150). To perform broad transcriptomic profiling on SARS-CoV-2 exposed HSPC-pDCs, an nCounter NanoString analysis was performed (NanoString Technologies, Seattle, WA, USA). HSPC-pDCs 10 from two donors were exposed to SARS-CoV-2 at 1 MOI (for 4, 24 or 48 hrs and mock treated samples at 4 and 48 hrs), after which cell pellets were collected and RNA extracted with the RNeasy mini kit (Qiagen). 30 ng of RNA was used as input for the analysis using the nCounter SPRINT Profiler (NanoString Technologies) and the nCounter PanCancer Immune Profiling Panel (cat# XT-CSO-HIP1-12) plus a custom made PanelPlus of the following genes: NFE2L2, 15 TMEM173, MB21D1, IFNLR1, IRF9, IFNL3, IFNL4, AIM2, TREX1, ENPP1, PCBP1, PQBP1, G3BP1, STIM1, LRRC8A, SLC19A1, NLRC3, NLRX1, ZDHHC1, TRIM56, TRIM32, RNF5, ULK1, TTLL4, TTLL6, AGBL5, To assign pathways to the gene clusters identified in pDCs from D high and D low 48 hrs after SARS-CoV-2 exposure using unsupervised hierarchical cluster analysis on the NanoString nCounter data, we utilized the Reactome Pathway Browser version 3.7, database release 75 (https://reactome.org/PathwayBrowser); a comprehensive web-based resource for curated human 15 pathways. Disease pathways were excluded from the analyses and we used UniProt as the source of entities (maximum pathway size was 400). Only six genes were not assigned to any pathways in Reactome. Reactome defines statistically significantly enriched pathways using a Binomial Test, followed by correction for multiple comparisons by the Benjamini-Hochberg approach (45). Differences between experimental conditions were analyzed using the ratio paired student T test with GraphPad Prism (Version 6). P-values ≤0.05 were considered significant: *p<0.05, **p<0.01, ***p<0.001. To determine correlation between IFNα production by pDCs and time of exposure to SARS-CoV-2, as well as to compare gene expression changes in D high and D low after 4 and 48 hrs after exposure to SARS-CoV-2, and determine correlation between pDC frequency and disease severity, simple linear regression analysis were performed using GraphPad Prism. The R squared and p-value are indicated in the figures. pDCs were either mock treated or exposed to the SARS-CoV-2 FR2020 early Wuhan-like strain or the SARS-CoV-2 alpha variant B.1.1.7 (0.1 MOI). Supernatants were collected at indicated time points and the production of type I IFNα (A) and CXCL10 (B) was quantified. The FR2020 strain was used in subsequent experiments where pDC were either mock treated (mock, grey), exposed to SARS-CoV-2 at 1 MOI (SARS-2, purple), TLR7 (2.5 μg/mL R837, blue) or TLR3 agonist (800 ng/mL poly(I:C), pink). Supernatants were collected after 24 hrs and analyzed for type I IFNα (C), (ICE) analysis (B). MyD88 KO and control pDCs were either mock treated (mock) or exposed to SARS-CoV-2 (SARS-2, 1 MOI), supernatants were collected at indicated time points and analyzed for CXCL10 (C) and type I IFNα (D). Type I IFNα production was then determined in cell culture supernatant from SARS-CoV-2 exposed TRIF KO or TRIG+MyD88 KO (E) and RIG-I KO or RIG- protein expression by ELISA. AAVS1 KO and TLR7 KO pDCs were either mock treated (mock) or exposed to SARS-CoV-2 (SARS-2, 1 MOI), supernatants were collected at indicated time points and analyzed for type I IFNα (D) and CXCL10 (E) proteins. Wild type pDCs were exposed to SARS-CoV-2 (0.5 MOI) in the absence or presence of an IRAK4 inhibitor (10 μM), 24 hrs after virus exposure the cell culture supernatants were analyzed for production of type I IFNα, CXCL10 and IL-6 proteins (F). IL-6 protein quantification in AAVS1 KO and TLR7 KO pDCs after SARS- grey), exposed to SARS-CoV-2 (SARS-2, 0.5 MOI, purple) or TLR7/8 agonist (2.5 μg/mL R848, red) and supernatants were collected after 24 hrs to quantify type I IFNα (C) and IL-6 (D) protein concentrations. To investigate if pDCs sensed the spike or envelope SARS-CoV-2 proteins, AAVS1 KO and TLR2 KO pDCs were exposed to SARS-CoV-2 (SARS-2, 0.5 MOI, purple) recombinant SARS-CoV-2 spike (S, 1 μg/mL, dark green) or envelope (E, 1 μg/mL, light green) 10 proteins and IL-6 (E) and type I IFNα (F) protein concentrations were quantified. Peripheral blood pDCs were isolated from PBMCs by negative selection and exposed to SARS-CoV-2 (1 MOI, purple), TLR7 agonist (2.5 μg/mL R837, blue) agonist, or E protein (1 μg/mL, light green) for 24 hrs and the concentration of IL-6 (G) and type I IFNα (H) was quantified in cell culture supernatants by ELISA. Bars represent mean values and equal symbols represent equal donors 15 (n=4). Statistical significance was determined using the ratio paired student T test for agonist or virus treated cells and compared to the mock treated condition, or by unpaired T test when comparing matched conditions between different KOs. *