Targeting pentose phosphate pathway for SARS-CoV-2 therapy

It becomes more and more obvious that deregulation of host metabolism play an important role in SARS-CoV-2 pathogenesis with implication for increased risk of severe course of COVID-19. Furthermore, it is expected that COVID-19 patients recovered from severe disease may experience long-term metabolic disorders. Thereby understanding the consequences of SARS-CoV-2 infection on host metabolism can facilitate efforts for effective treatment option. We have previously shown that SARS-CoV-2-infected cells undergo a shift towards glycolysis and that 2-deoxy-D-glucose (2DG) inhibits SARS-CoV-2 replication. Here, we show that also pentose phosphate pathway (PPP) is remarkably deregulated. Since PPP supplies ribonucleotides for SARS-CoV-2 replication, this could represent an attractive target for an intervention. On that account, we employed the transketolase inhibitor benfooxythiamine and showed dose-dependent inhibition of SARS-CoV-2 in non-toxic concentrations. Importantly, the antiviral efficacy of benfooxythiamine was further increased in combination with 2DG.

The rapid spread of SARS-CoV-2 resulted in millions of infected people worldwide. Though the majority of patients exhibit mild to moderate symptoms, around 20% of patients develop severe disease. Effective antiviral drugs against SARS-CoV-2 are currently lacking.
Emerging evidence suggests that metabolic disorders are associated with an increased risk of severe COVID-19. Increased amounts of glucose in the sera of COVID-19 patients were correlated with poor prognosis in individuals both with and without pre-existing diabetes (Shen et al., 2020). At the cellular level, SARS-CoV-2 induces a shift of the host cell metabolism towards glycolysis in infected cells, and the glycolysis inhibitor 2-deoxy-D-glucose (2DG), which targets hexokinase (the rate-limiting enzyme in glycolysis), interferes with SARS-CoV-2 infection in colon adenocarcinoma (Caco-2) cells and monocytes Codo et al., 2020). Hence, agents that antagonize virus-induced changes in the host cell metabolism, especially glycolysis, are candidates for COVID-19 therapy.
Using a proteomics dataset derived from SARS-CoV-2-infected Caco-2 cells , we identified further metabolic changes associated with SARS-CoV-2 infection. Cellular levels of key drivers of glycolysis including hexokinase 1 and 2 (HK1 and HK2) were increased in SARS-CoV-2-infected cells, while negative regulators such as fructose-1,6-bisphosphatase 1 and 2 (FBP1 and FBP2) were reduced ( Figure S1A). Changes in glycolysis can also affect other metabolic pathways. Metabolic intermediates generated by the glycolytic enzymes supply both the oxidative and the non-oxidative branch of the pentose phosphate pathway (PPP) with intermediates (Stincone et al., 2015). Notably, transketolase (TKT) and transaldolase 1 (TALDO1), two constituents of the non-oxidative PPP branch (Stincone et al., 2015), displayed increased levels in SARS-CoV-2-infected cells ( Figure S1A), suggesting that the non-oxidative PPP branch is also regulated in response to SARS-CoV-2 infection.
The non-oxidative PPP branch converts glycolytic intermediates into ribose-5-phosphate, which is required for the synthesis of nucleic acids, and sugar phosphate precursors, which are necessary for the synthesis of amino acids (Stincone et al., 2015). TKT, a key enzyme of the non-oxidative PPP, becomes activated by the binding of one molecule of thiamine diphosphate and of a bivalent cation (Zhao and Zhong, 2009). Recently, it has been shown that TKT activity may be strongly increased by the formation of a heterodimer with transketolase like 1 (TKTL1) (Li et al., 2019).
Irreversible inhibitors of TKT such as oxythiamine (an inactive analogue of thiamine) have been developed (Tylicki et al., 2018). To investigate whether TKT inhibitors affect SARS-CoV-2 replication, we used benfooxythiamine (BOT), a prodrug of the TKT inhibitor oxythiamine, currently evaluated in pre-clinical experiments (Coy, 2017). Indeed, non-toxic BOT concentrations inhibited the replication of two different SARS-CoV-2 isolates (FFM1 and FFM7) in a concentration-dependent manner, as indicated by immunostaining for the SARS-CoV-2 Spike (S) protein and the detection of genomic RNA by qPCR ( Figure S1B-S1D).
The antiviral effects of BOT were further increased by the glycolysis inhibitor 2DG (Figure S1D-S1F), indicating that the combined inhibition of ribose-5-phosphate production via the nonoxidative PPP branch and glycolysis is more effective in suppressing SARS-CoV-2 replication than the inhibition of either metabolic pathway alone. Mechanistically, 2DG inhibits the glycolysis enzymes hexokinase and phosphoglucose isomerase (PGI), both required for synthesis of fructose-6-phosphate, which serve as glycolytic intermediate for generating ribose-5-phosphate in non-oxidative PPP branch (Stincone et al., 2015).
Furthermore, ribose-5-phosphate can be also generated in the oxidative PPP branch through the metabolic conversion of glucose-6-phosphate (Stincone et al., 2015). Since 2DG decreases cellular levels of glucose-6-phosphate, it may additionally lower ribose-5-phosphate levels by reducing its availability for metabolizing in the oxidative branch of PPP. Importantly, the oxidative PPP branch seems also be activated in SARS-CoV-2-infected cells as indicated by increased 6-phoshogluconate dehydrogenase (PGD) levels ( Figure S1A). PGD catalyzes the conversion of 6-phosphogluconate into ribulose-5-phosphate, a precursor of ribose-5phosphate, in the oxidative PPP branch (Stincone et al., 2015).
Hence, 2DG could increase the effects of BOT by further reducing ribobose-5-phosphate synthesis through interrupting the supply of glycolytic intermediates into both oxidative and non-oxidative branch of PPP.
In conclusion, SARS-CoV-2 infection is associated with changes in the regulation of the PPP, which has potential implications for antiviral therapy. The TKT inhibitor BOT inhibited SARS-CoV-2 replication, indicating a role of the non-oxidative PPP branch in the virus replication cycle. The addition of the glycolysis inhibitor 2DG further enhanced the antiviral activity of BOT, which may result from the interruption of supply of glycolytic intermediates into PPP. The findings also demonstrate that the simultaneous inhibition of different metabolic pathways has potential as an antiviral treatment strategy for SARS-CoV-2-infected individuals ( Figure S1G).
Notably, uncontrolled glucose and pentose phosphate levels correlate with worse prognosis and higher mortality in COVID-19 patients (Deng et al., 2020;Thomas et al., 2020;Zhu et al., 2020), and reports show that a number of COVID-19 survivors suffer from metabolic disorders including abnormal glucose metabolism (Ayres, 2020). Therefore, targeting the host cell metabolism may prevent the metabolic derailment of immune cells in addition to the inhibition of SARS-CoV-2 replication and thus contribute to recovery in severe COVID-19 disease (Ayres, 2020).

Proteomics data
The LC-MS/MS proteomics data with the dataset identifier PXD017710 have been previously published . Proteins derived from Caco-2 cells infected with SARS-CoV-2 after 24 h post infection and matching search terms "glycolysis" and "pentose phosphate pathway" in KEGG database were selected for further analysis. Significantly deregulated proteins (p value ≤ 0.5) in mock versus infected cells were plotted using the heatmaps2 function of the gplots package of the R suite. Briefly, raw enrichment values were used to calculate per row Z scores. Maximum and minimum Z score were attributed the colour red and blue respectively. Scripts available upon request.

Antiviral and cytotoxicity assay
Caco-2 cells were seeded in 96-well plate. After reaching confluency, cell were pre-treated with benfooxythiamine (BOT, Zyagnum AG) for 24 h. Then the cells were additionally treated with 2-deoxy-D-glucose (2DG) or let untreated and infected with SARS-CoV-2 at MOI 0.01. 24 h later, cells and supernatants were harvested to determine viral load. The cytotoxic effect of BOT alone or in combination with 2DG was evaluated my MTT assay as previously described.

Statistics
All experiments were performed under similar conditions in three independent replicates.
GraphPad Prism 8 software was used to plot the graphical representation and to perform statistical analysis. Statistical significance was calculated by two sided unpaired t-test as described in the figure legends (ns -not significant, * p < 0.05, ** p < 0.01, *** p < 0.005).