Infection by SARS-CoV-2 has been modeled by making use of the interactions between viral proteins and human enzymes (5). Five of these enzymes (NDUFB9, ETFA, ATP6AP1, ATP6V1A and ATP1B1) are involved in oxidative phosphorylation. These interactions are interpreted as inhibitory or as relocating the involved proteins to RTCs instead of mitochondria, which would suppress their function. This has been modeled by constraining oxidative phosphorylation to have a zero flux in the model of infected cells. For the rest of metabolic enzymes expressed in the lung and interacting with viral proteins, the constraints on their rates have been removed (this is equivalent to assume that their rates could increase as a result of these enzymes being highly concentrated within RTCs). Possible changes in gene expression were modeled by identifying the transcription factors that interact with viral proteins. Only TCF12 was identified. This gene is known to be a transcriptional repressor. Thus, the genes regulated by TCF12 are assumed to be upregulated and the constraints on their reactions were also removed in the model. The model predicts a maximal rate of virion formation of 0.012 nmol-virions per hour and gram of dry infected cells. The metabolic model of SARS-CoV-2 infected cells and the model of non-infected lung cells are available at https://github.com/SergioBordel/COVID-19.
The effects of inhibiting the function of each of the enzymes in the virus-host interactome were tested with the function block of pyTARG. This restricts the rate of reactions catalyzed by the tested enzymes to 0.1 times its original value and computes the ratio by which the objective function (the rate of virion production or cellular biomass, for infected and for healthy cells respectively) decreases. This ratio takes values from 1 (indicating no effect) and 0.1 (indicating that the objective function decreases in the same proportion as the flux in the targeted reaction). The model of non-infected lung cells was used as a quantification of potential secondary effects. In order to select suitable targets, a difference between both ratios (non-infected and infected) of 0.3 was used as cutoff value. Using this method, 10 human enzymes were identified (see Table 1).
A literature search was carried out in order to identify inhibitors of the selected targets, and to evaluate potential secondary effects. When possible, drugs approved or under clinical trials were selected, but we also considered those in a pre-clinical status or under investigation (see references and structures of these compounds in supplementary files S2-S3). The literature survey yielded a total of 10 compounds (see Table 1). The affinity of the identified drugs for some the target enzymes (those for which a reliable 3D structure could be obtained) was further validated using docking. Docking was also used for the identification of putative inhibitors for enzymes without known inhibitors, as it is the case of Tipifarnib and Lonafarnib, predicted to bind FAR2 (Figure 1).
Table 1. Table with the selected target enzymes for SARS-CoV-2 treatment. See an extended version of Table 1 in Supplementary file S2. **These compounds have been identified in our docking analyses rather than in the literature revision approach.
Viral-host interaction
|
Ensembl ID
(human gene)
|
Definition
|
Cellular Process
|
Known Inhibitors
|
Orf3a-ALG5
|
ENSG00000120697
|
dolichyl-phosphate beta-glucosyltransferase
|
N-Glycan biosynthesis
|
Celgosivir; Tunicamycin
|
Orf9c-ALG8
|
ENSG00000159063
|
alpha-1,3-glucosyltransferase
|
N-Glycan biosynthesis
|
Celgosivir; Tunicamycin
|
Nsp4-ALG11
|
ENSG00000253710
|
alpha-1,2-mannosyltransferase
|
N-Glycan biosynthesis
|
Celgosivir; Tunicamycin
|
Nsp7-CYB5R3
|
ENSG00000100243
|
NADH cytochrome-b5 reductase
|
Sugar metabolism
|
Propylthiouracil; ZINC-39395747; ZINC-05626394
|
Spike-ZDHHC5
|
ENSG00000156599
|
palmitoyltransferase 5
|
Protein palmitoylation
|
2-bromopalmitate
|
Nsp2-SLC27A2
|
ENSG00000140284
|
solute carrier family 27 member 2
|
Lipid metabolism
|
Lipofermata
|
Nsp7-ACSL3
|
ENSG00000123983
|
long-chain acyl-CoA synthetase
|
Lipid metabolism
|
Triacsin C
|
Nsp7-AGPS
|
ENSG00000018510
|
alkylglycerone phosphate synthase
|
Lipid metabolism
|
Benzyl isothiocyanate; ZINC-69435460
|
Orf9c-FAR2
|
ENSG00000064763
|
alcohol-forming fatty acyl-CoA reductase
|
Lipid metabolism
|
Tipifarnib*; Lonafarnib*
|
Nsp7-NAT14
|
ENSG00000090971
|
putative N-acetyltransferase 14
|
Not determined
|
Not found
|
Some of the 10 putative targets are known oncogenes for which drugs are already in clinical trials (AGPS, SLC27A2 and ACSL3). AGPS is inhibited by benzyl isothiocyanate (BITC), the anti-fungal agent Antimycin A and the investigational new drug ZINC-69435460 [3-(2-fluorophenyl)-N-(1-(2-oxo2,3-dihydro-1H-benzo[d]imidazol-5-yl)ethyl)butanamide]. Antimycin A can be discarded due to its cytotoxicity, as it is also an inhibitor of cytochrome C reductase, essential for eukaryotic cells (7). The two other inhibitors could be promising alternatives. BITC inhibits the proliferation of human glioma U87MG and hepatic carcinoma HepG2 cells, as well as aggressive human breast tumors, and ZINC-69435460, makes multiple specific interactions with residues in the AGPS active site, and inhibits its activity (8). SLC27A2 (also known as FATP2) is a transmembrane transporter coenzyme that participates in the long-chain fatty acid beta-oxidation and peroxidase lipid metabolism. In a recent publication it has been shown that the selective pharmacological inhibition of FATP2 with lipofermata (5-bromo-5'-phenyl-spiro[3H-indole-3,2'(3'H)-[1,3,4]thiadiazol]-2(1H)-one) suppresses the activity of pathologically activated neutrophils and substantially reduces melanoma growth and invasion (9). ACSL3 is one of the five members of the long chain acyl-CoA synthetase (ACSL) family. Specifically, ACSL3 has been identified as a host factor required for Poliovirus replication, which, as SARS-CoV-2, is a positive-sense single-stranded RNA virus that replicates its genome in association with membranes (10). Inhibition of ACSL3 with inhibitors such as Triacsin C has been shown to reduce virion formation in human cells infected with different viruses, including cytomegalovirus, rotavirus and hepatitis C (11, 12). For these reasons, and according to our metabolic modelling results, this drug constitutes a very promising therapeutic option to treat SARS-CoV-2 infection.
ZDHHC5 has been found to interact with the Spike protein of SARS-CoV-2 (5). The protein ZDHHC5 is involved in the palmitoylation of proteins, which plays an important role in protein-membrane interactions, protein trafficking and enzyme activity by contributing to their membrane association. Palmitoylation of the SARS-CoV-Spike protein has been reported to facilitate its fusion with the host cellular membrane, which clearly suggests ZDHHC5 as potential target for therapeutic inhibition of SARS-CoV and other coronaviruses such as SARS-CoV-2. In vitro experiments have revealed some molecules that inhibit cellular processes associated with palmitoylation, with one of the best examples being 2-Bromopalmitate (13).
The enzyme NADH cytochrome-b5 reductase (i.e. CYB5R3) is inhibited by the approved drug, propylthiouracil (PTU) (14). PTU is an antithyroid medication used in the therapy of hyperthyroidism and the Graves disease. It is still unclear how PTU exerts its inhibitory effects on CYB5R3 at a molecular level. Nevertheless, a repurposing strategy for COVID-19 treatment may merit investigation. Other described inhibitors of CYB5R3 are the investigational new drugs ZINC-39395747 and ZINC-05626394, which have been shown to be more potent inhibitors of CYB5R3 than PTU (15); matching with our docking results (see Figure 2).
The rest of the identified targets are involved in N-glycan biosynthesis and glycosylation. Glycosylation is vital in the maturation process of viral proteins. The Spike protein of SARS-CoV-2 has 16 glycosylation residues (16). The importance of protein glycosylation is evidenced by the use of Celgosivir (an inhibitor of N-glycan synthesis) in treating cells that harbour Dengue virus, which prevents proper protein glycosylation, leading to protein misfolding and impaired replicative efficiency (17). Celgosivir (and also other iminosugars such as deoxynojirimycin or castanospermine) has been shown to be a very promising drug, not only for the treatment of Dengue virus or members of the same family (like for example ZIKV), but also for other viruses belonging to different families, such as the hepatitis C or Influenza viruses (18-20). As 3 of our 10 selected enzyme targets are involved in the N-Glycan biosynthesis pathway (i.e. ALG5, ALG8, ALG11) we propose treating SARS-CoV-2 with inhibitors of glycoprotein biosynthesis such as the previous mentioned iminosugars, or the nucleoside antibiotic tunicamycin, which also inhibits N-linked protein glycosylation. Interestingly, this compound has been shown to display strong antiviral activity against certain viruses like for example the Epstein-Barr virus (21). Similarly, drugs that disrupt the integrity of the Golgi apparatus (monensin, brefeldin A, bafilomycin, nocodazole) also alters glycoprotein synthesis and could be also effective for treating COVID-19.