Lipids serve multiple functions in the cell. They reside mostly in the cell membranes, where they can be distributed distinctly within membrane leaflets and in cholesterol rich nanodomains according to their biophysical and biochemical properties. These lipid nanodomains, (historically called “lipid rafts”), concentrate coronavirus receptors and their disruption affects the expression and proper localization of coronavirus receptors, impacting virus entry (1–7). Virus structural proteins are also found in these lipid domains due to their affinity to cholesterol and this localization seems to be important for the induction of cell-cell fusion observed during coronavirus infections (6, 8, 9).

At the plasma membrane we can also find caveolae, invaginations with lipid nanodomain-like organization. Although caveolae are not directly relevant for coronaviruses entry, one of its main components, caveolin-1, colocalizes with some coronavirus receptors, seems to be involved in virus infectivity and may participate in virus fusion to the plasma membrane in a cholesterol dependent manner (2, 10).

Besides its role in the membrane, cholesterol participates in coronavirus cycle and host response against virus infection in diverse ways. Key components of the cholesterol metabolism are altered and associated with virus infectivity, inflammatory response, and severity of COVID-19 and MERS (8, 11–13). Cholesterol esterification and triacylglyceride synthesis leads to the biogenesis of lipid droplets, organelles where coronavirus replicate and that contribute to the production of inflammatory mediators (14–18). Cholesterol oxidized products also modulate viral infection and host responses (19–31). Fatty acid metabolism is involved in the acylation of the spike protein, contributing to its localization in lipid nanodomains and trafficking, virus entry and assembly (32–36). Other relevant aspects of lipid metabolism for coronaviruses include the role of specific sphingolipids in virus infection and disease severity (37–45). In this review, we discuss some of these roles and their impact on infectivity and host responses. Besides, we present a large collection of results obtained using in vitro and in vivo coronavirus infection models that support the therapeutical benefits of using lipid metabolism drugs as strategies against coronaviruses.

We begin with a brief background on lipid domains and cholesterol metabolism to provide the basis for understanding their involvement in coronavirus infection.

Lipid nanodomains are nanoscale cell membrane domains rich in cholesterol, glycosylated, and saturated lipids, such as sphingolipids, assembled in a liquid-ordered (Lo) compact arrangement (46–48). These domains also segregate glycosylphosphatidyl (GPI)-anchored and lipidated proteins, forming signal transduction platforms (47). The existence of lipid nanodomains has been a matter of intense debate in the past, as their visualization can be challenging due to their transient character in quiescent cells. However, upon stimulation, receptors in these domains cluster, inducing lipid nanodomains coalescence, signaling protein recruitment, and interaction with the cortical actin filaments (49, 50). Solubilization of cell membranes with non-ionic detergents at cold temperatures (usually, Triton X-100 at 4°C) allows isolation of domains, segregating GPI-anchored proteins and Lo domain probes. Thus, this method is often used to separate membrane domains representing lipid nanodomains, although there is controversy about this correspondence. More precisely, this approach helps the isolation of Lo domains and the associated molecules, drawing the inference that these molecules would be associated with lipid nanodomains in the living cell (51). Proteins in these domains display specific characteristics that favor their compartmentalization. Moreover, proteins can be anchored in membranes by different lipid modifications by co- or post-translational modifications, some reversible while others are permanent. Such lipid anchors include glycophosphatidylinositol (GPI) anchors, N-terminal myristic acid tails, cysteine acylation, isoprenylation, and the addition of C-terminal sterol moieties. S-acylation is the reversible and enzyme-mediated attachment of a fatty acid to a cysteine residue via a thioester linkage (called palmitoylation because palmitic acid is frequently the added acyl chain). This process targets proteins to the inner leaflet of the cell membrane, preferentially to lipid nanodomains. Cholesterol-depleting and -sequestering drugs, such as MβCD and nystatin disrupt these lipid nanodomains hampering intracellular signaling triggered by molecules organized in these domains.

Caveolae are specialized lipid nanodomains that form invaginations of 50-200 nm in the plasma membrane. These structures likely serve as safeguards when sudden changes in membrane tension occur, for example, during hypotonic shock or mechanical stretch. In these situations, caveolae flatten, providing membranes required for cellular reshaping (52, 53). Caveolae are also crucial for transcytosis (54, 55) and cell signaling. Significantly, they are also involved in controlling lipid metabolism, the main subject of this review (56). Caveolae bear specific proteins, which are determinants for sculpting these structures and executing caveolar functions. These proteins belong to two prominent families: caveolins and cavins. The caveolin family comprises three members of integral membrane proteins named caveolin-1, caveolin-2, and caveolin-3. Caveolin-1 and -2 are ubiquitous but are enriched in endothelial cells, adipocytes, pneumocytes, and fibroblasts, whereas caveolin-3 is specific to muscle cells (57–61). Caveolin-1 and caveolin-3 share a high degree of identity. They are essential for caveolae formation, whereas caveolin-2, the most divergent member of the family, is not required but favors the formation of deeper caveolae (62, 63). The interaction of signaling molecules with caveolins contributes to their compartmentalization and functions. Although the absence of caveolins is not lethal (64–66), their deficiency is associated with various diseases (67–71). The cavin family comprises four cytosolic proteins: cavin1, cavin2, cavin3, and cavin4; the latter being specific to muscle cells. Cavin1 hetero-trimerizes with the other cavins and is essential for caveolae formation (72–74). The other cavins are not essential, except for cavin 2 in endothelial cells of specific tissues (75), but are important to shape and stabilize caveolae (76–79). Lipids play essential roles in caveolae dynamics and functions. Caveolae are enriched in cholesterol, sphingomyelin, and glycosphingolipids, like GD3 (80). Cholesterol binding to caveolin-1 promotes oligomerization (81) and trafficking from the Golgi to the plasma membrane (82). Cholesterol has a vital role in the caveolae structure, which are also disrupted by treatment with cholesterol-depleting drugs (83–85). Furthermore, not only do lipids affect caveolin expression and caveolae formation and motility, but caveolin expression also regulates lipid homeostasis. In the absence of caveolin-1, alterations in cholesterol metabolism occur, such as reduction of free cholesterol synthesis, enhancement of cholesterol esterification (86), and reduction of cholesterol in lipid droplets (87). Caveolin-1 can also be found in lipid droplets where it is believed to play a role in the stabilization, as well as in the lipolysis of these organelles (82, 88–90). Caveolin itself and caveolae as a whole, traffic to lipid droplets, increasing their size and lipid uptake (91). Taken together, all these data corroborate the relevance of caveolae to lipid metabolism.

The lipid droplet (LD) is an organelle where several hydrophobic reactions involved in lipid, energy, and redox homeostasis take place (92–94). Lipid droplets contain a hydrophobic core composed of neutral lipids, mainly triacylglycerol (TG) and cholesterol ester (CE) and are delimited by a monolayer of phospholipids (95–97) and associated proteins (98–100) Although LDs are ubiquitous, their protein and lipid composition strongly depend on the cell type and cellular metabolic state (101–103). Their numbers and sizes are also variable due to the balance between lipid synthesis and degradation (97, 104, 105). The biogenesis of LDs occurs in the ER (106), where critical enzymes involved in lipid synthesis are located, mainly diacylglycerol O-acyltransferases (DGAT1 and DGAT2) (107, 108) and acyl-CoA: cholesterol O-acyltransferases (ACAT1/SOAT1 and ACAT2/SOAT2) (109–111), responsible for the synthesis of TG and CE, respectively (94, 112). The lipid metabolism pathways leading to lipid uptake, cholesterol efflux, autophagy, β-oxidation, lipophagy, and lipid remodeling (109, 113–118) impact LD accumulation or consumption (119–122). Besides, LDs play a role in lipid availability and membrane biosynthesis (123, 124).

Cellular cholesterol content is regulated by uptake, de novo synthesis, storage, and export. Diet-derived cholesterol is absorbed by the enterocytes, transported to intracellular organelles, and released in the bloodstream (125), where it circulates associated with lipoproteins, like chylomicrons, very low-density proteins (VLDL), low-density lipoproteins (LDL), and high-density lipoproteins (HDL). LDL is recognized by LDL receptors on cell surface and is endocytosed by clathrin-coated vesicles (126). LDL-bound cholesterol ester (CE) is then hydrolysed by lysosomal acid lipase (LAL) in the lysosomes, from where vesicular and non-vesicular mechanisms transport cholesterol to other organelles (127). The plasma membrane holds 60-90% of this lipid (128). Studies using radiolabeled cholesterol binding probes indicated the existence of three distinct cholesterol pools: one that binds the probe, named the accessible pool, which is sensitive to cholesterol deprivation; the SM-sequestered pool that is not affected by cholesterol deprivation and binds the probe only after SMase treatment; and the essential pool which does not bind to the probe, even after SMase treatment and whose depletion causes rounding and dissociation of the cell from the culture dish. Interestingly, the excess of the accessible pool traffic to the ER, the organelle responsible for cholesterol synthesis, starting with acetyl-CoA and involving more than 20 enzymes (128).

Despite being the center of cholesterol synthesis, the ER disposes of only 1% of the total cellular cholesterol due to the fast transport of newly synthesized cholesterol to cellular membranes or lipid droplets (129). The master regulator of cholesterol synthesis is sterol regulatory element-binding protein 2 (SREBP2), a transcriptional factor synthesized as a precursor protein. In the ER, SREBP2 forms a heterodimer with SCAP (SREBP-cleavage activating protein), a sensor of cholesterol level (130). When ER’s cholesterol level is below 5% of total ER lipids, SCAP binds COPII in transport vesicles and escorts SREBP2 to the Golgi, where it is processed and activated by site 1 (S1P) and site 2 (S2P) proteases (131). The activated soluble N-terminus SREBP2 goes to the nucleus and binds to sterol regulatory elements (SRE) found in genes coding for proteins involved in cholesterol synthesis, like 3-hydroxy-3-methylglutaryl coenzyme A reductase (HMGCR) and squalene monooxygenase (SQLE). When cholesterol levels increase in the ER, cholesterol binds SCAP, blocking its interaction with COPII, thus inhibiting SREBP2 transport to the Golgi. Besides, cholesterol binding to SCAP induces its interaction with INSIG1 and INSIG2 (insulin-induced gene-1 and -2 proteins), which also interferes with SCAP’s association with COPII (132). Furthermore, the cell manages the cholesterol excess by storing or exporting it. Acyl-coenzyme A: cholesterol acyltransferase (ACAT) esterifies cholesterol that can be stored in LDs. CE from lipid droplets is released by the CEHs associated with the lipid droplet surface and is transported to other organelles or the extracellular milieu (133). ABC-binding cassette (ABC) transporters mediate the efflux of non-esterified apolipoprotein-bound cholesterol originated from diverse organelles ( Figure 1 ) (134).

In the next sections, we will systematize information about how distinct aspects of lipid metabolism favor coronavirus infection or the host response against it. Moreover, we will present results obtained with a wide variety of drugs that aim the lipid metabolism as a strategy to fight coronavirus infection and the harmful consequences of the triggered host inflammatory response. Finally, we will focus on drugs that were tested in in vivo animal models and clinical studies.

Coronaviruses bind to specific plasma membrane proteins, for instance, the angiotensin-converting enzyme-2 (ACE2) in the case of SARS-CoV and SARS-CoV-2 and dipeptidyl peptidase 4 (DPP4) in the case of MERS-CoV. Besides these receptors, other molecules facilitate the attachment of these viruses ( Table 1 ). After binding to cell receptors, coronaviruses can fuse directly to the plasma membrane or enter by endocytosis mediated by different mechanisms, depending on the virus and the cell type, and that involve lipid nanodomains, clathrin-coated vesicles, caveolae, or clathrin- and caveolae-independent pathways. Direct fusion with the plasma membrane relies on the proteolytic processing of Spike by proteases, like TMPRSS2 in the case of SARS-CoV-2. Otherwise, Spike can be processed by cathepsins activated by the lower pH found in endosomes (146). Thus, the cellular repertoire of receptors, proteases, and other cellular factors affect the virus entry route. For instance, in Calu-3 cells that express TMPRSS2, SARS-CoV-2 enters by fusion, whereas in Vero E6 cells, it enters by an endocytic pathway. The inhibition of infection by drugs that raise the endosomal pH, such as chloroquine, hydroxychloroquine, NH4Cl, and bafilomycin, indicate the requirement of endocytosis for successful infection (147–151).

In several studies, lipid nanodomains were implicated in coronavirus entry. Disruption of lipid nanodomains by cholesterol-depleting drugs, like MβCD, reduces coronavirus infectivity, whereas cholesterol replenishment restores virus infectivity (1–6). Nevertheless, the timing is crucial for this inhibition as MβCD pretreatment impairs SARS-CoV infection in Vero E6 cells, whereas a 3h post-treatment has no effect, indicating that cholesterol depletion hampers early steps of the virus cycle (3). Similarly, pretreatment, but not post-treatment with simvastatin, a cholesterol-reducing drug, inhibits SARS-CoV-2 replication in Vero E6 and Calu-3 cells. Accordingly, simvastatin impairs binding and entry of SARS-CoV-2 in Calu-3, possibly due to the dislodgment of ACE2 from lipid nanodomains caused by the drug and despite the increase in ACE2 expression that both simvastatin and infection provoke in these cells (7). MβCD treatment also induces relocation of ACE2 from nanodomains to membrane non-raft regions without impacting ACE2 expression levels in Vero E6 or Caco-2 cells (4, 5). Furthermore, a tagged-ectodomain of the SARS-CoV Spike also colocalizes with detergent-resistant membrane domains and with GM1 in fixed cells, which indicates the presence of ACE2 in lipid nanodomains upon Spike binding (4). In contrast to the view that the presence of ACE2 in lipid nanodomains favors SARS-CoV-2 infection, Wing and coworkers (2023) showed that avasimibe - an ACAT inhibitor - increases both free cholesterol in the plasma membrane and the size of GM1-enriched lipid nanodomains, as well as the ACE2 localization in these domains, while reducing SARS-CoV-2 attachment to Vero E6 cells (152) Structural proteins of the avian infectious bronchitis virus (IBV) also localize to detergent-resistant fractions and move to soluble fractions after MβCD treatment. MβCD, as well as mevastatin, another member of the statins’ drug class, also reduces the expression of these proteins (6). In contrast, Li and coworkers (2007) could not detect ACE2 in lipid nanodomains using Vero E6 cells and reported decreased ACE2 expression after MβCD treatment (3). This cholesterol-depleting drug also inhibits mouse hepatitis virus (MHV) replication. Still, the mechanisms cannot be attributed to the relocation of its receptor since CEACAM is detected in detergent-resistant membrane domains when a mild concentration of cold Triton X-100 (0.2%) is used to isolate this receptor in CEACAM-expressing HeLa cells membranes. However, an exogenous GPI-anchored CEACAM expression in detergent-resistant domains (isolated with 1% Triton X-100) fails to enhance MHV infection (1). Nevertheless, Choi and coworkers demonstrated that in the presence of MHV, CEACAM, and Spike redistribute into lipid nanodomains in CEACAM-overexpressing HEK293 cells. These data suggest that cholesterol depletion by MβCD does not affect virus binding but reduces virus entry drastically, possibly because viral proteins redistribute and interact with lipid nanodomains. Spike is detected only in detergent-soluble membranes of the Golgi, meaning that Spike is not associated with lipid nanodomains during virus assembly and budding but is associated with detergent-resistant membrane domains at the plasma membrane, which is important for Spike-induced cell-cell fusion (153). Accordingly, Spike of coronaviruses contains several cholesterol-binding motifs in their sequences (8, 9).

Cholesterol also participates in distinct entry pathways, affecting caveolae- and clathrin-mediated endocytic mechanisms (149, 154). Therefore, excluding alternative entry routes using specific tools is greatly valuable. Wang and coworkers report lipid raft-mediated entry of SARS-CoV in a mechanism independent of caveolae and clathrin in Vero E6 and HEK293-ACE2-GFP cell lines. Clathrin-mediated endocytosis was excluded based on experiments using chlorpromazine (an inhibitor of clathrin-mediated endocytosis), by silencing the clathrin coding gene, or by expressing a dominant-negative Eps15, which participates in clathrin-mediated endocytosis. Caveolae-dependent endocytosis was also excluded by experiments using filipin or nystatin, which inhibited endocytosis of cholera toxin B, used as a control but failed to inhibit SARS-CoV endocytosis. Moreover, caveolin-1 does not colocalize with the SARS-CoV Spike (155).

On the other hand, other studies found that SARS-CoV and SARS-CoV-2 invade HepG2 cells (a hepatocarcinoma cell line) and HEK-293-ACE2 cells, respectively, by clathrin-mediated endocytosis (156, 157). MHV-2 also uses clathrin-mediated endocytosis to enter the mouse astrocytoma DBT cells, which was demonstrated by using chlorpromazine, hypertonic sucrose medium, or by silencing the clathrin heavy-chain gene. Notwithstanding, the clathrin-mediated endocytosis of MHV-2 is Eps15-independent. Besides, the overexpression of a dominant-negative mutant of caveolin-1 did not affect the invasion of MHV-2 in DBT cells, ruling out the participation of this protein or caveolae in virus entry (158). Porcine hemagglutinating encephalomyelitis virus (PHEV) also uses clathrin-mediated endocytosis to invade mouse neuroblastoma cells (Neuro-2a). Entry by endocytosis is evidenced by virus presence in coated pits and EEA1-decorated vesicles and by the inhibition of invasion induced by pH-raising agents, such as chloroquine and NH4Cl. Clathrin’s involvement in PHEV entrance is further demonstrated by PHEV and transferrin coincidental labeling, chlorpromazine sensitivity, and low infectivity detected in a dominant-negative clathrin mutant-expressing or a clathrin heavy chain-silenced cell line. Additionally, silencing of caveolin-1 does not affect PHEV entry in Neuro-2a cells (150).

Other studies propose that caveolin-1, and not necessarily caveolae, are involved in coronavirus infection. The human coronavirus 229E binds to CD13, enriched in membrane detergent-resistant domains where it colocalizes with caveolin-1. MβCD does not hamper virus binding to the cell surface but reduces the colocalization of 229E with caveolin and virus infectivity, while cholesterol replenishment reverses these effects. The silencing of caveolin-1 does not impair 229E binding but reduces its entrance into the cell significantly. 229E particles are detected close to caveolae, but interestingly they are not found in the lumen of caveolae or endosomes. Thus, caveolin-1 may participate in the fusion of the virus particle with the plasma membrane but not in caveolae-mediated endocytosis of the 229E virus (2). Notably, epithelial cells, the primary targets for coronaviruses, do not present caveolae in their apical surfaces, although they express caveolin-1 on these sites (159, 160). The human coronavirus OC43, which uses HLA class I molecule or sialic acid as receptors, colocalizes to caveolin-1 in HCT-8 cells during the early phase of infection. Nystatin, MβCD, or CAV1 silencing inhibits virus invasion and infectivity. The requirement of dynamin was also proved using dynasore and MiTMAB (10). Zhou and colleagues (2022) further showed that SARS-CoV-2 pseudovirus uses CD147 to enter Vero E6 and Huh-7 cells and exploit an Arf-6-dependent caveolar/lipid raft pathway (161).

Enveloped viruses may also present cholesterol in their membranes (162, 163), and cholesterol depletion from the viral envelope of the transmissible gastroenteritis virus (TGEV) by MβCD reduces its infectivity drastically, whereas cholesterol replenishment reverts this effect (164). Curiously, the Spike protein, found in detergent-resistant domains of the TGEV envelope membrane, does not become detergent-soluble after MβCD treatment and does not segregate with the lipid raft marker flotillin-2, indicating that the viral membrane organization is different from that of the plasma membrane (165).

Structural and modeling studies identified that the N-terminal domain of the Spike protein of SARS-CoV-2 interacts with ganglioside sialic acids (166) (see Table 1 ). SARS-CoV-2 and other betacoronaviruses use these molecules to attach and invade host cells (140, 167). Interestingly, gangliosides are a group of sphingolipids enriched in lipid nanodomains.

In addition to functioning as alternative receptors, sphingolipids metabolism is involved in other aspects of SARS-CoV-2 infection. The central hub of sphingolipid metabolism is ceramide, which can be synthesized by sphingomyelin hydrolysis, de novo synthesis, or recycling from complex sphingolipids ( Figure 2 ). COVID-19 patients’ sera display increased levels of dihydrosphingosine, dihydroceramides, ceramides, sphingosine, and acid sphingomyelinase (ASM) correlating positively with disease severity. In contrast, the level of sphingosine-1-phosphate (S1P) correlates negatively with the disease severity (37). Accordingly, S1P can activate the S1P receptor and induce anti-inflammatory responses. Another study shows that patients with symptomatic COVID-19 have diminished serum sphingosine levels compared to asymptomatic counterparts, who also exhibit elevated acid ceramidase (AC), an enzyme that converts ceramide in sphingosine (38). Accordingly, in vitro treatment of Vero or human primary epithelial cells with exogenous sphingosine protected these cells from infection by pseudoviruses expressing the SARS-CoV-2 Spike protein, possibly due to the binding of the exogenous sphingosine to ACE2, hampering Spike’s interaction with its receptor (39). Notably, in this pseudovirus model, the interaction of ACE2 and Spike activates ASM. This enzyme metabolizes sphingosine into ceramide in acidic endolysosomes and at the plasma membrane at the physiological pH after being released by endolysosomes (168). In the plasma membrane, ASM is important for lipid raft formation in the outer leaflet. In contrast, neutral SM generates ceramide in the inner leaflet, where it interacts with intracellular signaling molecules (169). Increased ceramide in the outer leaflet of the plasma membrane favors SARS-CoV-2 infection. Accordingly, inhibition of ASM by various antidepressant drugs (either selective serotonin reuptake inhibitors (SSRI) or tricyclic class) or silencing of the ASM gene suppresses SARS-CoV-2 infection in Caco-2 cells. Importantly, nasal epithelial cells isolated from amitriptyline-treated (tricyclic class) volunteers were protected from SARS-CoV-2 infection (40). Fluoxetine, an SSRI, also suppresses the replication of different SARS-CoV-2 variants of concern in Vero and A549-hACE2-TMPRSS2 cells, human lung slices (170), and human airway organoid epithelia (41). Geiger and coworkers (2022) showed that fluoxetine and AKS466 - its ASM-independent derivative - target acid ceramidase, leading to the retention of SARS-CoV-2 in lysosomes and inhibiting virus replication in Huh.7, Vero and Calu-3 cells. Ceranib-2, a specific inhibitor of AC, also blocks SARS-CoV-2 replication (42) Further, fluoxetine treatment reduces IL-6 and NF-κB signaling in various cell lines (171). Gene silencing of glucosylceramide synthase (converts ceramide to glucosylceramide), and inhibition by Genz-123346 or Genz-667161, block SARS-CoV-2 early replication steps (43). The same group showed that SARS-CoV-2 infection raises the levels of glycosphingolipids, like dihydrosphingosine, sphingosine, GA1, and GM3, an effect reversed by Genz-123346. This inhibitor does not significantly alter the levels of ceramide, lactosyl ceramide, SM, and GM2 (44). 4-HPR, an inhibitor of dihydroceramide D4-desaturase 1 (DES1- catalyses dihydroceramide in ceramide), inhibits cell-cell fusion and SARS-CoV-2 infection. However, the addition of ceramide fails to revert the effects of 4-HPR. DES1 knockout cells exhibit a similar cell-cell fusion efficiency, despite the enhanced amounts of dihydrosphingosine-derived lipids compared to total sphingolipids. Therefore, the inhibitory effect of 4-HPR on SARS-CoV-2 infection does not rely on DES1, and it is also not related to ACE2 expression or localization. Still, it can be related to the drug effects on decreased membrane fluidity (45).

In conclusion, the metabolism of sphingolipids can determine the outcome of SARS-CoV-2 infection. Ceramide has a dual role in SARS-CoV-2 infection, promoting viral entry when it accumulates in the plasma membrane (40) and blocking viral egress of the viral particles that accumulate in the endolysosomes (42). The balance of ceramide and sphingosine seems crucial during the entry stage since engagement of Spike and ACE2 induces the conversion of sphingomyelin to ceramide favoring infection, whereas the conversion of ceramide to sphingosine blocks ACE2 and Spike’s interaction (39, 40). Patients’ shingolipidome shows that the severity of COVID-19 correlates to ceramide accumulation and S1P reduction (37, 38). S1P results from sphingosine phosphorylation and has anti-inflammatory effects, modulating endothelial cell chemotaxis and barrier integrity. While ceramide and sphingosine promote apoptosis, cell cycle arrest, and differentiation, S1P promotes proliferation and survival (172). Thus, S1P reduction may contribute to the exacerbated inflammation observed in severe COVID-19. For this reason, the S1P modulators, fingolimod, and siponimod, used to treat multiple sclerosis, have been proposed as therapy for treating COVID-19 patients ( Figure 2 ) (173, 174).

Beyond the essential role of cholesterol in structuring lipid nanodomains and involvement in coronavirus entry, its metabolism affects coronavirus infection and host defense mechanisms. Viral infection activates antiviral responses in the host cell, including the transcription of interferon-induced genes (ISGs), such as the cholesterol 25-hydroxylase (CH25H) gene, whose product modifies cholesterol into 25-hydrocholesterol (25HC). 25HC has several antiviral and anti-inflammatory activities in infection by diverse enveloped viruses, including coronaviruses. SARS-CoV-2 infection stimulates the expression of CH25H in Calu-3 and A549-ACE2 cells. The overexpression of CH25H or the treatment with exogenous 25HC abolishes infection by SARS-CoV, MERS-CoV, or SARS-CoV-2 pseudoviruses in Calu-3 cells. Mechanistically, 25HC induces cholesterol trafficking from the plasma membrane to the endoplasmic reticulum, where the ACAT esterifies it, resulting in the formation of lipid droplets and cholesterol depletion in the plasma membrane, restricting the entry of the virus by fusion with the cell surface (19). Heissler and collaborators (2023) report that 25-HC inhibits the common cold coronavirus hCoV-OC43 infection of CHO-K1 cells. They showed that 25HC suppresses SREBP-2 activation, inhibiting cholesterol biosynthesis and activation of ACAT. The latter results in cholesterol sinking into lipid droplets and depletion of the accessible cholesterol from the plasma membrane, which was confirmed by using anthrolysin O (ALOD 4) that binds specifically to this cholesterol pool. Interestingly, 25HC does not alter SM-sequestered cholesterol (20), indicating that 25HC acts independently of lipid nanodomains. In rabbit blood cells that are devoid of internal membranes, the authors show that 25HC is no longer able to induce a reduction of cholesterol at the plasma membrane, suggesting that 25HC effects are dependent on cholesterol transport from the plasma membrane to other organelles (20). Another study indicates that porcine epidemic diarrhea virus (PEDV) downregulates CH25H expression in Vero cells. However, the overexpression of CH25H or 25HC treatment also blocks PEDV infection. Interestingly, CH25H-M, a CH25H mutant lacking hydrolase activity and not producing 25HC, also inhibits PEDV infection, although to a lesser extent. CH25H affects the entry of PEDV and the replication of another porcine coronavirus, the TGEV (21). The infection of IPI-FX cells with porcine deltacoronavirus (PDCoV) upregulates CH25H expression, and 25HC treatment impairs PDCoV infection, not at the adsorption nor at the replication steps (22), but possibly post-entry steps of PDCoV infection such as replication or egress, at least using LLC PK1 cells (23). PDCoV infection induces a transitory increase in IFN-γ and the production of lipid droplets, while 25HC enhances both effects. Furthermore, infection activates the TGF-β signaling pathway, which 25HC reverses. Asiaticoside, an inhibitor of TGF-β, also inhibits PDCoV infection (23). The overexpression of CH25H or 25HC treatment also impairs the infection of HEK293-hACE2 cells by SARS-CoV or SARS-CoV-2 pseudoviruses. The authors show that 25HC accumulates in late endosomes and lysosomes and impairs Spike-mediated membrane fusion (24). The antiviral activity of 25HC was also demonstrated by Yuan and collaborators using the SARS-CoV-2 N antigen expression, viral load reduction, and plaque reduction assays in Vero E6 cells (25). Another relevant aspect of 25HC action on coronavirus infection is the modulation of the inflammatory response since severe cases are characterized by an exacerbated pro-inflammatory response that potentially progresses to organ failure and death. 25HC provides a negative feedback loop of SREBP-dependent activation of pro-inflammatory cytokines (26). Plasma samples of severe COVID-19 patients compared to mild cases show decreased levels of 25HC and increased cytokine levels that correlate with the disease’s severity. Accordingly, CH25H expression is also reduced at mRNA and protein levels in peripheral blood mononuclear cells (PBMC) of severe COVID-19 patients. In this context, treatment of PBMC with 25HC inhibits SREBP activity, the expression of pro-inflammatory genes, such as TNF-α, IL-1β, NOX2, and NLPR3, and the production of inflammatory cytokines, such as IL-6, TNF-α, IL-1b, IFN-γ, IL-8, CCL2, among others (27). In contrast, Zu and collaborators (2020) described a sudden increase of 25HC in the plasma of a severe COVID-19 patient two days before her death, as well as an increase in 25HC levels in serum samples of SARS-CoV-2-infected hACE2 mice compared to the control group. Caco-2 cells also exhibit CH25H upregulation upon SARS-CoV-2 infection, and the overexpression of CH25H inhibits SARS-CoV-2, as well as SARS-CoV-2 pseudovirus particle production in Vero cells, indicating an effect at the entry step of viral infection. 25HC also protects mice from infection by SARS-CoV-2 strain MASCp6.10 (28). Despite the controversial results regarding 25HC levels in the blood upon SARS-CoV-2 infection, both studies reveal a protective role of 25HC against either viral infection or the excessive production of inflammatory mediators caused by infection. In sharp contrast with all these papers reporting a protective role of 25HC, Fessler and collaborators (2022) show that 25HC is effective in controlling the infection caused by the common cold coronavirus (hCoV)-229E, but not the infection and associated damage inflicted by the SARS-CoV-2 (29). Regarding 229E infection, 25HC reduces viral plaque number and size in MRC-5 cells, indicating a post-entry effect. On the contrary, 25HC does not affect SARS-CoV-2 infection in TMPRSS2-expressing VeroE6 cells. Likewise, despite the increase of CH25H expression in K18-hACE2 mice, 25HC treatment does not change the viral load in these mice. Corroborating to the inefficacy of 25HC, CH25H-deficient mice show a similar viral load to wild-type mice. Furthermore, 25HC does not change inflammatory lung indicators to SARS-CoV-2 infection, such as the level of inflammatory mediators and total cell counts in bronchoalveolar lavage fluid (BALF). Of note, 25HC increases inflammatory chemokines in the plasma of infected treated mice and microvascular leakage, and it does not improve lung damage or the survival rate. These observations recommend caution regarding the therapeutic potential of 25HC (29). It is worth mentioning that Marcello and coworkers failed to find any correlation between 25HC in the blood of COVID-19 patients and disease severity. Instead, they observed a 50% decrease in 27HC in COVID-19 patients. In vitro, 27OH inhibits infection of SARS-CoV-2 and HCoV-OC43 (30). The role of other oxysterols in COVID-19 and other coronavirus pathologies still deserves more studies, such as the one by Ohashi and coworkers (2021) describing anti-SARS-CoV-2 activities (31). Finally, we would like to mention that 25HC inhibits caveolin-1 expression and the subsequent Cav-1+-endossome-dependent IFN-γ signaling in the context of Parkinson’s disease (175), an aspect of the regulation of the inflammatory response not investigated so far in coronavirus infection.

Wei and collaborators (2020) correlated SARS-CoV-2 infection with lipoprotein metabolism. First, they demonstrated that SARS-CoV-2 Spike can bind cholesterol and that HDL increases the entry of a pseudovirus expressing SARS-CoV-2 Spike-in Huh-7 cells in a dose-dependent manner. Furthermore, HDL also increases the entry and replication of SARS-CoV-2, a process dependent on the scavenger receptor B type 1 (SR-B1). ACE2 and SR-B1 are co-expressed in pulmonary and extrapulmonary tissue such as the retina, colon, and liver, indicating that SR-B1 facilitates SARS-CoV-2 infection and explaining the viral tropism for these organs (8).

The total cholesterol, HDL, and LDL levels in the blood of COVID-19 patients correlate negatively with infection and disease severity (11). On the other hand, SREBP and NFκB activation are higher in fatal cases compared to survivors. Moreover, the C-terminal fragment of SREBP-2 is present in the blood of COVID-19 patients and correlates to disease severity, cytokine levels, and vascular damage, being higher in septic patients. Treatment of PBMC from COVID-19 patients with fatostatin A (an SREBP-2 processing inhibitor) and SN50 (an NF-κB signaling inhibitor) suppress the production of pro-inflammatory TNF-α and IL-1β. Accordingly, the prototypical pro-inflammatory lipopolysaccharide (LPS) stimulation of HUVEC cells provokes an increase in the C-terminal SREBP content in cell culture lysates and supernatants, whereas the N-terminal SREBP increases only in the cell lysates. These results indicate that the liberation of SREBP C-terminal fragment correlates to inflammatory stimulation and may be a useful indicator of disease severity in COVID-19 patients and a possible target to reduce the exacerbated inflammation response in severe COVID-19 (12). Besides, the same study showed that SREBP-2 silencing blocks the LPS effect. Fatostatin, SN50, and SREBP-2 silencing all diminish sepsis outcomes of the cecal ligation and puncture (CLP) model of sepsis. The authors propose that SARS-CoV-2 infection leads to C-terminal SREBP release, which induces cytokine storm production and the vascular damage observed in COVID-19 septic patients (12). Pharmacological intervention on the transactivation of lipogenic enzymes by the N-terminal SREBP also suppresses MERS-CoV infection, whose primary target is the lipid metabolism pathway, at least in the Calu-3 cells infection model. Using a lipid metabolite library, Yuan and coworkers (2019) identified that AM580, a synthetic analog of retinoic acid receptor alpha (RARα), was protective against MERS-CoV infection in the hepatic cell line Huh7 (13). RARα regulates cholesterol metabolism by stimulating the expression of Apo-A1 in hepatic cells (176), as well as of cholesterol efflux transporters (ABCA1 and ABCG1) in macrophages (177). MERS-CoV infection triggers the expression of genes coding lipogenic enzymes like acetyl-CoA carboxylase (ACC), fatty acid synthase (FAS), and hydroxymethylglutaryl-CoA synthase (HMGCS), besides enhancing the production of lipid droplets. Not surprisingly, AM580 suppresses these effects and reduces MERS-CoV replication drastically in various cell lines, such as Calu-3, A549, Vero, THP-1, human primary monocyte-derived macrophages (MDMs), human primary small airway epithelial cells (HSAEC), as well in human intestinal organoids. AM580 binds to the N-terminus forms of SREBP1 and SREBP2, preventing lipogenic genes’ transactivation. The effect of AM580 is counteracted by exogenous palmitate, indicating that MERS-CoV subverts the fatty acid metabolism to its benefit. AM580 also abolishes the inflammatory response in Huh7 cells and MDMs (13). AM580 is further effective in vivo, as described in session 5 of this review. In the following sections of this review, we will contextualize other aspects of coronaviruses’ subversion of cholesterol metabolism.

Coronavirus infections induce extensive membrane remodeling in the host cell, creating double-membrane vesicles (DMVs) and convoluted membranes (CMs), which shelter replication-transcription complexes (RTCs). These complexes harbor viral proteins and host factors, maintaining a lipid microenvironment that protects newly synthesized RNA molecules from nucleases and innate immune sensors (201). The viral non-structural proteins nsp3, nsp4, and nsp6 are involved in the biogenesis of DMVs (the replication platforms of coronaviruses), which are connected and probably originate from the ER (202–207). The ectopic expression of nsp3 and nsp4 induces the formation of DMVs (203, 205, 208, 209), while the nsp6 bridges DMVs to the ER and to lipid droplets, allowing lipids, but not proteins, to flow from the ER to the DMVs (123). The nsp6 of alpha- and beta-coronaviruses also induces the formation of autophagosomes decorated with LC3 in an ATG5- and PI3K-dependent manner (210). Accordingly, treatment with K22, a drug that targets nsp6, inhibits infection of several coronaviruses (123, 211).

DMVs are like the autophagosomes, regarding their double-membrane constitution and the presence of LC3-I, and they emerge from the ER through mechanisms that involve host autophagic and the ERAD machinery (212, 213). The role of autophagy in coronavirus infection is controversial, though, and each coronavirus exploits the autophagic machinery in a particular way (214). Prentice and coworkers (2004) showed that deletion of the ATG5 gene inhibits the replication of murine hepatitis virus (MHV) strain A59 in stem cells (215). However, Zhao and collaborators (2007) reported that ATG5 is dispensable for efficiently replicating the same virus strain in murine bone marrow-derived macrophages or primary murine fibroblasts (216). Thenceforth, extensive literature shows that coronaviruses explore the autophagic process to produce the membrane system that allows its replication while hampering the completion of autophagy, avoiding its degradation in autophagosomes (193–195, 209, 217–222).

During autophagy, autophagosome membranes expand drastically, a process dependent on cellular factors, as will be briefly described. In nutrient-rich conditions, mTOR, a master suppressor of autophagy, interacts with and inhibits the ULK1 complex formed by ULK1, ATG13, FIP200, and ATG101. Stressing conditions relieve mTOR inhibition, and ULK1 translocates to the ER, recruiting the PI3K complex II (PI3KC3-C2), formed by VPS34, Beclin1-1, VPS15, and ATG14. PIP3KC3 originates from PIP3-rich domains in the ER, where the phagophore formation initiates. Assembly of PI3K3 complex depends on the recruitment of WD repeat domain phosphoinositide interacting 2b (WIPI2B) and double FYVE Containing Protein (DFCP1). WIPI2B binds and engages ATG12, ATG5, and ATG16L1 that lipidates microtubule-associated protein light chain 3 (LC3) and γ-aminobutyric acid receptor-associated proteins (GABARAPs) to membrane PE, originating LC3II, a signature of the autophagic process. LC3 II then regulates the elongation and closure of the autophagosomes carrying the material to be degraded after fusion with lysosomes (reviewed by Nakatogawa, 2020) (223).

Autophagosome formation also requires bulk lipid transfer from the ER’s cytosolic leaflet to the phagophore’s cytosolic leaflet. This process is associated with lipid movement between the cytosolic and the luminal leaflets of each organelle to rebalance the amount of lipids between leaflets. These tasks are executed by the ATG2 lipid transfer protein and the scramblases VMP1 and TMEM41B, respectively (224–228). VMP1 also regulates the contact sites of the ER with the phagophore and other organelles, such as mitochondria, endosomes, and lipid droplets (229). Similar to their role in the expansion of autophagosomes, they are also crucial for the formation of DMVs during coronavirus MHV-A59 infection or after ectopic expression of SARS-CoV-2 nsp3 and nsp4 non-structural proteins (230). VMP1 and TMEM41B were identified as important host factors for virus infection in genome-wide screens (231, 232), and genetic analysis revealed that deficiency of each of these proteins blocks DMV formation in distinct ways. In VMP1 KO cells, nsp3/4-bound DMVs are unable to close, while in TMEM41B KO cells, nsp3/4 complexes are unable to form. Interestingly, both proteins function as cholesterol, and PS scramblases (230), but the inhibition of PS synthesis partially rescues the defect of DMV production only in the absence of VMP1 (233). Deficiency in TMEM41B and VPM1 functions leads to the accumulation of accessible cholesterol in the cytosolic leaflet of the plasma membrane (227) and results in the formation of enlarged lipid droplets. In TMEM41B KO cells, this is due to a decrease in the mobilization of fatty acids from lipid droplets to mitochondria and their ß-oxidation (224). The absence of VMP1 also induces large lipid droplets because of the increased lipid flux from the ER to lipid droplets (229).

A genome-wide CRISPR screen of human coronaviruses-infected cells identified genes of the phosphatidylinositol phosphate biosynthesis and the cholesterol homeostasis host pathways, indicating the importance of lipid metabolism in coronavirus infection. Many of these genes were further validated by examining the effectiveness of virus infection in loss-of-function experiments, in which susceptible cells were manipulated to harbor mutant candidate genes. Moreover, several drugs targeting lipid metabolism exhibited inhibitory activity on viral replication: SAR405 (a selective and ATP-competitive inhibitor of class III PI3K - PIK3C3), YM201636 (a PIKfyve inhibitor), PF-429242 (a reversible, competitive aminopyrrolidineamide inhibitor of MBTPS1), 25HC (which promotes ER retention of the SCAP/SREBP complex), Fatostatin (which binds to SCAP and inhibits ER-to-Golgi translocation of SREBPs), Bardoxolone (an activator of the KEAP1-NRF2 complex) (232). Notably, apilimod (a PIKfyve inhibitor) also presents antiviral activity (234–236).

Class III PI3K (also known as VPS34), an essential player in the formation of autophagosomes and vesicular sorting and trafficking, is also central to the formation of DMVs. The VPS34 inhibitors, VPS34-IN1, PIK-III, and SAR405, all inhibit SARS-CoV-2 replication in Calu-3 cells. These drugs may act during early or later steps of the viral infection (14). Other inhibitors of fatty acid metabolism show antiviral activity in post-entry steps: orlistat (inhibits gastric lipases and FASN); triacsin C (targets long chain ACS); TOFA (targets acetyl-CoA) and C75 (inhibits FASN) (14).

Interestingly, both etomoxir (an inhibitor of carnitine palmitoyltransferase 1A -CPT1A), which blocks translocation of fatty acids into the mitochondria, and trimetazidine (inhibits long-chain 3-ketoacyl-CoA thiolase) do not impair SARS-CoV-2 infection, ruling out a role of β-oxidation in viral replication. Furthermore, the inhibitors VPS34-IN1, orlistat, TOFA, or A922500 inhibit the formation of replicative centers based on dsRNA labeling. Furthermore, TOFA and A922500 show much higher inhibition when the viral titers were analyzed, suggesting they have other additional effects besides inhibiting RNA synthesis (14).

The deletion of FASN in Caco-2 cells confirmed the central role of fatty acid metabolism in viral replication, while supplementation with palmitate and oleic acid was partially able to reverse FASN deficiency (14). Yuen and coworkers (2021) also reported an inhibitory effect of VPS34-IN1 in the replication and infectivity of SARS-CoV-2 in VeroE6 cells. VPS34-IN1 and its analog, compound 19, also impair viral infection of ex-vivo human lung tissue (221). They also analyzed the effects of other autophagy inhibitors (targets in parentheses) during viral infection of VeroE6 cells. They found that SBI-0206965 (ULK1), 3-methyladenine – 3-MA (class I PI3K), and hydroxychloroquine – HCQ (acidification of autophagolysosome) impaired viral replication. Interestingly, inhibition of ULK1, which is upstream of VPS34 in the autophagy pathway, stimulated SARS-CoV-2 infection instead of inhibiting it. Moreover, 3-MA showed a minimal inhibitory effect, while HCQ strongly inhibited SARS-CoV-2 replication. Twu and collaborators (2021) also report a central role of the class III PI3K in SARS-CoV-2 replication. The VPS34 inhibitor, VPS34-IN-2 (also known as PIK-III), inhibits nsp3/4-induced DMV formation and SARS-CoV-2 replication in A549/ACE2 and Calu-3 cells. Knockdown of VPS34 and Beclin 1 or the PI3P effector DFCP1 cells have the same effect in the A549 cell model (209). Taken together, these data reinforce the importance of lipid metabolism for SARS-CoV-2 infection and its potential as a therapeutic target.