Inflammasome activation in infected macrophages drives COVID-19 pathology

Severe COVID-19 is characterized by persistent lung inflammation, inflammatory cytokine production, viral RNA, and sustained interferon (IFN) response all of which are recapitulated and required for pathology in the SARS-CoV-2 infected MISTRG6-hACE2 humanized mouse model of COVID-19 with a human immune system1–20. Blocking either viral replication with Remdesivir21–23 or the downstream IFN stimulated cascade with anti-IFNAR2 in vivo in the chronic stages of disease attenuated the overactive immune-inflammatory response, especially inflammatory macrophages. Here, we show SARS-CoV-2 infection and replication in lung-resident human macrophages is a critical driver of disease. In response to infection mediated by CD16 and ACE2 receptors, human macrophages activate inflammasomes, release IL-1 and IL-18 and undergo pyroptosis thereby contributing to the hyperinflammatory state of the lungs. Inflammasome activation and its accompanying inflammatory response is necessary for lung inflammation, as inhibition of the NLRP3 inflammasome pathway reverses chronic lung pathology. Remarkably, this same blockade of inflammasome activation leads to the release of infectious virus by the infected macrophages. Thus, inflammasomes oppose host infection by SARS-CoV-2 by production of inflammatory cytokines and suicide by pyroptosis to prevent a productive viral cycle.


Infection of human macrophages and inflammasome activation
The MISTRG6 model of COVID-19 faithfully reflects many of the chronic immunoinflammatory features of the human disease and provides an opportunity to dissect the mechanisms of late immunopathogenesis in this disease 19 . As in severe human disease, COVID-19 in MISTRG6-hACE2 mice presents with persistent viral RNA, chronic IFN response accompanied with a chronic inflammatory state in macrophages that is initiated by infection of human macrophages and maintained by subsequent inflammasome activation 19 . These events may eventually contribute to the development of persistent pulmonary immunopathology and fibrosis, which is supported by histopathological and transcriptional analysis of lungs late in infection. Overall, our mechanistic study of this model defines a cascade of events, which, initiates with lung epithelial infection and is followed with infection of tissueresident macrophages in an ACE2 and CD16 mediated manner. SARS-CoV-2 replicates in these macrophages generating replicative intermediates which include dsRNA, subgenomic viral RNA, viral RNA polymerase (RdRp), and expression of a virally encoded fluorescent reporter gene (mNG), all of which is inhibited by Remdesivir, an inhibitor of viral replication. SARS-CoV-2 replication and replicative intermediates activate an inflammatory program which involves activation of inflammasomes, production, and release of inflammatory cytokines and chemokines, and finally pyroptosis. We established inflammasome activation by visualizing ASC speck formation, which colocalized with active caspase-1 and NLRP3; this led to maturation of inflammasome mediated cytokines IL-1β and IL-18, and results in pyroptosis assayed by gasdermin D (GSDMD) and LDH release. Inflammasome activation and downstream effectors in these infected macrophages are caspase-1 and NLRP3 dependent, as inhibitors of both caspase-1 and at NLRP3 block all downstream aspects of inflammasome activation and the inflammatory cascade both in vivo and in vitro. More importantly, targeting inflammasome mediated hyperinflammation prevented immunopathology associated with chronic SARS-CoV-2 infection in vivo.

Mechanisms of viral uptake
Consistent with the enhancing role for antiviral antibodies in macrophage infection, COVID-19 severity in patients was correlated with early, high levels of afucosylated IgG which enhanced the inflammatory response by monocytes and macrophages through Fc-mediated interactions with CD32 and CD16 47-49 . We observe a similar role for CD16 and antibodies in humanized mice infected with SARS-CoV-2-mNG. The frequency of infected macrophages which express high levels of CD16 correlated with the levels of anti-Spike antibodies in the lung tissue, particularly at 4dpi. mNG positivity in these cells also correlated with a strong proinflammatory cytokine signature as measured by elevated levels of IL-18, IL-1RA, and CXCL10, all of which contribute to severe disease in humans. CD16 blockade in vivo and in vitro prevented viral uptake and blocked this subsequent inflammatory response as measured by reduced levels of CXCL10, IL-18, and IL-1RA.
The ACE2 receptor that is utilized by SARS-CoV-2 to infect lung epithelium is also expressed preferentially by infected human macrophages in vivo 43 . Notably, CD14 hi CD16 hi cells and alveolar macrophages which had measurable viral RNA cells in patient samples did not appear to co-express the traditional viral entry factors, ACE2 and TMPRSS2 as measured by the relatively insensitive method of single cell RNA sequencing (scRNAseq) 7 . This may however be a technical limitation as we similarly could not detect measurable ACE2 transcript in alveolar macrophages or CXCL10+ macrophages (which is a proxy for mNG+ infected macrophages) by scRNAseq at any time point during infection.
Yet, ACE2 protein clearly colocalized with CD68, a marker of human macrophages and correlated with viral replication quantified by mNG in these cells. More importantly blocking ACE2 prevented viral uptake by macrophages. Interestingly in our system, ACE2 expression was inducible (data not shown, GSE186794), and its levels, correlated well with normalized viral RNA levels measured in the same samples (data not shown, GSE186794). To determine factors that could regulate ACE2 expression, we identified genes that correlate with ACE2. Interestingly, the top 100 genes that correlate with ACE2 expression (r>0.6) were enriched for interferon responsive genes (data not shown, GSE186794), further highlighting the importance of the interferon pathway in COVID19 patients.
Infection of macrophages in our system is therefore dependent on both ACE2 receptor as well as antibody-mediated uptake by CD16. Given the prevalence of antibodies increases as the disease progresses, it is likely that the latter mechanism plays a more important role later in infection.
However, there may be other mechanisms that enhance SARS-CoV-2 infection or the downstream inflammatory response in human macrophages that are not explored in this study. SARS-CoV-2-mNG lacks-Orf7a which can enhance proinflammatory cytokine production in monocytes via its interaction with CD14 41,77 . It is not clear whether CD14 expression in macrophages could mediate viral entry or enhance inflammatory cytokine production 77 . It has been noted that low molecular weight suggest that SARS-CoV-2 can infect these cells [80][81][82] . However, in our humanized mouse system it is clear that the majority of RNA found associated with host cells may be the result of phagocytic or other non-replicative uptake mechanisms. It was only by using SARS-CoV-2-mNG virus (see Fig. 2, 3) that we were able to distinguish these two processes at which point we could clearly distinguish infection from mere uptake of viral debris, which in fact is prevalent. Some SARS-CoV-2+ myeloid cells in humans also had distinct transcriptomes which were largely recapitulated in what we construe as Although viral uptake and the subsequent antiviral immune response, such as CXCL10, IL-18, IL-1RA production, is enhanced in presence of monoclonal antibodies, the outcome of this enhancement does not appear to be pathological when given early or late. This is in line with extensive clinical findings that show patients given convalescent plasma or monoclonal antibodies responded well to therapy and did not present with disease enhancement 83 . Several lines of evidence also suggest that FcRs are essential for antibody mediated protection and therapy 84,85 . A possible explanation for this conflicting role of antiviral antibodies is potentially explainable by the fact that the antibodies enhance infection of macrophages and thus inflammation but at the same time they neutralize virus and thereby attenuate disease leading to a net null effect consistent with the enhancing role of antiviral Abs on macrophage infection.

Infiltrating macrophages and the essential role of viral RNA-dependent type I IFN response in disease pathology
Infection causes human macrophages to preferentially produce CXCL10 which likely attract blood monocytes to the lung where they differentiate to inflammatory macrophages. These monocytes and monocyte-derived macrophages (MDM) eventually outnumber tissue-resident macrophages; they express higher levels of TLRs and may play a central role in viral RNA detection, possibly also released by pyroptosis of infected macrophages, and the ensuing inflammatory and IFN response. IFN production is critical for the antiviral response of the early phase of disease, as also evidenced in our model by drastically higher viral loads and precipitous decline in health when the antiviral response is disabled too early by dexamethasone treatment at the peak of infection 19 .
However, this same response when persistent can be pathogenic. We found that targeting either chronic viral replication or the late IFN response therapeutically in vivo attenuates many aspects of the overactive immuneinflammatory response, especially the inflammatory macrophage response.

Implications of our findings
Inhibition of viral replication, viral uptake and inflammasome activation in infected macrophages reduced lung hyperinflammation with high levels of inflammasome-induced cytokine IL-18, IL-1RA, and CXCL10 in infected MISTRG6-hACE2 mice. Patients with severe COVID-19 also have higher levels of IL-18, IL-1β, IL1-RA and CXCL10 2-6,14,34,79 . Inhibition of both caspase-1 and NLRP3 resolved lung immunopathology associated with chronic disease in MISTRG6-hACE2 mice. Given that multiple reports in patient samples also identify a role for inflammasome driven hyperinflammation in pathophysiology of COVID-19, targeting inflammasome sensors or downstream effector molecules in patients may provide alternative therapeutic options for resolving chronicity in COVID-19. However, the increased virus production seen upon inflammasome blockade could pose a significant risk to the benefit of wholesale inhibition of the pathway. The combination of Remdesivir and anti-IFNAR2 antibodies could be an effective therapy for chronic COVID-19 which spares the antiviral T cell response unlike dexamethasone. More generally, the findings from our study and its implications provide alternative therapeutic avenues to be explored in the clinic and may guide novel therapeutic developments and prompt clinical trials to investigate combinatorial therapies that target viral RNA, inflammasome activation or its products and sustained IFN response. Table S1: Human genes that are differentially regulated in lungs of infected MISTRG6-hACE2 in response to therapeutics.

Supplementary Tables
Genes that are upregulated in response to infection and downregulated in response to therapeutics (dexamethasone, anti-IFNAR+ Remdesivir) in these infected mice at 14dpi were included in the analysis (matched to Fig 1d). Normalized expression of duplicates. N=2 biologically independent mice examined over 2 -independent experiments. Differential expression analysis was performed with DESeq2 and statistical significance was deemed using Wald test.  1g).
N=2 biologically independent mice for each condition was pooled. Marker genes for each cluster of cells were identified using the Wilcoxon test with Seurat. For the adjusted P-values the Bonferroni correction was used.  Fig. 1g) during SARS-Cov-2 infection and their response to anti-IFNAR2 and Remdesivir therapy (matched to Fig. 1h). Normalized expression of duplicates analyzed.
N=2 biologically independent mice examined over 2 independent experiments. Differential expression analysis was performed with DESeq2 and statistical significance was deemed using Wald test. Table S4: Pearson and spearman correlation values calculated for each gene for its correlation with CXCL10, TNF or TLR7 in human monocytes and macrophages at 4dpi (based on Fig. 1g, matched to Fig. 3d).
For Pearson's test, significance was based on the t-test with statistic based on Pearson's productmoment correlation coefficient cor(x, y) and following a t distribution with length(x)-2 degrees of freedom. For Spearman's test, p-values are computed using algorithm AS 89 with exact = TRUE. Correlation values, p-values (two-tailed) and FDR-adjusted p-value are presented. Details of patient demographics for specimens use in IF staining: Age, gender, medication, time of death post-symptom onset (dps), co-morbidities, cause of death and histopathological findings. This table is presented as part of supplementary methods.