SARS-CoV-2 exploits host DGAT and ADRP for efficient replication

Abstract

Coronavirus Disease 2019 (COVID-19) is predominantly a respiratory tract infection that significantly rewires the host metabolism. Here, we monitored a cohort of COVID-19 patients' plasma lipidome over the disease course and identified triacylglycerol (TG) as the dominant lipid class present in severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2)-induced metabolic dysregulation. In particular, we pinpointed the lipid droplet (LD)-formation enzyme diacylglycerol acyltransferase (DGAT) and the LD stabilizer adipocyte differentiation-related protein (ADRP) to be essential host factors for SARS-CoV-2 replication. Mechanistically, viral nucleo capsid protein drives DGAT1/2 gene expression to facilitate LD formation and associates with ADRP on the LD surface to complete the viral replication cycle. DGAT gene depletion reduces SARS-CoV-2 protein synthesis without compromising viral genome replication/transcription. Importantly, a cheap and orally available DGAT inhibitor, xanthohumol, was found to suppress SARS-CoV-2 replication and the associated pulmonary inflammation in a hamster model. Our findings not only uncovered the mechanistic role of SARS-CoV-2 nucleocapsid protein to exploit LDs-oriented network for heightened metabolic demand, but also the potential to target the LDs-synthetase DGAT and LDs-stabilizer ADRP for COVID-19 treatment.

Fig. 1. Plasma lipidome of COVID-19 patients.

a Lipidome of COVID-19 patients’ plasma samples over the disease course. All blood samples were collected before the patients’ discharge from the hospital. Day 0 means the first day of patient hospitalization and blood taking. The hierarchical clustering analysis was based on the identified lipid metabolites with significant changes in quantity, comparing with Day 0. Each rectangle represents a lipid colored by its normalized intensity scale from blue (decreased level) to red (increased level). b Pie chart showing the relative ratio of eight lipid classes that are significantly perturbed. These lipids belong to BMP bismono-acylglycerophosphate. DG diacylglycerol, PC glycerophosphocholines, PE glycerophosphoethanolamines, PG glycerophosphoglycerols, PS glycerophosphoserines, SM sphingomyelin, TG triacylglycerol. c Overview of pathway analysis based on the identified lipids. The y-axis, “-log(p)”, indicates the log₁₀ transformed p-value after enrichment analysis; the x-axis, “Pathway Impact”, represents the value calculated from the pathway topology analysis. d Boxplots illustrate the top two dominant perturbed lipid class representatives in the time-course study, DG (18:1/18:2/0:0) (P = 0.0097, day 0 vs day 7) and TG (36:2, day 0 vs day 7) (P = 0.0084). y-Axis represents the peak height of selected lipids based on the LC–MS data.

Fig. 2. Comparative lipidome between SARS-CoV- and SARS-CoV-2- infections.

a Schematic flow chart showing the comparative analysis between SARS-CoV- and SARS-CoV-2- infections. Human lung Calu-3 cells were infected (1 MOI) by SARS-CoV or SARS-CoV-2, followed by lipid extraction at 8 and 24 hpi, respectively. b Hierarchical clustering analysis was generated based on all significantly increased/decreased lipids comparing either SARS-CoV- or SARS-CoV-2- infection with that of mock infection, respectively. The 3D principal component analysis (PCA) score plots showing the distribution pattern of the detectable lipid profile, which differentiate SARS-CoV-2-infection (red) from mock- nfection (yellow) and SARS-CoV-infection (blue) groups. The triangles represent the distribution of an individual sample dot within each group (n = 6).

Fig. 3. DGATs are potential therapeutic targets for COVID-19.

a, b DGAT genes are required for SARS-CoV-2 replication. siRNA knockdown of either DGAT1 (two distinct siRNA 1_1 and 1_2) or DGAT2 (siRNA 2_1 and 2_2) were performed on human colorectal Caco-2 (a) or lung Calu-3 cells (b) before virus infection for 48 h (0.1 MOI). Viral yields in the cell lysate were determined by RT-qPCR and normalized with human β-actin. One-way ANOVA was used for comparison with the scramble siRNA pre-treated group. c Dose–response analysis of the compound Xanthohumol is shown, depicting both antiviral activity (red) and cytotoxicity (black). The gray dash line indicates 50% of the mock-treated control with EC₅₀, CC₅₀, and chemical structure displayed. d Xanthohumol inhibited SARS-CoV-2 replication in Calu-3 cells that were infected by 0.1 MOI SARS-CoV-2. Viral loads in the cell supernatant and cell lysate were determined at 48 hpi by RT-qPCR assays, respectively. Data represent means ± SD. One-way ANOVA was used for comparison with the DMSO control group. e Xanthohumol inhibited SARS-CoV-2 replication in human embryonic stem cells-derived cardiomyocytes (hES-CMs) that were infected by 0.1 MOI SARS-CoV-2. Viral loads in the cell supernatant and cell lysate were determined at 24 hpi by RT-qPCR assays, respectively. Data represent means ± SD. One-way ANOVA was used for comparison with the DMSO control group. f siRNA-treated-Huh7 cells were infected with SARS-CoV-2 (10 MOI for 12 h) before staining with DAPI (blue), viral nucleocapsid protein (NP) (red), and BODIPY 493/503 lipid probe (green) for LD detection. Scale bar: 100 µm. g Knockdown of DGAT1/2 reduced viral yields in the cell culture supernatant but not cell lysate. A single-cycle SARS-CoV-2 replication assay was performed in Caco-2 cells transfected with the indicated siRNA. Viral yields in the cell lysate and supernatant were determined by RT-qPCR. Data collected at 2 hpi were taken as a baseline and at 10 hpi taken as the completion time for one virus life cycle. One-way ANOVA was used for comparison with the scrambled siRNA pre-treated group. For all statistical analyses, ^*P < 0.05, ^**P < 0.01, ^***P < 0.001, ^****P < 0.0001, n.s. indicates P > 0.05. h Cell lysate at 10 hpi was also utilized for western blotting detecting SARS-CoV-2 NP and host β-actin. Shown are triplicates (i.e., three different siRNAs) of each group.

Fig. 4. The interplay between SARS-CoV-2 and LDs-relevant host factors.

a Expression of individual SARS-CoV-2 protein when co-transfected with the DGAT1 or DGAT2 promoter-reporter plasmid. Shown is the Western blot detecting the viral proteins by anti-flag antibody. b SARS-CoV-2-NP trans-activates DGAT1/2 genes. Huh7 cells were transfected with the indicated reporter-gene plasmid and viral ORF clones, individually. The luciferase activity reflecting DGAT1 (upper panel) or DGAT2 (lower panel) gene expression was determined at 48 hpi. One-way AVONA was used for comparison with the control group. c SARS-CoV-2-NP does not interact with DGAT1 or DGAT2. A co-immunoprecipitation assay was conducted in 293 T cells transfected with NP and either DGAT1/2 plasmids. After pull-down, NP was detected by anti-flag while DGAT1/2 was detected by specific antibodies. d Flow chart showing the preparation of hamster lung extracts for identification of SARS-CoV-2 proteins associated with LDs. e ADRP interacts with viral NP but not spike. A co-immunoprecipitation assay was conducted in infected or non-infected hamster lungs. After pull-down using an anti-ADRP antibody, viral NP or spike (RBD) was detected by specific antibodies. f Knockdown of ADRP gene expression reduced SARS-CoV-2 replication in CaCo-2 cells (0.1 MOI, 48 hpi). Viral yields in the supernatant and cell lysate were determined by RT-qPCR, respectively. One-way ANOVA was used for comparison with the scrambled siRNA pre-treated group. ^***P < 0.001, ^**P < 0.01, ^*P < 0.05.

Fig. 5. Xanthohumol inhibits SARS-CoV-2 replication in vivo.

a Experimental design of the oral Xanthohumol regimen. b Viral yield in the hamster lung (n = 4) was harvested at 4 dpi and titrated by RT-qPCR assay. Data represent mean ± SD. One-way ANOVA was used for comparison with the DMSO control group. For all statistical analyses above, ^*P < 0.05, ^**P < 0.01, and ^***P < 0.001. c Representative images of viral NP distribution in lung tissues section from infected hamsters treated with DMSO or Xanthohumol or Remdesivir, at 4 dpi. SARS-CoV-2 NP (green) and cell nuclei (blue) were stained. Scale bar, 200 µm. These representative images were selected from a pool of 10 images captured in 4 hamsters per group. NP-positive cells per 50× field per the lung section of a hamster were quantified. One-way ANOVA followed by Dunnett’s post test and compared with DMSO control. ^****P < 0.0001. d The mRNA expression of IFN-γ and IL-10 in hamster lung was quantified at 4 dpi using qRT-PCR assays. One-way ANOVA was used for comparison with the DMSO group. e The cytokines IL-6 and TNF-α levels in hamster serum samples were quantified at 4 dpi using ELISA. One-way ANOVA was used for comparison with the DMSO group. f Representative images of H&E-stained lung tissue section from the hamsters treated with different groups indicated, illustrating the severity of alveolar infiltration (gold arrows). The lower panel are enlarged images corresponding to each black box of the upper panel. These representative images were selected from a pool of 10 images captured in 4 hamsters per group. Scale bar, 200 μm.

Fig. 6. Proposed model for host DGAT and ADRP in SARS-CoV-2 life cycle.

Virus manipulates host glycerophospholipid metabolism through upregulation of TG synthesis and LD formation. Specifically, viral NP trans-activates both DGAT1 and DGAT2 mRNA expression binding to their promoter regions, resulting in LD accumulation. Both DGATs and the other LD surface protein ADRP are required for SARS-CoV-2 replication. The identified compound Xanthohumol is a pan-DGAT1/2 inhibitor that exhibits anti-SARS-CoV-2 activity both in vitro and in vivo.