Innate immunity, cytokine storm, and inflammatory cell death in COVID-19

To date, 10 TLR family members, designated as TLR1–10, have been identified in humans. TLR expression can be detected in the respiratory tract, though the level of expression varies depending on the cell population. For example, TLR4 is expressed abundantly in macrophages, and TLR3 is more abundant in NK cells [15]. Downstream of TLR engagement by PAMPs and DAMPs, TLRs mediate their signaling through adaptor molecules. TLR3 exclusively signals via the adaptor protein Toll/interleukin-1 (IL-1) receptor (TIR) domain-containing adapter-inducing IFN-β (TRIF), while TLR4 signals through either myeloid differentiation primary response 88 (MyD88) or TRIF, and the remaining TLRs use MyD88 [16]. MyD88 signaling activates nuclear factor-κB (NF-κB), mitogen-activated protein kinase (MAPK), and IFN regulatory factor (IRF) signaling cascades for transcriptional upregulation of several pro-inflammatory cytokines, IFNs, and innate immune sensor genes, such as NOD-like receptor protein 3 (NLRP3). TRIF signaling is primarily involved in the production of IFNs to induce antiviral functions [16].

In the context of SARS-CoV-2 infection, multiple TLRs can sense diverse components of the infection, both in the form of viral PAMPs and host DAMPS, to induce innate immune activation (Fig. 1A). TLR1, TLR2, and TLR6 bind viral proteins, and TLR2 can form heterodimers with TLR1 and TLR6. Polymorphism in TLR1 is associated with the development of ARDS in sepsis [17], though its role in the development of COVID-19 requires further study. TLR2 senses the SARS-CoV-2 envelope (E) protein, and expression of TLR2 and its adaptor MyD88 are associated with COVID-19 disease severity [18]. TLR2-deficient murine macrophages or TLR2 inhibitor-treated human macrophages show reduced activation of pro-inflammatory signaling pathways and reduced production of cytokines in response to SARS-CoV-2 E protein [18]. The level of IL-6 in serum is decreased in Tlr2^–/– mice compared with wildtype mice when treated with the E protein [18], suggesting inflammatory responses triggered by the viral E protein depend on TLR2 signaling. Moreover, TLR2 inhibition protects against SARS-CoV-2–induced lethality in K18-human angiotensin-converting enzyme 2 (K18-hACE2) transgenic mice [18], indicating that TLR2 contributes to disease progression in COVID-19. Further, a single-cell computational analysis identified TLR2 as a key gene involved in exacerbating the hyperinflammatory response seen in patients with severe COVID-19 [19]. Additionally, decreased IL-6 production was found in E protein-stimulated human peripheral blood plasmacytoid dendritic cells (pDCs) where TLR2 had been deleted when compared to pDCs with intact TLR2 [20]. Further studies are required to establish the E protein as a ligand for direct binding to TLR2. In contrast to the pathogenic role of TLR2 in inducing excess inflammation, intranasal administration of INNA-051, a TLR2/TLR6 agonist, reduces the level of viral RNA in nose and throat swabs from SARS-CoV-2–infected ferrets [21], suggesting that stimulation of TLR2 in specific tissues may be beneficial in preventing the development of COVID-19 by reducing SARS-CoV-2 transmission. However, the administration of INNA-051 fails to induce inflammatory responses in the ferrets, indicating that INNA-051 may exert antiviral effects independent of TLR2 activation [21].

Roles for other TLRs in sensing SARS-CoV-2 have also been postulated based on their critical roles in other coronavirus infections. In the context of SARS-CoV, for example, TLR3 signaling has a protective role [22‐25]. Mice deficient in TLR3 show increased viral burden and impaired pulmonary function upon infection with mouse-adapted SARS-CoV (MA15) [25]. Consistently, TLR3 activation by poly I:C, a synthetic dsRNA used to mimic viral nucleic acids and model innate immune responses, improves the survival of aged mice and reduces pathological changes and viral loads in the lungs following infection with MA15 [23]. Moreover, inhibition of TLR3 leads to decreased production of inflammatory cytokines and IFNs in SARS-CoV-2–infected Calu-3/MRC-5 multicellular spheroids [26]. Similarly, administration of poly I:C improves survival in K18-hACE2 transgenic mice during SARS-CoV-2 infection by reducing viral load and inflammation in the lungs and brains [27]. These findings suggest there may be a protective effect of TLR3 signaling during COVID-19. Additionally, reduced expression of TLR3 in peripheral blood has been associated with worse outcomes in patients with severe COVID-19 [28]. This is further supported by a study which analyzed patients with severe COVID-19 and showed a positive correlation between the severity of disease and TLR3 inborn errors [29]. However, a follow-up study could not reproduce these results [30]. In addition, a TLR3 (rs3775290) polymorphism has been associated with an increased risk of pneumonia in patients with COVID-19 [31]. Further studies are needed to define the role of TLR3 in SARS-CoV-2 infection and the development of COVID-19.

Potential roles for TLR4 and TLR7/8 have also been suggested. In an in-silico study, TLR4 had the highest affinity for S1 of the SARS-CoV-2 spike (S) protein [32], which may be due to the interaction of TLR4 with oligomannose and glycan structures found on the surface of the S protein. Additionally, TLR4 polymorphisms are associated with COVID-19 severity; the TLR4 minor alleles 299Gly (G) and 399Ile (T) are associated with increased likelihood of severe COVID-19 and the risk of developing cytokine storm [33]. In murine models, Tlr4^–/– macrophages from mice had reduced Il1b gene expression compared to that of wildtype cells in response to S protein [34]. However, because S protein has an affinity for lipopolysaccharide (LPS) [35], it is possible that LPS contamination from the system used to express and purify the recombinant S protein in these studies contributed to the TLR4-dependent gene expression. Therefore, whether TLR4 can directly sense S protein or not requires further confirmation. In addition, TLR7/8 may also be involved in SARS-CoV-2 infection. TLR7/8 have long been studied for their role in antitumoral immune responses due to their ability to provoke rapid production of IFNs and cytokines [36]. A bioinformatic analysis reported that a large number of single stranded RNA (ssRNA) fragments are present in SARS-CoV-2, even more so than in SARS-CoV [37]. These ssRNA fragments can be recognized by endosomal TLR7/8 to generate antiviral immunity [38]. Furthermore, functional TLR7 genetic variants have been linked to COVID-19 severity in multiple studies [31, 39, 40]. Specifically, a TLR7 (rs179008) polymorphism has been associated with increased risk of pneumonia in patients with COVID-19 [31], and deleterious TLR7 variants are associated with more severe disease [39, 40]. Given that TLR7 is found on the X-chromosome, TLR7 may influence gender-based differences in COVID-19 susceptibility and severity. Males are nearly two times as likely to develop respiratory failure or die from COVID-19 as females are [41, 42]. TLR7/8 may also recognize antiphospholipid antibodies (aPL) [43, 44], which have been shown to be upregulated in patients with severe and critical COVID-19 [45, 46]. Furthermore, TLR7-deficient human peripheral blood pDCs show decreased IFN production compared with TLR7-sufficient pDCs following SARS-CoV-2 infection [20]. Additional investigations are needed to verify the roles of these and other TLRs in response to SARS-CoV-2 infection.

In addition to sensing by TLRs, viral RNA can be detected by RLRs such as melanoma differentiation-associated protein 5 (MDA5), retinoic acid-inducible gene (RIG)-I, and laboratory of genetics and physiology (LGP)-2 (Fig. 1B) [47, 48]. Sensing through RLRs leads to their interaction with the adaptor protein mitochondrial antiviral signaling (MAVS) to form a complex that activates IFN production via TNF receptor-associated factor 3 (TRAF3), TRAF-associated NF-κΒ (TANK)-binding kinase 1 (TBK1), IκB kinase (IKK), and IRF3 signaling [49, 50]. In addition, one study showed that RIG-I is required for the induction of NF-κB-sensitive genes, interleukin (IL)-6, and TNF, while MDA5 is not in Calu-3 cells during SARS-CoV-2 infection [51]. The earlier, more pronounced antiviral response to SARS-CoV-2 infection seen in children compared to adults has been suggested to be associated with higher basal expression of MDA5 and RIG-I in the epithelial cells of the upper airway as well as macrophages and dendritic cells in children [52]. This could be a possible explanation for why children have lower infection rates for SARS-CoV-2 and a lower risk for disease progression compared with adults. Screening of putative sensors involved in the sensing of RNA virus infection has found that MDA5 and LGP2 primarily regulate IFN induction in lung epithelial cells during SARS-CoV-2 infection [47]. Silencing MDA5, LGP2, and MAVS in lung epithelial cells reduces the expression of IFN-β during SARS-CoV-2 infection in vitro [47, 48]. Similarly, gene silencing of MDA5 in Calu-3 cells decreases IFN-β production compared to controls during infection with SARS-CoV-2 [51]. Abrogating SARS-CoV-2 sensing via MDA5 and MAVS depletion also reduces cell death, suggesting that cell death is mediated by the host response rather than by direct virus-induced damage [51]. In contrast, the role of RIG-I in sensing SARS-CoV-2 is currently inconclusive. One study found that silencing RIGI in Calu-3 cells abolishes the production of IFN-β and other pro-inflammatory cytokines during infection with SARS-CoV-2 [51]. In addition, SARS-CoV-2–infected RIG-I–deficient HEK-293 cells fail to induce IFN-β mRNA expression (53). In contrast, another study showed that silencing RIGI in Calu-3 cells does not inhibit the induction of IFN-β in response to SARS-CoV-2 infection [47]. RIG-I utilizes its helicase domain to recognize the 3’ untranslated region of SARS-CoV-2 genomic RNA, which differs from RIG-I’s canonical C-terminal domain sensing of viral RNA. This new mode of RIG-I recognition fails to activate its ATPase; it also does not activate the classical MAVS signaling pathway. These signaling differences could potentially explain the dispensable role of RIG-I in IFN-β production during SARS-CoV-2 infection [54]. Additional studies will be required to fully understand the role of RIG-I in response to SARS-CoV-2. However, it is known that SARS-CoV-2 encodes inhibitors of both RIG-I and MDA5, suggesting that RIG-I and MDA5 likely possess some cell type-specific functions. A better understanding of the roles of RIG-I and MDA5 in different cell types will aid in the development of more efficient COVID-19 therapeutics.

Several NLRs have been implicated in the production of IFNs and inflammatory cytokines during SARS-CoV-2 infection. For instance, silencing of an intracellular sensor of bacterial peptidoglycans, nucleotide-binding oligomerization domain-containing protein 1 (NOD1; also known as NLRC1), in Calu-3 cells during SARS-CoV-2 infection reduces the expression of IFN-β [47]. NLRC1 may mediate SARS-CoV-2 infection-induced NF-κΒ activation or directly bind viral RNA and regulate the MDA5-MAVS complex formation to modulate IFN-β production [55]. Several reports have demonstrated that NLRP3 is also involved in coronavirus infections (Fig. 1C). NLRP3 is a canonical inflammasome sensor, and it forms a multiprotein complex with the adaptor apoptosis-associated speck-like protein containing a caspase activation and recruitment domain (ASC) and caspase-1 in response to PAMP/DAMP sensing. Inflammasome formation results in the activation of caspase-1 to cleave its substrates, including the pro-inflammatory cytokines IL-1β and IL-18 to produce their bioactive forms. Caspase-1 also cleaves gasdermin D (GSDMD) to release its N terminal fragment to form pores in the membrane and induce cell death, pyroptosis [56]. The SARS-CoV-2 N protein is thought to bind to GSDMD to inhibit pyroptosis [57]. However, other studies have suggested that NLRP3 inflammasome activation and pyroptosis still occur during SARS-CoV-2 infections. NLRP3 deficiency inhibits caspase-1 and GSDMD activation in a murine coronavirus infection model with mouse hepatitis virus (MHV) [58], indicating that coronaviruses induce NLRP3 inflammasome assembly. Additionally, microscopy of monocytes and lung tissue samples from patients with COVID-19 show the formation of NLRP3 and ASC puncta, suggesting the formation of NLRP3 inflammasomes in these patients [59]. Moreover, SARS-CoV-2–infected human primary monocytes show NLRP3-dependent caspase-1 cleavage, GSDMD cleavage, and IL-1β maturation [59, 60]. NLRP3 is upregulated and activated by multiple SARS-CoV-2 PAMPs, including GU-rich RNAs, E, N, and open reading frame (ORF) 3a proteins [18, 61‐63]. In response to the SARS-CoV-2 S protein, NLRP3 expression and IL-1β release are upregulated in macrophages from patients with COVID-19 but not in macrophages from healthy patients [64]. Additionally, SARS-CoV-2 N protein interacts directly with NLRP3, thereby facilitating the binding of NLRP3 with ASC [62]. N protein-mediated lung injury, inflammation, and death in mice is reduced upon treatment with an NLRP3 inflammasome inhibitor, MCC950, or a caspase-1 inhibitor, Ac-YVAD-cmk [62]. MCC950 treatment also alleviates excessive lung inflammation and COVID-19–like pathology in adeno-associated virus (AAV)-hACE2 transgenic mice, indicating that the NLRP3 inflammasome induces excessive inflammatory responses during SARS-CoV-2 infection [65]. This detrimental effect of NLRP3 during SARS-CoV-2 infection is further supported by the finding that the severe pathology induced by SARS-CoV-2 in lung tissues is reduced in Nlrp3^–/– mice compared to wildtype mice [65]. Activation of the NLRP3 inflammasome is also seen in peripheral blood mononuclear cells (PBMCs) and tissues of postmortem patients with COVID-19 upon autopsy [59]. Moreover, higher levels of inflammasome-dependent products, such as IL-18 and active caspase-1, are associated with disease severity and poor clinical outcome [59]. Collectively, these data suggest that SARS-CoV-2 activates the NLRP3 inflammasome. Additionally, during SARS-CoV-2 infection, the expression of proteins involved in non-canonical NLRP3 inflammasome signaling, through caspase-11 and caspase-4, are upregulated in the lungs of mice and humans, respectively. Deficiency of caspase-11, but not GSDMD, an executioner of pyroptosis, reduces disease severity in SARS-CoV-2–infected mice [66], suggesting that caspase-11 (or caspase-4/5 in humans) may promote disease severity in COVID-19 independently of pyroptosis. Caspase-4/11 can be activated by oxidized phospholipids that are produced in damaged tissues (Fig. 1D) [67]. Thus, it is possible that this activation occurs during SARS-CoV-2 infection. Oxidized phospholipids are reportedly upregulated in patients with COVID-19 [68, 69], and outside the context of COVID-19, they induce cytokine release in a caspase-4/11–dependent manner without requiring GSDMD-dependent cell death [67].

Beyond NLRP3, other inflammasomes have also been implicated in SARS-CoV-2 infection. Colocalization of the AIM2 inflammasome with ASC specks in monocytes from patients with COVID-19 has been visualized using confocal microscopy [70]. AIM2 activation by SARS-CoV-2 is unexpected, although there is a report suggesting AIM2 activation by RNA viruses occurs in rare cases through an unclear mechanism [71]. AIM2 might sense host mitochondrial DNA, as mitochondrial membranes are damaged during cell death [72].

The presence of cytosolic DNA during infection activates the cyclic GMP-AMP synthase (cGAS) and stimulator of IFN genes (STING) signaling pathway, which is critical for limiting the replication of viruses (Fig. 1B) [73‐76]. Sensing of cytosolic DNA through cGAS generates the production of cyclic GMP-AMP, which acts as a second messenger to bind and activate STING [76]. Once activated, STING signals through TBK1 and IRF3 to regulate the transcription of innate immune genes, including cytokines and IFNs [73, 75, 76]. Analysis of the skin lesions of patients with COVID-19 shows a STING-dependent type I IFN signature that is primarily mediated by macrophages adjacent to areas of endothelial cell damage [77]. Moreover, cGAS-STING activity has been detected in lung samples from patients with severe COVID-19. The finding that STING regulates IFN production and endothelial cell death in response to mitochondrial DNA release correlates well with the studies of endotheliopathy and vascular damage due to gain-of-function mutations in STING [78] or after administration of highly potent STING agonists [79]. This possibly explains the presence of vascular damage and coagulopathy in patients with severe COVID-19. Moreover, another study has shown that the cGAS-STING pathway contributes to the production of NF-κB–dependent pro-inflammatory cytokines in SARS-CoV-2–infected epithelial cells [80]; however, cGAS-STING is not involved in IFN production in these cells [80]. This non-canonical function of the cGAS-STING pathway in epithelial cells may dysregulate the pro-inflammatory response.

Treatment of K18-hACE2 transgenic mice with a selective STING inhibitor before and throughout the course of SARS-CoV-2 infection or as a therapeutic agent given after the initiation of infection reduces inflammation and improves survival [77]. In contrast, STING agonists such as cyclic guanosine monophosphate-adenosine monophosphate (cGAMP) or diamidobenzimidazole (diABZI) have also been shown to limit SARS-CoV-2 infection in mouse models when given prophylactically or early in infection [81‐83]. Endotracheal administration of STING agonists in mice triggers a lung inflammatory response and cell death [84]. Together, these data suggest that early STING activation to induce IFN production may be beneficial, while later in infection, pro-inflammatory cytokines produced by epithelial cells, endothelial cells, and macrophages via the cGAS-STING pathway provoke pathological effects.

CLRs tailor their immune responses to pathogens by carbohydrate-based PAMP sensing (Fig. 1E) [85]. CLRs can have diverse functions in innate immune activation, ranging from the modulation of other PRR signaling pathways and phagocytosis for antigen presentation to the induction of inflammatory cytokine production [86]. An ectopic expression screen focused on myeloid cell receptors showed that several CLRs such as dendritic cell specific intercellular adhesion molecule-3-grabbing non-integrin (DC-SIGN), liver/lymph node-specific intercellular adhesion molecule-3-grabbing non-integrin (L-SIGN), liver sinusoidal endothelial cell lectin (LSECtin), asialoglycoprotein receptor (ASGR1), and c-type lectin domain family 10 member A (CLEC10A) are glycan-dependent binding partners of the SARS-CoV-2 S protein [87]. These receptors do not support active viral replication, but they can play a role in disease severity by contributing to robust inflammatory responses in myeloid cells [87]. Additionally, multiple studies show that LSECtin and DC-SIGN can serve as receptors for the S glycoprotein of SARS-CoV-2 [88, 89]. Anti-S nanobody treatment not only blocks ACE2-mediated infection but also inhibits myeloid receptor-mediated proinflammatory responses; this is likely because the S protein is recognized by both CLRs and ACE2 [87]. In addition, PM26, a glycomimetic antagonist of DC-SIGN, inhibits the interaction of the S protein with the lectin receptor and blocks DC-SIGN–mediated SARS-CoV-2 trans-infection of Vero E6 cells [88]. In addition to their roles in sensing to activate innate immune signaling, CLRs, especially DC-SIGN and L-SIGN, assist in the entry of viruses such as human immunodeficiency virus (HIV), cytomegalovirus, dengue, Ebola, and Zika virus into host cells [90‐93]. SARS-CoV-2 S protein binds with CLRs in a calcium-dependent manner to facilitate the internalization of the virus [94]. Single cell RNA sequencing analyses indicate that CLRs are highly expressed in the innate immune cells of patients with severe COVID-19 [87, 94], and a proteomic profiling study indicates that DC-SIGN is a mediator of genetic risk in COVID-19 [95]. Overall, the biological relevance of the interaction of CLRs with SARS-CoV-2, especially in the context of innate immune activation, remains understudied.