SARS‐CoV‐2 sensing by RIG‐I and MDA5 links epithelial infection to macrophage inflammation

Abstract SARS‐CoV‐2 infection causes broad‐spectrum immunopathological disease, exacerbated by inflammatory co‐morbidities. A better understanding of mechanisms underpinning virus‐associated inflammation is required to develop effective therapeutics. Here, we discover that SARS‐CoV‐2 replicates rapidly in lung epithelial cells despite triggering a robust innate immune response through the activation of cytoplasmic RNA sensors RIG‐I and MDA5. The inflammatory mediators produced during epithelial cell infection can stimulate primary human macrophages to enhance cytokine production and drive cellular activation. Critically, this can be limited by abrogating RNA sensing or by inhibiting downstream signalling pathways. SARS‐CoV‐2 further exacerbates the local inflammatory environment when macrophages or epithelial cells are primed with exogenous inflammatory stimuli. We propose that RNA sensing of SARS‐CoV‐2 in lung epithelium is a key driver of inflammation, the extent of which is influenced by the inflammatory state of the local environment, and that specific inhibition of innate immune pathways may beneficially mitigate inflammation‐associated COVID‐19.

The success of immunosuppressive corticosteroid dexamethasone in treating COVID-19 (Beigel et al, 2020) suggests the importance of immunopathology in disease, likely driven by immune activation in infected and virus-exposed cells. Intracellular innate immune responses have evolved to detect and suppress invading pathogens, but inappropriate responses can also contribute to disease (Blanco-Melo et al, 2020;Park & Iwasaki, 2020). Pathogenassociated molecular patterns (PAMPs) are detected by pattern recognition receptors (PRR), including cytoplasmic nucleic acid sensors, and Toll-like receptors (TLR) that sample extracellular and endosomal space. PRR activation triggers signalling cascades which activate downstream transcription factors, including interferon (IFN) regulatory factors (IRFs) and NF-jB family members, to initiate a defensive pro-inflammatory gene expression programme, principally mediated by IFN secretion from infected cells. Paracrine and autocrine IFN signalling can suppress viral replication and spread and, together with other secreted cytokines and chemokines, coordinates adaptive immune responses. Viruses have evolved countermeasures to innate defences and deploy a combination of evasion, and direct innate immune pathway antagonism, to promote replication (Sumner et al, 2017). The resulting virus-host conflict is often a significant cause of pathogenesis with PRR-induced inflammation driving disease at the site of replication and systemically (Park & Iwasaki, 2020).
Missense mutations in innate immune pathways (Pairo-Castineira et al, 2020;Zhang et al, 2020), and autoantibodies leading to deficient type 1 IFN responses , are associated with severe COVID-19, suggesting that intact innate immune responses are important in preventing disease, probably through controlling viral replication. Co-morbidities linked to severe disease are typically inflammatory in nature, suggesting that certain types of pre-existing inflammation influence disease severity (Paranjpe et al, 2020). However, the specific host-pathogen interactions that cause disease, and how these are impacted by existing inflammation, are not understood. Identification of the molecular events that link viral replication to inflammation and disease will be critical in the development of novel and more precise therapeutic agents. Moreover, such new knowledge will provide insights into the mechanisms by which the associated risk factors for severe COVID-19 impact immune homeostasis in general.
Here, we investigated early host-virus interactions to understand the mechanisms by which SARS-CoV-2 induces an innate response, whether it can escape consequent innate immune control and how it may propagate an immunopathogenic response. We focussed on lung epithelial cells and primary macrophages, which represent cells responsible for the earliest innate immune response to the virus (Bost et al, 2020;Chua et al, 2020). We found rapid replication and infectious virus release in lung epithelial cells prior to potent innate immune activation. Indeed, the cocktail of soluble mediators produced by infected cells strongly activated macrophages, which propagated a pro-inflammatory response. Critically, the production of an inflammatory secretome was directly downstream of RNA sensing by RIG-I and MDA5 because manipulation of sensing or signalling events in infected cells, using RNA interference or signalling pathway inhibition, suppressed subsequent macrophage activation and inflammatory gene expression. Furthermore, pre-exposure of epithelial cells or macrophages to exogenous inflammatory stimuli exacerbated inflammatory responses upon SARS-CoV-2 exposure. We propose that the innate immune microenvironment, in which sensing of SARS-CoV-2 infection occurs, determines the degree of virus-induced inflammation and has the potential to drive disease.

SARS-CoV-2 activates delayed innate immune responses in lung epithelial cells
In order to investigate innate immune responses to SARS-CoV-2, we first sought a producer cell line that did not respond to the virus, thereby allowing production of virus stocks free of inflammatory cytokines. As adaptive mutations have been reported during passage of the virus in Vero.E6 cells (Davidson et al, 2020;Ogando et al, 2020), we selected human gastrointestinal Caco-2 cells, which express the SARS-CoV-2 receptor ACE2 and entry factors TMPRSS2/ 4 ( Fig EV1A and B) and are naturally permissive (Stanifer et al, 2020). We found that Caco-2 support high levels of viral production ( Fig EV1C and D), but not virus spread (< 15% cells infected) ( Fig  EV1E and F). Importantly, they do not mount a detectable innate response to SARS-CoV-2 over 72 hpi at a range of multiplicities of infection (MOIs), as evidenced by a lack of interferon-stimulated gene induction (ISG) (Fig EV1G). They are also broadly less responsive to innate immune agonists than lung epithelial Calu-3 cells . Caco-2 cells were therefore used to produce SARS-CoV-2 stocks uncontaminated by inflammatory cytokines.
Comparatively, lung epithelial Calu-3 cells express high levels of receptor ACE2, and entry co-factors TMPRSS2 and TMPRSS4 ( Fig  EV1A and B) Zang et al, 2020), and are innate immune competent ( Fig EV1I) when stimulated with various PRR agonists. Consistently, Calu-3 cells supported very rapid spreading infection of SARS-CoV-2 followed by the activation of innate immune responses. SARS-CoV-2 replication displayed > 1,000-fold increase in viral genomic and subgenomic (envelope, E) RNA levels within 5 h post-infection (hpi) across a range of MOIs 0.08, 0.4, 2 TCID50/cell (Figs 1A and EV2A), with TCID50 determined in Vero.E6 cells. Genomic and subgenomic E RNA in Calu-3 plateaued around 10 hpi. Rapid spreading infection was evidenced by increasing nucleocapsid protein (N)-positive cells by flow cytometry and immunofluorescence staining, peaking at 24 hpi with 50-60% infected cells . Infectious virus was evident in supernatants by 5 hpi at the highest MOI and peaked between 10-48 hpi, depending on MOI (Figs 1E and EV2C). A pronounced innate immune response to infection followed the peak of viral replication, evidenced by induction of cytokines (IL-6, TNF), chemokines (CCL2, CCL5) and type I and III IFNs (IFNb, IFNk1/3) measured by . This was accompanied by an IFN-stimulated gene (ISG) expression signature (CXCL10, IFIT1, IFIT2, MxA) (Figs 1H and EV2D-F). Gene induction was virus dose-dependent at 24 hpi, but equalised across all MOIs by 48 hpi, as the antiviral response to low-dose virus input maximised. These data show that infected lung epithelial cells can be a direct source of inflammatory mediators.
We were surprised that SARS-CoV-2 replicated so efficiently in Calu-3 despite innate immune responses including IFN and ISG expression because coronaviruses, including SARS-CoV-2, are reported to be IFN sensitive (Stanifer et al, 2020). Indeed, recombinant type I IFN, but not type II or type III IFNs, effectively reduced SARS-CoV-2 replication if Calu-3 cells were treated prior to infection . However, type I IFN had little effect on viral replication when added 2 h after infection (Fig 1I-K). Thus, the IFN response induced in infected lung epithelial Calu-3 cells appears too late to suppress SARS-CoV-2 replication in this system. To determine whether viral exposure dose influences the race between viral replication and IFN, we infected cells at a 100× lower dose (MOI 0.0004 TCID50/cell) and observed a longer window of opportunity for exogenous type I IFN to restrict viral replication ( Fig 1I-K). This is consistent with the hypothesis that high-dose infection can overcome IFN-inducible restriction.

Peak SARS-CoV-2 replication precedes innate immune activation
To understand the apparent disconnect in the kinetics between innate immune activation and viral replication, we used single-cell imaging to measure nuclear localisation of activated inflammatory transcription factors NF-jB p65 and IRF3, which mediate multiple PRR-signalling cascades. NF-jB p65 nuclear translocation coincided with cells becoming N protein positive, and a change was evident from 5 hpi (Figs 2A and B,and EV3). The timing of NF-jB p65 translocation was dependant on the viral dose, from 5 hpi for the highest MOI (2 TCID50/cell, Fig EV3), between 5 and 10 hpi for MOIs 0.4 and 0.04 (Figs 2A and B,and EV3), and 24-48 hpi for MOI 0.004 ( Fig EV3). IRF3 activation was also virus dose-dependent but did not maximise until 72 hpi, later than NF-jB ( Fig 2C and D), and we observed a more modest shift in IRF3 nuclear intensity compared with NF-jB throughout infection. These data are consistent with the requirement of a threshold of viral RNA replication to induce transcription factor translocation and innate immune activation and suggest that SARS-CoV-2 may antagonise IRF3 activation to a greater extent than NF-jB. Although small variation in NF-jB p65 and IRF3 nuclear intensity was observed in N-negative cells, we did not see the same large increases sustained throughout the time course as in N-positive cells, consistent with direct activation of NF-jB p65 and IRF3 by virus replication (Fig EV3).

of 17
The EMBO Journal 40: e107826 | 2021 ª 2021 The Authors Supporting the observation of activation of NF-jB p65 and IRF3 activation by SARS-CoV-2 infection, single-cell fluorescence in situ hybridisation (FISH) analysis of IL-6 mRNA (a prototypic NF-jB regulated cytokine) showed increased IL-6 transcripts uniquely in N-positive infected cells, appearing at 6 hpi and peaking at 24 hpi (Figs 2E and F,and EV4A). IFIT1 transcripts (a prototypic ISG) measured by FISH also demonstrated rapid induction in N-positive cells with increased signal from at 6 hpi ( Fig 2G and H). Strikingly, IFIT1 mRNA was not highly induced in N-negative bystander cells consistent with defective interferon responses failing to induce ISGs and a timely antiviral state in uninfected cells (Fig 2H and I). As a control for these changes, we show that GAPDH transcripts did not change ( Fig EV4B). Secretion of pro-inflammatory chemokine CXCL10, and cytokine IL-6, followed gene expression and was detected from 24 hpi (Figs 2J and K, and EV4C). Further analysis revealed increases in lactate dehydrogenase (LDH) in infected cell supernatants from 48 hpi, equal across all MOIs, indicative of proinflammatory cell death (Figs 2L and EV4D). Importantly, cytokine secretion had also equalised across MOIs from 24 hpi (Fig 2J and K). LDH release paralleled loss of the epithelial monolayer integrity ( Fig 1C) and cell death (Figs 2M and EV4E and F) accounting for the reduction in cytokine secretion at 72 hpi ( Fig 2J and K).

SARS-CoV-2 is sensed by MDA5 and RIG-I
To determine the mechanism of virus sensing by innate pathways, we first confirmed that viral RNA replication is required for innate immune activation. Inhibition of viral RNA replication, with polymerase inhibitor Remdesivir, abrogated pro-inflammatory and ISG gene expression in a dose-dependent manner (Fig 3A-D). Critically, Remdesivir was only effective if added prior to, or at the time of infection, consistent with a requirement for metabolism to its active tri-phosphorylated form (Eastman et al, 2020) (Fig 3E-H). Inflammatory gene induction dependent on viral genome replication suggested that an RNA sensor activates this innate response.
Both genomic and subgenomic SARS-CoV-2 RNAs are replicated via double-stranded intermediates in the cytoplasm . Accordingly, we detected cytoplasmic dsRNA at 5 hpi in Calu-3 cells, preceding N positivity ( Fig 3I)  . This suggested sensing of multiple RNA species given the different specificities of RIG-I and MDA5 (Hornung et al, 2006;Kato et al, 2006;Rehwinkel et al, 2010;Wu et al, 2013). Intriguingly, unlike RIG-I, MDA5 was not required for induction of NF-jBsensitive genes IL-6 or TNF, consistent with differences in downstream consequences of RIG-I and MDA5 activation ( Fig 3N and O) (Brisse & Ly, 2019). Abrogating SARS-CoV-2 sensing via MDA5 and MAVS depletion also reduced cell death, suggesting cell death is mediated by the host response rather than direct virus-induced damage ( Fig 3P). Notably, sensor depletion did not strongly increase viral RNA levels ( Fig 3Q), or the amount of released infectious virus (Fig 3R), confirming that innate immune activation via RNA sensing did not potently inhibit viral replication.

NF-jB and JAK/STAT signalling drive innate immune responses
As a complementary approach to mapping SARS-CoV-2-induced innate immune activation, and to assess the potential of specific immunomodulators to impact inflammatory responses and viral replication, we examined the effect of inhibiting NF-jB activation using IK-b kinase (IKK-b) inhibitors TPCA-1 and PS1145. IKK-b is responsible for NF-jB p65 activation by phosphorylation following PRR signalling. Induction of ISGs and IL-6 was inhibited by TPCA-1, and with slightly reduced potency PS1145 (Figs 4A-C and EV5A and B). Inhibiting Janus kinase (JAK) with Ruxolitinib, to prevent JAK signalling downstream of the type I IFN receptor (IFNAR), also suppressed SARS-CoV-2-induced ISGs, but not NF-jB-sensitive IL-6 (Figs 4D-F and EV5C). Neither TPCA-1 nor Ruxolitinib treatment increased viral genome replication over a wide range of MOIs (Fig 4G and H) or N positivity or virion production after single dose infection (Fig EV5D-F). Importantly, NF-jB and JAK inhibition significantly reduced cell death in infected cultures ( Fig 4I). This is consistent with our earlier observation and with the notion that the innate immune response to infection is the main driver of lung epithelial cell damage. Our data, thus far, show that SARS-CoV-2 infection of Calu-3 lung epithelial cells results in multi-pathway activation, driving pro-inflammatory and IFN-mediated innate immune responses that are inadequate or arise too late to restrict virus. Critically, they also suggest that SARS-CoV-2-induced IFN and proinflammatory gene expression can be therapeutically uncoupled from viral replication.

Epithelial responses to SARS-CoV-2 drive macrophage activation
Resident and recruited pro-inflammatory macrophages in the lungs are associated with severe COVID-19 disease (Bost et al, 2020;Liao et al, 2020;Pairo-Castineira et al, 2020;preprint: Szabo et al, 2020). We therefore asked whether macrophages can support SARS-CoV-2 infection and how they respond indirectly to infection, through exposure to conditioned medium (CoM) from infected epithelial cells. Importantly, neither primary monocyte-derived macrophages (MDM) nor PMA-differentiated THP-1 cells (as an alternative macrophage model) supported SARS-CoV-2 replication, evidenced by lack of increase in viral RNA and by the absence of N-positive cells (Appendix Fig S1A-C). This is consistent with their lack of ACE2 and TMPRSS2 expression ( Fig EV1A and B). However, exposure of MDM to virus-containing conditioned medium from infected Calu-3 cells ( Fig 5A) led to significant macrophage ISG induction ( Fig 5B, E and H) and increased expression of macrophage activation markers CD86 and HLA-DR ( Fig 5C, D, F-G, I and J). Importantly, the immune stimulatory activity of conditioned media was dependent on RNA sensing and innate immune activation in infected Calu-3 cells because induction of inflammatory genes and macrophage activation markers was abolished by depletion of MAVS prior to Calu-3 infection (Fig 5B-D) or by inhibition of NF-jB (TPCA-1) or JAK activation (Ruxolitinib) in infected Calu-3 (Fig 5E-J). Note that in these experiments, MDM were exposed to equivalent numbers of viral genomes from the MAVS depleted, or inhibitor-treated conditioned media (Appendix Fig S1D-F). To confirm that soluble mediators produced by infected Calu-3 cells were key in driving MDM activation, we pre-treated MDM with either anti-IFN ab receptor 2 (IFNAR2) antibody or Ruxolitinib to inhibit IFN signalling during exposure to CoM. Both treatments reduced induction of ISG IFIT2 and CXCL10 in MDM. We also saw a trend towards decreasing CCL5 expression but this was not significant, suggesting other pro- . Together, these data demonstrate that production of IFNs and inflammatory mediators from infected lung epithelial cells, downstream of viral RNA sensing, can propagate potent proinflammatory macrophage activation.

Pre-existing immune activation exacerbates SARS-CoV-2dependent inflammation
Severe COVID-19 is associated with inflammatory co-morbidities, suggesting that pre-existing inflammatory states lead to inappropriate immune responses to SARS-CoV-2 and drive disease (Lucas et al, 2020;Mehra et al, 2020;Williamson et al, 2020;Wolff et al, 2020;Zhang et al, 2020). Macrophages in particular are thought to potentiate inflammatory responses in the lungs of severe COVID-19 patients (Nicholls et al, 2003;Liao et al, 2020) and so we investigated whether inflammatory stimuli might directly exacerbate macrophage responses to SARS-CoV-2 alone (Fig 6A-H). In these experiments, we produced virus in Caco-2, and therefore, it did not contain inflammatory cytokines ( Fig EV1G). We detected low-level innate immune activation after exposure of MDM to SARS-CoV-2 alone (Fig 6B-H). However, when MDM were primed with 100 ng/ml LPS prior to exposure to SARS-CoV-2, we observed an enhanced response compared with exposure to virus or LPS alone, evidenced by significantly increased levels of ISGs (Fig 6D and E) and proinflammatory CCL5 (Fig 6C). Of note, LPS alone induced IL-6 and inflammasome-associated IL-1b expression and secretion and this was unaffected by virus exposure (Fig 6F-H). Exposure of MDM to SARS-CoV-2, prior to stimulation with LPS ( Fig 6I-P), also enhanced macrophage inflammatory and ISG responses, but not IL-6 or IL-1b expression and secretion, compared to those detected with virus or LPS alone (Fig 6J-P). Importantly, LPS treatment of MDM, before or after virus challenge, did not alter SARS-CoV-2 permissivity of MDM, evidenced by no change in the level of detectable viral E gene in MDM supernatants (Fig 6B and J). Thus, the changes in MDM gene induction by virus after LPS treatment are due to differences in the MDM response to virus and not due to a difference in the amount of virus genome added or induction of viral gene expression. Finally, we modelled the lung epithelial cell response to the cytokines observed in activated macrophages. We first selected IL-1b, as it was produced by LPS-treated, LPS-primed virus-exposed and virus primed LPS-exposed MDM (Fig 6G, H, O and P) and has been observed in severe COVID-19 patient lungs (Laing et al, 2020;Rodrigues et al, 2021). Treatment of Calu-3 with IL-1b during infection significantly increased induction of both ISGs and proinflammatory cytokines, compared to their induction by virus alone (Fig 6Q-T). The exception was IL-6, which was highly induced by virus even in the absence of IL-1b pre-treatment ( Fig 6S). Next we treated Calu-3 cells with TNF, which is also produced by LPStreated or primed MDM (Appendix Fig S2A and B)  severe COVID-19 (Chua et al, 2020;Mahase, 2020), but found no enhancement of innate responses to SARS-CoV-2 (Appendix Fig  S2C). However, both IL-1b and TNF treatment increased virusinduced epithelial cell death (Fig 6U and Appendix Fig S2D), without impacting viral replication (Fig 6V and Appendix Fig S2E).
Together, these data suggest that SARS-CoV-2 infection of lung epithelium can promote immune activation of inflammatory macrophages, via secretion of cytokines, chemokines and virus from infected cells, and that this can be exacerbated by a pre-existing proinflammatory state. This is consistent with the hypothesis that chronic inflammatory states, rather than enhanced viral replication, drive detrimental immune activation and/or cell death.

Discussion
We found that SARS-CoV-2 can replicate and spread effectively in lung epithelial Calu-3 cells over a wide range of inoculum doses despite inducing potent IFN responses and ISG expression. We propose that in the model system used here, innate immune activation occurs too late to suppress replication and attribute this to the virus deploying innate immune evasion and antagonism strategies early in infection. Indeed, coronaviruses replicate inside membranous vesicles, thought to protect viral RNA species from cytoplasmic sensing, and have complex capacity to antagonise innate immunity, including inhibition of MDA5 activation (Liu et al, 2021;Xia et al, 2020) and preventing nuclear entry of inflammatory transcription factors (Totura & Baric, 2012;Banerjee et al, 2020;Miorin et al, 2020;Park & Iwasaki, 2020;Yuen et al, 2020). Consistent with the literature, our data indicate that SARS-CoV-2 more effectively antagonises IRF3 activation and nuclear translocation than NF-jB. Indeed, it is possible that the innate immune response and the secreted signals produced by infected cells are dysregulated by viral manipulation and that this imbalanced response contributes to disease particularly in the context of underlying inflammatory pathology (Blanco-Melo et al, 2020; Giamarellos-Bourboulis et al, 2020; Lucas et al, 2020). We demonstrate that SARS-CoV-2 can be sensed by both RIG-I and MDA5 and that, through their signalling adaptor MAVS, these sensors drive inflammatory responses in infected Calu-3 cells (Fig 7). Concordantly, both RIG-I and MDA5 have been implicated in sensing the murine coronavirus mouse hepatitis virus (Roth-Cross et al, 2008; Li et al, 2010) and MDA5 was recently shown to sense SARS-CoV-2 and trigger IFN production (Rebendenne et al, 2021;Yin et al, 2021). Likewise, the activation of dsRNA sensor PKR has also been observed during SARS-CoV-2 infection of other cell types . Interestingly, DNA sensing through cGAS-STING has also been reported to contribute to inflammatory responses (preprint: Neufeldt et al, 2020), likely through sensing of self-DNA or cellular damage. The eventual innate immune activation in Calu-3 cells is likely due to sensing of viral RNA when it accumulates to a level that overcomes sequestration and pathway inhibition by the virus, as well as to cellular stress responses to infection. Importantly, Calu-3 cells pre-treated with IFN resist infection illustrating that innate responses can suppress SARS-CoV-2 replication if an antiviral state is induced prior to infection, particularly with a low viral exposure dose.
Although SARS-CoV-2 RNA has been found associated with macrophages and monocytes from infected patients (Bost et al, 2020), we found that macrophages did not support SARS-CoV-2 replication. However, they were sensitive to conditioned media from infected Calu-3 containing virus, IFNs and pro-inflammatory mediators, inducing high levels of chemokine and ISG mRNA and expression of activation markers CD86 and HLA-DR upon exposure. Crucially, it is the response of the Calu-3 cells to virus infection, via RNA sensing, that drives macrophage activation in these experiments, evidenced by suppression of activation after either MAVS depletion or NF-jB (TPCA-1) or JAK inhibition (Ruxolitinib) in the infected Calu-3 cells. We found that IFNs produced by infected Calu-3 cells downstream of RNA sensing are key in driving MDM activation, evidenced by suppression of macrophage activation with IFNAR antibody, although we expect other soluble mediators to contribute. This inflammatory role for IFN may explain how delayed IFN response could contribute to pathogenicity rather than viral clearance (Park & Iwasaki, 2020).
A recent study suggested that sensing of abortive SARS-CoV-2 infection of macrophages may contribute to their activation (Zheng et al, 2021). Our data do not rule out a role for detection of abortive replication. However, they suggest that inflammatory mediators produced from infected cells, perhaps with responses particularly skewed towards pro-inflammatory pathways after viral manipulation, are key to macrophage activation. Notably, exposure of macrophages to infected Caco-2 supernatant, which contains virus but not significant levels of cytokine or IFN, did not strongly activate the macrophages. Indeed, our results show that it is important to distinguish between the effects of virus and the effects of cytokines in the ◀ Figure 5. Epithelial responses to SARS-CoV-2 drive macrophage activation.  (E) and (H) were tested side-byside with the same mock condition. Mean AE SEM shown, data from 4 to 6 independent MDM donors are shown. Statistical comparison by two-tailed paired t-test comparing MDM exposed to control infected CoM to siMAVS/inhibitor-treated infected CoM. * (P < 0.05), ** (P < 0.01), *** (P < 0.001). K-O MDM were treated with either anti-IFNAR antibody (2.5 µg/ml), an isotype control IgG antibody (IgG, 2.5 µg/ml), Rux (2 lM) or mock treated during 6 h of exposure to CoM from infected, unmodified Calu-3 cells, before washing and measuring MDM gene expression (K, L, M), and MDM activation markers (N, O) by flow cytometry 48 h later. Both gene expression and relative MFI are compared with mock-treated MDM exposed to CoM from uninfected Calu-3 cells.

A Schematic of experimental design. B-J Calu-3 cells were transfected with siRNA targeting MAVS or non-targeting control (siCtrl) (B-D) or treated with DMSO vehicle or inhibitors
Mean AE SEM shown, data from 7 to 8 independent MDM donors are shown. Statistical testing by one-way paired ANOVA, comparing treated MDMs to untreated control by Dunnett's multiple comparison test, * (P < 0.05), ** (P < 0.01), *** (P < 0.001).
ª 2021 The Authors The EMBO Journal 40: e107826 | 2021 viral prep. Here, we have achieved this by using Caco-2 cells to produce virus without significant inflammatory cytokines and interferons and Calu-3 to produce virus with a corresponding inflammatory secretome. Importantly, inhibiting RNA sensing or pathway activation did not particularly increase viral replication, consistent with our observation that, in this model at least, virus-induced innate immune responses do not significantly inhibit SARS-CoV-2 replication. These observations highlight the potential of immunomodulators in reducing SARS-CoV-2-driven inflammatory disease. Indeed, suppression of JAK1/2 signalling with Baricitinib, in SARS-CoV-2-infected macaques, significantly reduced macrophage recruitment and inflammatory signatures and preliminary data support its use in COVID-19 (Bronte et al, 2020). These studies are consistent with epithelial-driven inflammation contributing to myeloid cell infiltration and the role of macrophages in exacerbating immune responses in COVID-19 (Giamarellos-Bourboulis et al, 2020;Hoang et al, 2020;Liao et al, 2020). Our data provide a framework for dissecting  immunomodulators as therapeutics, and we propose that it is essential to test both immunomodulators, and direct-acting antivirals, in innate immune competent cells, rather than in Caco-2, Vero or other innate immune-inactive cell types, because the inevitable interactions between virus replication and innate immune pathways can influence drug efficacy and potency (Rasaiyaah et al, 2013;Kim et al, 2019;Sumner et al, 2020).
A key question is how our experiments in Calu-3 cells inform understanding of COVID-19. We propose that by studying virus replication in innate immune competent permissive host cells, we can probe the earliest interactions between the virus and the host that underpin subsequent inflammatory responses. Our data show that RNA sensing in infected Calu-3 cells creates a pro-inflammatory milieu capable of activating primary macrophages. Crucially, the combined profile of pro-inflammatory mediators in this system mirrors that observed in COVID-19 in vivo (Bost et al, 2020;Laing et al, 2020;preprint: Szabo et al, 2020) and primary airway epithelial cells (Fiege et al, 2021). We propose that in vivo it is the innate immune microenvironment in which the virus-host interaction occurs, and its consequent influence on immune activation, that determines disease outcome. This is consistent with our demonstration that exogenous inflammatory stimuli can drive a state in Calu-3 cells, and primary macrophages, that influences the response to virus, exacerbating inflammation. This link, between the immediate epithelial response to infection and external inflammatory signals, both amplified by macrophages, provides a plausible hypothesis to explain the association of severe COVID-19 with the presence of pro-inflammatory macrophages in bronchoalveolar lavage and patient lungs (Giamarellos-Bourboulis et al, 2020;Liao et al, 2020;preprint: Szabo et al, 2020) and inflammatory co-morbidities (Mehra et al, 2020;Williamson et al, 2020;Wolff et al, 2020), which could provide similar inflammatory stimulation.
It is remarkable how effective SARS-CoV-2 is in escaping human innate immune responses at the cellular level, despite being a recent zoonosis. Very low levels of adaptive change are consistent with adaptation to human replication prior to identification. Whether SARS-CoV-2 adapted in a non-human species prior to human infection, or whether adaptation in humans occurred before identification, remains unclear. One possibility is that coronaviruses replicate in a conserved niche, with regard to innate immune evasion, and thus are particularly good at zoonosis, perhaps evidenced by SARS-CoV-2 being preceded by SARS-CoV-1 and Middle Eastern Respiratory Syndrome virus (MERS), and apparent cross-species transfer and transmission in distantly related species including humans, bats (Boni et al, 2020), camels (Azhar et al, 2014), civet cats (Wang & Eaton, 2007) and mink (Koopmans, 2020).
Viral disease is often driven by host immune mechanisms that have evolved to protect the host from death, a paradox that is particularly evident in COVID-19. Here, we have taken a significant step towards explaining the consequences of SARS-CoV-2 infection of innate immune competent lung epithelial cells by illustrating how RNA sensing can directly drive potent inflammatory responses, irrespective of whether virus replication is suppressed. We propose that further studies addressing mechanisms of SARS-CoV-2 immune evasion and cytopathology, and the wider impact these have on epithelial-immune cell cross-talk, will inform development of effective therapeutics that are broadly active against zoonotic coronaviruses.

Isolation of primary monocyte-derived macrophages
Primary monocyte-derived macrophages (MDM) were prepared from fresh blood from healthy volunteers. The study was approved by the joint University College London/University College London Hospitals NHS Trust Human Research Ethics Committee, and written informed consent was obtained from all participants. Experiments conformed to the principals set out in WMA declaration of Helsinki and the Department of Health and Human Services Belmont Report. Peripheral blood mononuclear cells (PBMCs) were isolated by density gradient centrifugation using Lymphoprep (Stemcell Technologies). PBMCs were washed three times with PBS and plated to select for adherent cells. Non-adherent cells were washed away after 2 h and the remaining cells incubated in RPMI (Gibco) supplemented with 10% heat-inactivated pooled human serum (Sigma) and 100 ng/ml macrophage colony-stimulating factor (PeproTech). The medium was replaced after 3 days with RPMI with 5% FCS, removing any remaining non-adherent cells. Cells were infected or treated with conditioned media 3-4 days later.

Virus culture and infection
SARS-CoV-2 strain BetaCoV/Australia/VIC01/2020 (NIBSC) was propagated by infecting Caco-2 cells at MOI 0.01 TCID50/cell, in DMEM supplemented with 2% FBS at 37°C. Virus was harvested at 72 h post-infection (hpi) and clarified by centrifugation at 2,100 g for 15 min at 4°C to remove any cellular debris. Virus stocks were aliquoted and stored at À80°C. Virus titres were determined by 50% tissue-culture infectious dose (TCID50) on Vero.E6 cells. In brief, 96-well plates were seeded at 1 × 10 4 cells/well in 100 µl. Eight 10fold serial dilutions of each virus stock or supernatant were prepared and 50 µl added to 4 replicate wells. Cytopathic effect (CPE) was scored at 5 days post-infection, and TCID50/ml was calculated using the Reed & Muench method (Reed & Muench, 1938), and an Excel spreadsheet created by Dr. Brett D. Lindenbach was used for calculating TCID50/ml values (Lindenbach, 2009).
For infections, multiplicities of infection (MOI) were calculated using TCID50/cell determining on Vero.E6 cells. Cells were inoculated with diluted virus stocks for 2 h at 37°C. Cells were subsequently washed twice with PBS, and fresh culture medium was added. At indicated time points, cells were harvested for analysis.
MDM were infected with virus diluted in RPMI, 5% FBS (estimated MOI 0.02 TCID50/cell). MDM were harvested at 24 h or 48 hpi for gene expression analysis. For priming experiments, MDM were stimulated with 100 ng/ml of LPS (HC4046, Hycult Biotech) for 2 h. Media was replaced, and cells were exposed to SARS-CoV-2 as before, diluted in RPMI, 5% FBS. Cells were collected after 48 h for analysis. Alternatively, cells were mock exposed or exposed to SARS-CoV-2 for 3 days and then stimulated with 100 ng/ml of LPS. Cells were harvested after 24 h for analysis.
In macrophage experiments, a minimum sample size of six independent experiments using cells derived from separate donors was used to give 90% power in order for a two-sided test to detect > twofold differences between two groups with an estimated standard deviation of 0.5.

Treatment with cytokines, inhibitors and conditioned medium
Calu-3 cells were pre-treated with Remdesivir (Selleck Chemicals), TPCA-1 (Bio-Techne), PS1145 (Bio-Techne) or Ruxolitinib (Bio-Techne) at the indicated concentrations or DMSO control at an equivalent dilution for 1 h before SARS-CoV-2 infection unless otherwise stated. Inhibitors were maintained at the indicated concentrations throughout the experiments. For cytokine treatments, recombinant human IFNb, IFNk1, IFNk2, IFNc, IL1b or TNF (PeproTech) at a final concentration of 10 ng/ml was added at the indicated time points. To generate conditioned media (CoM), Calu-3 cells were mock-infected or infected with SARS-CoV-2 at 0.04 TCID50/cell and supernatants were harvested 48 hpi, clarified by centrifugation at 2,100 g for 15 min and 4°C and stored at À80°C. For conditioned media experiments, MDM were exposed to CoM as indicated, which was diluted 1:5 in RPMI, 5% FBS. After 6 h, conditioned medium was replaced with RPMI, 5% FBS and cells were harvested at 48 h for gene expression and surface marker expression analysis. MDM were treated where indicated during CoM exposure with either 2 lM ruxolitinib (Bio-Techne) or 2.5 lg/ml anti-IFNAR antibody (pbl Assay Science) or an isotype control IgG2A antibody (R&D).

RT-qPCR
RNA was extracted using RNeasy Micro Kits (Qiagen), and residual genomic DNA was removed from RNA samples by on-column DNAse I treatment (Qiagen). Both steps were performed according to the manufacturer's instructions. cDNA was synthesised using SuperScript III with random hexamer primers (Invitrogen). RT-qPCR was performed using Fast SYBR Green Master Mix (Thermo Fisher) for host gene expression or TaqMan Master mix (Thermo Fisher) for viral RNA quantification, and reactions performed on the QuantStudio 5 Real-Time PCR systems (Thermo Fisher). Host gene expression was determined using the 2-DDCt method and normalised to GAPDH expression. Viral RNA copies were deduced by standard curve, using primers and a TaqMan probe specific for E, as described elsewhere (Corman et al, 2020) and below.
The following primers and probes were used as follows: Target Sequence

Cytokine and LDH measurement
Secreted mediators were detected in cell culture supernatants by ELISA. CXCL10 and IL-6 protein were measured using DuoSet ELISA reagents (R&D Biosystems) according to the manufacturer's instructions. Secreted lactate dehydrogenase (LDH) activity was measured as a correlate of cell death in culture supernatants using Cytotoxicity Detection Kit PLUS (Sigma) according to the manufacturer's instructions. Culture supernatants were collected at the indicated time points post-infection, clarified by centrifugation and stored at 4°C until LDH measurement.

Antibodies
All antibody sources are cited with sample identifiers, and all antibodies were validated for their specific use by manufacturers or by previously published work as cited.

Flow cytometry
For flow cytometry analysis, adherent cells were recovered by trypsinising or gentle scraping and washed in PBS with 2 mM EDTA (PBS/EDTA). Non-adherent cells were recovered from culture supernatants by centrifugation for 5 min at 300 g and washed once in PBS/EDTA. Cells were stained with fixable Zombie UV Live/Dead dye (BioLegend) for 6 min at room temperature. Excess stain was quenched with FBS-complemented DMEM. For MDMs, Fc-blocking was performed with PBS/EDTA+10% human serum for 10 min at 4°C. Cell surface with CD86-Bv711 (IT2.2, BioLegend) and HLA-DR-PerCpCy5.5 or PE-Cy7 (L243, BioLegend) staining was performed in PBS/EDTA at 4°C for 30min. Unbound antibody was washed off thoroughly, and cells were fixed in 4% PFA prior to intracellular staining. For intracellular detection of SARS-CoV-2 nucleoprotein, cells were permeabilised for 15 min with Intracellular Staining Perm Wash Buffer (BioLegend). Cells were then incubated with 1 lg/ml CR3009 SARS-CoV-2 crossreactive antibody (a kind gift from Dr. Laura McCoy) in permeabilisation buffer for 30 min at room temperature, washed once and incubated with secondary Alexa Fluor 488-Donkey-anti-Human IgG (Jackson Labs). All samples were acquired on a BD Fortessa X20 or LSR II using BD FACSDiva software. Data were analysed using FlowJo v10 (Tree Star).

Image analysis
NF-jB, IRF3, IL-6 and GAPDH raw image channels were preprocessed using a batch rolling ball background correction in FIJI ImageJ software package (Schindelin et al, 2012) prior to quantification. Automated image analysis was carried out using the "Athena" HCS analysis software package (IDEA Biomedical IDEA Biomedical, Rehovot, Israel). For quantification of the percentage of nucleocapsid-positive cells within the population, the "Intracellular Granules" module was utilised. Nuclei were segmented using Hoechst33342 signal. Cell boundaries were determined by segmentation of CellMask signal. Infected cells were determined by thresholding intracellular N protein signal (Intracellular granules).
For all analysis, the N protein signal intensity was thresholded against the mock-infected wells to ensure no false segmentation of N +ve objects. Nuclear accumulation of NF-jB or IRF3 was carried out using the "Intranuclear Foci" module. Nuclei of cells were segmented using the Hoechst33342 stain. "Foci" of perinuclear N protein signal were identified and an "Infected" cell population determined based on the presence of such segmented foci objects. In all cells, the NF-jB or IRF3 signal present within segmented nuclei was quantified. For RNA-FISH quantification, the "Mitochondria" module was implemented. Nuclei were segmented using the Hoechst33342 stain. Cell cytoplasmic area was determined by segmentation of CellMask 647 signal. Intracellular N protein signal was segmented as "mitochondria" objects. IL-6/GAPDH RNA-FISH signal within segmented cells was then quantified. Infected cells were determined by the presence of N protein-segmented objects within the cell. Analysis parameters are detailed in Appendix Tables S1-S7.

Statistical analysis
Statistical analysis was performed using GraphPad Prism. As indicated, normally distributed data were analysed for statistical significance by t-tests (when comparing two groups) or one-way ANOVA with Bonferroni or Dunnett's post-test (when comparing more than two groups). Wilcoxon ranked paired non-parametric tests were performed for primary macrophage data that were not normally distributed. For imaging analysis, where appropriate, integrated intensities were normalised to the mean intensity of the mockinfected population for that respective time point. Comparisons were made using a Kruskal-Wallis test with Dunn's multiple comparison. Data show the mean AE the SEM, where appropriate the median is shown, with significance shown on the figures, and levels were defined as *P < 0.05; **P < 0.01 and ***P < 0.001, ****P < 0.0001.

Data availability
This study includes no data deposited in external repositories.
Expanded View for this article is available online.