key: cord-0753995-63ert5iz
authors: Mazzarella, Luca; Santoro, Fabio; Ravasio, Roberto; Massa, Paul E.; Rodighiero, Simona; Gavilán, Elena; Romanenghi, Mauro; Duso, Bruno Achutti; Bonetti, Emanuele; Pallavi, Rani; Trastulli, Deborah; Pallavicini, Isabella; Gentile, Claudia; Leonardi, Tommaso; Pasqualato, Sebastiano; Buttinelli, Gabriele; Martino, Angela Di; Fedele, Giorgio; Schiavoni, Ilaria; Stefanelli, Paola; Meroni, Giuseppe; Steinkuhler, Christian; Fossati, Gianluca; Minucci, Saverio; Pelicci, Pier Giuseppe
title: Inhibiting LSD1 suppresses coronavirus-induced inflammation but spares innate antiviral activity
date: 2021-05-03
journal: bioRxiv
DOI: 10.1101/2021.05.02.441948
sha: b66e7199ccad46462a6a20d76dadd945a56ee8f0
doc_id: 753995
cord_uid: 63ert5iz

Tissue-resident macrophages exert critical but conflicting effects on the progression of coronavirus infections by secreting both anti-viral type I Interferons and tissue-damaging inflammatory cytokines. Steroids, the only class of host-targeting drugs approved for Covid19, indiscriminately suppress both responses, possibly impairing viral clearance, and provide limited clinical benefit. Here we set up a mouse in vitro co-culture system that reproduces the macrophage response to SARS-CoV2 seen in patients and allows quantitation of inflammatory and antiviral activities. We show that the NFKB-dependent inflammatory response can be selectively inhibited by ablating the lysine-demethylase LSD1, which additionally unleashed interferon-independent ISG activation and blocked viral egress through the lysosomal pathway. These results provide a rationale for repurposing LSD1 inhibitors, a class of drugs extensively studied in oncology, for Covid-19 treatment. One-Sentence Summary Targeting a chromatin-modifying enzyme in coronavirus infections curbs tissue-damage without affecting antiviral response

Abstract: Tissue-resident macrophages exert critical but conflicting effects on the progression of coronavirus infections by secreting both anti-viral type I Interferons and tissue-damaging inflammatory cytokines. Steroids, the only class of host-targeting drugs approved for Covid19, indiscriminately suppress both responses, possibly impairing viral clearance, and provide limited 5 clinical benefit. Here we set up a mouse in vitro co-culture system that reproduces the macrophage response to SARS-CoV2 seen in patients and allows quantitation of inflammatory and antiviral activities. We show that the NFKB-dependent inflammatory response can be selectively inhibited by ablating the lysine-demethylase LSD1, which additionally unleashed interferon-independent ISG activation and blocked viral egress through the lysosomal pathway. 10 These results provide a rationale for repurposing LSD1 inhibitors, a class of drugs extensively studied in oncology, for Covid-19 treatment.

Targeting a chromatin-modifying enzyme in coronavirus infections curbs tissue-damage without 15 affecting antiviral response

Increasing evidence suggests that the severity of Covid-19 pathology is dictated by the profile of immune mediators produced in response to SARS-CoV2 infection. High levels of inflammatory cytokines typical of the innate immune response, such as IL6, TNFa and IL1, are associated with severe disease, whereas expression of molecules with direct antiviral activity, such as type I 5 interferon and its downstream targets (Interferon-Stimulated Genes, ISGs) are associated with better outcome [1] [2] [3] . Key transcriptional regulators of these innate responses to pathogens include Nuclear Factor-Kappa B (NF-κB), which regulates transcription of proinflammatory cytokines, and the family of Interferon-Regulated Factors (IRFs), which promote expression of Interferons and ISGs 4 . These responses are structured in self-amplifying paracrine loops where downstream 10 targets (e.g. IL1β and TNFa for NF-κB; type I Interferons for IRFs) are themselves activators of the same pathway, a logic architecture that allows the signal to propagate in space and time to rapidly arrest infection spread in the involved organ 5, 6 . Unbridled loop activation, however, can lead to extensive local or systemic damage, further amplified by the recruitment of additional effector cells like neutrophils or lymphocytes. 15 The initiation and duration of the innate transcriptional responses is further regulated by multiple cofactors, among which chromatin-modifying enzymes are thought to play a key role 7 . The Lysine Demethylase 1 (LSD1; also known as KDM1A), in particular, has been implicated in both NF-κB and IRF pathways and is pharmacologically targetable. Multiple LSD1 inhibitors have completed initial phases of clinical development for oncological indications with acceptable 20 toxicity profiles 8 and may be amenable to repurposing. In a mouse system of Lipopolysaccharide (LPS)-mediated NF-κB activation using bone marrow-derived macrophages (BMDM), genetic ablation of Lsd1 was found to be associated with decreased nuclear translocation of NF-ΚB and reduced binding to its target promoters, resulting in decreased transcription of proinflammatory cytokines and improved survival in a sepsis mouse model 9 . On the other hand, LSD1 has also been implicated in suppressing endogenous retroviral elements (ERVs) in mouse melanoma cells. Loss of LSD1 induced up-regulation of ERV expression with consequent accumulation of double-stranded RNA (dsRNA), recognition by the dsRNA-sensing 5 pathway MDA5-IRF3 and activation of an interferon-mediated response 10 .

Transcription factor dynamics in human and murine coronavirus infections have been mostly assessed using epithelial or fibroblast cell lines 11, 12 . Although epithelia and connective tissue are sites of active involvement in the pathogenesis of coronavirus infections, increasing evidence points to a central role for cells of the monocyte-macrophage compartment, which are strongly 10 modulated during Covid-19 in relation to disease severity [13] [14] [15] . These cells are equipped with a vast array of pathogen-sensing mechanisms and play a key role in the generation and amplification of inter-cellular signaling loops in innate immunity [16] [17] [18] . Notably, alveolar macrophages from Covid-19 patients or SARS-CoV2-infected African Green Monkeys contain viral RNA in significant amounts, including negative strand sequences indicative of active 15 replication 19, 20 . However, the study of the molecular mechanisms involved in macrophage responses to human SARS-CoV2 has been hampered by the lack of in vitro models of productive infection 21 , as also observed for SARS-CoV1 22,23 , suggesting that virus uptake by macrophages is mediated by elements of the in vivo environment that are missing in vitro, such as nonreceptor-mediated entry like antibody-mediated endocytosis or phagocytosis of infected cells 20 20, 24 . In this study we investigated the dynamics of NF-κB and IRF activation in macrophage coronavirus infections and the potential role of LSD1.

As model system, we used mono-or co-cultures of mouse macrophages and fibroblasts or lungepithelial cells infected with the Murine Hepatitis Virus (MHV), strain A59. MHV is a beta- 5 coronavirus phylogenetically close to human SARS-CoV1/2 4 and able to produce a disease highly similar to that triggered by human SARS-CoV 25,26 . MHV entry is mediated by the murine CEACAM1 receptor, which is highly expressed in mouse macrophages, both in vivo and in vitro 27 ECT and EAV can be biochemically separated, we applied progressive fractions of snBMDM obtained by size-exclusion chromatography to infected or uninfected L929 GFP cells and measured cell viability 48 hours post infection (hpi). As shown in figure 1D and F, the two activities were enriched in two clearly separated fractions, with ECT peaking in fractions 21-25 (corresponding to a predicted protein size of 50-60 kDa) and EAV peaking in fraction 27-31 (corresponding to a predicted protein size of ~20 kDa). Previous studies identified TNFa and type I interferon as 5 candidates for coronavirus-induced cytotoxic and antiviral activities respectively 25, 29 . ELISA clusters containing genes whose MHV-dependent upregulation is inhibited by DDP (clusters A1-3). These clusters were enriched for proinflammatory cytokines and NF-ΚB binding motifs (cluster A1), Interferon-stimulated genes (ISG) and IRF binding motifs (cluster A2) and genes 10 involved in granule formation and Sp1 binding motifs (cluster A3) (figure 3B). Cluster A1 included all cytokines currently implicated in the severe form of Covid-19 (Il1a, I1b, Il6, Tnf; supplementary table 3) and exhibited the strongest quantitative changes: it was the most highly upregulated in response to MHV (average of ~8 fold) and the most downregulated in response to DDP (reaching baseline levels at 10 µM). The impact of DDP on IRF-associated genes in cluster 15 A2 was significantly less pronounced: the extent of the MHV-induced upregulations in A2 was comparable to A1, but even with DDP 10 µM A2 cluster genes remained on average upregulated ~5 fold (figure 3C). In further support of a direct activity of DDP on NF-ΚB, we observed extensive overlaps between genes of the A1-A4 clusters and the reported LPS-induced and LSD1-regulated NF-κB targets 9 (supplementary figure 5B). 20 Gene expression analyses by qRT-PCR of representative NF-κB-dependent cytokines (Il1b, TNFa, Il6) showed rapid induction at 8-12 hpi and strong down-regulation by DDP ( figure 3D ).

In turn, the impact of DDP on the expression of IFNa and representative ISGs (Isg15 and Ifit1) was significantly less pronounced (figure 3E). Notably, IFNa transcription was detected after that of Isg15 and Ifit1, suggesting an Interferon-independent initiation of ISG expression, a phenomenon previously reported in other viral infection systems 31 .

Endogenous Retroviral Elements (ERVs), which lead to increased endogenous dsRNA and Activation of both NF-ΚB and IRFs is associated with nuclear localization 5, 33 . Thus, we investigated levels of nuclear NF-ΚB and IRFs by biochemical fractionation and immunofluorescence. Western blotting analyses of nuclear/cytoplasmic fractions showed rapid 15 nuclear re-localization of NF-ΚB after MHV-infection, as expected, already evident at 10 hpi (figure 6A-B). Same analysis for all 9 IRFs (supplementary figure 7A) showed upregulation and nuclear localization of IRF1 and, to a lesser extent, IRF2, whereas the other IRFs showed either no change (IRF5,9) or even decreased nuclear levels (IRF 3,4,6,7,8). Notable is the suppression of IRF3 and IRF7, typically involved in the response to dsRNA and ssRNA, thus confirming the 20 existence of cellular-evasive mechanisms specific to coronaviruses 12,34,35 . Results obtained by biochemical fractionation were confirmed by immunofluorescence analyses using anti-NF-ΚB and -IRF1 (figure 4C,D) antibodies, which showed nuclear localization of both factors upon MHV-infection. Notably, differential analyses of infected or uninfected cells (the former identified by NSP9 positivity) showed nuclear localization of both NF-ΚB and IRF1 also in NSP9-negative cells, suggesting activation of paracrine loops (figure 4C-D). Treatment of MHVinfected cells with DDP abrogated nuclear NF-ΚB but had a much less pronounced effect on IRF1, resulting in barely any difference at 24 hpi in fractionation experiments (Fig.4B ) and a 5 nuclear signal persistently above the baseline by immunofluorescence, both qualitatively and quantitatively ( fig.4C,D) . Chromatin Immunoprecipitation (ChIP) showed that MHV infection induced binding of both NF-ΚB and LSD1 at the promoter of NF-ΚB target genes, which was abrogated by DDP ( figure 4E ).

Collectively, these results suggest that LSD1 is specifically implicated in regulating MHV-10 induced NF-ΚB nuclear relocalization and NFKB-dependent inflammatory cytokine production, leaving relatively unaltered the IRF1 response and the ensuing Interferon secretion.

In addition to the secreted (extrinsic) antiviral activity induced by DDP in MHV-infected 15 BMDMs, we explored whether LSD1 inhibition also exerts a direct (intrinsic) antiviral activity, by measuring survival, titer and syncytia formation in L929, BMDMs and LA4 cells.

In all three cell types, treatment with DDP increased cell survival in a dose-dependent manner Together, these data suggest that LSD1 ablation allows the emergence of a cell-intrinsic antiviral response characterized by interferon-independent ISG activation and restoration of lysosomal acidification, resulting in reduced viral release. The precise point of regulation of this 5 activity remains to be elucidated. Dexamethasone is the only host-targeting medical treatment approved to date for Covid-19 as it moderately improves survival and effectively dampens the NF-κB-dependent response, but is also suspected to suppress interferon activity thus resulting in prolonged viral shedding 37-39 . All 15 three compounds inhibited the macrophage-secreted cytotoxic activity in a dose-dependent manner, as measured by restoration of cell viability of non-infected L929 cells exposed to BMDM-conditioned supernatants ( figure 6A ). When the same supernatants were added to infected L929 cells, even at the highest concentrations, neither of the two LSD1 inhibitors decreased the macrophage-secreted antiviral activity, as judged by restoration of cell viability 20 and lack of syncytia formation. Dexamethasone, instead, at cytotoxic-suppressing dosages failed to recover viability and to prevent syncytia (10 µM, figure 6B ,D). Similar results were obtained when the three drugs were applied directly on infected L929 cells: both DDP and Ory1001, but not dexamethasone, elicited intrinsic antiviral activity (figure 6C-D).

Then, we tested pharmacological interactions with type I Interferon, relevant for mechanistic understanding and potential pharmacological combinations. IFNa alone dose-dependently inhibited MHV activity and viral titer, and no synergism nor antagonism could be identified if 5 cells were co-treated with increasing doses of DDP (supplementary figure 10A,B) . Thus, although high-dose DDP does indeed reduce Interferon production by macrophages (figure 2I), it does not interfere with downstream interferon-dependent signaling.

As mentioned above, the possibility to directly study the response of human macrophages in vitro to SARS-CoV2 remains challenging. In our hands, as reported by others 21 , attempts to 10 infect monocyte-derived macrophages in vitro have been unsuccessful (supplementary figure   10C ). Thus, we took advantage of the transcriptomic data from broncho-alveolar lavage of in regulating innate responses to coronaviruses is depicted in figure 6G .

Based on our findings we propose LSD1 inhibition as a promising strategy to prevent or treat Covid-19, with potential advantages over the only host-targeting drug class approved for Covid- 5 19 to date, steroids, given the selective inhibition on NF-kB-dependent secretion of tissuedamaging proinflammatory with relative sparing or even enhancement of the Interferon response.

The viability of this strategy is further supported by findings in murine models of SARS-CoV respiratory diseases, in which direct pharmacological inhibition of NF-ΚB or generation of viral mutants inducing attenuated NF-ΚB activation led to decreased disease severity and prolonged 10 survival 40 . In agreement with our findings, the GSK inhibitor GSK-LSD1 was recently shown to reduce levels of proinflammatory cytokines in peripheral blood leukocytes isolated from severe COVID-19 patients 41 , although effect on tissue-resident macrophages and sparing of Interferon response were not shown. Oryzon has initiated the ESCAPE study with Vafidemstat in nonsevere Covid-19 (Eudract 2020-001618-39) but results are not yet known. 15 The need for cooperation between NF-κB and IRF factors in the so-called "enhanceosome" that regulates type I interferon expression 42,43 may provide a mechanistic explanation for the differential activity of steroids vs LSD1 inhibitors: whereas steroids inhibit NF-ΚB by preventing NF-ΚB nuclear translocation through upregulation of IkB 44 , LSD1 has been proposed to suppress the degradation of nuclear-translocated NF-ΚB 9 . Thus, LSD1 inhibition may allow 20 minute amounts of NF-ΚB to enter the nucleus for a restricted time window, allowing it to act as a "pioneer" transcription factor to render chromatin of the IRF1 target loci accessible for subsequent binding of canonical transcriptional activators 45 . Further research is required to elucidate if histone demethylase activity is required for the effect of LSD1in the NF-κB response, and if main targets are NF-ΚB itself or critical effector proteins encoded by NF-κB target-genes.

An intriguing finding is the existence of an LSD1-dependent, interferon-independent cellintrinsic response that results in the activation of a subset of ISGs and inhibition of lysosomal-5 mediated virus egress. The precise nature of this response, which may contribute to the initial, Interferon-independent trigger of IRF1, requires further investigation and may involve direct histone demethylation at regulatory elements of ISGs and/or genes involved in lysosomal activity. The finding that IRF1 and to a lesser extent IRF2 are the only IRFs that clearly change their intracellular localization in response to MHV infection is perhaps surprising, given the 10 widespread involvement of other IRF members in the antiviral response to many viruses.

However, although generally considered dispensable for the activation of antiviral responses 33 , IRF1 can lead to early activation of ISGs and type I Interferon, independently from other IRFs that typically occur as a later event, such as dsRNA-MDA5/RIG-I activated IRF3 31,46,47 . In the absence of other viral-induced effects, this would lead to Interferon-loop amplification, 15 inhibition of TNFa signaling or IL1b production and attenuation of NF-κB signaling 48,49 .

However, when IRF3-dependent amplification of type I Interferon is actively inhibited by coronavirus-specific mechanisms, cross-inhibition of the NF-κB loop is ineffective, thus favoring inflammation over Interferon-induced viral clearance as observed in severe Covid19 cases. Thus, sustaining interferon response while suppressing NF-kB through LSD1 inhibition may interrupt 20 the propagation of the proinflammatory loop and allow viral clearance with minimal tissue damage, providing clinical benefit even in earlier disease stages in which steroids seem to be ineffective or even detrimental.

Data and materials availability: All unique/stable reagents generated in this study are available from the corresponding authors with a completed Materials Transfer Agreement. 5 The RNAseq dataset (fastq files, raw and normalized counts and differentially expressed genes) generated during this study are available at GEO GSE169399

The full sequence of the MHV-A59 strain used in the present study is deposited at NCBI (submission initiated, will be finalized upon acceptance) 10 Viral stock. MHV strain A59 was kindly provided by Dr. Riccardo Villa at the IZSLER and propagated on L929 cells. Briefly, L929 cells were plated the day before infection at a density of 20 million cells in T175 flask. Cells were infected at MOI: 0,5 in 10 ml of free serum DMEM. After 1 hour incubation at 37 ˚C 40ml of DMEM supplemented with 3% FBS were added. Supernatant 15 containing virus was harvested when the virus-induced cytopathic effect was visible on more than 70% of cells, usually 36 hours after infection. The identity of the virus was confirmed by Illumina sequencing (see supplementary table 1) . For SARS-CoV2 experiments, a strain isolated from an Italian patient in February 2020 50 was used in the Biosafety level 3 facility of the Istituto Superiore di Sanita' in Rome. 20 The virus was propagated in Vero E6 cells cultured at 37 °C in 5% CO2 in minimal essential medium (MEM) supplemented with 10% fetal calf serum (FCS), 1% l-glutamine, and 1.4% sodium bicarbonate. Virus-infected cells were maintained at 37 °C in 5% CO2 in MEM supplemented with 2% FCS. Titers were measured by CCID50 system in Vero E6 cell. Briefly, samples were serially diluted 1/10 in medium. Then 100 µL of each dilution was plated into ten 25 wells of 96-well plates containing 80-90% confluent cells. The plates were incubated at 37 °C under 5% carbon dioxide for five days. Each well was then scored for the presence or absence of the virus. The limiting dilution end point (CCID50/ml) was determined by the Kärber equation.

Cell culture. 30 Raw264.7, L929, LA4, Calu3 and VeroE6 cell lines were purchased by American Type Culture Collection and grown according to ATCC recommendations. Cultures were maintained in a humidified tissue culture incubator at 37°C in 5% CO2. To assure mycoplasma-free conditions, all cells were routinely tested. BMDM were obtained from bone marrow of 6-10 week old female C57Bl6 mice (Charles River). 35 10 6 cells were plated in 10 cm untreated cell culture dishes, resuspended in 8 ml of Alpha MEM containing 20% FBS, 2 mM of L-glutamine, antibiotics, 40ng/ml of rm M-CSF (R&D System) and allowed to differentiate for 7 days. Human monocytes were purified from peripheral blood collected from healthy blood donors at the Centro Trasfusionale Policlinico Umberto I, University La Sapienza blood bank (Rome, Italy) 40 using Ficoll gradients (lympholyte-H; Cedarlane). CD14 cells were purified by anti-CD14 monoclonal antibody (mAb)-conjugated magnetic microbeads (Miltenyi Biotec) and then cultured for 6 days in RPMI 1640 medium (Life Technologies Invitrogen), supplemented with heatinactivated 10% lipopolysaccharide-free FBS, 1 mM sodium pyruvate, 0.1 mM nonessential amino acids, 2 mM L-glutamine, 25 mM HEPES, 100 U/ml penicillin, 100 µg/ml streptomycin (all from 45 EuroClone) in the presence of human recombinant M-CSF (100 ng/µL; Peprotech). Blood donors provided written informed consent for the collection of samples and subsequent analysis. Blood samples were processed anonymously.

L929 cells were plated at a density of 5X10 4 in 24 well plate a day before the infection. A spin 5 infection was used to infect L929 cells using a lentiviral vector carrying the H2B-GFP transgene 51 . Briefly, concentrated H2B-GFP lentivirus (MOI 4) was added to L929 cell in a 24-well nontissue culture-treated plate and centrifuge at 750 × g, 25 °C for 1 h. Infected cells were incubated for 3 h at 37 °C and replaced with fresh culture media (1 mL/well). After a recovery period of 2 days, GFP+ cells were sorted by fluorescence-activated cell sorting (FACS) and maintained in the 10 culture for further experiments.

In vitro pharmacological treatments. DDP38003 and Oryzon 1001 were synthesized as described in 30 . Dexamethasone (Sigma) was dissolved in DMSO. 15 PolyIC was purchased from Cytiva and dissolved in PBS to a concentration of 1mg/ml. LPS was purchased from Sigma and dissolved in water to a concentration of 1mg/ml.

BMDM were plated at a density of 500.000 cells per well in untreated cell culture 6 well plates, in 20 a total volume of 2ml. After one night incubation, BMDM were treated with the indicated compounds. 24 hours later the cells were infected with MHV at the desired MOI, by replacing the overnight medium with DMEM 3% FCS containing MHV. After 1 hour, the viral inoculum was removed and replaced with BMDM medium supplemented with the drugs. 24 hours later, supernatant was harvested, clarified by centrifugation and used for subsequent experiments.

500ul of BMDM supernatant were aliquoted in one 24 well and incubated on ice for 1 minute. UV inactivation was performed on ice, using Agilent Genomics/Stratagene Stratalinker 2400 UV Crosslinker, by delivering an energy dose equivalent to 0,3 Joules. 30 Antiviral and cytotoxicity assay. L929 cells were plated at a density of 5000/well in 96 well plate, in a total volume of 100 µl. The day after, the culture medium was removed and replaced with 50 µl of the serially diluted UV inactivated BMDM supernatant. For TNFα and IFNAR neutralization assay, 50 µl of the serially 35 diluted UV inactivated BMDM supernatants were prior treated for 30 minutes with different concentrations of anti-TNFα or anti-IFNAR and then added to the L929 cells as described above. For JAKi experiments, L929 cells were prior treated for 30 minutes with the specific inhibitor and then exposed to 50ul of the serially diluted UV inactivated BMDM supernatant, as described above. After 1-hour incubation -to test BMDM supernatant antiviral activity-cells were infected 40 by adding 50ul of virus, at the desired MOI, in DMEM supplemented with 3% FCS, or left uninfected -to test the BMDM supernatant cytotoxic effect-by adding 50 µl of DMEM supplemented with 3% FCS without virus. 48 hours later, the vitality of the cells was evaluated by CellTiter-Glo luminescent cell viability assays (Promega, Madison, WI, USA), following the manufacturer's instructions. 45 

Titration for MHV and SARS-CoV2 was performed using the TCID50 method, with some specifications. Briefly, supernatants from infected cells were 10-fold serially diluted and titrated on target cells plated at 80-90% confluence (for MHV, L929 cells; for SARS-CoV2, VeroE6 cells) in 96 wells in a total volume of 100 µl, (8 replicate wells for MHV, 10 replicate wells for SARS-CoV2) and incubated at 37 °C under 5% carbon dioxide. After a defined time period (2 days for 5 MHV, 5 days for SARS-CoV2), each well was scored for the presence of virus-induced cytopathic effects. The limiting dilution end point (TCID50) was determined using the Reed-Muench method for MHV and the Kärber equation for SARS-CoV2.

. 10 The sgRNA oligo (see Reagents table) 

To knock down LSD1, one short hairpin RNA (shRNA) sequences was tested. shLSD1#1 (see reagents table for sequence) was cloned into the pLKO vector by AgeI-Eco RI double digestion. 20 The plasmid was used to produce lentiviral particles in 293T cells. Live cell imaging. L929 (20k cells), BMDM (5k or 10k cells), BMDM:L929 1:1 (12.5k:12.5k cells) and BMDM:L929 2:1 (20k:10k cells) cocultures were seeded on a 96-wells plate at day 0, treated at 30 day1, infected with MHV 0.1MOI at day 2 for 1h, washed and kept in fresh medium plus treatments and 0.4ug/ml Propidium Iodide (PI) for the total duration of the time-lapse experiment. GFP, PI and bright field images were acquired on a Nikon Eclipse Ti microscope (Nikon Instruments S.p.A., Firenze, Italy) equipped with a xyz motorized stage, a Spectra X light engine (Lumencor, Beaverton, OR, USA), a Zyla 4.2 sCMOS camera (Andor, Oxford 35 Instruments plc, Tubney Woods, Abingdon, Oxon OX13 5QX, UK) a multi-dichroic mirror and single emission filters (Semrock, Rochester, New York, USA). Large images made by four partially overlapping (2%) fields of view (FOV) were acquired for each well with a 2x2 camera binning every hour from 4 hpi to 48 hpi using a 10x, 0.3 NA objective lens. Temperature, CO2 and humidity was controlled by a microscope cage incubator (Okolab, Napoli, Italy). 40 Bone marrow derived macrophages or L929 cells were seeded on a 12 glass-bottom wells (MatTek Corporation, Ashland, MA 01721, USA), coated with poly-D-Lysine 0.1% (w/v) in water (3x105 cells/well). After 24 hours cells were treated with DDP (10 µM) or DMSO and then infected with MHV for 12 hours. Cells were incubated with 1 µM of Lysosensor Green DND-189 (Thermo Fisher Scientific, Monza, Italy) for 90 min. and Hoechst 33342 (Euroclone 45 S.p.A., Pero, Italy) for the last 30 min. Cells were washed and fresh medium was added. Labelled live cells were imaged at 37 °C and 5% CO2 on a Leica Thunder Imager system (Leica Microsystems GmbH, Wetzlar, Germany), equipped with a xy motorized stage, 5 LED sources, a DFC9000 GTC sCMOS camera, a multi-dichroic mirror and 4 emission filters. Ninety-nine images were acquired for each condition using a 63x 1.4NA oil immersion objective lens.

Cells were seeded on Poly-D-lysine coated slides (10 5 cells/slide). After treatments cells were 5 fixed with methanol at -20°C for 6 min, blocked with 5% donkey serum for 60 min. Slides were stained with primary antibodies diluted in 1% BSA in PBS (NF-kB and IRF1 1:400, NSP9 1:1000) for 90 min. Secondary anti-rabbit (A488) and anti-mouse (Cy3) were used at 1:200 for 1 hour. Nuclei were stained with DAPI 1:1000 for 20 min. Mowiol was used as mounting solution. Cells labelled with DAPI and anti-NF-kB/NSP9 or anti-IRF1/NSP9 antibodies were imaged with 10 a 60x 1.4 NA oil immersion objective lens on a CSU-W1 Yokogawa Spinning Disk confocal system with a 50 µm pinhole disk mounted on an Eclipse Ti2 stative and equipped with a motorized xyz stage, 6 solid state lasers, a multi-dichroic mirror, single emission filters and a Prime BSI sCMOS camera (TELEDYNE PHOTOMETRICS, Tucson, AZ 85706, USA). Hundred FOV per condition were automatically acquired thanks to the JOBS application of the 15 NIS software (Nikon). Briefly, for each of the 100 positions defined in the JOB, an autofocus routine using the DAPI channel was used to define the acquisition focal plane and the 3 channels corresponding to the 3 stainings were acquired.

Bioinformatic analysis of imaging data: 20 

Time-lapse raw images were first corrected for uneven illumination (shading correction) and background (BG) variation in time thanks to the BaSiC Fiji/ImageJ plugin 53 . The corrections were done in batch using a custom made ImageJ macro. Briefly, for each 4-channel .nd2 time series, the channels were split, BaSiC-corrected and saved as tiff sequence in a new folder. 25 A custom-made ImageJ macro was used to identify syncytia formed by infected L929 GFP cells. Briefly, for each condition and in batch, the GFP time series was BG-corrected using a rolling ball radius of 50 pixels, filtered with a median filter (radius=2 pixels), then objects were identified using a threshold defined by the Otsu method, the objects with a size bigger than 10 um 2 were identified by the Analyze Particles ImageJ function and the region of interest (ROI) 30 added to the ROI manager. In each Field of View (FOV), the list of ROI was used to calculate the ROI area and the intensity of the PI channel. The mean area of single nuclei was calculated from the control condition (DMSO, not infected) at the first time point and the mean nuclei area resulted about 150 um 2 . Syncytia were arbitrarily considered as the union of at least 5 nuclei, using an object size threshold of 750 um 2 . This 35 threshold was visually confirmed to accurately capture the majority of syncytia. To calculate the death events of L929 cells, the mean PI intensity in each ROI obtained from the segmentation of the GFP channel was considered, and an intensity threshold of 2000 grey levels was used to define a PI-positive object. For both populations, the number of objects were transformed in nuclei number ("death events") multiplying every single object by a integer factor 40 (>=1) calculated dividing the area of each identified object by the mean nucleus area (150 um 2 ) and rounding to the minor integer. This correction was necessary to avoid underestimating death events associated with syncytia in-L929 infected cells. Live cells were calculated by dividing the total estimated nuclei by the number of death events, per each frame. In all figures, death events are expressed in absolute terms, whereas live cells are expressed relative to the number of live 45 cells at the earliest recorded time frame (4 hpi).

Confocal images of BMDM were analyzed by a custom-made ImageJ macro written in Jython. Briefly, the DAPI channel, after a median filter and BG subtraction, was used to automatically segment the nuclei. To the nuclei binary image a Voronoi filter was applied to roughly identify the area relative to each cell. For each cell identified by a nucleus, a band around the nucleus 5 with a thickness of 6 µm the cytoplasm was created and the intersection between the band and the cell area was considered as the "cytoplasm". The total NSP9 intensity inside the cytoplasm was measured for each cell. The signal corresponding to 2 standard deviations above the mean NSP9 signal in the non-infected cells was used as threshold to identify the NSP9 positive and NSP9 negative cell populations in infected cells. 10 The IRF1/NF-κB nuclear signal was quantified inside the ROI obtained from the DAPI segmentation as mean intensity.

Lysosensor assay. Bone marrow derived macrophages or L929 cells were seeded on a 12 glass-bottom wells 15 (MatTek Corporation, Ashland, MA 01721, USA), coated with poly-D-Lysine 0.1% (w/v) in water (3x10 5 cells/well). After 24 hours cells were treated with DDP (10 µM) or DMSO and then infected with MHV for 12 hours. Cells were incubated with 1 µM of Lysosensor Green DND-189 (Thermo Fisher Scientific, Monza, Italy) for 90 min. and Hoechst 33342 (Euroclone S.p.A., Pero, Italy) for the last 30 min. Cells were washed and fresh medium was added. Labelled live 20 cells were imaged at 37 °C and 5% CO2 on a Leica Thunder Imager system (Leica Microsystems GmbH, Wetzlar, Germany), equipped with a xy motorized stage, 5 LED sources, a DFC9000 GTC sCMOS camera, a multi-dichroic mirror and 4 emission filters. Ninety-nine images were acquired for each condition using a 63x 1.4NA oil immersion objective lens. Widefield images of the DAPI and Lysosensor were quantified using a custom-made ImageJ macro. Similarly to 25 what was done for the NSP9 quantification, the "cytoplasmic" Lysosensor total signal was calculated in a 6 µm thick band around the nucleus after BG subtraction.

Plates containing 15x10 6 cells were washed 3 times with PBS and fixed at RT with 1% 30 formaldehyde for 15 min. Cells were washed again 3 times with PBS, harvested with a cell lifter, collected into Falcon tubes and centrifuged at 424 rcf for 5 min at 4°C. Each pellet was resuspended in 3 ml of Lysis Buffer 1 (50 mM Hepes-KOH, pH 7.5, 140 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% NP-40 and 0.25% Triton X-100) and incubated on ice for 10 minutes. Nuclei were pelleted at 1600 rcf for 5 min. at 4°C, washed with 3 ml of Lysis Buffer 2 35 (10 mM Tris-HCl, pH 8.0 5M, 200 mM NaCl, 1 mM EDTA and 0.5 mM EGTA) and incubated at RT for 10 minutes. Nuclei were pelleted again at 424 rcf for 5 minutes at 4°C and the nuclear membrane was disrupted with 1,5 ml of Lysis Buffer 3 (10 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% Na-Deoxycholate, 0.5% N-lauroylsarcosine and protease inhibitors). Chromatin fragmentation was performed by sonication (Bioruptor® Plus 40 sonication device, 45-60 cycles, 30 seconds on/off, high power, at 4°C). Chromatin extracts containing DNA fragments with an average of 300 bp were then subjected to immunoprecipitation.

The immunoprecipitation was performed using magnetic Dynabeads Protein G. Beads were 45 blocked with 0,5% BSA in PBS and then mixed with the different antibodies (15 µg antibody:100 µl beads ratio) and incubated overnight on a rotating platform at 4°C. 1% of Triton X-100 was added to the sonicated lysates and lysates were centrifuged in microfuge (8000g, 10 min. at 4C) to pellet debris. Supernatants containing chromatin were subjected to immunoprecipitation and the 10% of the volume was used as input. Antibody-coated beads were added to each lysate and incubated overnight on a rotating platform at 4°C. Beads were washed 6 5 times (5 min each) with Wash Buffer (RIPA) and once with TE and 50 mM NaCl. The immunocomplexes were eluted in 100 µl of elution buffer (TE and 2% SDS) at 65°C for 15 minutes. To reverse cross-links immunocomplexes (and inputs) were treated with RNAase (0.5 mg/ml) at 37ºC, 20 min followed by proteinase K and 1% SDS at 65 C overnight. DNA was purified using AMPure XP beads following manufacturer's instructions. Isolated DNA was used 10 to analyze by quantitative PCR the expression of NF-κB targets (TNFα, IL1β, CXCL1, CCL5, CXCL10 and a negative control), with the Fast SYBR™ Green Master Mix on a thermocycler Viia7 (Life Technologies, Inc.)

Total RNA was purified using the RNeasy kit (Qiagen). RT-qPCR was performed using Luna® 15 Universal One-Step RT-qPCR Kit (New England Biolabs), following the manufacturer's instructions. Primers details are specified in reagents table.

RNA sequencing. mRNA-seq libraries were prepared according to the TruSeq low sample protocol (Illumina, San 20 Diego, CA, USA), starting with 1 µg of total RNA per sample. RNA-seq libraries were pair-end sequenced on an Illumina NovaSeq 6000 sequencing platform. RNA-seq data were mapped using STAR aligner 54 against the mouse genome (mm10). Counts were obtained by htseq-counts 55 and differential expression analysis was performed with DESeq2 package hosted in Galaxy online platform 56 using a false discovery rate (FDR) cut-off of 1 x 10 -4 9 . Hierarchical clustering 25 was performed on z-score across samples for each gene, using Ward's criterion with 1 -(correlation coefficient) as a distance measure.

In order to quantify the expression of Transposable Elements (TEs) while avoiding the biases 30 introduced by multimapping reads, we applied a clustering procedure that groups together TEs whose expression is supported by the same set of multimapping reads. Briefly, reads were mapped to the mouse reference genome (GRCm38 assembly) using Star and allowing an unlimited number of mapping locations. Then, BAM files for all samples were pooled and the coordinates of mapped reads were intersected (bedtools intersect 57 ) with those of TEs annotated 35 in RepeatMasker (v405). This operation allowed us to build a binary matrix associating each read to the TEs that it maps to. Such matrix was then transformed with mcxarray into a square matrix of Tanimoto distances between each pair of TEs and subjected to clustering using the Markov Cluster Algorithm (MCL, filtering parameter >=0.5. Tanimoto distance. Inflation parameter 1.2 58 ). This operation generated 514.866 clusters that contain a variable number of 40 TEs with common mappability profiles. The number of reads in each cluster was then quantified for each sample (discarding reads mapping to multiple clusters) and the resulting TE expression matrix was imported into R for differential expression analysis with DESeq2 59 . Cluster counts for each sample were first normalised using size factors estimated from the number of reads uniquely mapping to Ensembl genes, then differential expression analysis was performed using 45 the Wald Test and a design formula capturing the interaction between DDP treatment and MHV infection status (~Infection*Treatment). The p-values thus obtained were then corrected for multiple hypothesis testing using the Benjamini-Hochberg procedure. To generate fold change boxplots for LINEs, SINEs and LTR clusters, we first discarded all clusters containing TEs belonging to multiple Repeat Masker classes. For the remaining TE clusters, we then plotted the log2 fold changes estimated by DESeq2 as boxplots. In order to test for differences in 5 thesedistribution we used Welch's t test.

To identify the MHV viral strain we used by the SPADES software 60 . The longest contig obtained was compared with the most common strains of MHV (MHV1, MHV3, JHM, A59 - 1  3  5  7  9  11  13  15  17  19  21  23  25  27  29  31  33  35  37  39  41  43  45  47  49  51  53   0   50   100   150 200 Viability relative to nin ected ntreated t n ct n ct 1  3  5  7  9  11  13  15  17  19  21  23  25  27  29  31  33  35  37  39  41  43  45  47  49  51 as.numeric rame s nci ia 50 figure 1A) . Viability decreased at 48 hpi in both cell types, following comparable MOI-dependent kinetics (supplementary figure 1B) ; syncytia were consistently seen in L929 cells, but 10 not in BMDMs (supplementary figure 1C) .

To analyse cellular interactions between L929 fibroblasts and BMDMs, we set up a live-cell imaging system that allowed monitoring of cell death and syncytia formation of L929 and BMDM in mono-and co-cocultures. To distinguish L929 cells from BMDMs and to allow precise quantification of syncytia formation, L929 cells were engineered to express H2b-GFP (L929 GFP ), while cell death was monitored by adding Propidium Iodide (PI). 15 L929 GFP and L929 WT cells were equally permissive for infection (supplementary figure 1D ).

In Death events 

An inflammatory cytokine signature predicts COVID-19 severity and survival

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

Autoantibodies against type I IFNs in patients with life-threatening COVID-19. Science (80-. )

Human coronavirus: Host-pathogen interaction

NF-κB and the immune response

Type I interferons: crucial participants in disease amplification in autoimmunity

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

Severe COVID-19 Is Marked by a Dysregulated Myeloid Cell Compartment

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

We would like to thank dr Federica Facciotti for critical reading of the manuscript, dr Tiziana Bonaldi and dr Alessandro Cuomo for support with fractionation experiments, dr Chiara Soriani for assistance with imaging anlaysis, dr Mario Varasi at IFOM for assistance in compound synthesis, dr Riccardo Villa from IZSLER -Brescia for providing the

The authors declare no competing interests 10