Non‐steroidal anti‐inflammatory drugs, prostaglandins, and COVID‐19

Severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) is the cause of the novel coronavirus disease 2019 (COVID‐19), a highly pathogenic and sometimes fatal respiratory disease responsible for the current 2020 global pandemic. Presently, there remains no effective vaccine or efficient treatment strategies against COVID‐19. Non‐steroidal anti‐inflammatory drugs (NSAIDs) are medicines very widely used to alleviate fever, pain, and inflammation (common symptoms of COVID‐19 patients) through effectively blocking production of prostaglandins (PGs) via inhibition of cyclooxyganase enzymes. PGs can exert either proinflammatory or anti‐inflammatory effects depending on the inflammatory scenario. In this review, we survey the potential roles that NSAIDs and PGs may play during SARS‐CoV‐2 infection and the development and progression of COVID‐19. Linked Articles This article is part of a themed issue on The Pharmacology of COVID‐19. To view the other articles in this section visit http://onlinelibrary.wiley.com/doi/10.1111/bph.v177.21/issuetoc

and cerebrovascular disease are associated with higher incidence of mortality in COVID-19 patients (Docherty et al., 2020;Wang, Li, Lu, & Huang, 2020). Furthermore, post-mortem examinations reveal that tissue inflammation and organ dysfunction do not map to tissue/cellular dispersal of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) during fatal COVID-19 (Dorward et al., 2020). Thus it was concluded that immune-mediated as opposed to pathogen-mediated organ inflammation and injury contributed to death in COVID-19 patients (Dorward et al., 2020).
COVID-19 is caused by SARS-CoV-2, which generally has a 4-to 5-day average incubation period (prior to onset of symptoms), with 97.5% of symptomatic patients experiencing symptoms within 11.5 days (Lauer et al., 2020), although patients can also be asymptomatic. SARS-CoV-2 is normally transmitted via respiratory droplets to facial mucosal membranes (eyes, nose, and mouth) with high efficacy over short distances. SARS-CoV-2 belongs to a family of viruses named Coronaviridae, which are a broad range of single-stranded, positive-sense, RNA viruses that can cause respiratory, enteric, neurological, and hepatic diseases in multiple animal species and humans (Zumla, Chan, Azhar, Hui, & Yuen, 2016).
This relatively high genetic similarity with SARS-CoV-1 suggests that SARS-CoV-2 may exert its pathogenesis via similar mechanisms.
Pathophysiology of SARS-CoV-2 culminates in airway injury (possibly by direct effects on epithelial cells) and powerful host inflammatory responses (e.g., neutrophilic and macrophage inflammation and cytokine storm), as previously observed in SARS-CoV-1 (G. Chen, Wu, et al., 2020;Huang et al., 2020). Hence, both viral infection and host responses contribute to disease severity. Regarding epidemiology, the correlation between increasing severity with age of infected SARS-CoV-2 patients does correspond with epidemiology of SARS-CoV-1 and MERS-CoV (Tay, Poh, Rénia, MacAry, & Ng, 2020). SARS-CoV-1 infects various lung cells including ciliated airway epithelial cells, vascular endothelial cells, macrophages, and type II pneumocytes (AT2 cells) via binding to the host cell receptor, angiotensin-converting enzyme 2 (ACE2), and the transmembrane protease serine 2 (TMPRSS2) for S protein priming (Cui, Li, & Shi, 2019;Hamming et al., 2004;Jia et al., 2005;Qian et al., 2013;H. Xu, Zhong, et al., 2020). Conversely, MERS-CoV infects unciliated airway epithelial cells as well as type I (AT1 cells) and AT2 cells, via dipeptidyl peptidase-4 (DPP-4; CD26) de Wit et al., 2013;Raj et al., 2013). As the infection route of SARS-CoV-2 is also via the ACE2 receptor and TMPRSS2, it is expected that lung cells are infected as defined in SARS-CoV-1 (Hoffmann et al., 2020;Walls et al., 2020;Zhao et al., 2020;P. Zhou, Yang, et al., 2020). Moreover, as ACE2 expression in lung cells is lowered during SARS-CoV-1 infection and loss of ACE2 function has been shown to contribute to acute lung injury (ALI) (Imai, Kuba, & Penninger, 2008;Imai et al., 2005;Kuba et al., 2005;Kuba, Imai, Rao, Jiang, & Penninger, 2006), a reduction in ACE2 receptor expression may be a key mediator of COVID-19 pathogenesis, featured as hyperinflammation and thrombosis. So far, there remains no clear antiviral treatments or vaccine against COVID-19.  or making the disease more severe by using NSAIDs, suggestions on avoiding NSAIDs for COVID-19 patients were previously made based on numerous clinical trials and observations on non-COVID-19 pulmonary infectious diseases. For example, apart from the commonly adverse effects of NSAIDs such as gastrointestinal, renal, and cardiovascular complications (Bhala et al., 2013;Little et al., 2013), NSAIDs were found to cause more prolonged illness or complications when taken during respiratory tract infections (Le Bourgeois et al., 2016;Voiriot et al., 2019). It was reported that use of NSAIDs for fever or non-rheumatologic pain during the early stages of infection increased the risk of severe bacterial superinfection (Micallef, Soeiro, & Jonville-Béra, 2020). NSAIDs may increase hypercoagulation and the incidence of thrombosis due to decreased thrombomodulin (Rabausch et al., 2005;Schmidt et al., 2011), a particular concern given that COVID-19 patients often have coagulation abnormalities and increased vascular clotting (Levi, Thachil, Iba, & Levy, 2020). It is plausible that NSAIDs may possibly inhibit protective host immune reactions against coronavirus replication  and enhance the proinflammatory cytokine storm observed in lungs of COVID-19 patients, for example, through activation of inflammatory macrophages (Wu & Meydani, 2008).
Moreover, as SARS-CoV-2 can infect human gut enterocytes (Lamers et al., 2020), it is important to know if NSAIDs synergise with SARS-CoV-2 infection to potentiate severe intestinal damage.
However, do NSAIDs have any potential beneficial effects on the risk of infection with SARS-CoV-2 and/or COVID-19 severity? In a retrospective cohort study, Rentsch et al. (2020) found that exposure to NSAIDs (−365 to −14 days prior to baseline) was modestly associated with increased likelihood of COVID-19 infection (multivariable odds ratio [OR] 1.27, 95% confidence interval [CI] 1.02-1.58), was not associated with hospitalisation (P = 0.19), and had a negative trend of association with intensive care (P = 0.08). In this study, information of NSAIDs only obtained from pharmacy records was counted, but individuals assigned as non-NSAID users were still likely to get NSAIDs over the counter. In another study, Freites et al. (2020) found that use of NSAIDs was associated with reduced risk of hospital admission of COVID-19 patients who had chronic inflammatory rheumatic disease (OR 0.37, 95% CI 0.15-0.91, P = 0.03). Castro, Ross, McBride, and Perlis (2020) reported that prescription of ibuprofen and naproxen (both NSAIDs) was also associated with reduced hospitalisation among 2,271 individuals who tested positive for were tested COVID-19 positive) using a national wide claims database of South Korea. They found that prior exposure to aspirin significantly associated with reduced risk of developing COVID-19 (adjust OR 0.29, 95% CI 0.14-0.58, P < 0.001), while prior use of COX-2 inhibitors did not significantly reduce the risk (Huh et al., 2020). Interestingly, Hong et al. (2020) have recently performed a pilot trial to treat COVID-19 patients with celecoxib (Celebrex ® ; a selective inhibitor of COX-2) at either 0.2 g twice or once a day. The remission rates of COVID-19 disease for full/high (0.2 g twice a day) and half/medium (0.2 g once a day) doses of celecoxib and control groups were 100%, 82%, and 57%, respectively . Celecoxib treatment also improved pulmonary opacification and pneumonia faster than control group based on chest CT scan results. This study suggested that celecoxib may promote the recovery of ordinary and severe cases of COVID-19 and prevent the progression of severe disease to a critical stage . Low doses of celecoxib (i.e., 0.2 g once a day) used in this study was associated with a small increase in the risk (relative risk 1.35, 95% CI 1.00-1.82) of non-fatal myocardial infarction, but the high dose of celecoxib (i.e., 0.2 g twice a day) was not associated with increased risk of myocardial infarction (relative risk 1.05, 95% CI 0.33-3.35) (García Rodríguez, Tacconelli, & Patrignani, 2008). These findings may thus be indicative of the safety and efficiency for use of such doses of celecoxib in COVID-19 patients who have known cardiovascular conditions. These (pharma)epidemiological and clinical observations suggest that NSAIDs may be beneficial in controlling the development of severe COVID-19. The types and doses of NSAIDs determine their capability of inhibiting enzymic activities of COX-1 or COX-2 or both, which affects the production profile of downstream eicosanoids, resulting in varied side effects in the cardiovascular, gastrointestinal, renal, and other systems (Bhala et al., 2013;García Rodríguez et al., 2008;Little et al., 2013). However, the information for exposure of NSAIDs (e.g., drug types, doses, and exposure time) in the above epidemiological studies was unclear as they were mainly defined based on electronic health records. Lack of such information impedes further assessment of clinical benefits or harm of NSAIDs. Therefore, large-scale, double-blinded, randomised, and well-controlled clinical trials are warranted to thoroughly assess the effects of specific NSAIDs at appropriate doses in prevention and treatment of COVID-19 while also attempting to avoid their known side effects.
Moreover, another NSAID, indomethacin, has shown potential antiviral activity against human SARS-CoV-1 and canine coronavirus (Amici et al., 2006). Indomethacin does not affect coronavirus binding or entry into host cells but instead acts by blocking viral RNA synthesis in vitro. Oral administration of indomethacin (1 mgÁkg −1 Áday −1 [equivalent $40 mgÁkg −1 for a 70-kg human, less than the lowmedium dose of 75 mg daily used in adult patients] for 4 days starting on day 4 post-infection) markedly reduced (by $2-3 orders) shedding of canine CoV RNA in the faeces of dogs infected with canine CoV, but this antiviral effect was reversed upon suspension of indomethacin treatment (T. Xu, Gao, Wu, Selinger, & Zhou, 2020). Indomethacin has also been suggested to exhibit potent antiviral activity against SARS-CoV-2-infected Vero E6 cells in vitro and canine CoV-infected dogs in vivo (T. Xu, Gao, et al., 2020). A very recent study based on a multi-stage model-based approach showed that treatment with the sustained-release formulation of indomethacin at the dose of 75 mg twice a day (high dose used clinically according to García Rodríguez et al., 2008) is expected to achieve a complete response in 3 days for the treatment in patients infected by SARS-CoV-2, suggesting that indomethacin could be considered as a promising candidate for the treatment of COVID-19 (Gomeni, Xu, Gao, & Bressolle-Gomeni, 2020;Koch et al., 2020). The antiviral capacity of indomethacin is conferred by activation of protein kinase R, independently of interferons and double-stranded RNA (Amici et al., 2015) but may be via interactions with aldoketo-reductases, aldose reductases, PPAR-γ, and the cannabinoid CB 2 receptor (T. Xu, Gao, et al., 2020). Conversely, it is important to report that indomethacin and other NSAIDs (including coxibs) are associated with nephrotoxicity (Delaney & Segel, 1985;McCarthy, Torres, Romero, Wochos, & Velosa, 1982;Whelton, 1999). As a result, such NSAIDs are not recommended for use in patients with clinically complicated SARS-CoV-2 infections with deficiencies in renal function, where renal blood flow is maintained by the contributory effects of vasodilator PGs (Grosser, Fries, & FitzGerald, 2006) or with gastrointestinal risk factors (Capuano et al., 2020). Therefore, such clinically complex SARS-CoV-2 infections constitute a contraindication to the use of NSAIDs as their use may predispose to these well-known side effects of NSAIDs (Patrignani & Patrono, 2015). Importantly, indomethacin is also known to have coronary vasoconstrictor effects via blockade of vasodilatory PG synthesis or to a direct drug effect, suggesting that in patients with severe coronary-artery disease, indomethacin should be used with caution (Friedman et al., 1981). Using an original virtual screening protocol, celecoxib at 50 μM (much higher than its maximum serum concentration of 1.8 μM when 200 mgÁday −1 is used) was predicted to suppress the activity of the main chymotrypsin-like protease (a key target for antiviral drugs) of SARS-CoV-2 by $12% (Gimeno et al., 2020). In addition, naproxen has also been shown to possess antiviral activity against influenza A and B viruses by interfering with the RNA replication process (Zheng et al., 2019). Moreover, ibuprofen was reported to enhance ACE2 expression from diabetic rat cardiac tissues (Qiao et al., 2015), while celecoxib was shown to repress expression of TMPRSS2 in human prostate cancer cells (Kashiwagi et al., 2014). Given the protective action of ACE2 in ALI (Imai et al., 2008;Imai et al., 2005;Kuba et al., 2005;Kuba et al., 2006) and the negative correlation between ACE2 expression and SARS-CoV-2 severe outcomes (J. Chen, Jiang, et al., 2020), NSAIDs may help reduce COVID-19 disease severity due to SARS-CoV-2 infection if they can up-regulate and down-regulate ACE2 and TMPRSS2 in the lung, respectively.

| PGs IN THE HUMAN LUNG
NSAIDs serve to reduce inflammation by targeting cyclooxygenases (COXs, i.e., COX-1 and COX-2) and inhibiting biosynthesis of prostaglandins (PGs), a group of important lipid mediators. PGs are formed when arachidonic acid (AA) is released from cell membrane phospholipids by the actions of cytosolic PLA 2 (cPLA 2 ) and converted to PGH 2 by COXs. COX-1 is constitutively expressed in most cells, whereas COX-2 is induced upon the initiation of inflammation (Dubois et al., 1998). PGH 2 is unstable and converted into each PG by the corresponding specific synthases. Thus PGD 2 is produced by the PGD synthases, LPGDS and HPGDS); PGE 2 by the PGE synthases (mPGES-1, mPGES-2, and cPGES); PGF 2α by the PGF synthases, including AKR1C3 and AKR1B1; PGI 2 (prostacyclin) by the PGI synthase: and thromboxin A 2 (TXA 2 ) by TXA synthase (Figure 1). Single-cell RNA sequencing analysis (Du et al., 2017) reveals that cPLAs, COXs, and PG synthases are expressed in normal human lung cells from a healthy young individual (Figure 2). For example, COX-1 is expressed in immune cells such as dendritic cells (DCs) and mast cells while COX-2 is broadly but moderately expressed by epithelial (e.g., AT1), stromal (e.g., matrix fibroblasts), and immune (e.g., monocytes/macrophages, DCs, and mast cells) cells. Endothelial, stromal, mast, DC, and NK cells highly express PGDSs, but only matrix fibroblasts and lymphatic endothelial cells moderately express mPGES-1, the inducible enzyme that mediates PGE 2 biosynthesis in vivo, in the normal lung. PGs are found in most mammalian cells and tissues and can promote both the initiation and resolution of inflammation, depending on local concentrations, disease setting, and timing of action . Fluctuations in expression of these synthases within cells recruited to sites of inflammation govern the production profiles of PGs. Although PGE 2 is detectable in bronchoalveolar lavage (BAL) fluids (rather than bronchial wash fluids) taken from healthy individuals, its level (1.6 pM) is threefold to eightfold lower than levels of PGD 2 , PGF 2α , and TXB 2 (a stable TXA 2 metabolite) (Gouveia-Figueira et al., 2017). PG levels are usually elevated in response to inflammatory or noxious stimuli. For example, PGE 2 compared to other PGs was found to be significantly up-regulated in BAL fluids of individuals after exposure to biodiesel exhaust (Gouveia-Figueira et al., 2017).
Plasma levels of PGE 2 were increased in patients from early stage (within the first week, 1,966 pgÁml −1 ≈ 5.6 nM) of SARS-CoV-1 infection and lasted to the late stage (2-3 weeks, 2,170 pgÁml −1 ≈ 6.2 nM) of infection compared to that in plasma from control individuals (1,300 pgÁml −1 ≈ 3.7 nM) (Lee et al., 2004). This is associated with induction of COX-2 expression as the SARS-CoV-1 N protein binds directly to two regulatory elements (i.e., NF-κB and C/EBP binding sites) of the COX-2 promoter (Yan et al., 2006). By performing proteomic and metabolomic profiling of sera from COVID-19 patients and healthy individuals, Shen et al. (2020) found a remarkable elevation of serum amyloid A1 (SAA1), SAA2, SAA4, and other inflammation markers (e.g., C-reactive protein, SAP, and SERPINA3) in severe COVID-19 patients compared to non-severe COVID-19 patients and healthy individuals. SAAs can enhance COX-dependent AA metabolism and production of PGs, including PGE 2 , PGF 2α , and TXA 2 , in many types of human cells, such as monocytes, macrophages, and fibroblasts, through NF-κB activation (Li et al., 2017;Malle et al., 1997). Similarly, Yan et al. (2020)  and their metabolites present in the lung, for example, in BAL fluid, will also be highly informative.
Traditionally, PGs have been viewed as inflammatory mediators connecting innate immunity to phases of acute inflammation with proinflammatory cytokines (TNF-α and IL-1β) and LPS known to induce expression of inducible COX-2 and mPGES-1 (Díaz-Muñoz, F I G U R E 1 An overview of PG biosynthesis, receptors, and downstream signalling pathways. Arachidonic acid (AA) is released from membrane phospholipids via the actions of cytosolic PLA 2 (cPLA 2 ) following various stimuli and then metabolised to PGH 2 by COXs (COX-1 and COX-2). PGH 2 is unstable and subsequently converted into PGs, that is, PGD 2 , PGE 2 , PGF 2α , PGI 2 , and TXA 2 by the actions of their synthases PGDS (LPGDS and HPGDS), PGES (mPGES-1, mPGES-2, and cPGES), PGFS (AKR1B1 and PGFS/ABR1C3), PGIS, and TXAS, respectively. PGs bind to their receptors and activate different downstream signalling pathways. PGD 2 receptors, DP1 and DP2, activate the cAMP and PI3K pathways, respectively, while DP2 also represses the cAMP pathway. PGE 2 receptors EP2 and EP4 activate both cAMP and PI3K pathways, EP1 activates PKC and Ca 2+ pathways, and EP3 deactivates the cAMP pathway. Both PGF 2α receptor FP and TXA 2 receptor TP activate PKC and Ca 2+ pathways, whereas PGI 2 receptor IP triggers activation of cAMP signalling. On the other hand, non-steroidal anti-inflammatory drugs (NSAIDs) inhibit AA biosynthesis of all PGs by targeting COX-1 and/or COX-2 Osma-García, Cacheiro-Llaguno, Fresno, & Iñiguez, 2010). However, COX-2 and mPGES-1 expression is also observed in chronically inflamed tissues including joints of rheumatoid arthritis patients, colons of patients with inflammatory bowel disease, and cancerous tumours and their micro-environment (Ricciotti & FitzGerald, 2011;Wang & DuBois, 2018). Therefore, PGs appear to play integral roles during both acute inflammation and chronic inflammatory diseases.

Gordon et al. systemically mapped the interaction landscape between
SARS-CoV2 proteins, including four structural (S, E, M, N) proteins and non-structural proteins (Nsp), and human host cell proteins, and predicted numerous host proteins as potential drug targets. Remarkably, they found that non-structural proteins Nsp7 and Nsp14 interacted with PTGES2 (encoding mPGES-2) and PRKACA (encoding the catalytic subunit α of PKA), respectively (Gordon et al., 2020

| Potential roles of AA in COVID-19
AA is known to have potent antimicrobial capacity including leakage and lysis of microbial cell membranes, viral envelope disruption, amino acid transportation, inhibition of respiration, and uncoupling of oxidative phosphorylation (Das, 2018). It is reasonable to suggest Huh-7 cells and MERS-CoV-infected Huh-7/Vero cells (Müller et al., 2018), implying that anti-CoV treatments harnessing cPLA 2 α inhibition may be of potential therapeutic benefit. It remains unknown whether cPLA 2 α and its products (i.e., fatty acids) regulate CoV replication directly or indirectly through their further downstream metabolites of fatty acids, for example, PGs. Below, we will focus on discussion of possible functions of PGD 2 and PGE 2 even though other PGs may also influence SARS-CoV-2 infection and COVID-19. For example, PGI 2 and TXA 2 respectively can reduce and promote thrombosis, a typical complication that occurs in nearly half of critically ill COVID-19 patients and which contributes to mortality (Klok et al., 2020;Wise, 2020). Moreover, the vasodilatory and endothelial cell effects of PGs are well known.
However, here, we will focus on discussing the roles of PGs in COVID-19 immunopathology.

| Potential roles of PGs on thrombosis in COVID-19
The exact mechanisms behind the development of systemic coagulopathy and acquired thrombophilia defined in the majority of COVID-19 cases which can lead to venous, arterial, and microvascular thrombosis remain unclear (Becker, 2020). Indeed, clinical characteristics of COVID-19 include elevated D-dimer levels, prolonged thrombin time, and thrombocytopenia (platelet count <150,000Áμl −1 ), therefore suggesting an increased possibility of disseminated intravascular coagulation or pre-disseminated intravascular coagulation (Guan et al., 2020). Moreover, pooled analysis suggests that significant increases in D-dimer levels as a predictor of adverse outcomes were regularly observed in COVID-19 patient blood, implying the presence of underlying coagulopathy (Lippi & Favaloro, 2020). Although D-dimer levels can be altered by numerous inflammatory processes, in cases of COVID-19, it is almost certainly due to intravascular thrombosis (Cui, Chen, Li, Liu, & Wang, 2020;Leonard-Lorant et al., 2020). A retrospective cohort study found that on admission, increased D-dimer levels (>1,000 ngÁml −1 ) were associated with increased risk of death in hospitalised COVID-19 patients (F. Zhou, Yu, et al., 2020 for the production of platelets; therefore, the employment of antiplatelet agents may be of clinical benefit during COVID-19 pathogenesis. Aspirin is a broadly studied anti-platelet drug which exerts its cardioprotective effects via irreversible inhibition of platelet COX-1, thus blocking TXA 2 production in activated platelets, and so decreases prothrombotic events. However, aspirin does not confer platelet-specific effects, and in other cell types (via inhibition of COX-1 and in some cases COX-2), it can decrease prostanoid production, for example, PGI 2 , which serves to inhibit platelet aggregation (conversely to TXA 2 ). Interestingly, as determined by a pharmacodynamic interaction study in healthy volunteers, other NSAIDs including ibuprofen, naproxen, indomethacin, and tiaprofenic acid all block the anti-platelet effect of aspirin, whereas celecoxib and sulindac did not exhibit any significant anti-platelet effects (Gladding et al., 2008). Although aspirin can inhibit viral replication and confer anti-inflammatory and anti-coagulant effects, at present, it has not been thoroughly investigated in the treatment of thrombosis during COVID-19. However, COVID-19 clinical trials involving aspirin administration are ongoing. That said, PGs (whose production is blocked by NSAIDs) can also confer antiplatelet effects and therefore also merit clinical attention in the context of COVID-19. For example, it has been long understood that within the PG family, PGE 1 is the most potent inhibitor of ADP-induced platelet aggregation whereas PGE 2 possesses roughly a fifth of its activity (Irion & Blombäck, 1969). PGE 2 -EP4 receptor signalling does effectively inhibit platelet aggregation at high concentrations (>1 × 10 −6 M) (Macintyre & Gordon, 1975), and akin to PGI 2 , PGD 2 can also inhibit platelet aggregation (Smith, Silver, Ingerman, & Kocsis, 1974), but PGE 2 -EP3 receptor signalling augments platelet aggregation (Friedman, Ogletree, Haddad, & Boutaud, 2015).
Similarly, Vijay et al. (2015) found that expression of PLA 2 group IID (PLA2G2D) was increased in aged mouse lungs compared to younger mice, leading to augmented production of PGD 2 , PGE 2 , PGF 2α , and TXB 2 in lungs in response to SARS-CoV-1 infection.
Differential gene expression showed that the major source of PLA2G2D was lung-resident CD11c + cells, for example, alveolar macrophages and DCs (Vijay et al., 2015). Strikingly, aged mice with deficiency in PLA2G2D were protected from SARS-CoV-1 infection, exhibited enhanced virus-specific cytotoxic CD8 T cell responses, and increased migration of respiratory DCs to draining lymph nodes (Vijay et al., 2015). PGD 2 may contribute to SARS-

| PGE 2 on IFN signalling
Infection with SARS-CoV-2 ligates various pathogen recognition receptors, for example, TLRs and/or RIG-I-like receptors, and activates transcription factors such as IFN regulatory factor 3 (IRF3) and NF-κB that are responsible for expression of type I and III IFNs and proinflammatory mediators, including TNF-α, IL-6, and PGE 2 respectively. Secreted IFNs then activate the JAK-STAT1/2 pathway to trigger production of ISGs that directly recognise and execute antiviral functions (Park & Iwasaki, 2020

F I G U R E 3 Possible mechanisms for PGE 2 modulation of immune cell functions in COVID-19. PGE 2 is likely to modulate immune responses in various cell types during SARS-CoV-2 infection, influencing COVID-19 pathogenesis. In epithelial cells, attachment of SARS-CoV-2 with ACE2
and TMPRSS2 leads to endocytosis, viral replication, and cell damage, activating RLR (RIG-1 and MAD5)-dependent production of type I and III IFNs and the TLR-dependent NF-κB pathway. The NF-κB pathway induces expression of proinflammatory cytokines (e.g., IL-1b, IL-6, IL-8, and GM-CSF), chemokines (e.g., CCL2 and CXCL1), and other inflammatory mediators such as COX-2 and mPGES-1, resulting in PGE 2 secretion. Here, while it suppresses production of type I (and possibly type III) IFNs, PGE 2 further amplifies NF-κB signalling and production of cytokines and chemokines in a positive feedback loop. PGE 2 may also directly modulate ACE2 and TMPRSS2 gene expression, endocytosis, and viral replication. In monocytes/macrophages, activation of NF-κB and STAT3 mediates production of large amounts of inflammatory cytokines which contributes to development of cytokine secretion syndrome (the "cytokine storm"), chemokines that recruit monocytes and neutrophils, inflammatory biomarkers (e.g., SAAs, CPR, and D-dimer), and PGE 2 . Here, PGE 2 again represses IFN-induced expression of ISGs, contributing to delay of viral clearance. Importantly, PGE 2 context-dependently affects (either positively or negatively) not only NLRP3 inflammasome activation and related IL-1β maturation but also NF-κB-dependent monocyte/macrophage cytokine production. PGE 2 differentially regulates platelet aggregation via different receptors and probably inhibits NETosis associated with inflammation and thrombosis. PGE 2 also down-regulates IFN-γ production and cytotoxicity of NK and CD8 T cells that kill cells infected with SARS-CoV-2 but promotes differentiation of proinflammatory Th17 and Th1 cells, the chief cellular sources of the cytokine storm at late stages of COVID-19 produced from epithelial cells, monocytes, and macrophages, shortly after clinical symptoms appear (Mehta et al., 2020). Many clinical trials have been set up to treat COVID-19 by targeting pathways relating to proinflammatory cytokine production, for example, IL-1 and IL-6 (Bonam, Kaveri, Sakuntabhai, Gilardin, & Bayry, 2020; Merad & Martin, 2020). NF-κB is the key transcription factor responsible for induction of proinflammatory cytokines. Activation of NF-κB can stimulate gene expression of inducible COX-2 and mPGES-1 in many cell types, leading to production of COX-2-dependent PGE 2 . This PGE 2 acts autocrinally and/or paracrinally on NF-κB stimulation for expanding of proinflammatory cytokines and chemokines through the EP2 (maybe also EP4) receptors (Aoki et al., 2017;
Careful examinations are thus required to clarify the effects of PGE 2 upon NLRP3 inflammasome activation, for example, primed by different SARS-CoV-2 proteins in different cell types.

| PGE 2 on monocyte/macrophage functions
Monocytes and macrophages are main sources of proinflammatory and anti-inflammatory cytokines and generally function to eliminate pathogens. Expansion of IL-6-producing CD14 + CD16 + monocytes was observed in peripheral blood from severe COVID-19 patients (X. Zhang, Tan, et al., 2020;Y. Zhou, Fu, et al., 2020), but reduction of HLA-DR on CD14 + monocytes was found in COVID-19 patients with severe respiratory failure, which was associated with  2017) found an increase of CD14 + CD16 + monocytes in patients with severe sepsis or septic shock, which was positively associated with disease severity. After in vitro culture of monocytes from sepsis patients, PGE 2 diminished CD14 + CD16 + monocytes after 24 h, reduced TNF-α production, but enhanced anti-inflammatory IL-10 production (Qiu et al., 2017).
High amounts of IL-1β and PGE 2 are mainly produced from the classic inflammatory CD14 + CD16 − human monocytes after Candida albicans infection (Smeekens et al., 2011). Increase in inflammatory monocytes was indicated to be associated with increased survival rate at least in Gram-negative sepsis patients (Gainaru et al., 2018).
Hence, reduction of inflammatory CD14 + monocytes by PGE 2 may sustain immunosuppression and could be associated with poor clin- Single-cell RNA sequencing analysis of cells in BAL fluid from COVID-19 patients suggested that monocyte-derived macrophages, but not alveolar macrophages, contribute to lung inflammation and damage in severe COVID-19 patients . Bulk RNA sequencing analysis of BAL cells also suggested increased production of chemokines such as CCL2 and CXCL1 Xiong et al., 2020;Z. Zhou, Ren, et al., 2020), which recruit CCR2-expressing classical monocytes and neutrophils, respectively, to the lung from peripheral blood. Interestingly, PGE 2 -EP2 receptor signalling increases CCL2 and CXCL1 production from macrophages and other cells (Aoki et al., 2017;. During human monocyte/macrophage differentiation, cAMP-elevating reagents such as PGE 2 can cause a large increase in the mRNA and protein levels of several proinflammatory CCL and CXCL chemokines, contributing to the pathogenesis of lung disease (Hertz et al., 2009). Furthermore, alternatively activated M2 macrophages differentiated from monocytes promote tissue repair by secreting reparative cytokines such as TGF-β, amphiregulin (AREG) and VEGF (Wynn & Vannella, 2016). As PGE 2 facilitates generation of M2 macrophages, it may thus also contribute to lung fibrosis in severe COVID-19 patients.

| PGE 2 on NET release
The formation of neutrophil extracellular traps (NETs) is an evolutionarily ancient process which involves the release of de-condensed nuclear chromatin studded with various antimicrobial proteins, such as core histones, neutrophil elastase and MPO, to the extracellular space where they serve to trap and kill invading microorganisms (Brinkmann et al., 2004;Fuchs et al., 2007;Robb, Dyrynda, Gray, Rossi, & Smith, 2014). However, aberrant NET formation is involved in a wide range of NET-associated diseases. Hence, NETs are regarded as double-edged swords in innate immunity (Kaplan & Radic, 2012). Given the important role played by NETs in the pathogenesis of various respiratory diseases and thrombosis, many researchers also suggest NETs as key players in the pathogenesis of COVID-19, most probably via NET-mediated release of excessive amounts of IL-6 and IL-1β during cytokine storms in the COVID-19 milieu (Barnes et al., 2020;Mozzini & Girelli, 2020;Thierry, 2020;Thierry & Roch, 2020;Tomar, Anders, Desai, & Mulay, 2020). Indeed, there is a high degree of NET-IL-1β interplay during both venous and arterial thrombosis and severe asthma (Lachowicz-Scroggins et al., 2019;Liberale et al., 2019;Yadav et al., 2019), and it has been hypothesised that there may be therapeutic potential in targeting the IL-1β/NET feedback loop (Yaqinuddin & Kashir, 2020). It is proposed that upon SARS-CoV-2 infection, activated endothelial cells recruit neutrophils where they release NETs, which in turn activates the contact pathway of coagulation, subsequently trapping and activating platelets to potentiate blood clotting (Merad & Martin, 2020). NETs contribute to immunothrombosis in COVID-19 ARDS where pulmonary autopsies confirmed NET-associated microthrombi with neutrophil-platelet infiltration (Middleton et al., 2020). Such NETinduced immunothrombosis may help explain the prothrombotic clinical presentations observed in COVID-19 patients (Middleton et al., 2020). Indeed, NETs have been identified as potential markers of disease severity in COVID-19 , and elevated NET formation in hospitalised COVID-19 patients is associated with higher risk of thrombotic episodes . Serum samples of hospitalised COVID-19 patients contained greater cell-free DNA and hallmark NET-associated products, including MPO-DNA complexes and citrullinated histone H3, compared to healthy control serum samples . Furthermore, serum of COVID-19 patients requiring mechanical ventilation exhibited augmented cell-free DNA and MPO-DNA complexes, compared to patients breathing room air ). An additional NET marker, calprotectin, was found to be present at prominently elevated levels in the blood of 172 COVID-19 patients (H. Shi, Zuo, et al., 2020). PGs are reported to have inhibitory effect upon NET formation. PGE 2 inhibited NET formation after stem cell transplant (Domingo-Gonzalez et al., 2016) and via EP2-and EP4 receptor-mediated activation of cAMP (Shishikura et al., 2016). After co-culture of neutrophils with cAMP-elevating reagents, PMAinduced NET formation was significantly reduced (Shishikura et al., 2016). The adenylate cyclase toxin which vastly increases intracellular cAMP is also known to reduce NETs (Eby, Gray, & Hewlett, 2014). Interestingly, induction of intracellular cAMP production by PGE 1 markedly constrains NETs induced by the pancreatic cancer cell line AsPC-1 (Jung et al., 2019). Importantly, CGS21680, a selective agonist of the adenosine A 2A receptor (which increases intracellular cAMP), successfully diminished NET formation mediated by antiphospholipid antibodies, which increased the incidence of thrombotic events (Ali et al., 2019). Furthermore, in mice treated with antiphospholipid antibodies, CGS21680 impaired thrombosis within the inferior vena cava (Ali et al., 2019). Here, the authors also demonstrate similar inhibition of NETs via dipyridamole, an antithrombotic medication which increases extracellular adenosine and impedes cAMP breakdown (Ali et al., 2019). Together with the known inhibitory effects of PGE 2 on platelet function (Gross et al., 2007), such evidence suggests that there may be the potential for therapeutic gain in harnessing existing drugs that augment production of PGs and cAMP and thus reduce excess NET formation and the incidence of thrombotic events during SARS-CoV-2 infection.   that contribute to the cytokine storm and immunopathology, but viral infection may also activate regulatory T (Treg) cells which limit immunopathology through mechanisms such as the production of anti-inflammatory cytokines (e.g., IL-10). After infection with SARS-CoV-2, a significant reduction of peripheral blood CD4 + and CD8 + T

cells (a condition known as lymphopenia) results in moderate to
severe COVID-19 patients, which is correlated with disease severity and mortality (G. Chen, Wu, et al., 2020;Diao et al., 2020;Liu, Li, et al., 2020;Tan et al., 2020). In contrast to peripheral blood, mass lymphocyte infiltrations were observed in the lung as confirmed by post-mortem examination of a COVID-19 patient who suffered from ARDS . PGE 2 has multifaceted effects on modulation of T cell responses . PGE 2 suppresses T cell receptor-dependent T cell activation and proliferation via EP2/EP4 receptor-mediated cAMP-PKA pathway, but this suppressive effect is weakened by enhancing CD28 co-stimulation through augmentation of PI3K signalling (Yao et al., 2013). Following SARS-CoV-2 infection, the pathway related to CD28 signalling in T helper cells was significantly down-regulated, while the PKA pathway and PGE 2 biosynthesis pathway were significantly up-regulated (Z. Zhou, Ren, et al., 2020;Yan et al., 2020). Thus, PGE 2 -cAMP-PKA signalling is likely to inhibit antigen-dependent activation of antiviral T cell responses in COVID-19 patients. On this basis, use of NSAIDs by inhibiting endogenous PGE 2 production may enhance antiviral T cell responses in COVID-19 patients. Indeed, enhanced viral antigen-specific CD8 + and CD4 + T cell responses in lungs were found in PTGES-deficient mice where PGE 2 production was largely reduced, compared to WT mice, post-IAV infection, and this was associated with reduced viral load in PTGES-deficient animals (Coulombe et al., 2014). Moreover, increased expression of the immune checkpoint proteins, PD-1 (CD279) and TIM-3 (CD366), on CD8 + T cells was detected in severe and critical COVID-19 patients (Diao et al., 2020;Y. Zhou, Fu, et al., 2020), As for inflammatory T cells, PGE 2 regulates Th1 cell differentiation dependently on the strengths of T cell receptor and CD28 costimulation and timing of PGE 2 encounter (Yao et al., 2013). Given the down-regulation of CD28 signalling, immediate up-regulation of generelated PGE 2 synthases (i.e., PTGS1, PTGS2, and PTGES3) at the early stages post-SARS-CoV-2 infection, and the kinetics of PGE 2 secretion in COVID-19 patients Yan et al., 2020;Z. Zhou, Ren, et al., 2020), it is proposed that PGE 2 may inhibit the development of inflammatory IFN-γ-producing Th1 cells, although further investigations are required. This possibility is further supported by the findings that PGE 2 inhibits monocyte-derived DCs and macrophages to produce IL-12 (Kali nski, Hilkens, Snijders, Snijdewint, & Kapsenberg, 1997;van der Pouw Kraan, Boeije, Smeenk, Wijdenes, & Aarden, 1995), the key cytokine for generating IFN-γ-producing Th1 cells. Of note, IFN-γ production from CD8 + T cells and NK cells was not significantly different in blood of non-severe and severe COVID-19 patients, but IFN-γ + CD4 + T cells were likely to be reduced in severe, compared to non-severe, COVID-19 patients (G. Chen, Wu, et al., 2020;Qin et al., 2020). Th17 cells highly express IL-17A, IL-17F, IL-22, and GM-CSF that contribute to COVID-19 immunopathology.
There are mixed findings regarding Treg cells in COVID-19.
Reduction of Treg cell frequencies was observed in severe COVID-19 patients compared to non-severe patients (G. Chen, Wu, et al., 2020;Qin et al., 2020), but Y. Shi, Tan, et al. (2020) reported an increase in Treg cell numbers in peripheral blood of mild COVID-19 patients compared to the control group, and there was no difference in Treg cell numbers between mild/moderate and severe patients. Both stimulatory and inhibitory effects of PGE 2 on human Treg cell generation and suppressive function were observed Schiavon et al., 2019). The definitive functions of PGE 2 on Treg cells in peripheral blood and lungs of COVID-19 patients remain to be determined.

| PGE 2 on NK cell functions
Like T cells, NK cells are also depleted in peripheral blood of severe, but not mild, COVID-19 patients Wilk et al., 2020;Zheng et al., 2020). NK cell activation may also be impaired in COVID-19 patients due to down-regulation of CD107a and cytokines such as IFN-γ and TNF-α but increased expression of NKG2A/CD94 that inhibits NK cell cytotoxicity . These studies suggest that NK cell numbers and function are impaired in severe COVID-19 patients. PGE 2 was found to inhibit not only NK cell production of IFN-γ but also myeloid cell production of IL-12 that is required for IFN-γ production through EP4 receptors (Van Elssen et al., 2011). The PGE 2 -EP4 receptor-cAMP signalling pathway also suppresses the cytolytic activity of NK and CD8 + T cells by increasing expression of NKG2A/CD94 (Holt, Ma, Kundu, & Fulton, 2011;Park et al., 2018;Zeddou et al., 2005). Moreover, PGE 2 inhibits CXCR3 ligands such as CXCL9 and CXCL10 from antigen-presenting cells, preventing NK cell migration (Gustafsson et al., 2011). It is thus likely that PGE 2 both directly and indirectly inhibits NK cell migration to the lung, their activation and cytotoxic function, in the context of COVID-19. However, this remains to be confirmed.

| Potential roles of PGI 2 in COVID-19
Most human lung cells including stromal (fibroblasts), immune (monocytes, macrophages, and lymphocytes), and vascular endothelial cells express PGI 2 synthase (PGIS) and the PGI 2 receptor IP ( Figure 2). As PGI 2 also activates the cAMP-PKA signalling pathway, as does PGE 2 -EP2/EP4 receptor signalling, it is assumed that these two molecules may share similar effects on modulation of immune and inflammatory responses (see Dorris & Peebles, 2012) although PGI 2 has been less extensively studied, compared with PGE 2 . Firstly, PGI 2 is likely to suppress virus (e.g., RSV)-induced type I IFN production in the lung, which contributes to protection against viral infection (Hashimoto et al., 2004;Toki et al., 2013). Overexpression of PGIS in bronchial epithelium decreased viral replication and limited weight loss while IP deficiency exacerbated RSV-induced weight loss with delayed viral clearance and had greater IFN-α and β protein expression post-RSV challenge (Hashimoto et al., 2004;Toki et al., 2013).
Of note, despite inhibition of COX-2 by both NSAIDs and steroids, they have different in vivo actions in managing inflammatory diseases.
For example, while dexamethasone only had moderate effects on reducing mortality of severe/critical (but not non-severe) COVID-19 patients (Horby et al., 2020), celecoxib reduced disease severity and improved remission in both non-severe and severe COVID-19 patients , indicating different clinical outcomes from use of NSAIDs and other anti-inflammatory agents. Nevertheless, further studies are essential to clarify the clinical safety and efficiency of these two families of anti-inflammatory drugs and to understand the underlying immunomodulatory mechanisms.

| CONCLUSIONS
Here formation, inflammasome activation, and inflammatory cytokine production, whereas it can also contribute to inflammatory Th17 responses, NF-κB activation, and related inflammatory cytokine production. These effects of PGs rely, to a great extent, on the context and micro-environments such as strength (e.g., viral load) and timing of stimuli, organ locations, and responding cell types. Given the critical role of cytokine storms in COVID-19 immunopathology and the context-dependent regulation of cytokine production by PGs, it is imperative to understand the chief cellular sources of cytokines in the lung and peripheral blood, the key stimuli (e.g., SARS-CoV proteins or peptides) and to elucidate the kinetics of cytokine secretion. Results from such studies may be insightful for considerations on when

| Nomenclature of targets and ligands
Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018), and are permanently archived in the Concise Guide to PHARMACOLOGY 2019/20 (Alexander, Christopoulos, et al., 2019;Alexander, Fabbro, et al., 2019a, b;.